Knowledge Flows in European Industry (Routledge Studies in Business Organizations and Networks)

Knowledge Flows in European Industry Core European Policy concerns include raising the international competitiveness o...

Author: Y. Caloghirou

26 downloads 812 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Knowledge Flows in European Industry

Core European Policy concerns include raising the international competitiveness of European industry, developing a European economic ‘space’ and the European Research Area, narrowing the technology gap among EU member states and improving the economic and social cohesion of the region. These goals require politics to enhance linkages among knowledge-intensive activities within and across EU member states. Knowledge Flows in European Industry presents the results of an extensive research programme funded by the European Commission to empirically appraise the dissemination of knowledge relevant to the innovative activities of European manufacturing and service sectors. This includes knowledge ﬂows between industrial ﬁrms and sources of knowledge such as universities, technical institutes and government laboratories across Europe. This book explores the extent and density of innovation-related knowledge ﬂows affecting the innovative capacity of European industry, the mechanisms that support such ﬂows, the incentives to access and transmit results, and the determinants of knowledge transmission. Inferences are also made about the nature of the innovation systems that sustain and are inﬂuenced by such ﬂows as well as the tendency for these innovation systems to converge into larger European Innovation Systems. Yannis Caloghirou is Associate Professor of Economics of Technology and Industrial Strategy at the National Technical University of Athens, Greece. Anastasia Constantelou is Assistant Professor of Innovation Management at the University of the Aegean, Greece. Nicholas S. Vonortas is Professor of Economics and International Affairs at George Washington University, USA.

Routledge studies in business organizations and networks

1 Democracy and Efﬁciency in the Economic Enterprise Edited by Ugo Pagano and Robert Rowthorn 2 Towards a Competence Theory of the Firm Edited by Nicolai J. Foss and Christian Knudsen 3 Uncertainty and Economic Evolution Essays in honour of Armen A. Alchian Edited by John R. Lott Jr 4 The End of the Professions? The restructuring of professional work Edited by Jane Broadbent, Michael Dietrich and Jennifer Roberts 5 Shopﬂoor Matters Labor-management relations in twentieth-century American manufacturing David Fairris 6 The Organisation of the Firm International business perspectives Edited by Ram Mudambi and Martin Ricketts

7 Organizing Industrial Activities Across Firm Boundaries Anna Dubois 8 Economic Organisation, Capabilities and Coordination Edited by Nicolai Foss and Brian J. Loasby 9 The Changing Boundaries of the Firm Explaining evolving inter-ﬁrm relations Edited by Massimo G. Colombo 10 Authority and Control in Modern Industry Theoretical and empirical perspectives Edited by Paul L. Robertson 11 Interﬁrm Networks Organization and industrial competitiveness Edited by Anna Grandori 12 Privatization and Supply Chain Management Andrew Cox, Lisa Harris and David Parker

13 The Governance of Large Technical Systems Edited by Olivier Coutard

21 Workaholism in Organizations Antecedents and consequences Ronald J. Burke

14 Stability and Change in High-Tech Enterprises Organisational practices and routines Neil Costello

22 The Construction Industry An international comparison Edited by Gerhard Bosch and Peter Philips

15 The New Mutualism in Public Policy Johnston Birchall 16 An Econometric Analysis of the Real Estate Market and Investment Peijie Wang 17 Managing Buyer–Supplier Relations The winning edge through speciﬁcation management Rajesh Nellore 18 Supply Chains, Markets and Power Mapping buyer and supplier power regimes Andrew Cox, Paul Ireland, Chris Lonsdale, Joe Sanderson and Glyn Watson 19 Managing Professional Identities Knowledge, performativity, and the ‘new’ professional Edited by Mike Dent and Stephen Whitehead 20 A Comparison of Small and Medium Enterprises in Europe and in the USA Solomon Karmel and Justin Bryon

23 Economic Geography of Higher Education Knowledge, infrastructure and learning regions Edited by Roel Rutten, Frans Boekema and Elsa Kuijpers 24 Economies of Network Industries Hans-Werner Gottinger 25 The Corporation Investment, mergers and growth Dennis C. Mueller 26 Industrial and Labour Market Policy and Performance Issues and perspectives Edited by Dan Coffey and Carole Thornley 27 Organization and Identity Edited by Alison Linstead and Stephen Linstead 28 Thinking Organization Edited by Stephen Linstead and Alison Linstead 29 Information Warfare in Business Strategies of control and resistance in the network society Iain Munro

30 Business Clusters An international perspective Martin Perry 31 Markets in Fashion A phenomenological approach Patrik Aspers 32 Working in the Service Sector A tale from different worlds Edited by Gerhard Bosch and Steffen Lehndorff 33 Strategic and Organizational Change From production to retailing in UK brewing 1950–1990 Alistair Mutch

34 Towards Better Performing Transport Networks Edited by Bart Jourquin, Piet Rietveld and Kerstin Westin 35 Knowledge Flows in European Industry Edited by Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas 36 Change in the Construction Industry An account of the UK construction industry reform movement 1993–2003 David M. Adamson and Tony Pollington

Knowledge Flows in European Industry

Edited by Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas

First published 2006 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Simultaneously published in the USA and Canada by Routledge 270 Madison Ave, New York, NY 10016 Routledge is an imprint of the Taylor & Francis Group, an informa business © 2006 Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas, editorial matter and selection; individual chapters, the contributors.

This edition published in the Taylor & Francis e-Library, 2006.

“To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” All rights reserved. No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data A catalog record for this book has been requested ISBN10: 0-415-32707-5 (hbk) ISBN10: 0-203-35393-5 (ebk) ISBN13: 978-0-415-32707-7 (hbk) ISBN13: 978-0-203-35393-6 (ebk)

Contents

List of ﬁgures List of tables List of boxes List of contributors Foreword Acknowledgements Abbreviations and acronyms 1 By way of an introduction: knowledge ﬂows: the drivers for the creation of a knowledge-based economy

x xii xiv xv xviii xxii xxiv

1

YANNIS CALOGHIROU, ANASTASIA CONSTANTELOU AND NICHOLAS S. VONORTAS

PART I

Setting the agenda for the study of knowledge ﬂows 2 Theoretical foundations and key concepts

25 27

YANNIS CALOGHIROU, ANASTASIA CONSTANTELOU AND NICHOLAS S. VONORTAS

3 Conventional and experimental indicators of knowledge ﬂows

45

ANTHONY ARUNDEL AND ANASTASIA CONSTANTELOU

4 An operational framework for the study of knowledge ﬂows YANNIS CALOGHIROU, ANASTASIA CONSTANTELOU AND NICHOLAS S. VONORTAS

67

viii Contents 5 Knowledge ﬂows in European industry: an overview of evidence from surveys and case studies

76

YANNIS CALOGHIROU, ANASTASIA CONSTANTELOU AND NICHOLAS S. VONORTAS

PART II

Aspects of knowledge ﬂows 6 Facilitators and impediments to knowledge sharing: an exploration of different organizational forms

99

101

METTE PRAEST KNUDSEN

7 Knowledge ﬂows in the Danish ICT sector: the paradox of advanced demand and mediocre supply

115

CHRISTIAN R. PEDERSEN, MICHAEL S. DAHL AND BENT DALUM

8 Collaboration and innovation outputs

158

ANTHONY ARUNDEL AND CATALINA BORDOY

9 Firm size and openness: the driving forces of university– industry collaboration

183

ROBERTO FONTANA, ALDO GEUNA AND MIREILLE MATT

10 Self-selection and learning in European research joint ventures: a micro-econometric analysis of participation and patenting

210

LUCIA CUSMANO

11 The evolution of intra- and inter-sector knowledge spillovers in the EU Framework Programmes

242

MICHEL DUMONT AND AGGELOS TSAKANIKAS

12 Unveiling the texture of a European Research Area: emergence of oligarchic networks under EU Framework Programmes

268

STEFANO BRESCHI AND LUCIA CUSMANO

13 Small worlds and technology networks: the case of European research collaboration BART VERSPAGEN

299

Contents ix PART III

Policies for knowledge ﬂows

319

14 Overview of innovation policy affecting knowledge ﬂows in EU member states

321

ANTHONY ARUNDEL AND NICHOLAS S. VONORTAS

15 Towards a new agenda for enhancing knowledge ﬂows in Europe

332

YANNIS CALOGHIROU, ANASTASIA CONSTANTELOU AND NICHOLAS S. VONORTAS

Index

345

Figures

2.1 3.1 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14 6.1 6.2 6.3 7.1 7.2 7.3

Knowledge diffusion and absorption in the innovation system The complexity of knowledge ﬂows R&D activities Sources of new ideas for innovation Internet use Methods to protect innovations by country Most important innovation by country Source of new scientists/engineers hired to work on most important innovation by country Contributors to the original idea for most important innovation by country Contributors to the completion of most important innovation by country Importance of internal versus external knowledge sources for successful completion by country Location of most important external source of knowledge by country Underlying reason for decision to obtain knowledge from most important external knowledge source by country Contacting the most important external source of knowledge Type of knowledge received from most important external source Methods of communication to obtain external knowledge Functional organization of case A The organization of case B The organization of case C Danish ICT export performance of manufactured goods at the OECD market Export specialization in ICT and electro medical products, 1998 Municipality-level ICT employment specialization in 1992

41 51 80 81 82 83 84 85 86 87 88 89 90 91 92 93 105 108 110 121 123 140

Figures xi 7.4 A7.1 A7.2 9.1 9.2 9.3 9.4 10.1 10.2 11.1 12.1 12.2 12.3 12.4 12.5 12.6 12.7 12.8 12.9 13.1 13.2 13.3

Municipality-level ICT employment specialization in 1999 141 Share of the total Danish ICT employment at municipality level, 1992 152 Share of the total Danish ICT employment at municipality level, 1999 153 External sources of information for innovation ideas by NACE sector 189 External sources of information for innovation completion by NACE sector 190 Most important external source of information for innovation ideas by NACE sector 191 Most important external source of information for innovation completion by NACE sector 192 RJV participants’ average level of innovativeness. Patent stock (1978–96) of RJV participants by year t 218 RJV entrants’ innovative experience. Patent stock at time t 1 of new RJV participants at time t 219 EU clusters of the highest intra- and inter-sector spillovers in three FWP periods 255 Frequency of organizations by number of participations in RJV projects 276 Frequency of RJV membership: Prime Contractors versus Partners 277 Bipartite graph of RJV projects and organizations 279 Alternative representations of the unipartite graph 280 Degree distribution for the giant component of the RJV network 284–285 Preferential attachment in the RJV network 286 Resilience of the RJV network 289 Visual illustration of the RJV network 292 The k-core of the RJV network 293 The connected Caveman World and a local view of the Moore Graph 307 Length and clustering as a function of the graph tuning parameter 309 Theoretical predictions and actual values for clustering and characteristic path length, three network 316

Tables

3.1

Information on knowledge ﬂows collected by major surveys of manufacturing ﬁrms 53 4.1 A typology of knowledge ﬂows in an organizational context 73 5.1 KNOW survey – response rates 78 5.2 KNOW survey – country and sector distribution of sample 78 7.1 Segmentation of the ICT sector, 1997 119 7.2 Export specialization, 1990–96 – selected countries compared to OECD average 122 7.3 Export specialization in ICT and electro medical products, 1990–98 124–125 7.4 Export specialization in mobile phones, 1990–98 126 7.5 Percentage of population aged 25–64 by level of educational attainment, 1998 131 7.6 Percentage of employees in ICT with tertiary education, 1998 132 7.7 R&D specialization in manufacturing for selected small OECD countries 134 7.8 ICT patents granted at the United States Patent Ofﬁce 135 7.9 Relative number of product innovative ﬁrms in ﬁve EFS mega clusters, 1993–95 136 7.10 Overall performance of Danish ICT on a benchmark of statistical indicators 139 A7.1 Comparison of ICT deﬁnitions 145 A7.2 The IKE segmentation of the ICT sector in NACE codes 146 A7.3 ICT segments in SITC rev. 3 compared with NACE 147–149 A7.4 The used statistical indicators for the benchmark 150 A7.5 Regional employment specialization in 1998 in ICT segments and NACE codes 151 8.1 Innovative characteristics of the respondent ﬁrms 164 8.2 Percentage of ﬁrms making any use of each innovation method 165 8.3 Percentage of innovations developed via each innovation method 166

Tables xiii 8.4 8.5 8.6 9.1 9.2 9.3 A9.1 10.1 10.2 10.3 10.4 10.5 10.6 10.7 10.8 10.9 10.10 10.11 10.12 10.13 11.1a 11.1b 11.1c 11.2 11.3 11.4 11.5 A11.1 12.1 12.2 12.3 12.4 12.5 13.1

Mean innovation sales share by most important appropriation method 169 Logit results for introducing one or more product innovations developed through collaboration 174 Non-linear regression results for innovation sales share 176 Differences in responses between groups 193 Share of respondents for PROs contract classes 195 Regression summary – Probit and Truncated regressions 202 Descriptive statistics for selected variables 206 RTD priorities in Framework Programmes 213 EUREKA projects, by technological area 214 ICT sample: industry composition 215 MB sample: industry composition 215 Descriptive statistics: employment and patenting 216 Transition groups: RJV afﬁliation over time 220 Self-selection and learning: patenting differential and transition pattern 221 Self-selection and learning: patenting differential change 222 ICT sample – patenting differential and size: analysis of variance 224 MB sample – patenting differential and size: analysis of variance 225 ICT sample negative binomial random effects 230 MB sample negative binomial random effects 231 Size class analysis 232 FWP funding of the various technological areas 244 Funding of the ﬁfth FWP 246 Indicative funding of the priority key areas of the sixth FWP 246 Number of projects in different technological FWP ﬁelds 248 Sectors’ participation in FWPs 249 Country shift-share effects 259 Top ﬁve of highest and lowest total shift-share effects at country level 260–261 EU Framework Programmes included in the EU-RJV database 263–264 The EU RJV dataset: summary statistics 275 Number of partnerships between Prime Contractors 278 Collaboration with different Prime Contractors 278 Summary of results of the analysis of the unipartite RJV network 281 The core of RJV network 291 Basic statistics of the networks 313

Boxes

11.1 11.2

Mathematical deﬁnition of spillovers Shift-share analysis

252 257

Contributors

Anthony Arundel is a Senior Researcher at the Maastricht Economic Research Institute on Innovation and Technology (MERIT) of the University of Maastricht. His main research interest concerns the development, implementation and statistical analysis of questionnaire surveys on the innovative behaviour and strategies of ﬁrms, particularly with respect to intellectual property, knowledge ﬂows, environmental innovation and biotechnology. Catalina Bordoy is a Research Assistant at the Maastricht Economic Research Institute on Innovation and Technology (MERIT) of the University of Maastricht. She has worked on several projects on innovation indicators and the use of intellectual property rights at universities and public research institutes. In 2003 she obtained a PhD in Economics at the Universitat Autonoma de Barcelona (Spain). Stefano Breschi is Associate Professor of Industrial Economics and Deputy Director of CESPRI, Bocconi University, Milan, Italy. His main research interests are the economics of technical change, industrial organization and industrial dynamics, theory of the ﬁrm, and networks, economic geography and regional economics. Yannis Caloghirou is Associate Professor of Economics of Technology and Industrial Strategy at the National Technical University of Athens, where he is leading two research groups on information society and innovation studies at the Laboratory of Industrial and Energy Economics. He served as Secretary for the Information Society at the Greek Ministry of Economy and Finance (June 2002 to March 2004) and as Secretary-General for Industry at the Greek Ministry of Development (April 2000 to June 2002). He sat in a number of European Commission Expert Groups. He is co-editor of the recently published volume European Collaboration in Research and Development: Business Strategy and Public Policy (Edward Elgar, 2004). Anastasia Constantelou is Assistant Professor of Innovation Management at the University of the Aegean and has been conducting research for

xvi

Contributors

several years at the Laboratory of Industrial and Energy Economics, National Technical University of Athens. Her research interests focus on technology and innovation management, the economics of science and technology and the socio-economic implications of the information society. Lucia Cusmano is Assistant Professor at University dell’Insubria Varese, Italy, and has been conducting research for several years at CESPRI, University Luigi Bocconi, Milan. Her main ﬁelds of research relate to the economics of innovation and technical change, technology policy and cooperative research and development, national and local innovation systems, and innovation and development. Michael S. Dahl is Assistant Research Professor, at DRUID and Department of Communication Technology, Institute of Electronic Systems, Aalborg University, Denmark. His primary research interests concern the innovation system of the telecommunications industry and theoretical and empirical studies of new ﬁrm formation and knowledge diffusion. Bent Dalum is Associate Professor in Economics at the Department of Business Studies, Aalborg University, Denmark, and Deputy director of DRUID, the Danish Research Unit in Industrial Dynamics. He specializes in the economics of innovation with an emphasis on telecommunications. Michel Dumont is Assistant Professor at the Department of International Economics, International Management and Diplomacy of the University of Antwerp. He teaches international economic issues and international economic organizations. His current research focuses on the impact of international competition and technological change on the labour market position of low-skilled workers. Roberto Fontana is post-doctoral Research Fellow at the Centre for Research on Innovation and Internationalisation (CESPRI), Bocconi University. Current research interests are in the ﬁeld of the economics of information and communication technologies, emergence and evolution of industries in high-tech ﬁelds and university–industry relationships. Aldo Geuna is a Senior Lecturer at SPRU – Science and Technology Policy Research, University of Sussex. His research interests include the economics of knowledge production and distribution, economics of innovation and technological change, science and technology policy. Mette Praest Knudsen is Associate Professor at the Department of Marketing, University of Southern Denmark in Odense. Her main areas of research cover international knowledge ﬂows, strategic alliances, international market relations, e-business and international technology management.

Contributors

xvii

Mireille Matt is Associate Professor at the Bureau d’Économie Théorique et Appliquée (BETA) of the University Louis Pasteur, Strasbourg, France. Her research programme focuses on the analysis of technology policy, of research and development (R&D) cooperative agreements, the economic evaluation of large R&D programmes and the economics of science. Christian R. Pedersen is a Doctoral Student in DRUID/IKE Group, Department of Business Studies, Aalborg University, Denmark. His research interests focus on regional clusters and the development perspectives for the information and communication technology sector in North Denmark. Aggelos Tsakanikas is a Visiting Lecturer at the Department of Economics, University of Peloponnesus, and a Research Associate at the Foundation for Economic and Industrial Research (IOBE), Athens. His main research interests are found in the area of ﬁrm technology strategy and the economics of innovation. Bart Verspagen is the Scientiﬁc Director of the Eindhoven Centre for Innovation Studies, University of Eindhoven, the Netherlands. His work spans the broad area of the economics of technological change and innovation, and captures both empirical and theoretical developments. Nicholas S. Vonortas is Professor of Economics and International Affairs, Director of the Center for International Science and Technology Policy and Director of the Graduate Program on International Science and Technology Policy at the George Washington University, USA. His teaching and research interests are in industrial organization, the economics of technological change and science and technology policy. He specializes in strategic partnerships, innovation networks, technology transfer, technology and competition policy, and the appraisal (evaluation) of the economic returns of R&D programmes.

Foreword

The European Union has identiﬁed the transition to a knowledge-based society as one of its key political challenges. The Lisbon Summit of March 2000 famously ﬁxed the very ambitious objective of the European Union becoming the most competitive knowledge-based economy in the world by the end of the decade. Since then the so-called ‘Lisbon process’ has generated an impressive momentum, which has led to profound changes in European policies and, perhaps even more importantly, in the way in which these policies are designed. Despite this impressive progress, it must be acknowledged that our general understanding of the functioning of a knowledge-based economy is still somewhat rudimentary. While we know that knowledge is very important to the economy, we need to develop a deeper understanding of the processes through which knowledge plays a part in economic activities, and of the full implications of the transition towards a knowledgebased economy. The study of knowledge ﬂows provides important insights into how knowledge is incorporated into economic activities, and reveals some of the implications for policies that aim to develop a European knowledgebased society. The analysis of knowledge ﬂows engenders important lessons for policy makers of all kinds, and at all levels. But knowledge ﬂows are difﬁcult to grasp, in both an analytical and an empirical sense. One of the major endeavours that focused on this particular challenge was the ‘Innovation-related knowledge ﬂows in European industry’ (or KNOW) project. This project was supported by the European Union as part of the Fourth Framework Programme and involved leading research teams from seven European countries. The KNOW project was coordinated by Yannis Caloghirou of the Laboratory of Industrial and Energy Economics (LIEE) at the National Technical University of Athens, a research group with wide experience in the study of business strategy and public policy in the process of technological innovation. I am very pleased to see that this research team has gone an important step further, in presenting more focused, and at the same time more in-

Foreword xix depth, results in this volume. From the point of view of the European policy agenda, knowledge ﬂows in European industry are a particularly fascinating topic: •

•

Within the overall Lisbon strategy there is an urgent need to know more about the ways in which ﬁrms manage incoming and outgoing knowledge ﬂows. While there is a body of empirical evidence on large companies, data on the behaviour of small and medium-sized enterprises (SMEs) and on the situation in the more peripheral regions of the European Union are scant. In the speciﬁc context of the European Research Area, the chapters in this book address key questions regarding the link between the overall Lisbon agenda and European research policy: cooperation between different actors, and notably between ﬁrms and universities, is a crucial element of a European knowledge-based economy. At the same time, a careful analysis of the absorptive capacities of ﬁrms is needed in order to develop appropriate strategies.

The vital importance of efforts to build a European research area (ERA) is reﬂected in the future orientations for European research policy. Community research funding is set to double under the latest proposals for the ﬁnancial framework of the Union, and future action will be deﬁned by the Seventh Research Framework Programme. The guidelines currently being proposed for the next Framework Programme are oriented around six major objectives: • • • • • •

creating European centres of excellence through collaboration between laboratories; launching large scale European technological initiatives; stimulating the creativity of basic research through competition between teams at European level; making Europe more attractive to the best researchers; developing research infrastructures of European interest; improving the coordination of national research programmes.

These structural objectives represent a new phase in the building up of an ERA and constitutes important elements of Europe’s knowledge-based society. It is important here to note that social sciences and the humanities will ﬁgure largely in all the structural objectives of the next Framework Programme. In particular: •

Reinforced efforts to foster collaborative research in the social science disciplines could offer crucial insight into a whole range of key concerns highlighted in European policy discussions. A deeper understanding of such issues as competitiveness and growth, convergence

xx Foreword

•

•

•

•

•

and diversity, and demographic change, are central to the effective functioning of a knowledge-based economy and society, and to the overall economic competitiveness of the Union. The development of technology platforms could provide an opportunity for the social sciences and the humanities to interact more closely with technological research activities and relevant stakeholders. Greater integration should be the result, with more active participation in the deﬁnition of strategic research agendas and a more substantial role in their implementation. The priority given to fostering basic research will allow social scientists to adopt a bottom-up approach, building the socio-economic knowledge base in ways that will equip the Union to meet future policy challenges, and to address issues of immediate relevance. The strengthening of policies to make the EU more attractive to highly trained researchers is vital to the building of a ﬁrm foundation for European social sciences within the ERA. Capacity building is of particular relevance in a number of the new member states where historically the social sciences as scientiﬁc disciplines have suffered, and where some persistent weaknesses remain. A fully functioning and effective framework within the ERA for the social sciences and the humanities relies in particular on access to ‘non-physical’ research infrastructures such as databases and electronic archives. Efforts must be renewed to equip researchers with the appropriate resources – from fast, large capacity electronic networks, to information resources for funding schemes and career opportunities. Substantial gains are to be made through improved coordination of national research programmes in the social sciences. The lack of industrial funding means that national funds are the most signiﬁcant ﬁnancial source for the social sciences and the humanities throughout Europe: as a result the potential for true coordination across national borders has sometimes been limited.

The study of knowledge ﬂows is a major building-block in the analysis and understanding of innovation. Obviously, the more innovative European industry becomes, the more competitive will be our economies in the global market place. In the Fourth and Fifth EU Framework Programmes for research, the European Commission’s Directorate-General for Research supported a signiﬁcant number of research projects and thematic networks that directly addressed innovation issues. European research on innovation has thus seen impressive development in recent years. It has been strongly supported by the European Union and is an integral part of the European Union’s programme for research in the social sciences. The researchers involved in the KNOW project set out to appraise the

Foreword xxi diffusion of knowledge of relevance to the innovative activities of European industry, including both the manufacturing and service sectors. This appraisal focused on questions of interest to regional, national and panEuropean science, technology and innovation policy. The project launched a major empirical investigation of the traditional and emerging routes of innovation-related knowledge dissemination in European industry. This included knowledge ﬂows between industrial ﬁrms as well as between ﬁrms and other sources of relevant knowledge, such as universities, technical institutes and government laboratories. Recent advances in the theory of innovation systems guided the empirical investigation. The results of the project were used to evaluate the success of policies that had been implemented, to suggest improvements to existing policies and to identify future policy options regarding the creation and transmission of new technological knowledge in the European Union. The project encompassed the extent, density and mechanisms of innovation-related knowledge ﬂows affecting the innovative capacity of European industry and the mechanisms that support such ﬂows. It also examined the incentives to access and transmit knowledge, and the determinants of knowledge transmission. On the basis of the observed knowledge ﬂows, and the evolution of the determinants and transmission mechanisms of such ﬂows, the project made inferences about the nature of the innovation systems that sustain, and are inﬂuenced by, such ﬂows as well as about the tendency for these innovation systems to converge into a larger European innovation system. I am conﬁdent that this research effort is resulting in signiﬁcant advances in knowledge, which have the potential to support policy making in Europe and beyond, and will continue to have an impact long after the end of the Fourth and Fifth EU Framework Programmes. The results of the KNOW project should also support the implementation of a number of other EU activities and objectives. On behalf of the Commission, I would like to warmly congratulate the research teams involved for their commitment and for the accomplishment of this important piece of work. Dr Achilleas Mitsos, Director-General, DG Research at the European Commission Brussels, March 2005

Acknowledgements

This book is based on the outcome of a joint effort of fellow academics and researchers from seven European universities and research centres in the context of the research project ‘Innovation-Related Knowledge Flows in European Industry: Extent, Mechanisms and Implications’ funded by the Targeted Socio-Economic Research Programme (TSER) of the European Commission’s Research Directorate (Contract Number SOE1-CT981118). We are grateful to the European Commission for its generous ﬁnancial support and especially to the project ofﬁcer Virginia Vitorino for her diligence in overseeing the project and in providing helpful advice at critical moments. The project would not have been possible without the contribution of colleagues from the partnering universities and research centres, most of whom committed themselves to the preparation of this volume. In particular, we wish to thank all fellow academics and researchers that became involved at various stages of the project, namely Anthony Arundel, Catalina Bordoy, Stefano Breschi, Assimina Christoforou, Thomas Cleff, Lucia Cusmano, Dirk Czarnitzki, Aldo Geuna, Michael Dahl, Bent Dalum, Roberto Fontana, Stavros Ioannidis, Alexandra Karoutsou, Ioanna Kastelli, Yannis Katsoulakos, Georg Licht, Elisabeth Lykoyanni, Franco Malerba, Mireille Matt, Christian Pedersen, Mette Praest Knudsen, Aggelos Tsakanikas, David Ulph, Gert Villumsen, Bart Verspagen and Sandrine Wolff. We also acknowledge the beneficial involvement of EC officials Paraskevas Caracostas and Artemios Kourtesis who, as invited reviewers at various seminars and conferences organized in the context of the KNOW project, provided constructive criticism and valuable comments to participants. We are particularly grateful to Inderscience Enterprises Ltd, for the kind permission to reprint Chapter 12. This chapter under the same title by Breschi and Cusmano was ﬁrst published in the International Journal of Technology Management, 27(8): 747–772, Copyright (2004). Last but not least, our special thanks go to Terry Clague, Editorial Assistant at Routledge who bore with us patiently during the preparation of the

Acknowledgements xxiii book and to Cynthia Little who provided valuable assistance and editorial support in the preparation of the manuscript. Needless to say, the views expressed in this book as well as any errors or omissions remain the exclusive responsibility of the editors and of the authors of individual chapters.

Abbreviations and acronyms

AAU ACTS AMT ATP BDI BSE CATI CEO CIS CKO CMS CORDIS CPMR DTU EC EFS EMU EPO ERA ESPRIT ETSI EU EU-FWP EU-RJV FDI FP FWP GDP GSM ICT IP ISDN IST

Aalborg University Advanced Communications Technologies and Services Advanced Manufacturing Technology Advanced Technology Programme Bundesverband der Deutschen Industrie Bovine Spongiform Encephalopathy Cooperative Agreements and Technology Indicators Chief Executive Ofﬁcer Community Innovation Survey Chief Knowledge Ofﬁcer Carnegie Mellon Survey Community Research and Development Information Service Conference of Peripheral Maritime Regions (of Europe) Technical University of Denmark European Commission Danish Agency for Trade and Industry European Monetary Union European Patent Ofﬁce European Research Area European Strategic Programme for Information Technology European Telecommunications Standards Institute European Union European Union Framework Programme(s) The European Research Joint Ventures (database) Foreign Direct Investment Framework Programme(s) Framework Programmes Gross Domestic Product Global System for Mobile Communications Information and Communication Technologies Integrated Project(s) Integrated Services Digital Network Information Society Technologies

Abbreviations and acronyms xxv IT JRC KNOW LIEE MERIT MB MNE NIS NMT NOE NSI NTUA OECD PACE PC PRI PRO RACE R&D RJV RTD RTD&D SCI SI SME TFP TRIC TSER TTO UMTS WWW XDSL

Information Technologies Joint Research Centre Knowledge Flows in European Industries Laboratory of Industrial and Energy Economics Maastricht Economics Research Institute on Innovation and Technology Medical and Biotechnology Multinational Enterprise National Innovation Systems Nordic Mobile Telephony System Network(s) of Excellence National Systems of Innovation National Technical University of Athens Organisation for Economic Cooperation and Development Policies, Appropriation and Competitiveness for European Enterprises Personal Computer Public Research Institute Public Research Organization(s) Research in Advanced Communications for Europe Research and Development Research Joint Venture Research and Technological Development Research and Technological Development and Demonstration Science Citation Index System of Innovation Small- and Medium-sized Enterprises Total Factor Productivity Thematic Research and Innovation Centres (France) Targeted Socio-Economic Research Technology Transfer Ofﬁces Universal Mobile Telephony System World Wide Web xDigital Subscriber Line

1

By way of an introduction Knowledge ﬂows: The drivers for the creation of a knowledge-based economy Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas

Where is the information we have lost in data? Where is the knowledge we have lost in information? Where is the wisdom we have lost in knowledge?1

1.1 The context: the emergence of a knowledge-based economy The importance of knowledge, learning and innovation for industrial growth, productivity and the international competitiveness of individuals, ﬁrms, regions, nations and economic entities (such as the European Union (EU)) is acknowledged in the current academic literature and by policy and political decision makers. This increased importance of knowledge associated with learning is becoming evident in the spheres of economic life, both public and private. This interest in the role of knowledge cannot be attributed only to its accelerating growth; it stems from its untapped transformative potential. This is illustrated in the widely used umbrella term ‘the knowledge-based economy’. It can be argued that this term is overused and often perceived as rather vague in meaning. Perhaps, as sceptics legitimately claim (Smith 2002: 6), the term is ‘at best a widely used metaphor, rather than a clear concept’. Nevertheless, it reﬂects an increasing recognition from economists, analysts, policy makers, politicians, business people and citizens, of the role and signiﬁcance of knowledge in economic growth and its potential to produce new forms of economic organization that are both quantitatively and qualitatively different in terms of how they use knowledge. It is within this context that scholars as well as international institutions (OECD, EU, the World Bank) have linked the concept of the ‘knowledgebased economy’ with an emerging and gradual structural transformation of the advanced industrial economies and societies generally. For instance, as long ago as the mid-1990s, the OECD was claiming that the

2 Y. Caloghirou, A. Constantelou and N.S. Vonortas most contemporary developed economies were becoming knowledge based. This is generally being acknowledged by all sectors of society, as evidenced by the increasing investments in all forms of knowledge by individuals, ﬁrms and nations. In the year 2000, in the Lisbon Extraordinary European Council (23 and 24 March), the EU linked the strategic vision of the creation of ‘the most dynamic knowledge-economy in the world’ with a great transformation, namely reforming while maintaining the European social model. It is too soon for scholars to be absolutely certain about the ultimate outcome and ﬁnal character of the transformation being wrought by the new knowledge-based economy, although they may use the term. Dominique Foray (2004: ix), for instance, argues that the knowledgebased economy is a ‘plausible scenario of structural transformation, though still highly uncertain’, but supports the use of the term knowledge-based economy because it ‘enables readers to fully understand a qualitative innovation in the organization and conduct of modern economic life – namely the factors determining the success of ﬁrms and national economies are more dependent than ever on the capacity to produce and use knowledge’ (Foray 2004: x). Bengt-Åke Lundvall and others emphasize the decisive role of learning and the capability to learn in the ‘new economy’, stressing the learning element in the transformation, and introducing the learning economy perspective. In this context, Lundvall argues for the learning-economy as being the most appropriate concept for the study and the understanding of ‘the new historical period that our economies have entered, where the role of knowledge and learning is important . . . and where the success of individuals, ﬁrms, regions and national economies reﬂects their capability to learn’ (Lundvall 1996: 2). Moreover, Lundvall (2004) argues that it is useful to rethink the concept of the new economy as being a shift in techno-economic paradigm in line with the analytical schemes of Carlota Perez and Chris Freeman (Perez 1983; Freeman and Perez 1984). And he adds: Seen in this light the real challenge is much broader than that signalled by the new economy. The important message is that there is enormous untapped growth potential that could be mobilized to solve social and economic problems if our societies push for institutional reforms and organizational change that promote learning processes. (Lundvall 2004: 1) The idea of ‘an emerging new paradigm creating knowledge-based economies and societies’ (Rodrigues 2002: 3) is central to the intellectual thinking behind the Lisbon strategy. It seems to us that the knowledge-based economy or the learning economy is a reality-based conceptualization, which is deeper, richer and

By way of an introduction 3 historically better rooted than the notion of the new economy, which gained favour – especially in the USA – in the late 1990s. The new economy was largely associated with the dot.com boom and the accompanying stock market bubble.

1.2 The background: the interlocking of knowledge and the economy in a historical perspective The contribution of knowledge to the process of economic, technological and social change is not new: it has always played a part and often a crucial part in such change. Every economic activity is based on the use of knowledge in some way both in modern societies where ‘the central phenomenon is . . . that as an aggregate we know more . . . and that every aspect of our material existence has been altered by our new knowledge’ (Mokyr 2002: 2), and in pre-historic (Paleolithic and Neolithic) societies (Smith 2002: 9). Storytelling is one of the oldest forms of knowledge sharing used throughout human history. Thus, in this sense there has always been a knowledge dependent economy. Moreover, knowledge is a generic, general, but historically shaped notion, that grows and evolves with time. In this respect, modern distinctions between types of knowledge, such as Lundvall and Johnson’s (1994) taxonomy, which takes four categories – namely know what, know why, know how and know who – can be seen as being rooted in the Aristotelian categories of the three intellectual virtues (i.e. Episteme, Techni, Phronissis). In his book, The Gifts of Athena, the well-known economic historian Joel Mokyr (2002: 1) points to the deep historical roots of the notion of knowledge, in saying that: ‘The growth of human knowledge is one of the deepest and most elusive elements in history’. Furthermore, in this wellreceived book Mokyr explores the historical origins of the knowledge economy. He ﬁnds that the growth explosion in the modern west in the past two centuries was driven not just by the appearance of new technological ideas but also by the improved access to these ideas in society at large, as made possible by social networks comprising universities, publishers, professional sciences and kindred institutions. Mokyr (2002: Ch. 2) particularly stresses the importance of the ‘Industrial Enlightenment’. This phenomenon, which can be traced back to the mid-eighteenth century and produced the Industrial Revolution in Britain, had consequences for the whole of the western world. In Mokyr’s analysis, Industrial Enlightenment is related to a series of cultural and intellectual developments (free inquiry, intellectual tolerance, strong belief in the desirability and feasibility of technical and material progress, widespread dissemination of scientiﬁc and technical ﬁndings and solutions, challenges to established procedures) associated with the rapid growth of technological knowledge. This intellectual progress facilitated

4 Y. Caloghirou, A. Constantelou and N.S. Vonortas interactions among different types of knowledge. In particular, it enabled the connection between practical and theoretical knowledge, the linking between practitioners and theorists and, above all, made all types of knowledge more widely accessible. An environment conducive to the generation and diffusion of technological knowledge was thus created, leading to the sustained economic growth and continuing technological change that characterized the Industrial Age. The formation of a critical mass of knowledge, accompanied by its development and diffusion, was one of the main drivers of the Industrial Revolution of the nineteenth century.

1.3 The continuity: knowledge and information processing in the industrial economy Based on the ﬁndings from economic analysis, economic history and innovation studies, a number of stylized trends regarding the relationship between information and knowledge on the one hand, and the economy (and particularly the industrial economy) on the other, can be established. First, knowledge has, to a certain degree, always been present in economic activity and it has always been necessary for the functioning of an economy. Second, the organization of the production and diffusion of knowledge leading to the wide application of scientiﬁc theories is typical of an industrial economy. Third, industrial economies cannot and did not function without a massive information-processing and information-managing apparatus. A strong information dimension was a feature of an industrial economy. In this respect, Beniger (1986) supported the thesis that modern information and communication technologies (ICTs) began to take shape in the 1830s with the introduction of railroads for steampowered trains. In his book, The Control Revolution, Beniger (1986) shows that the great innovations in the area of information processing and handling were all introduced between 1840 and 1930. This applies both in terms of adoption of new procedures (standardization, printed forms, registries, trade catalogues, ﬁling systems, card indexes, incident reports, record-keeping, management techniques, organization of work, advertising, etc.) and the use of new technological systems (telegraphy, rotary power printing, typewriters, punch-card processing, postage stamps, telephony, motion pictures, radio, television, etc.). He argues, therefore, that the big push forward in terms of the workforce employed in these activities, at least in the US, took place between 1880 and 1930. During this period, the percentage of the workforce employed in the information processing and information handling industries rose from 6.5 per cent to 24.5 per cent while total industrial employment in the US in 1930 was of the order of 30 per cent. The creation and functioning of the new information handling industry, according to Beniger (1986), was very much associated with the Control Revolution, a social and economic response to the

By way of an introduction 5 necessity for managing the huge and speedy ﬂows that accompanied industrial development. In this context, the increase in the accessibility – in terms of technical ease and cost – of information in the modern industrial age allowed it to become a major driving force of the economy and society in general. Finally, the role of knowledge and learning had become very important for the manufacturing sector of industrial economies (e.g. Pasinetti 1981). So, if these stylized trends are characteristic of an industrial economy, what is new in the notion of a knowledge-based economy? What is special about knowledge in the information age? To what extent are we entering a new form of the knowledge-driven economy? What role do knowledge ﬂows and their relevant channels and mechanisms play in the transition towards a knowledge-intensive type of economic organization?

1.4 The break: ICT-related technological change It can be seen from the above that the knowledge-based economy did not begin with the emergence and current explosive development of ICTs and the Internet boom. Nor did it end when dot.coms became dot-bombs. But, ICT-related technological change provides a new and different operational environment for the generation, transmission and use of knowledge. Despite the fact that ICTs are ‘primarily an information management and distribution resource’ (Smith 2002: 12), the rapid growth in the use of the Internet and the diffusion of its culture signalled much more than the availability of a low cost and easily accessed platform for the distribution of information as a ‘conveyor’ of ideas, thoughts, concepts and knowledge. It is now clear that information society technologies (IST), the worldwide information networks, infrastructures and institutions associated with the new analytical and policy thinking perceptions and social attitudes regarding the communication and the use of information, have facilitated the storage, codiﬁcation, exchange, transmission, diffusion, use and re-use, combination and re-combination of information and knowledge to an extent – both quantitatively and qualitatively – that is unprecedented. The introduction in the ICT cluster of a new technology can not only enhance our ability to process data and to identify regularities and patterns, but also can inﬂuence our willingness to act and work accordingly. Data-mining technology is a good example of such an enabler. It allows, ‘for the intensive statistical analysis of large masses of structured, factual data and makes patterns of movement of all kinds buried in huge masses of data immediately obvious. Once a pattern is visible people can act’ (Hunter 2002: 4). It is in practice ‘the death of the distance’ (Cairncross 2001), the signiﬁcant reduction in the barriers to accessing information and codiﬁed knowledge, the possibility for greater codiﬁcation, transferability, re-use

6 Y. Caloghirou, A. Constantelou and N.S. Vonortas and re-combination of knowledge and the widespread user-friendly culture of the Internet – especially among the younger generation – that is facilitating the continuous interaction between creators and users of information and knowledge, and is allowing also to be information producers. The creation of a culture of, and the building of mechanisms and channels for, the exchange and sharing of information is leading to a new style of working and changed perceptions of the value of information and knowledge for economic activity and their importance to economic, industrial and business success. While Industrial Enlightment contributed signiﬁcantly to bringing about the Industrial Revolution and the creation of modern industrial economies, a new transformation and an accompanying intellectual climate is becoming manifest by the increasing accessibility to knowledge. Soete (2004) argues that it is the process of Knowledge (and Information) Enlightment that is one of the main drivers of a new structural transformation, already in progress, of both the economy and society. There is, then, a discernible trend towards a knowledge-based economy. But, as Mokyr warns, the transformation is inevitably going to be gradual. He also reminds us that throughout economic history, the effects of the most signiﬁcant changes do not surface for a generation.

1.5 Transformation in progress: the new structural trends We are currently experiencing a transition phase. In the perspective of the transformation already in progress, a number of new structural tendencies are visible in all spheres of economic, industrial and intellectual life. These structural changes have been brought about through a process of systemic interaction between technical change, new perceptions and policy-led reform. It is of the utmost importance to attempt seriously to identify areas where elements and islands of change are emerging. These changes can be seen in the role played by ICTs: they have become a driving force in the restructuring and growth of developed economies and in possibilities for the future. In addition, changes are occurring in the industrial structure, in the classiﬁcation of industries in relation to their technological level, in the character, structure and organization of the modern ﬁrm and in the paradigm of economic policy. Finally, changes can be seen in the new emerging and rapidly growing sub-disciplines and research ﬁelds within the disciplines of economics and management. 1.5.1 The role of ICTs: a universal, enabling and disruptive bundle of technologies The rapid formation and growth of an ICT cluster from the technologydriven digital convergence between information technology (IT) and telecommunications, created the technical background for the structural

By way of an introduction 7 changes described above. This ICT cluster enabled cheap and fast information processing, and cheap and easy access over long distances enabling increased international electronic networking. As Lundvall and Foray (1996: 14) argued: ‘The ICT system gives the knowledge-based economy a new and different technological base which radically changes the conditions for the production and distribution of knowledge as well as its coupling to the production system’. A number of speciﬁc features and characteristics identiﬁable with ICTs have been critical for their key role in the emergence of the knowledge-based economy. First, ICTs are considered to be the ﬁrst example of a truly global technology (Soete and ter Weel 1999; Soete 2002), based on their ability to codify vast amounts of knowledge over distance and in real time and transfer it as information. This technical development brings about the possibility of real global access to all sectors, institutions, economic agents and individuals that are networked or have the knowledge and the capability to access the networks. However, in order for someone to be able to access and make effective use of internationally available codiﬁed knowledge, it is necessary to possess the requisite tacit knowledge and the relevant skills. Thus, a general trend is observed. The increasing codiﬁcation, transferability and accessibility of knowledge increase the value of the uncodiﬁed or tacit knowledge that is often essential in order to make use of the codiﬁed information. In sum, knowledge is now internationally more accessible, but the importance of tacit knowledge (human capital, learning processes, and organizational knowledge) remains crucial. Second, ICTs are universal in character, i.e. they are useful and applicable to all activities and sectors in the economy and they are becoming the basis of new, more ﬂexible, open and productive styles of working. ICTs are used as an enabling platform to provide access to the information and knowledge that are so important for the development of all economic sectors. Moreover, the penetration of the so-called infrastructural technologies, such as biotechnology and advanced materials as well as ICTs, throughout the economy during the late 1990s, has drastically altered the meaning of high technology. Rather than referring to the output of research and development (R&D) intensive industries, as the OECD high-medium-low tech classiﬁcation developed in the 1980s did, high technology tends now to refer to a style of working that is applicable to just about any business activity, thus inﬂuencing the structure and the working of all sectors of the economy. Reﬂecting this is the increased information and knowledge component of practically every industry and business – a trend that will continue. Finally, the disruptive nature of ICTs is leading to the transformation of existing sectoral systems through its effect on their basic elements including products, agents, learning processes, interaction among agents and selection processes. This is promoting the emergence of new hybrid activities at the interface of technology-based, knowledge-intensive services and technology-intensive manufacturing (Caloghirou 2004).

8 Y. Caloghirou, A. Constantelou and N.S. Vonortas 1.5.2 Knowledge as the foundation for the transformation of an advanced industrial economy to a knowledge-based economy At the level of the economy, the main changes accompanying the globalization of economic activity are the increasing signiﬁcance of knowledge products, and the knowledge content of physical products and services, which are associated with the blurring of the borders between manufacturing and services. Overall, the role of services is increasing, while that of manufacturing as a discretely identiﬁed activity is decreasing. The well-known concept of the ‘weightless economy’ reﬂects the tendency towards a national product with comparably lower speciﬁc weight in physical terms, as the value per physical weight increases. At the same time, intangible investment is increasing much more rapidly than physical investment. In terms of internationalization/globalization trends, there are important differences in investment and trade. In addition, the process of de-localization, or more precisely re-localization, of global manufacturing from developed industrialized areas to less developed and rapidly industrializing countries is accelerating signiﬁcantly. The most striking example of this is the developments being made by China to become a world-manufacturing platform. However, here again, what is new (Soete 2004) is the increased tradability and global movement of intangibles (ﬁnance, international alliances between ﬁrms including international collaboration in RTD, diffusion of information and knowledge, migration and mobility of human capital, tacit knowledge). These types of technological and economic development have produced new structural conﬁgurations of industries and economies, and stimulated new ways of thinking about economic problems, all of which have had a direct effect on the perception of intangibles and knowledge and in particular on their perceived value as the drivers of economic growth. Even those who are sceptical about the notion of the knowledgebased economy agree that ‘it remains important therefore for scholars and policy makers to have an adequate view of the relevance, structure and characteristics of knowledge across industries, as knowledge has been and continues to be a core foundation of economic process’ (Smith 2002: 12). At the same time, enthusiasts claim that now more than ever, knowledge should be viewed as a core or key input for development, if not ‘the one factor of production, sidelining both capital and labor’ (Drucker 1998: 15). In this respect, knowledge is seen as being the primary source of economic productivity, as the new basis for wealth (Thurow 1999), or as the main source of the wealth of nations (Rodrigues 2002). Moreover, knowledge is viewed as being the cornerstone of the transformation of the economy in the advanced industrial countries (Foray 2004). But, what then are the deﬁning features of the knowledge-based economy? According to Foray’s (2004: ix) deﬁnition knowledge-based economies are ‘essentially, economies in which the proportion of knowledge-intensive jobs is high, the economic weight of information sectors is a determining factor,

By way of an introduction 9 and the share of intangible capital is greater than that of tangible capital in the overall stock of real capital’. So, in fact, what we are actually observing in economic evolution and the way we perceive it, is a gradual change of weights, a hybrid of the old and new structures, activities and practices, where symbiotic relations are being established, but which at the same time are very antagonistic. Through changing weights, the new structures will evolve continuously and the transformation will be one of degree. But, as Mokyr (2000: 13) reminds us, the ﬁrst lesson of economic history is that ‘degree is everything’. 1.5.3 Emergence of the creative sector, and increased knowledge intensity Industrial economies have always had a big information-oriented or information processing and handling sector, as Machlup (1962) (perhaps applying an overextended deﬁnition and a very generous estimation), Beniger (1986) and others have shown. The new thing is the fact that it is now becoming very evident that, at the industry level, a huge new sector is emerging, which can be described as the creative sector of the economy (Florida and Tinagli 2004). This sector has been burgeoning over the last 20 years. Based on Florida and Tinagli’s estimates (2004: 11): Today, from 25 to more than 30 percent of workers in the advanced industrial nations work in the creative sector of the economy, engaged in science and engineering, research and development, technologybased industries, in the arts, music, culture, arithmetic and design industries, or in the knowledge-based professions of health care, ﬁnance and law. In the United States, the creative sector accounts for nearly half of all wage and salary incomes, as much as the manufacturing and service sectors combined. This compares with the dawn of the twentieth century when ‘fewer than 10 percent of working people worked in the creative sector of the economy, while fewer than 15 percent of the workforce did so in 1950’ (Florida and Tinagli 2004: 11). Furthermore, ‘knowledge is becoming the main raw material in many manufacturing industries’ (Rodrigues 2002: 5), while the knowledge intensity of products and services is increasing in almost all sectors and industries (i.e. tourism, entertainment, health, agriculture, transport). Finally, many of the traditional industrial sectors have already built a rich and distributed knowledge base, which in an ICT environment can be more easily codiﬁed, integrated and redistributed. The introduction of a strong ICT element and a related style of working in the operation of organizations (business and non-business) is leading to increased knowledge bases and productivity gains accruing from the use of this enhanced and more accessible knowledge.

10 Y. Caloghirou, A. Constantelou and N.S. Vonortas 1.5.4 Knowledge and the ﬁrm At the ﬁrm level, knowledge management is becoming a vital emerging business function, especially for global enterprises, accompanied by a new top management position – that of Chief Knowledge Ofﬁcer (CKO). As Professor Thurow of the Massachusetts Institute of Technology (2002) anticipated: ‘It is probably the CKO who will become in the future as important as the Chief Financial Ofﬁcer was for decades’. Furthermore, ICTs and the new methods of working associated with them, are enabling the creation of high-tech ﬁrms even within traditional industries. Several policy and business analysts have gone on to claim that there are no longer high and low-tech industries, there are now high and low-tech ﬁrms. 1.5.5 A new knowledge-based political agenda for the EU All the above mentioned developments, both as real trends in the economy and as intellectual tendencies regarding the perception of its operation and growth, have had an influence on economic policy making and thinking. At the European level, the Lisbon Strategy was launched in 2000 as a grand long-term political project of the EU. This strategy supports knowledge and innovation-led growth. It is focused on the transformation of the Union into the most dynamic and competitive knowledge-based region in the world, while maintaining, through renewal, the European social model. The creation of a European Research Area (ERA), another component of the Lisbon strategy, is a major European research and technology policy initiative launched to respond to the innovation and knowledge gap of the EU vis à vis the USA. 1.5.6 New knowledge-oriented sub-disciplines in economics and management In the new economic and intellectual environment, new methods of inquiry into the nature and causes of things and new ﬁelds of research and scientiﬁc study such as the economics of knowledge and knowledge management are emerging. The economics of knowledge is a rapidly developing branch of economics that has grown out of the strand of multidisciplinary research focused on innovation studies. This community of innovation researchers has been working productively for over 40 years as a very diversiﬁed research galaxy of groups from all over the world with varied theoretical as well as policy interests. This tradition was established in the 1960s and owes much to the outstanding work of Professor Chris Freeman and the ethos of SPRU – Science and Technology Policy Research at the University

By way of an introduction 11 of Sussex. Their work contributed largely to the creation of a new scientiﬁc paradigm, which played a crucial role in the study and research of technological innovation as a social and economic process. Innovation research, as Soete (2003) pointed out, has been characterized by a spirit of ‘letting a thousand ﬂowers ﬂourish’. To return to the historical development of economic analysis, it should be remembered that, despite the fact that classical economists, in particular Karl Marx and Joseph Schumpeter, were very aware of the importance of knowledge and technology for long-term economic growth, technology was treated in the neoclassical analytical framework as an exogenous factor, a ‘black box’ variable or a residual. It is well documented that the strand of thought and inquiry on the economics of technology and innovation, from both an evolutionary and a mainstream perspective, has broadened the standard neoclassical microeconomic understanding of the concept of knowledge. From the mainstream standard economic theory perspective, the seminal contributions of Arrow in 1962 and Romer in 1990 have had the greatest inﬂuence on the model of new growth theory. Ioannides (2005), in his very informative and interesting paper on ‘Information and knowledge in twentieth-century economics: from prices, to contracts, to organizations’, says that: The role of information and knowledge in economic activity has ﬁgured prominently in most of twentieth-century economics. However, economists have relied upon a very narrow understanding of the two concepts, as illustrated by the fact that they have usually employed them as perfect substitutes. In that context, both were thought of as describing unambiguously codiﬁable – and, thus, decodiﬁable – signals about states of the world, which agents could use merely as inputs in their maximizing calculations. As a consequence, this view of information and knowledge excluded from the analysis issues like learning processes, ambiguous signals, socially shared knowledge, uncodiﬁable (i.e. tacit) knowledge, and so on. (Ioannides 2005: 2) However, in his account based on a survey of the most inﬂuential (in his view) debates and theoretical strands, he concludes that: My central argument in this paper is that there seems to be a clear shift in economic theory in the last couple of decades, away from this dominant view and towards more sophisticated accounts of knowledge . . .Thus, we ﬁnd consistently more elaborate contributions to the problem of knowledge as theory moved away from prices towards contracts and then towards institutions and organizations. (Ioannides 2005: 2)

12 Y. Caloghirou, A. Constantelou and N.S. Vonortas But of course, the development of the economics of innovation and also the economics of knowledge is rooted in the actual development and transformation of the economy itself. In this respect, Dominic Foray, a prominent scholar in the economics of knowledge, offers an analogy: ‘Just as industrial economics was founded with the advent of industrialization in around 1820, so the economics of knowledge developed as knowledgebased economies gradually came into being’ (Foray 2004: ix). In a sense, knowledge management is also not new, as the activities of all organizations depended on the management of the knowledge they were creating or using. However, knowledge management is being revisited and a more disciplined area of research is emerging. Although some believe that it is merely another consultancy fad, others see in its rise a way of meeting the challenge associated with the development of the Information Society and the move towards a knowledge-based economy, of managing tacit knowledge and knowledge resources, including skills and competences (Al-Hawamdeh 2002). 1.5.7 Change in the relative weight of knowledge To sum up this section, knowledge, whether embodied in human beings, technology or machines, has always been central to modern economic growth and development. What has transformed the role of knowledge in the Information Age is the gradual changes in its relative weight. Compared to other factors, such as natural resources, physical capital and lowskilled labour, knowledge has assumed greater importance, both in quantitative and qualitative terms. Advanced economies are now much more dependent on the generation, distribution, use and combination of knowledge which are increasing output and employment in high-tech industry segments in manufacturing (computers, electronics, aerospace, etc.) and even more so in services (software, banking, business services, telecommunications, transport), as well as making improvements in the creation and delivery of other goods and services. Knowledge is becoming the engine of economic development and the main driver of the possible great transformation – in Polanyi’s (1944) meaning of the term – already in progress. In addition, the growth and accumulation of knowledge has created the necessity for a whole new business and management function. Recently it has emerged as the focal point and the label of a grand industrial umbrella strategy for the EU in the new globalized competitive environment.

1.6 Increasing interdependence of knowledge generation and knowledge diffusion One of the features typical of the knowledge-based economy is the fact that the diffusion, exchange and use of information and knowledge are

By way of an introduction 13 just as signiﬁcant as is its creation. The dichotomy between knowledge production and its diffusion/application is disappearing. Their relationship is becoming more and more one of interdependence. This feature of the knowledge-based economy is directing increased attention to knowledge distribution networks and national/regional or supranational systems of innovation. These are the agents and structures that support the advance and use of knowledge in the economy, and the links between them. In this respect, it is widely agreed that the exchange, diffusion, adaptation and application of information and knowledge have become a major prerequisite for a modern advanced economy to extract productivity gains from technological development, to exploit untapped growth potential and to build competitive advantage in a globalized, highly competitive economic and industrial environment. In this context, knowledge ﬂows, and knowledge mechanisms and channels are the vehicles for the dissemination and diffusion of knowledge. Their importance for the development of the economy and society is rapidly increasing. The diffusion of ICTs accelerates a long-term tendency towards increasing codiﬁcation of knowledge, which also increases its transferability. Nevertheless, not all types of knowledge can be captured and codiﬁed as information. A vast proportion of knowledge, gained primarily from practical experience, resides in people, and in organization’s routines and functions. This tacit knowledge cannot be easily codiﬁed or transmitted. ICTs facilitate extensive electronic networking among individuals and groups, which enables signiﬁcant sharing of knowledge and experience through informal channels, such as e-mail, and facilitates possible face-to-face contacts. Among the different types of knowledge in Lundvall and Johnson’s taxonomy, know-what and know-why are the most easy to codify and transfer as information or even sold in the market. Know-how and know-who, acquired largely through practical experience and social interaction respectively, can only be captured and codiﬁed in part. Their use requires an additional tacit element, be it skills, rules of thumb, short-cuts, technical, engineering and market knowledge embodied in people. Thus, relationships and interactions have always mattered in the functioning and growth of an economy, but it seems that in the knowledgebased economy they are crucial. The importance of formal and informal networking, of systemic and continuous interaction between producers and users of knowledge, of continuous learning, of huge ﬂows and increasing and stable relationships among industry, academia and government was underlined in the mid-1990s in an OECD report, which stated that: In addition to knowledge investments, knowledge distribution through formal and informal networks is essential to economic performance. Knowledge is increasingly being codiﬁed and transmitted through computer and communications networks in the

14 Y. Caloghirou, A. Constantelou and N.S. Vonortas emerging Information Society. Also required is tacit knowledge, including the skills to use and adapt codiﬁed knowledge, which underlines the importance of continuous learning by individuals and ﬁrms. In the knowledge-based economy, innovation is driven by the interaction of producers and users in the exchange of both codiﬁed and tacit knowledge; this interactive model has replaced the traditional linear model of innovation. The conﬁguration of national innovation systems, which consist of the ﬂows and relationships among industry, government and academia in the development of science and technology, is an important economic determinant. (OECD 1996: 2) But as Smith (2002: 5) claims: ‘Knowledge for many key activities is distributed among agents, institutions and knowledge ﬁelds, and the problem is to understand the embodied and disembodied knowledge ﬂows between them’.

1.7 The content of the book: knowledge communication through disembodied ﬂows of knowledge between ﬁrms, universities and other research and education institutes This book is about knowledge ﬂows and, in particular, innovation-related knowledge ﬂows in European industry. The chapters in the book deal with disembodied ﬂows of knowledge between ﬁrms, universities and other research and education institutes. The book expands on the KNOW project, a European Commission funded research project under the Targeted Socio-Economic Research (TSER) Programme, which was coordinated by Yannis Caloghirou and the Laboratory of Industrial and Energy Economics at the National Technical University of Athens (LIEE/NTUA). Some of the most active research groups in the ﬁeld of innovation studies contributed to this research effort, including MERIT (University of Maastricht), CESPRI (Bocconi University), BETA (University of Strasbourg), IKE (Aalborg University), ZEW and SIRN. The KNOW project allowed for extensive quantitative and qualitative empirical work on the extent, mechanisms and implications of innovation-related knowledge ﬂows in European industry, and provided for intellectual stimulus and conceptual development related to the generation and diffusion of knowledge among diverse European research groups dealing with the study of innovation and its respective policies. The book goes fairly deeply into the investigation of different aspects of knowledge ﬂows in an attempt to relate research ﬁndings with policy ideas for the development of a knowledge-based economy in the EU. It was compiled with the aim of contributing to the understanding of knowledge ﬂows and the mechanisms of knowledge sharing and transfer that facili-

By way of an introduction 15 tate these ﬂows. The focus is on knowledge generation and exchange that take place in an environment of increasing interaction, collaboration and networking among different agents (ﬁrms, universities and other research and education institutes) involved in the knowledge process. Different aspects of knowledge ﬂows are explored in a number of different cases, and a variety of data sources that are not readily available are used. The multifaceted nature of knowledge, which makes it difﬁcult to capture its ﬂow, requires a variety of approaches, methodologies and means of empirical investigation. Our point of departure for the analysis of knowledge ﬂows is the systems of innovation approach. This choice is based on the fact that innovation and knowledge processes are interactive processes taking place in a contextual framework, which promotes strong interdependencies. Social network analysis is also employed. A microeconomic approach to the analysis of knowledge ﬂows is followed using the individual organization – ﬁrm, university, other public research organizations (PROs) – as the basic unit of analysis. From this angle, a wide range of key issues such as the nature and dynamics of knowledge networks, the patterns of participation, the sources of knowledge, patenting activity, intellectual property rights instruments and the beneﬁts arising from collaborative research efforts are examined. Moreover, the analytical focus is on disembodied knowledge ﬂows between ﬁrms, universities and other education and research establishments. It was felt that, while extremely important, embodied knowledge ﬂows require data and analytical techniques that differ signiﬁcantly from those used to explain disembodied ﬂows, and that the literature on them is already extensive. Disembodied knowledge ﬂows have received much less systematic empirical analysis. The evidence-based part of the research reported in this book draws on a multitude of data sources. Three of these deserve particular mention as they have been speciﬁcally compiled by the research partners in the KNOW project. The ﬁrst is a longitudinal dataset of company participation in European Framework Programmes (FWP) on Research and Technological Development and Demonstration (RTD&D) based on Cordis data (EU-RJV database). The version used for this research covers projects initiated between 1983 and 1998 by the ﬁrst four FWPs and includes programmes whose main focus was the creation of new technological knowledge. All the better known programmes such as ESPRIT, BRITE-EURAM, JOULE, RACE, BIOMED, BIOTECH, ENV, TELEMATICS were included as well as many others (64 programmes in all). The database records research projects that involve at least one agent from the private sector (ﬁrm). The second large dataset (RJV-EPO database) recorded patents and patent citations of the companies participating in the FWPs that were examined. The RJV-EPO dataset was built by matching the entries in the

16 Y. Caloghirou, A. Constantelou and N.S. Vonortas EU-RJV database mentioned above with EPO patents. The merging was carried out at the level of the ﬁrm, resulting in a large set of companies for which RJV participation, European patent applications and patent citations can be identiﬁed. The third large dataset (KNOW survey) was based on a focused survey of a large number of European ﬁrms in selected industries and gathered information on the speciﬁc internal and external mechanisms and institutions that support innovation-related knowledge ﬂows and of the procedures that facilitate learning. The KNOW survey covered seven EU member states – Denmark, France, Germany, Greece, Italy, the Netherlands and the UK – and ﬁve business sectors – food and beverages, chemicals (excluding pharmaceuticals), communication equipment, telecommunication services and computer-related services. Questionnaires from 558 companies with 10–1,250 employees were collected. The survey results were enriched with insights gained from in-depth interviews with a subset of 71 of these ﬁrms giving more or less equal representation of the seven countries. It would be an overstatement to claim that the set of papers comprising the chapters in this book exhaust the subject of disembodied knowledge ﬂows affecting European industry. The subject is simply too big for any single publication to cover. We do believe, however, that the work reported in the subsequent chapters is important, it provides a fresh look at several key issues and, on the whole, offers interesting policy insights. What stands out is that innovation-related knowledge ﬂows essentially deﬁne the links that make up production and innovation systems, thus relating to all policies that affect these systems. Governments wishing to be effective in leading their citizens into the new, knowledge-intensive, ‘learning’ era must be cognizant of the fact that their industry, science, technology and innovation policies will inevitably impact the channels, direction, and intensity of knowledge ﬂows affecting industry. As a result, the concern over knowledge ﬂows cannot be dissociated from – in fact, it is intimately linked with – the broader policies that affect production and innovation systems, be they science, technology and innovation policies, intellectual property protection policies, competition policies, procurement policies or whatever else.

1.8 The organization of the book The book is divided into three parts and consists of 15 chapters. In the ﬁrst part, an agenda and a relevant conceptual framework are set for the study of disembodied ﬂows of knowledge between ﬁrms, universities and other education and research institutes. Based on this framework, knowledge ﬂows in European industry are empirically investigated. In the second part, special aspects of knowledge ﬂows are examined on the basis of accumulated evidence. Finally, the third part provides an overview of

By way of an introduction 17 policies in the EU region that affect innovation-related knowledge ﬂows and draws a number of conclusions and implications for European policy makers. Part I starts with Chapter 2, authored jointly by Caloghirou, Constantelou and Vonortas, in which the theoretical foundations of the conceptual framework for the analysis of knowledge ﬂows are discussed. The discussion evolves around the notion of System of Innovation (SI) and four basic concepts associated with it, namely innovation, institutions, networking and learning. Knowledge is viewed as the outcome of learning processes within the SI. In addition, access to internal and external sources of knowledge is argued to be of vital importance and, consequently, the notion of absorptive capacity at the micro level is stressed. This introduces into the discussion the concept of knowledge ﬂows as linking mechanisms between different sources of scientiﬁc and technological information and knowledge and their potential users. The transfer of knowledge takes place within a network of various organizations of different nature, size and role. Therefore, special emphasis is given to the modes of interaction among actors. In this context, the degree of power distribution within the innovation system is related to the availability and intensity of knowledge ﬂows. In Chapter 3, by Arundel and Constantelou, a review of the methodology and of recent empirical studies on the capture and measurement of knowledge ﬂows is presented. Two main groups of indicators capturing instances of knowledge ﬂows among actors are identiﬁed, namely the ‘traditional’ and the experimental. Patent and bibliometric or scientometric indicators, classic input indicators (such as R&D expenses), trade of technological machinery and equipment, technology balance of payments, Foreign Direct Investment (FDI) indicators, research and technological collaboration and labour mobility indicators are considered among the traditional. Experimental indicators are derived from the investigation of knowledge ﬂows through surveys and other forms of detailed qualitative information. A brief presentation of some of the most important databases created for the construction of technology-based alliances and joint ventures as indicators of knowledge ﬂows is made. Taking into consideration both the advantages and the shortcomings of each category of indicators, a review of what innovation surveys have told us so far about knowledge ﬂows is pursued. A number of missing elements and key areas not covered by existing surveys are identiﬁed, such as channels for obtaining information, the purpose for which information is acquired, absorptive capacity, research cooperation and innovative behaviour, sectoral differences, appropriability conditions and their impact on innovation. The new survey conducted within the KNOW programme is an attempt to respond to the need of covering the above mentioned missing elements. The results are presented in Chapter 5 of this book. Chapter 4 by Caloghirou, Constantelou and Vonortas, summarizes the

18 Y. Caloghirou, A. Constantelou and N.S. Vonortas key elements of the conceptual framework that guide the empirical investigation of knowledge ﬂows at the micro and meso levels and develops a more detailed speciﬁcation of how these key elements can be operationalized. The operationalization of the concepts aims at framing the survey questionnaire and the in-depth interviews in a way that areas related to knowledge ﬂows that are usually missing from other large scale surveys will be covered. In this respect, sources, channels, speciﬁc mechanisms and types of knowledge are investigated through a new survey instrument speciﬁcally designed to collect the relevant information. In Chapter 5, Caloghirou, Constantelou and Vonortas present a ﬁrst round of the ﬁndings of a combined exploratory survey and interview exercise, undertaken in the context of the KNOW research project. The KNOW survey was addressed to two categories of enterprises: small- and medium-sized enterprises (SMEs) (10–250 employees) and larger ﬁrms (251–1,250 employees) active in ﬁve sectors (food and beverages, chemicals excluding pharmaceuticals, communication equipment, telecommunication services and computer-related services) in seven EU member countries (Denmark, France, Germany, Greece, Italy, the Netherlands and the UK). A total of 558 respondent ﬁrms, considered to be innovators, were included in the survey results. The results of the survey were enriched by information obtained from 71 in-depth interviews. The placement of this chapter at the end of Part I of the book is intended to highlight the diversity involved in the issue of knowledge ﬂows as well as of the approaches being taken by European companies. The tabulation of results indicates signiﬁcant differences between sectors, countries and ﬁrms of different sizes in terms of R&D orientation of the companies, sources of innovation-related knowledge (internal, external), access channels (including inter-organizational collaboration and use of the Internet), use of intellectual property protection mechanisms and innovation drivers (customers, suppliers, internal). Part II starts with two chapters (6 and 7) dealing with intra-ﬁrm and intra-sector knowledge ﬂows, respectively. In Chapter 6, Mette Praest Knudsen investigates how organizational structures can facilitate or impede intra-ﬁrm knowledge ﬂows. A number of facilitators and impediments to knowledge sharing are presented along with the three organizational structures examined: functional, disintegrated divisional and collaborative. Drawing evidence from the telecommunication sector, one case ﬁrm is analysed for each of the three organizational forms on the basis of the preceding theoretical principles. The analysis shows that the functional form is most appropriate for less knowledge-intensive activities and that incentive systems should be built in order to motivate the development team to take the innovation to production. In the case of a divisional form, a danger emerges that the knowledge base will evolve into a lock-in situation. A clear team-based organization is needed combined with an explicit focus on external sources of knowledge in order to avoid

By way of an introduction 19 knowledge lock-in. Finally, a pure collaborative relationship seems to be very competitive, but also vulnerable to internal changes. In Chapter 7, Pedersen, Dahl and Dalum examine the role and character of inter-sector knowledge ﬂows, focusing on the Danish ICT sector in an attempt to ﬁnd possible explanations for the apparent mismatch between the advanced demand and the rather weak supply in that country. They point out to a rather fundamental handicap for Denmark to achieve ambitious strategic goals in the ICT ﬁeld, known as the ‘lack-ofEricsson-Nokia’ effect. In other words, they strongly support the view that the lack of a large ‘domestic’ multinational ICT company is a serious weakness for a small country. In some cases, they argue, this lack can be compensated for by the emergence of small and medium-sized R&D oriented ICT ﬁrms, but too few have emerged in Denmark during the 1990s despite rather favourable economic conditions. Denmark appears to be well positioned to exploit the business opportunities from the point of view of demand, but not to the same extent as other Nordic countries such as Sweden and Finland. In Chapter 8, Arundel and Bordoy investigate the relationship between collaboration and innovation outputs. They use results from the KNOW survey, which, in contrast to other surveys, provides information on the effect of the ﬁrm’s collaboration on its innovative output. Two research questions are explored. First, the identiﬁcation of factors inﬂuencing the development of innovations through collaboration. Second, the link between the percentage of innovations that are developed through collaboration and the innovation sales share. Although the results of this particular study tend to be critical of the theory of ‘collaborative space’ in the joint development of innovative products, the authors are careful to point out the limitations. To conclude that external sources – among them collaboration – are not important for successful innovation would be hasty. For instance, ﬁrms may engage in collaboration in order to, among other things, gain access to other types of knowledge or learning capabilities whose effect may not show up directly as a change in the innovation sales share. In Chapter 9, Fontana, Geuna and Matt analyse the contribution of PROs – deﬁned as universities and other public research organizations – to the innovative processes of ﬁrms and especially SMEs. The authors use the data from the KNOW survey as the basis for their analysis. The descriptive analysis shows that for most ﬁrms participating in the survey, PROs were among the less important sources of information for developing innovations during the three years prior to the survey. Nevertheless, about half of the ﬁrms in the sample had developed collaborative research projects with PROs and a number of them had been engaged in periods of intensive interaction with PROs in order to meet speciﬁc needs. In addition, signiﬁcant sectoral differences were identiﬁed. Furthermore, the authors evaluated through an econometric model the effect of ﬁrmspeciﬁc, sector-speciﬁc and country-speciﬁc factors on the propensity for

20 Y. Caloghirou, A. Constantelou and N.S. Vonortas and the extent of collaboration between PROs and SMEs. Three major ﬁndings are worth noticing. First, larger ﬁrms show a much higher probability of R&D collaboration with PROs. However, the authors point out that ﬁrm size should not be considered in absolute terms, but rather in relative terms, i.e. in R&D employment, as a measure of R&D intensity. Second, the openness of the ﬁrm to the external environment – and its willingness to interact with it – has a signiﬁcant effect on both the propensity and the extent of PRO-ﬁrm collaboration. Third, a high degree of heterogeneity characterizes the relationship between ﬁrms and PROs. Therefore, public policies encouraging university–industry collaboration should take into account the way the various actors will respond depending on their speciﬁc characteristics, otherwise unintended inter-sectoral differences will arise. In Chapter 10, Cusmano undertakes an empirical investigation of RJV participation and patenting, by evaluating, empirically, the relationship between EU-funded RJVs and patenting activity in two high-tech research areas, namely: ICT, and Medical and Biotechnology (MB). A large data set of European ﬁrms engaged in collaborative research activities in these ﬁelds is used. The patenting performance of RJV participants is examined in comparison with the performance of other European ﬁrms. In the ICT area, the remarkable patenting performance of ﬁrms that are part of RJV compared to other ﬁrms, is mostly explained by self selection of innovative actors. This is in line with the main aim of ICT programmes, that is, attracting the industry big players and providing a common technological basis for IT applications. On the other hand, the analysis provides evidence of a positive effect of RJV afﬁliation in the emerging MB ﬁeld, where European consortia appear to have attracted ﬁrms with a high potential for innovation. The ﬁndings from this evaluation suggest that the analysis of industry dynamics should be central in the design of policy and evaluation of policy targets and achievements. In Chapter 11, Dumont and Tsakanikas compute knowledge spillovers that stem from R&D cooperation in the context of the EU FWPs (1984–98), examined both at the sectoral and the country levels. They attempt to evaluate the evolution of cooperation patterns in response to policy or technology shifts, thereby assessing the actual impact of a largescale network promoting policy such as the EU programmes. Their analysis points at the dominant role of various sectors that form the ICT area and their central position in the knowledge channels identiﬁed. Some traditionally ‘low-tech’ industries also have strong spillovers, but apparently most of their linkages are intra-sector, indicating their more selfreliant behaviour. Firms from smaller countries seem to rely more on foreign partners than ﬁrms from larger countries, reinforcing the importance of networking for such countries, to make up for the insufﬁcient scale and/or the lack of appropriate domestic partners. The proposed method of measuring knowledge spillovers can be useful not only in the

By way of an introduction 21 empirical analysis of spillovers, but also in the assessment of a public policy for which, at least implicitly, the creation of spillovers is an important rationale. Cooperation is not just a channel for transferring codiﬁed information, but more importantly a mechanism for sharing complementary tacit knowledge. Hence, it may be more effective than market transactions and more ﬂexible than full internalization through mergers and acquisitions. In Chapter 12, Breschi and Cusmano contribute to the debate about targets and effectiveness of networking policies at the EU level by presenting an analysis of the R&D network that emerged over FWPs. Social network analysis is employed to describe the structural properties and dynamics of the emerging network, which appears to be dense and pervasive, branching around a large ‘oligarchic core’, whose centrality and connectivity have strengthened over programmes. The chapter discusses the degree to which this network structure may respond to EU broad policy objectives and its implications for recent programmes aimed at shaping a European Research Area (ERA). Attention is placed on the recent focus by European institutions on networking centres of excellence. Since future initiatives are to build on the existing fabric, the authors argue that understanding how networks formed and evolved following previous stimuli is of great relevance for implementing and assessing the impact of the newly deﬁned network approach. It is expected that the emerging conﬁguration will exert signiﬁcant inﬂuence on the EU policy aiming at networking the centres of excellence as an instrument for creating ERA. In Chapter 13 Verspagen attempts to apply the ‘small worlds’ model in the case of European collaborative research and knowledge patent citations networks. Recent literature has related the features of small worlds to efﬁcient knowledge communications among network members. The empirical investigation shows that European research networks do indeed show small world properties. The network consisting of all FWP projects comes closest to the theoretical model of small worlds. Two other networks, the EUREKA projects network and the non-policy related network of patent citations, also resemble small worlds, but less closely than the FWP network. Overall, therefore, the author concludes that the European research networks analysed are relatively efﬁcient means of knowledge transfer. In Chapter 14, Arundel and Vonortas review the European policy instruments affecting the production and dissemination of innovationrelated knowledge and the ability of ﬁrms to beneﬁt from knowledge ﬂows. These concern (a) the absorption and use of externally developed knowledge, (b) the commercialization of the results of publicly funded R&D and (c) the ﬁnancing of innovation by private ﬁrms. The authors stress that many of these policies remain ﬁrmly entrenched at the local, regional or national level. It is argued that the increasing sophistication of science and technology (S&T) policy in Europe has led to concerted

22 Y. Caloghirou, A. Constantelou and N.S. Vonortas efforts to ensure that new ideas ﬁnd their way to ﬁrms that can apply them to their new products, processes and services. One consequence of this is that all EU member states emphasize the need to promote knowledge ﬂows between ﬁrms, and between ﬁrms and PROs. Many EU member states currently focus on encouraging universities and (PROs) to concentrate as much as possible in areas of interest to private ﬁrms and to improve the transfer of knowledge and expertise to ﬁrms, particularly smaller ones. Finally, in Chapter 15 Caloghirou, Constantelou and Vonortas present the main conclusions and policy implications for the EU. Not surprisingly, they are by and large reminiscent of the ongoing discussions on innovation systems and the knowledge-based economy. Perhaps the most basic and most important implication for policy decision makers is the overall conclusion that, since they deﬁne the links that make up production and innovation systems, innovation-related knowledge ﬂows are affected by every policy aimed at calibrating these systems. That is to say, industry, science, technology and innovation policies will impact the channels, direction and intensity of knowledge ﬂows affecting industry as will other general policies such as general investment policy, intellectual property protection policy, competition policy, procurement policy, regional policies and so forth. It is imperative that future European policy design is actually inﬂuenced by a ‘systems’ approach and builds effective links among all of these elements.

Note 1 Cited in Freeman and Louca, (2001: 328). According to the authors, the poet T.S. Eliot in his ‘Chorus’ for The Rock wrote the second and third questions. They argue that the ﬁrst question is a hypothetical one, in case Eliot had lived to see the Information Society.

References Al-Hawamdeh, S. (2002) ‘Knowledge management: re-thinking information management and facing the challenge of managing tacit knowledge’, Information Research – An International Electronic Journal, 8(1): 143. Arrow, K. (1962) ‘Economic welfare and the allocation of resources for invention’, in Nelson, R. (ed.) The Rate and Direction of Inventive Activity: Economic and Social Factors, Princeton: Princeton University Press. Beniger, J. (1986) The Control Revolution: Technological and Economic Origins of the Information Society, Cambridge: Harvard University Press. Cairncross, F. (2001) The Death of the Distance: How the Communication Revolution is Changing our Lives, London: Texere Publishing. Caloghirou, Y. (2004) TENIA (The Emergence of New Industrial Activities: Fusing services and manufacture), ﬁnal report, Athens. Drucker, P. (1998) ‘From capitalism to knowledge society’ in D. Neef (ed.) The Knowledge Economy, Woburn, MA: Butterworth, pp. 15–34.

By way of an introduction 23 Florida, R. and Tinagli, I. (2004) Europe in the Creative Age, DEMOS. Available at: www.demos.co.uk/creativeeurope (last accessed 19 April 2005). Foray, D. (2004) The Economics of Knowledge, Cambridge, MA: The MIT Press. Freeman, C. and Louca, F. (2001) As Time Goes By. From the Industrial Revolutions to the Information Revolution, Oxford: Oxford University Press. Freeman, C. and Perez, C. (1984) ‘Long waves and new technology’, Nordisk Tidsskrift for Politisk Economi, 17: 5–14. Hunter, R. (2002) World Without Secrets, New York: John Wiley & Sons. Ioannides, S. (2005) ‘Information and knowledge in twentieth-century economics: from prices to contracts, to organizations’, in G. Kouzelis, M. Pournari, M. Stoepler and V. Tselfas (eds) Knowledge in the New Technologies, Frankfurt: Peter Lang. Lundvall, B.-Å. (1996) ‘The social dimension of the learning economy’, DRUID Working Paper No. 96-1. Lundvall, B.-Å. (2004) ‘Why the new economy is a learning economy’, DRUID Working Paper No. 04-01. Lundvall, B.-Å. and Foray, D. (1996) ‘The knowledge-based economy: from the economics of knowledge to the learning economy’, in OECD (ed.) Employment and Growth in the Knowledge-Based Economy, Paris: OECD. Lundvall, B.-Å. and Johnson, B. (1994) ‘The learning economy’, Journal of Industry Studies, 1(2): 23–42. Machlup, F. (1962) The Production and Distribution of Knowledge in the United States, Princeton, NJ: Princeton University Press. Mokyr, J. (2000) ‘Economic history and the new economy’, presentation at the National Association for Business Economics 42nd Annual Meeting, 10–13 September, Chicago. Mokyr, J. (2002) The Gifts of Athena: Historical Origins of the Knowledge Economy, Princeton, NJ and Oxford: Princeton University Press. OECD (1996) The Knowledge-Based Economy, Paris: OECD. Pasinetti, L.L. (1981) Structural Change and Economic Growth, Cambridge: Cambridge University Press. Perez, C. (1983) ‘Structural change and the assimilation of new technologies in the economic and social system’, Futures, 15: 357–375. Polanyi, K. (1944) The Great Transformation: The Political and Economic Origins of Our Time (reprinted 1980), Boston, MA: Bacon Press. Rodrigues, M.J. (ed.) (2002) The New Knowledge Economy in Europe, Cheltenham: Edward Elgar. Romer, P. (1990) ‘Endogenous Technological Change’, Journal of Political Economy, 98(5): 71–102. Smith, K. (2002) ‘What is the “knowledge economy”? Knowledge intensity and distributed knowledge bases’, UNU/INTECH Discussion Paper Series (2002–6). Soete, L. (2002) ‘The challenges and the potential of the knowledge-based economy in a globalised world’, in Rodrigues, M.J. (ed.) (2002) The New Knowledge Economy in Europe, Cheltenham: Edward Elgar, pp. 28–53. Soete, L. (2003) ‘Innovation: the European research challenge’, presented at Roskilde workshop, 9th May. Soete, L. (2004) ‘The knowledge economy: the policy challenges’, paper presented at the World Bank Knowledge Economic Forum III on Improving Competitiveness through a Knowledge-based Economy, Budapest, 23–26 March.

24 Y. Caloghirou, A. Constantelou and N.S. Vonortas Soete, L. and ter Weel, B. (1999) ‘Schumpeter and the knowledge-based economy: on technology and competition policy’, MERIT working papers series, Maastricht. Thurow, L. (1999) Building Wealth, New York: HarperCollins Publishers. Thurow, L. (2002) Keynote speech at the Annual Conference of the Greek CEOs Society, Athens.

Part I

Setting the agenda for the study of knowledge ﬂows

2

Theoretical foundations and key concepts Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas

2.1 Introduction Economists and scholars of technical change are increasingly underlining the role of knowledge as being a determining factor in how successfully ﬁrms and national economies perform. This focus has led policy makers in Europe and the OECD countries to shift towards policies aiming at strengthening the ability of actors in a national context – particularly ﬁrms, universities and research establishments – to make efﬁcient use of new knowledge. The First Action Plan for Innovation in Europe (European Commission [EC] 1996: 6) explicitly state that the efﬁcient use of new knowledge depends on three factors: ‘the ability to produce knowledge, the mechanisms for disseminating it as widely as possible, and the aptitude of the individuals, companies and organizations concerned to absorb and use it’. Of particular interest in this book are the description and analysis of the mechanisms for knowledge sharing and transfer, which enable the ﬂow of knowledge among different actors. Several recent studies have sought to identify and empirically assess the impact of knowledge ﬂows in the economy. Outstanding in this respect, is the work published by the OECD under the National Systems of Innovation (NSI) initiative, which set out to establish a typology of knowledge ﬂows and to measure their availability and impact in different national settings. The ultimate goal of this effort was to establish a link between innovation systems and economic performance (OECD 1997). This chapter brings together key theoretical tools and concepts in order to build up a conceptual framework to underpin the investigation of the main issues surrounding knowledge ﬂows. The chapter starts with a discussion of the notion of a system of innovation (SI) and its basic concepts (innovation, institutions, networking and learning). The concepts of knowledge and knowledge ﬂows are deﬁned and discussed in relation to the ability of an innovation system to enhance its performance through its ‘distribution power’, that is, the system’s ability to support and improve the efﬁcient functioning of procedures for distributing and utilizing knowledge (David and Foray 1995). The chapter concludes with a discussion on

28 Y. Caloghirou, A. Constantelou and N.S. Vonortas the institutional- and ﬁrm-level factors that are considered to be critical for assessing the distribution power of innovation systems.

2.2 The concept and elements of a system of innovation Recent economic theorizing seeks to expand on our understanding of the relationships among economic actors from simple linear models of deterministic outcomes and rational behaviours to more complex models of social, institutional and political interactions. In particular, scholars of the economics of technical change identify diversity in the ways economic agents interact as both a cause and an effect of the enlargement of their knowledge base and increased competencies leading to the creation of novelty. Diversity is thus seen as a valuable asset that needs to be generated and preserved intentionally. The reasoning behind this is that qualitative differences in techniques, processes and organizational forms provide opportunities for economic agents to engage in learning processes. These learning processes strengthen the knowledge base of market participants and increase their potential for creativity and adaptability in times of uncertainty and ﬂux (Metcalfe 1993; Saviotti 1997; Cohendet and Llerena 1997). The mechanisms that generate and preserve diversity in economic processes cannot be considered in isolation from the national and international economic, political and institutional conditions. Therefore, the idiosyncratic ways in which learning processes unfold in distinct economic and social contexts need to be examined. The systems of innovation approach is an analytical concept that situates variations in economic, technological and institutional development within the wider national and supra-national context. A starting point for this approach is a recognition that historical, institutional and cultural factors affect the behaviour of economic agents by inﬂuencing their ability to learn and produce new knowledge or re-combine existing knowledge in new ways (Nelson and Rosenberg 1993). The idea of a ‘system’ is not new in the economics literature and several authors have employed the term ‘system(s) of innovation’ (Edquist 1997). However, the empirical and/or analytical focus is not the same in these different approaches.1 Recent contributions to the systems approach put emphasis on the national level and place innovation capabilities in technologies and organizational forms of production at the centre of analysis. For example, Nelson and Rosenberg (1993: 4) deﬁne a system as ‘a set of institutions whose interactions determine the innovative performance of national ﬁrms’ (emphasis added). Christopher Freeman introduced the concept of a national system of innovation in his study of the economic and technological development of the Japanese economy. He deﬁned it as ‘the network of institutions in the public and private sectors whose activities and interactions initiate, import, modify and diffuse new technologies’ (Freeman 1987: 1).

Theoretical foundations and key concepts 29 Along the same lines, Lundvall (1993: 277) argues that ‘what makes national systems of innovation important is that the organised markets of the real world may be organised differently in different national systems and that the behaviour of agents, rooted in different systems, may be governed by different rules and norms’. According to Lundvall (1992: 12), a NSI includes ‘all parts and aspects of the economic structure and the institutional set-up affecting learning as well as searching and exploring’. Within this deﬁnition, history, language and cultural elements are important determinants of national idiosyncrasies in ﬁrm behaviour, inter-ﬁrm linkages, industrial structures and public policy. Other authors use the terms ‘system of innovation’ and ‘technological system’ to refer to the set of distinct economic, industrial and institutional factors that contribute to the development and diffusion of new technologies and provide a framework within which governments form and implement policies to inﬂuence the innovation process (Metcalfe 1993; Carlsson and Stankiewicz 1991). Innovation lies at the heart of SI approach. The complex nature of the innovation process is probably one of the major reasons why there is no established and broadly accepted deﬁnition of a SI. Some authors have limited their notion of innovation to technical innovation: Nelson and Rosenberg endorse a view of innovation that is restricted mainly to technological and organizational advances within ﬁrms. In contrast, Lundvall’s definition of innovation is of interactive processes of ‘learning, searching, and exploring, which result in new products, new techniques, new forms of organization, and new markets’ (1992: 8). Such processes relate to research and development (R&D), but they can also be found in other economic activities such as marketing and procurement (Edquist and Johnson 1997). Along these lines, McKelvey (1991: 118) summarizes three different meanings of the term ‘innovation’. According to McKelvey, the term can denote (a) a speciﬁc stage in the process of technological change; (b) all kinds of organizational, social and institutional novelties; and (c) the process of creating, diffusing or using these various changes. From the above deﬁnitions it becomes clear that the notion of innovation encompasses much more than R&D departments. Thus, the need for direct and indirect measures of innovation activities is growing, which necessitates having a clearer view of what constitutes an innovation. A ﬁrst step in this direction was made in the Oslo manuals published by the OECD (1992/1997), which proposed guidelines for collecting and interpreting technological innovation data. These manuals distinguish between three types of innovation as follows: •

A technological new product is a product whose technological characteristics or intended uses differ signiﬁcantly from those of previously produced products. Such innovations can involve radically new technologies, can be based on combining existing technologies in new cases, or can be derived from the use of new knowledge.

30 Y. Caloghirou, A. Constantelou and N.S. Vonortas •

•

A technologically improved product is an existing product whose performance has been signiﬁcantly enhanced or upgraded. A simple product may be improved (in terms of better performance or lower costs) through use of higher-performance components or materials, or a complex product that consists of a number of integrated technical sub-systems may be improved by partial changes to one of the subsystems. A technological process innovation is the adoption of technologically new or signiﬁcantly improved production or delivery methods. These methods may involve changes in equipment, or production organization, or a combination of these changes, and may be derived from the uses of new knowledge. The methods may be intended to produce or deliver technologically new or improved products, which cannot be produced or delivered using conventional production methods, or essentially to increase the production or delivery efﬁciency of existing products. (OECD 1992/1997: 48–49)2

Also, the Oslo Manual (second edition) distinguishes between a ‘worldwide technological product or process innovation (TPP)’ and a ‘ﬁrm-only TPP innovation’. The former describes the very ﬁrst time a new or improved product or process is implemented, while the second occurs when a ﬁrm implements a new or improved product or process that is technologically novel for the unit concerned, but has already been adopted by other ﬁrms and industries (ibid, p. 52). Nevertheless, concerns have been raised regarding the extent to which deﬁnitions of concepts as amorphous as technologically new products and processes can be ubiquitously developed (Hansen 1999). There are two bases for these concerns. First, experience so far suggests that not only is it difﬁcult for ﬁrms to decide what is ‘new’, but also it is difﬁcult for them to make a distinction between ‘new’ and ‘improved’ products or processes (Hansen 1999). Second, the increasing importance in the economy of the service sector and the different notions of what constitutes a ‘new product’ or a ‘new process’ in this sector call for modiﬁcations to the deﬁnitions of innovation originally framed to apply to the manufacturing sector. The second Community Innovation Survey (CIS-II)3 took this point seriously and incorporated a deﬁnition of innovation in services: A new or improved service is considered to be a technological innovation when its characteristics and ways of use are either completely new or signiﬁcantly improved qualitatively or in terms of performance and the technologies used. The adoption of a production or delivery method, which is characterized by signiﬁcantly improved performance, is also a technological innovation. Such adoption may involve a change in equipment, or organization of production, or both, and may be intended to produce or deliver new or signiﬁcantly improved

Theoretical foundations and key concepts 31 services, which cannot be produced or delivered using existing production methods, or to improve the production of delivery efﬁciency of existing services. The introduction of a new or signiﬁcantly improved service, or production or delivery method can require the use of radically new technologies or a new combination of existing technologies or new knowledge. The technologies involved are often embedded in new or improved machinery, equipment or software. The new knowledge involved could be the result of research, acquisition or utilization of speciﬁc skills and competencies. (CIS-II Annex II.2, Core Questionnaire for the Service Sector) It should be noted that in the above deﬁnition, organizational and managerial changes in service sector ﬁrms are not considered to be technological innovations.

2.3 Learning, learning capability and the learning economy Learning is a central issue in economic development in the SI approach. Lundvall deﬁnes learning as ‘the generation, transfer, and distribution of knowledge’ (Lundvall 1999). This broad deﬁnition requires two qualiﬁcations. First, learning is interactive and interactive learning is central to the process of innovation (Edquist 1997: 5). Although almost all learning is interactive there are different kinds of learning, which involve different amounts of social interaction (Johnson 1992). For example, there is individual learning from isolated imprinting of immediate experiences of the memory, rote learning (learning by repetition, not necessarily understanding), learning via feedback and, lastly, systematic and organized searching for new knowledge in universities and R&D departments. The latter two types of learning require intense and complex forms of interaction and are assumed to be increasingly important. Second, learning is a socially embedded process. As Lundvall puts it, ‘without a minimum of social cohesion the capability to learn to master new technologies and new and more ﬂexible forms of organization will be weak’ (Lundvall 1999: 20). In modern economies, technical and organizational change has become increasingly endogenous. Learning processes have been institutionalized and learning by doing (and learning by using) have become increasingly important, leading to the emergence of the ‘learning economy’. In the learning economy ﬁrms choose organizational modes that enhance their learning capability. Networking with other ﬁrms, patterns of horizontal communication and mobility of people between different posts and departments are becoming more and more important. Firms in the learning economy are to a large extent ‘learning organisations’ (Lundvall and Johnson 1994). Learning organizations facilitate the learning activities of their members in the search for a process of continuous transformation. From a policy point of view, governments have an

32 Y. Caloghirou, A. Constantelou and N.S. Vonortas important role to play in providing the means for learning, the incentives to learn, enhancing the capabilities of public and private organizations to learn, and facilitating access to the relevant knowledge bases. However, how this is best achieved through government action remains a matter for consideration.

2.4 A discussion on knowledge and knowledge ﬂows The importance of learning in innovation points to the signiﬁcance of knowledge and knowledge ﬂows. In traditional neo-classical theory, knowledge is seen as a uniformly available public good that can be transferred and acquired at little cost. The normative approach to the creation of technological knowledge suggests a linear process in which ﬁrms endogenously seek out and apply knowledge inputs in the form of R&D to generate innovative output. Yet, this view has been criticized on both theoretical and empirical grounds. Modern economic theories are challenging the traditional rationalist view which regards knowledge as the outcome of a linear process of information accumulation leading to the building of a knowledge stock. Instead, a more complex, interactive process of knowledge formation is being proposed based on continuous, interactive processes of data exchange. At this point, a distinction between information and knowledge is necessary. David and Foray (1995) argue that information refers to ‘knowledge that has been reduced and converted into messages that can be easily communicated among decision agents’, while knowledge is understood as the ‘conceptual and factual contexts that enable agents to interpret and give meaning to “information”’ (David and Foray 1995: 26). Thus, whereas information is fragmented and can pass on from one individual to another, knowledge is structured, coherent and can be acquired even without new information being received as it depends very much on the cognitive abilities of the actors holding it (Ancori et al. 2000). This approach also attributes idiosyncratic and cumulative properties to knowledge formation since it views knowledge as the result of continuous and interactive context-speciﬁc learning processes experienced by individual agents who, in turn, put their ‘personal touch’ on these processes by bringing to them their own prior knowledge, competencies and experience. Several typologies of knowledge have been proposed in the literature. From an epistemological point of view two distinct dimensions of knowledge have been identiﬁed: codiﬁed or explicit knowledge and tacit knowledge.4 Codiﬁed or explicit knowledge refers to knowledge that has been reduced to a codiﬁed and transmittable form, while tacit knowledge refers to knowledge that resides in the subconscious of the human mind, and is acquired through experience, imitation and observation. Thus, it has a personal element, which makes it hard to formalize and communicate from one agent to another (David and Foray 1995; Nonaka and Takeuchi

Theoretical foundations and key concepts 33 1995). Tacit knowledge is highly individualistic and depends upon the cognitive capabilities of agents. In other words, even though several agents may receive the same message, the interpretation of the message depends on the very speciﬁc knowledge held and shaped personally by each individual. Examples of tacit knowledge are the know-how and skills possessed by employees. Another example is the tacit knowledge shared by a corporate environment, where intuitive and common perceptions and beliefs facilitate communication in the workplace. This example leads us to another typology of knowledge found in the literature: individual versus social or collective knowledge. Although individuals possess and generate knowledge, the way they learn is shaped by, and is the result of, a social process whereby members of a particular social group contribute with pieces of knowledge they collectively hold (Ancori et al. 2000).5 ‘Context’ – in the sense of the social, institutional, organizational and economic setting – matters in the process of knowledge creation. Organizations play a critical role in articulating and amplifying knowledge. The way knowledge is produced, stored and exchanged within an organization strongly inﬂuences the process of organizational learning and, eventually, the nature of the organization itself (Amin and Cohendet 2004). The interplay among the different types of knowledge (tacit, codiﬁed, individual, collective) is at the heart of the knowledge creation process. Nonaka and Takeuchi (1995) developed the idea that knowledge emerges out of a continuous dialogue between tacit and explicit knowledge held at different ‘ontological’ levels of social interaction (individual, group, organizational and inter-organizational). In their interpretation of knowledge formation, the interaction between tacit and explicit knowledge is ‘organizationally’ managed. It extends beyond the individual level and implies increasing social interaction at the collective level among individuals from different communities of practice of various origins and sizes, and internal or external to an organization. This view of knowledge creation has important implications for the learning processes at the ﬁrm level. In particular, given the general consensus that the production of scientiﬁc and technological knowledge is a key determinant of the innovative capabilities of ﬁrms, increasing emphasis is being given to the study of the various mechanisms at the level of the ﬁrm that contribute to the formation of its knowledge base. For example, Antonelli (1999) views technological knowledge as the outcome of a complex process that relies upon learning, the socialization of experience, the re-combination of available information, and formal R&D activities. He also distinguishes four different forms of knowledge along two dimensions: tacit or codiﬁed and internal or external to the ﬁrm. Internal tacit knowledge is generated through individual learning-bydoing and learning-by-using at the ﬁrm level, while informal links and social relations among members of scientiﬁc communities may result in the generation of external tacit knowledge that passes on from one agent

34 Y. Caloghirou, A. Constantelou and N.S. Vonortas to another in the form of ideas, skills and techniques. Similarly, internal codiﬁed knowledge is normally developed through the engagement of researchers in formal R&D activities. External codiﬁed knowledge, on the other hand, involves the re-combination of knowledge that already exists in embodied (that is, in artefacts and equipment or software) or disembodied forms, in sources external to the ﬁrm (universities, research institutes or other industrial organizations) or is generated as a result of formal cooperation among ﬁrms (Antonelli 1999). It is important to note that all activities surrounding the generation and/or acquisition of new knowledge may be enhanced or restricted by speciﬁc elements of the (national) institutional environment, such as the intellectual property rights (IPR) regime, the culture and norms that govern informal relations among scientists involved in research activities, or even the policies of journals in relation to publishing scientiﬁc papers (David and Foray 1995). Another property of knowledge that is particularly important in the study of ﬁrm behaviour and industry evolution concerns its domains (Malerba and Orsenigo 2000). Knowledge may differ across sectors in terms of technological domains, that is, in terms of the speciﬁc scientiﬁc and technological principles at the base of innovative activities in a sector (Malerba 1999). Furthermore, it is often the case that the knowledge underlying innovative activity in a particular industry changes over time, thus calling for new types of competencies to be developed for innovation. Another domain of knowledge at the level of the ﬁrm concerns the speciﬁc features of demand found in the market. Changes in demand structure and user preferences call for new knowledge to be acquired in related technologies, thus initiating a new cycle of learning processes in the ﬁrm. The view of knowledge as the outcome of learning processes and the emphasis on access to internal or external sources of knowledge point to the existence of knowledge ﬂows. Knowledge ﬂows link different sources of scientiﬁc and technological information and its potential users. They include technology transfer and the ﬂow of know-how and information, including both accidental spillovers and intentional transfers. There are many alternative routes for knowledge ﬂows. Knowledge ﬂows require a channel through which knowledge is communicated, for example, an established collaborative link between two scientists from different ﬁrms, informal discussions among employees or formal presentations of work in progress. Knowledge ﬂows within large ﬁrms that are active in several industrial sectors can also play a crucial role in the diffusion of knowledge across disciplines (Arundel et al. 1998). Following the emphasis on learning processes within the systems of innovation (SI) approach, David and Foray (1995) argue that what characterizes and determines the performance of different systems of learning in science and technology is not so much their ability to produce new know-

Theoretical foundations and key concepts 35 ledge as their ability to disseminate it effectively and allow it to become economically valuable to third parties. Thus, the intensity and variability of knowledge ﬂows among the constituents of a national system are critical determinants of its ‘distribution power’. It has therefore been suggested that policy makers should shift their interest from steady structures and absolute measures of innovative activities (such as R&D expenditure and patents) to the different types of interactions among actors within and beyond the boundaries of a national system. Four such types of interactions have been identiﬁed in the literature as embodying knowledge ﬂows (Smith 1994; OECD 1996), which are discussed below.

2.5 Modes of interaction embodying knowledge ﬂows The ﬁrst type of interaction refers to ‘inter-industry interactions among individuals/ﬁrms’. These occur through several channels such as formal and informal collaboration agreements among ﬁrms, the conduct of contract work, interactions among members of the different scientiﬁc communities in ﬁrms and user-producer interactions, as well as interactions with external-to-the-ﬁrm sources of information (ﬁrms providing training services, external consultants, etc.), all of which contribute to the accumulation of competitive intelligence. The empirical evidence suggests that these interactions involve knowledge ﬂows that may not always have a signiﬁcant impact on innovative outcomes (Arundel et al. 1998). In the case of collaborative R&D for example, the evidence regarding its impact on innovative performance is contradictory. Studies of the telecommunications and ofﬁce equipment sectors have found that ﬁrms that participate in cooperative R&D agreements are less innovative than those that do not (Arundel et al. 1996; Malerba et al. 1996). There are also ‘interactions among ﬁrms, universities and public research institutes’, which are representative of the second type of interaction. These include joint research activities and all other formal and informal linkages among the actors identiﬁed above, which are aimed at the acquisition by ﬁrms of generic knowledge and/or information from academic sources. ‘Inter-industry interaction through the purchase of machinery and equipment’ is an example of a third mode of interaction. Transactions involving technology, in the form of machinery and equipment, within and among sectors are regarded as contributing to intra- and interindustry ﬂows of the knowledge embodied in these products.6 Academics and policy makers have developed various methods for tracking these embodied inter-industry knowledge ﬂows. These methods make use either of quantitative information on the cross-sectoral use of patents or other specialized databases such as input–output tables and bilateral trade data. The results from empirical studies provide a picture of the relative intensity of embodied knowledge ﬂows across industrial sectors at aggregated

36 Y. Caloghirou, A. Constantelou and N.S. Vonortas levels of economic activity. The criticism levelled at this approach is that it is rather static in nature in the sense that it provides a snapshot of interindustry linkages at a particular point in time, but says nothing about the inter- and intra-ﬁrm dynamics of the process (Andersen 1992; Smith 1994). An in-depth examination of knowledge ﬂows and their magnitude and implications at more desegregated levels of economic activity requires a more ‘procedural’ approach that takes into account both quantitative and qualitative measures. Finally, the fourth type of interaction involves ‘personnel mobility and related interactions’. Although data on the number of scientists and engineers involved in research activities are widely available in most countries, details about mobility of personnel between industry and academia are very difﬁcult to trace. The most signiﬁcant contribution of universities to industry and research institutes is the continuous production of highly skilled personnel, trained to think critically. In addition, a whole range of mid-career and other training programmes offered by universities renew and reinforce the skills of industry employees. Often, industry reciprocates by offering practical training programmes for university graduates and employees and research institute personnel. Mobility of personnel could be considered as a channel through which knowledge is disseminated and/or diffused within and among ﬁrms, or among ﬁrms, universities and public research institutes. Depending on the circumstances, a variety of mechanisms can facilitate the ﬂow of knowledge through this channel: for example, an organized mobility programme designed for young researchers, or an enterprise policy of hiring academics to occupy top level positions in ﬁrms. It is important to note that the types of interactions identiﬁed above, as well as the channels through which they materialize, may have regional, national and even international dimensions. As already discussed, the SI approach, has a strong spatial dimension. The importance of regions, or more accurately distance, lies in the unique characteristics attributed to technological knowledge and learning, which are regarded as being evidently shaped by the opportunities for personal contact among the parties involved. When knowledge has a large tacit component, innovative activities tend to be regionally concentrated because economic agents beneﬁt from the externalities that appear either in the form of involuntary spillovers or as intentional information ﬂows. The existence of strong knowledge-intensive linkages with actors outside an agglomeration has been documented in the literature. The increasing globalization of economic activity through the internationalization of trade and investment has increased the opportunities for transnational interactions among economic agents and the subsequent ﬂows of knowledge beyond national boundaries. To this end, the intensity and impact of interactions within and across industries at the European level could be an indicator of the convergence of national innovation

Theoretical foundations and key concepts 37 systems and the possible emergence of a European SI. If domestic sources of information are considered to be more important than foreign sources then NSI are likely to retain their standing. However, it could be that national systems are losing out to intra-regional or pan-European systems (Arundel et al. 1998).

2.6 The institutional dimension The interactions among actors in a system should not be seen in isolation from the broader institutional environment within which they operate. Traditional economic and other regulation, such as competition and IPR protection, taxation, ﬁnance, education, national policies, European Union (EU)-level policies and so forth, can ease or block agents’ interactions and subsequent innovation-related knowledge ﬂows. This is particularly important in relation to the emergence of a European SI since there are signiﬁcant differences in the ways public sector institutions and research facilities supporting industrial innovation have been set up and operate in each country.7 Also, even within national borders it may be the case that the organization of public institutions differs and, therefore, their ability to support and promote innovative activities across industrial sectors will also differ (Nelson and Rosenberg 1993; David and Foray 1995). In this respect, some institutions may be more important than others for the organization of linkages and ﬂows of knowledge within particular industries. The range of institutions that are regarded as particularly relevant in shaping public and corporate strategies for science and technology and, thus, in inﬂuencing the ‘distribution power’ of a national system, varies according to the perspective adopted. For example, the OECD (1996) in considering best policy practice for the diffusion and adoption of new knowledge throughout an economy calls for actions which go beyond the narrow deﬁnition of innovation and technology diffusion. It argues for an extension to the boundaries of technology policy to include all measures targeting innovation and technology diffusion, irrespective of institutional arrangements and the division of labour among government agencies. Training and education policies, the ﬁnance structure and the broader macroeconomic and industrial context are inserted into the discussion as critical factors for inﬂuencing knowledge ﬂows. Furthermore, the availability of a modern communication infrastructure in a national system is seen as being particularly advantageous for establishing linkages among scientists and in allowing access to processed scientiﬁc information, electronic publishing in science, and science education and training (OECD 1998). However, adopting a less broad perspective, one might choose to focus on those institutions in the public and private sphere that are most directly science and technology related and regarded as being ‘critical’ for

38 Y. Caloghirou, A. Constantelou and N.S. Vonortas the distribution of knowledge. Smith (1994) identiﬁes three such types of institutions. a

b

c

Public sector institutions. These refer to all those formal establishments that have a direct or indirect effect upon the generation and diffusion of knowledge in a national system. Such institutions include universities, public research institutes and other private and non-proﬁt research organizations, research councils, standards setting bodies, patent ofﬁces and libraries. Public sector instruments. These are legal and regulatory measures and policy-related initiatives explicitly oriented towards the diffusion of knowledge. They include R&D collaborative programmes funded by domestic or foreign sources; technology-related legal and administrative regulations, such as mechanisms for protecting IPR; subsidies for new scientiﬁc structures and equipment, such as investments in the communications infrastructure; and public procurement policy. Technology infrastructure institutions. These encompass ‘soft’ measures, such as industry associations and conferences, training centres, trade and scientiﬁc publications, agencies and organizations supporting information exchange, etc.

The formal and informal institutions identiﬁed above differ from country to country and even from region to region. Thus, their role in and contribution to facilitating knowledge ﬂows will vary accordingly.

2.7

The ﬁrm-level dimension

At the corporate level, the intensity and effectiveness of inter- and intraindustry interactions are determined to a large extent by firms’ commitment to learning activities and their ability to recognize and appreciate the value of new information. Cohen and Levinthal (1990) call the latter the absorptive capacity of the firm and describe it as a firm’s ability to recognize the value of new external information ranging from generic science to new production equipment and to assimilate it, and its ability to exploit its economic potential through commercialization. A firm’s absorptive capacity largely depends on the level of related knowledge that already exists within the firm. Given that learning is a highly localized and history-dependent process, the set of skills and the expertise owned by a firm are critical for the nature and direction of learning processes aimed to enhance the knowledge base of the firm. Lack of tangible and/or intangible investment (in the form of human capital) in a particular area of expertise may inhibit the development of technological knowledge by the firm in that area at a later stage. Thus, the ability of a firm to make use of the results of research conducted by other firms or other public and/or private research establishments is

Theoretical foundations and key concepts 39 dependent on its ability to assess their economic potential and to build on these results. The second important factor in determining the intensity and effectiveness of interactions among actors and which affects knowledge ﬂows, is the intensity of the effort or commitment to learning displayed by the ﬁrms themselves. Strong commitment contributes to a ﬁrm’s absorptive capacity and reﬂects the intention of a ﬁrm to internalize the results of technology or knowledge purchased from third parties. Kim (1999: 115) describes this intensity of effort as ‘the amount of energy relinquished by organizational members to solve problems’. He argues that the greater the energy consumed in solving problems within a ﬁrm, the more intensive are the interactions and knowledge ﬂows among actors within and outside the ﬁrm, and thus the greater their effects are likely to be on the ﬁrm’s knowledge base. One implication of the need for ﬁrms to invest in learning in order to be able to effectively use external knowledge, even when it is freely available, is that as large ﬁrms are more likely to be involved in the types of activities that make it easier to absorb external technologies, they will likely be at an advantage in terms of both the production and use of new knowledge. Also, it has been suggested that the employees in large ﬁrms are often better placed and better ‘equipped’ to exploit external knowledge than those in smaller ﬁrms (Minne 1996). These facts are borne out by several surveys that show that there is a strong positive relationship between ﬁrm size and the probability that a ﬁrm conducts R&D, is involved in cooperative R&D and uses patent disclosure as a source of technical information (Malerba et al. 1996). The SI approach identiﬁes management practices as a signiﬁcant institutional inﬂuence on inter-ﬁrm learning and knowledge ﬂows, which are relevant to ﬁrm-level economic performance (Gjerding 1992). For example, an internal organizational structure that encourages interactions between the various departments and functions of the ﬁrm (R&D, production, sales and marketing) enhances the ﬁrm’s ‘distribution power’, that is, its ability to support and improve efﬁcient procedures for distributing and utilizing knowledge. Other aspects of the internal organization of ﬁrms, such as the geographic location of sub-units, the corporate communication infrastructure and network and the development of a corporate culture that recognizes the importance of the human factor have also been acknowledged as important in inﬂuencing learning processes and inter-ﬁrm ﬂows of knowledge (Odagiri and Koto 1993; Mowery and Rosenberg 1993). Human aspects, and especially those related to the properties that individuals bring to the workplace in terms of qualiﬁcations, scientiﬁc and technical knowledge, are receiving particular attention in the contemporary economic and business literature as a result of the current emphasis on the value of ﬁrms’ or nations’ intangible assets, that is, the non-material factors that contribute to their growth and

40 Y. Caloghirou, A. Constantelou and N.S. Vonortas performance that are not included in the traditional category of ﬁxed assets (EC 1999). In this context, managers and decision makers are increasingly being advised to think of economic development and growth as a process of knowledge rather than of capital accumulation (World Bank 1998). They are being advised to shift their strategic focus from power and resource allocation to ﬂows of knowledge, trust and communications. Investment in information and communication technologies (ICTs) and their infrastructures is regarded as a way to increase knowledge ﬂows and learning among individuals and organizations since they allow knowledge to be distributed across sectors, over long distances and at low cost. Also, it has been suggested that the availability of ICTs and communication infrastructures is shifting the boundaries between tacit and codiﬁed knowledge by making it possible, through the use of cost efﬁcient intelligent computing platforms, to extend the scope of knowledge codiﬁcation (EC 1999). Thus, new communication technologies not only facilitate knowledge ﬂows, but also contribute to the growth of accessible stocks of knowledge by transforming knowledge into information capable of being transmitted via information infrastructures.

2.8 An analytical framework for the study of knowledge ﬂows The arguments put forward in this chapter concerning the role of knowledge in innovation provide the analytical framework (see Figure 2.1) that underpins the investigation of innovation-related knowledge ﬂows in this book. The arrows in Figure 2.1 represent the interactions among actors, while the boxes represent the actors and their roles as facilitators in knowledge transactions. The performance of such a system depends on the level of its ‘distribution power’, which, in turn, is related to the availability and intensity of knowledge ﬂows. It is important to clarify that the links among these actors extend beyond national borders and the NSI. Firms can have linkages with local and foreign ﬁrms, and public sector instruments and infrastructure institutions encompass both domestic and foreign (for example, EU) initiatives and resources. Various empirical efforts have been made to characterize and quantify the role of the system elements in innovative performance. These efforts attempt to map and assess system interactions and thus require indicators of knowledge ﬂows. Chapter 3 reviews recent empirical studies on knowledge ﬂows and innovation capacity in Europe, the United States and Canada. From these studies we identify certain key areas which some of the analyses in the remaining chapters of this book examine in greater detail. By identifying key areas that have not been adequately addressed in the literature it provides a fertile ground for discussion in subsequent chapters.

Theoretical foundations and key concepts 41

Knowledge diffusion and absorption: linkages between main actors Firms’ absorptive capacity and learning capabilities The science system (public sector institutions)

Public sector instruments

Technology infrastructure institutions

Technological and innovative performance

Figure 2.1 Knowledge diffusion and absorption in the innovation system.

Notes 1 It is beyond the scope of this chapter to present in detail the various systems of innovation approaches. For a comprehensive account of different conceptions of NSI see McKelvey (1991), Lundvall (1992) and Edquist (1997) among others. 2 These deﬁnitions are intended to exclude changes to products that mainly provide subjective enhancement of customer satisfaction based on personal taste and aesthetic judgement and/or derived from fashion and/or brought about mainly as the result of marketing. Such changes are identiﬁed under the heading ‘other creative product improvements’. 3 The CIS is the main statistical instrument of the EU for the monitoring of innovation in enterprises and the analysis of its effects on the economy of Europe. The CIS has been carried out every four years since 1992 (CIS II took place in 1996 and CIS III in 2001) within the various Community RTD Framework Programmes. The ‘Oslo Manual’ provides a common methodological basis whereas data collection is undertaken by the statistical ofﬁces or competent research institutes in each member state. 4 The distinction between explicit and tacit knowledge was discussed in Polanyi (1962). 5 An interesting area of study for scholars in the economics of knowledge tradition is exactly this interplay between the individual and the collective levels: How will an individual beneﬁt from and contribute to common knowledge at the collective level? We do not deal with this aspect here. 6 Such knowledge ﬂows are beyond the scope of the empirical investigation on which this book is based. 7 Other studies have considered the issue of the convergence of national systems in Europe and the emergence of a pan-European system of innovation

42 Y. Caloghirou, A. Constantelou and N.S. Vonortas (Johnson and Gregersen 1997). These studies, however, have concentrated more on identifying convergence or divergence at the meso and macro dimensions of national systems across Europe, that is, their knowledge (education) infrastructure, production structures, institutional set-up, public and private consumer demand structures, public policies and aggregate innovative performance. In contrast, the contributions in this book seek to explore empirically the issue of convergence at the micro-level of the firm by (a) using the same structured questionnaire and identical methodology in seven EU countries, and (b) identifying the role and intensity of cross-European knowledge flows.

References Amin, A. and Cohendet, P. (2004) Architectures of Knowledge: Firms, Capabilities, and Communities, London: Oxford University Press. Ancori, B., Bureth, A. and Cohendet, P. (2000) ‘The economics of knowledge: the debate about codiﬁcation and tacit knowledge’, Industrial and Corporate Change, 9(2): 255–287. Andersen, E.S. (1992) ‘Approaching national systems of innovation’, in B.-Å. Lundvall (ed.), National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning, London: Pinter, pp. 68–92. Antonelli, C. (1999) ‘The evolution of the industrial organisation of the production of knowledge’, Cambridge Journal of Economics, 23: 243–260. Arundel, A. with Smith, K., Patel, P. and Sirilli, G. (1998) ‘The future of innovation measurement in Europe: concepts, problems and practical directions’, IDEA Paper, STEP Group, Norway, 31 July. Arundel, A. Steinmueller, E., Hawkins, R. (1996) ‘Strategies for the future; innovation in the European telecom equipment industry’, MERIT, University of Maastricht, July. Carlsson, B. and Stankiewicz, R. (1991) ‘On the nature, function and composition of technological systems’, Journal of Evolutionary Economics, 1: 93–118. Cohen, W. and Levinthal, D. (1990) ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35: 128–152. Cohendet, P. and Llerena, P. (1997) ‘Learning, technical change, and public policy: how to create and exploit diversity’, in C. Edquist (ed.) Systems of Innovation: Technologies, Institutions, and Organisations, London: Pinter, pp. 223–241. David, P.A. and Foray, D. (1995) ‘Accessing and expanding the science and technology knowledge base’, OECD STI Review, 16: 13–68. Edquist, C. (1997) ‘Systems of innovation approaches: their emergence and characteristics’, in C. Edquist (ed.) Systems of Innovation: Technology, Institutions and Organisation, London: Pinter, pp. 1–35. Edquist, C. and Johnson, B. (1997) ‘Institutions and organisations in systems of innovation’, in C. Edquist (ed.) Systems of Innovation: Technology, Institutions and Organisation, London: Pinter, pp. 41–63. European Commission (EC) (1996) First Action Plan for Innovation in Europe. Online. Available at: www.cordis.lu/innovation/src/ action.htm (last accessed 15 June 2004). European Commission (EC) (1999) Industrial Aspects of the Information Society, Report for DG XIII. Online. Available at: www.ll-a.fr/eu-epsilon/ (last accessed 12 June 2004).

Theoretical foundations and key concepts 43 Freeman, C. (1987) Technology Policy and Economic Performance: Lessons from Japan, London: Pinter. Gjerding, A.N. (1992) ‘Work organisation and the innovation design dilemma’, in B.-Å. Lundvall (ed.) National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning, London: Pinter, pp. 95–115. Hansen, J. (1999) ‘Technology innovation indicators: a survey of historical development and current practice’, SRI International (mimeo). Johnson, B. (1992) ‘Institutional learning’, in B.-Å. Lundvall (ed.) National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning, London: Pinter, pp. 23–44. Johnson, B. and Gregersen, B. (1997) ‘European integration and national systems of innovation’, ISE Report, Sub-project 3.1.3, SIRP, Linkopping University, 24 March. Kim, L. (1999) ‘Building technological capability for industrialisation: analytical frameworks and Korea’s experience’, Industrial and Corporate Change, 8(1): 111–136. Lundvall, B.-Å. (1992) ‘Introduction’, in B.-Å. Lundvall (ed.) National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning, London: Pinter, pp. 1–19. Lundvall, B.-Å. (1993) ‘User-producer relationships, national systems of innovation and internationalization’, in Foray, D. and C. Freeman (eds) Technology and the Wealth of Nations, London: Pinter. Lundvall, B.-Å. (1999) ‘Technology policy in the learning economy’, in D. Archibugi, J. and J. Michie (eds) Innovation Policy in a Global Economy, London: Cambridge University Press, pp. 19–34. Lundvall, B.-Å. and Johnson, B. (1994) ‘The learning economy’, Journal of Industrial Studies, 1: 23–42. Malerba, F. (1999) ‘Sectoral systems of innovation and production’, paper presented at the Druid Conference on National Innovation Systems, Industrial Dynamics and Innovation Policy, Rebild, Denmark, 9–12 June. Malerba, F. and Orsenigo, L. (2000) ‘Knowledge, innovative activities and industrial evolution’, Industrial and Corporate Change, 9: 289–314. Malerba, F. Lissoni, F. and Torrisi, S. (1996) The Ofﬁce and Computer Equipment Industry, EIMS Poject Report, Luxembourg: EIMS. McKelvey, M. (1991) ‘How do national systems of innovation differ? A critical analysis of Porter, Freeman, Lundvall, and Nelson’, in G. Hodgson and E. Screpanti (eds) Rethinking Economics: Markets, Technology, and Economic Evolution, Aldershot: Edward Elgar, pp. 117–137. Metcalfe, S. (1993) ‘Technology systems and technology policy in an evolutionary framework’, Department of Economics and PREST, University of Manchester, (mimeo). Minne, B. (1996) ‘Expenditures in relation to the knowledge-based economy in 10 OECD countries’, paper presented at OECD Conference on new S&T indicators for the knowledge-based economy, Paris, June. Mowery, D.C. and Rosenberg, N. (1993) ‘The US National Innovation System’, in R.R. Nelson (ed.) National Innovation Systems: A Comparative Analysis, New York: Oxford University Press. Nelson, R.R. and Rosenberg, N. (1993) ‘Technical innovation and national systems’, in R.R. Nelson (ed.) National Innovation Systems: A Comparative Analysis, New York, Oxford University Press, pp. 3–21.

44 Y. Caloghirou, A. Constantelou and N.S. Vonortas Nonaka, I. and Takeuchi, N. (1995) The Knowledge-creating Company, Oxford: Oxford University Press. Odagiri, H. and Koto, A. (1993) ‘The Japanese system of innovation: past, present, and future’, in R.R. Nelson (ed.) National Innovation Systems: A Comparative Analysis, New York: Oxford University Press, pp. 76–114. OECD (1992/1997) OSLO Manual: Proposed Guidelines for Collecting and Interpreting Technological Innovation Data, Paris: OECD. OECD (1996) The Knowledge-Based Economy, OECD/GD(96)102, Paris: OECD. OECD (1997) National Innovation Systems, PARIS: OECD. OECD (1998) Science, Technology and Industry Outlook, Paris: OECD. Polanyi, M. (1962) Personal Knowledge, London: Routledge. Saviotti, P.P. (1997) ‘Innovation systems and evolutionary theories’, in C. Edquist (ed.) Systems of Innovation: Technologies, Institutions, and Organisations, London: Pinter, pp. 180–199. Smith, K. (1994) ‘Interactions in knowledge systems: foundations, policy implications and empirical methods’, STEP Group Report No. 10, STEP, Norway. World Bank (1998) World Bank Development Report, Washington, DC: World Bank.

3

Conventional and experimental indicators of knowledge ﬂows Anthony Arundel and Anastasia Constantelou

3.1 Introduction The last decade has seen an enormous increase in both policy and academic interest in the ﬂows of knowledge between individuals, ﬁrms and institutions such as universities and government research labs and the role of such knowledge ﬂows in innovation. The third edition of the Oslo Manual, for example, includes a separate chapter on the role of ‘linkages’ in innovation and how to measure linkages in innovation surveys. European policy at both the supra-national and national levels includes a diverse range of programmes to encourage knowledge ﬂows, particularly between ﬁrms and the publicly-funded research infrastructure. This is based on a long-standing belief in a systematic failure of European ﬁrms to commercialize discoveries made by public universities and research institutes (European Commission [EC] 2001), although the causes of such a failure is the subject of a lively debate (Dosi et al. 2005). The policy interest in innovation-related knowledge ﬂows requires indicators of both the production of knowledge and also the extent and magnitude of knowledge-based transactions. The indicators used in the innovation economics literature fall into three categories: input or resource indicators, including R&D expenditures; output indicators; and progress indicators (Grupp 1998). The input indicators account for every type of resource involved in the innovation process including ﬁnancial, technological and human resources. Output indicators refer to all of the results (both tangible and intangible) from investment in innovation, and progress indicators refer to the economic effects of innovation at the micro- and macro-level (Grupp 1998: 142). Many of the widely-available input and output indicators are proxy measures of knowledge ﬂows because they give evidence of communications among individuals, ﬁrms and institutions during the innovation process (Leydersdorff and Scharnhorst 2003), but only a few indicators directly count or measure knowledge ﬂows. This chapter reviews the methodology and results of recent empirical studies on the measurement of knowledge ﬂows, with a focus on surveybased indicators. There are currently two main groups of indicators used

46 A. Arundel and A. Constantelou to measure knowledge ﬂows. The ﬁrst group includes the ‘traditional’ indicators, based on patents or citations, which have been available for decades. Some of their advantages and disadvantages are described in section 3.2 below. The second group includes indicators, obtained from innovation surveys, that have been under development since the early 1980s. Section 3.3 provides a brief overview of the methodology, advantages and disadvantages of innovation survey indicators of knowledge ﬂows. Section 3.4 reviews some of the insights gained from these innovation survey indicators with respect to the identiﬁcation of: (1) the major descriptive characteristics of knowledge ﬂows, such as their nature, extent and magnitude; (2) the effect of knowledge ﬂows on economic success; and (3) the role of knowledge ﬂows in national systems of innovation (NSI). Finally, Section 3.5 identiﬁes gaps in the innovation literature that call for further empirical investigation of the magnitude and impact of knowledge ﬂows, something we hope to partly ﬁll in in the subsequent chapters of this book.

3.2 Traditional indicators of knowledge ﬂows There are three types of traditional indicators that capture some aspects of knowledge ﬂows among actors in the public and private domain: science-based indicators, technology-based indicators and indicators of human capital. 3.2.1 Science-based indicators Science-based indicators essentially include patent and bibliometric indicators. Patents provide a wealth of information over time on the types of inventions, their inventors and the information sources used in their work. Analysis of patent applications and/or patent grants can be used to identify collaborative work between scientists and engineers (co-patents) from two or more collaborating entities (companies, universities, research institutes, etc.).1 Co-patents serve as useful indicators of the dissemination of economically useful knowledge, including between countries (OECD 2004: 32). Analyses of patent records from the European Patent Ofﬁce (EPO) database have been extensively used in innovation research to map collaboration activities in science and technology in Europe. The EPO patent database contains standardized information on the names of patenting organizations (ﬁrm, university, research institute, etc.). A typical record in this database contains information about the publication number of the patent, the name of the applicant(s), the code assigned to the applicant(s), the address of the applicant(s), and the main and supplementary technological classes of the patent. Bibliometric (sometimes called scientometric) indicators draw on statistics from the scientiﬁc literature to look at patterns of communication of

Indicators of knowledge ﬂows 47 scientiﬁc knowledge among research groups located in universities, research institutes and public and private companies. Typical bibliometric indicators are scientiﬁc publications and citations. Scientiﬁc journal articles are considered to be the outcome of scientiﬁc collaboration among researchers, which enhances science-related knowledge-ﬂows. For example, the increase in co-authored papers can be considered to be an indicator of the increasingly networked character of research. Analysis of scientiﬁc publications and co-publications broken down by scientiﬁc areas, types of institutions involved, geographical location of contributors, etc., can yield important information about the ﬂows of knowledge and the patterns of collaboration among participating entities over time and across scientiﬁc areas. Citations are frequently used as indicators of knowledge ﬂows. Analysis of the sources (citations) that scientists and engineers use in the production of new scientiﬁc and technical knowledge leading to scientiﬁc papers and patents can trace the dissemination of knowledge among actors in enterprises and academia. The Science Citation Index (SCI), a databank of scientiﬁc literature that covers all major ﬁelds of science, is a useful tool for mapping knowledge ﬂows in science located in universities and public research institutions. For example, one can focus on the addresses in patent literature and in scientiﬁc publications in order to examine the effect of geographical proximity on knowledge dissemination. A citation database could also be built using information from the EPO database on patent and non-patent literature cited in patent documents. A typical record in the SCI database consists of two variables: the publication number of the citing patent document and the publication number of the cited patent document.2 3.2.2 Technology-related indicators This group includes classic ‘input’ indicators such as research and development (R&D), which does not include information on knowledge ﬂows, and other indicators that provide some information, either aggregated at the national or sector level, or at ﬁner levels of disaggregation. Aggregate indicators include trade in machinery and equipment from one industrial sector to another and/or from one country to another. Both are classic indicators of technology diffusion and are indicators of the ﬂow of embodied technical knowledge. Input–output matrices that track the exchange of technology-related goods among industrial sectors and among countries have been used, respectively, to map inter-industry and inter-country technology ﬂows. The technology balance of payments is another aggregate indicator used by the OECD (2003: 128) to map the ﬂow of technological knowledge among OECD countries. This indicator covers knowledge-related transactions that do not necessarily involve the trade of technological

48 A. Arundel and A. Constantelou artifacts. Typically, national accounts of the technology balance of payments include fees for the licensing or sale of patents, expenditure on knowledge acquisition from external sources such as technical consultancies, R&D services, etc. These are indirect measures of the extent and magnitude of knowledge transactions among national economies. Quantitative data of this sort are not available for individual ﬁrms. Foreign direct investment (FDI) indicators have also been used to map international technology-related knowledge ﬂows. The underlying assumption is that the establishment of production facilities in a host country by a multinational enterprise (MNE) can transfer knowledge to the host country. For instance, the MNE may need to train its local workforce in the use of new production technology. In practice, the amount of knowledge that will be transferred through this mechanism will depend on the complexity of the technology, the capabilities of the domestic labour force and whether or not the MNE also establishes complementary R&D facilities in the host country. This information cannot be discriminated from the statistics on FDI that are available at the national level (Grupp 1998). 3.2.2.1 Technology-based alliances as indicators of knowledge ﬂows Indicators of technological collaboration among companies, universities and research institutes have been particularly popular in recent innovation studies. The quantitative and qualitative characteristics of technologybased alliances provide useful information for the mapping of knowledge ﬂows between the public and private sectors. Data on technology-based alliances and joint research activities between ﬁrms, universities and research institutes are collected from articles in the business and daily press and from information published by national and international authorities. An example of information derived from press articles is the Cooperative Agreements and Technology Indicators (CATI) database developed and maintained by the Maastricht Economics Research Institute on Innovation and Technology (MERIT) at the University of Maastricht. This database contains details on cooperative research agreements established among companies in technology-intensive sectors. In addition, several databases on joint research activities have been developed using the secondary data sources. Three such databases are extensively used in other chapters of this book in order to evaluate knowledge ﬂows. The European Research Joint Ventures database (EU-RJVs) is one of three core databases in the STEP-TO-RJVs databank.3 Some of these databases are used in the analyses in Part II of this book. The STEP-TO-RJV database records information on transnational collaborative research projects funded through the European Framework Programmes (FWPs) on Research and Technological Development (RTD) over the period 1983–98. This period corresponds to the ﬁrst four FWPs. The database

Indicators of knowledge ﬂows 49 includes programmes whose main focus was on the creation of new technological knowledge. All the better-known programmes, including ESPRIT, BRITE-EURAM, JOULE, RACE, BIOMED, BIOTECH, ENV, TELEMATICS and many more (64 programmes in all) satisfy this criterion. The database contains information on research projects that involve at least one agent from the private sector (ﬁrm) and identiﬁes individual participants in every project. The total number of recorded projects satisfying all selection criteria is 9,335. A total of 20,499 different organizations from 50 countries participated at least once in these projects. The sum of recorded memberships reaches 65,476. For a signiﬁcant number of the ﬁrms that participated in projects over the period 1992–96 the database also includes longitudinal ﬁnancial information (ﬁve years) obtained from the commercially available database AMADEUS. The EUREKA-RJV database is the second database in the STEP-TORJVs databank. This database records information on transnational collaborative research projects selected within the EUREKA initiative in the period 1985–96.4 The EUREKA-RJV database is structured in the same way as the EU-RJV database. There are 1,031 joint venture activities recorded in the database. These collaborative projects involve 6,233 memberships corresponding to 4,261 entities from 36 countries. The database records longitudinal ﬁnancial information for 1,250 ﬁrms derived from AMADEUS. The RJV-EPO database was jointly produced by partners in the ‘KNOW for Innovation’ research project and combines information from the EURJV database, the EPO patents database and the EPO patent citations database. The EPO patents database provides information about all patents applied for and granted by the EPO during the period 1978 to 1998; the EPO patent citations database contains standardized information on citing and cited patent documents for the period 1978–99. The merging of the three databases was carried out at the level of the ﬁrm, resulting in a large set of companies for which there are records of RJV participation, European patent applications and patent citations. 3.2.3 Human capital indicators Data on the stock of highly-skilled individuals have been available for many years and provide indicators for innovative capability. Recent work has focused on collecting data on the mobility of highly-skilled individuals. Indicators based on mobility data can provide a proxy measure for knowledge ﬂows, under the reasonable assumption that highly-skilled individuals moving from one ﬁrm to another, or from a university to a ﬁrm, will bring specialized knowledge with them. The Nordic countries have established statistical mechanisms to track the mobility of people from one sector to another and over time (OECD 1997).5 Other countries, such as the United States, Canada, France and Switzerland, obtain

50 A. Arundel and A. Constantelou some information on the career movements of individuals with doctorate level degrees (Auriol and Recotillet 2004). Since 2000, most OECD countries use their national census to collect data on the total stock of nonnational tertiary-educated individuals. Dumont and Lemaitre (2004) use this information to produce tables on the ﬂows of tertiary-educated individuals from one OECD country to another OECD country. 3.2.4 Shortcomings of traditional indicators Many traditional innovation indicators based on patents, bibliometrics and human capital stocks capture the output of knowledge creation (patents and journal articles) or its potential creation (human capital stocks). The growing literature in this ﬁeld has successfully used each of these indicators to substantially increase our understanding of knowledge ﬂows and spillovers. Furthermore, patents have considerable advantages for tracing knowledge ﬂows, including the long time series available and the consistency with which the information is collected. Nevertheless, there are a few drawbacks to the use of these indicators for evaluating knowledge ﬂows. The ﬁrst disadvantage is that they only provide indirect measures of the ﬂows of knowledge used in innovative activities. The use of patent citation data, for instance, is problematic because the cited patent (or research article) may not have contributed to the invention, with the citation only included to build the patent claim or it may have only been added by the patent examiner. Bibliometric citations are probably a more accurate trace of knowledge ﬂows, but their main value is for evaluating academic uses of knowledge. They are of less beneﬁt in tracking the types of information ﬁrms use to innovate. Some of these disadvantages could result in misinterpreting an unintended ﬂow of information from one actor to another as an intended ﬂow. Geroski (1995) points out that many apparent knowledge ﬂows might, in reality, be coincidental, as a result of the ‘more or less independent development of similar answers to commonly perceived problems which a group of competitors all arrive at by drawing on a pool of common scientiﬁc knowledge’ (p. 112). The second drawback is that traditional indicators, such as patents or bibliometric citations, are limited to codiﬁed knowledge. They cannot evaluate tacit and embodied knowledge easily because they do not provide a direct measure of these sources, although both can indirectly point to the possible transfer of tacit knowledge. Indicators for the mobility of the highly-skilled can proxy ﬂows of tacit knowledge, but it is difﬁcult to estimate the type of knowledge that is moved from one place to another with any precision. The third problem with traditional indicators is that they fail to capture the complexity of knowledge ﬂows, which can follow a range of alternative

Indicators of knowledge ﬂows 51

Firm A: Creative effort involved in producing an innovation

Protection of knowledge through secrecy

Codification or embodiment of knowledge used to produce the innovation

Market unpatented innovation

Information diffusion via conferences, trade fairs, journals

Patent innovation (Patent propensity)

Reverse engineering

Conference attendance, journal subscriptions, etc.

Access to patent databanks for disclosure of information

Non-codified knowledge

Undisclosed complementary information

Licensing agreements

Informal contacts, joint ventures

Firm B: which wants to acquire the knowledge used to produce the innovation

Figure 3.1 The complexity of knowledge ﬂows (source: Based on Arundel et al. (1998)).

paths in response to the strategic activities of different ﬁrms. This is illustrated in Figure 3.1, which shows the different routes that might be used by Firm B to obtain information about an innovation developed by Firm A.6 The speciﬁc routes available to Firm B depend on the strategic choices made by Firm A to appropriate its innovation. This includes whether or not the information behind the innovation is codiﬁed, although many innovations will probably be based on both codiﬁed and non-codiﬁed knowledge. For example, if ﬁrm A markets an innovation without patenting it, then Firm B may get access to this innovation through reverse engineering combined with other possible information sources, such as information disclosed in journals. If ﬁrm A decides to patent the innovation, information will be available through patent disclosures, but it might also be

52 A. Arundel and A. Constantelou available through a wide variety of other sources. In some cases the patent may not provide enough information for a competitor to successfully circumvent the patent if success depends on complementary knowledge on how to manufacture the product at low cost. Under these conditions, the patent could encourage Firm A to license the innovation, including the necessary ‘undisclosed complementary information’, to Firm B. Alternatively, patent protection might permit Firm A to disclose relevant information in other locations, such as at conferences or in journals. The complexity of the different methods of knowledge acquisition, and the effects of the strategic decisions made by Firm A, show that using patents to trace knowledge ﬂows will only be one part of the complete picture. Furthermore, patents are of no value at all if ﬁrms do not patent their innovations. Although this problem is less common in the pharmaceutical sector, where the majority of innovations are patented, in many sectors less than half of all innovations are patented (Arundel and Kabla 1998) and therefore invisible to patent statistics.

3.3 Survey indicators of knowledge ﬂows An alternative to using traditional indicators has been to investigate knowledge ﬂows through surveys and ask ﬁrms directly about their use of speciﬁc sources and types of knowledge. In theory, surveys can measure the ﬂows of all types of knowledge – embodied, codiﬁed and tacit – and investigate a wide range of sources, such as competitors, clients and public research organizations (PROs). Surveys can also evaluate the direct transfer of information via the purchase of licences, new instruments and production machinery, or the hiring of new staff. Table 3.1 provides an overview of the types of information of relevance to knowledge ﬂows that have been collected in several major surveys in Europe, the United States and Canada. Each of these surveys asks a different set of questions, at the level of the ﬁrm, on how it acquires information of value to its innovative activities. With the exception of the surveys on biotechnology and advanced manufacturing technology (AMT), each survey collects data on a sample of ﬁrms across all manufacturing sectors and in some service sectors. The biotechnology and AMT surveys are included here as an example of the types of data that can be collected for a sector or for a speciﬁc group of technologies. The Canadian Innovation Survey uses a hybrid design that collects information at the level of the ﬁrm and on a speciﬁc innovation. The number of stars (★) in Table 3.1 is a rough measure of the thoroughness with which the survey investigates knowledge ﬂows. For instance, surveys that collect interval data on the value of an information source are given more stars than surveys that collect only nominal data. Extra stars are also given for additional questions that provide more in-depth coverage.

★★

★

★

★★★ ★★★

★★ ★★★ ★★ ★★★

★★

★ ★

★ ★★

★ ★★★

CIS-1

★★★ ★★

★ ★★

PACE

Europe

★★ ★★★

★★

★ ★★

★ ★★

CIS-2

■

★★★ ★★★

★★

★ ★

★★★ ★★★ ★★ ★★ ★★★ ★

CMS

US

★★ ★★★

★ ★ ★★ ★ ★★★

★★★

★★★ ★★ ★★

★★★

★

★★ ★★

AMT

★★★

★

★★ ★

★★ ★ ★★★

★★ ★★★

Bio:1-2

★★ ★★

Canadian Innnovation Survey

Canada

Notes PACE (Policies, Appropriation and Competitiveness for European Enterprises) is a 1993 survey of Europe’s 500 largest manufacturing ﬁrms funded by DGXIII. CIS-1: 1993 Community Innovation Survey of ﬁrms of all sizes in 11 EU countries plus Norway. CIS-2: 1998 update of CIS-1; also includes the UK. CMS (Carnegie Mellon Survey): 1994 survey of R&D labs in the US and in Japan. Bio:1-2: Survey of Biotechnology Use – 1996; Biotechnology Firm Survey – 1997. AMT: Survey of Advanced Technology in Canadian Manufacturing – 1998. PROs: Public Research Organisations.

1. Sources of knowledge for the ﬁrm’s innovative activities (not speciﬁed as to type) Internal information sources External information sources By geographic location By purpose PRO information sources By geographic location 2. Codiﬁed knowledge Patents, publications, etc. 3. Tacit knowledge From PROs From other sources 4. Embodied technology Technology acquisition By geographic location Technology transfer By geographic location Adoption of speciﬁc technology 5. Interactive knowledge sharing Cooperative R&D/Alliances By geographic location 6. Channels for obtaining knowledge from: PROs Other sources

Type of knowledge

Table 3.1 Information on knowledge ﬂows collected by major surveys of manufacturing ﬁrms

54 A. Arundel and A. Constantelou Table 3.1 shows that existing innovation surveys provide some data on most types of knowledge ﬂows, although the best coverage to date is for sources of technical knowledge that do not specify the type of knowledge. The three main drawbacks to the type of data collected by innovation surveys summarized in Table 3.1 are: 1 2

3

the coverage of tacit knowledge is generally poor, since this type of knowledge is rarely separated out from other types; few data are collected on the ﬁrm’s objectives for using a particular source of knowledge, although basic information on the use of a speciﬁc source to suggest ideas for new projects or to assist with the completion of an ongoing project is gathered through the Carnegie Mellon Survey (CMS) and the Canadian Innovation Survey. However, none of the three European Community Innovation Surveys (CIS) investigate the purpose for which information is gathered. One means of overcoming this problem in analyses of the CIS is to use responses to questions on the objectives of innovation, but these will be difﬁcult to link directly to knowledge ﬂows; the information on knowledge cannot be readily linked to the three types of spillovers: market spillovers via embodied technology,7 knowledge spillovers via codiﬁed and tacit knowledge, and network spillovers, which can occur through all three types of knowledge.

Another obstacle is that none of the existing innovation surveys provide complete coverage of the different sources of knowledge available in a system, or the types of knowledge. The PACE (Policies, Appropriability and Competitiveness for European Enterprises) survey provides the best available data for PROs and it includes questions that can be used to provide proxy measures of the value of tacit and codiﬁed knowledge, but these measures are imperfect. CIS-1 includes extensive coverage of the acquisition and transfer of embodied technology, but it does not gather information on the geographic location of knowledge sources. CIS-2 and CIS-3 do not include questions on the ﬂows of embodied technology, but they do include extensive questions on innovation collaboration. Many of the limitations of innovation surveys are due to space constraints. It is simply not possible to cover all aspects of knowledge ﬂows in a single survey, particularly if the response is voluntary. The alternative is to use shorter, specialized surveys. This is the approach taken by PACE, which focuses on PROs and appropriation issues. The CMS gathered a wide range of interesting data of relevance to knowledge ﬂows, particularly on the use of the outputs of universities (Cohen et al. 2002a). A separate paper using the CMS data on appropriability conditions (Cohen et al. 2002b) found that differences in patent systems inﬂuence the dissemination of knowledge.8 The results suggest that the patent system in Japan plays a major role in disseminating know-

Indicators of knowledge ﬂows 55 ledge, while in the US this role is less evident. These differences are due to a combination of factors, including differences in the legal structure of the two systems, which has provoked different patenting strategies in the two countries. Generally, existing survey data provide relatively good coverage of the different types of knowledge ﬂows, whether among ﬁrms, or between ﬁrms, universities and PROs; and of the ﬁrm-level factors that inﬂuence ﬁrms’ abilities to use knowledge ﬂows. Conversely, surveys provide very little data in relation to the incentives for ﬁrms to participate in these ﬂows and the various channels and mechanisms that facilitate them. Strictly speaking, incentives should be limited to factors that improve the proﬁtability or efﬁciency of the ﬁrm’s innovative activities. For instance, the expectation that cooperative R&D reduces costs is an incentive to participate in these programmes. This type of information is rarely directly gathered in surveys, but output measures will provide an indication. For instance, are ﬁrms that participate in cooperative R&D more proﬁtable, or do they have a higher fraction of sales from innovative products? Also, surveys rarely explore in detail the different channels and the various mechanisms through which knowledge can ﬂow. An exception here is the CMS and PACE. PACE explores nine different channels for obtaining the results of research from PROs, while the CMS also looks at several channels for obtaining information from both PROs and from other ﬁrms. The scarcity of detailed investigations of knowledge channels is due to the fact that each speciﬁc knowledge source needs to be examined. This is only feasible when the source is of particular interest, either to policy or to economic theory.

3.4 What do innovation surveys tell us about knowledge ﬂows? This section reviews what innovation surveys have told us so far about knowledge ﬂows, focusing on the following three questions: •

•

What are the major descriptive characteristics of knowledge ﬂows, such as their nature, extent and magnitude? These include the mechanisms and channels for transferring information, the effect of ﬁrm characteristics such as size or R&D intensity on knowledge ﬂows, and the effect of external factors such as sector-wide institutional conditions. How do knowledge ﬂows affect innovative outputs or the economic success of a ﬁrm? Innovative output is measured in the CIS in terms of the percentage of total sales due to innovative products (innovative sales share). Other measures of economic success include positive changes in sales, employees, exports or proﬁtability.

56 A. Arundel and A. Constantelou •

What is the effect of the NSI and in particular geographical proximity, on the use of knowledge sources? How do proximity effects vary by type of knowledge (tacit, codiﬁed)? Are proximity effects, and/or the impact of national institutional conditions, greater in some sectors than in others?

3.4.1 Characteristics of knowledge ﬂows Most of the available analyses of the results of innovation surveys provide descriptive results about the relationship between knowledge ﬂows and several of the major characteristics of ﬁrms, such as size, sector of activity and R&D status. 3.4.1.1 Firm size One of the most robust results gleaned from innovation surveys is that there is a consistent, positive relationship between ﬁrm size and the importance of many external sources of information, including PROs, patent databases and cooperative R&D (Bosworth and Stoneman 1997; Schmidt 1997; Arundel and Steinmueller 1998; Autio et al. 1997; Arundel et al. 1995; Cohen et al. 2002a; Tether 2002; Laursen and Salter 2003). Furthermore, Laursen and Salter’s (2003) analyses of the CIS-3 data for the UK ﬁnd that large ﬁrms tend to ﬁnd more external knowledge sources of value than small ﬁrms. There are a few exceptions to this rule. Small ﬁrms attribute greater importance to trade fairs as an information source, and virtually all ﬁrms, regardless of size, ﬁnd publications to be of great value. There are several important aspects of the relationship between ﬁrm size and knowledge sources that have not been adequately explored. The fact that the importance of external information sources, or the number of sources used, increases with ﬁrm size tells us very little. What we really need to know is whether the number of external contacts, or the intensity of use of these contacts increases per employee, per R&D employee or per unit of R&D expenditure. Unfortunately, existing survey data only give us information on the variety of external sources used. An alternative would be to determine whether the R&D division(s) of large ﬁrms, and SMEs of similar size to them, use similar information-gathering strategies. 3.4.1.2 Sector effects Innovation surveys consistently report differences in the value of speciﬁc information sources by sector (Arundel et al. 1995; Autio et al. 1997; Tether 2002; Arundel and Geuna 2004). These differences are partly related to the technological intensity of the sector. There is also a wide variation by sector in the importance of customers and public research,

Indicators of knowledge ﬂows 57 while there is less variation within manufacturing in the value of reverse engineering as an information source. One challenge is to differentiate between true sector effects in the sourcing of knowledge and effects that actually occur at the ﬁrm level. For example, speciﬁc features of a technology that are shared by all ﬁrms active in a sector could result in a true sector effect. For example, almost all ﬁrms in the pharmaceutical sector give a high importance to universities as a source of knowledge, since innovation is strongly dependent on medical research. Conversely, ﬁrm speciﬁc conditions that are not sector speciﬁc, such as management strategies, can also have a large inﬂuence on knowledge sourcing strategies (Laursen and Salter 2003). 3.4.1.3 Internal versus external knowledge sources Survey research consistently shows that innovative ﬁrms rank information sources internal to the ﬁrm, such as their own R&D, more highly than external sources (Levin et al. 1987; Arundel 1997). The CIS-3 results for France report 46.6 per cent of ﬁrms give a ‘high’ importance to ‘sources within the company’ (Francois and Favre 1998), compared to the second highest rating of 31.8 per cent for customers, while the percentages of ﬁrms assigning high importance to one of the other nine sources suggested vary between 2 per cent and 13 per cent. These results are not surprising, since ﬁrms devote the major proportion of investment on new knowledge in-house. Whether or not this will change, due to an increase in cooperative R&D, is an empirical question that is discussed in greater depth in Chapter 8 in this book. Discussions of knowledge ﬂows need to obtain an assessment of the relative value to innovative activities of external versus internal information sources. One study based on the CIS-1 results for Belgium approaches this problem as the ‘make’ or ‘buy’ decision and ﬁnds, not surprisingly, that most ﬁrms use both strategies. Factors that increase the probability of only a buy decision include poor appropriability conditions (Veugelers and Cassiman 1999). Another aspect of interest is the relative importance of different external information sources and the factors that inﬂuence these differences. For instance, under what conditions are PROs a more important knowledge source than customers, or what types of ﬁrms rate suppliers as being more important than customers? Several factors, such as the technological content of innovation or strategic choices (Laursen and Salter 2003), might inﬂuence the relative importance of different information sources. For instance, using a speciﬁc information source may increase the risk of leaking conﬁdential information on the ﬁrm’s innovation projects to competitors. Some external sources, such as publications and patent databases, can be accessed in secret, without any information about a ﬁrm’s innovation strategies being

58 A. Arundel and A. Constantelou revealed to competitors. Another low risk external information source is reverse engineering, which Levin et al. (1987) found to be a comparatively important method for learning about new technological developments. Other external sources, such as fairs or conferences, are possibly low risk, while yet others could confer a high risk of leaking information to competitors. The location of innovative activity, either within or outside a ﬁrm, will also be inﬂuenced by characteristics of the technology itself. The internal expertise required to develop well-understood technologies is likely to be less than that required to develop complex technologies at the technological frontier. The main internal sources of information on the adoption of complex technologies are likely to be the ﬁrm’s research and experimental development departments, while well-understood technologies are more likely to involve assistance from operating staff or production engineers. The main external sources of information about complex technologies will be research-intensive institutions, such as universities, research consortia and joint ventures, while suppliers will likely be a major information source for established technologies. Well-understood technologies should also require the use of fewer internal and external information sources than new technologies under development. 3.4.1.4 Absorptive capacity The link between the type of technology under development and the adaptive capability of the ﬁrm is an underlying thread running through many of the studies based on innovation surveys. Very few of these studies address this problem directly, although several make the point that more R&D intensive ﬁrms are more likely to participate in cooperative R&D, and that large ﬁrms, which are also more likely to have high levels of absorptive capacity, are more likely than small ﬁrms to use many different external information sources. Although the concept of absorptive capacity has received a great deal of attention, it has only been studied empirically through the use of proxy indicators. One common assumption is that the amount of effort expended on innovation, for instance, the amount of R&D spending or scientists employed, is an indirect measure of absorptive capacity. This could be a reasonable assumption for large ﬁrms, but we do not know if ﬁrms intentionally perform speciﬁc activities, such as basic research, in order to build up absorptive capacity, or if it is largely a by-product of existing innovative activities. It becomes even more difﬁcult when looking at SMEs or ﬁrms that do not conduct R&D since there are few available indicators for measuring the ability of non R&D performing ﬁrms to adopt innovations that were developed outside of the ﬁrm.

Indicators of knowledge ﬂows 59 3.4.1.5 Public research organizations An important issue is why a particular knowledge source is used. This is of particular interest in the case of PROs, which are less useful to small rather than large ﬁrms. The CIS provide some evidence to show that there are differences in the innovation objectives among ﬁrms that use or do not use public research as an information source. Firms that stress development of new products or more technically advanced products as an objective are more likely to ﬁnd public research to be an important source of information than ﬁrms that innovate in order to improve their existing products (Arundel 1997; Tether 2002). On average, PROs do not rank as an important external information source to small- and medium-sized ﬁrms. This holds for most sectors. Unfortunately, the low value of PROs to small ﬁrms, consistently shown in many studies based on the CIS data, has partly led to a misinterpretation of the role of PROs in national innovation systems. PROs are substantially more important to large ﬁrms that account for the majority of R&D spending. Arundel and Geuna (2004) control for this effect by weighting the importance of different knowledge sources by the ﬁrm’s R&D expenditures, using both the PACE and CIS-2 data. This substantially increases the relative importance of PROs as a knowledge source compared to other sources. Another approach, ﬁrst used by Mansﬁeld (1991), is to ask ﬁrms for the percentage of innovations that would not have been developed without the presence of a speciﬁc factor (patents) or knowledge source. Beise and Stahl (1999) analysed the results of a similar question from the Mannheim Innovation Panel for Germany. They report that only 5 per cent of new product sales were dependent on publicly funded research conducted by PROs. 3.4.1.6 Collaborative R&D Collaborative or cooperative R&D is one mechanism that can be used by ﬁrms to obtain knowledge from external sources. This mechanism has received an enormous amount of interest since the 1990s from both innovation economists and European policy makers (Smith 2004). However, the value to ﬁrms of cooperative research needs to be estimated within the perspective of the value of the ﬁrm’s own internal innovative activities, which, as already discussed, have been shown to be of more value than external information sources. So far, we do not have good data on the relative value of cooperative R&D compared to internal activities, although analysis of the CIS-3 results shows that cooperation is widespread, with 7.1 per cent of all European SMEs (innovative and noninnovative ﬁrms combined) in the 25 EU countries participating in at least one cooperative R&D arrangement. This approach can be extended by further work on the percentage of R&D spent on external projects.

60 A. Arundel and A. Constantelou 3.4.2 Do knowledge ﬂows affect outputs? Two important questions are: 1) Are speciﬁc types of knowledge sources linked to an increase in innovative outputs and 2) Does cooperative R&D increase innovation outputs? The outputs of interest include the innovative sales share, sales growth, employment growth, exports and patents. These are two of the most important questions addressed in the context of this book, since the answer has immediate repercussions for both policy and the innovative strategies of private ﬁrms. Many of the analyses of the ﬁrst CIS show that the more innovative ﬁrms tend to use an above-average number of external knowledge sources (Bosworth and Stoneman 1997).9 Similar results were obtained by Autio et al. (1997) for the pulp and paper sector (NACE 21), although the differences between ﬁrms were rarely statistically signiﬁcant, due to the small number of ﬁrms in this sector that responded to the CIS. Some of these analyses do not control for the effect of ﬁrm size, which is an important confounding variable because larger ﬁrms tend to be more innovative than small ﬁrms. Other research using the ﬁrst CIS has looked at the association between innovative output and speciﬁc external sources of information by sector. Arundel et al.’s (1996) analysis of the CIS results for telecommunication ﬁrms ﬁnds that the use of PROs is positively associated with the share of total sales of innovative products. However, they also report a negative association between the amount of R&D conducted outside of the ﬁrm and successful innovation. Malerba et al. (1997) reported similar negative results for the ofﬁce equipment sector. Christensen et al. (1997) compared the use of major information sources among innovative leaders (in the top tercile for the percentage of innovative products) and laggards (bottom tercile) in ﬁrms in the German food and beverages sector. A higher percentage of leaders than laggards obtain information from PROs and journals, but there was little difference in the frequency of use for most other information sources. Many studies have looked at the relationship between cooperative R&D and innovation outputs, although few if any have concentrated on speciﬁc sectors. Nas et al. (1994) found that CIS-1 ﬁrms that participate in cooperative R&D have a higher share of new products in their product line than ﬁrms that do not. Conversely, two studies that looked at the ﬁrst CIS results for the telecommunication and ofﬁce equipment sectors found that ﬁrms that participated in cooperative R&D have a lower innovation sales share than those that did not (Arundel et al. 1996; Malerba et al. 1997). WZB (1996) reports that R&D cooperation among CIS respondents in the chemical sector is most prevalent in the more innovative subsectors of this industry, such as pharmaceuticals and agrochemicals. More recent research based on innovation surveys from the late 1990s has generally found a positive link between collaboration and innovation

Indicators of knowledge ﬂows 61 outputs. Mohnen and Therrien (2004), using the 1999 Canadian Innovation Survey, ﬁnd that innovative ﬁrms that collaborate on innovation have a higher innovation sales share than comparable ﬁrms that do not collaborate. Similar results are reported by Loof and Brostrom (2004) for Sweden, using the third CIS. Although an intermediate output indicator, Czarnitzki et al. (2004) ﬁnd that R&D collaboration has a positive effect on patenting activity for both Finnish and German ﬁrms. A study of Dutch ﬁrms that responded to CIS-2 and CIS-3 ﬁnds that the type of information source inﬂuences the outcome. Cooperation with competitors and suppliers improves ﬁrm level productivity, while cooperation with PROs is more likely to increase the innovation sales share (Belderbos et al. 2004). 3.4.3 National systems of innovation and geographical proximity Another key question is the extent to which the distributive power of knowledge ﬂows depends on regional or national innovation systems. So far, it has not been possible to empirically test the role of innovation systems in a satisfactory manner. The best method is to compare the behaviour of ﬁrms within the same industry, but in different countries. This technique was used in one study based on the CIS-1 data. A country effect was found after controlling for industry- and ﬁrm-level factors (Calvert et al. 1996). Another important issue linked to national systems of innovation (NSI) is the effect of proximity on knowledge ﬂows. Research using traditional indicators such as patents or citation analysis have identiﬁed strong proximity effects, but these results do not provide direct evidence of information ﬂows. Surveys have consistently reported proximity effects, although the strength of the effect varies with the source. The effect is strongest for PROs (Arundel and Geuna 2004). One would expect the proximity effect to be weakest for mechanisms such as reverse engineering, which can be done anywhere. Bosworth and Stoneman (1997), in an analysis of the CIS1 data, report that domestic partners are the most important source of and destination for technology transfer. Christensen et al. (1997), in an analysis of CIS-1 data for food and beverages ﬁrms, report that these ﬁrms engage in cooperative R&D with domestic PROs more frequently than with foreign PROs. Results from the French CIS-2 also show that cooperative agreements with French partners are more prevalent than those with foreign partners.10 Beise and Stahl (1999) did not ﬁnd that proximity to a PRO within Germany increased the probability that a ﬁrm would source a new product from a PRO. Also, more R&D intensive ﬁrms cited distant PROs more frequently than less R&D intensive ﬁrms.

62 A. Arundel and A. Constantelou

3.5 The way forward: missing elements in the analyses and indicators The discussion so far has highlighted several areas where further work on knowledge ﬂows would be of beneﬁt. This could be based either on more analyses of existing surveys, such as the European CIS and similar surveys in other countries, or on new surveys seeking previously unavailable data. Some key areas, which have not been adequately covered by existing surveys and where further investigations could be conducted, include the following. 3.5.1 Channels for obtaining information Very few of the existing surveys focus on the different channels available for obtaining information and the mechanisms that apply to each channel. As noted above, this is largely because of the number of questions that are required, which means that this is only feasible for specialized surveys. A detailed discussion on the different channels through which knowledge is communicated and the properties of each channel can be found in Chapter 4 of this book. 3.5.2 The purpose for which information is acquired Most innovation surveys provide no data on the type of information that ﬁrms are seeking from particular sources. This is unfortunate, because a better understanding of the type of information that is obtained could be of value to policy. For instance, why do ﬁrms access public research? Is it to get suggestions for new projects or to complete old ones? The former would indicate that PROs need to remain on the cutting edge of research, while the latter suggests that PROs would have more success in commercializing their research if they concentrated on applied research. The reasons why ﬁrms access public research could also vary by sector, ﬁrm size or other factors. For example, small ﬁrms are less likely to use PROs than larger ﬁrms, although we do not know why this is the case. This type of information could also explain why customers are consistently cited as one of the most important knowledge sources. It would be useful to know whether customers suggest new projects or technical solutions, or perhaps both. Suggesting new projects would act as a source of demand, spurring investment in in-house innovation activities. 3.5.3 Indicators of absorptive capacity There is a need for indicators of the prevalence of absorptive strategies. These could include questions on whether or not a ﬁrm conducts parallel research projects to try to replicate work done elsewhere, or R&D projects

Indicators of knowledge ﬂows 63 to help it understand discoveries made outside the ﬁrm. Furthermore, we need better information on the capacity of SMEs to use sophisticated technical information, such as public research results or patent disclosures. Policies to encourage SMEs to make use of this type of information could be misguided if these ﬁrms are incapable of using them. 3.5.4 Research cooperation and innovative behaviour Although the CIS and other surveys obtain a lot of data on cooperation, there are many relevant areas that are poorly understood. Information on the types of cooperative research agreements among and between ﬁrms and PROs, as well as additional information on the geographical location of partners involved in such agreements, could be combined to explore further the relationship between cooperation and proximity. Another interesting topic for study is whether or not participation in different types of cooperative research agreements is linked to innovative success. 3.5.5 Uncovering the cause of sector effects Differences in knowledge ﬂows by sector are likely to be caused by factors linked to speciﬁc sectors. For instance, pharmaceutical ﬁrms ﬁnd PROs of greater value than ﬁrms in other sectors because PROs conduct valuable research in the areas of chemistry, health and biology. It would be worthwhile to try to open the black box to uncover the particularities of speciﬁc sectors. 3.5.6 Appropriability conditions and their impact on innovation Appropriability conditions could inﬂuence the type of external knowledge sources used by ﬁrms and explain some of the sectoral differences in knowledge ﬂows. Although some limited data on appropriation have become available, additional measures of appropriability conditions could enable comparison of their inﬂuence on the decision to source knowledge from internal versus external sources. This chapter has reviewed the typical (traditional) and experimental indicators of knowledge ﬂows used in large-scale innovation surveys. Its main objective is to inform the reader about the state of the art in innovation measurement and to identify areas where further empirical investigation and quantitative work are needed. We would argue that the KNOW survey, conducted under the auspices of the KNOW Programme and funded by the European Commission with the purpose to empirically appraise the diffusion of knowledge of relevance to the innovative activities of European industry, has been a serious effort to tackle many of the elements identiﬁed above. Some of the results are discussed in Part II of this book.

64 A. Arundel and A. Constantelou

Notes 1 Patent applications and patent grants are different sides of the same coin. Which of them is used in innovation research depends on the circumstances and the research questions involved. For a debate on this topic see Grupp (1998). 2 Such a database has been developed by MERIT. The EPO–MERIT database contains 894,103 records that correspond to 482,687 citing patents and 388,986 cited patents. 3 The creation of this database was funded by the European Union under the Targeted Socio-Economic Research (TSER) Programme ‘Science and Technology Policies Towards Research Joint Ventures’ (Caloghirou and Vonortas 2000). 4 EUREKA was designed in the mid-1980s to complement the European FWPs on RTD by selecting collaborative research projects focusing on the development of ﬁnal products and processes. The selected projects are not funded by any central agency; individual partners seek funding from their respective national governments. In contrast to the RTD FWPs, EUREKA does not prespecify technology areas for competition. 5 See, for example, Jacobsson et al. (1996) and Sternberg et al. (1996) for indicators of human resources. 6 This example is quoted from Arundel et al. (1998). 7 Note that these types of spillovers are beyond the scope of the present analysis since examination of embodied knowledge ﬂows is not within the remit of this book. 8 See Chapter 2 for the different types of institutions identiﬁed as being critical for the distribution of knowledge. 9 Part of this effect is probably because more innovative ﬁrms also tend to be larger, and larger ﬁrms simply have more opportunities to form external contacts. 10 For example, 79 per cent of French ﬁrms that report cooperative research agreements with PROs cooperate with a French PRO, while only 21 per cent cooperate with a foreign PRO. The only type of cooperative agreements where French ﬁrms have more foreign than domestic partners is agreements with competitors where 34 per cent of ﬁrms report an agreement within France compared to 66 per cent outside of France (Francois and Favre 1998).

References Arundel, A. (1997) ‘Enterprise strategies and barriers to innovation’, in A. Arundel and R. Garrelfs (eds) Innovation Measurement and Policies, Luxembourg: European Commission, pp. 101–108. Arundel, A. and Geuna, A. (2004) ‘Proximity and the use of public science by innovative European ﬁrms’, Economics of Innovation and New Technology, 13: 559–580. Arundel, A. and Kabla, I. (1998) ‘What percentage of innovations are patented? Empirical estimates for European ﬁrms’, Research Policy, 27(2): 127–141. Arundel, A. and Steinmueller, W.E. (1998) ‘The use of patent databases by European SMEs’, Technology Analysis and Strategic Management, 10: 157–173. Arundel, A., Soete, L. and van der Paal, G. (1995) Innovation Strategies of Europe’s Largest Industrial Firms, Brussels: European Commission. Arundel, A, Steinmueller, W.E. and Hawkins, R. (1996) ‘Strategies for the future: innovation in the European telecom equipment industry’, EIMS Publication No. 39, Luxembourg: EIMS.

Indicators of knowledge ﬂows 65 Arundel, A., with Smith, K., Patel, P. and Sirilli, G. (1998) ‘The future of innovation measurement in Europe: concepts, problems and practical directions’, IDEA Paper, Oslo: STEP. Auriol, L. and Recotillet, I. (2004) ‘Conclusions of the workshop on user needs for indicators on careers of doctorate holders’, OECD DSTI/EAS/STP/NESTI (2004)28, Paris: OECD, 27 September. Autio, E., Dietrichs, E., Führer, K. and Smith, K. (1997) Innovation Activities in Pulp, Paper and Paper Products in Europe, STEP Report 4/97, Oslo: STEP. Beise, M. and Stahl, H. (1999) ‘Public research and industrial innovations in Germany’, Research Policy, 28: 397–422. Belderbos, R., Carree, M. and Lokshin, B. (2004) ‘Cooperative R&D and ﬁrm performance’, Paper presented to the DRUID 2004 Conference on Industrial Dynamics Innovation and Development, Denmark: Elsinore. Bosworth, D. and Stoneman, P. (1997) ‘Information and technology transfer in Europe’, in A. Arundel and R. Garrelfs (eds) Innovation Measurement and Policies, Luxembourg: European Commission, pp. 114–120. Caloghirou, Y. and Vonortas, N. (2000) ‘Science and technology policies towards research joint ventures’, Final Report to the European Commission of the TSER Project ‘Science and Technology Policies Towards Research Joint Venture’, DGXIII, Contract Number SOE 1-CT97-1075. Calvert, J., Ibarra, C., Patel, P. and Pavitt, K. (1996) Innovation Outputs in European Industry: Analysis from CIS, EIMS Publication No. 34, Luxembourg: EIMS. Christensen, J., Rama, R. and von Tunzelmann, N. (1997) ‘The European food products and beverages industry’, in A. Arundel and R. Garrelfs (eds) Innovation Measurement and Policies, Luxembourg: European Commission, pp. 176–181. Cohen, W., Nelson, R. and Walsh, J. (2002b) ‘Links and impacts: the inﬂuence of public research on industrial R&D’, Management Science, 48: 1–43. Cohen, W., Goto, A., Nagata, A., Nelson, R. and Walsh, J. (2002a) ‘R&D spillovers, patents and the incentive to innovate in Japan and the United States’, Research Policy, 31(8–9): 1349–1367. Czarnitzki, D., Ebersberger, B. and Fier, F. (2004) ‘The relationship between R&D collaboration, subsidies and patenting activity: empirical evidence from Finland and Germany’, ZEW Discussion Paper No. 04-37, ZEW, Mannheim. Dosi, G., Llerena, P. and Labini, M.S. (2005) ‘Science-technology-industry links and the “European paradox”: some notes on the dynamics of scientiﬁc and technological research in Europe’, Working paper, Santa Anna School of Advanced Studies, Pisa. Dumont, J.-C. and Lemaitre, G. (2004) ‘Counting immigrants and expatriates in OECD countries: a new perspective’, OECD, Directorate for Employment Labour and Social Affairs, Paris. European Commission (EC) (2001) ‘Benchmarking industry–science relations: the role of framework conditions’, Final report, DG Enterprise, Brussels. Francois, J.-P. and Favre, F. (1998) Technological Innovation is Progressing in Industry: Le 4 Pages des Statistiques Industrielles, Paris: SESSI. Geroski, P. (1995) ‘Markets for technology: knowledge, innovation and appropriability’, in P. Stoneman (ed.) Handbook of the Economics of Innovation and Technological Change, Oxford: Basil Blackwell, pp. 90–131. Grupp, H. (1998) Foundations of the Economics of Innovation: Theory, Measurement and Practice, Cheltenham: Edward Elgar.

66 A. Arundel and A. Constantelou Jacobsson, S., Oskarsson, C. and Philipson, J. (1996) ‘Indicators of technological activities: comparing educational, patent and R&D statistics in the case of Sweden’, Research Policy, 25: 573–585. Laursen, K. and Salter, A. (2003) ‘Searching low and high: what types of ﬁrms use universities as a source of innovation?’, Paper presented to the DRUID summer conference, ‘Creating Sharing and Transferring Knowledge’. Levin, R., Klevorick, A., Nelson, R. and Winter, S. (1987) ‘Appropriating the returns from industrial research and development’, Brookings Papers on Economic Activity, 3: 783–820. Leydersdorff, L. and Scharnhorst, A. (2003) ‘Measuring the knowledge base: a program of innovation studies’, Report for the Förderinitiative ‘Science Policy Studies’ of the German Bundesministerium für Bildung und Forschung. Amsterdam, March. Loof, H. and Brostrom A. (2004) ‘Does knowledge diffusion between university and industry increase innovativeness?’, CESIS Electronic Working Paper Series No. 21, CESIS, Stockholm. Malerba F., Lissoni, F. and Torrisi, S. (1997) ‘Computer and Ofﬁce Machinery – Firms External Growth and Technological Diversiﬁcation: Analysis During CIS’, EIMS Publication No. 47, Luxembourg: EIMS. Mansﬁeld, E. (1991) ‘Academic research and industrial innovation’, Research Policy, 20: 1–12. Mohnen, P. and Therrien, P. (2004) ‘Comparing the innovation performance of Canadian ﬁrms and those of selected European countries: an econometric analysis’, in F. Gault and L. Earl (eds) Innovation in Canadian Industry, Montreal: McGill-Queens University Press, pp. 313–339. Nas, S.O., Sandven, T. and Smith, K. (1994) ‘Inovasjon og teknologi I norsk industri: En oversikt (Innovation and New Technology in Norwegian Industry: An Overview)’, STEP Report 4/94, Oslo: STEP. OECD (1997) ‘National innovation systems’, Paris: OECD. OECD (2003) ‘OECD science, technology and industry scoreboard’, Paris: OECD. OECD (2004) ‘Compendium of patent statistics’, Paris: OECD. Schmidt, U. (1997) ‘Innovation in small ﬁrms’, in A. Arundel and R. Garrelfs (eds) Innovation Measurement and Policies, Luxembourg: European Commission, pp. 222–228. Smith, K. (2004) ‘Measuring innovation’, in Fagerberg, J. (ed.) The Oxford Handbook of Innovation, pp. 148–178. Sternberg, L., Gustafsson, E. and Marklund, G. (1996) ‘Use of human resource data for analysis of the structure and dynamics of the Swedish innovation system’, Research Evaluation, 6: 121–132. Tether, B.S. (2002) ‘Who co-operates for innovation and why?: an empirical analysis’, Research Policy, 31: 947–967. Veugelers, R. and Cassiman, B. (1999) ‘Make and buy in innovation strategies: evidence from Belgian manufacturing ﬁrms’, Research Policy, 28: 63–80. WZB (1996) ‘Innovation in the European Chemical Industry’, EIMS Publication No. 38, Luxembourg: EIMS.

4

An operational framework for the study of knowledge ﬂows Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas1

4.1 Introduction A major challenge for researchers in the empirical investigation of knowledge ﬂows is to design questions that probe some of the more difﬁcult aspects of these ﬂows that were identiﬁed in Chapter 3. The contributors to this book have in various ways tried to address this. A necessary ﬁrst step towards doing so is to decompose the concepts and constructs discussed in Chapter 2 in such a way that they can be empirically detected and assessed. Chapter 4 tackles this issue. It operationalizes the concepts discussed so far with the aim of identifying insightful ways of framing survey questions about knowledge ﬂows that would be of interest to managers and policy makers in the public and private sectors. Diffusion and absorption of knowledge within the innovation system is a complex process. Figure 2.1 in Chapter 2 shows that the transfer of knowledge takes place within a network of differentiated units of varying size, nature and role. The study of the knowledge ﬂows occurring within such a network can be conducted at three different levels of analysis: the nodal level (that is, concentrating on the behaviour of individual units), the dyadic level (that is, concentrating on the joint behaviour of two units which transact with one another) and the systemic level (that is, studying the behaviour of the entire network) (Gupta and Govindarajan 2000: 474). For practical reasons, when designing a survey instrument, investigation usually has to be limited to the nodal level, that is, to the level of an organization.2 This is because the type of information that can usually be obtained from surveys is generally under the control of senior staff, either in their heads or in their keeping (Padmore et al. 1998). The emphasis on ﬁrm behaviour allows for an investigation of the knowledge inﬂows from all other nodes in the network. At the level of the organization, innovation occurs as a result of existing or newly acquired knowledge being utilized. Identifying where the knowledge comes from (the knowledge sources) is important for an organization’s strategy and is useful for policy makers. At an operational level, transmission channels facilitate the ﬂow of knowledge from one unit to

68 Y. Caloghirou, A. Constantelou and N.S. Vonortas another. Also, the properties of the transmission channels within and outside the boundaries of an organization, such as, for example, the level of their integration into an overarching structure or the level of their informality and openness, are likely to affect their performance. Last, but not least, identifying the type of knowledge being transferred allows for a better understanding of an organization’s needs, in terms of knowledge, to complement its activities. In the following paragraphs we discuss each of these components.

4.2 The sources of knowledge Of particular interest to the study of innovation are the sources of information and/or knowledge for innovation. The Oslo Manual (OECD 1997) identiﬁes sources of information both internal and external and classiﬁes them into the following broad categories: internal sources within the ﬁrm or business group, external market/commercial sources, education/research institutions and generally available information (OECD 1997: 71). The approach proposed here is somewhat different in that it identiﬁes actors (individuals and institutions) as being the repositories of knowledge. These actors also act as dispatchers of knowledge. Below we describe the main actors/sources of knowledge identiﬁed. 4.2.1 Individuals Firm employees working in-house in the various departments and interacting with each other play a signiﬁcant role in the ﬂow of knowledge for innovation. Also, individual customers can be an important source of marketing information for the ﬁrm through the feedback they provide on their tastes and needs. Expert professionals and consultants may bring valuable knowledge into the ﬁrm, while inventors and individuals who act as ‘technology gatekeepers’ and keep abreast of new developments in scientiﬁc and technological ﬁelds close to the ﬁrm’s knowledge base, also facilitate knowledge ﬂows. 4.2.2 Other ﬁrms This category is very broad and encompasses a variety of suppliers, customers, competitors, consulting ﬁrms and collaborators. Supplier ﬁrms are important since some types of knowledge may be impossible to strip out from the products in which they are embodied. Embodied knowledge can be acquired through purchase of the product in which the knowledge is embedded. Therefore, supplier ﬁrms often provide ﬁrms with knowledge embodied in the goods and services purchased. Similarly, customer ﬁrms often develop close user–producer relationships with their suppliers, which allow valuable knowledge to be trans-

A framework for the study of knowledge ﬂows 69 ferred. This is particularly so in the case of large ﬁrms which often dictate to their suppliers about the appearance, quality and performance speciﬁcations of the products they want to purchase from them thereby providing another means for ﬁrms to gain information required for innovation. Also ﬁrms can learn from their competitors in a variety of ways, including copying competitors’ products and reverse engineering. Formal contractual agreements among ﬁrms are opening up additional routes for the ﬂow of innovation-related knowledge. Depending on their type, these contractual agreements may involve one-way or two-way knowledge ﬂows between ﬁrms. Typical examples of such agreements include arm’s length buy/sell contracts, franchising, licensing and cross-licensing. In addition, ﬁrms interact with other ﬁrms in the context of cooperative agreements that may involve the development of a new product, joint research and development (R&D) efforts, joint manufacture of a product line, cooperation in marketing and logistics, etc. Participation in such cooperative agreements can be an important source of innovation-related knowledge for ﬁrms. 4.2.3 The academic sector (universities and public research institutes) Traditionally, universities have been regarded as repositories of knowledge and thus are a critical source of information for innovation. More precisely, their role in the innovation system is twofold. First, through their education and training programmes they provide ﬁrms with qualiﬁed scientists and engineers. Second, in the course of their research activities they make signiﬁcant contributions to the innovative activities of ﬁrms as a result of R&D spillovers. 4.2.4 Government agencies Government agencies cover a wide array of public institutions that have an indirect impact on the innovative activities of ﬁrms. Their primary role is to set and oversee the rules of the economic, business and social environment within which ﬁrms operate. By determining the rules of the game these agencies often act as ‘key customers’ who pass on to ﬁrms the information necessary for them to meet the standards and objectives laid down by the customer. Typical examples of agencies of this sort include regulatory commissions, standards-setting institutes and government departments dealing with public procurement contracts. Patent ofﬁces by deﬁnition facilitate the transfer of knowledge. In the same way, government departments responsible for the design and management of policy initiatives related to the diffusion of knowledge, such as R&D collaborative programmes, also act as facilitators in the transfer of knowledge, thereby indirectly affecting the innovative behaviour of ﬁrms. In recent years, the proximity between the source of knowledge and

70 Y. Caloghirou, A. Constantelou and N.S. Vonortas the recipient organization has been seen as being of particular advantage to knowledge ﬂows. In particular, the literature suggests that geographical proximity between university and industry increases the beneﬁts to ﬁrms of academic research. However, there are various dimensions of such proximity that merit investigation. First, geographical proximity refers to the level of spatial closeness between the transacting organizations. Various levels of geographic proximity can be identiﬁed, for example, proximity within a district, a region or a country, or proximity to a neighbouring country. Second, there may be cultural proximity among entities, which implies common perceptions, values and beliefs about business conduct and modes of behaviour. Finally, there may be proximity of the knowledge base, that is, proximity in terms of the nature of the knowledge that underpins ﬁrms’ innovative activities. This type of proximity is usually identiﬁed through the application of econometric techniques to databases containing innovation-related performance indicators (patents, participation in research consortia, patent citations, etc.).

4.3 The channels of knowledge Knowledge ﬂows among different units require the existence of transmission channels. A channel is understood as the means by which knowledge is transferred. We can distinguish six different types of channel. Written: This category refers to a range of codiﬁed forms of communication and includes access to journals, papers, reports, patents, licences, copyrights, etc. Verbal: This category covers interpersonal forms of communication that facilitate knowledge transfer, such as takes place in meetings, conferences, telephone conversations, etc. Electronic: This is a new category that has appeared as the result of the expansion of electronic means of communication within and among organizations. It includes all forms of communication (written or verbal) that are transposed into an electronic environment. Typical examples include e-mail exchanges, electronic newsletters addressed to groups of practitioners, interactive websites, etc. The detection of this type of channel opens up new areas of investigation regarding the potential of electronic forms of communication vis-à-vis more traditional channels. Transfer of personnel: This is by far the most important channel for knowledge transfer partly because knowledge transferred in this way is mainly tacit and would therefore be difﬁcult to acquire by other means. Personnel might move from one department to another within a ﬁrm through formal or informal procedures and temporary transfers. Hiring skilled people from the market is another way of personnel transfer. Also, formal and informal training in house are channels for knowledge acquisition common to most ﬁrms.

A framework for the study of knowledge ﬂows 71 Transfer of a product or service: This category includes all forms of embodied knowledge that are transferred, that is, knowledge obtained by ﬁrms from their suppliers or competitors through the purchase of a technological product (equipment, material) and/or service (piece of software or software-based instrument). Joint practice: This channel refers to formal and informal methods of technological collaboration among individuals belonging to the same or different groups, departments, ﬁrms or organizations. Normally, the greater the familiarity and the more socializing that occurs between personnel from different units, the greater is their willingness to become engaged in joint practice. However, the density of communication and the richness of the knowledge exchanged in joint practice (both in quantitative and qualitative terms) depend on a number of factors beyond the type of channel used, for example, the strategic and competitive positioning of participating entities, the scope of joint practice, the technological knowledge base of the industry concerned and the cognitive perceptions of the agents involved (Dodgson 1993). Thus, formal collaborative efforts, research joint ventures (RJVs), participation in task forces and permanent committees and informal teamwork all fall into this category.

4.4 Properties of channels The properties of the channels identiﬁed so far are expected to affect the performance of channels as well as the extent of knowledge ﬂows. The following types of properties can be identiﬁed and empirically detected using qualitative techniques, such as case studies and participant observation. 4.4.1 Authority structure The distinction here is between a command structure, where one agent just tells another to transmit information, and a voluntary transfer structure, where agents are free to choose whether to transfer knowledge. The ﬁrst type of structure usually applies in hierarchical structures. 4.4.2 Internalized/non-internalized This concerns whether information ﬂows between agents with a common objective or with different independent objectives. Thus, an internalized mechanism can either be a single ﬁrm operating on non-hierarchical lines, or a group of ﬁrms acting under a cooperative agreement. This distinction only matters when the authority structure is voluntary since, by deﬁnition, a hierarchical structure is internalized.

72 Y. Caloghirou, A. Constantelou and N.S. Vonortas 4.4.3 Priced/unpriced Knowledge can be made available either for free, or for a fee. The cost of knowledge acquisition is different from the costs of using a particular channel. While this distinction can apply whether the mechanism is hierarchical, internalized or non-internalized, it only really signiﬁes in the case of non-internalized mechanisms. 4.4.4 Restricted/unrestricted access This refers to whether the information that is made available is known to be available to everyone, or to only a limited number of entities (individuals, ﬁrms). Table 4.1 presents a typology of the different mechanisms put in place for the ﬂow of knowledge and information based on the transmission channel(s) they use and the properties of these channels. The table presents a matrix of channels and channel properties with the various mechanisms of knowledge and information ﬂows in the appropriate cells. The majority of mechanisms in the table operate at the boundaries between ﬁrms and other entities. At the same time, in most cases, the information and/or knowledge available within the ﬁrm is restricted. The only unrestricted mechanisms identiﬁed, where information is available to everyone at no cost to the individual, relate to public sources of information, such as the Internet, patent databases, access to scientiﬁc journals and open conferences.4 The proposed typology of knowledge ﬂow mechanisms contributes to the development of a methodological framework for the empirical study of knowledge ﬂows at the organization level. The importance of each mechanism depends, of course, on the value assigned to it by respondents within each organization. It may be that the same mechanisms are valued differently by different organizations, either because they contribute to different stages in the innovation process or because of the nature of knowledge that underpins a particular organization’s innovative activities.

4.5 Types of knowledge acquired Another important aspect in the empirical investigation of knowledge ﬂows is the type of knowledge exchanged. This describes what the knowledge is about or alternatively how the knowledge is going to be used by the organization in its innovative activities. Typical knowledge types include: Marketing knowledge: This refers to knowledge that helps ﬁrms deﬁne more clearly possible proﬁtable types of innovation.

Internal meetings

Verbal

Joint practice

Product

Internal product exchange Project meetings Team work

Electronic Intra-ﬁrm e-mails Personnel Internal staff transfer In-house training

Internal reports

Internalized unpriced restricted

Hierarchical

Properties

Written

Channel

■

Consultation

RJV reports etc. Internal meeting RJV meeting

Inter-ﬁrm e-mails News alerts Internal staff transfer Staff poaching External staff exchange RJV single lab Formal training Internal product exchange Reverse engineering RJV product exchange Observations Project meetings Team work

Consultancy report

Non-priced restricted

Internal reports

Internalized unpriced restricted

Non-hierarchical

Table 4.1 A typology of knowledge ﬂows in an organizational context

Patents

Priced unrestricted

Scientiﬁc journal Closed meeting Open conference Closed conference Telephone call Electronic newsgroups Internet and web access Quits/hires Informal learning processes

Limited circulation paper

Internalized unrestricted

74 Y. Caloghirou, A. Constantelou and N.S. Vonortas Scientiﬁc knowledge: This refers to the knowledge acquired by the ﬁrm’s own R&D department in a bid to generate new processes/products, for example, information about new developments in genetic engineering might open up new ways in which drugs can be produced. Technological knowledge: This is knowledge about new products and/or processes that the ﬁrm may acquire from other ﬁrms and which can either duplicate or supplement its own R&D efforts. Strategic knowledge: This type of knowledge informs the ﬁrm about the progress of its competitors in issues that concern the ﬁrm’s strategies towards future actions and areas of interest.

4.6 Organizational characteristics and knowledge ﬂows An empirical investigation of knowledge ﬂows at the organizational level should take into account that the extent and intensity of knowledge ﬂows also depend on organizational characteristics such as: • • • • • • •

ownership status; corporate culture and type of management; strategic orientation; means of internal communication; corporate strategy; technology strategy; absorptive capacity.

Absorptive capacity is the ability of an organization to make use of knowledge acquired from external sources. The following could be regarded as proxy indicators of an organization’s absorptive capacity: • • • • •

number and skills of scientiﬁc and technical personnel; time spent and methods used by scientiﬁc and technical personnel to keep informed of technical developments outside the organization; investments in training programmes; organization’s own parallel efforts to keep abreast of scientiﬁc and technical developments; level of ICT use and sophistication.

4.7 Conclusion In this chapter we have made an effort to operationalize into more concrete and empirically identiﬁable elements the key constructs of the analytical framework for the study of knowledge ﬂows discussed in Chapter 2. The effort is mainly conceptual and reﬂects upon what participants in the ‘KNOW for Innovation’ research project considered to be a useful framework for the empirical investigation and quantitative measurement of knowledge ﬂows.

A framework for the study of knowledge ﬂows 75 Participants in the KNOW research consortium used this framework to devise a new survey instrument designed to collect information on the sources, channels, and speciﬁc mechanisms supporting innovation-related knowledge ﬂows in selected sectors of European industry. The results of this effort are discussed in detail in Chapter 5.

Notes 1 This chapter has beneﬁted extensively from comments from all partners in the ‘KNOW for Innovation’ project. In particular, the authors wish to acknowledge the contribution of Professor David Ulph for his insightful comments on the ﬁrst version of this chapter. 2 In the ‘KNOW for Innovation’ project we chose to focus on the ﬁrm level. By ‘ﬁrm’ we mean a physical site or establishment. The ﬁrm can thus be a division or subsidiary of a large company or a subsidiary of a multinational. 3 The assumption here is that institutions and organizations cover the cost of access to information.

References Gupta, A. and Govindarajan, V. (2000) ‘Knowledge ﬂows within multinational corporations’, Strategic Management Journal, 21: 473–496. Dodgson, M. (1993) Technological Collaboration in Industry: Strategy, Policy, and Internationalization in Innovation, London: Routledge. OECD (1997) OSLO Manual: Proposed Guidelines for Collecting and Interpreting Technological Innovation Data, Paris: OECD. Padmore, T., Schuetze, H. and Gibson, H. (1998) ‘Modelling systems of innovation: an enterprise-centred view’, Research Policy, 26: 605–624.

5

Knowledge ﬂows in European industry An overview of evidence from surveys and case studies Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas1

5.1 Introduction The information available in the research joint ventures (RJV) patent citation databases used in other chapters of this book is extremely valuable for characterizing innovation-related knowledge ﬂows in European industry, for empirically estimating the effect of knowledge spillovers and interﬁrm linkages on ﬁrm innovative capabilities and competitiveness, and for addressing the spatial aspect of knowledge ﬂows. These data sources do not, however, allow investigation of other issues that may be important to policy decision making. For example, the available databases do not enable investigation of how knowledge ﬂows are triggered in individual ﬁrms, the purpose to which the knowledge is put – for example to suggest an innovation project or to complete one in progress, or the organizational factors that determine the effectiveness of the information transfer (including learning processes in individual organizations). Most existing surveys measure information ﬂows, but information must be turned into knowledge in order to be of use to a ﬁrm’s innovative activities. It is this process of turning the information, including both its formal and its tacit elements, into knowledge that is generally referred to as learning. Mastering the process of learning is vital for all organizations. The issue is of critical importance to policy since simply enhancing the ﬂow of information may not result in equivalently enhanced levels of knowledge. It is the latter that policy makers ultimately aim for. It was for these reasons that the consortium of research teams working on the KNOW project decided to undertake a focused survey of ﬁrms in selected industries to gather supplementary information on the speciﬁc mechanisms and channels that support innovation-related knowledge ﬂows. In combination with several company interviews, the focused survey was intended to facilitate an in-depth examination of why ﬁrms use speciﬁc formal and informal knowledge ﬂows, of the relative importance of internal and external transfer mechanisms, of the institutional and organi-

Knowledge ﬂows in European industry 77 zational environment in which transfers materialize and of the procedures that facilitate learning. This chapter presents the ﬁrst round of ﬁndings from a combined survey and interview exercise conducted in the realm of KNOW among innovative small- and medium-sized enterprises (SMEs) in seven European countries operating in ﬁve selected sectors. The results allow for a ﬁrst assessment of ﬁrms’ experiences with knowledge ﬂows in relation to innovative activities, and identify differences and similarities in the behaviour of ﬁrms across Europe.

5.2 The KNOW survey database A new focused survey of ﬁrms in selected industries was carried out during the course of the KNOW Research Project to gather supplementary information on the speciﬁc internal and external mechanisms and institutions that support innovation-related knowledge ﬂows and of the procedures that facilitate learning. Five manufacturing and service sectors were selected as representative of the European industry including: food and beverages (NACE 15), chemicals (excluding pharmaceuticals) (NACE 24 excluding 24.4), communication equipment manufacturing (NACE 32), telecommunication services (NACE 64.2) and computer related activities (NACE 72).2 The survey covered Denmark, France, Germany, Greece, Italy, the Netherlands and the UK. The KNOW survey: •

•

•

obtained information on the percentage of each ﬁrm’s new or improved product innovations that were developed in-house through buying in, or via collaboration with other divisions of the same ﬁrm or with independent ﬁrms or PROs; obtained information on each ﬁrm’s use of three appropriation methods – secrecy, patents and lead times – and asked about which method was the most important to the ﬁrm for protecting its innovations; distinguished between research spending in-house, in other divisions of the ﬁrm and external to the ﬁrm. It also asked about the number of employees with an academic qualiﬁcation in science or engineering.

To ensure comparability among countries it was decided at the outset to include ﬁrms with between ten and 1,000 employees. Two size classes were used: one for small ﬁrms with 10–250 employees and one for ﬁrms with 251–1,000 employees. The information was collected by computer aided telephone interviews (CATI). Table 5.1 summarizes the effort and the achieved response rates by country. Although the design of the questionnaire allowed the collection of data from non-innovators as well, their share in the returned responses turned out to be disproportionately low. These companies were consequently

78 Y. Caloghirou, A. Constantelou and N.S. Vonortas Table 5.1 KNOW survey – response rates Country

Number contacted

Responses

Response rate (%)

Innovators

Greece Italy Denmark The UK France Germany The Netherlands

260 278 170 1,003 613 470 331

110 97 130 96 79 101 151

42.3 34.9 76.5 9.6 12.9 22.0 45.6

100 80 98 46 76 94 138

Total

3,017

764

25.3

632

Table 5.2 KNOW survey – country and sector distribution of sample Country

Sectors 15

Germany France Italy The Netherlands The UK Denmark Greece Total % Weighted %

% 24

18 13 19 35 6 20 21

16 12 20 33 19 11 21

132 23.7

132 23.7

48.2

17.9

32 21 14 13 13 6 15 19

64 4 13 2 4 4 11 10

101 48 18.1 8.6 8.0

2.0

72

Weighted %

Total

20 13 24 29 9 21 29

79 65 78 114 44 78 100

14.2 11.6 14.0 20.4 7.9 14.0 17.9

145 26.0

558 100

100.0

23.9

100

36.0 21.2 24.9 3.3 5.8 6.9 2.0

excluded from subsequent analysis. Responses from ﬁrms that ex post proved to belong to non-targeted sectors were also excluded. A signiﬁcantly larger number of ﬁrms ultimately were outside the size range that had initially been set, so in order to increase the total sample for the empirical analysis it was decided to increase the upper limit by 25 per cent, thereby including ﬁrms with up to 1,250 employees. A total of 558 respondent ﬁrms met the criteria of an innovator and belonged to the selected sectors and size class. Table 5.2 shows the country and sectoral distribution of the ﬁnal sample, both unweighted and weighted, in terms of employment. The set of responding ﬁrms in Table 5.2 were classiﬁed as innovative, that is, ﬁrms that had introduced one or more innovations in the last three years. More than half (55 per cent) of the set correspond to small ﬁrms (<50 employees) and 40 per cent correspond to middle-sized and larger ﬁrms (>250 employees). More than a quarter (27 per cent)

Knowledge ﬂows in European industry 79 reported research and development (R&D) intensity of between 10 and 25 per cent; about a quarter reported R&D intensity between 1 and 5 per cent. More than 40 per cent had never cooperated with a partner outside the private sector (universities, PROs); about a third had cooperated three or more times with a non-private sector partner. About 9 per cent of these ﬁrms did not employ scientists. In 16 per cent of them, numbers of scientiﬁc personnel outweighed low skilled personnel.

5.3 The in-depth interviews The results of the survey were greatly enriched by insights from in-depth interviews with 71 of the 558 companies. The decision about which ﬁrms to interview was not made randomly. The aim was to have equal representation of all ﬁve sectors examined and of seven countries involved. Thus, interviews were allocated as two per sector per country. These in-depth interviews expanded the available information from the survey on the following subjects: • •

• •

• •

extent of patent use to obtain ideas for innovation as well as a means for protecting proprietary information; extent of Internet use in the search for scientiﬁc and technological information, information regarding market dynamics and consumer behaviour, as well as a means of communication with suppliers, customers, collaborators; collaboration with universities and its beneﬁts; the ways in which the source of information indicated in the survey as being the most important, interacted with the ﬁrm regarding a speciﬁc innovation; evaluation of internal versus external sources of information; research cooperation: types of partnerships, types of partners, perceived returns, problems in managing the partnership(s).

5.4 Descriptive analysis of knowledge ﬂows Five hundred and ﬁfty-eight ﬁrms were surveyed regarding their experience with knowledge ﬂows, ﬁrst with respect to their general innovation activities and second, with respect to their economically most important innovation in the last three years. Figure 5.1 depicts the signiﬁcant differences between sectors across all countries and between countries across sectors in terms of R&D orientation of the surveyed ﬁrms. Figure 5.2 shows that attending trade fairs and conferences, and scientiﬁc and business journals are the most important sources of new ideas for innovation in the countries examined. About half the innovating ﬁrms perform technical analyses of competitors’ products (reverse engineering) to gain new concepts for their own

80 Y. Caloghirou, A. Constantelou and N.S. Vonortas

nm

UK

ce ee

ar

k

Gr

UK

nm

De

s

ly

nd rla

the Ne

Fra

nc

y an rm Ge

ee Gr

ar nm

De

nd

Ita

rla

Ne

the

nc Fra

an rm

Continuous R&D activities

Ita

0% ce

20%

0% k

40%

20% UK

40%

s

60%

ly

80%

60%

e

80%

y

100%

e

ICT services

100%

Ge

ar k Gr ee ce

Ge

nm

De

y Ne

th

er lan

ce

Ita l

an

Fra n

rm Ge

Telecom equipment

De

0% rm

20%

0%

ar k Gr ee ce

40%

20%

UK

60%

40%

ds

80%

60%

y

80%

Ne Italy the rla nd s

Chemical sector 100%

an y Fra nc e

Food and beverages sector 100%

Occasional R&D activities

No R&D activities

Figure 5.1 R&D activities (source: KNOW Survey).

innovations. Danish and Dutch chemical companies report this most often. With the exception of Germany and France, reverse engineering is most important for chemicals ﬁrms. In Germany and France the telecommunication and computer services (ICT services) sector favours the practice of reverse engineering. Searching patent databases for creative ideas is not very popular. Firms generally ﬁnd it too time consuming and not sufﬁciently rewarding. Only ﬁrms located in the Netherlands seem to use this method frequently: more than one-third of ﬁrms reported searching patent databases regularly, especially in the chemicals sector where about 70 per cent of innovators used this method. Overall, about 80 per cent of ﬁrms use the Internet regularly in everyday business. Only in the food and beverages sector is the dissemination of Internet technology still poor: this sector had the lowest penetration of ICT in all seven countries (Figure 5.3). Most ﬁrms use the Internet both for searching for scientiﬁc and technical information and for communicating with other companies. However, the possibility of communication via the Internet seems to be the ﬁrst step into the virtual world. In most

Knowledge ﬂows in European industry 81

0%

Ne

th

rm Ge

ee

m

Gr

UK

De n

er Ne

th

rm

Chemicals Computer/telecom services

Food and beverages ICT equipment

Technical analysis of competitors’ products 100%

80%

80%

60%

60%

40%

40%

20%

20%

x

ce

k ar

ee Gr

UK

nm

lan

th Ne

De

ds

ly Ita

nc Fra

Ge

rm

an

y

ce

ar

ee

k

Gr

UK

nm De

ds lan

Ne

th

er

e

ly Ita

Fra

nc

y an rm

Food and beverages Telecom equipment

e

0%

0%

Ge

Chemicals ICT services

Reading scientific or business journals

100%

er

Ge

Food and beverages Telecom equipment

U De K nm ar k Gr ee ce

0%

er

20%

ce

40%

20%

ar k

40%

lan ds

60%

Ita ly

80%

60%

an y Fra nc e

80%

Ita ly lan ds

Searching patent databases 100%

an y Fra nc e

Attending trade fairs and conferences 100%

Chemicals Computer/telecom services

Food and beverages ICT equipment

Chemicals ICT services

Figure 5.2 Sources of new ideas for innovation (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

ﬁrms, e-mail communication has already replaced traditional communication channels like regular mail or fax. In Greece, the Internet is used by only 58 per cent of ﬁrms – the lowest dissemination of this technology in the countries in the sample. In the Netherlands and Denmark almost 100 per cent of ﬁrms use the Internet. In most countries, secrecy is the preferred protection strategy (Figure 5.4). More than 80 per cent of German innovators favour this strategy. In Italy and Greece lead-time advantages are most important, and patenting is less frequently used. While in the Netherlands four out of ten ﬁrms apply for patents when innovating, in Greece only one in ten does so. In the German and Dutch chemicals sectors patents are of high importance. British and Italian manufacturers of communication equipment as well as

82 Y. Caloghirou, A. Constantelou and N.S. Vonortas

20%

0%

0%

th Ne

Ge

er Ne

th

rm Ge

U De K nm ar k Gr ee ce

20%

er

40%

an y Fra nc e

60%

40%

rm

60%

U De K nm ar k Gr ee ce

80%

Ita ly lan ds

100%

80%

an y Fra nc e

100%

Ita ly lan ds

Searching for scientific and technical information

General use of the Internet

Exchanging information with customers, suppliers or collaborators by e-mail 100%

Food and beverages Telecom equipment

80% 60%

Chemicals Computer/telecom services

40% 20% x

ce

k

ee

ar nm

Gr

UK De

ds lan

er Ne

th

e

ly Ita

nc Fra

Ge

rm

an

y

0%

Figure 5.3 Internet use (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

French and British ICT service ﬁrms use patents frequently. Dutch ﬁrms rely on lead-time advantages less than ﬁrms in other countries. Secrecy is the favoured knowledge protection strategy in the Netherlands. The economically most important innovations are usually new products or the combination of new products and processes (Figure 5.5). Process and service innovations are less frequently mentioned by ﬁrms as being economically most important. In Italy 25 per cent of ﬁrms regard a new process as most valuable, while only 5 per cent of French ﬁrms do so. Figure 5.6 shows the sources used for hiring new scientists by countries and industries. About 50 per cent of ﬁrms in France and the Netherlands hired highly skilled personnel from other divisions or units of their own companies. In France, suppliers and customers are more frequently used as a source of qualiﬁed employees than in other countries: more than 40

Knowledge ﬂows in European industry 83

20%

0%

0%

rm

Ne

Ne

Ge

er th

rm Ge

er

20%

th

40%

an y Fra nc e

60%

40%

U De K nm ar k Gr ee ce

60%

Ita ly lan ds

80%

an y Fra nc e

80%

U De K nm ar k Gr ee ce

Secrecy 100%

Ita ly lan ds

Patents 100%

Lead time advantages 100%

Food and beverages Telecom equipment

80% 60%

Chemicals Computer/telecom services

40% 20%

ce

k ar

ee Gr

nm De

ds lan

er Ne

th

e

ly Ita

nc

Fra

y an rm Ge

UK

x

0%

Figure 5.4 Methods to protect innovations by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

per cent of French innovators hired personnel from their suppliers and 20 per cent from their customers. Universities or PROs are utilized the most by Italy with a proportion of 46 per cent. In the UK, Denmark and Greece only one in every ten ﬁrms hired personnel from universities or PROs to work on the most important innovations. Italy is the leading country in terms of acquiring new personnel directly from universities or PROs. Figures 5.7 and 5.8 clearly show that the economically most valuable innovations are driven by demand: customers dominate as the source of original ideas for innovations. This result was conﬁrmed by the in-depth interviews (reported later), where many interviewees mentioned that their ﬁrms implement the lead-user concept in their innovation projects. Only Italian ﬁrms reported that suppliers and competitors are more important than customers. Nevertheless, competitors seem to be a reasonable source of innovation for all countries, a fact supported by the importance of reverse engineering mentioned above. Supplier, and also universities or PROs and consultancies, are relatively more important in the Netherlands.

84 Y. Caloghirou, A. Constantelou and N.S. Vonortas Product innovation 100%

Process innovation 100%

x

80%

80%

60%

60%

40%

40% x

20%

0%

lan ds

Ita ly

th

x

Ne

Combined product and process innovation 100%

Service innovation 100%

x

80%

60%

60%

40%

40% 20%

k

ee

ar

lan

Gr

ds

ly Ita

er Ne

th

e nc Fra

Ge

rm

an

y

ce ee

k ar nm

Gr

UK De

ds lan

Ne

th

er

e

ly Ita

Fra

nc

y an rm

Food and beverages

ce

x x

0%

UK

x

0%

nm

20%

De

80%

Ge

er

an y Fra nc e

rm Ge

U De K nm ar k Gr ee ce

lan ds

Ita ly Ne

th

er

an y Fra nc e

rm Ge

x

0%

U De K nm ar k Gr ee ce

20%

Chemicals

Telecom equipment

Computer/telecom services

Figure 5.5 Most important innovation by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

In addition to serving as frequent sources of original ideas, customers and suppliers are most frequently mentioned as being important contributors to innovation completion (Figure 5.8). Customers or suppliers account for a very large percentage of innovation completion in Dutch ﬁrms’ projects, followed by French and Italian ﬁrms. Internal knowledge is valued highly in all countries for its contribution to innovation, and especially by German and British ﬁrms (Figure 5.9). Italy is ranked lowest; Italian ﬁrms seem to have the most balanced approach to internal versus external sources of innovation. Dutch ﬁrms seem to be more open to external sources of innovation than ﬁrms in the other six countries. They also seem more open to intercontinental sources of knowledge (see below). Globalization has for many years been a hot topic for business analysts and policy makers. The importance given to this topic is not, however, supported by the evidence on the location of important external sources

Knowledge ﬂows in European industry 85 Other divisions/units of own company 100% 80%

80% x

60%

60%

x

40%

x

40%

x

20%

20% x

U De K nm ar k Gr ee ce

lan ds

Ne

th

er

e

Ita ly

Fra nc

an y rm Ge

U De K nm ar k Gr ee ce

Ita ly

0%

Ne

th

er

an y Fra nc e

rm

lan ds

M

0%

Customers

Universities or public research institutes

100%

100%

80%

80%

60%

60% 40%

40% 20%

x

20%

x

ce

k ar

ee Gr

UK

nm

lan er

th Ne

De

ds

ly Ita

Fra

Ge

rm

an

y

ce

k

ee

ar

De

th Ne

Gr

UK

lan

nm

ds

ly Ita

er

e nc Fra

y an rm Ge

e

0%

0%

nc

Ge

Suppliers 100%

Consultants 100%

Food and beverages Telecom equipment

80% 60% x

40%

Chemicals Computer/telecom services

x

x x x

20%

ar k Gr ee ce

UK

nm De

Ne Italy th er lan ds

Fra nc e

Ge rm

an y

0%

Figure 5.6 Source of new scientists/engineers hired to work on most important innovation by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

of innovation-related knowledge: national sources still dominate (Figure 5.10). While this result is striking, the potential inﬂuence of size in this survey must be remembered: by and large, our sample is made up of small- and medium-sized companies. Firms in the smaller countries – Greece, Denmark and the Netherlands – tend to be more internationally orientated than those located in bigger countries. Greek chemical ﬁrms,

Competitors

Suppliers

100%

100%

80%

80%

60%

60%

40%

40%

20%

x

20%

x x

x

U De K nm ar k Gr ee ce

lan ds

Ne

th

er

e

Ita ly

Fra nc

an y rm Ge

U De K nm ar k Gr ee ce

lan ds

er Ne

th

rm Ge

Ita ly

0%

an y Fra nc e

0%

Universities or public research institutes

Customers 100%

100%

80%

80% 60%

ce

k ar

ee Gr

UK

nm De

ds lan

Ne

th

er

e

ly Ita

an

y

x x

rm Ge

Gr

De

Ne

th

ee

ar

UK

lan

nm

ly Ita

er

Fra

nc

an rm Ge

ce

0%

k

20%

0%

ds

20%

e

40%

y

40%

nc

x

Fra

60%

Consultants 100%

Food and beverages Telecom equipment

80%

Chemicals Computer/telecom services

60% 40% 20%

ar

ee ce

Gr

UK

nm De

Ne

Ita ly th er lan ds

nc e Fra

y an Ge rm

k

x x

0%

Figure 5.7 Contributors to the original idea for most important innovation by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

Competitors

Suppliers

100%

100%

80%

80%

60%

60%

40%

40% x

20%

x

Ne

U De K nm ar k Gr ee ce

lan ds

Ita ly

er Ne

th

rm

an y Fra nc e

0%

U De K nm ar k Gr ee ce

lan ds

Ita ly

er th

rm

an y Fra nc e

0%

Ge

20%

Customers

Universities or public research institutes

100%

100%

80%

80%

60%

60%

40%

40%

x

x

Ne

ce

k ar

ee Gr

UK

lan

nm De

ds

ly Ne

th

er

nc Fra

an rm Ge

e

x

y

ce ee

ar

De

th

Gr

UK

lan er

nm

ly Ita

nc Fra

an rm Ge

k

0%

ds

0%

e

20%

y

20%

Ita

Ge

x

Consultants 100%

Food and beverages Telecom equipment

80% 60%

Chemicals Computer/telecom services

40% x

20%

ar k Gr ee ce

nm De

nd s

ly

er la

Ne th

Ita

Fra nc e

an y Ge rm

UK

x

0%

Figure 5.8 Contributors to the completion of most important innovation by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

88 Y. Caloghirou, A. Constantelou and N.S. Vonortas Internal most important

External most important

100%

100%

80%

80% x

U De K nm ar k Gr ee ce

lan ds

Ita ly

er

e

th

rm Ge

Ne

Ne

th

er

Ita ly

Fra nc

rm Ge

x

Fra nc

0%

an y

0%

U De K nm ar k Gr ee ce

20%

lan ds

40%

20%

e

60%

40%

an y

60%

Internal and external of equal importance 100%

Food and beverages Telecom equipment

80% 60%

Chemicals Computer/telecom services

x

40% 20%

ce

k ar

ee Gr

UK

nm De

ds

ly Ne

th

er

lan

Ita

e nc Fra

Ge

rm

an

y

0%

Figure 5.9 Importance of internal versus external knowledge sources for successful completion by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

for example, mention companies elsewhere in Europe as their most important innovation source. Of the larger countries, about 30 per cent of French enterprises report ﬁrms in other European countries as being important sources, followed by British ﬁrms. The United States is mentioned by about 15 per cent of Dutch ﬁrms as a very important source of innovation-related knowledge. A great variety of reasons were given for obtaining knowledge from the most important external sources (Figure 5.11). The dominant reasons reported were to reduce development costs and risks, to increase the technical expertise of the ﬁrm and to build on the research ﬁndings of others. German, Dutch, French and British ﬁrms report a host of other reasons as well. Previous experience is seen as by far the most effective way of accessing the most important external source of knowledge, followed by participation in trade fairs and conferences (Figure 5.12). Business and professional associations seem to play a quite distinct role in that respect in the

Knowledge ﬂows in European industry 89 Same country

Europe 100%

x

80%

x

60%

0%

0%

Ge

er

x x

x

Ne

Ne

th

United States

Other countries

100%

100%

80%

80%

60%

60%

40%

40%

ce

k ar

ee Gr

UK

nm

lan

Ne

De

ds

ly er

Ge

th

an

nc

e

y

ce

k ar

ee Gr

UK

nm De

ds lan

Food and beverages

xx

x x

x

0%

Fra

x

Ne

th

er

e

ly Ita

nc Fra

rm

an

y

0%

Ge

20%

x x

rm

20%

Ita

rm Ge

x

rm

20%

U De K nm ar k Gr ee ce

20%

Ita ly lan ds

40%

an y Fra nc e

40%

U De K nm ar k Gr ee ce

x

Ita ly lan ds

x

er

60%

th

80%

an y Fra nc e

100%

Chemicals

Telecom equipment

Computer/telecom services

Figure 5.10 Location of most important external source of knowledge by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations.

UK. British followed by French, Dutch and Italian ﬁrms also use the Internet for this purpose. German ﬁrms seem to behave differently. Scientiﬁc and technical information is the dominant type of knowledge obtained from the most important external source, followed by knowledge relevant to market introduction (Figure 5.13). By far the most frequent method of communication with external sources of knowledge is informal personal contacts, followed by research cooperation (Figure 5.14). Exchange of personnel and other methods are used in some countries (for example, France and the Netherlands) more than in others. In-depth interviews expanded the information from the survey in several subjects. The responses are summarized here by subject area. •

Use of patents to obtain ideas for innovation. The majority of ﬁrms do not use patent databases to obtain ideas for innovation. None of the German or Dutch ﬁrms interviewed did. Reported reasons include:

To reduce development costs or risk

To update firm’s technical expertise

100%

100%

80%

80%

x

x

60%

x

x

60%

x

40%

xx x

40% 20%

20% M

U De K nm ar k Gr ee ce

lan ds

Ne

th

er

e

Ita ly

Fra nc

an y rm Ge

U De K nm ar k Gr ee ce

lan ds

Ita ly

er

0%

Ne

th

rm

an y Fra nc e

0%

To build on innovations or research findings of others

To meet government regulations

100%

100% x x x

80% 60%

x

nm

ar

UK

k

M

De

ds

ly

lan

x

Ne

th

er

nc

an rm Ge

e

MxM

y

ce

Gr

De

Ne

th

ee

ar

UK

lan

nm

ly Ita

er

Fra

nc

an rm

k

0%

ds

0%

e

20%

y

20%

Ge

xx

40%

xx

Fra

40%

ce

60%

ee

x x

Gr

80%

Ita

Ge

x

Other reasons 100% 80%

Food and beverages Telecom equipment

x x

60%

Chemicals Computer/telecom services

x

40%

x

20%

ar k Gr ee ce

nm De

Ne

Ita ly th er lan ds

nc e Fra

y an Ge rm

UK

M

0%

Figure 5.11 Underlying reason for decision to obtain knowledge from most important external knowledge source by country (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations, ‘M’ indicates a missing value.

Previous experience x

100%

x

x

80% x

th Ne

Trade fairs and conferences

The Internet

100%

100%

80%

80%

40%

x

20%

20%

xx

De

nm

ds

ly Ne

th

er

lan

Ita

e nc

Fra

y Ge

rm

an

ce

k

Gr

De

Ne

th

ee

ar

UK

nm

ds

ly

lan er

Fra

Ita

e nc

an rm Ge

xxxx

M

0%

y

0%

ce

xx

k

x

ee

x

ar

x

Gr

40%

x

60%

x x

UK

60%

lan ds

rm Ge

U De K nm ar k Gr ee ce

Ne

th

er

e

Ita ly

Fra nc

rm

xxx

xx

er

0%

e

20%

0%

an y

20%

lan ds

40%

an y

40%

Ge

x

60%

Ita ly

x

60%

Fra nc

80%

Business and professional associations

x

x

U De K nm ar k Gr ee ce

100%

Other way 100% 80%

Food and beverages Telecom equipment

x x

60%

Chemicals Computer/telecom services

x

40% x

20%

x

ar k Gr ee ce

nm De

nd s

ly

er la

Ne th

Ita

Fra nc e

an y Ge rm

M

UK

M

0%

Figure 5.12 Contacting the most important external source of knowledge (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations, ‘M’ indicates a missing value.

92 Y. Caloghirou, A. Constantelou and N.S. Vonortas Technical or scientific knowledge

60%

60%

40%

40%

20%

x

MMMM

lan ds

er

an y Fra nc e

rm

Ne

Ge

U De K nm ar k Gr ee ce

lan ds

Ita ly

x

0%

Ne

th

er

an y Fra nc e

rm

x x

x

20% MMMM

0%

Ge

x

80%

x

Ita ly

80%

Knowledge related to market introduction 100%

x

U De K nm ar k Gr ee ce

x

x

th

100%

Other type of knowledge 100%

Food and beverages Telecom equipment

80% 60% x

40%

x

20%

x

ce

k ar

ee Gr

UK

lan

nm

ds

ly Ita

er

M

Ne

th

e nc Fra

y an rm

M

De

MMMM

0%

Ge

Chemicals Computer/telecom services

Figure 5.13 Type of knowledge received from most important external source (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations, ‘M’ indicates a missing value.

– The technological proﬁle of the ﬁrm does not ﬁt the type of information that can be found in patents. – Respondent is not R&D intensive. – The ﬁrm is not aware of such databases. – There is sufﬁcient know-how in house. – The information available in patents is often of marginal importance to the ﬁrm. – Small ﬁrms cannot maintain the human capital required to utilize these databases. – Information is obtained through informal contacts. – Firm relies on the parent company for patent searches. – The search for innovating ideas is done through scientiﬁc journals, trade fairs and exhibitions, reverse engineering and the Internet. – Firm obtains the necessary information from suppliers. The minority of ﬁrms that do search patent databases mentioned the following reasons:

Knowledge ﬂows in European industry 93 Informal personal contacts

Research co-operation

x

100%

U De K nm ar k Gr ee ce

M

Ne

Ge

er

xx

x

Ne

th

Fra nc

rm Ge

x

lan ds

0%

xx x

rm

20%

0%

U De K nm ar k Gr ee ce

20%

lan ds

40%

e

40%

Ita ly

60%

an y

80%

60%

Ita ly

x

80%

er

x

th

xx

an y Fra nc e

x

100%

Exchange of students or other personnel

Other method

100%

100%

80%

80%

60%

60% 40%

ce

k ar

ee Gr

nm

Ne Italy th er lan ds

e nc

an

xx

UK

M

y

ce

k

ee

ar

xx

0%

Gr

UK

nm De

ds lan

ly Ne th

er

Ita

e nc Fra

y an rm Ge

xxxx

rm

x

0%

De

20%

xx

Ge

20%

Fra

40%

x

Figure 5.14 Methods of communication to obtain external knowledge (source: KNOW Survey). Notes ‘x’ indicates that the value for the corresponding stratum is based on a low number of observations, ‘M’ indicates a missing value.

– – – – – – •

Explore market trends and novelties. Assist technicians to access the state-of-the-art in each scientiﬁc topic. Identify similar products in the market and, more generally, study what has been done in a speciﬁc area. Identify innovative companies and then try to reverse engineer their products. Identify the patent owner who will provide the technical solution. Take ideas for developing new applications.

Use of patents to protect innovation-related knowledge. The vast majority of interviewed ﬁrms do not use patents to protect inventions. The few that do reported the following reasons: – –

Patents represent the most effective method for protecting intellectual property (IP). Patents are preferred for products with an expected long life-cycle.

94 Y. Caloghirou, A. Constantelou and N.S. Vonortas The reported reasons for avoiding patents include: – Lead time and secrecy are more effective for protecting IP. – Company does not introduce new products. – Cost of patents, including application and administration, is too high. – Cost and degree of protection from infringement out of balance. – Signiﬁcant delays introduced by applying for a patent. – Too much information is disclosed when applying for a patent. – Increasingly difﬁcult to patent due to invention overlap. – Patents are not customary in the speciﬁc activity (software). – Patents are applied for centrally by the company’s headquarters. – The speciﬁc national patent system is considered insufﬁcient for protecting against closely-related spin-offs of the original idea. •

Internet utilization. The majority of interviewed ﬁrms used the Internet.3 A signiﬁcant number of them had developed their own websites. They reported the following reasons for using the Internet: – – – – – – –

– – – – –

Communicate with customers, suppliers and collaborators. Search for technical characteristics of competing and new products. Contact overseas customers. Get information about patents and about training programmes. Save time, decrease costs. Access technical information. Business-to-business (B2B) and other e-business considered a good opportunity to reduce distribution cost and to facilitate market access without intermediate channels. Obtain competitive intelligence on prospective collaborators. Obtain ideas on new potential services to market. Search for new suppliers. Search literature online. Maintain databases of customer proﬁles or technological information.

Reported reasons for not using the Internet or for only limited use include: – Need to ﬁlter information. – Perceived problem with updating information on the Internet. – Not useful. – Firm does not know how to use it. •

Collaboration with universities and PROs. The majority of interviewed ﬁrms in Italy, the Netherlands and the UK had collaborated at least once with universities and PROs. The converse was true for ﬁrms in France, Germany and Greece. The reported incentives for R&D collaboration with these institutions include:

Knowledge ﬂows in European industry 95 – – – – – –

Obtain EU funding. Access state-of-the-art scientiﬁc knowledge. Access students, well-trained human capital. Access specialized instruments and facilities. Cost effectiveness. The ﬁrm’s main clients view collaboration with universities and PROs positively. – Obtain more reliable results than those from consultants. Many interviewees also expressed signiﬁcant reservations about collaborating with universities and PROs, even those that have had such experience. Reservations were based on the following arguments: – Firm is too small to attract the interest of the university. – Firm did not have the chance to develop such collaboration. – Firm does not do basic or applied work; it is involved in development which does not require academic input. – The applicability of the research results is questionable; universities too theoretical for industrial needs. – Firm lacks the necessary ﬁnancial resources. – Firm believes in internal development of innovations. – Firm prefers informal relationships with academics based on personal relationships. – Service industry ﬁnds such collaboration not very useful. – Universities often sluggish and inactive in cooperative research. – Firm fears loss of secrets; cultural differences with universities. – Universities often lag behind industry; graduates unaware of latest market developments. – Other divisions of the company have the necessary capabilities in basic research. – Firm prefers to collaborate with technical centres and consultants; their competencies are perceived to be closer to the needs of industry. •

Cooperation with other private sector ﬁrms. The majority of ﬁrms interviewed in Italy, the Netherlands and the UK have developed research partnerships with other companies. The reverse is true for ﬁrms in Greece and Germany. In France it was fairly balanced. The reasons reported for participating in ﬁrm-to-ﬁrm partnerships include: – – – – –

Access to accumulated knowledge and experience of partners. Achieve synergies. Access information about new market trends. Increase own capabilities. Access markets and technologies.

96 Y. Caloghirou, A. Constantelou and N.S. Vonortas – Collaboration with platform providers brings useful information for developing applications. – Collaborate in anticipation of national safety regulations. – Share costs of innovation. – Supplier–customer collaboration strengthens lead-user concept. – Collaborate in non-core technical activities. – Collaborate only when the technology is mature, mainly to lower the cost of technological improvement. – Handle increasing technological complexity and provide integrated solutions. – Firm focuses on hybrid activities that make collaboration necessary. – Stay abreast of technological frontier. Market size often insufﬁcient to justify the necessary investment. – Acquire competencies that do not exist in the company and create radically new domains of activity. In contrast, the reported reasons for not doing collaborative R&D with other companies include: – – – – – – – – – –

•

The type of products and the sectoral characteristics do not encourage cooperation. Lack of appropriate organizational structure, and lack of information about suitable partners and how to contact. Lack of appropriate ﬁnancial and human resources. Inward looking strategy. Risk of passing vital information to competitors. Market not big enough for more than one company. Cooperation is time consuming and costly. Parent company engages in partnerships; unnecessary for interviewed company. Never cooperate with other companies in core business. Cultural differences with prospective partners increase costs of cooperation.

The most important source of knowledge contributing to the most economically important innovation. Answers varied considerably.4 A minority of ﬁrms replied that the most important source of knowledge was internal to the ﬁrm. Regarding external sources, ﬁrms mentioned suppliers, customers, competitors, universities, PROs and consultants. Suppliers were important for: –

Collaborating with the interviewed ﬁrm; provide necessary assets that innovating ﬁrm lacks. – Giving technical information and suggesting ways to use the provided materials. – Helping implement new pieces of machinery.

Knowledge ﬂows in European industry 97 Customers were important for: – Offering new ideas. – Bringing forward their preferences and informing the interviewed ﬁrm about the success of similar products from competitors. – Providing the very specialized information necessary for the successful completion of the innovation. – Operating as lead users; testing prototypes. – Exchanging personnel with the ﬁrm. Competitors were important for: – Collaborating to reduce the cost and risk involved in innovation. – Introducing similar products serving as benchmarks and sources of ideas. Universities and PROs were important for: –

Providing ideas (for example, through research undertaken by students) and new solutions. – Providing technical support. Consultants were important for: – Providing information for new market trends and consumer needs. – Assisting in the completion of the innovation. – Providing technical support. •

Relative importance of internal and external sources of information. In line with the information from the survey, there was a clear leaning towards assigning more importance to internal sources of information for innovation. This was particularly so for the implementation of the innovation. A much smaller number of interviewees considered internal and external sources to be of equal importance or considered external sources of information as being more important. The importance of internal sources of information was the result of: – Familiarity and experience of company employees with the business. – Importance of ﬁrm-speciﬁc and industry-speciﬁc information for innovation. – Greater reliability. – Difﬁculty in ﬁnding the appropriate expertise in external sources. – Speciﬁcations given by the parent company; development undertaken internally. External sources of information were important for: –

Providing technical expertise and helping to build on the research of others.

98 Y. Caloghirou, A. Constantelou and N.S. Vonortas – Helping to internalize external knowledge. – Offering the original idea. – Obtaining specialized know-how. – Acquiring new competencies and opening up new commercial opportunities. – Cooperation with other organizations crucial for keeping up with rapid changes in technological trajectories. – Assisting with quality management standards and complex production processes. – The output obtained from the external source assists the ﬁrm in exploring different aspects of the same innovation. – Sharing risks and costs. – Collaborating to offer more complex solutions to satisfy customer needs. The descriptive results summarized in this chapter of the focused survey of European ﬁrms in selected industries regarding the use of particular mechanisms and channels for the ﬂow of knowledge provide a ﬁrst glimpse into the procedures and tactics followed by ﬁrms for gaining access to knowledge necessary for innovation. These ﬁndings are further explored in Part II of the book where contributors often use sophisticated data analysis techniques to guide the empirical investigation of key research questions.

Notes 1 This chapter is based on the collective work of a large number of researchers in several European countries. We thank them all, especially Thomas Cleff and Dirk Czarnitzki, for their valuable insights and hard work on the data. 2 For the purpose of this chapter, the last two sectors have been merged and called ICT services. 3 All interviewed French ﬁrms reported using the Internet; nine out of ten had their own websites. 4 The Netherlands did not report results on this item.

Part II

Aspects of knowledge ﬂows

6

Facilitators and impediments to knowledge sharing An exploration of different organizational forms Mette Praest Knudsen

1 Introduction A multiplicity of research papers have focused on ﬂows of knowledge as a source of sustained competitive advantages (see, for example, the special issue by Argote et al. 2000). Intra-ﬁrm knowledge ﬂows seem unproblematic as they are both the most important type of knowledge ﬂows (Schrader 1991), and they are hindered less by conﬁdentiality and legal obstacles than external transfers (Szulanski 1996: 27). But why do we need to study intra-ﬁrm knowledge ﬂows? One aspect of this question can be addressed by discussing the limits to integration of organizational knowledge and, subsequently, utilization of the knowledge to achieve sustained competitive advantage. Becker (2001: 1039) argues that dispersed knowledge can never be fully centralized in organizations thereby making knowledge sharing of vital interest to the ﬁrm. Key challenges to knowledge management that arise out of the debate on the dispersedness of knowledge include opaqueness, higher resource requirements, learning and competence potentials, differences in cognitive frame and decision problems. The possibility of coping with each of these challenges and thus ensuring knowledge sharing in the organization, is restricted by the structural characteristics of the organization. The structure of an organization should, according to Tushman and Nadler (1978: 613–614), ﬁt characteristics both inside and outside of the ﬁrm, which may include the technologies of an organization, the nature of the environment and interdependencies that exist among the units within an organization. However, the decision about an appropriate organizational structure cannot include all possible variables; the structure chosen will have an impact on the functioning of them. This chapter investigates the research question: How do organizational structures facilitate or impede intra-ﬁrm knowledge ﬂows?1 Several supplementary questions follow from this: Do formal vertical structures impede knowledge ﬂows because their formal hierarchical structure tends to decrease the motivation and creativity of the individual members of the research and development (R&D) department? Do ﬂat

102 M.P. Knudsen organizational structures such as networks create knowledge adaptation problems as a result of the informal routes of the knowledge ﬂows, which may produce better search activities, but induce problems over integration back into the ﬁrm? To explore these questions, this chapter draws on a combination of survey data and case studies. Section 6.2 outlines some basic themes surrounding knowledge ﬂows and organizational structures. Section 6.3 presents the methodological considerations and the background to the case studies. Section 6.4 presents and discusses the three case studies, on the basis of the theoretical discussion in Section 6.2. Finally, Section 6.5 outlines the conclusions and raises some general points for management regarding the problems of choosing particular forms for the organization of R&D.

6.2 Knowledge ﬂows and organizational structures In order for knowledge ﬂows or, more precisely, knowledge sharing, to exist certain requirements must be fulﬁlled. The ﬁrst requirement is that there must be two actors: the transferer, who should be willing to reveal, in some form, his knowledge, and the recipient of this knowledge. These actors may be within a single ﬁrm or in two independent ﬁrms: resulting respectively in inter- or intra-ﬁrm knowledge ﬂows. The two actors may engage in either a cooperative or a competitive relationship. Depending on the nature of the relationship, the challenges for knowledge sharing differ.2 The second requirement, the willingness or not to share knowledge, may reveal a number of impediments to knowledge sharing, such as lack of motivation (Kalling 2003), symptoms of ‘not-invented-here’ syndrome (Katz and Allen 1982) and lack of organizational openness towards knowledge sharing (Goh 2002). Finally, ability of the transferer to reveal the knowledge and ability of the recipient to perceive the knowledge and make use of it must exist. Based on such simple observations, the literature has pointed to some facilitators of, and impediments to, knowledge ﬂows (Mangematin and Nesta 1999; Lane and Lubatkin 1998; Stock et al. 2001; Szulanski 1996). A ﬁrst condition for successful knowledge sharing is the existence of absorptive capacity on the part of the recipient (Cohen and Levinthal 1989, 1990). Another factor relates to the characteristics of the knowledge to be transferred. If the knowledge is of a complex and tacit type then the transfer requires face-to-face interaction, but remains burdensome (Rocha 1999). On the other hand, codiﬁed knowledge, for instance prototypes, can easily be transferred and likely at lower cost. However, there is a problem in distinguishing tacit and codiﬁed knowledge: knowledge in organizations typically represents a mixture of both. An alternative way to characterize knowledge is to consider whether the knowledge from the transmitting ﬁrm is supplementary or complementary to that of the receiving ﬁrm. See, for example, Rindﬂeisch and Moorman (2001), Echambadi

Facilitators and impediments to knowledge sharing 103 et al. (2001) and Dussauge et al. (2000) for deﬁnitions and discussions. Supplementary knowledge is deﬁned as knowledge where there is a degree of redundancy or overlap with existing knowledge, which makes it easier to transfer, whereas complementary knowledge offers a larger potential for learning, but due to the smaller overlap in the knowledge bases, may be difﬁcult to share among colleagues. Organizational structures must be seen as distinct from organizational design: the former can be viewed as structural characteristics that remain unchanged over a long period of time and inﬂuence or constrain important aspects of an organization’s behaviour. Organizational design comprises organizational structures, some aspects of which can be prescribed, or at least are acceptable, to formal authority (Weick 2003: 93–94). When it is argued that the structure of the organization can facilitate or impede knowledge ﬂows, this is based on the idea that the structure acts as a framework for activities in the organization constituted by employees, resources and products. This chapter does not deal with the organizational design decision and the changing nature of structures. Below we review three distinct organizational forms based on Burton and Obel (1995: Ch. 2.4).3 The functional form is deﬁned as a unit grouped by functional specialization, for instance, production, marketing, R&D. A distinction is drawn between line and staff departments where the line departments include R&D. The line departments handle coordination and control and, hence, the information ﬂow tends to follow the hierarchical lines in the organization leading to a highly involved top management. The scope and ﬂexibility of knowledge sharing is rather limited as the coordination of knowledge rests with the top management. A key challenge for management is to engage project leaders in strategizing and implementation of knowledge sharing initiatives. On the other hand, the functional form allows greater efﬁciency from economies of scale and skills that may ensure knowledge production of the supplementary type. In relation to R&D, less developed capabilities in knowledge absorption may limit the production of new knowledge that is complementary to existing knowledge sources. The disintegrated divisional form is a sub-group of the divisional form, where self-contained and somewhat autonomous units are coordinated by a headquarters unit. The disintegrated form implies a certain degree of independence of the R&D unit from the other divisions. The R&D division is in itself ﬂexibly organized allowing for teams to be set up in response to new projects. The information ﬂow is typically restricted to the division itself and allows only limited sharing of information across divisions. It is therefore to be expected that problems related to knowledge transfer primarily exist between divisions, whereas knowledge ﬂows from external relationships are facilitated through direct access via the divisions.

104 M.P. Knudsen Accordingly, it is crucial that the organization has in place the appropriate knowledge management tools to allow knowledge to ﬂow across the divisions, for example, through rotation of employees or a centralized intranet, which captures project experiences. Finally, the collaborative form is to some degree similar to the ad hoc conﬁguration, in which a number of experts dominate their own knowledge domain and no hierarchical levels are involved. Equally, there is no centralized decision making. In the collaborative form, the experts may include external partners that act within the ﬁrm boundaries allowing for active and spontaneous knowledge-sharing processes. The decision making is related to single projects and is typically carried out by project leaders who, themselves, are experts. Within teams of experts, knowledge may ﬂow freely, but require extensive coordination among project members. The organization of the ﬁrm may result in ﬂuid and changing boundaries impeding the management of production and R&D. In relation to knowledge ﬂows, one challenge concerns the coordination of tasks, projects and employees. The experts may expose symptoms of the not-invented-here syndrome (Katz and Allen 1982) if incentive schemes are not developed and implemented. The collaborative form is strongly focused on existing relationships and their maintenance. Channels are well established and knowledge ﬂows are facilitated by trust, mutual working practices and rotation of employees. However, problems may occur in the creation of new relationships and stimulation of knowledge sharing practices.

6.3 Methodological considerations This chapter draws on data on the telecommunications sector collected in Denmark in the period 1998–2000 relating to innovative activities. Based on questionnaire results, ten follow-up case studies were conducted which revealed further details on both internal and external knowledge ﬂows. The interview structure was guided by the results of a mail survey conducted approximately six months earlier. Detailed information concerning knowledge ﬂows within the ﬁrm and in cooperation with external partners was gleaned through a focus on the most important innovations within the ﬁrm and their development. We describe here three out of the ten cases studies. The three distinct organizational forms are examined within the context of these three case studies and discussed in terms of their differences in facilitating and impeding knowledge sharing within knowledge-based organizations. The characteristics of the knowledge-sharing processes involved knowledge similarities and the dispersed nature of the knowledge, which were identiﬁed and discussed for each of the organizational forms.

Facilitators and impediments to knowledge sharing 105

6.4 Knowledge ﬂows in distinct organizational structures In Section 6.2 a number of facilitators and impediments to knowledge sharing were presented along with three organizational structures. In this section, one case ﬁrm is presented for each of the three organizational forms and their associated facilitators or impediments factors are described and discussed. Some guidelines for management are outlined. 6.4.1 Case A: Functional form Firm A is a large telecommunications equipment manufacturer with core competencies in optics. Of the 280 employees in the ﬁrm, 10.7 per cent are employed in R&D. The most important innovation was based on the ﬁrm’s core competencies and was developed using a combination of existing technologies. For the development of this innovation, internal knowledge sources were more important than external ones. The company’s organization is depicted in Figure 6.1. The interviewee (the chief executive ofﬁcer – CEO) took up his position in 1999. This change of manager had resulted in fairly extensive organizational changes. First, the new manager brought with him a network of contacts and experience from international operations. In practical terms, this has led to an increased openness towards strategic collaborations: approximately 40 per cent of the ﬁrm’s innovations since that time are the results of collaboration with external partners. However, these alliances have produced only inﬂows of knowledge based on restrictive knowledge sharing processes with the partners. These processes are not restricted by formal procedures, but depend entirely upon individual employees knowing what they are allowed to share. The R&D department receives knowledge from universities, suppliers and customers, and especially customer input is considered vital for the initiation of new projects. Hence, the organization is highly marketoriented and driven by end-user needs. The knowledge-absorption processes are tightly controlled, and the transferred knowledge is required to be codiﬁed even though typically the initial transfer takes place face to face. The very complex knowledge that has been received Administration, finance, HRM

R&D

Testing and prototypes

Figure 6.1 Functional organization of case A.

Production

Marketing and sales

106 M.P. Knudsen has become quite important to the department, but has been difﬁcult to exploit fully. As stressed by Cohen and Levinthal (1990), absorption involves three processes – identiﬁcation, assimilation and exploitation – each sub-process being necessary if the new knowledge is to be fully absorbed. In the case of ﬁrm A, their main problem has been ﬁnding ways to exploit the knowledge. The unproblematic processes of knowledge identiﬁcation and assimilation are expected to be facilitated by the results of the high training share, where more than 75 per cent of the employees participate in training on an annual basis. The organizational form is traditional in terms of its hierarchical set up, with a ﬂow of inventions being converted into innovations through production, sales and marketing. Bottlenecks occur quite frequently in the link between the development and production departments. In the development and prototyping phases, developers are tightly focused on producing a functional product, but engineers consider their job is done once the products leave the department. The engineers lack any sense of ownership of the process and, therefore, any incentive for cooperation and knowledge sharing within the ﬁrm, which could reduce time to product and, subsequently, time to market. This lack of cooperation can be mediated by the use of cross-functional teams and job rotation schemes to smooth the process from prototype to production. At the time of interview, the ﬁrm was planning to introduce a job rotation scheme. Finally, the knowledge workers in R&D were more interested in pursuing their own new ideas than conﬁning their activities in line with the strategy of top management. The CEO, as a result of the organizational structure, plays an important part in the initiation of new projects, but the engineers involved do not necessarily follow his instructions. 6.4.2 Discussion The recent change in top management has resulted in a more openminded culture towards external cooperation. Several new alliances have been formed, but the CEO is more interested in knowledge receipt and absorption than in creating an environment conducive to mutual knowledge exchange with external partners. Larsson et al. (1998) discuss the problem of mutuality in knowledge sharing by focusing on the intentionality on the part of both the sender and the receiver of knowledge. How can the sender be motivated to send the knowledge with the least efﬁciency losses, and similarly how can the receiver be motivated to be receptive towards knowledge inﬂows. The authors argue that the sender should be transparent and the recipient should be receptive in order to ensure successful knowledge transfer. In the case of ﬁrm A, it has to learn to be transparent and allow the sender also to receive knowledge. The result of imbalances in intentionality may lead to competition-based knowledge sharing and, ultimately, to learning races.

Facilitators and impediments to knowledge sharing 107 Internally, these problems may result in symptoms of the not-inventedhere syndrome, where the R&D department is unwilling to accept knowledge from collaborating partners. Katz and Allen (1982: 7) deﬁne the syndrome as ‘the tendency of a project group of stable composition to believe it possesses a monopoly of knowledge of its ﬁeld, which leads it to reject new ideas from outsiders to the detriment of its performance’. Ultimately, the problems may escalate and result in loss of both inﬂows and outﬂows of knowledge. One possible solution to these problems is to establish a development facility abroad with new engineers and to foster close collaborative contacts with relevant foreign partners. This facility might then act as the driver of a new R&D culture among the existing engineers and create competitive pressure for greater cooperation within the ﬁrm. This suggestion ﬁnds some support in Tsai (2002: 186), who ﬁnds that social interaction stimulates more knowledge sharing among organizational units that are in competition with each other than among units that are not. Internally, knowledge ﬂows are also impeded, ﬁrstly by the R&D engineers who are eager to work on their own ideas, but less keen to expend effort on the projects formulated by the organization. Furthermore, earlier failures with development projects have led to a lack of internal cooperation and competition-based knowledge sharing. Sherman and Smith (1984) suggested that the structural characteristics of the organization may constitute one possible explanation. They hypothesized about, and found support for, a strongly negative effect from the structural characteristics on the level of intrinsic motivation (Sherman and Smith 1984: 882–883). Increased centralization, formalization and standardization resulting in less personal freedom, reduction in autonomy and implementation of formal rules reduce the motivation of employees. This was conﬁrmed by Kalling (2003: 123), who added that ﬁtting the knowledge to the existing knowledge base appeared to be a more efﬁcient way of ensuring that the recipients of the knowledge understood its value. Currently, attempts have been made to solve these problems of motivation by installing intranets, setting up document management systems and installing groupware (Goh 2002: 25), but the success of these attempts has been limited (Hendriks 1999: 91). In an attempt to ﬁnd motivators for knowledge sharing, Hendriks (1999: 95) builds on Herzberg’s two-factor theory (Herzberg 1968). Herzberg distinguishes between ‘hygiene’ factors and ‘motivators’, where the presence of the former group of factors does not motivate behaviour, but their absence leads to dissatisfaction. Hygiene factors include salary, status, company policy and interpersonal relations. Incentive schemes offering additional payments, bonuses, air miles or salary penalties may increase knowledge sharing, but not the motivation to share (see Severinov (2001) for a formal model on incentive schemes and Hendriks (1999) for a discussion on bonuses and air miles). Motivators include achievement, responsibility, recognition, operational autonomy,

108 M.P. Knudsen promotional opportunities and challenges. In attempting to resolve their internal motivation problems, ﬁrm A should focus on the group of motivators, that is, increase the sense of ownership of the knowledge-sharing process by revealing the strategy behind it or by offering promotional opportunities. The organizational structure of ﬁrm A has made it impossible to avoid the problems described in Section 6.2 that impede knowledge production and utilization. An obvious challenge for management is to mediate the motivational problems and the competition-based knowledge ﬂows by establishing appropriate incentive schemes and creating cross-functional teams (Goh 2002: 26), which should foster an internal organizational culture towards knowledge sharing and joint development efforts. 6.4.3 Case B: Disintegrated divisional form Firm B operates in the computer services market supplying information technology (IT) infrastructure solutions based on standard products. There are 360 employees, 19.4 per cent of whom are engaged in R&D. The R&D department serves as an independent unit – partly self-ﬁnanced. The most important innovation produced by ﬁrm B is the integration of the Internet into the existing solutions. This integration facilitates both a broader use of the Internet and enables distribution of new software and updates. Internal and external sources of knowledge were equally important in the development of this innovation. In 1999, ﬁrm B was acquired by a large company. This led to drastic changes in the organization in terms of strategy, product development and customer relationships. Their most important innovation can be traced back to development projects that were initiated before the acquisition took place. The company’s organization is illustrated in Figure 6.2. The organization of the R&D department is such that approximately 50 per cent of R&D costs are covered through customer-related projects, with the remaining 50 per cent being devoted to general development projects, such as the development of new methods, etc. In principle, a typical

Market division

Market division R&D

Admin. Market division

Figure 6.2 The organization of case B.

Market division

Facilitators and impediments to knowledge sharing 109 R&D project is initiated by one of the market divisions that reacts to a request from a customer, feeding the idea into the R&D department. R&D then assembles a team of knowledge workers with the necessary competencies. The team is physically located in the division that produced the idea and works there for the duration of the project, cooperating with both the customer and the market division. Customers providing input to development projects ensure a high degree of market orientation, which in turn ensures a high rate of ﬁnished and successful projects. However, too great a customer focus could result in a narrow knowledge base that is not renewed, which threatens the long-term production of knowledge. The phenomenon of continued production of knowledge based on a stable knowledge base may lead to lock-in: a change in customer needs may demand external knowledge, which may not be forthcoming. The advantage of this structure, however, is that teams can be set up to satisfy speciﬁc needs and that these teams are in close contact with the customer. Teams can be assembled from the members of the R&D department according to the needs of the market divisions. The R&D division acts as a knowledge platform from which the knowledge required for a particular project can be sourced. This format ensures ﬂexibility, less overlap in competencies, feedback of experience to all necessary agents in R&D and possible utilization of market knowledge for future development purposes. The disadvantage of this format is the high start-up costs involved in setting up a new team and the members getting to know their colleagues. However, this disadvantage seems to be outweighed by the advantages of closeness to the customer and the cross-disciplinarity in the R&D and division interface. 6.4.4 Discussion To ensure smooth knowledge ﬂows and feedback into the system a very structured project evaluation process is required at the end of each project. A loosely-coupled organization, operating with ad hoc teams, creates a need for tight knowledge management to ensure that experience passes on into the system. The project-based teams and the cross-divisional cooperation stimulates the ﬂow of knowledge from the market to R&D, and vice versa. Such teams may therefore stimulate cooperation-based knowledge sharing. By cooperation-based knowledge sharing we mean knowledge ﬂows that are based on a willingness to cooperate to ensure the best project outcome, making the not-invented-here syndrome obsolete. However, even where this operates, the maintenance of the knowledge base requires deliberate openness towards external knowledge sources. One way to ensure this is through training. Training of employees can bring new knowledge to the organization, ﬁrst through the content of the

110 M.P. Knudsen training courses themselves, and second through an increased motivation to search for and capture new knowledge. In ﬁrm B, training receives a lot of emphasis with more than 70 per cent of the employees annually being involved in training of some kind. Finally, ﬁrm B must stimulate knowledge production through the use of external cooperation partners. As internal and external sources of knowledge were of equal signiﬁcance in the development of their most important innovation, there is the danger that the change in ownership of the company will lead to a reduced focus on external sources and ultimately problems of knowledge access. Firm B must beware of falling into this trap. In conclusion, ﬁrm B seems to have an organizational structure that stimulates knowledge production and knowledge sharing to ensure swift and successful ﬁnalization of development projects. A clear role for management is to focus on continued access and utilization of external knowledge sources to stimulate a broadening of the knowledge base and adaptation of complementary knowledge sources. Both these aspects will be vital for the long-term survival of the company. 6.4.5 Case C: Collaborative The last of our case study ﬁrms was a small ﬁrm with less than 30 employees, which was established in 1994. Originally the ﬁrm was a distributor of entertainment and telecommunication products, but lately it has started to develop its own products. Its trading activities have acted as a cash cow for the development of its own products. The share of turnover from its own products is currently very small, but the expectation is that this share will rise and eventually become the main source of its proﬁts. The ﬁrm operates as a traditional trading company for its distribution activities, but its product development activities are organized as a collaborative network as illustrated in Figure 6.3.

Distribution of electronics and multimedia products R&D and production of own products

Collaborative foreign partner

Figure 6.3 The organization of case C.

Facilitators and impediments to knowledge sharing 111 A particular feature of this ﬁrm is the organization of its R&D department. Product development activities are carried out in cooperation with an Asian ﬁrm. The Asian ﬁrm has strong technological competencies, but no manufacturing, sales or marketing skills. The knowledge of the two companies is therefore complementary in nature. Overall, for ﬁrm C there is a ﬁnancial inﬂow from the distribution of activities to R&D and a technological knowledge ﬂow from the Asian partner. Thus, even though the two ﬁrms are legally independent, they are mutually dependent. The knowledge ﬂows between them usually occur through e-mail and the Internet, and also through extended visits to the Asian engineering ﬁrm. Thus, periods of close personal interaction are followed up and reinforced by periods of electronic contact. The knowledge ﬂows have so far been unproblematic even though the knowledge is complementary in nature. 6.4.6 Discussion The relationship of ﬁrm C with its Asian collaborator is characterized by trust, openness and high levels of competence on both sides. To ensure that trust is maintained and strengthened, employee rewards based on joint success could be instigated (Goh 2002: 26). Management should also continue to communicate plans for future projects and new strategies to employees. The present organizational cooperative structure is more sensitive to change than the structures in ﬁrm A and ﬁrm B. This sensitivity has to do with the mutual dependence of ﬁrm C and its Asian partner. If one of the ﬁrms were to decide to try to build competencies that currently reside in the other, these efforts could result in a learning race and competitionbased knowledge sharing, which eventually might destroy the present organizational structure. Possible safeguards against a learning race developing include assets exchange, buying up of shares and other legal controls. Firm C is embarking on developments to build technological competencies that match those of its partner and has bought the intellectual property (IP) behind their jointly-owned technologies. Coordination of knowledge sources is another aspect that is important for ﬁrm C bearing in mind its structure. De Silva and Agustí-Cullel (2003: 51) deﬁne knowledge coordination as ‘the design and use of organized and purposeful strategies to control knowledge distribution and dissemination across organizations’. Both internal and external coordination of knowledge sources are important to maintain the stability of the present structure. Internal coordination is important to keep employees focused on company strategy and the role of the external partner in the achievement of this strategy. Intra-ﬁrm coordination of knowledge is therefore as important as inter-ﬁrm coordination for maintenance of the organizational structure.

112 M.P. Knudsen In conclusion, the chosen organizational structure seems viable and able to withstand external competitive pressures, but requires extensive knowledge management and coordination strategies to maintain the balance of mutual dependence between the partners. As already stressed, this dependence can be maintained through adopting certain safeguards and also through informal mechanisms like joint strategic planning, increased cooperation and strengthening of social relationships among employees, all of which can add to the stability of the organizational structure.

6.5 Conclusions The theoretical discussion of organizational forms shows that the functional organizational structure is expected to induce a discrepancy between top management and the knowledge workers in R&D. This is conﬁrmed by the case study, and the conclusion must therefore be drawn that the functional form is more appropriate for less knowledge-intensive activities. If an organization is moving towards becoming more dependent on new product development, then the organizational form should be reconsidered and a reorganization implemented. In terms of the divisional form it was expected that external knowledge ﬂows in particular would be facilitated through the divisions. For ﬁrm B, the result was the opposite: the knowledge ﬂowing from the divisions to R&D was too specialized to facilitate knowledge production. Hence, there is a danger that the knowledge base will evolve into a state of lock-in leading to the need for a reorganization of the knowledge base. In the case study, ﬂows between the divisions were seen to present problems, which were resolved through the use of cross-functional teams. Finally, the collaborative organizational structure ran the risk that too few new relationships would be established. The case of ﬁrm C conﬁrmed this, as the partners were primarily interested in maintaining the balance of mutual dependence and ensuring internal and external coordination of knowledge. These problems, however, had been anticipated and ﬁrm C was endeavouring to safeguard its dependence on the Asian partner by controlling intellectual property rights and building its own technological competencies. As can be seen from the above cases, each of the organizational forms brings different requirements for the knowledge transfer processes. In case A, incentive systems had to be built which would motivate the development team to follow up and participate in taking the innovation to production. In case B, a clear team-based organization required not only the capture of project experiences, but also an explicit focus on external knowledge sources to avoid knowledge lock-in. Finally, in case C, a pure collaborative relationship proved to be highly competitive, but also vulnerable to internal changes. A number of problems of a more general charac-

Facilitators and impediments to knowledge sharing 113 ter were identiﬁed in relation to organizational forms and knowledge sharing processes. These problems were addressed by suggestions for management. Further research could be directed towards a more explicit approach to the challenges posed by internal and external knowledge ﬂows in different organizational forms. To do this we would need to know more about the mechanisms that stimulate knowledge sharing, and how these mechanisms work under different organizational constraints.

Notes 1 The term ‘knowledge ﬂows’ is used as a generic and broad notion that includes among others knowledge sharing and knowledge transfer. The chapter does on several occasions include inter-ﬁrm knowledge ﬂows, but these are not at its core. 2 A combination of cooperation and competition may also be envisaged, which is known as coopetition; see, for example, Loebecke et al. (1999) and Tsai (2002). In the present chapter, the situation is characterized as either competitive or cooperative. 3 Further organizational forms include the simile, the matrix, the international conﬁgurations as well as the bureaucracies, which are not included in the present chapter.

References Argote, L., Ingram, P., Levine, J.M. and Moreland, R.L. (2000) ‘Knowledge transfer in organizations: learning from the experiences of others’, Organizational Behavior and Human Decision Processes, 82(1): 1–8. Becker, M.C. (2001) ‘Managing dispersed knowledge: organizational problems, managerial strategies and their effectiveness’, Journal of Management Studies, 38(7): 1037–1051. Burton, R.M. and Obel, B. (1995) Strategic Organizational Diagnosis and Design: Developing Theory for Application, Dordrecht: Kluwer Academic Publishers. Cohen, W.M. and Levinthal, D.A. (1989) ‘Innovation and learning: the two faces of R&D’, Economic Journal, 99: 569–596. Cohen, W.M. and Levinthal, D.A. (1990) ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35: 128–152. Dussauge, P., Garrette, B. and Mitchell, W. (2000) ‘Learning from competing partners: outcomes and durations of scale and link alliances in Europe, North America and Asia’, Strategic Management Journal, 21: 99–126. Echambadi, M., Cavusgil, S. and Aulakh, P. (2001) ‘The inﬂuence of complementarity, compatibility and relationship capital on alliance performance’, Journal of the Academy of Marketing Science, 29(4): 358–373. Goh, S.C. (2002) ‘Managing effective knowledge transfer: an integrative framework and some practice implications’, Journal of Knowledge Management, 6(1): 23–30. Hendriks, P. (1999) ‘Why share knowledge? The inﬂuence of ICT on the motivation for knowledge sharing’, Knowledge and Process Management, 6(2): 91–100. Herzberg, F. (1968) Work and the Nature of Man, London: Granada Publishing.

114 M.P. Knudsen Kalling, T. (2003) ‘Organization-internal transfer of knowledge and the role of motivation: a qualitative case study’, Knowledge and Process Management, 10(2): 115–126. Katz, R. and Allen, T.J. (1982) ‘Investigating the not invented here syndrome: a look at the performance, tenure and communication patterns of 50 R&D project groups’, R&D Management, 12(1): 7–19. Lane, P.J. and Lubatkin, M. (1998) ‘Relative absorptive capacity and inter-organizational learning’, Strategic Management Journal, 19: 461–477. Larsson, R., Bengtsson, L., Henriksson, K. and Sparks, J. (1998) ‘The interorganizational learning dilemma: collective knowledge development in strategic alliances’, Organization Science, 9(3): 285–305. Loebecke, C., Fenema, P.C.V. and Powell, P. (1999) ‘Co-opetition and knowledge transfer’, The DATA BASE for Advances in Information Systems, 30(2): 14–25. Mangematin, V. and Nesta, L. (1999) ‘What kind of knowledge can a ﬁrm absorb?’, International Journal of Technology Management, 18(3/4): 149–172. Rindﬂeisch, A. and Moorman, C. (2001) ‘The acquisition and utilization of information in new product alliances: a strength of ties perspective’, Journal of Marketing, 65(April): 1–18. Rocha, F. (1999) ‘Inter-ﬁrm technological cooperation: effects of absorptive capacity, ﬁrm size and specialization’, Economics of Innovation and New Technology, 8: 253–71. Schrader, S. (1991) ‘Informal technology transfer between ﬁrms: cooperation through information trading’, Research Policy, 20: 153–170. Severinov, S. (2001) ‘On information sharing and incentives in R&D’, RAND Journal of Economics, 32(3): 542–564. Sherman, J.D. and Smith, H.L. (1984) ‘The inﬂuence of organizational structure on intrinsic versus extrinsic motivation’, Academy of Management Journal, 27(4): 877–885. de Silva, F.S.C. and Agustí-Cullel, J. (2003) ‘Issues on knowledge coordination’, Knowledge and Process Management, 10(1): 37–59. Stock, G.N., Greis, N.P. and Fischer, W.A. (2001) ‘Absorptive capacity and new product development’, Journal of High Technology Management Research, 12: 77–91. Szulanski, G. (1996) ‘Exploring internal stickiness: impediments to the transfer of best practice within the ﬁrm’, Strategic Management Journal, 17(Winter special issue): 27–43. Tsai, W. (2002) ‘Social structure of “Coopetition” within a multiunit organization: coordination and introrganizational knowledge sharing’, Organization Science, 13(2): 179–190. Tushman, M.L. and Nadler, D.A. (1978) ‘Information processing as an integrating concept in organizational design’, Academy of Management Review, 3(3): 613–624. Weick, K.E. (2003) ‘Organizational design and the Gehry experience’, Journal of Management Inquiry, 12(1): 93–97.

7

Knowledge ﬂows in the Danish ICT sector The paradox of advanced demand and mediocre supply Christian R. Pedersen, Michael S. Dahl and Bent Dalum

7.1 Introduction The point of departure of this chapter is part of the general worry about the rather weak European performance in the information and communication technology (ICT) sector (Begg et al. 1999; Dalum et al. 1999). In the context of the European project on Knowledge Flows in European Industries (the KNOW project), three ﬁelds were studied: the food industry, ﬁne chemicals and the Information and Communication Technologies (ICT) sector. The main focus in the ICT study is that, on the one hand, Denmark is one of the most advanced user nations, measured by conventional indicators for user penetration (such as number of personal computers (PCs), mobile phones and Internet access per capita). On the other hand, however, it is fairly obvious that the country is not one of the major players in the international ICT markets, even allowing for country size. The nearby and somewhat similar countries of Sweden, Finland and the Netherlands are living proof that small countries can be very visible in this ﬁeld. This chapter will focus on the role and character of knowledge ﬂows in the ICT sector in a bid to ﬁnd possible explanations for this apparent lack of match between the advanced demand, but rather weak supply side of the ICT sector in Denmark. A tradition of analysing broad business sectors based on a cluster approach gradually evolved from policy makers in Denmark during the 1990s.1 The idea was to include important interactions, which may not be captured by traditional industrial classiﬁcation schemes. The explicit goal was to promote discussion of the policy implications at national level on speciﬁc groups of related industries. The ﬁrst round of studies of this type was conducted at the beginning of the 1990s. The entire spectrum of industries was divided into resource areas, more recently referred to as mega clusters (Dalsgaard 2001). More recently, the Danish Agency for Trade and Industry (EFS) has launched a second round of studies of a revised division of mega clusters, starting with a so-called ‘benchmarking’

116 C.R. Pedersen, M.S. Dahl and B. Dalum report giving an overview of the basic statistical trends of the clusters.2 This was followed by studies of some of the mega clusters, for example, construction and ICT (EFS 2001). The aim was to incorporate the general structural characteristics of the mega cluster with a particular focus on the interactions between the use of new technologies, and their development and manufacture. Given that the application of new technologies very often is located in industries other than the one in which they were developed, the combination of the user and producer industries is emphasized. In the context of ICT it might be more difﬁcult than for other mega clusters to draw the boundaries, taking into account the swift development of the underlying basic technologies. The focus of Danish policy makers on the level of mega clusters changed during the production of the second round of studies when they introduced a new level of analysis. Now the focus is on small scale clusters. These are more narrowly deﬁned as national or regional clusters that are current high performers or have considerable future potential. The overall aim is the same, but based on a narrower group of target industries. These public policy deﬁnitions of small-scale clusters and mega clusters on both national and regional levels, increase the confusion about how a cluster is actually deﬁned. In this chapter, a sector deﬁnition rather than a cluster deﬁnition of the ICT area is chosen. When we talk about the Danish ICT sector, this is the industry in a national context. This equates to the ICT mega cluster in the public policy deﬁnition. The concept of clusters should not be used on the national level, but rather for regional or geographically limited concentrations of ﬁrms in a given location. Such ﬁrms have three common factors, which qualify them for this label: (1) they are coherent, that is, their activities evolve around a limited segment of a large industry (for example, wireless communication); (2) they have a common technological knowledge base in a similarly limited area of the industry; (3) they have a common pool of labour experienced in the speciﬁc technology, market and industry segment in question. There are some problems in deﬁning the ICT sector, since it differs from what are usually termed ‘high tech’ or R&D intensive industries, that is, industries developing and producing the ICTs (hardware), pharmaceuticals/ biotechnology and aerospace. Originally the major focus in analyses of high-tech industries concentrated on the physical output (hardware). In the case of ICTs, however, it has become more and more evident that the hardware and software aspects are fundamentally intermingled – in many cases they can be regarded as two sides of the same coin. This has obviously led several analysts and researchers to use – rather than the electronics industry – the concept of an ICT sector, which comprises the electronics industry as well as software development and IT services. In a broader perspective, it may even include such industries as broadcasting, electronic and printed media, publishing, movie production and advertising, which, to an increasing degree, are converging with what tradition-

Knowledge ﬂows in the Danish ICT sector 117 ally has been conceived of as the ‘core’ ICT industries. The present chapter takes as its point of departure a narrower deﬁnition, close to the statistical deﬁnition used by the OECD (2000c).3 The EFS (2001) study contains a very broad speciﬁcation of the ICT industry (which includes publishing, media and printing). The present study has applied the narrower OECD speciﬁcation in order to focus on the ‘core’ ICT activities. The OECD approach was used and further developed in a cooperative project by the Nordic Statistical Institutes (2000, 1998),4 which produced inter-Nordic comparisons of, for example, ICT production and employment patterns (Nordic Statistical Institutes 2000, 1998). The purpose of this chapter is to investigate the apparent paradox of advanced demand from the Danish population and Danish ﬁrms versus a rather weak supply in terms of production of ICT equipment and services in the country. This is examined by studying the structure, strengths and weaknesses of the industry in Denmark compared to the rest of western Europe and the US. The remainder of this chapter is structured as follows. ICT is examined within an innovation system perspective in Section 7.2. Section 7.3 investigates the structure and performance of the Danish ICT sector in an international perspective. Some features of the weak Danish position in ICT manufacturing are highlighted in Section 7.4. The consumption pattern is described in Section 7.5. Section 7.6 presents introductory case studies of two strong niches in the industry, which are geographically clustered. The ﬁnal conclusions and discussion are presented in Section 7.7.

7.2 ICT within an innovation system perspective Among the analytical ancestors of the innovation system literature are cluster studies, such as the Lundvall et al.’s (1984) analysis of the Danish ‘agro-industrial complex’. The microeconomic foundations for this study were stated with explicit reference to the economics of innovation (Lundvall 1985). This approach became one of the central foundations for the Danish contribution to Porter’s major competitive advantage of the nations project 1987–90, published as Moeller and Pade (1988). There was a clear Scandinavian inﬂuence behind two of the components in the ‘diamond model’ (Porter 1990: Ch. 4) – that is, ‘demand’ and ‘related and supporting industries’. In parallel with, but closely related to the proliferation of cluster studies during the 1990s surveyed by Porter (1998: Ch. 7), a stream of literature has emerged under the heading of ‘systems of innovation’. Taking as its point of departure the concept of national systems of innovation (Freeman 1995, 1987; Lundvall 1992; Nelson 1993), research has developed concerning sectoral systems (Breschi and Malerba 1997; Malerba 2002), technology systems (Carlsson and Jacobsson 1997) and regional systems (Braczyk et al. 1997).5 The common thrust of this work is an emphasis on interaction

118 C.R. Pedersen, M.S. Dahl and B. Dalum between actors leading to conceptualization of innovation as a process, which often is highly embedded in a given social context. Compared to the large number of cluster studies published in the last ten years, the innovation systems literature is a more specialized line of research, which focuses on the economics of innovation. But the 1990s saw a clear trend towards convergence between the innovation systems approach and the discipline of economic geography, for example, Storper (1992, 1995), Saxenian (1994) and Maskell and Malmberg (1999) (and see also the broad survey of the entire ﬁeld by Clark et al. 2000), a trend that is continuing in the current decade. In the present context a certain emphasis will be given to the phenomenon of regional clustering of pools of labour and knowledge, which plays a prominent role in the rather few success stories from the Danish ICT sector. In a knowledge ﬂow context, these are intimately related to university–industry links.

7.3 Structure and performance of the Danish ICT sector in an international perspective In the OECD’s (2000b) publication Measuring the ICT Sector Denmark is ranked high in terms of ICT employment in comparison with 24 other OECD countries. Based on 1997 employment data Denmark had more than 96,000 ICT employees in almost 12,000 ﬁrms. The share of ICT employment in the business sector was 5.1 per cent compared to an OECD (24) average of 3.6 per cent and a European Union6 average of 3.9 per cent (OECD 2000b). The only countries with a higher share were Sweden, Switzerland, Hungary, Finland and Norway. A comparison of the industrial distribution of ICT employment among the OECD countries reveals that Denmark has a low share of employment in manufacturing and telecommunications services and consequently a high share in other ICT services (OECD 2000b). A closer analysis of ICT employment in Denmark, based on a narrower deﬁnition using data from Statistics Denmark, reveals more of the structure of the ICT sector. In Table 7.1 the 1997 data on employment and exports are assigned to six different ICT segments. Services account for approximately 80 per cent of ICT employment and exports. However, the wholesale segment distorts the pattern. The large employment share of 41 per cent reﬂects the problems of distinguishing sales and distribution activities for ICT equipment visà-vis other kinds of electrical equipment, such as various types of domestic electrical appliances. This may result in an overestimation of the ICT employment data. However, the wholesale segment for ICT exports in large part is exports of ICT hardware, although registered as a service sector activity by Statistics Denmark. These features of the data make them hard to compare in a broader international context (than the Nordic countries). These problems with international statistical comparisons are familiar

41% 16% 22%

38,869 15,242 20,280 74,391 93,749

Total ICT services

Total ICT sector

8.42%

6.68%

3.49% 1.37% 1.82%

0.24% 1.01% 0.49% 1.74%

Share of total Danish employment

Note Deﬁnition of the segments is shown in the appendix in Table A7.2.

Source: Based on data from Statistics Denmark (The Danish Mega Cluster Statistics).

100%

79%

3% 12% 6% 21%

Share of ICT employment

2,725 11,213 5,420 19,358

Number of full time employees

Manufacturing Computers Communication equipment Instruments Total ICT manufacturing Services Wholesale Telecommunication services IT services and consulting

Segment

Table 7.1 Segmentation of the ICT sector, 1997

32,533,056

26,814,247

21,760,539 1,596,634 3,457,074

1,007,853 1,458,853 3,252,103 5,718,809

Exports (DKK 1,000)

100%

82%

67% 5% 11%

3% 4% 10% 18%

Share of ICT exports

7.60%

6.27%

5.09% 0.37% 0.81%

0.24% 0.34% 0.76% 1.34%

Share of total Danish exports

120 C.R. Pedersen, M.S. Dahl and B. Dalum and are discussed in more detail in OECD (2000c, Annex 2). The work in this chapter could be seen as something of a ‘patchwork’ that tries to combine various indicators, which, to a reasonable extent, are internationally comparable. This is at the expense of a single coherent ICT deﬁnition applied throughout the chapter. According to Table 7.1, ICT manufacturing accounts for less than 2 per cent of total Danish private sector employment and an even smaller share of exports. Communication equipment accounts for more than half of ICT manufacturing employment and less than 25 per cent of its exports. In comparison, the Finnish and Swedish employment shares are higher, 4 per cent and 3 per cent of total private sector employment, respectively, but the difference in total ICT employment shares between these three Nordic countries is, however, not so great. In Sweden it is almost 10 per cent, while the Danish and Finnish ICT employment shares are around 8.5 per cent (Nordic Statistical Institutes 2000). The service segment of Danish ICT is by far the largest measured by employment. The Information Technology (IT) services and consulting segment is a subset of total ICT services. In terms of employment this segment is larger than the total ICT manufacturing activities, but in terms of exports it is less than twothirds of manufacturing, which indicates that IT services and consulting are directed towards the domestic market. The Danish ICT sector experienced substantial growth in the period 1992 to 1998 in terms of turnover per employee (35 per cent), employment (20 per cent) and value added per employee (25 per cent) – all in current prices. Generally, value added per employee has increased faster than employment, indicating increased productivity. The increase in employment can primarily be assigned to ICT services while ICT manufacturing has shown below average growth. However, the size and competitive performance of the aggregate ICT sector is weak in the international context. For a group of 15 OECD countries in 1997 only one, Iceland, had a GDP share of the ICT industries lower than Denmark’s (OECD 2000c, Table 1).7 In terms of competitive performance, Danish ICT manufacturers have lost market share within the OECD market 1988–96, as indicated in Figure 7.1. OECD imports of ICT equipment have grown more than 120 per cent, while Danish exports of ICT goods have increased by 68 per cent. 7.3.1 International specialization of ICT manufactured goods The structure of Danish ICT manufacturing exports is shown in Table 7.2 which presents OECD trade statistics to compare the export structures of various industries across 12 OECD countries. Export specialization is a convenient measure for analysing the relative export structure of a country vis-à-vis the average pattern of a relevant group of countries. Among the countries shown Ireland was the only small country with a specialist export trade, which was for the high-technology industries in

Knowledge ﬂows in the Danish ICT sector 121 240 220 200 180 160 140 120 100 80

1988

1989

1990

1991

Danish exports

1992

1993

1994

1995

1996

OECD average imports

Figure 7.1 Danish ICT export performance of manufactured goods at the OECD market (source: based on EFS (2000)). Note OECD average imports of ICT goods are a weighted average of 22 member countries.

1990.8 During the 1990s this pattern changed somewhat, to include Sweden, the Netherlands and Finland.9 In the Swedish and Finnish cases two ICT companies, Ericsson and Nokia, were the major forces behind the increased high-tech specialization (particularly with production of mobile communications equipment). In Sweden, pharmaceuticals were also important in increasing the specialization in high-tech. Exports from the Netherlands were specialized in pharmaceuticals and the computer industry. For the Dutch ICT industry export pattern this can be related directly to Philips in terms of non-Dutch production (and thus exports) of consumer electronics and lack of success in telecommunications hardware. Ireland stands out as by far the most high-tech export intensive OECD country in 1996 (44.3 per cent of manufacturing exports).10 At the same time, all of the small countries listed in Table 7.2 have the common characteristic of being specialized, though generally decreasingly so, in exports from the low-tech industries. However, the so-called lowtech industries in these countries are often based on very advanced process technologies. The technology intensities used in these standard OECD classiﬁcations are international averages that may hide substantial national differences. From Table 7.2 the lack of Danish specialization in ICT hardware at the aggregate level is striking. Among the high-tech industries Denmark’s only

0.8

Denmark

0.7

0.9 1.1 1.1 1.0 0.7 0.2 2.5 0.4 1.4 1.5 1.4

1996

■

2.6

0.4 1.6 0.8 1.4 0.9 0.6 3.1 0.6 1.6 0.9 0.2

1990

2.7

0.2 1.9 1.1 1.4 1.0 0.4 3.1 0.9 1.7 0.7 0.3

1996

Pharmaceuticals ■

0.5

0.3 0.6 1.2 0.7 0.6 0.0 5.0 0.6 1.5 1.8 1.8

1990

■

0.8

0.9 1.0 0.7 0.7 0.7 0.1 1.2 0.4 0.9 1.4 2.6

1990

0.6

1.4 1.5 0.8 0.7 0.6 0.1 1.8 0.3 1.1 1.3 1.9

1996

Radio, TV and communication ■

2.1

2.3 1.3 1.5 1.2 0.7 3.0 1.5 1.5 0.8 0.8 0.2

1990

Low-tech

2.1

1.9 1.1 1.4 1.1 0.7 2.9 1.1 1.5 0.8 0.8 0.1

1996

■

3.7

0.3 0.3 2.6 1.6 0.6 3.0 3.1 0.7 0.9 1.0 0.1

1990

3.5

0.4 0.3 2.6 1.6 0.7 3.4 2.3 0.8 0.9 1.0 0.1

1996

Food processing ■

3.0

3.8 2.7 0.6 0.7 0.8 0.3 0.3 1.7 0.3 0.8 0.1

1990

3.2

3.3 2.7 0.5 0.7 0.6 0.2 0.2 1.9 0.3 0.7 0.1

1996

Wood and furniture

Notes 1 Western Germany in 1990, uniﬁed Germany in 1996. The export data only contain manufactured goods (in current US$). ‘Specialization’ is the relative share of national exports from a given industry compared to the OECD average. If the indicator is above 1.0 the country has an above-average export share – i.e. is ‘specialized’.

0.5

0.5 0.3 2.1 0.7 0.5 0.1 4.9 0.4 1.6 1.6 1.6

1996

Computers

Source: Adapted from OECD (1999b: Appendix Table 12.1.1).

0.5 0.8 0.8 0.9 0.6 0.1 2.2 0.5 1.4 1.8 1.5

Finland Sweden The Netherlands France Germany1 Greece Ireland Italy The UK The US Japan

1990

High-tech

Table 7.2 Export specialization, 1990–96 – selected countries compared to OECD average

Knowledge ﬂows in the Danish ICT sector 123 period of specializing in pharmaceuticals was 1990–96. A more ﬁnegrained analysis may, however, modify this picture slightly. Figure 7.2, based on OECD trade data according to the SITC classiﬁcation, shows that Denmark was specialized in telecommunications equipment in 1998. Electro-medical equipment also shown in Figure 7.2, is not included in the OECD and Nordic deﬁnitions of ICT. It is, however, closely connected to what is usually considered part of the electronics industry and is shown here because it is the only segment of electronics in which Denmark has been specialized since the early 1960s (Dalum et al. 1988). The 1998 snapshot depicted in Figure 7.2 is extended backwards for 1990–98 in Table 7.3. Again, the seven countries studied in the KNOW project plus Sweden, Finland and the USA are represented. The general absence of export specialization in ICT equipment is shown not to be conﬁned only to Denmark: it also applies to Italy, France and Greece. Even Germany is shown only to have been specialized in the instruments group. The United Kingdom and more particularly the USA are strong in ICT with specialization in several segments. 6 5 4 3 2 1

A US

UK

ed en Sw

Ita ly Ne th er lan ds

ee ce Gr

y an Ge rm

nc e Fra

lan d Fin

De

nm

ar

k

0

Electro-medical

Telecommunications equipment

Consumer electronics

Computers

Electronic components

Office machinery

Instruments, etc.

ICT total

Figure 7.2 Export specialization in ICT and electro medical products, 1998 (source: based on data from OECD (2000e) International Trade by Commodities Statistics, No. 1. Note Table A14.1–4 in the appendix contains the speciﬁcation of the various ICT products. Electro-medical equipment is not included in the ICT total.

0.45 1.18 0.62 2.08 3.31 0.33 0.77 0.26 0.13 0.63 0.90 0.42 1.02 0.68 1.71

1.96 2.80 0.36 0.78 0.24 0.11 0.68 0.81 0.47 0.85 0.61 1.73

0.53 0.60

0.32

0.44 1.24 0.62

0.64 0.63

0.64 0.59

0.32

0.93 0.86

0.55 0.47 0.51 1.19 1.16 1.26 0.68 0.68 0.65 Finland Electro medical 1.57 1.91 1.93 Telecommunications 2.38 1.72 2.16 equipment Consumer 0.54 0.45 0.34 electronics Computers 0.39 0.46 0.74 Electronic 0.26 0.24 0.25 components Ofﬁce machinery 0.05 0.07 0.07 Instruments etc. 0.74 0.76 0.69 ICT total 0.66 0.57 0.70 United Kingdom Electro medical 0.72 0.62 0.69 Telecommunications 1.09 1.07 1.05 equipment Consumer 0.70 0.77 0.61 electronics Computers 1.61 1.64 1.62

0.38

0.47

0.54

0.89 0.84

1994

0.57 0.55

0.87 0.83

Germany 0.73 0.71 0.74 0.83

1993

0.59 0.57

Electro medical Telecommunications equipment Consumer electronics Computers Electronic components Ofﬁce machinery Instruments etc. ICT total

1992

1991

1990

1.77

0.71

0.43 1.22

0.09 0.64 1.00

0.64 0.23

0.30

1.93 4.10

0.50 1.17 0.62

0.52 0.63

0.31

0.96 0.81

1995

1.63

0.81

0.40 1.39

0.11 0.58 0.99

0.60 0.26

0.27

2.35 4.06

0.50 1.20 0.61

0.49 0.61

0.34

0.95 0.78

1996

1.63

0.72

0.41 1.39

0.11 0.65 1.06

0.65 0.29

0.32

2.73 4.01

0.46 1.11 0.61

0.50 0.63

0.28

0.90 0.81

1997

1.60

0.67

0.41 1.57

0.16 0.65 1.24

0.55 0.37

0.33

2.46 4.99

0.46 1.14 0.60

0.53 0.59

0.26

1.06 0.70

1998

Table 7.3 Export specialization in ICT and electro medical products, 1990–98 1991

0.80 0.54

0.01 0.04

0.01 0.02

0.05

0.01

0.03 0.30

0.59 0.72 0.65

0.81 0.54

0.39

0.35 0.84

0.42 0.93 0.43

0.19 0.98 0.54

0.50 0.15

0.69

1.82 0.95

1992

0.21 0.96 0.54

0.51 0.15

0.72

0.54 0.60 0.77 0.74 0.67 0.67 Greece 0.08 0.03 0.21 0.25

0.76 0.56

0.42

0.42 1.07 0.57 France 0.41 0.94

0.42 0.16

0.76

Denmark 2.20 1.90 1.16 0.93

1990

0.06

0.03

0.10 0.51

0.57 0.71 0.63

0.72 0.61

0.35

0.34 0.80

0.22 0.98 0.49

0.49 0.20

0.50

1.74 0.78

1993

0.04

0.03

0.03 0.33

0.55 0.67 0.62

0.68 0.63

0.33

0.37 0.76

0.20 0.98 0.53

0.59 0.18

0.56

1.81 0.79

1994

0.04

0.03

0.15 0.30

0.64 0.69 0.67

0.68 0.74

0.36

0.32 0.80

0.21 0.94 0.51

0.48 0.16

0.72

1.87 0.72

1995

0.04

0.10

0.11 0.20

0.68 0.69 0.72

0.76 0.83

0.37

0.33 0.81

0.17 0.85 0.52

0.48 0.17

0.74

1.68 0.83

1996

0.07

0.16

0.07 0.26

0.61 0.70 0.75

0.78 0.85

0.37

0.37 0.92

0.19 0.89 0.56

0.47 0.20

0.76

2.06 0.95

1997

0.09

0.19

0.10 0.43

0.50 0.64 0.77

0.77 0.81

0.43

0.28 1.07

0.15 0.94 0.58

0.41 0.21

0.78

1.70 1.14

1998

0.21 0.47 0.18 0.66 0.53 0.37 0.53 0.49 0.63 3.48 0.12 0.39 0.31 0.21 1.00 0.80

0.20 0.47

0.19

0.66 0.55

0.41 0.50 0.50

0.59 3.12

0.12

0.59 0.30

0.22 1.02 0.80

0.21 0.97 0.85

0.28 0.34

0.16

0.49 3.75

0.32 0.49 0.45

0.57 0.53

0.17

0.27 0.40

0.63 1.32 1.18

1.02

0.24 0.98 0.93

0.23 0.33

0.18

0.66 4.23

0.39 0.49 0.45

0.51 0.52

0.17

0.27 0.44

0.65 1.34 1.29

1.15

0.26 0.89 1.11

0.23 0.50

0.19

0.51 5.12

0.32 0.50 0.41

0.44 0.51

0.17

0.26 0.42

0.79 1.35 1.28

1.06

0.21 0.79 1.20

0.21 0.48

0.17

0.46 5.39

0.27 0.46 0.37

0.35 0.45

0.12

0.24 0.44

0.85 1.43 1.22

0.85

0.21 0.78 1.19

0.20 0.43

0.28

0.98 5.21

0.28 0.44 0.34

0.32 0.40

0.12

0.22 0.44

0.76 1.40 1.23

0.82

0.17

0.31

1.18 0.42

0.40

0.62 1.55 1.37

1.96 1.38

0.37

0.63 1.58 1.32

1.91 1.25

0.35

2.27 2.27 0.53 0.57 0.85 0.79 The USA 3.37 3.30 1.40 1.39

1.33 0.43

0.44

0.56 1.57 1.28

1.71 1.32

0.33

3.43 1.49

2.25 0.58 0.91

1.41 0.75

0.39

0.77 0.59

0.03 0.13 0.18

0.29

0.51 1.56 1.31

1.64 1.48

0.36

3.51 1.47

2.72 0.57 0.91

1.41 0.73

0.37

1.00 0.50

0.06 0.15 0.16

0.31

0.52 1.55 1.34

1.56 1.62

0.35

3.60 1.46

2.73 0.55 0.95

1.46 0.80

0.40

1.30 0.52

0.06 0.21 0.18

0.36

0.62 1.55 1.35

1.60 1.66

0.38

3.68 1.30

2.75 0.61 1.15

1.94 0.91

0.52

1.21 0.53

0.07 0.28 0.16

0.29

0.64 1.57 1.34

1.54 1.71

0.38

3.50 1.26

2.85 0.64 1.23

2.19 1.04

0.47

1.35 0.40

0.05 0.30 0.19

0.30

0.88 1.58 1.39

1.52 1.84

0.40

3.37 1.19

2.73 0.70 1.35

2.60 1.04

0.47

1.32 0.34

0.05 0.22 0.21

0.22

Note Electro medical is not included in the ICT total. Instruments etc. are instruments and equipment for detecting, measuring, checking and controlling physical phenomena or processes.

0.64 1.56 1.31

1.92 1.27

0.35

3.38 1.41

2.40 0.61 0.81

1.24 0.44

0.39

0.02 0.01 0.03 0.06 0.07 0.08 0.10 0.10 0.15 The Netherlands 0.58 0.82 0.77 0.73 0.66 0.64

0.22

Source: Based on data from OECD (2000e) International Trade by Commodities Statistics, No. 1.

0.54 1.32 1.12

0.74 1.37 1.04

0.78 0.74 1.49 1.44 1.09 1.08 Italy Electro medical 0.13 0.17 Telecommunications 0.56 0.53 equipment Consumer 0.24 0.19 electronics Computers 0.75 0.72 Electronic 0.49 0.51 components Ofﬁce machinery 0.35 0.40 Instruments etc. 0.48 0.46 ICT total 0.53 0.51 Sweden Electro medical 0.69 0.58 Telecommunications 3.08 3.15 equipment Consumer 0.16 0.14 electronics Computers 0.70 0.66 Electronic 0.33 0.29 components Ofﬁce machinery 0.27 0.25 Instruments etc. 0.99 1.05 ICT total 0.83 0.82

0.95

0.62

0.61

0.57

Electronic components Ofﬁce machinery Instruments etc. ICT total

126 C.R. Pedersen, M.S. Dahl and B. Dalum It should be borne in mind that these results are sensitive to the methods used, such as level of aggregation, choice of year, etc., but that the specialization patterns are generally fairly stable or ‘sticky’ over time (Dalum et al. 1998). The indicator is, for example, below one in the case of telecommunications for France, except for 1998 where it is above one (Table 7.3). The main feature of Danish telecommunications in the 1990s is that, apart from 1990 and 1998, it was generally below one. However, the two examples from telecommunications indicate that further disaggregation of the data might reveal segments where these countries are specialized. Table 7.4 presents export ﬁgures for the large and rapidly growing mobile communications equipment segment of telecommunications in the 1990s. In this segment Denmark has been persistently specialized in the 1990s, while France and Germany only embarked on a phase of specialization in the late 1990s. From this international comparison of the manufacturing segment of ICT it cannot be concluded that the relative weakness of Denmark is an isolated phenomenon in the EU. It also applies to such diverse countries as Germany (except for instruments), France, Italy and Greece. The cases of Sweden, Finland and the Netherlands illustrate that small country size is not in itself an inhibiting factor. What appears to be important is whether historical evolution of their respective national systems of innovation has led to the formation of large ‘domestic’ multinationals of the Ericsson–Nokia–Philips type. This was evidently not the case in Denmark. Denmark has generally been characterized by small- and medium-sized ICT ﬁrms, as discussed in Dalum et al. (1988). In the mid-late nineteenth century and the early decades of the twentieth century the Great North-

Table 7.4 Export specialization in mobile phones, 1990–98

Denmark Finland France Germany Greece Italy The Netherlands Sweden The UK The USA

1990

1991

1992

1993

1994

1995

1996

1997

1998

3.38 4.55 0.62 0.68 0.03 0.57 0.08 3.83 1.82 1.99

2.02 4.69 0.67 0.64 0.05 0.67 0.06 4.45 2.10 2.07

1.99 7.90 0.35 0.55 0.03 0.65 0.09 4.96 1.71 2.36

1.87 8.95 0.53 0.60 0.05 0.49 0.11 6.17 0.90 2.45

1.30 8.68 0.48 0.95 0.11 0.25 0.10 6.08 1.30 2.00

1.07 9.50 0.65 1.08 0.11 0.27 0.11 7.23 1.26 1.68

1.29 9.59 0.85 1.11 0.19 0.20 0.08 7.75 1.84 1.26

1.59 8.20 1.11 1.21 0.26 0.19 0.11 8.05 1.67 1.16

2.02 9.72 1.24 1.00 0.48 0.20 0.08 7.50 2.11 0.95

Source: Based on data from OECD (2000e) International Trade by Commodities Statistics, No. 1. Note Mobile phones are deﬁned as SITC Rev. 3 code 764.32. The specialization index is based on OECD ( 2000e).

Knowledge ﬂows in the Danish ICT sector 127 ern Telegraph Company was a large international player in telegraph technology, which could have become an embryo ‘Danish Ericsson’. However, Great Northern never managed to enter telephone technology with the thrust that characterized Ericsson in Sweden and Siemens in Germany. In the early post World War II period Great Northern entered the emerging radio communications ﬁeld through the acquisition of a small start-up, Storno. This sector became the stronghold in Europe of Ericsson and Storno alongside Motorola and General Electric in the US. In the 1950s and 1960s Storno was the third largest producer worldwide, after these two US companies. Storno produced what is today known as ‘closed’ radio communications systems for the police forces, airports, transport companies, etc. Storno and Ericsson were heavily involved in developing the path-breaking Nordic Mobile Telephony (NMT) system launched in 1981 as the world’s ﬁrst cross international public (or ‘open’) mobile communications system, a predecessor of the successful GSM (Groupe Spéciale Mobile) system. While the early 1980s witnessed phenomenal growth of Nokia in Finland, originally based on mobile phone terminals, and later on entire systems, a ‘counterfactual’ discussion was ongoing in Denmark about whether or not, at a certain stage, a ‘Danish Nokia’ had been a real option. There seem to be no ‘basic laws’ of economics that prevented another path of development for the Danish telecommunications industry. Although small, Storno was a leading international company, and a distinct research tradition in the wireless ﬁeld had already emerged at the Technical University of Denmark in Copenhagen in the early decades of the twentieth century. The Great Northern group sold Storno to General Electric (US) in 1976, and the company was passed on to Motorola in 1986. In 1976, Great Northern, like many other companies, did not recognize the enormous market potential of mobile telephony. If it had, the Danish ICT manufacturing landscape would be signiﬁcantly different. Motorola used Storno as an entry port in order to become a more central player in the early stages of the GSM standardization process, responsibility for which was assumed in 1987 by a new body, the European Telecommunications Standards Institute, ETSI. To become a player in the ETSI context required a presence in Europe. As Motorola’s European development arm, Storno managed to develop a GSM terminal in 1992 in line with Ericsson, Nokia and the north Denmark DC Development, a joint venture between two domestic mobile phone producers in the Aalborg region. Of the ﬁrst four GSM phones developed, two were developed in Denmark, one in Sweden and one in Finland. Motorola subsequently moved its GSM development activities to the US; the two Aalborg ﬁrms were ﬁnancially drained by the development efforts, preventing them from entering a growth process of Ericsson–Nokia dimensions. However, mobile phone equipment has continued to be an area of specialization in Denmark as can be seen from Table 7.4. The patterns for

128 C.R. Pedersen, M.S. Dahl and B. Dalum Sweden and Finland in the 1990s are remarkable. Germany (Siemens) and France (Alcatel) have also become specialized. The USA, in spite of being home to one of the top three mobile phone producers, Motorola, had experienced a decreasing specialization in mobile phones in the last half of the period. The lack of success of US manufacturers in second generation (2G) mobile phone technology compared to European and Asian manufacturers is apparently responsible for this loss of specialization. In addition to mobile communications Denmark is specialized in the electro-medical equipment ﬁeld, which is dominated by small- and medium-sized enterprises (SMEs), who have developed highly specialized niche products, often through intense collaboration with hospital doctors. This structure also characterizes the instruments segment in the 1960s to the 1980s. The most important knowledge ﬂows for these niche ﬁrms have generally been a combination of strong interaction with the users in the innovation process and strong links with university research. While the trade data for ICT hardware can be used as a reasonable structural indicator in an international context, it is much more difﬁcult to make international comparisons of the ICT service industries. During the 1980s and 1990s standardized – and machine independent – packages of software became very widespread. The US (and Canada) have generally dominated this market through Microsoft (Windows, Word and Excel), Corel (WordPerfect), Lotus, SAS, SPSS and Oracle. At the less standardized end of the market, US companies such as EDS, Computer Associates and the omnipresent IBM have dominated the ICT services market through semi-standardized software packages, which have then undergone extensive modiﬁcation by each company. ICT services and consulting has been domestically oriented in most countries. Although software packages (and the hardware) have become standardized, implementation and modiﬁcation of large ICT service systems continued to be a bespoke business activity well into the 1990s. Packages such as Oracle and the German SAP have emerged as more standardized solutions for larger ﬁrms. In the 1980s there were two main segments in ICT services and consulting. Mainframe-based service suppliers comprised a mix of afﬁliates of large US hardware companies (especially IBM) and US (such as EDS) and national consultancy companies. In the Danish case, the latter segment was composed either of large domestic companies (such as Maersk Data, ØK Data and LEC) or banks (PBS, Bankdata, SDC) and government organizations (Kommunedata and Datacentralen). Following the emergence of microcomputers an entire segment of personal computer (PC) distributors evolved. Due to the typically rather centralized administrative procedures of government in Denmark, the mainframe-based service suppliers (both private and government owned) tended to be dominant during the 1980s. But during the 1990s application of the ﬁrst stand-alone PCs and then PC

Knowledge ﬂows in the Danish ICT sector 129 networks and the Internet gave rise to the emergence of a dot.com segment. In Denmark, this segment never matched that of Sweden in the 1990s (represented by such high ﬂyers as Framfab, Icon Media Lab, Cell Network and Adcore). The government-owned Kommunedata and Datacentralen in Denmark were privatized at the end of the 1990s. Kommunedata (IT solutions for the municipal administrations) underwent privatization but without being changed fundamentally. Datacentralen was sold to CSC (a large US IT consultancy company). These two companies are quite big, even in an international context, but very much oriented towards the domestic market. A small segment for software applications emerged in the 1990s. Two ﬁrms, Navision and Damgaard Data, managed to develop total IT solutions for SMEs in terms of standardized packaged software with a certain international market presence. These two ﬁrms merged to become NavisionDamgaard, which became internationally known and was subsequently acquired by Microsoft, who wanted to enter this software segment. The mainframe-based IT service ﬁrms experienced tremendous pressure during this period from PCs and network distributors. During the 1990s, solutions based on PCs connected in networks to an increasing extent, were substituted for mainframe based solutions. However, by the end of the 1990s and at the beginning of the 2000s, these two groups of actors increasingly converged. A trend towards IT solutions based on PCs in networks with a centralized structure for servers and software, has changed the direction of development once again. Due to the heavy ﬁnancial pressures on the sector from the beginning of 2000, substantial restructuring took place and this is still ongoing in 2004. The trend would appear to be towards a small group of ﬁrms capable of delivering, and even managing, total IT solutions for private ﬁrms as well as government authorities. The ﬁrms in the segment have been converging with the old mainframe-based IT consultants to become a group of complete service and network suppliers. Many of these ﬁrms, for example, Aston Group, Eterra and WM Data, which all grew out of the distribution of PCs and standardized (US) software packages, have experienced severe economic problems and some have ﬁled for bankruptcy. The dot.com segment has been shrinking rapidly due to the crisis in the IT sector and the cut in IT spending. The structural change is still ongoing, but without reservation it can be concluded that the Danish IT service and consulting segment is not, and will not be, a major international player. This would hold even if the telecommunications service segment were included as assumed by the term ICT services. The telecom service sector accounts for 16 per cent of total ICT employment in Denmark (Table 7.1), but is domestically oriented. The one-time government monopoly company, TDC, employed more than 16,000 persons, but was not a major international player. It was acquired by US Ameritech in 1998, which in its turn was taken over by SBC.

130 C.R. Pedersen, M.S. Dahl and B. Dalum The mobile communications segment has similar characteristics. The biggest company is TDC, followed by the Aalborg-based company, Sonofon, owned by Norwegian Telenor. The mobile service providers (also including afﬁliates of large foreign companies, such as Telia and Orange) are technically fairly advanced in an international context. But they are not large companies and they have basically become dominated by foreign multinationals. To summarize, the general impression of a rather weak Danish ICT sector appears well founded.

7.4 What lies behind the relative weakness of the Danish ICT sector What are the major factors behind this pattern? As a ﬁrst step towards answering this, we analyse data on the ICT education structure. The Danish higher education system and especially the system for training skilled workers, are generally considered to be of a high standard. Table 7.5 presents some basic data from the Education at a Glance 2000 report (OECD 2000a). As can be seen from the table, a large proportion of Denmark’s population have upper secondary as well as post-secondary practical-oriented education. However, the shares of graduates with at least three years of study, and of graduates in science and engineering are rather low. But these data can be disputed: Education at a Glance 1998 used a different method of classiﬁcation, categorizing into non-university and university level education. In both these categories, Denmark performed better than the OECD average.11 The Nordic Statistical Institute’s (2000, 1998) efforts to produce comparable data on ICT activities represents a step towards international comparisons at the detailed level, as shown in Table 7.6. The high numbers of university educated employees in the Nordic ICT industries are striking. The close similarities in their education systems are probably the reason for this. However, Table 7.6 shows that Denmark is systematically ranked below the other Nordic countries. Of these four countries, Finland has by far the biggest proportion of higher educated employees in ICT. The relative small and internationally weak Danish ICT sector is explained in part by the small numbers of engineers and computer scientists being produced. There are two technical universities in Denmark producing engineers in electronics, at the bachelors (diploma engineers), masters and doctorate levels: the Technical University of Denmark (DTU) and Aalborg University (AAU). DTU has been internationally known in electrical engineering for two centuries. AAU, founded in 1974, became quickly established in electronics, and was responsible for educating half of the Danish masters students in electronics from the early 1990s. There is also

32 24 22 39 16 54 49 36 19 46 38

39 48 53 40 61 30 30 40 57 57 44

Upper

17 15 20 10 9 4 10 0 8 10 8

least two years, focusing on practical skills

Notes 1 1997. 2 1995. 3 1996. 4 1993.

13 13 5 11 14 11 11 24 15 12 14

Education of at least three years theoretical duration

Post-secondary tertiary education

■Education of at

Source: Adapted from OECD (2000d: Appendix Table 7), based on OECD (2000a).

Finland1 Sweden Denmark France Germany Greece1 Ireland The Netherlands The UK EU OECD

Below upper

Primary and secondary education

Table 7.5 Percentage of population aged 25–64 by level of educational attainment, 1998

0.082 0.073 0.042 0.164 0.092 0.064 0.253 – 0.192 0.12 0.09

Graduates in science and engineering (% of employment)

17 20 16 13 31 8 11 9

8 7

8 4

6 5 4 4

University level tertiary

37 13 15 13

24 17

25 27

Total tertiary

■

22 7 12 10

16 19

14 19

Nonuniversity tertiary

Norway3

35 7 11 9

13 21

24 24

University level tertiary

57 14 23 19

29 40

38 43

Total tertiary

59 24 31 28

51 44

45 52

Finland ■ Non-university and university tertiary

62 18 24 21

35 28

37 47

Sweden ■ Non-university and university tertiary

Notes 1 Tertiary education is equal to ISCED level 5 (Non-university tertiary education) and ISCED Level 6 and 7 (University level tertiary education). 2 1996 level. 3 1999 level. 4 Total manufacturing (NACE 15–37). 5 Total services activities (NACE 50–74, 92). 6 Total private sector (NACE 15–37, 45, 50–74, 92, 93).

Source: Based on Nordic Statistical Institutes (1998, 2000).

ICT manufacturing ICT services Wholesale of ICT products Telecommunications ICT services and consultancy Total manufacturing4 Total services activities5 Total private sector6

Nonuniversity tertiary

Denmark2

Table 7.6 Percentage of employees in ICT with tertiary education,1 1998

Knowledge ﬂows in the Danish ICT sector 133 a system of decentralized engineering schools providing bachelors-level training in electrical engineering, originally set up to offer a higher education opportunity for skilled workers to follow a tailor-made admissions programme. There are six such schools. Computer scientists are trained at the Copenhagen, Aarhus and Aalborg universities, and the University of Southern Denmark has recently started a programme in electrical engineering. The country is thus well provided with institutions capable of educating the relevant manpower; the problem is that too few students want to take these degrees – in spite of increased job opportunities in the second half of the 1990s.12 The relative distribution of R&D expenditure within manufacturing industries is shown in Table 7.7 for a group of small advanced OECD countries, as well as for the entire OECD group. The data also include a comparison according to four levels of R&D intensity. The Danish (and Dutch) emphasis on food processing related R&D is outstanding with a share three times as large as the OECD average of 2 per cent. The relative specialization in pharmaceuticals R&D is also approximately 3, but this represents 30 per cent of Danish R&D expenditure in manufacturing. Conversely, Danish R&D efforts in ICT manufacturing are generally very small in international terms. A more direct indicator is the OECD measure of ICT research in total business enterprise expenditure on R&D (OECD 2000c, Table 14). Denmark lagged behind the 14 most developed OECD countries in 1997 with a share of 7.2 per cent and was also at the lower end in terms of business sector share of R&D for communications (telecommunications) services. The high degree of concentration of R&D expenditure on a few industries for the small high-income OECD countries shown in Table 7.7 is quite clear. In Finland the increase in specialization from 1.3 to 3.1 in telecommunications equipment and semiconductors reﬂects that this industry accounted for 50 per cent of Finnish manufacturing R&D in 1998, basically due to the very rapid growth of Nokia. For Sweden, the dominance of Ericsson is less outstanding, although this industry contributes with more than 20 per cent of Swedish manufacturing R&D.13 From the innovation data, the available evidence on Danish ICT patenting in the 1990s presented in Table 7.8 also indicates a rather low relative share vis-à-vis the OECD countries considered, and also the entire group of OECD and EU countries. The share and the average annual growth rate for 1992–98 are very low. However, comparison of the 1992–98 and 1992–99 growth rates indicates underlying data problems – probably the small numbers – which may disturb the pattern substantially. Finland has experienced the highest growth among the countries shown with 30 per cent average annual growth in the 1990s. The ICT sector accounted for almost one-third of total patenting in 1999. The Swedish pattern is somewhat similar, but less pronounced. The reason is evident: there are no ‘domestic’ large ICT multinationals in Denmark.

2.0 0.6 0.2 11.5 21.2 7.6 1.2 2.1 71.1 6.0 9.8 14.5 0.1 12.6 14.4 4.9 0.6 100.0 46.5 40.7 9.0 3.7 100.0

2.0 0.6 0.4 1.3 21.6 9.6 1.1 1.8 70.7 7.1 7.1 16.3 0.2 14.4 10.7 7.3 0.6 100.0 43.8 44.1 8.0 4.2 100.0

1995

■ 3.5 1.0 1.8 0.5 1.4 3.1 1.8 0.7 0.7 2.6 0.3 0.7 21.5 0.0 0.0 2.4 14.4 1.0 0.8 0.9 1.9 2.3 1.0

1990

Denmark2

3.1 0.2 0.8 0.3 1.6 3.1 0.7 0.3 0.7 3.2 0.2 0.6 13.0 0.0 0.0 1.2 12.1 1.0 0.9 0.9 1.8 1.7 1.0

1998

■ 1.0 0.6 0.5 3.6 0.9 1.8 0.5 1.0 1.0 2.3 0.3 1.9 2.9 1.3 0.4 0.2 0.4 1.0 1.1 1.0 0.6 1.6 1.0

1990

Sweden

0.7 0.4 1.0 2.4 0.9 1.7 0.5 0.8 1.0 1.7 0.2 1.4 0.5 1.3 0.5 1.1 0.2 1.0 1.1 1.0 0.6 1.2 1.0

1995 2.9 0.8 0.0 0.4 1.8 1.1 0.3 0.7 0.7 0.5 0.4 1.1 1.0 0.4 0.1 0.2 0.0 1.0 0.7 1.4 0.8 1.8 1.0

3.7 0.8 1.4 0.6 1.5 1.2 0.5 1.2 0.8 0.5 0.7 0.9 1.1 0.5 0.2 0.2 0.2 1.0 0.8 1.1 1.0 2.2 1.0

The Netherlands ■ 1990 1996 ■ 3.4 1.6 5.1 8.7 1.0 0.7 1.8 1.7 0.8 2.3 0.3 1.3 10.2 0.1 0.0 1.1 0.8 1.0 0.6 1.0 1.7 4.8 1.0

1990

Finland

1.2 1.2 2.3 3.2 0.6 0.4 0.9 0.8 1.1 1.7 0.1 3.1 1.4 0.0 0.0 0.6 1.0 1.0 1.3 0.6 1.1 2.0 1.0

1998

Notes 1 Distribution of OECD R&D manufacturing expenditure. 2 The national distribution divided by the OECD distribution. The weighted average per country is equal to 1.0. If above 1 the countries are ‘R&D specialized’ in the given industry and vice versa if below 1. 3 Not all sub-groups are shown. Sub-groups are shown in italics.

Source: Adapted from OECD (2000d: Appendix Table 28).

Food, beverages, tobacco Textiles and clothing Wood and furniture Paper and printing Chemicals Pharmaceuticals Non-metal mineral products Basic metals Metals and machinery3 Non-electrical machinery Computers Telecommunications and semiconductors Shipbuilding Motor vehicles Aerospace Scientiﬁc instruments Other manufacturing Total manufacturing High-tech industries Medium-high technology Medium-low technology Low technology Total manufacturing

1990

OECD-141

Table 7.7 R&D specialization in manufacturing for selected small OECD countries

Knowledge ﬂows in the Danish ICT sector 135 Table 7.8 ICT patents granted at the United States Patent Ofﬁce Share of ICT patents in total

Finland Sweden The USA Ireland The Netherlands Japan Denmark European Union Total OECD

■

Growth rate

Growth rate ■ 1992–99

1992

1998

1999

1992–98

6.0 7.3 8.8 14.2 10.2 14.1 6.4 6.2

29.0 16.8 18.4 24.4 16.6 21.0 3.1 11.0

30.4 16.9 17.5 16.5 14.4 18.5 6.3 10.4

41.7 28.2 21.5 17.3 15.2 13.2 0.0 15.7

31.8 23.2 16.6 11.7 10.3 8.9 13.0 12.5

9.5

17.6

16.4

18.6

14.1

Source: Adapted from OECD (1999b: Appendix Table 11.3.1 for 1992 and 1998); OECD (2000d: Appendix Table 36 for 1999).

Nokia and Ericsson have been exceptionally aggressive in patenting in mobile communications-related technologies, while the Danish ICT sector is dominated by small R&D oriented ﬁrms with a low propensity for patenting. Denmark could be described as a development ‘hub’, especially for terminals for wireless (mobile, cordless and satellite) communications, equipment for optical communications and highly specialized electronic measurement equipment. This type of equipment is typically developed by local R&D units of large multinationals, which probably result in patents being taken out by the headquarters outside Denmark. Alternatively, the innovations may come from small ﬁrms contracted by large multinationals. In this case the process does not typically result in a patent application. However, innovation data based on a Danish questionnaire carried out in relation to the DISKO project (Lundvall 1999) contain a dataset on innovative activity, on which Table 7.9 is based, for a sample of 1,910 ﬁrms. As a rough illustration of variations in propensities to innovate across different clusters, ﬁrms reporting product innovations in the DISKO panel data have been aggregated into ﬁve of the EFS mega clusters. According to this source, the ICT mega cluster has by far the highest rate of product innovation, as more than 75 per cent of the ICT ﬁrms have produced a product innovation. Only half of the DISKO ﬁrms have produced product innovations. The reported product innovations might not all have been patented. This indicates that the product innovation dynamics of the ICT sector are quite strong, at least compared with other parts of the Danish economy.

136 C.R. Pedersen, M.S. Dahl and B. Dalum Table 7.9 Relative number of product innovative ﬁrms in ﬁve EFS mega clusters, 1993–95 N = 1,910

Product innovation

No product innovation

Total

N

ICT MC Medico/health MC Agro-food MC Construction MC

75.1 57.8 52.4 38.3

24.9 42.2 47.6 61.7

100 100 100 100

Total economy

50.4%

49.6%

100% 1,910

249 93 248 517

Source: Based on data from the DISKO panel data 1993–95. Note The DISKO data set contains questionnaire data from a sample of 1,910 ﬁrms, representative of the private sector of the Danish economy.

7.5 ICT consumption pattern A substantial amount of data on the use of ICT equipment indicates that the markets of the small rich countries – as well as the US market – represent the most advanced demand. This is typically illustrated by per capita penetration ratios in such ﬁelds as Internet connections, mobile phones, PCs in homes, etc. (see, for example, OECD 1999a, 2000c: Appendix Tables 4–6). Among the small countries this especially applies to the Nordic area. The mobile communications ﬁeld represents a now classic case of close interaction between advanced demand characteristics, institutional set up, regulation and international competitiveness of the supplier industries, that is, the very essence of the factors emphasized in cluster analysis. The establishment of the Nordic Mobile Telephony (NMT) system in 1981, through cooperation between the Nordic telecom service providers and the regulatory authorities, created the ﬁrst ever cross-border market for public mobile telephony, based on a common standard. This was decisive in the Nordic ﬁrms (Ericsson, Nokia and the US–Danish Storno) becoming world leaders in mobile communications equipment from the early 1980s. This strong competitive position was further reinforced when the pan-European GSM standard, now the dominant world standard, was implemented in 1992. In the late 1980s the Nordic companies and authorities, with Nokia and Ericsson acting as the leading architects to a substantial degree, inﬂuenced the GSM standardization process. The biggest producer of mobile terminals at the time, Motorola, acquired Storno in Denmark to avoid being excluded from this important process. The story to a large extent was repeated in the 1997–98 speciﬁcations for the third generation (3G) mobile communications system, Universal Mobile Telecommunication System (UMTS), which was also dominated by Ericsson and Nokia. The development of the mobile communications

Knowledge ﬂows in the Danish ICT sector 137 industry is probably one of the most clear textbook examples of cluster dynamics. Previously, the Nordic countries were the lead-users, which no longer will be the case. The allocation of UMTS licences in Europe in 2000 was very turbulent, resulting in escalating costs for telecom service providers to become licence holders. The ensuing ﬁnancial crisis in the telecommunication services sector, accompanied by large scale redundancies and reduced expectations for 3G communications, has marred the introduction of 3G in Europe. The massive investments required to build the 3G infrastructure combined with the deep ﬁnancial problems of the telecommunications sector in general (the equipment hardware industry as well as telecommunications services) has caused the establishment of 3G networks and introduction of 3G phones to be delayed and postponed all over Europe. Nevertheless, in spite of these problems, 3G services have been introduced in some European countries: in Denmark services were introduced in October 2003, but it seems that their previous advantage of being an early advanced user has been lost. Given the advanced nature of demand in the Nordic region concerning the use of ICT products and services, it might be expected that a similar position could emerge in relation to the new software development and IT service industries, including the dot.com segment. On the international scene, the Nordic area in this respect has been characterized as ‘leading Europe in the technology revolution that will dominate the early years of the 21st century’ (Financial Times 2000). Besides the Nordic lead in mobile communications consumption, the Swedish Internet service segment grew very fast in the late 1990s, both on the consumption and the supply side. A large number of new dot.com ﬁrms with some international visibility emerged. The Swedish government gave more encouragement to the diffusion of high-speed Internet connections than governments in other comparable countries. Subsequently, there was a rather dramatic international downturn for the dot.com industry, which hit Sweden especially hard. The general characteristics of the leading role of the Nordic countries may still persist, but they would appear to be basically founded on their strengths in the wireless segment. Denmark has been lagging behind on the supply side reﬂecting the lack of an ‘Ericsson-Nokia’ effect, and on the consumption side in relation to high-speed Internet access. The Danish Government published a Network Report (Ministry of Research and Information Technology 2000), which contains policy reviews as well as a series of internationally comparable data on ICT consumption patterns. Concerning high-speed Internet access for households and small ﬁrms, Denmark ranks half way in the group of advanced countries.14 In international comparisons of ICT performance, the supply and demand side indicators are often combined with general macroeconomic indicators. This may give rise to confusion concerning the competitiveness of the various sub-segments of the ICT sector per se. One example of this

138 C.R. Pedersen, M.S. Dahl and B. Dalum can be found in the consultancy report produced by PLS Rambøll and Børsens Nyhedsmagasin (2000b). Ten western countries were benchmarked, using statistical data and the peer judgements of 170 experts in the ten countries. Table 7.10 shows the benchmark of the statistical indicators. The data contain information on ICT applications, R&D efforts, expenditure on IT education, macroeconomic stability and performance, characteristics of the capital and labour markets, etc.15 There is a group of clear ‘front runners’ (USA, Sweden and Finland); Denmark is among a group of ‘followers’ along with UK, Norway and the Netherlands, while the ‘main group’ is comprised of France, Germany and Spain. Bearing in mind the caveats that apply to this kind of methodology, it seems to ﬁt the pattern described above. Denmark appears to be fairly well positioned to exploit the business opportunities in ICT from the point of view of demand, but not to the extent say of Sweden and Finland.

7.6 Regionally clustered strong niches in the ICT sector Generally there are two factors that determine the geographical distribution of the ICT sector. It should be expected that one of the major location factors is a ‘metropolis’ effect. Many of the ICT service activities, such as software development, IT services and telecommunication services, are typically concentrated in cities (see Figure 7.3). This also applies to the supply of skilled labour, such as engineers, computer scientists, business economists, etc., which again are expected to be a function of location of universities and business schools, often determined by government decisions. The second factor is the somewhat ‘random’ location of ﬁrms, which is based on the preferences of their original founders. This emphasizes the combination of a random initial triggering event (such as individual founder preferences) and subsequent cumulative causation effects (mainly via external economy effects on the labour market) leading to concentrations of related industries, that is, geographical clusters. As an indicator of relative specialization, employment share of a given activity in a region may be compared with the national average. A value higher than 1 indicates an above-average employment share – that is, the region is ‘specialized’ in that activity, and vice-versa. Measured in this way, only two regions are seen to be specialized in ICT in 1992 and these are the regions around the two largest cities, Copenhagen and Aarhus.16 At municipality level, the specializations in ICT are again around Aarhus, Copenhagen and north of Copenhagen, but with additional single dots spread across the country, shown in Figure 7.3 as the grey and black regions. The third and fourth largest cities, Aalborg and Odense, are not specialized, but a map showing the share of total ICT employment (Figure A7.1 in the appendix) shows that ICT employment is concentrated in the larger cities.17

Total score

68.5 67.4 63.2 58.8 53.2 49.5 47.6 45.7 38.1 28.0

Un-weighted average

7.3 7.2 6.7 6.2 5.6 5.3 5.1 4.8 4.0 2.9

ICT labour force 0.83

9.0 3.3 7.3 5.0 6.3 4.0 2.0 7.8 5.3 5.3

ICT infrastructure 0.82

6.8 8.0 5.4 6.8 6.2 6.2 7.6 3.0 4.2 1.0

ICT education 0.82

3.0 6.5 7.5 8.5 4.5 2.5 5.5 7.0 6.0 4.0

Technocommercialization 0.77

8.5 8.5 8.0 6.5 3.5 5.0 2.5 6.0 2.5 4.0 0.77

9.0 8.0 6.0 5.0 10.0 4.0 7.0 3.0 1.0 2.0

Technology absorption 0.76

2.7 9.7 6.3 6.7 5.0 4.0 7.3 6.7 3.7 3.0

Capital markets 0.75

8.3 4.5 6.5 6.3 3.3 8.8 2.5 6.3 4.8 4.0

B2B e-commerce 0.70

10.0 9.0 7.0 3.0 6.0 8.0 5.0 2.0 4.0 1.0

Network culture 0.68

10.0 7.0 5.0 9.0 6.0 8.0 4.0 3.0 2.0 1.0

ICT consumers 0.66

8.4 8.8 7.4 5.4 6.2 5.8 5.0 1.8 4.6 1.6

0.64

5.6 7.1 7.3 5.9 5.7 3.6 6.0 4.9 5.6 3.9

Macro economy

Note See appendix Table A7.4 for explanation of the benchmark indicators. Total scores weighted according to the global importance of the indicators.

Source: Adapted from PLS Rambøll Management and Børsens Nyhedsmagasin (2000b: 31).

Weight

The USA Sweden Finland The Netherlands Norway The UK Denmark France Germany Spain

Globalization

Table 7.10 Overall performance of Danish ICT on a benchmark of statistical indicators

0.62

6.3 5.7 6.7 6.7 5.0 3.3 6.3 6.0 5.0 4.7

Flexible labour market

140 C.R. Pedersen, M.S. Dahl and B. Dalum

Pandrup Aabybro Støvring Nørager Skive

Aalborg

Aarhus Copenhagen

Rosenholm Rønde

Struer

Århus Hørning Them Horsens

Rinkøbing Videbæk

Juelsminde

Hedensted

Aabenraa

Hundested Helsinge Skævinge Hillerød Søllerød

Allerød Hørsholm Farum Berkerød Ølstykke Stenløse Værløse Lyngby-Taarbæk Ledøje-Smørum Ballerup Herlev Gladsaxe København Rødovre Glostrup Abertslund Brøndby Vallensbæk Høje-Taastrup Roskilde Dragsholme Jernløse Præstø

Sønderborg

Aarup Bogense Odensa Ørbæk

ICT specialization in Denmark 1992 2– 1.25–2 1–1.25 0–1

Figure 7.3 Municipality-level ICT employment specialization in 1992 (source: based on total employment data from Statistics Denmark).

During 1992–99 employment in Danish ICT grew by almost 34 per cent to 109,000. In comparison, total private sector employment grew by 6.3 per cent in the same period. Figure 7.4 depicts the regional ICT employment specialization pattern for 1999. The picture is complex, but there seems to be a grouping around three of the Danish university cities, Copenhagen, Aarhus and Aalborg, but not Odense. Figure 7.4 also shows several localities with specialization mainly based on single ﬁrms with individual strengths, for example, Bang & Olufsen (high end consumer electronics) in Struer, and Jamo (loudspeakers) in Sallingsund. The development from 1992–99 shows a fairly persistent geographical pattern of specialization, but also a tendency towards higher specialization in the larger cities. The concentration in cities can also be seen in Figure A7.2 in the appendix. It shows the share of total ICT employment at municipality level in 1999, which has been generally stable, located in the large cities. In 1999, Aalborg, the third largest city in Denmark, became specialized in ICT employment, partly due to the emergence of a coher-

Knowledge ﬂows in the Danish ICT sector 141 Pandrup

Aabybro

Skævinge

Aalborg Sallingsund Støvring Struer

Hadsten Rønde

Them Horsens Hedensted

Århus

Borgense

Juelsminde Jernløse

Sønderborg

Dragsholm

Skanderborg

Ølstykke

Allerød Hillerød Karlebo Birkerød Hørsholm Lyngby-Taarbæk Søllerød Værløse København Farum Gentofte Gladsaxe Herlev Rødovre Ballerup Glostrup Ledøje-Smørum

Høje-Taastrup Roskilde Brøndby Albertslund Fladså Vallensbæk Præstø ICT specialization in Denmark 1999 2– 1.25–2 1–1.25 0–1

Figure 7.4 Municipality-level ICT employment specialization in 1999 (source: based on total employment data from Statistics Denmark).

ent telecommunications cluster in North Jutland, centred around wireless communications equipment (see Figure 7.4). The presence of the cluster can also be seen in Figure A7.2 and Figure 7.4 where the small municipality of Pandrup is marked with more than 1 per cent of the total ICT employment due to the localization of a large manufacturer of mobile phones. This strong specialization emerged from the mid-1960s in maritime communications and diversiﬁed into mobile communications at the beginning of the Nordic Mobile Telephone System boom from 1981. The cluster accounts for more than half of total ICT employment in North Jutland County. The ﬁgures and the development in 1992–99 clearly point to ICT being mainly, but not exclusively, a ‘city’ industry. A closer look at the different ICT segments (computers, communication equipment, instruments, wholesale, telecommunication service and ICT services and consulting, see deﬁnitions in Table A7.1) reveals that both the Copenhagen and the Aarhus regions are specialized in ﬁve of the six ICT segments (lacking only communication equipment). The two further observations of high specialization among the segments are communication equipment and telecommunications in North Jutland and communication equipment in Ringkoebing County. Furthermore, Western Zeeland is specialized in

142 C.R. Pedersen, M.S. Dahl and B. Dalum computers and Vejle is specialized in computers and communication equipment. Further details from the regional employment data in terms of specialization for each of the four-digit NACE industries, points to a somewhat more differentiated geographical specialization pattern when the more aggregated county-level is considered, see Table A7.5. For the greater Copenhagen region the strength is in the area north of the inner core of the City (Copenhagen County) with specialization in 18 of the 19 fourdigit NACE codes. The rest of the areas in the greater Copenhagen region are also specialized in three to 11 of the NACE codes. The Aarhus region is specialized in seven followed by North Jutland in six. The 4-digit data also points to the existence of other interesting concentrations of ﬁrms. In the county of Vejle (around the town of Horsens) there is a strong specialization in the manufacture of components (3,210). Ringkoebing County specializes in the manufacture of radios and televisions, etc. (3,230) with the location of the big (in the Danish context) consumer electronics ﬁrm, Bang & Olufsen. Lastly, there is a strong specialization in Western Zeeland in the manufacture of wire and cable (3,130), which mainly can be attributed to the location of the large producer, NKT. Telecom service provision (NACE 6420) is concentrated in Copenhagen, Aarhus and Aalborg. This is a direct effect of the location of the major service providers (cable as well as wireless) – Tele Danmark, Sonofon, Telia and Orange, all with headquarters in Copenhagen, although most of Sonofon’s employment is located in North Jutland. In the Copenhagen area another cluster has emerged for optical communications equipment. This optical communication cluster is based on research at DTU. This led to the formation of NKT Elektronik, which, together with DTU, produced a group of ﬁrms and spin-offs. The largest company was Lucent Technologies Denmark, which was sold to a Japanese company in 2001, as a result of the ﬁnancial crisis being experienced by the large international telecommunications companies, and which hit AT&T-Lucent rather hard. In spite of this, the Copenhagen cluster appears to have growth potential. This can be seen at the most detailed level of the OECD export specialization pattern. Denmark was specialized in optical ﬁbre cables in 1990 (SITC Rev. 3 773.18, c.f. in the appendix), and then in 1993–98 in SITC 884.19.18 Such patterns are hard to detect in the statistical sources because optical ﬁbres have not traditionally been considered to be electronics. As can be seen, a metropolis effect is prevalent in the regional specialization pattern of ICT activities. At the more detailed industry level, however, there is a certain degree of geographical diversiﬁcation, which may be related to the rather decentralized nature of the public education system. There is, thus, a close correlation between the distributions of basically government ﬁnanced R&D and higher education institutions in ICT and

Knowledge ﬂows in the Danish ICT sector 143 the regional distribution of private employment in Denmark. Engineers and computer scientists typically choose jobs close to these institutions. More specialized small-scale clusters usually emerge around these. This pattern has major implications for policy. To further encourage the development of ICT activities, a coordinated policy approach to such ﬁelds as research and higher education, specialized venture and seed capital, and regional development of the necessary infrastructure facilities (science parks, telecommunications networks, general transport facilities, etc.) are important.

7.7 Conclusion The most obvious constraint on the future development of ICT in most western countries is the lack of appropriately qualiﬁed labour. During the early 1990s the intake of new students to engineering schools and business schools in Denmark decreased. However, between 1995 and 2000 the intake of students to IT-related university programmes increased from 1,750 to 5,500. In 1999 two new institutions, the IT-High School in Copenhagen and IT-West, started R&D and education with an enrolment in 2000 of more than 800 students. A dedicated science park and buildings for the IT-High School are part of the new Oerestad project in Copenhagen.19 At the national level a major initiative has been launched by the Ministry under the label ‘The Digital Denmark’. This initiative was ﬁrst presented in a report in 1999 with ﬁve ambitious goals (Ministry of Research and Information Technology 1999). One of these goals is the establishment of ‘IT-lighthouses’ in two different regions in Denmark. These types of initiatives are inspired by several major investments in regional growth centres in other countries.20 Massive investments in ambitious projects in other countries have resulted in the establishment of special high growth regions for ICT businesses. These regions are now powerful magnets for high technology businesses, which agglomerate and cluster in these regional growth centres. A strong physical concentration of innovative environments of knowledge and education institutions has determined these developments. The goal of the lighthouse projects is to stimulate the formation of ICT businesses in connection with knowledge and education institutions. The two Danish ICT lighthouses are located in North Jutland and the Copenhagen Oerestad region. The presence of the small-scale cluster in wireless communications in North Jutland is the main reason for the location of one of these projects; the major concentration of ICT activities in the Oeresund region and the deliberate, and high proﬁled, efforts to let this region become much more visible at the international scene, was decisive in the second case.21 Another striking feature of the ICT industry in Denmark is the

144 C.R. Pedersen, M.S. Dahl and B. Dalum dominance of foreign ownership in a signiﬁcant part of the Danish ICT sector, not least in those sub-industries displaying very fast technological development. This is apparently an indication of a lack of investors willing to take the high risks that are inevitable in the ICT ﬁeld. The amount of seed and venture capital and the structure of their supply are issues that need to be inﬂuenced through policy, perhaps via government guarantees for high-risk investments in new technology start-ups.22 However, among the governments of most of the well-developed countries there has been an avowed intention to become ‘the IT power house of the next millennium’. The Network Report (Ministry of Research and Information Technology 2000) contains a rather extensive arsenal of already initiated activities as well as a series of proposals. The political intentions of government have been restated in two reports from the Ministry of Research and Telecommunications in 2001. However, this chapter has pointed to some serious handicaps to Denmark being able to achieve such a goal related to the supply side. The problem appears to be rather fundamental: the lack of a large ‘domestic’ multinational ICT company is a serious problem for a small country. In some cases this might be compensated for by the emergence of small- and medium-sized R&D oriented ICT ﬁrms, but too few have emerged during the 1990s despite fairly favourable economic conditions.23 The challenges for Denmark appear to be great, given the present structure of the ICT sector. Among the factors from a cluster perspective is the rather weak Danish infrastructure for broadband Internet access. The heritage from the earlier telecommunications monopoly, TDC, still has a dampening effect on the diffusion of the necessary technologies. Denmark could learn much from its neighbour, Sweden.

Knowledge ﬂows in the Danish ICT sector 145

Appendix Table A7.1 Comparison of ICT deﬁnitions Nordic Statistical 3001 3002 3130 3210 3220

Description

KNOW Project

Manufacture of ofﬁce machinery Manufacture of computers and other information processing equipment Manufacture of insulated wire and cable

N/A

Manufacture of electronic valves and tubes and other electronic components Manufacture of radio and television transmitters and apparatus for line telephony/ telegraphy

N/A N/A 3210 3220

3230

Manufacture of radio and television receivers, sound/video recording or reproducing apparatus and associated goods 3320 Manufacture of instruments and appliances for measuring, checking, testing, navigating and other purposes (except 3330) 3330 Manufacture of industrial process control equipment 5143 + 5164 + 5165 Wholesale of ICT equipment

N/A N/A

6420

Telecommunications

6420

7133

Rental and leasing of ICT equipment

N/A

7210 7220 7230 7240 7250

Hardware consultancy Software consultancy and supply Data processing Database activities Maintenance and repair of ofﬁce, accounting and computing machinery Other computer related activities

7210 7220 7230 7240

7260

Source: Based on Nordic Statistical Institutes (1998). Note Sectors not included in the KNOW Project are shaded grey.

3230 N/A

7250 7260

Telecommunications ICT Services and consulting

Wholesale

Instruments

Source: Adapted from Dahl and Dalum (2001).

Services

Computers

Hardware manufacturing

Communication equipment

Segment

Activity

5164 5165 6420 7133 7210 7220 7230 7240 7250 7260

3330 5143

3320

3230

3001 3002 3130 3210 3220

NACE

Manufacture of ofﬁce machinery Manufacture of computers and other information processing equipment Manufacture of insulated wire and cable Manufacture of electronic valves and tubes and other electronic components Manufacture of radio and television transmitters and apparatus for line telephony/telegraphy Manufacture of radio and television receivers, sound/video recording or reproducing apparatus and associated goods Manufacture of instruments and appliances for measuring, checking, testing, navigating and other purposes (except 3330) Manufacture of industrial process control equipment Wholesale of electrical household equipment appliances and radio and television goods Wholesale of ofﬁce machinery and equipment Wholesale of other machinery for use in industry, trade and navigation Telecommunications Rental and leasing of ICT equipment Hardware consultancy Software consultancy and supply Data processing Database activities Maintenance and repair of ofﬁce, accounting and computing machinery Other computer related activities

Description

Table A7.2 The IKE segmentation of the ICT sector in NACE codes

Knowledge ﬂows in the Danish ICT sector 147 Table A7.3 ICT segments in SITC rev. 3 compared with NACE NACE

SITC rev. 3

Product

Telecommunications equipment 3130 773.18 Optical ﬁbre cables 3220 764.11 Telephone sets 3220 764.13 Teleprinters 3220 764.15 Telephonic or telegraphic switching apparatus 3220 764.17 Other apparatus for carrier-current liner systems 3220 764.19 Other telephonic or telegraphic apparatus 3220 764.31 Transmission apparatus 3220 764.32 Transmission apparatus with reception apparatus 3220 764.82 Television cameras 3220 764.91 Parts and accessories for apparatus of heading 7641 3230 764.93 Parts and accessories of 761, 762, 7643, 7648 3320 764.83 Radar, radio-navigation aid, remote control apparatus 3320 874.77 Other instruments and apparatus for telecommunications Consumer electronics 3230 761.1 3230 761.2 3230 762.11 3230 762.12 3230 762.21 3230 762.22 3230 762.81 3230 762.82 3230 762.89 3230 763.31 3230 763.33 3230 763.35 3230 763.81 3230 763.82 3230 763.83 3230 763.84 3230 764.21 3230 764.22 3230 764.23 3230 764.24 3230 3230 3230

764.25 764.26 764.81

3230 3230

764.92 764.99

Computers 3002 752.1 3002 752.2 3002 752.3 3002 752.6

Television receivers, colour, whether or not combined Television receivers, monochrome, combined or not Radio, external source of power, vehicles, combined Radio, external source of power, vehicles, non-combined Radio, without external source of power, combined Radio, without external source of power, non-combined Other radio receivers, combined with sound reproduction Other radio receivers, combined with a clock Other radio-broadcast receivers, non-combined Record-players, coin- or disc-operated Other record players Turntables (record-desks) Video recording or reproducing apparatus Transcribing machines Other sound reproducing apparatus Sound recording apparatus Microphones and stands therefore Loudspeakers, mounted in their enclosures Loudspeakers, not mounted in their enclosures Headphones, earphones and combined microphone/speaker Audio-frequence electric ampliﬁers Electric sound ampliﬁer sets Reception apparatus for radio-telephone, -telegraph, not elsewhere speciﬁed Parts and accessories for apparatus of heading 7642 Parts and accessories for apparatus of group 763 Analogue or hybrid data processing machines Digital automative data processing machines, cent. proc. unit Digital processing units with: storage, input, output Input or output units, whether or not with storage continued

148 C.R. Pedersen, M.S. Dahl and B. Dalum Table A7.3 continued NACE

SITC rev. 3

Product

3002 3002 3002

752.7 752.9 759.97

Storage units, with the rest of a system or not Data processing equipment, not elsewhere speciﬁed Parts, accessories of the machines of group 752

Electronic components 3130 773.11 Winding wire 3130 773.12 Co-axial cable and other co-axial conductors 3130 773.14 Other electric conductors, for a voltage <80 volts 3130 773.15 Other electric conductors, 80 volts 1 w 3210 776.35 Thyristors, diacs and triacs 3210 776.37 Photosensitive semi-conductor devices; light emittance 3210 776.39 Other semi-conductor devices 3210 776.41 Digital monolithic integrated circuits 3210 776.43 Non-digital monolithic integrated circuits 3210 776.45 Hybrid integrated circuits 3210 776.49 Other electronic integrated circuits, microassemblies 3210 776.81 Piezo-electric crystals, mounted 3210 778.61 Fixed capacitors for 50/60 hz circuit, capacitor >0, 5kvar 3210 778.62 Tatanlum ﬁxed capacitors 3210 778.63 Aluminium electrolytic ﬁxed capacitors 3210 778.64 Ceramic dielectric ﬁxed capacitors, single layer 3210 778.65 Ceramic dielectric ﬁxed capacitors, multilayer 3210 778.66 Paper or plastic dielectric ﬁxed capacitors 3210 778.67 Other ﬁxed capacitors 3210 778.68 Variables or adjustable capacitors Ofﬁce machinery 3001 751.13 3001 751.21 3001 751.22 3001 751.24 3001 751.28 3001 3001

751.31 751.32

Automatic typewriters; word processing machines Electronic without external source of power Other calculating machines Cash registers, incorporating a calculating device Postage-franking and similar machines, with calculating device Electrostatic photocopying apparatus, direct process Electrostatic photocopying apparatus, indirect processing

Knowledge ﬂows in the Danish ICT sector 149 Table A7.3 continued NACE

SITC rev. 3

Product

3001 3001 3001 3001 3001 3001

751.33 751.34 751.35 759.1 759.91 759.95

Non-electrostatic photocopying apparatus, optical Non-electrostatic photocopying apparatus, contact Thermo-copying apparatus Parts, accessories of the apparatus of heading 7513 Parts, accessories of the machines of sub-group 7511 Parts, accessories of the machines of sub-group 7512

Instruments and equipment for detecting, measuring etc. 3320 871.31 Microscopes other than optical and diffraction apparatus 3320 873.11 Gas meters 3320 873.13 Liquid meters 3320 873.15 Electricity meters 3320 873.21 Revolution counters, mileometers and the like 3320 873.25 Speed indicators and tachometers; stroboscopes 3320 874.11 Compasses; navigational instruments and appliances 3320 874.13 Surveying, hydrological, etc., instruments and appliances 3320 874.25 Measuring or checking instruments, machines, not elsewhere speciﬁed 3320 874.31 Apparatus for measuring the ﬂow or level of liquids 3320 874.35 Instruments and apparatus for measuring pressure 3320 874.37 Other instruments and apparatus for measuring, checking 3320 874.41 Gas or smoke analysis apparatus 3320 874.42 Chromatographs and electrophoresis instruments 3320 874.43 Spectrometers, spectrographs using optical radiation 3320 874.44 Exposure meters 3320 874.45 Other instruments and apparatus using optical radiation 3320 874.46 Apparatus for physical or chemical analysis, not elsewhere speciﬁed 3320 874.53 Machines and appliances for testing materials 3320 874.61 Thermostats 3320 874.63 Pressure regulators and controllers 3320 874.65 Other regulating or controlling instruments and apparatus 3320 874.71 Instruments and appliances for measuring ionizing radiation 3320 874.73 Cathode-ray oscilloscopes and oscillographs 3320 874.75 Other apparatus for measuring electricity, without recording 3320 874.78 Other apparatus for measuring electricity quantities 3320 874.9 Parts and accessories for machines, appliances etc., not elsewhere speciﬁed 3340 871.91 Telescopic sight for 7, 87, 881, 884, 8996 3340 871.92 Lasers (other than laser diodes) 3340 871.93 Other optical devices, appliances and instruments 3340 881.21 Cinematographic cameras 3340 881.22 Cinematographic projectors 3340 884.19 Optical ﬁbres, bundles, cables; polarizing material 3340 884.39 Mounted optical elements, not elsewhere speciﬁed Source: Based on OECD (2000e) International Trade by Commodities Statistics, No. 1.

150 C.R. Pedersen, M.S. Dahl and B. Dalum Table A7.4 The used statistical indicators for the benchmark Driver

Indicators

ICT labour force

Migration of highly educated labour Number of highly educated engineers Number of labour with ICT education Number and the quality of engineers and scientists ICT infrastructure Number of PCs per 100 inhabitants ICT expenditure per 100 inhabitants Internet hosts per 1,000 inhabitants International Internet bandwidth, kbps per 100 inhabitants Business websites per 1,000 inhabitants ICT education Government expenditure on education per capita Education quality in mathematics and natural science Techno-commercialization Nation’s share of the total ICT sector in OECD Nation’s share of the total R&D expenditure The ICT sectors share of GDP Technology absorption Number of PCs per 100 white collar workers Globalization High-technology trade ratio (exports/imports) International strategic alliances in per cent of the total number of alliances (1990–99) Exports per capita Capital markets Growth in venture capital per year per capita Venture capital per capita Growth in the ﬁnancial assets of the institutional investors Financial assets of institutional investors in per cent of GDP B2B e-commerce Secure servers per one million inhabitants Network culture Strategic alliances per one million inhabitants ICT consumers Value of business-to-consumer (B2C) trade per capita Number of consumers on the Internet in per cent of total inhabitants Internet customers in per cent of the total number of banking customers Penetration ratio in per cent of the total retail sale Mobile phones per 100 inhabitants Macro economy Growth in GDP per capita Growth in exports (percent) Projected surplus of the balance of trade in 2001 compared with 1999 GDP per capita Change in inﬂation rate in per cent 1997–99 Labour productivity (GDP per employee per hour) Flexible labour market Growth of the labour force Frequency of employment Changes in real salaries 1995–99 Source: Adapted from PLS Rambøll Management and Børsens Nyhedsmagasin (2000b: 31).

Copenhagen County

Frederiksberg Municipality

Copenhagen Municipality

NACE

0.66 4.26 0.36 0.24 0.50 0.26 0.33 0.32 0.26 0.12 0.04 0.11 0.11 0.46

0.22 0.74 0.76 0.55 1.25 0.72 0.44 0.46 0.51 2.02 0.41 0.41 0.18 0.80

2.38 1.37

1.04 2.42 1.24

0.42

1.99 0.98 1.66 1.34 0.33 1.02 0.26

1.39

0.37

0.29 0.31 0.23 0.04 0.16 0.07 0.02

0.37

0.40 0.26 0.48

0.08 1.11

1.68 0.00 0.37

0.72 0.00 0.15

2.54 0.53 0.65

1.46 0.69 0.83

0.00 1.80 0.79

Storstroem County

0.00 0.13 8.52

Frederiksborg County 0.00 0.53 0.15

Roskilde County

0.00 2.03 0.43

Western Zeeland

Bornholm County 0.22

1.56 0.00 0.06 0.00 0.00 0.18 0.83

0.59

0.03 0.16 0.38

0.00 0.00

0.04 0.00 0.00

0.00 0.00 0.00

Fyn County 0.50

0.28 0.29 0.51 0.44 0.07 0.59 0.32

0.78

0.28 0.32 0.70

0.27 0.82

0.26 0.08 0.01

0.91 1.37 0.65

0.48

0.50 0.47 0.32 0.11 0.00 0.28 0.17

0.67

1.22 0.27 0.44

0.99 2.53

0.01 0.69 0.94

0.00 0.47 0.00

Ribe County 0.33

0.79 0.15 0.11 0.13 0.09 0.53 0.24

0.19

0.09 0.31 1.00

0.03 0.06

0.03 0.00 0.56

0.00 0.04 0.00

Vejle County 0.86

0.53 2.18 0.59 0.99 0.34 0.98 0.24

0.53

0.36 0.55 0.91

0.31 0.55

4.40 0.92 2.14

0.03 0.23 2.12

Ringkoebing County 0.74

0.43 0.16 0.40 0.36 0.00 0.29 0.00

0.28

0.34 0.48 0.35

0.15 0.10

0.05 2.44 6.96

0.00 0.05 0.03

Aarhus County 1.17

0.07 0.76 1.25 1.23 0.02 0.85 0.17

1.84

0.70 1.14 1.07

2.07 0.70

0.12 0.23 0.52

0.16 1.46 0.86

0.41

0.17 0.20 0.21 0.10 0.23 0.18 0.14

0.14

0.39 0.22 0.57

0.86 0.06

0.00 0.04 2.69

0.00 0.03 0.01

Viborg County

0.86

0.47 0.39 0.82 0.19 0.00 0.36 0.05

1.05

0.78 0.50 0.55

0.56 1.84

1.72 4.98 1.17

6.50 0.32 0.02

Note The specialization indicator is the share of ICT employment in every industry of the total employment of a given county compared with the national average. A value above 1 indicates an above average employment share – i.e. the county is specialized. Specialized sectors in bold.

Source: Based on total employment data from Statistics Denmark.

Computers 3001 0.00 0.00 2.40 3002 0.55 0.08 2.75 3130 0.01 0.00 1.76 Communication equipment 3210 0.06 0.00 1.75 3220 0.45 0.00 1.36 3230 0.03 0.04 0.23 Instruments 3320 1.50 0.01 1.62 3330 0.18 0.18 1.11 Wholesale 5143 1.50 0.56 3.02 5164 0.76 0.79 3.12 5165 0.81 0.23 2.55 Telecommunication 6420 1.94 0.96 1.56 ICT services and consulting 7133 2.39 1.05 2.56 7210 3.16 0.33 1.40 7220 1.39 1.29 2.49 7230 1.47 0.81 2.72 7240 1.46 3.33 5.21 7250 0.26 0.62 4.30 7260 2.95 0.30 3.95 Total ICT sector – 1.13 0.65 2.29

Southern Jutland County

Table A7.5 Regional employment specialization in 1998 in ICT segments and NACE codes North Jutland County

Odense

Århus

Share of the ICT employment in Denmark 1992 5% 3%–5% 1%–3% 0%–1%

Hillerød Allerød Birkerød Søllerød Lyngby-Taarbæk Værløse Gladsaxe Ballerup København Herlev Frederiksberg Glostrup Brøndby Albertslund Høje-Tasstrup Roskilde

Note One per cent of the total ICT employment in Denmark is equal to 816 persons in 1992.

Figure A7.1 Share of the total Danish ICT employment at municipality level 1992 (source: Based on total employment data from Statistics Denmark).

Kolding

Vejie

Horsens

Herning

Struer

Aalborg

Århus

Aalborg

Share of the ICT employment in Denmark 1999 5% 3%–5% 1%–3% 0%–1%

Hillerød Allerød Birkerød Sollerød Lyngby-Taarbæk Gentofte Gladsaxe Herlev Ballerup Glostrup Roskilde Brøndby Albertsland Høje-Taastrup København Frederiksberg

Note One per cent of the total ICT employment in Denmark is equal to 1,091 persons in 1999.

Figure A7.2 Share of the total Danish ICT employment at municipality level 1999 (source: Based on total employment data from Statistics Denmark).

Sønderborg

Odense

Kolding

Vejie

Horsens

Struer

Pandrup

154 C.R. Pedersen, M.S. Dahl and B. Dalum

Notes 1 For more detailed information on the mega clusters and the Danish cluster studies, see Drejer et al. (1999) and Dahl and Dalum (2001). 2 Danish Agency for Trade and Industry referred to as EFS (2000). In 2002, this agency was renamed the National Agency for Enterprise and Housing. We have chosen to keep the old acronym EFS, which is the label on the publications referred to here. 3 More speciﬁcally the ‘sector based’ deﬁnition in OECD (2000b) Annex 2, p. 249. 4 Table A14.1 in the appendix contains the deﬁnition in terms of a list of the 4digit NACE codes. This table also presents a comparison with the deﬁnition used in the TSER KNOW project. Further, Table A14.2 contains a division into six segments of which communication equipment and telecommunications, as well as the services and consulting segments, are included in the KNOW project. 5 Surveyed by Edquist (1997). For a recent overview, see Lundvall et al. (2002). 6 Excluding Greece, Luxembourg and Spain. 7 ICT has been deﬁned as ISIC Rev. 2 classes 3825 (computers), 3832 (radio, TV and communications equipment) and 72 (communications services). 8 Among the entire group of OECD countries only the US, Japan, Ireland and the UK were specialized in high-tech industries in 1990. For a deﬁnition of the categories of technology intensity, see OECD (1999b, Annex 1). 9 In 1996 also the newcomers Mexico and Korea joined the group of high-tech specialized exporters. 10 See also Green et al. (2001). 11 See Appendix Table 2.6.1 in OECD (1999b). 12 A phenomenon also experienced in other western European countries in this period. 13 The relation between the patterns of R&D versus export specialization (Figure 14.3 and Figure 14.7) is rather striking. 14 See Ministry of Research and Information Technology (2000) addendum to the Network Report, p. 12. 15 Speciﬁed in more detail in appendix Table A14.4. 16 There are 16 counties in Denmark. Employment data (full time and part time) for 1992 are available from Statistics Denmark for 4-digit and 6-digit NACE codes. The ‘Copenhagen Region’ is interpreted here as the counties of Copenhagen, Frederiksborg and Roskilde as well as the municipalities of Copenhagen and Frederiksberg. The latter municipality was not specialized in ICT in 1992. 17 The ICT sector in Figure 14.3, Figure 14.4, Figure A14.1 and Figure A14.2 is deﬁned as NACE/DB (93): 3001, 3002, 3130, 3210, 3220, 3230, 331020, 331030, 331090, 3320, 3330, 514320, 516410, 516520, 6420, 713310 and 7200. 18 The indicator was above 2 for 1997–98. A reclassiﬁcation of the export data has apparently been made between 1992 and 1993. 19 For further details see Ministry of Research and Information Technology (2000). 20 Silicon Valley (US), Information Age Town (Ennis, Ireland), Oulu Technopolis (Northern Finland) and Kista Science Park (Near Stockholm, Sweden). 21 An indication of the success of these efforts is illustrated in a ranking of the Oeresund region among the high tech valleys worldwide in Wired (2000: July, pp. 259–271). 22 Two new major seed funds (organized as private companies) have recently been announced, ﬁnanced jointly by government funds and private ﬁnance.

Knowledge ﬂows in the Danish ICT sector 155 One is based in the Copenhagen region and oriented towards biotechnology and pharmaceuticals. The other is part of the Aalborg University Science Park, NOVI, with an emphasis on ICT projects. 23 In Danmark.com PLS Rambøll Management and Børsens Nyhedsmagasin (2000a) have mapped a large amount of new Internet start up ﬁrms. The ﬁeld appears to ﬂourish, but nonetheless very few major international players have appeared.

References Begg, I., Dalum, B., Guerrieri, P. and Pianta, M. (1999) ‘The impact of specialisation in Europe’, in J. Fagerberg, P. Guerrieri and B. Verspagen (eds) The Economic Challenge for Europe: Adapting to Innovation Based Growth, Olso/Rome/ Maastricht: Edward Elgar Publishing, pp. 21–45. Braczyk, H.-J., Cooke, P. and Heidenreich, M. (eds) (1997) Regional Innovation Systems, London: University College London Press. Breschi, S. and Malerba, F. (1997) ‘Sectoral innovation systems: technological regimes, Schumpeterian dynamics, and spatial boundaries’, in C. Edquist (ed.) Systems of Innovation: Technologies, Institutions and Organizations, London: Pinter Publishers, pp. 130–156. Carlsson, B. and Jacobsson, S. (1997) ‘Diversity creation and technological systems: a technology policy perspective’, in C. Edquist (ed.) Systems of Innovation: Technologies, Institutions and Organizations, London: Pinter Publishers, pp. 266–294. Clark, G.L., Feldman, M.P. and Gertler, M.S. (eds) (2000) The Oxford Handbook of Economic Geography, Oxford: Oxford University Press. Dahl, M.S. and Dalum, B. (2001) ‘The ICT cluster in Denmark’, in OECD (ed.) Innovative Clusters, Paris: OECD, pp. 65–90. Dalsgaard, M.H. (2001) ‘Danish cluster policy: improving speciﬁc framework conditions’, in OECD (ed.) Innovative Clusters: Drivers of National Innovation Systems, Paris: OECD, pp. 347–360. Dalum, B., Fagerberg, J. and Jørgensen, U. (1988) ‘Small open economies in the world market for electronics: the case of the Nordic countries’, C. Freeman and B.-Å. Lundvall (eds) Small Countries Facing the Technological Revolution, London: Pinter Publishers, pp. 113–138. Dalum, B., Freeman, C., Simonetti, R., von Tunzelmann, N. and Verspagen, B. (1999) ‘Europe and the information and communication technologies revolution’, in J. Fagerberg, P. Guerrieri and B. Verspagen (eds) The Economic Challenge for Europe: Adapting to Innovation Based Growth, Oslo/Rome/Maastricht: Edward Elgar Publishing, pp. 106–129. Dalum, B., Laursen, K. and Villumsen, G. (1998) ‘Structural change in OECD export specialisation patterns: de-specialisation and “Stickiness” ’, International Review of Applied Economics, 12(3): 423–443. Drejer, I., Kristensen, F.S. and Laursen, K. (1999) ‘Studies of clusters as the basis for industrial and technological policy in the Danish economy’, in OECD (ed.) Boosting Innovation: The Cluster Approach, Paris: OECD, pp. 293–314. Edquist, C. (ed.) (1997) Systems of Innovation: Technology, Institutions and Organisation, London: Pinter. EFS (2000) ‘Benchmarking af dansk erhvervsliv [Benchmarking Danish Businesses]’, Copenhagen: Danish Agency for Trade and Industry [EFS].

156 C.R. Pedersen, M.S. Dahl and B. Dalum EFS (2001) IT/Kommunikation – en erhvervsanalyse [IT/Communication – A Business Analysis], Copenhagen: Danish Agency for Trade and Industry [EFS]. Financial Times (2000) ‘Survey on Nordic Information Technology’, 11 May, pp. 1–8. Freeman, C. (1987) Technology Policy and Economic Performance: Lessons from Japan, London: Pinter. Freeman, C. (1995) ‘The national innovation systems in historical perspective’, Cambridge Journal of Economics, 19(1): 3–24. Green, R., Duggan, I., Giblin, M., Moroney, M. and Smyth, L. (2001) ‘The boundaryless cluster: information and communications technology in Ireland,’ in OECD (ed.) Innovative Clusters, Paris: OECD, pp. 47–64. Lundvall, B.-Å. (1985) ‘Product innovation and user–producer interaction’, Serie om industriel udvikling, 31: 1–70. Lundvall, B.-Å. (ed.) (1992) National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning, London: Pinter Publishers. Lundvall, B.-Å. (1999) Det danske innovationssystem [The Danish Innovation System], Aalborg: Danish Agency for Trade and Industry. Lundvall, B.-Å., Johnson, B., Andersen, E.S. and Dalum, B. (2002) ‘National systems of production, innovation and competence-building’, Research Policy, 31(2): 213–231. Lundvall, B.-Å., Olesen, N.M. and Aaen, I. (eds) (1984) ‘Det landbrugsindustrielle kompleks [The agro-industrial complex],’ Aalborg: Aalborg University Press. Malerba, F. (2002) ‘Sectoral systems of innovation and production’, Research Policy, 31(2): 247–264. Maskell, P. and Malmberg, A. (1999) ‘Localised learning and industrial competitiveness’, Cambridge Journal of Economics, 23: 167–185. Ministry of Research and Information Technology (1999) Det digitale Danmark [The Digital Denmark], Copenhagen: Ministry of Research and Information Technology. Ministry of Research and Information Technology (2000) Netværksredegørelse [Network Report], Copenhagen: Ministry of Research and Information Technology. Moeller, K. and Pade, H. (eds) (1988) Industriel Succes [Industrial Success], Copenhagen: Samfundslitteratur. Nelson, R.R. (ed.) (1993) National Innovation Systems: A Comparative Analysis, New York: Oxford University Press. Nordic Statistical Institutes (1998) ‘The information and communication technology in the Nordic countries: a ﬁrst statistical description’, TemaNord, 587: 1–69. Nordic Statistical Institutes (2000) The Information and Communication Technology in the Nordic Countries, Copenhagen: Nordic Statistical Institutes. OECD (1998) Education at a Glance 1998, Paris: OECD. OECD (1999a) OECD Communications Outlook, Paris: OECD. OECD (1999b) Science, Technology and Industry Scoreboard: Benchmarking Knowledgebased Economies, Paris: OECD. OECD (2000a) Education at a Glance 2000, Paris: OECD. OECD (2000b) Measuring the ICT Sector, Paris: OECD. OECD (2000c) OECD Information Technology Outlook: ICTs, E-Commerce and the Information Economy, Paris: OECD. OECD (2000d) Science, Technology and Industry Outlook: Science and Innovation, Paris: OECD.

Knowledge ﬂows in the Danish ICT sector 157 OECD (2000e) International Trade by Commodities Statistics, No. 1, CD-ROM, Paris: OECD. PLS Rambøll Management and Børsens Nyhedsmagasin (2000a) Danmark.com – Kortlægning af de danske internetpionerer [Denmark.com – Mapping the Danish Internet Pioneers], Copenhagen: PLS Rambøll Management and Børsens Nyhedsmagasin. PLS Rambøll Management and Børsens Nyhedsmagasin (2000b) Frontløberlande i den nye økonomi – Benchmarking Danmark.com [Frontier Countries in the New Economy – Benchmarking Denmark.com], Copenhagen: PLS Rambøll Management and Børsens Nyhedsmagasin. Porter, M.E. (1990) The Comparative Advantage of Nations, New York: Free Press. Porter, M.E. (1998) On Competition, Boston: Harvard Business School Press. Saxenian, A. (1994) Regional Advantage: Culture and Competition in Silicon Valley and Route 128, Cambridge, MA: Harvard University Press. Storper, M. (1992) ‘The limits to globalization: technology districts and international trade’, Economic Geography, 68(1): 273–305. Storper, M. (1995) ‘The resurgence of regional economics, ten years later’, European Urban and Regional Studies, 2(3): 191–221. Wired Magazine (2000) ‘Venture Capitals’, July, pp. 259–271.

8

Collaboration and innovation outputs Anthony Arundel and Catalina Bordoy

8.1 Introduction The role of collaboration in the creation and diffusion of innovations has received a considerable amount of interest in the past decade from both innovation economists and European policy makers. The European Commission (EC), for example, requires collaboration for research funded under the Framework Programmes (FPs) and encourages national policies to increase the rate of collaboration between ﬁrms and between ﬁrms and public research organizations (PROs) such as universities and publicly-funded research institutes (Hagedoorn et al. 2000). Part of the increased attention given to collaboration is due to a shift in economic theories of innovation from a linear to an interactive model and to a recognition of the importance of tacit knowledge to innovation and technology transfer. The ﬁrst shift places greater importance on information and knowledge ﬂows. This is partly because there are more linkages within an interactive or chain-link innovation system and partly because modern innovation theories stress the diffusion of knowledge among many different actors (Kline and Rosenberg 1986; Rothwell 1992; Freeman 1987; Nelson 1993; Lundvall 1992; Edquist 1997; Asheim and Smith 1999; Lundvall et al. 2002). Within this framework, the literature has focused on two types of linkages: inter-ﬁrm relationships, such as networks of user-producers or competing ﬁrms (von Hippel 1988; Lundvall 1991, 1992), and linkages between ﬁrms and PROs (Mansﬁeld and Lee 1996; Pavitt 1991). Although many of these linkages do not require collaboration, many researchers have argued that collaboration plays a central role in the ﬂow of knowledge among different actors. One perspective is that the role of collaboration in innovation has increased in importance over the past few decades. As an example, Gibbons et al. (1994) argue that knowledge production has shifted from a mode 1 model, in which problems are identiﬁed and solved in academic and corporate research laboratories, to a mode 2 model, in which knowledge is created in a great variety of networked organizations, including large ﬁrms, small high-technology based ﬁrms, government institutions,

Collaboration and innovation outputs 159 universities, research laboratories and institutes. Georghiou (1998) further develops their mode 2 model of innovation and suggests that contract research, based on cooperation and enabled by intellectual property rights, plays a central role in the exchange and production of new knowledge among government, industry and university participants. Furthermore, the new mode of knowledge production includes a wider and more heterogeneous set of actors that collaborate temporarily on speciﬁc problems. One feature is the development of a ‘collaborative’ space that links PROs, ﬁrms and other institutions. Similarly, Antonelli (1999) refers to ‘the systematic use of technological co-operation’ as the ‘dominant form’ of producing new knowledge. The second theoretical development with implications for collaboration is derived from research on the nature of knowledge and information. This posits a continuum between highly codiﬁed knowledge, which is widely accessible in publications and patents, and highly tacit knowledge, which is only held in the minds of science and engineering staff (Saviotti 1998; Cowan et al. 2000). Much of the knowledge developed within companies, particularly those active at the technological frontier, results from learning-by-doing or learning-by-using and may not be available in a codiﬁed form. Compared to other forms of knowledge ﬂows based on codiﬁed information, collaboration provides better opportunities for the transfer of non-codiﬁed and tacit knowledge through interpersonal interactions (Senker 1995). In addition to providing access to non-codiﬁed knowledge, collaborative research can have several other advantages, including a reduction in technological and market uncertainty, cost-sharing, risk spreading, reduced duplication of research, economies of scale and an ability to combine different types of expertise (Tether 2002). These advantages could improve the economic efﬁciency of innovation. Most of the evidence in support of the growing importance of collaboration is derived from case studies or analyses of data on patent citations (Narin et al. 1997), formal research partnerships or external research and development (R&D) expenditures. Hagedoorn and Schakenraad (1992) identiﬁed an increase in R&D partnerships over the 1980s and Howells (1999) found that the percentage of business R&D in the UK that is spent outside the ﬁrm increased from 5 per cent in 1985 to 10 per cent in 1995. However, most of these studies focus on indicators, such as patents, that are dominated by a small number of leading technologies, primarily biotechnology and information and communication technologies (ICT). The role of collaboration in the development of these technologies could differ from that in many other areas of innovation. Recent innovation surveys, based on guidelines set out in the OECD’s Oslo Manual, provide some empirical data on the prevalence of collaboration among innovative ﬁrms in all manufacturing sectors. Results from the 1997 Community Innovation Survey (CIS) for seven European countries

160 A. Arundel and C. Bordoy show that only 28 per cent of innovative ﬁrms developed one or more product innovations together with ‘other enterprises or institutes’ during the three years between 1994 and 1996.1 Several major disadvantages of collaboration could partly explain why the majority of innovative ﬁrms that responded to the CIS survey do not develop innovations with other ﬁrms or institutes. One is the diversion of energy and talents (Geroski 1995), while another is concern over appropriating the results, plus the need to reveal information to a cooperation partner. The latter can be an essential beneﬁt of collaboration when most knowledge is in tacit form, but a serious disadvantage if either or both partners are concerned about revealing information that they would prefer to keep secret. The cost savings to a ﬁrm from collaboration could also be reduced by the cost of developing an appropriate level of absorptive capacity (Cohen and Levinthal 1990) to be able to successfully incorporate the results into new products and processes. The need for ﬁrms to collaborate could also depend on the characteristics of the technologies that are used by the ﬁrm. Well-understood technologies will probably be developed in house, while ﬁrms should be more likely to cooperate on innovations that are based on complex technologies that require a range of expertise or those at the technological frontier. An example of a complex technology is telecommunications equipment, which includes software, electronics and computer parts. However, telecom equipment ﬁrms, as well as all other ﬁrms, face three options for acquiring technologies outside their area of expertise. They can develop the necessary capabilities in-house, collaborate on their development with other organizations or they can ‘buy-in’ sub-components from other ﬁrms. Veugelers and Cassiman (1999) characterize these choices as ‘make’ or ‘buy’ decisions and ﬁnd that most ﬁrms use several strategies. If the beneﬁts of collaboration to a ﬁrm outweigh the disadvantages, as suggested by Antonelli (1999) and others, we should expect collaboration to inﬂuence innovation outcomes: more collaboration should lead either to more innovations or to more ‘valuable’ innovations. In the ﬁrst case, collaboration could lead to economic beneﬁts to the ﬁrm in terms of higher proﬁts or an increase in the proportion of sales from new or signiﬁcantly improved products (the innovation sales share). In the second case, collaboration could lead to more ‘radical’ or complex innovations that are technically superior to the types of innovations that could be developed if the ﬁrm only drew on its in-house capabilities. These radical or complex innovations could also lead to economic beneﬁts for the ﬁrm, although ﬁrms active in sectors with rapid technical change may need to collaborate simply to remain in business. Several studies have used the CIS and similar innovation surveys to investigate the relationship between collaboration and a measure either of the quality of the ﬁrm’s innovations or the ﬁrm’s innovation sales share. All survey research on the relationship between collaboration and innova-

Collaboration and innovation outputs 161 tion quality, usually proxied by a measure of the ‘innovativeness’ of the ﬁrm, ﬁnds a positive correlation between the two. Albach et al. (1996) ﬁnd that R&D cooperation among ﬁrms in the chemical sector is most prevalent among the more innovative sub-sectors of this industry, such as pharmaceuticals and agrochemicals. Similarly, Fritsch and Lukas (2001) report that more ‘innovative’ German ﬁrms were more likely to be involved in cooperation than less innovative ﬁrms. Tether (2002), in an analysis of the 1997 CIS for the UK, ﬁnds that ﬁrms that introduced ‘new-to-market’ innovations were more likely to cooperate than ﬁrms that did not. Two studies using the 1999 Canadian innovation survey also ﬁnd that collaboration was more common among ﬁrms that developed ‘world-ﬁrst’ innovations compared to those that innovated via imitation (Therrien and Chang 2003; Landry and Amara 2003). The results of survey research on the effect of collaboration on economic outcomes, such as the innovation sales share, are more ambiguous. Nas et al. (1994) report that participation in cooperative R&D increases the innovation sales share. Similarly, Mohnen and Therrien (2003), using the 1999 Canadian innovation survey, ﬁnd a positive relationship between collaboration and the innovation sales share. Conversely, two studies using the 1993 CIS results for the telecom and ofﬁce equipment sectors found that ﬁrms that participate in cooperative R&D have a lower innovation sales share compared to ﬁrms that do not cooperate (Arundel et al. 1996; Malerba et al. 1997). This raises the possibility that ﬁrms in these two sectors cooperate as part of a catch-up strategy. Vonortas (1997) reports that ﬁrms that did not cooperate in research joint ventures (RJV) had higher proﬁts than ﬁrms that did participate, which raises many questions about the value of collaboration and the reasons why ﬁrms participate in such projects. It should be noted that collaboration is only one method of sourcing knowledge from outside the ﬁrm. The use of other methods of acquiring external knowledge, such as searching patent databases or reading trade journals, is often positively correlated both with collaboration and with the quality of the ﬁrm’s innovations. To summarize, the empirical research to date shows that both collaboration and other methods of sourcing knowledge outside the ﬁrm are positively linked to different measures of the technical quality of the ﬁrm’s innovations, or its ‘innovativeness’. At the same time, survey research shows that the majority of ﬁrms attach greater importance to their own inhouse activities than to external knowledge sources, including the use of collaboration. The effect of collaboration on economic outcomes, such as the innovation sales share, is ambiguous, with some studies ﬁnding a positive effect and others a negative effect. The main drawback to much of the empirical research, largely due to limitations in questionnaire design, is that it is not possible to directly link collaboration with either the quality of the ﬁrm’s innovations or to

162 A. Arundel and C. Bordoy economic outcomes such as the ﬁrm’s innovation sales share. For example, we often know whether or not a ﬁrm collaborates, but we lack information on the effect of the ﬁrm’s collaborations on its innovative outputs. In contrast, the KNOW survey overcomes this limitation through a question on the percentage of each ﬁrm’s product and process innovations that were developed through collaboration. We use the results from this survey question to explore two research issues. First, what factors inﬂuence the use of collaboration as a method of developing innovations? Possible factors include ﬁrm size, R&D efforts, appropriation conditions and the ﬁrm’s ability to source external knowledge without resorting to collaboration. Second, is there a positive correlation between the percentage of innovations that are developed through collaboration and the innovation sales share?

8.2 Methodology The KNOW survey was conducted in the spring of 2000 in seven European Union (EU) countries: the UK, Denmark, the Netherlands, France, Germany, Italy and Greece. The survey was limited to ﬁve sectors: food and beverages (NACE 15), chemicals excluding pharmaceuticals (NACE 24 minus NACE 24.4), telecom equipment (NACE 32), telecom services (NACE 64.2) and computer services (NACE 72). These speciﬁc sectors were chosen to provide a range of low, medium and high technology manufacturing and to include two innovative service sectors. In addition, the survey was limited to small and mid-sized ﬁrms to avoid problems with deﬁning the sector of activity of large ﬁrms, many of which are active in several sectors. For each country, a random sample of ﬁrms from two size classes (10–250 employees and 251–1,000 employees) within each of the ﬁve sectors was drawn from a national business registry. Useable responses were obtained from 558 ﬁrms, for a response rate of 25.3 per cent (noninnovative ﬁrms and ﬁrms that did not ﬁt the sampling criteria for size and sector were excluded).2 This chapter uses a maximum of 507 ﬁrms because we exclude 50 ﬁrms in the telecom services sector (see below).3 The ﬁrst half of the questionnaire asked about the total innovative activities of the ﬁrm while the second half focused on the ﬁrm’s most economically important innovation. This chapter addresses the ﬁrms’ total innovation activities. There are two key questions. The ﬁrst question concerns an estimate of the innovation sales share, deﬁned as the percentage of each ﬁrm’s sales, in the previous year, from products or services that were ‘signiﬁcantly improved or new to your ﬁrm in the last three years’. The CIS uses a similar question, as noted in the introduction. The second key question produces an estimate of the percentage of each ﬁrm’s innovations in the last three years that were introduced through ‘buying in’, ‘in-house development’ or ‘collabo-

Collaboration and innovation outputs 163 ration with external partners’. The question is asked separately for product and process innovations.4 The questionnaire deﬁnes ‘buying-in’ as ‘purchasing, licensing or contracting out’. Of note, ‘external’ partners include other divisions of the same ﬁrm in addition to PROs and independent ﬁrms, while ‘in-house’ activities are restricted to the site of the ﬁrm itself or, where relevant, to the site of the respondent’s division or subsidiary. Preliminary analyses of the distribution of the results for the two key questions by country and sector found that the results for telecom services (NACE 64.2) varied widely by country, whereas there was considerably less variation for the other four sectors. For example, the average innovation sales share for telecom services ranged by 50 percentage points between countries, compared to 23 points for computer services, 19 points for food and beverages and 12 points each for chemicals and telecom equipment. National differences in the date and degree of the liberalization of the telecommunications sector probably explains differences by country, but these differences will distort the results for the innovation sales share. For this reason, all analyses exclude telecom service ﬁrms. Some of the ﬁrms did not complete all survey questions. For a few variables, an estimated value could be computed using regression techniques and information on the ﬁrm’s sector, country, size and other characteristics. When used, information on the number of ﬁrms with estimated values are noted in the relevant tables.

8.3 Descriptive results Three hundred and forty-nine ﬁrms in the sample (68.8 per cent) have less than 250 employees, while 158 (31.2 per cent) of the ﬁrms are midsized, with between 251 and 1,250 employees. The mid-size ﬁrms account for 82.3 per cent of total employment among the sample. Almost half (46.1 per cent) of the 507 ﬁrms are a division or subsidiary of a larger ﬁrm, while the remainder are single independent entities. Table 8.1 provides basic results for the innovative characteristics of the respondent ﬁrms. Most of the ﬁrms perform well on a range of indicators of innovative activity. All but 4.0 per cent of the small ﬁrms and 6.0 per cent of the midsize ﬁrms perform R&D on either a continuous or occasional basis. In addition, 57.3 per cent of the small ﬁrms and 67.1 per cent of the mid-size ﬁrms had contracted-out some R&D to independent ﬁrms or organizations in the previous year, while 46.5 per cent of the small ﬁrms and 61.3 per cent of the mid-size ﬁrms had one or more research projects with a PRO in the previous three years. The innovative characteristics of the respondent ﬁrms could vary by sector, due to differences in technological opportunities or the role of science in innovation (Malerba 2002); and by country, due to differences

Notes Standard errors in parentheses. a The number of respondent ﬁrms varies due to missing values. b <250 employees. c 250–1,250 employees. d P value for the difference between small and mid-size ﬁrms. e Includes estimated values for the R&D status of 16 ﬁrms (3.2%). f Limited to small and mid-size ﬁrms that are part of a group.

Performs R&De: Continuously Occasionally Never Positive R&D expenditure in the last ﬁscal year: In-house In other divisions of same ﬁrmf In other independent ﬁrms/organizations One or more research projects with a PRO in the previous three years Mean percentage of employees with an academic degree

Table 8.1 Innovative characteristics of the respondent ﬁrms

69.3% (2.47) 26.7% (2.38) 4.0% (1.09) 94.0% (1.27) 60.9% (4.17) 57.3% (2.65) 46.5% (2.89) 28.1% (1.56)

349 138 349 299 348

Small b

349 349 349

Na

158 82 158 124 158

157 157 157

Na

94.3% (1.85) 68.3% (5.17) 67.1% (3.75) 61.3% (4.39) 20.4% (1.60)

77.1% (3.37) 16.9% (3.02) 6.0% (1.96)

Mid-size c

1.000 0.311 0.040 0.007 0.003

0.088 0.017 0.217

Pd

Collaboration and innovation outputs 165 in national innovation systems (Lundvall et al. 2002). For some indicators, such as the percentage of ﬁrms with positive R&D spending in an external ﬁrm or organization, most of the variation is across countries rather than across sectors, with little difference between sectors in the same country. As an example, the percentage of ﬁrms that have funded R&D in independent ﬁrms or organizations is similar in all ﬁve sectors in Germany, ranging from 77.3 per cent of the telecom equipment ﬁrms to 90 per cent of computer services ﬁrms (p = 0.735). The only exception is for Italy, where by sector there are signiﬁcant differences in this variable. Conversely, there are statistically signiﬁcant differences by country for this indicator for all four sectors. As an example, the percentage of computer service ﬁrms that have funded R&D in independent ﬁrms or organizations ranges from 27.3 per cent of these ﬁrms in Italy to 90 per cent in Germany.

8.4 Innovation methods As noted by Veugelers and Cassiman (1999), ﬁrms innovate through a variety of methods, ranging from purchasing new technology ‘off the shelf’, with no further adjustment or alteration to the technology, to conducting a substantial amount of R&D in-house in order to commercialize a new idea. A ﬁrm can also collaborate with external partners, with some of the R&D being done in-house and some by external partners. Table 8.2 provides the percentage of ﬁrms that report developing one or more product or process innovations within the last three years through each of these three main methods: in-house development, buying-in and collaboration. For all respondent ﬁrms combined, 94.9 per cent report developing at least one innovation in-house. Almost equal percentages of the ﬁrms have innovated through buying-in (61.9 per cent)

Table 8.2 Percentage of ﬁrms making any use of each innovation method

Number of innovative ﬁrms In-house Buying-in Collaboration

Products/service innovators

Process innovators

Product or process innovators

484 93.6 (1.1) 45.9 (2.3) 50.8 (2.3)

426 90.9 (1.4) 58.0 (2.4) 58.5 (2.4)

507 94.9 (1.0) 61.9 (2.2) 62.5 (2.2)

Notes Standard errors are given in parentheses. The results exclude 23 ﬁrms that did not introduce a product/service innovation and 81 ﬁrms that did not introduce a process innovation. All ﬁrms introduced at least one product/service or process innovation. The percentages for product/service innovation methods include estimated values for four ﬁrms, while the percentages for process innovation methods include estimated values for 14 ﬁrms. The results, after excluding estimated values, are almost identical.

166 A. Arundel and C. Bordoy and through collaboration (62.5 per cent). Both buying-in and collaboration are more frequently used for process innovations than for product innovations. For 54 per cent of the ﬁrms, the main collaboration partners for both process and product innovations were located in the same country as the ﬁrm itself. Table 8.3 gives, for each sector, the average percentage of product and process innovations that are developed through each innovation method. For all sectors combined, 71.4 per cent of product innovations are developed in-house, 13.2 per cent are bought-in and 15.4 per cent are developed through collaboration. In all sectors, a higher percentage of product than process innovations is developed in-house. The differences by sector are not statistically signiﬁcant for the percentage of both product and process innovations developed via collaboration, or for the percentage of product innovations developed in-house or through buying-in. In contrast, there are signiﬁcant differences by sector in the share of process innovations developed in-house (p = 0.001) and through buying-in (p < 0.000). For each of the three innovation methods, there are statistically signiﬁcant differences in the mean by country. In most cases, these differences hold after controlling for the sector. National differences are greatest for the percentage of product innovations developed via buying-in and collaboration, and lowest for the percentage developed in-house. This suggests that national innovation systems, insofar as they play a role, have a greater effect on the choice between collaboration and buying-in than they do on the decision to innovate in-house.

Table 8.3 Percentage of innovations developed via each innovation method N

In-house

Buying-in

Collaboration

Product/service innovations Food and beverages 126 Chemicals 120 Telecom equipment 98 Computer services 140

68.5 (2.84) 75.0 (2.74) 67.4 (3.34) 73.8 (2.14)

15.4 (2.23) 11.7 (2.17) 16.5 (2.56) 10.0 (1.41)

16.1 (2.15) 13.2 (1.81) 16.1 (2.68) 16.2 (1.65)

100% 100% 100% 100%

Total

484

71.4 (1.36)

13.2 (1.03)

15.4 (1.02)

100%

Process innovations Food and beverages Chemicals Telecom equipment Computer services

118 111 79 118

50.1 (3.14) 58.5 (3.20) 60.2 (3.43) 66.3 (2.47)

35.2 (3.28) 22.3 (2.86) 20.3 (3.02) 13.5 (1.67)

14.7 (1.87) 19.2 (2.37) 19.4 (2.73) 20.2 (1.90)

100% 100% 100% 100%

Total

426

58.7 (1.55) 23.1 (1.43)

18.3 (1.09)

100%

Notes Standard errors in parentheses. See Table 8.2 for other details.

Collaboration and innovation outputs 167 In the KNOW questionnaire, collaboration is deﬁned to include innovation projects that involve other groups or divisions within the same ﬁrm. Not surprisingly, subsidiaries or divisions of multi-location ﬁrms, compared to single entity ﬁrms, develop a higher percentage of their innovations through collaboration (20.1 per cent versus 11.7 per cent, p < 0.000). Conversely, single entity ﬁrms develop more innovations inhouse (74.4 per cent versus 67.6 per cent, p = 0.013). There is no difference between the two types of ﬁrms in the percentage of innovations that are bought-in. The number of competitors could inﬂuence the use of collaboration if increased competition forces ﬁrms to look for more novel or substantial innovations. However, there is no evidence from the survey results that the number of competitors inﬂuenced the percentage of innovations developed through cooperation.

8.5 Innovation sales share On average, 37.4 per cent of sales in the most recent ﬁscal year were from products or services that were ‘signiﬁcantly improved or new to the ﬁrm’ within the last three years. There is no signiﬁcant difference in this mean innovation sales share by country, which varies from 32.2 per cent for the Netherlands to 40.4 per cent for Denmark. In contrast, there are statistically signiﬁcant differences by sector that match common assumptions about low, medium and high technology sectors. The mean innovation sales share is lowest in the food sector, at 21.9 per cent, increases to 29.9 per cent for the chemical sector and is highest for the telecom equipment (47.9 per cent) and computer services (51.0 per cent) sectors. Within each sector, there is no statistically signiﬁcant difference between small and mid-size ﬁrms. If external information sources provide valuable information that the ﬁrm can use to develop innovations, as suggested in much of the literature, then ﬁrms that access external information sources should be more successful innovators than ﬁrms that do not. Greater success in innovation should, in turn, lead to a higher innovation sales share. A ﬁrm can acquire external knowledge of value to its innovative activities either through collaboration or through other methods that do not require the development of personal relationships between the ﬁrm and the producer of the knowledge. In addition to the question on the percentage of innovations developed collaboratively, the survey obtained information on four mechanisms that could involve collaboration. These include whether or not the ﬁrm had R&D projects with PROs, spent R&D on projects conducted by other divisions of the ﬁrm’s group, spent R&D on projects with completely independent ﬁrms and organizations, or received subsidies for innovation projects. Of note, the two questions on the location of R&D expenditures

168 A. Arundel and C. Bordoy do not require active collaboration because the ﬁrm may have contracted out the R&D. Subsidies could proxy for collaboration because many national and European innovation subsidies require some type of collaboration.5 None of these ﬁve measures of actual or possible collaboration are signiﬁcantly correlated with the ﬁrm’s innovation sales share. For example, the mean innovation sales share for ﬁrms with and without R&D projects with PROs is 39.3 per cent and 35.4 per cent, respectively (p = 0.14). The second method of acquiring external knowledge does not require personal contacts with individuals from other ﬁrms or organizations. The survey asks if the ﬁrm regularly seeks ideas for innovation from searching patent databases, reading the scientiﬁc and business literature, attending trade fairs and conferences and reverse engineering of competitor’s products. A ﬁfth question asks if the ﬁrm uses the Internet to search for scientiﬁc and technical information. There is no difference in the innovation sales share for four of these ﬁve methods, but the innovation sales share is 38.7 per cent for ﬁrms that regularly obtain ideas for their innovation projects from reading scientiﬁc or business journals, compared to 31.5 per cent of ﬁrms that do not (p = 0.02). The methods used by the ﬁrm to appropriate its investments in innovation could also inﬂuence the innovation sales share through two mechanisms. Firms that are able to use patents to appropriate their investments could be more willing to collaborate with other ﬁrms because the patent could clarify ownership of jointly developed inventions. In addition, irrespective of collaboration, patents could encourage more investment in innovation, thereby leading to higher innovation sales shares. The questionnaire asked each respondent to identify the most important appropriation method to their ﬁrm: patents, secrecy, lead-time advantages or ‘other’ methods. The average innovation sales shares by the most important appropriation method are given in Table 8.4. For all sectors combined, ﬁrms that ﬁnd patents to be their most important appropriation method have the lowest average innovation sales share (31 per cent), while ﬁrms that ﬁnd lead-time advantages to be their most important appropriation method have the highest average innovation sales share (45.2 per cent). The positive effect of lead-time advantages is logical, since these ﬁrms must continually introduce improved or new products. The result for patents is unexpected, although it is largely due to the computer services and chemicals sectors. Patents are of little value in the former. In chemicals, it is possible that patents, by providing good protection from competitors, reduce the competitive pressure to innovate.

8.6 Econometric results We use regression models to evaluate the two main research questions, as outlined in the introduction: (1) What factors inﬂuence the use of

37.4

All sectors

31.0

19.3 24.6 56.7 31.3 0.03

0.62 0.07 0.17 0.15

P*

34.6

20.3 29.7 46.7 45.2

Mean

Secrecy

0.13

0.58 0.85 0.77 0.12

P*

45.2

22.8 38.2 48.2 59.3

Mean

Lead-times

0.00

0.65 0.02 0.93 0.00

P*

Note * Signiﬁcance of the difference for the appropriation method compared to the three other appropriation methods combined.

21.7 30.3 47.9 51.1

Mean

Overall mean Patents

Mean innovation sales share by most important appropriation method

Food Chemicals Telecom equipment Computer services

Table 8.4

33.0

22.8 29.8 39.7 44.6

Mean

Other

0.05

0.66 0.91 0.23 0.12

P*

170 A. Arundel and C. Bordoy collaboration as an innovation method? (2) Does collaboration inﬂuence the innovation sales share? Both regressions include independent variables for a range of factors that could inﬂuence the dependent variables. These include ﬁrm size, sector of activity, country of location, measures of the ﬁrm’s internal innovative capabilities, the use of external information sources that do not require collaboration such as scanning journals (JOURNAL), indicators for collaboration, the ﬁrm’s ownership status and appropriation conditions. With three exceptions (ﬁrm size, R&D employee share and collaboration product share), all independent variables are dummies, coded as 1 when the characteristic of interest is present and zero otherwise. Both regressions include the log of the number of employees (LOGEMP) because innovation surveys consistently show that large ﬁrms are more likely to collaborate than small ﬁrms, although this could be because larger ﬁrms have more opportunities and more resources for collaboration. The descriptive analyses ﬁnd that the ﬁrm’s sector of activity has a strong effect on the innovative product share, but there are no signiﬁcant differences by sector in the percentage of product innovations developed through collaboration. Nevertheless, other research suggests that collaboration assists the development of advanced technology and should therefore be more prevalent in high technology sectors such as telecom equipment and computer software, after controlling for other factors. We therefore include dummies for three sectors (CHEMICALS, TELECOMEQ, COMPSER) in all regression models, with the food sector as the reference category. The descriptive results found marked differences by country for the percentage of innovations developed through collaboration, but no relationship between the ﬁrm’s country of location and the innovation sales share. We include country dummies in both sets of regressions, with the Netherlands as the reference category, to evaluate the effect of different national systems of innovation. An alternative explanation for differences by country is that the samples are not comparable, even after controlling for ﬁrm size or other variables that could be related to country in the regressions. We cannot address the alternative explanation, except indirectly. We expect that the country of location will have a greater effect on whether or not a ﬁrm has developed an innovation via collaboration than on the innovation sales share. The former should be inﬂuenced by the nexus of linkages that have developed over decades between ﬁrms and between ﬁrms and PROs. In contrast, with the development of the single European market, similar market pressures should apply to all ﬁrms, no matter where they are based. This should reduce national differences in innovation sales shares, with the exception of the effect of indeterminate differences in national innovative capabilities. We expect both the probability of collaboration and the innovation sales share to increase with different measures of the innovative capability

Collaboration and innovation outputs 171 of the ﬁrm, although the effect should be stronger for the innovation sales share. More innovative ﬁrms will have more expertise to exchange with other ﬁrms and therefore could be more likely to collaborate, but conversely ﬁrms with few innovation resources in-house may be forced to collaborate. It is therefore impossible to predict the direction of the relationship, or if the two factors will simply cancel each other out. We use two measures of innovative capability: whether or not the ﬁrm performs R&D on a continuous basis (CONTR&D), with occasional or no R&D as the reference category, and the ﬁrm’s R&D intensity, measured as the share of employees active in R&D (R&DEMPSHARE). Firms with an outward-looking approach to innovation could have a higher innovation sales share if these external contacts improve development efﬁciencies or provide access to new ideas. The model includes two relevant dummy variables: one or more research projects with a PRO (PRIPROJ) and positive R&D spending outside the ﬁrm in independent organizations (EXTR&D). Both variables also provide a proxy for possible collaborative research with independent organizations. They should therefore have a positive effect on the probability of developing at least one innovation via collaboration. Another possible indicator for collaboration is whether or not the ﬁrm received a subsidy for innovation (SUBSIDY). In Europe, many innovation subsidies are conditional on collaboration either with other ﬁrms or with PROs. Firms that are part of a larger group can collaborate both with other divisions of the same group, or with independent organizations and ﬁrms. The models include a dummy variable (INDEPENDENT) that is equal to 1 if the ﬁrm is a single entity. Since these ﬁrms will have zero opportunity to collaborate within their ﬁrm, they could be less likely than group ﬁrms to develop a collaborative innovation. They could also have a lower innovation sales share if collaboration increases this share and if they collaborate less than group ﬁrms. Finally, appropriation preferences could inﬂuence both the willingness to collaborate and the innovation sales share. Firms that ﬁnd patents to be their most important appropriation method (PATENT) could be more willing to collaborate because patents could solve frictions over ownership of the research results, while ﬁrms that ﬁnd lead-time advantages (LTA) to be their most important appropriation strategy should have a higher innovation sales share, based on the descriptive results. 8.6.1 Model speciﬁcations The dependent variable for the innovation sales share for product or service innovations can vary from 0 to 1, with all possible values falling within this range.6 Since the dependent variable cannot take a negative value or exceed unity, an ordinary least square (OLS) model will provide biased coefﬁcients. We use a non-linear model that assumes an ‘S’-shaped

172 A. Arundel and C. Bordoy distribution in the dependent variable. This assumption is reasonable because ﬁrms will need to develop a minimum level of internal expertise before their innovation sales share can increase notably from zero. Conversely, several factors will constrain the innovation sales share from reaching 100 per cent, such as sales from existing products and appropriation and competition effects that prevent ﬁrms from dominating the market with their innovative products. The model takes the following form: exp('xi i) yi 1 exp('xi i) where yi is the innovation sales share and xi is a vector of independent variables. This functional form guarantees that for any value of both the variables and the parameters, the value of the conditional mean lies between 0 and 1. We also applied the non-linear model to the percentage of innovations developed via collaboration. However, the non-linear model did not ﬁt these data, possibly because of the high percentage of ﬁrms with zero innovations developed via collaboration. Consequently, we reverted to a simpler binomial logit in which the dependent variable equals 1 when the ﬁrm introduced at least one innovation developed through collaboration and zero otherwise. The binary logit assumes that there is an unobservable variable yi* such that y 1 if y *i > 0 y 0 if otherwise, and yi* is deﬁned by y *i 'Xi i , where εi has a distribution function F derived from the logistic cumulative distribution function: 1 F(x) . 1 exp(x) Then, given the characteristics Xi of individual i, we have: exp('Xi) Prob (yi 1)1F('Xi) . 1exp('Xi)

Collaboration and innovation outputs 173 For both the model of collaborative innovation and the innovation sale share, we provide results for all ﬁrms, for ﬁrms that are part of a group only and for single entity ﬁrms. Several independent variables are not available for a notable percentage of the respondent ﬁrms and therefore we provide two versions of the model for all ﬁrms: a reduced model that maximizes the number of ﬁrms and a complete model that results in a loss of about 18 per cent of the ﬁrms, due to missing values for some variables. Separate results are given for single entity and group ﬁrms because the deﬁnition of collaboration in the questionnaire includes collaboration both with a ﬁrm that is part of a group and one that is external to the group. Therefore, the variable for collaboration product share among ﬁrms that are part of a group can include an indeterminate mix of innovations developed within the ﬁrm and with external partners. This problem does not occur among the single entity ﬁrms. The full models for both dependent variables use similar vectors of independent variables. The main difference is that the model for innovation sales share includes a variable for the collaboration product share (COLLABSHARE). The vector for this model is as follows: ISSHARE F (0 1COLLABSHARE 2R&DEMPSHARE 3LOGEMP 4CHEMICALS 5TELECOMEQ 6COMPSER 7INDEPENDENT 8EXTR&D 9LTA 10PATENT 11JOURNAL 12PRIPROJ 13SUBSIDY 14CONTR&D jCOUNTRYj i) 8.6.2 Collaboration as an innovation method Table 8.5 provides logit model results for the introduction of one or more innovations developed through collaboration. Firm size, measures of innovative capabilities (R&D employee share, CONTR&D), the ﬁrm’s preferred method of appropriation, the receipt of subsidies for innovation and a practice of regularly scanning journals, have no effect in any of the four models. Four of the ﬁve variables that increase the possibility of developing an innovation via collaboration are positive and signiﬁcant in at least one model. These include whether or not the ﬁrm is independent (negative, indicating that ﬁrms that are part of a group are more likely to introduce a collaborative innovation), any expenditures on external R&D, any expenditures on R&D in independent ﬁrms or organizations (among independent ﬁrms only) and R&D projects with PROs. The exception is SUBSIDY, which has no effect in any of the models. Computer services ﬁrms are also signiﬁcantly more likely than the reference category of food ﬁrms to innovate via collaboration. As noted in the descriptive results, the ﬁrm’s country of location has a substantial impact on collaboration. Compared to the reference category

0.744 0.006 0.263 0.351 0.381 0.002 0.313 0.301

0.287 0.378 0.342 0.274 0.302 0.326

0.561 0.006 0.137

0.405 0.534 1.19***

0.560* 0.740**

0.373 0.470 0.140 0.726*** 0.038 0.190

1.84*** 0.498 0.306 0.440 0.053 0.492 1.560** 0.622 1.040** 0.477 0.323 0.550 358 88.5*** 0.29

Constant R&D employee share Log employees

Dummy variables for sector Chemicals Telecom equipment Computer services

INDEPENDENT Any external R&D R&D in other part same ﬁrm R&D in independent ﬁrms/ organizations Lead time advantages PATENT JOURNAL R&D projects with PROs SUBSIDY CONTR&D

Country dummies GERMANY DENMARK FRANCE ITALY GREECE UK Number of cases Chi-square Pseudo R-square

Notes * 0.05 p 0.10 ** 0.01 p 0.05 *** p 0.10

B

Std error

B

0.272 0.286

0.143 0.152 1.58*** 0.458 0.440 0.411 0.058 0.437 1.17** 0.465 0.949** 0.438 0.490 0.476 498 92.2*** 0.26

0.255 0.349 0.301

0.281

0.319 0.346 0.328

0.638 0.005 0.225

Std error

0.457* 0.022 0.049

0.668**

0.240 0.267 1.10***

0.079 0.006 0.230

Reduced model

Full model

Variables

0.399 0.370 0.636 0.455 0.392 0.410 0.441

0.556 0.559 0.545

1.13 0.010 0.380

Std error

2.12*** 0.844 0.093 0.832 1.36 0.902 1.66* 0.860 1.30* 0.770 0.016 0.905 195 36.8*** 0.24

0.800** 0.218 0.399 0.113 0.572 0.119 0.004

0.128 0.025 1.03*

0.534 0.001 0.077

B

Independent ﬁrms

Table 8.5 Logit results for introducing one or more product innovations developed through collaboration

0.427 0.513 0.537 0.582 0.424 0.503 0.573

0.437

0.508 0.628 0.645

1.10 0.403 0.403

Std error

1.24* 0.685 0.630 0.612 1.66* 0.938 6.57 15.73 0.592 1.26 0.790 0.869 161 40.8*** 0.30

0.287 0.589 0.432 0.313 1.13*** 0.213 0.300

0.038

0.103 1.03 1.21*

0.993 0.136 0.136

B

Group ﬁrms

Collaboration and innovation outputs 175 of the Netherlands, ﬁrms located in Germany, Italy and Greece are signiﬁcantly less likely to develop an innovation via collaboration. There is no signiﬁcant difference for ﬁrms in Denmark, although the coefﬁcients in all regressions are positive. These results suggest that ﬁrms in Denmark and the Netherlands are the most likely to collaborate. This effect could be due to national innovation systems, although it is impossible to tell for sure. 8.6.3 Innovation sales share Table 8.6 provides the non-linear model results for the innovation sales share as the dependent variable. Of note, the percentage of innovations developed via collaboration (collaboration product share) has no effect in any of the four models. Although the results are not shown here, collaboration also has no effect when introduced as a dummy variable (any products developed via collaboration versus no products developed via collaboration). Collaboration can also occur through ﬁrms that spend part of their R&D in divisions of the same ﬁrm (R&D in related ﬁrms), in independent ﬁrms or organizations or as a result of R&D projects with PROs. All three of these variables either have no effect in all four models, or else the effect is negative. As an example of the latter, ﬁrms that are part of a larger group and which have R&D projects with PROs have a lower innovation sales share (coefﬁcient of 0.47, p < 0.05) than group ﬁrms without R&D projects with PROs. One measure of external knowledge sourcing, however, has a signiﬁcant, positive effect on the innovation sales share. Firms that regularly read scientiﬁc or business journals to obtain ideas for innovations (JOURNAL) have a higher innovation sales share than ﬁrms that do not. This effect holds for three of the four models and is in the same direction in the fourth model for independent ﬁrms only. After controlling for other characteristics, particularly the ﬁrm’s sector of activity, the different measures of innovative capability have no effect in the full model or in the separate models for single entity and group ﬁrms. This could partly be because all of the ﬁrms in the model are innovative, which reduces the possible range of innovative capabilities within each sector. In the reduced model, ﬁrms that perform R&D on a continuous basis (CONTR&D) are more innovative than those that only perform R&D on an occasional basis or not at all. Firm size has no effect on the innovation sales share in all four models. As expected, ﬁrms that rely on lead-time advantages have a higher innovation sales share than ﬁrms that do not. Conversely, reliance on patents has no effect on the innovation sales share, except in the model for ﬁrms that are part of a group, where the effect is negative. In all regressions, the ﬁrm’s sector of activity has a signiﬁcant impact on

Notes * 0.05 p 0.10 ** 0.01 p 0.05 *** p 0.10

Constant Collaboration product share R&D employee share Log employees Dummy variables for sector Chemicals Telecom equipment Computer services INDEPENDENT Any External R&D spending R&D in related ﬁrms R&D in independent ﬁrms/ organizations Lead time advantages PATENT JOURNAL R&D projects with PROs SUBSIDY CONTR&D Country dummies GERMANY DENMARK FRANCE ITALY GREECE UK Number of cases F R-square

Variables

0.392 0.003 0.003 0.129 0.200 0.196 0.192 0.156 0.147

0.138 0.202 0.176 0.139 0.155 0.163 0.249 0.224 0.242 0.312 0.244 0.292

1.82*** 0.001 0.004 0.09

0.43** 1.05*** 1.14*** 0.06 0.34*

0.28** 0.08 0.43** 0.15 0.001 0.24

0.09 0.12 0.43* 0.20 0.20 0.24 344 5.79*** 0.26

B

Std error

B

419 7.35*** 0.25

0.228 0.211 0.222 0.228 0.224 0.252

0.136 0.142

0.15 0.31** 0.07 0.14 0.16 0.14 0.31 0.21

0.122 0.189 0.159

0.186 0.179 0.172 0.138

0.350 0.003 0.003 0.109

Std error

0.28** 0.09 0.37**

0.40** 1.03*** 1.15*** 0.003

1.89*** 0.002 0.003 0.02

Reduced model

Full model

Table 8.6 Non-linear regression results for innovation sales share

190 3.49*** 0.28

0.478 0.474 0.502 0.498 0.449 0.529

0.192 0.177 0.308 0.236 0.196 0.206 0.215

0.33* 0.48*** 0.41 0.37 0.009 0.09 0.33 0.43 0.44 0.57 0.20 0.51 0.71

0.294 0.283 0.285

0.628 0.004 0.005 0.186

Std error

0.43 0.83*** 0.99***

2.03*** 0.000 0.005 0.02

B

Independent ﬁrms

0.13 0.350 0.05 0.314 0.58* 0.339 1.47 1.213 1.13* 0.623 0.39 0.434 151 3.23*** 0.33

0.224 0.238 0.308 0.305 0.219 0.262 0.285

0.216

0.26 0.07 0.002 0.604** 0.62** 0.47** 0.15 0.34

0.313 0.339 0.313

0.607 0.005 0.004 0.197

Std error

0.36 1.42*** 1.07***

2.38*** 0.004 0.005 0.31

B

Group ﬁrms

Collaboration and innovation outputs 177 the innovation sales share, as expected. Compared to the reference category of the food sector, the highest innovation sales shares are in the two high technology sectors of telecom equipment and computer services. Compared to the results for collaboration in Table 8.5, the ﬁrm’s country of location has less of an effect on the innovation sales share. There are no signiﬁcant differences by country for the reduced model and for independent ﬁrms. The greatest effect of country is in the model for ﬁrms that are part of a group. In this model, the innovation sales share for ﬁrms located in Greece and France is higher (although only weakly signiﬁcant) than the share for the reference category of the Netherlands. One possible explanation is that group ﬁrms located in Greece and France are introducing innovations that were developed by divisions located in other countries.

8.8 Conclusions Small and mid-size ﬁrms in all four of the sectors examined here develop a large majority of their product and process innovations in-house. Inhouse innovation is less important for process innovation, but most of the difference is due to a higher rate of buying-in than of collaboration. For product innovations, collaboration and buying-in account for almost equal shares of all innovations, at 15.4 per cent and 13.2 per cent, respectively. These results highlight the need for surveys to carefully distinguish between innovative activities based on buying-in versus collaboration. Otherwise survey respondents might assume that buying-in is a form of collaboration because it could require some discussion with technology suppliers. The regression analyses for the probability of developing at least one innovation through collaboration show that computer service ﬁrms are most likely to collaborate, probably because many of them develop customized software and computer solutions that require close contacts with their customers. Not surprisingly, the different proxies for opportunities to collaborate (R&D projects with PROs, positive external R&D spending) also increase the probability of developing at least one innovation via collaboration. Appropriation preferences, ﬁrm size and R&D intensity do not increase the probability of developing at least one innovation via collaboration. The most interesting result of these regressions is the strong effect of location on the probability of developing an innovation through collaboration. Firms located in Germany, Italy and Greece are signiﬁcantly less likely to collaborate than ﬁrms located in the Netherlands. This could be due to differences in national innovation systems, although the survey data do not permit a full exploration of this result. Finally, ﬁrms that are part of a group are more likely than independent ﬁrms to innovate via collaboration, possibly because they can collaborate with other divisions of the same ﬁrm.

178 A. Arundel and C. Bordoy The regression analyses for the innovation sales share fail to ﬁnd any positive effect from collaboration, either for the share of innovations developed via collaboration or for the four proxies for possible collaboration, other than the independent status of the ﬁrm. These include spending part of their R&D in related ﬁrms, in external organizations, through R&D projects with PROs or the receipt of subsidies for innovation. In some models, the effect of external R&D and R&D projects with PROs is negative, suggesting that some forms of collaboration reduce the innovation sales share. A possible explanation is that the two proxies for collaboration, using the location of R&D spending, are limited to activities in the last year, while the innovation sales share is based on innovations introduced within the previous three years. Our model makes the reasonable assumption that ﬁrms with any external R&D in the previous year are likely to have had external R&D spending in earlier years, but this assumption may not be valid. In this case, the negative result for external R&D could be due to ﬁrms turning to collaboration in order to increase their innovation sales share. However, this argument does not hold for the variable for any R&D projects with PROs, which covers all such arrangements over the previous three years. In contrast to the lack of a positive effect for collaboration, ﬁrms that regularly read scientiﬁc or business journals to seek ideas for innovations have a signiﬁcantly higher innovation sales share than ﬁrms that do not. This effect is signiﬁcant (p<0.05) in three of the four models and in the right direction in the fourth. This result is not due to sectoral effects, for example, if journals are a substantially more important source of knowledge in high technology than in low technology sectors, because the model controls for the effect of the ﬁrm’s sector on the innovation sales share. These sector effects are also substantial, as expected. In addition, ﬁrms that ﬁnd lead-time advantages to be their most important appropriation method have a higher innovation sales share, while a reliance on patents for appropriation either has no effect on the innovation sales share or reduces it in one model. In contrast to the regression results for developing at least one innovation via collaboration, the country of location has very little effect on the innovation sales share, with only three weakly signiﬁcant country effects across the four regressions. This suggests that the strong country effects observed in Table 8.5 for collaboration are due to differences in the national innovation system and not to differences in the sub-samples drawn from each country. In general, the results of this study provide no evidence in support of the theory that a ‘collaborative space’ or network based on collaboration plays a key role in innovation or in the innovation sales share. However, this does not mean that external knowledge sources are unimportant, as shown by the strong positive effect of regularly reading journals on the innovation sales share. Access to external knowledge clearly plays a role in successful innovation, even though collaboration may not be the main

Collaboration and innovation outputs 179 mechanism. This conclusion is also supported by a companion study, using the same survey data, on the ﬁrm’s most economically important innovation. Over 70 per cent of ﬁrms stated that external knowledge sources contributed to their most economically important innovation, while almost half (46 per cent) stated that external sources were either more important or of equal importance to their own in-house resources to the development of this innovation (Arundel and Bordoy 2002). In addition to collaboration, there are many other mechanisms for knowledge ﬂows, including the tradition of ‘open science’ (Cohen et al. 2002), informal contacts and job mobility. These could play a more central role than collaboration in a new mode 2 innovative space based on networks. Although we ﬁnd no evidence in support of a key role for collaboration in innovation, there are several important limitations to this study. First, the time constraint for the innovation sales share of three years could limit the opportunity for collaboration to have an effect. This could be particularly true if collaborative innovations are more radical and therefore take more time to develop a market. However, this is less likely in the telecom equipment and computer services sectors, where rapid adoption of new technology is the norm. Second, our sample only covers small and mid-size ﬁrms. Collaboration could be much more important for innovation by large ﬁrms. Third, our results are only applicable to four sectors, although we intentionally selected a low, medium and high technology manufacturing sector and one service sector. There is extensive evidence from other research to show that collaboration is more important in biotechnology and pharmaceuticals. Fourth, we have no information about the technical or economic quality of the innovations developed through collaboration or other methods. It is possible that some ﬁrms in our sample collaborated to develop radical innovations with large economic payoffs, whereas other ﬁrms collaborated to develop minor improvements to existing products. In this case, the effect of these two types of collaboration on the innovation sales share could cancel each other out. Fifth, we lack data on the cost of developing innovations via collaboration versus in-house. In some cases collaborative innovations could be less expensive, which means that they could be more proﬁtable to the ﬁrm, even if they do not take a relatively higher share of innovation sales. Sixth, the innovation sales share is only one measure of the success of innovative activities. We have no data for other possible measures, such as the proﬁtability of different innovation methods. Finally, we implicitly assume that ﬁrms collaborate with other ﬁrms or organizations in order to jointly develop innovative products. However, ﬁrms may have other motivations for cooperation, such as learning about the technical capabilities of their partners or learning how their partners organize their innovation routines. Although this type of knowledge could be beneﬁcial to the ﬁrm and result in higher aggregate sales and proﬁts, it may not show up as a change in the innovation sales share.

180 A. Arundel and C. Bordoy

Notes 1 The seven countries are Germany, France, Italy, Sweden, Norway, Portugal and Ireland. The results vary very little by ﬁrm size. The analyses were conducted by Anthony Arundel. 2 The response rate for the population of innovative ﬁrms, the actual target of the survey, is much higher. For the Netherlands, the response rate is 45.6 per cent, based on the entire population of ﬁrms in the eligible sectors and size classes. However, limiting the target population to innovative ﬁrms only, using the second CIS estimates of the share of innovative ﬁrms in each sector and size class, increases the Dutch response rate to 72.7 per cent. 3 For ﬁrm size, this chapter uses a maximum cut-off of 1,250 employees to allow for natural employment growth between the time that the data in the business registries were collected and the time of the survey. This adds an additional 11 ﬁrms to the sample population. Further details on the survey methodology are available in Chapter 5 of this volume. 4 Respondents had no difﬁculty in answering these questions, with fewer than 1 per cent of respondents failing to answer the question for product innovations. However, we do not expect respondents to have answered the questions in terms of the actual number of innovations, which would be very difﬁcult to estimate. Instead, we expect the answers to refer to a general idea of the relative role of each innovation method to the development of each ﬁrm’s innovations. 5 See the Chapter 14 by Arundel and Vonortas in this volume. 6 We favour this speciﬁcation over other models that assume that values outside the observed range are possible but are unobserved due to truncation or censoring. This is not logically possible for the innovation sales share. The nonlinear model, which also does not permit values outside the range of 0 to 1, is therefore more realistic than other speciﬁcations.

References Albach, H., Audretsch, D.B, Fleischer, M., Greb, R., Hofs, E., Roller, L.-H. and Schulz, I. (1996) ‘Innovation in the European Chemical Industry’, EIMS Project Report, Luxembourg, May. Antonelli C. (1999) ‘Industrial organisation and the production of knowledge’, Cambridge Journal of Economics, 23: 243–260. Arundel, A. and Bordoy, C. (2002) ‘In-house versus ex-house: the sourcing of knowledge for innovation’, in J. de la Mothe and A.N. Link (eds) Networks, Alliances and Partnerships in the Innovation Process, Dordrecht: Kluwer Academic, pp. 67–87. Arundel, A., Steinmueller, E. and Hawkins, R. (1996) ‘Strategies for the future: innovation in the European telecom equipment industry’, Maastricht, MERIT, July. Asheim, B. and Smith, K. (1999) Regional Innovation Systems, Regional Networks and Regional Policy, Cheltenham: Edward Elgar. Cohen, W. and Levinthal, D. (1990) ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35: 128–152. Cohen, W.M., Nelson, R.R. and Walsh, J.P. (2002) ‘Links and impacts: the inﬂuence of public research on industrial R&D’, Management Science, 48: 1–23. Cowan, R., David, P.A. and Foray, D. (2000) ‘The explicit economics of knowledge codiﬁcation and tacitness’, Industrial and Corporate Change, 9: 211–254.

Collaboration and innovation outputs 181 Edquist, C. (ed.) (1997) Systems of Innovation: Technologies, Institutions and Organisations, London: Pinter. Freeman, C. (1987) Technology Policy and Economic Performance: Lessons from Japan, London: Pinter. Fritsch, M. and Lukas, R. (2001) ‘Who cooperates on R&D?’, Research Policy, 30(2): 297–312. Georghiou, L. (1998) ‘Science, technology and innovation policy for the 21st century’, Science and Public Policy, 25: 135–139. Geroski, P. (1995) ‘Markets for technology: knowledge, innovation and appropriability’, in P. Stoneman (ed.) Handbook of the Economics of Innovation and Technological Change, Oxford: Basil Blackwell, pp. 90–131. Gibbons, M., Limoges, C., Nowotny, H., Schwartzman, S., Scott, P. and Trow, M. (1994) The New Production of Knowledge, London: Sage. Hagedoorn, J. and Schakenraad, J. (1992) ‘Leading companies and networks of strategic alliances in information technologies’, Research Policy, 21(2): 163–190. Hagedoorn, J., Link, A.N. and Vonortas, N. (2000) ‘Research partnerships’, Research Policy, 29(4–5): 567–586. Howells, J. (1999) ‘Research and technology outsourcing’, Technology Analysis and Strategic Management, 11: 17–29. Kline, S. and Rosenberg, N. (1986) ‘An overview of innovation’, in R. Landau and N. Rosenberg (eds) The Positive Sum Strategy, Washington, DC: National Academic Press, pp. 275–306. Landry, R. and Amara, N. (2003) ‘Effects of sources of information on novelty of innovation in Canadian manufacturing ﬁrms’, in F. Gault (ed.) Understanding Innovation in Canadian Industry, Kingston: McGill-Queen’s University Press, pp. 67–110. Lundvall, B.-Å. (1991) ‘Innovation, the organised market and the productivity slowdown’, in Technology and Productivity: The Challenges for Economic Policy, Paris: OECD, pp. 447–458. Lundvall, B.-Å. (1992) (ed.) National Systems of Innovation, London: Pinter. Lundvall, B.-Å., Johnson, B., Andersen, E.S. and Dalum, B. (2002) ‘National systems of production, innovation and competence building’, Research Policy, 31: 213–231. Malerba, F. (2002) ‘Sectoral systems of innovation and production’, Research Policy, 31: 247–264. Malerba, F., Lissoni, F. and Torrisi, S. (1997) ‘The ofﬁce and computer equipment industry’, EIMS Project Report, Luxembourg: European Commission. Mansﬁeld, E. and Lee, J.-Y. (1996) ‘The modern university: contributor to industrial innovation and recipient of industrial R&D support’, Research Policy, 25(7): 1047–1058. Mohnen, P. and Therrien, P. (2003) ‘Comparing the innovation performance of manufacturing ﬁrms in Canada and in selected European countries: an econometric analysis’, in F. Gault (ed.) Understanding Innovation in Canadian Industry, Kingston: McGill-Queen’s University Press, pp. 313–339. Narin, F., Hamilton, K.S. and Olivastro, D. (1997) ‘The increasing linkage between US technology and public science’, Research Policy, 26: 317–320. Nas, S.O., Sandven, T. and Smith, K. (1994) ‘Inovasjon og teknologi I norsk industri: En oversikt’ (Innovation and New Technology in Norwegian Industry: An Overview), STEP Report 4/94, STEP, Oslo.

182 A. Arundel and C. Bordoy Nelson, R. (ed.) (1993) National Innovation Systems: A Comparative Study, New York: Oxford University Press. Pavitt, K. (1991) ‘What makes basic research economically useful?’, Research Policy, 20(4): 109–119. Rothwell, R. (1992) ‘Successful industrial innovation: critical factors of the 1990s’, R&D Management, 22(3): 221–239. Saviotti, P.P. (1998) ‘On the dynamics of appropriability, of tacit and of codiﬁed knowledge’, Research Policy, 26: 843–856. Senker, J. (1995) ‘Tacit knowledge and models of innovation’, Industrial and Corporate Change, 4(2): 425–447. Tether, B. (2002) ‘Who co-operates for innovation, and why: an empirical analysis’, Research Policy, 31: 947–967. Therrien, P. and Chang, V. (2003) ‘Impact of local collaboration on ﬁrms’ innovation performance’, in F. Gault (ed.) Understanding Innovation in Canadian Industry, Kingston: McGill-Queen’s University Press, pp. 111–138. Veugelers, R. and Cassiman, B. (1999) ‘Make and buy in innovation strategies: evidence from Belgian manufacturing ﬁrms’, Research Policy, 28: 63–80. Vonortas, N. (1997) Cooperation in Research and Development, Boston, MA: Kluwer Academic. von Hippel, E. (1988) The Sources of Innovation, Oxford: Oxford University Press.

9

Firm size and openness The driving forces of university–industry collaboration1 Roberto Fontana, Aldo Geuna and Mireille Matt

9.1 Introduction A large number of works have studied university–industry relationships either from a qualitative point of view or relying on a case study of a single university. The aim of this chapter is to provide some statistical evidence at the cross-country–cross-industry level to verify some of the hypotheses put forward in the qualitative literature. On the basis of the results of the KNOW survey carried out in seven European Union (EU) countries in 2000, we examine two main issues. First, the contribution made by public research organizations (PROs) to the innovative process of ﬁrms is analysed. Second, the existence and the extent of cooperative R&D projects between ﬁrms and PROs are examined. A two-equation econometric model evaluates the effect of ﬁrm-speciﬁc, sector-speciﬁc and country-speciﬁc factors (such as ﬁrm size, appropriation and signalling, searching of knowledge sources, government support) upon the propensity for and the extent of collaborations between PROs and ﬁrms. The analysis in this chapter provides some preliminary evidence which allows a better understanding of the ﬁrm and industry characteristics that affect the contribution of PROs to ﬁrms’ innovative activities and to their involvement in R&D collaborations with ﬁrms. The estimations produce some evidence that highlights how the size of the ﬁrm and its openness to the external environment have a signiﬁcant effect on both the extent of and propensity for PRO-ﬁrm collaboration. The discussion of university–industry relationships, which entered the policy arena in the early 1980s, has become the property of both academics and the general public. An enormous number of contributions to academic writings and articles in the business and public press have come from policy makers in the last few years in a bid to explain, justify and regulate the interactions between universities and ﬁrms. At the European level, very few of these works have been supported by systematic data analysis. If we exclude the results of the Policies, Appropriability and Competitiveness for European Enterprises (PACE) questionnaire (which focused on large EU research and development (R&D) intensive ﬁrms),

184 R. Fontana, A. Geuna and M. Matt and the scant information on the role of universities and public research centres available from Community Innovation Surveys (CIS) I, II and III, there is little other evidence.2 In a few European countries in recent years, country-speciﬁc data have been gathered and analysed. For example, the studies of Meyer-Krahmer and Schmoch (1998) and Beise and Stahl (1999) provide interesting evidence of the contribution of public research to industrial innovation in Germany. Several works have studied university–industry relationships from a qualitative point of view or by relying on a case study of a single university.3 The aim in this chapter is to provide some statistical evidence at the cross-country–cross-industry level to verify some of the hypotheses put forward in the qualitative literature. The analysis here provides preliminary evidence of ﬁrm and industry characteristics that affect the contribution of PROs (deﬁned here as universities and other public research centres) to ﬁrms’ innovative activities and that inﬂuence ﬁrms’ involvement in R&D projects with PROs. We use the results of the 2000 KNOW survey covering seven EU countries, including the four largest. The survey was limited to ﬁve sectors: food and beverages, chemicals (excluding pharmaceuticals), communications equipment, telecommunications services, and computer services and focused on small- and medium-sized enterprises (SMEs). We examine two main issues: the contribution made by PROs to the innovative process within ﬁrms and the existence as well as the extent of cooperative R&D projects between ﬁrms and PROs. The descriptive analysis aims at distinguishing the relationship between sources of knowledge and speciﬁc phases of the innovation process. Also, although PROs are rarely seen as being the most important source of innovation completion for ﬁrms, we examine what are the speciﬁc patterns characterizing PRO-ﬁrm relationships at sector level when PROs are the most important sources of knowledge. The analytical part of this chapter is based on direct measurement of the extent of collaborations between ﬁrms and PROs. Unlike previous studies we have information both on the importance of university research for the innovative process of ﬁrms and, most signiﬁcantly, on the number of research and development projects conducted jointly with PROs in the three years before the KNOW survey (1997–2000). A two-equation econometric model evaluates the effect of ﬁrm-speciﬁc, sector-speciﬁc and country-speciﬁc factors (such as ﬁrm size, appropriation and signalling, searching knowledge sources, government support) upon both the probability of an R&D collaboration developing with PROs, and the number of R&D projects developed by the ﬁrm in the previous three years. Particular attention is devoted to the role of ﬁrm size with measures of both total employment and R&D employment being used. Also, the idea that the openness of the ﬁrm to the external environment has an important effect on the development of collaboration with a PRO is tested via a set of proxies for this phenomenon, such as the extent to which ﬁrms

University–industry collaboration 185 actively search for relevant scientiﬁc information in publications, hold patents and participate in government-funded projects. The chapter is organized as follows. Section 9.2 brieﬂy reviews the literature on university–industry relations. Section 9.3 presents a descriptive analysis of the contribution of PROs to the innovation process in ﬁrms. The propensity for and extent of PRO-ﬁrm collaborations is examined in Section 9.4 using an econometric model. Finally, Section 9.5 presents the conclusions and summarizes the limitations of current policy actions.

9.2 University–industry cooperation: a review of the literature based on survey analysis Since the 1980s, many countries have implemented policies to facilitate the transfer of knowledge from universities to companies via establishment of legal frameworks; creation of technology transfer ofﬁces inside universities; increasing the mobility of researchers to industry; large cooperative R&D programmes; etc. Some analyses show that these policy measures have contributed greatly to increasing the number and the scope of links between the two worlds.4 However, there is no clear evidence on their economic impact. The relationship between university and industry is a complex and heterogeneous phenomenon. Actually, the channels used by ﬁrms to draw on knowledge developed by PROs are diverse. The intensity of links varies across ﬁrms, sectors and countries. The extensive literature on university–industry is empirical and based on case studies, patent and bibliometric analyses, or large surveys. Rather than trying to be exhaustive, our literature review aims at linking this contribution to a speciﬁc approach: we will mainly present the results from survey analysis and progressively focus on formal cooperation as being the speciﬁc channel of interaction used by both worlds. On the basis of these criteria, we will examine ﬁve important and inter-connected issues addressed by this literature. One part of the literature analyses the impact of scientiﬁc results on the economic sphere – regardless of which channel of interaction is used. A second issue, from the ﬁrm’s point of view, concerns the relative importance of PROs as an external source of information for new ideas and innovation completion. Other contributions study the variety and importance of channels (that is, publications, informal contacts, conferences, recruitment of students, formal collaborative contracts, etc.) used by both actors to exchange knowledge. We will pay speciﬁc attention to the relative importance of formal collaborative agreements. Finally, a set of econometric models, which draw upon the results of innovative research carried out in PROs highlights the characteristics of ﬁrms. Very few analyses based on large surveys focus on formal agreements. One of our aims is to shed new light on this speciﬁc topic. One strand of publications analyses the impact and inﬂuence of scientiﬁc research results on the economic system. Some of these studies are

186 R. Fontana, A. Geuna and M. Matt based on surveys of ﬁrms in different industry sectors. They show that without results developed by academics, many innovations could not have been realized or would have come much later (Mansﬁeld 1991; Beise and Stahl 1999). Cohen et al. (1998) underline that academic research has positively inﬂuenced ﬁrms’ sales, research productivity and patenting activity. These positive impacts are conﬁrmed by studies based on bibliometric data and regression analysis. For instance, Narin et al. (1997), using citations in patents to non-patent literature (such as journal articles, books and abstracts), conclude that the knowledge ﬂow between the two worlds increased threefold in the United States (US) between the end of the 1980s and the mid-1990s. More generally, US studies highlight that geographical proximity between university and industry increases the beneﬁts of academic research. These studies conclude that PROs produce substantial R&D spillovers, but they do not analyse the channels through which university research impacts on industrial innovation. The latter is the aim in this study. Using the Carnegie Mellon Survey on industrial R&D, Cohen et al. (2002) examine a broad range of information sources used by ﬁrms to innovate, of which one is the R&D conducted in PROs. They distinguish between the sources of information that contribute to innovative ideas and to the completion of innovation. With the exception of a few industries (pharmaceuticals, petroleum, etc.), PROs do not play a central role in suggesting new ideas. Overall, PROs seem to be more important for innovation completion than for suggesting new ideas. Although in both phases public research is less important than contributions from the vertical chain of production (suppliers, buyers, the ﬁrm itself), among the sources not in the production chain (competitors, consultants, joint ventures) PROs are signiﬁcant. The analysis also highlights that different sectors behave differently. In Europe, the CIS I, II and III found PROs were not considered to be a major source of information, but made no distinction between the different phases of innovation. The results of the KNOW survey used in this chapter allow us to take account of both the innovation idea and the innovation completion phases and thus to make comparisons with the results of Cohen et al. (2002). Moreover, we attempt to identify, from the list of possible sources, the most important contributors to both phases of innovation. What are the key channels through which PROs affect industrial innovation? Cohen and his colleagues found that in the US the channels of open science, especially publications, public meetings and conferences and also informal information exchange and consulting, were the most important. Cooperative ventures do not seem to have been so important as other channels for industrial R&D. These results are controversial in relation to European contributions. For instance, based on a survey of ﬁrms and universities, Meyer-Krahmer and Schmoch (1998) found collaborative research and informal contacts to be the most important channels

University–industry collaboration 187 of communication. The signiﬁcance of formal collaboration is conﬁrmed by the CIS I and II surveys. These surveys show that ﬁrms consider universities to be important partners in technological cooperation. According to the European Commission (EC) benchmarking study, universities are increasingly involved in cooperative R&D. However, the results of the PACE survey highlight that large companies class recruitment of new graduates, informal contacts and contract research as the most important ways to access academic knowledge, and that ‘low tech’ sectors favour formal collaboration much more than do ‘high tech’ sectors (Arundel and Geuna 2004).5 The European evidence would lead one to conclude that formal collaboration is also becoming an important channel for accessing knowledge. In our sample, although ﬁrms do not deem that PROs play an important role in the innovation process, about half of the ﬁrms have formal collaborations with PROs. What increases the propensity of ﬁrms to draw upon public research (all channels considered)? In a regression analysis, Cohen et al. (2002) take size and age of the ﬁrm as the two explanatory variables. Larger ﬁrms and start-ups have a higher probability of beneﬁting from academic research. Other studies (Arundel and Geuna 2004; Schartinger et al. 2001) incorporate additional explanatory variables, such as level of R&D expenditure, degree of ﬁrms’ innovativeness. A more recent study (Laursen and Salter 2003) introduced into this literature the concept of ﬁrms’ open search strategies. Firms that adopt open search strategies have a higher probability of considering the knowledge produced by universities as important for their innovation activities. As cooperation is considered both by empirical studies and by policy makers to be a central channel, it seems important to analyse it more deeply. Paradoxically, studies based on surveys rarely focus on R&D cooperation. Hall, Link and Scott (2000) analysed Advanced Technology Programme (ATP) cooperative agreements and concentrate on those involving universities. Collaborative projects with universities are more innovative and risky than those not involving universities: they encounter more difﬁculties, but they are more stable. Drawing on CIS II, Mohnen and Hoareau (2002) ﬁnd that ﬁrms that cooperate with universities are generally large, are active in scientiﬁc sectors, patent and receive government support. Firms that are part of a group that cooperates rely less on collaborations with universities than with independent ﬁrms. Mohnen and Hoareau hypothesized that in a conglomerate, collaborations with universities are established at the headquarters level. Our econometric model focuses on cooperation and conﬁrms the results obtained by Mohnen and Hoareau (2002). However, we go a step further and use the concept of openness of ﬁrms, ﬁrst introduced by Laursen and Salter (2003). Their measure of the degree of openness depends on the number of external channels6 of information used by ﬁrms to innovate. It is based upon the idea that the strategy regarding

188 R. Fontana, A. Geuna and M. Matt openness is a search strategy. Our concept of openness is based upon the idea that besides different channels there might be different mechanisms used for knowledge exchange. For instance, knowledge exchange may involve the combination of a screening strategy and a signalling activity. First, searching and screening actions correspond to the process of looking for knowledge outside the border of the ﬁrm. If the external knowledge is codiﬁed, the ﬁrm will devote resources to screening the information contained in it, for example, in publications databases. If the external knowledge is tacit, it becomes strategic to look for potential partners to increase the sharing possibilities. Participation in governmentfunded R&D projects is an appropriate way to meet new partners and ‘screen’ them (to learn about them, their competencies and their networks) (Matt and Wolff 2003). Disclosure of public information constitutes an important signalling strategy (Matt and Wolff 2003), the second important element of openness. The signalling activity is how ﬁrms inform the outside environment about their range of competencies. For example, patenting activity, especially for small ﬁrms, has the dual role of protecting results and signalling domains of competences. Finally, we use a direct measure of the extent of collaborations between ﬁrms and PROs. In contrast to earlier work we have information both on the importance of PRO research and on the number of research and development projects with PROs. This allows us to study both the propensity of a ﬁrm to cooperate with a university (do they cooperate or not?) and the extent of this cooperation (number of R&D projects).

9.3 Descriptive analysis of the contribution of public research organizations to the innovation process The relevance of external contributors in the innovation process may change depending on whether the early or the late stage of innovation is considered. In this section we present some descriptive statistics concerning the role of external information sources in the innovation process. The aim is to separate the relationship between the sources of knowledge and the particular phases of the innovation process and to see whether speciﬁc patterns characterizing the role of PROs emerge at sector level. The analysis, based on the results of the KNOW survey carried out in 2000, covered seven EU countries, including the four largest,7 and focused on ﬁve sectors: food and beverages (NACE 15), chemicals, excluding pharmaceuticals (NACE 24 minus NACE 24.4), communications equipment (NACE 32), telecommunications services (NACE 64.2) and computer services (NACE 72). These speciﬁc sectors were chosen to provide a range of low, medium and high technology manufacturing and to include two innovative service sectors. In each country, a random sample of ﬁrms from two size classes (10–249 employees and 250–999 employees) within

University–industry collaboration 189 each of the ﬁve sectors was drawn from a national business registry. The response rates by country varied from a minimum of 9 per cent in the UK to a maximum of 76 per cent in Denmark. The average response rate was 25 per cent (and 33 per cent if the UK is excluded). Of the 675 ﬁrms that responded, 558 – all innovators – were retained for the analysis (non-innovative ﬁrms were excluded).8 9.3.1 The external sources of information for innovative ideas and innovation completion For each innovator ﬁrm, the assessment of the role of external information sources in different stages of the innovation process was made by distinguishing between the contribution in the early phase of ideas generation and in the late phase of ﬁnalization. The analysis presented here is based on responses that refer only to ﬁrm’s most economically signiﬁcant innovation introduced in the previous three years. Figure 9.1 depicts the weighted percentages of respondents that answered positively to the question about whether a speciﬁc external information source had contributed to the original idea behind the innovation. Results by sector may sum to more than 100 per cent because more than one answer was allowed. What is ﬁrst suggested by the distribution of responses is the high relevance, for all sectors, of customers, competitors and suppliers as sources of innovative ideas (with the exception of telecommunications services, these sources were all rated at higher than 20 per cent), compared to Computer service Telecomm. service Communications equipment Chemical Food

Consultants

PRIs and universities

Customers

Suppliers

Competitors 0

20

40

60

80

Weighted percentage of respondents

Figure 9.1 External sources of information for innovation ideas by NACE sector.

190 R. Fontana, A. Geuna and M. Matt consultants and PROs. In the case of the communications equipment sector, suppliers are reported as being the major contributors to ideas for innovation. The sector that relies more than any other on PROs as a source of ideas is the chemicals sector followed by communications equipment and food and beverages. Another peculiarity of the chemicals sector is that it relies very heavily on competitors, while all sectors rely heavily on customers for ideas. In the case of innovation completion the results tend to mirror those for innovative ideas with one major difference (see Figure 9.2). While customers and suppliers are still the most relevant categories (between 20 per cent and 40 per cent of the answers in each sector), competitors, PROs and consultants are considered relevant by less than 20 per cent of the respondents. Although a large number of respondents indicated that customers are important contributors to innovation completion, in this stage of the innovation process, sector differences seem to dominate. It should be noted that the relevance of PROs as a source of information for innovation completion is small in the case of food and beverages, but with respect to ideas, it is signiﬁcant for the telecommunication services sector. Our results are comparable to those of Cohen et al. (2002) for the US, but with a small difference. As in their study, PROs never score higher than the actors in the vertical chain of production and sale (that is, cus-

Computer service Telecomm. service Communications equipment Chemical Food

Consultants

PRIs and universities

Customers

Suppliers

Competitors

0

20

40

60

80

Weighted percentage of respondents

Figure 9.2 External sources of information for innovation completion by NACE sector.

University–industry collaboration 191 tomers and suppliers in our case) for both phases of innovation. They are nevertheless comparable to other sources (that is, consultants and competitors). Our results also conﬁrm that rivals are a more important source for innovation ideas and PROs dominate competitors for innovation completion (except in chemicals). In Cohen et al. (2002) PROs are ranked higher than consultants in both phases, while according to our analysis, in all sectors except chemicals consultants are preferred. 9.3.2 The most important contributors to innovative ideas and innovation completion After identifying external contributors to innovation, we made an attempt to identify, from the options listed, those most important for both ideas and innovation completion. In both phases, and in almost all sectors, customers were identiﬁed as the most important source. As contributors to innovative ideas suppliers ranked second and competitors third, but ranked lowest as contributors to innovation completion.9 PROs generally were ranked immediately below consultants both in terms of contributing to ideas and in terms of contributors to innovation completion (see Figure 9.3). However, results did differ by sector. For instance, for chemical ﬁrms’ suppliers and competitors play a major role in both phases of innovation. Results are similar in both phases for consultants in the case of

Computer service Telecomm. service Communications equipment Chemical Food

Consultants

PRIs and universities

Customers

Suppliers

Competitors

0

20

40

60

80

Weighted percentage of respondents

Figure 9.3 Most important external source of information for innovation ideas by NACE sector.

192 R. Fontana, A. Geuna and M. Matt Computer service Telecomm. service Communications equipment Chemical Food

Consultants

PRIs and universities

Customers

Suppliers

Competitors 0

20

40

60

Weighted percentage of respondents

Figure 9.4 Most important external source of information for innovation completion by NACE sector.

telecommunication services. In the case of innovation completion, telecommunication services display the highest percentages for PROs, while food and beverages display the highest percentages in the PRO category in the case of innovation ideas. Of particular note is that a higher share of chemical ﬁrms indicated PROs as ‘most important’ in innovation completion than considered PROs to be the most important for innovative ideas (see Figure 9.4). 9.3.3 Why and how do ﬁrms in the chemicals and food industries approach universities? The descriptive statistics presented above suggest that, for the ﬁrms included in our sample, PROs are not generally considered to be important for innovation completion. Nevertheless, the subset of ﬁrms that do consider PROs as the most important source for innovation completion can be characterized by a speciﬁc behaviour in the way they approached this source of external knowledge. Analysing whether there are differences in this respect will enable us to identify some of the determinants of this behaviour. We divided the sample into two groups: G1 is made up of ﬁrms that identiﬁed PROs as the most important contributors to innovation completion and G0 includes ﬁrms that identiﬁed other sources as the most important. We analysed whether the characteristics of the G1 ﬁrms were

University–industry collaboration 193 Table 9.1 Differences in responses between groups Chemicals

Food

Motivation for knowledge acquisition

G1

G0

G1

G0

Cost and risk reduction Update of technical expertise Building on others innovation Meet government regulation Mode of contact Previous experience Business or professional associations Trade fairs and conferences Internet Communication methods Informal contacts R&D cooperation Exchange of students Type of knowledge acquired Technical and scientiﬁc Linked to the market

YES YES (+) NO NO ()

NO YES NO NO

YES (+) YES (+) NO (+) YES

YES YES NO NO

YES (+) NO () YES NO (+)

YES NO NO NO

YES (+) NO (+) YES NO

YES NO NO NO

YES () YES NO

YES NO NO

YES (+) NO (+) NO (+)

YES NO NO

YES () NO

YES YES

YES (+) YES (+)

YES YES

different from those of G0 respondents based on the answers to questions in four modules: motivation for knowledge acquisition, mode of contact, communication method, and type of knowledge acquired. The sample involved ﬁve sectors, but for only two (chemicals and food and beverages) was there a sufﬁcient number of observations to develop this analysis. The results are presented in Table 9.1. A YES in the cell indicates that more than 50 per cent of the respondents in that group replied positively to the question; similarly a NO indicates an over 50 per cent negative response. For example, in the chemicals sector, for the ﬁrst question more than 50 per cent of the respondents who identiﬁed PROs as the most important contributors to innovation completion had decided to obtain knowledge from PROs in the interests of cost and risk reduction; less than 50 per cent of the respondents who identiﬁed other actors as the most important contributors to innovation completion had based their decision on these aspects. So, if G1 answered in a signiﬁcantly different way from G0 – that is, moving from YES to NO, or vice versa – this would indicate that G1 attributed a speciﬁc role to PROs compared to other possible partners (customers, suppliers, consultants, competitors). When G1 and G0 are both YES or are both NO, the sign between brackets indicates whether the response rate of the G1 group differs by 10 per cent or more from the response rate of G0. No sign means that the share of the two groups’ answers was approximately the same. In the remainder of this section we summarize the results of the chemicals and the food and beverages sectors.

194 R. Fontana, A. Geuna and M. Matt 9.3.3.1 Chemicals sector Respondents from the chemicals sector who identiﬁed PROs as the most important contributors to innovation completion seem to adopt a knowledge sourcing strategy that speciﬁcally taps PROs. The motivation for chemicals ﬁrms to exploit PROs rather than other external partners is related to cost and risk reduction. The connection is generally based on long-term relationships (previous experience is important), which have become formalized via cooperative agreements. Nevertheless, informal contacts play a part in information exchange. Trade fairs and conferences tend to be the preferred places for chemicals ﬁrms to meet PROs. Chemicals ﬁrms mainly acquired technical and scientiﬁc knowledge from PROs. Other information from the ten interviews that were conducted in each country in the survey also conﬁrms that relationships with universities are established via public programmes and are reinforced by hiring university researchers.10 9.3.3.2 Food and beverages sector In the case of the food and beverages sector, the importance of PROs compared with other partners is much less clear. What distinguishes this sector is the overriding necessity to meet government regulations. Respondents who identiﬁed PROs as the most important contributors to innovation completion all seemed to have developed links with universities to enable them to meet the requirements of government regulation. G1 respondents gave more positive answers than G0s for motivations related to cost and risk reduction and updating of technical expertise. G1s have links with known partners (previous experience), but establish mainly informal contacts (formal R&D agreements are the exception rather than the rule). G1 food and beverage companies acquired technical and scientiﬁc information from their academic partners. The SMEs in the food and beverages sector regarded universities as the experts able to deal with many of the major issues they face: BSE (bovine spongiform encephalopathy), food quality, safety constraints in food production, etc. These constraints are often established by government, but can also be imposed by large distributors, and might require evidence, for instance, of the hygiene standards in the production process. The statistical evidence and the responses from the interviews seem to suggest that PROs have a speciﬁc role in the food and beverages sector. They provide reliable and upto-date testing facilities to show that products meet the regulations (imposed by government or other institutions). Such activities (testing and expert advice) do not necessarily involve formalized agreements.

University–industry collaboration 195

9.4 Identifying the factors explaining the propensity and the extent of pro-ﬁrm cooperation The analysis in the previous section suggests that there are different reasons why ﬁrms interact with PROs. However, a speciﬁc communication channel was not considered. In this section we look further into what determines the willingness of ﬁrms to establish formal cooperation with a university. More particularly, we provide a quantitative assessment of the propensity for and the extent of ﬁrms’ engagement in collaborations with PROs. In Section 9.4.1 we focus on identiﬁcation and selection of the variables to include in an econometric model. In Section 9.4.2 we estimate the model. Consistent with the ﬁndings from other surveys of ﬁrms’ innovative activity (Klevorick et al. 1995; Arundel et al. 2000; Cohen et al. 2002; Swann 2002), the ﬁrms included in our sample infrequently rated PROs as the most important source of information. About 50 per cent of them had had some cooperation with PROs in the three years before the questionnaire; of the 458 ﬁrms that responded 222 said they had been involved in one or more R&D cooperation with PROs in the previous three years. Participation in cooperative projects varied depending on which industry ﬁrms belonged to. Food and beverages, and chemicals are the industries with the largest share of ﬁrms collaborating with PROs while telecommunication services is the industry least involved with PROs. A relatively large number of computer services ﬁrms never cooperate with PROs, although some have conducted a number of research and development projects with PROs (more than six in the last three years).11 Table 9.2 shows a subdivision of the number of cooperative projects broken down by sectors.12 Overall, the ﬁrms surveyed had an average of 1.6 R&D contracts with PROs; they had collaborated with PROs from a minimum of 0 to a maximum of 25 times and the distribution of their cooperation is very skewed (see Appendix A9.1 for the descriptive statistics). The population of ﬁrms carrying out R&D projects with PROs can be described as being composed of a large number of organizations cooperating in only a small way, and a small group of ﬁrms involved in a large number of cooperative

Table 9.2 Share of respondents for PROs contract classes Contract classes

Food

Chemicals

Communications equipment

Telecomm service

Computer service

0 1 2 3

44.7% 43.0% 7.9% 4.4%

44.5% 37.3% 16.4% 1.8%

52.3% 30.2% 15.2% 2.4%

75.6% 22.0% 0% 2.4%

56.1% 28.0% 7.5% 8.4%

196 R. Fontana, A. Geuna and M. Matt agreements. Although PROs are rarely the most important source of information either for innovative ideas or innovation completion, they have developed cooperative relationships with ﬁrms with different frequencies. Two questions stand out. Why did certain ﬁrms collaborate with PROs during the three years before the questionnaire while others did not? And, what are the characteristics of the ﬁrms that might explain the different levels of cooperation with PROs? 9.4.1 The econometric model To answer these questions we developed an econometric model that facilitates evaluation of the effect of ﬁrm-speciﬁc, industry-speciﬁc and countryspeciﬁc factors upon the number of cooperations between ﬁrms and PROs. The aim of the regression analysis is twofold. The main purpose is to test for the existence of a relationship by analysing the propensity for ﬁrms to engage in R&D projects with PROs and identifying some ﬁrmspeciﬁc, industry-speciﬁc and country-speciﬁc characteristics. In addition, we aim to measure the extent of the relationship as proxied by the number of R&D projects that ﬁrms have been engaged in with PROs. To highlight the determinants that could affect the relationship we focused on ﬁrm characteristics. In particular we identiﬁed four broad classes: (1) ﬁrm size; (2) openness of the ﬁrm; (3) ﬁrm activity; (4) type of innovation process. From the questionnaire, speciﬁc questions designed to elicit information regarding each of these classes were selected in order to construct independent variables. In this section we discuss the choice of these variables. Descriptive statistics are reported in Appendix A9.1. 9.4.1.1 Accounting for ﬁrm size The role of ﬁrm size in inﬂuencing the propensity of ﬁrms to collaborate with PROs is one of the basic tenets of the literature on university– industry relationships as acknowledged in recent empirical investigations (Arundel and Geuna 2004; Mohnen and Hoareau 2002; Cohen et al. 2002; Laursen and Salter 2003). The rationale underlying the role of ﬁrm size in affecting the progress of R&D collaboration is that big ﬁrms have more resources to help them to establish their relationships with PROs, while the smaller the ﬁrm, the fewer the resources that are available to develop multiple relationships.13 As a measure of ﬁrm size we considered the number of employees (EMPLOYEES). Beside this measure we relied also on another measure of size: R&D employment (R&D). This is an indicator of the research size of the ﬁrm rather than of its overall size, which is accounted for by the number of employees.

University–industry collaboration 197 9.4.1.2 Openness of the ﬁrm We deﬁne openness as the attitude of ﬁrms towards establishing a relationship with PROs. As mentioned in Section 9.2, the concept of openness we propose focuses on the mechanisms through which knowledge can be imported from outside the ﬁrm rather than on the channels used. These mechanisms can be proxied by different ‘enablers’. For instance, to get access to external knowledge ﬁrms also have to undertake an in-depth screening activity. Screening entails selection from sources of codiﬁed as well as tacit knowledge. In our contribution, the screening activity involves both analysis of publications databases and participation in public funded R&D programmes. Moreover, ﬁrms may combine the screening activity with a strategic signalling of their range of competencies to the external world. By signalling their competencies, ﬁrms will attract potential partners and thus open up new opportunities for collaboration. Among other things, patenting is part of this signalling strategy. Firms regularly tap different sources to obtain ideas for innovation. Indeed, it seems that public research plays only a minor role in suggesting new ideas for innovation projects. This was one of the most important ﬁndings of the Yale Survey (Levin et al. 1987). More recently, these ﬁndings have been conﬁrmed by other investigations. Surveying a sample of US ﬁrms included in the Carnegie Mellon R&D survey of industrial manufacturers, Cohen et al. (2002) found that sources of information more directly linked to the production and sales chain, such as suppliers and customers, are regarded as most important by ﬁrms. On the basis of the ﬁndings of CIS III for UK ﬁrms, Swann (2002) argues that the extent to which ﬁrms rely on external information channels other than private and public research institutes is a consequence of the type of innovative activity in which the ﬁrms are engaged. Firms engaged in process innovations for instance are more likely to collaborate with universities and PROs than to use them only as sources of information. This suggests that additional external sources of information are complements to rather than substitutes for collaboration with PROs. Among the external channels of information generally considered, there is participation in trade fairs and conferences, searching patent databases and reading scientiﬁc and business publications. As a determinant of the propensity for collaboration with PROs, publications as a source of ideas seem to be particularly important since reliance on them indicates the relevance of academic research for the innovative process. Thus we have constructed a dummy variable (PUBLICATIONS), which takes the value 1 when the ﬁrm screens information from scientiﬁc and business journals and 0 when it does not. We expect this variable to positively affect participation in collaborative projects with PROs. Secrecy and lead time are generally considered to be the preferred methods for exploiting the beneﬁts of process innovation. Patenting is

198 R. Fontana, A. Geuna and M. Matt required to protect product innovations from imitation (Levin et al. 1987). However, patents constitute a way of both protecting innovations from imitation and ‘conveying public research to industry’ (Cohen et al. 2002). Therefore, the possibility of patenting should induce ﬁrms to engage in collaborations with PROs in order to implement and bring to market novel ideas based on the knowledge developed within universities and public research. However, because it is the outcome of a research process, patenting is one way for ﬁrms to communicate the extent of their engagement in research, to signal their competencies. Czarnitzki and Fier (2003) offer an example of this. They compare the number of patents applied for by publicly funded R&D consortia and by privately funded consortia in Germany. They found that ﬁrms in publicly funded networks are more likely to apply for patents than ﬁrms in private networks. One interpretation that they give for this result is that ﬁrms want to impress government and other actors and gain reputation to inﬂuence future grants or partnerships. Patents are thus being used as a deliberate signalling strategy. We would expect that appropriation and signalling strategy affect the existence and the extent of R&D projects with PROs. Speciﬁcally, the use of patents to protect innovation and signal competencies should have positive effects on participation in collaborative projects with PROs. A dummy variable (PATENT) is employed to capture this effect. Government policies are also likely to positively inﬂuence both the propensity for ﬁrms to develop R&D collaborations with PROs and the intensity of collaboration. The inﬂuence is obviously direct in the case of policies that entail subsidies speciﬁcally targeted to the setting up of projects with PROs. Empirical evidence in support of this relationship has come from both CIS I and CIS II (Arundel et al. 2000). However, it should be noted that government policies could also affect the propensity for ﬁrms to engage in collaborations with PROs in two indirect ways. Matt and Wolff (2003) show that participating in public programmes allows ﬁrms to acquire complementary knowledge and then screen partners by learning about their environment and their technological competence to become part of a broader network. From this viewpoint, participating in public research programmes is also an indicator of the extent of openness of the ﬁrm. More open ﬁrms should be more likely to engage in collaborative agreements with PROs. Participating in public research programmes also constitutes a screening strategy for ﬁrms. Very often these programmes impose limits on disclosure of information about the partners and the research topic. To account for both the direct and the indirect inﬂuence, we have created a dummy variable (SUBSIDIES), which takes the value 1 if a ﬁrm has received public subsidies from regional, national or EU authorities for R&D activities in the three years preceding the questionnaire.

University–industry collaboration 199 9.4.1.3 Firm activity and relationship with PROs Activity in terms of scientiﬁc intensity can inﬂuence both the existence and the extent of a ﬁrm’s relationships with PROs. Firms that invest heavily in R&D are likely to possess a high capacity to absorb the knowledge developed outside the ﬁrm (Cohen and Levinthal 1990). If ‘absorptive capacity’ does play a major role we would expect that the larger the ﬁrm the higher the probability of a relationship with a PRO being established and the greater the number of collaborative R&D projects. R&D intensive ﬁrms might be more likely to develop R&D cooperations with PROs as they are active at the technological frontier and thus are more reliant than other ﬁrms on scientiﬁc developments. To test for this effect, we included a variable for R&D intensity of the ﬁrm (R&DINT), based on the ratio between R&D employment and total employment. We also acknowledged that ﬁrm activity may be inﬂuenced by the ‘legal status’ of the ﬁrm. It is generally accepted that R&D activities tend to be concentrated at the ﬁrm’s headquarters. However, empirical studies by and large have failed to explicitly include this determinant among the independent variables – mainly because of lack of information on the location of the respondent with respect to the company headquarters. Mohnen and Hoareau (2002) found that ﬁrms that collaborate with PROs and are part of large units rely less on this collaboration than do independent ﬁrms. To explain this result, they hypothesized that lower levels of involvement are a consequence of the fact that within big ﬁrms collaboration is mediated by the headquarters. The level of detail provided by the KNOW survey enables us to test this hypothesis by checking whether there is a relationship between the legal status of the ﬁrm and the propensity for engagement in collaborations with PROs. In particular, a dummy variable (HEADQ) is used to take into account whether the respondent is located within the central headquarters of the company. We expect this dummy to positively affect the development of R&D collaboration. 9.4.1.4 Types of innovative activities and processes Typically ﬁrms carry out different types of activities which inﬂuence their opportunity to innovate. They can engage in product innovations, process innovations or both (Klevorick et al. 1995). Although it is very likely that there is a link between the type of innovative activities carried out by ﬁrms and the propensity for and the extent of ﬁrms’ collaboration with PROs, recent investigations provide mixed results concerning the direction and the extent of the relationship. Using data from CIS II for a sample of European countries, Mohnen and Hoareau (2002) found a positive relationship between the introduction of radical product innovations and the extent of reliance on PROs.

200 R. Fontana, A. Geuna and M. Matt In the case of UK ﬁrms included in CIS III, Laursen and Salter (2003) found only partial support for the hypothesis that the more innovative ﬁrms in terms of product innovations are those that rely more on public sources. Using the same UK data, Swann (2002) stressed that companies involved in process innovation are more likely to cooperate with PROs than those engaged in product innovation. In an attempt to shed additional light on both the direction and the extent of the relationship between the type of innovative activity of the ﬁrm and the propensity for ﬁrms to collaborate with universities we decided to include in the regression two dummy variables – one to capture whether the ﬁrm has introduced process innovation (PROCINN) and one focused on product innovation (PRODINN). They are a test for the effects of different innovative processes on the development of collaboration with PROs. In addition to the variables described so far, we decided to include in the regressions two additional dummy variables. First, a dummy variable (COUNTRY), which accounts for country ﬁxed effects, and second, a control dummy (SECTOR) for sector-speciﬁc effects. 9.4.2 Model estimation We model the participation and level of participation in cooperative projects with PROs using a Probit and a Truncated regression model. The Probit model enables estimation of the probability of occurrence of a certain phenomenon in terms of a binary dependent variable. Our speciﬁcation includes the decision to participate in R&D projects with PROs as a discrete dependent variable that takes the value 1 when the ﬁrm has participated in a project and the value 0 when it has not. The Truncated regression model allows the level of participation for the non-limit observations – that is, the number of R&D projects greater than zero – to be estimated. The advantage of a two-equation model is that it separates the analysis of collaboration with PROs or not, from the analysis of multi-collaboration. The Probit model reveals the relevance of the factors considered to the choice to cooperate, while the Truncated model explains the different levels of cooperation. Before analysing the factors that affect the development of cooperative R&D projects with PROs it should be noted that the two separate equation models referred to above may give rise to inconsistent results because of the ‘problem of truncation’ (Greene 1993). This problem arises because the Truncated regression describes the number of R&D projects between ﬁrms and PROs, but the actual number is observable only if ﬁrms decide to engage in the R&D process. The dependent variable in the Truncated equation is therefore incidentally truncated according to the values taken by the Probit equation which acts as a ‘selection equation’. An estimation of the Truncated equation based on observed data only may produce inconsistent estimates because a selection bias is introduced. Checking

University–industry collaboration 201 whether Probit and Truncated regressions can be run separately or must be run simultaneously to avoid giving rise to inconsistent estimates due to selection bias, can be done by applying the Heckman procedure (Heckman 1979). The Heckman procedure is a two-step method involving estimation of a Probit regression of the observations from the sample followed by an ordinary least-squared (OLS) regression on both the observed values for the independent variables and the new values constructed from the previous estimates. To establish whether simultaneous estimation is necessary or whether the two equations can be estimated separately, a signiﬁcant value of the rho is required. In our case, the application of the Heckman procedure produced a ρ (rho) of 0.3098 with a value of 0.2815 for the standard error enabling us to reject the need for the two equations to be estimated simultaneously. Moreover, comparison between the values of the coefﬁcients estimated by Truncated regression and by OLS in the Heckman procedure highlights the absence of substantial differences between the two estimates. The Probit equation was estimated in the following form: y 1 ln R&DINT 2 ln EMPLOYEES 3 PUBLICATIONS 4PATENTS 5HEADQ 6SUBSIDIES 7PROCINN 8PRODINN n

m

i1

j 1

冱iSECTOR 冱jCOUNTRY 1 where y = 1 if number of projects is >0 and y = 0 if number of projects is 0. The second equation is an estimated Truncated regression. 1n(1P) 1 1n R&DINT 2 1n R&D 3PUBLICATIONS 4PATENTS 5HEADQ 6SUBSIDIES 7PROCINN 8PRODIN n

m

i1

j 1

冱iSECTOR 冱jCOUNTRY 21 where P is observed when number of projects is >0. Table 9.3 presents the results of the parsimonious estimations.14 All the chosen independent variables, excluding PRODINN, have a positive effect on the propensity for ﬁrms to engage in R&D projects with PROs and all the coefﬁcients of the variables are signiﬁcantly different from zero. There is a ‘size effect’ on the propensity for ﬁrms to engage in projects with PROs as represented by the positive coefﬁcients for EMPLOYEES, our proxy for ﬁrm size. R&D employment does not signiﬁcantly affect the propensity to be involved with PROs in R&D projects and therefore it was not included in the ﬁnal estimation model. R&DINT positively affects the propensity for ﬁrms to engage in R&D projects with PROs. These results suggest that larger ﬁrms that are heavily engaged in

1.201** (2.23) 0.386* (1.84) 0.028 (0.08) 2.549** (4.96) No No 205.503 72.61** 0.169 357

0.625** (3.95) 0.785** (4.05) 0.384** (2.40) 0.555** (3.66) 0.178** (2.82)

Notes * Indicates signiﬁcant at 10% conﬁdence interval. ** Indicates signiﬁcant at least at 5% conﬁdence interval. t-value in brackets.

SUBSIDIES PUBLICATIONS PATENT HEADQ ln EMPLOYEES ln R&D ln R&DINT PROCINN PRODINN Intercept Sector dummies Country dummies Log likelihood Wald Chi-square Pseudo R-square No Obs 1.327** (2.27) 0.226 (0.29) 0.051 (0.14) 2.023** (3.06) Yes Yes 193.248 77.87** 0.218 357

0.634** (3.58) 0.859** (4.07) 0.388** (2.22) 0.494** (2.77) 0.162** (2.44)

184

0.124** (2.86) 0.198 (0.72) 0.149 (1.19) 0.250** (2.47) 0.230 (0.69) Yes Yes 125.883 66.88**

0.124** (3.18) 0.468* (1.79) 0.131 (1.11) 0.178 (1.57) 0.508** (2.52) No No 129.982 43.89** 184

0.150* (1.79) 0.035 (025) 0.191** (2.00) 0.075 (0.88)

0.143* (1.78) 0.037 (0.24) 0.126 (1.41) 0.104 (1.25)

Coefﬁcient

Coefﬁcient

■

Coefﬁcient

Coefﬁcient

Truncated (2) Number of projects with universities

Probit (1) Discrete variable

Table 9.3 Regression summary – Probit and Truncated regressions

University–industry collaboration 203 R&D activities (high R&D intensity) have a greater propensity than small ﬁrms to become involved in projects with PROs. However, the R&D size of the ﬁrm, proxied by R&D employment, does not affect this tendency.15 There are differences in both the inﬂuence and the signiﬁcance of the independent variables present in the Truncated regression when compared to the Probit regression. The main difference between the two regressions is that in the Truncated regression, R&D employment exhibits a positive and signiﬁcant coefﬁcient while EMPLOYEES, which is a proxy for the ‘absolute size’ of the ﬁrm, does not affect the level of participation. Other things being equal we can argue that while there is indeed an absolute size effect determining the propensity for a ﬁrm to engage in R&D projects with PROs, there is a signiﬁcant relative size effect, as captured by R&D employment, in explaining the extent of participation in the projects. R&D intensity, which is a proxy for the position of the ﬁrm with respect to the technological frontier rather than being a proxy for ﬁrm size, is still a signiﬁcant explanatory variable, though with a lower probability than in the Probit regression. In terms of the effect of the other independent variables, the positive inﬂuence of subsidies as an incentive to engage in R&D activities is conﬁrmed to be as signiﬁcant in determining the level of collaboration as it was for determining the propensity to collaborate. PUBLICATIONS, PATENTS and HEADQ variables change in signiﬁcance between the Probit and the Truncated regression. Other effects being equal, searching in scientiﬁc or business journals for ideas has a positive impact on the propensity to engage in R&D projects with PROs, while it is not signiﬁcant in explaining the level of participation in R&D projects. Similarly, patenting, which had a positive and signiﬁcant effect in the Probit estimation, is no longer signiﬁcant in the Truncated regression. Respondents located in the headquarters of a ﬁrm are more likely to develop R&D collaborations with PROs than other respondents, but this characteristic does not affect the level of cooperation. Finally, process innovation does not signiﬁcantly affect the extent of the collaboration. The introduction of sector ﬁxed effects does not change the signiﬁcance of the Probit estimation except with regard to propensity to engage in projects with PROs, which is not affected by the ﬁrm being a process innovator once country dummies are included. Instead, sector dummies affect the estimates of the Truncated model. When sector dummies are included, R&D intensity is no longer signiﬁcant. We can argue that, since sectors differ in terms of R&D intensity, the presence of these differences affects the level of cooperation with PROs. The inclusion of sector dummies makes patents signiﬁcant and positive. This result can be interpreted as capturing the effect of signalling and, thus, ﬁrm openness, on the extent of collaboration rather than appropriation because sectoral dummies account, at least partially, for appropriability regimes. When country dummies are introduced, the relevance of both R&D intensity

204 R. Fontana, A. Geuna and M. Matt and the dummy variables for subsidies is affected. Country speciﬁc factors related to both the scientiﬁc proﬁle of innovating ﬁrms and their reliance on subsidies inﬂuence the level of interaction between ﬁrms and PROs. Some of our ﬁndings can be summarized as follows. The propensity for ﬁrms to engage in collaborations with PROs is positively affected by their size and openness. We deﬁne openness as the attitude of ﬁrms towards establishing a relationship with PROs. Large ﬁrms with a high absorptive capacity are more likely to engage in R&D cooperation with the academic world. However, absorptive capacity loses its signiﬁcance if the ﬁrm does not proactively screen the scientiﬁc and technological environment in which it works. In other words, the mechanisms through which ﬁrms can import knowledge from outside their boundaries are important explanatory variables of R&D cooperation. As mentioned above, there are different enablers of these mechanisms. Seeking information in scientiﬁc and business journals and also participation in government-funded projects are two proactive means used to relate to the socio-economic environment. Patents and public programmes can signal in which domains the ﬁrm has competencies; this is especially so in the case of SMEs for whom secrecy is the usual way to approach appropriability and thus patents could be interpreted as a proxy for signalling. These three variables positively affect the propensity for ﬁrms to collaborate with PROs. In other words, larger ﬁrms with higher learning abilities and proactive screening and signalling strategies are the most likely partners for universities. Openness affects the level of cooperation to a lesser extent.

9.5 Conclusions The focus of the KNOW questionnaire on SMEs has produced a unique dataset for the researcher to use for analysis of the impact of internal and external knowledge sources upon the innovative process of SMEs. This chapter has analysed the contribution of PROs to the innovative process of SMEs and has examined the ﬁrm speciﬁc, sector speciﬁc and country speciﬁc factors that explain the existence and extent of cooperative R&D projects between SMEs and PROs. The descriptive results provide direct evidence that PROs were among the less important sources of information for both innovative ideas and innovation completion in relation to the most important innovations developed during the three years prior to the survey for SMEs in the food and beverages, chemicals, communications equipment, telecommunications services and computer services sectors in seven EU countries. Surprisingly, the contribution of PROs to the phase of completion of the innovation (time period during which an innovation is being developed up to ﬁnalization) is similar to the innovative ideas phase. If the most important external source of information is considered, for certain sectors PROs are contributing more to the completion phase than to the inno-

University–industry collaboration 205 vative ideas phase. Overall, signiﬁcant sectoral differences were found. For example, respondents from the food and beverages sector assign particular importance to government regulation as a driver of relationships with PROs. The interviews conﬁrmed the results of the descriptive analysis pointing to the fact that most ﬁrms do not have the resources to spare (they are not big enough) to develop relationships with PROs although a few ﬁrms have periods of intensive interaction with PROs to meet speciﬁc needs. Although PROs do not play a central role in the innovative process of SMEs, about half of the ﬁrms in the sample had developed R&D cooperative projects with PROs. The econometric model developed allows estimation of the impact of ﬁrm-speciﬁc factors, controlling for sector and country ﬁxed effects, upon both the probability of developing an R&D project with PROs and the number of R&D projects developed by the ﬁrm in the previous three years. The results of this analysis point to two major phenomena. The ﬁrst concerns the relationship between the probability of R&D collaboration with PROs and ﬁrm size. Our results suggest that the probability depends on the absolute size of the ﬁrm. Larger ﬁrms have a much higher probability of R&D collaboration. This result is consistent with many previous empirical investigations of determinants of university– industry relationships (Arundel et al. 2000; Mohnen and Hoareau 2002; Cohen et al. 2002; Laursen and Salter 2003). However, the number of R&D cooperations is not affected by the absolute size of the ﬁrm but rather by the relative size as measured by R&D employment. This aspect is not highlighted in previous contributions. R&D intensity affects both the propensity for and the extent of engagement in R&D projects. The second phenomenon concerns the openness of ﬁrms, that is, their willingness to search, signal and screen the outside world by searching publications databases, by patenting and by participating in government subsidized programmes. Our ﬁndings suggest that the reliance on publications for acquiring knowledge affects the probability of collaboration with a PRO, but not the level of collaboration developed. Rather, ﬁrms that patent to protect innovation (and signal competencies) have a higher probability of collaborating and a higher level of collaboration. In addition, the results of the estimation suggest that those ﬁrms that have received public subsidies have both a higher probability of developing R&D cooperation with PROs and a higher number of collaborations, although the impact of subsidies on the extent of the collaboration is mediated by country speciﬁc effects. This means that screening is an important precondition for the development of relationships between SMEs and PROs, but that other factors should be taken into account when the focus is on the extent of the relationships. In both equations sectoral and country-ﬁxed effects are signiﬁcant and important. Overall, the results of this analysis support the view that relationships between ﬁrms and PROs are characterized by a high degree of

206 R. Fontana, A. Geuna and M. Matt heterogeneity. To consider university–industry relationships in a general way and develop policies on the basis of such generalization will lead to unintended inter-sectoral differences and the various actors will react to these policies in different ways depending on their speciﬁc characteristics. Furthermore, it is extremely important to take into account that policies in support of collaboration between PROs and ﬁrms should create incentives for both sets of actors to cooperate. Current policies are mainly directed to forcing PROs into these types of relationships with no acknowledgement that without appropriate ‘demand’ little will be achieved. This chapter provides strong evidence that, after controlling for ﬁrm size and other factors, the openness of ﬁrms to the external environment (and therefore their willingness to interact with it) is very important in explaining the probability of their collaborating with PROs. Without willingness on both sides satisfaction will not be achieved.

Appendix Table A9.1 Descriptive statistics for selected variables (all variables) Variable

Obs

Mean

Std Dev.

Min.

Max.

Number of projects R&D R&DINT EMPLOYEES PUBLICATIONS PATENTS HEADQ PROCINN PRODINN SUBSIDIES

458 491 485 546 552 551 554 543 553 492

1.62 13.53 0.15 194.82

2.84 32.52 0.23 261.52

0 0 0 2 0:99 0:354 0:241 0:95 0:22 0:341

25 300 1 1,200 1:453 1:197 1:313 1:448 1:531 1:151

Notes 1 The authors would like to thank Anthony Arundel, Gustavo Crespi, Lionel Nesta, W. Edward Steinmueller and one anonymous referee for their comments and suggestions. We would also like to thank Joseph Heili, Mohammed Dif, Isabelle Terraz and Lucio Aparicio for their contribution during the KNOW for INNOVATION project. Financial support from the Commission for the European Communities, TSER Programme, project KNOW is acknowledged. 2 See Arundel et al. (1995) and Arundel and Geuna (2004) for an analysis based on the PACE data. See, among others, Mohnen and Hoareau (2002) for an analysis based on CIS II. 3 See, among others, Faulkner and Senker (1995) for a qualitative technologyspeciﬁc study. Geuna et al. (2004), among others, for a university speciﬁc case (University Louis Pasteur of Strasbourg). 4 See, for instance, Link (1996), Hall et al. (2000), Cohen et al. (1998), Caloghirou et al. (2001).

University–industry collaboration 207 5 It is important to note that the unweighted results of the PACE survey show that publications are the most important method for learning about public research output. 6 They use 15 different external sources of information to construct the openness variable. The more that ﬁrms use different external sources, the more open they are. 7 The countries are: Denmark, France, Germany, Greece, Italy, the Netherlands and the UK. 8 See Chapter 5, this volume, (2002) for a description of the KNOW survey’s methodology and main results. 9 The exception is the case of food and beverages for which competitors rank second in the case of innovation ideas. In the case of innovation completion suppliers are identiﬁed as the most important contributors to innovation by communications equipment ﬁrms. 10 The minimum selection criteria were to cover the ﬁve sectors and in each sector to choose one large and one small company. The main questions tackled during the interviews concerned the competition strategy of the ﬁrm, their cooperative research behaviour, their patenting behaviour and the speciﬁc innovation detailed in the survey. 11 The highest number of research and development projects with PROs reported is 25. Two respondents answered 80 and two responded 100. They were excluded from the analysis because we considered their answer was either incorrect or that the numbers included informal contacts. 12 In Table 9.2 the following codiﬁcation is employed. 0 = zero contracts; 1 = maximum of 1 contract; 2 = maximum of 2 contracts; 3 = more than 2 contracts. 13 Whether a higher propensity for big ﬁrms to collaborate with PROs corresponds to a better capability to exploit the beneﬁts deriving from the collaboration is controversial. Link and Rees (1990) and Acs et al. (1994) argue that big ﬁrms have lower R&D productivity than small ﬁrms and are therefore less efﬁcient at exploiting beneﬁts deriving from interactions with PROs. Cohen and Klepper (1996) argue instead that the lower productivity of big ﬁrms is not related to R&D efﬁciency linked to ﬁrm size but is rather the consequence of the presence of high ﬁxed costs. 14 All the estimations have been corrected for heteroschedacity with the STATA robust estimation procedure. 15 Several attempts to include other variables in the list of independent variables have been made. In particular we checked for the inﬂuence strategic alliances might have on the propensity for ﬁrms to engage in projects with PROs. Developing external formal R&D collaborations and partnerships with other ﬁrms is one of the possible strategies followed by ﬁrms to establish collaborative relationships. Firms involved in strategic alliances may also have a higher propensity for participation in R&D cooperative projects with PROs. One of the possible reasons for this is that once they have developed the skills needed to manage cross-boundary relationships, ﬁrms become more willing to cooperate with external partners in the development of a core strategic activity. To analyse the possibility that ﬁrms involved in strategic alliances are more likely to participate in R&D cooperative projects with PROs we introduced in the regression a dummy variable (RJV), which takes the value of 1 when the ﬁrm is involved in a research joint venture and 0 when it is not. While the effect of this variable on propensity is generally positive, the coefﬁcient of the variable was not signiﬁcant.

208 R. Fontana, A. Geuna and M. Matt

References Acs, Z.J., Audretsch, D.B. and Feldman, M.P. (1994) ‘R&D spillovers and recipient ﬁrm size’, Review of Economics and Statistics, 76: 336–340. Arundel, A. and Bordoy, C. (2002) ‘In-house versus ex-house: the sourcing of knowledge for innovation’, in J. de la Mothe and A.N. Link (eds) Networks, Alliances and Partnerships in the Innovation Process, Boston, MA: Kluwer Academic, pp. 67–87. Arundel, A. and Geuna, A. (2004) ‘Proximity and the use of public science by innovative European ﬁrms’, Economics of Innovation and New Technology, 13: 559–580. Arundel, A., Cobbenhagen, J. and Schall, N. (2000) The Acquisition and Protection of Competencies by Enterprises, Final Report for EIMS Project 98/180, Maastricht: MERIT. Arundel, A., van de Paal, G. and Soete, L. (1995) Innovation Strategies of Europe’s Largest Firms: Results of the PACE Survey, European Innovation Monitoring System, Report No. 23, Brussels: European Commission. Beise, M. and Stahl, H. (1999) ‘Public research and industrial innovation in Germany’, Research Policy, 28: 397–422. Caloghirou, Y., Tsakanikas, A. and Vonortas, N.S. (2001) ‘University–industry cooperation in the context of the European Framework Programmes’, Journal of Technology Transfer, 26(1–2): 153–161. Cohen, W.M. and Klepper, S. (1996) ‘A reprise of size and R&D’, The Economic Journal, 106: 925–951. Cohen, W.M. and Levinthal, D.A. (1990) ‘Absorptive capacity: a new perspective of learning and innovation’, Administrative Science Quarterly, 35: 128–152. Cohen, W.M., Florida, R., Randazzese, L. and Walsh, J. (1998) ‘Industry and the academy: uneasy partners in the cause of technological advance’, in R. Noll (ed.) Challenges to the University, Washington, DC: Brookings Institution Press, pp. 171–199. Cohen, W.M., Nelson, R.R. and Walsh, J. (2002) ‘Links and impacts: the inﬂuence of public research on industrial R&D’, Management Science, 48: 1–23. Czarnitzki, D. and Fier, A. (2003) ‘Publicly funded R&D collaborations and patent outcome in Germany’, paper presented at the 3rd European Meeting on Applied Evolutionary Economics, Augsburg, Germany, 10–12 April. Faulkner, W. and Senker, J. (1995) Knowledge Frontiers, Oxford: Oxford University Press. Geuna, A., Llerena, P., Matt, M. and Savona, M. (2004) ‘Collaboration between a research university and ﬁrms and other institutions’, in F. Cesaroni, A. Gambardella and Walter A. Garcia-Fontes (eds) (2004) R&D, Innovation and Competitiveness in the European Chemical Industry, Dordrecht: Kluwer Academic Publishers. Greene, W.H. (1993) Econometric Analysis, Englewood Cliffs: Prentice Hall. Hall, B.H., Link, A.N. and Scott, J.T. (2000) ‘Universities as research partners’, NBER Working Papers No. 7643, Cambridge, MA: NBER. Heckman, J. (1979) ‘Sample selection bias as a speciﬁcation error’, Econometrica, 47: 153–161. Klevorick, A., Levin, R., Nelson, R.R. and Winter, S.G. (1995) ‘On the sources and signiﬁcance of inter-industry differences in technological opportunities’, Research Policy, 24: 195–205.

University–industry collaboration 209 Laursen, K. and Salter, A. (2003) ‘Searching low and high: why do ﬁrms use universities as a source of innovation?’, paper presented at the 3rd European Meeting on Applied Evolutionary Economics, Augsburg, Germany 10–12 April. Levin, R., Klevorick, A., Nelson, R.R. and Winter, S.G. (1987) ‘Appropriating the returns from industrial R&D’, Brookings Papers on Economic Activity, 2: 783–831. Link, A.L. and Rees, J. (1990) ‘Firm size, university based research and the returns to R&D’, Small Business Economics, 2: 25–31. Link, A.N. (1996) ‘Research joint ventures: evidence from federal register ﬁlings’, Review of Industrial Organization, 11: 617–628. Mansﬁeld, E. (1991) ‘Academic research and industrial innovation’, Research Policy, 26: 1–12. Matt, M. and Wolff, S. (2003) ‘ “EU sponsored” versus “spontaneous” R&D collaborations: towards a micro-analysis’, paper presented at the conference Evaluation of Government Funded R&D Activities, Vienna, 15–16 May. Meyer-Krahmer, F. and Schmoch, U. (1998) ‘Science-based technologies: university–industry interactions in four ﬁelds’, Research Policy, 27: 835–852. Mohnen, P. and Hoareau, C. (2002) ‘What type of enterprise forges close with universities and government labs? Evidence from CIS 2’, MERIT-Infonomics Research Memorandum Series, August. Narin, F., Kimberly, K.S. and Olivastro, D. (1997) ‘The linkages between US technology and public science’, Research Policy, 26: 317–330. Schartinger, D., Schibany, A. and Gassler, H. (2001) ‘Interactive relations between universities and ﬁrms: empirical evidence for Austria’, Journal of Technology Transfer, 26: 255–269. Swann, G.M.P. (2002) Innovative Business and the Science and Technology Base: An Analysis using CIS 3 Data, report prepared for the Department of Trade and Industry, October.

10 Self-selection and learning in European research joint ventures A micro-econometric analysis of participation and patenting1 Lucia Cusmano 10.1 Introduction This chapter presents an empirical evaluation of the relationship between participation in European research joint ventures (RJVs) and patenting activity, in two high tech ﬁelds that are receiving increasing attention and funding from European institutions: information and communication technology (ICT) and medical and biotechnology (MB). In the ICT area, the remarkable performance in patenting of RJV members compared to non-participating ﬁrms can be mostly explained by self selection of innovative actors. This is in line with the main aim of ICT programmes, that is, attracting the industry big players and providing a common technological basis for IT application. On the other hand, there is clear and robust evidence of a positive effect of RJV afﬁliation in the emerging MB ﬁeld, where European consortia appear to have attracted ﬁrms with a high potential for innovation. The ﬁndings from this evaluation suggest that the analysis of industry dynamics should be central in the design of policy and evaluation of policy targets and achievements. Cooperative technological activities have been extensively investigated in the theoretical literature within the broader framework of technology policy analysis (for example, Stoneman and Vickers 1988; Geroski 1995; Llerena and Matt 1999; Cassiman 2000), game theoretical models (for example, Spence 1984; Katz, 1986; d’Aspremont and Jacquemin 1988; Sinha and Cusumano 1991; Kamien et al. 1992; Morasch 2000) and strategic management studies (for example, Teece 1986; Contractor and Lorange 1988; Hagedoorn 1993; Hagedoorn and Narula 1996). Despite the increasing attention being paid to the rationale for and beneﬁts and costs of R&D agreements, there has been little empirical work on their efﬁcacy. Empirical investigations have been conducted mostly at the level of case studies (for example, Lazaric and Marengo 2000) or one-time surveys of participants (for example, Sakakibara 1997), while the emphasis of technology policy on extensive collaborative programmes has not been matched by systematic investigations on large samples. In particular, there have been few empirical studies that have

Self-selection and learning in European RJV 211 attempted to estimate the effects of policy-supported cooperative R&D on ﬁrms’ innovative potential, as measured by effective patenting output.2 The gap between theoretical support and empirical assessment is all the more critical at European level, bearing in mind the central role assigned to cooperative R&D in European Research and Technological Development (RTD) programmes. The study described in this chapter aims to assess the relationship between participation in European RJVs and patenting activity, by elaborating a large panel of data on ﬁrms that have taken part in European R&D consortia in two high tech areas that are critical to the development of a competitive European Innovation System and, accordingly, have received increasing attention and ﬁnancial resources from EU institutions: that is, ICT and MB. In this chapter we examine the patenting performance of RJV participants compared to the performance of other European ﬁrms, and explore the most relevant differences between the two macro areas. The presentation of econometric results is linked to the discussion of speciﬁc methodological issues, which are extremely relevant when analysing economic count data characterized by a strongly skewed distribution, as is the case for patent applications from European industry. The work is organized as follows. Section 10.2 brieﬂy describes the Framework Programmes (FWPs) and the EUREKA initiative, commenting on the trends and composition of European ﬁnancial support for cooperative research. Section 10.3 introduces the dataset on which the empirical analysis is based and describes the sampling methodology. The descriptive analysis of RJV members and non-participants provides some ﬁrst insights, and highlights the need for the issue of self-selection by innovative RJV members to be addressed. Section 10.4 statistically evaluates the relevance of self-selection and proposes control variables to distinguish between self-selection and learning effects. Section 10.5 describes the application of non-linear count data models for assessing the effect of RJV afﬁliation on patenting behaviour. The ﬁnal section, Section 10.6, comments on the main results and discusses avenues for further empirical research.

10.2 European cooperative policies European technology policy assigns a central role to cooperative activities, which have been actively supported through a variety of programmes and, currently, account for the majority of R&D funding at European level. Although the Single European Act (1987) provided the ﬁrst ofﬁcial basis for the implementation of science and technology policy in the EU through the ﬁrst Framework Programme, the European Union (EU) had been supporting R&D consortia since the early 1980s, mainly in response to the declining competitiveness of European industry vis à vis US and Japanese companies, and the weakness of national champion policies.

212 L. Cusmano The ﬁrst genuinely collaborative and EU-wide initiatives concerned the electronics sector and led, in 1984, to the creation of ESPRIT – the European Strategic Programme for Information Technology. ESPRIT served as a model for subsequent programmes, that went beyond the ICT area and were implemented within the Framework Programmes from 1984; thus it became the EU’s medium-term planning instrument for RTD. ICT has emerged over subsequent FWPs as the main broad area of research investment at European level, replacing the Energy sector at the top of EU RTD priorities (Table 10.1).3 In relative terms, however, the focus of EU ﬁnancing since the early 1990s has gradually shifted to Industrial and Material Technologies, and Life Sciences and Technologies, the latter including biomedical programmes.4 ICT and MB represent the most relevant research areas within EUREKA – another Europe-wide programme, at the heart of the 1980s upsurge in technological collaboration among European ﬁrms. Unlike FWPs, EUREKA is a non-EU initiative, that is, it is transnational and Europe-based, but exists outside of the formal institutional borders of the EU.5 Initially designed to support industry-led grand projets, EUREKA has evolved into an umbrella mechanism for mostly small market-oriented R&D projects, which include some pre-competitive activities and also involve non-European countries. Over time, the participation of small and medium-sized enterprises (SMEs) has increased and the spread of projects across technology sectors has broadened, although ICT and MB have remained prominent (Table 10.2). Following different approaches – top-down or bottom-up – and being subject to different political constraints, FWPs and EUREKA have become the main instruments for the Commission and national governments to give selective support to companies seeking to undertake collaborative R&D with ﬁrms or research institutes in other European countries. The sections in this chapter focus on RJVs promoted under these programmes in the two largest and fast growing research ﬁelds, ICT and MB, in order to explore the relationship between ﬁrms’ participation and innovative performance.

10.3 Data and descriptive analysis The investigation is performed on a panel of data on European businesses over the period 1992–96, comprising both RJV participants and a control group of non-participants. The RJV_EPO dataset represents the main source of information concerning RJV participants in ICT and MB consortia. This dataset combines information about RJVs promoted under FWPs and EUREKA with details about ﬁrms’ patenting activity retrieved from the European Patent Ofﬁce (EPO).6 The analysis refers to participants in the third and fourth FWPs for which sector and employment information is available.7 This is a less strict requirement than the selection rule based

100

Total

Note * Adjusted to 1992 prices.

Source: European Commission (1994).

25 11 7 5 50 – – – – 2

Budget share (%)

3,750

I 1984–87

Information and communication technologies Industrial and materials technologies Environment Life science and technologies Energy Transport Socio-economic research International cooperation Dissemination and exploitation of results Human capital and mobility

Total million ECUs*

Framework Programme Years

Table 10.1 RTD priorities in Framework Programmes

100

42 16 6 7 22 – – 2 1 4

5,396

II 1987–91

100

38 15 9 10 16 – – 2 1 9

6,600

III 1990–94

100

28 16 9 13 18 2 1 4 3 6

12,300

IV 1994–98

214 L. Cusmano Table 10.2 EUREKA projects, by technological area (number and value, 1989–97) Area

Information technology Communications Robotics Environment Transport Medical and biotechnology Energy Lasers New materials Total

Number 1989

1997

50 19 70 32 21 55 14 13 23 297

■

Value* (MECU) 1989

1997

119 22 96 138 47 127 29 14 76

1,512 1,194 1,078 607 591 542 526 271 153

3,296 665 434 492 522 443 125 64 76

668

6,474

6,223

Sources: Peterson (1993), Peterson and Sharp (1998). Note * Projects costs including public and private investments.

on availability of ﬁrm-level R&D data, which is frequently the norm in the empirical literature investigating patenting behaviour. Such criteria would result in a rather small sample of RJV participants. In this study, bearing in mind the advantage of full availability of patent data and the aim of achieving a more comprehensive overview of the impact of RJV working, we prefer to use less strict criteria so as to be able to include in the investigation a wider range of participants. However, this implies the use of variables other than R&D measures, whose identiﬁcation and construction will be extensively discussed in Section 10.4. A further selection criterion concerns industry classiﬁcation. The RJV samples comprise ﬁrms from the main industries that are involved cooperating in the ICT and MB sectors, that is, we have dropped from the dataset ‘outlier sectors’, which only occasionally join European RJVs in the two areas of interest. The resulting RJV samples include 551 ﬁrms in the ICT area and 164 ﬁrms in the MB area. Since the evaluation of the impact of RJV membership on ﬁrms’ patenting activity requires a control, we constructed a control group (CG) of ﬁrms that have not participated in a European RJV. The CG was extracted from AMADEUS by way of a stratiﬁed sampling methodology, in order to mimic the RJV sample in terms of both sector and country composition. In other words, for each industry, the share of ﬁrms included in the CG is close to the share included in the RJV sample, and within each industry, the country composition of the CG reﬂects the country composition of RJV participants.8 The resulting CG samples are composed of 491 ﬁrms from the IT area and 198 ﬁrms from the MB area. Tables 10.3 and 10.4 report the industry composition of the estimated samples.

Self-selection and learning in European RJV 215 For both technological areas, the descriptive statistics (see Table 10.5) point to a high degree of heterogeneity between RJV participants and non-participants: ﬁrms that enter European R&D consortia are, on average, bigger and more innovative than non-participants. This relevant heterogeneity represents a cause of concern for the interpretation of results when estimating the effect of RJV participation on patenting activity. In fact, self selection of innovative RJV entrants may be an important source of bias in the estimation. If there is evidence that neither the ﬁrms joining consortia nor those not participating can be considered random draws, constructing and interpreting an econometric assessment relating participation to innovative activities requires caution. To what extent does the observed higher innovativeness of RJV participants reﬂect a greater propensity for innovative actors to join? Does participation widen the ex

Table 10.3 ICT sample: industry composition Dummy variable

Industry (US SIC Code)

MED_INS COMP COMP_SER E_EQUIP COMM COMM_SER P_EQUIP TRADE

Medical Instruments (384) Ofﬁce, Computers and Accounting Equipment (357) Computer Services (737) Electric Equipment and Supplies (36, exc. 365, 366, 367) Communication Equipment (365, 366, 367) Communication (48) Professional and Scientiﬁc Equipment (38, exc. 384) Wholesale Trade Durable and Nondurable Goods (504, 506, 508, 514) Engineering and Management Services (871, 873)

ENGMAN

Total

Frequency 23 111 255 56 130 71 73 200 123 1,042

Table 10.4 MB sample: industry composition Dummy variable

Industry (US SIC Code)

Frequency

FOOD CHEM DRUGS MED_INS P_EQUIP

Food and Kindred Products (20) Chemicals (excluding Drugs) (28, excluding 283) Drugs (283) Medical Instruments (384) Professional and Scientiﬁc Equipment (38, excluding 384) Wholesale Trade Durable and Nondurable Goods (504, 506, 508, 514) Engineering and Management Services (871, 873)

89 39 115 19

Total

362

TRADE ENGMAN

20 46 34

ICT Total (N = 1042) RJV (N = 551) CG (N = 491) MB Total (N = 362) RJV (N = 164) CG (N = 198)

Sample

5,315.190 7,083.830 1,314.954 3,028.980 4,007.690 1,671.650

1,166.380 1,858.250 593.315

Std Dev.

1,485.200 2,392.680 466.812

Mean

Average employment level (1992–96)

Table 10.5 Descriptive statistics: employment and patenting

7.2 7.2 10.0

1.7 1.7 4.8

Min.

26,915.0 26,915.0 18,051.0

74,098.0 74,098.0 23,974.4

Max.

■

0.558 1.091 0.112

0.736 1.041 0.395

Mean

4.498 6.631 0.506

4.775 5.653 3.514

Std Dev.

0 0 0

0 0 0

Min.

80.8 80.8 4.0

85.4 85.4 69.0

Max.

Patent applications (yearly average, 1992–96)

Self-selection and learning in European RJV 217 ante innovative gap? Is it possible to identify and distinguish the learning effects of cooperation from the level of innovativeness prior to afﬁliation? We suggest some statistical tests for assessing the relevance of self-selection and discuss the use of variables that should allow control for the relevant heterogeneity of RJV members and non-participants.

10.4 Self selection and learning effects As commented on by Klette et al. (2000) with respect to studies on the impact of subsidies on commercial R&D, in which systematic differences between supported and non-supported ﬁrms are often underlined, group differences in control group analysis do not make the evaluation results less interesting, but do limit the kind of questions the study can answer. Therefore, assessing whether self selection is relevant appears as a necessary ﬁrst step in the evaluation of European consortia contribution to innovativeness. However, the issue of self selection is not explicitly addressed by the scant econometric literature on R&D cooperation. For the purposes of such investigation, our analysis follows the approach of very recent empirical work in the area of international trade, which deals with evidence of a positive correlation between ﬁrm productivity and exporting activity. Bernard and Jensen (1995), Clerides et al. (1998) and Aw et al. (2000) address the complex issue of whether exports play a causal role in generating higher productivity since exporters gain knowledge and expertise in the export market (learning hypothesis), or whether the correlation simply reﬂects the decision to export by the most productive ﬁrms (selfselection hypothesis). These authors underline that both mechanisms are plausible, though their actual importance most likely varies across countries and industries, and they propose several econometric tests for evaluating their relative importance. We here adapt and implement a few tests developed in this recent literature, with particular attention to Aw et al.’s (2000) analysis. First, we further characterize the innovativeness differential between RJV participants and non-participants in terms of patent stock. Figure 10.1 depicts, for the period 1986 to 1996, the average number of patent applications over the whole EPO horizon (1978–96), from ﬁrms that were part of at least one RJV at time t. Except for the ﬁrst years of the MB programmes, the average number of participants’ patent applications is a great deal higher than the average patent stock for the CG. In both the ICT and MB areas, there is clear evidence that early participants were more innovative on average than later entrants. In fact, the average level of innovativeness of ﬁrms involved in European projects decreases over time, as if those programmes had attracted the most innovative entities in the early stages and later extended to include less innovative ﬁrms. The decline is more evident in the ICT area, whereas in the MB area the very

218 L. Cusmano early pioneers were not very innovative, but entry in the RJV group in 1996 of highly innovative ﬁrms has reversed the negative trend. We can achieve greater insight by considering the average level of innovativeness of RJV members before their ﬁrst RJV participation, rather than up to 1996. Figure 10.2 depicts the average number of patent applications up to time t 1 from ﬁrms that joined a European RJV for the ﬁrst time at time t. The cumulated innovative experience of new entrants is very heterogeneous over time. In the ICT area we can identify a few peaks, which roughly correspond to the starting years of new waves of projects, in particular 1986 for EUREKA and 1992 for the third FWP, which directed a conspicuous share of the budget to cooperative R&D in ICT. The last wave of MB programmes has attracted highly innovative new members, whereas, except for the peak of innovativeness in 1989, early RJV participants had not innovated signiﬁcantly prior to their ﬁrst RJV experience.

Average number of patents

ICT sample 45 40 35 30 25 20 15 10 5 0

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1993

1994

1995

1996

MB sample Average number of patents

25 20 15 10 5 0

1986

1987

1988

1989

1990

RJVs participants

1991

1992

Control group

Figure 10.1 RJV participants’ average level of innovativeness. Patent stock (1978–96) of RJV participants by year t.

Average number of patents in t1

Average number of patents in t1

Self-selection and learning in European RJV 219 ICT sample 12 10 8 6 4 2 0

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1992

1993

1994

1995

1996

MB sample 20 18 16 14 12 10 8 6 4 2 0

1986

1987

1988

1989

1990

1991

Figure 10.2 RJV entrants’ innovative experience. Patent stock at time t 1 of new RJV participants at time t.

From this ﬁrst-level descriptive analysis, we may expect the self-selection bias to be more signiﬁcant in the ICT than in the MB area, but there also appears to be some evidence that, even in ICT, the innovation gap widened after entry. In order to evaluate in statistical terms the relationship between the level of patenting activity and RJV participation and, in particular, the relevance of self-selection, we fully exploit the time series dimension of the dataset and group ﬁrms into ‘transition groups’ (Table 10.6). At each time t, we compare the status of ﬁrms, taking into account their ‘backward’ and ‘forward’ transition patterns. The control group represents the benchmark, that is, innovativeness (number of patent applications per year) in the other groups is deﬁned in terms of differential in innovativeness of non-members. Therefore, we distinguish ﬁrms that never entered consortia (CG) from ﬁrms that participated in a project for the last time before t 1 (EARLY), or that are about to enter a consortium in a later time (LATE). As the year of ﬁrst afﬁliation approaches, ﬁrms are ﬂagged by the dummy variables ENTRY (entry in

220 L. Cusmano Table 10.6 Transition groups: RJV afﬁliation over time

Control group CG Earlier participants EARLY EXIT Entrants LATE ENTRY NEW Frequent participants INCUMB NAP REST AWAKE

Earlier

t1

t

t+1

Later

no

0

0

0

no

yes yes/no

0 1

0 0

0 0

no no

no no no

0 0 0

0 0 1

0 1 1/0

yes yes/no yes/no

yes/no yes/no yes/no yes yes yes

1 1 1 0 0 0

1 0 0 0 0 1

1/0 0 1 1 0 1/0

yes/no yes yes/no yes/no yes yes/no

Note RJV membership = 1, non-membership = 0.

t + 1) and NEW (entry in t). Active participants are identiﬁed differently, depending on whether they have been taking part in a project for at least two years up to time t (INCUMB), or are not currently part of a consortium, but have been earlier and will be in the future involved in cooperative activities. These ‘sleeping partners’ are ﬂagged by the dummy variables NAP if their previous afﬁliation only recently terminated, and REST if their most recent cooperative project was completed before t 1, that is, it is at least two years since participation in a cooperative project. AWAKE refers to ﬁrms that enter a consortium at time t after a sleeping break. Finally, EXIT describes ﬁrms whose last participation to European RJVs recently terminated. If self selection is at work, we expect the relative innovative level in period t to be positively correlated with future RJV participation, that is, with an entrant status. Therefore, we ran a regression of ﬁrms’ innovative differential, with respect to the average CG level [PAT_DIFFit = Patit Mean (PatCGit)], on the variables describing ﬁrms’ transition status at time t.9 Table 10.7 presents the results of the GLS Random Effects estimation, where ﬁrms speciﬁc effects (ηi) are assumed to be random and uncorrelated with explanatory variables. The overall ﬁndings are consistent with the self-selection hypothesis, since current patenting activity and RJV participation at a later time are signiﬁcantly and positively correlated. In this respect, the remarkable signiﬁcance of the variable ENTRY appears all the more interesting: more innovative ﬁrms at time t are clearly more likely to enter a European RJV

Self-selection and learning in European RJV 221 Table 10.7 Self-selection and learning: patenting differential and transition pattern (1986–96) Dep. Var. PAT_DIFF

IT (N = 1042) Coeff.

EARLY EXIT LATE ENTRY NEW INCUMB NAP REST AWAKE _Cons

0.400 0.310 0.310 0.736 0.658 0.738 0.759 0.103 0.305 0.000

Wald chi2 (Pr > chi2) Nr Obs

17.01 11,462

(Std Err.) (0.299) (0.328) (0.328)** (0.245)*** (0.244)*** (0.237)*** (0.406)** (0.449) (0.406) (0.162) (0.048)

MB (N = 362) Coeff. 0.867 0.866 0.673 0.874 0.683 1.172 0.874 1.024 0.765 0.000 30.39 3,982

(Std Err.) (0.423)** (0.434)** (0.375)* (0.389)** (9.389)* (0.380)*** (0.659) (0.748) (0.659) (0.251) (0.000)

Notes * Signiﬁcant at 10% level. ** Signiﬁcant at 5% level. *** Signiﬁcant at 1% level.

at time t + 1. Assuming cooperative investments are not immediately reﬂected by learning and patenting, the signiﬁcance of the variable NEW may be interpreted along the same lines. The result is similar across technological areas, although a greater level of conﬁdence is recorded in the ICT ﬁeld. The strong evidence in favour of the self-selection hypothesis does not imply, however, that the learning hypothesis is to be entirely dismissed. Indeed, continuous active participation (INCUMB) is highly signiﬁcant in both areas, although the coefﬁcient is considerably higher in the MB ﬁeld, where the variable EXIT also appears signiﬁcantly correlated with current patenting performance, supporting the hypothesis that the innovative differential may be positively affected by the full completion of cooperative experiences.10 Overall, it appears that high innovativeness is associated with continuous or frequent membership in both areas, whereas it is only in the MB ﬁeld that there is a correlation with fully completed RJV activity. We investigated self selection and learning assumptions further by considering the change in innovative differential over time. We can interpret the evidence of a widening post-entry differential as favourable to the learning hypothesis. Therefore, we relate the average ﬁrm innovative differential (from the CG level) over the entire RJV period (1986–96) to the pre-entry level of the differential and the post-entry change in the differential where, for frequent participants, the threshold for distinguishing between ‘pre’ and ‘post’ is represented by the ﬁrst year of afﬁliation to an RJV.

222 L. Cusmano Table 10.8 Self-selection and learning: patenting differential change Dep. Var. AV_DIFF

IT (N = 551) Coeff.

DIFF_PRE D_POST _Cons F

1.125 0.188 0.040 (Prob > F)

2,618.35

MB (N = 164) (Std Err.) (0.021)*** (0.008)*** (0.038) (0.000)

Coeff. 1.178 0.372 0.037 9,323.00

(Std Err.) (0.016)*** (0.005)*** (0.025) (0.000)

Note AV_DIFF = Mean1986–96 [Patit MeantCG]. DIFF_PRE = Mean1986–entry [Patit MeantCG]. D_POST = Meanentry–1996 [(Patit MeantCG) DIFF_PRE]. *** Signiﬁcant at 1% level.

The evidence from this cross-sectional analysis points in the same direction for the two ﬁelds (Table 10.8): the participants’ higher average level of innovativeness (AV_DIFF) is mainly explained by higher innovative activity prior to entry (DIFF_PRE). However, there also appears to be a signiﬁcant contribution from the increase in differential taking place after the entry into the ﬁrst RJV (D_POST).11 10.4.1 Control variables for disentangling self-selection and learning Self-selection is clearly an important explanation for the outstanding innovative performance of RJV participants compared to the standard for European industry, but, by itself, it does not rule out learning. Rather, if the two effects were related differently to some exogenous factor, this could be used as a control variable to disentangle their relevance and impact. If we take the view that production of knowledge is a cumulative process, we should expect past innovative experience to play a central role in explaining current innovative activity alongside speciﬁc current investments, such as afﬁliation to EU consortia. Cumulativeness in the production of knowledge has been discussed in the innovation literature since Arrow (1962) and has been speciﬁcally addressed in the literature about technological regimes (Nelson and Winter 1982; Winter 1984; Malerba and Orsenigo 1993, 1997), emphasizing that the existing knowledge base represents the main input in the generation of new knowledge. Most empirical works estimating a patent function include a measure of R&D capital in the list of explanatory variables to enable innovative efforts to be traced over several years (Griliches 1998). However, as Henderson and Cockburn (1996) point out, if we assume that successful patent applications are driven by the available stock of knowledge capital, a better measure of this stock is innovative output rather than innovative input. In addition, we expect the patent stock to pick up, besides innovative

Self-selection and learning in European RJV 223 experience, ﬁrm-speciﬁc propensity to patent and, to some degree, an endogenous response to past success. Adopting this perspective, we can view the patent ﬂow as an output measure of knowledge-producing activities, and the (lagged) patent stock as a measure of input in the same process. Accordingly, we consider the stock of patent applications from time 0 (that is, 1978, the ﬁrst year of EPO operation) to time t 1 as a proxy of the ﬁrm’s ‘cognitive’ capital, which results from other cooperative experiences as well as in-house R&D investments or informal activities, and fuels the generation of knowledge at time t. Following Henderson and Cockburn (1996), we assume that knowledge stock, like physical capital, is subject to obsolescence and depreciation. Hence, a perpetual inventory method and a 15 per cent or, alternatively, a 20 per cent depreciation rate should be employed to construct the patent stock variable.12 The relevant heterogeneity between participants and non-participants in terms of employment level prompts us to consider size as another important factor accounting for exogenous propensity to innovate. However, the empirical literature investigating the relationship between size and innovation has produced neither unambiguous evidence nor unambiguous interpretation of identiﬁed patterns.13 Hence, rather than choosing size as a control variable based on a priori beliefs, using simple variance analysis we can perform a preliminary test on the extent to which innovative differential is explained by size differences. Firms are assigned to ﬁve size classes related to their average employment level over the period 1992 to 1996.14 The signiﬁcance of size variance is tested with respect to average innovative differential, pre-entry differential and postentry differential change. In the ICT ﬁeld, high signiﬁcance (1 per cent level) is detected in relation to average and pre-entry differential (Table 10.9). That is, the employment level appears to have relevant explanatory power when considering the general higher innovativeness of participants and, more speciﬁcally, their better patenting performance up to the time of their ﬁrst afﬁliation. Size also explains the post-entry change in the differential, but the level of signiﬁcance strongly reduces. In other words, size appears to be a relevant variable accounting for the greater propensity for the more innovative actors to join consortia, but we are less conﬁdent that the change in the innovative gap following cooperative investments, if any, depends on the employment level. For the MB sample, evidence of the signiﬁcance of size in explaining patenting differences is weaker (Table 10.10). Employment class affects the pre-entry differential, but, contrary to the ICT case, it is not signiﬁcant in explaining the overall difference and widening post-entry gap. This evidence supports the idea that the size variable should control for the prior-to-entry innovative level and identify the possible beneﬁts of RJV participation from the follow-up of self selection.

Total

2,086.399

73.106 2,013.293 549

4 545

4.95

0.001

Pr > F

Total

Size Residual

D_POST

Dep. Var.

16,030.927

233.981 15,796.946

Partial SS

549

4 547

df

2.02

F

0.091

Pr > F

Post-entry patenting differential change (N=551)

Note AV_DIFF = Mean1986–96 [Patit MeantCG]. DIFF_PRE = Mean1986–entry [Patit MeantCG]. D_POST = Meanentry–1996 [(Patit MeantCG) DIFF_PRE].

550

Size Residual

4,578.214

5.43

0.000

Total

4 546

F

175.029 4,403.185

df

Size Residual

Partial SS

DIFF_PRE

Pr > F

Partial SS

AV_DIFF

F

Dep. Var.

Dep. Var.

df

Pre-entry patenting differential (N = 550)

Average patenting differential (N = 551)

Table 10.9 ICT sample – patenting differential and size: analysis of variance (1986–96)

Total

471.592

27.798 443.795 163

4 159

2.49

0.045

Pr > F

Total

Size Residual

D_POST

Dep. Var.

4,121.562

34.647 4,086.915

Partial SS

163

4 159

df

0.34

F

0.853

Pr > F

Post-entry patenting differential change (N = 164)

Notes AV_DIFF = Mean1986–96 [Patit MeantCG]. DIFF_PRE = Mean1986–entry [Patit MeantCG]. D_POST = Meanentry–1996 [(Patit MeantCG) DIFF_PRE].

163

Size Residual

1,782.291

1.53

0.197

Total

4 159

F

65.949 1,716.342

df

Size Residual

Partial SS

DIFF_PRE

Pr > F

Partial SS

AV_DIFF

F

Dep. Var.

Dep. Var.

df

Pre-entry patenting differential (N = 164)

Average patenting differential (N = 164)

Table 10.10 MB sample – patenting differential and size: analysis of variance (1986–96)

226 L. Cusmano The following sections apply non-linear count data models to explore the relationship between RJV afﬁliation and patenting behaviour. Patent application at ﬁrm level is the dependent variable and participation in EU consortia is one of the explanatory factors, controlling for knowledge capital, patent stock, industry and, in alternative speciﬁcations, testing for the relevance of other characteristics, such as relationship with large industrial groups or afﬁliation to both FWPs and the EUREKA programme.

10.5 Econometric assessment: patenting and participation in ICT and MB consortia 10.5.1 Econometric models for patent data Generalizations of the Poisson distribution are normally made to estimate equations where patent is the dependent variable, since its integer property is handled directly and the zero value represents a natural outcome of the distribution. The innovative output is conceived as the result of a very large number of Bernoulli trials, for each of which the probability of success is very small. The probability function is deﬁned by: e Pit f(Pit) Pit! it

it

Pit 0, 1, 2, . . .

The process is completely speciﬁed by the parameter λ, which is both mean and variance of the distribution, and can be modelled as an exponential function of the explanatory variables: E(Pit) Var(Pit) it e x'

it

The assumption that events occur independently over time and the mean–variance equality are important weaknesses when analysing patenting behaviour, which typically is characterized by serial correlation and ‘overdispersion’, that is, the variance-to-mean ratio is greater than one.15 Hausman et al. (1984) developed a generalization of the Poisson distribution, the Negative Binomial model, which accounts for inter-personal heterogeneity and overdispersion. The model allows for unexplained randomness in λit, which is captured by a gamma-distributed term ηit: E(Pit) it e x'

it

it

In the Random Effects speciﬁcation of the model, randomness is introduced both across ﬁrms and across time, since the dispersion term is ﬁrmspeciﬁc and stochastic (Hausman et al. 1984; Cameron and Trivedi 1986). In all tests on our samples, evidence of overdispersion was found, there-

Self-selection and learning in European RJV 227 fore negative binomial models are estimated by maximum likelihood. In addition, only the Random Effects speciﬁcation is employed since, contrary to the Fixed Effects speciﬁcation, it allows units for which the dependent variable is always equal to zero, that is, non-innovators, to be taken into account.16 10.5.2 Estimation results We estimated a patent equation at ﬁrm level, where patenting performance of ﬁrm i at time t, measured by yearly patent applications to the EPO, is related to participation of European RJVs at time t j (testing for different values of j), controlling for cumulated innovative experience, as proxied by the depreciated stock of patents at time t 1 (LAGST),17 size, measured by average employment over the estimated period (in logarithmic terms, LEMP) and several dummies, including industry and time dummy variables.18 The signiﬁcance of other control variables was tested over different speciﬁcations. Tables 10.11 and 10.12 present the estimates of the Negative Binomial model for patenting, illustrated above, when alternative speciﬁcations of the vector of explanatory variables are adopted. The tables refer, respectively, to the ICT and MB samples. The basic speciﬁcation (Column 1) does not include dummies other than industry and time. In the basic estimation, we use a three-period lagged measure of RJV participation (RJV_3), that is, we assume that RJV afﬁliation produces its effects with a short-medium lag. In Branstetter and Sakakibara (1998), a two-period lag is indicated as a sensible estimation of the time required for the ﬁrm to absorb and process the knowledge that ‘spills out’ from R&D consortia and employ it for patenting purposes. In the case of the Japanese RJVs that the authors analyse, research personnel are typically rotated into consortia and rotated back to parent ﬁrms on a two-year cycle, bringing to the ﬁrms a substantial amount of explicit and tacit knowledge about the new technology developed within consortia (Sakakibara 1997). It is reasonable to assume that similar routines occur in ﬁrms that contribute to European RJVs. Moreover, we take into account the medium-term scale of European projects, which, for the most part, last between two and three years with only a very few, concentrated in the EUREKA programme, going on for more than four years. Using a three-period lag amounts to assuming RJV capital and knowledge investments are being turned into patent applications mostly during the ﬁnal stages of a project, or in its immediate aftermath.19 In the other speciﬁcations, we test the signiﬁcance of participation in both the Framework and the EUREKA programmes (Column 2), substitute yearly participation ﬁrst with the count of cumulated project-years (Column 3) and then with a frequent-participant dummy (Column 4), and assess the relevance of ﬁrms’ afﬁliation to a larger industrial group (Columns 5 and 6).

228 L. Cusmano In all ﬁelds and speciﬁcations, the patent stock variable is highly signiﬁcant. This evidence supports the ‘technological cumulativeness’ idea: ﬁrms with the biggest stock of accumulated knowledge and the highest recorded patenting activity, are most likely to be innovating.20 In general, we detect a greater effect in the ICT ﬁeld, which may be explained by the different stages in the lifecycles of ICT and MB technologies and to the structure and dynamics of European industry. European research in ICT has mostly involved large industrial groups with signiﬁcant experience in both generic technologies and applied research. The hierarchy of innovators in this ﬁeld appears to be relatively stable, as discussed in the empirical research into technological regimes (Malerba and Orsenigo 1993, 1997). European programmes have attracted ﬁrms that were already remarkably more dynamic than the average European level. However, the development of biotechnologies has introduced ‘instability’ in the hierarchy of innovators, so that the effect of the total number of patents, which is nevertheless very important, is generally smaller. In both areas, the industry classiﬁcation is signiﬁcantly associated with different patenting levels, mostly reﬂecting the coexistence of a large share of mostly non-patenting service ﬁrms with R&D-intensive component producers.21 As far as the lagged RJV measure is concerned, there are some interesting differences between the two ﬁelds. For the ICT ﬁrms, RJV participation never appears to result in a signiﬁcant rise in patenting levels. In terms of the self-selection analysis presented above, it appears that the average higher level of innovativeness of RJV participants is mostly explained by intensive patenting activity prior to entry, which is strongly related to size and industry, but no signiﬁcant impact of RJV afﬁliation on this ‘in-built’ innovative attitude was detected. In contrast, in the MB area, the RJV variable coefﬁcient is positive and signiﬁcant, although the estimate is sensitive to the lag speciﬁcation: RJV afﬁliation is signiﬁcant when speciﬁed with a lag of at least three years. In terms of the incidence rate ratio, an increase of one project per year at time t 3 is associated with a 13 per cent increase in the number of patent applications at time t. This result is consistent with the medium-term horizon of MB consortia: nearly half of the RJVs went on for 31–36 months, the average duration being 42 months. Another explanation relates to the early development of RJVs in the MB area, where the largest number of projects was conducted in the second half of the 1980s, within the EUREKA programme, speciﬁcally aimed at fostering close-to-market technological applications. The timing of the projects combines with the peculiar technological evolution of the ﬁeld, especially biotechnologies, in explaining why frequent early participants have reaped the greatest innovative beneﬁts.22 No signiﬁcant effect is detected when testing the impact of participation in both FWPs and EUREKA, which is done by entering (Column 2)

Self-selection and learning in European RJV 229 both an additional intercept term (EU_EK) and an interaction term for RJV_3 (R3_EUEK). In the MB sample, it would appear that frequent and diversiﬁed involvement in European programmes affects the ability to reap patenting beneﬁts from the yearly projects, but the interaction variable, R3_EUEK, which is signiﬁcant at the 10 per cent level, picks up the positive effect of RJV_3, whose coefﬁcient becomes negative. Results of the signiﬁcance of RJV afﬁliation do not qualitatively change when measuring consortia involvement in terms of accumulated cooperative experience rather than yearly participation. In Column 3, we substitute the measure of projects per year with the ‘stock of RJV participation’ (SRJV), that is, the count of accumulated project-years up to time t.23 However, we still assume the effect of RJV experience shows up with some lag and employ a three-period lagged measure of the RJV stock, SRJV_3. Results conﬁrm the existence of technology ﬁeld differences that are emphasized when referring to RJV participation ﬂows. The variable is non signiﬁcant in the ICT sample, but is signiﬁcant (5 per cent level) in the MB area. The coefﬁcient, however, is much smaller than in the case of the ﬂow variable and corresponds to a 1.05 incidence rate ratio. We may suspect cumulative participation to reﬂect self-selection to a substantial degree, even after controlling for size and knowledge capital, as large and highly innovative ﬁrms are also frequent participants. However, in both ﬁelds, the quality of being a frequent (or a nonfrequent) participant, over the entire programme horizon (1986–96), measured by a dummy variable (FR), does not signiﬁcantly contribute to explaining innovative ability. Column 4 presents the analysis for a cut-off number of three project years. In the ICT ﬁeld, the coefﬁcient becomes signiﬁcant when considering a cut-off number of ten,24 whereas in the MB area the result is robust to different speciﬁcations of the variable. The evidence in this respect is particularly interesting: yearly patenting is positively related to the (lagged) number of projects per year or a measure of the cooperative experience accumulated up to a speciﬁc time t, rather than to the overall membership frequency, which may partly capture the effect of self-selection. In Column 5, we extend the basic speciﬁcation by including another control factor, afﬁliation to large industrial groups, which may play a relevant role in explaining the patenting attitude of the observed entities.25 A signiﬁcant and positive effect for the subsidiary dummy (G_S) is detected in the MB area. However, we expect the size variable to capture part of the group effect, since independent establishments are, on average, smaller than group members. In fact, when omitting the size variable, the relevance of the group dummies increases (Column 6). The positive result for RJV participation in the MB ﬁeld is robust to the inclusion of these additional control factors. Indeed, the change in the value of the coefﬁcient over the different speciﬁcations is negligible.

(0.000)

(0.000)

0.834 0.034 0.129 0.704 0.577 0.007 0.852 0.536 0.139 0.019 0.680 2.599 1.844 0.118 0.213 – – – –

Coeff.

211.57

Yes 1,667.384 1,424.12

(0.041)*** (0.031) (0.019) (0.343)** (0.239)*** (0.287) (0.247)*** (0.219)*** (0.379) (0.292) (0.239)*** (0.182)*** (0.304)***

(1) (Std err.) 0.840 0.041 – 0.696 0.622 0.014 0.849 0.560 0.121 0.032 0.698 2.592 1.884 – – 0.002 – – –

Coeff.

(0.000)

214.12

Yes 1,669.017 (0.000) 1,408.69

(0.042)*** (0.032) (0.099) (0.341)** (0.242)** (0.286) (0.246)*** (0.219)** (0.377) (0.290) (0.239)*** (0.182)*** (0.308)*** (0.099) (0.188)

(2) (Std err.)

(0.000)

(0.000)

(0.008)

0.851 0.055 – 0.694 0.653 0.075 0.859 0.601 0.161 0.087 0.730 2.595 1.967 – – – 0.208 – –

Coeff.

206.37

Yes 1,667.984 1,430.49

(0.343)** (0.239)*** (0.287) (0.247)*** (0.219)*** (0.379) (0.292) (0.239)*** (0.182)*** (0.304)***

(0.041)*** (0.031)*

(3) (Std err.)

– 0.045 0.048

0.845 0.042 0.009 0.717 0.634 0.007 0.859 0.564 0.094 0.020 0.703 2.587 1.882 – –

Coeff.

(0.000)

212.11

Yes –1,668.80 (0.000) 1,416.40

(0.142)

(0.342)** (0.238)*** (0.289) (0.246)*** (0.219)*** (0.379) (0.292) (0.239)*** (0.181)*** (0.309)***

(0.041)*** (0.033)*

(4) (Std err.)

(0.000)

(0.000)

0.857 – 0.009 0.669 0.643 0.050 0.831 0.565 0.156 0.004 0.660 2.597 1.655 – – – – 0.103 0.017

Coeff.

215.09

Yes 1,669.685 1,424.49

(0.041)*** (0.032) (0.019) (0.344)** (0.241)*** (0.288) (0.247)*** (0.219)* (0.383) (0.294) (0.241)*** (0.183)*** (0.312)*** – – – – (0.216) (0.149)

(5) (Std err.)

(0.000)

(0.000)

(0.210) (0.146)

–

(0.019) (0.340)** (0.239)*** (0.284) (0.244)*** (0.217)*** (0.376) (0.293) (0.236)*** (0.183)*** (0.257)***

(0.040)***

(6) (Std err.)

Notes * Signiﬁcant at 10% level. ** Signiﬁcant at 5% level. *** Signiﬁcant at 1% level. Industry dummy variables (US SIC Code): MED_INS (384), COMP (357), COMP_SER (737), E_EQUIP (36, excluding 365, 366, 367), COMM (365, 366, 367), COMM_SER (48), P_EQUIP (38, excluding 384), ENGMAN (871, 873).

213.89

Yes 1,668.934 1,413.68

0.842 0.042 0.009 0.697 0.633 0.018 0.851 0.560 0.123 0.031 0.700 2.589 1.896 – – – – – –

@tabt:LAGST LEMP RJV_3 MED_INST COMP COMP_SER E_EQUIP COMM COMM_SER P_EQUIP ENGMAN LAGNOST CONS R3_EUEK EU_EKA SRJV_3 FR3 G_P G_S

Time dummies Log likelihood Wald chi2 (Pr > chi2) LR test – pooled – (Pr > chi2)

Coeff.

Dep. Var. PAT

Table 10.11 ICT sample (N = 1,042) negative binomial random effects

(0.000)

33.96

Yes 511.259 402.33

0.797 0.128 0.620 1.664 1.230 1.209 1.650 0.848 2.824 2.512 0.739 0.146 – – – –

Coeff.

(0.000)

(0.000)

(0.089)*** (0.088) (0.332)* (0.606)*** (0.540)** (0.666)* (0.608)*** (0.635) (0.334)*** (0.827)*** (0.326)** (0.304)

(2) (Std err.)

322.980

Yes 515.063 398.184

0.809 0.121 – 1.692 1.223 1.246 1.713 0.871 2.922 2.567 – – 0.053 – – –

Coeff.

(0.000)

(0.000)

(0.022)**

(0.603)*** (0.539)** (0.659)* (0.598)*** (0.633) (0.332)*** (0.818)***

(0.086)*** (0.087)

(3) (Std err.)

27.77

Yes 517.58 399.88

0.799 0.125 – 1.691 1.211 1.202 1.777 0.905 2.896 2.743 – – – 0.053 – –

Coeff.

(0.000)

(0.000)

(0.240)

(0.592)*** (0.529)** (0.651)* (0.588)*** (0.620) (0.332)*** (0.811)***

(0.087)*** (0.087)

(4) (Std err.)

27.51

Yes 513.393 409.30

0.531 0.471

0.806 0.069 0.124 1.608 1.284 1.394 1.934 0.879 2.914 2.111 – – –

Coeff.

(0.000)

(0.000)

– (0.356) (0.266)*

(0.085)*** (0.089) (0.053)** (0.589)*** (0.527)** (0.644)** (0.596)*** (0.617) (0.330)*** (0.922)

(5) (Std err.)

Notes * Signiﬁcant at 10% level. ** Signiﬁcant at 5% level. *** Signiﬁcant at 1% level. Industry dummy variables (US SIC Code): CHEM (28, excluding 283), DRUGS (283) MED_INS (384), ENGMAN (871, 873), FOOD (20).

30.64

Yes 515.202 400.85 (0.000)

Time dummies Log likelihood Wald chi2 (Pr > chi2) LR test – pooled – (Pr > chi2)

(0.085)*** (0.086) (0.054)** (0.598)*** (0.535)** (0.653** (0.594)*** (0.628) (0.331)*** (0816)***

0.805 0.121 0.126 1.679 1.209 1.261 1.709 0.870 2.920 2.556 – – – – – –

LAGST LEMP RJV_3 CHEM DRUGS MED_INST ENGMAN FOOD LAGNOST CONS R3_EUEK EU_EKA SRJV_3 FR3 G_P G_S

(1) (Std err.)

Coeff.

Dep. Var. PAT

Table 10.12 MB sample (N = 362) negative binomial random effects

29.80

Yes 513.699 412.74

0.830 – 0.126 1.614 1.285 1.325 1.835 0.917 2.928 2.240 – – – – 0.608 0.513

Coeff.

(0.000)

(0.000)

(0.340)* (0.259)**

(0.079)*** – (0.053)** (0.578)*** (0517)** (0.630)** (0.575)*** (0.603) (0.330)*** (0.641)***

(6) (Std err.)

232 L. Cusmano 10.5.2.1 Analysis by size class The evidence from both the self-selection analysis and patent equation estimation points to an interesting line of investigation, which concerns the role played by size in determining the ability, or interest, of ﬁrms to reap advantage from RJV afﬁliation. Accordingly, we tried to assess whether the relationship between patenting activity and our explanatory variables changes over different size classes. We estimated the basic speciﬁcation on three subsamples, grouping ﬁrms in three size classes based on average number of employees over the period 1992–96. Table 10.13 presents the estimated coefﬁcients for the participation variable when employing the basic speciﬁcation (Column 1 in Tables 10.11 and 10.12). Even in the MB area, small ﬁrms do not appear to beneﬁt substantially from R&D cooperation in terms of patenting activity. In this case, the measure of output – patent applications – is all the more important for the interpretation of results. Small ﬁrms exhibit a low propensity to patent and are likely to acquire beneﬁts from RJVs activity that are not measurable by patent output, for instance, access to best practice, cost and risk sharing, exchange of technological information that would otherwise not be accessible, opportunities for product and process development that do not translate into formal innovations. In other words, we can expect participation in RJVs to produce ‘patenting’ beneﬁts for ﬁrms that already exhibit a signiﬁcant propensity to patent, rather than to ﬁrms that, due to their managerial capabilities or attitudes, or their internal resources and competencies, do not generally orient their more innovative investments and activities towards patenting. At the other dimension, for large ﬁrms the impact of participation in a RJV is positive and signiﬁcant in the MB area only. Indeed, the result Table 10.13 Size class analysis (RJV 3 estimates)

IT CLASS 1 (N = 485) CLASS 2 (N = 324) CLASS 3 (N = 233) MB CLASS 1 (N = 181) CLASS 2 (N = 100) CLASS 3 (N = 81) MEMP > 250 (N = 181) MEMP > 500 (N = 125)

Coeff.

Std Err.

0.141 0.153 0.007

(0.125) (0.895) (0.020)

0.455 1.662 0.108 0.149 0.136

(0.313) (1.124) (0.065)* (0.059)** (0.062)**

Note MEMP = average number of employees over the period 1992–96. CLASS 1: MEMP < 250. CLASS 2: 250 < MEMP < 1,000. CLASS 3: MEMP > 1,000.

Self-selection and learning in European RJV 233 becomes even more signiﬁcant if a larger sub-sample of medium-large ﬁrms is selected, that is, if we exclude from the sample small and mediumsized ﬁrms (that is, those with less than 250 or less than 500 employees). A possible interpretation of the non signiﬁcance for large ICT ﬁrms may relate to the minor importance, in relative terms, of RJV projects for ﬁrms with large R&D departments and substantial R&D budgets. Moreover, strategic issues can be extremely important in affecting the attitude of large ﬁrms to information sharing and disclosure of technological developments through patenting. As noted earlier in this chapter, large ﬁrms entering RJVs in the ICT area are highly innovative prior to entry. We can expect, then, that for those ﬁrms participation is not aimed primarily at increasing patent output, but rather is due to a keenness to join key European networks involving the main sector competitors and public R&D institutions. This analysis is certainly not conclusive, but does open up some interesting questions for further research regarding the relationship between ﬁrm size and reason for cooperative R&D investment.

10.6 Conclusions Cooperative programmes have gained centre stage in European RTD policies. Shared-cost R&D consortia currently represent the main source of R&D funding at European level. Both policy attention and funding have been increasingly oriented towards the broad ﬁelds of ICT and MB, areas in Europe where the research infrastructure is in urgent need of more effective policies. In this chapter we have presented an assessment of the effect of RJV participation on patenting activity, which elaborates a large panel of data of European ﬁrms engaged in cooperative research activities in these areas. Some interesting results emerge from the empirical analysis in terms of the characteristics of participants and the differences between the two technological macro areas studied, which have important policy implications. The evidence about patenting activity shows that RJV participants are signiﬁcantly more innovative than non-participants. In the ICT area, the difference mostly reﬂects self selection. The most important waves of European programmes have attracted highly research-intensive ﬁrms that were already innovating at a higher rate than the average European level, whereas in the MB area, except for a peak in 1989, early RJV members do not exhibit high levels of patenting prior to entry. In this respect, the target and nature of the European consortia appear to be relevant explanatory factors. In fact, ICT projects are oriented towards more generic, pre-competitive research, with the aim of providing a common technological basis for IT applications and support for the development of a European market for information services. All the major

234 L. Cusmano telecommunication operators and key European equipment manufacturers participate in these projects. In the MB ﬁeld, a high share of consortia deals with applied projects within the EUREKA programme and attracts, together with a few major players, a large number of ﬁrms that are not endowed with a signiﬁcant ‘innovative platform’ in terms of patents. The results of the econometric investigation, based on a Negative Binomial modelling of the patent function, are consistent with the preliminary ﬁndings of the self-selection analysis. There is no evidence of a positive correlation between patenting activity and RJV afﬁliation in the ICT area, whereas a positive and signiﬁcant effect is detected in the MB ﬁeld, though cooperative investments appear to show their effects with a rather signiﬁcant lag. This ﬁnding is consistent with the nature and medium- to long-term duration of the MB projects supported by the EU. Applied market-oriented projects represent a relevant share of European MB consortia, especially within EUREKA. In this sense, if we take patents as indicators of performance, it appears that the programmes’ objective to promote research in emerging ﬁelds and innovative application has been achieved. Only SMEs do not appear to beneﬁt from RJV participation in terms of increased patent applications. It is certainly the case that the patent measure underestimates their innovative ability, but we may also relate the negligible impact to the requirement of a minimum amount of resources and extension of productive activities in order to be able to exploit joint investments and the externalities generated in the course of the interaction. The overall positive result partly reﬂects the large number of virtually non-innovative ﬁrms that embarked on the innovation track only after the ﬁrst entry into a EU-supported RJV. Hence, the positive effect might be related to the role of RJVs in opening up ‘innovative networks’ to new members. In the case of ICT consortia, which are more oriented towards precompetitive research and development of generic technologies, it appears that they have attracted the major leaders in the area, thereby reinforcing their role rather than opening up innovative networks to new members with a high, but still unexpressed, innovative potential. The results of the empirical investigation open up interesting questions for further studies. In particular, the differences between technological macro sectors and the role of cumulated innovative experience and size in determining ability to reap beneﬁts from cooperation should be further investigated. In terms of policy implications, the empirical results should be interpreted bearing in mind that the ICT and MB industries are at different stages in their life cycles. It appears that cooperative policies have not affected patenting behaviour in technological areas where a network of excellence has already emerged and the hierarchy of innovators is fairly

Self-selection and learning in European RJV 235 stable, whereas they have favoured the exploitation of innovative potential by new actors in the case of emerging technologies. This ﬁnding suggests that scrupulous attention should be given to sector dynamics during policy design and in the evaluation of policy targets and achievements.

Appendix 10.1: Missing values imputation Missing data represent a common and pervasive problem in panel analysis, especially when data are collected through waves of surveys. Standard statistical methods are designed for complete datasets and rectangular data matrices. In the presence of missing values, the rectangular form can be restored either by discarding incomplete cases or imputing values in place of the missing ones. Complete-case analysis, which simply discards cases with any missing values, has two major drawbacks. First, ‘listwise deletion’ implies loss of information and less efﬁcient estimates than anticipated. If the number of variables is very large, then even a sparse pattern of missing Xs can result in a substantial number of incomplete cases. Second, if the units with missing values are systematically different from the units with complete data, ignoring these differences and discarding incomplete cases lead to biased analysis (Little and Schenker 1995). In the case of our panel, discarding cases because of sparse and missing data about the employment level amounts to reducing the number of observations by more than half. We might therefore prefer to evaluate RJV impact on a larger sample of ﬁrms and incorporate some of the incomplete cases in the analysis, after imputing missing values. Imputation methods are generally based on the predictive distribution of the missing values, given the other observed values for a unit.26 A very simple technique is the mean imputation, which substitutes missing values with means from the observed units. The use of unconditional sample means, however, yields inconsistent estimates, since variances are understated and association between variables is distorted (Little and Schenker 1995). Conditional mean imputation replaces each missing value with an estimate of its conditional mean, given the other observed values of the same unit. Conditioning is based on forming ‘adjustment cells’ that group individuals, or units, according to some observed variables. Missing values are then replaced by the relevant mean of the complete cases in the same cell. Hot deck imputation is also based on adjustment cells. Individuals are grouped into categories and imputation is made by substituting missing values with observed values drawn from individuals within the same cell. The imputation technique we adopted is based on the idea of adjustment cells, but also exploits the time-series dimension of the data. In this regard, it follows the suggestion by Lepkowski (1989), who deals with the case of knowledge non-response in panel surveys and underlines that, when responses to an item are highly correlated over time, responses in

236 L. Cusmano one wave are powerful auxiliary variables for imputing the missing responses for the same item in another wave. In our case, the employment variable is highly correlated across time. Therefore, we exploit both the time-series information on the same unit and information about the population (or ‘cell’) to which the unit belongs. We form adjustment cells by considering the rate of growth of employment in a given sector, in a given country and in a given year. In more detail, employment missing values are imputed taking into account ﬁrm i time-series information (employment at time t + 1 or t 1), and extra-sample information (AMADEUS database) about the average employment rate of growth at each time period for each sector k in each country w. Moreover, in order to reduce the problem of limited sampling variability, which typically arises when adopting imputation techniques, we add to the sector- and country-speciﬁc rate of growth a stochastic component, ε, normally distributed with zero-mean and standard deviation equal to the estimated standard deviation of the sector- and country-speciﬁc rate of growth at time t. EMPikwt EMPikwt1 (1 g– kwt kwt) where i = ﬁrm, k = sector, w = country, t = time 1 N EMPikwt EMPikwt1 g– kwt 冱 N i1 EMPikwt1 and kwt N(0, ˆg kwt)

Notes 1 I wish to thank Franco Malerba, Francesco Lissoni, Stefano Breschi, Fabio Montobbio and Maria Luisa Mancusi for their encouragement and helpful comments. I also beneﬁted from valuable feedback from the participants in the 2001 KNOW FOR INNOVATION ﬁnal workshop (Athens) and the 2001 Nelson and Winter Conference (Aalborg). The usual disclaimers apply. 2 The seminal work by Branstetter and Sakakibara (1998) on Japanese government-sponsored research consortia represents in this respect a notable exception. 3 The ICT research has been organized in several broad rolling-work programmes, covering information technology, electronics and telematics (e.g. RACE, ACTS, ESPRIT, TELEMATICS). 4 EU investments in biomedical R&D have been channelled, since the ﬁrst half of the 1990s, through two related programmes, BIOMED and BIOTECH. BIOMED had the objective of improving the efﬁcacy of medical and health

Self-selection and learning in European RJV 237

5

6

7

8

9 10

11

12

R&D in the EU Member States, to be achieved in particular, by way of better coordination of their research activities and application of results through Community cooperation and pooling of resources. The aim of BIOTECH research has been to improve the basic biological knowledge of living systems and extend its application to agriculture, health, nutrition and environmental issues. EUREKA was launched in 1985 as a loose inter-governmental, decentralized and industry-led initiative. France’s initiative was of the greatest importance for the design of the programme, which was proposed in response to the US Strategic Defence Initiative (SDI) perceived as having major industrial policy implications. The political process which led to its creation represents an extremely interesting case of tension between the need to strengthen European industrial cooperation and the unwillingness of national governments to transfer power to the Commission (Sandholtz 1992; Peterson 1993). The dataset represents the output of research conducted within the EC TSER project ‘KNOW FOR INNOVATION’ (Innovation-Related Knowledge Flows in European Industry: Extent, Mechanism, Implications), Contract No. SOE1-CT981118, DGXII G4, coordinated by Professor Y. Caloghirou of the National Technical University of Athens (Greece). In more detail, the RJV_EPO dataset results from the merging of three databases: CORDIS (Community Research and Development Information Service), the EUREKA central database, and the EPO-CESPRI database, which provides full details about patents applied for and granted by the European Patent Ofﬁce since the beginning of its operations (1978). Sector and employment details are extracted from the AMADEUS database, which gives longitudinal ﬁnancial and sector information for approximately 200,000 ﬁrms in Europe. Firms are included in the samples if employment information is available for at least one year more than the estimation period 1992–96. Employment missing values have been imputed using a technique based on sector- and country-speciﬁc employment average rate of growth, with the addition of a stochastic component (see Appendix 10.1). Again, the AMADEUS dataset presents sparse and missing values for the employment level over the period 1992–96. Only units for which employment detail is given for at least one year have been taken into account and missing values have been imputed according to the technique speciﬁed in Appendix 10.1. The CG dummy variable provides the basis for the estimation. In interpreting results we should also take into account the limits of the patent variable in depicting innovative potential. That is, there may be cases where ﬁrms have not patented very much before entry, but in fact there is self selection by ﬁrms with a high innovative potential. Other factors, such as prestige of researchers, number of publications, participation in meetings and conferences, could be considered as signals of innovative capacity, which are perceived by possible partners and exploited during or after participation in a consortium. When excluding from the analysis the highly innovative ﬁrms that entered in 1996 only (see Figure 10.1), results do not signiﬁcantly differ. In the MB area only, the coefﬁcient of DIFF_PRE strongly reduces, whereas D_POST coefﬁcient increases, but still both variables are signiﬁcant at the 1 per cent level. Henderson and Cockburn (1996) employ a patent stock variable to measure knowledge capital accumulated by major R&D performing ﬁrms in the pharmaceutical industry and ﬁnd, over the period 1961–88, a signiﬁcant effect of such a variable on grants of ‘important’ patents, i.e. patents that have been granted in two of the three main jurisdictions: Japan, Europe and the United States.

238 L. Cusmano 13 Scherer (1980), Kamien and Schwartz (1982) and Cohen (1995) provide overviews of studies, based around the Schumpeterian debate, on the relationship between market concentration, ﬁrm size and innovative effort. 14 Size classes: 1) 0–50; 2) 51–250; 3) 251–500; 4) 501–1,000; 5) >1,000. 15 If the equality assumption is violated, parameters are consistently estimated, but standard errors are typically underestimated, implying spuriously high levels of signiﬁcance. 16 In a Fixed Effects count-data model, units for which the dependent variable is always equal to zero do not contribute to maximizing the likelihood function. 17 We measure the knowledge capital stock in logarithmic terms, adding 1 to zero-value observations in order to get a deﬁnite logarithmic value, and including an adjustment dummy variable (LAGNOST), which is equal to 1 if the stock of patents at time t 1 is equal to zero. By taking the natural log of the (depreciated) patent stock, we still implicitly assume there are diminishing effects of accumulated knowledge on patent activity. A 15 per cent depreciation rate is employed in the basic speciﬁcation. We performed estimates employing a non-logarithmic stock variable. However, this variable captures most of the size effect (LEMP is always non-signiﬁcant) and the performance of the model, in terms of log-likelihood, worsens signiﬁcantly. 18 Time dummies are included in all regressions to control for the ﬂuctuating and slightly declining trend of patents in our data. 19 We assessed the validity of this assumption and tested the sensitivity of results to both shorter and longer lags. We do not include various lags of the RJV participation variable in the same equation in order to avoid problems related to multicollinearity. In fact, consortia afﬁliation is highly correlated over time. This high value of the correlation coefﬁcient is partly due to the time horizon of projects, whose duration ranges from less than one year to ﬁve years. Estimations may be also performed employing a contemporaneous RJV participation measure. Branstetter and Sakakibara (1998) use the contemporaneous RJV variable, as well as alternative lagged measures, and tackle the endogeneity problem employing Two Stages Least Squares. However, this speciﬁcation strongly implies that consortia have an immediate effect on ﬁrms’ patenting activity. We prefer to assume that the full impact of RJV participation comes only after a lag and expect the patent stock variable (LAGST) to account for generic differences between ﬁrms’ innovative abilities, so that controlling for those differences (and controlling for size and industry), the lagged RJV participation variable mostly captures the effects that are due to different levels of involvement in European R&D consortia. 20 The use of a 20 per cent depreciation rate in the calculation of the stock variable leads to higher log-likelihood and greater coefﬁcient, as if faster depreciation would better explain the relationship between current patenting activity and cumulated innovative experience. 21 TRADE represents the basis for estimation (omitted dummy). 22 In this sense, when the lag is very high there is a risk that the RJV variable does not really represent the intensity of RJV participation, but rather picks up the pioneer ﬁrms that entered the ﬁrst European R&D projects in the early years of the EUREKA programme and already had a high propensity to innovate. This issue is similar to the self-selection issue, although it is more speciﬁc and concerns a clearly identiﬁed group of ﬁrms that could play the role of outliers. We have assessed the relevance of this effect by ﬂagging the greatly innovative ﬁrms, which entered MB RJVs in 1989, with a dummy variable, PIONEER, and re-estimating the basic speciﬁcation. The coefﬁcient of RJV_3 is still positive and signiﬁcant, whereas the pioneer dummy is non-signiﬁcant. 23 We deﬁne project-years as the sum of the years ﬁrm i has been afﬁliated to

Self-selection and learning in European RJV 239 R&D consortia, where independent RJVs in the same year account for one project-year each. 24 We expect this variable to ﬂag the biggest participants. 25 We have chosen 1992, the ﬁrst year of our analysis, as the reference year for classifying ﬁrms as Parent (dummy variable G_P), Subsidiary (G_S) or Independent (G_I). Firms that are not classiﬁed by Dun and Bradstreet (1992) as parent or subsidiary are labelled ‘independent’. This last category serves as the basis for our estimation (omitted dummy variable). 26 Little and Rubin (1987) provide a broad taxonomy of imputation procedures that are common practice in survey analysis.

Bibliography Arrow, K.J. (1962) ‘Economic welfare and the allocation of resources for invention’, The Rate and Direction of Inventive Activity: Economic and Social Factors, Princeton NJ: NBER, pp. 609–625. Aw, B.Y., Chung, S. and Roberts, M.J. (2000) ‘Productivity and turnover in the export market: micro-level evidence from the Republic of Korea and Taiwan’, The World Bank Economic Review, 14(1): 65–90. Bernard, A.B. and Jensen, J.B. (1995) ‘Exceptional exporter performance: cause, effect or both?’, Journal of International Economics, 47: 1–25. Branstetter, L. and Sakakibara, M. (1998) ‘Japanese research consortia: a microeconometric analysis of industrial policy’, Journal of Industrial Economics, XLVI: 207–233. Cameron, A.C. and Trivedi, P.K. (1986) ‘Econometric models based on count data: comparisons and applications of some estimators and tests’, Journal of Applied Econometrics, 1: 29–53. Cassiman, B. (2000) ‘Research joint ventures and optimal R&D policy with asymmetric information’, International Journal of Industrial Organization, 18(2): 283–314. Clerides, K.S., Lach, S. and Tybout, J.R. (1998) ‘Is learning by exporting important? Micro-dynamic evidence from Colombia, Mexico and Morocco’, Quarterly Journal of Economics, 113: 903–947. Cohen, W.M. (1995) ‘Empirical studies of innovative activity’, in P. Stoneman (ed.) Handbook of the Economics of Innovation and Technological Change, Oxford: Basil Blackwell, pp. 182–264. Contractor, F.J. and Lorange, P. (eds) (1988) Cooperative Strategies in International Business, Lexington, MA: Lexington Books. d’Aspremont, C. and Jacquemin, A. (1988) ‘A note on cooperative and non-cooperative R&D in duopoly’, American Economic Review, 5: 1133–1137. Dun and Bradstreet (1992) Linkages Database, Murray Hill, NJ: Available at: www.dnb.com. European Commission (1994) The European Report on Science and Technology Indicators, Brussels: European Commission, DG XIII. Geroski, P. (1995) ‘Markets for technology’, in P. Stoneman (ed.) Handbook of the Economics of Innovation and Technological Change, Oxford: Basil Blackwell, pp. 90–131. Griliches, Z. (1998) R&D and Productivity: The Econometric Evidence, Chicago, IL: University of Chicago Press.

240 L. Cusmano Hagedoorn, J. (1993) ‘Understanding the rationale of strategic technology partnering: interorganizational modes of cooperation and sectoral differences’, Strategic Management Journal, 14: 371–385. Hagedoorn, J. and Narula, R. (1996) ‘Choosing modes of governance for strategic technology partnering: international and sectoral differences’, Journal of International Business Studies, 27: 265–284. Hausman, J., Hall, B. and Griliches, Z. (1984) ‘Econometric models for count data with application to patents – R&D relationship’, Econometrica, 52: 909–938. Henderson, R. and Cockburn, I. (1996) ‘Scale, scope, and spillovers: the determinants of research productivity in drug discovery’, Rand Journal of Economics, 27(1): 32–59. Kamien, M.I. and Schwartz, N.L. (1982) Market Structure and Innovation, Cambridge: Cambridge University Press. Kamien, M.I., Muller, E. and Zang, I. (1992) ‘Research joint ventures and R&D cartels’, American Economic Review, 82(5): 1293–1306. Katz, M.L. (1986) ‘An analysis of cooperative R&D’, Rand Journal of Economics, 17(4): 527–543. Klette, T.J., Møen, J. and Griliches, Z. (2000) ‘Do subsidies to commercial R&D reduce market failures? Microeconometric evaluation studies’, Research Policy, 29: 471–495. Lazaric, N. and Marengo, L. (2000) ‘Towards a characterisation of assets and knowledge created in technological agreements: some evidence from the Automobile Robotics sector’, Industrial and Corporate Change, 9(1): 53–86. Lepkowski, J.M. (1989) ‘Treatment of wave nonresponse in panel surveys’, in D. Kasprzyk, G.J. Duncan, G. Kalton and M.P. Singh, Panel Surveys, New York: John Wiley & Sons, pp. 1–24. Little, R.J.A. and Rubin, D.B. (1987) Statistical Analysis with Missing Data, New York: John Wiley & Sons. Little, R.J.A. and Schenker, N. (1995) ‘Missing data’, in G. Arminger, C.C. Clogg and M.E. Sobel (eds) Handbook of Statistical Modelling for Social and Behavioral Sciences, New York: Plenum Press, pp. 39–75. Llerena, P. and Matt, M. (1999) ‘Inter-organizational collaboration: the theories and their policy implications’, in A. Gambardella and F. Malerba (eds) The Organization of Economic Innovation in Europe, Cambridge: Cambridge University Press, pp. 179–201. Malerba, F. and Orsenigo, L. (1993) ‘Technological regimes and ﬁrm behavior’, Industrial and Corporate Change, 2(1): 45–74. Malerba, F. and Orsenigo, L. (1997) ‘Technological regimes and sectoral patterns of innovative activities’, Industrial and Corporate Change, 6(1): 83–118. Morasch, K. (2000) ‘Strategic alliances as Stackelberg cartels: concept and equilibrium alliance structure’, International Journal of Industrial Organization, 18(2): 257–282. Nelson, R.R. and Winter, S.G. (1982) An Evolutionary Theory of Economic Change, Cambridge, MA: Harvard University Press. Peterson, J. (1993) High Technology and the Competition State: An Analysis of the EUREKA Initiative, London: Routledge. Peterson, J. and Sharp, M. (1998) Technology Policy in the European Union, London: Macmillan Press. Sakakibara, M. (1997) ‘Evaluating government-sponsored R&D consortia in Japan: Who beneﬁts and how?’, Research Policy, 26: 447–473.

Self-selection and learning in European RJV 241 Sandholtz, W. (1992) High-Tech Europe: The Politics of International Cooperation, Berkeley, CA: University of California Press. Scherer, F.M. (1980) Industrial Market Structure and Economic Performance, 2nd edn, Chicago, IL: Rand McNally. Sinha, D.K. and Cusumano, M.A. (1991) ‘Complementary resources and co-operative research: a model of research joint ventures among competitors’, Management Science, 37(9): 1091–1105. Spence, M. (1984) ‘Cost reduction, competition and industry performance’, Econometrica, 52: 101–121. Stoneman, P. and Vickers, J. (1988) ‘The assessment: the economics of technology policy’, Oxford Review of Economic Policy, 4(4): 1–16. Teece, D. (1986) ‘Proﬁting from technological innovation: implications for integration, collaboration, licensing and public policy’, Research Policy, 15: 285–305. Winter, S.G. (1984) ‘Schumpeterian competition in alternative technological regimes’, Journal of Economic Behaviour and Organisations, 5: 287–320.

11 The evolution of intra- and intersector knowledge spillovers in the EU Framework Programmes1 Michel Dumont and Aggelos Tsakanikas

11.1 Introduction In this chapter we investigate knowledge spillovers stemming from research and development (R&D) cooperation in the context of the EU Framework Programmes (EU-FWPs). We measure spillovers in a direct way, following a limited number of assumptions. Intra- and inter-sector spillovers are computed for the ﬁrst four Framework Programmes (FWPs) (1984–98), in order to analyse the evolution of cooperation patterns in response to policy and other (that is, technology) shifts. We adopt a ‘learning by networking’ perspective, which assumes that knowledge ﬂows are not limited to ﬂows of codiﬁed information and argue that measuring spillovers is essential in assessing the rationale and the impact of networkpromoting policies. In the burgeoning (game-) theoretical literature on spillovers and R&D cooperation, the magnitude and speciﬁc nature of spillovers is found to be crucial for the rationale of a policy to promote and subsidize R&D cooperation. However, there has been relatively little empirical investigation of policy-induced spillovers in general and spillovers within the EUFWPs in particular. In this chapter we apply the method proposed in Dumont and Tsakanikas (2001a) to measure knowledge spillovers stemming from cooperation in R&D. Contrary to most other methods, we clearly deﬁne the spillover mechanism and compute knowledge spillovers in a direct way, based on a limited number of prior assumptions. By focusing on spillovers through subsidized R&D cooperation within the EU-FWPs we can assess the impact of a large-scale network promoting policy. As Jaffe (1998) argued in his assessment of the US Advanced Technology Program (ATP), evaluating a programme the rationale of which is to create spillovers, implies an attempt to measure spillovers. We use data on inter-ﬁrm linkages established in the context of the EUFWPs, the major mechanism for subsidizing cross-national ‘pre-competitive’ R&D collaboration. Intra- and inter-sector spillovers are computed for each of the ﬁrst four FWPs, covering an extensive period of 15 years

Knowledge spillovers in EU FWPs 243 (1984–98), in order to analyse the evolution of cooperation patterns in response to policy and other (that is, technology) shifts. We endorse a ‘learning by networking’ perspective, which assumes that knowledge ﬂows are not limited to ﬂows of codiﬁed information, hence, redundant linkages can be justiﬁed, whereas an efﬁciency approach, focusing on the ﬂows of codiﬁed knowledge, would favour a strategic network position with as few (redundant) linkages as possible. The chapter is structured as follows: in Section 11.2 we discuss the policy priorities set by EU funding in the various technological areas and subsequent changes as the FWPs evolved. Section 11.3 describes the dataset used for computing spillovers and in Section 11.4 the methodology is brieﬂy presented.2 Section 11.5 shows the clusters of the strongest intra- and inter-sector spillovers at aggregate EU level. In Section 11.6 we apply a shift–share analysis to the ‘market’ of FWP cooperative linkages to analyse the shift in spillovers over the consecutive FWPs. Section 11.7 summarizes our conclusions and proposes avenues for possible future research.

11.2 Evolution of the EU-FWP funding At the beginning of the 1980s concerns were raised that the European economy was falling behind the US and Japan in emerging high-tech areas like information and communications technologies (ICTs). The establishment of a programme promoting collaborative research in this area was the ﬁrst action the EU took in order to face decreasing market share. ESPRIT 1 was established in 1984 in collaboration with 12 large European ICT ﬁrms3 in order to strengthen the scientiﬁc and technological basis and IT capabilities of European industry and eventually improve its competitiveness in global markets (Molina 1996). However, the ofﬁcial basis for the implementation of science and technology policy in the EU was only created in 1987, when the ﬁrst FWP was established under the Single European Act. ESPRIT 1 served as a model and motivation for a general ‘umbrella type’ framework which included programmes in other technological areas. The implementation of European Research and Technological Development (RTD) policy was achieved through three major mechanisms: (1) shared cost contract RTD projects; (2) concerted actions; and (3) own research activities taking place in a network of European research centres, the Joint Research Centre (JRC). The ﬁrst two mechanisms represent the so-called ‘indirect measures’, involving actors from all over Europe, whereas the third action is a more direct measure, referring to the EU’s own research activities. However, shared cost contract RTD projects can be considered the main mechanism, since they acquire the major proportion of EU funding. Projects are coordinated and implemented through the FWPs, which act as planning instruments, laying down budget allocations in the various technology

244 M. Dumont and A. Tsakanikas areas.4 FWPs have gradually become the main instrument of the European technology policy, aiming at enhancing European competitiveness by promoting transnational cooperation in research at a pre-competitive level, and encouraging the dissemination of information (Lucchini 1998). The FWP budget allocations reﬂect the shift in EU policy priorities regarding the various research ﬁelds. These allocations are the result of a consultation procedure involving the EU, the Council of Ministers, the European Parliament and the Economic and Social Committee. Table 11.1a sums up the picture over the main technology ﬁelds in the second, third and fourth FWPs. Since there have been changes over the years in the names and the content of these ﬁelds, the table is based on an ofﬁcial harmonization (European Commission 1997: 503). The budget of the ﬁrst FWP was classiﬁed in broadly deﬁned categories and therefore could not be incorporated in this table. The focus of the ﬁrst FWP was mainly on ICTs, since this was the main reason for its formation. In the second FWP the total budget was increased, although it still represented less than 3 per cent of EU member states’ total government spending in RTD and about 1 per cent of all (public and private) RTD spending (Peterson and Sharp 1998). ICTs have always been the main technological ﬁeld in terms of funding, presenting a continuous increase in absolute numbers. However, due to the diversiﬁcation of priorities, their share in the total FWP budget Table 11.1a FWP funding of the various technological areas (in million C) Second FWP

Third FWP

Fourth FWP

ICT Information technologies (Tele)communications Other (e.g. telematics) Energy Thermonuclear fusion Nuclear ﬁssion Non-nuclear energy Industrial and materials technology Life sciences Agro-industry (incl. ﬁshery) Biotechnology Biomedical/Health Environment Environment Marine science

2,275 (45.6%) 1,600 550 125 1,173 (23.5%) 611 440 122 845 (16.9%) 390 (8.3%) 190 120 80 311 (6.6%) 261 50

2,491 (42.7%) 1,517 548 426 1,052 (18.1%) 562 231 259 997 (17.1%) 706 (13.5%) 373 184 149 581 (11.1%) 464 117

3,405 (33.0%) 1,932 630 843 2,256 (21.9%) 840 414 1,002 1,995 (19.4%) 1,572 (17.0%) 684 552 336 1,080 (11.7%) 852 228

Total*

4,994 (100%) 5,827 (100%) 10,308 (100%)

Source: European Commission (1994: pp. 214–216) Note * Total of the main technological ﬁelds.

Knowledge spillovers in EU FWPs 245 decreased steadily from a peak of 45.6 per cent in the second FWP to 33 per cent in the fourth FWP. The focus in the ICT budget shifted from strengthening international competitiveness of European IT ﬁrms (ﬁrst FWP) to technological support for the development of ICT infrastructure, with an increased budget for telematics applications (European Commission 1997: 506). Energy showed the sharpest decrease over the FWPs, with its budget share dropping from almost 50 per cent in the ﬁrst FWP to 21.9 per cent in the fourth FWP. Life Sciences showed the sharpest increase in budget share, which, apart from the emergence of biotechnology as an important research ﬁeld, also reﬂects the link with the EU policy regarding Agriculture, Health and Nutrition. The budget share for environment increased from 6.6 per cent in the second FWP to 11.7 per cent in the fourth FWP, although the EU claims that almost all technological areas include environmental aspects (for example, encouraging environment-friendly technologies) (European Commission 1997: 507). In recent FWPs, efforts are directed towards more international cooperation, dissemination and the valorization of project results and human capital, training and mobility. The FWP objectives are ﬁtted into a more general socio-economic policy framework with, for example, attention on the user-friendliness of ICT and on sustainable growth (European Commission 1997: 507–509). Furthermore, in the fourth FWP, efforts were made to link EU RTD policy with regional development policies and with the general aim of socio-economic cohesion. The objective of reducing the scientiﬁc and technological gap between member states gained priority (Georghiou 2001: 893). These policy shifts resulted in signiﬁcant changes in the structure of the ﬁfth FWP (1998–2002). This programme adopted a problem-solving approach and was structured differently. It consists of four focused Thematic programmes and three wide-ranging Horizontal programmes. The former cover a series of well-deﬁned problems with a clear European focus, while the latter respond to common needs across all research areas including policies for external relations, small and medium-sized enterprises (SMEs), human resources and social and employment issues. This structure is a blurring of the two previously distinct indirect actions of EU policy, while direct actions are still performed through the JRC. Regarding budget allocations, ICTs (through the Information Society Technologies (IST) Programme) still receive the major proportion of EU funding (Table 11.1b). The sixth FWP still under way when this investigation was conducted (2003–06) has adopted a totally different structure. Major innovations include the use of two basic instruments for implementing the research activities: Networks of Excellence (NoE) and Integrated Projects (IP). The overall budget of the sixth FWP is C17.5 billion representing an increase of 17 per cent from the ﬁfth FWP and accounting for 3.9 per cent of the

246 M. Dumont and A. Tsakanikas Table 11.1b Funding of the ﬁfth FWP (in million C) Budget (in million C) RTD Thematic programmes Quality of life and management of living resources User-friendly information society (IST) Competitive and sustainable growth Energy, environment and sustainable development Total

2,413 3,600 2,705 2,125 10,843

Horizontal programmes Conﬁrming the international role of Community research Promotion of innovation and encouragement of SME participation Improving human research potential and the socioeconomic knowledge base

1,280

Total

2,118

475 363

Source: www.cordis.lu/fp5.

Table 11.1c Indicative funding of the priority key areas of the sixth FWP (in million C) Thematic priorities

Budget

Life science, genomics and biotechnology for health Information society technologies Nanotechnologies and nanosciences, multi-functional materials, production processes Aeronautics and space Food safety and quality Sustainable development, global change and ecosystems Citizens and governance in a knowledge-based society

2,255 3,625

20.0 32.1

1,300 1,075 685 2,120 225

11.5 9.5 6.1 18.8 2.0

11,285

100.0

Total

%

Source: www.cordis.lu/fp6.

Union’s total budget (2001) and 6 per cent of the Union’s public research budget. Seven key areas for the advancement of knowledge and technological progress were chosen and are being funded by more than C11 billion.5 As Table 11.1c shows, IT still ranks ﬁrst in terms of funding with 32 per cent of the total budget for the seven areas being allocated to it.

11.3 Data This chapter draws information from the EU-RJV database, a new and extensive data source that has been developed by the Laboratory of Industrial and Energy Economics (LIEE) at the National Technical University

Knowledge spillovers in EU FWPs 247 of Athens (NTUA). The speciﬁc database contains information on crossnational R&D collaboration established in Europe through the ﬁrst four FWPs, covering up to 64 programmes.6 Programmes that have a focus on industrial research and are aimed at the creation of new technological knowledge are included in the database, while those that simply provide support (for example, disseminate information) or concern other EU actions were excluded. Furthermore, at the project level, only R&D consortia with at least one ﬁrm participant were included.7 The dataset that is used in this chapter includes 9,335 research consortia that were formed between 1984 and 1998. Almost 20,500 organizations from 53 countries account for 64,476 participations in these collaborations. However, the real value-adding aspect of this dataset is the conjunction of this information with ﬁrm level data. More speciﬁcally, drawing on AMADEUS (see note 5), data for 2,607 European ﬁrms (almost 40 per cent of the ﬁrm participants) were retrieved, including sectoral information that is basically used in the subjoined analysis. Table 11.2 shows the number of projects classiﬁed in the various technology areas of the four FWPs. The table therefore reﬂects the relative diversiﬁcation of policy priorities over the FWPs. The area of information processing and information systems is always in ﬁrst place in terms of funding, although in terms of projects it dropped from ﬁrst to fourth place in the last FWP. Projects in aerospace technology more than doubled in each successive FWP, and aerospace technology is the area with the highest number of projects in the fourth FWP. Materials technology and industrial manufacture maintain a relatively stable and strong position in all FWPs. For environmental protection the number of projects tripled in each consecutive FWP, reaching seventh position in the fourth FWP, while biotechnology also underwent a sharp increase in the fourth FWP following considerable decline in the third FWP. It should be mentioned that projects in the ﬁelds of biotechnology– biomedicine, environmental protection and energy, promote relatively more basic research than those in other ﬁelds and are thus characterized by a high degree of cooperation between universities and research institutes and low involvement of industrial partners. This explains their low inclusion rate in the database. On the other hand, programmes in the ﬁelds of ICTs, materials technology and industrial manufacture entail more applied research and therefore involve ﬁrm participants to a greater extent. However, it seems that in the fourth FWP the relative involvement of industrial partners in more basic research programmes like BIOMED, BIOTECH and ENV increased, leading to an overall increase in the projects included in the database. In Table 11.3, all NACE sectors appearing in the dataset are classiﬁed by the number of ﬁrms’ membership in the FWPs. The ﬁrst ﬁve sectors represent over 40 per cent of all ﬁrm membership. Furthermore, sectors 34 (motor vehicles), 35 (other transport equipment), 73 (R&D) and 64

331 279 206 186 177 163 162 160 160 143 143 136 107 77 76 76 64 57 48 40 34 10 4

Aerospace technology Renewable sources of energy Other technology Telecommunications Measurement methods Biotechnology Waste management Reference materials Standards Energy saving Fossil fuels Other energy topics Radioactive waste Agriculture Safety Transport Resources of the sea, ﬁsheries Nuclear ﬁssion Environmental protection Medicine, health Food Meteorology Life sciences

435.0 95.7 173.0 560.8 47.8 144.3 72.5 35.2 35.2 65.8 65.8 29.9 49.4 97.3 69.1 83.3 53.8 32.3 49.9 19.0 16.9 11.8 2.1

2,686.9 2,629.7 623.8 602.7 Electronics, microelectronics Energy saving Fossil fuels Renewable sources of energy Telecommunications Resources of the sea, ﬁsheries Agriculture Food Environmental protection Meteorology Safety Standards Transport Measurement methods Reference materials Medicine, health Biotechnology Regional development

Information processing, information systems Aerospace technology Industrial manufacture Materials technology 601 286 286 286 276 218 184 184 159 125 125 70 59 57 57 38 36 16

812 633 604 604

Number of projects

1,753.0 154.4 154.4 154.4 784.5 201.4 158.2 158.2 114.4 71.3 71.3 60.0 110.9 21.7 21.7 78.7 40.9 11.8

2,038.1 701.4 656.2 656.2

1,453

2,358.9 1,746.9 1,274.7 747.5 612.3 458.1 612.0 411.8 412.6 188.8 188.8 188.8 280.8 173.9 284.2 80.3

1,293.2 1,293.2

1,293.2

Number Funding of (in projects million C)

Industrial manufacture 1,453 Materials technology 1,453 Information processing, information systems 1,417 Electronics, microelectronics 986 Telecommunications 583 Environmental protection 581 Agriculture 514 Safety 445 Education, training 431 Biotechnology 420 Resources of the sea, ﬁsheries 325 Measurement methods 242 Reference materials 242 Standards 242 Food 240 Transport 223 Meteorology 222 Medicine, health 146

Aerospace technology

Funding Fourth FWP (in million C)

Note a As EU classiﬁes projects in two or three technological areas there is an unavoidable double counting, which explains why the total number of projects here exceeds the total number of projects available in the database.

Source: Adapted from EU- RJV database.

800 717 669 607

Number Funding Third FWP of (in projects million C)

Information processing, information systems Electronics, microelectronics Materials technology Industrial manufacture

First and second FWP

Table 11.2 Number of projects in different technological FWP ﬁeldsa

Radio, television and communication equipment and apparatus Other business activities Computer and related services Other transport equipment Chemicals Motor vehicles and trailers Machinery and equipment n.e.c. Ofﬁce machinery and computers Research and development Medical, precision and optical instruments; watches and clocks Wholesale trade (except motor vehicles) Electricity, gas, steam and water supply Electrical machinery and apparatus n.e.c. Post and telecommunications Basic metals Fabricated metal products (except machinery and equipment) Construction Non-metallic mineral products Food and beverages Textiles Sale, repair and maintenace of motor vehicles Cokes, reﬁned petroleum products and nuclear fuel Land transport Recreational, cultural and sporting activities Supporting and auxiliary transport activities Retail trade Manufacture of rubber and plastic products Extraction of crude petroleum and natural gas Publishing, printing, reproduction of recorded media Other mining and quarrying Other

32 74 72 35 24 34 29 30 73 33 51 40 31 64 27 28 45 26 15 17 50 23 60 92 63 52 25 11 22 14

Totals

Description

NACE

Table 11.3 Sectors’ participation in FWPs

12,178

1,387 1,133 925 911 787 704 679 504 499 482 462 434 423 414 295 273 207 161 149 140 131 124 107 91 85 74 71 64 59 52 351

Number of memberships

100.0

11.4 9.3 7.6 7.5 6.5 5.8 5.6 4.1 4.1 4.0 3.8 3.6 3.5 3.4 2.4 2.2 1.7 1.3 1.2 1.1 1.1 1.0 0.9 0.7 0.7 0.6 0.6 0.5 0.5 0.4 2.9

%

2,607

152 261 142 66 175 57 231 54 38 139 184 57 82 33 109 106 64 75 80 68 16 23 26 22 18 28 41 18 38 16 188

Number of ﬁrms

100.0

5.8 10.0 5.4 2.5 6.7 2.2 8.9 2.1 1.5 5.3 7.1 2.2 3.1 1.3 4.2 4.1 2.5 2.9 3.1 2.6 0.6 0.9 1.0 0.8 0.7 1.1 1.6 0.7 1.5 0.6 7.2

%

4.7

9.1 4.3 6.5 13.8 4.5 12.4 2.9 9.3 13.1 3.5 2.5 7.6 5.2 12.5 2.7 2.6 3.2 2.1 1.9 2.1 8.2 5.4 4.1 4.1 4.7 2.6 1.7 3.6 1.6 3.3 1.9

Average memberships per ﬁrm

250 M. Dumont and A. Tsakanikas (posts and telecommunications) have the highest average number of participations per ﬁrm. The strong position of sector 35 is explained by the large number of projects in aerospace technology (sub-sector 353 being the manufacture of aircraft/spacecraft).

11.4 The computation of spillovers within the FWPs Dumont and Tsakanikas (2001a) argued that a public policy that only draws conclusions from input/output-based methods which are often used to analyse (embodied) spillovers (for example, OECD 1999), would probably be more inward and backward looking than a policy that also considers indications of the direction in which the international technological space is evolving. Furthermore, such a policy may run the risk of promoting inefﬁcient and collusive lock-in situations. As input/outputbased methods are acknowledged to be rather unreliable for small open economies (De Bresson and Hu 1999), and international spillovers are often found to be more signiﬁcant for these types of countries, an approach that overcomes these shortcomings seems even more necessary. The method proposed here to estimate disembodied knowledge spillovers8 focuses on a speciﬁc spillover mechanism that seems to have been somewhat neglected, that is, cooperation in R&D. The reason for this neglect, despite the increasing cooperation in R&D, is probably that most scholars stick to the traditional strict deﬁnition of spillovers as externalities (for example, Grossman and Helpman 1992; Branstetter 1998). Some recent deﬁnitions, however, comprise the voluntary exchange of useful knowledge (Llerena and Matt 1999; Rycroft and Kash 1999; De Bondt 1999). If in cooperations between ﬁrms, the largest proportion of the knowledge ﬂows between partners can be regarded as intended, spillovers – sensu stricto – should be considered as an inevitable result of cooperation, as pointed out by Inkpen (1998). Inter-ﬁrm collaboration in R&D can be seen as an agreement in which mutual spillovers are, if not intended, at least implicitly accepted. Hagedoorn et al. (2000: 571) argue that cooperative agreements turn the hostage situation of a market transaction into a mutual hostage situation. More speciﬁcally, according to transaction cost economics, these agreements have a cost advantage over the market or hierarchical mode of operation due to the high cost of writing and enforcing speciﬁc contracts. It is exactly the type of research undertaken through the partnerships examined in this chapter (generic technologies, pre-competitive research) that is expected to suffer from severe appropriability problems (knowledge spillovers) (Caloghirou and Vonortas 2000). In this chapter we focus only on knowledge ﬂows between ﬁrms.9 In contrast to other methods of measuring disembodied spillovers (for example, Jaffe 1986, 1989; Branstetter 1998) we do not assume an unambiguous decreasing relationship between technological distance and

Knowledge spillovers in EU FWPs 251 spillovers. If technological proximity can be seen as a good proxy for absorptive capacity, then, following Katsoulacos and Ulph (1998), we consider the total spillover as the interaction of a knowledge ﬂow component with an absorptive capacity component. Collaboration involving knowledge sharing between close competitors is, for obvious reasons, more problematic than between more distant partners. Moreover, as pointed out by Katsoulacos and Ulph (1998), research by (technologically) close competitors may involve a high degree of overlap and therefore a lower degree of useful knowledge ﬂows – intended or not – than research efforts by partners that are more complementary. Verspagen (1997) ﬁnds signiﬁcant inter-sector spillovers and doubts whether the magnitude of spillovers is related to technological proximity. Luukkonen (2001) reports, for Finnish FWP participants, that vertical consortia had the lowest additionality, that is, resulted in little additional own R&D efforts. Kesteloot and Veugelers (1997) ﬁnd, for private R&D alliances, that the search for complementarity is more important than technological relatedness. Methods of similarity-based measurement in general can only result in symmetric spillovers. It is also not very clear which spillover mechanisms are implied, unless patents are considered as the main source of knowledge externalities. Although patents are supposed to allow ﬁrms to appropriate their research efforts, they might prefer other ways (for example, secrecy) of appropriation. If so, of course, externalities need not be linked to patent-based proximity. The main hypothesis for the analysis that follows, is that the number of cooperative links between ﬁrms is a proxy measure for the underlying knowledge ﬂows and thus for the spillovers stemming from these ﬂows, which are assumed to be proportionate to the knowledge ﬂows. Hagedoorn and Duysters (2000) argue forcefully that, from a learning perspective, multiple contacts in inter-ﬁrm networks will be more effective than pursuit of non-redundant contacts, dictated by strict maximizing efﬁciency rules. They found strong evidence that, in a dynamic environment, the absolute number of network links is more relevant than the network position (Hagedoorn and Duysters 2000: 23). We furthermore assume that in R&D projects more knowledge actually ﬂows from the partners to the prime contractor than vice versa, while ﬂows between other partners are assumed to be fairly balanced. A survey of Finnish FWP participants shows that project coordinators more often assessed a project as successful than did other participants (Luukkonen and Hälikkä 2000: 52). Research on Greek FWP ﬁrm participants has also shown that the project coordinator seems to beneﬁt from the collaboration more than the other partners, especially in terms of upgrading its knowledge base (Tsakanikas 2002).10 Finally, we assume knowledge ﬂows to be inversely related to the total number of participants in each project, to account for the fact that knowledge exchange is swifter in small rather than in large consortia. However,

252 M. Dumont and A. Tsakanikas Box 11.1 Mathematical deﬁnition of spillovers Spillover from sector j (all EU countries) to sector i in country c: SPc,ij = Σn Σk Σl≠k (Inkl PCc,nk)/NPn c: country i, j: sector n: project number k,l = 1 . . . NPn NPn: Number of participants in project n Inkl = 1 if ((actor k belongs to country c and sector i) and actor l belongs to sector j) and 0 if not

PCc,nk = 2 if actor k is prime contractor and 1 if not

in spite of the lower ﬂow awarded to each bilateral collaborative link, the sum of knowledge ﬂows in a project with n partners, for each single participant being given by (n 1)/n, projects with a large number of participants will have a higher total of knowledge spillovers. This accounts for the fact that large projects tend to be more important, both in ﬁnancial and strategic terms, than projects with a small number of participants. Box 11.1 summarizes the mathematical spillover deﬁnition. The analysis is performed in the 25 most active sectors, in terms of participations, representing almost 95 per cent of all FWP participations (see Table 11.3). The most commonly-used way to estimate international spillovers is to construct weighted foreign R&D stocks of all countries considered.11 Weights account for a speciﬁc spillover mechanism or proxy thereof. Coe and Helpman (1995), for instance, use bilateral import shares to account for the importance of international trade in enhancing spillovers. Regressing total factor productivity (TFP) on own R&D efforts and the weighted foreign R&D stock, Coe and Helpman (1995) found signiﬁcant spillover effects. Keller (1996) constructed R&D stocks using randomly computed bilateral import shares rather than actual shares. Estimations using these counterfactual shares as weights result in more signiﬁcant and higher spillovers than those estimated with actual shares. Keller (1998) rightly argued that this result casts some doubt on import-related spillovers.12 A more far-reaching conclusion from this result, however, would be that any inference from signiﬁcant spillover estimates about the impact of a given mechanism accounted for in the weights of the R&D stocks, may simply be spurious. Actual spillover effects cannot easily be distinguished from counterfactual effects when R&D expenditures are used. Dumont and Tsakanikas (2001a) used R&D stocks, computed from R&D expenditure data, to weight spillovers. Domestic sector R&D stocks were taken as a proxy for absorptive capacity (Cohen and Levinthal 1989) and foreign intra- and inter-sector R&D stocks as a proxy of the amount of knowledge that can spill over. However, here we preferred to run the

Knowledge spillovers in EU FWPs 253 analysis on unweighted spillovers. This was not primarily prompted by the aforementioned possible spurious effect of using R&D expenditures, but rather by a lack of those data for a relatively large number of countries, sectors and years. As our aim was to analyse the evolution of spillovers for the period 1984–98 for most EU countries and sectors, computing weighted spillovers would severely hamper this analysis. Certain EU countries, and all services sectors (which, as will be shown, are important knowledge suppliers in the cooperations examined), would simply have to be excluded from the analysis.13 Based on the computed 25 25 matrices, we derived clusters of the most important intra- and inter-sector spillovers, using cut-off criteria, similar to Hauknes (1999: 63–64). The ﬁrst cut-off (link strength) restricts the linkages on the basis of the fraction of the spillover from a sector to a given sector in the total spillover that the given sector receives. The second cut-off (signiﬁcant sectors) restricts the linkages on the basis of the fraction of spillover from a sector to a given sector in the total spillover for all sectors. Three degrees of linkage are considered. The strongest linkages are those representing more than 30 per cent of the total spillover ﬂowing from a supplier-sector (aggregated over all countries) to the given country recipient-sector and more than 2 per cent of the total spillover for the given country (SPij 30% of SPi AND SPij 0.02*Σj SPj). The weaker linkages represent, respectively, fractions of 20 per cent to 1 per cent (SPij 20% of SPi AND SPij 0.01*Σj SPj) and 10 per cent to 0.5 per cent (SPij 10% of SPi AND SPij 0.005*Σj SPj). Although all spillovers are computed bilaterally, albeit the distinction between project coordinators and other participants is not always symmetrical, the spillovers in the clusters are often unilateral (for example, strong spillover link from 64 to 72, but no signiﬁcant spillover from 72 to 64 in the cluster of the ﬁrst and second FWP). This results from the cut-off criteria which relate each bilateral spillover to the total spillovers of the respective sectors. So, elaborating the example in the ﬁrst and second FWP, sector 72 (computer and related services) receives a substantial part (more than 20 per cent) of its knowledge from sector 64 (post and telecommunications), whereas the latter sector receives less than 10 per cent of its knowledge from sector 72, or the total spillovers received by sector 64 are less than 0.5 per cent of total spillovers for all sectors. The cut-off criteria are obviously arbitrary, but yield convenient cluster representations and in addition the further analysis does not depend on these criteria. As the number of projects increased considerably in each consecutive FWP there is a signiﬁcant size difference. The ﬁrst FWP represents 793 project participations in our ﬁle, the second FWP 2,122, the third FWP 3,374 and the fourth 5,678 participations. To ensure some degree of comparability and to preclude a size bias, the ﬁrst and second FWPs are merged.

254 M. Dumont and A. Tsakanikas

11.5 The pattern of intra- and inter-sector spillovers Figure 11.1 shows the clusters of the strongest linkages, aggregated at EU level, in the three periods considered: FWP 1 + 2 (1984–91), FWP 3 (1990–94) and FWP 4 (1994–98).14 The supplying sectors are depicted in square boxes whereas the recipient-sectors are circled. The arrows run from the supplying to the recipient-sectors and are for the reasons mentioned above (Section 11.4) not necessarily symmetrical. Stronger spillover linkages are represented by dotted arrows, the weaker linkages by normal arrows and dashed arrows. Figure 11.1 gives only a ﬁrst graphical impression of the intra- and inter-sector linkages at EU level and to a large extent simply reﬂects (changes in) policy priorities. We will therefore not dwell upon the depicted clusters except to highlight a few marked features. The dominance of ICT is revealed by the central position of sectors 30 (ofﬁce machinery and computers), 32 (radio, television and communication equipment), 64 (post and telecommunications) and 72 (computer and related services). A number of ‘low-tech’ sectors like 15 (food and beverages), 17 (textiles), 27 (basic metals) and 28 (fabricated metal products) appear in the clusters, but apparently all of their linkages are intrasector. Pre-competitive R&D collaboration is clearly not limited to ‘high-tech’ sectors, but the amount of knowledge received by ‘low-tech’ sectors from other sectors as a result of the FWPs, seems to be small. In Dumont and Tsakanikas (2001b) low-tech sectors such as 15 (food and beverages) and 17 (textiles) were found to cooperate relatively highly with universities and research institutes whereas the level of cooperation of high-tech sectors such as 32 (electronic equipment) and 33 (instruments) with these research institutes was relatively small. The degree of intra-sector linkages slightly decreased from 27 per cent in the ﬁrst and second FWP to 24 per cent in the fourth. The percentage of domestic intra-sector linkages increased from 22 to 26. The reluctance of ﬁrms to cooperate with domestic competitors has apparently reduced over time. The convergence of both these percentages over time may be an indication of European integration having blurred the distinction between domestic and foreign competitors. Furthermore, the degree of intra-sector collaboration, in spite of some strong intra-sector linkages, seems relatively low at the two-digit analysis level. Katsoulacos and Ulph (1998) argue that subsidizing R&D for the purpose of information sharing probably results in little additional information sharing between less close competitors as, at least in theory, they would be inclined to share all their information anyhow. Since most of the collaboration in the FWPs is between ﬁrms from different industries these authors conclude that such subsidies probably have not provoked much additional effort or information sharing. From a ‘learning by networking’ perspective, this claim regarding the beneﬁts of collaboration is

32 FWP 12 27

30 31

27 35

29

35

29

30

33

72

32

72

34

73

74

74

28

24

24

28

34 64

64

73 31

72

74

40

72

74

40

FWP 3 30

31 32

30 64

32

51

33 64 15

15

17

17

24

24

50

35

29

35

35

34

34

FWP 4

63

34

35

29

29

29

74

27

27

60

60

64 72

64

40

72

40

74

31

34

50

32 24

73

32 33 24

Supplier sector (all countries)

User sector (country)

SPij 30% of SPi and SPij 0.02*Σj SPj

SPij 20% of SPi and SPij 0.01*Σj SPj

SPij 10% of SPi and SPij 0.005*Σj SPj

Figure 11.1 EU clusters of the highest intra- and inter-sector spillovers in three FWP periods.

256 M. Dumont and A. Tsakanikas perhaps overly pessimistic. It rests on the assumption that information sharing is independent of cooperation, that is, ﬁrms can fully share information without cooperating. However, if the tacit component of knowledge is important, cooperation could – in our view – result in knowledge ﬂows between partners that could never be attained between noncooperating ﬁrms.15

11.6 Analysis of FWP spillovers In this section a more dynamic analysis is performed. We avoid the somewhat tautological activity of simply representing budget shares in a number of different ways, by looking at the performance of individual countries in relation to the entire group of EU countries. We ﬁrst discuss a number of correlations between aspects of spillovers of individual countries and some country characteristics. Dumont and Tsakanikas (2001a) found a positive correlation between a country’s percentage of spillovers from domestic partners and country size (gross domestic product (GDP)). This correlation seems to have intensiﬁed over time. The correlation between countries’ percentage of spillovers from domestic partners, and countries’ respective sizes for the three FWP periods is 0.63, 0.75 and, again, 0.75.16 This highly signiﬁcant correlation supports the view that international cooperation is important to small countries in allowing ﬁrms from these countries to cooperate with foreign partners, thereby compensating for the lack of appropriate domestic partners. The FWPs may thus have been extremely useful for ﬁrms from small countries to establish contacts with foreign partners with complementary capabilities, which might never have happened without the FWPs. This contention is reinforced, for example, for Greek ﬁrms, establishing new relationships and accessing complementary resources are among the major objectives of participating in EU-FWPs (Tsakanikas 2002). Greece, Ireland, Portugal and Spain all have a low level of domestic linkages and thus are highly dependent on foreign partners for collaboration. This ﬁnding shows the increasing importance of the FWPs for participating ﬁrms located in the ‘cohesion countries’. Sharp (1998) pointed to the ‘small but steady rise in the shares of cohesion countries in partnerships’ (p. 583). There are also indications that the FWPs have been instrumental in organizations from lagging EU regions becoming, or becoming part of, European centres of excellence, although examples of these are relatively rare (Ireland, Lisbon and Attiki) and restricted to a small number of universities and research centres, with very few local ﬁrms beneﬁting from spillovers (European Commission 1997: 379–398). For a dynamic and more informative analysis of intra- and inter-sector spillovers we computed shift-share effects. A shift-share analysis is traditionally used to analyse the dynamics of export market shares (for example, Fagerberg and Sollie 1987), but here we consider the total

Knowledge spillovers in EU FWPs 257 Box 11.2 Shift-share analysis SSPc =Σs SPc,ij/Σc Σs SPc,ij =Σs((SPc,ij/Σc SPc,ij)(Σc SPc,ij/Σc Σs SPc,ij))=SSPc,ij SSPij c: country i, j: sector s: 1 . . . 25 (number of sectors) SSPc: Share of country c in total EU FWP spillover pool SPc,ij: Total of spillovers from sector j (all EU countries) towards sector i in country c SSPc,ij: Share of country c in spillovers from sector j (all EU countries) towards sector i SSPij: Share of spillovers from sector j towards sector i in total of linkages From SSPc = SSPc,ij SSPij we can decompose the change in the total share of country c (see e.g. Fagerberg and Sollie 1987): ∆SSPc = ∆(SSPc,ij SSPij) = ∆SSPc,ij SSPij + SSPc,ij ∆SSPij + ∆SSPc,ij ∆SSPij ∆SSPc,ij SSPij: Share effect (competitiveness measure) SSPc,ij ∆SSPij: Composition effect ∆SSPc,ij ∆SSPij: Adaptation effect (interaction effect)

spillover pool in a given FWP as a market. For each EU country (except Luxembourg) and each possible intra- or inter-sector spillover link, we compute a market share (SSPc,ij). The change in market share (between the ﬁrst and second FWPs and the third FWP and between the third and fourth FWPs) can be decomposed into three effects. The deﬁnition and decomposition of the share effects is presented in Box 11.2. The share effect (∆ SSPc,ij SSPij) measures the extent to which the total change in the country share can be explained by a change in the country share measured at a ﬁxed share composition at the EU level and can therefore be seen as a measure of competitiveness,17 as it reveals how the country share has increased (or decreased) at the expense (to the beneﬁt) of the other EU countries. The composition effect measures the part of the total share explained by a change in the composition of the EU matrix, keeping the individual shares ﬁxed (SSPc,ij ∆ SSPij). Finally, the adaptation effect, which is an interaction effect ∆ SSPc,ij ∆ SSPij, reﬂects the proportion of total change that is due to a move towards or away from links that become more or less important at the EU level. The analysis shows the degree to which the share of a given EU country in the ‘market of FWP cooperation’ increased or decreased and whether this can be explained by the competitiveness, the specialization and the adaptability of intra- and inter-sector linkages. Overall, sectors 30, 32 and 72 show the sharpest decline, obviously as a result of the diversiﬁcation of policy priorities in the third FWP compared to the ﬁrst and second FWP; but whereas sector 32 recovers in the fourth FWP, the position of sectors 30 and 72 decline further. Transport sectors 34, 35 and 63 (supporting

258 M. Dumont and A. Tsakanikas and auxiliary transport activities) and energy sector 40 sharply increased their market shares from the third FWP on. Table 11.4 shows the overall shift-share effects of individual EU countries. Comparing the change of total shares between the third FWP and the period of the ﬁrst two FWPs, shows that large countries like France, Germany and the UK suffered the sharpest decline in their shares due to a large extent to a negative share effect, and small countries like Austria, Denmark and Sweden had the highest increase, again explained to a large extent by a (positive) share effect. In the fourth FWP large countries like Italy and Germany lost share, while small countries like Sweden, Finland and Austria gained share. The sharp gain in the shares of the latter group should not come as a surprise, given that they only ofﬁcially joined the EU in 1995 (that is, during the fourth FWP). A somewhat more surprising ﬁnding is the signiﬁcant positive correlation of 0.51 (two-tailed signiﬁcance = 0.062) between a country’s GDP per capita growth in the period 1992–96 on the one hand, and a country’s total FWP market share growth between the ﬁrst period (1984–91) and the third FWP (1990–94) on the other. However, although this result is an interesting avenue for future research, due to the crudeness of the link, we would not point to this as evidence of a positive impact of FWP participation on economic growth. Table 11.5 presents the top ﬁve most-positive and the top ﬁve mostnegative total shift-share effects for individual EU countries. This list shows which intra- and inter-sector spillovers contributed most to the overall change in the share of a given country. Belgium, Germany and Ireland are the only countries for which the share decreased over all FWPs (Table 11.4). Table 11.5 shows that for Germany and Ireland this is, to a large extent, due to the tenacity of sector 30 (ofﬁce machinery and computers), which lost ground at the EU level. Van Essen and Verspagen (1997) also found that Germany and Belgium suffered the sharpest decline in their high-tech market shares in the period 1988–94. Furthermore, the shares of Austria, Finland, Greece, the Netherlands, Spain and Sweden increased in all periods due to their strong positions in sectors 24 (chemicals), 32 (radio, television and communication equipment), 40 (electricity, gas, steam, water supply) and 72 (computer and related services).

11.7 Conclusions In this chapter, the method proposed in Dumont and Tsakanikas (2001a) for computing knowledge spillovers stemming from inter-ﬁrm R&D cooperation was applied to analyse the changing pattern of intra- and intersector collaboration in the ‘pre-competitive’ EU-FWPs. The evolution reﬂects not only the shift in policy priorities of the EU, but also other possible shifts in the European techno-economic space.

Austria Belgium Denmark Finland France Germany Greece Italy The Netherlands Portugal Spain Sweden The UK Ireland

0.00713 0.01271 0.01566 0.00097 0.02174 0.00661 0.00310 0.02358 0.00023 0.00260 0.00127 0.00548 0.04390 0.00001

Share

0.00000 0.00030 0.00004 0.00001 0.00225 0.00106 0.00005 0.00071 0.00003 0.00001 0.00016 0.00004 0.00095 0.00003

Composition

Table 11.4 Country shift-share effects

0.00056 0.01292 0.00133 0.00022 0.03901 0.00514 0.00162 0.02275 0.00284 0.00227 0.00146 0.00009 0.00285 0.00025

Adaptation

0.00765 0.00051 0.01694 0.00073 0.06300 0.00253 0.00144 0.00013 0.00258 0.00488 0.00257 0.00561 0.04769 0.00029

Total

FWP3– FWP12

0.01491 0.00699 0.00862 0.01352 0.00468 0.01478 0.00257 0.02932 0.00773 0.00365 0.00680 0.01730 0.00528 0.00008

Share

0.00000 0.00000 0.00002 0.00000 0.00015 0.00008 0.00000 0.00004 0.00002 0.00000 0.00000 0.00000 0.00009 0.00000

Composition

0.00374 0.00043 0.00167 0.00114 0.01207 0.00583 0.00064 0.00343 0.00421 0.00120 0.00009 0.00002 0.00472 0.00035

Adaptation

0.01108 0.00656 0.01031 0.01466 0.00754 0.02069 0.00193 0.03278 0.00350 0.00485 0.00689 0.01732 0.00992 0.00044

Total

FWP4– FWP3

51 51 51 32 74

15 34 51 35 32

30 74 72 32 30

74 74 74 51 31

27 33 34 72 34

15 34 72 34 92

72 24 30 32 30

NL 40 51 (+) 27 72 31

()

(+)

FR

()

29 73 32 26 74

FWP123

AT 34 32 (+) 32 26 32

FWP123

24 34 27 29 72

74 15 30 72 34

35 32 34 63 72

30 29 32 26 32

32 33 72 32 40

FWP34

24 34 28 28 74

74 15 30 30 34

35 32 35 35 32

73 45 74 26 73

32 32 32 29 32

FWP34

PT

DE

BE

24 32 64 24 24

74 32 64 72 30

34 35 34 40 24

74 30 45 74 34

64 51 51 74 31

FWP123

40 64 64 24 51

24 30 32 30 30

34 34 35 40 24

27 32 72 73 72

32 32 72 32 29

FWP123

40 26 74 63 72

35 32 29 17 34

35 32 32 50 60

73 51 30 32 17

34 35 27 64 31

FWP34

Table 11.5 Top ﬁve of highest and lowest total shift-share effects at country level

40 45 29 51 40

34 31 29 17 34

35 35 74 34 60

32 32 72 32 17

72 35 27 64 32

FWP34

ES

GR

DK

40 29 40 32 34

72 26 45 24 35

32 72 29 72 30

33 24 35 35 30

31 29 40 40 29

FWP123

40 40 74 73 35

33 51 26 27 35

72 32 27 72 72

35 28 33 28 72

31 31 40 28 26

FWP123

74 74 32 74 33

35 32 17 72 29

64 32 64 40 30

40 40 31 17 29

35 64 24 73 74

FWP34

74 72 72 26 28

72 72 17 72 27

64 32 51 72 32

28 40 31 17 31

35 64 24 24 35

FWP34

24 34 72 40 24

30 64 64 32 30

29 31 32 29 31

72 32 72 32 72

UK 24 34 (+) 72 40 40

32 32 72 32 30

40 34 29 35 74

64 30 32 32 30

()

(+)

IT

()

()

29 28 74 74 28

72 29 45 74 28

17 72 30 32 72

35 64 64 24 32

72 30 32 72 34

35 35 34 40 74

31 40 17 27 28

24 30 72 32 72

35 45 72 24 72

30 30 92 72 34

63 35 63 40 29

31 29 17 74 74

SE

IE

29 74 35 35 34

40 32 72 72 35

33 40 74 72 30

35 51 34 64 30

51 30 72 51 45

34 34 34 74 34

40 32 64 74 35

30 30 30 30 30

29 74 72 30 72

17 33 33 32 24

73 27 64 72 40

35 34 45 32 35

40 74 34 51 35

72 72 64 24 74

24 64 32 24 24

32 29 32 64 40

34 34 74 32 72

30 30 72 74 29

30 72 72 15 63

24 64 64 40 51 FI

30 51 72 64 32

24 45 29 40 74

24 72 64 72 32

33 33 33 33 34

24 24 40 23 23

35 64 64 45 64

31 74 40 29 45

64 24 24 72 64

34 64 30 32 40

40 27 23 40 24

51 24 23 33 64

35 64 40 64 40

262 M. Dumont and A. Tsakanikas Despite the diversiﬁcation of technology ﬁelds, ICT remained the dominant area over all FWPs, with a budget share of still more than 30 per cent in the fourth FWP. Sector 30 (ofﬁce machinery and computers) and sector 72 (computer and related services) have steadily lost ground, indicating a shift away from hardware and to a lesser extent from software to equipment. Some low-tech industries like 15 (food and beverages), 17 (textiles), 27 (basic metals) and 28 (fabricated metal products) have strong spillovers, but apparently most of the FWP linkages are intra-sector, indicating that precompetitive R&D collaboration is not restricted to high-tech sectors and also that these low-tech sectors are extremely self-reliant and do not receive much knowledge through the FWPs from other sectors. It is notable that ﬁrms from small countries rely more on foreign partners than ﬁrms from larger countries, reinforcing the importance of networking for small (cohesion) countries, to make up for the insufﬁcient scale and/or the lack of appropriate domestic partners. We here applied a more informative shift-share analysis, which, by comparing the performance of individual countries with the performance of the entire group, avoids complex representation of something that can be gleaned simply by looking at the FWP budgets. We found a surprisingly signiﬁcant positive correlation between the total change in the share of a country in the FWP market over the period 1984 to 1994 and in GDP per capita growth over the period 1992 to 1996, but the crudeness of this link constrains us from drawing any conclusion other than that it is deserving of further consideration. We believe that the method proposed to measure knowledge spillovers, following a limited number of assumptions and focusing on a speciﬁc and important spillover mechanism, could be useful in the empirical analysis of spillovers in general and could be instrumental in the assessment of a public policy for which, at least implicitly, the creation of spillovers is an important rationale. Our main argument is that cooperation should not be seen simply as a channel for the transfer of codiﬁed information: it is also, if not predominantly, a mechanism for sharing complementary tacit knowledge. As such, it may be more effective than market transactions and more ﬂexible than full internalization through mergers and acquisitions. The analysis described in this chapter only concerns funded, technologically targeted, pre-competitive R&D collaboration, which is of course just one type of inter-ﬁrm collaboration. These partnerships do not necessarily represent the preferred choice of the ﬁrms. The innovation process entails far more than this rather speciﬁc topic and, therefore, conclusions cannot easily be extrapolated. We have brieﬂy discussed problems of data limitations, the possible spurious effects of using R&D stocks and different lag structures that may thwart the estimation of spillover effects and assessment of the importance of different spillover mechanisms. Furthermore, apart from the linkages created and the spillovers that are generated, ﬁrms need to develop proper absorptive capacity (Cohen and

Knowledge spillovers in EU FWPs 263 Levinthal 1989, 1990) in order to fully beneﬁt from this effect. This is why further research based on ﬁrm level data could link the ‘external’ knowledge transfer phenomenon captured by the spillover ﬂow component with the ‘internal’ processes that each ﬁrm uses in order to absorb and use this knowledge. Further research could also include comparison between cooperation within the FWPs and more market-oriented cooperation (for example, EUREKA or private technology alliances) to establish the impact of a public policy promoting R&D collaboration on private spontaneous R&D networking. Spillovers computed on these extended data on R&D collaboration could then be compared more satisfactorily with other mechanisms. It would also be interesting to analyse the importance of geographic and cultural distance between partners for the sharing of (tacit) knowledge through R&D networking, following for example, Caniels (2000) by replacing patents with collaborative links as a measure of knowledge ﬂows.

Appendix Table A11.1 EU Framework Programmes included in the EU-RJV database Programme acronym

FWP

ACTS AERO 0C AERO 1C AGRIRES 3C AIM 1 AIM 2 AIR BAP BCR 4 BIOMED 1 BIOMED 2 BIOTECH 1 BIOTECH 2 BRIDGE BRITE BRITE/EURAM 1 BRITE/EURAM 2 BRITE/EURAM 3 CAMAR CLIMAT 3C CRAFT DECOM 2C DECOM 3C DRIVE 1 DRIVE 2 ECLAIR ENNONUC 3C ENS ENV 1C ENV 2C EPOCH ESPRIT 1

Fourth FWP Second FWP Third FWP First FWP Second FWP Third FWP Third FWP First FWP Second FWP Third FWP Fourth FWP Third FWP Fourth FWP Second FWP First FWP Second FWP Third FWP Fourth FWP Second FWP First FWP Third FWP First FWP Second FWP Second FWP Third FWP Second FWP First FWP Third FWP Third FWP Fourth FWP Second FWP First FWP

Number of projects

154 28 34 113 43 44 436 366 265 274 674 156 492 97 219 378 472 2,058 80 108 539 74 73 69 66 42 789 14 560 715 34 241

Budget (million C)

671 35 53 50 20 97 377 75 59 151 374 186 596 100 185 500 770 1,833 55 17 57 12 32 60 124 80 175 41 319 914 40 750

Average funding per project

Number of included projects

4.36 1.25 1.56 0.44 0.47 2.20 0.86 0.20 0.22 0.55 0.55 1.19 1.21 1.03 0.84 1.32 1.63 0.89 0.69 0.16 0.11 0.16 0.43 0.87 1.88 1.90 0.22 2.95 0.57 1.28 1.18 3.11

152 28 29 1 36 35 184 69 160 3 146 33 274 49 206 303 388 1,453 21 0 216 6 31 67 59 41 136 13 125 222 10 234 continued

264 M. Dumont and A. Tsakanikas Table A11.1 continued Programme acronym

FWP

ESPRIT 2 ESPRIT 3 ESPRIT 4 EURAM EURET FAIR FAR FLAIR FOREST HYMGEN C JOULE 1 JOULE 2 LIBRARIES LRE MAST 1 MAST 2 MAST 3 MAT MATREC C MHR 4C NNE-JOULE C ORA RACE 1 RACE 2 RADWASTOM 3C RADWASTOM 4C RAWMAT 3C REWARD SMT TELEMAN TELEMATICS 2C TRANSPORT

Second FWP Third FWP Fourth FWP First FWP Second FWP Fourth FWP Second FWP Second FWP Second FWP Second FWP Second FWP Third FWP Third FWP Third FWP Second FWP Third FWP Fourth FWP Third FWP Second FWP Second FWP Fourth FWP Third FWP Second FWP Third FWP First FWP Second FWP First FWP Second FWP Fourth FWP Second FWP Fourth FWP Fourth FWP

Totals

Number of projects

Budget (million C)

Average funding per project

435 605 1,599 87 9 632 127 34 38 29 267 401 51 25 48 93 157 178 71 211 577 19 94 123 217 121 236 13 394 21 641 336

1,600 1,532 2,084 30 25 740 30 25 12 15 122 217 23 23 50 118 243 67 45 65

3.68 2.53 1.30 0.34 2.78 1.17 0.24 0.74 0.32 0.52 0.46 0.54 0.44 0.90 1.04 1.27 1.55 0.38 0.63 0.31

14 550 554 62 80 70 6 307 19 913 263

0.74 5.85 4.50 0.29 0.66 0.30 0.46 0.78 0.90 1.42 0.78

17,596

18,710

Number of included projects 380 483 834 62 9 240 16 17 14 4 143 286 35 18 48 34 85 57 67 0 475 16 83 118 30 40 84 11 242 20 431 223 9,335

Source: Adapted from CORDIS, CD-ROM (1999 III).

Notes 1 The authors would like to thank the members of the OECD Focus Group on Innovative Firms and Networks for their valuable feedback at the workshops in Rome and Vienna. 2 A detailed methodological discussion and the ﬁrst application of this method can be found in Dumont and Tsakanikas (2001a). 3 The so-called ‘big 12’ being Bull (FR), CGE (FR), Thomson (FR), AEG (DE), Nixdorf (DE), Siemens (DE), Olivetti (IT), STET (IT), Philips (NL), GEC (UK), ICL (UK) and Plessey (UK). 4 Shared cost projects refer to subsidy from the Commission of 50 per cent of the total costs of joint research for companies and up to 100 per cent of marginal and additional costs for universities and research institutes. 5 The Programme is made up of three main blocks of activities: (1) Focusing and Integrating European Research (Area) – (ERA); (2) Structuring the ERA; and (3) Strengthening the Foundations of the ERA. The main bulk of the Programme is the ﬁrst block, which includes the seven key areas mentioned. However, explaining the full details of the Programme goes beyond the needs of this chapter (see www.cordis.lu./fp6 for further information).

Knowledge spillovers in EU FWPs 265 6 The EU-RJV database is part of the STEP TO RJVs Databank which includes several databases with detailed information on collaborative R&D in Europe. It was constructed in the context of an EU funded TSER project entitled ‘Science and Technology Policies Towards Research Joint Ventures’. The primary source of information for the EU-RJV database was the CORDIS database provided by the European Commission (1999) and AMADEUS, a commercial database, provided by Bureau Van Dijk, which contains ﬁnancial and sectoral ﬁrm level data on approximately 200,000 European ﬁrms. 7 A complete list of the Programmes, the total number of projects and their budgets along with the number of projects that meet the criterion of at least one ﬁrm participant and are thus included in the database, is shown in the appendix. It was decided to exclude organizations, which, due to the poor quality of information, could not straightforwardly be identiﬁed as ﬁrms. 8 Griliches (1992) explains the distinction between embodied and disembodied R&D spillovers. Spillovers are embodied if they relate to the purchase of equipment or goods/services (that is, rent spillovers). Disembodied spillovers, which he assumes to be more important than embodied spillovers, are deﬁned as ideas resulting from the research of one actor, used by another actor. 9 The important academia–industry linkages within the FWPs are analysed, to some extent, in Dumont and Tsakanikas (2001b). 10 Mothe and Quelin (1999) reached the same conclusion. 11 See Branstetter (1998) for a recent review of empirical literature on international spillovers. 12 Keller (2001) reviews some comments on his counterfactual work and refers to recent evidence on the importance of trade for international technology diffusion. 13 Future research could involve using data on R&D expenditures to weight spillovers as well as to compute foreign R&D stocks using bilateral FWP collaboration shares as weights. We could then compare the results of regression estimations to estimations of other mechanisms, keeping in mind that the different mechanisms may have a different lag structure (for example, spillovers enhanced by trade compared to spillovers from pre-competitive R&D collaboration). 14 Cluster representations at the level of individual EU countries for the three periods considered are available from the authors. 15 Geographically close ﬁrms may also beneﬁt from considerable knowledge (tacit) spillovers, even if they have no formal collaboration agreements. The (micro-economic) evidence on localized spillovers (for example, Jaffe et al. 1993; Branstetter 1998) suggests that this may indeed be the case. Thompson and Fox-Kean (2001) argue, using ﬁner technological (patent) classiﬁcations than Jaffe et al. (1993), that the evidence for geographical knowledge spillovers may be spurious. 16 Two-tailed signiﬁcance is, respectively, 0.016, 0.002 and 0.002. 17 In our case, this means that we keep the weights of each intra- and inter-sector link in the EU matrix (spillovers aggregated over all EU countries) ﬁxed at the start of the period value.

References Branstetter, L.G. (1998) ‘Looking for international knowledges: a review of the literature with suggestions for new approaches’, Annales d’Economie et de Statistique, 49–50: 517–540. Caloghirou, Y.D. and Vonortas, N. (2000) Science and Technology Policies Towards Research Joint Ventures, DGXII, Final Report, European Commission TSER Programme, EC: Brussels.

266 M. Dumont and A. Tsakanikas Caniels, M. (2000) Knowledge Spillovers and Economic Growth: Regional Growth Differentials Across Europe, Cheltenham: Edward Elgar. Coe, D.T. and Helpman, E. (1995) ‘International R&D spillovers’, European Economic Review, 39: 859–887. Cohen, W.M. and Levinthal, D.A. (1989) ‘Innovations and learning: the two faces of R&D’, The Economic Journal, 99(September): 569–596. Cohen, W.M. and Levinthal, D.A. (1990) ‘Absorptive capacity: a new perspective on learning and innovation’, Administrative Science Quarterly, 35: 128–152. De Bondt, R. (1999) ‘Spillovers and innovative activities’, International Journal of Industrial Organization, 15: 1–28. De Bresson, C. and Hu, X. (1999) ‘Identifying clusters of innovative activities: a new approach and a tool box’, in T. Roelandt and P. Den Hertog (eds) Cluster Analysis and Cluster-Based Policy Making in OECD Countries, Paris: OECD, pp. 27–59. Dumont, M. and Tsakanikas, A. (2001a) ‘Knowledge spillovers through R&D networking’, in: OECD (ed.) Innovative Firm Networks, Paris: OECD, pp. 209–231. Dumont, M. and Tsakanikas, A. (2001b) ‘Knowledge spillovers through R&D networking’, paper presented at the OECD Innovative Firms and Networks Focus Group meeting, Vienna, 8–9 March. European Commission (1994) The European Report on S&T Indicators 1994, DG XII, Luxembourg: Ofﬁce for the Ofﬁcial Publications of the European Commission. European Commission (1997) The Second European Report on S&T Indicators 1997, DG XII, Luxembourg: Ofﬁce for the Ofﬁcial Publications of the European Commission. European Commission (1999) CORDIS – Community R&D Information Services, Brussels: European Commission. Fagerberg, J. and Sollie, G. (1987) ‘The method of constant market shares analysis reconsidered’, Applied Economics, 19: 1571–1583. Georghiou, L. (2001) ‘Evolving frameworks for European collaboration in research and technology’, Research Policy, 30(6): 891–903. Griliches, Z. (1992) ‘The search for R&D spillovers’, Scandinavian Journal of Economics, 94(Supplement): 29–47. Grossman, G.M. and Helpman, E. (1992) Innovation and Growth in the Global Economy, Cambridge, MA: MIT Press. Hagedoorn, J. and Duysters, G. (2000) Learning in Dynamic Inter-ﬁrm Networks – The Efﬁcacy of Multiple Contacts, Research memorandum 2000/9, Maastricht: MERIT. Hagedoorn, J., Link, A.N. and Vonortas, N. (2000) ‘Research partnerships’, Research Policy, 29(4–5): 567–586. Hauknes, J. (1999) ‘Norwegian input–output clusters and innovation patterns’, in OECD (ed.) Boosting Innovation: The Cluster Approach, Paris: OECD, pp. 61–90. Inkpen, A.C. (1998) ‘Learning and knowledge acquisition through international strategic alliances’, Academy of Management Executive, 12(4): 69–80. Jaffe, A.B. (1986) ‘Technological opportunity and spillovers of R&D: evidence from ﬁrms’ patents, proﬁts and market value’, American Economic Review, 76(5): 984–1001. Jaffe, A.B. (1989) ‘Characterizing the “technological position” of ﬁrms, with application to quantifying technological opportunity and research spillovers’, Research Policy, 18: 87–97. Jaffe, A.B. (1998) ‘The importance of “spillovers” in the policy mission of the Advanced Technology Program’, Journal of Technology Transfer, 23: 11–19.

Knowledge spillovers in EU FWPs 267 Jaffe, A., Trajtenberg, M. and Henderson, R. (1993) ‘Geographic localisation of knowledge spillovers as evidenced by patent citations’, Quarterly Journal of Economics, 108(3): 577–598. Katsoulacos, Y. and Ulph, D. (1998) ‘Endogenous spillovers and the performance of research joint ventures’, Journal of Industrial Economics, 46(3): 333–359. Keller, W. (1996) ‘Are international R&D spillovers trade-related? Analyzing spillovers among randomly matched trade partners’, SSRI Working Paper No. 9607, Madison: University of Wisconsin. Keller, W. (1998) ‘Are international R&D spillovers trade-related? Analysing spillovers among randomly matched trade partners’, European Economic Review, 42: 1469–1481. Keller, W. (2001) International Technology Diffusion, NBER Working Paper No. 8573, Cambridge, MA: NBER. Kesteloot, K. and Veugelers, R. (1997) ‘Stable R&D cooperation between asymmetric partners’, in S. Poyago-Theotoky (ed.) R&D Cooperation: Theory and Evidence, London: Macmillan Press, pp. 97–125. Llerena, P. and Matt, M. (1999) ‘Inter-organizational collaboration: the theories and their policy implications’, in A. Gambardella and F. Malerba (eds) The Organization of Economic Innovation in Europe, Cambridge: Cambridge University Press, pp. 179–201. Lucchini, N. (1998) ‘European technology policy and R&D consortia: the case of semiconductors’, International Journal of Technology Management, 15(6/7): 542–553. Luukkonen, T. (2001) ‘Networking impacts of the EU Framework Programme’, mimeo, TEKES, Helsinki. Luukkonen, T. and Hälikkä, S. (2000) Knowledge Creation and Knowledge Diffusion Networks, Helsinki: TEKES. Molina, A.H. (1996) ‘Innovation in the context of European R&D collaborative Programmes: the case of multimedia and the newspaper industry’, International Journal of Technology Management, 12(3): 271–281. Mothe, C. and Quelin, B. (1999) ‘Creating new resources through European R&D partnerships’, Technology Analysis and Strategic Management, 11(1): 31–44. OECD (1999) Boosting Innovation: The Cluster Approach, Paris: OECD. Peterson, J. and Sharp, M. (1998) Technology Policy in the European Union, London: Macmillan Press. Rycroft, R.W. and Kash, D.E. (1999) Innovation Policy for Complex Technologies, Issues in Science and Technology Online, Fall 1999, Available at: www.nap.edu/issues/ 16.1/rycroft.htm. Sharp, M. (1998) ‘Competitiveness and cohesion: are the two compatible?’, Research Policy, 27(6): 569–588. Thompson, P. and Fox-Kean, M. (2001) ‘An improved method of measuring localization of knowledge spillovers by patent citations’, Carnegie Mellon Working Paper. Tsakanikas, A. (2002) ‘Corporate strategy and research joint ventures’, PhD Thesis, Laboratory of Industrial and Energy Economics, National Technical University of Athens. Van Essen, M. and Verspagen, B. (1997) The Dutch Technology Position: The Specialization in High-Tech Trade, Research Report, Maastricht: MERIT–Dutch Ministry of Economic Affairs. Verspagen, B. (1997) ‘Estimating international technology spillovers using technology ﬂow matrices’, Weltwirtschaftliches Archiv, 133, pp. 226–248.

12 Unveiling the texture of a European Research Area Emergence of oligarchic networks under EU Framework Programmes* Stefano Breschi and Lucia Cusmano 12.1 Introduction This chapter1 provides a contribution to the recent debate about targets and effectiveness of network policies at the European Union (EU) level by presenting an analysis of the research and development (R&D) network that has emerged over Framework Programmes (FWPs). Social network analysis is employed to describe the structural properties and dynamics of the emerging network, which appears to be dense and pervasive, branching around a large ‘oligarchic core’, whose centrality and connectivity have strengthened over programmes. In this chapter we discuss the degree to which this network structure may respond to EU broad policy objectives and its implications for recent programmes aimed at shaping a European Research Area (ERA). Attention is paid to the late focus by European institutions on networking centres of excellence. Since future initiatives are to build on the existing fabric, we argue that understanding how networks formed and evolved following previous stimuli is of great relevance for implementing and assessing the impact of the newly deﬁned network approach. Over the last couple of decades, the promotion of consortia involving ﬁrms, universities, research centres and public agencies has gained centre stage in science and technology policy (STP) in Europe. Cooperative programmes, in the form of shared-cost R&D consortia, have become the most important source of EU funding and institutional support for innovation, international competitiveness and, by way of knowledge exchange and diffusion, intra-European cohesion. Extensive support for research networks dates back to the early 1980s when cooperative initiatives at the continental level represented a response to the decline of the European industry competitiveness vis à vis US and Japanese companies and to the weakness of national champion policies. Accordingly, the ﬁrst cooperative programmes concerned those ﬁelds such as information and communication technology (ICT) where

Unveiling the texture of an ERA 269 the European innovative gap was perceived to be large and widening. However, with the full institutionalization of FWPs, the EU’s medium-term planning instrument for research and technological development (RTD), the cooperative approach has gradually been extended to a wide range of industries and institutions. The aim of fostering competitiveness in ‘high tech’ ﬁelds, by pooling the most advanced resources and capabilities, has been accompanied by the other major European objective: ‘cohesion’, that is, the integration of national research communities and the linking of marginal actors to the main component of the EU R&D network. Cooperative policies have certainly been pervasive and effective in aggregating public and private institutions from national research communities. However, concerns have been expressed about their effectiveness in raising the level of innovative investments, supporting European competitiveness and providing an efﬁcient mechanism for creating a critical mass of knowledge and competencies whose beneﬁts may extend to the laggards. Following recent political debate and the challenges posed by future enlargement, the implementation of cooperative policies by way of widespread support for a large variety of projects and institutions is to undergo signiﬁcant changes. The European Commission has called for a change in approach that responds to the need to reinvigorate the European research infrastructure and reﬂects the most recent theoretical and empirical debate about R&D networks. Starting with the sixth FWP (2002–06), policy actions have been more focused on identifying crucial ‘nodes’ and networking ‘centres of excellence’, that would represent the backbone of a truly ERA and act as catalysts for smaller components or backward areas. The aim in this chapter is to provide a contribution to the debate about targets and effectiveness of network policies at EU level by presenting a thorough analysis of the large R&D network that has emerged over successive FWPs. The concerns expressed by the European Commission (2000a) in its milestone communication Towards a European Research Area and the recent focus on networking centres of excellence appear to reﬂect dissatisfaction with the limitations of past cooperative policies in structuring a robust and efﬁcient knowledge and research network. However, little empirical research on the overall structure and evolution of European networks has yet been produced. We would argue that identiﬁcation and characterization of networks that have emerged from early European programmes represent a fundamental step for the assessment of past achievements and an important benchmark for future policy design. Indeed, a widespread and robust network, branching around a large ‘oligarchic core’, has already emerged as a more or less intended consequence of early FWPs. Since future initiatives are to build on the existing fabric of science and technology in Europe, understanding how networks formed and evolved following previous stimuli may be of great relevance for implementing and assessing the impact of the newly deﬁned network approach.

270 S. Breschi and L. Cusmano In this chapter, social network analysis and graph theory are employed to describe the structural properties and dynamics of the EU-wide network stemming from the R&D consortia promoted under the third and fourth FWPs. The analysis provides empirically grounded elements for discussing the degree to which this network structure may respond to the EU’s broad policy objectives of competitiveness and cohesion and its implications for recent programmes aimed at shaping the ERA. The chapter is organized as follows. Section 12.2 discusses the aims and articulation of EU FWPs, commenting on the recent debate about refocusing the whole European technology policy, to better coordinate the efforts of the EU with national strategies and create a truly European research area. Section 12.3 provides a description of the EU RJV dataset, while Section 12.4 discusses some methodological issues arising in the attempt to analyse the R&D consortia by means of network analysis. Section 12.5 provides an analysis of the research joint venture (RJV) network and Section 12.6 offers a summary and concluding remarks in relation to the debate about the creation of a European Research Area.

12.2 The cooperative approach of Framework Programmes: changing priorities towards a European research area The Single European Act and the Maastricht Treaty gave full and clear competence to European Institutions in the area of RTD, although the ﬁrst genuinely collaborative and EU-wide initiatives, such as ESPRIT, date back to the early 1980s. The setting up of FWPs, which gave greater coherence to existing initiatives, was intended to strengthen the scientiﬁc and technological basis of European industry and to encourage it to become more competitive at the international level. When the ﬁrst FWP was launched in 1984 the ‘technology gap’, which was perceived to be of the greatest relevance in explaining declining European competitiveness, was the main concern driving policy action. Along the lines of the Single Market approach, the focus of European RTD policies was primarily directed towards overcoming the fragmented national structure of European industry and markets, permitting economies of scale that could not be achieved at national level. Accordingly, a preference was indicated for research conducted on a vast scale, projects addressing common interests that could best be tackled through a joint effort, research contributing to the cohesion of the common market promoting the setting of uniform laws and standards (European Commission 1994). FWPs have provided a systematic procedure for discussing and agreeing upon priorities, guidelines and budget allocation, and have become the Commission’s main instrument for offering selective support to European companies seeking to undertake collaborative R&D with ﬁrms or research institutes in other European countries. In fact, RTD policy has been

Unveiling the texture of an ERA 271 implemented mostly by supporting shared-cost contractual research, that is, multinational consortia (or RJVs) grouping ﬁrms, public agencies, research centres and universities focused, in principle, on pre-competitive research projects.2 The pre-competitive requirement is designed to avoid potential conﬂict with EU competition policy which forbids collaboration at the stage of developing products for an immediate market, whereas Articles 85 and 86 of the Treaty of Rome allow collaboration for precompetitive research (Ants and Lee 1997). The collaborative approach has been extended to structure new areas of Union intervention as the budget of FWPs has increased and priorities changed. If competitiveness has remained the primary goal of European technology policy (Peterson and Sharp 1998), a wider set of objectives has been pursued over time, as the communitarian approach to RTD gained new momentum with the Maastrich Treaty. Since the third FWP (1990–94), EU RTD policy has been recognized as playing an important, though complementary role to other EU policies, in reducing unemployment, accelerating structural change and, at the same time, ensuring greater cohesion. From this perspective, the latest programmes have shifted the emphasis from supply-side factors, central to the design of the ﬁrst policies, to diffusion-oriented projects and greater learning skills and increased knowledge among Europeans. After two decades of active policy making the fundamental role of European institutions in promoting scientiﬁc advance and technological innovation, and the centrality of the collaborative approach have been fully acknowledged. However, the degree to which the existing instruments and schemes of intervention can meet the ever more diversiﬁed objectives and the challenges posed by future enlargement has been closely questioned. Indeed, the current debate about the future of European technology programmes is characterized by the concern that Europe might not successfully achieve the transition to a knowledge-based economy. The communication of the European Commission Towards a European Research Area (European Commission 2000a) started a broad discussion on the development of research cooperation in Europe, centred around two main concerns: the persistent lower level of innovative investment in Europe compared to the US and Japan, and the fragmentation of research efforts, which EU investments often exacerbate, creating an ineffective ‘15 + 1’ static conﬁguration. The fragmentation, isolation and compartmentalization of national research systems and the disparity of regulatory and administrative frameworks further worsen the effect of the low investment in RTD, limiting European capacity to produce knowledge and ability to innovate. Hence, a more concerted effort for the development of a real European Research Area has been called for by the Commission as an urgent priority within the EU agenda. The collaboration networks promoted by FWPs, and the indirect forms

272 S. Breschi and L. Cusmano of cooperation to which they have given rise, represent a considerable achievement, but in their current form they do not appear to be sufﬁcient to correct the structural weaknesses of European research or to be a viable instrument for an enlarged Union. According to the Commission, for cooperative efforts to produce long-lasting effects, they should aim primarily at changing the organization of research in Europe, rather than simply adding resources and facilities. The focus of European programmes is thus set to change, from direct support for a large variety of projects and organizations, which often overlap national incentive schemes without forming a coherent whole, to a more limited number of priorities and measures that exert a ‘co-ordinating, structuring and integrating’ effect on European research (European Commission 2000b). Investing in infrastructures, strengthening relations between existing organizations and programmes, improving conditions for political consultation, establishing a common system of scientiﬁc and technical reference and promoting greater mobility of researchers represent the priorities of the newly-designed European technology policy. Interventions in these areas are meant to create favourable conditions for greater public and private RTD investments and development of a ‘critical mass’ in major research ﬁelds. In other words, action at the European level is justiﬁed and aims to create a more supportive and coherent European framework for research. From this perspective, policy intervention is primarily directed at stimulating resources and expertise to converge in strategic research areas; self-organization of ‘clusters of excellence’ is expected to follow and will be supported. Within priority areas, clearly deﬁned scientiﬁc and technological objectives are to be pursued by mobilizing a critical mass of activities and resources, and allowing greater ﬂexibility than in previous programmes for the allocation of resources and management of speciﬁc research activities. To this end, starting from the sixth FWP, a differentiated range of instruments that reﬂect a ‘variable geometry’ approach has been set in place. Among these, networks of excellence are to play a prominent part in overcoming the fragmentation of the European research system and strengthening the European position in speciﬁc research areas. The premise of the European Commission is that world class centres of excellence already exist in Europe in a wide range of research ﬁelds. However, they are often scattered and only loosely connected, and their areas of expertise are not always widely known about across Europe, especially by ﬁrms that could usefully join forces with them. The integration of these centres into long-term R&D consortia, ﬁnancially supported by the European Union and focused on leading-edge research, would contribute to enhancing the European position in strategic ﬁelds, attracting new resources and expertise and, more particularly, restructuring the way research is carried out in Europe, favouring the development of an overall more collaborative attitude by public and private actors (European Commission 2000b).

Unveiling the texture of an ERA 273 According to the agenda set out by the Commission, the ﬁrst step towards networking the critical mass of resources and expertise that excel in speciﬁc research ﬁelds is mapping these centres of excellence, in accordance with transparent and competitive mechanisms. The selected institutions are then required to adopt a joint work programme in a ﬁeld representing a substantial proportion of their activities, in which their expertise is complementary, and develop interactive working methods, including staff exchange and intensive use of electronic networks. Consortia are therefore expected to carry out a ‘cluster’ of integrated projects in basic or generic research areas, possibly of a long-term and multidisciplinary character. The creation of these networks is to be supported with European ﬁnancing, but their activities should not become dependent on this support. In fact, the European funding is meant to complement resources deployed by the participants and should take the form of a ﬁxed grant for integration. Compared to previous programmes, consortia will enjoy greater freedom in managing their projects, and the follow-up by the Commission services will move from detailed monitoring of inputs to a more strategic monitoring of outputs (European Commission 2000a, 2000b). In other words, the European action is meant to provide a stimulus for centres of excellence to ‘cluster around’ common long-term precompetitive objectives, network on a permanent basis and self-organize the division of task and information ﬂows. The policy is clearly oriented towards the original EU objective of strengthening European competitiveness. However, speciﬁc requirements will be imposed on consortia in order that this policy also serves the objective of cohesion. These may include dissemination and communication activities, training of researchers and systematic networking efforts aimed at transferring knowledge to external teams. In other words, the dual role of technology policy that has emerged over FWPs will be maintained, although the role of European institutions appears to be one of guiding rather than leading the organization of research networks, which are expected to produce leading-edge knowledge and best practice and spread them to peripheral actors. Indeed, the impact on cohesion of this excellence policy has emerged as an important concern in the debate that followed the Commission’s proposal. The European Committee on Legal Affairs and the Internal Market (2000) underlined the risk of a concentration of facilities to the detriment of peripheral areas. The General Secretariat of the Conference of Peripheral Maritime Regions (CPMR) of Europe (2000) called for a greater emphasis on the integration of peripheral regions into niches of excellence in a number of highly specialized ﬁelds by way of partnerships between top-level researchers working in peripheral areas and identiﬁed centres of excellence. However, the more controversial among those issued debated by

274 S. Breschi and L. Cusmano European institutions and Member States appears to be precisely the identiﬁcation or mapping of these centres of excellence. Both the Italian and Finnish governments, for instance, pointed to the need for selection to be based on open competition and periodic evaluation and renewal, in order to avoid the risk of a pre-determined, static conﬁguration. The Economic and Social Committee (2000) recommended that the number of these consortia be limited and, at least during a preliminary pilot phase, that their funding be restricted to a well-deﬁned period to prevent them becoming permanent institutions. The Committee of the Regions (2000) urged that excellence should be based more on knowledge and cooperation than on competition between geographic areas. Along the same lines, the European Science Foundation (2000) underlined the need for care in avoiding a selection based on a ‘juste retour’ to meet national sensibilities, thereby clearly separating the ‘excellence’ and ‘cohesion’ objectives. More generally, the Commission’s proposal has been welcomed by the scientiﬁc and industrial community as an attempt to set new priorities and stimulate integration for leading-edge research, but several institutions have expressed doubts about its implementation. The idea of ‘identifying’ centres of excellence according to top-down procedures has raised doubts and criticisms. According to the Academia Europaea (2000), dividing a priori the research community into various classes of excellence could be counterproductive. A more positive approach from the Commission would be to support virtual centres of excellence using broadband communication between units identiﬁed by national bodies and to let self-reinforcing mechanisms in the research community lead to the strengthening of intra-European networks. A similar position has been adopted by the Bundesverband der Deutschen Industrie (BDI) (2000), which deems special administration for networking centres of excellence to be neither sensible nor necessary, since European researchers generally are members of comprehensive networks with existing centres of competence. Indeed, the Commission’s proposal appears to acknowledge the existence of EU-wide networks and the importance of ﬂexible management, but hints at the loose connectedness of existing webs and lack of coordination, setting priorities and instruments for creating a more rational structure and orienting their focus to precise goals, so that dispersion of resources and duplication of effort can be reduced. However, European programmes themselves have contributed in the past to the creation of very dense and pervasive networks, whose structure and dynamics are likely to affect the response to the new policy. The networks of excellence approach aims at creating the backbone of an ERA by stimulating emergence and self-organization of dynamic consortia, but, we would argue, a very dense texture of direct and indirect linkages is already in place as a result of previous actions. Although policy design and instruments differ in part, we may expect these collaborative patterns to be replicated to some degree, or, at least to affect future structures and

Unveiling the texture of an ERA 275 dynamics. In this respect, understanding how networks formed and evolved following previous stimuli can provide useful guidelines for evaluating the impact of the newly-deﬁned network strategies. Excellence consortia are to emerge, or be selected, from an existing ‘fabric’, which will inﬂuence their ability to acquire and spread knowledge to and from other research nodes, within or outside the excellence core. In the following sections, we examine the network that formed under previous programmes, focusing on its topological features and evolution over time.

12.3 The EU RJV dataset This empirical section examines the network of RJVs funded by the European Commission in the time period 1992–96, within the third and the ﬁrst part of the fourth FWP.3 The dataset provides detailed information on 3,874 research projects and 9,816 organizations.4 The 3,874 projects are distributed across 30 technological programmes, with a clear preponderance of ICT-related technological areas. The two leading programmes – ESPRIT 3/4 and BRITE-EURAM 2/3 – account, respectively, for 23 per cent and 17 per cent of all projects, whereas the share of biotech and biomedical programmes (BIOMED 1/2 and BIOTECH 1/2) is relatively small (less than 5 per cent of all RJVs). Firms account for the majority (around 64 per cent) of all RJV members, while research and education organizations together account for about 21 per cent of RJV members. The average number of organizations per RJV project is 7.09 (Table 12.1). Around 41 per cent of all RJVs had less than ﬁve organizations and slightly more than 85 per cent of them had less than ten organizations. As far as the time-scale of projects is concerned, the majority of RJVs were conducted over the medium term: about 44 per cent of them lasted between 31 and 36 months, 31 per cent for less than two years and less than 13 per cent lasted more than three years. Looking at the frequency of participation in RJVs, the average number of projects per organization is 2.79 (Table 12.1). However, the variance across organizations is quite wide: 91 per cent of all organizations have participated in less than ﬁve RJVs and most (68 per cent) have been Table 12.1 The EU RJV dataset: summary statistics (standard deviations in parentheses) Third FWP

Fourth FWP

Total

Number of projects 2,131 1,743 3,874 Total number of organizations 6,291 5,335 9,816 Average organizations per project 7.10 (5.12) 7.08 (3.65) 7.09 (4.52) Average projects per organization 2.40 (5.17) 2.31 (4.51) 2.79 (7.46)

276 S. Breschi and L. Cusmano 70 60

Frequency of organizations

50 40 30 20 10 0 10 1

2

3

4

5 6 7 Number of RJV projects

8

9

10

>10

Figure 12.1 Frequency of organizations by number of participations in RJV projects.

occasional participants in the sense that they have been a member in only one RJV (Figure 12.1). According to these ﬁgures, for the majority of organizations membership of EU-supported R&D consortia is not a frequent event, even though there are a few organizations (mainly large ﬁrms and universities) that participate extensively and continuously in shared cost activities.5 This pattern partly reﬂects the broad ‘political’ role of FWPs, whose generic and horizontal objectives lead to rather frequent and intense participation by European technological leaders, which are represented in all the most important EU R&D consortia. This type of interpretation seems to be further corroborated by the fact that the so-called Prime Contractors (that is, the organizations that take the leading role within each R&D consortium and coordinate the activities of the participants) have, on average, participated in many more RJVs than Partners (Figure 12.2). Almost 80 per cent of all Partners have participated in only one RJV project, while the percentage is only 40 per cent in the case of Prime Contractors. Moreover, about 15 per cent of all Prime Contractors have participated in more than ten RJV projects. It is also important to note that, even though 1,495 Prime Contractors (39 per cent) have acted as Prime Contractor only once, 963 (46 per cent) of them have acted as Coordinator in three or more RJV projects, and

Unveiling the texture of an ERA 277 90 80 70 Prime Contractors (N=2,094) Partners (N=7,722)

Frequency

60 50 40 30 20 10 0 1

2

3

4

5

6

7

8

9

10

>10

Number of RJV projects

Figure 12.2 Frequency of RJV membership: Prime Contractors versus Partners. Note Prime Contractors are the organizations which have played the leading role in the coordination of at least one R&D project. Partners are the organizations which never assumed that role.

1,778 (85 per cent) have also participated as Partners in RJV consortia led by other Prime Contractors (Table 12.2). At the same time, it is also signiﬁcant that although the vast majority of Partners are occasional members of R&D consortia, those Partners that participate in more than one RJV project tend to collaborate with different Prime Contractors. For example, of the 989 Partners that have participated in two RJV projects, 91 per cent of them worked to a different Prime Contractor in each venture (Table 12.3). Hence, it appears that a signiﬁcant number of Partner organizations is not simply ancillary to leading institutions, but take advantage of these European projects to connect with different key actors. Overall, this evidence seems to suggest that different organizations play a fundamentally different role in the R&D network. On the one hand, organizations that participate only occasionally in R&D consortia contribute little or nothing to the networking activity taking place in the RJV network. On the other hand, most networking activity seems to occur among Prime Contractors, and among them and the non-occasional Partners. These two types of actors represent the backbone of the RJV network and it is on the analysis of the topological features of this network that the next section is focused.

278 S. Breschi and L. Cusmano Table 12.2 Number of partnerships between Prime Contractors Number of times each Prime Contractor has cooperated as a Partner with other Prime Contractors. Number of partnerships with other Prime Contractors

Number of organizations

Frequency of organizations

0 1 2 3 4 5 6–168

316 359 360 283 164 124 488

15.09 17.14 17.19 13.51 7.83 5.92 23.30

Total

2,094

100.00

Table 12.3 Collaboration with different Prime Contractors Number of different Prime Contractors with which each Partner collaborated as a function of the number of RJV participations (frequency distribution). Number of Prime Contractors

Number of RJV participations 2

1 2 3 4 5 6 7 8 9–43 Total

9.4 90.6

100

3 2.3 15.9 81.8

100

4 0.0 2.2 25.4 72.4

100

5 0.0 1.1 4.4 17.8 76.7

100

6–48 0.0 0.0 0.0 2.2 6.2 19.5 11.1 13.1 47.9 100

12.4 The RJV network as a bipartite graph The network formed by RJV projects and member organizations can be studied, using the tools of graph theory, as an afﬁliation network (or bipartite graph). An afﬁliation network is a network in which actors (that is, organizations) are joined together by common membership of groups (that is, RJV projects) of some kind. Afﬁliation networks can be represented as a graph consisting of two kinds of vertices, one representing the actors and the other the groups (see top part of Figure 12.3). In order to analyse the patterns of relations among actors, however, afﬁliation net-

Unveiling the texture of an ERA 279 1

A

B

2

C

D

3

E

A

F

4

G

H

I

J

K

H B

F I

C D E

G

K J

Figure 12.3 Bipartite graph of RJV projects and organizations. Top: bipartite graph of organizations (A to K) and projects (1 to 4), with lines linking each organization to the project in which it participated. Bottom: the one-mode projection of the same network onto just organizations.

works are often represented simply as unipartite (or one-mode) graphs of actors joined by undirected edges – two RJV members who participated in the same project, for example, being connected by an edge (see bottom part of Figure 12.3). In what follows, we analyse the unipartite graph of organizations involved in R&D consortia, although this representation may miss some relevant information. Before proceeding to analysis of this network, however, we need to discuss in detail a crucial problem arising in its construction. Unlike other afﬁliation networks that have been recently examined (for example, scientiﬁc co-authorship, company chief executives), the dataset used here contains some additional information about the role played in each RJV project by different organizations. More speciﬁcally, the EU RJV dataset allows us to identify for each R&D consortium the organization acting as Prime Contractor (that is, the coordinator of the activities undertaken within the project). In recognition of the importance of these organizations (see Section 12.2 above), and given the fact that it is likely that organizations involved in large R&D consortia know the coordinator better than they know each other, an alternative way of representing the unipartite graph is the ‘star’ network (Figure 12.4). According to this hypothesis, Prime Contractors would then act as intermediaries in the ﬂows of knowledge between partners in the same RJV and no direct edge would exist between partners. Both assumptions about the role played by Prime Contractors are, of course, rather strong and somewhat arbitrary. However, as they seem

280 S. Breschi and L. Cusmano B

F

B

F

D

G

D

G

(a) Unipartite graph as a (b) Unipartite graph as a ‘star’ completely connected network (Prime Contractor subgraph (no specific role as co-ordinating agent) for Prime Contractor)

Figure 12.4 Alternative representations of the unipartite graph.

equally reasonable, in the absence of other reasons to adopt either of the two, we will explore the main topological characteristics of the RJV network with respect to both, by referring to them, where appropriate, as hp. a) and hp. b), respectively.

12.5 Analysis of the RJV network In this section, we examine the main topological features of the unipartite graph of R&D consortia, by employing a number of indicators that have been widely applied to the study of a number of other networks (Newman 2001). The indicators are reported in Table 12.4 and results are shown for the cumulative RJV network up to the third FWP (1992–94) and for the cumulative RJV network up to the fourth FWP (1992–96). 12.5.1 Density of network and size of the giant component As far as the density of the network is concerned (that is, the ratio between the number of actual links and the maximum theoretical number of possible links), the ﬁrst result to note is that the value of the indicator differs under the two hypotheses adopted here. This is not surprising given the different way the network is constructed. Overall, the results indicate that the network is quite dense. On average, each RJV organization is directly linked to the other 21 (hp. a)) and the other four (hp. b)) organizations. Even more interestingly, the results also show that the density of the network is reducing over time, as more organizations join the network and new links are forged between existing organizations, and among them and new participants. A quite important result to note is that the network is highly connected. The largest component found in the network (that is, the largest connected subgraph of the network) ﬁlls the major proportion of the graph (96.3 per cent), all the other components are very small. The second largest component for example, contains only ten organizations (that is, ≈0.1 per cent of all organizations). This indicates that the vast majority of organizations involved in EU-sponsored programmes are

hp. a) hp. b) hp. a) hp. b) hp. a) hp. b) hp. a) hp. b)

hp. a) hp. b) hp. a) hp. b)

6,291 65,712 12,123 0.3321 0.0612 105 5,964 94.8 13 20.9 (40.6) 3.8 (8.9) 3.16 (0.44) 4.66 (0.72) 12 14 0.812 (0.29) 0.043 (0.16)

Third FWP (1992–94)

9,816 103,687 21,308 0.2152 0.0442 114 9,455 96.3 10 21.8 (49.1) 4.5 (11.9) 3.16 (0.39) 4.53 (0.67) 8 12 0.816 (0.28) 0.048 (0.16)

Third and fourth FWP (1992–96)

Notes The indicators marked with the label * have been calculated with reference to the largest component only. Numbers in parentheses are standard errors. The connectivity of the network, in terms of size and number of connected components, is clearly not affected by the hypothesis used to construct the graph.

Clustering coefﬁcient*

Maximum distance*

Average distance*

Second largest component Average degree*

Number of components Size of largest component as a percentage of all organizations

Density (100)

Number of nodes (organizations) Number of edges

Table 12.4 Summary of results of the analysis of the unipartite RJV network

282 S. Breschi and L. Cusmano directly or indirectly connected via collaboration, thus providing a ﬁrst indication of the effectiveness of such programmes in promoting the integration of the European R&D network. At the same time, it should also be noted that the size of this major component does not increase much from the third to the fourth FWP. In other words, the network seems to be highly connected from the outset.6 This might, of course, reﬂect the growth of a giant cluster of connected organizations that occurred during the ﬁrst and the second FWP, for which we do not have information. However, an alternative hypothesis is that the R&D network we are studying simply reﬂects a network of relationships that developed over time and well before the launch of EU-sponsored programmes. Given its size and presumably its importance, in what follows we restrict our analysis to the giant cluster. 12.5.2 Average degree and skewed degree distribution The average degree (that is, the number of other nodes to which a node is directly connected) of organizations in the RJV network is around 22 according to hp. a) and around 4.5 according to hp. b). The average degree also increases, albeit slightly, from the third to fourth FWP. A quantity of interest that has been studied recently for various networks is the degree distribution, P(k), giving the probability that a randomly selected node has k links. The distribution of degrees for the giant component of the RJV network is reported separately in Figure 12.5, for the two hypotheses used here. Once again, the distribution of degrees looks slightly different under the two hypotheses used here to construct the network (see top part of Figure 12.5). The distribution peaks around k = 6 under hp. a), whereas it peaks at k = 1 under hp. b). Moreover, the largest observed degree is higher in the former case (max k = 933), than in the latter (max k = 262). Apart from these rather obvious differences however, the distribution of degrees appears to be highly skewed in both cases. A very large number of organizations have a very small number of direct links with other organizations, but there is fat tail of organizations with a very large number of connections. Networks showing a skewed degree distribution are quite common, including the Internet, the world wide web (WWW) and scientiﬁc publications among others (Albert et al. 1999: 130–131). In general, the probability distribution of the number of links that connect a certain node in these networks decays following a power-law P(k) ~ k with scaling exponent in the range between 2 and 3 (Barabási and Albert 1999). The power-law implies that nodes with few links are the most numerous, but the probability of larger numbers of links falls off so gradually that nodes with several hundred links are to be expected. The power-law tail in the case of the RJV network is quite evident from the raw data (see top part of

Unveiling the texture of an ERA 283 Figure 12.5). However, because of the large variance in the downward tail of the distribution, a better estimate of the parameter is obtained by performing logarithmic binning of the data and normalizing data by bin width (see bottom part of Figure 12.5). Fitting the data in this way indicates that = 2.120 under hp. a) and = 2.032 under hp. b)7. This result indicates that the overall network connectivity is dominated by a few highly connected organizations. It is thus interesting to examine the process through which such a distribution is generated. A common assumption in this respect is that nodes link with higher probability to those nodes that already have a larger number of links, a phenomenon labelled ‘preferential attachment’ (Barabási et al. 2001). In other words, the probability with which a new node connects to an existing node is not uniform, but there is a higher probability that it will be linked to a vertex that already has a large number of connections. Highly connected nodes become more and more connected, thus generating a power-law distribution of degrees. In order to test this conjecture, we considered the distribution of degrees of organizations already in the RJV network in the period 1992–94, corresponding to the third FWP. Then we calculated the number of new links established in the period 1995–96, that is, during the fourth FWP. In the absence of preferential attachment, the probability that a link added at time t connects to an organization that has collaborated previously with k others is therefore given by nk(t)/N(t), where nk(t) is the number of organizations with degree k immediately before the addition of this link and N(t) is the total number of organizations in the network (Newman 2001). To the extent that the proportion of new links added to organizations with degree k exceeds nk(t)/N(t) and increases with k, the assumption of preferential attachment should ﬁnd support. Results for the RJV network seem to support the theory (see Figure 12.6). Organizations with a greater number of previous links tend to acquire a disproportionately higher number of new links. For example, organizations with numbers of previous links k < 10 accounted for about 87 per cent of all organizations existing in the third FWP and only 44 per cent of all new links added in the fourth FWP. On the other hand, organizations with numbers of previous links between k = 50 and k = 60 accounted for only about 0.2 per cent of all organizations existing in the third FWP, and accounted for more than 4 per cent of all the new links added in the fourth FWP. The explanation for the preferential attachment in the case of the RJV network must stem, at least in part, from the so-called Matthew effect (Merton 1968): institutions that are successful in getting funds for their research have a higher probability of producing exploitable research, which improves their probability of joining other projects (and therefore increase their number of links) in the future.8,9

103 (a)

P(k)

102

101

100 100

101

102

103

k

101

(b)

102

P(k)

103 4

10

P(k) = 25k2.120 R2 = 0.99

105 106 100

101

102

103

k

Figure 12.5 Degree distribution for the giant component of the RJV network. Note Histogram of the degree distribution for the giant component of the RJV network. a) Raw distribution under hp. a); b) Raw distribution under hp. b); c) Degree distribution with logarithmic binning under hp. a); d) Degree distribution with logarithmic binning under hp. b). All histograms shown in double-log scales.

(c)

P(k)

103

102

101

100 100

101

102 k

100 (d) 101

P(k)

102 103

P(k) = 1.472k2.032 R2 = 0.98

104 105 100

101

102 k

Figure 12.5 continued.

286 S. Breschi and L. Cusmano

Relative probability of new links

45 40 35 30 25 20 15 10 5 0

10

20

30

40

50

60

70

80

90

100

110

120

130

Number of previous links (k)

Figure 12.6 Preferential attachment in the RJV network. Note The relative probability of new links is deﬁned by the ratio between the proportion of new links added to organizations with k previous links, and the proportion of organizations with k previous links on all organizations existing in the network immediately before the addition of the link.

12.5.3 Average distance A frequently used measure to quantify the efﬁciency of a network in connecting different organizations and facilitating the ﬂows of information and knowledge is the so-called average distance. For any pair of nodes, i and j, in the network, their ability to communicate with each other depends on the length of the shortest path lij (that is, the minimum number of edges) which links them. The average over all pairs of nodes, denoted as d = , is called the average separation (distance) of the network, characterizing the network interconnectedness. In other words, the average distance measures the number of steps that have to be taken in order to connect two randomly selected nodes. An alternative measure that is also used to evaluate the degree of connectedness of a network is its diameter, which is deﬁned as the maximum separation of pairs of nodes in the network, namely the greatest distance one will ever have to go to connect two nodes together. Once more, it turns out that the values of the average distance and the diameter of the RJV network differ under the two hypotheses used to construct it (Table 12.4). Apart from this rather obvious result, however, the most interesting ﬁnding is that the average separation between pairs of organizations is relatively small. It takes an average of only 3.1 steps (4.7 under hp. b)) to reach one randomly chosen organization from any other. Signiﬁcantly, the observed value of the average distance is approximately

Unveiling the texture of an ERA 287 equal to (or even lower for hp. b) than) the value it would assume in a classical random graph – that is, a network in which nodes are connected to one another uniformly at random – with the same number of nodes and same average degree of nodes.10 The phenomenon of relatively short paths connecting randomly selected pairs of vertices has been noted in several other networks (Watts and Strogatz 1998), and has been labelled the ‘small world effect’. In the context of the RJV network, the small world effect apparently implies that the EU supported R&D consortia work as effective means of knowledge diffusion. 12.5.4 Clustering A further important characteristic of many real-world networks is that they are clustered, meaning that they possess local clusters of nodes in which a higher than average number of nodes are connected to one another. More precisely, a network shows clustering if the probability of a tie between two nodes is much greater if the two actors in question have one other mutual acquaintance, or several. Formally, one can deﬁne a clustering coefﬁcient C as follows: for any node i one picks the ki other nodes with which the node in question is linked. If these nodes are all connected to one another (that is, they form a fully connected clique), there will be ki(ki 1)/2 links between them, but in reality there will be much fewer. If one denotes with Ki the number of links that connect the selected ki nodes to each other, the clustering coefﬁcient for node i is then Ci = 2Ki/ki(ki 1). The clustering coefﬁcient for the whole network is obtained by averaging Ci over all nodes in the system. The clustering coefﬁcient C thus tells how many of a node’s collaborators are, on average, willing to collaborate with each other. Looking at the values of the clustering coefﬁcient in the RJV network, one observes that C takes extremely high values under hp. a), implying that, on average, about 82 per cent of the collaborators of a certain organization also collaborate with each other (Table 12.4). However, this rather extreme value partly reﬂects the bipartite nature of the graph (see Section 12.3 above). In the one mode projection of a bipartite graph in fact, cliques are automatically formed, thus contributing to increasing the value of the clustering coefﬁcient. For example, in Figure 12.4 above, all possible triangles among actors are formed, thus enhancing the value of the clustering coefﬁcient of each node. However, it is also clear that such a clustering is reﬂecting the fact that the four agents have all participated in the same project, rather than a true clustering. Adopting hp. b) allows us to disentangle true clustering from the trivial effects arising from the speciﬁc features of the bipartite graph. The value of C drops dramatically to 0.048, being rather stable over time. This implies that, on average, 5 per cent of the collaborators in a certain organization also collaborate with each other. This value is certainly lower

288 S. Breschi and L. Cusmano than the one found for the unipartite projection of the bipartite graph, but is still much higher than the value one would observe in a classical random graph.11 This result thus suggests that organizations involved in R&D consortia tend to introduce pairs of their collaborators to each other, engendering new collaborations. 12.5.5 Resilience of the RJV network The analysis carried out so far seems to indicate that the overall connectivity of the RJV network is dominated by a few important organizations. In order to test how the topological features of the network examined above depend on the activities of a few major actors, we have studied the changes in the diameter, average distance and size of the largest component when a small fraction f of nodes is removed. More precisely, we have compared how the quantities mentioned above vary, respectively, by removing the most important organizations in terms of number of connections (that is, degree k) and by removing the same fraction of organizations randomly (Figure 12.7). The results show that the network is extremely robust with respect to the random removal of a very high fraction of organizations. Thus, even when as many as 30 per cent of all organizations are removed from the network, the communication between the remaining nodes in the network is virtually unaffected. The largest component still accounts for more than 95 per cent of the remaining organizations, and the diameter and the average vertex–vertex distance remain stable. The reason for this result has to lie in the extremely skewed degree distribution (see above). This implies that the majority of organizations have only a few links and, therefore, organizations with small connectivity will be selected with much higher probability. The removal of these ‘small’ nodes does not alter the path structure of the remaining organizations, and thus has no impact on the overall network topology (Albert et al. 2000). On the other hand, the results also show that the RJV network is very ‘vulnerable’ with respect to the removal of the most connected nodes. When such nodes are eliminated, the diameter as well as the average distance among organizations increase rapidly, almost doubling in value when 4 per cent (430) of the most connected organizations are removed. In addition, the size of the largest component drops to less than 50 per cent of the remaining organizations when 5 per cent (488) of the most connected actors are removed from the network. Once again, the explanation for this result is related to the lack of homogeneity of the connectivity distribution: the connectivity of the network is maintained by a few highly connected organizations whose removal drastically alters the network’s topology, and decreases the ability of the remaining organizations to communicate with each other. In this context, of course, the term ‘removal’ must be taken with a grain of salt. The most important

80 70

Diameter

60 50 40 30 20 10 0 0.0

0.5

1.1

1.6

2.1

2.7

3.3

3.9

4.5

5.2

6.0

8.6

f (%)

Size of largest component

100

75

50

25

0 0.0 35.0

0.5

1.1

1.6

2.1

2.7

3.3

3.9

4.5

5.2

6.0

8.6

f (%)

Average distance

30.0 25.0 20.0 15.0 10.0 5.0 0.0 0.0 0.5 1.1 1.6 2.1 2.7 3.3 3.9 4.5 5.2 6.0 8.6 f (%)

Figure 12.7 Resilience of the RJV network. Notes Changes in the diameter, average distance and size of the largest component. a) Diameter; b) Size of largest component as a percentage of the network; c) Average distance. In all ﬁgures, diamonds show the response when the most connected nodes are removed, circles show the response when a fraction f of nodes are removed randomly.

290 S. Breschi and L. Cusmano message emerging from this analysis is that the overall connectivity of the RJV network is maintained by a few important organizations, whereas the vast majority of the other partners play no role in this respect. 12.5.6 Visualizing the RJV network Many of the results reported above could also be captured by visualizing the RJV network. It is almost impossible to draw a meaningful network of nodes and edges for the size considered here. Therefore, in order to provide a visual illustration of the network, we shrank the network by clustering some nodes, using the following procedure. As noted above, the majority of organizations participated in only one RJV project or in more than one RJV project, but collaborated with the same Prime Contractor. In the RJV network, there are 5,762 such organizations. These organizations contribute nothing to the overall connectivity of the network. Moreover, if we take all the organizations that collaborated with one, and only one, Prime Contractor these organizations can be considered as structurally equivalent, that is, as actors that have identical ties, so that it is possible to ‘collapse’ them into a single node. The remaining 3,693 organizations12 comprise Prime Contractors as well as Partners that have collaborated with more than one Prime Contractor.13 This subgraph comprises (by deﬁnition) a single component, that is, all organizations are directly or indirectly connected, and it can be considered as the core of the RJV network. In other words, if we remove these organizations the RJV network would break up into a very large number of disconnected components, whereas the removal of the other nodes would not affect the connectedness of the remaining nodes. Table 12.5 reports the indicators shown in Table 12.4 for the subgraph consisting of the 3,693 organizations, which represent the core of the network. It is quite important to note that the density of connections in this subgraph is remarkably higher than for the RJV network as a whole (compare with Table 12.4). This suggests, not only that the organizations in this subgraph account for the overall connectivity of the RJV network, but also that they maintain a dense web of relationships among each other. This is evident too from the value of the average vertex–vertex distance, which is much lower than for the network as a whole, and from the value of the clustering coefﬁcient, which is higher (under the hp. b)) than the value for the entire network. In order to visualize the network, the 3,693 organizations have been further grouped using the number of connections as a clustering principle. More speciﬁcally, we partitioned the subgraph of 3,693 organizations using the notion of k-core. A subset of vertices of the graph is called k-core or core of order k, if every vertex from the subset is connected to at least k vertices from the same subset. The notion of k-core, therefore, points to areas of the network where interaction among actors

Unveiling the texture of an ERA 291 Table 12.5 The core of RJV network The core of the RJV network comprises the Prime Contractors and the nonoccasional Partners (i.e. Partners that have collaborated with more than one Prime Contractor) belonging to the largest component. The subset thus deﬁned includes 3,693 organizations, 1,978 Prime Contractors and 1,715 Partners. Third and Fourth FWP (1992–96) Number of nodes (organizations) Number of edges Density (100) Average degree – All core Average degree – Prime Contractors Average degree – Partners Average distance Maximum distance Clustering coefﬁcient

hp. a) hp. b) hp. a) hp. b) hp. a) hp. b) hp. a) hp. b) hp. a) hp. b) hp. a) hp. b) hp. a) hp. b) hp. a) hp. b)

3,693 60,203 15,298 0.882 0.224 32.6 (57.5) 8.28 (15.5) 40.2 (74.3) 12.6 (20.0) 23.9 (24.5) 3.33 (2.8) 2.79 (0.37) 3.66 (0.50) 8 10 0.55 (0.27) 0.13 (0.25)

is particularly intense.14 In the network formed by the 3,693 organizations, there are 15 cores. For the sake of network visualization, we have grouped into a single node all organizations belonging to a k-core with k > 6. This node contains 674 organizations. Each organization included in this node was connected to at least seven of the other 673 organizations. As for the remaining 3,019 (that is, 3,693 674) organizations belonging to a k-core with k = 6, we further grouped them according to p-cliques. A p-clique consists of a subset of nodes such that each node has a proportion p of all its connections inside the clique. In order to group the organizations, we have set p = 1.15 The overall RJV network partitioned following the procedure described above is depicted in Figure 12.8. Visual inspection of the network illustrates many of the properties noted above. a

b

First, it is possible to appreciate the short chain of links that connect organizations in the RJV network. Starting from any node in the graph it takes a relatively small number of steps to reach any other node in the network. Second, the picture also gives an account of the highly skewed distribution of degrees. Most organizations involved in the R&D consortia are included in the ‘ring’ of nodes with only one link to the network.

292 S. Breschi and L. Cusmano

c

d

On the other hand, a handful of organizations have a very high number of connections, both with ‘occasional’ participants and among each other. Third, it is also quite clear why the connectivity of the RJV network is maintained by few highly connected organizations. For example, by removing the 674 organizations in the higher-order k-cores (box symbol in Figure 12.8), it is possible to see how the overall network breaks up into a myriad of smaller components, and how the average distance of the surviving largest component tends to increase. Fourth, the picture also allows us to appreciate that clustering tends to be more intense among the most connected organizations. In order to better evaluate this, we have visualized separately the subgraph formed by the 674 organizations included in the higher-order k-cores (Figure 12.9). One can observe the highly dense web of connections linking these organizations (density = 30 per cent). Moreover, the clustering coefﬁcient C for this subgraph takes also a very high value (0.17

Figure 12.8 Visual illustration of the RJV network. Notes The node marked with the box symbol contains 674 organizations belonging to a k-core with k > 6. The nodes marked with the diamond symbol contain 3,019 organizations. Each node is a p-clique, i.e. nodes included in it have 100 per cent of their connections within the clique (excluding links pointing externally). The nodes marked with circles contain 5,762 organizations. Each node contains all the organizations that participated in only one RJV project or in more than one RJV project, but always with the same Prime Contractor.

Unveiling the texture of an ERA 293

Figure 12.9 The k-core of the RJV network (k > 6, N = 674). Note The ﬁgure reports the internal connections among the 674 organizations belonging to the kcores with k > 6 (box symbol in Figure 12.8).

under hp. b), meaning that on average 17 per cent of the collaborators in a given organization also collaborated with each other). In other words, this subgraph of highly connected organizations also shows a very high degree of ‘transitivity’, that is, those collaborating with a certain organization tend also to collaborate with each other.

12.6 Implications of oligarchic integration for a European research area In the design of European institutions, an ERA should gradually emerge from the current fragmented and duplicative national innovation systems and supra-national programmes, once virtual mechanisms of coordination, networking and excellence promotion are set in place. Naturally, for collaborative endeavours to be effective, they must be heavily supported by measures that create ‘enabling conditions’, such as investment in research infrastructures, better coordinated implementation of national policies, development of a common system of scientiﬁc and technical reference, incentives for the training and mobility of researchers, support for risk capital investment.

294 S. Breschi and L. Cusmano Much importance is placed on the use of ‘variable geometry’ instruments, mentioned in the Treaty but seldom exploited, for gearing a critical mass of resources to strategic research ﬁelds, achieving world class excellence and spreading knowledge beyond the leading clusters to peripheral actors. A paramount instrument for achieving these ambitious goals is represented by networks of excellence, which would represent the backbone of a competitive ERA. Top-level resources will be encouraged to cluster, and leading institutions to tightly connect in order to carry out long-term integrated projects in particular ﬁelds of expertise, leaving much room for these networks to self-organize. The ideal position that emerges from the current proposals and debate is one in which centres of excellence cooperate intensively within stable consortia, which represent primary nodes linked directly to each other, forming a highly connected core, which serves the objective of enhancing the organization and quality of research and, ultimately, European competitiveness. The cohesion objective would be pursued by connecting peripheral institutions to the nodes of this core, which could therefore act as catalysts and sources of spillovers for smaller components or backward areas. In some sense, European technology policy should be oriented towards facilitating an ‘enlightened’ oligarchic integration, so that the best European talents and resources receive appropriate incentives and converge on issues of common interest, providing beneﬁts to the whole science and technology system. However, the empirical evidence we have presented in this chapter suggests that oligarchic networks have already emerged as a consequence of previous cooperative programmes. In particular, the network formed by RJVs promoted under the third and fourth FWPs appears to be fairly dense and extremely pervasive, while presenting hierarchical topological features. As shown in Section 12.5, organizations are, on average, connected by short chains of links, but this high connectivity is strongly dependent on a core of central actors that take part in a great number of projects, frequently acting as Prime Contractors. Around this pivotal group of very frequent participants who exhibit a high degree of intra-connectedness, we can identify two ‘lower’ layers. A minor group of fairly frequent but low-proﬁle participants, who enter consortia as Partners and take advantage of the programmes to establish links with several leading actors, and an extremely large number of Partners for whom participation is an exceptional event. In purely topological terms, this three-layer structure should ensure cohesion and efﬁcient transmission of knowledge since, even for the most peripheral agents, the network core, where most interaction, and possibly most knowledge production and information exchange take place, is only a few edges away. The small-world characteristics imply apparently that the European network is likely to diffuse knowledge in an efﬁcient manner. The core of the network carries the greatest interest for the researcher, as the amount and quality of effective knowledge production and trans-

Unveiling the texture of an ERA 295 mission within the overall network clearly depend on the resources deployed by the members of this core, by their expertise and by their degree of integration. In the present analysis, we have emphasized the vulnerability of the emergent ERA from this pivotal group and focused on its connectivity. The term ‘oligarchic’ conveys the idea of a leadership that emerged in the very early stages and has strengthened over time, such that removing those actors would cause the current conﬁguration to collapse and the ability of remaining organizations to communicate with each other to be dramatically decreased. Members of the core entered early RJVs as Prime Contractors and gained in terms of connectedness and centrality over time by way of repeated participation and preferential attachment. As we have noted, organizations with a larger number of previous links tend to acquire a disproportionately higher number of new links, particularly attracting peripheral actors. However, a high share of cooperative links is directed towards other members of the core, whose density is therefore remarkably higher than for the RJV network as a whole. This highly dense and connected core comprises industry and technology leaders and a signiﬁcant number of outstanding public research agencies and academic institutions. In short, it includes many entities that are expected to be natural candidates for the networks of excellence to come, in response to the call from the Commission for the creation of the excellence backbone of a European Research Area. To a certain degree, the newly designed policy is then going to replicate existing patterns of interaction, although it is intended to give to these consortia a more focused orientation on leading-edge research and less of a political role. Indeed, the current analysis has emphasized that connectedness is not the real problem. The limited effectiveness of previous programmes can hardly be related to lack of interactive opportunities for key actors. Rather, questions arise about the aims and content that have generally characterized consortia involving technology leaders. The dynamics that characterized the RJV network pose additional questions about the impact of policies that aim at introducing novelty to the way research is carried out in Europe. Self-reinforcing mechanisms and structural inertia appear in fact to have played an important role in determining the network conﬁguration. That is, the core of the network formed in the early stages and closed quite quickly around frequent Prime Contractors. Self-organization of networks could only strengthen this feature and imply risks of lock in. Late members would be unlikely to acquire hub roles and the network would soon become resistant to signiﬁcant changes in the structure of relationships that may, at times, reorient the network towards more productive research areas. This points to the need to carefully evaluate the role European and national institutions might still play in setting priorities and ‘creating’ room for new actors within the stratiﬁed core.

296 S. Breschi and L. Cusmano

12.7 Conclusions This chapter has provided a preliminary view on the emergent ERA, exploring topological features and dynamics of the network which stemmed from RJVs promoted under FWPs. From the empirical analysis, the emergence of a dense and hierarchical network can be inferred. A highly connected core of frequent participants, taking leading roles within consortia, is linked to a large number of peripheral actors, forming a giant component exhibiting the characteristics of a ‘small world’. We expect this conﬁguration to have important implications for the policy aimed at networking centres of excellence, partly because the same dynamics may replicate, and partly because the existing ‘fabric’ will exert a signiﬁcant inﬂuence on the creation of new network structures.

Notes * This paper was ﬁrst published in the International Journal of Technology Management, 27: 8, and the copyright belongs to Inderscience Enterprises Ltd. 1 The research for this chapter received ﬁnancial support within the EC TSER Programme’s KNOW FOR INNOVATION Project (Innovation-Related Knowledge Flows in European Industry: Extent, Mechanism, Implications), Contract No. SOE1-CT98-1118, DGXII G4, coodinated by Professor Y. Caloghirou of the National Technical University of Athens (Greece). An earlier version of the chapter was presented at the Athens ﬁnal workshop of the research group. The authors wish to thank Aldo Geuna, Bart Verspagen, Nicholas Vonortas and all participants at the meeting for useful comments. None of them are, of course, responsible for any possible errors. Umberto Fuso provided extremely valuable research assistance. 2 The composition of consortia is ruled by criteria which normally require the involvement of actors from at least two Member States and preferably located at different stages of the broadly deﬁned ‘production chain’. The EU covers up to 50 per cent of the costs of the projects, whereas members of the consortium share the burden of the remaining costs. Organizations which do not have cost accounting making it possible to reveal total costs, such as universities or colleges, receive up to 100 per cent of costs. 3 The dataset has been compiled by NTUA, Athens, based on the EU CORDIS database. We wish to thank Yannis Caloghirou for kindly allowing us to use the database. 4 The dataset examined contains information on all EU-sponsored RJVs which include at least one participant from the private sector. 5 More speciﬁcally, 39 organizations, including 13 ﬁrms and 17 universities, have participated in more than 50 but less than 100 RJVs; ten organizations (two ﬁrms, two universities and six research centres) have participated in more than a hundred R&D consortia. 6 For example, at the beginning of the third FWP in 1992, the largest component accounted for about 95 per cent of all organizations already in the network. 7 Quite notably, the value of the parameter is close to the value found for other networks. For example, Barabási and Albert (1999) found that the probability that k documents point to a certain WWW page follows a power law distribution with = 2.1.

Unveiling the texture of an ERA 297 8 Merton (1968) ﬁrst introduced the concept in relation to the allocation of credit for scientiﬁc work. The term refers to the Gospel according to St Matthew: ‘For unto every one that hath shall be given, and shall have abundance: but from him that hath not shall be taken away even that which he hath’. 9 Garcia-Fontes and Geuna (1999) found evidence of the Matthew effect when examining the BRITE-EURAM I and II programmes, as 13 per cent of all organizations involved in the programmes received 52 per cent of contracts. 10 The average vertex–vertex distance in a random graph with the same parameters as the RJV network would be equal to [log N/log z = 2.79], where N is the total number of organizations and z is the average degree of nodes (Newman et al. 2001). Under hp. b), the average distance in a random graph with similar parameters would be equal to about 6, which is remarkably higher than the observed value. 11 In a classical random graph, the clustering coefﬁcient is C = z/N where z is the average degree of nodes and N is the total number of nodes. In a random graph with the same parameters as the RJV network the C coefﬁcient would then be equal to 0.0047. 12 We are considering only the giant component of the RJV network, comprising 9,455 organizations. 13 Remember that we have deﬁned Prime Contractors as those organizations that have played that role at least once, and Partners as those organizations that have never acted as Coordinators, but have collaborated with more than one Prime Contractor. 14 Moreover, one important property of k-cores is that they are nested. For example, all nodes in the 2-core are included in the 1-core. 15 It must be noted that a p-clique with p < 1 is not necessarily connected. Setting p 1 simply means that a given clique will contain nodes that are connected only to other nodes inside the clique.

References Academia Europaea (2000) Towards a European Research Area: Responses to the Communication from the European Commission, London: Academia Europaea. Albert, R., Jeong, H. and Barabási, A.L. (1999) ‘Diameter of the World-Wide-Web’, Nature, 401: 130–131. Albert, R., Jeong, H. and Barabási, A.L. (2000) ‘Error and attack tolerance of complex networks’, Nature, 406: 378–381. Ants, M. and Lee, N. (eds) (1997) The Economics of the European Union: Policy and Analysis, Oxford: Oxford University Press. Barabási, A.L. and Albert, R. (1999) ‘Emergence of scaling in random networks’, Science, 286: 509–512. Barabási, A.L., Jeong, H., Néda, Z., Ravasz, E., Schubert, A. and Vicsek, T. (2001) ‘On the topology of the scientiﬁc collaboration networks’, Physics A, 311: 590–614. Bundesverband der Deutschen Industrie (BDI) (2000) BDI Position Paper on the Communication from the Commission ‘Towards a European Research Area’, Berlin: BDI. Committee of the Regions (2000) Opinion on the Communication from the Commission Entitled ‘Towards a European Research Area’, Brussels: Committee of the Regions. Conference of Peripheral Maritime Regions of Europe General Secretariat (2000) Technical Note on the European Commission’s Communication Entitled ‘Towards a European Research Area’, Rennes: CPMR General Secretariat.

298 S. Breschi and L. Cusmano Economic and Social Committee (2000) Opinion on the Communication from the Commission to the Council, the European Parliament, the Economic and Social Committee and the Committee of the Regions – Towards a European Research Area, Brussels: Economic and Social Committee. European Commission (1994) The European Report on Science and Technology Indicators I, Brussels: EC DG XIII. European Commission (2000a) Towards a European Research Area, Communication from the Commission to the Council, the European Parliament, The Economic and Social Committee, the Committee of the Regions, Brussels: EC. European Commission (2000b) Making a Reality of the European Research Area, Communication from the Commission to the Council, the European Parliament, the Economic and Social Committee, the Committee of the Regions, Brussels: EC. European Committee on Legal Affairs and the Internal Market (2000) Opinion on the Communication from the Commission Entitled ‘Towards a European Research Area’, Brussels: European Committee on Legal Affairs and the Internal Market. European Science Foundation (2000) Response to the Communication of the Commission of the European Communities ‘Towards a European Research Area’, Brussels: European Science Foundation. Garcia-Fontes, W. and Geuna, A. (1999) ‘The dynamics of research networks in Europe’, in A. Gambardella and F. Malerba (eds) The Organization of Innovation in Europe, Cambridge: Cambridge University Press. Merton, R.K. (1968) ‘The Matthew effect in science’, Science, 159: 56–63. Newman, M.E.J. (2001) ‘Clustering and preferential attachment in growing networks’, Physical Review E, 64(2), 025102 (2001). Newman, M.E.J., Strogatz, S.H. and Watts, D.J. (2001) ‘Random graphs with arbitrary degree distributions and their applications’, Physical Review E, 64(2), 026118 (2001). Peterson, J. and Sharp, M. (1998) Technology Policy in the European Union, London: Macmillan Press. Watts, D.J. and Strogatz, S.H. (1998) ‘Collective dynamics of “small-world” networks’, Nature, 393: 440–442.

13 Small worlds and technology networks The case of European research collaboration1 Bart Verspagen 13.1 Introduction European Research and Technology Development (RTD) Actions are an increasingly important part of the activities of the European Commission. With the Economic and Monetary Union (EMU), the degrees of freedom for national governments in applying traditional policy instruments have certainly become smaller. At the same time, it has been generally recognized that the key for European competitiveness vis-à-vis the United States and parts of Asia lies in the technology domain. The European RTD policy has traditionally been strongly aimed at the formation of research networks between ﬁrms and other organizations engaged in research and development (R&D). In the road leading up to the most recent wave of European Union (EU) ﬁnancing of RTD networks, the so-called sixth Framework Programme (FWP), the formation of these RTD networks has been put central in the process designed to lead to the so-called European Research Area (ERA). The main goal of the ERA is the further integration and focusing of community research (EU 2002). Networks of strategic technology alliances as institutions emerging from the free market are also assuming more importance (for example, Hagedoorn and Schakenraad 1992). By its very nature, knowledge is a commodity displaying increasing returns: a single piece of knowledge can be used by many people or organizations at the same time. However, there is still a major difference between knowledge and information. To master knowledge takes more than just reading information about how a particular thing works. The user of knowledge needs a background in the speciﬁc ﬁeld of science or technology, involving signiﬁcant effort in the form of study and buildingup of experience. Also, important parts of knowledge are tacit, implying that face-to-face contact is necessary to transfer it from one person to the other. RTD activities and learning in networks are primary ways of taking advantage of the increasing returns from the properties of knowledge. The majority of these networks are formed by means of decentralized

300 B. Verspagen interaction of agents (researchers, organizations). Two agents might link up with the aim to develop a speciﬁc piece of knowledge and, in the process, they exchange knowledge. Both agents might also form links with other agents, and in this way a network is formed. Knowledge ﬂows through the network, not just directly between two agents, but also in the form of indirect connections where an agent acts as an intermediary between two other agents that are not directly linked. The structure and nature of these networks has been the traditional ﬁeld of study of social network theory and graph theory (for example, Wasserman and Faust 1994). The tools and concepts in this ﬁeld have been adopted in the analysis of RTD networks, both from the point of view of policy and strategic management within ﬁrms (for example, Vonortas 1997). More recently, the self-organized nature of networks of RTD cooperation has been studied using the concepts of so-called scalefree networks (for example, Newman 2001) and small worlds models (Watts 2000). These self-organized networks have attracted the attention of European policy makers (for example, Johnston et al. 2003) because of the speciﬁc opportunities they offer for efﬁcient distribution and integration of knowledge. It may thus be argued that the self-organized networks of researchers are good models for thinking about the ERA and the future of European RTD policy (for example, Johnston 2003). The literature on self-organized networks argues that speciﬁc topologies of collaboration networks may form as the result of decentralized interaction between researchers and/or organizations and that these topologies have consequences for the way in which the networks operate. A scale-free network, for example, is one in which the distribution of each agent’s links follows a so-called power-law: a straight line displays a double log plot of the number of links each agent has against the rank-order of these links. In practical terms, this means that in these scale-free networks, the number of nodes (that is, researchers) with only a few links is relatively high, while the number of nodes with many links is relatively small. The ‘transition’ of the ﬁrst type of node to the second type is linear in the double log plot mentioned. Such a network may emerge as the result of a mechanism of reputation formation: everybody wants to cooperate with the most prestigious agents in the network, but reputation is in part dependent on the number of cooperations an agent has. Based on this self-reinforcing mechanism, some agents emerge as very important in terms of the number of nodes they have. The theory of small worlds (Watts 2000), on the other hand, argues that there is a relationship between the local clustering of links in a network and its efﬁciency in terms of the distances that information in the network has to travel between nodes. Cowan and Jonard (2003) developed a theoretical model of networks of inventors exchanging knowledge and conclude that the small worlds topology is a rather efﬁcient one in terms of optimizing global inventive performance.

Small worlds and technology networks 301 This chapter will attempt to apply the small worlds theory to the case of European technology networks. In line with Cowan and Jonard, it will be argued that the small worlds topology can be seen as an efﬁcient way of sharing knowledge. Speciﬁcally, it is argued that this topology can be seen as a way of uniting two distinct perspectives found in the strategic management literature on network formation. These are the notions of social capital and structural holes (Walker et al. 1997). Using the small worlds idea in this way, it is argued here that one may use this concept to judge the overall efﬁciency of a network of technology alliances. This argument is applied to empirical data on European RTD networks. The next section brieﬂy reviews the social capital and structural holes theory of network formation in the strategic management literature. Section 13.3 sketches the context of European research networks and formulates two speciﬁc research questions. Section 13.4 outlines the theory of small worlds and argues how it can be applied to the theory discussed in Section 13.2. Section 13.5 presents the empirical results. The argument is summarized and concluded in Section 13.6.

13.2 Strategic technology alliances and networks Hagedoorn and Schakenraad (1992) and many others since, observed a strong increase in the number of strategic technology alliances between ﬁrms. Explanations have been offered from a variety of theoretical perspectives for the emergence of these alliances and their increasing role. Hagedoorn et al. (2000) discuss three ﬁelds of theory: transaction cost economics, strategic management and microeconomic industrial organization. Many of the contributions to this literature focus on a single alliance as the unit of observation. Various arguments, including the exploitation of economies of scale, the sharing of risks, access to resources complementary to the ﬁrms’ own resources, etc., have been put forward to explain why ﬁrms are willing and eager to enter into such partnerships. Recently, however, the strategic management literature has begun to pay attention to the broader context of alliances (for example, Gulati 1995; Walker et al. 1997). This follows a trend that can be seen in the sociological literature dealing with innovation and networks (for example, Alter and Hage 1993; Granovetter 1983). The central notion in this literature is what Nooteboom (1999) has termed indirect access. Because a partner with whom a ﬁrm collaborates may also collaborate with others, the ﬁrm does not only gain access to the resources of its immediate partner, but also, to a certain extent, to those of others with whom the partner collaborates. The issue of strategic collaboration then becomes one of picking partners based not only on their own immediate capabilities and resources, but also on those of the other actors they gain (partial) access to through the network.

302 B. Verspagen In the recent theory of strategic technology alliances, the notion of ‘structural equivalence’ has played a central role (Hagedoorn and Duysters 2002). If two network actors have a high degree of structural equivalence they have (direct) ties to the same broad set of other actors. The phenomenon of structural equivalence has given rise to two different perspectives on the dynamics governing the formation of networks and their efﬁciency as information or knowledge transmitters. The ﬁrst perspective is that of social capital (Bourdieu 1980; Coleman 1990). Social capital can be deﬁned as ‘the sum of the resources, actual or virtual, that accrue to an individual or a group by virtue of possessing a durable network of more or less institutionalized relationships of mutual acquaintance and recognition’ (Bourdieu and Wacquant 1992 cited in Walker et al. 1997: 109). To put the concept of social capital into a network context, Walker et al. (1997) argued that the density of network relations is a good proxy: ‘Some ﬁrms occupy positions that are embedded in regions ﬁlled with relationships, indicating a high level of available social capital, but other positions are located in regions with few relationships, suggesting a low social capital’ (Walker et al. 1997: 111). According to them, social capital and network structure must be considered as variables that inﬂuence each other jointly. They argue that high social capital (that is, being embedded in a dense part of the overall network) leads to a higher propensity of new partnerships. The analysis in this chapter will, however, not be concerned very much with this dynamic relationship between the two variables. The second perspective on networks stems from work by Burt (1992) who argued that a ﬁrm interested in using networks as a source of knowledge should attempt to minimize structural equivalence with the partners that it links up to directly. In other words, in strategically selecting partners for cooperation, one should look for partners that have direct links to other actors with whom one does not yet have strategic links, in order to try to ﬁll the structural holes in one’s own network. Burt’s (1992) idea is that this strategy will provide access to knowledge or information that has a high yield. An important role then is reserved to nodes acting as socalled ‘bridges’ between two relatively unconnected parts of a network. Confronting these two perspectives, one can see two distinct tendencies emerging. Strategies built solely on the idea of enhancing social capital would lead to networks with very dense local environments, with relatively few connections between them. On the other hand, when ﬁrms mainly pursue a strategy of ﬁlling structural holes in their own network, the tendency is for less densely connected local environments, but with the average distance between the agents in the network being smaller because in trying to access parts of the network ﬁrms seek out bridges.

Small worlds and technology networks 303

13.3 European policy and RTD networks2 Technology is an economic good that is characterized by externalities. The presence of externalities implies two things. First, relative to a situation where all beneﬁts are internal, there are disincentives for a ﬁrm investing in the development of new knowledge because it cannot appropriate all the beneﬁts from the investment. Second, from the perspective of the economy as a whole, investment in the development of new knowledge may be too small to achieve a social optimum, even if, at the level of individual ﬁrms, investment levels are ‘optimal’. In other words, externalities associated with the development of new knowledge by ﬁrms will lead to market failure. Market failure has been the reason for technology policies, both at the level of national (or even regional) governments and the EU as a whole. European technology policy has mainly taken the form of RTD policy. This set of policies is aimed at stimulating partnerships between ﬁrms and other organizations (such as universities, institutes, etc.) undertaking R&D work. Such partnerships are generally aimed at undertaking joint work towards a speciﬁc goal that ﬁts a set of predeﬁned areas of work. Usually, the work undertaken is pre-competitive, that is, the fact that two (or more) ﬁrms engaged in the same project jointly develop knowledge does not implicitly lead to joint product development (which would conﬂict with competition policy). This implies that most of the work undertaken in European RTD projects is aimed at the early stages of technology development. With the movement towards the new phase of the European RTD policies (sixth FWP), it has been recognized that European research is still to a large extent scattered across geographical space and interaction is hindered by various barriers, including cultural differences, differences in the scientiﬁc level of researchers, differences in national policies towards universities and public research institutes (for example, in funding mechanisms), still-existing problems in the inter-country mobility of researchers, etc. The ERA has been presented as a programme aimed at eliminating these barriers thereby increasing the level of integration of European research and in this way meeting critical mass requirements in more areas at the world’s scientiﬁc and technological frontier. The goal of establishing an ERA can be seen as the major effort towards meeting the goals formulated at the Lisbon Summit, that is, making Europe the most competitive region in the world knowledge economy. One may devise many different methods of evaluating the overall outcomes of European RTD policy in the light of the goals of the ERA. From the point of view of networks, one criterion for evaluation would be to quantify and measure their efﬁciency as transmitters of knowledge between the network partners (as suggested in a theoretical network by Cowan and Jonard 2003). This chapter will attempt to provide such an assessment using empirical data on European research networks resulting

304 B. Verspagen from the RTD policies. The two perspectives on network formation (social capital and the theory of structural holes) that were discussed above will be used as broad guidelines in the formation of the research strategy of individual organizations (ﬁrms, public research institutes and universities). Comparison of these two perspectives will lead to two main theoretical questions: 1

2

Taking the theory of social capital as the starting point, what can be said about the implications of highly dense networks for the overall efﬁciency of RTD networks as a means of transmitters of knowledge? Do highly dense networks lead to efﬁcient or inefﬁcient transmission of knowledge ﬂows through the network? Can the perspectives of social capital and structural holes be related into one coherent theoretical model? Using such a model, is it possible to judge the relative impacts of both mechanisms on the structure of empirically observed networks?

It will be argued that the recent literature on ‘small worlds’ (Watts 2000, see also Cowan and Jonard 2003) provides a theoretical model such as that referred to in the second question. Moreover, using this model provides some hints about the answers to the ﬁrst question. Section 13.4 brieﬂy explores the theory of small worlds.

13.4 Knowledge ﬂows and network structure3 13.4.1 Concepts and deﬁnitions In order to quantify the relative efﬁciency of networks as devices for the diffusion of knowledge and information, we need to introduce some terminology as well as two central concepts. We use the term ‘edges’ to refer to the connections in a network between two nodes (the actors operating in the network). The central concepts are the characteristic path length of a network and clustering. 13.4.1.1 Characteristic path length We assume that knowledge ﬂows in the network by means of direct and indirect linkages. This means that although two ﬁrms or organizations may not be directly connected, it is still interesting to look at the way in which they are connected through intermediaries in the form of other ﬁrms or organizations. For any pair of actors in the networks, the path length between them is deﬁned as the minimum number of intermediaries necessary to connect them, plus one. It follows from this deﬁnition that actors that are directly connected have a path length of one between them. The particular indicator that has been used in much theoretical

Small worlds and technology networks 305 work in order to arrive at a single measure of path length for the network as a whole is called characteristic path length and is deﬁned as the median of the average path length of all actors in the network. The average path length of an actor is deﬁned as the average length of the (shortest) paths to all other actors in the network. Characteristic path length of the network is interesting because it gives an indication of the relative (potential) efﬁciency of the network. The shorter the path length in a network, the easier and quicker knowledge or information may diffuse through the network. From the point of view of a single actor in the network, shorter path length implies easier access to the knowledge of other actors in the network. Characteristic path length will depend on a number of network characteristics. The number of actors in the network is one prime determinant, as is the average number of direct connections that each actor has (the latter is termed ‘average degree of the network’). In addition to these quantities, which can be easily measured for any network, it is the network topology that determines characteristic path length. The prime factor of interest with regard to network topology is the way in which the edges are distributed across the network as a whole. 13.4.1.2 Clustering The concept of network clustering can also be used to quantify some interesting factors in network topology. In order to deﬁne network clustering, one must ﬁrst deﬁne the neighbourhood of actor i. The term will be used to describe the subset of actors that have a direct connection to actor i. Obviously, the number of actors in the neighbourhood of actor i is identical to the degree of actor i (i is deﬁned not as a member of its own neighbourhood). Now deﬁne clustering of the neighbourhood of i as the number of edges in neighbourhood i as a fraction of the maximum possible edges in that neighbourhood. This last is simply the number of combinations of two distinct actors one can draw from a subset of k actors, where k is the degree. Clustering at the level of the network as a whole is then deﬁned as the average of clustering of all neighbourhoods i. One immediately recognizes the relevance of the clustering measure for the theory of social capital in the formation of networks as was brieﬂy explored above: it is a direct measure of the amount of social capital in a local environment. The theory of small worlds is essentially about the relationship between clustering (network density) and other characteristics of the network. This is explored in the next sub section. 13.4.2 The theory of small worlds In order to illustrate the importance of the concept of clustering, consider a simple example. Suppose one is given the task of transmitting a message

306 B. Verspagen to a person not known to oneself, and located at some far (geographical) distance. Suppose also that the only allowed way of transmitting this message is by means of personal contacts with one’s own friends. It is allowable for these friends to tell their friends, etc., until the target-person is ultimately reached. Assuming the target person for the message is not a member of one’s own personal network (‘friends’), one question of interest in this regard is how clustered the total network is. Suppose that all one’s friends’ friends, are also one’s own friends (‘everybody you know, knows the same people that you know’). This corresponds, by deﬁnition, to a high value of clustering. In this case, it will obviously not be easy to reach the unknown target person by means of these personal relations because the target person is not already in one’s circle of friends. The challenge is to ﬁnd the odd friend that can set into motion the chain that will ultimately reach the target person. If, on the other hand, one’s friends know completely different persons, it will be easier to imagine a short path to the unknown target person, by simply probing one’s own network in a chain-letter type fashion. The example shows by intuition that high clustering goes hand in hand with long characteristic path length, and vice versa. The theoretical work in this area (Watts 2000) tends to support this intuition. In fact, two speciﬁc network typologies have attracted much attention, each representing one extreme of the average clustering–characteristic path length relation. The ﬁrst typology is the so-called ‘connected caveman world’. Such a world consists of a number of distinct ‘caves’ that are connected to each other by a single edge. A cave is a subset of actors that is fully connected. Each cave is connected to exactly two other caves by two distinct members, each of whom is connected to one other actor in a different cave. In such a network, which is depicted in Figure 13.1, all paths to actors in the same cave have the minimum length of 1, but paths to actors outside the cave are much longer because they depend on only a few actors that can act as intermediaries. Since the cave is almost fully connected internally, the degree of clustering of this neighbourhood is for all practical purposes equal to one. This high clustering leads to a relatively high value for characteristic path length. The connected cavemen world is one that can be interpreted as emerging from a purely social capital theory-inspired world. The individual caveman obviously has a high amount of social capital, because (s)he is connected to a relatively large number of individuals in the immediate direct environment. However, the world is larger than a single local environment (cave), and the high resulting characteristic path length implies that the high social capital comes hand-in-hand with a large average distance to other parts of the world. The other network typology of interest is that of the Moore graph. This is a network in which every vertex is adjacent to exactly k other vertices

Small worlds and technology networks 307

Figure 13.1 The connected Caveman World (left) and a local view of the Moore Graph (right).

and none of these other vertices is connected to another. The Moore graph is a purely theoretical construction and cannot even be obtained for many combinations of n and k. This is why Figure 13.1 displays only a local view around a single node for the Moore graph. The Moore graph is of interest here because it puts a lower limit on characteristic path length. An approximate expression for characteristic path length of the Moore graph with given n and k can be obtained, and it can be shown (Watts 2000: Section 4.1.2 and the references there) that it is impossible to obtain shorter characteristic length for any graph with identical n and k provided that the degree of the individual nodes does not differ too widely. Hence, the Moore graph has much lower characteristic path length than the Connected Caveman World. By deﬁnition, the Moore graph has zero clustering. In other words, the much shorter path length of the Moore graph has been obtained by means of a much lower (in fact zero) clustering coefﬁcient. One may thus say that the Moore graph represents an extreme case that results when social capital (dense local relations) plays no role at all in network formation, but instead networks are constructed with the aim of leaving no structural holes. In practice, one can only approximate the perfect Moore graph by means of a random graph, which will yield values for the clustering coefﬁcient that are in fact positive, but very close to zero. It is possible to deﬁne a class of random graphs that provide a more or less smooth transition between the highly clustered and long connected caveman world and the short and not very highly clustered Moore graph. This is illustrated below. To summarize, two main typologies of graph have been introduced:

308 B. Verspagen one for which the characteristic path length is high and, hence, information takes a long time to travel from one (average) network member to another; and one for which the characteristic path length is much lower and, hence, information travels much more rapidly. But the ‘long’ network is characterized by high clustering, which implies a high degree of social capital and may thus increase the quantity and quality of knowledge transferred in the local environment. The ‘short’ network, on the other hand, is characterized by low clustering and hence low degrees of social capital. Thus, the extreme case of high clustering and long path length can be seen as a representation of a social capital-inspired world view, while the other extreme (short path length and low clustering) can be seen as emerging from the desire to ﬁll structural holes in the network. Is the negative relationship between clustering and length that results from comparing these two stereotypes a general phenomenon? It can be shown that high clustering and long characteristic path length need not always go together. A special class of networks can be identiﬁed in which clustering is relatively high, but characteristic path length is relatively short. This type of network has been called ‘small worlds’ (Watts 2000). The term has been derived from the hypothesis that although most people in the world mainly know other people that belong to a fairly clustered set of friends, only ﬁve to six intermediaries are necessary to connect the largest part of the population of the globe. In other words, even though a person knows mainly people in her own environment, she will have indirect links to most other people in the world through only a small number of steps. Speciﬁcally, a small world has been deﬁned as a network with n actors and average degree k that displays characteristic path length approximately equal to a Moore graph with the same n and k, but has much larger clustering than this Moore graph. The relevance of the idea of small worlds becomes immediately obvious in the context of the comparison of the structural holes perspective and the social capital perspective. Start from a situation of the connected cavemen world, that is, a world in which social capital dominates the formation of networks completely. Then imagine that in a controlled experiment, networks members are allowed to ﬁll in structural holes in their personal network by engaging in partnerships with ﬁrms in other parts of the network. How would this affect the characteristic path length of the network and, hence, the efﬁciency of knowledge transfer? How ‘many’ structural holes would have to be ﬁlled before a certain degree of network efﬁciency would be reached? The theory of small worlds provides a very clear-cut answer to these questions. In order to show how the small world is located inbetween the connected caveman world and the Moore graph, Watts (2000) proposed a number of formal models describing the construction of (random) graphs. The exact construction of these graphs need not concern the reader here. What is interesting, however, is that these networks can be

Small worlds and technology networks 309 1,000

0.7

900 0.6 800 700 600

0.4

500 0.3

Length

Clustering

0.5

400 300

0.2

200 0.1 100 0

0 0.000001 Clustering

0.00001

0.0001 Length

0.001

0.01

0.1

1

φ

Figure 13.2 Length and clustering as a function of the graph tuning parameter φ.

tuned by a single parameter to either side of the connected caveman world or the Moore graph. Moreover, this parameter has a clear interpretation in terms of the structural holes perspective: it measures the tendency for network connections to ﬁll in structural holes. Figure 13.2 illustrates the model. The horizontal axis of Figure 13.2 displays the parameter φ, which tunes the graph. The parameter gives the fraction of edges in a graph that are shortcuts. Shortcuts are edges that complete triads or, alternatively, shortcuts are edges that connect two vertices that would otherwise be widely separated. Clearly, then, shortcuts are exactly the ‘bridges’ that ﬁll in structural holes in a network member’s environment. The fraction of such shortcuts or bridges in the total of all connections can thus be seen as a degree of presence of the structural holes strategy in overall network formation. Instead of using φ, one may also use a slightly different parameter denoted by ψ. This parameter measures the fraction of all pairs of edges in the network that are not connected and have one, and only one, common neighbour. Such pairs of edges are called ‘contractions’ and can be seen as an alternative conceptualization of the idea of bridges ﬁlling in structural holes. Watts (2000: 73) argues that ‘ψ is an analogous parameter to φ,

310 B. Verspagen although it is more general, as most shortcuts result in contractions, but not the reverse’. In the networks depicted in Figure 13.2, there is an exact quadratic relationship between φ and ψ, which is governed by only one additional parameter. This is the so-called bundle size, which is roughly equivalent to the number of parties involved in a single partnership or alliance. However, because the quadratic term in this relationship adds little to the values of n and k that characterize the networks of interest in this chapter, using φ or ψ is more or less equivalent from a theoretical point of view. In the empirical applications below, however, which of the two parameters is used may matter. Figure 13.2 shows that for a low fraction of shortcuts in the set of the network’s edges, one observes a high value for characteristic length of the network and also a high value for the clustering coefﬁcient of the network. This, in other words, corresponds to the connected caveman world network or, alternatively, to a world in which social capital completely dominates network formation. At the other extreme, where the large majority of all edges are shortcuts and the structural holes perspective dominates network formation, one observes low values for clustering and characteristic length. This corresponds to the Moore graph. What is interesting, however, is that the path from high to low values or φ, that is, the path from a pure social capital perspective to a pure structural holes perspective, is quite different for the two variables in the graph. The clustering coefﬁcient does not descend very much at the beginning, but shows a rapid decline towards the end. Characteristic path length shows more or less the opposite trend: it declines rapidly at ﬁrst and reaches very low levels already at relatively low levels of φ. The small world phenomenon is found at the limited range (say between 0.01 and 0.1) for which characteristic path length is already at levels comparable with the Moore graph (right extreme), but the clustering coefﬁcient is still relatively high. Note that the curves in Figure 13.2 are drawn for some speciﬁc values of other network parameters. These are the size of the network (number of vertices) and the average number of direct connections that each of them has (average degree). Also, the underlying model assumes that the vertices do not differ too widely in their actual degree (see Watts 2000: Chs 3 and 4 for more detail). When applying the theory to a speciﬁc realworld case, as is done in the next section, these assumptions will need to be taken into account. This model of small worlds thus provides an answer to both questions formulated in Section 13.3. The model shows that very dense networks in which social capital dominates network formation do indeed lead to inefﬁcient networks in terms of the overall speed of transmission of knowledge ﬂows through the network. Compared to this extreme case, the application by network members of a strategy of trying to ﬁll structural holes will generally decrease characteristic path length and, hence, facilitate know-

Small worlds and technology networks 311 ledge ﬂows in the network as a whole. The theory also shows how frequently the structural holes strategy needs to be applied in order to approach the theoretical lower boundary on characteristic path length. Speciﬁcally, the model suggests that the small worlds range of the parameter φ is relatively efﬁcient. Forming shortcuts beyond the upper limit of the small worlds range does not lead to a noticeable further decrease in characteristic path length. Note that this is a characteristic of the network as a whole, and that in any network there will be some ﬁrms with shorter (or longer) path length. In other words, even in the small worlds range, there will be an incentive for some ﬁrms to invest further in alliances that ﬁll in structural holes in their own local environment. We next investigate how the real world networks of European research cooperation compare to the model depicted in Figure 13.2 and whether these networks are near to the small worlds range. In this way, the answer to the two questions posed in Section 13.3 can be derived for the real world case of European RTD networks.

13.5 Empirical results The three datasets that are considered in this chapter can all be envisaged as networks that facilitate knowledge ﬂows. These networks consist of actors (the nodes) and the relations between them (the edges). Most theory in the ﬁeld of network analysis considers the edges in a network as binary: they are either on or off. The analysis here follows this representation and contends that a relation between two actors exists when these two actors participate in the same project at least once (in the case of RJV (research joint venture) networks), or where there is at least one citation relation between the two actors (in the case of the patent citation network). Two of the networks are related to data from European research networks. The ﬁrst, labelled RJV, contains data from the European FWPs. It contains collaborative research projects that were (partially) funded from the budget of the European Commission (up to the end of the 1990s). FWPs are organized by topics, which are broadly deﬁned in terms of technological areas and possible applications. Funding is awarded on the basis of evaluation by a group of independent experts of written proposals from the applicants. Here, we consider each individual project in this database as a completely connected subgraph. The second dataset on European collaborative research projects concerns the EUREKA programme (EUR dataset). EUREKA is not primarily a funding device, but acts as a broker. Projects in EUREKA focus around a speciﬁc technological issue and bring together parties interested in this issue for exchange of information and knowledge and possible joint work. Again, EUREKA projects are treated as a completely connected subgraph. Data for both the EUR and RJV datasets were taken from the European Commission’s CORDIS database.

312 B. Verspagen The third and ﬁnal dataset contains patent citations (PC) network. These are based on European patent applications only. The set of ﬁrms in this database was limited to the ﬁrms that were present in either one of the two other databases. Selection of the patents of these ﬁrms was made by researchers at CESPRI (Bocconi University, Milan). It is important to note that the patent citations dataset is inherently different from the other two datasets. First, at the ﬁnest level of disaggregation, it contains binary relations only: citations occur in pairs of cited and citing patents. This is an important difference, which has implications for the clustering of the network. In the case of RJV and EUR, each project is a completely connected subgraph, which, by deﬁnition, implies a more clustered network. Second, not all ﬁrms that participate in EUR or RJV ﬁle European patent applications. As a result, the patent citations dataset has far fewer ﬁrms than the two other networks (more details below). Finally, patent citations, in contrast to the two other datasets, have a direction associated with the ﬂow of information or knowledge. The citing ﬁrm can be considered as the receiver of knowledge, the cited ﬁrm as the originator. This, in principle, would allow for the construction of directed graphs. However, because the theory explained above only considers undirected graphs, it was decided to treat all relations in the patent citation network as two-way relationships. In most cases, this is completely appropriate because there are indeed citations between the edges in the network in both directions. Table 13.1 gives some basic statistics about the networks. The RJV network is the largest. It comprises almost 4,000 projects, jointly undertaken by almost 10,000 distinct organizations. The average number of participants per project is close to seven. Note that the set of 3,874 projects does not lead to a completely connected network. The largest connected network comprises 9,455 organizations, the other approximately 400 organizations are not connected to the rest of the network. The analysis is undertaken for the set of 9,455 connected organizations. Nodes in this network are, on average, directly connected to almost 22 other edges (k). The EUR network is smaller – almost half the size of the RJV network. This is mainly due to a much smaller number of projects (close to 1,000). The number of participants per project is only slightly lower than for the RJV network. One can deduce that on average organizations in the EUR network participate in fewer projects than those in the RJV network. The EUR network also has a larger fraction of organizations that are unconnected to the rest of the network. The total number of nodes in the connected part of the network is still quite high, that is, close to 5,000. Average degree (k) for EUR is much larger (36) than for the RJV network. It has already been mentioned that the PC network is rather different from the other two networks. Here ‘projects’ are in fact citation pairs of two patents and, hence, the number of participants per ‘project’ is exactly

Clustering Characteristic path length

Number of projects in database Average number of participants per project Number of organizations in dataset n k φ ψ

Table 13.1 Basic statistics of the networks

0.65 3.09

3,874 7.2 9,816 9,455 21.8 0.0010 0.081

RJV

0.002 2.95

9,455 22

Moore graph

0.43 3.45

1,031 6.0 4,261 3,414 36.2 0.0012 0.077

EUR

0.010 2.61

3,414 36

Moore graph

0.32 2.98

4,753 2.0 897 813 8.46 0.13 0.171

PC

0.009 3.35

813 8

Moore graph

314 B. Verspagen two, which is of course much lower than for the other two projects. The total number of citation pairs, however, is larger than the total number of projects in either one of the other two networks. The number of organizations in the network is now, however, much smaller. Average degree is also much smaller. How then do the three networks compare relative to a Moore-graph, that is, the right hand side of Figure 13.2? This is the question that is answered by the last two rows in Table 13.1. The RJV network displays the highest clustering, followed by the EUREKA network and then the patent citations network. All three clustering values, however, are much bigger than one would expect for a random network (Moore graph) with similar n and k. Characteristic path length is similar to what is expected for the Moore graph, except perhaps for the EUREKA network, which shows a somewhat higher value than a similar random network. However, this value is still of an order of magnitude smaller than the value for a connected caveman world with similar parameters (not documented in the table, but approximately equal to 45). The RJV network shows characteristic path length just above the corresponding Moore graph and the patent citations network has a shorter path length than the Moore graph. Hence, it can be concluded that the three networks investigated here can be considered to be small worlds. They have degrees of clustering of the order of magnitude of the connected caveman world, but characteristic path length of the order of magnitude of the Moore graph. In terms of the main topic of interest in this chapter, this implies that the observed networks can be considered as relatively efﬁcient knowledge transmitting devices (Cowan and Jonard 2003). Following Watts (2000), one may proceed to investigate whether or not the model used to depict the networks in Figure 13.2 provides a reasonable approximation of the three cases that have been analysed. In order to do this, one can simply plot the curves as in Figure 13.2 for the speciﬁc parameters (n, k) that apply in the real-world case. This can be done for φ and ψ. One can then measure these two parameters for the real-world case and plot the observations for clustering and characteristic path length in the same graph. When these two points are located near to their respective lines in the graph, the models can be considered to be adequate predictions. This is depicted in Figure 13.3. For each of the three networks, φ and ψ are both depicted. For the RJV network, the observed points are in the right direction, but still at some distance for the φ-graph (left). For the ψgraph, the actual points lie right on the theoretical relationships described by the two lines. This means that, in general, the model used to construct the theoretical curves can be considered as a reasonable or even good approximation of reality in the case of the RJV network. The theoretical relationship between φ and ψ, however, is not entirely adequate, and contractions (ψ) appear to be a much more useful concept in the case of the RJV network.

Small worlds and technology networks 315 In the case of the EUR network, the approximation is clearly much less precise. In particular, the EUR network turns out to be less clustered than the theoretical model would predict. The prediction for characteristic length is reasonable (for φ) or good (for ψ). Note that clustering of the EUR network is still much higher than for a random network, so that one may still argue that this network adheres to the small-world deﬁnition. But it is not true that the model used to construct the theoretical curve is a completely adequate approximation of reality in this case. The implications of this ﬁnding are discussed below. Finally, the actual observations for PC network are located a bit further to the right in both graphs. These observations already come fairly close to the range for which the network is comparable to a random network, although clustering is still somewhat higher. Now the φ-graph appears as the most reasonable approximation, but again observed clustering is lower than the predicted value. How can these results be interpreted? In the case of the RJV network, the random graph model of Figure 13.2 above is a very good approximation for the structure of the underlying graph, at least when one uses the share of contractions in total edges as the tuning parameter of the random graph. This means that the relative frequency of contractions in the network fully explains its characteristics in terms of length and clustering. For the other two networks, the theoretical models in large part offer the same explanation, but other factors (not accounted for in the theoretical exposition here) need to be included to explain the exact properties of the graph. Nevertheless, these networks can be considered to be small worlds, that is, they are relatively efﬁcient information transmitting devices. Speciﬁcally, the clustering coefﬁcients of the EUR and PC network are somewhat lower than one would expect in purely random networks with similar φ or ψ. This deviation from the predicted value may be the result of the unevenness in the number of connections (k) between members of the network. It may also be the case that strategic behaviour by some network members has led to the lower value of the clustering coefﬁcient.

13.6 Conclusions This chapter has argued that the recent theories of self-organized networks, speciﬁcally the small worlds model, provide a natural way of uniting two perspectives found in the strategic management literature on networks and strategic alliances. One perspective, that of the theory of social capital, leads to networks consisting of densely connected local environments corresponding to high values of social capital. Firms are considered to seek such dense local environments in building their strategic alliances. The theory of small worlds shows that in the extreme case of very dense

0.8

250 200 150

0.4 100

0.1 0

φ

0. 00 1

1

0. 1

0. 00 00 0 0. 1 00 00 1 0. 00 01 0. 00 1 0. 01

0

Clustering

1

0

0.3 0.2

50

0.1

0.4

0. 1

0.2

0.5

0. 01

0.3

0.6

Clustering

0.5

Length

0.6

Clustering

200 180 160 140 120 100 80 60 40 20 0

0.7

0.7

Length

0.8

Ψ

Length

0.1

0.

0.1 0

1

00 0.

0.

1

0.2

0.

1

01 0.

01

00 0.

1 00

00 0.

00 0.

00

00

01

0

0.3

φ

Clustering

Length

1

0.2

0.4

1

0.3

0.5

0.

0.4

0.6

01

0.5

1

Clustering

0.6

200 180 160 140 120 100 80 60 40 20 0

0.7

Clustering

0.7

0.8

Length

0.8

Length

RJV network 50 45 40 35 30 25 20 15 10 5 0

Ψ

φ

1 0.

01 0.

1 00 0.

00 1 0. 00 01

00 0.

0.

00

00

01

0

0.2 0.1 0

Clustering

Length

1

0.1

0.3

0. 1

0.2

0.4

01

0.3

0.5

Clustering

0.4

200 180 160 140 120 100 80 60 40 20 0

0.6

0.

Clustering

0.5

0.7

1

50 45 40 35 30 25 20 15 10 5 0

0.6

Length

0.7

Length

EUR network

Ψ

PC network

Figure 13.3 Theoretical predictions and actual values for clustering and characteristic path length, three network.

Small worlds and technology networks 317 local environments, the ‘average’ path length for knowledge ﬂows between two actors in the network becomes very large. This can be considered as an impediment to the efﬁcient spread of knowledge and information throughout the whole network rather than its local environs and, thus, is an undesirable property from a policy point of view. The other perspective in the strategic management literature is that of structural holes. This theory argues that in seeking partnerships, ﬁrms will try to connect to bridges between their ‘own’ part of the network (the local environment which may be densely connected) and other interesting parts of the network. Rather then extending their network based on social capital (that is, seeking dense local relations), ﬁrms are expected to pick and choose partnerships based on the strategic access that a new alliance will give them to other parts of the network, embodying other types of knowledge. The theory of small worlds is a useful way of representing the tendency for this ‘bridging’ behaviour in terms of a single parameter. It shows how the characteristics of the network in terms of clustering (deﬁned as density of local environments) and average path length of the network develop as a function of this tendency. For very dense local environments (that is, a world dominated by network ties resulting from a social capital view of the world), average path length is very long, while for extreme levels of ‘bridging’ (a ‘structural holes world’) path length approaches a lower limit. Inbetween these two extreme cases one ﬁnds a small range of the bridging parameter for which clustering is still relatively high, but average path length is already close to the lower limit implied by the extreme structural holes world. This is the range of ‘small worlds’ as deﬁned by Watts (2000). Cowan and Jonard (2003) found that these small worlds can be characterized as relatively efﬁcient networks in terms of knowledge transfer. In an empirical application, it has been shown that European research networks do indeed show small worlds properties. The network consisting of all FWP projects comes closest to the theoretical model of small worlds. Two other networks, the network of EUREKA projects and the non-policyrelated network of patent citations, also resemble small worlds, but in these cases the theoretical model proposed by Watts performs less well. Overall, therefore, it can be concluded that the European research networks analysed in this chapter are relatively efﬁcient means of knowledge transfer.

Notes 1 This chapter is based on part of the MERIT contribution to the (KNOW) project, and builds on the work by the NTUA and CESPRI teams in preparing the databases used here. I am grateful for the opportunity to use these databases. I also thank Nick Vonortas, Stefano Breschi, Robin Cowan, Jerald Hage and Geert Duysters for comments and discussions.

318 B. Verspagen 2 This section necessarily takes a rather simplistic and generalistic view of EU RTD policies. 3 The exposition here draws heavily on Watts (2000).

References Alter, C. and Hage, J. (1993) Organizations Working Together, London: Sage. Bourdieu, P. (1980) ‘Le capital sociale: Notes provisaires’, Actes de la Recherche en Sciences Sociale, 3: 2–3. Bourdieu, P. and Wacquant, L. (1992) An Invitation to Reﬂexive Sociology, Chicago, IL: University of Chicago Press. Burt, R.S. (1992) Structural Holes: The Social Structure of Competition, Cambridge, MA: Harvard University Press. Coleman, J. (1990) ‘Social capital in the creation of human capital’, American Journal of Sociology, 94: S95–S120. Cowan, R. and Jonard, N. (2003) ‘The dynamics of collective invention’, Journal of Economic Behavior and Organization, 52: 513–532. Cowan, R. and Jonard, N. (2004) ‘Invention on a network’, Structural Change and Economic Dynamics, forthcoming. EU (2002) ‘Council decision of 30 September 2002 adopting a speciﬁc programme for research, technological development and demonstration: “Integrating and strengthening the European Research Area” (2002–2006)’, Ofﬁcial Journal of the European Communities, Brussels: European Commission, 29 October. Granovetter, M.S. (1983) ‘The strength of weak ties: a network theory revisited’, Sociological Theory, 1: 1360–1380. Gulati, R. (1995) ‘Social structure and alliance formation patterns: a longitudinal analysis’, Administrative Science Quarterly, 40: 619–652. Hagedoorn, J. and Duysters, G. (2002) ‘Learning in dynamic inter-ﬁrm networks: the efﬁcacy of multiple contacts’, Organization Studies, 23: 525–548. Hagedoorn, J. and Schakenraad, J. (1992) ‘Leading companies and networks of strategic alliances in information technologies’, Research Policy, 21: 163–190. Hagedoorn, J., Link, A.N. and Vonortas, N.S. (2000) ‘Research partnerships’, Research Policy, 29: 567–586. Johnston, P. (ed.) (2003) Complexity Tools in Evaluation of Policy Options for a Networked Knowledge Society, Brussels: European Commission, DG Information Society. Johnston, P., Pestel, R., Mladenic, D. and Grobelnik, M. (2003) European Research Co-operation as a Self-Organising Complex Network: Implications for Creation of a European Research Area, Brussels: European Commission, DG Information Society. Newman, M.E.J. (2001) ‘The structure for scientiﬁc collaboration networks’, Physical Review E 64 016131 and 016132. Nooteboom, B. (1999) Inter-Firm Alliances, London: Routledge. Vonortas, N. (1997) Cooperation in Research and Development, Boston: Kluwer Academic Publishers. Walker, G., Kogut, B. and Shan, W. (1997) ‘Social capital, structural holes and the formation of an industry network’, Organization Science, 8: 109–125. Wasserman, S. and Faust, K. (1994) Social Network Analysis, Cambridge: Cambridge University Press. Watts, D.J. (2000) Small Worlds: The Dynamics of Networks between Order and Randomness, Princeton, NJ: Princeton University Press.

Part III

Policies for knowledge ﬂows

14 Overview of innovation policy affecting knowledge ﬂows in EU member states Anthony Arundel and Nicholas S. Vonortas 14.1 Introduction In Europe, public policies affecting the production and dissemination of innovation-related knowledge are implemented at various levels of governance: supranational (European Union – EU), national (member states) and local. Supranational policies have been especially useful for establishing dense science and technology intensive networks across the continent. Primary policy tools serving that purpose have been the international cooperative research programmes such as those organized through the Research and Technology Development (RTD) Framework Programmes and EUREKA.1 The Framework Programmes (FPs) encourage knowledge ﬂows by limiting ﬁnancial subsidies to collaborative research projects that involve multiple partners, drawn from either the private sector or from public research institutions (PRIs). The latter includes both universities and specialized research organizations such as the Fraunhofer Institutes in Germany and TNO in the Netherlands. Special emphasis has been placed on small and medium-sized enterprises (SMEs), universities and PRIs, considered key for innovation and yet facing particular problems with regards to their efﬁcient integration in the European technical enterprise. In addition, the EU’s Structural Funds subsidize infrastructural projects in Europe’s less economically-developed regions. Some of the Structural Funds are used to develop the public research infrastructure and networks with other European research institutions. The main responsibility for innovation policy and consequently for programmes to encourage innovation-related knowledge ﬂows lies, however, with the EU member states. In the 1980s, national innovation policies often focused on supporting a few innovative leaders or ‘national champions’ via direct economic subsidies for research and development (R&D). During the 1990s, innovation policies in many EU countries shifted in response to three factors: the need to reduce direct R&D subsidies to ﬁrms both for budgetary reasons and to satisfy European competition policy; the adoption of evolutionary theories and system views of the

322 A. Arundel and N.S. Vonortas innovation process; and the widespread conviction that European ﬁrms failed to translate European strengths in basic research into economically successful innovations. These three factors increased the popularity of policies to develop networks of innovative ﬁrms and PRIs in order to encourage knowledge ﬂows between different parts of the innovation system. The system perspective also encouraged member states to establish framework conditions to support innovation and to reduce innovation subsidies for the private sector that were targeted towards strategic technologies such as information and communication technologies (ICT) or biotechnology. Instead, the trend in countries such as the Netherlands, the UK and Denmark was towards innovation programmes that did not favour speciﬁc technologies. However, this trend conﬂicted with the goal of increasing the commercial applications of public sector research, which has partly been met in several EU states through the use of technology forecasting techniques to target funding for public research towards technologies with commercial applications. From the late 1990s until 2003, innovation policy in many EU countries was going through a period of readjustment to bring policies for the private and public sectors into alignment. This has partly been achieved through increasing support for technology-speciﬁc networks and clusters, and through a concerted effort in many EU countries to increase linkages between the public research sector and private ﬁrms. The KNOW survey, conducted in the Spring of 2000, occurred when this realignment was well underway in many EU countries, with the possible exception of Greece. Therefore, where applicable, we illustrate our policy overview with the results of survey questions that may have been inﬂuenced by policy. Our overview of current policy is largely based on national reports on innovation policy that are available from the TrendChart website.2 TrendChart is funded by the European Commission and provides a wealth of data on innovation performance and on innovation policies for each EU member state. At the time of writing, the most recent reports cover national innovation policy as of March 2003, plus major policy issues and programmes that are under development or consideration. This overview focuses on policies within the seven countries covered by the KNOW survey. However, since KNOW includes the four largest European economies, a representative country from the Nordic region, and two other smaller economies, this overview covers almost all current EU policy themes at the national level. There are three main categories of innovation policies of relevance to knowledge ﬂows. The ﬁrst set of programmes is designed to improve the innovative capabilities of ﬁrms, usually SMEs, that lack internal research capabilities. The second set supports research collaboration between private ﬁrms, while the third group encourages linkages between the public research sector and private ﬁrms.

Innovation policies affecting knowledge ﬂows 323

14.2 Improving the absorptive capacity of SMEs Many EU member states maintain a range of programmes to improve the ‘absorptive capacity’ of ﬁrms, or their ability to successfully adopt and implement new technology and to develop innovations in-house. The absorptive capacity of a ﬁrm has two components. First, ﬁrms can innovate by adopting and modifying technologies developed by other organizations, including other ﬁrms and PRIs. This is often seen as an issue of diffusion or technology transfer from one organization to another. An example is the purchase of new computer-controlled manufacturing equipment. The ability of a ﬁrm to introduce this equipment into its production line depends on its understanding of the advantages and disadvantages of the new technology for its own needs and strategies. Second, ﬁrms can innovate by undertaking creative activities such as R&D to develop new or improved products and processes. Much of this work can beneﬁt from discoveries made by other ﬁrms or PRIs. The capacity of a ﬁrm to use these discoveries depends on its ability to understand them and to assess their commercial applications. Any activity that a ﬁrm undertakes to deepen and widen its scientiﬁc and technological skills will also improve its capacity to absorb knowledge from external sources. Most innovation programmes in this category are focused on the ﬁrst type of absorptive capacity since the target audience largely consists of SMEs that only innovate when necessary and therefore have a low absorptive capacity. Many of the target ﬁrms, for example, will only perform R&D occasionally, if at all, and the occasional R&D performers will have low R&D intensities. Only 15 per cent of the KNOW respondent ﬁrms ﬁt this description. The main goal of many of these programmes is to encourage SMEs to adopt existing technology, but many programmes also try to create a long-term improvement in the innovative capabilities of these ﬁrms. The front-line programme in most EU member states to support innovative capabilities in SMEs is a system of regionally-based technology transfer or innovation ofﬁces to provide support and technical advice, such as the Manufacturing Advisory Service and the Innovative Manufacturing Research Centres (IMRCs) in the UK, ANVAR in France, and the TIC-net regional information and consulting centres in Denmark. Greece is currently establishing a network of 13 regional technology centres. Such ofﬁces may provide general educational programmes, customized assistance and consulting services, or provide ﬁrms with information on national assistance programmes. General educational programmes include demonstration projects, courses on innovation management and visits to successful innovative ﬁrms. Demonstration centres provide information on, and demonstrations of, the speciﬁc technologies in use. The goal is to reduce the risk involved in their adoption by helping the ﬁrm make an informed

324 A. Arundel and N.S. Vonortas decision. These centres are usually located at research institutes with the relevant expertise. An example is the PEPER programme in Greece. The UK provides extensive educational programmes on how to manage innovation, using forums, seminars, conferences and workshops that focus speciﬁcally on this topic. Several countries run programmes where staff from SMEs can visit successful innovative ﬁrms in order to learn about best practice in their industry. The leading example, which has been copied by several other EU countries, is the Teaching Company Scheme in the UK. Customized assistance programmes include evaluations of a ﬁrm’s general innovation needs and subsidies to hire recent science and technical graduates. Evaluations often include technology audits or subsidies to conduct technology feasibility studies. Several of these programmes involve visits by a consultant to the ﬁrm. A ﬁxed number of days of consultancy are usually provided for free, while the cost of additional days has to be partly paid for by the ﬁrm. Customized evaluations are provided by expert consultants who assess the ﬁrm’s technical problems and evaluate how innovation ﬁts in with the ﬁrm’s management and business plans. These services are provided by the Teknologisk Servicesystem in Denmark. In Italy, integrated aid programmes that cover both innovation consultancy and equipment investment are provided to ﬁrms in the Mezzogiorno region. Technology audits focus speciﬁcally on technical problems within the ﬁrm and make recommendations on how to solve the problem. Several of these programmes are linked to expertise at a public research institute (PRI). Technology feasibility programmes, such as SMART in the UK, subsidize the cost of evaluating the feasibility of adopting or developing an innovative technology. By reducing risk, they provide an incentive for SMEs that generally innovate infrequently to innovate, or an incentive to innovative SMEs to move into new areas. In addition to evaluating the technology, most programmes require the ﬁrm to develop a business plan for the use of the technology. A programme common to many EU countries to improve the absorptive capacity of ﬁrms is a hiring subsidy for technical staff. In most EU countries the programme and pays up to 50 per cent of the wage costs, for between one and three years, to SMEs involved in employing a recent university graduate to assist the ﬁrm to innovate. Examples include the CORTECHS and CIFRE programmes in France and HERON in Greece. Several countries also design the subsidy so that the new employee provides a direct link between their university or technical institute and the ﬁrm. In Denmark, the subsidy pays 50 per cent of the cost of hiring a PhD student, whose doctoral research is in an area of interest to the ﬁrm. The student’s university also receives state funding. The UK has gone the farthest in this direction. It subsidizes higher education institutions to place graduates in ﬁrms to transfer technology during a two-year project. Supervision is provided jointly by the ﬁrm and the education institute.

Innovation policies affecting knowledge ﬂows 325 The question in the KNOW questionnaire on whether or not the ﬁrm hired new staff from a university or public research institute to work on its most important innovation is of relevance to this policy. Although we do not know whether the hiring was subsidized by a government programme, 21 per cent of KNOW respondents reported hiring a new scientist or engineer from the public research sector to work on this innovation. This was approximately the same as the percentage of ﬁrms that reported hiring from a supplier (22 per cent) and more than the percentage of ﬁrms that reported hiring from a consultant (16) or from a customer (10). Most programmes to build absorptive capacity are not linked to speciﬁc technologies. However, a few countries offer programmes to encourage ﬁrms to adopt targeted technologies or even offer ﬁnancial subsidies for this purpose. For example, France provides soft loans to SMEs for the adoption of computer integrated manufacturing equipment. An increasingly popular area consists of programmes to encourage SMEs to introduce e-commerce or otherwise develop Internet skills. Relevant programmes are currently available in Germany, Italy, France (PAGSI) and Greece. According to the results of the KNOW survey, the intensity of Internet use is lower in these four countries than in Denmark, the UK and the Netherlands. For example, the percentage of KNOW respondents that use the Internet for searching for scientiﬁc and technical information ranges between 73 per cent and 83 per cent in the four countries with programmes to encourage Internet use, and between 88 per cent and 91 per cent in the other three countries.

14.3 Research collaboration Several programmes support knowledge ﬂows by either encouraging or subsidizing technical collaboration and networking between ﬁrms, or between ﬁrms and PRIs. To the best of our knowledge, all EU member states subsidize the creation of sectoral or regional networks of ﬁrms. All of these programmes are designed to ‘connect’ different actors within the innovation system. Several countries also subsidize collaborative research between ﬁrms. All EU countries also offer subsidies for collaborative research between PRIs and ﬁrms, but these programmes constitute a major part of current innovation policy and are therefore discussed separately in Section 14.4 below. Policies to promote networks and regional or sectoral clusters have been increasing in popularity in Europe over the last decade. The April 2002 German White Paper on innovation policy particularly stressed the value of networks, which are now explicitly recognized and constitute a ‘signiﬁcant change in innovation policy making in recent years’.3 Relevant German programmes include InnoRegio, EXIST and BioRegio. The Italian programme PIA provides subsidies for the establishment of networks among ﬁrms in a similar sector. ANVAR in France promotes

326 A. Arundel and N.S. Vonortas networks between SMEs and large ﬁrms. Another programme, RRIT, supports research and innovation networks in strategic technologies. The Dutch policy to support clusters was established in the early 1990s. Furthermore, the Dutch government’s procurement programmes for innovative technology favour networks between contractors. The UK gives a high priority to encouraging clusters, primarily at the local level, with most support provided by the Regional Development Agencies. These provide forums and workshops where staff from different ﬁrms can meet. A powerful method for creating knowledge ﬂows between ﬁrms is through collaborative research. Many collaborative research programmes are funded directly by the ﬁrms themselves without any government subsidy. Whether or not a member state provides a subsidy for interﬁrm collaboration depends on its general approach to supporting private R&D. Countries that primarily subsidize R&D through tax credits, such as the UK, the Netherlands and Denmark, rarely provide direct grants to subsidize cooperative R&D among ﬁrms. As an example, the UK only provides direct subsidies for collaborative pre-competitive R&D between ﬁrms in the aerospace and the pharmaceuticals sectors.4 Otherwise, its innovation policy stresses support for a favourable economic framework for business rather than direct ﬁnancial subsidies for private R&D. Other EU countries provide direct grants for collaborative R&D between ﬁrms, though often with some limitations. Direct grants provide cash to fund part of the costs of an innovation project. They are usually limited to 50 per cent or less of the costs, with the ﬁrm required to fund the remaining 50 per cent. The major concern with direct grants is that ﬁrms may use them to replace private funds, for research that they would conduct anyway. For this reason, direct grants are often targeted to research that ﬁrms would be unlikely to conduct without a subsidy, such as basic research in Belgium, or they are targeted to SMEs that lack the ﬁnancial resources to perform R&D. Germany, Italy and France provide direct grants, but tend to limit them to collaborative programmes that include SMEs. The largest innovation ﬁnance programme in Germany provides direct grants for research within 250 thematic areas. Several countries provide soft loans for private R&D. In most countries, soft loans do not require collaborative research, although they could promote knowledge ﬂows by increasing the innovative capabilities of ﬁrms. France, Germany, the Netherlands and the UK only offer soft loans to SMEs. In total, 30 per cent of KNOW respondents report receiving some form of public subsidy for R&D. Subsidies are more frequently reported by ﬁrms in the three manufacturing sectors (37 per cent) than by ﬁrms in the two service sectors (21 per cent). Even though more subsidies are available to SMEs than to large ﬁrms, larger ﬁrms in the manufacturing sector are more likely to report a subsidy (43 per cent of ﬁrms with over 250 employees compared to 34 per cent of ﬁrms with less than 250 employees). There is no difference by ﬁrm size in the service sector.

Innovation policies affecting knowledge ﬂows 327

14.4 Commercializing publicly-funded research Current innovation policy in ﬁve of the seven KNOW countries stresses the need for greater collaboration and knowledge ﬂows between PRIs and ﬁrms in order to help turn public investment in research into successful innovations.5 The two exceptions, Italy and Greece, recognize the importance of PRI-ﬁrm collaboration, but place greater emphasis on other areas due to the structure of their innovation system. For example, the focus of Greek innovation policy is on innovation ﬁnance and supporting start-ups in order to build up basic levels of innovative capabilities, while Italy’s efforts are focused on major reforms to the public education and research system and developing a strategic vision for R&D that will meet Italy’s future needs. The increased policy attention to PRI-ﬁrm collaboration is partly due to the fact that PRIs receive the largest slice of government funding and have no choice about implementing government policy. In contrast, it is both difﬁcult and undesirable for governments to try to coerce ﬁrms to perform R&D in speciﬁc areas that the government may view as strategic, or to take part in collaborative research. Two main types of policies are widely used to encourage the commercialization of publicly funded research. The ﬁrst consists of incentives for PRIs to conduct research of value to the private sector. These incentives are often designed to inﬂuence the activities of universities or institutions where the research agenda has traditionally been determined by academic criteria rather than by the needs of government or industry. The second policy area, which has attracted an enormous amount of attention and funding over the last decade, consists of ﬁnancial support for collaboration between ﬁrms and PRIs.

14.4.1 Incentives for PRIs Many member states support institutions with a speciﬁc mandate to conduct research of value to industry. The classic example is the Fraunhofer Institutes in Germany. Nevertheless, many of these institutions are under pressure to further increase the commercial relevance of their work, the efﬁciency with which technology is transferred to ﬁrms and the percentage of their operating costs that is funded by contract research. Long-established public research institutes often specialize in innovation of relevance to low or medium technology sectors such as agriculture or machinery with many SMEs. These ﬁrms often lack the ﬁnancial resources or expertise to solve technical problems in-house. The applied research institutes offer SMEs basic technical services for free or for a small fee. Ongoing concerns in Europe about being left behind in strategic or enabling technologies have led to the establishment of new research

328 A. Arundel and N.S. Vonortas institutes in advanced technologies such as ICT, nanotechnology and biotechnology, where commercial applications are fed by scientiﬁc advances. Many of them are virtual research institutes that link researchers from several universities, PRIs and ﬁrms. This results in considerable savings and is expected to increase the efﬁciency of existing expertise by improving knowledge ﬂows and cooperation. Virtual research institutes can also encompass both basic and applied research, since there is no existing ‘research culture’ that must be overcome. Examples include Denmark’s ‘Large Cross-Disciplinary Research Groups’ and the Thematic Research and Innovation Networks (RRIT) in France. In addition, basic and pre-competitive research institutes are usually established in strategic technologies such as biotechnology or microelectronics. Other programmes to encourage PRIs to conduct research of relevance to business include both programmes that actively direct research into business relevant research and passive programmes that establish the potential for contacts between academic researchers and ﬁrms. As an example of the latter programme, all EU countries now provide technology transfer ofﬁces (TTO) that can assist academics with establishing a commercial spin-off, patenting an invention or arranging a licensing agreement with a ﬁrm (OECD 2003). Both TTOs and science parks can also provide opportunities for contacts between industry researchers and academics. A few EU member states have introduced mechanisms to deliberately target academic research funds towards areas of value to industry. For example, the research councils in the UK are responsible for distributing funds for academic research. They use two mechanisms to target research towards areas of value to industry. First, they include representatives from industry who take part in the funding decisions and second, they use the results of the Technology Foresight reports to identify promising technologies with potentially large markets. Technology Foresight programmes have recently been established in the six KNOW countries, including FUTUR in Germany and QuickScan in the Netherlands. In some countries, such as Denmark, the Foresight exercises directly inﬂuence research priorities in PRIs, while in other countries the link is either not yet worked out or is indirect. In the Netherlands, PRIs can receive extra public funding if they have projects that are partly funded by a private ﬁrm. Over time, PRIs are expected to fund a percentage of their research from ‘third stream’ or private sources. Both the Netherlands and Denmark have revised legislation covering the mission of universities to include the dissemination and application of knowledge. 14.4.2 PRI-ﬁrm research collaboration All KNOW countries subsidize ﬁrms to either contract out research to PRIs or conduct collaborative research with PRIs. Examples include

Innovation policies affecting knowledge ﬂows 329 InnoNet and Prolnno in Germany, with a focus on SME-PRI linkages, MURST in Italy, IOP in the Netherlands, LINK in the UK and the Innovation Consortiums in Denmark. In many programmes, the ﬁrms pay all of their own research costs, but government funds the PRI costs (up to 50 per cent of the total). This type of subsidy is justiﬁed by the need to overcome some of the disadvantages of contracting out research or collaborating with universities and PRIs. These include concerns over conﬁdentiality, higher risks for the basic and pre-competitive research where the expertise of many PRIs lies, and a preference for ﬁrms to keep more applied and commercial research in-house. In addition to producing research output of value to industry, these programmes can assist in developing expertise within universities and PRIs on problems of importance to industry. With such a diversity of programmes to encourage and ﬁnance PRIﬁrm linkages, one would expect PRIs to be an important source of knowledge for the innovative projects of ﬁrms. An encouraging result from KNOW is that 38 per cent of the respondent ﬁrms report one or more research project that involved PRIs. However, other KNOW questions show that PRIs are relatively unimportant, compared to other information sources. For example, only 11 per cent of ﬁrms report that PRIs contributed to the idea behind their most important innovation, compared to 31 per cent that cite their suppliers, 32 per cent that cite competitors and 49 per cent that cite customers. The results are similar in terms of who contributed to the completion of this innovation. Twelve per cent of the ﬁrms cite PRIs, compared to 35 per cent that cite customers and 39 per cent that cite suppliers. Overall, only 5 per cent of the KNOW respondents stated that PRIs were their most important contributor to the completion of this innovation. The lack of importance of PRIs could be due to the recent implementation of many policies to encourage PRI-ﬁrm links, or these links may not be particularly valuable for smaller ﬁrms or for ﬁrms in the ﬁve sectors covered by KNOW.

14.4 Conclusions The increasing sophistication of science and technology (S&T) policy in Europe has led to a gradual refocusing of the overall policy target from passive support for the creation of new ideas to a concerted effort to ensure that these ideas ﬁnd their way to ﬁrms that can apply them to their new products, processes and services. One consequence of this change in S&T policy, due to the adoption of a systems approach to innovation, is that all EU member states emphasize the need to promote knowledge ﬂows between ﬁrms and between ﬁrms and PRIs. The current policy focus in many EU member states is to encourage universities and government research institutes to direct their research efforts to areas that are of interest to private ﬁrms and to improve the transfer of their expertise and

330 A. Arundel and N.S. Vonortas knowledge to ﬁrms, particularly SMEs. Nevertheless, the KNOW results show that these policies do not appear to have had much of an effect on the use of PRIs as a source of knowledge for SMEs. Another thrust of innovation policy in many EU countries is to support regional and sectoral networks of innovation that include ﬁrms and PRIs. Other programmes maintain a longstanding technology transfer infrastructure to improve the absorptive capacity of SMEs. These provide basic educational courses on innovation management and technology audits and are designed to help ﬁrms identify technical problems and develop innovative solutions.

Notes 1 See Caloghirou and Vonortas (2000), Caloghirou et al. (2004) and Peterson and Sharp (1998) for extensive discussions of cooperative R&D and related policies in Europe. 2 The country policy reports are available at www.trendchart.org under ‘Country Pages’ (last accessed 7 January 2004). See EC (2002) for a policy overview for 2002. Diederen (1999) also provides a policy overview for EU member states in the late 1990s. 3 TrendChart Country Report for Germany, October 2002. 4 This is stressed by UK policy documents as part of the UK’s ‘free market’ approach to innovation. However, it is interesting to note that aerospace and pharmaceuticals account for almost half of all manufacturing R&D in the UK. Therefore, the UK still manages to provide direct subsidies to its most innovative manufacturing industries. 5 Such cooperation is a major emphasis of recent policy documents in Germany (BMBF 2002), France (Journal Ofﬁciel 1999) and the Netherlands (EZ 2002).

References BMBF (Federal Ministry of Economics and Technology) (2002) Innovation Policy: More Dynamic for Competitive Jobs, Berlin, April. Available at: www.bmbf.de/pub/ innovation_policy.pdf, last accessed 7 January 2004. Caloghirou, Y. and Vonortas, N. (2000) ‘Science and technology policies towards research joint ventures’, ﬁnal report of the Project SOE1-CT97-1075, funded under the TSER Programme, DGXII, European Commission, Brussels. Caloghirou, Y., Ioannides, S. and Vonortas, N.S. (eds) (2004) European Collaboration in Research and Development – Business Strategy and Public Policy, Cheltenham: Edward Elgar. Diederen, P. (ed.) (1999) Innovation and Research Policies: An International Comparative Analysis, Cheltenham: Edward Elgar. European Commission (2002) ‘Innovation Policy in Europe 2002’, Innovation Papers No. 29, Ofﬁce for Ofﬁcial Publications of the European Communities, Luxembourg. Available at: www.trendchart.org/reports/annual_home.html. EZ (Ministry of Economic Affairs) (2002) ‘Samenwerken en stroomlijnen: opties voor een effectief innovatiebeleid’, May, The Hague. Journal Ofﬁciel, Loi 99-587, 13 July 1999. For information in English, see

Innovation policies affecting knowledge ﬂows 331 www.france-science.org/innovation/innovationlaw.asp, last accessed 7 January 2004. OECD (2003) Turning Science into Business: Patenting and Licensing at Public Research Organisations, Paris: OECD. Peterson, J. and Sharp, M. (1998) Technology Policy in the European Union, London: Macmillan Press.

15 Towards a new agenda for enhancing knowledge ﬂows in Europe Yannis Caloghirou, Anastasia Constantelou and Nicholas S. Vonortas 15.1 Introduction As innovation has become more international, pervasive and distributed, the interest of strategists and policy makers in the extent and type of knowledge ﬂows as well as in the channels and mechanisms of dissemination that facilitate such ﬂows has increased dramatically. The chapters in this book constitute an extensive theoretical and empirical appraisal of this multi-faceted phenomenon, with a focus on innovation-related knowledge ﬂows affecting European industry. The analysis in this book has been based entirely on disembodied ﬂows of knowledge between ﬁrms, universities and other education and research institutes. Disembodied knowledge ﬂows have been the subject of much less systematic empirical analysis in economics, reﬂecting in part the interests of received theory and in part the lack of appropriate data. Since the mid 1970s, however, our understanding of knowledge spillovers, technological opportunities and the market processes underlying technological progress in different settings has advanced considerably, which has created the necessary preconditions for exploiting newly available sources of empirical information. Our analysis has also followed an innovation systems approach. Elaborated in Chapter 2, this approach allows a more holistic examination of the various angles of the phenomenon being studied and especially the interactions among the different players involved in knowledge exchange. Although we do not pretend to have discovered a grand unifying theory, the evidence-based chapters in this volume (Chapters 5–13) cover a remarkable breadth of issues, including the nature and dynamics of alliance and knowledge networks, patterns of participation, instruments for intellectual property rights (IPR) protection, learning sources, and the beneﬁts to be derived from cooperative research and development (R&D). These chapters investigate knowledge ﬂows from a microeconomic point of view, using the individual organization – ﬁrm, university, other public research organization (PRO) – as the basic unit of analysis. The international research team working on KNOW created several

Knowledge ﬂows – a new research agenda 333 rich, longitudinal sources of empirical information reﬂecting: (1) company participation in European Framework Programmes (FWPs) on research and technological development (RTD) based on data from the CORDIS database; (2) patents and patent citations based on data from the European Patent Ofﬁce (EPO); and (3) company behaviour regarding innovation-related knowledge ﬂows based on a survey of more than 600 European companies and a large set of interviews. In light of the Community’s objective to create the foundations of a European Research Area (ERA), this concluding chapter summarizes the results of the collective effort of contributors in this volume to amass a comprehensive dataset of quantitative and qualitative information to support wide-ranging analysis on knowledge ﬂows in European industry. Parallel to this, it highlights key messages addressed to policy makers and industry practitioners regarding measures to promote and enhance national and cross-national channels of knowledge ﬂows in the light of the Lisbon objectives to transform the continent into the most advanced knowledge economy in the world within a decade.

15.2 Framework Programme networks Social network analysis has been employed in this book to describe the structural properties and dynamics of the cooperative network supported through FWP RTD between 1992 and 1996 across all technology areas. This work has characterized the network as dense and pervasive, branching around a large ‘oligarchic core’ (of organizations) whose centrality and connectivity have strengthened over the years. In this network, high connectivity is strongly dependent on a core of central actors that take part in many research projects, frequently as Prime Contractors. Around this pivotal group, two ‘lower layers’ of actors are identiﬁed: (1) a minor group of fairly frequent but low-proﬁle participants, which enter consortia as partners and take advantage of the FWPs to establish links with several leading actors; and (2) an extremely large number of partners for whom participation in a FWP is an exceptional event. In purely topological terms, it can be claimed that this three-layer structure ensures cohesion and efﬁcient transmission of knowledge since the network core is never further than two or three steps away from even the most peripheral agents. The ‘small world’ characteristics of this network imply that it must be an efﬁcient channel of knowledge transmission. Hence, while on the one hand the overall inter-organizational network developed through FWPs has been shown to have acted as the foundation for efﬁcient channels of knowledge communication, on the other hand, this network is very heterogeneous, and is built around a dense core including some of the largest and most innovative corporations and a number of outstanding academic institutions and other PROs. Were several of those core actors to be removed, the current network

334 Y. Caloghirou, A. Constantelou and N.S. Vonortas conﬁguration would collapse and the ability of the remaining organizations to communicate with each other would dramatically diminish. This organizational core seems to have for long been established and has been perpetuated and strengthened in the course of successive FWPs through the typical processes of network self-organization such as preferential attachment and reputation effects. It is important that policy makers recognize the danger that such self-organizing processes might also produce lock-in phenomena and network ossiﬁcation. Related to the above, there is also evidence of some degree of selfselection in participation in European FWPs, based on the positive correlation between patenting activity and the participation rates of individual organizations. While this correlation might be an indication of the FWPs beneﬁting the participating organizations, it could also be a reﬂection of the more innovative agents showing a higher propensity to participate frequently (and become part of the network core mentioned above). However, if a distinction is made between different technological areas then evidence of company self-selection in FWPs is less clear cut and is even more difﬁcult to detect in different size classes. Comparison between the ﬁeld of information and communication technology (ICT) and medical and biotechnology (MB) is stark: while the set of cooperative research programme participants examined tended to be more innovative than non-members in both cases, ICT programmes have attracted highly R&D-intensive ﬁrms that were already signiﬁcantly more innovative than the average European level, whereas MB programme participants did not exhibit high levels of patenting prior to entry. And, while in ICTs size was found to be positively related to patenting activity, no such clear relationship emerged for MB. That is to say, FWPs seem to have reinforced existing leaders and networks in the ICT area, a relatively more ‘mature’ ﬁeld with an established ‘network of excellence’ and a fairly stable hierarchy of innovators. At the same time the FWPs seem to have favoured the exploitation of the innovative potential of new actors in MB, a more ‘ﬂuid’, emerging ﬁeld.

15.3 Intellectual property protection Regarding intellectual property protection, only a small fraction (less than 15 per cent) of all participating entities in the cooperative research network of the FWPs had registered patent applications with the EPO during 1978–98. However, the knowledge-intensive network emerging from their patent citations has a clear European bias. About half of all citations were directed to European organizations of all types, with the remaining directed primarily to organizations in the United States or Japan. In most countries, secrecy is the preferred strategy for the protection of intellectual property. Developing lead-time advantages is also very import-

Knowledge ﬂows – a new research agenda 335 ant. With the exception of chemicals, patenting ranks quite low in terms of frequency of use. The value of patents is particularly low in information technology (IT) service sectors. Firm size seems to be related partly to the reported low priority of patents as a mechanism of intellectual property protection: a larger share of mid-sized ﬁrms (above 250, below 1,250 employees) than smaller ﬁrms cite patents as being important; smaller ﬁrms are more likely to rate secrecy highly. Almost twice as many midsized companies as small ﬁrms have patented their most important innovation. Other important factors related to patenting activity include the nature of the technology (chemicals patent more), industrial structure (ﬁrms with many competitors prefer secrecy) and the innovation activity in the ﬁrm (R&D continuity positively inﬂuences the propensity to patent). The tendency to patent also varies across countries.

15.4 Sources of learning The KNOW survey, based on small and medium-sized enterprises (SMEs) in manufacturing and service sectors in seven European countries, indicated that traditional activities such as attending trade fairs and conferences, and reading scientiﬁc and business journals are the most important sources of new ideas for innovation. Reverse engineering is also important. European SMEs do not use patent database searches to obtain ideas for innovation. Customers, suppliers and competitors are very important sources of innovation-related information for the surveyed SMEs, which is in line with the picture that emerged from the pan-European Community Innovation Surveys. According to the KNOW survey respondents, the economically most valuable innovations are pulled by demand: customers are the dominant sources of the original ideas for innovations. On the whole, suppliers and competitors are also seen as important sources of knowledge for innovation. There are signiﬁcant differences between countries in this regard. In addition to serving as frequent sources of original ideas, customers and suppliers are most frequently cited as being contributors to innovation completion. As expected, internal knowledge as a contributor to innovation is highly valued in all countries, and especially in Germany and Britain. Italian ﬁrms seem to have the most balanced approach to internal versus external sources of information. Dutch ﬁrms seem to be more open to external sources of innovation than their counterparts in other countries. It is interesting that for the surveyed SMEs national sources dominate as the most important external sources of innovation-related knowledge. Firms from the smaller countries such as Greece, Denmark and the Netherlands tend to be more internationally orientated than those located in the larger countries. The most important reasons for obtaining knowledge from external

336 Y. Caloghirou, A. Constantelou and N.S. Vonortas sources reportedly include reducing development costs and risks, increasing the technical expertise of the ﬁrm, and building on the research ﬁndings of others. Scientiﬁc and technical information is the dominant type of knowledge obtained from external sources, followed by knowledge relevant to market introduction. By far, the most frequent method of communication with external knowledge sources is informal personal contact, followed by research cooperation. Exchange of personnel and other methods are more important in some countries (for example, France and the Netherlands) than in others. The large majority of surveyed ﬁrms use the Internet regularly in everyday business (a tendency that we would assume will have increased since the KNOW survey was conducted). Firms in the computer services sector lead in Internet use, followed by chemical ﬁrms. Internet access is very poor only in the food and beverages sector. The lowest use of the Internet by countries was reported for Greece. Almost all Internet users used it to access scientiﬁc and technical information and to communicate with their suppliers, customers and collaborators. Internet use was found to be positively related to the level of scientiﬁc personnel, R&D intensity and the size of the ﬁrm.

15.5 Beneﬁts from cooperative R&D Previous research has shown that the beneﬁts from cooperative R&D are positively related to the ﬁrm’s in-house technical capabilities, especially the ability to undertake R&D on its own. Cooperation, therefore, seems to complement rather than substitute for internal technical capabilities. In order to beneﬁt from cooperative R&D, a ﬁrm must continually upgrade its knowledge base and technical capabilities. Work based on the KNOW survey also indicates that location strongly affects the probability of developing an innovation through collaboration. Firms in Germany, Italy and Greece are signiﬁcantly less likely to enter into collaborations than ﬁrms located in the Netherlands. This might reﬂect cultural differences – e.g. Dutch ﬁrms were found to be the most open to external sources of information – or differences in national innovation systems, but the survey does not enable a full exploration of this result. In all ﬁve sectors surveyed for KNOW, almost two-thirds of product and service innovations were developed in-house, between 9 and 13 per cent were bought in, and around 20 per cent had been developed via collaborations. Empirical results show that the share of innovations developed inhouse has a positive and statistically signiﬁcant effect on the innovative sales share of the surveyed ﬁrms, as opposed to the share of innovations developed via collaborations. Although there are beneﬁts to collaboration, this result raises questions regarding the advantages of too frequent collaboration, implying some kind of threshold. More work is needed in this area.

Knowledge ﬂows – a new research agenda 337 Finally, although according to the KNOW survey PROs do not play a central role in the innovative process of SMEs, many such ﬁrms do develop cooperative R&D projects with PROs. The probability of R&D collaboration with PROs has been shown to depend on ﬁrm size: larger ﬁrms are much more likely to be involved in collaborative R&D. However, the number of R&D cooperations is affected by the ‘relative size’ of the ﬁrm as measured by R&D employment: R&D intensity affects both the propensity for and the extent of engagement in R&D projects. It must be noted that the ‘openness’ of SMEs, that is, their willingness to signal their competencies and screen the outside world for information, positively affects the likelihood of collaboration with a PRO, but not the level of collaboration developed. SMEs that patent to protect innovations (and signal competencies) and ﬁrms that have received public subsidies have both a higher probability of taking part in cooperative R&D with PROs and a higher number of collaborations, although the impact of subsidies on the extent of the collaboration is mediated by country-speciﬁc effects.

15.6 Policy implications The research results reported in the preceding chapters have important policy implications, including the following. •

The channels and mechanisms of knowledge ﬂows deﬁne some of the most important connecting links between the various parts of innovation systems. They relate directly or indirectly to all policies that affect such systems. At minimum, they relate to the entire spectrum of science, technology and innovation policy, being particularly akin to policies affecting knowledge access and dissemination and to policies affecting learning processes. Knowledge ﬂows are also directly related to intellectual property protection and competition policies that create the infrastructure supporting formal interaction among economic agents in production and innovation systems.

Chapter 14 brieﬂy summarized EU member state policies affecting innovation-related knowledge ﬂows. This review has referred only to technology and innovation, and limited its coverage to policies directed to enhancing the absorptive capabilities of SMEs, policies addressing collaboration in research and policies relating to the commercialization of publicly-funded research. This coverage could obviously be greatly expanded to include all policies of relevance to the ‘knowledge’ or ‘learning’ economy (Lundvall and Borras 1998; OECD 2001). Governments should be aware that most of their science, technology and innovation policies will have an impact on the channels, direction and intensity of knowledge ﬂows affecting industry.

338 Y. Caloghirou, A. Constantelou and N.S. Vonortas •

The importance of policies concentrating on national channels of knowledge ﬂows remains high. International channels are, however, developing fast and will attract increasing policy attention. Coordinating the two will be essential, especially in close-knit country groupings like the European Union. This is acknowledged in the concept of the ERA.

National channels for knowledge ﬂows are still more important than international channels: a long series of surveys, including the KNOW survey and in-depth interviews, has underlined a continuing dominance of national over international channels. While these results reﬂect in part the ‘localization’ of knowledge, they should be interpreted with care as they may often over-emphasize the importance of national channels for SMEs. Results may also be sector-sensitive. Globalization is encouraging increased international links, which tend to be exploited ﬁrst by the more innovative and/or larger ﬁrms.1 Most importantly, long streams of European RTD programmes have created very strong networks among economic agents across Europe. The current efforts directed to the creation of an ERA, which call for more extensive coordination of supranational (European), national and regional science, technology and innovation programmes and policies, seem to be most appropriate. •

Policies to enhance the absorptive capabilities of ﬁrms are key and probably more important than at any other time.

Available survey work indicates that, on average, ﬁrms rate the contribution of internal knowledge sources to their innovative activities more highly than the contribution of external sources. While empirically correct, this statement needs to be qualiﬁed in terms of (a) ﬁrm size and sector – larger ﬁrms and those operating in certain sectors tend to be more open to external sources of innovation, and (b) degree of innovativeness – more innovative ﬁrms are above the industry average in their use of external knowledge sources. The importance of openness to external sources of knowledge notwithstanding, the relationship between internal and external sources must be stressed. Rather than being substitutes, it is now clearly understood that the two are complementary. A ﬁrm must be internally competent to be able to tap into and beneﬁt from external knowledge ﬂows. For example, the beneﬁts of cooperative R&D are positively related to the ﬁrm’s internal capabilities, particularly with respect to conducting R&D and developing relationships with other organizations. In order to gain from R&D cooperation, a ﬁrm must continuously upgrade its knowledge base and capabilities. Hence, all sorts of policies that enhance industry’s absorptive capacity – understood in its broad sense as incorporating the ability both to access

Knowledge ﬂows – a new research agenda 339 and to utilize knowledge – necessarily enhance the beneﬁts to industry from a given level of knowledge ﬂows. •

SME’s innovation efforts are strongly dependent on large, important customers and suppliers.

The lifeblood and great advantage of SMEs lie in their ability to respond quickly to demand. Ensuring continuing demand for advanced products from their large customers, and strengthening their links with sophisticated suppliers, will be extremely effective incentives for SMEs to innovate. SMEs consistently report that the most important sources of information and innovation ideas are customers, suppliers and competitors. With the exception of certain industries such as biotechnology, large companies generally use universities and PROs much more than do their smaller counterparts. •

The Internet has not replaced traditional channels of knowledge ﬂows nor is it expected to do so in the foreseeable future. However, it is proving to be an additional, very important channel for communication and knowledge exchange. Most ﬁrms have embraced it enthusiastically. Policy can increase access and facilitate utilization of the Internet, a goal that falls squarely within the Lisbon objectives.

The majority of ﬁrms surveyed note that external information sources contribute both to the original idea behind the economically most important innovation and to its completion. They access these sources of information mainly through traditional mechanisms of knowledge communication and transfer such as trade fairs, conferences, scientiﬁc and business journals, and reverse engineering. Nevertheless, the vast majority of the SMEs surveyed in KNOW use the Internet regularly in everyday business, particularly to get scientiﬁc and technical information and to communicate with their suppliers, customers and collaborators. Other beneﬁts with respect to Internet use by SMEs include the ability to benchmark the performance of competitors, create new business opportunities and access information rapidly. Government policy could help transform the Internet from an information medium into an integrated strategy tool. The Lisbon objectives of creating in Europe the most advanced knowledge-based society in the world certainly include such a goal. And the Commission’s i2010 initiative2 as well as several areas of concentration of DG Information Society clearly aim at making this goal reality. •

Despite the contemporary climate for stronger IPR protection, most European SMEs do not search patent databases for creative ideas or strive to apply for patents.

340 Y. Caloghirou, A. Constantelou and N.S. Vonortas In most countries, secrecy remains the preferred intellectual property rights protection strategy for SMEs along with lead-time advantages which are also seen as very important. With the exception of the constellation of chemicals and pharmaceuticals sectors, patenting as a mechanism for IPR protection is used by industry less frequently than commonly understood. The value assigned to patents is particularly small for the services sector, as demonstrated by the results from the ICT service sectors surveyed in KNOW. Firm size seems to be related to the reported low priority given to patenting. Smaller ﬁrms tend to apply for patents less often than larger ones. Other contributing factors are the nature of the technology, the industrial structure and the innovation activity in the ﬁrm. It is still debatable whether stronger intellectual property protection is desirable across the board. •

The implementation of policies since the mid 1980s to promote cooperative R&D has resulted in the formation of impressive knowledge communication networks across Europe. This thrust is expected to be maintained through the establishment of the ERA.

R&D collaboration involving ﬁrms, universities and other PROs has become an important source of knowledge ﬂows. There are strong indications that the European FWPs for RTD have created highly connected and dense networks with efﬁcient structures for knowledge communication. The FWPs were found to have interconnections in the form of common participants. Organizations other than ﬁrms (universities, consultants, etc.) have played a very important role in establishing links among cooperative R&D programmes. While most organizations participate in these programmes only infrequently, there is a cadre of technology and innovation leaders who have become part of a tightly-knit core in these networks. •

A relatively small number of organizations, primarily large companies, universities and PROs, have emerged as core players in European cooperative R&D activities and play a disproportionately more important role in maintaining channels of communication. It is anticipated that these same organizations will emerge as the core players in the ‘networks of excellence’ and ‘integrated projects’ currently being implemented in the context of the sixth FWP for RTD.

The FWP research network is characterized by a very large number of peripheral agents, with one or a few formal connections, coexisting with a relatively small number of central players with large numbers of connections that play an extremely important role in maintaining communication among distant nodes. Prime Contractors have, on average, participated in a much larger number of cooperative projects compared to other partners. Most networking activity in this cooperative R&D network occurs mainly among Prime Contractors who are the central

Knowledge ﬂows – a new research agenda 341 actors. The frequency distribution of Prime Contractors is biased towards universities and large industrial groups. In addition, the distribution of innovative output on the basis of the patenting activity of research partners is highly skewed in favour of Prime Contractors, and especially companies. Marginal participants are, on average, the least innovative. It would not be unreasonable to expect that the same or similar organizations emerge as the core players in the new instruments of the sixth FWP. There are positive and negative aspects to this. On the positive side is that these organizations possess the capabilities to play leading roles. The negative aspects are the reduced potential for newcomer entry in the inner circle of the network and the phenomenon of negative lock-in. Policy must ensure that there are no unreasonable barriers to entry. •

The knowledge-intensive network that the innovative participants in collaborative R&D projects refer to and benchmark against, is strongly rooted in Europe.

Intra-network patent citations in FWP cooperative R&D projects have tended to go to European patent holders. The most central actors receive the most intra-network patent citations. This result runs counter to the more pessimistic arguments concerning the ability of European industry to be at the forefront in state-of-the-art technologies relative to their counterparts in the United States and Japan. •

Cooperative R&D programmes will have differential effects across industries and technology ﬁelds depending on the degree of maturity of the industry. Attention to sector dynamics is warranted during programme design and evaluation.

Evidence of these differential effects comes from the Information and Communication Technologies (ICT) and Medical and Biotechnology (MB) areas of the FWPs. On the one hand, in the ICT area European programmes have largely attracted R&D-intensive ﬁrms that were already signiﬁcantly more innovative than the average European level (self-selection), whereas in the MB area early research partnership members did not exhibit high levels of patenting prior to entry. On the other hand, while there is no clear and robust evidence of a positive correlation between patenting activity and research partnership participation in the ICT ﬁeld, there is evidence of this in the MB ﬁeld. Such differences could well be related to the respective life cycles of these industries. Cooperative policies seem to have attracted and reinforced existing leaders and networks in the more mature of the two industries, where a network of excellence exists and where the hierarchy of innovators is generally stable, while they seem to have favoured the exploitation of innovative potential by new actors in an emerging technology area.

342 Y. Caloghirou, A. Constantelou and N.S. Vonortas If this is actually the case, then the early stages of technological development and competition in an industry policy should be directed towards creating new networks of excellence and opening up existing networks to potential innovators by promoting R&D-intensive programmes that are strongly technology-oriented. In later stages of the life cycle, when the industry is technologically mature and networks of leading actors are well established, it would be more effective for policy to try to link peripheral actors to extant networks in order to achieve a broader diffusion of knowledge and safeguard against the use of collaboration to create barriers to entry. •

Policy must recognize the heterogeneity in the relationships between industry and PROs.

Relationships between ﬁrms and PROs are characterized by a high degree of heterogeneity. Policy must be responsive to such differences. It is also essential that policies supporting collaboration between PROs and ﬁrms should be based on real incentives for both parties. In the past, they have been mainly directed to forcing PROs into establishing these types of relationships with no acknowledgement that without constructive ‘demand’ little can be achieved. This demand is not always forthcoming depending, as was shown in earlier chapters, on the type of industrial structure, technology characteristics and ﬁrm characteristics. •

Geographical proximity matters for knowledge ﬂows and can strongly inﬂuence the localization of production and innovative activity. The reasons for this are complex and call for equally complex policy approaches to create competitive advantage for regions.

Pioneering work in Europe (including the KNOW survey) and the United States has pointed to geographical clustering in knowledge-related activities. Technological knowledge and spillovers seem to be (geographically) localized. Weighted data for Europe’s largest ﬁrms indicate that the sourcing of technical knowledge from universities and other PROs is subject to localization effects: domestic public research is typically rated more important than foreign-sourced research. There are preliminary indications of clustering among organizations in neighbouring countries for participation in FWP cooperative R&D ventures. The preconditions for dynamic knowledge-related clusters are not easy to achieve. For example, the geographical distribution of a sector such as ICT can be reasonably expected to depend upon: i A ‘metropolis’ effect resulting from the fact that many of the ICT service activities are typically concentrated in cities.

Knowledge ﬂows – a new research agenda 343 ii The supply of skilled labour, which is expected to be a function of the location of universities and business schools, often determined by government decisions. iii ‘Random’ location of manufacturing ﬁrms based on the personal preferences of their founders. Such patterns can be seen in Denmark where a strong ‘metropolis’ effect on regional ICT specialization is counterbalanced by the rather decentralized nature of the public education system. On the whole, in ICT there seems to be a close correlation between the distribution of government-ﬁnanced R&D, higher education institutions and the regional distribution of private employment. Engineers and computer scientists typically choose to work near to these institutions, around which more specialized small-scale clusters often emerge. In other words, creating local competitive advantages requires complex policy initiatives.

15.7 A ﬁnal word In conclusion, we want to stress again the multi-faceted nature of the subject matter of this book and its importance for policy. The theoretical part of the book has argued strongly that innovation-related knowledge ﬂows deﬁne the links that make up production and innovation systems.3 As such, they relate directly or indirectly to every policy that affect such systems. Governments wishing to be effective in leading their citizenry into the new, knowledge-intensive, ‘learning’ era must be cognizant of the fact that their industrial, science, technology and innovation policies will inevitably impact the channels, direction, and intensity of knowledge ﬂows affecting industry. As a result, the concern over knowledge ﬂows cannot be disassociated from – in fact, it is intimately linked with – the broader policies that affect production and innovation systems, be they science, technology and innovation policies, intellectual property protection policies, competition policies, procurement policies, and so forth. Such a message is in concert with contemporary policy thinking in Europe as reﬂected in the discussions over the ERA, the seventh FWP for RTD, and the i2010 agenda.

Notes 1 This becomes more important when considered in conjunction with the observation that customers and suppliers are the most important source of information for SMEs for new technologies and products. If major customers or suppliers look abroad, small ﬁrms that supply them or buy from them will do the same. 2 The i2010 initiative launched by the European Commission in June 2005 is a comprehensive strategy for modernizing and deploying all European policy instruments to encourage the development of the digital economy: regulatory instruments, research and partnerships with industry. 3 See Figure 2.1 for an illustration.

344 Y. Caloghirou, A. Constantelou and N.S. Vonortas

References Lundvall, B.-Å. and Borras, S. (1998) The Globalising Learning Economy: Implications for Innovation Policy, Brussels: European Commission. Organization for Economic Cooperation and Development (OECD) (2001) The New Economy: Beyond the Hype, Paris: OECD.

Index

Aalborg 140, 142; companies 127, 130; Science Park (NOVI) 155; University 133 absorptive capacity 67, 74, 102, 160, 199, 204, 251–2, 262, 323–5, 330, 338–9; indicators 62–3 AMADEUS database 49, 214, 236–7, 247 appropriation 54, 63, 178, 183–4, 203–4, 250–1; conditions 162, 170; method 77, 168–9, 171; preferences 173, 177; signalling strategy 198 Austria 258–9 barriers to entry 303, 341–2 Belgium 57, 64, 258–9; basic research in 326 bibliometrics 46, 50, 185–6; citations 50 biotechnology 7, 52, 116, 155, 159, 212, 228, 322, 328; collaboration 179; research 49, 236–7, 245–7, 275 bridges 302, 309, 317 business 16, 115, 118; registries 180, 193; schools 138, 143, 343 buying in 160–3, 165–6, 186; rate 177 Canada 40, 49, 128; Innovation Survey 61, 161 Carnegie Mellon Survey R&D survey (CMS) 54, 186, 197 centres of excellence 256, 272–4, 294 channels of knowledge 71, 76, 185–7, 337; acquisition 197; communication 332, 334, 340 citations 46, 50, 61, 64, 311; pairs 47, 12, 314 cluster 116, 137, 142, 144, 253, 273, 292, 297, 300, 304–10, 312–17; analysis 115, 136

clustering: coefﬁcient 287, 290–1, 297; organizations 281–2; sectoral 325–6 codiﬁed information 21, 51, 70, 242–3, 262 codiﬁed knowledge 7, 11, 13–14, 32–4, 40, 50–2, 54, 56, 102, 159; sources of 197 cohesion 269–71, 273–4, 294, 333; countries 256, 262; objective 273, 294 collaboration 18–20, 46–7, 77–9, 94, 96–8, 128, 158–62, 165–8, 170–8, 197, 251, 254–6, 277–8, 282–3, 325–7; advantages 207, 337; agreements 35, 104, 270–2; disadvantages 329; external 106, 203; formal 185, 187, 194–5; intensity 198, 203; partners 105–7, 110, 163, 194; with PROs 188, 197–9, 201–7; strategic 301; transnational 48–9, 244; with universities 187, 197, 200 collaborative research programmes 15, 46, 64, 171, 186, 243, 321, 325–8; European 21, 311 collaborators 68, 79, 111, 159, 287–8, 293 commercial applications 98, 165, 322–3, 327–8 Commission for the European Communities 206 communications 40, 45, 80, 89, 154, 214, 339; channels of 70, 73, 186, 195; infrastructures 38 Community Innovation Survey (CIS) 30, 41, 55, 60–3, 75, 159–62, 184, 186, 198; CIS I 54, 57, 61; CIS II 59, 187, 199; CIS III 56, 197, 200 competition 37, 76, 102, 111–13, 137, 172, 270, 294; international 245, 268, 270, 273; policies 337, 343; pressure 107, 168

346 Index competitive advantage 13, 101, 117; local 343 competitors 52, 57–8, 64, 68–9, 96–7, 186, 189–93, 207, 254, 329, 335, 339; cooperation with 61; main sector 233; products 79 complementary knowledge 102–3, 110–11, 198 computer services sector 124–5, 141–2, 146–7, 162–8, 173–9, 184, 188, 195, 204, 253–4, 262, 325, 336; market 108 connected caveman world 306–10, 314 consortia 220, 268, 273, 295; afﬁliation 238; composition of 296; European 217; limitation of 274; multinational 135, 271; research 58 consumer electronics 121, 124–5, 140–2, 147 contractions 310, 314; relative frequency of 315 cooperation 21, 102, 111–13, 187, 196, 247, 262–5, 272–4, 295, 302, 328, 336; beneﬁts 234; distribution 194–5; international 213, 245, 256, 321; level of 106, 203–4; partners 104, 160; patterns 242–3; research 17, 63, 79, 89, 336 cooperative agreements 63–4, 69, 187, 194, 250 cooperative research 95, 186, 211, 233–4, 330; behaviour 207; beneﬁts from 333, 336; between ﬁrms 326; in ICT 218; programmes 183–5, 220, 233, 268, 334; with PROs 200, 337; R&D 39, 55–9, 60–1, 161, 187, 195, 204, 250–1, 339–41 CORDIS database 15, 237, 265, 296, 311, 333 core 291–2, 294–7, 333–4, 340; competencies 105; density 334; players 340–1; of RJV network 290–1, 293–4; strategic activity 207 country 170, 178; developed 144; effects 163, 170, 178, 196, 200–4, 205; policy reports 330; rate of growth 236; share 257; shift-share effects 259; size 256 customer 52, 56–7, 68, 79, 83–4, 108–9, 177, 187–91, 193, 329, 339, 343; communications 94–7, 335–6; key 69; as knowledge sources 62, 105, 109; satisfaction 41

degree of nodes 253, 287, 297, 305, 308, 310, 314; distribution 282–5 Denmark 16, 18–19, 64, 77–93, 104, 122–44, 152–4, 162, 167, 174–6, 189, 207, 258–9, 322–8, 336, 343; companies 80, 117, 127–9, 136, 140–3, 324; consultancies 128–9, 323; Copenhagen and region 140–3, 155; private sector employment 120; research groups 328; specialization 127–8; universities 127, 130, 133, 140–2 Denmark ICT sector 115–55; employment 118, 140; employment shares 120, 152–3; export performance 121 development costs and risks 88; reduction 94, 97–8, 193–4, 232, 250 dissemination 35, 273; of information 244–5; of knowledge 54; of results 213 distance 36, 281, 290; average 286, 288, 292; between agents 302; between nodes 300 domestic partners 21, 61, 254, 256; lack of 262 dot.com segment 3, 5, 129, 137 econometric model 70, 168, 183–5, 187, 195–6, 205, 234 economic framework 28, 36, 144, 326 economic growth 1, 11, 258; potential 38–9 economies of scale 103, 270, 301 edges 281, 304, 309–12 education 37, 42, 143; decentralized 69, 132–3, 343; level of 131; programmes 323–4; report 130; and research 142, 327, 332 electronics 116, 123–5, 148, 160, 212, 236, 254 employees 77, 104–5, 107, 162–3, 176, 188, 196, 201–2; graduate 130–1, 324; highly skilled 49; in ICT 130, 132; job rotation schemes 106; mobility of 179; number of 170, 232; qualiﬁed 82; in R&D 171; rewards for 111; rotation of 104; value added 120; see also personnel staff employment 184, 199, 237, 245, 271; class 223; data 117–18, 154, 223–7; growth 60, 236; increase 120; level 223; private 143, 343; regional 142, 151; share 138; of students 185

Index 347 enablers 197, 204, 293 ERA 21, 269–70, 274, 293–6, 300, 303, 338, 340, 343 ESPRIT 49, 212, 243, 270, 275 EUREKA 64, 211–14, 218, 226–8, 234, 237–8, 263, 311, 321; projects network 21, 314, 317 Europe 40–2, 88, 137, 186, 237, 321, 325, 327, 329, 341–3; economy of 41; national systems 42; surveys in 52; western 117, 143, 154 European Commission 14, 63, 158, 187, 244–5, 256, 265, 269–74, 275, 295, 299, 311, 322 European Committee on Legal Affairs and the Internal Market 273 European industry 14, 16, 63, 76–7, 128, 228, 243, 270, 332; competitiveness 211, 268; cooperation in 237; standard for 222 European Patent Ofﬁce 46, 212, 237, 333–4; database 15–16, 49, 76, 212, 237 European policy 22, 45, 303; instruments 21; policy makers 17, 59, 300 European research 227–9, 233, 272, 274, 277; centres 243; in ICT 228; infrastructure 269; networks 21, 303, 311, 317; structural weaknesses of 272 European Research Area (ERA) 10, 268, 270–1, 293, 295, 299, 333 European Union (EU) 1, 64, 118, 126, 131, 135, 183, 211, 299, 303, 321, 338; countries 59, 188, 204, 253, 257–8, 324, 326, 328, 330; Framework Programmes (EU-FWPs) 242–3, 256, 271; funding 95, 212, 243, 245–6, 268, 272; initiatives 270; institutions 21, 210–11, 271, 273–4, 293; member states 22, 237, 274, 296, 321–5, 329, 337; policy objectives 10, 22, 37; R&D network 269, 276, 279, 287, 294 explicit knowledge 32–3, 41, 227; see also codiﬁed knowledge export 60, 118, 120, 122, 124–5, 142, 217; data 154; market 217; market shares 256 external knowledge 39, 161–2, 168, 263; access to 197; codiﬁed 188; ﬂows 104, 112–13 ﬁnancial issues 8, 37, 45, 111, 142, 211, 326–7; research costs 329; support 299, 321, 327

Finland 19, 115, 118, 121–35, 138–9, 258–9; employment shares 120; ﬁrms 61; FWP 251; government 274; Oulu Technopolis 154 ﬁrm networks 325; dense 302, 310, 340; innovative 322; topological features of 280, 288, 294, 296 ﬁrm size 39, 56, 60–2, 85, 170, 177, 180, 183–4, 205–6, 276, 340; differences 223, 326; medium large 158, 162–3, 187, 196, 199, 201, 204–7, 233, 335, 340; research size 196; small 110, 188, 203, 340 ﬁrms 27, 35, 49, 52, 68–71, 78–80, 105, 116, 179, 186, 195–9, 204, 212, 219, 234–5, 256, 268, 299–304, 315, 323, 327–33; absorptive capacity 38–9, 58; age of 187; characteristics 55–6, 184; European 16, 20, 211–12, 233, 247, 322; incentives 55, 206, 339; information 97, 205, 223; innovative abilities 57, 76, 160–2; 170, 184, 238; legal status 199; multinational 48; niche 128; non-innovative 77, 227; productivity 217; research-intensive 58, 233; sector 175; strategies 51, 74, 207; technology proﬁle 92, 161; transition status 220 ﬁrms, location 138, 256, 336; country 173, 177; multi-location 167; random 343 ﬁrms, private sector 22, 47, 67, 95, 321–2, 327, 329; employment 140; privatized 129 food and beverages 60–1, 77, 80, 115, 134, 162–3, 170, 173, 177, 184, 188–95, 204–7, 254, 262 Foreign Direct Investment 17, 48, 61, 144 Framework Programmes (FWPs) 15, 48, 158, 211–12, 218, 226–8, 245–50, 253–7, 265, 268–75, 280–3, 294–6, 299, 303, 311, 321, 333–4, 341; budget allocations 243, 262; cooperative links 342; funding 244–6; market share growth 258; political role of 276; for RTD 340, 343 France 16–18, 49, 64, 77–93, 122–8, 138–9, 162, 174–6, 180, 237, 258–9, 326, 336; CIS II 61; ﬁrms 98, 177; programmes 323–8; research partnerships 95; soft loans to SMEs 325

348 Index funding 158, 233, 246–7, 283; European 210–11, 273; mechanisms 303; private 154, 326, 328; for public research 322; R&D 165; restriction of 274–5 geographical distribution 140–2, 263, 303, 343; clusters 117, 138, 342; location 54, 63, 116; proximity 47, 56, 61, 70, 186, 342 Germany 16–18, 59–61, 77–8, 80–93, 122–4, 126–8, 138–9, 162, 165, 174–7, 180, 184, 198, 258–9, 325–6, 335–6; ﬁrms 61, 94; Fraunhofer Institutes 321, 327; programmes 328–9; Report 330; research partnerships 95; governments 31, 144, 338; agencies 37, 69; decisions 343; institutions 158; monopoly company 129; national 299, 303; organizations 128; policies 198, 327, 339; regulations 193–4, 205; research institutes (GRIs) 22, 45, 329; support 154, 183–5, 187, 327 Greece 16, 18, 77–8, 80–93, 122–4, 126, 131, 154, 162, 174–7, 207, 256, 258–9, 322–7, 336; ﬁrms 94, 177, 256; FWP ﬁrms 251; research partnerships 95; university 246, 296, 317 gross domestic product 256; per capita growth 258, 262 heterogeneity 206, 215, 217, 223, 342 high tech sectors 116, 177–9, 187–8, 210 higher education institutions 142–3, 343 high-tech sectors 12, 121, 123, 154, 158, 254, 262, 269; market shares 258 human capital 7, 38–9, 92, 213, 245; indicators 46, 49; stocks 50 human resources 45, 245; lack of 96 ideas for innovation 34, 84, 97–8, 189, 197, 203 incentives 94, 203, 293–4, 327, 342; for cooperation 206; disincentives 303; to invest in alliances 311; lack of 106; to learn 32; national 272; for PRIs 327 independent ﬁrms 163, 165, 174–5, 177–8, 199 industrial sectors 34–5, 37, 47; innovation 184, 186; traditional 9 industrial technologies 212–13, 244, 247

industry 34–8, 97, 186, 196, 226, 228, 295, 327; characteristics 183–4; classiﬁcations 214, 228; dynamics 210; policy 342; sectors 116, 195 informal contact 89, 95, 179, 185–7, 193–4, 336 information 51–2, 62, 67–9, 77, 96, 159, 184, 299, 302, 305, 308, 317; access to 5, 72, 75; channels 75, 186; disclosure of 188, 233; exchange 194, 300, 311; external channel 187, 197; ﬂows 50, 103, 286; gathering strategies 56, 185; leak 57–8; purpose 54; sharing 254; 256; transfer 55, 76; transmitters 315 information and communication technology (ICT) 4, 10, 13, 20, 40, 115–17, 141–3, 147, 154, 159, 210–14, 243–5, 247, 254, 268, 275, 322, 328, 334, 340–1, 343; area 219, 233–4; budget share 262; businesses 138; cluster 5–7; 135; consortia 234; consumers 136–7, 150–1; deﬁnitions 120, 145; equipment 118, 123; ﬁeld 221, 223, 229; ﬁrms 19, 81–2, 121, 228; manufacturing 119, 132; market 128; niches 138; projects 155; research 236; services 80, 98, 120, 128–9, 132, 138, 146 information and communication technology employment 118, 120, 138, 140–1, 150 information and communication technology samples 215, 224, 227, 230 information processing 4, 7, 9; equipment 145; systems 247 information sources 35, 46, 51–2, 56–61, 68–9, 79, 186, 195–7, 329, 332, 339; commercial 68; external 56, 89, 96–8, 167, 170, 185, 188–9, 204, 335; most important 190–2; public 72 Information Technology sector 6, 108, 129, 137–8, 214, 236, 335; applications 233; capabilities 243; European ﬁrms 245; Microsoft 128–9; solutions 108 in-house development 70, 62, 162–3, 165–6, 171, 177, 179; of capabilities 160, 336; research 223, 329; solutions 327; spending 77 innovation 13–21, 28–37, 50–2, 55, 57–9, 63, 68–9, 72, 76–9, 105–6, 110, 135, 158, 162–3, 177, 183, 186–7, 195–9, 210, 215, 228, 324–5, 332,

Index 349 339–40; beneﬁts 228; bought-in 167; capacity 40; collaboration 54, 170–3, 175–8, 336; completion 84–6, 97, 184–6, 189–93, 196, 204, 207, 329; development 167; economically signiﬁcant 82, 96, 108, 160–2, 179, 189, 322, 335, 339; economics of 12, 118, 159, 179; experience 218, 222, 227, 234, 238; funding 168, 326–7; gap 10, 217, 219, 223, 269; ideas 79, 83–4, 189, 191–2, 335, 339; in-house 95, 104, 323; investments in 232, 269, 271; management 63, 330; methods 165–6, 179–80; national systems 27–9, 40–1, 56, 117, 126; networks 301; objectives 59–60; performance 321–2; policies 16, 343; prior to entry 222; quality 160; research 11, 46, 64, 197; sales share 55, 61, 161–3, 167–80, 337; surveys 46, 56, 58, 62, 159; theories 158 innovation systems 14, 16, 17, 22, 27–8, 41, 325, 327, 337; approach 117–18, 128; chain-link 158; national 36, 59, 61, 165–6, 170, 175, 177–8, 293, 321, 336 innovative ﬁrms 60–1, 64, 78, 80, 136, 160, 162, 167, 171, 180, 189, 200, 220, 229, 238, 324, 341; level 220, 223; networks 234; outputs 60, 162, 222, 226; performance 28, 42, 212; platform 234; potential 211, 237, 334, 342; products 55, 170; research 185; RJV entrants 215; technology 324, 326 institutions 2, 17, 34, 37, 42, 68, 76, 136, 283, 303; inﬂuence 39; interactions 28; internal and external 40, 77; public and private 269; support from 268; technical 324 intellectual property rights 15, 34, 93, 111–12; protection 18, 37–8, 94, 333–5, 337, 340, 343 inter-ﬁrm links 29, 76, 242, 251; collaboration 35, 38–9, 102, 111–13, 158, 250, 262, 326 internal expertise 58, 172, 232, 336, 339 internal knowledge 104, 113; contribution to innovation 335, 338; impact of 204; sources 57, 63, 96, 105, 108, 110 Internet 5–6, 18, 72, 79–81, 89, 92, 98, 108, 111; 115, 129, 168, 193, 336, 339; access 144; broadband 274; communication 94; connections 136;

e-commerce 94, 325; high-speed access 137; interactive websites 70; skills 325; start-up ﬁrms 155; technology 80; world wide web (www) 282 intranets 107; centralized 104 inventions 106; contributions to 50; overlap 94; types of 46 investment 38, 57, 143, 168, 303; current 222; in developments 303; in information 40; in R&D 199, 327 Ireland 64, 121–2, 131, 135, 154, 180, 256–9 Italy 16, 18, 77–93, 122–6, 162, 165, 174–7, 180, 207, 258–9, 324–7, 336; ﬁrms 94, 335; government 274; organizations 17, 312, 317, 329; research partnerships 95 Japan 122, 135, 154, 237, 243, 268, 335, 341; companies 211; patent system 54; RJVs 227 joint research 35, 48, 69, 71, 73; Centre (JRC) 243, 245; and development 108, 112, 168, 270, 272, 303; ventures 58, 186, 234, 303 journals 34, 50–2, 60, 170, 173–4, 176, 186; papers published in 34, 279; scientiﬁc or business journals 47, 72, 79, 92, 168, 175, 178, 197, 203–4, 335, 339; statistics from 46 KNOW FOR INNOVATION survey 16, 19, 63, 162, 167, 183–8, 199, 206–7, 322, 325, 329–30, 333–42; countries 327–8; KNOW project 115, 154, 265, 333; questionnaire 204; research project 14–18, 49, 74–8, 123, 236–7, 296, 317; respondents 326 knowledge 11, 32–3, 50, 223, 269, 274, 299, 300–2, 305; absorption 41, 103–6; accessibility 4, 6–9, 19, 337; acquisition 48, 52, 70–2, 193, 205; base 18, 28, 70, 109–12, 222; capital 226–9, 237–8; channels 20, 55, 73; communications networks 21, 340; distribution 12–14, 17, 36–41, 47, 63, 67–9, 101, 104, 271, 294, 300, 342 economy 3, 22, 27, 303, 333, 338; exchange 106, 188, 251; infrastructure 42; networks 15, 332; spillovers 20–1, 34, 76, 242, 250, 252, 258, 262, 265, 332; suppliers 253; workers 106, 112

350 Index knowledge ﬂows 13–18, 27, 32, 34–42, 45–57, 60–3, 67–77, 79, 101–11, 115, 118, 128, 158–9, 186, 242–3, 250–3, 256, 263, 279, 300, 304, 310–12, 317, 321, 325–9, 332–3, 342; affecting industry 338; between divisions 112; between ﬁrms 22; channels for 98, 338; embodied 64; innovation-related 21–2, 85, 88, 335–7, 343; inter-sector 19; traditional channels of 339 knowledge management 10, 12, 101, 109, 112; tools 104 knowledge production 13, 33, 110, 112, 158–9; activities 223; by universities 187 knowledge sharing 15, 27, 101–2, 109–10, 113, 251; barriers to 18, 104–5, 107–8; competition- based 106–7, 111; scope and ﬂexibility 103 knowledge sources 15, 34, 54–60, 67–9, 161, 178–9, 183–4, 188, 340; coordination of 111; intercontinental 84; internal and external 17, 19, 59, 63, 74, 88–93, 108, 112, 175, 192, 198–9, 204, 251, 323, 336–8; sourcing strategies 194 knowledge transfer 48, 67, 70, 112–13, 308, 317; problems related to 103 knowledge transmission 294, 303–6, 310, 333; channels 67–8, 70, 72 knowledge, new 29, 31, 34, 106, 110; diffusion of 37; in-house 57 knowledge, specialized 49, 98; strategic 74; supplementary 102–3 labour 143; common pool 116; lowskilled 12; market 138, 150; mobility 17 lags 228, 234, 238; structures 262, 265 leaders 295, 342; European 276; major 234 lead time 77, 94, 197; advantages (LTA) 81, 168, 171, 175–6, 178, 335 lead user concept 83, 96–7 learning 2, 13–17, 31–3, 38–9, 101, 159, 221–2; capabilities 19; by networking 217, 242–3, 254; processes 7, 11, 28, 33–4, 38–9, 76–7; race 111; skills 271; sources 333 licence 69–70, 163, 328; holders 137; purchase of 52; UMTS 137 linkages 37, 45, 170, 253, 257, 262, 280, 302, 322; direct and indirect 282,

304; extra-sector 20; between ﬁrms and PROs 158 Lisbon Extraordinary European Council 2, 10; objectives 333, 339; Summit 303 lock in 18–19, 112, 250, 295, 334, 341 low tech sectors 20, 178–9, 187–8, 254, 262; exports from 121 Luxembourg 64, 154 Maastricht Treaty 270–1 management 40, 67, 74, 103, 106, 111–13, 232; and business plans 324; changes in 31, 105; practices 39; of production 104; of research activities 272; strategies 57 manufacturing sectors 52–3, 118, 326, 335; ICT 126; R&D 330 market 28, 94–5, 116, 136, 198, 243, 250, 271; capital and labour 138; concentration 238; demand 34; development 179; divisions 109; domestic 129; domination 172; dynamics 79; failure 303; free 299; information 68; international 129; knowledge 13, 72, 109; links 193; organized 29; orientation 109; processes 332; relevant knowledge 336; share 120, 257; size 96; transactions 21, 262; trends 92, 95 marketing 29, 69, 103, 106; cooperation 263; organization 105, 234 medical and biotechnology (MB) 20, 210, 214, 217–21, 229, 231–4, 237, 334, 341; consortia 212, 228; industry composition 215; patenting differential 225; sample 223, 227 metropolis effect 142, 343 Ministry of Research and Information Technology 137, 143–4, 154 mobile communications 127–30, 135–7, 141; phones 115 mobility of researchers 8, 31, 49–50, 185, 213, 245, 272; incentives for 293; inter-country 303; programme 36 Moore graph 306–10, 314 motivation 101, 106–8, 110, 112, 193–4; for cooperation 179 NACE 77, 162; codes 142, 147–9, 151, 154; sectors 188, 191–2, 247, 249 Netherlands, the 16, 18, 64, 77–8, 80–95, 98, 115, 121–6, 131, 134–5, 138–43,

Index 351 162, 167, 170, 175–7, 180, 207, 258–9, 322, 325, 336; ﬁrms 61, 321, 328; IOP 329; Maastricht University 48; MERIT (Maastricht Economics Research Institute on Innovation and Technology) 14, 48, 64, 317; procurement programmes 326 networks 7, 15, 17, 20, 31, 67, 102, 105, 178–9; 262, 273–7, 293, 300–4, 307, 311–17, 321, 338, 342; afﬁliation 227–9, 278–9; approach 269; centres of excellence 21, 234, 245, 268–9, 272, 294–6, 334, 340; cluster 305; conﬁguration 295; connections 283, 286–7, 292, 309, 341; contractors 326; density 280, 305, 333; diameter 288; distributors 129; edges 310; links 251, 282; members 308; organizations 158; policies 242, 268–9; Report 137, 144; research 299; self-organized 295, 300, 315, 334; typologies 305–6; visualization 290–1 nodes 67, 281–3, 286–8, 291, 294, 297, 300, 302, 311–12; crucial 269; distant 341; and edges 290; primary 294; random connection of 287; research 275; ring of 291 Nordic countries 49, 136–7, 332; ﬁrms, world leaders 136; ICT industries 130; Statistical Institutes 117, 120, 130 Nordic Mobile Telephony (NMT) 127, 136; boom 141; service providers 136 Norway 64, 118, 130, 132, 138–9, 180 not-invented-here syndrome 102–4, 107–9, 323 ofﬁce equipment sector 35, 60, 124–5, 148, 161; computers 254, 258, 262 oligarchic core 268–9, 293–5, 333 openness 111, 187–8, 196–8, 204–7; to collaborators 105; to external sources 20, 183–4, 336–8 organizations 1, 13, 31–3, 40, 67–8, 71–4, 179, 247, 277, 279–80, 288, 300, 304; connected 273, 283, 286, 290–2, 295, 312; European 335; structures 96, 101–13; unconnected 312 Oslo Manual 29–30, 41, 45, 68, 159 ownership status 74, 170–1 PACE survey 54, 59, 183, 187, 206–7 participation 161, 220–3, 228, 232–3,

256–8, 276, 294, 340; in consortia 226; in cooperative agreements 69, 195; level of 200, 203; marginal 341; patterns of 15, 332; per project 251, 312–13; rates 334; repeated 295; research 49, 70 partners 79, 198, 262, 276–9, 290, 294, 297, 302–3, 310, 333, 341; academic 193–4; complementary 251; foreign 20, 107, 256; multiple 200, 321; prospective 95–6 patent 222–3, 232–4; analyses 185; databases 56–7, 72, 80, 92, 161, 168, 197–8, 214, 335, 340; disclosures 39, 51, 63; grants 46, 64; holders 341; indicators 17, 46; national system 94; ofﬁces 38, 69; protecting innovations 79, 81, 83, 198, 205, 337; proximity 251; stock 217, 223, 226–8, 237–8; see also patents patent applications 46, 64, 217–18, 227, 232–4, 334; European 16, 49, 312; successful 222 patent citations 16, 21, 49–50, 70, 333, 335; network 311–12, 314, 317, 341; pairs 49 patenting 15, 20, 61, 81, 188, 197–8, 203–7, 210–11, 214–15, 221, 226–9, 232–4, 328, 334–5, 341; current activity 220, 238; differences 223; differential 222; level 219; performance 223; prior to entry 228, 334; strategies 55 patents 15, 35, 46, 50, 59–61, 70, 77, 81–2, 89, 92, 171, 185–8, 198, 204, 263, 333–5, 340; cost of 94; declining trend of 227, 237–8; licensing or sale of 48; output 233–4; owner 93; protection 52; reliance on 175; system 54 path length 304; average 305, 317; characteristic 305–11, 314–16 performance 34, 68–9, 138; of competitors 339; countries 262; economic 27; enhanced 30 peripheral actors 273, 294–5, 341–2 personal computers (PCs) 115, 128–9, 136 personnel 70–1, 83; exchange of 89, 336; mobility 36, 73; skilled 79; transfer of 70 pharmaceutical sector 52, 57, 60, 63, 116, 121, 123, 134, 155, 162, 179, 237, 326, 330, 340

352 Index policies 69, 233, 242, 343; cooperative 234, 269, 342; design 235, 274; implications 337; intervention 272; makers 27, 35, 67, 84, 183, 334; national level 37, 115, 158, 303; priorities 243, 254, 257–8; private and public 42, 262, 321–2; targets and achievements 210 Portugal 180, 256, 259 power law 300; distribution 17, 296; tail 282–3 pre-competitive research 234, 244, 250, 303; collaboration 242, 247, 254, 262, 271; EU- FWPs 258; higher risks for 329 Prime Contractors 251, 276–9, 290–1, 294–5, 297, 333, 341 procurement 29, 69; public policies 38, 343 products and processes 4, 30, 48, 74, 103, 106, 108, 112, 160, 168; development 110–11, 232; innovations 77, 135, 163, 165–6, 170–1, 174, 177, 197–9, 200, 203, 336; organization 30, 42, 194; sales chain 186, 190, 197, 296 project 314; classiﬁed 247; coordination of 104; coordinators 251, 253; costs of 296; evaluation process 109; leaders 103–4; members 104; participations 253; per organization 275; teams 109; years 238 public research 197–8, 321–2; agencies 295; budget 246; centres 183–4, 188; institutions (PRIs) 19, 35–6, 38, 47, 69, 303–4, 321–5, 327–30; organizations (PROs) 20–2, 52–64, 79, 83, 94–7, 158, 163, 170–3, 183–8, 190–6, 204, 333–4, 337–42; output 63, 207; programmes 194 public research publications 56–7, 185–8, 202, 207, 237; databases 197, 205; economic and business 28, 39; trade and scientiﬁc 38, 161 regulatory authorities 69, 136, 271 relationships 13, 196, 272; new 256; with other organizations 339; with PROs 197, 199, 205 research 49, 275; costs 329; in Europe 212, 271, 295; institutes 34, 36, 46–8, 68, 159, 212, 247, 254, 324; national systems 269, 271; organizations 38,

275, 321; personnel 36; private establishments 38, 197; strategic areas 272, 294; virtual institutes 328 research and development (R&D) 9, 29, 47, 60, 79, 101–5, 107–9, 111–12, 143, 165, 210, 214, 299, 323, 327, 336; activities 33–4, 80, 198–9, 203; budgets 233; collaboration 20, 61, 94, 167, 205, 339–40; consortia 227, 239, 247, 270, 277, 279–80, 288, 291; contracted out 163, 168; cooperation 161, 187, 193, 195, 204–5, 217, 232, 242; departments 31–2, 39, 74; efforts 48, 69, 138, 251; employee share 170, 174, 176; employment 56, 184, 196, 199, 201, 203, 205, 303; European consortia 211, 215, 238, 282; expenditure 35, 45, 56–9, 164, 167, 171, 178, 187, 252–3; ICT ﬁrms 18, 19; incentives 203; industrial 186, 247; in-house 164–5; intensity 55, 171, 177, 203, 205, 323, 335–7; intensive ﬁrms 58, 61, 116, 134, 144, 175–6, 199, 228, 334, 341–2; interﬁrm cooperation 258; market oriented 212; medical and health 57, 236; network 21, 263, 268, 282, 326; participation 188, 197–8; partnerships 159; private 251, 326; productivity 207; programmes 38, 69; projects 62, 184, 205, 251; PROs 167, 175–8, 188, 195–9, 201, 203, 205, 207; public policy 263; size of ﬁrm 203; stocks 262, 265 research and development funding 233, 321; European 211, 276, 296; government 142, 188, 204, 343; private 198; public 59, 197–8, 326, 328, 337–8 research and technological development (RTD) 48, 64, 269–71, 299, 333, 340; Framework Programmes (FWPs) 213; investments 271–2; networks 300, 303, 312–14; policies 233, 245, 304, 311, 318, 338; spending, public and private 244; sponsored programmes 282 research joint ventures (RJVs) 15–16, 48–9, 71, 76, 161, 210, 223, 246, 265; afﬁliation 211, 226–9, 232–4; early members 217–19, 228, 233; in ICT 233; independent 239; members 275, 279; networks 270, 279–91, 297, 311;

Index 353 participation 20, 212–15, 219–22, 227–9, 232–4, 238, 275–6; projects 271, 275–9, 294–6 resources 45, 115, 196, 234, 272–3, 294, 301–2; for collaboration 170 reverse engineering 57–8, 61, 69, 79–80, 83, 92–3, 168, 335, 339 science and technology 9, 14, 17, 77, 96, 204, 210, 299; collaboration 48, 71, 187, 212, 325; competencies 111–12, 198, 323; development 13, 144, 342; expertise 88, 97, 160, 193–4, 336; gap 245; information 34, 80, 89; innovation policies 337–8; networks 272–4, 321; policy 243, 268–9, 294, 329; science park 143 Science Citation Index (SCI) 47 screening 188, 197, 204–5; of partners 198 secrecy 77, 81–2, 168, 197, 204, 251, 335, 340 sector 170; differences 191, 204–5; dynamics 235–6, 341; effects 57, 63, 178, 200, 202, 205 self-selection 217, 220–3, 228–9, 233; analysis 232, 234; bias 219; issue 238 service sector 8, 31, 52, 118, 146, 179, 326, 340 share 257–9; of contractions 315; of a country 262; decrease 258; of ICT employment 140; of innovations 337 shared cost projects 243, 264, 276; contractual research 271; R&D consortia 233, 268 sharing 188; information 13, 103; risks 301 shift-share 258; analysis 243, 256–7, 260–2 signalling 11, 188, 204–5; competencies 337; effect 203; strategic 197–8 Single European Act 211, 243, 270 size 223, 228–9, 232–4; classes 162, 188; effect 201–3; ﬁrms 187, 203–5, 334–7; network 310 skilled personnel 70, 74, 82, 130, 138, 343 small and medium-sized enterprises (SMEs) 19– 20, 56–9, 128, 144, 167, 179, 184, 194, 233–4, 245, 321–7, 330, 335–9, 344; capacity of 63 ICT ﬁrms 126, 325; innovative ability 234; participation of 212; SME-PRI linkages 329

small worlds 21, 287, 296, 300–1, 304–5, 308–11, 314–17; characteristics 294, 333 social capital 301–2, 304–8, 310, 315, 317 software 128–9, 145, 160; development 137–8 Spain 138–9, 154, 256, 258–9 specialization 140–2, 154; international 120; in IT 138; pattern 126 spillover 50, 54, 227, 251–2, 256, 262–5, 294, 342; cut-off criteria for 253; disembodied knowledge 250, 265; inter- and intra-sector 242–3, 254–8; R&D 69, 186; types of 64 staff 52, 103; exchange 273; graduates 187; hiring subsidy 324 start-ups 187; costs 109; support for 327 structural holes 301–2, 304, 307–11, 317; theory of 304 subsidies 174–6, 198, 203, 205, 324–5; for innovation 167, 171–3, 178; for R&D 217, 242, 254; reliance on 204 suppliers 57–8, 61, 68, 79, 83–4, 96, 136–7, 186, 189–93, 197, 207, 253–4, 271, 329, 335, 339, 343; communication with 94, 105, 336 support 97, 247, 268–72, 322–3 Sweden 19, 115, 118, 120–32, 134–5, 137–9, 144, 180, 258–9; Science Park 154 Switzerland 49, 118 System of Innovation (SI) 17, 27–8, 34, 117; approach 36, 39 tacit knowledge 7, 8, 11–14, 32–3, 36, 40–1, 50–6, 70, 102, 158–60, 227, 263; complementary 21, 262; component 256; sources of 197 Targeted Socio-Economic Research (TSER) Programme 14, 64, 115, 237, 296 technological knowledge 3–7, 74, 116, 179, 193–4, 342; innovation 11, 29, 199, 203, 271; new 30, 49; transfer 13, 22, 69, 71, 185, 273 technology 7, 9, 35–7, 58, 95, 105, 116, 121, 154, 160, 325, 335, 340; absorption 150; audits 330; balance of payments 47–8; commercial applications 322; development 30, 71, 137, 303; differences 229; gap 270; opportunities 163, 188, 332; policies 210, 273, 303; projects 248–9,

354 Index technology – contd. 275; range of 162, 167; strategic alliances 74, 299, 301–2; suppliers 177; transfer 34, 61, 158, 323–4, 330; transfer ofﬁces (TTO) 185, 328 telecommunications sector 6, 35, 77, 80, 104–5, 119, 126–9, 132–4, 142–7, 154, 161–3, 184, 188–9, 195, 253; cluster 141; equipment 123–5, 160, 165–7, 174–9; funding 137–8; hardware 121; operators 234; products 110 Towards a European Research Area 269, 291 trade fairs and conferences 56–8, 70, 79, 88, 92, 168, 185–6, 193–4, 197, 237, 335, 339

189, 207, 258–9, 322–6, 335; ﬁrms 197, 200; free market 330; research councils 328–30; research partnerships 95 United States 3–4, 9–10, 40, 49, 88, 117, 122–9, 135, 138–9, 154, 186, 190, 242–3, 335, 341–2; companies 197, 211, 268; market 136; Patent Ofﬁce 135; patent system 55 university 19, 27, 31, 34–8, 45–9, 54–8, 77, 79, 94–7, 183–7, 190–2, 247, 268, 276, 295, 303–4, 324–5, 328–9, 332–3, 343; education; 50, 105, 130, 133, 143; industry links 20, 118, 194–6, 205–6, 263; locations 138; research 128, 194 user-producer relationships 35, 68, 116

United Kingdom 16, 18, 56, 77–93, 122–6, 131, 138–9, 154, 162, 174–6,

wireless communications 135, 137, 141, 143

eBooks – at www.eBookstore.tandf.co.uk

A library at your fingertips!

eBooks are electronic versions of printed books. You can store them on your PC/laptop or browse them online. They have advantages for anyone needing rapid access to a wide variety of published, copyright information. eBooks can help your research by enabling you to bookmark chapters, annotate text and use instant searches to find specific words or phrases. Several eBook files would fit on even a small laptop or PDA. NEW: Save money by eSubscribing: cheap, online access to any eBook for as long as you need it.

Annual subscription packages We now offer special low-cost bulk subscriptions to packages of eBooks in certain subject areas. These are available to libraries or to individuals. For more information please contact [email protected] We’re continually developing the eBook concept, so keep up to date by visiting the website.

www.eBookstore.tandf.co.uk

Organization and Identity (Routledge Studies in Business Organizations and Networks)

Knowledge and Innovation in Business and Industry: The Importance of Using Others (Routledge Studies in Innovation, Organizations and Technology)

Flows in transportation networks

Information Warfare in Business - Strategies of Control and Resistance in the Network Society (Routledge Studies in Business Organizations and Networks)

Decisions, Risk and Reward (Routledge Studies in Business Organizations and Networks)

Managing Professional Identities: Knowledge, performativity and the 'new' professional (Routledge Studies in Business Organization and Networks)

Technological Communities and Networks: Triggers and drivers for innovation (Routledge Studies in Business Organization and Networks)

AIDS and Business (Routledge Advances in Management and Business Studies)

Authority and Control in Modern Industry: Theoretical and Empirical Perspectives (Routledge Studies in Business Organization and Networks, 10.)

Innovation, Knowledge and Power in Organizations (Routledge Studies in Global Competition)

Interfirm Networks: Organization and Industrial Competitiveness (Routledge Studies in Business Organization and Networks, 11)

Knowledge Management in Modern Organizations

Shopfloor Matters: Labor-Management Relations in 20th Century American Manufacturing (Routledge Studies in Business Organizations and Networks, 5)

Knowledge Networks for Business Growth

Privatization and Financial Collapse in the Nuclear Industry: The Origins and Causes of the British Energy Crisis of 2002 (Routledge Studies in Business Organizations and Networks)

Knowledge Flows in European Industry (Routledge Studies in Business Organizations and Networks)

Organization and Identity (Routledge Studies in Business Organizations and Networks)

Democracy and Efficiency in the Economic Enterprise (Routledge Studies in Business Organizations and Networks)

Towards better Performing Transport Networks (Routledge Studies in Business Organizations and Networks)

Knowledge and Innovation in Business and Industry: The Importance of Using Others (Routledge Studies in Innovation, Organizations and Technology)

Flows in transportation networks

Information Warfare in Business - Strategies of Control and Resistance in the Network Society (Routledge Studies in Business Organizations and Networks)

Decisions, Risk and Reward (Routledge Studies in Business Organizations and Networks)

Privatization and Supply Chain Management (Routledge Studies in Business Organizations and Networks, 12)

The Corporation: Growth, Diversification and Mergers (Routledge Studies in Business Organizations and Networks)

Formula Funding of Public Services (Routledge Studies in Business Organizations and Networks)

Managing Professional Identities: Knowledge, performativity and the 'new' professional (Routledge Studies in Business Organization and Networks)

Technological Communities and Networks: Triggers and drivers for innovation (Routledge Studies in Business Organization and Networks)

AIDS and Business (Routledge Advances in Management and Business Studies)

Authority and Control in Modern Industry: Theoretical and Empirical Perspectives (Routledge Studies in Business Organization and Networks, 10.)

Innovation, Knowledge and Power in Organizations (Routledge Studies in Global Competition)

Interfirm Networks: Organization and Industrial Competitiveness (Routledge Studies in Business Organization and Networks, 11)

Knowledge Management in Modern Organizations

Shopfloor Matters: Labor-Management Relations in 20th Century American Manufacturing (Routledge Studies in Business Organizations and Networks, 5)

Knowledge Networks for Business Growth

Knowledge Networks for Business Growth

Privatization and Financial Collapse in the Nuclear Industry: The Origins and Causes of the British Energy Crisis of 2002 (Routledge Studies in Business Organizations and Networks)

Complexity, Organizations and Change (Routledge Studies on Complexity in Management)

Building Chaos An International Comparison of Deregulation in the Construction Industry (Routledge Studies in Business Organization and Networks, 22)

Business Clusters: An International Perspective (Routledge Studies in Business Organization and Networks)

The Organisation of the Firm: International Business Perspectives (Routledge Studies in Business Organisation and Networks, 6)

Measuring Business Excellence (Routledge Studies in Business Organization and Networks, 21)

Social Networks and Organizations

The New Mutualism in Public Policy (Routledge Studies in Business Organization and Networks, 15,)

Working in the Service Sector (Routledge Studies in Business Organization and Networks)

Inclusion and Exclusion in European Societies (Routledge European Sociological Association Studies in European Societies, 5)

Knowledge Flows in European Industry (Routledge Studies in Business Organizations and Networks)

Organization and Identity (Routledge Studies in Business Organizations and Networks)

Democracy and Efficiency in the Economic Enterprise (Routledge Studies in Business Organizations and Networks)

Towards better Performing Transport Networks (Routledge Studies in Business Organizations and Networks)

Knowledge and Innovation in Business and Industry: The Importance of Using Others (Routledge Studies in Innovation, Organizations and Technology)

Flows in transportation networks

Information Warfare in Business - Strategies of Control and Resistance in the Network Society (Routledge Studies in Business Organizations and Networks)

Decisions, Risk and Reward (Routledge Studies in Business Organizations and Networks)

Privatization and Supply Chain Management (Routledge Studies in Business Organizations and Networks, 12)

The Corporation: Growth, Diversification and Mergers (Routledge Studies in Business Organizations and Networks)

Formula Funding of Public Services (Routledge Studies in Business Organizations and Networks)

Managing Professional Identities: Knowledge, performativity and the 'new' professional (Routledge Studies in Business Organization and Networks)

Technological Communities and Networks: Triggers and drivers for innovation (Routledge Studies in Business Organization and Networks)

AIDS and Business (Routledge Advances in Management and Business Studies)

Authority and Control in Modern Industry: Theoretical and Empirical Perspectives (Routledge Studies in Business Organization and Networks, 10.)

Innovation, Knowledge and Power in Organizations (Routledge Studies in Global Competition)

Interfirm Networks: Organization and Industrial Competitiveness (Routledge Studies in Business Organization and Networks, 11)

Knowledge Management in Modern Organizations

Shopfloor Matters: Labor-Management Relations in 20th Century American Manufacturing (Routledge Studies in Business Organizations and Networks, 5)

Knowledge Networks for Business Growth

Knowledge Networks for Business Growth

Privatization and Financial Collapse in the Nuclear Industry: The Origins and Causes of the British Energy Crisis of 2002 (Routledge Studies in Business Organizations and Networks)

Complexity, Organizations and Change (Routledge Studies on Complexity in Management)

Building Chaos An International Comparison of Deregulation in the Construction Industry (Routledge Studies in Business Organization and Networks, 22)

Business Clusters: An International Perspective (Routledge Studies in Business Organization and Networks)

The Organisation of the Firm: International Business Perspectives (Routledge Studies in Business Organisation and Networks, 6)

Measuring Business Excellence (Routledge Studies in Business Organization and Networks, 21)

Social Networks and Organizations

The New Mutualism in Public Policy (Routledge Studies in Business Organization and Networks, 15,)

Working in the Service Sector (Routledge Studies in Business Organization and Networks)

Inclusion and Exclusion in European Societies (Routledge European Sociological Association Studies in European Societies, 5)

Recommend Documents