Advances in Spatial Science Editorial Board Manfred M. Fischer Geoffrey J.D. Hewings Peter Nijkamp Folke Snickars (Coordinating Editor)
Titles in the Series M. M. Fischer, J. Revilla Diez and F. Snickars Metropolitan Innovation Systems VIII, 270 pages. 2001. ISBN 978-3-540-41967-9
L. Anselin, R.J.G.M. Florax and S. J. Rey Advances in Spatial Econometrics XXII, 513 pages. 2004. ISBN 978-3-540-43729-1
L. Lundqvist and L.-G. Mattsson (Eds.) National Transport Models VIII, 202 pages. 2002. ISBN 978-3-540-42426-0
R.J.G.M. Florax and D. A. Plane (Eds.) Fifty Years of Regional Science VIII, 400 pages. 2004. ISBN 978-3-540-22361-0
J. R. Cuadrado-Roura and M. Parellada (Eds.) Regional Convergence in the European Union VIII, 368 pages. 2002. ISBN 978-3-540-43242-5
D. Felsenstein and B.A. Portnov (Eds.) Regional Disparities in Small Countries VI, 333 pages. 2005. ISBN 978-3-540-24303-8
G. J. D. Hewings, M. Sonis and D. Boyce (Eds.) Trade, Networks and Hierarchies XI, 467 pages. 2002. ISBN 978-3-540-43087-2
A. Reggiani and L.A. Schintler (Eds.) Methods and Models in Transport and Telecommunications XIII, 364 pages. 2005. ISBN 978-3-540-25859-9
G. Atalik and M. M. Fischer (Eds.) Regional Development Reconsidered X, 220 pages. 2002. ISBN 978-3-540-43610-2 Z. J. Acs, H. L. F. de Groot and P. Nijkamp (Eds.) The Emergence of the Knowledge Economy VII, 388 pages. 2002. ISBN 978-3-540-43722-2 R. J. Stimson, R. R. Stough and B. H. Roberts Regional Economic Development X, 397 pages. 2002. ISBN 978-3-540-43731-4 S. Geertman and J. Stillwell (Eds.) Planning Support Systems in Practice XII, 578 pages. 2003. ISBN 978-3-540-43719-2 B. Fingleton (Ed.) European Regional Growth VIII, 435 pages. 2003. ISBN 978-3-540-00366-3 T. Puu Mathematical Location and Land Use Theory, 2nd Edition X, 362 pages. 2003. ISBN 978-3-540-00931-3 J. Bröcker, D. Dohse and R. Soltwedel (Eds.) Innovation Clusters and Interregional Competition VIII, 409 pages. 2003. ISBN 978-3-540-00999-3 D. A. Griffith Spatial Autocorrelation and Spatial Filtering XIV, 247 pages. 2003. ISBN 978-3-540-00932-0 J. R. Roy Spatial Interaction Modelling X, 239 pages. 2004. ISBN 978-3-540-20528-9 M. Beuthe, V. Himanen, A. Reggiani and L. Zamparini (Eds.) Transport Developments and Innovations in an Evolving World XIV, 346 pages. 2004. ISBN 978-3-540-00961-0 Y. Okuyama and S. E. Chang (Eds.) Modeling Spatial and Economic Impacts of Disasters X, 323 pages. 2004. ISBN 978-3-540-21449-6
H.W. Richardson and C.-H.C. Bae (Eds.) Globalization and Urban Development X, 321 pages. 2005. ISBN 978-3-540-22362-7 G. Arbia Spatial Econometrics XVII, 207 pages. 2006. ISBN 978-3-540-32304-4 B. Johansson, C. Karlsson, R. Stough (Eds.) The Emerging Digital Economy X, 352 pages. 2006. ISBN 978-3-540-34487-2 H. Westlund Social Capital in the Knowledge Economy X, 212 pages. 2006. ISBN 978-3-540-35364-5 A.E. Anderssson, L. Pettersson, U. Strömquist (Eds.) European Metropolitan Housing Markets VI, 363 pages. 2007. ISBN 978-3-540-69891-3 A.T. Murray, T.H. Grubesic (Eds.) Critical Infrastructure VIII, 311 pages. 2007. ISBN 978-3-540-68055-0 R. Cooper, K. Donaghy, G. Hewings (Eds.) Globalization and Regional Economic Modeling XIII, 475 pages, 2007. ISBN 978-3-540-72443-8 R. Capello, R. Camagni, B. Chizzolini, U. Fratesi Modelling Regional Scenarios for the Enlarged Europe: European Competiveness and Global Strategies XX, 337 pages, 2008. ISBN 978-3-540-74736-9 C. Jensen-Butler, B. Sloth, M.M. Larsen, B. Madsen and O.A. Nielsen Road Pricing, the Economy and the Environment XVI, 425 pages, 2008. ISBN 978-3-540-77149-4
Ugo Fratesi
•
Lanfranco Senn
Editors
Growth and Innovation of Competitive Regions The Role of Internal and External Connections
123
Dr. Ugo Fratesi Politecnico di Milano Piazza Leonardo da Vinci, 32 20133 Milano Italy
[email protected]
Prof. Lanfranco Senn Università Bocconi Via Roentgen, 1 20136 Milano Italy
[email protected]
Advances in Spatial Science ISSN 1430-9602 ISBN 978-3-540-70923-7 e-ISBN 978-3-540-70924-4 DOI: 10.1007/978-3-540-70924-4 Library of Congress Control Number: 2008931509 c 2009 Springer-Verlag Berlin Heidelberg This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Cover design: SPi Publishing Services Printed on acid-free paper 9 8 7 6 5 4 3 2 1 springer.com
Contents
Part I Foundations of Growth in Interconnected Territories Regional Growth, Connections and Economic Modelling: An Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ugo Fratesi and Lanfranco Senn
3
Sustainable Interrelated Growth: A Phenomenal Approach . . . . . . . . . . . . 29 Alberto Bramanti and Massimiliano R. Riggi A Model of Local Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Giuseppe Folloni The Dynamics of an ‘Innovation Driven’ Territorial System . . . . . . . . . . . . 59 Alberto Bramanti and Ugo Fratesi Part II Innovation and Entrepreneurship in Regions Faced with External Competition The Co-Evolution of Entrepreneurship and Clusters . . . . . . . . . . . . . . . . . . 95 Christian Garavaglia and Stefano Breschi Learning, Innovation and Growth Within Interconected Clusters: An Agent-Based Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 Mario A. Maggioni and Stefano N. Roncari Knowledge-Based Economy and Knowledge Creation: The Role of Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Roberto Camagni and Roberta Capello Systems of Innovation and Regional Growth in the EU: Endogenous vs. External Innovative Activities and Socio-Economic Conditions . . . . . . 167 Riccardo Crescenzi and Andr´es Rodr´ıguez-Pose
v
vi
Contents
Some Conjectures on the Tie Between Digital Divide and Regional Disparities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 Marco Alderighi Part III The Factors of Development in Advanced Regions Interconnection-Infrastructure as a Prerequisite for the Development of Territories – The Role of Network Externalities . . . . . . . . . . . . . . . . . . . . 217 Anna Creti Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 Massimiliano R. Riggi and Mario A. Maggioni Intra-National Disparities, Regional Interactions and the Growth of Countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 Marco Alderighi and Marco Percoco Creativity, Cultural Investment and Local Development: A New Theoretical Framework for Endogenous Growth . . . . . . . . . . . . . . . 281 Pier Luigi Sacco and Giovanna Segre Part IV Methods and Theories for the Study of Interconnected Territories Modelling Individual Behaviour of Firms in the Study of Spatial Concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 Giuseppe Arbia, Massimiliano Copetti, and Peter Diggle The New Approach to Regional Economics Dynamics: Path Dependence and Spatial Self-Reinforcing Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 Domenico Marino and Raffaele Trapasso Part V Conclusions What Policy for Interconnected Territories? Conclusions and Openings . . 351 Lanfranco Senn and Ugo Fratesi Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
Contributors
Marco Alderighi is assistant professor of economics at Universit`a della Valle d’Aosta, Aosta. e-mail:
[email protected] Giuseppe Arbia is full professor of economic statistics and econometrics at the Department of Business, Statistical, Technological and Environmental Sciences, University G. d’Annunzio, Chieti and Chairman of the Spatial Econometric Association. e-mail: arbia.unich.it Alberto Bramanti is associate professor of applied economics at Bocconi University, Milan, and coordinator of the reasearch area ‘Regional Economics’ within CERTeT (Centre for Regional Economics, Transport and Tourism). e-mail:
[email protected] Stefano Breschi is associate professor of industrial economics at Bocconi University, Milan, and research fellow at KITES-CESPRI, Bocconi University, Milan. e-mail:
[email protected] Roberto Camagni is full professor of urban economics at Politecnico di Milano and past president of the European Regional Science Association. e-mail:
[email protected] Roberta Capello is full professor of regional economics at Politecnico di Milano and incoming president of the Regional Science Association International. She is editor in chief of the “Italian Journal of Regional Science“ and co-editor of “Letters in Spatial and Resource Sciences”. e-mail:
[email protected] Massimiliano Copetti holds a post-doctoral position at University of Foggia and is also a biostatistical consultant at CSS Hospital in San Giovanni Rotondo. e-mail:
[email protected] Riccardo Crescenzi is lecturer in economic geography at the Department of Geography and Environment of the London School of Economics. e-mail:
[email protected]
vii
viii
Contributors
Anna Cret`ı is assistant professor of industrial organization at Bocconi University, Milan, and research director at IEFE (Center of Research on Energy and Environmental Economics and Policy). e-mail:
[email protected]. Peter Diggle is professor of statistics and associate dean for research at Lancaster University School of Health and Medicine, adjunct professor of biostatistics at Johns Hopkins University School of Public Health and co-editor of “Biostatistics”. e-mail:
[email protected] Giuseppe Folloni is full professor of applied economics at the University of Trento. e-mail:
[email protected] Ugo Fratesi is assistant professor of regional economics at Politecnico di Milano and co-editor of the “Italian Journal of Regional Science”. e-mail:
[email protected] Christian Garavaglia is assistant professor of economics at University of Milan-Bicocca, Faculty of Statistics, and Research Fellow at KITES-CESPRI, Bocconi University, Milan, and LIUC-Carlo Cattaneo University, Castellanza. e-mail:
[email protected] Mario A. Maggioni is full professor of economics at the Catholic University of Milan, director of the Research Centre on Cognitive Science and Communication. He is associate editor of “Networks and Spatial Economics” and “Economia Politica-Journal of Analytical and Institutional Economics”. e-mail:
[email protected] Domenico Marino is associate professor of economics at Mediterranean University of Reggio Cablabria and member of the scientific commitee of the “Fuzzy Economic Review”. e-mail:
[email protected] Marco Percoco is research fellow at the at the Department of Institutional Analysis and Public Management, Bocconi University, Milan, and fellow of CERTeT (Centre for Regional Economics, Transport and Tourism). e-mail:
[email protected] Massimiliano R. Riggi is senior researcher at the Economic Research Unit of UniCredit Retail Division. e-mail:
[email protected] Andr´es Rodr´ıguez-Pose is professor of economic geography and head of the Department of Geography and Environment at the London School of Economics. He is also programme director at the Spatial Economics Research Centre at the same institution and managing editor of “Environment and Planning C: Government and Policy”. e-mail:
[email protected] Stefano Roncari is CEO of EGGSIST LTD (IT consulting firm based in Hong Kong and Bejing, PRC) and member of the scientific committee of the Research Centre on Cognitive Science and Communication. e-mail:
[email protected]
Contributors
ix
Pier Luigi Sacco is professor of economic policy at IUAV University, Venice. He is member of the editorial board of “Quality & Quantity”, “Mind and Society” and “Economia della cultura”. e-mail:
[email protected] Anna Segre is assistant professor of public finance at the Faculty of Economics of Turin University, where she teaches Cultural Economics. e-mail:
[email protected] Lanfranco Senn is full professor of regional economics at Bocconi University, Milan; director of CERTeT (Centre for Regional Economics, Transport and Tourism), and past president of the Italian section of the Regional Science Association International. e-mail:
[email protected] Raffaele Trapasso is a regional economist at the Urban Development Programme of the OECD in Paris. e-mail:
[email protected]
Regional Growth, Connections and Economic Modelling: An Introduction Ugo Fratesi and Lanfranco Senn
1 Framing the Problem With increasing globalization over recent decades, the impacts of economic stimuli at the national level have diminished in terms of their importance for economic processes; the stimuli are increasingly originating at the international level. As markets integrate, the competitors of firms are generally firms from other countries and the domestic market is no longer and not necessarily the most important one. At the same time, the internal resources of firms are no longer sufficient for their competitiveness in a globalized World, and, to sustain their growth, they have to rely increasingly on external resources, knowledge in particular, which are normally accessible at the local and regional level (Audretsch 1988). The regional scale, therefore, has increased in importance for economic growth as a result of globalization forces; competition is now centered on region-region interactions with the regions often located in different countries. This book is concerned with the study of regional economic growth in advanced countries. The focus is essentially on the dynamics of regional performance and on the mechanisms that allow some regions to grow more rapidly than others, to become more competitive and to remain so in the long run. This book therefore draws on regional science literature that is concerned with the growth of regions with the primary purpose of clarifying which mechanisms are at work and the secondary purpose of clarifying which development policies ought to be applied. The distribution of economic activities, involving the detection and explanation of location and agglomeration, its efficiency and its evolution, is the other traditional body of economic geography literature from which the analysis in this book also draws heavily (Isard 1956; Gabszewicz et al. 1986; McCann 2002). These two strands of literature cannot be considered as separate; on the contrary, they are clearly complementary since no agglomeration takes place without growth differentials; nor can the dynamics and development patterns of regions be studied or influenced without knowing what drives location decisions and what U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
3
4
U. Fratesi, L. Senn
agglomeration/dispersion forces are at work. However, the study of regional growth remains essentially a study of the dynamics within and between regions, where elements of change are more important than comparative equilibria. The study of the determinants of regional growth is not only interesting from a speculative point of view because it sheds further light on some under-investigated aspects of the phenomena; it is much more interesting when it provides a theoretical basis for more effective development policies. In this respect, it is clear that the situation and characteristics of regions belonging to advanced countries are different from those of regions belonging to emerging or under-developed countries. Since only a small set of the theories developed for the study of the former are applicable to the latter, this book limits itself explicitly to advanced countries and the word development will hereafter only refer to lagging regions of these countries. Having been conceived with a strong interest in theories with policy applications, this book starts from concepts deriving from the observation of real facts, and analyses the different ways in which it is possible to enhance regional growth in a competitive world. It is therefore concerned with: (a) The growth of regions (both advanced and lagging) in the developed world, in particular the causal variables that could be an endogenous trigger to growth. In this sense, particular attention is paid to innovation because, in developed countries, competition is no longer achievable through cost reduction alone, but through a continuous process of upgrading production in both manufacturing and services. Innovation itself is seen as a process based on knowledge and learning; (b) Regional competition, because in an integrated world the processes taking place locally are not independent of external ones. One of the main pre-conditions for regional competitiveness is regional attractiveness, because firms look for the best production conditions. Competition is a dynamic process and implies continuous change and adjustment; this is why innovation (of product, process and organization) is so important; (c) The agents of innovation and competition. They are not anonymous and may consist of individual agents (such as firms, entrepreneurs, skilled workers and public institutions) but more and more frequently these agents integrate in networked and interacting systems. The individual agents often have some sense of belonging to their localities but, even more importantly, the interaction among agents takes place more easily when they agglomerate. This keeps the local scale fundamental in the consideration of all economic processes notwithstanding the presence of globalization forces that have generated larger and larger numbers of footloose firms. The approach of this book is neither microeconomic, investigating individual firms, nor macroeconomic, investigating the system as a homogeneous aggregate; the most fruitful approach is a meso-economic one, investigating the interactions between both micro and macroeconomic forces. To illustrate the approach, consider the representation of space presented in Fig. 1, where the three main dimensions of the economy at a macro level, the
Regional Growth, Connections and Economic Modelling: An Introduction
Time
Sectors
Sectors
Time
a
Space
b
Space
Time
Sectors
Sectors
Time
c
5
Space
d
Space
Fig. 1 Different possible scales for growth and innovation analyses
spatial dimension, the sectoral dimension and the time dimension are represented. In Fig. 1a, regional development analyses are revealed, where the economy is sliced vertically into regions and their dynamics are investigated. The study of the evolution of industries, typical of evolutionary industrial economics, is represented in Fig. 1b, where the economy is divided horizontally into sectoral slices. This approach has progressed considerably in recent years (see Malerba 2006, for a recent survey). Modifications of industries have important spatial implications, which however are not normally at the core of these analyses even though spatial patterns of innovation differ greatly from sector to sector (Breschi 2000). Our approach operates in the manner of Fig. 1a and we will focus on regions, extending the analysis to industries only where this is regionally and structurally relevant. Hence, the approach in the book belongs to the tradition of regional development theories, but, in contrast to the more traditional analyses, we will not consider the region as an economic unit per se. Rather, interactions between and within regions are very relevant to the performance of individual regions in an integrated world and will be at the core of the analyses of the following chapters.
6
U. Fratesi, L. Senn
In addition, the definition of region itself is not always constant in spatial analyses and it has been taken for granted for a long time that different intersections of geographical, administrative, economic, historical or social definitions are possible (Peare and Thomas 1968; Dawkins 2003). For example, in the literature on Regional Innovation Systems a region is defined as “a meso-level political unit set between the national or federal and local levels of government that might have some cultural or historical homogeneity but which at least has some statutory powers to intervene and support economic development, particularly innovation” (Cooke 2001, p.953). Administrative powers are not necessary for our purposes, while the presence of people in a territory, with some economic or social contiguity is fundamental for our operational concept of region. What we have in mind when we talk about regions is thus an economic region not very dissimilar to the conception of Boudeville (1966) as a continuous economic space a` la Perroux (1950). In principle, the territorial scale of this argument should not be relevant because any territorial scale could present problems of growth, development triggers, exposure to competition, they could be endowed with the same types of agents and so forth. However, to set in motion a sustainable competitive pattern of growth, a minimum threshold is needed, either in terms of geographic concentration or in terms of specialized/diversified structure. The number, characteristics and behaviour of agents are also important. Therefore, although a predetermined scale of analysis is not necessary, we will refer implicitly or explicitly to regions which, by and large, tend to converge towards the size of an administrative region that might approximate to an EU NUTS2 Region. Moreover, some chapters end up referring to administrative regions not because of theoretical necessity, but because policy issues are absolutely relevant throughout the book and policies, even those generated by higher administrative levels, are normally implemented within administrative boundaries. Finally, another obvious reason, especially relevant for the empirical chapters, is that statistical data normally refer to administrative boundaries. The focus on regions does not imply that clusters are irrelevant. On the contrary, the interest in clusters is explicit in some chapters of the book. However, this interest is not direct but comes from the influence of clusters on the development of regions and countries (Porter 1998; OECD 2001; Rosenfeld 2002, and, on a more critical approach, Martin and Sunley, 2003). A dynamic analysis based on the cluster (Fig. 1c) can be considered as working in an interception between the regional and the sectoral dimension. In fact, among the most diffused at present and most general cluster definitions, that of Porter (1998) focuses on the geographical concentration of interconnected companies.1 In theory, a region may not have a cluster defined as such but, generally, one or more will be present. When an economic region coincides with a cluster, as is the case with some industrial districts (Becattini 1990), innovative milieus (Aydalot 1986), new 1
In less recent works, such as and Czamanski and De Ablas (1979), the cluster concept was “devoid of any spatial connotation” (ibid, p. 62), being a subset of industries with closer than average connections, whereas the ‘industrial complex’ was meant to be a cluster with spatial concentration.
Regional Growth, Connections and Economic Modelling: An Introduction
7
economic spaces (Scott 1998), or local production systems (Crouch et al. 2001), the cluster itself should be considered as a particularly homogeneous economic region. When more than one cluster is present in a region, the dynamics of all of them predominantly determine the regional performance. The encompassing definition of a cluster and the eagerness of policy makers to imitate success stories has given rise to a cluster relevance in policies that is considered excessive by Asheim et al. (2006), in particular due to the theoretical shortcomings of the broad and static definition of Porter (1998) and to excessive reliance on case studies. The analyses of clusters in this book, on the contrary, will always be explicitly dynamic (in terms of generation and evolution) and involving theoretical argumentations instead of case studies. Moreover, our interest in clusters comes from the fact that it is often argued that the clustering of firms in related industries enhances knowledge creation (Maskell 2001; Maggioni 2002). The approaches adopted in this book ought to apply, with different degrees of relevance depending on the chapter, to economic regions, administrative regions and clusters; however, static analyses of regions and/or clusters (Fig. 1d) will not be featured, since they provide only a modest contribution to the explanation of growth and, even more importantly, because regional competitiveness does not last forever and its pre-requisites have to be continuously renewed and evaluated over time. It is hence normal that sustainability, the condition for persistence of growth over time is relevant to any analysis focused on long-term dynamics. Sustainability has many different definitions and measures, often involving environmental and socio-economic aspects (Hanley 2000). In particular, environmental aspects have gained increasing attention in regional science (Batabyal and Nijkamp 2004) but they need a different toolbox; hence the use of sustainability in this book will be socio-economic in nature. For our purpose, a process is deemed sustainable if the growth of today does not negatively affect the basis for growth tomorrow. Strictly linked with sustainable growth is the concept of resilience, i.e. the ability of a territorial socio-economic region to react to the modification of the external environment without losing its internal cohesion or its ability to grow.
2 Innovation As the Key Determinant of Sustainable Regional Performance In the Advanced World In older theories, the performance of regions, as well as the performance of countries whose theories were adapted to the study of regions, was basically due to demand in the Keynesian tradition, and supply, in the neoclassical tradition (Capello 2007). Theories such as the economic base recognized the presence of external demand factors as the source of regional growth. On the other hand, other theories, based on supply aspects, put the trigger of growth on internal factors, but these were mainly seen as the endowment of production factors, focusing especially on labour, capital and land. Other theories explored the role of geographical location and accessibility, important because of their effects on transport costs.
8
U. Fratesi, L. Senn
Only more recently, in the last 20 years, has the focus on material resources been deemed less important than the presence of less material or completely immaterial resources (human capital, knowledge, trust, embeddedness, etc.), thus addressing issues that help explain the capacity to innovate. Within neoclassical mainstream macroeconomics, starting with the seminal work of Romer (1986), the presence of knowledge, modelled as a public good, allows for long-run endogenous growth. In more complex models (Romer 1990; Grossman and Helpman 1991), some agents/firms/sectors produce innovation as an output, which is afterwards used as an input by other firms; this makes innovation non-rival but partially excludable rather than public and this is enough to assure endogenous long run growth. These developments have also strongly influenced regional theories, although they have limitations for this use (Martin and Sunley 1998). The New Economic Geography, which applies mainstream economic techniques to the analysis of spatial aspects and which started (Krugman 1991; Venables 1996) with the study of the static aspects of location and agglomeration, is now also developing (Martin and Ottaviano 1999; Baldwin and Martin 2004) dynamic models in which growth and location take place simultaneously, in order to detect the growth effects of agglomeration and the agglomeration effects of growth. In these models, growth takes place through innovation modelled as in the new growth theory. Evolutionary theories, starting with the seminal work of Nelson and Winter (1982), have pointed out the importance of innovation as a dynamic socio-economic process which leads to the economic performance of firms, sectors and nations. In particular, a broad set of literature has developed that investigates the National Systems of Innovation (NSI) (Lundvall 1992; Nelson 1993; Edquist 1997). The NSI theory has rapidly found its counterpart at the regional level with the introduction of the notion of a Regional Innovation System (e.g. Cooke 1996; Braczyk et al. 1998; Cooke et al. 1998) and the interpretative scheme of the Triple–Helix (Leydesdorff and Etkowitz 1998) which sees innovation as the outcome of the interaction of three main groups of local actors: firms, government and research institutions. The Regional Innovation System does not constitute the only paradigm in which space and innovation co-evolve. Innovation is seen as a territory-specific feature in a large number of other approaches, where it is no longer a process fully internal to the firm. Starting with Aydalot (1986), the GREMI Innovative Milieu school (Aydalot 1986; Camagni 1991; Maillat et al. 1993; Ratti et al. 1997) has investigated how the co-presence of economic actors in the territory and the relationships between firms in particular, can act to reduce dynamic uncertainty and hence favour innovation locally. These relationships have to be competitive and cooperative at the same time, so that firms will not fear opportunistic behaviour and could also act together in the common interest. The interaction of economic and social aspects and the focus on intra-firm and within territory relationships as a way to achieve external economies able to compensate for the lack of internal economies of scale, are not new; they were present for example in industrial districts (Becattini 1990). However, the focus of the milieu school is essentially dynamic and the innovative characteristics of the
Regional Growth, Connections and Economic Modelling: An Introduction
9
territory are essential in the definition of the milieu itself, in addition to fostering competitiveness. There is a series of other approaches and contributions that, even if they do not exactly constitute a school, move in the same direction and at the same time see innovation as the main driver of competitiveness and an essentially local process (see Simmie 2005 for a review). Among these are the new industrial spaces (Piore and Sabel 1984; Scott 1988), the learning region (Lundvall and Johnson 1994; Florida 1995; Morgan 1997) and Porter’s (1990) ‘competitive diamond’, which has been defined by its author so that the interactions between it four sets of factors are more effective when the firms are clustered in space. It should be remembered that innovation is a peculiar economic concept/entity, especially because it includes ‘public’ and ‘private’ aspects (Dosi 1988). For instance, it is characterized by a degree of appropriability which varies across industries and time (Dosi 1988). Technological innovation itself is not to be seen as a purely public good since acquiring the technology of others through imitation is costly, and only part of knowledge is codifiable in handbooks (Archibugi and Michie 1998). On the other hand, new knowledge can be utilized by different firms and regions at the same time. Finally, although it can be accumulated like capital, knowledge accumulation is highly path-dependent and innovations can be introduced only on the basis of knowledge previously possessed in the same or other closely related fields. Innovation remains a territorial process despite claims from authors such as Cairncross (1997) of the so-called ‘death of distance,’ since the ability to understand does not flow as easily as information does (Morgan 2004). In fact, learning normally takes place at the regional level (Storper 1995; Florida 1995, Malmberg and Maskell 2006) and geographical proximity normally enhances the other types of proximity (Boschma 2005). In particular, the territorial level allows types of learning different from those internal to the firm, such as collective learning which takes place through the mobility of qualified workers within the local labor market or through the relationships between customers and suppliers (through which technical and managerial information flows) or through imitation processes by firms which did not innovate in the first place, and generates what may be referred to as local spin-offs (Camagni 1991b). Since, as we have illustrated above, territories have to compete in the global economy through a continuous processes of innovation, the main aim of this book is to investigate the regional factors and mechanisms that are needed for these processes to take place.
3 The Phenomena and Factors of Growth in Innovative Regions To investigate the conditions which help a regional system to innovate and compete, we start from actual observation of regional growth phenomena. First of all, any regional growth process undergoes some dynamic structural change. We
10
U. Fratesi, L. Senn
observe that a regional economy grows over time either quantitatively (through the enlargement of its economy) or qualitatively (through the change or adjustment of its structure and sectoral mix). This change is either endogenously pushed or exogenously pulled by competition that, also for regions, is no longer national but global. Hence, growth normally involves structural change. However, this may occur disruptively or gradually with a smooth process. Creative destruction has been seen to be a common feature of innovation processes which need to replace old routines with new ones and in the process face resistance; all regional agents, public and private, prefer the gradual option. Not all structural changes, however, can take place gradually and this provides the opportunity for the local systems that are most keen to change to operate in a better position to exploit any window of opportunity. At the same time, strong discontinuities can interfere with the social fabric and lead to the de-structuring of the local production system; hence, sustainable growth processes tend to limit strong discontinuities as much as possible. Innovation is the broad conceptual category and factor which is recognized in the literature as allowing growth and structural change to occur; in the various approaches that have been described, innovation may take place through different processes (accumulation of knowledge, research and development, invention, productivity increases), and may be:2 • Radical and creative. In this case, significant discontinuities will probably be generated and major changes in the regional production systems will be needed in order to introduce it. • Incremental. In this second case, radical change is not needed to implement it. At the same time, especially in this case, path dependence phenomena may occur and regions may react to radical external innovations with incremental innovations and minor adaptations. In the short run, this strategy may be fruitful, but over a longer period, the detachment from state-of-the-art technology will harm competitiveness. Innovation may affect products, processes, organizations and may take place in two ways: either it is strategic, i.e. endogenously determined for the purpose of competing, or it comes from the adjustment to external influences or challenges. The most competitive regions are strong in strategic innovation, whereas decline can be postponed by adaptive innovation. Adaptive innovation occurs through the absorption of someone else’s ideas. The ability of regions to do this depends on their attitude to change, which depends in turn on a large number of structural socioeconomic factors, and on their ability to access external ideas, which depends on the role and connections of the region in inter-regional networks. The factors that are believed to affect the competitiveness of regions can be classified in groups. First, there are a number of microeconomic factors, some of which 2
In other cases it is more fruitful to add a third category and distinguish between radical, adaptive and incremental innovations depending on whether the innovation is completely new or applies to innovations from other fields originally developed for other purposes or it is the expansion of something that already exists (Bramanti and Senn, 1991).
Regional Growth, Connections and Economic Modelling: An Introduction
11
are traditional while others were more recently acknowledged. Investment is needed for regional firms to grow and renew their productive possibilities, hence the importance of the availability of capital. Moreover, capital ought also to be available for risky investments, since every innovation involves uncertainty; venture capital enables this, and is hence important, especially for innovative companies and high-tech clusters (Bottazzi and Da Rin 2002; Bresnahan and Gambardella 2004). Further, these opportunities are especially important for small and medium firms, which experience greater difficulties in finding traditional loans (Pollard 2003) and which, according to some studies, tend to be more often involved in radical innovation (Almeida and Kogut 1997). Another main factor of dynamic competitiveness for regions is the presence of tacit and codified knowledge (Fisher 2003). The latter can be easily transmitted with new communication technologies, but needs people able to understand and use it. The former, on the other hand, cannot be blueprinted but can only be transferred through common experience (Fisher 2001). In both cases, what is relevant is the number and the type of agents, their ability to interact and their previous knowledge. Defined as such, this microeconomic factor encompasses the traditional role of labor and also the more recently introduced developments in human capital theory (Lucas 1988). Obviously, the quality of schooling assumes an important role in this case (Wossmann 2002). The second group of factors are traditional macroeconomic ones, but often presented in a more modern framework. One factor that remains very important, despite the reduction in transport costs, is accessibility, which depends on the geographical location of the region and its endowment of infrastructure. Starting with Krugman (1991), it has been shown that reducing transport costs may lead to increasing agglomeration; moreover, transport costs remain one of the few factors that can be almost fully controlled by footloose firms (Vanhove 1999). Finally, starting with Aschauer’s (1989) seminal contribution, a broad literature has developed to measure the effects of infrastructure, and especially transport and mobility infrastructure on regional and national growth. Beyond transport infrastructure, research infrastructure, either public or private, has become particularly important for innovative regions, especially when it interacts with firms. Business-university relations, for example, have been shown to be a key factor for competitiveness (OECD 2002) and university decentralization has also been used successfully as a policy instrument (Andersson et al. 2004) even though the effects of academic expenditure depend on the existence of a critical mass of high-tech employment (Varga 2000). Considering innovation as an interactive process needing regional and external knowledge and actors, networking infrastructure (a theme which is expanded by Cret´ı, 2009, in this book) is even more important than just traditional infrastructure. The last macroeconomic factor is the industrial structure of the region. To compete worldwide, firms need to have strong complementarities with other firms, possibly within the same territory. The characteristics of the local productive fabric, and in particular the presence of local input-output relations, affect not only the costs of firms, but their innovative performance as well, due to the need for
12
U. Fratesi, L. Senn
interaction between different types of firms. This aspect is further investigated in the chapters by Folloni (2009) as well as by and Bramanti and Fratesi (2009), both in this book. The third group of factors influencing regional competitiveness is institutional and cultural. Institutional economics, an approach first introduced by North and developed by many others, conceives growth as determined simultaneously by economic mechanisms and actual actors (Davis and North 1971; North 1990). As mentioned in the definition of the first paragraph, the region in this book is not only a territory, but is mainly composed of the set of economic, social, political or administrative actors that live within the territory. These actors interact inside and outside the territory both as individuals and through higher-level structures, or institutions, that are one of the main characteristics of regions. Institutions can include very different things as one can see from the partitioning of Parto (2005) into five types: associative, behavioural, cognitive, regulative and constitutive. The study of the formation and evolution of regional institutions, especially those involved with innovation, is a main point of interest of the book itself, since it is generally thought to influence the performance of territories (Cooke and Morgan 1998). The mechanisms of governance are an important determinant of the regional innovation processes and for example it is possible to classify regional innovation systems as ‘grassroot’, ‘network’ and ‘dirigiste’ depending on which agents are the initiators of the innovation process (Cooke et al. 2004). The local market itself is not only a way of organizing the relationship between the actors, but can be a competitiveness factor for innovative firms. Porter (1990, 1998) identifies the conditions of local/national demand as a major stimulus of the innovativeness of firms, which increases their global competitiveness. Also from the point of view of pure demand, in the presence of increasing returns to scale, the size of the local market is important in determining the profitability, number or dimension of local firms, a point which is emphasized by new economic geographers when they talk of the ‘home market effect’ (Krugman 1980). The other factors belonging to this group are even less material. In particular, there is considerable evidence to support the role of entrepreneurship in territorial development and cluster formation (Audretsch and Keilbach 2004, 2005; Feldman and Francis, 2006; Garavaglia and Breschi, 2009, in this book), since not all agents share the same risk-awareness nor have they the same ability to perceive economic opportunities. Moreover, the local culture (or ‘atmosphere’) can be more or less supportive of potential entrepreneurs and in this way contribute to determining whether a potential entrepreneur will actually become one. Recently, the role of the creative environment for local growth and the attraction of innovative persons has been highlighted (Florida 2002, 2005). In particular, the city and the region can be seen as the main geographical scale for creativity processes to take place because they allow face-to-face contacts (Scott 2000, Storper and Venables 2004; Scott, 2006; see also Sacco and Segre, 2009, in this book). The factors of spatial organization constitute a fourth group, and mainly come from economic geography. Local spillovers are a strictly territorial factor which affect the performance of regions. The literature on spillovers is broad and basically
Regional Growth, Connections and Economic Modelling: An Introduction
13
consists of contributions investigating growth or knowledge spillovers (Audresch and Feldman 2004), the process in either case is unintentional, so that Maier and Sedlacek (2005) classify them as externalities. In the first case, the growth of neighbouring regions directly affects the local performance. In the second, spatial proximity is a factor that allows easier transfer of information from one firm to others that are geographically close; the mechanisms through which this may occur are various and need careful investigation (Breschi and Lissoni 2001a,b). Local knowledge spillovers are important because they may allow regional firms to achieve better innovative performance than they would have done by themselves and in this way boost regional performance. At the same time, the firms which developed the original innovation may find it harder to internalize the results of their R&D investment and this may discourage them from pursuing it. Many empirical studies have been done on the relationship between knowledge spillovers and regional growth (D¨oring and Schnellenbach 2006), estimating the effects of spillovers (e.g. Van Stel and Nieuwenhuijsen 2004; Fritsch and Franke 2004) and characterizing them, e.g. the maximum distance beyond which their effects are no longer significant (which Bottazzi and Peri 2003, estimate as 300 km). The quality and magnitude of local spillovers depends on the characteristics of internal networks, an issue which will be further developed in the next section. A second spatial factor is the presence of agglomeration economies, either sectorspecific (localization) or not (urbanization). The co-location of a large number of firms and workers in the same place generates externalities which are beneficial for the innovativeness of firms located in the region. If the concentration exceeds a certain threshold, however, congestion dis-economies may arise and reduce the advantages; with respect to this, the threshold is not fixed but the provision of adequate infrastructure and the modifications in the economic structure, for example with an economy becoming less material, can increase it. The advantages of agglomeration may be static, measured in terms of higher regional income or higher firm productivity, or dynamic, appraised by regional or firm growth, (Rosenthal and Strange 2004). The regional urban structure is very closely related to this. In addition to being a place where the traditional sources of external economies (labour market pooling, social overhead capital, diversified providers) are strong, the city assumes an even more important role when competitiveness is based on the above-mentioned creativity and on innovation (Simmie 2001). There is evidence that cities play an important role as incubators of new innovative enterprises, due to the technological uncertainties associated with the early stages of innovation and to the role that a diversified urban environment plays in facilitating search and experimentation in innovation (Feldman and Audretsch 1999). At the same time, the city can act as a milieu (Camagni 2001). Finally, the cities act as hubs for the international and international networks of transfer of that knowledge which is not diffused but rather organized in urban hierarchies and networks. The last group of factors affecting regional performance can be broadly defined as relational. Territorial economies can be thought of as stocks of relational aspects, and untraded interdependencies are even more important than market-traded ones
14
U. Fratesi, L. Senn
for the performance of some very successful advanced regions (Storper 1995, 1997). Social networks, and in particular the presence of trust relationships, are a major cause of the agglomeration of economic activities, even though they are neithert necessary nor present in every case (Gordon and McCann 2000). The importance attributed to social capital in the theories of regional growth has been increasing since its re-discovery by Putnam (1993). The relationships between the agents in the territory may allow collective learning (Capello 1999), which makes it possible for territories where firms are small to be as efficient in the creation and renovation of knowledge as larger firms are. All the factors described above have to be inter-connected and are complementary. All may be internal or external, but at the same time not all aspects are present in all (successful) regions at all times and with the same intensity.
4 Innovation, Networks and the Importance of Balanced Connections Among all the factors that shape regional growth and that facilitate innovation, we want to direct attention in this book to the role of networking. Already in Marshall, the possibility of facilitating knowledge flows was cited among the causes of agglomeration, but it is only recently that networking has been definitively recognized as a crucial factor both for economic growth and competition, and for the internalization and diffusion of innovation (Fisher 2003). Many recent theories of regional growth are based on the role of networks (innovative milieus, industrial districts, local systems of production, regional innovation systems, learning regions, etc.). The emergence and consolidation of networks stems both from hard support (infrastructure and telecommunications) and from soft factors, specifically the development of economic, social and political-administrative relationships. Networking takes place either between the regional economy and the global economy or within the regional economy itself. These two cases will hereafter be referred to as openness and embeddedness/robustness. They take different forms and occur through different channels. First, inter-regional, external networks and connections take place through interregional trade, which allows the exchange of goods and, less rapidly, of the technology incorporated into these goods (learning through reverse engineering). Also interregional input-output relationships for the exchange of intermediate goods allow for the same consequences: the fragmentation of production increases interregional dependence and theoretically places production networks in a much more complex and integrative structure (Jones and Kierzkowski 2005). A second major channel is international foreign direct investments (both inward and outward) which generate stable interregional economic relationships and are a very important channel for technology transfer (Baldwin et al. 2005) since, for example, a subsidiary controlled by a multinational corporation brings new
Regional Growth, Connections and Economic Modelling: An Introduction
15
products, new routines, new managerial expertise into the local economy. These can be acquired more or less rapidly by the local system through labour market mobility (Fosfuri et al. 2001), input-output linkages and reverse engineering. At the regional level, national capital flows are equivalent to FDI. The movement of workers between regions acts as a third important channel through which interregional networks are set up. In fact, workers embed knowledge of a large number of technological and organizational aspects. Moreover, in order to establish relationships with other regions, some knowledge of the local customs is often essential and is facilitated by people with personal experience. Also the personal relationships generated by interregional movement of people are often essential in the establishment of business relationships. Interregional networks can be directly related to innovation and knowledge transfer, for example when firms collaborate in consortia, or when firms within a region interact with external research institutions or universities. Finally, interregional relationships can also be mediated through higher level organizations such as a regional government which is directly involved in economic agreements with other regions as counterparts. Within regions, internal networks are set up through channels which are generally quite different, often being more informal. In this second case in fact, we deal with phenomena such as governance, social capital, collective learning, productive interdependence and labour market integration. All these channels are essential for local spillovers to occur. Local input-output relationships are one major formal channel for internal network to be set up. The importance for the regional economy of firms located in a region being linked to local suppliers (so that they become more embedded in the local economy and hence less footloose) has been emphasized. The local labor market is another main channel for internal networks to be established. The mobility of workers and labour market pooling and poaching have been traditionally interpreted as major factors of agglomeration, allowing firms to find the skills they need for their activities locally. Moreover, the mobility of workers is a channel for local knowledge spillovers to take place from firm to firm, and allows for collective learning. Formal and informal institutions are the third major channel for internal networks. They allow firms, entrepreneurs, managers and qualified workers to exchange knowledge and information about the market, business opportunities, innovations invented outside the region, and best practices. Some of the connections are directly dedicated to knowledge exchange through business-university relationships, the creation of consortia between firms, and joint research projects. It has to be emphasized here that internal networks go far beyond pure exchange to involve non-material aspects. For example, internal networks are particularly facilitated by a sense of belonging to the local community, which facilitates the formation of trust relationships between local economic agents. Trust implies that firms do not fear opportunism when undertaking joint-ventures, their relationships can be re-organized without fear of reprisals and that they are able to act collectively, which brings advantages to the whole group (Gordon and McCann 2000). The existence of trust relationships has been shown to be fundamental in many successful
16
U. Fratesi, L. Senn
areas, such as the Silicon Valley and the Italian industrial districts, and is one of the factors which the innovative milieu school advances as fundamental evidence for facilitating innovation. Both types of interconnections are needed: within the region and between regions. The first ones, involving firms, workers and institutions, are needed for knowledge and innovation to circulate within the regional economy and enhance the strength of the regional system, otherwise the firms would be competing in an atomistic and, eventually, footloose way. The link between innovation and external knowledge, which is another main topic of the book, is also fundamental at the regional level (Malecki and Oinas 1999; Bathelt et al. 2004). In fact, the process of innovation creation, which starts from local knowledge, is a very important mechanism but only for a very few pioneering regions and only in a few specific sectors is the regional component of creation of knowledge large with respect to the acquisition, adaptation and mastering of external, mainly international, knowledge. Inter-regional networks are hence needed for innovation, since the region is too small an entity to produce internally all the knowledge needed to compete globally and needs to learn and import external knowledge continuously. The good balance of the two mechanisms is a pre-requisite for sustained competitiveness (see Fig. 2) and, although research on each is extensively reported in the literature, the operation of a good balance of internal connections and interregional connections is an aspect into which further research is needed. In fact, the External Connections Openness
Too strong External Connections: risk of Disintegration
Right mix of openness and embeddedness. Sustainable growth
Too strong Internal Connections: risk of Localism
Internal Connections Embeddedness
Fig. 2 The right mix of internal (Embeddedness) and external connections (Openness). Source: our adaptation from Bramanti and Miglierina (1995)
Regional Growth, Connections and Economic Modelling: An Introduction
17
prevalence of one type of connections entails risks for the region (Bramanti and Miglierina 1995). If external networks are much stronger than the internal ones, the region risks disintegration. When firms are only marginally embedded in the territory, for example because they have few local suppliers, because they get most knowledge from outside the region, because the entrepreneurs are not personally locally networked, the firms themselves can easily relocate to follow better business opportunities such as lower production costs. If the internal networks are strong and the region is weak on external connections, there is the opposite risk of localism, which implies a regional economy is unable to acquire and master external knowledge and is hence likely to be less innovative and potentially less competitive. Moreover, overstrong internal ties, resulting from very tight social networks, can make it difficult for firms to explore different modes of production and to change their organization to meet external challenges. Finally, overstrong internal networks can make knowledge spillovers too great and in this way lead to the re-location of innovative firms away from the local system (Alsleben 2004).
5 Methodologies to Formalize Regional Theories of Growth: The Contribution of This Volume Regional economic modelling can be used for two different purposes: the first is directly linked to policy and aims at calculating ex-ante the effects of policies. The second is the formalization of theories. This may happen through models that are descriptive or causal; the formalization of theories, however, is often used as a first step in mastering the mechanics of the regional economy and in this way helps to build the economic representation of the regions that is needed to calculate impacts. In this book, the approach is essentially theoretical, hence we are not interested in case studies, nor in models directly involved with the calculation of impacts in a given regional economy; however, some contributions include empirical analyses whose scope goes beyond the estimation database to entail theoretical consequences of general interest. The approach of this book is eclectic regarding the methods used to analyse regional economies. We believe in fact that, despite many advances in each methodological approach and increasing convergences between them, none is yet able to represent with rigour all the complex aspects of the region. For this reason, depending on which specific aspects are the focus of investigation, one type of model can be more suitable than another. Within this book, and in the literature, we basically encounter three broad classes of methodological approaches to writing models: (a) Macro: The macro approach is used when the regional economy is modelled with variables. This used to be the rule until the 1980s in Keynesian, economic base and circular cumulative growth models. Simulation models, used in some chapters of the book, have made the macro approach a viable alternative today
18
U. Fratesi, L. Senn
for representing loops and feedbacks of complex systems and for this reason, some of the chapters of the book involve the creation of models which are complex with respect to more traditional models, but essentially macro in their characteristics. These models are sometimes defined as ‘ecological’ since their approach is based on the dynamics of different types of populations. Moreover, an added value in macro causal models is the ability to represent circular cumulative causation processes (following the tradition of Myrdal, Kaldor, Dixon-Thirlwall); the simulation allows these processes to be represented in a more realistic way and the phenomena to be better understood. Three contributions of the book use a macro approach, the chapters by Bramanti and Riggi (2009), by Folloni (2009), and by Bramanti and Fratesi (2009). (b) Micro-founded: This group of models incorporates the most recent (the last 20–25 years) mainstream models studying growth, agglomeration, localization or a combination of all three. These models have the advantage that macrobehaviour arises from the micro-behaviour of economic agents and hence the hypotheses on agent behaviour can be discussed before their consequences obtained. Moreover, these models have the advantage (which is in some cases also a limitation) that the behaviour of agents is optimizing and the outcome is an equilibrium, making them strictly rigorous. Although all contributors to this book are well aware of the developments in micro-based regional growth models, only two chapters (Cret´ı 2009, and Alderighi 2009) introduce new models of this type. The starting points are strictly territorial effects that are very complex to disentangle and, although the methodologies for solving micro-based models have advanced considerably in the last decades, it is still difficult to represent the complexity of a territorial system without making severly restrictive assumptions. (c) Agent-based: This is a class of models that has only recently become available and is still developing. It involves the building of a virtual economy in which various agents have a pre-determined but possibly differentiated behaviour; the computer simulation allows the researcher to make the resulting macro or population-based consequences of complex hypotheses evident. This is a very promising line of development for the representation of territorial processes, and is adopted by some chapters of the book. The advantages of this approach are made explicit in the chapter by Marino and Trapasso (2009), and an application to the dynamics of clusters can be found in the chapter by Maggioni and Roncari (2009). Finally, another approach may involve the use of econometric models, which start with theory and empirics at the same time in the identification of causal relationships between economic variables. This allows one to refer to them as systems of relationship models, since the procedure is different from hypothesis-testing which involves the use of econometric techniques to test whether some hypothesis are verified in reality. In this sense, alongside the study of theory, the book also aims at verifying with actual data the hypotheses that the literature and the book formulate about the behavior of regional economies. For this reason, some contributions to the book
Regional Growth, Connections and Economic Modelling: An Introduction
19
(by Camagni and Capello 2009, by Crescenzi and Rodr`ıguez-Pose 2009, by Riggi and Maggioni 2009 by Alderighi and Percoco 2009) also involve the setting-up and estimation of econometric relationships. Unfortunately, none of these is yet able to implement the advances in techniques suggested by Arbia, Copetti and Diggle (2009) in their chapter. Finally, the chapters by Garavaglia and Breschi (2009) and by and Sacco and Segre (2009) are essentially theoretical.
6 Book Synopsis The book further deepens these intuitions and is organized accordingly. It is divided into four parts, according to the main focus of the various chapters. The first part is the most general and aims at providing the framework for the rest of the book. The second part goes more deeply into the processes of innovation and entrepreneurship at regional level. The third part investigates some major factors of development for innovative regions. The fourth part is devoted to the methodologies which are to be used for these analyses. The rest of this first part of the book continues the intuitions of this introductory chapter with the aim of acting as a framework for the rest. The chapter “Sustainable Interrelated Growth: A Phenomenal Approach”, by Bramanti and Riggi, looks at the right mix of openness and robustness in terms of human capital, focusing on the role of internally-trained human capital and externally-trained human capital as the means for the region to connect both with inside and external regions. It is analytically shown with a Solowian model that the optimal growth rate depends on a balance of the two. Moreover, various mechanisms for the regional accumulation of human capital are compared and it is shown that an endogenous accumulation process is more effective only with a favourable initial situation and if the TSPI (Territorial System of Production and Innovation) has a significant size, otherwise exogenous accumulation is preferable. The chapter “A Model of Local Development”, by Folloni, introduces an analytical model of local growth that is able to represent the patterns of sustainable growth in a cumulative framework. This model focuses on the role of two types of firms in the local economy, of which only one, the leaders, are able to compete in external markets and are hence the key of regional competitiveness. At the same time, the local identity is also fundamental, since it allows the local system to grow, pulled by leader firms. For different values of the parameter, different types of local systems can be represented. It is finally shown that one vulnerable situation for a local system is to be characterized by firms that are too small and unable to compete externally. The chapter “The Dynamics of an Innovation-Driven Territorial System”, by Bramanti and Fratesi, investigates the dynamics of an innovation-driven territorial system with a macro-population model. In this contribution, the key to competitiveness is innovation, which depends on the interplay of internal and external factors.
20
U. Fratesi, L. Senn
The model adds to the two types of firms discussed in the Folloni chapter by introducing a third type of firm that is able to export but does not invest in R&D. This framework allows various simulations to show the theoretical existence of optimal values of parameters such as internal spillovers or the openness of the local system, maintaining the case for a balanced local system. The second part of the book goes more deeply into innovation and entrepreneurship, considered as the key aspects of regional competitiveness. In chapter “The Co-evolution of Entrepreneurship and Clusters”, Garavaglia and Breschi review the extensive literature on entrepreneurship within clusters. They start from the mechanics of the entrepreneurial process and then investigate what influences entrepreneurial activity in clusters, from the demand and the supply side. It is claimed that entrepreneurial activities are the real engine of cluster formation, growth and persistence in time, so that entrepreneurship and clusters co-evolve. Moreover, it is argued that pecuniary and non-pecuniary, as well as economic and sociological and psychological factors have to be jointly considered. The chapter “Learning, Innovation and Growth within Interconnected Clusters: An Agent Based Approach”, by Maggioni and Roncari, introduces an agent-based model for the interactions and the clustering of firms in a territory. The focus is on innovation and knowledge, which can come from either internal R&D or from two external sources: unintentional spillovers or intentional knowledge barter. Bounded rationality firms can relocate (with some cost) to take advantage of better opportunities and in this way give rise to very different results at the regional level. Furthermore, it is possible to show the drawbacks of several policies which attempt to support regional innovation systems and that differentiated clusters tend to be more resilient than homogeneous ones. In chapter “Knowledge-Based Economy and Knowledge Creation: the Role of Space”, Camagni and Capello analyze the role of space in knowledge creation and diffusion. They find three approaches to the knowledge economy that basically followed each other, starting with sector-based definitions, followed by functionbased definitions and, more recently, relationship-based definitions which focus on cognitive capabilities. If knowledge comes from a cognitive process, the territory becomes fundamental and three interacting relational preconditions have to integrate for regional competitiveness. This is shown by an empirical analysis that provides evidence that territories with a more efficient transfer and transcoding system show higher innovativeness. The chapter “Systems of Innovation and Regional Growth in the EU: Endogenous vs. External Innovative Efforts and Socioeconomic Conditions”, by Rodr´ıguezPose and Crescenzi, focuses on the socioeconomic conditions that make some regions more innovative than others. It is argued that regional innovation, fostered by innovative activities and benefiting from external connections, is not sufficient to translate into regional growth; the socio-economic conditions of the region and the regional innovation system act as a “social filter.” From an empirical analysis of EU Regions, the capacity of the local population to assimilate research generated locally or in neighbouring regions is shown to be more important than R&D expenditure.
Regional Growth, Connections and Economic Modelling: An Introduction
21
The third part of the book goes deeper into the main factors of regional growth and competitiveness. The first two contributions introduce two micro-based economic models to investigate ICTs and interconnection infrastructure. In the chapter “Some Conjectures on the Tie Between Digital Divide and Regional Disparities”, Alderighi introduces a new principal-agent model to investigate theoretically the relationship between the ICT divide and regional disparities. His analysis reveals that the link between investment in ICTs and regional performance may be indirect and mediated by the capabilities of local workers. In addition to explaining why performance may bear little correlation to investment, this chapter supports the case for policies aimed at the regional characteristics. In particular, if the cost of ICTs is higher in lagging regions, the gap can be reduced by providing investment incentives. On the other hand, if the lagging region is poorly endowed with skills, supporting ICT investment generates distortions and is less efficient than training. Network infrastructure is fundamental for the circulation of knowledge. For this reason, Cret`ı in “Interconnection Infrastructure as a Prerequisite for the Development of Territories: The Role of Network Externalities” investigates the decisions of firms to invest in connections with other firms. The complex relationship between network effects, network extent, TFP and aggregate welfare is analysed using a model in which the factor technologies usage directly affects Total Factor Productivity. In particular, the most favourable condition for a firm to increase its investment in network technologies is revealed. However, since externalities are present, some conflict between the private incentives to telecommunications usage and their effects on total welfare also arises. The chapter “Regional Growth and the Co-evolution of Clusters: The Role of Labour Flows”, by Riggi and Maggioni, works on the three dimensions of Fig. 1 and analyses the effects of industry-specific interactions and region-specific interactions on the dynamics of clusters. Using an ecological approach, they focus on labour flows as the main channel for these interactions, which is specifically consistent with their empirical analysis of US data. Depending on which of the two types of interactions are positive, cluster growth patterns are determined. Moreover, in high skill sectors, wage growth and employment growth appear to be complementary, whereas in low skill sectors, inward-migration negatively affects wages. External connections have been shown to be necessary for regional competitiveness. For this reason, in “Intranational Disparities, Regional Interactions and the Growth of Countries”, Alderighi and Percoco investigate the effects on growth of regional interactions, which are assumed to depend on regional differences. An analytical model is introduced to show how national performance is affected by regional interaction and then tested using the Gini index of the polarization of production as a proxy of interaction. The index is found to be positive and significant in a wide number of alternative econometric specifications. The last factor of regional competitiveness worthy of detailed analysis in the book is creativity. The chapter “Creativity, Cultural Investment and Local Development: A New Theoretical Framework for Endogenous Growth”, by Sacco and Segre, first analyses and summarizes the recent but already extensive literature on
22
U. Fratesi, L. Senn
the growth effects of creativity. Four main topics are addressed from a theoretical point of view: the core vs. non-core components of the creative class; the relationship between creativity, knowledge spillovers and local clusters of activities; the genius loci, stemming from the observation that a deepening tension between local and global is evident in culture; and the relationship between history, science, math and technology on the one hand and the arts on the other which lay the basis for culture-led local economic development. The four topics are finally brought together in the last part of the chapter in the foundation of a novel approach to endogenous growth. The fourth part of the book is more methodological. Faced with the complex phenomena of regional growth and competitiveness, many previously developed methodologies are inadequate. For this reason, two chapters analyse and support the most recent advances in the theoretical and empirical toolboxes. The chapter “Modelling the Individual Behaviour of Firms in the Study of Spatial Concentration”, by Arbia, Copetti and Diggle, presents a review of the methods proposed in the recent literature to study the location of economic agents, overcoming the limitations of the old approaches and looking at models that are empirically testable with statistical models. First, in the case of the measures of spatial concentration based on regional partitions, improvements are presented which consider an explicit measure of spatial agglomeration (or spatial correlation) together with the a-spatial characteristics of concentration. Secondly, methods are reviewed that are aimed at overcoming the problems linked to the arbitrariness of partitioning by considering the data mapped onto a continuous space. The chapter “A New Approach to Regional Economics Dynamics: Path Dependence and Spatial Self-Reinforcing Mechanisms”, by Marino and Trapasso, analyses the advances in the modelling of regional economies as complex evolving systems arising from self-reinforcing mechanisms due to the heterogeneity of agents, multiple equilibria, lock-in and path-dependency. Three main approaches, adaptive landscapes, complex networks and percolation, are presented and discussed. Faced with complexity, traditional economic policy loses effectiveness in promoting and sustaining economic development; moreover, the national dimension loses importance in favour of the local and global ones. Finally, a central planner is unable to govern all the underlying relationships between economic agents at any given time according to linear type response procedures. In “What Policy for Interconnected Territories? Conclusions and Openings”, which also concludes the book, Senn and Fratesi suggest some of the possible fields of application of the intuition that regions grow with a virtuous process if the public and private agents of its territory maintain an equilibrated set of connections both inside and outside the region. Seven fields are exemplified, related to as many important areas of policy (among the many possible) in order to suggest that proving the intuition would have important policy implications. The issues are: promotion of interregional and international trade; attraction of foreign direct investments (FDI); development of R&D and innovation; tourism attraction and marketing; management and governing of the impact of large infrastructural projects; promotion and
Regional Growth, Connections and Economic Modelling: An Introduction
23
consolidation of clusters, industrial districts and local production systems; regional and urban strategic planning.
References Alderighi M (2009) Some conjectures on the tie between digital divide and regional disparities. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Alderighi M, Percoco M (2009) Intranational disparities, regional interactions and the growth of countries. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Almeida P, Kogut B (1997) The exploration of technological diversity and the geographic localization of innovation. Small Bus Econ 9:21–31 Andersson R, Quigley JM, Wilhemson M (2004) University decentralization as regional policy: the Swedish experiment. J Econ Geogr 4:371–388 Arbia G, Copetti M, Diggle P (2009) Modelling the individual behaviour of firms in the study of spatial concentration. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Archibugi D, Michie J (1998) Technical change, growth and trade: new departures in institutional economics. J Econ Surv 12–3:247–332 Aschauer DA (1989) Is public expenditure productive? J Monetary Econ 23:177–200 Asheim B, Cooke P, Martin R (eds) (2006) Clusters and regional development, critical reflections and explorations. Routledge, London Asleben C (2004) The downside of knowledge spillovers: an explanation for the dispersion of high-tech industries. J Econ 84(3):217–248 Audretsch DB (1998) Agglomeration and the location of innovative activity. Oxford Rev Econ Pol 14(2):18–29 Audretsch D, Keilbach M (2004). Entrepreneurship and regional growth: an evolutionary interpretation. J Evol Econ 14:605–616 Audretsch D, Keilbach M (2005). Entrepreneurship capital and regional growth. Ann Reg Sci 39:457–469 Aydalot Ph (ed) (1986) Milieux innovateurs en Europe. Paris, GREMI, C3E Baldwin R, Martin P (2004) Agglomeration and regional growth. In: Henderson JV, Thisse J-F (eds) Handbook of regional and urban economics. North Holland, Amsterdam Baldwin R, Braconier H, Forslid R (2005) Multinationals, endogenous growth, and technological spillovers: theory and evidence. Rev Int Econ 13(5):945–963 Batabyal AA, Nijkamp P (2004) The environment in regional science: an eclectic review. Pap Reg Sci 83:291–316 Bathelt H, Malmberg A, Maskell P (2004) Clusters and knowledge: local buzz, global pipelines and the process of knowledge creation. Prog Hum Geogr 28(1):31–56 Becattini G (1990) The Marshallian industrial district as a socio-economic notion. In: Pyke F, Becattini G, Sengenberger W (eds) Industrial districts and inter-firm co-operation in Italy. ILO, Geneva Boschma RA (2005) Proximity and innovation: a critical assessment. Reg Stud 39.1:61–74 Bottazzi L, Da Rin M (2002) Venture capital in Europe and the financing of innovative companies. Econ Pol v.34:229–269 Bottazzi L, Peri G (2003) Innovation and spillovers in regions: evidence from European patent data. Eur Econ Rev 47:687–710 Boudeville R (1966) Problems of regional economic planning. Edinburgh University Press, Edinburgh
24
U. Fratesi, L. Senn
Braczyk HJ, Cooke P, Heidenreich M (eds) (1998) Regional innovation systems. UCL Press, London Bramanti A, Fratesi U (2009) The dynamics of an innovation driven territorial system. In: Fratesi, U. and Senn, L. (eds.) Growth and innovation of competitive regions: the role of internal and external connections, Springer, Berlin Bramanti A, Miglierina C (1995) Alle radici della crescita regionale: fattori, fenomeni agenti. L’Industria XVI(1) Bramanti A, Riggi MR (2009) Sustainable interrelated growth: a phenomenal approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Bramanti A, Senn L (1991) Innovation, firm and milieu: a dynamic and cyclic approach. In: Camagni R (ed) Innovation networks spatial perspectives, GREMI, London, pp 89–104 Breschi S (2000) The geography of innovation: a cross-sector analysis. Reg Stud 34.3:213–229 Breschi S, Lissoni F (2001a) Knowledge spillovers and local innovation systems: a critical survey. Ind Corp Change 10–4:975–1005 Breschi S, Lissoni F (2001b) Localised knowledge spillovers vs. innovative milieux: knowledge “tacitness” reconsidered. Pap Reg Sci 80:255–273 Bresnahan T, Gambardella A (2004) Building high tech clusters: silicon valley and beyond. Cambridge University Press, Cambridge Cairncross F (1997) The death of distance. Harvard Business School Press, Cambridge Camagni R (ed) (1991) Innovation networks spatial perspectives. GREMI, London Camagni R (2001) The economic role and spatial contradictions of global city-regions: the functional, cognitive and evolutionary context. In: Scott AJ (ed) Global city-regions: trends, theory, policy. Oxford University Press, Oxford, pp 96–118 Camagni R, Capello R (2009) Knowledge-based Economy and Knowledge Creation: the Role of Space. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Capello R (1999) Spatial transfer of knowledge in high-technology milieux: learning vs. collective learning processes. Reg Stud 33:353–365 Capello R (2007) Regional economics. Routledge, London Cooke P (1996) The new wave of regional innovation networks: analysis, characteristics and strategy. Small Bus Econ 8:159–171 Cooke P (2001) Regional innovation systems, clusters and the knowledge economy. Ind Corp Change 10(4):945–974 Cooke P, Morgan K (1998) The associational economy: firms regions and innovation. Oxford University Press, Oxford. Cooke P, Heidenreich M, Braczyk H-J (2004) Regional innovation systems, the role of governance in a Globalized World, 2nd edn. Routledge, London Cooke P, Uranga MG, Extebarria G (1998) Regional systems of innovation: an evolutionary perspective. Env Planning A, 1998/30 Crescenzi R, Rodr`ıguez-Pose A (2009) Systems of innovation and regional growth in the EU: endogenous vs external innovative efforts and socioeconomic conditions. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Cret´ı A (2009) Interconnection infrastructure as a prerequisite for the development of territories: the role of network externalities. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Crouch C, Le Gales P, Toglia C, Voelzkow C (2001) Local Production Systems in Europe: Rise or Demise? Oxford University Press, Oxford Czamanski S, De Ablas LA (1979) Identification of industrial clusters and complexes: a comparison of methods and findings. Urban Stud 16:61–80 Davis LE, North DC (1971) Institutional change and American economic growth. Cambridge University Press, Cambridge
Regional Growth, Connections and Economic Modelling: An Introduction
25
Dawkins CJ (2003) Regional development theory: conceptual foundations, classic works, and recent developments. J Planning Literature 18(2):131–172 D¨oring T, Schnellenbach J (2006) What do we know about Geographical knowledge spillovers and regional growth? A survey of the literature. Reg Stud 40.3:375–395 Dosi G (1988) Sources, procedures and microeconomic effects of innovation. J Econ Lit XXVI:1120–1171 Dosi G, Freeman C, Nelson R, Silvenberg G, Soete L (eds) (1988) Technical change and economic theory. Pinter, London Edquist C (ed) (1997) Systems of innovation. Pinter, London Feldman M, Audretsch DB (1999) Innovation in cities: science-based diversity, specialization and localized competition. Eur Econ Rev 43:409–429 Feldman M, Francis JL (2006) Entrepreneurs as agents in the formation of industrial clusters. In: Asheim et al (eds) Clusters and regional development. Routledge, London, pp 115–136 Fisher M (2001) Innovation, knowledge creation and systems of innovation. Ann Reg Sci 35– 2:199–216 Fisher M (2003) The new economy and networking. In: Jones DC (ed) New economy handbook. Academic, New York, pp 343–367 Florida R (1995) Towards the learning region. Futures 27(5):527–536 Florida R (2002) The rise of the creative class. Basic Books, New York Florida R (2005) The flight of the creative class: the new global competition for talent. Harper Collins, London Folloni G (2009) A model of local development. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Fosfuri A, Motta M, Ronde T (2001) Foreign direct investment and spillovers through workers’ mobility. J Int Econ 53:205–222 Fritsch M, Franke G (2004) Innovation, regional knowledge spillovers and R&D cooperation. Res Policy 33:245–255 Fujita M, Mori T (1996) The role of ports in the making of major cities: self-agglomeration and hub effect. J Dev Econ 49:93–120 Gabszewicz JJ, Thisse J-F, Fujita M, Schweizer U (1986) Location theory. Harwood, London Garavaglia C, Breschi S (2009) The co-evolution of entrepreneurship and clusters. In: Fratesi, U. and Senn, L. (eds.) Growth and innovation of competitive regions: the role of internal and external connections, Springer, Berlin Gordon IR, McCann P (2000) Industrial clusters: complexes, agglomeration and/or social networks? Urban Stud 37(3):513–532 Grossman GM, Helpman E (1991) Quality ladders in the theory of growth. Rev Econ Stud 58(1):43–61 Hanley N (2000) Macroeconomic measures of sustainability. J Econ Surv 14–1:1–30 Henderson JV, Thisse J-F (eds) (2004) Handbook of Regional and Urban Economics. North Holland, Amsterdam Henderson JV, Shalizi Z, Venables AJ (2001) Geography and development. J Econ Geogr 1:81–105 Isard W (1956) Location and space-economy. MIT, Cambridge Jacobs J (1969) The economy of cities. Random House, New York Jones RW, Kierzkowski H (2005) International fragmentation and the new economic geography. N Am J Econ Finance 16(1):1–10 Krugman P (1980) Scale economies, product differentiation, and the pattern of trade. Am Econ Rev 70:950–959 Krugman P (1991) Increasing returns and economic geography. J Polit Econ 99(3):483–499 Leydesdorff L, Etkowitz H (1998) The triple-helix as a model for innovation studies. Sci Public Policy 25(3):195–203 Lucas R (1988) On the mechanics of economic development. J Monetary Econ 22:3–42 Lundvall B-A (ed) (1992) National systems of innovation. Pinter, London Lundvall B-A, Johnson B (1994) The learning economy. J Ind Stud 1(2):23–42
26
U. Fratesi, L. Senn
Maggioni MA (2002) Clustering dynamics and the location of high-tech-firms. Physica-Verlag, Berlin Maggioni MA, Roncari S (2009) Learning, innovation and growth within interconnected clusters: an agent based approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Maier G, Sedlacek S (2005) Spillovers and innovation, environment and space, an introduction. In: Maier G, Sedlacek S (eds) Spillovers and innovations, space, environment and the economy. Springer, Berlin, pp 1–18 Maillat D, Qu´evit M, Senn L (1993) Reseaux d’innovation et milieux innovateurs: un pari pour le d´eveloppement r´egional Gremi-Edes, Neuchˆatel Malecki EJ, Oinas P (1999) Making connections: technological learning and regional economic change. Ashgate, Aldershot Malerba F (2006) Innovation and the evolution of industries. J Evol Econ 16:3–23 Malmberg A, Maskell P (2006) Localized learning revisited. Growth Change 37–1:1–18 Marino D, Trapasso R (2009) A new approach to regional economics dynamics - Path dependence and spatial self reinforcing mechanisms. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Markusen J, Venables A (1999) Foreign direct investment as a catalyst for industrial development. Eur Econ Rev 43(2):335–356 Martin P, Ottaviano GIP (1999) Growing locations: industry location in a model of endogenous growth. Eur Econ Rev 43:281–302 Martin R, Sunley P (1998) Slow convergence? The new endogenous growth theory and regional development. Econ Geogr 74–2:201–227 Martin R, Sunley P (2003) Decostructing clusters: chaotic concept or policy panacea? J Econ Geogr 3:5–35 Maskell P (2001) Towards a knowledge-based theory of the geographical cluster. Ind Corp Ch, 10–4:921–943 McCann P (ed) (2002) Industrial location economics. Edward Elgar, Cheltenham Morgan K (1997) The learning region: institutions, innovation and regional renewal. Reg Stud 31.5:491–503 Morgan K (2004) The exaggerated death of geography: learning, proximity and territorial innovation systems. J Econ Geogr 4:3–21 Nelson R (1993) National innovation systems: a comparative analysis. Oxford University Press, New York Nelson R, Winter SG (1982) An evolutionary theory of economic change. Harvard University Press, Cambridge North DC (1990) Institutions, institutional change and economic performance. Cambridge University Press, Cambridge OECD (2001) Innovative clusters: drivers of national innovation systems. OECD, Paris OECD (2002) Benchmarking industry-science relationships. OECD, Paris Parto S (2005) Economic activity and institutions: taking stock. J Econ Issues 39(1):21–52 Peare CWD, Thomas H (1968) Regional economic statistics. J Roy Statistical Society 131(3): 330–339 Perroux F (1950) Economic space: theory and applications. Q J Econ. 64(1):89–104 Piore MJ, Sabel CF (1984) The second industrial divide. Basic Books, New York Pollard JS (2003) Small firm finance and economic geography. J Econ Geog 3:439–452 Porter M (1990) The competitive advantage of nations. Billing and Sons, Worchester Porter M (1998) On Competition. Harvard Business School Press, Harvard Putnam RD (1993) Making democracy work: civic traditions in modern Italy. Princeton University Press, Princeton Ratti R, Bramanti A, Gordon R (eds) (1997) The Dynamics of Innovative Regions: the GREMI Approach. Ashgate, Adelrshot
Regional Growth, Connections and Economic Modelling: An Introduction
27
Riggi MR, Maggioni M (2009) Regional growth and the co-evolution of clusters: the role of labour flows. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Romer (1986) Increasing returns and long-run growth. J Polit Economy. 94:1002–1037 Romer P (1990) Endogenous technological change. J Polit Econ 98.5:part II, S71–S102 Rosenfeld SA (2002) Creating smarty systems, a guide to cluster strategies in less favoured regions, European Commission, DG-REGIO, Brussels Rosenthal SS, Strange WC (2004) Evidence on the nature and sources of agglomeration economies. In: Henderson JV, Thisse J-F (eds) Handbook of regional and urban economics. North Holland, Amsterdam Sacco PL and Segre A (2009) Creativity, cultural investment and local development: a new theoretical framework for endogenous growth. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Scott AJ (1998) New industrial spaces, Pion, London Scott AJ (2000) The cultural economy of cities. Sage Publications, London Scott AJ (2006) Entrepreneurship, innovation and industrial development: geography and the creative field revisited. Small Bus Econ 26:1–24 Senn L, Fratesi U (2009) What policy for interconnected territories? Conclusions and openings. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Simmie J (2001) (ed) Innovative cities. Spon Press, London Simmie J (2005) Innovation and space: a critical review of the literature. Reg Stud 39.6:789–804 Storper M (1995) The resurgence of regional economies, ten years later: the region as a Nexus of untraded interdependencies. Eur Urban Reg Stud 2(3) Storper M (1997) The regional World – territorial development in a global economy. Guilford Press, New York Storper M, Venables AJ (2004) Buzz: face-to-face contacts and the urban economy. J Econ Geogr 2004:351–370 Van Stel AJ, Nieuwenhuijsen HR (2004) Knowledge spillovers and economic growth: an analysis using data of Dutch regions in the period 1987–1995. Reg Stud 38(4):393–407 Varga A (2000) Local academic knowledge transfers and the concentration of economic activity. J Reg Sci 40.2:289–309 Venables AJ (1996) Equilibrium locations of vertically linked industries. Int Econ Rev 37:341–359 Vickerman RW (1991) Regions’ infrastructure in a region’s development. In Vickerman RW (ed) Infrastructure and regional development. Pion, London Wang J, Blomstrom M (1992) Foreign direct investment and technology transfer: a simple model. Eur Econ Rev 36:137–155 Wossmann L (2002) Schooling and the quality of human capital. Springer, Berlin
Sustainable Interrelated Growth: A Phenomenal Approach Alberto Bramanti and Massimiliano R. Riggi
1 ‘Sustainable Growth’ and ‘Endogenous Development’: a Preamble It is strikingly clear that local contexts are more and more important as providers of rich external economies, shaping the form and sustainability of territorial competitiveness (Camagni 1991; Markusen 1996; Bramanti and Maggioni 1997; Crounch et al. 2001). The process of globalisation has been an important driver of change exposing even the most remote areas to competition, but cognitive and normative resources, as well as social capital and trust, are deeply embedded in the territorial social fabric. It is thus quite natural to identify the two driving forces of dynamic competitiveness in: (a) the robustness of local connections (frequently considered as the ‘milieu effect’) and (b) the openness of the territorial system – made up of actors, firms, and institutions – to the global world, to the long range networks of finance, information and globalised markets. Robustness refers to a relational space (Camagni 1991) which is the field of social interaction, interpersonal synergy, social collective action, imitation of successful managerial practice, interpersonal face-to-face contact, informal cooperation between firms, tacit circulation of information and knowledge. It is worth noting that the TSPI1 (Territorial System of Production and Innovation), as a relational space, empowers and guides innovative agents to innovate and coordinate each other. TSPI is the territorial counterpart of what the American economic sociologist Mark Granovetter has labelled the ‘embeddedness’ of social and economic 1 This is a multi-faceted concept deeply rooted in regional analysis. Territorial Production Systems (TPS) have been widely used by a number of scholars and in a dynamic view it brings together a ‘production system’, a ‘technical culture’ and a set of actors (enterprises, professional associations, local authorities, universities and laboratories, individuals). The TSPI adds the concept of innovation to the previous TPS, and there are two main mechanisms at work: the first is the creation of innovation and the second is the diffusion of the innovation, which is always a social type of communication (see Bramanti and Fratesi, 2009, in this book).
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
29
30
A. Bramanti, M.R. Riggi
processes (Granovetter 1973). We are far beyond the notion of external economies as simply scale-related: they are complex outcomes of interaction between scale, specialization and flexibility in the context of proximity. Openness, on the other hand, refers to the linkages that a TSPI establishes with: leading worldwide institutions (supra-national institutions, major investment banks, global business consultants); state-of-the-art information and knowledge producers (first top universities, research centres, techno parks); global champion firms (market leaders and very demanding clients) and communities of practices (knowledge workers operating within learning organizations). Through the long-range linkages a TSPI can get in touch and learn ‘newness’ (techniques, styles of life, needs) wherever it springs up. So, within a TSPI ‘gate’ institutions and actors become more and more central to the system success. The presence of a major university, an international fair, a global cultural event are all basic components of the creative economy and a huge source of competitive advantage. The relevance of the role played by local and global relationships in determining the economic performance of TSPI is witnessed by their current centrality both in the theoretical and empirical literature, as well as in the policy debate. Following Maggioni and Bramanti (2002) we can look at local identity (proximity) and global relationships (networks) as two complementary elements useful to ensure a stable path of ‘sustainable growth’ and ‘endogenous development’. “These two terms seem to mimic and mix the two buzzwords of mainstream economics: ‘endogenous growth’ and ‘sustainable development’. On the contrary, by sustainable growth we refer to a given set of production and innovation dynamics concerning the existence of enabling conditions for the re-production of the local endowment of resources allowing positive performance and the persistence of the TSPI in the long run. Endogenous development refers to the governance of the TSPI and concerns the capability of local agents to control and guide (at least partially) the patterns of qualitative and quantitative expansion of the system.” (p. 248).
In order to investigate the relational features of development2, we look at the TSPI analysed along a ‘phenomenal approach’. Phenomena represent a display of reality within which factors and actors behaviours interact: phenomena are therefore the result of underlying causal connections and actors behaviour (Bramanti and Miglierina 1995). Phenomena are more elaborate concepts than factors, even though they are a bit more ‘vague’; to some extent we can tackle phenomena as ‘meta factors’ and in the present chapter we want to model ‘sustainable growth’ as deriving from the strict interplay of two specific meta-factors, or phenomena: robustness and openness. This working hypothesis is far from being neutral, but we are convinced that this approach is a convenient approximation of how sustainable regional growth relies 2
That are human relations essential to many types of economic coordination and, specifically, human relations, rules and conventions that are at the heart of the economic process today. Many ‘non-economic’ forces – such as institutions, cultures, and social practices – exert a relevant role in economic life; they are central to the economic process as they underpin the mobilization of economic resources and the organization of production systems, the exchange and convergence in international best practices.
Sustainable Interrelated Growth: A Phenomenal Approach
31
on the process of production, accumulation and exchange of knowledge supported and fostered by the interplay of the two cardinal dimensions of ‘internal synergy’ and ‘external energy’. We address this issue in two steps: first we explain, in the space of phenomena, how TSPI growth relies on innovation, which remains the main trigger of territorial growth, and is fostered externally by open worldwide networks and diffused internally and exploited by local dense relations; second we model the growth relation by operationalising ‘robustness’ and ‘openness’ phenomena through human capital. Further on in the chapter we deal with alternative ways of looking at human capital accumulation and at the deriving alternative policies.
2 From Innovation to Growth: A Phenomenal Approach The logical frame adopted to explain sustainable territorial growth is quite simple. Innovation is a must in TSPIs operating in OECD countries3 facing a global competition. We do not have to digress to justify this starting assumption as the theme is fully developed in the Bramanti–Fratesi chapter within the book. We can only remind the reader that innovation is a complex phenomenon in which external and internal information, knowledge, competence and creativity mix together. While the production and exchange of information are mainly related to a-spatial networks in which codified knowledge is exchanged on a global scale, knowledge and competence feed on both internal and external circuits despite clear evidence of a major role played by proximity and face-to-face contacts. Creativity emerges from the synergetic meeting of the previous stages – information, knowledge and competence – combined in a specific cultural and territorial context strongly embedded in people. Therefore, an over-emphasis on ‘internal elements’ – and sometime on strong ties4 – produces an inward looking TSPI in which the innovative potential is reduced to learning-by-doing and learning-by-using activities, generating incremental innovations. When radical innovations are required weak access to external sources (loose external linkages and weak ‘gate’ actors) reduces the spectrum of technological opportunities.
3
This is an approximation to indicate TSPIs belonging to an already developed economic macro context. The model supports a wide range of applications but it is not consistent with developing countries in their early stages context. 4 Strong ties refer to family membership, close friends and long time neighbours or co-workers. They are highly demanding and very selective, marked by trust and reciprocity in multiple areas of life. According to sociologists who study networks most people can make and manage between five and ten strong-tie relationships. But weak ties are indeed more important following the seminal article by Granovetter (1973). Weak ties are a key mechanism for mobilizing resources, ideas and information; weak ties require less investment and we can use them more opportunistically, they allow for rapid entry of new people and rapid absorption of new ideas offering fundamental support to the creative process (Florida, 2002).
A. Bramanti, M.R. Riggi
Openness: long-range networks, external energy, strategic alliances
The outcome of innovation is total output growth
Collective learning mechanisms
Information The mix of internal and external circuits gives origin to a production, Knowledge Creativity accumulation and exchange of information, knowledge, competence and Competence creativity
Windowing for innovation
TSPI innovation process
Creative capital vs. social capital
Robustness: milieu connections, internal cohesion, local synergy Strong ties vs. weak ties
Part of the output is invested in openness and robustness
32
The output of knowledge circuits and learning mechanisms is innovation (in the firm and in the system)
Fig. 1 A diagrammatic view of TSPI sustainable growth mechanism
The reverse occurs with an over-emphasis on external networks and codified knowledge. Here the TSPI will not learn through the interaction of local actors and the most likely result it the loss of local identity, a weakening process of the ‘genius loci’ and a deriving lesser attractiveness of TSPI towards external assets, as far as the possible delocalization of the production process. The mechanisms at work within the TSPI are captured by the following diagram (Fig. 1): • Local robustness or internal relationships is – in the ‘phenomenal space’ – a proxy of the absorptive capacity of the system in terms of innovation, it is here that the so called ‘genius loci’ is at work; • Openness or external linkages refers to the general innovation stimuli and refers to information, knowledge and capabilities that affect the innovative capacity; • The process of blending and mutually reinforcing robustness and openness generates truly innovative products, strengthens the competitiveness of TSPI and reinforces the learning mechanisms5 of the local milieu; • The outcome of a truly innovative TSPI is increasing total output according to the ‘productivity of innovation’, i.e. the capacity to transform advances in the 5
Learning may be understood as a qualitative change in a person’s way of seeing, experiencing, understanding, and conceptualizing something in the real world. In addition, learning is not something that requires time out from productive activities; it is the hearth of productive activities. To put it simply learning in the new form of labour.
Sustainable Interrelated Growth: A Phenomenal Approach
33
technological field into new products for which there is demand in the market and generate more revenues. In this frame the ‘fundamentals’ of the TSPI (from the production processes to the governance routines) define its ability to exploit innovation within the production function. If the region benefits from a thick institutional setting, valuable entrepreneurs, friendly networks, cooperation between actors (all determinants of relational development), it is more able to put innovation at work. On the other hand, if the region suffers from rarefied connections, predatory behaviour, and weak socio-economic environment, potential innovation is difficult to transform into territorial growth. Openness can be thought of as a fundamental building block of growth because it plays the gate role through which innovation flows into the TSPI; robustness represents the other side of the coin: it strengthens the absorptive capacity of different actors at work within the TSPI and enables them to fully exploit the external stimuli. There are certainly many different ways of looking at ‘openness’ and ‘robustness’ but we consider the human capital approach a priority. The loops and feed-backs of knowledge production, accumulation and exchange (sketched in Fig. 1) are deeply rooted in organizations and institutions that, in turn, depend heavily on people: skilled human capital, workforce, entrepreneurs. People are at the core of the creative process which marks a wide range of knowledge-intensive industries; people are the subject of learning process, people: “apply or combine standard approaches in unique ways to fit situations, exercise a great deal of judgement, perhaps try something radically new from time to time (. . . ) engage in work whose function is to create meaningful new forms.” (Florida, 2002: 69).
For the sake of simplicity we distinguish between people (i.e. human capital in the standard economic approach) between ‘internally trained’ human capital (ITHC) – the local production factor – and ‘externally trained’ human capital (ETHC), i.e. people6 coming from the outside, frequently endowed with superior education, experience, and with different competences from those available within the TSPI.
3 The Core Model: A Solowian Approach Having described the process of innovation (technological, organizational and social) of TSPI – rooted in the balancing of the two ‘building blocks’ (robustness and openness) – we are now able to introduce a simple ‘neoclassical’ model in which the regional total output – the outcome of innovation – is the result of internal robustness (hinging on the ‘genius loci’) and external openness of the system to global phenomena and stimuli (spurring innovation patterns). Openness is measured in terms of ETHC – i.e. personnel with higher formal education and international experience in the field of innovation – while robustness 6
We are here referring to the ‘creative class’ of Florida (2002) distinguishing it from the ‘service class’ – frequently referring to immigrants – which includes workers in lower-wage, lower-autonomy service occupation.
34
A. Bramanti, M.R. Riggi
is explained in terms of ITHC – i.e. a workforce well acquainted with local routines and exchanging knowledge tacitly. The basic resource in this model is human capital, which we use/measure as units of input within the regional production function. We choose to represent the regional production function as a Solowian model of capital accumulation. We therefore define a Cobb Douglas production function with constant returns to scale, such that openness (Et ) and robustness (Ht ) are the two main inputs expressed in terms of externally and internally trained workers respectively, both contributing to the definition of Yt that represents the flow of output produced at time t: Yt = Htα Et1−α where α (with 0 ≤ α ≤ 1) is the technological parameter expressing the share of total output invested in the local factor (ITHC) and its complement to unity 1 − a represents the share of total output invested in the external factor (ETHC, a proxy for innovation coming from outside). As in the standard Solowian model, output is a homogeneous good that can be consumed Ct or invested It ; if we look at total product from an expenditure approach we can therefore state that it is allocated to consumption and internal investment: Yt = Ct + IHt where IHt is the investment in the local factor (Ht − Ht−1 ). Assume that openness is given and is defined by an exogenous dynamic at the growth rate gE according to the relation Et = Et−1 egE (with E0 = 1 for simplicity). The evolution of human capital deserves further attention; we need to specify the process of its accumulation over time with a law of motion: Ht = Ht−1 + sH Yt − δ Ht so that the net increase in the stock of human capital at a point in time equals gross investment less depreciation. sH Yt is the share of output invested in local factor accumulation, which depreciates at a constant rate δ > 0. We aim to show that if more output is invested in the local factor (ITHC) the regional development path is biased toward local strong ties while, investment in innovation (ETHC bridging technological advances) is limited: when the level of consumption is sub-optimal, the implication is that the balance between openness and robustness is inappropriate. In order to do that, we express the model in relative terms, dividing by Et : when comparing two regions in terms of output, we are not interested in their absolute size, but in the output they are able to produce with internal factor. The Cobb-Douglas α production function becomes: Yt = Ht which is analogous with the ‘intensive form’ of the Solowian Et Et production function. Using lower case for relative variables we can write: yt = htα where y may be considered as the total attainable output provided a given level of external innovation and ha the internal/external ratio in terms of knowledge embedded in human capital (something like a tacit/codified ratio in the production and exchange of knowledge).
Sustainable Interrelated Growth: A Phenomenal Approach
35
The equation for the dynamics of human capital then becomes: Ht−1 Ht Ht Ht Yt Ht H˙ Et = Et +sH Et − δ Et or, Et = sH Et − δ Et (dot indicates the variation over time of the underlying variable). Since the variation over time of HEtt can be expressed as: ˙ E˙ , the accumulation of human capital over time in relative terms becomes h˙ = HE−H E2 h˙ = sH f (h) − (gE + δ )h, which has the advantage of depending only on ht . Both Et and Ht show decreasing returns to scale, so that an equilibrium is reached when 1 α −1 (* indicates variables in equilibrium). h˙ = 0, which implies h∗ = gEsH+δ Notice that this level of h∗ may be optimal for different levels of sH (share of output invested in local factor), each one corresponding to a different value of h∗ . In particular, for growing sH , the optimal level of h∗ grows, as shown in Fig. 2. In order to show the optimal levels of investment – split between openness and robustness –we have to consider that the total output produced allows an optimal level of consumption to be defined, which is the real variable to be maximised in equilibrium to define a wealthy TSPI. The dynamics of consumption in absolute terms is monotonic in domestic investment, the local factor of production: the more that is invested, the more output grows and the more consumption increases. Nevertheless, this occurs at a declining rate, because both inputs show decreasing marginal product. This means that consumption itself can be maximised by choosing from among the different equilibria that h∗ defines. Since output is spent in either consumption or investment in local factor (sH + c = 1) – where c is the propensity to consume – we can say that in equilibrium C∗ = (H ∗ )α − IH∗ , but in equilibrium investment is only made to replace depreciated human capital, then c∗ = (h∗ )α − (gE + δ )h∗ is the optimal level of equilibrium consumption in relative terms, which can be represented in graphical terms in Fig. 3. Steady State Investment (gE + d ) h
y
y shigh
slow
hlow
Fig. 2 Low and high equilibria
hhigh
h
36
A. Bramanti, M.R. Riggi
c*
⊕
h
h(s)*
Fig. 3 Optimal levels of consumption
Even at this stage an important implication can be drawn. This simple presentation of a model of exogenous growth is consistent with the findings of relational development. Optimal levels of equilibrium consumption stem from a balance between openness (Et ) and robustness (Ht ) or, in the model, human capital and innovation. Any deviation from h⊕ implies a reduction of the equilibrium level of consumption, so unbalanced openness and robustness patterns can provide equilibria, but not all of them are identical. Investments in local factor and innovation are both necessary in a balanced path for an optimal equilibrium (Golden Rule), such as that depicted in Fig. 3. In order to assess how much to invest in the two different factors, regional policy makers should consider the technological parameters of their production processes, here indicated by α and 1 − α . When a region shows higher α , this means that the region is able to gain more advantage from investment in local factor than from investment in innovation, and the optimal balance between Ht and Et is biased towards α . Policies aimed at promoting local robustness are then suitable; this does not mean that benefits from opening the TSPI to the outside must be neglected. In fact, policies devoted to ITHC can promote faster growth towards a steady state, but higher levels of consumption in the steady state imply a trade-off between openness and robustness. Optimal policies7 are identified if sH = α . The simple Solowian model presented, examines the main building blocks of sustainable interrelated growth in a neoclassical framework and, interestingly, the main result still holds good, giving significance to the topic of borders in territorial systems (how local is local development?), what are internal vs. external factors, what are strong vs. weak ties, what are short-range vs. long-range networks, and their implication on growth performances. This confirms that a sustainable growth pattern must balance dynamically the degree of internal robustness – the social ∗
C ∗ α −1 L1−α − (g + δ ) = 0, then α Y = δ H. But Proof: C=Y–I. By the Golden Rule ∂∂ H ∗ = α (H ) E we also know that in the steady state sH Y = δ H, then the steady state is compatible with the Golden Rule iff sH = α . 7
Sustainable Interrelated Growth: A Phenomenal Approach
37
embeddedness of the innovation process – with external openness to the rest of the world, the participation in global networks and long-range relations. Moreover, solutions biased towards internal or external investment, engender a crisis in the regional system in the long run. These two extreme cases can be called, after previous work on relational development (Bramanti and Miglierina 1995) death by entropy or disintegration of TSPI. In the first case – a system totally oriented inwards – entrepreneurs only follow established routines and imitate each other, weakening the system. In case of an external shock – implying a drop in the international price of traded goods – the consequence is a firm with no competitive tools in the short run, the only focus being on innovation whose horizon is long. In the second case – a system totally oriented outwards – delocalisation of important parts of the production value chain is chosen to exploit the opportunities provided by global markets. The end result is a weak case for the territory in attracting and rooting firms and the emergence of ‘isolated’ firms competing as ‘monads’ in the global market.
4 Human Capital and Robustness We have assumed that human capital is central in explaining growth patterns. In the current section, we will discuss its importance for regional development and then we will specify the general formulation for its dynamics expressed in the Golden Rule previously reported. In the twentieth century, a new paradigm of production emerged, characterised by a sharp shift from the relevance of physical elements to the relational dimension of the structure and the dynamics of the economic system (Castells 2000). Those relational aspects, as already discussed, constitute the basic components of regional development. It is currently recognised that disparity in productivity and growth has far less to do with the abundance of natural resources and much more to do with the ability to improve the quality of human capital and factors of production, to create new knowledge and good ideas and embed them in equipment and embody them in people (David and Foray 2002). This perspective stresses the fact that the most valuable assets are intangible investments (human, social and creative capital) and that knowledge, competence and creativity are key factors. The rise of the knowledge economy is due to information–, knowledge–, and skills–based activities playing an increasingly significant role in economic growth. Human capital is widely acknowledged as one of the main boosters of economic growth; furthermore, large differences exist within and between regions in terms of both quantity and quality of educational structure and institutions (Wossmann 2002). Accepting this view, we have stressed the role of human capital – both internally and externally trained – in the process of territorial development. In this respect, the formulation of the ‘Golden Rule’ presented above is quite general and, in particular, it does not distinguish between different ways of setting up human capital.
38
A. Bramanti, M.R. Riggi
Two broad families can be identified: formal human capital accumulation (education and accumulation of more general codified knowledge) and informal human capital accumulation (stemming from learning-by-doing processes or, broadly speaking, from tacit knowledge accumulation). Actually, these two ways coexist and human capital is generated through a mixed process of tacit and codified knowledge accumulation. Which features distinguish these two kinds of learning processes? Basically, codified knowledge can be exogenously transmitted to a region with no need of interaction, whereas personal interaction is peculiar for the process of learning to set up tacit knowledge. In the next sections the two forms of accumulation are presented and then a mix of the two is discussed. In order to operationalise the resulting three ways of accumulation and make them comparable, the law of motion of human capital accumulation will be made explicit in a slightly different way following to different extents, Mahajan and Peterson (1987), Maggioni, (1997), Maggioni and Riggi (2002), Riggi (2004), Maggioni and Nosvelli (2005).
4.1 Exogenous Accumulation Human capital is developed mainly through education. This may be represented by a mechanism of exogenous learning in which codified knowledge is spread across the region. Mahajan and Peterson (1987) identify the following specification to proxy this phenomenon: dN = a(K − N) dt Where N is the skilled population (the share of TSPI population endowed with human capital), whose rate of growth is driven by the parameter a and K represents the total amount of human capital that the region can support (carrying capacity). Let us say that the existing population can reach a maximum fixed level of education. In this dynamic, each actor only learns what he/she observes from the outside, regardless of interactions with other firms within the region that have already innovated. The parameter a describes the speed at which people learn and innovate through the process of technological windowing, external networking and weak ties. The external learning mechanism is positively influenced by (K–N), the distance between the maximum level of human capital and the degree already attained within the population.
4.2 Endogenous Accumulation If education is the main driver for human capital accumulation, it is far from explaining the underlying process in an exhaustive way. In fact, learning-by-doing (Arrow 1962), learning-by-interacting and any other form of interaction between
Sustainable Interrelated Growth: A Phenomenal Approach
39
workers or different actors (between firms, research centres and public institutions) – occurring in the workplace or at the regional level – are essential channels to build human capital. Internal connections and tacit knowledge are at the base of this process and appear in the formal model for innovation diffusion captured by a parameter b, across a population (N) of dimension K. dN = bN(K − N) dt The speed of the innovative process in this second version is then proportional to the quality and quantity of social, economic, technological and productive interactions that take place within a region between actors. Interactions between the population endowed with human capital and potential adopters stem from the product of bN and (K − N).
4.3 Comparing Exogenous and Endogenous Accumulation Although the previous mechanisms for human capital accumulation simplify the two main channels driving the process, both are actually relevant in interpreting the effectiveness of human capital as a production factor. We need, therefore, to include both kinds of accumulation processes. The simultaneous consideration of the two may give rise to two different features: first it could be argued that their combination can generate positive synergic effects in terms of the creation of new knowledge. On deeper analysis, however, it emerges that the two forms of accumulation echo the two crucial dimensions of territorial sustainable interrelated growth: • Endogenous accumulation is based on the generation of collective learning processes that flourish within a TSPI, where tacit knowledge, strong ties and social embeddedness can be indicative of original capabilities of the social and economic environment (citizens and not just workers are very familiar with the routines, norms and conventions on which the TSPI hinges); • The exogenous accumulation process is based on the external openness of TSPI and innovations are mainly grasped from the outside through a process of technological windowing, fostered by codified knowledge exchange and weak ties. The relative efficiency of the two processes can be analysed using a simulative approach after Maggioni and Riggi (2002). Depending on cumulative interactions and the number of actors involved, the evolution of human capital is better described by either form of learning being a Pareto superior situation. In analytical terms, the objective is to identify an accumulation process that, at each point in time, is more efficient in producing human capital at territorial level. In the next sections, we first simulate the case the two processes which are identical in terms of the parameter values (a and b), and then we relax this hypothesis.
40
A. Bramanti, M.R. Riggi
4.3.1 The Case of Identical Parameters Even in the case of identical parameters, the initial conditions in both cases need to be discussed. For null initial values, the endogenous model leads in fact to a meaningless case in which no human capital is produced; an initial critical mass is required for the process to show significant results. The exogenous process, instead, allows the accumulation process to start up, even where there is initially no human capital in the region. Apart from the special case just mentioned, identical TSPIs with different accumulation processes differ only in the adjustment process towards the ceiling K; no intersection between the two lines is possible (see Fig. 4). They have in common the initial value (N = 1) and the ceiling K towards which they both converge. The reasons for this dynamics appear when the slopes of the two lines representing the two accumulation processes8 are compared. In case of exogenous process dN dt = beso (K − N), whereas in the case of endogenous process the equation becomes dN dt = bendo N(K − N). It is clear that for each finite value of N < K, we have: N θendo < N θeso therefore, at each moment in time, the human capital accumulated in an exogenous way is always greater than the one accumulated in an endogenous way. That is, the exogenous accumulation process, in the case of identical parameters, is more efficient in accumulating human capital.
1: N1 1: 2:
2: N 2 50.00
1
1
2
1
2
1: 2:
25.00
2 1
1: 2:
2
0.00 0.00
7.50
15.00
22.50
30.00
Fig. 4 Paths of accumulation of human capital from exogenous (line 1) to endogenous (line 2) processes
8
With reference to the previous section here we indicate a = bexo and b = bendo .
Sustainable Interrelated Growth: A Phenomenal Approach
41
4.3.2 The Case of Different Parameters Let us now relax the initial hypothesis of identical parameters and verify the case in which the parameters assume more realistic values in terms of the system that we want to represent. The basic case consists of different speeds of diffusion of the two processes. In particular, we assume that the endogenous accumulation process is faster than the exogenous process (in the simulation illustrated in Fig. 5 beso = 0.3 < bendo = 0.9). This workable hypothesis accounts for the fact that we consider it easier to accumulate human capital through learning-by-doing and learning-by-interacting mechanisms rather than by formal education or the technological windowing process. Internal accumulation, in addition, on the other hand, consolidates existing routines and is more effective in terms of productivity. The dynamics of accumulation in the two forms shows a point in which the dominance of either process reverses: first the exogenous process prevails, then the endogenous process is Pareto superior since it relies on higher levels of human capital that make the interaction the best way to accumulate human capital.9 The point at which the endogenous human capital (Nendo ) reaches the exogenous human capital (Nexo ) can be expressed as: a−
1: 2:
bendo bendo 2 eso + a = Nt−1 a K
1: N eso 50.00
2: N endo 2 1
1
1: 2:
exo Nt−1
1
2
2
25.00
1
1: 2:
2
0.00 0.00
5.00
10.00
15.00
20.00
Fig. 5 Different propensities to accumulate human capital 9
For the sake of clarification, we decided to simulate an endogenous parameter three times the exogenous one, so that the dynamics are clearer. The results discussed above, however hold for bexo < bendo . It is also clear that the greater the gap, the quicker the endogenous process becomes preferable in terms of human capital accumulation.
42
A. Bramanti, M.R. Riggi
where a=
exo Nt−1 bexo K + 1 + bexo 1 + bexo
exo Thus, the higher the levels of exogenous human capital, the higher is the ratio bbendo exo – of which Nt−1 is an increasing non-linear function – and the higher is K. Summing up, the human capital is higher in the case of the endogenous process when the process reaches a considerable size. So, the endogenous accumulation process is more effective if there is a favourable initial situation and if the TSPI is large enough, otherwise accumulation through the exogenous process is more effective.
5 Conclusion The Solowian model presented here offers an interpretation of sustainable interrelated growth in terms of the openness-robustness equilibrium as the only way to pursue growth in the long run. Its sequential and linear structure, nevertheless, causes stationarity, leaving open the question of how innovation is generated within the system (the ‘black box’ of learning mechanisms at work within the TSPI). Extending the model, a process creating human capital is presented. This involves the acknowledgement of a new way of introducing innovation within the region, and this way has important consequences for the development patterns of the variables involved. The property of stationarity is lost, and the system is able to boost a positive dynamics internally. Anyway, even before the extension of the model, this process is not autarchic, because tight interactions are at work between internal and external forces. If the blending of internal synergy and external energy (robustness and openness; strong and weak ties; social and creative capital; milieu relations and a-spatial networks) has gained a definite place in the theory of territorial development, economic theory suggests that the equilibrium derives from a trade-off (alternative cost choices), and the present chapter has shown how to reach an ‘optimum’ within a simplified, elementary frame. As far as the alternative human capital accumulation processes examined in the present frame are concerned, it is not possible to identify a priori a ‘superior’ option, since it is contingent on the grounds of some empirical features (the value of the parameters capturing different structural features of the TSPI) that are not examined here. An interesting contribution in this direction is provided by Maggioni and Nosvelli (2005) who apply a similar framework to learning processes aimed at producing and diffusing innovations at regional level. In the case of industrial districts, for example, given the limited size of firms and their strong specialisation, endogenous and exogenous forms of learning (leading to different quality of human capital) are shown to be substitutes: territories benefit from endogenous factors, setting up a regional specific human capital that represents a competitive factor for the success of the TSPI.
Sustainable Interrelated Growth: A Phenomenal Approach
43
From a policy point of view, it is worth distinguishing policies directed to openness from robustness, which are frequently very difficult to implement, as they need targeted, integrated actions not only on the economic side but also on the structure of the socio-economic system as a whole. The non-complementarity between the different modes of human capital accumulation suggests that the selection and development of policies suitable to either process of accumulation become crucial for the policy maker. On a future research agenda, a further step will be a deeper examination of parameter gE (the exogenous dynamics of openness) which is certainly determined by the strong ‘white noise’ of the global relations outside the TSPI, but also coinfluenced by the internal process of collective learning and the specific presence of ‘gate actors’. The final goal is to capture at least some of the feedback on robustness and openness, up to the endogeneisation of the parameter.
References Arrow KJ (1962) The economic implications of learning by doing. Rev Econ Stud 80:155–173 Becattini G (2004) Industrial districts: a new approach to industrial change. Cheltenham, Edward Elgar Bramanti A (1999) From space to territory: relational development and territorial competitiveness. ´ Revue d’Economie R´egionale et Urbaine 3:634–657 Bramanti A (1991) Il modello dello sviluppo endogeno interrelato. Rivista economica del Mezzogiorno N. 2 Bramanti A, Fratesi U (2009) The dynamics of an innovation driven territorial system. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Bramanti A, Maggioni M (1997) La dinamica dei sistemi produttivi territoriali: teorie, tecniche, politiche. Franco Angeli, Milano Bramanti A, Miglierina C (1995) Alle radici della crescita regionale: fattori, fenomeni, agenti. L’Industria, a. XVI(1) Brusco S (2004) Industriamoci. Capacit`a di progetto e sviluppo locale. Donzelli Editore, Roma Camagni R (ed) (2001) Innovation networks: spatial perspectives Belhaven Press, London Castells M (2000) The rise of the network society. In the information age: economy, society and culture. Blackwell, Oxford Cohen MW, Levinthal DA (1990) Absorptive capacity: a new perspective on learning and innovation. Admin Sci Q 35:128–152 Corpataux J, Crevoisier O (2005) Increased capital mobility/liquidity and its repercussions at regional level. Eur Urban Reg Stud 4:315–334 Crounch C, Le Gal´es P, Trigilia C, Voelzkow H (eds) (2001) Local production systems in Europe: rise or demise. Oxford University Press, Oxford David P, Foray D (2002) An introduction to the economy of the knowledge society. Int Soc Sci J Florida R (2002) The rise of the creative class. Basic Books, New York Gambardella A, Mariani M, Torrisi S (2006) How provincial is your region? Openness and regional performance in Europe. WP, Bocconi University Granovetter M (1973) The strength of weak ties. Am J Sociol 6:1360–1380 Maggioni MA, Bramanti A (2002) Local and global networks in the economics of SMEs. Is proximity the only thing that matters? In: McNaughton RB, Green MB (eds) Global competition and local networks. Ashgate, Aldershot, pp 247–277
44
A. Bramanti, M.R. Riggi
Maggioni MA, Nosvelli M (2005) L’apprendimento collettivo tra saperi locali e reti globali. In: Bruzzo A, Occelli S (eds) Le relazioni tra conoscenza ed innovazione nello sviluppo dei territori. FrancoAngeli, Milano Maggioni MA, Riggi M (2002) Forme alternative di collective learning: un approccio sistemicopopolazionista ed alcune simulazioni. In: Camagni R, Capello R (eds) Apprendimento collettivo e competitivit`a territoriale. FrancoAngeli, Milano, pp 109–135 Mahajan V, Peterson RA (1985) Models for innovation diffusion. Sage University Press, Beverly Hills Maillat D, Perrin JC (eds) (1992) Entreprises innovatrices et d´eveloppement territorial. GREMIEdes, Neuchatel Markusen A (1996) Sticky places in slippery space: a typology of industrial districts. Econ Geogr 3:293–313 Ratti R, Bramanti A, Gordon R (eds) (1997) The dynamics of innovative regions. The GREMI approach, Ashgate, Aldershot Riggi M (2004) Labour market dynamics and the evolution of industrial clusters: towards a microfoundation of the ecological approach. PhD thesis, Universit`a di Bologna Rogers EN (1995) Diffusion of innovations. The Free Press, New York Scott AJ (1988) New industrial spaces: flexible production and regional development in North America and Western Europe. Pion, London Solow R (1956) A contribution to the theory of economic growth. Q J Econ Storper M, Venables AJ (2005) Buzz: Face-to-face contact and urban economy. In: Breschi S, Malerba F, (eds) Clusters, networks, and innovation. Oxford University Press, Oxford, pp 319– 342 Todorova G, Durisin B (2007) Absorptive capacity: valuing a reconceptualization. Acad Manage Rev 3:774–786 Wossmann L (2002) Schooling and the quality of human capital. Springer, Berlin Zahra SA, George G (2002) Absorptive capacity: a review, reconceptualization, and extension. Acad Manage Rev 27:185–203 Zimmermann J-B (1998) La prossimit`a nelle relazioni imprese-territori: nomadismo e ancoraggio territoriale. L’Industria 3:613–632
A Model of Local Development Giuseppe Folloni
1 Introduction The paper presents a model which describes the growth dynamics of a local system in which there are two populations of firms. The first population consists of the local firms which respond to demand generated within the system by final consumers or by the demand from local exporter firms for intermediate goods. The second population consists of leader firms able to compete on external (foreign) markets and which are the source of opportunities for the local system. The model’s basic structure has Keynesian features because the implicit exogenous variable on which the model’s dynamics depends is external demand to which the system is able to respond. In this sense the model adopts an approach a` la DixonThirlwall (1975). The difference is that in the Dixon-Thirlwall model the exogenous lever, i.e. the growth of income worldwide, is explicit. In our model, which relates to a small region, external exogenous demand is present (there is no possibility of surplus supply) but it is not explicitly modelled. Rather, it is internal conditions – the competitiveness that the system is able to achieve – which determine the system’s ability to meet such demand. In this regard, the model resembles another classic approach to regional analysis: the Economic Base model (Tiebout 1956a, b; Williamson 1975), in which regional growth is driven by specialization responding to external demand, by local subcontracted activities induced by that specialization and by the presence of firms serving the local population. Unlike these models, however, the one presented here emphasises supply-side factors (the size of the different groups of firms, their dimension, their ‘quality’ in terms of entrepreneurial capital). In the literature, the productive vitality of a local system arises from the internal division of labour among firms (as reported by studies on industrial districts) and from innovative capacity (Camagni 1991a, b; Ratti et al. 1997). The first of these dimensions (coherent division of labour) entail that firms have a certain technological capacity deriving from history, repeated cooperation, favourable local U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
45
46
G. Folloni
institutions, a shared culture and tacit knowledge present in the local system: all of these factors, together with physical proximity, combine to reduce specialization transaction costs. Secondly, the system’s innovative capacity – the increased knowledge incorporated into agents’ behaviour so that they are able to maintain and increase their competitiveness – springs from the same factors. Mutual understanding, trust and reciprocity – that is, the set of untraded interdependencies described by Storper (1995) – are even more important in guaranteeing the system’s innovative capacity than in simply determining a vertical division of labour and inter-firm specialization. Our model incorporates innovative capacity into the presence of exporter firms, or leader firms. The population of such firms thus comprises two elements: firstly these firms permit the generation of local interdependencies (traded and untraded) which strengthen the local system of firms (also non-leader) in terms of productive capacity, human capital, the ability to exploit spillovers, etc. On the other hand stands the ability of leader firms to obtain, through competition on markets, the information necessary to maintain and increase that capacity. The model therefore explicitly recalls the Bramanti-Miglierina model (1995)1 in that it also models the two populations of firms (leaders and non-leaders) and the dynamic interdependency between them as a decisive factor in the system’s capacity for growth. Figure 1 is a graphic representation of the Bramanti-Miglierina model.
E firms population
Competitiveness
L firms population
Per capita Income
Fig. 1 Graphic description of Bramanti-Miglierina model
1
See also Bramanti 1992.
A Model of Local Development
47
The purpose of the 1995 Bramanti-Miglierina model was to show that an imbalance between the two populations gave rise either to a low-level equilibrium, in the absence of leader firms able to capture innovative and competitive opportunities for the system, or to disintegration in the absence of induced subcontracted and outsourced activities able to contribute to the system’s competitiveness. In the latter case, the leader firms would have looked elsewhere for the sub-suppliers and services that they needed and would finally emigrate from the system. In the end, therefore, the system lapsed into the first of the two states described. This study intends to model the fact that the absence of a population of competitive firms (absence of leader firms) generates low levels of income because there is no stimulus to the local system to incorporate new knowledge in order to ‘follow’ the innovative dynamics of the leader firms. The hypothesis that competitive firms abandon the system because of a lack of induced goods and services activities is not considered – and our model differs from Bramanti-Miglierina 1995 in this –. We emphasise instead that features of the local system not favourable to competitiveness and innovation impede the formation of leader firms. The absence of the latter hampers the specialization and innovation of local induced activities, and this blocks growth. The approach taken by this study also differs from that by Bramanti and Fratesi 2009, in this book, who distinguish three populations of firms (leaders, followers, and cooperative sub-contractors) resulting from interactions between the system and external knowledge and R&D circuits. The Bramanti-Fratesi model has LotkaVolterra features in that one of the three populations – that of the ‘followers’ – ‘preys on’ the innovation and knowledge opportunities present in the system, thus weakening the system of leader firms and inducing sub-optimal R&D investment decisions. The differences with respect to the model presented here are obvious. Our model does not comprise a population of followers which free ride on knowledge and innovation producers. Moreover, the competitive capacity that drives growth is here tied to the leader firms – both directly and indirectly – because the leader firms induce the local production system to follow dynamics of specialization, innovation and new knowledge acquisition appropriate for growth.
2 Local Identity The model describes the evolution of two populations of firms. The first population consists of the set of local activities that respond to demand generated within the system by final consumers and by the leader firms which use local firms as subsuppliers. The set of local firms will be denoted by F. The second population consists of leader firms able to export and to give competitiveness to the system. The set of exporting leader firms will be denoted by E. The ratio between the two populations is an indicator of the system’s ability to grasp opportunities that arise on global markets. A too low ratio (absence of leaders)
48
G. Folloni
indicates purely local forms of agglomeration. A too high ratio indicates the presence of competitive and isolated large firms without an adequate local productive system to support them. A local system in which the competitive enterprises are large firms with few connections with the rest of the territory, and in which the local population of activities is simply an aggregate of non-base activities is only marginally considered here. A local productive system’s ability to grow depends on the co-presence of the two populations and on their qualitative characteristics. These characteristics and the interrelations between the two populations define what we call ‘local identity’. 1. Local identity is defined primarily by the type and quality of social, human and entrepreneurial capital in the area; it is the fundamental factor in the development of the specialization, division of labour and inter-firm networks that enhance the system’s productivity. The expression ‘social, human and entrepreneurial capital’ may denote very different phenomena. (i) There are systems in which economic activity is essentially a response to local demand (non-base activities in a typical Economic Base model). The micro enterprises in many urban areas of the developing countries, without a division of labour, specialization or productive cooperation internal to the area, with low human capital and equally low endowments of capital and technological knowledge, constitute a system of this type. Agglomerates of firms like these do not have normally propulsive capacity and follow the trend in demand; they have high birth and death rates, and record tendentially nil average long-period growth rates. (ii) There are systems in which the quality and level of human and entrepreneurial capital are higher and constitute the ‘local network of relations and sub-supply’ of competitive firms, the leaders, as in a classic industrial district. (iii) There are then systems in which the information generated by the presence of leader exporting firms feeds back into the system, enhancing its ability to monitor opportunities, foresee challenges, and innovate in order to respond to them, as in an innovative milieu. 2. The other dimension of local identity is the capacity to compete. Competitiveness captures the area’s comparative advantages a` la Porter and depends on the existence of leader firms able to translate the information available on markets into opportunities to innovate and export, and therefore into growth. The competitiveness thus generated will be denoted in the model with C, and it obviously depends on the presence and importance of leader firms. Local identity (the E and F populations of firms, in number and quality) gives rise to competitiveness. The system’s growth of output (Y in total, y in per-capita terms) depends on the level of competitiveness.
A Model of Local Development
49
3 The Model 3.1 Assumptions The hypotheses are the following. ¯ so that growth in the 1. The area’s population, denoted by P, is constant (P = P), two populations of firms produces a variation in the intensity of the production network with respect to the population. Moreover, the dynamics of the local system’s total output, Y, and that of per-capita output, y, are exactly proportional. There is no public sector, so that the area’s output and income coincide. 2. The specializations that the local system specifically assumes are not modelled: output is aggregate and there is no distinction between intermediate and final production. 3. The model does not explicitly consider either exports or imports. Both these aggregates are defined indirectly. In fact, the growth of the leader firms coincides with an increase in competitiveness on international markets, and therefore with the capacity to export. 4. Imports are implicitly defined by the fact that demand (measured by the system’s total income which coincides with output, Y) is not entirely matched by supply from local firms. 5. The expenditure, in local output and imports, represents primarily consumption by the population. However, one may consider that a proportion of this income saved and transferred to firms, represents investments made by households in those firms. However, this aspect is not highlighted in this simple basic version of the model. 6. The level of competitiveness is assumed to depend on the presence of leader firms (C = f (E)). 7. The level of per capita income at time t+1 depends on competitiveness of the area at time t (yt+1 = f (Ct )). 8. There are diminishing returns both in the capacity of leader firms to acquire competitiveness for the system, and in the relation between higher competitiveness and the level of income (the elasticities of competitiveness to the number of leaders and of income to competitiveness are less than 1). 9. The unit size of local firms is also fixed and is denoted by δ . 10. Consequently, the total number of local firms, substantially parametered with the need to serve the local system, depends on the area’s income and on the size of each firm: F = f (y, P, δ ). 11. The ratio between the two populations of firms, the parameter σ , is an indicator of the density of leaders in the system and measures therefore the system’s vitality. Moreover, from what we said about the function of leaders in generating local interdependencies, it is also an indicator of the network between these leaders and local firms.
50
G. Folloni
The model can therefore be described by the system set out in [1]. ⎧ Ct = Et λ (0 < λ < 1) ⎪ ⎪ ⎪ ⎨y = C α = E αλ (0 < α < 1) t t t+1 θ ⎪ δ Ft = Yt (0 < θ < 1) ⎪ ⎪ ⎩ Et = σ Ft (0 < σ < 1)
(1)
with Yt = P¯ yt (where yt is per capita income), so that: P¯ θ θ y = ω ytθ , δ t
Ft =
(ω > 1).
The similarity with the Bramanti-Miglierina model in Fig. 1 is evident.
3.2 A Particular Solution In the particular case where λ = α = 1 (constant marginal returns) and θ = 1 (a system with no imports), with appropriate substitutions, system [1] reduces to the following equation: yt+1 − σ ω yt = 0 whose solution is: yt = (σ ω )t y0 → gy = σ ω − 1
(2)
(with gy constant across time). In other words, the growth rate, gy , is positive if
σ ω > 1, i.e. σ
P¯ >1 δ
(2a)
and negative in the opposite case. In this particular case, in the steady state per capita income tends to zero if gy is negative, and goes to infinity if gy is positive. Moreover, equation [2a] coincides with the following: Et = Ct > yt
(2b)
that is, the numerousness of leader firms, and therefore their capacity to give competitiveness to the system, must exceed the level reached by per capita income if it is to induce further growth in the system. ¯ Condition [2b] consists of two parameters, σ eω where ω = δP . The second of the two parameters, which compares the population of the system to the dimension of local firms, is an indicator of the system’s productive potential, and therefore of the possible forms of labour division. There are two limiting cases.
A Model of Local Development
51
If δ is very low, and P is relatively high, we have the case already-described of urban systems made up of micro enterprises. It is likely, as happens in many such cases, that parameter σ will be very low and tend to zero. These are local systems unable to express competitive leadership on broader markets. The present model does not endogenize the relation between firms’ size, δ , and presence of leaders, σ . However, it is plausible that this relation exists and is positive: as the size of the firms making up the local productive system increases, so does the density of leader firms in it, at least within a certain range. A system of micro enterprises or of self-employed workers will find it difficult to create both local leaders and local networks of efficient and innovative firms. When both δ and P are low, we have small systems of micro enterprises, which are not commented upon in this analysis. If, instead, δ is very high, we have a system of large isolated enterprises (where σ is also probably very high) unable to find locally the network of sub-suppliers that they need, and forced to look elsewhere for it, or a system of firms acting as growth poles in the Perroux/Chenery sense (if σ is not so high) that could create their own local network of suppliers.
3.2.1 The General Case The general case is that in which: λ , α , θ , σ < 1 and ω > 1. Brief discussion is required on the meanings of the various parameters. We begin with parameters λ , α , θ . The other parameters, σ and ω , will be analysed below. From [1] it is possible to show that:
λ=
gyt gCt gFt ,α = ,θ = gEt gCt gyt
The fact that λ is positive and <1 implies that the growth rate of competitiveness is less than that of the population of leader firms. One may accordingly talk of decreasing returns in the relationship between leaders and competitiveness. The fact that α is positive and <1 implies that the growth rate of per capita income (which coincides with that of total income, on the hypothesis of a constant population) is less than the growth rate of the system’s competitiveness (diminishing returns on competitiveness). Finally, 0 < θ < 1 implies that the proportion of income, i.e. demand, matched by local supply, equal to Ytθ −1 , diminishes over time if gyt is positive. In the general case, with substitutions system [1] yields [3]: yt+1 = Etλ α = (σ ω )λ α ytθ λ α
(3)
If we redefine: ln(yt ) = zt , θ λ α = a, λ α ln(σ ω ) = c, we obtain: zt+1 = c + azt
(3b)
52
G. Folloni
whose solution is: c c at + zt = z0 − 1−a 1−a
Substituting zt with its definition and deriving with respect to t: d ln(yt ) c = ln(y0 ) − at ln(a) gyt = dt 1−a
(4)
(5)
And substituting a and c with their definitions in terms of the model’s parameters: ⎧ 1−θ λ α ⎪ P¯ θ λα ⎪ g > 0 i f σ ω = σ > y yt ⎪ 0 δ ⎨ gyt = 0 ⎪ ⎪ ⎪ ⎩ gyt < 0
¯θ
1−θ λ α λα
P¯ θ
1−θ λ α λα
i f σ ω = σ Pδ = y0 i f σω = σ
δ
(6)
< y0
3.3 Steady State In the long run, the rate of growth of output tends to zero for any value of σ and ω , given the hypothesis of decreasing returns. In fact, from [5]: c a∞ ln(a) = 0 (a < 1) (7) lim gyt = ln(y0 ) − t→∞ 1−a It is therefore possible to calculate the steady-state value of per capita income from [4]: λα c ⇒ y∞ = (σ ω ) 1−θ λ α ln(y∞ ) = (8) 1−a In other words, the growth rate is positive if the initial per capita income is less than the steady-state income. In fact, it follows from [6] that: ⎧ ⎪ ⎨gyt > 0 i f y∞ > y0 (9) gyt = 0 i f y∞ = y0 ⎪ ⎩ gyt < 0 i f y∞ < y0. Per capita output will therefore grow towards the steady state in the first case; it will be constant in the second; and it will diminish towards the steady-state level in the third case. Figure 2 shows the dynamics of per capita output in the three cases described. The decisive parameters for establishing whether growth will be positive or negative are σ and ω . These require more detailed analysis.
A Model of Local Development
53
100
10
1
t
Fig. 2 Dynamics of per capita income in the three cases (logarithmic scale), with the following values of the parameters: α = 0.9, λ = 0.9, θ = 0.8, δ = 10, P = 2000 and σ = 0.1/0.04/0.03 in the three cases Pθ
δ
Growth area
Decline area
σ
Fig. 3 The regions of growth and decline
Growth is positive if:
σω = σ
1−θ λ α Pθ > y0 λ α . δ
(10) 1−θ λ α
y λα
From [10] follows that the steady-state function ω (σ ) = 0 σ defines the zero growth line and marks out two regions: one of growth, and one of decline (see Fig. 3).
54
G. Folloni
By analysing the two critical parameters σ and δ (under the hypothesis of fixed population), it is therefore possible to insert the different types of local production systems into the two regions defined – that of growth, and that of decline. Parameter σ , the ratio between leader firms and the number of local firms, is an indicator of the system’s vitality and of the quality of the local firms’ system: this is able to generate leaders and the presence of the latter enhance the robustness of the local network of firms supporting them.
3.4 Very Low σ As said, the distinctive feature of many areas in the developing countries is the absence of firms able to compete on markets (very low σ ). The presence of a population of local firms, which may even be large (because also δ is low with respect to the population), represents a fragmented set of micro enterprises whose aim is to survive, not to achieve market capacity. In Fig. 4, which shows the various types of local production system, the aggregations of micro enterprises of the type described are located in the area at high risk of decline, either absolute (negative rates of income growth) or at least relatively to innovative areas with a strong growth capacity.
Pθ
δ
CIuster of micro firms Systems of survival Microenterprises
Protodistricts
Industrial districts • dynamic
Systems of large firms
• in decline
σ* Fig. 4 Different types of local systems
σ
A Model of Local Development
55
3.5 Low/Moderate σ A large body of literature (Rabellotti 1995, 1999; Humphrey and Schmitz 1998; Schmitz 1999; Visser 1997) emphasises that many developing economies show areas (especially urban or suburban) where clusters of small firms are beginning to assume the features of actual networks, with an incipient division of labour among firms, the growth of specializations of a certain importance, and an increased capacity to export. These are the local systems indicated in Fig. 4 as ‘clusters of firms’ and ‘protodistricts’. As regards the former (clusters of firms), there are still relatively few leader firms (low level of σ ), the great majority of firms are of very little size (low δ ) and the capacity to maintain dynamic competitiveness suffers, as a consequence. Strong foreign competition may stunt these incipient forms of clustering in the area of decline (for example, an invasion by low-cost competitive products will reduce parameter θ and operational spaces within the system). By ‘proto-districts’ are meant local systems in which entrepreneurial capital and innovative knowledge are considerably more robust and may bring about – in an appropriate policy context – transition to a genuine district with the typical characteristics described in the literature.
3.6 Moderate σ In typical Marshallian industrial districts, a higher σ indicates a balance between local firms and leader firms where the two populations are connected by the forms of specialization and cooperation that constitute the typical strengths of this organizational form. Figure 4 also shows that Marshallian districts may lose their propulsive force either because local production network disintegrates due to relocation (with a lowering of the system’s position on the vertical axis, i.e. a decrease in the value ¯θ of ω = Pδ , which measures the system’s size and indicates the potential number of firms in the local production system) or because of a decrease in the number of leader firms (this too due to relocation to more central markets, with a fall in the level of σ ).
3.7 High σ In a system of large firms (very high σ ), the system consists of exporter firms without a local system of subcontracted and outsourced activities. At the limit (σ = 1), we have a system of pure leader firms of large size (high δ ) without local supporting activities.
56
G. Folloni
4 Non-concluding Remarks A recurrent feature evinced by the foregoing analysis has been the existence of functional relations between certain parameters of the model, and this raises the problem of endogeneity. It has been noted, in fact, that a very low δ is associated with a strong likelihood that the value of σ will be low as well: under-sized firms or self-employed workers are often unable to emerge as leaders with their own networks of subcontracted and outsourced activities. It is also likely that, in these conditions, the returns from the capacity to export on competitiveness, and from the latter on income, will be lower (and the hypothesis of decreasing returns – low λ and α – is more stringent). Finally, low values of these parameters increases the system’s vulnerability to ‘invasions’ by products from other areas (decrease in the θ parameter, and a consequent diminution in the system’s size). These are therefore forms of endogeneity of the parameters which the simple basic model presented above is unable to capture. Further development of the model should consider at least the most important of these functional relations. Secondly, the intention here is not to suggest policy directions, given that the purpose of the analysis has been to furnish a simple model illustrating the original idea behind the Bramanti/Miglierina model. However, it is clear that both the model and the functional relations among its parameters evidence the ‘low level equilibrium trap’ which characterized many areas of small and medium-sized firms in the developing countries. The decisive policy instrument would therefore be one which fosters the growth of leader firms able to consolidate the overall system through dynamic knowledge spillovers and by inducing opportunities throughout the local production system. A typical Industrial District of an advanced economy can also experience a decline towards a low level equilibrium (as shown in Fig. 4). Moreover, the described forms of interdependence between the parameters can further accentuate this decline, in a cumulative way. Consequently only positive “shocks” that let the value of some fundamental parameter grow, will permit the system to cross the zero-growth boundary and become, again, vital.
References Andrews RB (1953) Mechanics of the urban economic base: historical development of the base concept. Land Econ 29:161–167 Archer BH (1976) The anatomy of a multiplier. Reg Stud 10:71–77 Arrow KJ (1962) The economic implications of learning by doing. Rev Econ stud 29:155–173 Blumenfeld H (1955) The economic base of a community. J Am Inst Planners 21:114–132 Boltho A, Holtham G 1992 The assessment: new approaches to economic growth. Oxf Rev Econ Policy 8:1–14
A Model of Local Development
57
Bramanti A (1991) Il modello dello sviluppo endogeno interrelato. Rivista economica del Mezzogiornon. 2 Bramanti A, Fratesi U (2009) The dynamics of an innovation driven territorial system, In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Bramanti A, Miglierina C (1995) Alle radici della crescita regionale: fattori, fenimeni, agenti. L’Industria 17:5–31 Camagni R (1989) Cambiamento tecnologico, milieu locale e reti di imprese: verso una teoria dinamica dello spazio. Economia e politica industriale no 64 Camagni R (1991a) (ed.) Innovation networks: spatial perspectives. Belhaven, London Camagni R (1991b) Local milieu, uncertainty and innovation networks: towards a new dynamic theory of economic space. In: Camagni R (ed.), Innovation Netwoks. Belhaven, London, pp 121–144 Chatterji M (1992) Convergence clubs and endogenous growth. Oxf Rev Econ Policy 8:57–69 Cheshire PC, Malecki EJ (2004) Growth, development and innovation: a look backward and forward. Papers Reg Sci 83:249–267 Dei Ottati G (1992) Fiducia, transazioni intrecciate e credito nel distretto industriale. Note Economiche 23(1–2):1–30 Dixon R, Thirlwall AP (1975) A Model of Regional Growth-Rate Differences on Kaldorian Lines. Oxf Econ Papers27(2):201–214 Esposito GF, Labia M (1993) I distretti industriali tra conservazione e mutamento strategico. Studi e informazionia. 15(1) Folloni G, Maggioni M (1994) Un modello dinamico di crescita regionale: leader e attivit`a indotte. Quaderno Ricerca di Base N. 3/94. Milano, Universit`a Bocconi Humphrey J, Schmitz H (1998) Trust and inter-firm relations in developing and transitino economies. J Dev Stud 34(4):32–61 Martin RC, Miley HW Jr (1983) “The stability of economic base multipliers: some empirical evidence. Rev Reg Stud 13:18–27 Maskell P et al. (1998) Competitiveness, localised learning and regional development. Routledge, London Onida F, Viesti G, Falzoni AM (ed.) (1992) I distretti industriali: crisi o evoluzione? E G E A, Milano Rabellotti R (1995) Is there an ‘Industrial District’ model? Footwear districts in Italy and Mexico compared. World Dev 23(1):29–41 Rabellotti R (1999) Recovery of a Mexican cluster: devaluation bonanza or collective efficiency? World Dev 27(9):1571–1586 Ratti R, Bramanti A, Gordon R (1997) The dynamics of innovative regions. The GREMI approach. Ashgate, Aldershot Romer P (1990) Endogenous technological change. J Polit Econ 98:71–102 Rullani E (1992) Sviluppo industriale: nuove realt`a e nuove opportunit`a per gli anni ‘90. Commento, in Tolomelli C (ed.), op. cit Schmitz H (1999) Global competition and local cooperation: success and failure in the Sinos Valley, Brazil. World Dev 27(9):1627–1650 Scott M (1992) “A new theory of endogenous economic growth”, Oxf Rev Econ policy 8(4):29–42 Storper M (1992) I processi di riorganizzazione produttiva dei sistemi industriali negli anni ‘90, In: Tolomelli C (ed.) Le politiche industriali regionali. Esperienze, soggetti, modelli. CLUEB, Bologna Storper M (1995) The resurgence of regional economics, ten years later: the region as a nexus of untraded interdependencies. Eur Urban Reg Stud 2(3):191–221 Tiebout X, Charles M (1956a) Exports and regional growth. J Polit Econ 64:160–164 Tiebout X, Charles M (1956b) The urban economic base reconsidered. Land Econ 31:95–99 Tolomelli C (1992) Le politiche industriali regionali. Esperienze, soggetti, modelli. CLUEB, Bologna
58
G. Folloni
Visser E (1997) The significance of spatial clustering: external economies in the peruvian smallscale clothing industry. In: Van Dijk, Rabellotti (eds.) Enterprise clusters and networks in developing countries. EADI Book Series 20, Frank Cass London, pp 61–92 Williamson X, Robert B (1975) Regional growth: predictive power of the export base theory Growth and Change 6:3–10
The Dynamics of an ‘Innovation Driven’ Territorial System Alberto Bramanti and Ugo Fratesi
“There is nothing more difficult to plan, more doubtful to success, nor more dangerous to manage than the creation of a new order of things. . . Whenever the enemies have the ability to attack the innovator they do so with the passion of partisans, while the others defend him sluggishly. So that the innovator and his party alike are vulnerable.” Niccol`o Machiavelli, The Prince
1 Introduction: Innovation and Local Competitiveness The capacity to generate and implement advances in technology or information technology – and even better in both – is increasingly regarded as the single most important force driving economic growth. This is true on a macro global scale (large competing economies) as well as at a ‘meso’ level, i.e. the level of a territorial system of production and innovation (TSPI).1 There are a number of approaches today – the Italian district school (Becattini 1990, 2004), the Californian school of new industrial geography (Scott 1988; Storper 1995), the European GREMI school of milieux innovators (Camagni 1991; Ratti et al. 1997), and the French school of proximity (Rallet 1993; Rallet and Torre 1995) – which show how the complex interaction of agent behaviour, increasing returns and learning processes yield spatially differentiated performance from territories which become winners or losers in the new competitive environment. 1 This is a multi-faceted concept deeply rooted in regional analysis. Territorial Production Systems (TPS) have been widely used by a number of researchers and in a dynamic view it brings together a ‘production system’, a ‘technical culture’ and a set of actors (enterprises, professional associations, local authorities, universities and laboratories, individuals). The TSPI adds the concept of innovation to the previous TPS, and there are two main mechanisms at work: the first is the creation of innovation and the second is the diffusion of innovation, which is always a social type of communication (see Bramanti and Riggi, 2009, in this book).
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
59
60
A. Bramanti, U. Fratesi
These approaches merge into various issues centred on the notion of territory as (see Bramanti et al. 1997, p. 4): (a) “the birthplace of technology and innovation — i.e. the progress from given resource allocation processes to a collective build-up of specific resources (Gaffard 1990)” (b) “a place for co-ordinating industrial activities, a link between external territorial economies and organisational and inter-organisational firm trajectories (Veltz 1993)” (c) “a political decision-making unit governing localisation, able to create and redistribute resources, and expressing specific governance structures in the relations between actors (Storper and Harrison 1991)” (d) “a place in which untraded inter-dependencies (means through which the actors grow technologically and organisationally, and co-ordinate themselves) form, express themselves, and evolve (Storper 1995)” The present work focuses on the dynamics of a TSPI supported by two complementary circuits: one internal to the firm and its closest environment where production is rooted (milieu effect or TSPI ‘robustness’); the other external, arising from transnational networks of production, science and technology (global effect or TSPI ‘openness’). The primary cause and driver of growth within the TSPI is innovation defined as an interactive, collective process stemming from a creative combination of generic know-how and specific competencies (Dosi 1988; Freeman and Soete 1997). The two main ‘building blocks’ of the model developed are, therefore, local agents – addressed as three different populations of firms (leader, follower and cooperative sub-contractors) – and R&D/knowledge circuit. The passage from external knowledge plus R&D to innovation and finally to actual economic performance is, however, not always a smooth and easy process because getting a new idea adopted is often very difficult. While some areas outperform their investment in R&D in economic terms, other territories find it difficult to make successful links between technology and growth. To evolve positively, a TSPI should be able to balance dynamically the degree of internal strength within its web (internal synergy) with that of its opening to the world (external energy) and any ‘unbalanced’ solution leads to a dissolution of the TSPI as such: alternatively to ‘death from entropy’ or to a disintegration of the system, along with the elimination of the spatial effects of proximity (Bramanti 1999; see also the Introduction to the volume). The paper will thus investigate the dynamics of a TSPI using a system dynamics simulation model which will represent systematically the circuit from external knowledge, to proprietary and diffused local knowledge, to innovation and regional competitiveness. This process flows through the interactions and articulations of the main groups of local agents. With respect to this, our model identifies three main types of firms, the first two of which were already described in the previous chapter by Folloni (2009): leaders, the only ones able to innovate, which bridge external end internal knowledge and compete in the external market; and co-operative
The Dynamics of an ‘Innovation Driven’ Territorial System
61
sub-contractors, which don’t compete externally but co-operate with the leaders in production and innovation. Finally, we have a third group of firms which we called followers, which only exploit the local opportunities without contributing to the renewal of the local knowledge fabric. We added this third group in order to study the pre-conditions and the effects of local opportunistic behaviour. With the model (which is calibrated on standardised values due to the unavailability of actual data) our investigation will focus on the analysis of system performance as a function of a number of parameters characterising the TSPI which are mainly: R&D infrastructure, openness, local spillovers, and milieu effects. We will show that for some parameters, at least openness and local spillovers, there is a threshold beyond which their positive effects are reversed and become negative, so that there is scope for policies to achieve the best values. The rest of the chapter is organised as follows: in Sect. 2 the path from external knowledge to R&D to productive activity and competitiveness will be analysed in the literature. In Sect. 3, the main characteristics of a ‘system dynamics’ approach will be presented. Section 4 will set the scene by explaining our selection of which interactions between local agents to include. Sections 5–8 will complete the set up of the model and illustrate the dynamics within a TSPI. Section 9 will simulate alternative growth patterns for different values of the parameter in search of threshold effects. Section 10 will use simulations to show the evolution of the TSPI subjected to some stress. Section 11 concludes with our policy remarks and some directions for future research.
2 Transforming R&D into Productive Activity Once R&D related innovations are developed, research spillovers begin to operate giving access to a broader range/larger number of firms (and most of all within a relationally dense TSPI). In addition, there is a significant risk of free-riding – and this is the case with the ‘predator’ archetype of follower firm in our model – so why should an individual firm spend time and effort on R&D when similar results could be achieved by poaching the technological innovation of others? A number of works have shown that an innovation is likely to be applied first in the firm or within the area where the innovation is achieved and that private innovation benefits from economies of agglomeration and of scale, so enhancing the total investment in R&D (private plus public), which is likely to increase the local capacity to innovate which, in turn, will lead to better economic performance. Among the various causes, local social conditions act as a social filter which determine the pace at which any society adopts innovation and transforms it into real economic activity (see Crescenzi and Rodr´ıguez-Pose 2009, in this volume). Innovation, communication channels, time and the social system are the four main elements in every diffusion process (Rogers 1995). Beyond R&D, knowledge diffusion and accumulation is the determining factor for TPSI development raising the productivity of the other local inputs such
62
A. Bramanti, U. Fratesi
as labour, physical capital and human capital. Given the impact of knowledge on the total factor productivity, knowledge diffusion acts as a direct scale effect with respect to territorial productivity. As in any market providing ‘public goods’, in the presence of spillovers and free-riding, there exists a range of optimal knowledge diffusion. Too much diffusion may weaken a firm’s R&D efforts, while too little knowledge diffusion may prevent the accumulation of ‘tacit’ knowledge. This tradeoff is not fixed but is a dynamic equilibrium in the interplay between innovators and imitators. From R&D to innovation, and from innovation to diffusion and adoption, the next step, not surprisingly, is feedback of learning mechanisms reinforcing innovation at work within TSPI. But in order to appreciate these, and specifically collective learning (Capello 1999), we should consider ‘tacit knowledge’ due to the frequent claim that: “knowledge transmission is mostly a matter of face-to-face contacts and labour mobility, i.e., that the most important knowledge carriers are people, in particular people who know and possibly trust each other, meet frequently, and trade job offers very often.” (Breschi and Lissoni 2001, p. 256). Rosenberg (1982, p. 143) encapsulated the essence of tacit knowledge as “the knowledge of techniques, methods and designs that work in certain ways and with certain consequences, even when one cannot explain exactly why”. To the extent that “all knowledge is either tacit or rooted in tacit knowledge” (Polanyi 1969, p. 144) it is worth distinguishing between skills and tacit knowledge. The first implies being able to make something happen, it involves cognition but also other aspects such as manual dexterity or sensory ability. Knowledge implies understanding, it is a perceptual, cognitive process. Tacit knowledge can be acquired by working closely with experts, learning through example and experience; this is what normally happens in TSPI, particularly within the labour market for unskilled and skilled workers. Imai and Baba (1991) suggest that personal interaction is necessary for information exchange between design, test, production and distribution, and this chain, largely based on firm-specific accumulated technological knowledge, is the most common in innovation (Pavitt 1987, 1999). This is even more true in the new areas of science, where personal interaction is required to access the tacit knowledge associated with the originating scientists.2 This is not to deny the rising role of codified knowledge in nurturing learning mechanisms as stressed in recent development in the economics of knowledge codification (Cowan et al. 2000; Steinmueller 2000) but to recognise the full complementarity of both dimensions and the importance of face-to-face contacts, which have four major properties (Storper and Venables 2004, p. 353): “[Face-to-face contact] is an efficient communication technology; it allows actors to align commitments and thereby reduces incentive problems; it allows screening of agents; and it motivates effort”. 2
Field research has shown that tacit knowledge is required in at least three areas of advanced technology: biotechnology in large pharmaceutical companies, advanced ceramics engineering and parallel computing (Senker, 1995) or, more generally, in high-tech sectors (MacKenzie and Spinardi, 1995).
The Dynamics of an ‘Innovation Driven’ Territorial System
63
In this chapter, the two sides of knowledge exchange are both at work and interact to produce conversion of knowledge into innovation and, in turn, into successful new products leading to superior economic performance. Moreover, by setting the parameters of knowledge spillovers, we can depict and deal with different kinds of TSPI as strongly ‘codified’ knowledge-oriented3 as well as looser and highly ‘tacit’ knowledge-oriented. In order to bring together considerations already developed up to now, consider a TSPI as a system based on the horizontal expansion of a given stock of knowledge, historically accumulated, plus the activation of an absorption process of external knowledge. So the stock of knowledge possessed by local firms is formed by a pool of tacit and codified knowledge. Not only are local firms active agents in producing knowledge, but local institutions also contribute to the process of information and knowledge diffusion. Successful management of knowledge transfer needs firstly a re-contextualisation of external knowledge and secondly an operational encoding of this new knowledge into the internal capabilities and organisational routines of firms. A third building block of the innovative TSPI is undoubtedly the emergence of a ‘local system of innovation’ (Braczyk et al. 1998; de la Moethe and Paquet 1998). After long research into ‘national systems of innovation’ (Freeman and Lundvall 1988; Nelson 1993) regional scientists began to put together some of the elements they had been researching separately in view of some interesting questions about the systemic nature of innovation that where intractable at a national level (Cooke and Morgan 1998). Competitiveness and innovativeness have become inextricably linked and in the light of a rising number of new ‘external’ competitors, local governments are more and more interested in assisting their industries to compete. An innovation system consists of elements and relationships that interact in the production, diffusion and deployment of new and economically useful knowledge (Lundvall 1992), but innovations are the result of social interaction between economic actors in which the feedback mechanism is strongly determinant. This social interaction is most effective where there are institutions able to maintain a learning culture within the TSPI. The management of the system increasingly becomes a matter of creating the institutional setting (rules, routines, habits) for stimulating interactive learning. A region (a TSPI) endowed with mutual understanding, trust and reciprocity within the collective economic community shows robust system characteristics which, in turn, can channel regular flows of information to the members of the regional innovation community. Regions and TSPI have their own ‘social filter’ in which innovative and conservative components are combined: the fewer the innovative components, the lower the capacity to accomplish high returns from R&D activity. But this social filter may be strongly influenced and shaped by 3
The more knowledge is codified, the more a certain degree of internal R&D is required, even for technological windowing and imitation from outside. The absorptive capacity of a firm strongly depends on its ability to manage informational flows and external technical knowledge.
64
A. Bramanti, U. Fratesi
appropriate local innovation institutions (see Crescenzi and Rodr´ıguez-Pose 2009, in this volume). The idea of a convergence and crossing-over of three actors – research, business and government – has been represented and explained through the Triple Helix Model (THM) (Leydesdorff and Etkowitz 1998). The model refers to a spiral dynamic of innovation that captures multiple reciprocal relationships in institutional settings (public, private and academic) at different stages in the capitalisation of knowledge. The THM could be represented by three factors: the actors, the institutions and the rules and regulations. A number of researchers suggested analyzing co-evolution between technologies and institutions in addition to co-evolution between markets and technologies (Dodgson and Rothwell 1994) and the THM focus on the interaction between these different interfaces. The sources of innovation in the model are not synchronized a priori, so different outcomes are possible, generating differences between territories and sectors. The evolutionary interpretation of the THM assumes that universities, government and industry within TSPI are learning to encourage economic growth by means of loosely coupled reciprocal relationships and joint undertakings persisting over time. R&D of individual firms alone is no longer enough for technological effectiveness. Internal R&D must be complemented by external sources of innovation spreading from the other two spokes of the helix. The central idea is that, while R&D still matters, it is only a part of a larger system including education, vocational training, government support and linkage between actors. It is therefore worth underlining that the rationale of a spatial innovation system (regional or local) stems from the combination of local and distant connections (Malecki and Oinas 1999). For the effectiveness of a TSPI, the quality of the interactions taking place between firms and between the firms and the other economic agents is very important. These interactions are facilitated by the social networks and milieu effects. Networks are socially bounded and learning through networking and by interacting is frequently grounded in face to face contact. In the literature on social capital (hereafter SC), (National Statistics 2001) we commonly find three basic forms: bonding SC, bridging SC and linking SC. There is more than one reason for considering that successful and dynamic interaction between bonding and bridging forms of social capital is likely to be a key component of sustained knowledgegeneration (Schuller 2006). If we think of bonding SC as the relationships within relatively homogeneous groups – such as intra-disciplinary or intra-professional affiliation – and of bridging SC as the relationships between relatively heterogeneous groups – such as interdisciplinary or inter-professional connections – we can see that the interplay between the two SCs – and the two professional groups – may give rise to effective inter-disciplinary and complementary interaction that promotes technological innovation. Social networks can easily support: (i) the exchange of information and knowledge; (ii) the enforcement of trust and loyalty reducing dynamic uncertainty; (iii) the feeding of creativity; (iv) the strengthening of technological problem-solving. When
The Dynamics of an ‘Innovation Driven’ Territorial System
65
social proximity overlaps and combines with institutional, organizational and technical proximity, it fosters processes of collective learning. Ideas and creativity can be picked up from outside the firm, but within (as well as from outside) the TSPI and actors can increase their ability not only to react to change but also to promote it.
3 System Dynamics and the Modelling of Innovative TSPI System dynamics (SD) is rooted in the notions that the feedback approach and the endogenous perspective generate explanatory power. Forrester, the pioneer of SD, suggested that feedbacks imply that “decisions are not entirely ‘free will’ but are strongly conditioned by the environment” (1961, p. 17). Typically, firms within a TSPI exhibit aggregate behaviour resulting from selforganization of local decisions. That implies that simple algebraic mathematical analysis does not apply, while simulations provide a promising tool. System dynamics models seem appropriate to highlight the feedbacks and loops arising between firms through the interaction of individual behaviour. The system dynamics modelling in economics implies an ecological approach, i.e. basically macro. However, it is on an intermediate level between microeconomic solvable models and traditional macroeconomic models and this has some advantages: – it has the advantage over the former of being able to represent more complex behaviour since the model can adapt to a broader view of the problem at hand and is not constrained to a small number of analytically solvable equations; the drawback is that normally the optimisation of individual behaviour is lost, since agents tend to follow rules instead of maximizing case by case. But this is not implausible in a world of imperfect information; – this approach has the advantage over traditional macroeconomic models of allowing much more complex and realistic behaviour to be modeled and of finding an equilibrium for a system in cases in which assumptions would be needed. The main advantage of an approach based on dynamic simulation is the possibility of representing loops and feedbacks and of evaluating their prevailing effect. On the other hand, this approach also permits the patterns which lead to equilibrium and also to dis-equilibrium patterns which may be the rule for real world TSPIs to be studied. Changing one part of the system may have un-expected and counter intuitive effects; many times this sheds new light on complex problems and suggests new ways of analyzing them. So, the system dynamics approach develops all its potential in dominating the uncountable ‘what if’ arising when dealing with a complex dynamic process. While the stock-and-flow diagram provides a powerful focal point for understanding TSPI evolution, the real power of SD is achieved/shown by running computer simulations and examining the resultant dynamic behaviour.
66
A. Bramanti, U. Fratesi
4 Setting the Scene: The Agent Interaction Putting together the theoretical hypothesis illustrated in Sect. 2, we introduce two different ‘circuits’ at work within an innovative TSPI. The first is the producers’ network – and specifically we model three kinds of firms, leader, follower and cooperative sub-contractor – and the second is the innovation process rooted in the circuit of knowledge production, accumulation and exchange. Let us briefly describe the three different classes of agents competing within the TSPI. The first type of agent that we characterise is the leader firm. It has been shown that, even in districts with small and medium enterprises, the performance of those firms which take a leading role over the others is considerably better (Bellandi 2007). There are firms which, due to their structure and entrepreneurial initiative, are natural gateways to the economic world beyond the TSPI and hence acquire knowledge on markets and technology which can be exploited to build product which are competitive world-wide. End-markets for a TSPI are basically external, since it is small with respect to the potential economies to which its products can be sold. Being an open space on a regional or infra-regional scale, the TSPI buys inputs on the national and international markets and needs to deliver products of sufficient quality to be sold externally. Consistent with the view expressed above, knowledge is not a purely internal and endogenous asset for a TSPI: the global pace of innovation is fast, and a large number of competitors are doing R&D at the same time. Since innovation is also an incremental process (Dodgson and Rothwell 1994; Breschi and Malerba 2005), no TSPI can compete in an innovation-led economy if it is not able to (i) acquire the newly generated external knowledge quickly and effectively and (ii) generate some workable improvements itself leading to a competitive advantage over the competitors in at least one business area. The leader firms are therefore those firms with developed ‘in house’ R&D activity and – due to this investment – a capacity to absorb external knowledge which becomes available on the external networks. Without the leaders, the TSPI will not be able to manage innovation in the long run, since the innovative effort of local institution would be weakened by the lack of a strong absorptive attitude of the system. In particular, innovation does not directly and necessarily transform into competitive products and performance. By definition, the leader firms are the ones on the technology frontier, the only ones able to interact with the external firms through market or technological networks and in this way the only ones able to signal the technological evolution that the TSPI itself will have to undergo in the future to the internal firms of the TSPI. The absence of leader firms will thus prevent the TSPI from having high-wage skilled activities and will condemn it to remain a laggard. In this limit case, however, it is even hard to define the territory as a TSPI. The second group of economic agents are the firms here defined as co-operative sub-contractors. This modelling choice is deeply rooted in the extensive literature on the role of customer-supplier networks (CSN) (Nonaka et al. 2001; Lane 2001).
The Dynamics of an ‘Innovation Driven’ Territorial System
67
Here the term network means two or more firms involved in a long-term relationship. CSNs combine relatively loose coupling relations between legally independent units. Due to new competitive pressure on external markets, leader firms begin to concentrate on ‘core business’ and externalise activities which are performed at the same quality level but more cheaply by co-operative sub-contractors. At least a part of this network shift to new reciprocal and co-operative relations is based on trust. The reliability of the suppliers as well as the possibility of collaborating with suppliers in the development of new products are extremely important. The development of new products needs new components which are not yet available in the market. It is therefore of paramount importance for a leader firm to have a networks of suppliers which collaborate with it in order to invent and industrialise the components needed. A successful TSPI has therefore some of the characteristics of an industrial complex. The firms of the co-operative sub-contractors are hence essential for the effectiveness of a TSPI: allowing and supporting the leader firms to innovate and produce goods that are competitive world-wide, representing an ‘anchorage’ factor which prevents leader firms from delocalizing. Within the TSPI, however, not all firms contribute to the generation of innovation; some firms located in the TSPI could assume free riding habits. These firms are substantially imitators, and we choose to call them followers, since they follow the leader firms with products which are similar in technological content after the time needed to copy it, which is shorter for firms located in the same area. These firms behave as predators in the Volterra-Lotka prey-predator model.4 Obviously, we can illustrate different and further tasks in the TSPI, but in the frame of methodological parsimony – typical of any SD model – we are convinced that the core structure of the innovative system is duly captured by the three types of firms. The model allows us to consider ‘residual’ – everything that is not exactly represented by the three archetypes – through a factor of labour force not employed by any of the previous three categories and which is considered to be employed by generic sub-contractors (which are not modelled).
5 An All-Encompassing Model On the basis of the theory outlined in the previous sections, it is possible to build a simulation model that represents the behaviour of the various types of actors described in Sect. 4 and their interaction nurtured by the main flows of knowledge exchange – generating innovation and successful performance – within the TSPI. No data exist at the moment to calibrate a model of innovative behaviour in a real TSPI, for this reason, in this phase, we chose to build a model which is purely theoretical, an abstraction of a generic real TPSI. For this reason, all variables are 4 It is obviously possible that these firms are not brand new but are old leader or co-operative subcontractors firms which at a certain moment find convenient to limit themselves to the imitation and no longer participate in the circuits of knowledge creation.
68
A. Bramanti, U. Fratesi
standardized, i.e. their initial values are fixed at 1 for the parameters and at 100 for the status variables. When, on the contrary, two or more variables need to be directly compared, one of them is fixed at 100 and the others are fixed as a fraction or multiple of 100 (see also the model equations in the Appendix). The model is subject to a process of calibration which, through the choice of the constants in the equations, consists in making the inflows and outflows of each node identical, when the value of the node is standard, with a procedure similar to that applied in a different (multiple equilibria) framework by Fratesi (2004). This process imposes ex-ante to the model the presence of at least one equilibrium, whose stability can be tested ex-post by introducing small perturbations and observing whether they adjust or not and, if not, if they make the system adjust to other equilibria. Since the purpose is to represent the sustainable patterns of innovation within the TSPIs, the presence of one equilibrium is a minimum requirement for these patterns to be represented. Under normal parameterisation, the model is stable, and the initial calibration equilibrium is reached again after any exogenous temporary perturbation. With a stress analysis – omitted for reasons of space – it is possible to show that the original equilibrium is very stable and that the exponents of the equations would need to be set implausibly high to render it unstable. The model hence has one equilibrium for each combination of the parameters and this allows us to represent the different sustainable patterns for the TSPIs associated with the different parameterisations of the system, by a one-to-one correspondence. The approach is typically populationist, with three populations (leader firms, co-operative sub-contractors and followers) which can increase or decrease and an intuitive positive judgement applies to positive growth while the opposite applies to a decrease. Instead of representing the number of firms, we will use their aggregate size in terms of employment. This will allow us to consider, as a first approximation, all firms belonging to a group as contributing to the total, and to ignore considering their specific size. The elements needed are hence the birth and death functions of all these firms. Before coming back to this issue, however, it is convenient to illustrate the knowledge circuits among the firms and the agents of the TSPI.
6 Sources, Accumulation and Exchange of Knowledge Consistently with the difference outlined in Sect. 2, knowledge internal to the TSPI is modelled in two groups: the first is ‘firm-specific’ and belongs to the leader firms, the second is ‘TSPI-specific’ and belongs to all firms of the territory including the leaders. The latter is in fact non-proprietary, a kind of local public property. Internal knowledge, however, does not exhaust all the knowledge available worldwide, which is an exogenous parameter in the model. The literature on localized knowledge spillovers (LKS) is very broad and increasing at a very fast pace. This is not the context to review it even in outline (for a survey and a discussion see, inter alia, Breschi and Lissoni 2001). In the model we
The Dynamics of an ‘Innovation Driven’ Territorial System
69
simply assume that the phenomenon exists and is relevant as it feeds TSPI-specific knowledge from firm-specific knowledge, but it is regulated by a parameter.5 The main source of knowledge creation in the TSPI is R&D and, in turn, R&D depends upon three different sources: investment by leader firms; the local presence of universities and research institutions; and the degree of openness of the system. We will not need to distinguish the parts of R&D which are brand new and the parts which are acquisition of external knowledge. R&D broadens the proprietary stock of knowledge of leader firms. This may seem a bit unusual but is actually a further specification of the existing theory along the following line of reasoning: in order to develop and accumulate knowledge you have to produce at least a part of it ‘inhouse’; R&D carried out by universities cannot reach co-operative sub-contractors directly because in this model those firms do not invest in direct R&D and therefore do not have the right absorptive capacity to enable them to exploit it. Cohen and Levinthal (1990) clearly show that firms must invest in basic research to create the capability to recognize, exploit and assimilate knowledge produced elsewhere. The openness of the system is equally important for R&D, since the more open a TSPI is to the international environment, the more easily ‘state-of-the-art’ knowledge becomes locally available and, hence, the faster the pace of innovation becomes within the TSPI. Obviously, in the presence of spillovers effects, the stock of local knowledge that is not firm-specific rises and is also incremented by the different learning mechanisms already mentioned. These spillovers are mainly due to the presence of co-operative sub-contractors and are facilitated by the presence of strong positive milieu effects, i.e. of the possibility for firms to co-operate in networks. Both stocks of internal knowledge – firm-specific and system-specific – are subject to obsolescence, because there is a continuous process of innovation outside the TSPI, making older techniques obsolete: knowledge is overtaken by new knowledge in an endless process of creative destruction. Possessing knowledge – even in the presence of an appropriate absorptive capacity – is not enough to gain market share. The innovations have to be transformed into saleable products. From the continuous accumulation of knowledge, the leader firms are able to introduce new products whose invention is the outcome of a stochastic process designed to model the associated uncertainty. In this way, the leader firms are able to increase the demand for their products on the international market. Notice that the TSPI is assumed to be small with respect to the international economy. For this reason, the only constraint on international demand for local products is local productive capability, provided the products are sufficiently technologically advanced. The approach is, therefore, one of partial equilibrium, leaving the extension to general equilibrium to future work. However, even this partial equilibrium approach fits well such cases as the introduction of successful new innovative products onto international markets: in the first stage the firms don’t know the shape of the demand very 5
Parameters are relevant in the model from two points of view. Setting them at different starting values we are able to describe different regional contexts. Moving them within a simulation we can answer some of the numerous ‘what ifs’ arising when dealing with agent interactions and feedback.
70
A. Bramanti, U. Fratesi
well and are thus unable to maximize their revenue with the price; the first constraint for the new and successful product is productive capability. Similarly to the accumulation of knowledge, the process of product innovation also has to be continuous since imitators are constantly at work internationally and locally. The magnitude of international imitation depends on the openness of the system6 as well as on the state of technology inside and outside the TSPI, whereas local imitation – which increases the share of products manufactured by the followers to the detriment of those manufactured by the leaders – depends (i) negatively on how large the ratio of proprietary local knowledge to non-proprietary local knowledge is, since only the second is available to the followers and (ii) negatively on the strength of milieu connections, since they are presumed to keep opportunistic behaviour under control. The hypothesis here is that, once a follower firm becomes able to produce the same products as a leader firm, it can do so with lower costs, since it doesn’t spend on R&D and hence makes the leader gives up this production; a mechanism similar to a product life cycle. The market shares of the followers can also be eroded externally, depending on the openness of the system. The co-operative sub-contractors, on the other hand, are not on the international market, and do not have own products but are heavily dependent on the leaders. We can state that the follower firms are not bad for the TSPI in the short run, being involved in export production, but are harmful for development in the long run since, because they do not invest in innovation, they hamper the accumulation of knowledge and hence sustainable competitiveness.
7 Supply Constraint and the Carrying Capacity of TSPI A TSPI has limited resources: the skilled workforce is small; as are social overhead capital (infrastructure) and land. This raises the problem of the carrying capacity: no system can bear excessive weight, due to the supply constraint and the consequent rationing due to increasing prices. A special role in the model is therefore attributed to the skilled workforce, which is a pre-requisite for the activity of innovative firms. We assume that the number of skilled workers is a parameter of the model and migration is not modelled, although it might be interesting to direct attention to it, as other chapters of this book do (Bramanti and Riggi 2009; Riggi and Maggioni 2009, in this volume). Innovative firms can also use unskilled workforce, but this is not constrained in the model, since unskilled workers can be cheaply poached from the non-modelled generic sub-contractors. Assuming that price in innovative products is a mark-up and that some substitutability between skilled and unskilled labour is possible, all three types of firms 6
Which in this way has a twofold counterbalancing effect: on the one hand it facilitates the flows of knowledge between the TSPI and the external inflows making easier the R&D process; on the other hand it facilitates to external imitators to copy the products of the leader and hence decrease their market share.
The Dynamics of an ‘Innovation Driven’ Territorial System
71
Table 1 The evolving structure of TSPI
Falling number Rising number of firms of firm
Decreasing employees
Increasing employees Average size decreasing
3 Growth with fragmentation (Average size decreasing)
2b Growth with expansion
2a Average size increasing
Average size decreasing
4b Downsizing of TSPI
4a Average size increasing
Growth with concentration (Average size increasing)
1
modelled compete on the same local labour market to hire skilled workers, whose importance is greater for the leader. This endogenously generates a wage for the skilled workers which depends positively on the quantity of firms present in the territory and negatively on the skilled workforce available (see the equations in Appendix 1). The wage of immobile skilled workforce, multiplied by the number of skilled workers, is a good performance indicator for the model: the higher the wages a TSPI is able to pay to its skilled workforce, the more successful is its economy. High wages, however, affect all types of firms negatively, and therefore act as a cause of mortality (or relocation, which is equivalent for the TSPI) for all. Having set-up all the elements of the model it is now time to introduce the equations of birth and mortality rates for the three types of firms: in the model, these are not necessarily to be interpreted as births and deaths of firms, since the hiring of workers by new firms and the increase of employment by incumbent firms have the same aggregate effect. Here we are not interested in firm size. Saying ‘leaders are growing’ in the model means the total employment of leader firms is rising independently of the number of firms. To some extent, we can say that situation [1] and [2a] (see Table 1) are more conducive of innovation – as claimed for the role of medium firms in Italian TSPI (Varaldo 2006) – but it not always true and we do not have a unique trend in the change of firm size within successful TSPI, and anyway, we are assigning a positive growth role to the right column of Table 1.
8 Interactions Between Firms and Feedback Mechanisms The birth rate of leader firms is positively influenced by the ratio of demand for leader products to the number of leaders combined with the presence of co-operative sub-contractors in the area. The functional form chosen (see Appendix 1) is an exponential one, which allows us to set up exponents according to the strength of
72
A. Bramanti, U. Fratesi
adjustment mechanisms (with higher exponents implying a speedier adjustment). Note that since these are not production functions, the sum of the exponents is not very important because the model is not calibrated on real data. The sum of the exponents being equal to one, or more or less than one, is not a signal of, respectively, constant, decreasing or increasing returns to scale nor does it influence the stability of the calibration equilibrium, which needs implausibly high exponents to be unstable. This process, like many others, does not have an upper ceiling, but, since each variable in these processes has decreasing returns to scale, the joint effects of the inflows and outflows tend to return the system to the equilibrium. The same functional form has been used for the other equations. For instance, the death rate is also negatively dependent on the ratio of demand for leader products to the number of leaders, and also on the wages and the openness of the system, which would make de-location easier (see Appendix 1 for the full set of equations). The co-operative sub-contractors do not have own products, but depend on the leaders and hence tend to adjust to the number of the latter present in the TSPI. The birth rate therefore depends on the number of leaders per sub-contractor and on two factors which ought to strengthen proximity ties: the milieu effect and the quantity of existing local knowledge, which both help leaders to find productive partners within TPSI instead of in the international market. Because followers are imitators, their birth rate does not depend on either of the other types of firms, but only on the ratio of demand for their own products to the number of followers. The death rate depends negatively on the same ratio and on the price of skilled labour.7 In summary, the following five adjustment mechanisms are at work in the model: • • • •
leaders tend to adjust to demand for leader products followers tend to adjust to demand for follower products sub-contractors tend to adjust to leaders the total employment in all firms tends to adjust to the carrying capacity of the system through the price mechanism • and, obviously, all the three kinds of firms compete on the same local labour market for skilled workforce The model is dynamically stable due to these mechanisms, but can assume different equilibria depending on the parameters.8 Figure 2 represents the employment pattern in the three types of firms in a stochastic simulation. Leaders and co-operative sub-contractors vary in phase with a very short adjustment lag, whereas the followers move like typical predators in prey-predator models, i.e. increase after an increase in leader firms and, at the same time, cause the leaders to decrease until they themselves decrease. 7
Followers also need skilled labour to imitate leader products. Based on the absorptive approach, an economic agent would not be able to innovate without the proper workforce. Technology may be free (through LKS) but producing innovative products is costly. 8 The strength of the model is its universality. All the TPSI share the same general structure but each may have its own set of parameters.
G
Qualified Workers
Follower Predator
11
INDICATOR Qualified Wages
Follower IN
Fig. 1 Diagrammatic representation of the model
Follower OUT
3
5
Demand for Follower Products
international imitation 2
10
Local Imitation
4 4bis
Product Innovation
F
Subcontractors OUT
Milieu effects
Leader OUT
Demand for Leader Products
International Imitation
9
B
Local Knowledge
Strenght of spillovers
D
Knowledge Leader
External knowledge
Specific knowledge ratio
Leader IN
Cooperative Subcontractors IN subcontractors
2
1
Leader
A
Openness
7
6 Obsolescence rate
Obsolescence 2
Learning
8
Knowledge spillovers
Obsolescence
R&D
C
Local Universities & Insitutions
E
The Dynamics of an ‘Innovation Driven’ Territorial System 73
74
A. Bramanti, U. Fratesi 1: Leader
2: Cooperativ e subcontractors
3: Follower Predator
102 100 50
1: 2: 3:
3
2 2
1
3
100 100 50
1: 2: 3:
2
3
2
1 1
1 3
99 100 49 0.00
1: 2: 3:
100.00
Page 1
200.00
300.00
400.00
Time Untitled
Fig. 2 The co-variations between the various types of firms present in the territorial system
Price of resources effects
Leader
–
+
Direct effects
+
Leaders
+ Direct effects
–
Cooperative subcontractors
–
Cooperative subcontractors
+
Price of resources effects
3a
Imitation
+ Local knowledge
3b
Fig. 3 Two examples of the feedbacks and loops present in the model
The model also exhibits a number of loops and feedbacks, both positive and negative, whose aggregate effect is complex and depends on the parameters. The system dynamics approach has been chosen precisely because it allows these feedbacks to be modeled in a way consistent with actual dynamics. The first example of these feedbacks is in Fig. 3a, which shows the relationship between leaders and sub-contractors. The direct effects of one on the other is positive: leader firms are the source of demand for sub-contractors firms and, at the same time, the competitiveness of leaders is strengthened by the local availability of sub-contractors which can co-operate in production and in the development of new
The Dynamics of an ‘Innovation Driven’ Territorial System
75
Table 2 Different performance according to alternative set of parameters Weak endowment Strong of R&D endowment of R&D institutions institutions
Low degree of openness
3 ‘Innovative TSPI’ but very self-centered innovation may decrease. It needs to strength external networks
High degree of openness
2 ‘Internationalised TSPI’ very innovative, highly competitive, gaining leaders robust positive feedbacks
‘Dominated TSPI’ ‘Localistic TSPI’ Outflows easily overcome inflows, or scarcely competitive, loosing leaders not able to cut down vicious circles inflows are not fully exploited (due to weakly positive feedbacks lack of absorptive capacity) 4 1
products. There is, however, also a negative indirect effect: because local resources are constrained, the presence of more leader firms makes resources more expensive for the sub-contractors and vice-versa (Table 2). A second, less immediate example is in Fig. 3b: the direct effects of leader firms on the co-operative sub-contractors are positive, and the presence of more sub-contractors makes an increase in local knowledge possible. However, the more knowledge becomes ‘public property’ – that is, knowledge in non-idiosyncratic and ‘firm specific’ form that is not fully appropriated by leaders – the easier it is for other local firms to imitate leader products, raising the weight of followers to the detriment of leader firms.
9 Simulating Alternative Growth Patterns This section shows the ability of the model to illustrate economic performance by different TSPIs depending on their specific set of parameters. In the following sub-sections, we explore the performance of different TSPIs (all with the same productive structures but differing in openness, endowment of research infrastructure, and alternative spillovers effect. They show different paths and locate in different places in the ranking of regions.
9.1 TSPI Endowed with different R&D Infrastructure The first experiment relates to the weight of universities and other public or private research facilities within the TSPI. These have direct positive effects on the innovation of leader firms, whereas their direct effects on co-operative sub-contractors and follower firms is negligible and therefore excluded from the model.
76
A. Bramanti, U. Fratesi
160
250
140 200
120 100
150
Leader
Follower
80 60 40
100
Cooperative subcontractors (secondary axis
50
Wages of qualified workers (secondary axis)
20 0 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270
0 Endowment of R&D infrastructure
Fig. 4 Behaviour of TSPI depending on the endowment of R&D infrastructure
However, the endowment of R&D infrastructure has indirect effects on all other economic variables of the TSPI. Figure 4 reports the equilibrium values of the model variables representing employment in the various types of firms, plus the wages of skilled workers, which can be used as a performance indicator. The interval chosen for the horizontal axis is from the minimum value of the parameter which allows an equilibrium to exist and, on the right side, a value after which nothing significant happens. The numbers are not significant either, since all variables are standardized at 100 at the beginning. The qualitative behaviour is more interesting: there is a clear increase in employment in the three groups of firms up to the carrying capacity of the system. However, the patterns seen in the simulation differ from group to group: the firms which gain first from an increase in R&D infrastructure are the leaders, and the cooperative sub-contractors which gain from local spillovers. The followers grow very little: they do not benefit directly from the presence of the leaders through an inputoutput relationship, but only from imitating their products. In addition, they don’t gain from an increase in R&D infrastructure, which instead makes the products of the leaders more technologically advanced and hence less easily imitated. From a systemic point of view, an increase in R&D infrastructure is always desirable, since the wages of workers always increases with them.
9.2 TSPI with Different Degrees of Openness The second experiment deals with the optimal degree of openness for the TSPI. The openness of a system has various effects: it influences positively the possibility of local firms innovating, through technological windowing, but it also makes the imitation of local products by external producers easier and, finally, it can exert a
The Dynamics of an ‘Innovation Driven’ Territorial System
77 120
250
100
200
Leader
80 150 60
Cooperative subcontractors
100 40 50
20
0
Follower
Wages of qualified workers (secondary axis)
0 Openness of TSPI
Fig. 5 Behaviour of TSPI depending on its openness
de-localising effect on the leader firms which are less embedded in the territory and more mobile internationally in their search for production opportunities. As in the previous case, Fig. 5 plots the different equilibria in terms of employment of firms for different values of openness. The first relevant result is ‘death from entropy’ of a system whose openness is close to zero. Since local innovation cannot take place without contacts with the external world, this makes it impossible to have competitive leader firms and, consequently, sub-contractor firms, for two self-reinforcing reasons: without a leader, there isn’t enough demand for their products and because the non-proprietary local knowledge depends heavily on spillovers, which start from the proprietary knowledge resulting from R&D by leader firms. The follower firms, on the other hand, are not affected very negatively by the lack of openness of the system, since the main effect on them is a positive one, the absence of international imitation. From a systemic point of view, the graph of the wages of skilled workers shows that there is an optimum degree of openness for a TSPI.9 After openness has reached its optimal level, in fact, the advantages in terms of further innovation are more than offset by the leader firms being foot-loose, and a number of them relocating to external sites. It is interesting to see how the leader firms carry the co-operative sub-contractors, which grow after them and which are less negatively affected by international imitation and less likely to become foot-loose, which explains why the sub-contractors tend to decrease only for very large values of the parameter.
9
We define here an operational social optimum which is consistent with the set-up of the model and is the value of the parameter which maximizes the sum of the skilled wages.
78
A. Bramanti, U. Fratesi
9.3 TSPI Showing Different Strength of Internal Spillovers The strength of spillovers, defined as the speed at which firm-specific knowledge becomes available throughout the TSPI, can have a twofold effect on the performance of the system. On the one hand, the existence of spillovers makes knowledge more widely available, so that more local firms benefit from the innovation of the others and are hence able to be competitive. On the other hand, however, spillovers make the knowledge of imitators more similar to that of the original innovators, and lead to innovative firms which cannot take advantage of (a part of) their innovative effort in the markets and are hence negatively affected. The model is able to address this behaviour and to compare the equilibria, i.e. the economic situation of a TSPI depending on the strength of spillovers. As can be seen from Fig. 6, there is also a social optimum for the strength of spillovers which maximizes the prices/wages. The strength of spillovers which is most favourable to leader innovative firms, is lower than the social optimum, because for low spillovers they are able to maintain the advantages of innovation internally. However, it is worth noting that the spillovers optimum for the leader is not zero, because the presence of co-operative sub-contractor firms that collaborate in the development of products is very useful to them, and the latter firms need to have access to a rich knowledge system. The strength of spillovers which is most favourable to co-operative sub-contractor firms is higher than the social optimum. These firms benefit directly from spillovers, which make them more technologically endowed, so if spillovers are high, they are more innovative and more useful to leader firms which will find the skilled suppliers they need more easily locally. The optimum spillovers for the co-operative sub-contractors, however, is also bounded, since the leader firms suffer too much from imitation when spillovers become too strong and their weakness is transferred to their sub-contractors. 200
120
180 100
160
Leader
140
80
120 60
100 80
40
60 40
20
20 0
0 Strenght of spillovers
Fig. 6 Behaviour of TSPI depending on the strength of internal spillovers
Cooperative subcontractors
Follower
Wages of qualified workers (secondary axis)
The Dynamics of an ‘Innovation Driven’ Territorial System
79
The followers, finally, are negatively but very weakly affected by too high spillovers: In fact, strong spillovers enable the firms which act as predators to imitate more easily the products invented by the innovative leaders; however, this is counterbalanced by the weakness of leaders which provides them with fewer products to imitate.
9.4 TSPIs Showing Different Levels of Milieu Effect In Fig. 7, the performance of the TSPI is plotted against different magnitudes of milieu effect. When milieu effects are very low, the TSPI is characterized by nonco-operative behaviour. For this reason, only followers/imitators are widespread, whereas leaders and subcontractors are very few. Moreover, the low milieu effects have negative effects on learning and on the choice of local subcontractors by leader firms. When allll this is considered, the performance of the system in terms of wages of skilled workers is very low. When the milieu effects are higher, leader firms gain from less imitation and from the presence of more sub-contractors locally. Learning effects favour the latter, so that the systemic performance of the TSPI is higher. In the long-run, milieu effects favour both leader and local co-operative subcontractors, but the latter which are chosen more frequently by the leaders are most favoured. The followers, being imitators, are very scarce in an economy with strong milieu effects. What is remarkable, though, is that, economies characterized by low milieu effects – where there are more leaders – also have more followers (these firms need leaders to imitate) than economies in which the milieu effect is very low.
250
250
200
200 Leader
150
150 Cooperative subcontractors
100
100 Follower
50
50
0
0 Milieu Effects
Fig. 7 Behaviour of TSPI depending on the strength of milieu effects
Wages of qualified workers (secondary axis)
80
A. Bramanti, U. Fratesi
10 TSPI Fitness and Policy Experiments Simulation methodology enables the study of equilibria, as was done in the previous section, but the dynamics leading to equilibria are also very interesting. In particular, the strand of economics usually defined as “evolutionary” makes it obvious that economic systems spend most of their time out of equilibria, and so the dynamics by which the equilibrium is reached are as important as the equilibrium itself. In this section, three experiments are conducted in which the economy of the TSPI, starting at the calibration equilibrium is perturbed by various external stresses or behavioural changes, and the subsequent adjustments are described. We start by simulating a sharp decline in the presence of the leader firms (§ 10.1); then we examine a scenario of easy imitation which discourages leaders from innovative investment (§ 10.2); and finally we look at a de-localisation choice of leader firms (§ 10.3). We have not chosen three ‘negative’ shocks by chance; we are specifically interested in the conditions for TSPI survival in a harsh global competitive scenario. Nevertheless, there are no logical difficulties in studying positive exogenous shocks whose outcome is to make the TSPI perform better.
10.1 The Resilience of the System In the first experiment, the external shock is the sudden loss of their international market share (let us say 50%) by the leader firms due to the emergence of new strong competitors such as the Chinese case in a number of consumer goods markets. The immediate effect (Fig. 8) is the adjustment of employment by the leader to their present market strength, by reducing their productive capacity. The reduction of leaders brings a decrease in R&D expenditure with a consequent fall in the stock of knowledge of leaders. However, this fall in leader production implies a decrease in wages. This explains why the fall of co-operative sub-contractors is not as large as that of the leaders and, consequently, the reduction in local knowledge due to lower spillovers from the leader is also not as strong (learning mechanisms are still at work). The followers, which pay lower wages, might initially gain from the loss of leaders. However, since they rely on the imitation of leader products, they are negatively affected as soon as international imitation reduces the demand for their products, which is not filled with the same strength as before. Notice that due to continuing product innovation springing from previously accumulated knowledge, the demand for leader products starts increasing immediately after being hit by the shock. Moreover, the stock of leader knowledge has decreased more slowly than their market share due to the cumulative nature of knowledge. Finally, the presence of a co-operative sub-contractors is not directly affected and hence not so seriously hit. For all these reasons, the leader firms are those which are most affected by the crisis but also the first to recover. This implies that the stock of knowledge of these
The Dynamics of an ‘Innovation Driven’ Territorial System
81
Fig. 8 The resilience of the system
firms also starts recovering, followed by the sub-contractor firms and also by the products of the followers, which experience increasing opportunities for imitation. Wages also reflects this pattern and begins recovering as does, finally, the stock of non-proprietary local knowledge and follower employment.
82
A. Bramanti, U. Fratesi
This experiment has shown the mechanics of what is known in the literature as the “resilience”10 of the system, i.e. the capacity of a system due to internal resources to recover from external negative shocks, provided that some minimum threshold level is not exceeded and the social tissue and the co-operative attitudes of the agents are not undermined. In this case, the accumulated stock of knowledge, that can be used to produce further innovations, and the presence locally of a set of co-operative and skilled suppliers, has allowed the leader firms, the engine of wealth, to recover from a large, though temporary, external shock. The system is not resilient when the stress, instead of being temporary and external, affects the internal structure permanently, as is the case in the following two examples.
10.2 An Increase in Local Imitation This second experiment looks at local imitation assuming that firms that do not innovate, increase their imitation activity substantially. This can happen due to a loosening of local ties, the weakening of trust, the rise of more aggressive and opportunistic behaviour by newcomers and/or an increase in turnover in the local labour market with the poaching of skilled workers. The result is a reduction in the social reprisals for copying another’s products, or a change in the behaviour of firms that once invested in internal R&D and decide that it is cheaper for them to reduce these expenses and rely on imitation and ‘reverse engineering’. In the model, rising copying attitude is represented by the permanent doubling of the pace of local imitation by follower firms. Figure 9 reports the dynamic patterns within a TSPI. The immediate effects of the increase in local imitation is rising market demand for the followers to the detriment of the leaders. This also implies an increase in the production of the former and a fall in that of the latter. The production, of the co-operative sub-contractors also starts to fall, though less rapidly (Fig. 9). These effects on demand and production are not the only ones, however. Fewer leader firms imply less R&D and, consequently, a reduction in the stock of leader knowledge and their Product-Innovation process. The local knowledge of the TSPI also starts to decrease, due to a reduced presence of sub-contractor firms and lower spillovers. Wages fall, but the positive effects of this on leader and sub-contractor firms is not able to compensate for the fact that they obtain less from their innovation and are hence less competitive. The follower/imitator firms, finally, have a large advantage only in the short run: after the reduction leader firms, in fact, they have fewer products to imitate and this 10
‘Resilience’ is a mechanical property of metals allowing them to deform but not to break. Applied to social aggregates it refers to the capacity of the system to modify itself without loosing its identity.
The Dynamics of an ‘Innovation Driven’ Territorial System 1: Leader
2: C ooperat iv e s ubc ont rac t ors
83
3: F ollower Predat or
4: I N D I C ATOR Qualif ied W ages
120
1: 2: 3: 4:
100
1
1: 2: 3: 4:
2
80
4
2 2
85 1 3
2
1
3
1 3
3
4 1: 2: 3: 4:
4
4
40 70 0. 00
50. 00
1: Knowledge Leader 1: 2: 3: 4:
1000 2000 120 100
100. 00
2: Local Knowledge 1
150. 00
3: Demand f or Leader Products
200. 00
4: Demand f or Follower Products
2
2
1: 2: 3: 4:
800 1600 80 75
1
2 1
3
3
2
3
3 4
1: 2: 3: 4:
1
4
4
4
600 1200 40 50 0. 00
50. 00
Page 2
100. 00
150. 00
200. 00
Time
Fig. 9 An increase in local imitation
also reduces their market share, until it reaches an equilibrium at a new value which is higher than the starting one but far lower than the peak it had reached. The new equilibrium of the model is characterized by lower wages for skilled workers, less production in both leader and co-operative sub-contractor firms and higher production in the follower firms, due to the relative cheapness of skilled labour, but not enough to compensate for the fall in the first two groups. Both leader firm-specific knowledge and local knowledge are lower.
84
A. Bramanti, U. Fratesi
The conclusion that can be drawn from this exercise is that TSPIs characterized by higher incidences of opportunistic behaviour are less innovative and less able to create wealth for their inhabitants.
10.3 An Increase in De-localisation The third experiment looks at how the TSPI is affected by a change in behaviour of the leader firms. These firms, for external reasons which make it cheaper for them to relocate, even in spite of losing the advantages of embeddedness, experience an increasing rate of de-localisation (Fig. 10). In the model, this is reflected by a permanent increase in the mortality rate of leader firms. For the TSPI it is in fact identical, if a firm closes or relocates elsewhere. The direct effect of this is a fall in the production of leader firms, while the first indirect effect is a reduction in the R&D of these firms (which is assumed to be a fixed share of their revenue in the model). The local co-operative sub-contractor firms also begin to suffer as a consequence of the decreasing demand. Since there is less demand for local skilled workers, wages also start to fall. The behaviour of the follower firms is slightly different from the others: due to the reduced competition for local workers and the fact that the leader firm-specific knowledge decreases more quickly and to a greater extent than local knowledge, the follower firms gain from the exit of leaders from the TSPI. Afterwards, with fewer leaders, the products the followers can imitate locally also decrease, and this impacts negatively on the products of the followers. For this reason, after the initial increase, the followers decrease to a level which is higher than the starting one but lower than the peak. The final equilibrium is hence an equilibrium with fewer leaders, fewer subcontractors and slightly more followers. Wages of skilled workers have fallen. There is less firm-specific knowledge and local knowledge has decreased even more in absolute terms but not in per-firm terms. The market shares of leaders and followers have also decreased, characterizing a TSPI which is less wealthy than before.
11 Concluding Remarks and Looking Ahead This chapter has outlined the complex processes of creation, accumulation, exchange and exploitation of knowledge and the derived circuits of innovation within a TPSI in an advanced economy. The model – according to the theoretical view which accepts the death of the Fordist ‘one best way’ of production and its replacement by different possible ‘worlds of production’ (Storper 1993) – has outlined a number of possible sustainable development patterns for TSPI, basically rooted in the same components.
The Dynamics of an ‘Innovation Driven’ Territorial System
85
Fig. 10 An increase in delocalisation
The main competitiveness factor has been identified as the presence of innovative leader firms, which however need to be embedded in a system of local collaboration in order to prosper and remain innovative. Leaders – with the support of their regional innovation system – are the main agents of knowledge accumulation (through R&D investment) and exploitation (via product innovation), originating innovative products which, in turn, assure the economic performance and dynamic competitiveness of the system (see Table 3).
86
A. Bramanti, U. Fratesi
Table 3 The process of knowledge accumulation and exploitation
A number of parameters regulate the flows of knowledge, and the degree of innovation and learning within TSPI. These parameters have been identified as the openness of the system, the presence of local innovative institutions, the speed of local knowledge spillovers, the extent of co-operative social network mechanisms and the endowment of local skilled labour. The TSPI exhibits different working equilibria, depending on the values assumed by these parameters, and with a variable mix of the three types of firms operating in it: innovative leader firms, co-operative sub-contractor firms, and opportunistic follower/imitator firms. The model also shows that some sets of behaviour parameters are better than others in terms of aggregate performance (which is measured by production of leader firms and wages paid to skilled labour). This is certainly an important result because it implies that policies matter, and should have a role in enhancing TSPI competitiveness. First of all, policies may foster the process of speeding up the innovation process by influencing the spillover effect as well as the learning processes within TSPI. The ways of organising public research activities within the TPSI strongly influences the costs of transferring the knowledge that has been produced. Moreover, the institutional setting – expressed by norms, rules and standards – dramatically affects the ‘epistemic communities of researchers’ at work producing R&D. Secondly, there is a range of balanced levels of openness and robustness (i.e. milieu effects and local spillovers in the model) whereas an unbalanced mix may damage the survival of the TSPI as such (see Table 4). Thirdly, the right balance among the different types of firms is equally important; specifically, leaders need local sub-contractors to compete internationally and so decide not to relocate and, at the same time, local subcontractors need leaders to exist. Due to the fact that they all depend on the same local resources and which are
The Dynamics of an ‘Innovation Driven’ Territorial System
High degree of robustness (milieu effect)
High degree of openness
‘Potentially implosive TSPI’ Very inward looking milieu, the circuits of innovation are no longer fed by external channels and fade
Low degree of robustness (milieu effect)
Table 4 Balanced vs. unbalanced evolution of TSPI Low degree of openness
87
‘Paralyzed TSPI’ ‘Potentially explosive TSPI’ Leader are no longer supported by local All the positive feedbacks are very sub-contractors and are rapidly attracted weak, scarcely innovative tissue may by external locations meet with difficulties on the demand side
3
4
2 ‘Leading TSPI’ Very innovative, highly competitive, strong regional innovation system local milieu conducive of pro-active innovative behaviour
1
‘limited’ by the carrying capacity of the system, too high a presence of any two of the types of firms is detrimental to the other. Openness and robustness are thus at the very heart of TSPI dynamics. The first is decisive, acting as a two-way connection: on the one hand it enhances R&D (through technological windowing), on the other, it acts as a de-localisation force exerting constant pressure on all the three types of firms. As innovative leaders are always screening alternative locations, this competitive challenge urges the TSPI to develop a strong regional innovation system and to produce effective local public goods. Equally important is the role of the local fabric. A strong milieu effect means trust11 , social cohesion, a sharing of rules and norms; in short, a complex mixture enhancing learning mechanisms and facilitating the birth of new co-operative subcontractor firms. The analysis contained in this chapter has been conducted theoretically with a simulation model due to the impossibility of getting the necessary data for an actual TSPI, hence it does not provide implementable policy prescriptions directly. However, theoretical concepts such as the degree of openness and the robustness of a TSPI could possibly be measured and in this way they may constitute a basis on which to run policies with the aim of improving both the innovative performance and the well being of territories. Looking ahead on the agenda, three main improvements seem to be closer than the calibration to real-World cases. The first regards the way of modelling some parameters – starting with openness and robustness – so that they are determined endogenously within the model. Both are a measure of some structural characteristics of the TSPI and so may be targeted by policies and certainly brought closer, to some extent, to a more desirable level. The second is the evolution of firms which, as a result of internal competition and system performance, may transform from one type into another. 11
Trust reduces coordination costs – other things being equal – but “trust cannot be bought: and if it could be bought it would have no value whatsoever.” (Arrow, 1971).
88
A. Bramanti, U. Fratesi
Thirdly, it is theoretically challenging to include at least a second TSPI in the frame and to explore the new scenarios arising from inter-regional competition. Or, the reasoning could be pushed even further into the logic of general equilibrium by endogenising the demand for innovative products, possible once the world is composed of a number of different competing and interacting TSPIs. But this is a task for a whole new book.
Appendix 1 — The model equations 1 – Leader(t) = Leader(t – dt) + (Leader IN – Leader OUT) * dt INIT Leader = 100 Leader IN = constant * Leader * (Demand for Leader Products/Leader) * (Cooperative Sub contractors/Leader)0.8 Leader OUT = constant * Leader * (Leader/Demand for Leader Products) * Skilled Wages* (Leader/Co-operative Sub contractors)0.8 Openness0.7 2 – Co-operative Sub contractors(t) = Co-operative Sub contractors(t – dt) + (Sub contractors IN – Sub contractors OUT) * dt INIT Co-operative Sub contractors = 100 Sub contractors IN = constant *Co-operative Sub contractors * (Leader/Co-operative Sub contractors)0.8 * Mileu Effect * Local Knowledge Sub contractors OUT = constant * Co-operative Sub contractors *(Co-operative Sub contractors/Leader)0.8 * Skilled Wages 3 – Follower Predator(t) = Follower Predator(t – dt) + (Follower IN – Follower OUT) * dt INIT Follower Predator = 50 Follower IN = constant * Follower Predator * (Demand for Follower Products/ Follower Predator) Follower OUT = constant * Follower Predator * (Follower Predator/Demand for Follower Products) * Skilled Wages 4 – Demand for Leader Products(t) = Demand for Leader Products(t – dt) + (Product Innovation – Local Imitation – International Imitation) * dt INIT Demand for Leader Products = 100 4bis – Product Innovation = constant * (Knowledge Leader/External Knowledge)0.5 *normal(1,σ 2) 5 – Demand for Follower Products(t) = Demand for Follower Products(t – dt) + (Local Imitation – international imitation 2) * dt INIT Demand for Follower Products = 50 6 – Knowledge Leader(t) = Knowledge Leader(t – dt) + (R&D – Knowledge spillovers – Obsolescence) * dt INIT Knowledge Leader = 1000 R&D = constant * External Knowledge * Local Universities & Institutions * (Leader)0.8 * Openness
The Dynamics of an ‘Innovation Driven’ Territorial System
89
7 – Local Knowledge(t) = Local Knowledge(t – dt) + (Learning + Knowledge spillovers – Obsolescence 2) * dt INIT Local Knowledge = 2000 Learning = constant * (Co-operative Sub contractors)0.8 * Mileu Effect 8 – Knowledge spillovers = constant * Knowledge Leader *Strength of spillovers Obsolescence = Knowledge Leader*Obsolescence rate Obsolescence 2 = Local Knowledge*Obsolescence rate 9 – Specific knowledge ratio = Knowledge Leader/Local Knowledge*2 10 – Local Imitation = constant * Demand for Leader Products * (1/Specific knowledge ratio).25 * (1/Mileu Effect) International Imitation = constant * Demand for Leader Products * Openness International Imitation 2 = constant * Demand for Follower Products * Openness 11 – Skilled Wages = constant * numeric solution to (Skilled Workers = (.6/.4)ˆ.4 * Leader *(1/Skilled Wages)ˆ.4 + (.4/.6)ˆ.6 * (Co-operative Sub contractors + Follower Predator) * (1/Skilled Wages)ˆ.6)
Calibration Parameters A-Openness = 1 B-External knowledge = 10,000 C-Local Universities & institutions = 100 D-Strength of spillovers = 1 E-Obsolescence rate = .05 F-Milieu effect = 1 G-Skilled workers = 100
References Arrow K (1971) Essays in the theory of risk bearing. North-Holland, Amsterdam Becattini G (1990) The Marshallian industrial district as a socio-economic notion. Pyke F, Becattini G, Sengenberger W (eds.) Industrial district and inter-firm co-operation in Italy. International Institute for Labour Studies (ILO), Geneva Becattini G (2004) Industrial districts: A new approach to industrial change. Edward Elgar, Cheltenham Bellandi M (2007) Industrial Districts and Waves of Industrialization. A Rich and Contested Terrain SR. Ital J Reg Sci 6(2):7–33 Braczyk HJ, Cooke P, Heidenreich M (1998) Regional innovation systems. UCL, London Bramanti A (1999) From space to territory: relational development and territorial competitiveness. Revue d’Economie R´egionale et Urbaine 3:633–658
90
A. Bramanti, U. Fratesi
Bramanti A and Riggi MR (2009) Sustainable interrelated growth: a phenomenal approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: The role of internal and external connections. Springer, Berlin Breschi S, Lissoni F (2001) Localised knowledge spillovers vs. innovative milieu: knowledge tacitness reconsidered. Papers Reg Sci 80:255–273 Breschi S, Malerba F (eds.) (2005) Clusters, networks, and innovation. Oxford University Press, Oxford Camagni R (ed.) (1991) Innovation networks: Spatial perspectives. Belhaven, London Capello R (1999) Spatial transfer of knowledge in high-technology Milieux: learning vs. collective learning processes. Reg Stud 4:353–365 Cohen MW, Levinthal DA (1990) Absorptive capacity: a new perspective on learning and innovation. Administr Sci Quart 35:128–152 Cooke P, Morgan K (1998), The associational economy. firms, regions, and innovation. Oxford University Press, Oxford Cowan R, David PA, Foray D (2000) The explicit economics of knowledge codification and tacitness. Ind Corporate Change 2:211–253 Crescenzi R, Rodr´ıguez-Pose A (2009) Systems of innovation and regional growth in the EU: endogenous vs external innovative efforts and socioeconomic conditions. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: The role of internal and external connections. Springer, Berlin de la Moethe J, Paquet G (eds.) (1998) Local and regional systems of innovation. Kluwer, Boston Dodgson M, Rothwell R (eds.) (1994) The handbook of industrial innovation. Edward Elgar, Aldershot Dosi G (1988) Sources, procedures and microeconomic effects of innovation. J Econ Lit 26:1120– 1171 Folloni, G. (2009) A model of local development, in Fratesi, U. and Senn, L. (eds.) Growth and innovation of competitive regions: the role of internal and external connections, Springer, Berlin. Forrester JW (1961) Industrial dynamics. MIT Press, Cambridge, Mass Fratesi U (2004) Regional economies, innovation and competitiveness in a system dynamics representation. Working Paper No. 204, Universit`a Politecnica delle Marche, Department of Economics Freeman C, Lundvall B (eds.) (1988) Small countries facing the technological revolution. Pinter, London Freeman C, Soete L (1997) The economics of industrial innovation. The MIT Press, Cambridge, MA ´ Gaffard JL (1990) Economie Industrielle et de l’Innovation. Dalloz, Paris Imai K, Baba Y (1991) Systematic innovation and cross-border networks. transcending markets and hierarchies to create a new techno-economic system. Technol Product, OECD, pp. 389–405 Lane C (2001) Organizational learning in supplier networks. Dierkes M, Berthoin Antal A, Child J, Nonaka I (eds.) Handbook of organizational learning & knowledge. Oxford University Press, New York, pp 699–715 Leydesdorff L, Etkowitz H (1998) The triple-helix as a model for innovation studies. Sci Public Policy 3:195–203 Lundvall BA (1992) National systems of innovation. Pinter, London Malecki EJ, Oinas P (eds.) (1999) Making Connections. Technological learning and regional economic change. Ashgate, Aldershot MacKenzie D, Spinardi G (1995) Tacit Knowledge, weapons design, and the uninvention of nuclear weapons. Am J Sociol 101:44–99 National Statistics (2001) Social capital. A review of the literature. Office for National Statistics, London, UK Nelson R (1993) National innovation systems: A comparative analysis. Oxford University Press, New York
The Dynamics of an ‘Innovation Driven’ Territorial System
91
Nonaka I, Toyama R, Byosi`ere P (2001) A theory of organizational knowledge creation: Understanding the dynamic process of creating knowledge. Dierkes M, Berthoin Antal A, Child J, Nonaka I (eds.) Handbook of organizational learning & knowledge. Oxford University Press, New York, pp 491–518 Pavitt K (1987) International patterns of technological accumulation. Hood N, Vahlne JE (eds.) Strategies in global competition. Croom Helm, London Pavitt K (1999) Technology, management and systems of innovation. Edward Elgar, Cheltenham Polanyi M (1969) Knowing and being. Edited with an introduction by Marjorie Grene. University of Chicago Press, Chicago Rallet A (1993) Choix de proximit´e et processus d’innovation technologique. Revue d’Economie R´egionale et Urbaine 3:365–385 ´ ´ Rallet A, Torre A (eds.) (1995) Economie Industrielle et Economie Spatiale. Economica, Paris Ratti R, Bramanti A, Gordon R (Eds.) (1997) The dynamics of innovative regions. GREMI, Ashgate, Aldershot Riggi MR, Maggioni M (2009) Regional growth and the co-evolution of clusters: the role of labour flows. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: The role of internal and external connections. Springer, Berlin Rogers EN (1995) Diffusion of Innovations. The Free Press, New York Rosenberg N (1982) Inside the black box: Technology and economics. Cambridge University Press, Cambridge Schuller T (2006) Social capital, networks, and communities of knowledge. Kahin B, Foray D (eds.) Advancing knowledge and the knowledge economy. The MIT Press, Cambridge, MA, pp 91–110 Scott AJ (1988) New Industrial Spaces. Pion, London Senker J (1995) Tacit knowledge and models of innovation. Ind Corporate Change 2:425–447 Steinmueller WE (2000) Will new information and communication technologies improve the codification of knowledge? Ind Corporate Change 2:361–376 Storper M (1993) Regional ‘Worlds’ of production: learning and innovation in the technology districts of France, Italy and the USA. Reg Stud 5:433–455 Storper M (1995) The resurgence of regional economies, ten years later: the region as a nexus of untraded interdependencies. Eur Urban Reg Stud 3:191–221 Storper M, Harrison (1991) Flexibility, Hierarchy and regional development: the changing structure of industrial production systems and their forms of governance in the 1990s. Res Policy 5:407–422 Storper M, Venables AJ (2004) Buzz: face-to-face contact and the urban economy. 4:351–370 Varaldo R (2006) Il nuovo modello competitivo e aziendale dei distretti industriali. Economia e politica industriale 1:25–42 Veltz P (1993) D’une g´eographie des coˆuts a` une g´eographie de l’organisation. Quelques th`eses sur l’´evolution des rapports entreprises/territoires. Revue e´ conomique 4:671–684
The Co-Evolution of Entrepreneurship and Clusters Christian Garavaglia and Stefano Breschi
1 Introduction1 Clustering of firms and spatial concentration of economic activities have received increasing attention over the last two decades from regional, international and industrial economists. Agglomeration economies, network effects and knowledge spillovers were believed to influence the growth of a territory and were perceived as key engines of cluster development in the economic literature. Furthermore, entrepreneurship has been increasingly considered as a crucial factor in determining the economic growth of a territory (Audretsch 2002; Acs and Audrestch 2003; Audretsch and Keilbach 2004). The generation, evolution and persistence of clusters seem to be strictly related to the entrepreneurial activities taking place in a given area. In this paper we investigate the co-evolution of entrepreneurship and emergence of clusters, we highlight the contributions of different theoretical perspectives in explaining the processes behind entrepreneurial activity and its influence on the process of cluster formation. On the one hand, entrepreneurial activity in a geographical area is perceived to be the means by which firms take advantage of the positive external spillovers generated in the area. On the other, some contributions recognise entrepreneurship to be strictly related and influenced by the social ties available in a spatially bounded region. The purpose of the paper is to explore theoretically the relationship between entrepreneurship and the development of spatial clusters of firms. How do clusters emerge and thrive? What is the relationship between entrepreneurship and cluster formation? What is the nature and impact of social networks and external economies in cluster development? In this paper, our first claim is that we find both supply and demand-side conditions in clusters for stimulating entrepreneurship. Our second 1 Christian Garavaglia wishes to acknowledge the financial support of the Italian Ministry for Education, Universities and Research (FIRB, Project RISC - RBNE039XKA: “Research and entrepreneurship in the knowledge-based economy: the effects on the competitiveness of Italy in the European Union”). Stefano Breschi gratefully acknowledges financial support from Bocconi University.
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
95
96
C. Garavaglia, S. Breschi
claim is that both pecuniary and non-pecuniary aspects influence entrepreneurship and cluster formation. The final claim argues that entry by spin-off represents a significant factor for cluster development and persistence. In discussing these questions, we do not intend to give an exhaustive survey of the literature, but to examine the relationships of major contributions of two different areas of research: entrepreneurship and cluster formation. We develop our analysis from an evolutionary perspective. Both entrepreneurship and cluster formation are dynamic phenomena that exhibit clear dynamics and coevolve with related environmental conditions. Looking at the structure of industrial districts and clusters at one point in time may therefore be quite misleading. In order to understand the relationship between entrepreneurship and cluster formation we follow a dynamic approach. The characteristics of the cluster emerge over time and co-evolve with the development of entrepreneurial activities and local supporting institutions. Cluster formation is a complex self-organising process: the symbiotic relationship between entrepreneurs and local resources is the core of cluster evolution. Consequently, we analyse the process of cluster formation in a dynamic setting and focus specifically on entrepreneurship.
2 The Entrepreneurial Process Entrepreneurship is crucial in influencing the economic growth and development of a territory. New firms and new entrepreneurial activities largely develop in geographically defined areas. Entrepreneurship, then, seems to be bound up with space and clusters. Once established, these geographical clusters seem to follow a selfreinforcing process (Feldman 2001). The first important task is to explain how entrepreneurial activities originate and develop. This section argues that it is crucial to take into account explicitly that entrepreneurial activities involve both pecuniary and non-pecuniary aspects. If it is obvious that entrepreneurs are profit-seeking economic agents, and consequently explore and exploit opportunities that facilitate the realisation of profits, it is less acknowledged that entrepreneurship also involves non-pecuniary aspects that relate to the cognitive and social spheres of the individual. This argument is developed in the present section. Cluster formation also shows both pecuniary and non-pecuniary factors. In the next sections, then, we consider the process of cluster formation and investigate how the processes behind entrepreneurship might influence and be influenced by the local context.
2.1 Defining Entrepreneurship Entrepreneurship is a polyhedric phenomenon. Different fields of study are involved in understanding and examining the essence of entrepreneurial activities: despite the potential richness that a mixture of disciplines implies, a major weakness is that
The Co-Evolution of Entrepreneurship and Clusters
97
in several cases “researchers from one discipline tended to ignore entrepreneurship studies by researchers in other disciplines” (Wortman 1992). Consequently, no widely accepted single definition of the concept of entrepreneurship has been given. Moreover, different types of entrepreneurial activities may be defined, such as the creation of new firms, the processes of “corporate entrepreneurship”, the concept of “academic entrepreneurship” and so forth. We restrict our analysis in what follows to those entrepreneurial activities that create new firms. In this paper, we define entrepreneurial processes in a given socio-economic context as all those activities, functions, and actions related to the perception of opportunities and the mobilization of resources related to generating new firms to pursue these opportunities (see Bygrave and Hofer 1991, among others, for a similar definition). Traditional theories of entrepreneurship basically restrict their attention to the profit-seeking motivation behind entrepreneurs. The neoclassical tradition considers market economies as systems in which equilibrium is achievable and represents them as such. The role of entrepreneurs is then merely a function of coordination of resources and calculation of the profit maximising output. Accordingly, the core of the theory focuses on ‘demand for entrepreneurship’, mainly determined by profit opportunities available in the market (Sorrentino 2003). In other words, the traditional explanation of entrepreneurial activities merely refers to the existence of some unexplored opportunities for profit (Eckhardt and Shane 2003). Knight (1921) improves this view by making a clear distinction between the roles of business manager and entrepreneur, and by highlighting the importance of the ability of entrepreneurs to act strategically in order to generate profits for their firms. The basic idea of these models rests on the choice of individuals to become entrepreneurs rather than being employed as workers. The higher expected income associated with entrepreneurial activity stimulates individuals to decide rationally to become entrepreneurs, and this decision depends on the individual’s self evaluation of their own entrepreneurial abilities. In this definition, there are no entrepreneurs in equilibrium. Only if the equilibrium conditions change because of exogenous factors (change in preferences and technology), then, will there be a role for entrepreneurs who adjust to the new economic situation and establish a new equilibrium (Glancey and McQuaid 2000). The Austrian view of the story is different. Market economies are in a constant state of flux: markets are seen as a discovery process in which a critical role is played by entrepreneurs in disequilibrium conditions (Kirzner 1997). Entrepreneurs are profit-seeking speculators characterised by “[. . . ] superior knowledge” that enables them to “[. . . ] benefit from the ignorance of others” (Glancey and McQuaid 2000). Disequilibrium is characterised by the presence of some kind of imperfect knowledge: some actors are better informed than others who lack full comprehension of market conditions. Entrepreneurs are able to discern price differentials in this situation and, exploiting arbitrage conditions, to take advantage of market opportunities to make profits. Schumpeter, on the contrary, considers economic change as endogenously determined. The exploitation of unnoticed market opportunities is not Schumpeter’s focus of analysis: it is the creation of new possibilities that represents the real
98
C. Garavaglia, S. Breschi
essence of the entrepreneurial behaviour. Entrepreneurs are revolutionary agents of change who drive the evolution of the economy (‘creative destruction’). Entrepreneurial action typically takes the form of innovation and profit opportunities are created by entrepreneurial innovation: “[. . . ] the new commodity, the new technology, the new source of supply, the new type of organization” (Schumpeter 1942) determine the opportunity for entrepreneurs to disrupt the market and gain profits. The core of the analysis, here, clearly rests on the genius, the personality and the talent of the innovator-entrepreneurs. Entrepreneurship, then, has a critical role for the evolution of economies and knowledge development. In other words, entrepreneurship is considered as an endogenous mechanism which overturns the status quo, and by trying out new ideas, generates the dynamics and the evolution of the economic system (Glancey and McQuaid 2000; Fumagalli 1990; Wu 1989). To sum up, in this context, the core of the analysis relates mainly to the ‘supply of entrepreneurship’, given by the ability to discern speculative opportunities, the innovative attitude of the entrepreneurs and the creative behaviour of individuals. This short review of the literature about the meaning of entrepreneurship highlights a few important points. First of all, understanding entrepreneurship requires that one identifies and explains the nature of the opportunities available to entrepreneurs. Second, entrepreneurship is a dynamic process, a key aspect of which is the development of the entrepreneur’s knowledge and ability to perceive the opportunities and act upon this perception (Verheul et al. 2002). Third, entrepreneurship takes place within a well-defined spatial context since it both influences the evolution of the surrounding area and is influenced by its characteristics. It follows that, on the one hand, entrepreneurs are active agents who stimulate the local context to further innovate and develop a system of localised learning (Feldman and Francis 2004a). On the other hand, the economic context, and more significantly the local context, influences the conditions in which the opportunities and knowledge emerge and thus stimulate the formation of new firms. To conclude, entrepreneurship is a much more complex phenomenon than simple profit-seeking behaviour. Beside pecuniary incentives, many other factors play a key role in influencing the entrepreneurial activity. Accordingly we now develop our analysis in our following discussion by examining some concepts that we believe to be important.
2.2 Behind the Entrepreneurial Process Entrepreneurs display idiosyncratic patterns in identifying and pursuing opportunity. A possible explanation lies in opportunity recognition and development processes being directly influenced by (Ardichvili et al. 2003): (1) (2) (3) (4)
The type of opportunities; Entrepreneurial alertness and personality traits; Information asymmetry and prior knowledge; Social networks.
The Co-Evolution of Entrepreneurship and Clusters
99
2.2.1 Type of Opportunities Opportunities may be defined as the chance to meet undefined market needs or underutilised or unemployed resources and capabilities (Kirzner 1973; Ardichvili et al. 2003). Opportunity identification involves processes of development, perception, recognition, creation and evaluation that overlap and interact with each other. Traditional theories (Kirzner 1973) see opportunity development and recognition as processes of recognising and acting upon something already formed. Other more recent approaches (Ardichvili et al. 2003) consider these concepts to be more complex and dynamic, that become more articulated as the entrepreneurs develop them. For our purposes it is also important to recognise that there may be different types of opportunities relating to the degree of identification of market needs and valuecreation capabilities. These categories influence the way opportunities are identified and developed by entrepreneurs. On the one hand, market needs (problems) may be known or unknown. On the other hand, value-creation capabilities are identified by financial, physical and intellectual resources and may translate into defined or undefined capability solutions. When known market needs combine with specified capability solutions, the opportunity development requires the process of matching identified resources with needs in order to create businesses that create value. The better the needs and solutions are known, the more likely will be the generation of new and successful firms (Ardichvili et al. 2003). As argued above, it is clear that needs and solutions appear to be linked strictly to the local area, similarly to opportunities for innovations and their related solutions, which originate in many cases from efforts directed at matching local customer requests as well as from their feedbacks.2
2.2.2 Alertness As argued above, entrepreneurs display heterogeneous attitudes in responding to the perception of opportunities. The term “alertness” refers to the entrepreneurial ability of responding to opportunities. Both personal traits and the characteristics of the local context (its history, the system of values, the ease of communication. . . ) contribute in shaping the conditions that influence entrepreneurial alertness. Higher alertness translates into greater ability to recognise opportunities. The role of entrepreneurs clearly departs from the role of business managers because of the important capabilities of the former to strategically recognise and act upon opportunities in order to generate profits (Knight 1921; Glancey and McQuaid 2000). This behaviour is both the outcome of deliberate conscious techniques and the result of intuitive, unpremeditated and spontaneous decision-making rules linked to the entrepreneur’s personal traits. 2
Lissoni (2001) presents the case of the Brescia mechanical engineering cluster, in which specific requests of key customers and their feedback were crucial in developing innovations in the cluster.
100
C. Garavaglia, S. Breschi
2.2.3 Information Asymmetry and Prior Knowledge Following the Austrian tradition, the existence of entrepreneurship is explained by the disequilibrium conditions of the market, which is characterised by the presence of some kind of imperfect knowledge and information asymmetry in which entrepreneurs are better informed than others who are deficient in comprehension of market conditions, and are therefore are able to take advantage of market opportunities to achieve profits (Kirzner 1973). Prior knowledge, then, generates a filter in the opportunity recognition and discovery processes (Shane 1999). The decision to start a new firm clearly depends on the founder’s prior experience, human capital and social ties. Pre-entry experience and knowledge affect the mode of entry, which markets are entered and the timing of entry (Helfat and Lieberman 2002). Moreover, the features of the founder determine the characteristics of the firm’s initial strategy (Boeker 1989; Burton et al. 2001).
2.2.4 Social Networks Active interaction with a network of people improves an entrepreneur’s ability to identify opportunities (Granovetter 1973; Hills et al. 1997). Moreover, the network ties improve entrepreneurial alertness and creativity (De Koning 1999). Granovetter’s concept of weak ties represents the possibility of acquiring and identifying relevant information by interacting with a network of people. An entrepreneur who is part of a group (usually a family or an ethnic group) faces the uncertainties of such activity with the security of being embedded in a social network that shares common values, beliefs and culture. Such support is also crucial when introducing innovations: in that case, establishing links and relationships with other individuals in a certain group enables the entrepreneur to understand the direction that will receive legitimisation as time goes by and the innovation becomes the natural way of doing things. Potential new entrepreneurs in the early phases of the evolution of an industry, a cluster or an innovation, face start-up problems: they lack informed customers, efficient suppliers, availability of capital and services, a skilled workforce (Van Wissen 2004). The development of a social structure of the cluster (Carroll and Hannan 2000), which creates a network of firms and people with a common social background, helps in overcoming the problems and stimulating entrepreneurial actions.
2.3 Factors Shaping the Supply Of and Demand for Entrepreneurship Some models that develop entrepreneurship as a complex phenomenon have been proposed in the literature. The main conclusion of the discussion above emphasizes the need to consider both ‘supply’ and ‘demand’ factors in understanding
The Co-Evolution of Entrepreneurship and Clusters
101
the entrepreneurship process (see also Storey 1994; Blanchflower 2000; Verheul et al. 2002). The OECD Employment Outlook (2000) also claims that selfemployment decisions are a function of market conditions as well as skills and spirit of potential entrepreneurs. Verheul et al. (2002), Audrestch (2002), Van Stel et al. (2003, 2004) and Shane (2003) propose a model in which the rate of entrepreneurship is determined by micro and macro factors. Although it is the individual who in the end chooses to become an entrepreneur or not after evaluating the risks and rewards associated with entrepreneurial activity, there is a broad range of social, economic, technological, political and cultural factors which crucially influence the supply and demand factors for entrepreneurship. At the micro level, the individual weighs alternative types of occupational choice. Alternative forms of employment, wage-employment and unemployment, are compared with entrepreneurial self-employment. The individual’s risk-reward profile represents the process of evaluation of these alternatives and results, in the end, in the entrepreneurial decision. On the one hand, we recognize the opportunities generated by the demand side, i.e. the demand for goods and services. On the other, the supply side determines both the characteristics of a given population and the abilities, preferences, personal traits, available resources of individuals, which finally define the supply of potential entrepreneurs. Supply and demand conditions determine the grounds on which individual choices are made. We are able to ascertain the rate of nascent entrepreneurship by aggregating individual choices. The supply side characteristics generate the conditions for the entrepreneurial decisions of individuals, thus creating the propensity to entrepreneurship among individuals of a given population. On the supply side it is also crucial to analyze the entrepreneurial phenomenon at the individual-micro level. Personal traits, preferences, values, attitudes, beliefs, abilities, skills, educational level of the individual are essential factors that ultimately influence the decision of the potential entrepreneur to identify and capture the opportunity created by the market. Our first claim consists of the recognition of the existence in clusters of both supply and demand-side factors that shape the entrepreneurial activities that we discuss in sect. 4.
3 Agglomeration Economies and External Positive Spillovers There has been a resurgence of interest in the agglomeration of firms in defined geographical clusters in the economic literature recently.3 The emergence of industrial clusters has been widely analysed and discussed in a Marshallian sense. 3
Despite substantial research on clusters, there is still much confusion and lack of consensus about what a geographical cluster is and how it should be conceptualised. Concepts like spatial agglomeration, clusters, industrial districts, innovative milieu, local and regional innovation systems, industrial complexes are often used in a rather interchangeable way by different researchers.
102
C. Garavaglia, S. Breschi
Traditionally, the availability of a skilled and specialized workforce, the localization of specialized suppliers and the ease of transmission of knowledge and information flows have been considered the most relevant causes for the existence of “external economies” in a region. Firms want to locate in a local area where they are likely to find the specialized skilled workforce they need. A geographically concentrated industry, then, gives firms an advantage by offering them highly skilled human capital. Consequently, employees logically move to areas where employers look for such specific skills, contributing to the self-reinforcement of this process. Moreover, customer firms and suppliers gain by locating close to each other both because of savings in transportation costs and because of backward and forward linkages that generate positive feedbacks. Finally, the process of clustering enables the firms to profit from some knowledge diffusion. “Mysteries of trade become no mystery and knowledge is in the air”, after the well-known Marshallian saying. Also the concept of “local technological infrastructure” (Carlsson and Stankiewicz 1991; Black 2004) indicates how networks of economic agents that interact with each other represent the source of agglomeration and knowledge spillovers, which in turn foster the development of the local area and the rate of innovative activities (Bramanti and Fratesi 2009, in this book). In particular, the existence and magnitude of ‘localized knowledge spillovers’ has received much attention in recent years (Breschi and Lissoni 2001). The role of innovation opportunities as a locational factor has been forcefully stressed by a vast and heterogeneous literature, comprising some influential case studies on US high-tech clusters (Saxenian 1994), Italian industrial districts, and local innovation systems (innovative milieux) in Europe and elsewhere (Markusen 1996; Keeble and Wilkinson 1999). A common theme linking these various contributions is the claim that knowledge relevant for innovative and entrepreneurial activities is mostly tacit and that therefore knowledge transmission is a matter of face-to-face contacts and labour mobility, i.e. that the most important knowledge carriers are people, in particular people who know and possibly trust each other, meet frequently and trade job offers very often. To the extent that having access to relevant knowledge inputs requires co-location, it becomes quite obvious to expect that the propensity for innovative and entrepreneurial activities to cluster spatially will be highest in industries in the early stages of their industry’s life cycle and where tacit knowledge plays an important role (Audretsch and Feldman 1996). The recent contributions of the “new economic geography” (Krugman 1991a, b; Puga and Venables 1996; Ottaviano and Thisse 2001, 2004) are more sceptical about the role of knowledge externalities and tend to explain the processes of geographic concentration of firms on the basis of pure pecuniary externalities, especially those related to the labour market and to the demand side. Rather than focusing on agglomeration economies external to the firms, this literature stream points out the role of increasing returns to scale internal to the firms, together with the importance of transport costs and labour mobility across regions. The emergence of clusters is thus explained as a cumulative and self-reinforcing process in which the location In this chapter, we adopt a broad and general notion by defining a geographical cluster as a non-random spatial concentration of firms and economic activities.
The Co-Evolution of Entrepreneurship and Clusters
103
of new firms in a region attracts workers from other areas to benefit from relatively higher wages; this in turn stimulates the further entry of new firms into the same region to benefit from a larger local market and save on transport costs. The basic point to notice in this cumulative process is that the emergence of clusters is entirely explained on the basis of the locational choices made by entrepreneurs and that nonmarket external economies, such as knowledge spillovers, do not play any role in driving such locational choices. Even though the arguments put forward by the “new economic geography” have been much criticised by most traditional regional economists and economic geographers (who have reacted by emphasising the importance of non-market mechanisms and social interactions as crucial explanatory factors of the emergence of spatial clusters of firms), a review of this debate is beyond the scope of this paper (for a review, see Breschi and Lissoni 2001). Here we limit ourselves to observing that the process of firms clustering in a local area may be considered the result of dynamic agglomeration economies, both pure and pecuniary ones: economies of scale and spillovers are “[. . . ] generated by the localisation in the same geographically bounded space as other firms working on similar technologies or products” (Feldman and Francis 2004a). Moreover these forces follow an evolutionary path. The process of firms agglomerating shows path-dependence and lock-in. Selfreinforcing mechanisms operate in the direction of intensifying the advantages of locating in the same area. Once established, the agglomeration process seems to be robust to changes. The cumulative nature of knowledge and its stickiness to a local area suggest the existence of relevant differences in clustering and competitiveness over space from region to region (Fritsch 2004).
4 What Influences the Entrepreneurial Activity In a Cluster? The emergence of clusters of firms is a cumulative process that involves both the supply and the demand side. In particular, we claim that the entrepreneurial activities are the real engine of cluster formation, growth and persistence. Entrepreneurship and clusters are interrelated factors that co-evolve over time. In this section we focus on what we believe to be the relevant factors that influence entrepreneurship in a cluster. Obviously, each cluster has its own history and the factors we study may play a critical role under given conditions while not being significant in other circumstances. Researchers, then, need to investigate the evolution of any given cluster with regard to its own specificities, in order to identify the factors in play.
4.1 External Pecuniary Economies Whatever the reasons behind the existence of external economies, the theoretical contributions of the traditional economic geography approach claim that the
104
C. Garavaglia, S. Breschi
economic opportunities and positive externalities generated in a local area stimulate new firms to localise in that area, thus reinforcing the agglomeration process and the persistence of clusters. According to this view, the formation of clusters is explained with regard to pure economic factors. The existence of external pecuniary economies is able to explain endogenously the agglomeration of firms independently of the exogenous natural and physical characteristics of the location (Ottaviano and Puga 1998; Rosenthal and Strange 2004; Ottaviano and Thisse 2001). The level of transport costs, the increasing returns to scale at the plant level and factor mobility, involving a circular causation mechanism, play a critical role in triggering the process of agglomeration of firms. The original family of models of the “new economic geography” assume the existence of two identical regions, two sectors (“manufacturing” producing under returns to scale technology and “agriculture” producing under constant returns to scale) and two types of input, one of which is employed in manufacturing and is mobile across regions (workers) and the other (land) is utilised in agriculture and is immobile. A large number of producers in a region implies a large number of product varieties produced, and consequently a rise in the demand for labour and wages thus inducing workers to move to this region. The concentration of workers (who are also consumers) generates a larger market in this region (the “home market effect”). Because of economies of scale and transport costs, producers benefit from concentrating production in only one region and shipping the product to the other. Forward linkages due to the incentive for workers to be where the producers of final goods are located and backward linkages that represent the incentive of producers to cluster in the larger market create a centripetal force that results in firms and workers clustering in a region (Krugman and Venables 1990; Krugman 1991a, b; Ottaviano and Puga 1998; Ottaviano and Thisse 2001, 2004). According to this line of research, the essence of cluster formation is then the existence of these pecuniary externalities, while other non-pecuniary factors are believed to be too vague (Krugman 1991a).
4.2 Opportunities and Demand Following on from the discussion above, in this subsection we emphasise the role of demand and the opportunities related to demand. To keep it simple, let us think about the classical Hotelling model (1929). Under given conditions, namely little differentiation and low transport costs, the outcome of two firms consists in clustering in the same point in space. This result is due to the so-called “direct effect” (or “market area effect”) that directly relates the location choice of the firm to that portion of the market to which the firm has privileged access (Ottaviano and Thisse 2004). Both the emergence and growth of clusters are related to some kind of demand that will sustain the process of development of firms. The importance of the existence of a sizeable and growing source of demand in sparking off the growth of a cluster has been analysed by Bresnahan et al. (2001) in research into the sources of success of regional clusters of innovation in a number of international cases. The
The Co-Evolution of Entrepreneurship and Clusters
105
most common similarity among the examples under examination was the connection between cluster development and the existence of a new technological market opportunity that had not been exploited before. This explains the emergence of clusters in Ireland, India, Israel and Taiwan which were complementary to the existing ICT technologies. We claim that the existence of opportunities and demand also play a critical role in the development phase of a cluster. Market needs are usually well known in a cluster and profit-seeking entrepreneurs create new firms, both in competition and cooperation with established firms, to satisfy these needs. Opportunities often develop locally. Entrepreneurs identify the opportunities in a local area and exploit them in order to generate profits. Often, the development of firms and activities in a local area creates new opportunities for other new businesses in connection with the support and complementary activities that are not directly developed by the existing firms but purchased externally. Italian industrial districts are a clear example of a highly fragmented production process. Firms specialise in one activity of the production chain. In fact, firms tend to cluster in those areas where complementary activities and industries that share common knowledge are relevant (Feldman and Audretsch 1999). Most of the new firms in a cluster are subcontractors of established firms. In some cases, it is the incumbents themselves that finance the start up of former employees in order to economize on previous investments in some related activities, thus generating new opportunities (Hanson 1996). What emerges is a population of vertically disintegrated small firms. Moreover, technological progress and inventions are often highly localised due to the nature of knowledge creation and its initial applications (Feldman and Francis 2004b). Technological change is directly regarded as a driving force in determining the demand for entrepreneurship (Wennekers and Thurik 1999). Developments in technology stimulate the demand for entrepreneurship through the incentive to develop new products and services (Casson 1995).
4.3 Information and Knowledge Spillovers Tacit knowledge is a non-rival good that is best transmitted through frequent personal contact, so the higher the frequency of interaction, the lower the marginal cost of transmitting knowledge. The creation of a local industrial environment facilitates such interaction and the flows of information and ideas in the spatial structures. Spatial proximity of knowledge owners is crucial for the transmission of tacit and sticky knowledge (Von Hippel 1994; Acs and Varga 2004). The individual decision to become an entrepreneur and explore new opportunities is influenced by the degree of uncertainty of this choice. Figueiredo et al. (2002) find evidence of information costs as a centripetal force of agglomeration: clustering reduces uncertainty and information costs and thus influences the location decisions of new entrepreneurs. In fact, the spreading of information and knowledge at the local level contributes to lowering the risk and uncertainty related to entrepreneurial activities.
106
C. Garavaglia, S. Breschi
Accordingly, an informational cascade effect may influence the process of development of new firms in a cluster. Entrepreneurs who initially may be pessimistic about the future profitability of their entry may decide to enter the market by observing the experience and the performance of established firms. As more firms enter, more information becomes available to potential entrants thus speeding up the resolution of uncertainty and stimulating additional entry (Horvath et al. 2001). Moreover, skilled labour and the proximity to sources of knowledge and expertise are crucial for small firms and high technology firms. Small firms in particular lack the resources of large firms and are thus more dependent on resources in their local environment (Feldman and Francis 2004a). In industrial clusters, small firms are able to reduce their size disadvantage as they benefit from localised knowledge spillovers (Pyke and Sengenberger 1990). In this context, entrepreneurship may be considered as the vehicle that drives the spillover of knowledge in the process of economic evolution (Audretsch and Feldman 2003; Audretsch and Keilbach 2004). Entrepreneurship is the mechanism that governs the selection process by which economically exploitable knowledge emerges from a diversity of ideas. New firms, then, represent a source of generation of diversity, knowledge spillovers and new approaches that otherwise would not have been explored. Those regions and clusters that exhibit higher rates of entrepreneurial activity, then, also show higher degrees of competitiveness and growth rates. Empirical evidence on knowledge spillovers is ambiguous, however. Some recent contributions (Breschi and Lissoni 2001; Gordon and McCann 2000) refer to overestimation of knowledge spillovers in explaining cluster formation. The most important point to note is that most of the econometric literature that has attempted to estimate the existence and magnitude of localised knowledge spillovers has not been able so far to shed light on the actual mechanisms by which knowledge is transmitted. As a consequence, most of the results obtained on the spatial concentration of innovative and entrepreneurial activities might be observationally equivalent to other explanations, particularly those that deny any importance to knowledge spillovers and attribute the role to more traditional market-based pecuniary externalities arising in the labour market and in the market for intermediate inputs. Proximity may not imply any significant advantage from knowledge spillovers. The reverse may be true: innovative activity (knowledge) is geographically concentrated in some industries simply because the location of production is concentrated and not vice versa (Audretsch and Feldman 1996).
4.4 External Non-Pecuniary Factors In addition to external purely pecuniary economies, the theories of industrial districts emphasise the role of socio-economic and cultural aspects in influencing the competitiveness and growth of a given area (Becattini 1987, 1989; Brusco 1989; Sforzi 1990). The Italian tradition of industrial districts developed on a Marshallian tradition. Industrial districts are defined as a socio-territorial entity characterised by
The Co-Evolution of Entrepreneurship and Clusters
107
the coexistence of a community of people and a set of firms in a well-specified area (Becattini 1989). A shared system of values and thought, combined with a local entrepreneurial culture, shapes the industrial atmosphere and facilitates relationships between firms, people and institutions. This field of research emphasises social relationships in a cluster, but it mainly confines itself to the behaviour and values of the local community and the institutions that constitute the industrial environment in which the productive system is embedded. According to these contributions, the local structures of the social community and institutions increase the willingness and the opportunities to become self-employed (Bagnasco and Trigilia 1984, 1985; Becattini, 1997; Tappi 2000). Moreover, the cultural homogeneity of the district favours cooperative and trusting behaviour among people and firms thus lubricating social relationships, consensus and loyalty among economic and non-economic actors. In this context, economic activities are coordinated by both explicit and tacit rules and social conventions (Lazerson and Lorenzoni 1999a; Colli 1999). Public institutions and private organisations contribute to enforcing these mechanisms. On the other hand, this cultural homogeneity presents some drawbacks. The exploration of new technologies, new investments and new ideas become limited in such contexts. Inertial effects of lock-in may cause firms to stick to the existing knowledge, skills and technology and to move along a sub-optimal path (Nelson and Winter 1982; Hannan and Freeman 1984; Carroll 1988; Lazerson and Lorenzoni 1999a, b).
4.5 Entrepreneurial Alertness The literature on Italian industrial districts points out that most of the entrepreneurs in the districts were previously artisans or m´etayers and developed organisational skills, flexibility and the culture of reinvesting the revenues in business. The agricultural and artisan background of these areas generate a pool of people whose entrepreneurial alertness is particularly vivid. People’s personal traits in those areas where artisan activities and m´etayers were widespread developed distinct and reactive attitudes towards opportunity-recognition, self-employment behaviour, risk-taking, and investment inclination (Paci 1980; Forni 1987; Lazerson and Lorenzoni 1999a; Daumas 2006). In this sense, the local context contributes to shaping the pre-conditions that influence entrepreneurial alertness and the supply of entrepreneurship among people living in a given area and thus the consequent emergence of clusters.
4.6 Prior Knowledge In addition to the supply of entrepreneurship and alertness generated by the rural and artisan backgrounds of some areas, many researches highlight the importance
108
C. Garavaglia, S. Breschi
of prior technical knowledge for the take off and development of a cluster. Italian examples are full of situations in which both large vertical integrated firms as well as small artisan firms were the incubator of the specialised knowledge, technical and manufacturing skills for many employees.4 Many small firms were set up by larger incumbents after factory closures, either by deliberate strategy of the incumbents to create subcontractors or in a spin-off process by ex-employees. Italy is not unique. The Mexican garment industry shows a similar story (Hanson 1996). Firstly, the prior knowledge of some immigrants who had been textile and garment merchants in their countries of origin created the knowledge pre-conditions for the setting up of the first garment firms in the country. Secondly, in a process of industry dispersion, the relocation of the production in other areas was initiated by pioneer firms whose founders were former employees and had prior experience in the capital’s garment industry. Prior knowledge and experience may be gained from existing large firms, as mentioned above, as well as from public institutions, universities and research centres (Bresnahan et al. 2001). For example, Feldman et al. (2005) claim that in the case of the biotechnology and ICT cluster in the Washington DC/State area in the USA, the skilled labour originated from government institutions, federally-funded laboratories and agencies.
4.7 Exposure Effect In a local area where a given type of industrial activity is concentrated and where the percentage of self-employed people is accordingly higher than in other areas, the supply of entrepreneurship may become a self-sustaining process. At the micro level, the individual weighs alternative types of occupational choice. Alternative forms of employment, wage-employment and unemployment are compared with entrepreneurial self-employment. The individual’s risk-reward profile represents the process of evaluation of these alternatives and results in the the entrepreneurial decision. We claim that, according to the “mere exposure effect”, the supply of potential entrepreneurs in clusters is higher than in other areas, ceteris paribus. A lot of psychology research has been done on the consequences of the “mere exposure effect” (Zajonc 1968). This suggests that the more an individual is exposed to a stimulus, the more he will tend to appreciate it. People develop their tastes and preferences over time with repeated exposure, and when asked to take decisions will usually choose the familiar over the unfamiliar. We claim that in clusters, other things being equal, people are exposed to entrepreneurial activities more often than in other areas, so when individuals weigh alternative types of occupational choice, they tend to choose entrepreneurial self-employment more often. Therefore, we find both more potential entrepreneurs and higher supply of entrepreneurship in clusters than in other areas. 4
Among others see the cases of Prato, Castel Goffredo, Carpi, Belluno, Lumezzane, Manerbio, Putignano, Teramo, Murgia, Barletta, Corsano, Lavello, Lecco (Lazerson and Lorenzoni, 1999a; Lazerson and Lorenzoni, 1999b; Viesti, 2001; Colli, 1999).
The Co-Evolution of Entrepreneurship and Clusters
109
4.8 Local Social Networks A recent wave of studies argues that agglomeration economies and spillovers of ideas alone can not explain the location of new entrepreneurs (Klepper 2004; Sorenson 2003). According to these authors, entrepreneurial activities require the mobilisation and the organisation of the resources necessary to start a new firm. In this respect, geography and location matter because they help in establishing and maintaining those network ties that are vital to mobilise the resources needed to found a new venture. In fact, individuals do not interact with other agents at random, but tend to form social relationships with others living in the same region and with whom they share a common culture, interests and background and maintain these over long periods. These local social networks in turn play a crucial role in deciding both who engage in entrepreneurship and which regions are most likely to attract new firms. In the first place, social ties affect the awareness of entrepreneurial opportunities. As argued above, evaluating market opportunities is often a first step in the entrepreneurial process and it requires access to private information and data. Social links are a powerful means of accessing this valuable data, by connecting potential entrepreneurs with other individuals working in the same industry, e.g. employees of incumbent firms, and living in the same local community. Once an opportunity has been identified, social networks also constrain where individuals can build new firms successfully. The success of a new company depends on having access to some key factors, notably human capital, financial capital, knowledge capital and trust. Once again, social relationships are likely to increase the likelihood of mobilising the financial and human capital needed for its operations and bringing them to the nascent firm. Given the high risks involved and the fundamental information-asymmetry problem afflicting new ventures, entrepreneurs may find it difficult to collect sufficient financial and human capital. Strong and close social ties help to overcoming this constraint by reducing the uncertainty and creating the trust necessary to convince investors and potential employees of the real prospects of the nascent firm. Typically, these aspects could not be easily replicated outside the local community, outside the geographically bounded area of the cluster (Figueiredo et al. 2002). An important consequence of this perspective on entrepreneurship and clustering is that locational inertia is likely to emerge. The need to draw on social networks to identify entrepreneurial opportunities and to mobilise financial and labour resources, together with the fact that individuals prefer to remain close to family and friends, tend to bind potential entrepreneurs to the regions in which they have contacts, i.e. the regions in which they have worked and lived, even though other locations seem to be more attractive. This implies that once clustered in a given location, industry tends to remain located in the same area, even when co-location implies disadvantages. Empirical evidence by Figueiredo et al. (2002) shows that entrepreneurs display a strong tendency to locate in their family home environment and will pay employees more than three times the going rate before considering moving from their home base on a par with staying.
110
C. Garavaglia, S. Breschi
5 Entrepreneurship and Cluster Formation The picture that emerges from the discussion above tells us that the process of cluster formation and development is a complex and multifaceted phenomenon. Supply as well as demand-side factors are in play. Economic incentives as well as sociological and psychological explanations may shed some light on the geographical agglomeration processes of firms. Both exogenous and endogenous forces shape and trigger cluster formation and persistence. Various strands of literature focus more on some of these aspects than others. What we claim here is that the truth is probably in between and, more importantly, that entrepreneurship seems to be the real engine of cluster formation. Feldman and Francis (2004a) and Feldman et al. (2005) propose a three-stage model5 of cluster formation in which entrepreneurs shape and determine the evolution and success of a cluster. An exogenous shock to the system of production, such as downsizing, lay-offs, exogenous policies, corporate mergers and acquisitions, or bankruptcy of large firm (see also Leslie and Kargon 1994; Lazerson and Lorenzoni 1999a; Viesti 2001), which lowers the opportunity cost of entrepreneurship, sparks off the latent entrepreneurship in a local area. A complex self-regulating process emerges in which new entrepreneurial experiences and agglomeration economies move symbiotically. At least three aspects of this process are worth noting. First of all, agglomeration external economies (such as the presence of venture capitalists, infrastructure, services, higher education, common interest organisations) are not pre-requisites for cluster emergence but rather the outcome of multi-agent interaction in an area that further sustains and incentivates subsequent entrepreneurship. Secondly, an entrepreneurial tradition is not a pre-condition for cluster emergence in the earliest stage of cluster formation but skilled labour and intellectual experience is. Thirdly, the process of entrepreneurial spin-offs plays a key role in the development phases of the cluster. Increased opportunities, local recognition, lower risk and serial entrepreneurs give birth to new firms and activities. In this process, then, an informational cascade effect and trial-and-error experiences offer a learning mechanism that eases subsequent entrepreneurship. The local aspect of entrepreneurship is determining in influencing cluster formation. People tend to start new firms in the locations where they have social ties and business networks (Feldman 2001; Feldman and Francis 2004a; Sorenson 2003; Stuart and Sorenson 2003; Sorenson and Audia 2000; Figueiredo et al. 2002). In the absence of the explicit combination of this important sociological and psychological phenomenon with the economic incentives and opportunities, we believe that cluster formation cannot be explained. We then conclude that the dynamic process of cluster formation is considerably influenced by both the economic externalities and the social system in terms of institutions, state policies, membership organisations and supporting local environment in determining a self-sustaining system, and
5
Stam (2005) also distinguishes different phases in the spatial development of new firms.
The Co-Evolution of Entrepreneurship and Clusters
111
in particular we believe the role of social ties in determining the location choice of entrepreneurs is crucial. To advance our argument, we believe that individuals have their own history and experience before actually becoming entrepreneurs (Freeman 1986). Therefore, the decision to become an entrepreneur depends on the personal experience, human capital and social capital of the individual and on the external culture and system of values (Colli 1999; Helfat and Lieberman 2002). It follows that particular spin-offs bring with them the memories, experience, tacit knowledge, routine and characteristics of the founder. Spin-off formation and its links to the social capital of the founder combined with a hereditary theory of organizational competence may well explain cluster formation even without reference to any kind of external pecuniary economies. As stated above, some recent authors sustain that agglomeration economies and knowledge spillovers are insufficient explanation of the location of new entrepreneurs and focus attention on social networks and spin-off formation in explaining the geographical agglomeration processes of firms (Klepper 2004). Looking at the growth of the automobile and television receiver industries in US, Klepper (2004) claims that the existence of external economies can hardly explain the evolution patterns of these industries. On the one hand, the automobile producers were geographically dispersed initially. Over time the industry evolved into an oligopoly highly concentrated around only one city, Detroit. On the other hand, most of the firms in the television industry, from the start in the 1940s and 1950s were clustered around New York, Chicago and Los Angeles. After some years the industry became more concentrated in terms of market structure but remained dispersed geographically. These two patterns of evolution are hardly compatible with the existence of positive external economies that induce firms to agglomerate around a local area. There is an alternative explanation. In addressing the question of why industries agglomerate geographically, Klepper rejects the theories that feature increasing returns to scale –both internal and external to firms– and proposes an interpretative framework, based on the inheritance of organizational competences. Briefly, the explanation begins by assuming that knowledge of particular routines and specific technologies accounts for a much of the heterogeneity across firms in profitability and performance. Replicating this scarce and valuable knowledge is extremely difficult –if not impossible– for potential entrepreneurs whose background is in a different line of business or completely outside the industry. On the other hand, entrepreneurs from related industries or from the leading incumbents in the industry are likely to bring some the organizational competences of the originating companies from their pre-entry experience. Entrants having access to this knowledge and capability, then, enjoy a potentially large advantage over other firms. If it is accepted that the most competent entrants in an industry are those founded by entrepreneurs who were former employees of firms in the same or in a related industry, then it is possible to explain the clustering of firms within a few regions, even in the absence of locational effects.6 Given that social and economic forces induce entrants to locate 6
Please note that this theoretical approach is not necessarily in contrast with the conceptualisations of clusters and spatial agglomerations proposed in the regional and economic geography literature,
112
C. Garavaglia, S. Breschi
geographically close to their origins, regions that (by chance) contain a number of the early leaders will also host spin-offs generated by such companies. By virtue of their superior heritage, these spin-offs will outperform start-ups founded by inexperienced entrepreneurs, possibly located in other areas, thereby causing the industry to become heavily agglomerated around the regions of the early incumbents.
6 Conclusions This paper attempts to explain theoretically the factors that induce firms to cluster geographically. We claim that the process of cluster formation is highly complex and self-organising, involving both external pecuniary and non-pecuniary factors. The characteristics of the cluster develop over time and co-evolve in a symbiotic relationship with local entrepreneurial activities and supportive local institutions. Also, entrepreneurship is a multifaceted phenomenon related to pecuniary, sociological and psychological factors. In the paper we argue that it is crucial to investigate the co-evolution of entrepreneurship and the emergence of clusters in order to explain cluster formation. Different theoretical contributions explain the processes behind entrepreneurial activities. The relevant aspects emerging in a cluster that we believe to be important in influencing entrepreneurship are explicitly analysed. We start by considering external pecuniary economies and then proceed to analyse the role of opportunities and demand, information and knowledge spillovers, external non-pecuniary factors, entrepreneurial alertness, the importance of prior knowledge, the consequences of the “mere exposure effect”, and the role of local social ties. The picture that emerges identifies cluster formation and development as a dynamic and complex evolutionary process in which economic incentives as well as sociological and psychological factors may explain the geographical agglomeration of firms. In this process both exogenous and endogenous forces are in play. Various strands of literature focus more on some of these aspects than others. We believe here that the truth is probably in between. What we stress is that the real engine of cluster formation is related to local entrepreneurial activities: the ‘home bias’ effect of the locational choice of entrepreneurs in a cluster is the key force at the basis of its formation. Our work may be improved along different lines. First of all, other more systematic theoretical taxonomies of the relevant factors that influence entrepreneurship in a cluster may overlap with our categorisation and benefit from interaction. Further theoretical discussions are left for future research. More importantly, it is which stress the importance of external economies. The point is that in the early stages of emergence and formation of a new cluster, none of the conditions generating localisation economies is yet in place. In this respect, the entrepreneurial process has to be understood as the primary and most fundamental driver behind the birth and emergence of new clusters of economic activities. Once the cluster has reached a sufficiently mature stage, it may well be that conventional external economies set in and help to explain the differential performance across regions.
The Co-Evolution of Entrepreneurship and Clusters
113
necessary to recognise that each cluster has its own history and the factors we analysed may play a critical role under certain conditions while not being significant in other circumstances. Empirical evidence that witnesses the evolution of any given cluster with regard to its own specificities, is then needed to identify the factors in play. Finally, highlighting the relevant factors behind the entrepreneurial activity that spark off and sustain cluster development and growth is of prime interest to the policy maker. Moreover, the analysis of the complexity of the co-evolution of entrepreneurial activities and agglomeration mechanisms encourages the policy makers to re-examine their tool box (Marino and Trapasso 2009, in this book). We believe this topic deserves specific attention.
References Acs Z, Audrestch D (2003) Handbook of entrepreneurship research: an Interdisciplinary Survey and Introduction, Kluwer Academic, Dordrecht Acs Z, Varga A (2004) Entrepreneurship, agglomeration and technological change. Discussion Papers on Entrepreneurship and Public Policy, MPI Jena Ardichvili A, Cardozo R, Ray S (2003) A theory of entrepreneurial opportunity identification and development. J Bus Venturing 18:105–123 Audrestch D (2002) Entrepreneurship: a survey of the literature. Paper Prepared for the European Commission, Enterprise Directorate General, July Audretsch D, Feldman M (1996) Innovative clusters and the industry life cycle. Rev Ind Organ 11:253–273 Audretsch D, Feldman M (2003) Knowledge spillovers and the geography of innovation, In: The handbook of urban and regional economics, vol 4 Audretsch D, Keilbach M (2004) Entrepreneurship and regional growth: an evolutionary interpretation. J Evol Econ 14:605–616 Bagnasco A Trigilia C (1984) (a cura di) Societ`a e politica nelle aree di piccola impresa. Il caso di Bassano. Venezia, Arsenale Editrice Bagnasco A Trigilia C (1985) (a cura di) Societ`a e politica nelle aree di piccola impresa. Il caso della Valdelsa. Milano, Franco Angeli Becattini G (1987) Mercato e forze locali: il distretto industriale. Bologna Becattini G (1989) Riflessioni sul distretto industriale marshalliano come concetto socioeconomico. Stato e Mercato, 25 Becattini G (1997) Prato nel mondo che cambia (1954–1993). In: Prato storia di una citt`a, vol IV. Firenze, Le Monnier Black G (2004) The geography of small firm innovation. Kluwer Academic, Boston/Dordrecht Blanchflower D (2000) Self-employment in OECD countries. NBER Working Paper 7486, National Bureau of Economic Research Boeker W (1989) Strategic change: the effects of rounding and history. Acad Manage J 32:489–515 Bramanti A, Fratesi U (2009) The dynamics of an innovation driven territorial system. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Breschi S, Lissoni F (2001) Knowledge spillovers and local innovation systems: a critical survey. Ind Corporate Change 10:975–1005 Bresnahan T, Gambardella A, Saxenian A (2001) ‘Old economy’ inputs for ‘new economy’ outcomes: cluster formation in the new silicon valleys. Industrial and Corporate Change 10:835–860 Brusco S (1989) Piccola impresa e distretti industriali. Rosenberg & Sellier, Torino
114
C. Garavaglia, S. Breschi
Burton M, Sorensen J, Beckman C (2001) Coming from good stock: career histories and new venture formation. In: Lounsbury M, Ventresca M (eds) Research in the sociology of organizations, vol 19, Social Structure and Organizations Revisited, JAI, Greenwich Bygrave WD, Hofer CW (1991) Theorizing about entrepreneurship. Entrepreneurship Theory and Practice 16:13–22 Carlsson B, Stankiewicz R (1991) On the nature, function and composition of technological systems. J Evol Econ 1:93–118 Carroll GR (1988) Ecological models of organizations. Ballinger Publishing Company, Cambridge, MA Carroll GR, Hannan MT (2000) The demography of corporations and industries. Princeton University Press Casson M (1995) Entrepreneurship and business culture. Cheltenam, UK, Edward Elgar Publishing Colli A (1999) Legami di ferro. Storia del distretto metallurgico e meccanico lecchese tra Otto e Novecento. Meridiana Libri Daumas JC (2006) Districts Industriels: le Concept et l’Histoire. XIV International Economic History Congress, Helsinki De Koning A (1999) Conceptualizing Opportunity Recognition as a Socio-Cognitive Process. Centre for Advanced Studies in Leadership, Stockholm Eckhardt JT, Shane SA (2003) Opportunities and entrepreneurship. J Manage 29:333–349 Feldman MP (2001) The entrepreneurial event rivisited: firm formation in a regional context. Industrial and Corporate Change 10:861–891 Feldman MP, Audretsch D (1999) Innovation in cities: science-based diversity, specialization and localized competition. Eur Econ Rev 43:409–429 Feldman MP, Francis J (2004a) Fortune favors the prepared region: the case of entrepreneurship and the capitol region biotechnology cluster. Eur Plann Stud 11:765–788 Feldman MP, Francis J (2004b) Home grown solutions: fostering cluster formation. Econ Dev Q 18:127–137 Feldman MP, Francis J, Bercovitz J (2005) Creating a cluster while building a firm: entrepreneurs and the formation of industrial clusters. Reg Stud 39.1:129–141 Figueiredo O, Guimaraes P, Woodward D (2002) Home-field advantage: location decisions of portuguese entrepreneurs. J Urban Econ 52:341–361 Forni M (1987) Storie familiari e storie di propriet`a: itinerari sociali nell’agricoltura italiana del dopoguerra. Rosenberg & Sellier, Torino Fritsch M (2004) Entrepreneurship, entry and performance of new business compared in two growth regimes: east and west germany. J Evol Econ 14:525–542 Freeman J (1986). Entrepreneurs as organizational products: semiconductor firms and venture capital firms. In: Libecap G, (ed), Advances in the study of entrepreneurship, innovation and economic growth. Entrepreneurship and innovation: the impact of venture capital on the development of new enterprise. vol I, JAI Press, Greenwich, CT Fumagalli A (1990) L’Imprenditore nella Storia dell’Analisi Economica. In: (a cura di) Mussati G (ed) Alle Origini dell’Imprenditorialit`a. La Nascita di Nuove Imprese: Analisi Teorica e Verifiche Empiriche. Etaslibri, Milano Glancey KS, McQuaid RW (2000) Entrepreneurial economics. Macmillan Press, London, UK Gordon I, McCann P (2000) Industrial clusters: complexes, agglomeration and/or social networks?. Urban Stud 37:513–532 Granovetter M (1973) The strength of weak ties. Am J Sociol 78(6):1360–1380 Hannan M, Freeman J (1984) Structural inertia and organizational change. Am Sociol Rev 49: 149–164 Hanson G (1996) Agglomeration, dispersion, and the pioneer firm. J Urban Econ 39:255–281 Helfat C, Lieberman M (2002) The birth of capabilities: market entry and the importance of prehistory. Ind Corporate Change 11(4):725–760 Hills G, Lumpkin GT, Singh RP (1997) Opportunity recognition: perceptions and behaviors of entrepreneurs. Frontiers of entrepreneurship research. Wellesey, MA, Babson College
The Co-Evolution of Entrepreneurship and Clusters
115
Horvath M, Schivardi F, Woywode M (2001) On industry life-cycles: delay, entry, and shakeout in beer brewing. International J Ind Organ 19:1023–1052 Hotelling H (1929) Stability in competition. Econ J 39:41–57 Keeble D, Wilkinson F (1999) Collective learning and knowledge development in the evolution of regional clusters of high technology smes in Europe. Reg Stud 33:295–303 Kirzner IM (1973) Competition and entrepreneurship. University of Chicago Press, Chicago Kirzner IM (1997) Entrepreneurial discovery and the competitive market process: an Austrian approach. J Econ Lit 35(1):60–85 Klepper S (2004) The geography of organizational knowledge. Mimeo Knight FH (1921) Risk, uncertainty and profit. Houghton Mifflin Company, Boston, MA Krugman P (1991a) Increasing returns and economic geography. J Polit Econ 99(3):483–499 Krugman P (1991b) Geography and trade. MIT Press, Cambridge Mass Krugman P, Venables A (1990) Integration and the competitiveness of peripheral industry, In: Bliss C, Braga de Macedo J (eds) Unity with diversity in the European community. Cambridge University Press Lazerson M, Lorenzoni G (1999a) The firms that feed industrial districts: a return to the italian source. Ind Corporate Change 8:235–266 Lazerson M, Lorenzoni G (1999b) Resisting organizational inertia: the evolution of industrial districts. J Manage Governance vol 3, December Leslie S, Kargon R (1994) Electronics and the geography of innovations in post-war america. Hist Technol 11:217–231 Lissoni F (2001) Knowledge codification and the geography of innovation: the case of brescia mechanical cluster. Res Policy 30:1479–1500 Marino D, Trapasso R (2009) A new approach to regional economics dynamics - Path dependence and spatial self reinforcing mechanisms. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Markusen A (1996) Sticky places in slippery space: a typology of industrial districts. Econ Geogr 72:293–313 Nelson RR, Winter S (1982) An evolutionary theory of economic change. Harvard University Press, Cambridge MA OECD Employment Outlook (2000) Directorate for employment, labour and social affairs Ottaviano G, Puga D (1998) Agglomeration in the global economy: a survey of the ‘new economic geography’. World Econ 21:707–731 Ottaviano G, Thisse JF (2001) On economic geography in economic theory: increasing returns and pecuniary externalities. J Econ Geogr 1:153–179 Ottaviano G, Thisse JF (2004) New economic geography: what about the N? CORE Working Paper, January Paci M (1980) (ed) Famiglia e mercato del lavoro in un’economia periferica. Franco Angeli, Milano Pyke F, Sengenberger W (1990) Introduction. In: Pyke F, Becattini G, Sengenberger W (eds) Industrial districts and inter-firm co-operation in italy. International Institute for Labour Studies, Geneva, pp 1–9 Puga D, Venables AJ (1996) The spread of industry: spatial agglomeration in economic development. CEPR Discussion Paper Series, 1354, London Rosenthal SS, Strange WC (2004) Evidence on the nature and sources of agglomeration economies. In: Henderson JV, Thisse JF (eds) Handbook of regional and urban economics. North Holland Saxenian A (1994) Regional advantage. culture and competition in silicon valley and route 128, Harvard University Press Sforzi F (1990) The quantitative importance of marshallian industrial districts in the italian economy. In: Pyke F, Becattini G, Sengenberger W (eds) Industrial districts and inter-firm co-operation in italy. International Institute for Labour Studies, Geneva, 1–9 Shane S (1999) Prior knowledge and the discovery of entrepreneurial opportunities. Organ Sci 11(4):448–469
116
C. Garavaglia, S. Breschi
Shane S (2003) A general theory of entrepreneurship: the individual-opportunity nexus. Edward Elgar Schumpeter JA (1942) Capitalism, socialism and democracy. New York, Harper & Row Sorenson O (2003) Social networks and industrial geography. J Evol Econ 13:513–527 Sorenson O, Audia P (2000) The social structure of entrepreneurial activity: geographic concentration of footwear production in the united states, 1940–1989. Am J Sociol 106:424–462 Sorrentino M (2003) Le nuove imprese. economia delle nuove iniziative imprenditoriali. Cedam Stam E (2005) Why butterflies don’t leave: spatial development of new firms. Paper Presented at EMAEE Conference, Utrecht 2005 Storey D (1994) Understanding the small business sector. London/New York, Routledge Stuart T, Sorenson O (2003) The geography of opportunity: spatial heterogeneity in founding rates and the performance of biotechnology firms. Res Policy 32:229–253 Tappi D (2000) Evolutionary path of italian industrial districts: the importance of formal and informal institutions in the background of technological innovation. Paper presented at the Schumpeterian Conference 28th June -1st July 2000, Manchester Van Stel A, Wennekers S, Thurik R, Reynolds P, de Wit G (2003) Explaining Nascent Entrepreneurship Across Countries. SCALES-Paper N200301 (Scientific Analysis of Entrepreneurship and SMEs) Van Stel A, Wennekers S, Thurik R, Reynolds P (2004) Explaining Variations in Nascent Entrepreneurship. Research Report H200401, SCALES (Scientific Analysis of Entrepreneurship and SMEs) Van Wissen L (2004) A spatial interpretation of the density dependence model in industrial demography. Small Business Econ 22:3–4 Verheul I, Wennekers S, Audretsch D, Thurik R (2002) An eclectic theory of entrepreneurship. In: Audretsch D, Thurik R, Verheul I, Wennekers S (eds) Entrepreneurship: Determinants and Policy in a European–US Comparison. Kluwer Academic, Boston/Dordrecht Viesti G (2001) Come nascono i distretti industriali. Laterza Von Hippel E (1994) Sticky information and the locus of problem solving: implications for innovation. Manage Sci 40:429–439 Wennekers S, Thurik R (1999) Linking entrepreneurship and economic growth. Small Business Econ 13(1):27–55 Wortman MS (1992) The state of the art in entrepreneurship and opinion. Unpublished monograph presented at the National Academy of Management Conference, Las Vegas, August 10 Wu S (1989) Production, entrepreneurship, and profits. Basil-Blackwell, New york, NY Zajonc R (1968) Attitudinal effects of mere exposure. J Pers Soc Psych
Learning, Innovation and Growth Within Interconected Clusters: An Agent-Based Approach Mario A. Maggioni and Stefano N. Roncari
1 Introduction The aim of the paper is to model the interaction between firms within a territory using an agent-based model (Lane, 1993; Axelrod, 1997; Luna and Stefansson, 2000) in order to observe the emergence of spatial clustering, the evolution of different learning processes arising from the interaction between firms, and the development of innovative behaviour within interconnected clusters. The genesis, emergence and growth of innovative clusters and the relation between the dynamics of these clusters, the process of regional development and the level of regional welfare has been thoroughly discussed in recent contributions (Bresnahan et al., 2001; Braunerhjelm and Feldman, 2006; Karlsson, 2007). Without going into detail, it has been extensively demonstrated that the existence of thriving clusters is positively correlated with the prosperity of a region measured in terms of both employment a/o income. In the chapter, the emergence of phenomena of spatial concentration of firms and its related effects on the knowledge endowment of the region is studied as a simplified simulation model of the interactions arising in the real world between the rise of innovative industrial cluster and the increase in the level of regional income and welfare. Firms are modelled as heterogeneous individual agents endowed with different levels of two distinct types of knowledge (technology-specific and market-specific). Total knowledge is obtained as the product of these levels (Winter, 1987). The two kinds of knowledge are imperfect substitutes, hence a knowledge set where both are present in a satisfying level will be preferred to a set endowed with a high-level of one and a low-level of the other. Knowledge can be improved incrementally by an “individual” in-house process of R&D (or, more simply, through learning-by-doing) and by two alternative “collective” learning processes involving interactions between firms (learning-byusing a/o interacting).
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
117
118
M.A. Maggioni, S.N. Roncari
The first refers to “knowledge spillover”, i.e. unintentional transfer of knowledge from “learned” firms to “non-learned” ones due to physical/sectoral/technological proximity (Freeman, 1988; Stoneman, 1995; Audretsch and Feldman, 1996b); the second refers to intentional knowledge barter exchange, i.e. an intentional exchange of knowledge which takes place if and only if the barter is mutually, although not necessarily equally, profitable (Cowan and Jonard, 1999). The knowledge of a firm becomes obsolete over time. In order to keep their stock of knowledge above a survival threshold, they must interact with other firms. Interactions are obtained/facilitated through proximity (Audretsch and Feldman, 1996a; Gaspar and Glaeser, 1998; Storper and Venables, 2004). Thus firms relocate by searching for local concentration of “learned firms”. Failure to maintain their knowledge above the threshold results in the death of the firm. Innovation is modelled as a random process (which assigns the status of “potential innovator” to a given firm with a probability proportional to the firm’s knowledge level) but, in order to become an actual innovator, the firm must also have a sufficient level of total knowledge. Firms can re-locate (i.e. they move within the territory). The relocation decision can offer new development opportunities although it involves inevitable relocation costs, and is based on an approximate evaluation of the possibilities of success (Maggioni, 2002a and 2002b). The evaluation is based on the local concentration of production factors (specialised input producers and workers) around “successful” firms. A series of simulations – conducted through Clusterbugs, an original agent-based simulation model developed in the Swarm environment1 (Bonabeau et al., 1999; Terna, 1998, Luna and Stefansson, 2000) – shows the effect of different prevailing interaction modes (which can be interpreted as different regimes of intellectual property rights enforcement) and of the efficiency of different learning processes on the clustering process, on the average level of knowledge accumulated by firms and on the variance of individual knowledge levels. Finally the existence of inter-industry relationships within the same region is modelled by the introduction of two distinct firm populations (Maggioni, 2005 and 2006; Maggioni and Riggi, 2007) which interact differently as far as the learning processes are concerned. The effects of two distinct spinoff dynamics (one mainly determined by the parent firm; the other mainly driven by local environment conditions) on the structure and dynamics of clusters are also investigated. In the following pages we summarise the main characteristics of the model with regard to its main development areas: the birth of new firms, the exit of inefficient firms, the increase and decrease in knowledge levels, the innovation and relocation processes.
1
An open-source framework for agent-based simulation: www.swarm.org. For a detailed discussion, see Maggioni and Roncari, 2005.
Learning, Innovation and Growth within Interconected Clusters
119
2 Entry Dynamics When does the birth of new firms occur? The criterion of birth distribution over time is determined at the beginning of each technological era. In fact, the simulation time (which is measured in tics (elementary time units), is also divided into periods, called eras, each lasting around 200 ± 20% tics and representing the beginning of new “technological paradigms”, that is, the beginning of a new stage of development of technical and scientific knowledge, which allows the emergence of radical inventions and the success of new entrepreneurial initiatives (Dosi, 1982; Van Duijn, 1983 Kleinknecht, 1986; Geroski, 1995; Freeman, 1996). At that very moment, when and how new firms will be set up is decided: in particular, the new firms enter the market continuously, throughout the era with probability 0.8, all new firms appear in the first tic of the current era with probability 0.2, to represent the event where a new competence-destroying technology produces a sudden possibility of success for new “entrepreneurial talents”. The number of “newborn” firms is defined at the beginning of the era (or in each time unit, depending on the birth criterion described above) by the following S shaped function: newBorn = [bugsNow∗ (maxBugs − bugsNow)/maxBugs]∗ [0 > n > 1] This means that the firm population growth starts slowly, increases rapidly and, finally comes to a halt when the population level reaches a ceiling (maxBugs), which is exogenously fixed. Once the total number of newborn firms has been determined, they are generated according to two different criteria: 50% of the newborn population will be composed of absolute beginners: the new firms will appear in randomly-extracted site, and will be provided with a random knowledge-set (0.1 < KnowAi < 10 e 0.1 < KnowBi < 10); the remaining 50% will be made of spin-offs: in this case the new firms will be placed in the neighbourhood of the “parent firm”, and will inherit part of their knowledge set from it.2 Not all firms in the territory will generate spin-offs. The choice of potential “parents” is based on the knowledge set of each agent: one can easily infer that the best-performing firms will have more chances to sustain the start-up of new entrepreneurial activities; and, since Clusterbugs mainly concentrates on knowledge performance, the spin-off probability is determined with regard to knowledge levels. A list of potential parents, whose global knowledge is higher than a minimum threshold, is therefore compiled. After that, actual parent firms will be extracted from the list: each agent has a certain probability of generating spin-offs, which derives from the ratio between the previously-calculated total number of spin-offs and the number of potential parents. Thus in this paper, although we do not focus strictly on entrepreneurship – defined by Garavaglia and Breschi (2009, in this volume) as: those “activities which 2
One of the knowledge levels will be the same as the parent’s, whereas the other level will be determined randomly, from 0.1 to 10.
120
M.A. Maggioni, S.N. Roncari
give birth to new firms” – our simulation exercise confirms their main theses: (i) within clusters there are better supply and demand conditions for stimulating entrepreneurship; (ii) entrepreneurship and cluster formation are co-evolving dynamic phenomena; (iii) entry by spin-off is a significant factor for cluster development and persistence.
3 Exit Dynamics Every firm risks being expelled from the market if it becomes “inefficient”. This condition has been modelled with respect to the global knowledge of each agent, which directly affects the production costs, in our simplified representation. In order to survive, firms must achieve a minimum knowledge level (minKnowledge), which is calculated according to the average market level, since the efficiency of every actor can be defined only in comparison with its competitors’ (Klepper, 1992). minKnowledge = avgKnowledge− (minKnowledgeParam∗ knowVarCoeff) Exit from the market does not occur immediately. From the moment the firm’s knowledge goes below the market average, the model starts a sort of countdown – defined by the parameter bugHardiness, exogenously fixed by the researcher to be equal to 250 tics – which is reset any time the firm is able to exceed the threshold, so that every actor has some time to make up the gap.3 Inefficiency is not the only cause of exit from the market, though. Clusterbugs includes a random algorithm simulating bankruptcy for reasons not directly dealing with the firm’s knowledge performances (e.g. because of wrong financial choices, legal problems, contrasts between owners and managers).
4 Knowledge Creation and Depletion Every firm’s knowledge endowment is subject to continuous changes, due to the action of two opposite phenomena: learning and obsolescence, both particularly important in high-tech sectors investigated in this work. In the following paragraphs the term “primary knowledge” will be used to refer to the kind of knowledge a certain agent owns at the higher level, whereas the “secondary knowledge” will refer to the kind of knowledge which is owned at a lower level. Therefore, there will be no stable correspondence between these two terms and the two knowledge types (technical and commercial), since every firm will continually modify both levels over time thus changing their rankings. 3
We assume that every firm can survive for some time without making profit, thanks to the “wealth” (i.e. profits) accumulated in the previous periods and to the implementation of new activities which may allow efficiency to improve in the future (e.g. relocations).
Learning, Innovation and Growth within Interconected Clusters
121
4.1 Learning Processes In general, Clusterbugs allows for two different learning modes: social and individual learning. In the first case, knowledge increases as a result of the interaction with other firms; in the second case, learning is a by-product of the usual production activity of each firm
4.1.1 Social Learning Every tic, after eventually moving (see Section 5), each firm builds up a map of its neighbourhood in order to choose the agent to get in touch with, and hopefully learn something from the interaction. If more than one firm is present in the neighbourhood, the firm chooses between them randomly, since it cannot know the knowledge endowment of its “coopetitors” in advance i.e. before it has established a connection with them. In other words, it cannot decide who the best partner is, without communicating.4 Having established a connection, the firm compares its knowledge endowment with its fellow’s and, if it discovers a significant gap5 in at least one knowledge type, then the social learning process may take place and that knowledge type will be incremented by a certain amount. The efficacy of this knowledge transfer, however, depends on the interest of both actors, therefore two different ways of “interactive learning” have been modelled and labelled: knowledge barter exchange and pure knowledge spillover. The knowledge barter exchange occurs if and only if both firms profit from the interaction, and are therefore interested in supporting the knowledge transfer. In this case, the learning process will be extremely efficient for both of them. Moreover, the knowledge gap will gradually decrease every time the same actors interact.6 In other words, they will be more likely to share their knowledge as their connection becomes closer over time. The knowledge increment pursued through barter exchange is determined by the following formula: knowIncrement = [−0.1∗(bestBug.knowX-knowX)2 +1.1∗(bestBug.knowX-knowX)] This function describes a concave non-monotonic learning curve (Fig. 1), whose efficacy improves progressively as the knowledge gap gets wider, but only up to a certain value. Beyond that ceiling, the efficacy starts decreasing, to underline the 4 This is true even in cases where the firm had communicated with some neighbours in the past, since every actor’s knowledge level is continually changing. 5 As we will clarify in the next pages, this gap represents the minimum “knowledge difference” for the agents to perceive the learning opportunities connected with the current interaction. As Preissl Solimene (2003, p. 44) point out, “It often happens that the recipient of a piece of information has not searched for it, because he/she had no idea of its existence”. 6 Each agent keeps track of all the subjects it exchanges knowledge with.
122
M.A. Maggioni, S.N. Roncari y 3 2.5 2 1.5 1 0.5 x
0 0
1
2
3
4
5
6
7
8
9
10
11
? 0 .5
Fig. 1 Knowledge transfer as non monotonic function of the knowledge gap
fact that, when the knowledge gap is too wide, the firm interested in learning will not be able to exploit the data shared by the partner (at least not in that very moment), because of the lack of the basic knowledge needed to exploit the new information.7 The pure knowledge spillover, instead, takes place when the imitator “has nothing to offer”, that is, when one agent has lower levels in both of the knowledge types than the other agent. In this case, the learning process will be less efficient, since the firm generating the knowledge spillover will not voluntary support the transfer, which might well fail (Jaffe et al., 1993). Finally, the knowledge gap required to spark a learning process is wider and does not depend on the number of interactions: since there is not frequent and easy communication with the target company. The efficiency of a merely imitative learning ( pure spillover) is defined as a quota of the increment determined by the knowledge barter exchange function. This quota can vary from 0 to 1, according to the research purpose, and is fixed by the spillEfficacy parameter.
4.1.2 Individual Learning The second way of improving a firm’s knowledge endowment consists in individual activity based on learning-by-doing procedures (Arrow, 1962). The constant use of a certain knowledge allows an economic agent to improve it over time, through the usual production activity. In a sense individual learning is a simple unintended by-product of being in business. Every firm has thus been provided with a simple capacity to learn from itself. Moreover, assuming that the “primary” knowledge type is practiced in a more intensive way, it will be improved more quickly. In the function below, KnowX is right the primary knowledge type, whereas learnBDEfficacy is a parameter defined by the researcher, representing the efficacy of the 7
In this way, we are modelling the role of absorptive capacity firstly described by Cohen and Levinthal (1990).
Learning, Innovation and Growth within Interconected Clusters
123
learning-by-doing process. knowX+ = (knowX ∗ learnBDEfficacy) The secondary knowledge type (knowL) will be subject to a less efficient improvement, because of a half-weighting learnBBDEfficacy: knowL+ = (knowL∗ learnBDEfficacy/2)
4.2 Knowledge Obsolescence If, on the one hand, every firm is learning continuously; on the other hand, the value of its knowledge “stock” decreases over time – and this is especially true in a highly innovative and/or highly competitive context – thus facing the firm with progressive obsolescence. This phenomenon, although independent of the learning process defines every firm’s stock of knowledge and, consequently, its chances of survival in the market. If the firm cannot overcome the process of knowledge degradation through a combination of different learning modes, its “countdown” will be started and it will risk being expelled from the market, because of competitive selection (which is based on the average knowledge level). In Clusterbugs, obsolescence is programmed as a progressive decrease in the knowledge levels, which affects every firm in every time unit. Its impact has then been diversified according to the knowledge type: the secondary knowledge will be subject to a larger decrease, because it is used in a less systematic way and therefore characterised by higher “volatility”. By default, the rate of decrease per tic is 0.1% of the stock of secondary knowledge and 0.05% of stock the primary knowledge.
5 Innovation Innovation – in such a pure Schumpeterian framework – represents the driving force of competition and interaction between companies, especially in high-tech industries (Nelson and Winter, 1982; Metcalfe, 1998). For this reason it plays a core role in our model, driving cognitive development and the consequent cognitive flows which constitutes the essence of all inter-firm relationships within our artificial world. The innovation process in Clusterbugs is modelled as a change of status of individual firms; in other words, every firm can acquire or lose the “innovator” status over time. It is useful to remember that every firm is an imitator itself, that means that each one tries to draw as much knowledge as possible from the interaction with
124
M.A. Maggioni, S.N. Roncari
“coopetitors”, especially in highly competitive and highly innovative industries, in which firms need to constantly update their stock of knowledge in order to be profitable and remain in the market. Some of these firms will be able to become innovators, while others will remain imitators (Iwai, 1984a, b). In Clusterbugs, progress is based on Schumpeter’s traditional taxonomy, which considers invention, innovation and diffusion as the basic phases of the innovation process. These refer respectively to the starting formulation of the “innovative idea”, to the actual realisation of this idea through the development of new products and processes, and to the diffusion of the innovation on the market.8 In our model, a firm needs to meet two basic requirements in order to acquire innovator status: • It must develop an invention, i.e. an innovative idea or “potential innovation”.9 • It must be endowed with adequate global knowledge level, which is the product of the two knowledge levels. In any moment, every firm has some chance of developing an invention, according to its own knowledge level. This means that it is more likely (although not totally taken for granted) for a firm which possesses a modern and advanced information background to develop an idea in line with the opportunities of the new “technological era”. However, the invention might still also be developed by firms with a limited global knowledge: The following function describes the probability of invention: inventProbability = ideasQuota ∗ invCoeff ideasQuota is an exogenously set parameter, which defines a “rough probability” and which is the same for every firm; invCoeff is a parameter (included between 0 and 1) which will define the actual invention probability of every single firm, according to everyone’s global knowledge. However, the development of an invention is not enough for a firm to acquire innovator status. The firm needs to transform the new idea into something valuable for the market, i.e. converting it into an innovation. This requires the availability of a solid basis which, in our model, is defined by the firm’s global knowledge. In particular, the level of knowledge needs to be at least as high as the value of the variable innovMinKnow, the same for every firm and defined as a multiple of the
8
“The Schumpeterian trilogy that divides the technological change process into three stages is often considered to provide a useful taxonomy. The first stage is the invention process, encompassing the generation of new ideas. The second stage is the innovation process encompassing the development of new ideas into marketable products and processes. The third stage is the diffusion stage, in which the new products and processes spread across the potential market. The impact of new technology occurs at the diffusion stage and thus the measurement of impact is very much a measurement of how the economy changes as new technologies are introduced and used” (Stoneman, 1995, p. 2). 9 As regards the first aspect, the formulation of innovative ideas takes place at the beginning of every “technological era”.
Learning, Innovation and Growth within Interconected Clusters
125
average knowledge: innovMinKnow = innovMinKnowParam ∗ avgKnowledge where innovMinKnowParam is an exogenous parameter and avgKnowledge is the average knowledge of the reference world. Hence, if the firm has global knowledge level which exceeds the average global knowledge by innovMinKnowParam10 and if in the current “technological era” it has developed an invention, this firm will become an innovator. Once a firm has acquired innovator status, it will be able to progress immediately in both the types of available knowledge (even though to a different extent). The reason for such a dynamics is that, by becoming an innovator, the firm will benefit from a profitable knowledge and this will give it a competitive advantage over the rival firms. Indirectly we are thus assuming the existence of a certain regime of intellectual property rights which ensure a certain degree of appropriability (i.e. through a temporary monopoly right granted by the patent system) As in the real world, in Clusterbugs too, innovators will not be thus forever, in spite of their performance and their ability to update and produce new inventions over time. At the beginning of each era, firms create the innovative ideas able to exploit the new technological paradigm, making it possible to become innovators; at the same time, those who are not able to generate new ideas will be excluded from the pool of the possible innovators, even though they possess a high level of knowledge.11 As a consequence, an innovative firm could lose its status at the beginning of a new “technological era” if it is not able to develop new inventions.
6 Relocation Firms can move in the space (which in Clusterbugs is modelled as a lattice of 100 × 100 cells) in the sense that they can decide where to locate and re-locate. In general, in every time unit the firm evaluates opportunities to move according to the distribution of the production factors in the space. This is represented by a virtual “ring” of resources that surrounds every firm. During its usual production activity, every firm contributes to the support of the stock of production factors which defines the area where it operates. While staying in a certain position, the firm generates a high level of positive localisation externalities which, according to Marshall, deal with the following three factors: labour market pool, specialised input supply and knowledge spillover (Krugman, 1991). If we focus our attention on the effect of the labour market pool, we can say that the firm, just by producing its goods in a certain place, creates a “pool” of specialized labour which stays there and which possesses a certain degree of inter-firm 10
Which, in the benchmark case, is set at 10%. It must be noted, however, that previous innovators will have a high probability of being redrawn because their knowledge endowment will be probably higher than the population average.
11
126
M.A. Maggioni, S.N. Roncari
mobility within the local labour market. A similar process applies to the process of generation and co-location of firms specialised in the production of specific inputs (Saxenian, 1990). Finally, knowledge has some relevant local component: tacit knowledge can hardly be transmitted without face-to-face interactions, thus limiting the geographical scope of this kind of spillover (Storper and Venables, 2004). More specifically, the relocation process is divided into three different stages: the analysis of the surrounding space, the evaluation of the convenience of moving and the actual move.
6.1 Scanning the Space In the first stage of this process, the firm locally explores a part of the space, looking for a preferable place where it can find a high concentration of production factors. While observing the surrounding space, the firm takes into consideration not only the adjacent cells, but all the cells that lie in its visual field (a parameter defined by the researcher and, by default, corresponding to a 10-cell radius). As a consequence, every agent starts to check the factor concentration in every free cell which falls within its visual field. If the firm doesn’t spot any “better” cell than the existing one, it stays where it is. If it spots a better one, the agent temporarily memorizes its space coordinates. At the end of the scanning process the best cell among the ones that have been memorized, will undergo a deeper analysis (see next section). If the firm detects more than one cell with the same factor concentration, and all better than the existing one, it chooses one randomly.
6.2 Assessing the Options After it has detected a possible target cell, the firm carries on a detailed analysis in order to estimate the actual benefit and costs of a relocation. As far as benefits are concerned, the firms takes into consideration the availability of resources that, in the following period, will be present in the cell in which it is currently located (source cell) and in the destination (or target cell). In particular, when evaluating the “factor concentration” that characterises the source cell in the following tic, the firm will take into consideration the existing concentration, to which it will add its own “factor output capacity”, also considering the incremental learning which influences it. With respect to the concentration of production factors in the target cell, the firm will add the existing concentration to its own output capacity which will be reduced, according to moving costs and to the loss of production factors. As far as costs are concerned, three elements are significant in the evaluation process: the pure moving costs, the reduction of its own contribution to the development of the local production factors and the interruption of the learning by doing process.
Learning, Innovation and Growth within Interconected Clusters
127
6.2.1 Moving Costs When the firm decides to move, it will incur some additional costs and therefore, its profits will decrease. Since profits are not explicitly modelled but are assumed to be directly proportional to the level of the global knowledge of every single agent (as product between the levels of the two types of knowledge), these costs were shaped through a reduction of the levels of both knowledge types: knowA− = knowA∗ movCost knowB− = knowB∗ movCost knowA and knowB here represent the levels of the two knowledge types, while movCost is an exogenous parameter (positive and less than one) which defines the extent of the decrease. Two different reasons explain the correlation between relocation and decrease in the knowledge levels. The first relates to an alternative interpretation of “space” within the model; the other refers to the geographical specificity of the knowledge endowments. According to the first interpretation the lattice in which firms exists and locate should be interpreted in a sectoral/technological connotation.12 According to such a perspective, moving costs are reflected in the partial depletion of the stock of existing knowledge since it loses part of its value, because it can only partially adapt to the new production context. The second refers to the local geographical specificity of the knowledge acquired by firms during their life span. If we assume that both technology-related and market-related knowledge have a relevant local idiosyncratic component (i.e. that their value decreases with the distance from the original location in which that specific knowledge has been produced) then, measuring moving costs as reduction in the knowledge endowment becomes straightforward.
6.2.2 Reduction of the Development of Local Production Factors When the firm moves, it will have to give up some production resources which are based in the current location. Some workers, some suppliers will not follow the firm toin the new location;13 some of the accumulated knowledge (because of its spatial and technological specificity) will have a lesser value in the new location.
12
The vertical axis may refer to different technologies while the horizontal refers to different industries. Therefore each cells corresponds to a given combination of technology and industry. Contiguous cells on the same row but in different columns may refer to the use of the same technology by different industries; contiguous cells on the same column but in different rows may refer to the use of different technologies by the same industry. 13 Note that by “location” we may also consider a given technology and/or industry if the “space” we are considering has a geographical dimension and a technological/industrial dimension.
128
M.A. Maggioni, S.N. Roncari
In our model, the loss of production factors is reproduced through a sudden reduction of the “factor output capacity” at the very instant of the move: output = output∗ outputQuotaWhenMoving where outputQuotaWhenMoving represents the part of the resources that the firm will be able to maintain after it moves (this is exogenously set and, by default, equal to 0.1). 6.2.3 Interruption of Learning by Doing The last cost connected to the moving process is a temporary stop of the selfincremental process of learning. Indeed, when the firm decides to move, it will have to momentarily interrupt its production activities, and therefore to concentrate its resources on the management of logistic and organization problems that the act of moving implies. As a consequence, the organization will not be able to benefit from the learning form which develops in the field of production.
6.3 Moving After a careful evaluation the firm finally decides to move. As expected, it will incur some moving costs, it will not be able to benefit from the knowledge improvement produced by individual learning, and it will finally increase th e concentration of production factors in the target cell, with part of its output capacity. But, if the firm decides to move, it means that the availability of production factors in the target cell justifies the relocation risk. In other words, the firm reckons that there is a reasonable chance of increasing its profits in the new position. By using the words “risk” and “chance”, we want to highlight the fact that, although the firm cautiously evaluates the opportunity of moving, its decision will still be taken with limited rationality (Simon, 1984). In the model, agents use all the available information in order to take a decision, and – while doing that – they act in an entirely rational way (Ottaviano and Puga, 1998). But still, the information they have is purely perceptual: there is no certainty that the new location will offer chances of higher profit (read: global knowledge) and, consequently, better chances of “surviving”. The only information which the agent can rely on is represented by the availability of production factors in a given target area. In general it is very likely that, within an area characterised by above average concentration of resources, there will be located firms with an above average level of global knowledge (which offer better chances for interactive learning procedures), but it cannot be assumed that this will always be the case. For example, the concentration of production factors might be due to the presence of a very compact cluster of “average” firms, or of few very knowledgeable firm, but with which the moving firm cannot exchange knowledge (through the knowledge barter mode, the more
Learning, Innovation and Growth within Interconected Clusters
129
efficient learning mode), because these local firms have higher level of both types of knowledge, thus forcing the newcomer to learn through the difficult and less efficient spillover mode. In other words, firms in Cluserbugs use all the available information in a rational way, but – because their information set is very limited, they are not able to foresee all the effects of their choices. The overall effect at the regional level of the individual movements of the firms is the eventual formation of stable (or unstable) clusters, here defined as spatial agglomeration of firms located in neighbouring cells and measured by a specific index.14 Given the way the learning process is modelled in Clusterbugs, it is not surprising to observe clustering dynamics in the regional population of firms. What is, however, interesting is the different result (also from a spatial perspective) which may be obtained by changing the value of relevant parameters defining some basic characteristics of firms.
7 A sectorally Heterogeneous Population In all sections above, every firm acting within the Clusterbugs belongs to the same industry. Therefore the same social learning mechanisms describe every interaction between any pair of firms. In order to better represent the real interactions emerging in a region, with specific reference to the existence of different industries, a radical change in the model is introduced by defining two populations of firms which cannot change their industrial features over time and we assume that social learning mechanisms (both pure spillover and barter exchange) are more efficient when performed between firms belonging to the same industry than when performed across different industries (Fig. 2). However, in order to produce results comparable with the benchmark case, the average level of social learning was kept constant while modelling two cases to study the effects of increasing technological heterogeneity between the two industries: we set the simulation parameters so that the average learning efficiency within the whole world remained constant. In the first case intra-industry learning is 1.5 times more efficient than the benchmark while inter-industry learning is 0.5 of the benchmark case; in the second case intra-industry learning is 1.8 times more efficient than the benchmark while inter-industry learning is 0.2 of the benchmark case. double familyEfficacy = (family == bugToLearnFrom.family)?1.5 : 0.5; We also allow spin-off mechanisms to take place in this two-industry case. In particular, we performed two sets of simulations with different “generation” rules. In the first set (genetic spin-off), parent firms belonging to a certain industry were able 14
The clustering index is calculated as a rate of positive spatial correlation between each cell and the neighbouring cells (for more details see note 18).
130
M.A. Maggioni, S.N. Roncari
Fig. 2 Spatial distribution of firms and surrounding production factors Table 1 Possible scenarios arising in the multi-population case Advantages of Intra-population Knowledge transfer
Spin-off types
Genetic spin-off Probabilistic spin-off
Intra 1.5
Intra 1.8
2Pg1.5 2Pp1.5
2Pg1.8 2Pp1.8
to produce spin-offs belonging only to their own industry; in the second set, parent firms were able to produce spin-offs belonging either to the same or to the other industry, according to a probability coefficient which depends on the relative share of industries in the neighbourhood. For this reason, the spin-off of a firm located in a multi-industry cluster has a higher probability of belonging to a different industry than its parent’s. Table 1 illustrates the four possible scenarios derived from the interaction of the different simulation criteria.
8 The Simulation Results Having described how the artificial world of Clusterbugs works we can now illustrate the result of a simulation set that we conducted by performing a series of sensitivity analyses (i.e. by changing values of key parameters) in order to understand the macro effects of micro variations, within a systemic perspective (Ballot and Weisbuch, 2000; Conte et al., 1997).
Learning, Innovation and Growth within Interconected Clusters
131
In particular, our simulations test the benchmark situation (which has been described above) against seven alternative scenarios which will be described in the following sections. Our aim is to discover, through the model, the existence of unexpected connections between the phenomena considered. The goal of Clusterbugs, just like any other agent-based model, is to record the emergence of counter-intuitive results, of aggregated effects which cannot be calculated in advance. In fact, the model includes so many variables and rules that it is practically impossible to foresee its final behaviour: the power of this kind of simulation models is thus the opportunity to observe how a micro-founded world evolves over time, and how single changes at the agent level can produce very complex systemic effects (Schelling, 1978). In our research, for each scenario 10 different simulations (with a length of 5.000 ticks) were performed and evaluated on the basis of the value (and, sometimes, of the variation coefficient over different simulations) of about 20 variables (described in Appendix 1). The most relevant are the following: • The average level of knowledge in the firm population (in order to compare the effectiveness of different institutional arrangements in terms of their ability to sustain a high level of knowledge in the population) • The variance (or, better, the coefficient of variation) in the level of knowledge within the firms population (to measure the degree of inequality existing between different firms within the same population) • The amount of knowledge transferred between agents (as a measure of the active interrelationships of agents) • The relative share of the different learning types (to understand whether cooperative interactions – barter exchanges – prevail over pure imitative behaviour) • The average success-rate of inventions (that is, the proportion of the total inventions which have been transformed into innovations) • The turn-over rate of the population of firms (which is an inverse measure of average life expectancy of firms in the population) • The number of clusters generated (measured through an agglomeration index) • The relative number of specialised versus diversified clusters, when two types of firms are considered (in order to understand the prevalence of urbanisation vs. localisation external economies) • Finally, in order to measure the relative stability of the simulation results a “longitudinal” coefficient of variation across all runs of regressions for each variables has been computed The simulation results in Figs. 3–5 show the evolution (within a specific simulation) of three main variables: average Knowledge level in the total population of firms, Clustering index and Innovators quota. Figures 6–8 show the evolution of the same variables in the two-population cases. These figures must be interpreted in a very descriptive way since the variables shows a significant variance in their behaviours and they refer to one specific simulation only. More significant insights may be derived from Table 2 which shows the mean value, the variance and the coefficient of variation for each single simulation
132
M.A. Maggioni, S.N. Roncari Average Knowledge
60 50 B
40
HCM
30
`
20
LIP HBI SKO NS
10 0 runs
Fig. 3 Average knowledge Clustering Index
10 9 8
B
7
HCM
6
LIP
5
HBI
4 3
SKO
2
NS
1 0 runs
Fig. 4 Clustering index
(of 5,000 ticks) in the columns for each relevant variable and, the mean and the coefficient of variation of the previously mentioned results across 10 different simulations in the rows.
8.1 Highly Competitive Market (HCM) A highly competitive market is obtained in the simulation by increasing the knowledge threshold needed to “survive” in the market. Since in Clusterbugs prices, costs, revenues and profits are not explicitly modelled, knowledge is the only variable which, in a sense, represents both output and profits. As expected, an increase in the knowledge thresholds, produce both an increase in the death rate firms as well as
Learning, Innovation and Growth within Interconected Clusters
133
Innovators quota
0.45 0.40 0.35
B
0.30
HCM
0.25
LIP
0.20
HBI
0.15
SKO
0.10
NS
0.05 0.00 runs
Fig. 5 Innovators quota Average Knowledge 2 populations
80 75 70 65
2Pp1,5
60
2Pp1,8
55
2Pg1,5
50
2Pg1,8
45
B
40 35 30 runs
Fig. 6 Average knowledge in the two population scenarios
an increase in the average knowledge endowment of firms. However, this obvious results is accompanied by a counterintuitive one: One might think that by raising the threshold a more select and homogenous entrepreneurial e´ lite would emerge in the population. The simulation results contradict such a hypothesis and show an increase in the heterogeneity of the knowledge level within the population (as suggested by the value of knowVarCoeff, the coefficient of variation of the knowledge level, which is sensibly higher than the benchmark case). Therefore, raising the level of competition in this artificial region does not produce a population of “fitter” firms (as suggested by the extreme advocates of free-market); on the contrary, heterogeneity increases while the agglomeration propensity of firms decreases marginally.
0.0562 0.0709 0.06512 0.0004
0.1366 0.1642 0.14555 0.0008
0.430 0.459 0.14486 0.0002
0.415 0.499 0.44978 0.0022
Lower invention probability (LIP) min 2 0.703 0.096 MAX 5 0.723 0.107 mean 3.4 0.71320 0.10240 Cv 0.3059 0.0001 0.0001
Higher barrier to innovation (HBI) min 1 0.689 0.098 MAX 3 0.756 0.122 mean 2.2 0.72067 0.10482 Cv 0.2545 0.0008 0.0008
Stronger knowledge obsolescence (SKO) min 3 0.650 0.101 0.468 MAX 4 0.694 0.112 0.516 mean 3.4 0.67685 0.10815 0.48609 cv 0.0706 0.0004 0.0001 0.0006
0.0644 0.0709 0.06729 0.0001
0.1501 0.1747 0.16632 0.0005
0.411 0.440 0.42445 0.0003
Highly competitive market (HCM) min 2 0.716 0.094 MAX 4 0.746 0.101 mean 3.4 0.73256 0.09665 Cv 0.1882 0.0002 0.0001
0.1236 0.1453 0.13728 0.0004
0.398 0.450 0.42272 0.0009
0.085 0.103 0.09390 0.0003
0.698 0.754 0.72519 0.0003
0.0014 0.0033 0.00231 0.0002
0.0010 0.0036 0.00198 0.0004
0.0007 0.0011 0.00090 0.0000
0.0020 0.0032 0.00267 0.0001
0.0014 0.0030 0.00191 0.0001
0.2248 0.3975 0.32903 0.0112
0.4830 0.8571 0.66553 0.0225
0.3818 0.5748 0.46982 0.0100
0.2564 0.3312 0.31019 0.0024
0.2733 0.4131 0.31711 0.0044
0.2222 0.3286 0.27954 0.0051
-0.4372 -0.3617 -0.40016 -0.0022
0.1538 0.4167 0.27655 0.0314
0.2513 0.3902 0.33445 0.0099
0.1827 0.2794 0.23200 0.0050
23.67 36.13 28.54239 0.6552
39.17 46.04 41.99484 0.1855
34.66 43.62 38.89893 0.2440
41.37 49.98 45.28344 0.2718
36.78 44.98 40.31 0.1597
44.71 362.83 169.89224 66.3368
171.01 393.51 241.58160 27.4221
173.70 388.30 249.85728 15.7229
239.46 785.08 426.13606 85.0038
147.00 425.83 256.60 31.6178
0.2825 0.5272 0.42573 0.0187
0.3287 0.4444 0.36408 0.0048
0.3417 0.4725 0.40471 0.0069
0.3740 0.5607 0.44227 0.0092
0.3173 0.5611 0.39397 0.0151
Cluster Spillover- Spillover- Spillover- Innovators- Innovators- Innovators- Ideas- Knowledge Knowledge Knowledge number Quota Quota Quota Quota Quota Quota Success mean variance coeffVar mean variance coeffVar mean variance coeffVar mean
Benchmark(B) min 3 MAX 6 mean 3.8 CY 0.2000
Sim
Table 2 Simulation results
3415.44 4114.51 3681.56728 18.3090
2497.43 3262.89 2871.21196 23.1761
2567.86 2911.93 2778.68419 5.4083
3079.34 3475.45 3234.28604 8.2644
2539.35 3069.06 2850.58 11.2859
KnowTransfer total
0.6820 0.8214 0.73700 0.0036
0.4995 0.6513 0.57398 0.0046
0.5134 0.5824 0.55393 0.0011
0.6159 0.6944 0.64672 0.0016
0.5079 0.6138 0.5701 0.0023
45 98 79.8 4.2476
66 86 76.4 0.7047
86 99 91.2 0.3439
63 89 74.8 1.0289
70 101 89.3 0.7974
Know- Bugs Transfer last mean tick
76.82 83.65 79.98 0.0602
74.41 84.36 80.67 0.1411
77.12 82.39 79.67 0.0636
69.81 78.65 74.31 0.1086
73.82 86.81 81.93 0.1462
Bugs mean
589 672 625 1.1949
511 628 589 3.0689
578 617 598 0.3592
715 811 755 1.4559
518 627 588 1.3882
630 667 644.6 0.2523
532 643 608.6 2.9475
588 618 602.6 0.1564
756 830 788.0 0.9396
525 634 599.4 1.5273
1.66 1.72 1.696 0.0003
1.46 1.70 1.603 0.0041
1.46 1.68 1.569 0.0031
2.29 2.42 2.344 0.0008
1.49 1.64 1.54 0.0013
Bugs Number Death birth of mean deaths
24 27 25.8
Eras
4.5389 25 5.9528 26 5.25299964 25.4 0.0532
6.4465 24 7.5257 26 6.83122367 25 0.0281
5.2249 25 5.5797 28 5.3962365 26.2 0.0027
4.4221 25 6.2088 27 5.27791087 26 0.0644
4.4302 6.6117 5.4580 0.1261
Cluster index mean
134 M.A. Maggioni, S.N. Roncari
0.1469 0.1731 0.51546 0.0006
0.1506 0.1697 0.16040 0.0004
0.1551 0.1679 0.16164 0.0001
Sectorally heterogeneous population (2Pg1.5) min 1 0.500 0.092 0.459 MAX 4 0.690 0.100 0.626 mean 2.3 0.60513 0.09513 0.09587 cv 0.3522 0.0042 0.0001 0.0044
Sectorally heterogeneous population (2Pg1.8) min 1 0.465 0.072 0.537 MAX 3 0.521 0.079 0.577 mean 2.2 0.49492 0.07546 0.55546 cv 0.2545 0.0007 0.0001 0.0004
Sectorally heterogeneous population (2Pp1.B) min 2 0.447 0.070 0.528 MAX 3 0.524 0.088 0.594 mean 2.2 0.48959 0.07802 0.57084 cv 0.0727 0.0016 0.0006 0.0010
0.1352 0.1605 0.14405 0.0004
0.1323 0.1585 0.14979 0.0004
0.000 0.000 0.0 −
Sectorally heterogeneous population (2Pg1.5) min 1 0.525 0.087 0.455 MAX 3 0.668 0.098 0.594 mean 2.4 0.62976 0.09292 0.48651 cv 0.1833 0.0031 0.0001 0.0031
NO spillover allowed (NS) min 2 0.000 0.000 MAX 5 0.000 0.000 mean 3.7 0.0 0.0 cv 0.2730 − −
0.0014 0.0027 0.00214 0.0001
0.0019 0.0023 0.00209 0.0000
0.0018 0.0047 0.15881 0.0003
0.0018 0.0035 0.00228 0.0001
0.0014 0.0024 0.00191 0.0000
0.2385 0.3082 0.28341 0.0032
0.2590 0.3152 0.28542 0.0012
0.2692 0.4014 0.00250 0.0152
0.2815 0.3960 0.31750 0.0030
0.2666 0.3626 0.30316 0.0027
0.3203 0.4141 0.36050 0.0026
0.3081 0.3720 0.34569 0.0015
0.2053 0.4436 0.31125 0.0175
0.2127 0.3790 0.29786 0.0089
0.1536 0.3515 0.25447 0.0157
52.25 67.29 61.10301 0.7889
53.12 70.35 62.71183 0.7166
47.07 65.66 0.32762 0.6164
37.57 70.34 52.22154 1.4402
35.45 51.38 42.01341 0.4671
474.69 972.74 644.99353 46.7412
407.15 1025.24 772.59238 66.1517
271.47 1048.07 55.49335 101.1367
143.10 849.27 496.49161 115.2500
95.04 462.64 236.82219 51.6722
0.3646 0.4639 0.41360 0.0035
0.3555 0.5489 0.43991 0.0114
0.3132 0.5258 578.54517 0.0121
0.2619 0.5226 0.41429 0.0178
0.2297 0.4304 0.35564 0.0138
5260.02 7119.07 6379.38004 85.3332
6071.83 7135.35 6569.44646 31.3999
3170.31 4594.70 0.42330 29.7375
3579.64 4785.24 4031.50946 30.9181
2259.92 3099.83 2621.58578 18.2613
1.0497 1.4238 1.27495 0.0173
1.2064 1.14271 1.31157 0.0065
0.6325 0.9120 4069.46713 0.0058
0.7159 0.9450 0.80462 0.0058
0.4513 0.6120 0.52488 0.0031
68 88 78.6 0.6697
84 96 88.6 0.2194
61 99 0.81139 1.5542
73 98 88.0 0.4773
55 99 77.5 2.8135
77.25 86.35 82.63 0.1231
83.57 86.45 85.21 0.0182
76.58 87.89 80.1 0.1053
81.19 87.45 84.33 0.0372
78.25 88.10 83.166 0.1353
502 584 534 1.5311
553 579 568 0.1697
492 603 83.28 1.8268
537 586 568 0.5046
549 634 589 0.9851
524 616 565.4 1.6524
572 601 589.4 0.1579
503 367 557 2.3201
554 610 590.3 0.4435
585 651 614.9 0.7915
1.41 1.48 1.445 0.0006
1.40 1.50 1.450 0.0010
1.34 1.68 586.9 0.0051
1.34 1.59 1.471 0.0024
1.43 1.72 1.564 0.0076
6.4616 23 9.3933 26 7.933 25.2 0.1218
7.3071 25 8.3843 27 7.944 25.8 0.0200
5.8398 24 9.9315 27 1.485 7.447 256 0.1643
6.0134 25 9.0657 27 7.254 26.2 0.1342
4.2081 25 6.4495 27 5.455 26 0.0716
Learning, Innovation and Growth within Interconected Clusters 135
136
M.A. Maggioni, S.N. Roncari Innovators quota 2 populations
0.35 0.30 0.25
2Pp1,5 2Pp1,8
0.20 0.15
2Pg1,5 2Pg1,8
0.10
B
0.05 0.00 runs
Fig. 7 Innovators quota in the two population scenarios Clustering index 2 populations
12 10 2Pp1,5 2Pp1,8 2Pg1,5 2Pg1,8 B
8 6 4 2 0 runs
Fig. 8 Clustering index in the two population scenarios
9 Learning Learning is therefore a crucial engine in the differentiation of individual firms. When the environment is made artificially more homogeneous by the introduction of a higher benchmark, which selects a smaller share of (more knowledgeable) firms, heterogeneity is reintroduced through a sort of equilibrium dynamic which – as an effect – raises the level of heterogeneity (around a higher average level of knowledge). Learning mechanisms in a select e´ lite are more efficient and tend to produce even further selection by raising the average level of knowledge on which the selection device is based. The total amount of knowledge transferred in the simulation between the agents is higher than the benchmark case. This is due to the share of barter exchange being higher than the pure spillover phenomenon.
Learning, Innovation and Growth within Interconected Clusters
137
One may observe that such dynamics describe a virtuous circle of “competition fostering further competition” leading to a population of more and more knowledgeendowed firms; however one may also observe that such a privileged situation is obtained at the expense of an increase in the variance of the population. In other words, highly innovative and knowledge-endowed firms compete in the market with weaker firms which are doomed to extinction in the long run, even though they are endowed with higher knowledge levels than the best performers in the benchmark case. At the macro level, such dynamics translate into an increase in the turn-over rate of firms with obvious negative effects on the regional welfare level (if one assumes reasonable levels of information and transaction costs leading to frictional unemployment). The clustering index is slightly lower than the benchmark case and this may well be explained in terms of the higher turnover rate of firms (almost 1.5 times the benchmark value) which makes any stable aggregation of firms extremely difficult to establish.
9.1 Lower Invention Probability (LIP) This scenario is characterised by a reduction in the development rate of new inventive ideas (through a reduction in the value of the parameter ideasQuota. The simulations show a decrease in the share of innovators over the total number of firms; however, despite the fact that invention is the first seed of the innovation process, the average amount of knowledge and the total amount of knowledge transferred between firms is only slightly less than in the benchmark case. Even more surprisingly, the success rate of invention (i.e. the ratio between the number of inventions – new ideas – and the number of innovations – and of innovators) increased. In other words, a reduction in the number of new ideas is accompanied by a higher probability that those ideas become marketable innovations. Such a phenomenon may be interpreted as a sort of spontaneous equilibrium dynamics of the population of firms which grant a higher probability of success to a lower number of initial inventions in order to keep the rate of technological change and knowledge advancement roughly constant. Alternatively, one could interpret these results in a different perspective (by looking at a symmetric simulation in which the invention rate is increased). The results show that a larger number of inventions does not necessarily lead to a higher number of innovations. The consequence of this last observation – on the design of science and technology policy intervention by public authorities (at a local, regional, national level) – is really clear: supporting the creation of new ideas through the funding of basic research is certainly a necessary but not sufficient condition to achieve a higher level of innovative performance. Fewer inventions but more focussed on the technological and productive characteristic of the economic system may give better results.
138
M.A. Maggioni, S.N. Roncari
9.2 Higher Barrier to Innovation (HBI) A third scenario compares the benchmark case with a situation in which firms find it more difficult to translate an original idea or an invention into a marketable innovation. This scenario represents, in a region in which innovative dynamics are hindered by a set of exogenous constraints (which may be financial – as in the case of a stricter credit rationing situation –, legal – as in the case of a broader patent scope, etc.). In this scenario, only firms with very high levels of global knowledge are able to transform an innovative idea into a proper marketable innovation. One would expect that, given the lower innovation activity, the average knowledge level would decrease. However, to our surprise, we find an increase in a series of indicators of technological “progress”, such as the average level of knowledge and the total amount of transferred knowledge. In other words, it seems that hardship stimulates positive innovation-based competition which not only leads to a feasible equilibrium level but even an increase in the performance of the whole regional technological system. By comparing this scenario with the previous ones (see Sections 1 and 2), both “hyper-liberalistic” polices based on price competition and “hyper-interventionist” policies based on indiscriminate support for “blue sky” basic research appear not to be conducive to better technological performance of the regional innovation system. On the contrary, a harsh technological environment seems to produce an environment more conducive to innovation and knowledge. Agglomeration dynamics are also enforced within such a scenario (as signalled by the increase in the clusterIndex) suggesting that external economies play a compensating role for firms exposed to such innovation hardship. Thus, fewer, larger clusters seem to grant firms the knowledge advantages (deriving from easier knowledge spillovers and barter exchanges) to transform original ideas into marketable innovations. “United we stand” may thus be the motto of such an advanced technological scenario, even though the value of the total amount of transferred knowledge is less stable throughout the different simulation runs as signalled by the “longitudinal” coefficient of variation of such a variable.
9.3 Stronger Knowledge Obsolescence (SKO) The results obtained when the obsolescence rate of the cumulated knowledge is increased (simulating a faster rate of competence disrupting technological and organisational change), suggest some further considerations on the process of knowledge creation and destruction within a regional economic system. Obviously, in this case, the average level of knowledge of the population of firms is lower than the benchmark case. However, such a reduction is not so dramatic due to the counteracting effect of the social learning mechanisms. The total level of transferred knowledge is in fact higher than the benchmark and such a result is
Learning, Innovation and Growth within Interconected Clusters
139
obtained by a relevant increase of the “barter exchange” over the “pure spillover” dynamics. Once again the system seems to possess some homeostatic tendencies which tend to counteract each exogenous shock we impose. Surprisingly, the number of innovations (as measured by innovatorsQuota) and the innovation success rate is also higher than the benchmark case. This may be explained in terms of the higher heterogeneity which characterises the population of firms, which gives a stronger competitive advantage to more knowledge-endowed firms. The turnover rate is significantly higher than in the benchmark case while the clustering coefficient is slightly lower.
9.4 No Spillover Allowed (Strict Enforcement of Intellectual Property Rights) (NS) A further interesting scenario arises in the case where social learning mechanisms are limited to barter exchange, given that no pure spillover is allowed. Such a scenario simulate an institutional framework in which the enforcement of intellectual property rights (obtained through a patent system) is absolute. In such a situation only voluntary exchange of knowledge happens since this are is the only one where both sides of the exchange gain something. The first observation regards the extreme variability of the simulation results across the different runs: the average knowledge of the firms population varies between 35.5 and 51.4 (against an average value of the benchmark case of 40.3). The total amount of transferred knowledge varies from 2.260 to 3.100 over the different runs against an average of about 2.800 in the benchmark case. The coefficient of variation of such variables is between 1.5 and 3 times higher than the benchmark case. In other words, the absence of the spillover dynamics seems to generate a highly unstable situation in which very positive results are generated together with other very poor ones. One may therefore observe a contrario that the spillover mechanism produces a sort of auto-organising mechanism of the system – possibly through an increase in the turnover rate – which is able to reduce the variability of the global performance of the system.
9.5 Sectorally Heterogeneous Population (2Pg1.5; 2Pp1.5; 2Pg1.8; 2Pp1.8) When different populations (proxy of different industries: i and j) are introduced into the model, the possibility of specialised versus diversified clusters emerges. In general specialized clusters are more likely to emerge than diversified ones but diversified clusters display a longer life (on average). The average knowledge level
140
M.A. Maggioni, S.N. Roncari
of the population increases with respect to the benchmark case and is higher the more technologically heterogeneous are the two populations of firms.15 One may interpret these results as suggesting that industrial diversification drives economic development (Glaeser et al., 1992; Swann et al. 1998). Consequently, the number of innovators and the share of successful inventions also increase. When the difference in the efficiency of the interactive learning mechanisms (barter exchange and pure spillover) between intra-industry and inter-industry is increased (by keeping the average constant and equal to the benchmark case), the average knowledge level increases thus showing a potential beneficial effect of higher industrial diversification as the engine of growth as in Jacob (1969). Pure spillover is lower than in the single population case; however, it is worth noting (as illustrated in Fig. 1) that as the difference between intra- and inter-population knowledge transfer mechanisms increases, the spillover quota decreases. Lower values of the index are also recorded in the probabilistic with respect to the genetic spin-off generation rule. By referring to Table 2, it also easy to show that not only is the innovators quota significantly increased in the 2 population case compared to the benchmark but also that the innovators quota follows the increase in the index of industrial heterogeneity. This result may be interpreted as a confirmation of the Jacobs (1969) hypothesis which links innovation to the ability in the re-combination of previously existing knowledge within highly heterogeneous environments (Glaeser et al., 1992). One may further add that the innovators quota is also higher in the probabilistic spin-off case than in the genetic one, thus showing that heterogeneity is an advantage when local firms are able to cope with it and to cross industrial boundaries at least in the medium-long term. The relative share of barter exchange over spillover in the total social learning activities increases, thus showing a positive relation between industrial diversification and mutually beneficial relationships between firms. One should also note that these results are amplified in the “probabilistic” with respect to the “genetic” spinoff scenario since the “probabilistic” scenario allows offspring to be more similar to the neighbouring firms, therefore more efficient in the learning process. Furthermore, compared with the benchmark case, the agglomerative behaviour of firms within the model becomes stronger as evidenced by both fewer clusters and a higher clustering coefficient. This means that agglomeration economies are stronger and produce a smaller number of larger clusters. Finally, throughout the simulation runs, specialised clusters (i.e. cluster composed of firms belonging to the same industry) prevail over differentiated ones. However differentiated clusters tend to be more stable along each simulation run (i.e. their average “life” is longer).
15
While the coefficient of variation of the same variable decreases, thus signalling a more equal distribution of knowledge endowment within the whole region.
Learning, Innovation and Growth within Interconected Clusters
141
10 Conclusion The paper deals with the complex relationship between spatial clustering and innovation within a region. Through an agent-based approach it has been possible to show how the autonomous behaviour of a number of firms (with limited rationality) may produce very different results at the regional level both in terms of clustering patterns and in terms of innovative output. The simulation model (built within the Swarm environment) allowed the analysis of the effects of a series of economic and institutional changes in the regions. In particular, it has been possible to show that several policy instruments commonly used to support regional innovation systems (by spurring competition, funding basic R&D, raising the degree of appropriability) have significant drawbacks in terms of higher turnover rate of firms, higher uncertainty of outcomes, lower knowledge transfer. Furthermore the analysis of multi-industry scenarios has shown that, while specialised clusters are more common than differentiated ones, the latter tend to be more resilient. Innovation is far more common and knowledge is transferred at a faster rate when different industries are established in the region. These results are reinforced when the difference in the learning efficiency between intra-industry and inter-industry relations is higher. The comparison of two different scenarios in the industrial determination of spinoff firms shows better results from “probabilistic” scenarios, in which the industrial characteristics of a new-born firm partly depend on the parent firms and partly on the neighbouring firms, thus establishing a relationship between the spatial distribution of firms and the structural evolution of industry over time. Further research is needed to investigate these relevant phenomena. We believe that an agent-based approach will complement analytical models and case studies by showing the plurality of different macro patterns which can emerge from very similar micro foundations when modelling the locational decision and the learning procedure of firms within interconnected clusters.
Appendix 1 (Variables Used to Evaluate the Simulation Results) clusterNum: number of clusters16 existing at the end of simulation (last tick); avgSpillQuota: average value (over 5.000 ticks) of the ratio of “pure spillover” knowledge transfers to total knowledge transfers; spillQuotaVariance: average value (over 5.000 ticks) of the variance of SpillQuota in each tick;
16
A cluster is defined as a spatial agglomeration of firms located in neighbouring cells.
142
M.A. Maggioni, S.N. Roncari
spillQuotaVarCoeff: average value (over 5.000 ticks) of the coefficient of variation17 of the ratio of “pure spillover” knowledge transfers to total knowledge transfers; avgInnQuota: average value (over 5.000 ticks) of the share of innovators out of the total number of agents; innQuotaVariance: average value (over 5.000 ticks) of the variance of innQuota in each tick; innQuotaVarCoeff: average value (over 5.000 ticks) of the coefficient of variation of the share of innovators out of the total number of agents; ideasSuccess: share of successful innovation out of the total number of inventions developed during the simulation; avgKnowledge: average value (over 5.000 ticks) of the global knowledge of the population of firms (in each tick); knowVariance: average value (over 5.000 ticks) of the variance of Knowledge in each tick; knowVarCoeff: average value (over 5.000 ticks) of the coefficient of variation of the knowledge level of the firms population; totalKnowTransfer: total amount of knowledge transferred through “barter exchange” or “pure spillover” (over 5.000 ticks); avgKnowTransf: average amount of knowledge transferred in each tick; bugsNow: number of firms in each tick; totalBirths: number of new firms “generated” within 5.000 ticks; avgBugs: average number of firms (over 5.000 ticks); totalDead: total number of firms expelled from the market (over 5.000 ticks); avgDeathQuota: average death rate (over 5.000 ticks) calculated as the ratio of eliminated firms to the current population of firms in each tick; avgClusterIndex: agglomeration index calculated on the basis of the “queen contiguity” rule;18 technoEra: number of “technological eras” occurring within 5.000 ticks (useful for horizontal comparison of the data and better framing of the simulation results). Acknowledgments Financial support from Universit`a Cattolica, D.1 Research Project 2005 “Learning dynamics, innovation and growth within an open local system”, is gratefully acknowledged.
References Arbia G, Espa G (1996) Statistica economica territoriale. CEDAM, Padova Arrow KJ (1962) The economic implications of learning by doing. Rev Econ Stud 29(3):155–173 Audretsch DB, Feldman MP (1996a) Innovative clusters and the industry life cycle. Rev Ind Org 11(2):253–273 17
The coefficient of variation (the ratio between variance and mean) has been used in the analysis since it allows the comparison of different variables with extremely different mean values. 18 The clustering index (following Arbia and Espa, 1996) is built as a rate of positive spatial correlation between each cell and the neighbouring cells defined through a “queen” criterion.
Learning, Innovation and Growth within Interconected Clusters
143
Audretsch DB, Feldman MP (1996b) R&D spillovers and the geography of innovation and production. Am Econ Rev 86(3):630–640 Axelrod R (1997) The complexity of cooperation: agent-based models of competition and collaboration. Princeton University Press, Princeton, NJ Ballot G, Weisbuch G (eds.) (2000) Applications of simulation to social sciences. Hermes Science, Paris Bonabeau E, Dorigo M, Theraulaz G (1999) Swarm intelligence. From natural to artificial systems. Oxford University Press, Oxford Braunerhjelm p, Feldman m (eds.) (2006) Cluster genesis: technology-based industrial development. Oxford University Press, Oxford Bresnahan TF, Gambardella A, Saxenian A (2001) Old economy’ inputs for ‘new economy’ outcomes: cluster formation in the new silicon valleys. Ind Corp Change 4:835–860 Cohen W, Levinthal D (1990) Absorptive capacity: a new perspective on learning and innovation. Admin Sci Quart 35(1):128–152 Conte R, Hegselmann R, Terna P (eds.) (1997) Simulating social phenomena. Springer, Berlin Cowan R, Jonard N (1999) Network structure and the diffusion of knowledge. J Econ Dyn Control 28(8):1557–1575 Dosi G (1982) Technological paradigms and technological trajectories–A suggested interpretation of the determinants and directions of technical change. Res Policy 11(3):147–162 Freeman C (1988) Diffusion: the spread of new technology to firms, sectors and nations. Heertje A (ed.) Innovation, technology and finance. Basil Blackwell, Oxford, pp 38–70 Freeman C (ed.) (1996) Long wave theory. Edward Elgar, Aldershot Garavaglia C, Breschi S (2009) The co-evolution of entrepreneurship and clusters. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Gaspar J, Glaeser EL (1998) Information technology and the future of cities. J Urban Econ 43(1):136–156 Glaeser EL, Kallal HD, Scheinkman JA, Shleifer A (1992) Growth in cities. J Polit Econ 6:1126– 1152 Geroski PA (1995) what do we know about entry? Int J Ind Org 4:421–440 Iwai K (1984a) Schumpeterian dynamics: an evolutionary model of innovation and imitation. J Econ Behav Org 5(2):159–190 Iwai K (1984b) Schumpeterian dynamics, part II: technological progress, Firm growth and ‘Economic selection’. J Econ Behav Org 5(3):321–355 Jacobs J (1969) The economy of cities. Random House, New York Jaffe AB, Trajtenberg M, Henderson R (1993) Geographic localization of knowledge spillovers as evidenced by patent citations. Q J Econ 108(3):577–598 Karlsson C (2007), Handbook of research on clusters: theories, policies and case studies. Edward Elgar, Aldershot Kleinknecht A (1986) Long waves, depression and innovation. De Economist 134(1):84–108 Klepper S (1992) Entry, exit, growth, and innovation over the product life cycle, paper presentato alla conferenza della International Joseph A. Schumpeter Society, Kyoto, August 1992 Krugman P (1991) Geography and trade. The MIT Press, Cambridge, MA Luna F, Stefansson B (eds.) (2000) Economic simulations in swarm: agent-based modelling and object oriented programming. Kluwer Academic Publishers, Norwell, MA Maggioni MA (2002a) Clustering dynamics and the location of high-tech firms. Springer, Heidelberg Maggioni MA (2002b) The development of high-tech clusters: theoretical insights and policy implications. In: Feldman MP, Massard N (eds.) Institutions and systems in the geography of innovation. Kluwer, Dordrecht, pp 309–340 Maggioni MA (2005) The dynamics of high-tech clusters: competition, predation, synergies. In: Quadrio Curzio A, Fortis M (eds.) Research and technological innovation. The challenge for a new Europe. Phisica Verlag, Heidelberg
144
M.A. Maggioni, S.N. Roncari
Maggioni MA (2006) Mors tua, vita mea? The rise and fall of innovative industrial clusters. In: Braunerhjelm P, Feldman M (eds.) Cluster genesis- The origins and emergence of technology. Oxford University Press, Oxford, pp 219–242 Maggioni MA, Riggi MR (2007) High-tech firms and the dynamics of innovative industrial clusters. In: Karlsson C (ed.) Handbook of research on clusters: theories, policies and case studies. Edward Elgar, Aldershot Maggioni MA, Roncari SN (2005) Artificial silicon landscapes: innovative/imitative behaviours and clustering dynamics within a heatbugs-like model. paper presented at the II Workshop di Vita Artificiale, Roma, 2–5 Marzo 2005 Metcalfe SJ (1998) Evolutionary economics and creative destruction. Routledge, London Nelson RR, Winter SG (1982) An evolutionary theory of economic change. Harvard University Press, Cambridge, MA Ottaviano G, Puga D (1998) Agglomeration in the global economy: a survey of the ‘new economic geography’. World Econ 21(6):707–731 Preissl B, Solimene L (2003) The dynamics of clusters and innovation: beyond systems and networks. Physica Verlag, Heidelberg Saxenian A (1990) Regional networks and the resurgence of silicon valley. California Manage Rev 33(1):89–112 Schelling TC (1978) Micromotives and macrobehavior. W.W. Norton, New York Simon HA (1984) On the behavioral and rational foundations of economic dynamics. J Econ Behav Org 5(1):35–55 Stoneman P (1995) The handbook of economics of innovation and technological change. Blackwell, Cambridge, MA Storper M, Venables AJ (2004) Buzz: face-to-face contact and the urban economy. J Econ Geogr 4(4):351–370 Swann GMP, Prevezer M, Stout D (eds.) (1998) The dynamics of industrial clusters: international comparisons in computing and biotechnology. Oxford University Press, Oxford Terna P (1998) Simulation tools for social scientists: building agent based models with swarm. J Artif Soc Soc Simulation 1(2):http://jasss.soc.surrey.ac.uk/1/2/4.html Van Duijn JJ (1983) The long wave in economic life. Allen and Unwin, London Winter SG (1987) Knowledge and competence as strategic assets. In: Teece DJ (ed.) The competitive challenge: strategies for industrial innovation and renewal. Ballinger, Cambridge, MA, pp 159–184
Knowledge-Based Economy and Knowledge Creation: The Role of Space Roberto Camagni and Roberta Capello
1 Introduction For some decades there has been a wide consensus among economists on the role played by non-material resources in economic growth. Knowledge has been especially pinpointed as the main discriminating element in economic performance. In the globalising economy, even regional competitiveness – and consequently regional growth – is no longer dependent on the traditional production resource endowment, capital and labour. The hyper-mobility that nowadays characterises these factors reduces their geographical concentration, and shifts the elements on which competitiveness rests from the availability of material resources to the presence of immobile local resources like local culture, competence, innovative capacity; in general knowledge. Much research has been produced since the eighties on the idea of a knowledge-based economy, and on the preconditions for knowledge creation. However, when one looks carefully into the existing literature, two striking aspects emerge. On the one hand, it appears evident that the knowledge-based economy does not have a unique interpretative paradigm, but has been defined on the basis of different approaches ranging from the earliest sectoral, through a more recent functional to the latest relation-based one. As a consequence, the term is still vague and not precisely defined and rather different policy suggestions have been highlighted. On the other hand, it appears quite evident that the different approaches to the concept share one common element, that of the central role played by spatial elements in the creation and diffusion of knowledge, both evidenced by empirical analyses or deductively derived from theoretical elements. Different reasons were given for the importance of space in the creation of a knowledge-based economy: externalities stemming from urban environment, knowledge spillovers subject to strong and visible distance-decay effects, collective learning based on a relational space where economic and social interactions take place and are embedded into geographical space. U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
145
146
R. Camagni, R. Capello
This new approach links to the idea that knowledge develops and accumulates through slow individual and collective learning processes, and grows through information, interaction and local knowledge. Knowledge creation is therefore a local process, rooted in the historical development of the area, accumulated over time through experience, local culture in local labour market and local context, and therefore difficult to transfer to somewhere else. The aim of the paper is twofold. First of all, to present a review of the literature on the knowledge-based economy and the preconditions for knowledge creation. The paper provides a classification of theoretical approaches to the concept and highlights the role of space in each approach. An innovative reflection on the preconditions for a knowledge-based economy is presented (Sects. 2 and 3) especially for the last approach which takes the role of space more directly into consideration. Secondly, the paper aims at validating the new theoretical reflections on the preconditions for a knowledge-based economy through an empirical analysis (Sects. 4 and 5). The paper ends with some concluding remarks and some policy implications (Sect. 6).
2 Approaches and Definitions: The Role of Space Although early use of the term goes back to the work of Fritz Machlup (1962), only in recent years has the concept of the knowledge-based economy begun to spread in the scientific and political literature. This is, mainly due to work sponsored by the OECD (David and Foray 1995; Foray and Lundvall 1996). The European Union set itself the goal in 2000 of becoming the most competitive and dynamic knowledgebased economy in the world. It subsequently confirmed that goal in 2005, submitting its Structural Fund resources to achieve it. But what does this concept really mean? Vaguely, we know that research, human capital, creative utilisation of scientific concepts and information should merge, giving rise to continuing innovation and advanced production. The OECD suggested using about sixty indicators – among which R&D and high technology activities play a dominant role – to measure the knowledge-based economy (OECD 2004; Van Oort and Raspe 2006). In the history of the concept, sector-based definitions and function-based definitions were successively proposed and held for long times (Table 1). While human capital has always been considered as a basic condition for any knowledge-based development, different factors were indicated as the driving forces of change. In an early stage, that can be located in the late 1970s and 1980s, most attention was directed to “science-based” (Pavitt 1984) or high-technology sectors; regions hosting these sectors were considered as “advanced” regions leading the transformation of the economy. New jobs were expected mainly from these new sectors, while more traditional sectors were expected to restructure or even to flow off-shore, giving rise to serious tensions in the local labour markets.
Knowledge-Based Economy and Knowledge Creation: The Role of Space
147
Table 1 Alternative approaches to the knowledge-based economy Sector-based definitions (’70–’80)
Function-based definitions (’80–’90)
Relation-based definitions (’90–2000)
Science-based sectors, High-tech sectors
High education, R&D
Cognitive capability
Advanced regions Invention-innovation short circuit Radical innovation, Schumpeterian profits
Scientific regions Spin-offs, spatial spillovers Technological breakthrough, royalties on patents
Spatial context
High-tech clusters
Role of space
Proximity economies, specialisation advantages
Science parks, large city-regions Proximity and agglomeration economies
Learning regions Collective learning, local synergies Continuing innovation, productivity increases Innovative milieux, large cities Uncertainty reduction, relational capital
Driving forces of the knowledge-based economy Location regions Path towards innovation From innovation to performance
It soon became evident that the dichotomy was too simplistic, and that many knowledge-based advances were possible and were actually introduced by “traditional” sectors – such as textiles and car production – in their path towards rejuvenation. Furthermore, complexification of technological fili`eres inside the value chain increasingly underlined the relevance of advanced tertiary sectors. These supplied producer services mainly in the form of consultancy for process innovation (proper acquisition and use of advanced technologies, tailor-made software, systems integration in production, administration and logistic processes, organisational support) and for product innovation (marketing, design, testing, advertising, finance, distribution). In the second stage which developed mainly during the 1980s and 1990s, a function-based approach was preferred (even though it overlapped conceptually with the previous one), which stressed the importance of pervasive and horizontal functions like R&D and high education. “Scientific” regions, hosting large and well-known scientific institutions, were studied deeply and relationships between these institutions and the industrial fabric were analysed, with some disappointment as far as an expected but not often visible direct linkage was concerned (MacDonald 1987; Massey et al. 1992; Monk et al. 1988; Storey and Tether 1998). Indicators of R&D inputs (like public and private research investment and personnel) and increasingly indicators of R&D output (like patenting activities) were used in order to understand the engagement of firms and territories on knowledge, intended as a necessary long term precondition for continuing innovation (Dasgupta and Stiglitz 1980; Antonelli 1989; Griliches 1990). This approach, equating knowledge and scientific research, was recently relaunched by the well-known European strategy defined in the Lisbon and Luxembourg Ministerial meetings (2000 and 2005), that engages the Union to
148
R. Camagni, R. Capello
Fig. 1 The “Lisbon Performance” of European regions
become the most competitive and dynamic knowledge-based economy in the world. A complex indicator for regional achievement of the Lisbon performance was circulated in the Luxembourg meeting, concentrating on private R&D investment and expenditure, educational level of the labour force and productivity level (Fig. 1). An increasing flow of public resources into the scientific research system is requested at present (Mid 2007) and likely to be consented to by public authorities, giving rise to a huge scientific engagement into the measurement of the internal efficiency, productivity and impact of the research system itself (Okubo 1997; Joly 1997; Bonaccorsi 2003). It is difficult to escape the impression that both the sector-based and the functionbased approaches to the knowledge-based economy, both driven by the need to measure and quantify, resulted in a reductive and simplified picture of the complex nature of knowledge creation and its relation to inventive and innovative capability. The presence of advanced sectors and advanced functions like R&D and higher education are special features of only some of the possible innovation paths and, though relevant, cannot be considered as necessary or sufficient preconditions for
Knowledge-Based Economy and Knowledge Creation: The Role of Space
149
innovation. Furthermore, emphasising the stock of human capital, advanced functions and sectors may risk overlooking the interactive process between the different actors of knowledge development, which is increasingly seen as the crucial element in knowledge creation and evolution. This element is typical of production contexts characterized by the presence of SMEs but also of contexts where big firms develop their own internal knowledge, culture and know-how by enhancing internal interaction and boosting selective external interaction with industrial partners, schools, professionals and research centers. Therefore, a rather different approach should be utilised, a cognitive one, stressing the relational, cultural and psychological elements that define the preconditions for knowledge creation, development, transmission and diffusion. A third stage of reflection exists, typical of the present in which a relation-based approach is preferred, concentrated on the identification of a “cognitive capability” (Foray 2000): the ability to manage information in order to identify and solve problems, or, more precisely in the economic sphere, the ability to transform information and inventions into innovation and productivity increases, through co-operative or market interaction. The “learning” region is identified as the place where such cognitive processes play a crucial role, combining existing but dispersed know-how, interpretations of market needs, information flows with intellectual artifacts such as theories and models and allowing exchange of experiences and co-operation (Lundvall and Johnson 1994). Especially in contexts characterised by a plurality of agents – like cities or industrial districts – knowledge evolution “is not the result of individual efforts in R&D within individual firms, but rather the combination of complementary capacities and of widespread interactive learning processes, which involve many ‘customers’ and ‘suppliers’ along a well-defined fili`ere or supply chain” (Cappellin 2003a). The third approach is very different from the previous ones as far as the path towards innovation is concerned. In the first, the sector-based approach, attention is focused on an invention-innovation short circuit taking place inside individual firms (or their territories) operating on advanced sectors. R&D facilities are strictly linked to production facilities, while firms tend to cluster inside high-tech districts in order to take advantage of all sorts of proximity externalities. In the function-based approach, a sort of division of labour operated between R&D/higher education facilities on the one hand and innovating firms on the other. Their interaction produced academic spin-off or knowledge spillover flowing from the former to the latter (Acs et al. 1994; Audretsch and Feldman 1996; Anselin et al. 2000). In the third, more recent, approach, attention is focused mainly on the construction of knowledge through cooperative learning processes, nourished by spatial proximity (“atmosphere” effects), network relations (long-distance, selective relationships), interaction, creativity, and recombination capability. A collective learning process of this kind was first hypothesized by the GREMI group (Camagni 1991; Perrin 1995) and subsequently widely adopted as a sound theoretical concept for the interpretation of knowledge-based development and innovation (Keeble and Wilkinson 1999, 2000; Capello 1999; Cappellin 2003a).
150
R. Camagni, R. Capello
What is striking in all these approaches is the central role played by spatial elements, both evidenced by empirical analyses or deductively derived from theoretical elements. The tendency of high-technology activities to cluster along valleys, corridors, glens, and high-tech districts was early empirical evidence: externalities coming from the presence of advanced education facilities were invoked to explain these facts, but international accessibility, advanced urban atmosphere, traditional industrial competencies under reorientation (Malecki 1980; Saxenian 1996) were also suggested. The role of space in function-based approaches is relevant in two respects. Firstly, space acts as a strong concentration mechanism of advanced facilities, which are mainly located inside large agglomerations or city-regions in order to benefit from scale effects in both input markets (human capital, private financial capital) and output markets (higher-education services, research services) (Scott 2001); secondly, space acts as a driver of knowledge spillover from R&D clusters, which are subject to strong and visible distance-decay effects. In both cases, space is treated in a widely abstract, indirect and stylised way: concentration and agglomeration of main R&D facilities are by and large assumed to be acceptable starting points for empirical analyses, while knowledge diffusion processes are analysed in terms of pure probability functions, decreasing with physical distance. Spatial spillover effects are considered as a black box, with no reference to real, territorialized, channels of direct knowledge interaction (Capello and Faggian 2005). The cognitive approach takes up this last challenge more directly. Knowledge flows and information channels are investigated, and the role of the local milieu becomes clear: abstract space becomes real territory, a relational space where functional and hierachical, economic and social interactions take place and are embedded into geographical space. The local milieu – a “territory” identified by both geographical proximity (agglomeration economies, district economies) and cognitive proximity (shared behavioural codes, common culture, mutual trust and sense of belonging) – supplies the socio-economic and geographical substrate on which collective learning processes can be incorporated, mainly due to two main processes (Camagni and Capello 2002): – The huge mobility of professionals and skilled labour – between firms but internally to the local labour market defined by the district or the city, where this mobility is maximal), and – The intense co-operative relations among local actors, and in particular customer– supplier relationships in production, design, research, and finally knowledge creation The milieu becomes therefore a “cognitive engine” and possibly an innovation place: its characteristics enhance interaction and co-operation, reduce uncertainty (especially concerning the behaviour of competitors and partners), reduce information asymmetries (therefore reducing mutual suspicion among partners) and reduce probability of opportunistic behaviour under the threat of social sanctioning (Camagni, 1991, 2004), all elements that are confirmed by many regional economics schools (Bellet et al. 1993; Rallet and Torre 1995; Cappellin 2003b).
Knowledge-Based Economy and Knowledge Creation: The Role of Space
151
The foregoing concerning the role of territorial variables and the centrality of local conditions should not be taken as suggesting a return to an anti-historical localism or territorial autarchy. On the contrary, local milieux should be perfectly accessible, open and receptive to external flows of information, knowledge, technologies, organisational and cognitive models, and always be ready to recombine local knowledge and external knowledge anew. What is really meant by referring to the importance of local territories is the fact that, while some relevant production factors like financial capital, general information, consolidated technologies and codified knowledge are readily available virtually everywhere nowadays, the ability to organise these “pervasive” factors into continuously innovative production processes and products is by no means pervasive and generalised, but exists selectively only in some places where tacit knowledge is continuously created, exchanged and utilised and business ideas find their way to real markets.
3 Knowledge Creation: Processes and Policies Let us now look deeper into the new approach to the knowledge economy. From the foregoing discussion, knowledge development appears as: – A cognitive process involving the whole society – An interactive process within the firm, involving its main departments and functions – A collective process within the local milieu – A process intrinsically non linear and subject to synergetic effects – Based on what may be called the local “relational capital” (Camagni 1999; Cappellin 2003a), made up of social networks and cohesion, variety of actors, external cooperation networks, external accessibility and integration, tradition of public/private partnership, receptiveness to external stimuli While codified knowledge is transmitted through information networks, without relevant spatial impedence, tacit knowledge is created by chains of personal links, highly sensitive to proximity. The traditional and the new view of knowledge and its use in the economy may be summarised as in Table 2. According to the traditional view, knowledge is mainly privately owned, developed through R&D investments and incorporated into a technology or a product. Diffusion follows a predictable probability function, subject to a distance decay effect. On the other hand, the modern view puts learning and interaction processes at the forefront, and considers knowledge as complex semi-public or co-operative, its diffusion is subject to strong spatial barriers and follows widely unpredictable creative processes. While according to the traditional view technological adoption creates huge adjustment costs in the local communities, in the new view, on the other hand, it depends exactly on the willingness and intentions of the local community. The effects of the new approach are very clear and visible as far as the policy implications are concerned: while the first approach put emphasis on reduction of
152
R. Camagni, R. Capello
Table 2 Traditional and new views about knowledge The traditional view: knowledge as a product
The modern view: knowledge as a cognitive process
Incorporated in a technology Produced through R&D investments Bought on the market
Developed through a learning process Nourished by interaction Transferred through complex co-operation processes Mainly publicly or co-operatively owned Subject to strong spatial barriers and unpredictable innovative processes Territorially important: local preconditions (the local milieu model) Development depending on territorial context and communities Policy goals: improving spatial context for knowledge development
Mainly privately owned Diffusing according to predictable probability functions Space important o the interregional scale (core/periphery model) Adoption impacting on workers and communities Policy goals: speed up adoption and reduce social costs Source: Adapted from Cappellin (2004)
social resistence and costs, the new approach emphasizes the improvement of local relational capital, internal interaction, accessibility and absorptive capacity, identity and creativity (Cappellin 2003a). The cycle of knowledge creation and diffusion may be sketched in the following phases, where spontaneous processes interact with appropriate education and vocational training policies: 1. General education and receptiveness: understanding and internalisation of codified knowledge 2. Research/Invention, or the creation of new codified knowledge through the original combination of existing elements of codified knowledge 3. Transcoding of codified knowledge for the use of entrepreneurs and firms 4. Application of codified knowledge to the solution of specific application problems 5. Vocational training and targeted education: transformation of codified knowledge into individual competencies and skills 6. Interactive learning processes through the exchange of tacit knowledge among individuals 7. Development of tacit knowledge within organisations, institutions 8. Collective learning and territorial diffusion of tacit knowledge through co-operation, imitation, mobility of skilled labour, chains of professional mobility 9. Codification of tacit knowledge through socialised activities (agencies for technolology transfer, co-operative research policies at territorial level) Source: adapted and integrated from Cappellin (2004)
Knowledge-Based Economy and Knowledge Creation: The Role of Space
153
Integrated R&D and higher education institutions
Society, Competence and General Education
Knowledge - oriented entrepreneurship and learning oriented labour market
Science and Technology
Economy and Entrepreneurship
Efficient transcoding and transfer system
Fig. 2 An integrated approach to the knowledge society: the three relational pre-conditions
In a simplified scheme, some functional preconditions are needed for knowledge creation and diffusion, namely: – – – –
Competence (“civilisation mat´erielle”) General education and higher education of human capital R&D investment, and investment in science Dynamic entrepreneurship
But more importantly, these preconditions, which are embedded in the three main sub-systems of society – the education system, the research system and the economic system – have to integrate and interact with each other, giving rise to three crucial “relational” preconditions (Fig. 2): – Integration between R&D institutions and the higher-education system – Efficient transcoding and transfer system to translate the research output into a language that firms can understand and use – Knowledge-oriented entrepreneurship and a learning oriented labour market In the policy sphere, the recent French experience of the “pˆoles de competitivit´e” (competitiveness poles) is in our opinion the best example of a policy intervening on our three relational preconditions, namely the integration area between the three subsystems, and not directly on each sub-system: public resources are allocated to projects developed on local territories through cooperative agreement between universities, research centres and firms, with general monitoring by local public authorities.
154
R. Camagni, R. Capello
This conceptual approach is tested empirically in the next part of this paper; through an econometric exercise. The idea is to describe and interpret the relevance of the relational preconditions for knowledge development and the more general relevance of local relational capital – in the form of local linkages oriented towards the innovation process.
4 Database, Indicators and Methodology for the Empirical Analysis As is always the case when dealing with theoretical frameworks, the statements made should achieve empirical validation. From the above framework of analysis, the following main assumptions require to be proved empirically: 1. The simple and traditional functions like investment in science, entrepreneurship, education and general culture are not the main determinants for the development of a knowledge society. The integration of such functions as efficient transcoding and transfer system or/and knowledge oriented entrepreneurship is a much better precondition for local knowledge accumulation and local science-based development. We expect, therefore, to find a link between the existence of such integration and processes of knowledge accumulation. 2. The role played by such “integrated functional preconditions”, as we label them, increases in territories where the spontaneous attitude of local economic actors to integrate and to cooperate is higher. Therefore, we expect to find a greater role played by integrated functional preconditions on knowledge accumulation for those actors (and territories) having higher relational capital – the latter intended as the set of norms and values which govern interactions among people, the institutions where they are incorporated, the relationship networks set up among various social actors and the overall cohesion of society (Camagni 1991). The aim of this part of the work is, therefore, to present the results of the empirical analysis testing these two hypotheses. This is not a simple task, given the high degree of abstraction of the assumptions made, and the low measurability of the phenomena described. The task was carried out by the construction of a primary source database, which allowed us to define some proxies for the phenomena we were interested in. The database was built on an ad-hoc questionnaire to 160 firms located in two provinces in Italy, Pisa and Genova, chosen for their nationally-known specialisation in hightech sectors, and therefore seen as two knowledge societies. The questionnaire contained questions regarding: – Firm’s characteristics: year of establishment, number of employees, a variety of economic performance indicators (e.g. sales, exports), competitive position, knowledge base – Firm’s innovative capacity: output (patents, percentage of firms’ turnover related to product and process innovation) indicators of innovation
Knowledge-Based Economy and Knowledge Creation: The Role of Space
155
– Knowledge-oriented entrepreneurship. As has been widely emphasised in the literature, a firm’s openness to new innovation (which depends on its ability to exploit new knowledge) depends to a significant extent on the level of prior related knowledge stored within the firm, which enables the value of new information to be recognised, assimilated and applied for commercial purposes. These abilities collectively constitute what has been labelled “absorptive capacity” (Cohen and Levinthal 1990).1 Measures of knowledge-oriented entrepreneurship are therefore measures of absorptive capacity: R&D expenditures over the last five years – Territorial relational capital: information about the degree of importance of local and external knowledge sources (e.g. competitors, providers, customers, universities, knowledge facilitators, etc.) to recent product innovations of local firms – Local science park efficiency, i.e. of the efficiency of the transfer and transcoding system: the role of science parks in transferring strategic knowledge (i.e. knowledge useful for the innovation activities of firms), either directly, by transferring their knowledge to firms, or indirectly by connecting their customers with relevant actors able to develop important innovation for their customers. For the former case, information about frequency, typology of information/knowledge accessed through science parks, obstacles to knowledge acquisition, has been collected The survey covering 160 firms distributed equally between the two areas of Pisa and Genova (80 firms each) was conducted over a period of three months.2 Table 3 reports the description of variables included in the empirical analysis. A set of conventional variables on innovation capacity, firm size and relational capital Table 3 Description of variables Description
Construction
Product innovation
Binary dummy coded 1 if firm introduced at least one product innovation over the previous five years, 0 otherwise; Binary dummy coded 1 if firm had more than 20 employees, 0 otherwise; Binary dummy coded 1 if the degree of importance of local links to innovative activity was above the mean, 0 otherwise; Binary dummy coded 1 if a firm’s R&D expenditure was above the sample mean, 0 otherwise; Binary dummy coded 1 if firms valued the role of the science park as a relevant factor for searching and identifying local sources of knowledge to be applied in innovative processes, 0 otherwise.
Firm size Relational capital Absortive capacity Efficient transfer and transcoding system
1
More recently absorptive capacity has been reinterpreted in terms of knowledge-relatedness in order to take account of technological diversification processes occurring within firms (Breschi et al. 2003). 2 Questionnaire were submitted through face-to-face interviews with firms’ technical staff and owners. The survey was administrated between November 2003 and January 2004.
156
R. Camagni, R. Capello
were built; to this set, a group of variables was added, with the aim of capturing the key integrated functional preconditions to knowledge development, i.e. an efficient transfer and transcoding system, knowledge-oriented entrepreneurship and territorial relational capital oriented to higher order functions. The following dummy variables were built as indicators of the phenomena to be described in the empirical part, namely (Table 3): – An indicator of breakthrough innovation (or product innovation), equal to 1 if a firm had introduced at least one product innovation over the previous five years, 0 otherwise – An indicator of firm size, coded 1 if a firm had above 20 employees, 0 otherwise – An indicator of relational capital, which assumed value 1 if the degree of importance of local links to innovative activity was above average, 0 otherwise – An indicator of knowledge-oriented entrepreneurship or of absorptive capacity, with a value 1 if a firm’s R&D expenditure was above the sample mean, 0 otherwise – An indicator of an efficient transfer and transcoding system, coded 1 if firms value the role of the science park as a relevant factor for searching and identifying local sources of knowledge to be applied in innovative processes, 0 otherwise These variables have been used in the empirical analysis. The analysis was first run on the whole database in order to test the above-mentioned hypotheses at the firm level (sect. 5.1). In the second stage, the database was split into the two geographical areas, which differ greatly in terms of sectoral and firm characteristics, and particularly in terms of relational capital. In this way, the above assumptions could be tested in areas with different degrees of relational capital (sect. 5.2). The methodology chosen was generally descriptive, given the low number of observations available. When we are able to use the whole database, some interpretative analyses can be applied.
5 Integrated Functional Preconditions and the Creation of a Knowledge Society: The Role of Space 5.1 An Empirical Investigation at the Firm Level In this section, the results of the empirical investigations at the firm level are presented. The first descriptive results are obtained by running a cluster analysis, which allows homogeneous groups of firms presenting heterogeneous behaviour in the phenomena under analysis to be singled out. In our case, firms were grouped according to similar behaviour in terms of knowledge creation, relational capital, knowledge acquisition from science parks and internal accumulation of knowledge.
Knowledge-Based Economy and Knowledge Creation: The Role of Space
157
Table 4 Cluster analysis results Clusters
1
2
3
4
Size (1 large firms, 0 small firms) Product innovation Absorptive capacity Relational capital Efficient transfer and transcoding system Firms location (1 = Pisa; 0 = Genova)
74.3 31.4 20 17.1 8.7 28.6
28.9 52.6 44.7 44.7 31.4 65.8
0 30.6 33.3 31.9 14.8 45.8
13.3 73.3 73.3 60 80 80
24.4 40 40 34.4 25.4 50
Sectorsa : Others Specialised suppliers Supplier dominated Scale intensive Science based Low skilled services High skilled services
5.9 5.9 17.6 41.2 8.8 2.9 17.6
2.6 2.6 10.5 31.6 10.5 2.6 39.5
1.4 5.6 13.9 29.2 11.1 1.4 37.5
0 26.7 6.7 0 6.7 6.7 53.3
2.5 6.9 13.2 29.6 10.1 2.5 35.2
Number of observations within each cluster
35
38
72
15
a We
Mean
160
introduce two additional classes to the Pavitt taxonomy (Pavitt 1984).
Each of the four groups obtained through the cluster analysis is characterised by firms showing similar behaviour within the group, and having instead very different behaviour to firms belonging to other groups (Table 4). The four groups can be described as follows: 1. The first group includes large firms specialised in traditional sectors with a product innovation capacity lower than average: this cluster is characterised at the same time by lower than average absorptive capacity, very low relational capital and relatively low acquisition of knowledge from science parks 2. The second group is populated by medium-sized innovative firms: interestingly enough, these firms are also characterised by high (higher than the average) absorptive capacity, high relational capital and, last but not least, by a high acquisition of knowledge through science parks 3. The third group contains small firms with a very low innovative capacity. In this group, relational capital, absorptive capacity and knowledge acquisition from science parks are lower than average 4. The fourth group is characterised by the most innovative firms (mostly small firms) of the whole sample. These firms are also defined by very high absorptive capacity, very high relational capital and a high capacity for acquiring knowledge from science parks A descriptive way to analyse the first of our assumptions, i.e. whether product innovation is linked to specific integrated functional preconditions, is to represent the four clusters in a chart where innovation capacity is plotted against absorptive capacity of firms or transfer and transcoding system efficiency, depicted in our analysis as the role played by science parks in generating strategic knowledge for innovation activities of firms.
80 70 60 50 40 30 20 10 0
R. Camagni, R. Capello cluster 4 cluster 2 cluster 3 cluster 1
0
20
40
60
80
Product innovation
Product innovation
158
100
80 70 60 50 40 30 20 10 0
cluster 4 cluster 2 cluster 3
cluster 1
0
10
20
30
40
50
60
70
Absorptive capacity
Efficient transfer and transcoding system
(a)
(b)
Fig. 3 Functional preconditions for a knowledge society: comparison of clusters of firms: (a) Efficient transfer and transcoding system and product innovation, (b) Absorptive capacity and product innovation Table 5 Logit results on the whole sample Dependent variable: breakthrough innovation Independent variables: Efficient transfer and transcoding system Absortive capacity Constant
Coefficienta
Std. Err.
Wald-test
0.94 (0.024) 1.04 (0.040) −0.67 (0.002)
0.417
5.07
0.512
4.19
0.222
9.29
Pseudo R-sq. 0.10 Probability of innovation 79% a P-values
into brackets.
Figure 3 shows the relationship between breakthrough innovation and, transfer and transcoding system (part a) and absorptive capacity (part b) respectively. The positive relationship between the two phenomena emerges. Cluster 4, characterised by firms with higher innovation capacity, is also characterised by firms with high absorptive capacity; at the same time, cluster 4 is the one making the most efficient use of the transfer and transcoding system. By the same token, clusters 2, 3 and 1 have, in decreasing order of magnitude, lower levels of innovation capacity, accompanied by lower levels of absorptive capacity, and by less efficient use of transfer and transcoding systems. A confirmation of the positive and significant relationship between these variables and the innovative capacity of firms is provided by a logit analysis presented in Table 5, where the control variable of the size of firms, not significant in the regression, is not reported. Breakthrough innovation in this case is explained by the presence of both the absorptive capacity and the efficiency of the transcoding system. Both interpretative variables have a positive and significant sign, showing the role they play in explaining the innovative capacity of sample firms. From the coefficients, one can calculate the probability that a firm will develop a breakthrough innovation: the results show that the probability of a firm innovating is 79% if a firm
Knowledge-Based Economy and Knowledge Creation: The Role of Space
159
both exploits an efficient transcoding system and has a well developed absorptive capacity (Table 5). The second assumption to be tested was whether the relationship between innovation capacity and the presence of integrated functional preconditions is more evident in firms with a higher relational capital. In descriptive terms, this can be shown by Fig. 4, in which relational capital shows a clear relationship to product innovation processes. In order to prove a relationship between innovation capacity and integrated functional preconditions with an interpretative methodology, a logit regression model was run inserting the interaction terms between absorptive capacity and relational capital and between an efficient transfer and transcoding system and relational capital.3 The results are shown in Table 6, with the exception of the firm size variable,
80 cluster 4
Product innovation
70 60 cluster 2
50 40 cluster 3
cluster 1
30 20 10 0 0
10
20
30
40
50
60
70
Relational capital Fig. 4 Relational capital and product innovation: comparison of clusters of firms Table 6 Logit results on firms with high relational capital Dependent variable: breakthrough innovation Independent variables: Efficient transfer and transcoding system × high relational capital Absortive capacity × high relational capital Constant
Coefficienta
Std. Err.
0.939 (0.091) 1.26 (0.077) −0.59 (0.001)
0.55
2.86
0.71
3.12
0.182
Wald-test
10.8
Pseudo R-sq. 0.07 Probability to innovate 84% a P-values
into brackets.
3 The logit model estimated with both the single variables and the interactive terms of the single variables with the relational capital show that single variables lose part of their explicative power while interactive variables do not achieve significant values.
160
R. Camagni, R. Capello
which was not significant. The interactive variables are both positive and significant, and indicate that the probability that a firm develops a product innovation is higher (84% versus 79%) than in the model with no interaction terms (Table 5). Relational capital reinforces the role played by both variables, increasing the probability of a breakthrough innovation occuring in firms.
5.2 An Empirical Investigation at the Territorial Level An additional interesting empirical test is to see whether the above-mentioned positive relationship between product innovation and the integrated functional preconditions holds is to repeat the analysis by splitting the database into the two territorial areas, Pisa and Genova. It is widely known that the two areas differ greatly in terms of firm composition, sectoral composition and local industrial atmosphere. Genova is a typical large firm area, operating in heavy industrial sectors, undergoing extensive restructuring after the serious industrial crisis in these sectors. Pisa is a small firm area, highly specialised in high-tech sectors, and with an industrial atmosphere typical of an industrial district: a large number of small and highly specialised firms operate in cooperation with other local firms, and a feeling of belonging to a particular area is rooted in local people and agents. With the exception of the latter, for which no indicators exist in official statistics, the different size and sectoral composition is shown by official data. Table 7 provides these data, and shows the large discrepancy between the two local economies. Our sample reflects this discrepancy, too (Table 8). The highest share of large firms in our sample is located in Genova, mainly belonging to scale-intensive sectors. Small firms are instead more present in Pisa; 85% of the sample firms in Pisa have less than 20 employees, against 66% of the sample in Genova; 62.5% of science-based firms and 76.8% of high skilled service firms in the sample are located in Pisa. Moreover, our sample shows that the relational capital is much higher in Pisa than in Genova, the share of firms with strategic (innovative) linkages with their local suppliers being much higher in Pisa (45%) than in Genova (8.8%). From these data, which reflect reality, we can define Pisa as a territory with higher relational capital than Genova. Given that this is the case, it is interesting to see what the relationship between product innovation and functional preconditions is in the two areas. Table 9 shows the descriptive results of our database split into the two areas. It is easy to envisage a positive relationship between an efficient transfer and transcoding system and product innovation capacity; by the same token, higher degrees of absorptive capacity are related to higher product innovation. Moreover, Table 6 shows that innovation capacity is greater in Pisa, where relational capital is much more prominent than in Genova. The results of the logit model run on the Pisa and Genova samples provide prima facie evidence that the role played by the integrated functional preconditions on
Knowledge-Based Economy and Knowledge Creation: The Role of Space
161
Table 7 Characteristics of the two areas analysed – 2001 Variables
Pisa (NUTS 3 level) Genova (NUTS 3 level)
Absolute percentage share: Firmsize: <20 employees Between 20 and 50 >50 employees
66.1 12.5 21.4
62.0 10.0 27.8
Employees in: Others Specialised suppliers Supplier dominated Scale intensive Science based Low skilled services High skilled services
41.7 6.04 18.2 2.8 2.3 17.2 11.4
38.4 4.02 16.1 6.5 0.9 19.5 14.2
Firm size: <20 employees Between 20 and 50 >50 employees
1.11 1.04 0.80
1.00 0.84 1.04
Employees in: Others Specialised suppliers Supplier dominated Scale intensive Science based Low skilled services High skilled services
1.24 0.74 0.96 0.41 1.42 0.96 0.88
1.14 0.49 0.85 0.95 0.60 1.09 1.10
L.Q. (relative to the national average):
a We
introduce two additional classes to the Pavitt taxonomy (Pavitt 1984).
firms’ innovativeness is a positive and significant one. As Tables 10 and 11 show, the two independent variables have the expected positive sign. Moreover, the results of the logit models show that Pisa has a higher probability of firms innovating (84%) than Genova (68%) (Tables 10 and 11 respectively). The difference in the probability of innovating is due to the difference in relational capital measured in the two areas: Pisa, a higher milieu area, in which relational capital is more widespread, demonstrates a higher probability than a more urban area like Genova.
6 Conclusions The aim of the paper was to present some reflections on the main preconditions for a local knowledge-based economy to develop, i.e. the main preconditions upon which the competitiveness of regions depends heavily.
162
R. Camagni, R. Capello
Table 8 Sample distribution by territories Variables Firm size distribution within areas: <20 employees Between 20 and 50 >50 employees Total within areas Firm size distribution between areas: <20 employees Between 20 and 50 >50 employees
Pisa (NUTS 3 level) % 85 10 5 100
Genova (NUTS 3 level) % 66.3 12.5 21.2 100
56.2 44.4 19
43.8 55.6 81
Sectors within areas a : Specialised suppliers Supplier dominated Scale intensive Science based Low skilled services High skilled services Others Total within areas
10.1 3.8 11.4 12.7 2.5 54.4 5.1 100
3.8 22.5 47.5 7.5 2.5 16.3 0 100
Sectors between areas a : Specialised suppliers Supplier dominated Scale intensive Science based Low skilled services High skilled services Others
72.7 14.3 19.1 62.5 50 76.8 100
27.3 85.7 80.9 37.5 50 23.2
45
8.8
Relational capital b a We
introduce two additional classes to the Pavitt taxonomy (Pavitt 1984). of firms that have replied that local links are helpful for their innovative activity.
b Share
Table 9 Functional preconditions for a knowledge society: Comparison of two territories Variables of functional preconditions Absorptive capacity Efficient transfer and transcoding system Relational capital Product innovation
Pisa (NUTS 3 level) %
Genova (NUTS 3 level) %
21.3 82.6 45 76.7
7.5 27.4 8.8 35
The reflections started from the analysis of the existing literature on the knowledge-based economy; the conceptual framework provided suggested that the determinants of a local knowledge society are not embedded in investment in innovation, general education or local entrepreneurship, as widely believed. The paper instead put much emphasis on the importance of interaction, integration and synergy among
Knowledge-Based Economy and Knowledge Creation: The Role of Space
163
Table 10 Logit results on the sample of Pisa Dependent variable: breakthrough innovation Independent variables: Efficient transfer and transcoding system Absortive capacity Constant
Coefficienta
Std. Err.
0.98 (0.034) 0.66 (0.21) −0.63 (0.001)
0.46
4.48
0.54
1.6
0.18
11.55
Wald-test
Pseudo R-sq. 0.06 Probability of innovation 84% a P-values
into brackets.
Table 11 Logit results on the sample of Genova Dependent variable: breakthrough innovation Independent variables: Efficient transfer and transcoding system Absortive capacity Constant
Coefficienta
Std. Err.
1.91 (0.016) 1.20 (0.016) −2.33 (0.000)
0.79
5.81
0.64
3.48
0.32
Wald-test
50.9
Pseudo R-sq. 0.11 Probability of innovation 68% a P-values
into brackets.
the well-known functional preconditions, since these are the ones which make the difference in the process of knowledge creation. An efficient local transfer and transcoding system of knowledge between the scientific world and the industrial sphere stemming from a cooperative attitude of the scientific and the industrial worlds is one strategic precondition. By the same token, entrepreneurship open to new culture and new competence acquisition is more important than the mere presence of entrepreneurs. Last, but not least, a scientific world integrated with the needs of an evolving society, where R&D activity is carried out in cooperation between scientific and academic institutions, and where higher education is fed by scientific experience, is the most important ingredient for a local knowledge society to develop. These integrated elements need a strong cooperative and synergic attitude by local actors (population but also private and public economic agents) in order to happen and for this reason, they are expected to develop more easily among actors and in territories which have already a high relational capability. The paper presents an analysis in which these elements are empirically described. The empirical results of the paper show that in the presence of a more efficient
164
R. Camagni, R. Capello
transfer and transcoding system and of a more open attitude of firms towards innovation, the innovation capacity of firms is much higher. This also holds true for territories; those territories with a more efficient transfer and transcoding system and an attitude more open to innovation show a higher degree of innovation. The analysis also shows that this is even more true for those firms and in those territories where there is a greater relational capacity. Policy implications can be drawn from this analysis. Mere investment in R&D activities, in general education or in entrepreneurial activity alone is not expected to give much impetus to the development of a knowledge society. The latter is instead more dependent upon the development of integrated activities between society, the scientific world and the economy for knowledge development. Policy interventions should therefore be devoted to the support of learning-oriented entrepreneurship, learning-oriented labour markets, efficient transcoding and transfer systems, and integrated R&D and higher education institutions. These policies, coupled with a strategy of increasing local synergy and cooperation among local actors, could be useful in helping a local economy move towards a knowledge society, and therefore towards more competitive conditions for growth.
References Acs Z, Audretsch D, Feldman M (1994) R&D spillovers and recipient firm size. Rev Econ Stat 76(2): 336–340 Anselin L, Varga A, Acs Z (2000) Geographic and sectoral characteristics of academic knowledge externalities. Pap Reg Sci 79(4): 435–443 Antonelli C (1989) A failure inducement model of research and development expenditure: Italian evidence from the early 1980s. J Econ Behav Org 12(2): 159–180 Audretsch D, Feldman M (1996) R&D spillovers and the geography of innovation and production. Am Econ Rev 86(3): 630–640 Bellet M, Colletis G, Lung Y (1993) Introduction au Num´ero Sp´ecial sur Economie et Proximit´e, Revue d’Economie R´egionale et Urbaine 3: 357–361 Bonaccorsi A (ed) (2003) Il Sistema della Ricerca Pubblica in Italia. Franco Angeli, Milano Breschi S, Lissoni F, Malerba F (2003) Knowledge-relatedness in firm technological diversification. Res policy 32(1): 69–87 Camagni R (1991) Local milieu, uncertainty and innovation networks: towards a new dynamic theory of economic space. In: Camagni R (ed) Innovation networks: spatial perspectives. Belhaven-Pinter, London, pp 121–144 Camagni R (2004) Uncertainty, social capital and community governance: the city as a milieu. In: Capello R, Nijkamp P (eds) Urban dynamics and growth: advances in urban economics. Amsterdam, Elsevier, pp 121–152 Camagni R, Capello R (2002) Milieux innovateurs and collective learning: from concepts to measurement. In: Acs ZJ, de Groot HLF, Nijkamp P (eds) The emergence of the knowledge economy. Springer, Berlin, pp 15–46 Capello R (1999) Spatial transfer of knowledge in high-technology milieux: learning vs. collective learning processes. Reg Stud 33(4): 353–365 Capello R, Faggian A (2005) Collective learning and relational capital in local innovation processes. Reg Stud 39(1): 75–87 Cappellin R (2003a) Territorial knowledge management: towards a metrics of the cognitive dimension of agglomeration economies. Int J Technol Manage 26(2–4): 303–325
Knowledge-Based Economy and Knowledge Creation: The Role of Space
165
Cappellin R (2003b) Networks and technological change in regional clusters. In: Br¨ocker J, Dohse D, Soltwedel R (eds) Innovation clusters in interregional competition. Springer, Berlin, pp 53– 78 Cappellin R (2004) International knowledge and innovation networks for European integration, cohesion, and enlargement. Int Soc Sci J 56(180): 207–225 Cohen WM, Levinthal DA (1990) Absorptive capacity: a new perspective on learning and innovation. Adm Sci Q 35(1): 128–152 Dasgupta P, Stiglitz J (1980) Uncertainty, industrial structure and the speed of R&D. Bell J Econ 11(1): 1–28 David P, Foray D (1995) Accessing and expanding the science and technology knowledge base. STI Rev 16: 16–38 Foray D (2000) L’Economie de la Connaissance. La D´ecouverte, Paris Foray D, Lundvall BA (eds) (1996) Employment and growth in the knowledge-based economy. OECD, Paris Griliches Z (1990) Patent statistics as economic indicators: a survey. J Econ Lit December 1661– 1707 Joly P-B (1997) Chercheurs et Laboratoires dans la Nouvelle Economie de la Science. Revue d’Economie Industrielle 79(1): 77–94 Keeble D, Wilkinson F (1999) Collective learning and knowledge development in the evolution of regional clusters of high-technology sms in Europe. Reg Stud 33(4): 295–303 Keeble D., Wilkinson F (2000) High technology clusters, networking and collective learning in Europe. Ashgate, Aldershot Lundvall BA, Johnson B (1994) The learning economy. J Ind Stud 1(1): 23–42 MacDonald S (1987) British science parks: reflections on the politics of high technology. R&D Manage 17(1) 25–37 Machlup F (1962) The production and distribution of knowledge in the United States. Princeton University Press, New York Malecki E (1980) Corporate organisation of R&D and the location of technological activities. Reg Stud 14(3): 219–234 Massey D, Quintas P, Wield D (1992) High tech fantasies: science parks in society, science and space. Routledge, London Monk CSP, Porter RB, Quintas P, Storey D, Wynarczyk P (1988) Science parks and the growth of high technology firms, Croom Helm, London OECD (2004) Global knowledge flows and economic development. OECD, Paris Okubo Y (1997) Bibliometric indicators and analysis of research systems: methods and examples. STI Working Papers 1/97, OECD, Paris Pavitt K (1984) Sectoral patterns of technical change: towards a taxonomy and a theory. Res Policy 13(4): 343–373 Perrin J-C (1995) Apprentissage Collectif, Territoire et Milieu Innovateur: un Nouveau Paradigme pour le D´eveloppement. In: Ferr˜ao J (ed) Pol´ıticas de Inovac¸a˜ o e Desenvolvimento Regional et Local, Edic¸a˜ o do Instituto de Ciencias Sociais de Universidade de Lisboa, republished in Camagni R, Maillat D (eds) (2006), Milieux innovateurs. Economica-Anthropos, Paris, pp 99–128 Rallet A, Torre A (eds) (1995) Economie Industrielle et Economie Spatiale. Economica, Paris Saxenian A (1996) Regional advantage: culture and competition in silicon valley and route 128. Harvard University Press, London Scott AJ (ed) (2001) Global city-regions: trends, theory, policy. Oxford University Press, Oxford Storey DJ, Tether BS (1998) Public policy measures to support new technology-based firms in the European union. Res Policy 26(9): 1037–1057 Van Oort F, Raspe O (2006) Economic growth and the urban knowledge economy. Utrecht University Papers in Evolutionary Economic Geography (PEEG), no. 06/07
Systems of Innovation and Regional Growth in the EU: Endogenous vs. External Innovative Activities and Socio-Economic Conditions Riccardo Crescenzi and Andr´es Rodr´ıguez-Pose
1 Introduction The ability of the model of “sustainable economic growth” put forward by the Lisbon Agenda1 to deliver its benefits evenly to the EU regions depends essentially on the capacity of each region to produce and access innovation. In this context the uneven geographical distribution of R&D activities has been regarded as a crucial source of competitive advantage for some areas and the promotion of R&D investment has become a key ingredient of EU regional development policies. EU policies to promote the “knowledge-based” economy have mainly been focused on various forms of support for R&D activities not only for the “production” of new knowledge but also for the economic exploitation of existing knowledge. This emphasis on R&D – a distinctive feature of the “function-based” approach to the process of innovation (see the critical analysis by Camagni and Capello 2009, in this book) – overlooks, however., at least one other important factor which influences the translation of innovation into regional growth i.e. the underlying regional system of innovation conditions. A successful process of innovation depends on “localised structural and institutional factors that shape the innovative capacity of specific geographical contexts” (Iammarino 2005, p. 499), as highlighted by the systems of innovation (Lundvall 2001), regional systems of innovation (Cooke et al. 1997), and learning regions approaches (Morgan 2004; Gregersen and Johnson 1996). The idea of “innovation prone” and “innovation averse” societies (Rodr´ıguez-Pose 1999) also plays an important part. The literature on regional systems of innovation has provided relevant insights into the role of institutional and socio-economic contextual factors for the production 1
The European Council which met in Lisbon in 2000 set forth the goal of making the EU “the most competitive and dynamic knowledge-based economy in the world, capable of sustainable economic growth with more and better jobs and greater social cohesion” (Presidency Conclusions, Lisbon European Council, 23 and 24 March 2000, par. 5).
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
167
168
R. Crescenzi, A. Rodr´ıguez-Pose
of economically successful innovation. However, this literature has mainly adopted a case-study perspective, thus acting as a “focusing device” for factors relevant for the process of innovation. This chapter aims at translating such insights into a quantitative framework by focusing on the socio-economic conditions that make some regions more open to innovation than others: while regional growth is fostered by local innovative activities and benefits from external knowledge flows (which, in turn, are enhanced by the increased interconnection between territories), the underlying socio-economic conditions play a crucial role as well. Not only is the “production” of innovation itself shaped by the system of innovation in place in the local economy but this latter factor also influences the extent and the effectiveness of the diffusion of (un-codified) knowledge. Our analysis, by combining R&D, knowledge flows, and innovation systems approaches in a single model, will show that contextual socio-economic factors need to be addressed when designing innovationbased regional development policies. In addition, the theoretical and empirical tools developed for the analysis of the “social-filter” conditions of the EU regions provide interesting insights into the geography of socio-economic disadvantage in Europe. After a careful quantitative analysis of the regional systems of innovation of the EU-25 regions, a multiple regression analysis is conducted, including measures for R&D investment and proxies for regional innovation systems to reveal the importance of both endogenous innovative efforts and socio-economic conditions for the creation of economically productive innovation and regional growth. In addition the model includes, for each region, a proxy for the innovative efforts and socio-economic conditions of neighbouring areas, assessing the potential location advantage determined by an “innovation prone” interconnected neighbourhood. Our results show that while innovative output spills over into interconnected areas, the regional system of innovation benefits only the local economy. As a consequence, endogenous socio-economic conditions are crucial for the successful translation of innovation into economic growth, as the lack of an adequate system of innovation cannot be compensated for by external factors. The chapter is organised into four sections. In the first section the theoretical framework of the analysis is outlined, by combining the formal regional growth theories with the systems of innovation approach. The second part introduces the empirical analysis of the socio-economic conditions of the regions of the EU. In the third section our model (and its theoretical justification) is outlined and the empirical results are discussed. The final section concludes with some economic policy implications.
2 Innovative Efforts, Knowledge-Flows and Innovation Systems: A Set of Complex Links In this section we focus on the linkage between space and geography, on the one hand, and the production, diffusion, and translation of innovation into economic growth on the other. We highlight how these processes interact with the underlying
Systems of Innovation and Regional Growth in the EU
169
geography of local socio-economic conditions in order to show how the understanding of their spatial structure is crucial for an in-depth understanding of the growth dynamics of the increasingly interconnected system of the regions in Europe.
2.1 Innovative Efforts, Systems of Innovation and Regional Growth The “endogenous growth” (Grossman and Helpman 1991) perspective, by recognising endogenous innovation as a key engine for economic growth, provides a first explanation for persistent differential economic performance and also for localised knowledge spillovers. “It is now increasingly accepted, even among many neoclassical economists, that models that do not include (. . . ) innovation as contributed by intentional activities in private firms overlook one of the most important sources of technological progress in capitalist economy” (Fagerberg 1994, p.1170). However, the “catch-up” theory (Abramovitz 1985, Fagerbeg 1988, Verspagen 1991) with its “descriptive” nature “and strong emphasis on historical analysis” (Fagerberg 1994, p.1160) is the theory that definitely challenges the linear model of innovation. In this context the role of endogenous “contextual” factors in the process of generation and “absorption” of innovation is introduced into a formal analytical framework that provides interesting insights in the context of EU regional growth and employment dynamics by taking into account differences across regions in terms of innovative efforts (Fagerberg et al. 1997). In this perspective local R&D expenditure is a proxy for both local innovative efforts and local capability to exploit exogenously produced innovation (Cohen and Levinthal 1990; Maurseth and Verspagen 1999). However, where our aim is the understanding of place-specific influences on economically-useful innovation – in line with the “modern view” on the links between innovation and regional growth put forth by Camagni and Capello (2009, in this book) we need to take into account the insights from a variety of “holistic and interdisciplinary” contributions that have focused on various kinds of learning processes embedded in ordinary economic activities (Edquist 1997) and in their socio-institutional environment The “territorially-embedded” factors influencing the process of innovation have become the focus for differentiated theoretical perspectives (see also Bramanti and Fratesi 2009, in this book, for an analysis of these factors as components of the local Territorial System of Production and Innovation – TSPI): from innovative milieus (Camagni 1995) and industrial districts (Becattini 1987) to learning regions (Morgan 1997) and systems of innovation (Cooke et al. 1997; Cooke 1998). However, the explanatory power of such theoretical tools is part and parcel of the unit of analysis they have been designed to analyse. As a consequence, all these approaches (even though to different degrees) show a strong explanatory power where “interpreted as meta-models that show the relevance of proximity elements, local synergy and interaction” (Camagni 1995, p.317, for innovative milieus) as also of “inter-organization networks, financial and legal institutions, technical agencies
170
R. Crescenzi, A. Rodr´ıguez-Pose
and research infrastructures, education and training systems, governance structures, innovation policies” (Iammarino 2005, p.499 for systems of innovation). However, these concepts need to be carefully refocused where the unit of analysis is widened in order to reveal the “regularities” of the bond between innovation and regional growth by developing a model of territorial development. A contribution in this direction is provided by the “evolutionary integrated view of the regional systems of innovation” (Iammarino 2005) where the integration of two apparently opposing views, the macro/national and the micro/local perspectives, is put forward as a better tool for the identification of differences between regional systems of innovation. Where the top-down perspective of the national systems of innovation (focusing on the macro institutional and organisational economic tissue) is contrasted with the bottom-up view of micro-local interactions between the individual innovative agents, a meso-level emerges. This meso-level “shapes and constrains” new growth opportunities “by the inheritage of local structural regularities from past knowledge accumulation and learning” (p. 503). The identification and measurement of the “structural regularities” of regional systems of innovation is made difficult by the intrinsically unique and dynamic nature of the systemic interactions between (local) actors. As a consequence, for the purposes of the specification of our analysis, we further develop this meso-level concept by focusing on the socio-economic infrastructure of each regional space. This is first shaped by the macro-level (the national system of innovation) but, in turn, acts as a substratum for micro-level interactions of the individual agents: the focus of our analysis is thus on the local structural preconditions for such interactions. In other words, by cross-fertilising the systemic approach with the “learning regions” approach, we attempt to proxy the “external conditions in which externalised learning and innovation occur” (Cooke et al. 1997, p. 485), as these conditions are the true regional policy target. It is precisely in this sense that we make use of the concept of “social filter” as the unique combination “of innovative and conservative components, that is, elements that favour or deter the development of successful regional innovation systems” (Rodr´ıguez-Pose 1999, p. 82). Thus, the “social filter” represents the set of “conditions that render some courses of action easier than others” (Morgan 2004), making “innovation prone” interactions and institutions more likely in certain localities than in others.
2.2 The Role of Accessibility to Extra-Regional Innovation and Innovation-Prone Areas Within the framework outlined in several other chapters of this book and further discussed in the previous sections of this essay, innovation is not the outcome of a linear process. On the contrary it has been shown to be the result of a complex set of interactions between innovative units (R&D departments within firms, universities, research centres etc.) and their external environment through the “network” structure of the regional economy. As Bramanti and Riggi (2009, in this volume) point
Systems of Innovation and Regional Growth in the EU
171
out, a fully functional territorial system of innovation needs not only to be internally robust (in terms of social interactions and synergies), but also open to external (possibly worldwide) linkage. From this perspective, the spatial dimension of the economy becomes an important factor in explaining differential innovative performance at the regional level. On the one hand, the spatial organisation may influence the spatial extent and intensity of knowledge flows. On the other hand, the geography of the underlying socio-economic conditions shapes such flows and their effects on the local economy. The key concept on which the mechanics of both these processes relies is the differential communication technology of codifiable information and tacit knowledge. “Codifiable information is cheap to transfer because its underlying symbol systems can be widely disseminated through information infrastructure” (Leamer and Storper 2001, p. 650). However, information is not completely codifiable due to some specific features which, in some cases, make codification impossible or too expensive. “If the information is not codifiable, merely acquiring the symbol system or having the physical infrastructure is not enough for the successful transmission of a message” (Storper and Venables 2004, p. 354), thus making “face-to-face” contacts the most effective communication technology. Face-to-face contacts exhibit at least two features relevant for the processes under analysis: they are an intrinsically “spatial” communication technology and they are “socially” shaped. The space-sensitivity of such contacts is the mechanism through which geography and distance exert their influence over the process of innovation. The combination of physical geography, communication and transport infrastructure, and urbanisation patterns determines how easy (difficult) and dense (sparse) such contacts will be, thus emphasizing (hampering) their “potential” as communication technology. However, while face-to-face contacts as a communication technology make the transmission of innovation possible, they also pursue other functions that make communication more effective. Among these functions Storper and Venables (2004) cite the following: trust and incentives in relationships, screening and socialising, and motivation. These functions are clearly influenced by the socio-institutional environment in which they take place. As a consequence not only is the “production” of innovation itself shaped by the “social filter” in place in the local economy but this latter factor also influences the extent and the effectiveness of the diffusion of innovation and knowledge. For this reason the role of interconnection between territories and the corresponding knowledge flows cannot be fully understood unless they are associated with the underlying socio-economic conditions. It is reasonable to expect that face-to-face contacts are maximised within the region due to the effect of closer proximity and common socio-institutional infrastructure and networks. However, part of the “uncodifiable” knowledge produced in a region, could overcome the limits of the “institutionally defined” region thus “flowing” into neighbouring interconnected territories (Adams 2002; Adams and Jaffe 2002; Anselin et al. 1997; Jaffe 1986; Sacco and Segre 2009, in this book). Not only the local innovative efforts have an influence on the innovative output and on economic performance, but also the ability to access external sources of “uncodified” knowledge. Such “accessibility to extra-regional innovation” must be, in turn, compared with the endogenous social-filter conditions which make communication
172
R. Crescenzi, A. Rodr´ıguez-Pose
more effective and determine to what extent knowledge is translated into economic growth. In addition, when assessing the growth opportunities associated with interconnected territories, it is worth casting some light on the ability of the regional economy to benefit not only from exogenously produced (codified and un-codified) knowledge, but also from the favourable social filter conditions in neighbouring areas. In this sense it is necessary to understand whether the connection with more favoured regions, by means of positive externalities, could compensate for poor local socio-economic conditions. In other words, we aim at understanding whether the mechanics of the social filter are similar to that of local innovative efforts whose benefits, as we discussed above, tend to spill over into neighbouring areas or, on the other hand, local conditions are the only relevant variable. This issue has significant policy implications when designing innovation-based development strategies in lagging regions, as we will discuss in further detail when drawing some economic policy implications from our results. Where a territory is shown to be unable to benefit from the connection with extra-regional “innovation prone” areas, addressing endogenous conditions becomes an essential ingredient for a successful strategy.
3 The Socio-Economic Conditions of EU Regions and Their Innovative Performance 3.1 Scale of Analysis and Data Availability When the theoretical concepts outlined in the previous section have to be tested empirically, the selection of an appropriate scale of analysis, within the constraint of data availability, becomes crucial. In the context of our analysis, the focus is on the “institutionally-defined region”, the sub-national level which maximises the level of internal coherence in terms of socio-institutional features while being associated with a meaningful political decision-taking level. By coherently applying such criteria to the EU-25 regions, we rely on NUTS1 regions for Germany, Belgium, and the UK and NUTS2 for the remaining countries (Spain, France, Italy, the Netherlands, Greece, Austria, Portugal, Finland, Czech Republic, Hungary, Poland, Slovakia). Countries without a relevant regional articulation (Denmark, Ireland, Luxembourg, Estonia, Latvia, Lithuania, Slovenia, Malta, and Cyprus) were necessarily excluded from the analysis.2 In addition, at 2
As far as specific regions are concerned, no data are available for the French D´epartments d’Outre-Mer (Fr9). Uusimaa (Fi16) and Etela-Suomi (Fi17) were excluded from the analysis due to the lack of data on socio-economic variables. Etela-Suomi (Fi17) and Trentino-Alto Adige (IT31) were excluded from the analysis as they have no correspondent in the NUTS2003 classification, thus preventing us from matching data available only in the new NUTS classification. Islands (PT2 Ac¸ores, PT3 Madeira, FR9 Departments d’Outre-Mer, ES7 Canarias) and Ceuta y Melilla (ES 63) were excluded from the analysis as time-distance information, necessary for the computation of spatially lagged variables, is not available.
Systems of Innovation and Regional Growth in the EU
173
the moment regional data on R&D expenditure for Sweden are not available in the Eurostat databank. In our analysis, EUROSTAT data (stored in the REGIO databank on which we largely relied for our empirical analysis) have been complemented with Cambridge Econometrics (CAMECON) data for GDP. Table 1 in the appendix provides a detailed definition of the variables included in the analysis.
3.2 An Empirical Definition of the Concept of “Social Filter”: The Social Filter Index When the case-study approach of the systems of innovation literature is to be integrated in a cross-section analysis aimed at singling out some regularities in the socio-economic structures of “innovation-prone” regions in the EU, the concept of social filter needs to be “proxied” by means of a set of variables available for all the regions under analysis. In particular, the variables which seem to be more relevant for shaping the social filter of a regional space are those related to three main domains: educational achievements (Lundvall 1992; Malecki 1997; Bramanti and Riggi 2009, in this book), productive employment of human resources (Riggi and Maggioni 2009, in this book) and demographic structure (Fagerberg et al. 1997; Rodr´ıguez-Pose 1999). From the first domain, tertiary educational attainment (of both the population and the labour force) and participation in lifelong learning programmes can be assumed as a measure for the accumulation of skills at the local level. From the second area, the percentage of labour force employed in agriculture and the long-term component of unemployment are included in the analysis. On the one hand, long term unemployment represents the incidence of people whose possibilities of being productively involved in the labour market are persistently hampered by inadequate skills (Gordon 2001). On the other hand, in particular in the new members of the European Union, agricultural employment is frequently synonymous with “hidden unemployment”. From the third area, the percentage of population aged between 15 and 24 is considered as a proxy for the flow of new resources entering the labour force, potentially “renewing” the existing stock of knowledge and skills. Each of these variables is assumed to represent one “face” of the intrinsically multifaceted concept of social filter. While keeping in mind this composite nature of the social-filter concept and the autonomous role played by each domain, Principal Component Analysis (PCA) is applied to the set of variables discussed above, to achieve two distinct objectives. First, Principal Component Analysis allows us to represent a large part of the global information provided by the original set of variables in two-dimensional space. Such a representation yields a first “classification” of the EU regions in terms of homogeneous categories based on their social-filter conditions. Second, PCA enables the development of a “joint” measure for the social-filter conditions of each regional space by “reducing” them to an
174
R. Crescenzi, A. Rodr´ıguez-Pose
individual variable able to preserve as much of the initial information (variability) as possible. Such a measure not only makes it possible to represent the social-filter conditions of the EU regions by means of a single variable, but also enables us to deal with the problem of multicollinearity which would prevent the simultaneous inclusion of the individual (highly correlated) social-filter variables in a regression model. Tables 1a, b show the output of the Principal Component Analysis. Table 1a Principal component analysis: Eigenanalysis of the correlation matrix
Eigenvalue Proportion Cumulative
PC1
PC2
PC3
PC4
PC5
PC6
2.5886 0.431 0.431
1.2723 0.212 0.643
0.9083 0.151 0.795
0.6418 0.107 0.902
0.5661 0.094 0.996
0.0229 0.004 1
Table 1b Principal component analysis: Coefficients of principal components Variable Education population Education labour force Life-long learning Agricultural labour force Long term unemployment Young people
PC1
PC2
0.576 0.554 0.395 −0.43 −0.14 0.019
−0.224 −0.313 0.26 −0.285 −0.459 0.701
Eigenanalysis of the Correlation Matrix (Table 1a) shows that the first principal component (PC) alone accounts for around 43% of the total variance with an Eigenvalue significantly larger than 1, the second PC accounts for an additional 21% of the total variability with an Eigenvalue still larger than 1. The first two principal components therefore explain a significant part of total variability (64%). The coefficients of the first PC emphasize the educational dimension of the social filter by assigning a large weight to the educational achievements of the population (0.576) and of the labour force (0.554) and to the participation in lifelong learning programmes (0.395). A negative weight is, as expected, assigned to the agricultural labour force (−0.430) and, with a smaller coefficient, long-term unemployment (−0.140). The weight of the young population (0.019). is much smaller, in the first principal component. The second PC, instead, emphasizes precisely this demographic dimension by assigning a relatively small and negative value to educational (except lifelong learning) and “resources employment” variables in contrast with the large and positive coefficient of the “young population” variable. The first principal component will provide us with the “joint measure” for each region’s social filter and will be referred to as the “social-filter index”. Consequently, the first principal component scores are computed from the standardised3 value of 3
Standardised in order to range from zero to 1.
Systems of Innovation and Regional Growth in the EU
175
the original variables by using the coefficients listed under PC1 in Table 1b. The second principal component will be scattered against the first in order to represent the social filter variables in a two-dimensional space. As discussed above, this second PC contrasts the incidence of young people in the total population with the other features of the regional social filter, and for explanatory purposes can be referred to as the “demographic dynamism index”.
3.3 The Geography of the Social Filter Conditions in the EU
Demographic Dynamism Index (PC2)
A first representation of the socio-economic “infrastructure” of the EU regions is provided by the PCA, which yielded the scatter plot shown in Fig. 1. For each region this graph plots the score of the first principal component on the x-axis and that of the second PC on the y-axis. As discussed above, the first two principal components account for a high percentage of the global “information” of the original variables. Consequently, the scatter in Fig. 1 accurately represents the relative position of the EU regions in terms of their socio-economic conditions. The more favourable the socio-economic conditions (societies more “prone” to innovation) the higher the scores of the first PC (Social-Filter Index). This means that the regions are ordered on the x-axis by the value of their social filter index so that “innovation prone” regions appear furthest to the right of the graph. The value of the second principal component (Demographic Dynamism Index) is higher for the regions where the percentage of the young population is higher (the only variable
Q1(0.053)
0.50
Q3(0.6) AT33 AT34
NL21
AT11
AT32 NL34 PT11 NL12 IT32AT31 AT21 NL31 NL23 NL11 ES42 HU02 ES61 ES62 CZ02 IT33 AT22 NL41 GR42CZ05 ES52 NL13 FI15 HU03 CZ04 PT14 HU07HU06 CZ03 NL33 CZ06 NL22 ES43 ES53 IT4 CZ07 AT12 IT2 NL32 HU04 CZ08 UKE ITB NL42 PL04 SK02IT91 IT11 HU05 UKD UKH ES51 UKC UKG UKF UKL UKKUKM SK04 SK03 PL0C PT13 FI13 PL08 IT53 PT12IT92 UKJ IT12 DE2 IT51 PL0F PL02 UKN FI14 ITA CZ01 FR3 PL01 IT8 FR42 FI2 PT15 FR22 FR43 FR21 FR41 HU01SK01 DEB IT52 PL0B PL0G PL06 AT13 IT93FR83 FR23 DEF DE1 PL0E FR51 DE9 ES24 IT71 GR22 IT60 DE7 FR63 ES22 PL03 FR25 FR26 DEAGR3 DE6 IT72 IT13 PL07 PL05 FR24 GR12 ES23 ES41ES3 GR41PL09 DE8 GR11 FR81 FR82 DE5 ES13 FR52 FR53DEC FR72 FR61 DEE DEGDED PL0A GR24PL0D FR62 ES12FR71 ES21 GR43 GR13 ES11 BE2 DE3 DE4 GR23 BE3 GR25 GR14 GR21 FR1
0.25 0.00 -0.25 -0.50
A
-0.75
BE1
-1.00 -0.5
0.0
Ave(0.336) 0.5 1.0 Social Filter Index (PC1)
Fig. 1 Social Filter: first vs second PC
C UKI
Ave(-0.21)
B
1.5
176
R. Crescenzi, A. Rodr´ıguez-Pose
associated with a large and positive coefficient). Thus the y-axis contrasts with the x-axis by differentiating the regions essentially on the basis of their demographic dynamics: e.g. two regions with a similar value for the social filter index (plotted in a very close position on the x-axis) might have differential demographic dynamics and thus would be plotted far away from one another on the y-axis. Figure 1 suggests that the variability in terms of the Demographic Dynamism Index tends to be lower for intermediate values of the social filter index (thus highlighting a more homogenous demographic structure for these regions) while increasing with low and high values of the Social-Filter Index. Low values (below the first quartile – Q1) and high values (above the third quartile – Q3) of the Social-Filter Index (PC1) the Demographic Dynamism Index (PC2) enable three interesting clusters of regions, circled in the graph, to be identified. Conversely, when intermediate values of the Social-Filter Index operate, no particular pattern is apparent. Some of the lowest social filter index regions, which are grouped together on the left-hand side of the graph, show particularly weak demographic dynamics thus suggesting that an ageing population is a prominent characteristic of their generally unfavourable social filter conditions (group A in the graph). At the opposite extreme of the distribution of the Social-Filter Index, the Demographic Dynamism Index (PC2) allows us to distinguish two further groups: a) the regions where generally favourable social filter conditions are associated with a relatively older population (group B) and b) those, on the contrary, where the demographic dynamics are relatively more favourable (group C). While group C regions benefit from a globally favourable social filter and also benefit from favourable demographic dynamics that keep the young percentage of the population relatively high, group B regions compensate for less favourable demographic trends with a better performance in the other social-filter indicators. Conversely, in group A regions the process of ageing of the population is not compensated for by other more favourable aspects, showing the most critical situation for their chance of success in the knowledge-based economy. As the European Commission (2006) has highlighted, the most effective tool to compensate for the negative impact of ageing population on productivity and employment would be investment in human capital which, however, in this last group of regions is low as well. In general, when interpreting the meaning of the Demographic Dynamism Index, it must be borne in mind that a large part of the demographic dynamism of the European Union is generated by migration flows which far exceed the natural increase in the European population (European Commission (2006). After 1990, net migration has become the major component of population growth in the EU-25: “since 2000 more than three-quarters of the total population growth in the EU-25 is due to net migration. However, while the former EU-15 countries fully account for the population growth by international migration, net migration is (still) negligible in the new Member states” (Eurostat 2006, p. 46). As a consequence, the Demographic Dynamism Index (PC 2) mainly highlights the differential migration trends: some regions have been able to benefit from inward migration flows that have accentuated the incidence of young people in the total population (those in the higher part of the graph), while in other regions low birth rates and the natural ageing of the
Accessibility to Innovation Prone Extra-Regional Areas
Systems of Innovation and Regional Growth in the EU
177
0.6 UKH NL23 NL41UKG NL33NL32 NL31 BE2 UKF NL22 UKJ UKE UKD NL34NL13 UKLUKK NL12NL21 FR3 UKC NL42 BE3 BE1 DEA NL11 FR22 DE5 UKM DE9 FR21 DEF DE6 DEC ES23 DEB UKN FR23 DE7 DEE FR25 FR24 FR41 FR1 FR51 DE1 DE8DEGDE4ES22 DE3 FR53 FR42 FR26 FR52 FR63 ES24 ES21 FR61 ES13 FR43 FR72 DED FR62 ES42ES41FR81ES12 ES51 FR71 AT34 ES62 ES52 CZ02 CZ04 FI2 PL0G ES3 ES53 DE2 FR82 FI13 FI14 FI15 ES11ES43 AT33 IT12 CZ03 ES61 PT11 PT12PT15 AT31 AT32 PL04 PT14 CZ05 ITBFR83 IT11 AT12 IT13 AT21 AT22 IT2 AT11 CZ06 PL0EPL0F PT13 CZ01 sk02 IT33 HU03 PL0B ITA HU04 IT32 IT51 HU07 HU02 AT13 CZ07 IT4 PL08 CZ08 PL0APL02 IT52PL01 IT53 HU06 IT93 GR41 HU05 sk03 sk01 IT71 GR43 GR42 IT91 IT6 HU01 IT92 IT72 sk04 PL05 IT8 GR11 PL03 PL0D PL07 PL0C GR22 PL09 PL06
0.5
0.4
0.3
GR25 GR23 GR24
UKI
GR13GR12 GR14 GR21
0.2
GR3
-0.5
0.0
0.5 Social Filter Index
1.0
1.5
Fig. 2 Endogenous vs External Socio-Economic conditions
population prevail (those in the lower part). Such differential trends in the capacity to attract external human resources may have a significant impact on the long-term innovative capabilities of a local system of innovation as extensively discussed in Riggi and Maggioni (2009, in this book). This “classification” of the EU regions according to their social filter conditions reflects the underlying socio-economic geography of the union which is more directly analysed in Fig. 2. Figure 2 scatters the social filter index of each region against the corresponding accessibility to “innovation prone” extra-regional areas i.e. the distance-weighted average of the social filter index4 of the neighbouring regions. In other words, the graph compares the local social filter conditions with those of neighbouring regions in order to highlight the presence of a spatial pattern in the distribution of regional competitive disadvantage. This spatial pattern, which is clearly highlighted in Fig. 2, shows that less well-endowed regions (left-hand part of the graph) tend to have disadvantaged neighbours compared to innovation-prone areas (right-hand side) whose neighbours tend to be equally favourable to the development and absorption of innovation. In other words, the regions with more favourable socio-economic conditions also benefit from better “accessibility” to “innovation prone” extra-regional areas while the more disadvantaged ones, due to the spatial concentration of the factor of (dis)advantage, tend to have less direct access to more advantaged territories. The empirical model presented in Sect. 4 will assess whether this particular geography
4
The calculation of this last variable will be discussed in further detail when justifying its inclusion in the regression model.
178
R. Crescenzi, A. Rodr´ıguez-Pose
of these socio-economic factors has an influence on local economic performance or not.
3.4 The Social Filter Index and Its Association with R&D Expenditure: Four Typologies of Regions and Their Differential Growth Performance Socio-economic conditions, as discussed when outlining the theoretical framework of the analysis pursued in this chapter, exert a special influence on regional economic performance when interacting with local innovative activities. Fig. 3 attempts a visual representation of this relationship by analysing all the possible combinations between local innovative efforts (as proxied by R&D expenditure) and social filter conditions (proxied by the Social-Filter Index).
Fig. 3 Local innovative efforts and social filter conditions
Systems of Innovation and Regional Growth in the EU
179
The theoretical analysis suggests that the “optimal conditions” for economic success are met where a considerable amount of innovative effort is pursued in an “innovation prone” socio-economic environment (top right quadrant of the graph). Conversely, economic performance seems to be significantly hampered by the coexistence of “inadequate social-filter conditions and inadequate innovative efforts”. In this latter case, endogenous efforts are reduced and their effects on the local economy are negatively affected by the inadequate environment. In addition to these two opposite cases, a variety of intermediate situations are possible. In particular, a region with a favourable “social filter” might pursue “sub-optimal research efforts”, or relevant investment in R&D might be undertaken in an unfavourable socio-economic environment i.e. where a “dissociation between R&D and local conditions” is recorded. In the former case the theory suggests the possibility of benefiting from externally produced innovation which might partially compensate for the inadequacy of local efforts. In the latter case, instead, R&D investment runs the risk of producing the “cathedral in the desert” effect. Consistent with this analytical framework, Fig. 4 plots the observed R&D expenditure in the EU regions against the corresponding value of the social filter index. As a result, we notice that the large majority of the EU lagging regions are clustered in the lower right-hand corner of the graph: a combination of underinvestment in R&D and adverse social-filter conditions might be contributing to their economic disadvantage. Conversely, the innovation leaders are clustered in the top right-hand area: in these regions the innovative efforts find the most appropriate socio-economic environment. This group includes: the German regions of Berlin, Baden-W¨urttemberg, Bayern, and, to a lesser extent, Bremen, the entire South East England, the ˆIle de France and Midi-Pyr´en´ees in France. Between these two “poles” there are a variety of intermediate possible combinations of local efforts and social 0.02
1.5 UKI
Social Filter Index
BE1
1.0
0.5
0.0
NL31 DE3 UKJ UKM FR1 NL32 ES21 UKK NL33 DED DE4 ES3 CZ01 DEG UKD UKE NL11 UKL DE8 BE2 UKGUKF NL22 DEE BE3 FI15 NL41 UKC UKN SK01 DE7 NL42 DE6 NL21 FR71 FI2 NL23 HU01FI14 DE2 FR62 FI13 ES24 FR82 DEF GR3 ES51 NL12 ES12 ES41 NL13 ES13 FR42 ES52 AT13 DE5 DEADEB FR61FR81 DE9 FR43 NL34 ES61 FR41 FR52 ES23ES62AT32 DEC FR72 FR24 FR3 FR51 ES53 FR23 FR26 FR53 FR25 PT13FR22 ES42 AT34 FR63 AT33 FR21 IT60 ES43 CZ06 HU02 AT21 PL07 HU05 AT12 AT31 PL0B GR12 PL01 IT4 IT2 PL0C HU03 ES11 PL06 IT33 AT22 IT52 IT53 AT11 HU06 IT32 CZ03 CZ08 IT13 IT51 CZ04 CZ05 HU04 IT12 CZ07 CZ02 IT11 FR83 PL0G IT8 PL04 PL0F ITA IT71 GR42 HU07 IT91 ITB IT72 GR13 IT93 SK03 PL08 PL0E GR21 PL09 PL05 PL02 SK04 SK02 PT11 IT92 PT15 PL0A PL03 PT14 PT12 GR41 GR22 GR14 GR43 PL0D GR11 GR24 GR23 GR25 ES22
UKH DE1
0.5
-0.5 0.00
0.01 0.02 0.03 R&D expenditure (% Regional GDP)
Fig. 4 R&D expenditure and “Social Filter” conditions
0.04
180
R. Crescenzi, A. Rodr´ıguez-Pose
conditions associated with different economic outcomes. This kind of evidence suggests that the large majority of the EU regions are simultaneously affected (even if with place-specific combinations of each factor) by inadequate social-filter conditions and insufficient innovative efforts. Such heterogeneity of situations and their association with regional growth performance deserves more in-depth investigation by means of regression analysis.
4 An Empirical Model for EU Regional Growth Performance 4.1 Model Specification In the previous sections we discussed the mechanisms by which distance and space influence the innovation process. While technological development has increased accessibility to codified information we pointed out that an important part of the process of innovation relies on three other crucial factors: “endogenous” innovative efforts, socially-embedded factors and spatially-bound knowledge spillovers. In this section such factors are introduced in a formal model to study how local innovative efforts are translated into regional growth and assess the role of systems of innovation and knowledge spillovers in this process. As succinctly presented in Table 2, the model combines the mechanisms discussed in the theoretical section in a common analytical framework in order to assess their empirical relevance and their reciprocal interactions. In line with this, the standard specification of an endogenous growth model for the EU regions (where innovation is considered as the key determinant for regional growth performance) is extended in order to take into account both the “endogenous” characteristic of the regional system of innovation and the “spillover effects” due not only to innovative activities of neighbouring regions but also to their structural socio-economic features.
Table 2 Structure of the empirical model
R&D Regional systems of innovation GDP per capita National effect
Endogenous factors
External conditions and spillovers
Investment in R&D in the region Regional system of innovation As a proxy for initial conditions and potential Controlled for by a set of National Dummies
Investment in R&D in neighbouring regions Regional system of innovation in neighbouring regions Initial conditions in neighbouring regions
Systems of Innovation and Regional Growth in the EU
181
The full estimated model is: 1 Yi,t ln = α + β1 ln(yi,t−J ) + β2RDi,t− j + β3SocFilteri,t−J J Yi,t−J + β4 Spillovi,t− j + β5ExtSocFilteri,t−J + β6ExtGDPcapi,t−J + β7 D + ε where: Yi,t 1 ln J Yi,t−J
α ln(yi,t−J ) RDt− j SocFilteri,t−J Spillovi,t− j ExtSocFilteri,t−J ExtGDPcapi,t−J D ε
Is the usual logarithmic transformation of the ratio of regional per capita GDP in region i at the two extremes of the period of analysis (t-J, t) Is a constant Is the log of the GDP per capita of region i at the beginning of the period of analysis (t-J) Is expenditure in R&D as a % of GDP in region i at time (t-J) Is a proxy for the socio-economic conditions of region i representing its ‘social filter” Is a measure of accessibility to extra-regional sources of innovation Is a measure of the “social filter” of neighbouring regions Is a measure of the GDP per capita in neighbouring regions Is a set of national dummy variables Is the error term
Initial level of GDP per capita – As customary in the literature on the relationship between innovation and growth, the initial level of the GDP per capita is introduced in the model in order to account for the region’s stock of existing knowledge. R&D expenditure – The percentage of regional GDP devoted to R&D is a measure of the economic input in order to generate innovation in each region. Spillovers – This variable proxies the capability of exogenously produced innovation to influence regional economic performance in addition to “endogenous” innovation. For this purpose a measure for the accessibility of “extra-regional” innovative activities has been developed and introduced in the analysis by means of a standardised “accessibility of innovation index”. The index is a potential measure of the “innovative activities” (in terms of nationally weighted millions of Euros invested in R&D activities) that can be “reached” from each region at a “cost” which increases with distance. Our index is based on the usual formula for accessibility indices: Ai = ∑ g(r j ) f (ci j ) j
where Ai is the accessibility of region i, rj is the activity R to be reached in region j, cij is the generalised cost of reaching region j from region I and g(·) and f(·) are “activity” function (i.e. the activities/resources to be reached) and “impedance”
182
R. Crescenzi, A. Rodr´ıguez-Pose
function (i.e. the effort, cost/opportunity to reach the specific activity) respectively. In our index the “activity” to be reached is R&D expenditure and the “impedance” is the bilateral trip-time distance between region i and region j: f (ci j ) = wi j =
1 di j ∑ d1i j j
where dij is the average trip-length (in minutes) between region i and j. We base our analysis on the travel time calculated by the IRPUD (2000) for the computation of the Peripherality Indicators, made available by the European Commission.5 We chose road distance, rather than straight line distance, as (in particular on a smaller scale) it gives a more realistic representation of the real “cost” of the contacts across distance. In addition, the use of trip-length rather than kilometres allows us to take account of “different road types, national speed limits, speed constraints in urban and mountainous areas, sea journeys, border delays (. . . ) as well as congestion in urban areas” (IRPUD 2000, p. 22) which significantly affect real-world interactions. Thus, the amount of knowledge flowing from outside the region is proxied by the average R&D expenditure of all the other regions weighted by the inverse of the bilateral time-distance. The resulting variable is then standardised by making it range from zero to one, in order to be comparable with the social filter index. Extra regional social filter – Following a similar procedure we calculated, for each region, the distance-weighted average of the social-filter index of all other regions thus attempting to assess the existence of spillovers from “innovation prone” regions towards their neighbourhood as far as their social filter is concerned. GDP in neighbouring regions – Again the same weighting procedure is pursued in order to introduce the initial economic conditions (GDP per capita) of neighbouring regions. This variable accounts for the advantage of proximity to relatively “advanced” regions which could potentially increase the scope for imitation.
4.2 Estimation Issues and Data Availability In this section we estimate the model outlined above by means of HeteroskedasticityConsistent OLS (Ordinary Least Square). In order to minimize the effect of spatial autocorrelation (i.e. the lack of independency among the error terms of neighbouring observations), we include in the analysis a set of national dummy variables accounting for the “national fixed effect” which, in turn, accounts for a consistent part of the similarities between neighbouring regions. Furthermore, by including 5
As the time distance-matrix is calculated either at the NUTS1 or at the NUTS2 level, in order to make it coherent with our data which combine different NUTS levels we relied on the NUTS distance matrix using the NUTS 2 regions with the highest population density in order to represent the corresponding NUTS1 level for Belgium, Germany and the UK.
Systems of Innovation and Regional Growth in the EU
183
spatially lagged variables in our analysis, we attempt to model explicitly the interactions between neighbouring regions, thus minimizing their effect on the residuals. Endogeneity is another major problem, which we dealt with by including6 in the model the value of the explanatory variables as the mean over the period (t-J-5) – (t-J), while the average growth rate was calculated over the period from t-J to t. In addition, in order to resolve the problem of different units of account, explanatory variables are expressed, for each region, as a percentage of the respective GDP or population. The empirical model is estimated by considering a 1995–2003 time span, in order to include all the EU-25 members for which regional data are available.
4.3 Innovation, Spillovers and Social Filter: Empirical Results The estimation results for the empirical model outlined in the previous section are presented in Table 3. In regressions 1–3 the variables for “social filter” and “accessibility to external sources of innovation” are progressively introduced. In regressions 4–9 the various components of the social filter are autonomously introduced in the model. In equations 10–12 the effect of the endowment of neighbouring regions in terms of social filter and economic wealth is assessed. The r-squared confirms the overall goodness-of-fit of all the regressions presented and in all cases the probability of the F-statistics lets us reject the null hypothesis that all of the regression coefficients are zero. The V.I.F test has been pursued for the variables included in all the specifications of the model excluding the presence of multicollinearity. Additionally, we test for the presence of spatial autocorrelation in the residuals and the Moran’s I statistic allows us to exclude this possibility for all the regressions presented in the table. From a general inspection of Table 3 we notice that the initial level of the GDP per capita is significant in a few cases only, thus suggesting that for the period under analysis, neither regional convergence nor divergence can be recorded. Only when social conditions are explicitly controlled for (regressions 3, 10, 11 and 12) is there evidence of a weak degree of regional convergence. Local R&D expenditure seems to exert a positive influence on economic growth, findings in line with similar research (Fagerberg et al. 1997; Rodr´ıguez-Pose 1999, 2001; Cheshire and Magrini 2000; Bilbao-Osorio and Rodr´ıguez-Pose 2004; Crescenzi 2005). In addition, for the European regions considered, investing in R&D seems to be a more important source of economic growth than relying on knowledge spillovers from neighbouring regions. When considering both factors together
6
In the case of the New Member States data availability has prevented us from calculating the mean of the explanatory variables over the 5 year period (t-T-5) forcing us to use a shorter time span. For some EU 15 countries slightly differential time spans have been used according to data availability for each variable. The specification of each individual case would take up too much space.
Education labour force
2
3
0.1126∗∗∗ (0.02563)
4
6
7
8
9
10
0.10707∗∗∗ 0.09655∗∗∗ 0.08491∗∗∗ 0.08989∗∗∗ 0.10777∗∗∗ 0.12054∗∗∗ (0.02561) (0.02671) (0.03019) (0.0292) (0.02709) (0.02802)
5
0.12187∗∗∗ (0.02805)
11
0.12059∗∗∗ (0.02809)
12
x
0.013236 (0.008148)
0.2682∗∗ (0.1174)
0.1791 (0.1218)
x
0.1366 (0.1212)
x
0.017003∗∗∗ (0.005341)
x
0.01387∗ 0.013157∗ (0.008031) (0.007908)
0.01052∗∗ 0.010787∗∗ (0.004626) (0.004598)
0.1424 (0.1207)
0.2556∗∗ (0.1229)
0.2664∗∗ (0.1177)
0.2653∗∗ (0.1182)
0.2548∗∗ (0.1172)
0.1883 (0.1213)
0.019224∗∗∗ (0.006986)
x
x
x
x
x
x
0.013733∗ 0.012717∗ 0.012262 0.013353 0.013807∗ 0.014184∗ (0.007975) (0.0083) (0.008336) (0.008182) (0.008119) (0.008052)
0.166 (0.1208)
x
0.013936∗ (0.008059)
0.010538∗∗ (0.004682)
0.177 (0.1223)
x
0.014229∗ (0.008067)
0.011422∗∗ (0.004713)
0.1909 (0.1234)
−0.003098 −0.005756 −0.00663∗ −0.00574∗ −0.005112 −0.003359 −0.00196 −0.002733 −0.004345 −0.006577∗ −0.006349∗ −0.007705∗ (0.003255) (0.00353) (0.003543) (0.003267) (0.003268) (0.003346) (0.003803) (0.003478) (0.003339) (0.003571) (0.003668) (0.003929)
0.09406∗∗∗ 0.12284∗∗∗ 0.12182∗∗∗ (0.02572) (0.02814) (0.02796)
1
Social filter individual components: Education population
National dummies
Accessibility to extra regional innovation
Social filter index
R&D expenditure
Log GDP 95
Constant
Dependent variable: regional growth rate (1995–2003)
Table 3 H-C OLS estimation of the empirical model. R&D, social filter and knowledge spillovers
184 R. Crescenzi, A. Rodr´ıguez-Pose
0.00385 (0.01076) 0.003802 (0.006528) 0.001892 (0.006205) −0.009089 (0.005882)
∗ , ∗∗
and ∗∗∗ denote significance at a 10%, 5% and 1% level respectively. SE in parentheses – n. of observations = 166 in all regressions
Extra-regional social filter Total accessibility 0.012617∗∗∗ to innovation (0.005656) prone space Accessibility −0.00808 to innovation prone (0.0261) extra-regional areas Accessibility to wealth 8.8E-07 neighbouring regions (0.00000138) R-Sq 0.659 0.665 0.672 0.681 0.676 0.66 0.66 0.659 0.665 0.67 0.672 0.672 R-Sq (adj) 0.62 0.626 0.631 0.642 0.636 0.618 0.618 0.618 0.624 0.63 0.629 0.63 F 16.84 17.27 16.7 17.45 17.03 15.82 15.85 15.81 16.19 16.61 15.72 15.77 Moran’s I −0.0193012 −0.0185667 −0.0189041 −0.0194612 −0.0198153 −0.0193265−0.0198503 −0.0195195 −0.0199182 −0.0188243 −0.0188376 −0.0189403
Young people
Long term unemployment
Agricultural labour force
Life-long learning
Systems of Innovation and Regional Growth in the EU 185
186
R. Crescenzi, A. Rodr´ıguez-Pose
(Regression 1) the coefficient of local R&D expenditure is positive and significant, while the impact of innovation generated in interconnected regions is insignificant. Relying exclusively on local R&D inputs is, however, not a guarantee of achieving greater growth, as such relationship does not always prove robust when controlling for social conditions (the ‘social filter’ variable). As highlighted in Regression 2, the local socio-economic conditions are a better predictor of economic growth than investment in R&D. The social filter variable is always positively associated with economic growth and statistically significant. The relevance of the ‘social filter’ is enhanced when R&D investment and exposure to knowledge spillovers are considered in conjunction with local conditions (Regression 3). The results indicate that having a good social filter increases the potential of European regions to assimilate spillovers, making local R&D expenditure irrelevant. Knowledge spillovers, by increasing the “amount of knowledge” available in the region, reinforce the effect of local innovative activities, and, up to a certain extent, they may even compensate for a weak contribution of locally pursued innovative activities. Thus, other things being equal, a region in an innovative neighborhood is more advantaged than one located close to less innovative areas. However, local socio-economic conditions may prove the true differential competitive factor by enabling the translation of all source of knowledge into successful innovation and economic growth. The analysis of the individual sub-components of the social filter uncovers the specific importance of the educational endowment of both the population and the labour force (regressions 4 and 5). On the other hand, other social filter variables are not statistically significant when autonomously introduced in the model (regressions 6–9). While the interconnection with innovative activities pursued in other territories has a positive impact on the local economy by means of knowledge spillovers, both the socio-economic endowment (regr.11) and the level of economic wealth (regr.12) of neighboring regions seem to have no significant effect on local economic performance. Extra-regional social filter (i.e. the inverse-distance weighted average of neighboring regions’ Social Filter Indexes) is significant only when considered jointly with “endogenous” features, as in regression 10 where the total accessibility to “innovation prone” space is considered by including in a single variable both the region’s features and that of its neighborhood. On the basis of these results, the potential of a region in terms of economic performance is maximized when an appropriate social filter is in place. The reception of R&D spillovers from neighbouring regions is an important additional source of advantage which, in any case, requires an appropriate social infrastructure in order to be productively translated into new innovation and economic growth.
5 Conclusions and Policy Implications This chapter has combined the theoretical insights of various strands of literature by focusing on different, though complementary, aspects of the process of innovation. This synthesis has enabled us, on the one hand, to explain the role of socio-economic
Systems of Innovation and Regional Growth in the EU
187
contextual factors in the process of innovation and, on the other hand, to overcome the case-study approach of many empirical analyses. In this perspective it has been possible to analyse the regularities of the innovative dynamics of the EU regions and their “social-filter” preconditions, and draw a quantitative map of the EUwide sources of competitive disadvantage. Furthermore the empirical analysis has confirmed, in a rigorous regression framework, the importance not only of “endogenous” innovative efforts but also of local socio-economic conditions for the genesis and assimilation of innovation and its transformation into economic growth across European regions. Our results show that, while innovative output spills over into interconnected areas, the regional system of innovation only benefits the local economy. As a consequence, endogenous socio-economic conditions are crucial for the successful translation of innovation into economic growth, as the lack of an adequate system of innovation cannot be compensated for by external factors. In other words, our results, departing slightly from a neo-Schumpeterian approach, support the idea that the capacity of the local population to assimilate whatever research is being generated locally or in neighbouring regions and transform it into innovation and economic activity matters more than the threshold of expenditure. This piece of evidence has important regional policy implications. When innovation is recognized as the key source of sustained economic growth, the mechanics of its contribution to economic performance becomes crucial for effective policy targeting. Policies based on innovation may deliver, at the regional level, very differentiated results, according to the possibility of simultaneously benefiting from endogenously and externally produced knowledge and favorable underlying socioeconomic conditions. Referring to Fig. 5, we can imagine two alternatives strategies for the development of lagging, less innovative areas. The first strategy, ideally represented by the white arrow, has the objective of increasing the innovative competitiveness mainly by increasing R&D activities and expecting contextual conditions to adjust autonomously. However, the empirical evidence discussed in this paper suggests that detachment of innovative efforts from social filter conditions may severely hamper the growth-enhancing impact of innovation-based policies. Furthermore, in Chap. 11, Riggi and Maggioni showed that regional disparities in terms of skilled-labour endowment – an important component of the social filter – are likely to be persistent due to the absence of any automatic adjustment mechanism in the labour market. As a consequence, a more effective innovation-based development strategy should follow the pattern indicated by the black arrow in the graph: address social-filter conditions and stimulate innovative activities jointly. In less innovative regions, complementary policies are necessary in order to simultaneously address the underlying social filter conditions. Incentives for local innovative activities have to be complemented by reinforcing the local endowment in terms of education and skills in order to guarantee the greatest returns from policies based on R&D. Furthermore, as discussed in Chap. 7 by Camagni and Capello, the performance of a local knowledge society also crucially relies on the interaction, integration and synergy among the socio-institutional “functional preconditions”. As a consequence, balanced development strategies will also need
188
R. Crescenzi, A. Rodr´ıguez-Pose
Fig. 5 Alternative innovation-based economic development strategies
to address the dynamic nature of the regional system of innovation – in terms of its institutional and relational dimension. Our empirical analysis of that system by means of the “social filter” concept has been confined only to the socio-economic pre-conditions. In this book, various chapters have highlighted different mechanisms underlying the generation and diffusion of innovation and its impact on regional growth. Here we have highlighted the importance of favourable endogenous socio-economic factors as a fundamental pre-condition for the success of any innovation-based growth strategy. Consequently, policies targeted towards the improvement of regional social filter conditions should, in their turn, be part and parcel of a balanced development strategy addressing simultaneously the whole set of these complementary factors.
Systems of Innovation and Regional Growth in the EU
189
Appendix Table A. 1 Description of the variables Variable
Definition
Innovation R&D
Expenditure on R&D (all sectors) as a % of GDP
Social filter Life-long learning Education labour Force Education population Agricultural labour force Long term unemployment Young people
Rate of involvement in Life-long learning – % of adults (25–64 years) involved in education and training % of employed persons with tertiary education (levels 5–6 ISCED 1997). % of total population with tertiary education (levels 5–6 ISCED 1997). Agricultural employment as % of total employment Long term unemployed as % of total unemployment. People aged 15–24 as % of total population
References Abramovitz M (1985) Catching up and falling behind, Economic research report n.1. Trade Union Institute for Economic Research, Stockholm Adams JD (2002). Comparative localization of academic and industrial spillovers. J Econ Geogr 2:253–278 Adams JD, Jaffe AB (2002) Bounding the effects of R&D: an investigation using matched firm and establishment data. Rand J Econ 27:700–721 Anselin L, Varga A, Arcs Z (1997) Local geographic spillovers between university research and high technology innovations. J Urban Econ 42:422–448 Becattini G (1987) Mercato e forze locali. il distretto industriale. Bologna, il Mulino Bilbao-Osorio B, Rodr´ıguez-Pose A (2004) From R&D to innovation and economic growth in the EU. Growth Change 35(4):434–55 Bramanti A, Fratesi U (2009) The dynamics of an innovation driven territorial system. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Bramanti A, Riggi MR (2009) Sustainable interrelated growth: a phenomenal approach. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Camagni R (1995) The concept of innovative milieu and its relevance for public policies in European lagging regions. Pap Reg Sci 74(4):317–340 Camagni R, Capello R (2009) Knowledge-based Economy and Knowledge Creation: the Role of Space. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Cheshire P, Magrini S (2000) Endogenous processes in European regional growth: convergence and policy. Growth Change 31:455–479 Cohen, W, Levinthal D (1990) Absorptive capacity: a new perspective on learning and innovation. Adm Sci Q 35:128–152
190
R. Crescenzi, A. Rodr´ıguez-Pose
Cooke P (1997) Regions in a global market: the experiences of Wales and Baden-Wurttemberg. Rev Int Polit Econ 4(2):349–81 Cooke P (1998) Origins of the concept. In: Braczyk H, Cooke P, Heidenreich M (eds) Regional Innovation Systems. UCL Press, London Cooke P, Gomez Uranga M, Etxeberria G (1997) Regional innovation systems: institutional and organizational dimensions. Research Policy 26:475–91 Crescenzi R (2005) Innovation and regional growth in the enlarged Europe: the role of local innovative capabilities, peripherality and education. Growth Change 36(4):471–507 Edquist C (1997) Systems of innovation approaches – their emergence and characteristics. In: Edquist C (ed) Systems of Innovation: technologies,iInstitutions and organizations. Pinter Publishers/Cassell Academic, London European Commission (2006) The demographic future of Europe – from challenge to opportunity, COM(2006) 571 final, Brussels EUROSTAT (2006) Population statistics, Official Publication of EC, Luxemburg Fageberg J (1988) Why growth rates differ. In: Dosi G, Freeman C, Nelson R, Silveberg G, Soete L (eds) Technological change and economic theory. Pinter, London Fagerberg J (1994) Technology and international differences in growth rates. J Econ Lit 32:1147– 1175 Fagerberg J, Verspagen B, Caniels M (1997) Technology, growth and unemployment across European Regions. Reg Stud 31(5):457–466 Gordon IR (2001) Unemployment and spatial labour markets: strong adjustment and persistent concentration. In: Martin R, Morrison P (eds) Geographies of Labour Market Inequality. Routledge, London Gregersen B, Johnson B (1996) Learning economies, innovation systems and European integration. Reg Stud 31(5):479–490 Grossman GM, Helpman E (1991) Innovation and Growth in the Global Economy. MIT Press, Cambridge, MA Iammarino S (2005) An evolutionary integrated view of regional systems of innovation: concepts, measures and historical perspectives. Eur Plan Stud 13(4):497–519 IRPUD (2000) European Peripherality Indicators (E.P.I.). IRPUD GIS database. Institute of Spatial Planning, Dortmund Jaffe AB (1986) Technological opportunity and spillovers of R&D: evidence from firms’ patents, profits and market share. Am Econ Rev 76:984–1001 Leamer E, Storper M (2001) The economic geography of the internet age. J Int Bus Stud 32(4):641–665 ˚ (1992) National systems of innovation: towards a theory of innovation and interactive Lundvall BA learning. Pinter, London ˚ (2001) Innovation policy in the globalising learning economy. In: Archibugi D, Lundvall BA ˚ (eds) The globalising learning economy. Oxford University Press, New York Lundvall BA Malecki E (1997) Technology and Economic Development: The Dynamics of Local, Regional and National Competitiveness, 2nd edn. Addison Wesley Longman, London Maurseth PB, Verspagen B (1999) Europe: one or several systems of innovation? An analysis based on patent citations. In: Fagerberg J, Guerrieri P, Verspagnen B (eds) The economic challenge for Europe. Edward Elgar, Cheltenham Morgan K (1997) The learning region: institutions, innovation and regional renewal. Reg Stud 31.5:491–503 Morgan K (2004) The exaggerated death of geography: learning, proximity and territorial innovation systems. J Econ Geogr 4:3–21 Riggi MR, Maggioni M (2009) Regional growth and the co-evolution of clusters: the role of labour flows. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Rodr´ıguez-Pose A (1999) Innovation prone and innovation averse societies. Economic performance in Europe. Growth Change 30:75–105
Systems of Innovation and Regional Growth in the EU
191
Rodr´ıguez-Pose A (2001) Is R&D investment in lagging areas of Europe worthwhile? Theory and Empirical evidence. Pap Reg Sci 80:275–295 Sacco PL, Segre A (2009) Creativity, cultural investment and local development: a new theoretical framework for endogenous growth. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Storper M (1995) Regional technology coalitions. An essential dimension of national technology policy. Res Policy 24:895–911 Storper M, Venables AJ (2004) Buzz: face-to-face contact and the urban economy. J Econ Geogr 4:351–370 Verspagen B (1991) A new empirical approach to Catching up and falling behind. Struct change econ dyn 12:374–97
Some Conjectures on the Tie Between Digital Divide and Regional Disparities Marco Alderighi
1 Introduction In the mid-1980s there was much debate on the impact of information and communication technologies (ICTs) on productivity. The famous words of Robert Solow (1987) summarized the view of most skeptics: “You can see the computer age everywhere but in the productivity statistics.” In other words, according to Solow, there was no empirical evidence from the existing data that ICTs were generating productivity gains, but many firms were devoting considerable resources to ICT investment at the same time. This was known in the literature as the Solow paradox or productivity paradox. Many studies showed that ICTs contributed very little to productivity growth in the 1980s and in the early 1990s. Roach (1991) analyzed the productivity of white- and blue-collar workers between the mid-1970s and the mid-1980s. He showed that output per blue-collar worker grew by 16.9% between the mid-1970s and the mid-1980s, while output per white-collar worker, who mainly benefited from strong investment in ICTs, decreased by 6.6%. Loveman (1990) studied the impact of ICTs on the productivity of 60 business units estimating a production function by ordinary least squares. He showed that the contribution of ICT capital to output was approximately zero over the 5 year period studied in almost every subsample he examined. Barua, Kriebel and Mukhopadhyay (1991) found that there is a link between ICTs and intermediate indicators of productivity performance but not between ICTs and productivity. Morrison and Berndt (1990) using data from the US Bureau of Economic Analysis found evidence that every dollar spent on ICT in the manufacturing sector, delivered only about $0.80 of value on the margin, indicating general over-investment in ICT. Cron and Sobol (1983) studied the wholesale sector. They found no ICT impact on productivity even though they show that ICT use increases the variance of the performance, seeming to suggest that ICT adoption favors well-organized firms and damages the others. In the following years, many empirical and theoretical explanations were offered for the Solow paradox. From an empirical point of view, explanations for the Solow U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
193
194
M. Alderighi
paradox were found in error measurement issues (Wolff 1999), or in the fact that ICTs affect flexibility and product variety more than productivity (Brynjolfsson and Bharadwaj et al. 1999; Brynjolfsson and Hitt 1996). Other researchers have argued that the Solow paradox can be solved by looking at total factor productivity instead of labor productivity (Gera et al. 1999; Lehr and Lichtenberg 1999), by looking at quality improvements (Siegel 1994), by showing that it is difficult to identify the impact of ICTs since it accounts for small shares in the production (Oliner and Sichel 1994) and by composite explanations (Triplett 1999). From a theoretical point of view, it has been shown that ICT adoption requires a period of learning before it leads to improvements in productivity (David and Wright 1999; Mehmet 1998) and that ICT improvements are principally devoted to customer satisfaction, and that productivity gains related to ICT adoption are between but not within firms (Davis et al. 1993). At the end of the twentieth century, the Solow paradox was substantially solved as many papers identified a positive relation between ICT and productivity.1 However, among regional economists, it became evident that ICT were potentially able to produce new sources of differences between social groups, regions and countries, hence increasing the North-South, East-West dichotomy. For example, Martin (1998, 1999) sustained that differences in regional access, endowment and use of ICTs, known in literature as the “digital divide,” might account for differences in regional productivity. Especially in Europe, the relationship between digital divide and regional disparities became a central issue of the productivity debate (European Commission and DG Enterprises 2003). The President of the Committee of the Regions, Albert Bore (2003), said that “Fighting the digital divide is a particularly important objective given Europe’s changing age structure, the challenges of lifelong learning and changes in the workplace and economic activity.” More in general, the reduction of digital divide as a tool to increase social cohesion became one of the Lisbon goals (European Commission 2000). The relevance of this issue is clarified by the fact that EC allocated about 60% of Structural Funds to finance physical infrastructure to counterbalance existing development gaps. This paper aims to enter the debate on digital divide and regional disparities and at the same time to provide a new interpretation of the Solow paradox. In the model we are going to present, the two phenomena are strongly related as the causes explaining the productivity paradox are the same that link the digital divide to regional imbalances. A key element of our model is the assumption that the impact of ICT investment on productivity is not direct, but is mediated by the workers’ ability to manage ICTs, which determines the use of ICTs. In other words, ICTs increase the firm’s productivity when there are workers who can “use” them. Thus, productivity gains 1
It is not possible to provide a full summary of the empirical literature on the impact of ICT on economic growth as it is very extended. Only to mention the most relevant contributions we have Bartelsman and Doms (2000), Colecchia and Schreyer (2002), Gordon (2000), Jorgenson and Stiroh (1999, 2000a, b), Oliner and Sichel (2000), van Ark et al. (2002a, b). See also: http://www.nsf.gov/sbe/srs/seind02/c8/c8b.htm for a selected review on ICT. For a recent review, see: Alderighi et al. (2005).
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
195
are the combination of two factors: ICT investments and the workers’ ability to manage ICTs. From an empirical point of view, there are many important contributions that show the link between ICT skills and productivity. For example, Forth and Mason (2004) provide empirical findings that the skill gap is responsible for lower than average performance in UK manufacturing firms. O’mahony et al. (2003) show that the highly skilled workers experienced the largest gain in wages as a result of increases in ICT capital. Similarly, Card and DiNardo (2002) and Kahn and Lim (1998) identify skill-biased technical change due to ICT even where other authors, as Chun (2003), find short-term skill bias but not long-term. Finally, many authors (see, for example: Brynjolfsson and Hitt 2000, 2002; Brynjolfsson and Yang 2001; Greenwood and Yorokoglu 1997) find evidence that there are productivity gains especially when ICT investment is followed by organizational change, thus emphasizing the importance of organizational skills and networking skills for the exploitation of the potential provided by the ICTs. The “mediated” nature of the linkage between ICT investment and productivity may be an additional explanation to the Solow paradox. Even assuming perfect correlation between ICT use and productivity gains, we can find weak correlation between ICT investment and productivity provided that investment and use do not coincide. We will investigate later why an ICT investment is not always followed by a corresponding ICT usage. Now, it is worth noting that the mismatch between ICT investment and usage may provide additional elements to the debate on digital divide and regional disparities in terms of policies. In fact, many authors believe that in order to decrease the regional disparities (e.g., the differences in firm productivity between regions) it is necessary to reduce the digital divide (the difference in ICT investment between regions). However, as we have previously argued, regional disparities (such as difference in productivity) arise from differences in ICT usage and not in ICT investment, the reason why a simple investment in a laggard region may produce poor results. In a context of certainty, there is no discrepancy between investment and usage. Firms acquire exactly the technology that they are going to use. However, uncertainty is an important element in the decisions of firms, especially for ICT investment. Thus, we provide a model which includes uncertainty. In particular, we model the relationship between productivity, ICT investment and ICT skills using agency theory. Principal-agent theory started with the contributions of Baron and Myerson (1982), Guesnerie and Laffont (1984), and Mirrlees (1971). This stream of literature assumes that there is asymmetrical information between the principal (the management) and the agent (the worker). We assume that the worker has more private information on his level of ICT skills than the management. This hypothesis can be sustained by the following arguments. First, a worker can effectively retain more information than the management on his ICT skills (as he knows how good he really is). Second, as technology continually changes, firms do not know if their workers will be confident with the new generations. The basic framework is as follows. A firm – representative of a given regional economy – makes two sequential choices: first, it decides the investment level in
196
M. Alderighi
ICT, and second, it hires a worker offering an incentive-based contract.2 We suppose that investment is not per se productive but the use of ICT is productive, and output depends on the worker’s usage. This assumption is a key feature of the model. It means that if a technology is not utilized, it has no impact on the productivity. But at the same time, an investment is required in order to obtain productivity gains. Thus, the ICT investment choice limits the worker’s usage. In this model, output is the result of the use of ICTs. From the point of view of the worker, the usage of ICT and consequently production requires effort. We assume that the use of ICTs can be more or less costly in terms of worker’s effort conditionally on his ICT skills: high skilled workers are more accustomed to new technologies and exert lower effort than low skilled workers. Therefore, different workers may decide different levels of ICT use and hence produce different levels of output. Empirical studies seem to confirm our model assumptions. The one closest to our theoretical model is a study by Forth and Mason (2004), performed on micro data of the International Benchmarking Surveys (1997, 1998 and 1999) on UK firms. The authors find that only 18–24% of enterprises gave a positive answer to the following question: “Do your employees have sufficient understanding of the ICT available in your company to enable them to maximize the competitive advantage that these technologies bring?” They also find that ICT deficiencies are related to lower utilization of ICT and hence that skill constraints restrict the extent of ICT adoption and the intensity of use of ICT. In the same vein, OECD (2003) observes that “Other factors, such as, the regulatory environment, the availability of appropriate skills, the ability to change organizational set-ups, as well as the strength of accompanying innovations in ICT applications, affect the ability of firms to seize the benefits of ICT. Consequently, countries with equal ICT diffusion will not always have similar impacts of ICT on economic performance.” Despite the simple structure of the model, the set of results emerging from the analysis appears quite interesting. We will summarize: (1) The optimal contract offered by the firm is increasing in transfer and effort for lower types and flat for higher types (Proposition 1). (2) The investment decision is increasing in the ICT skill distribution and is decreasing in costs of investment (Proposition 2). (3) Provided that costs of ICT are sufficiently low, there is under-use and overinvestment with respect to a case of full information (Proposition 3). (4) Concerning the ICT paradox, the model predicts little correlation between ICT investment and regional performance as detected in the early studies on the productivity paradox (Proposition 4). The main policy implications are: (1) Sponsoring ICT investment is more effective when there are differences in costs than when there are differences in skills. 2
In this paper a region coincides with a local labor market area. In fact, we assume that a “regional” firm can hire workers only within its local labor market.
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
197
(2) Sponsoring ICT training and other skill enhancing is more effective when there are differences in costs than when there are differences in skills. The remainder of this paper is organized as follows: in Sects. 2 and 3, we will present the model and the main results; in Sect. 4 we show some of the policy implications; Sect. 5 concludes the paper.
2 The Model We set the model in a principal-agent setup. ICT technology. We consider the ICT investment as a constraint on the agent’s possible actions (use of ICTs). For example, the acquisition of a mobile telephone allows employees to be contacted away from the workplace; dedicated lines allow computers to be linked in two different buildings that otherwise could not share information, and so on. Formally, ICT investment, I, is modeled as an upper limit on the ICT usage, q, i.e., q ≤ I.3 Let C(I) be the adoption cost, with C strictly increasing and convex. The level of output Y only depends on the ICT use, i.e., Y = S(q), with S strictly increasing and concave. For simplicity, we only focus on the increase of output generated by the use of ICTs. The output is sold on a competitive market at price p, that is normalized to one. Note that the ICT investment does not automatically increase output; but it needs some commitment on the employees part. Agent. The agent has private information on his type θ ∈ [θ , θ ] (where θ denotes an agent with high ICT skills and θ denotes an agent with low ICT skills), and the principal is only acquainted with his type distribution, F. We assume that F has a monotone hazard rate ∂ [F(θ )/ f (θ )]/∂ θ ≥ 0. The agent uses ICTs with an effort level ψ = ψ (θ , q), which depends upon his ICT use q and his type θ . We assume that ψq , ψθ , ψθ θ , ψqq , ψθ θ q , ψθ qq ≥ 0. The utility of the agent depends positively on the extra-bonus he receives and negatively on his additional effort; we adopt an additive separable utility function (as, e.g., in Kofman and Lawarr´ee 1996) for analytical convenience: U = t − ψ (θ , q)
(1)
Principal. The principal faces a sequential problem: first, she has to decide the investment level in ICTs, and second, she has to hire a worker in the local labor market. Her objective is profit maximization: V = S(q) − t − C(I)
(2)
3 The investment decision is a multidimensional and discrete variable, but for our intents we assume it is continuous and one-dimensional. So, we assume that it is possible to order all possible levels of ICT investment and correspondingly different levels of usage.
198
M. Alderighi
3 Complete Information In case of complete information, the principal solves her optimization problem knowing the agent’s type. P1
max S(q(θ )) − C(I) − t(θ )
q(θ ),I
subject to U(θ ) = t(θ )− ψ (θ , q) ≥ 0 q(θ ) ≤ I,
(3) (i) (ii)
where (i) is the participation constraint and (ii) is the constraint on ICT use. The problem P1 can be rewritten in an easier form after these considerations. First, knowing θ , the principal offers a transfer of an amount exactly equal to the reservation payoff to induce the agent’s participation and extracts all the rent, hence: t(θ ) = ψ (θ , q(θ )). Secondly, she can choose the investment level without uncertainty and so fix it exactly equal to the level of ICT use by the agent: I = q(θ ). Hence P1 is equivalent to the simpler unbounded optimization problem P1 : P1
max S(q) − ψ (θ , q) − C(q). q(θ )
(4)
The first order condition is: S (q) − C (q) = ψq (θ , q)
(5)
With the aforementioned assumptions, equation (5) is also a sufficient condition for P1 . Hence in a world without informational asymmetries, the agent receives only his reservation payoff and the principal induces an effort level that matches the marginal benefits to the marginal costs. Using Dini’s Theorem, we can derive the slope of q(θ ): q (θ ) =
ψqθ (θ , q) S (q(θ )) − C (q(θ )) − ψqq(θ , q)
(6)
As expected, the level of ICT use that corresponds to the level of investment depends negatively on θ . In case of incomplete information, θ corresponds to the agent’s private information. The revelation principle (see, e.g., Laffont and Tirole 1993) implies that we can restrict attention to contracts that ensure that an agent truthfully announces his type. In this paper we focus on piecewise differentiable solutions. The principal behaves as a Stackelberg leader offering a contract, in which she has committed herself, specifying I, q and t for all possible configurations (θ ) reported by the agent and contracts are enforceable by the court.
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
199
4 Contractible Investment When information asymmetry is present, we distinguish two cases: contractible investment and non-contractible investment. In this paragraph, we examine the first. In the case of contractible investment the principal can simultaneously offer a contract to the agent based on t, q and I. The timing of the problem can be summarized as follows: (1) Nature determines θ , then agent knows it (2) Principal offers a set of incentive compatible contracts including the investment decision (t(θ ), q(θ ), I) where I ≥ q(θ ) (3) Agent announces θ , then (4) Principal invests I = q(θ ) and agent produces S(q(θ )) choosing q(θ ) ≤ I (5) Game stops. Agent gets t(θ ), and the principal gets S(q(θ )) − t(θ ) − C(I) The solution of the principal’s and the agent’s maximization is represented by the following optimization program:
θ
P2 subject to
max
q(θ ),I θ
[S(q(θ )) − t(θ )]dF(θ ) − C(I)
θ ∈ arg max t(θ˜ )− ψ (θ , q(θ˜ )) ∀θ θ˜
(7) (i)
U = t(θ ) − ψ (θ , q(θ )) ≥ 0∀θ
(ii)
q(θ ) ≤ q¯ ∀θ ,
(iii)
where (i) is the incentive-compatibility condition, requiring that truth telling be a Bayesian-Nash equilibrium, (ii) represents a participation constraint reflecting that agent knows his cost prior to contracting, and (iii) is the investment constraint on ICT use. This problem is consistent with the standard incentive problems and may be restated as follows (see Appendix 8): P2
θ
max q(θ )
θ
S(q(θ )) − ψ (θ , q(θ )) − subject to
F(θ ) ψθ (θ , q(θ )) dF(θ ) − C(q(θ )) (8) f (θ )
q (θ ) ≤ 0
∀θ
(a)
This is a calculus of variations problem (see, e.g., Kamien and Schwartz 1991), that can be solved supposing that constraint (a) is not binding and then verifying whether the obtained solution respects the constraints. The first order condition is: S (q(θ )) − C (q(θ )) = ψq (θ , q(θ )) +
F(θ ) ψqθ (θ , q(θ )) f (θ )
(9)
The Legrende condition is respected. Using Dini’s theorem we can derive the slope of q(θ ) that is non positive as required by (a):
200
M. Alderighi
q (θ ) =
ψqθ (θ , q(θ )) + ∂∂θ
F(θ ) f (θ )
θ) ψqθ (θ , q(θ )) + F( ψ (θ , q(θ )) f (θ ) q θ θ
S (q(θ )) − C (q(θ )) − ψqq(θ , q(θ )) −
F(θ ) f (θ ) ψqqθ (θ , q(θ ))
≤ 0 (10)
In fact the numerator is positive due to the assumptions made on the effort function and on the hazard rate and the denominator is negative by the hypothesis on the output function, effort function and the cost function. As we expected, the introduction of asymmetric information adds an extra term to the solution – compare equation (5) with (10) – which induces a reduction of the level of q for each value of θ . In fact, fix θ arbitrarily. In the first order condition for problems P1 and P2 the left-hand sides (marginal revenue net of ICT costs) are the same in both equations and are a decreasing function in q. The right-hand sides of the two equations (marginal payment to the agents) are both increasing functions of q and the term of equation (10) is greater than in (5). Therefore, the equilibrium quantity in problem P2 is lower or equal to that in P1. Considering also that at θ = θ the problem P1 and P2 have the same solution (because the two equations coincide), we have a clear characterization of the solution of these two problems that may be represented in Fig. 1. We call Q1 and Q2 respectively the curve that describes the ICT use by agents coming from the solution of problems P1 and P2; and I1 and I2 the investment decision of the firm related to Q1 and Q2. Note that in the case of perfect information or imperfect information with contractible investment, the investment curve and the usage curve coincide.
Fig. 1 Adoption and use of ICT in case of complete information and incomplete information with contractible investment
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
201
5 Noncontractible Investment In the case of noncontractible investment, there are some substantive differences in the results. Timing of the problem can be summarized as follows: (1) Nature determines θ , then agent knows it. (2) Principal invests I, then, (3) She offers a set of incentive compatible contracts to the agent considering the restriction imposed by her investment decision: (t(θ ), q(θ )) where q(θ ) ≤ I. (4) Agent announces θ ; then he produces S(q(θ )) choosing q(θ ) ≤ I. (5) Game stops. Agent gets t(θ ), and the principal gets S(q(θ )) − t(θ ) − C(I). The main difference with the previous case is due to the fact that the principal has to choose the level of ICTs before knowing the agent’s type. We have already shown that, provided there are no binding constraints, the use of ICTs increases with the skills of the agent. This means that if the principal wants to offer an incentive contract to high-skill agents, she has to invest more in ICT (in step 2) than if she wants to offer incentive contracts only to low-skill agents. Given these considerations we can imagine that the principal, to save money, may decide to offer incentive compatible contracts only to a subset of types. Consider now dividing the agents into three segments. Those with low ICT skills (θ L , θ ], those with intermediate skills [θ H , θ L ] and those with high skills [θ , θ H ). The principal can offer a contract not to all segments but only to agents with intermediate skills. Clearly the principal has to choose θ L and θ H . If she chooses θ L = θ this means that she offers incentives also to the lower segment. If not, low-skill agents will not be offered incentives to use ICTs. On the other hand, the principal has no interest in offering incentives to agents with high skills (i.e., those who have higher than the required skills), but in any case they can choose the contract offered to θ H . For the low-skill segment (θ L , θ ], the principal offers a contract that asks to produce nothing, giving nothing in return, so the participation constraint is respected. For the high-skill segment [θ , θ H ) the principal offers a contract asking to produce q(θ H ) and giving t(θ H ). This contract is accepted by the agent of this type because it is the best available option. So, the problem becomes
θL
P3
max
q(θ ),I θ H H
(S(q(θ )) − ψ (θ , q(θ )) − U(θ ))dF(θ )
+ F(θ ) S(q(θ H ) − ψ (θ H , q(θ H )) − U(θ H ) − C(I)) subject to
(11)
θ ≤ θ H ≤ θ L ≤ θ¯ (∗)
θ ∈ argmax t(θ˜ )− ψ (θ , q(θ˜ )) for θ ∈ [θ H , θ L ] θ˜
U = t(θ ) − ψ (θ , q(θ )) ≥ 0 q(θ ) ≤ I
∀θ ∀θ
(i) (ii) (iii)
202
M. Alderighi
In P3 the principal maximizes her expected payoff by choosing the level of investment I and by deciding who (i.e., the types) should be offered the incentive contract determining the values of θ H , θ L . The objective function (11) is composed of three terms. The first one (the integral) refers to the expected profit related to intermediateskill segment (as in P2). The second is related to the high-skill segment, which we have assumed will choose the contract of the type θ H . The third term captures the costs of investment. The term related to the low-skill types does not appear in the equation because it is nil (the principal offers nothing and pays nothing). Condition (*) indicates that θ H , θ L are free to assume the values of the interval [θ , θ ]. Conditions (i), (ii), and (iii) are the same as in problem P2 and are referred exclusively to the intermediate segment. Problem P3, after substitution and integration by parts (as in P2 solution), can be restated as follows:
θL F(θ ) S(q(θ )) − ψ (θ , q(θ )) − ψ (θ , q(θ )) f (θ )d(θ )+ P3 max f (θ ) q(θ ) θ H (12) + F(θ H )[S(q(θ H ) − ψ (θ H , q(θ H ))] − C(q(θ H )) subject to q (θ ) ≤ 0
(a)
This is a problem of calculus of variations with salvage term. As in the previous case, we solve the problem as if constraint (a) is not binding and then we check. The solution of P3 can be summarized by equations (13) and (14): S (q(θ )) = ψq (θ , q(θ )) + S (q(θ H )) −
F(θ ) ψθ q (θ , q(θ )), f (θ )
θ H ≤ θ ≤ θ L = θ¯
1 C (q(θ H )) = ψq (θ H , q(θ H )) F(θ H )
(13) (14)
Equation (13) is the Euler equation, θ L = θ¯ comes from the transversality condition on θ L and equation (14) is the transversality condition for θ H . The Legrende condition is respected. Using Dini’s theorem we can derive the slope of q(θ ) that is non positive as required by (a):
θ) θ) ψqθ (θ , q(θ )) + ∂∂θ F( ψqθ (θ , q(θ )) + F( ψ (θ , q(θ )) f (θ ) f (θ ) q θ θ q (θ ) = , θ H ≤ θ ≤ θ¯ θ) S (q(θ )) − ψqq(θ , q(θ )) − F( ψ ( θ , q( θ )) q θ q f (θ ) (15) The solution of P3 presents some elements of novelty compared with the solutions of P1 and P2. First, on the left-hand side of equation (13) the term C (q(θ )) does not appear. This means that the principal, after deciding the investment level, does not consider in her optimization problem the cost of investment in ICT; the cost is sunk. Second, this cost, discounted by 1/F(θ H ) ·C (q(θ H )) − a factor greater than one t enters the transversality condition (equation (14)) (which describes the investment decision). Let us rewrite equation (14) as in (16) to obtain an interpretation of 1/F(θ H ).
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
203
F(θ H ) S (q(θ H )) − ψq (θ H , q(θ H )) = C (q(θ H ))
(16)
θH.
The principal spends Consider the marginal expenditure of the principal at C (q(θ H )) in order to offer an incentive contract to the type θ = θ H and to allow segment [θ , θ H ) to get the contract of θ H . The marginal expected revenue (net of the incentive) is given by S (q(θ H )) − ψq (θ H , q(θ H )) times the probability that the agent belongs to [θ , θ H ]. Coeteribus paribus, when (marginal) investment cost shrinks, the segment of high skilled types [θ , θ H ) without incentive contract reduces and θ H gets closer to θ . Combining equations (5), (9), (13), and (14), we can characterize the solution of P3 compared with the solutions of P1 and P2. Consider first equations (5) and (14) at θ H . Because the marginal cost of ICT is higher in P3 than P1 (analogously with what was done at the end of the previous paragraph for equations (5) and (9)) the quantity q(θ H ) in P3 is lower than in P1. Comparing equations (9) and (13) at θ H , it is also possible to show (by the same reasoning) that q(θ H ) in P3 is greater than in P2. Hence q(θ H ) in P3 is between q(θ H ) in P1 and P2. For θ H ≤ θ ≤ θ¯ , using equations (9) and (13) again, we get q(θ ) in P3 greater than in P2. In Fig. 2, we show the result in a graph q(θ ) × θ . The solution to problem P3 requires differentiating between curve I3 (the ICT investment curve) and curve Q3 (the ICT use curve). The ICT investment curve is constant and parallel to the x-axis. It passes through the intersection of equations (13) and (14) that we call A. Use curve Q3 is flat for the value θ H ≤ θ ≤ θ¯ and coincides with the investment curve as far as point A, then it slopes down.
Fig. 2 Adoption and use of ICT in case of complete information and incomplete information with contractible and non-contractible investment
204
M. Alderighi
6 Main Results In this section, we provide the proofs of the results presented in the introduction assuming that firms behave according to case P3. As a result of the solution of P3 we obtain: Proposition 1. In case P3, the optimal contract offered by the firm, is increasing in transfer and effort for lower types and flat (at higher values) for higher types. Proof. See the discussion in previous section.
The following theorem concerns the determinants of regional disparities and digital divide (notice that we define respectively the region more endowed with skilled labor; i.e., the region that has an higher value for the average skill level) as “rich.” Proposition 2. In case P3, the investment decision is increasing in the ICT skill distribution (i.e., the higher the average value for ICT skills, the higher the ICT investment) and is decreasing in marginal costs of investment, more formally: (a) Assuming ψ (θ , q) = θ q, and that FR is a positive shift of FP (where R and P stay for rich and poor) such that FR (x) = FP (x + h), ∀x, h > 0 then IR > IP . (b) Assuming different costs in the two regions CR and CP such that CR < CP , then IR > IP . Proof. Part a) The monotone hazard rate and positive shift assumptions imply that FR / fR < FP / fP and from equation (13) that qR (x) < qP (x). Now, by contradiction, assume that IR < IP that means qR (θRH ) < qP (θPH ). As C is convex, from equation (16) we have that FR (θRH )[S (qR (θRH )) − θRH ] < FP (θPH )[S (qP (θPH )) − θPH ]. Replacing equation (13) we obtain FR (θRH )[FR (θRH )/ fR (θRH )] < FP (θPH )[FP (θPH )/ fP (θPH )] that holds if θRH + h < θPH . But using equation (13) it means we are requiring S (qR (θRH )) = θRH +FP (θRH +h)/ fP (θRH +h) < θPH +FP (θPH )/ fP (θPH ) = S (qP (θPH )). From concavity of S it means that qR (θRH ) > qP (θPH ) which contradicts. Part b) First, note that equation (13) is the same for the two regions so that qR = qP . The value of the incentive thresholds may vary. Combining equations (13) and (14), we obtain F(θ )2 / fR (θ ) = Cj (q(θ )) where j = R, P. Noting that the LHS of the equation is increasing in θ , the RHS is decreasing in θ , and CP > CR for each θ by assumption yields to θRH < θPH .
Finally, the wedge introduced by non-contractability explains the high levels of investment in ICT but the lower use in some circumstances. This phenomenon is captured by theorem 3. Proposition 3. In case 3, provided that ICT costs are sufficiently low, there is underuse and over-investment of ICT with respect to (a case of full information). Proof. The sketch of the proof can be done assuming that (marginal) costs tend to zero. We restate equations (5), (9), (13) and (14) with this assumption:
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
205
New solution of P1 (complete information): S (q) = ψq (θ , q)
(17)
New solution of P2 (incomplete information with contractible investment): S (q(θ )) = ψq (θ , q(θ )) +
F(θ ) ψqθ (θ , q(θ )) f (θ )
(18)
New solution of P3 (incomplete information with non-contractible investment): S (q(θ )) = ψq (θ , q(θ )) +
F(θ ) ψθ q (θ , q(θ )) f (θ )
S (q(θ )) = ψq (θ , q(θ ))
(19) (20)
If costs are nil, the main changes are in the solution to P3: in fact equation (20) describes a situation in which the principal decides to invest in ICT for all types of agents, and the ICT use (equation (19)) coincides with the case of problem P2 (equation (9)). Figure 3 represents the new situation. The level of investment in P3 is always higher than or equal to the level in P2 or P1, hence there is over-investment. The level of use in P3 is the same as in P2 and is always lower or equal to the one in P1, hence under-use. This result is due to the fact that when investment is non contractible and its cost is low, the principal does not care to over-invest because her goal is to offer incentive compatible contract to all the types of agents. In any case, in a context of incomplete information the agent under-uses ICT.
Fig. 3 Complete and incomplete information without investment costs
206
M. Alderighi
This model helps to shed light on the productivity paradox. We sustain that the puzzle may arise from the fact that empirical analyses usually investigate the correlation between ICT investment and performance but not between ICT use and performance. We show that even if we assume that there is perfect correlation between ICT use and performance however there can be scarce correlation between ICT investment and performance. This reasoning seems to be confirmed for example by Bergman et al. 1999. The following theorem sustain that as the relation between adoption and performance is mediated by the workers’ use, it is possible to show that there is scarce correlation between ICT investment and productivity.4 Proposition 4. Under case 3, assume that θ is uniformly distributed on [0, 1], S(q(θ )) = q(θ ),C(q(θ )) = q(θ )2 and ψ (q(θ ), θ ) = Dθ q and that D is uniformly distributed on [ 3.5,4.5], then there is scarce correlation (the correlation coefficient is about 0.12) between ICT investment and productivity (Productivity paradox). Proof. See Appendix 2.
7 Policy Implications at Regional Level The model proposed in the previous sections identifies at least three channels in which ICT produce regional disparities. The first one concerns the difference in the access to ICT (i.e., cost differentials), the second one is the difference in ICT skills between regional labor markets and the third one is the asymmetric information between managers and workers concerning workers’ ICT skills. All these channels produce similar results in terms of regional disparities as well as in terms of digital divide (lack of endowment) as we have demonstrated at the beginning of this section. This fact has often induced policy makers to conclude that there is a direct link between digital divide and regional disparities and that the remedy to the regional differences must be to fill the gap in ICT endowment. The reasoning is as follows. Since we observe a positive relation between digital divide and regional disparities and noting that ICT can have a positive effect on productivity, we could be induced to conclude that there is a direct causal link between digital divide and regional disparities. Consequently, reducing the digital divide will reduce the regional disparities. In this paper we sustain that the causality nexus between digital divide and regional disparities is not direct but it is mediated by the ICT skills possessed by the local workforce and hence the conclusions drawn by the previous reasoning must be re-modulated in order to keep it into account. First of all, we have to notice that there are many policy tools that can be used in order to reduce regional disparities such as subsidization of ICT investment, improvement of labor market flexibility and incentives for the acquisition of ICT skills. 4
The following results are obtained by simulation. In Appendix 2 we discuss the details.
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
207
We believe that the choice of the appropriate policy depends upon the origin of regional disparities. In particular, we sustain that subsidizing ICT investment is more effective when there are high differential costs or at least when the costs of investing in ICT are quite high, but it is not so effective when there are profound ICT skills differences or costs are low. When there are differences in costs5 and the same ICT skill distribution in the rich and poor regions, the subsidization of ICT access in the poor region permits to fill the infrastructural gap. On the other side, when the costs are similar but there are differences in ICT skills, subsidization produces distortions in the markets. In fact, this policy favors ICT investment in those regions where ICT is less productive. Nevertheless, when ICT investment costs are quite high, so that the number of workers belonging to the incentive segment (θ H , θ¯ ) in both regions is low, subsidizing the poor region can be effective in terms of reducing regional disparities even though this policy is quite expensive and not efficient. However, this strategy has temporary effects because the richer regions benefit more than the poor region from each reduction in ICT costs and regional disparities can re-emerge. Consequently, this policy should be expected only to induce an initial stimulus in the poor region to generate learning-by-doing effects and to start implementing policies directed at improving the ICT skills of the population. Educational or training policies are more effective when regional differences come from differences in ICT skills. This point needs some clarification: assume for instance that the two regions are characterized by strong regional disparities due to strong differences in the ICT skill distribution. Moreover, assume that available resources permit investing in X% of the workers of the poor region in order to increase their ICT skills. It is simple to show that the model predicts that the higher the ICT skills possessed by this X%, the higher will be production and hence the lower will be the regional disparities. Note, however, that as we increase the skills of a subgroup of the population, the incentive segment expands especially when we are close to θPH but the effect of an additional increase in the skills is very small when we move very far from θPH . This result leads us to say that it is not very effective to make only a small part of population very skilled but it is more effective to increase the average level. A second point (that does not emerge from the basic model but we can clearly obtain by endowing the model with some migration dynamics) is that the small group of workers with high skills in the poor region will migrate to the rich region because they can gain more. So, investment in high skills does not generate returns for the poor region. This phenomenon is well-known in literature as “brain-drain” i.e., richer regions will attract better workers. (Fratesi and Riggi 2007, Kanbur and Rapoport 2005). 5
Firms located in peripheral regions often suffer of additional costs when they choose to an ICT investment. First, the most advanced technology is not available or it is available with a temporal lag (for example, the broad band). Second, due to their peripheral location, there are high costs when they ask a software customization. Third, the quality of infrastructure can be lower (bottlenecks in Internet). Finally, some providers ask higher fees.
208
M. Alderighi
8 Conclusions In this work, we aimed at providing some suggestions for thinking about the digital divide and regional disparities, contrasting the widespread opinion that there is a direct causal nexus between the two variables. In fact, we assign a substantive role to workers’ skills and in particular to ICT skills and we base our analysis on an indirect relationship between digital divide and regional disparities mediated by the role of ICT skills. Under this assumption, that fits with the empirical findings of the recent literature (for example, Forth and Mason 2004), we derive some explanations for regional disparities and we propose some indications: (a) sponsoring the ICT investment is more effective when there are differences in costs than there are differences in skills, (b) sponsoring ICT skill improvements is more effective when there are differences in costs than there are differences in skills. The simple structure of the model makes it possible to generalize these results to keep into account many aspects. We sketch some feasible extensions and we provide our comments on the possible outcomes. First, the model can accommodate for migratory effects. High skilled workers of poor regions prefer to move to rich regions to increase their wage. Migration increases regional disparities and modifies the allocation of the resources. Second, the model could be extended to take into account externalities and standard issues. In the same fashion of O-ring theory (Kremer 1993), production is the combination of the effort of different workers so that similar abilities are required to reach a target level at lower costs. This induces firms to seek particular skill standards among workers and hence to exclude low skilled workers from the labor market. This context can produce segregation of the lower segments in the richer regions. Third, the model lacks dynamic and does not keep into account learning-bydoing issues. The former introduces difficulties in introducing incentive schemes as described by the rachet effect and the latter works in the opposite direction, inducing workers to exert greater efforts (with respect to the static case) to gain higher skills for the future. The balance of the two forces affects the evolution of the regional gap. Fourth, dynamic effects and migration can be also combined to generate path dependency and agglomeration. It is an open question whether the laggard regions can catch up the leading one or initial advantages persists for long time.
Appendix 1 The solution to problem P2 can be derived following Laffont and Tirole (1993): first, we characterize the transfer scheme that respects the incentive and participation constraint, second we replace it in the objective function so as to obtain an unconstrained optimization problem. The incentive constraint (9.i) requires that agent θ reports his type truthfully. We define φ (θ , θ˜ ) as the utility of agent θ when he announces θ˜ :
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
φ (θ , θ˜ ) ≡ t(θ˜ ) − ψ (θ , q(θ˜ )), ∀θ
209
(21)
To respect the incentive constraint we require this function to have a maximum at θ = θ˜ , ∀θ . So, the first order condition is:
φθ˜ (θ , θ˜ )|θ =θ˜ = 0, ∀θ
(22)
This means: t (θ ) − ψq (θ , q(θ )) · q (θ ) = 0, ∀θ . It can be shown (see for example Laffont and Tirole 1993), that when q (θ ) ≤ 0, the first-order condition, given by equation (22), is necessary and sufficient for the optimum. Let us define Φ(θ ) ≡ φ (θ , θ ) as the indirect utility function of type θ . Differentiating totally equation (21) (when θ = θ˜ ) and substituting equation (22), we obtain: Φθ (θ ) = φθ˜ (θ , θ ) + φθ (θ , θ ) = φθ (θ , θ ). Note that this is an application of envelope theorem to the maximization of equation (21).This yields Φθ (θ ) = −ψθ (θ , q(θ ))
(23)
Integrating this expression in [θ , θ¯ ], considering that the participation constraint imposes Φ(θ¯ ) = 0, we obtain: Φ(θ ) =
θ¯ θ
ψθ (θ , q(θ ))d θ
(24)
Hence the transfer, which is the sum of the effort function and utility function is given: t(θ ) = ψ (θ , q(θ )) +
θ¯ θ
ψθ (θ , q(θ ))d θ
The optimization problem may be written as P2 :
P2
θ
max q(θ )
θ
S(q(θ )) − ψ (θ , q(θ )) − subject to
θ¯
θ
ψθ (τ , q(τ ))d τ dF(θ ) − C(q) ¯ (26)
q¯ ≤ q(θ )
q (θ ) ≤ 0
(25)
(27) (28)
Condition (a) is the sufficient condition for the agent’s maximization problem of the agent. Then, integrating the integral by parts and substituting (iii), as for problem P1, q¯ = q(θ ), we obtain P2 .
Appendix 2 In this section we show how a model that may shed some light on the productivity paradox from a theoretical view. From our point of view, the puzzle arises from a problem in the measurement of the impact of IT due to the fact that empirical studies
210
M. Alderighi
Table A. 1 Case
Investment
Output
ρ (q, S(q(θ )))
ρ (q(θ ), S(q(θ )))
1 2 3
q = q(θ ) q = q(θ ) q independent of q(θ )
S(q(θ )) S(q(θ )) S(q(θ ))
1 1 0
1 1 1
measure the level of investment in ICT and not its the effective use. This hypothesis is confirmed by a recent study (Bergman et al. 1999). In what follows we show that even if there is a strong relationship between use and productivity, there can be a weak relationship between investment and productivity. In fact, in our model with a single agent (employee), productivity is given by output level S(q(θ )) and the investment is given by I. We consider a non-linear correlation measure ρ (R´enyi 1959; see also Granger and Terasvirta 1993) with the following proprieties: 1. 2. 3. 4. 5. 6.
ρ (X ,Y ) = ρ (Y, X ) 0 ≤ ρ (X ,Y ) ≤ 1 ρ (X ,Y ) = 0 if and only if X and Y are independent ρ (X ,Y ) = 1 if Y = f (x), where f is a one-to-one function ρ ( f (X), g(Y )) = ρ (X ,Y ) if f and g are one-to-one functions ρ ( f (X), g(Y )) = r(X,Y ) if the joint density of X and Y is Gaussian, where r(X ,Y ) is the linear correlation coefficient
Table 1 presents the main differences between the three models and the implied non-linear correlation coefficient. The model predicts perfect non-linear correlation in the first two cases (complete information and incomplete information with contractible investment) and nil correlation in the last case (incomplete information with non contractible investment). In the three cases the correlation between use and output is perfect: ρ (q(θ ), S(q(θ ))) = 1. In the remainder of this paragraph we present a very simple simulation that underlines the spurious relationship generated by case 3. Assume: θ uniform on [0, 1], S(q(θ )) = Aq(θ ), C(q(θ )) = Bq(θ )2 and ψ (q(θ ), θ ) = Dθ q. Table 2 presents the solution of the problem in the three cases. We have simulated 50.000 times a situation in which 1.000 firms have parameter A = 10, B = 1 and chosen D uniformly on [3.5, 4.5]. The parameter D reflects a difference in the effort function between agents in the simulated sample. Note that S(q(θ )) remains unchanged. Figure 4 shows the distribution of the correlation coefficient. Its mean is 0.117472861 and its standard error is 0.009919051. This result is summarized in Theorem 4. Some considerations can be drawn. First, introducing small differences between firms, we obtained a positive correlation between the level of investment and productivity (also in this case the output and the use are perfectly correlated). Second, similar results are obtained, keeping the parameter D constant and varying A and B.
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
211
Table A. 2 Case
Investment
Use
ρ (q, S(q(θ )))
ρ (q(θ ), S(q(θ )))
1
A 2B + 2Dθ
A 2B + 2Dθ
1
1
2
A 2B + 2Dθ
A 2B + 2Dθ
1
1
0
1
A B √ if θ ∈ 0, D 4 BD
A √ 4 BD
3
A otherwise 2B/θ + 2Dθ
9% 8% 7% 6% 5% 4% 3% 2% 1%
54
48
42
6
60 0. 1
0. 1
0. 1
0. 1
13 0.
0.
13
0
4
8
12 0.
12
06
00
8
94
11 0.
0. 1
0. 1
0. 1
0. 0
08 0.
08 2 0.
0.
07 6
0%
Fig. 4 The distribution of the correlation coefficient
Third, in case of ICT, it is important to focus on the link between the adoption and use. Hence, a first clarification of ICT paradox derives from the fact that both the principal’s over-investment decision and an agent’s under-use comes from rational behavior under asymmetrical information whereas it cannot be explained in a context of complete information. A second consideration on the ICT paradox is that even though we have assumed perfect correlation in the model between output and use, under no contractibility there is no correlation between output (productivity) and ICT investment. Moreover, by simulating a context with small differences between firms, we can reproduce situations which present small values for the correlation between productivity and investment. This result has to be interpreted as a spurious correlation.
212
M. Alderighi
When empirical analyses try to measure the impact of ICT on productivity, in reality they are testing two hypotheses jointly: first, that the ICT adoption and ICT use are correlated; second, that ICT use and productivity are positively correlated. If the data show low correlation, one or both of these relations are false.
References Alderighi M, Biagi F, Riggi M (2005) Growth, information and communication technologies and regional disparities: a review of theory and evidence. mimeo Bartelsman EJ, Doms M (2000) Understanding productivity: lessons from longitudinal microdata. J Econ Lit 38:569–594 Baron DP, Myerson RB (1982) Regulating a monopolist with unknown costs. Econometrica 50:911–30 Barua A, Kriebel C, Mukhopadhyay T (1991) Information technology and business value: an analytic and empirical investigation. Inf Syst Res 6:3–23 Bergman EM, Feser EJ, Kaufmann A (1999) Lean production systems in regions: conceptual and measurement requirements. Ann Reg Sci 33:389–423 Bharadwaj A, Bharadwaj S, Konsynski B (1999) Information technology effects on firm performance as measured by Tobin’s q. Manag Sci 45:1008–1024 Bore A (2003) Recommendation, Official Journal of the European Union, 10 April 2003 Brynjolfsson E, Hitt LM (1996) Paradox lost? Firm-level evidence on the returns to information systems spending. Manag Sci 42:541–58 Brynjolfsson E, Hitt LM (2000) Beyond computation: information technology, organizational transformation and business performance. J Econ Perspect 14:23–48 Brynjolfsson E, Hitt LM (2002) Computing productivity: firm-level evidence. Rev Econ Stat 85(4):793–808 Brynjolfsson E, Yang S (2001) Intangible assets and growth accounting: evidence from computer investments. mimeo Card D, DiNardo JE (2002) Skill biased technological change and rising wage inequalities: some problems and puzzles. NBER working paper, 8769. J Labor Econ 20(4):733–783 Chun H (2003) Information technology and the demand for educated workers: disentangling the impacts of adoption versus use. Rev Econ Stat 85:1–8 Colecchia A, Schreyer P (2002) ICT investment and economic growth in the 1990s: is the United States a unique case? Rev Econ Dyn 5:408–442 Cron WL, Sobol MG (1983) The relationship between computerization and performance: a strategy for maximizing the economic benefits of computerization. J Inf Manag 6:171–181 David P, Wright G (1999) General purpose technologies and surges in productivity: historical reflections on the future of the ICT revolution. mimeo European Commission, Enterprise DG (2003) 2003 European innovation scoreboard: technical paper No 3, Regional innovation performances. Available online at: http://trendchart.cordis.lu/ Davis G, Collins R, Eierman M, Nance W (1993) Productivity from information technology investment in knowledge work. In: Barker R, Kauffman R, Mahmood (eds.) Strategic information technology management. IDEA Group, Harrisburg, PA, pp. 327–343 European Commission (2000) Presidency conclusions. Lisbon European Council, March 2000 Forth J, Mason G (2004) Information and communication technology (ICT) skill gaps and company-level performance: evidence from the ICT professionals survey 2000-01. National Institute of Economic and Social Research (NIESR, London), Discussion Paper No. 236 Fratesi U, Riggi M (2007) Does migration reduce regional disparities? The role of skill-selective flows. Rev Urban Region Dev Stud 19:78–102
Some Conjectures on the Tie Between Digital Divide and Regional Disparities
213
Gera S, Gu W, Lee F (1999) Information technology and productivity growth: an empirical analysis for Canada and the United States. Can J Econ 32(2) Gordon RJ (2000) Does the new economy measure up to the great inventions of the past? J Econ Perspect 14:49–74 Granger CWJ, Terasvirta T (1993) Modelling nonlinear economic relationships. Oxford University Press, Oxford Greenwood J, Yorokoglu M (1997) 1974. Carnegie-Rochester Conference Series on Public Policy 46:49–95 Guesnerie R, Laffont JJ (1984) A complete solution to a class of principal-agent problems with an application to the control of a self-managed firm. J Pub Econ 25:329–69 Jorgenson DW, Stiroh KJ (1999) Information technology and growth. Am Econ Rev 89:109–115 Jorgenson DW, Stiroh KJ (2000a) U.S. economic growth at the industry level. Am Econ Rev 90:161–167 Jorgenson DW, Stiroh KJ (2000b) Raising the speed limit: U.S. economic growth in the information age. Brookings Papers on Economic Activity 2000(1):125–235 Kahn JA, Lim J (1998) Skilled labor-augmenting technical progress in US manufacturing, Q J Econ 113:1281–1308 Kamien MI, Schwartz NI (1991) Dynamic optimization, the calculus of variations and optimal control in economics and management. North-Holland, New York. Kanbur R, Rapoport H (2005) Migration selectivity and the evolution of spatial inequality. J Econ Geogr 5(1):43–57 Kofman F, Lawarr´ee J (1996) Collusion in hierarchical agency. Econometrica 6:629–56 Kremer M (1993) The O-Ring theory of economic development. Q J Econ 108:551–75 Laffont JJ, Tirole J (1993) A theory of incentives in procurement and regulation. MIT Press, Cambridge, MA Lehr W, Lichtenberg F (1999) Information technology and its impact on firm-level productivity: evidence from government and private data sources, 1977–1995. Can J Econ 32(2):335–362 Loveman G (1990) An assessment of the productivity impact of information technologies. Massachusetts Institute of Technology, MA. mimeo Martin P (1998) Can regional policies affect growth and geography in Europe? World Econ 21:757–774 Martin P (1999) Are European regional policies delivering? CERAS-ENPC, Universite de lille-1, Paris, France Mehmet J (1998) The information technology productivity paradox. Rev Econ Dyn 1:551–92 Mirrlees J (1971) An exploration in the theory of optimum income taxation. Rev Econ Stud 38:175–208 Morrison CJ, Berndt ER (1990) Assessing the productivity of information technology equipment in the U.S. manufacturing industries. National Bureau of Economic Research Working Paper 3582. RESTAT 79(3):471–481 O’Mahony M, Robinson C, Vecchi M (2003) The impact of ICT on the demand for skilled labour: a cross-country comparison. mimeo OECD (2003) ICT and Economic Growth, Evidence from OECD countries, industries and firms. Oliner S, Sichel D, (1994), Computers and output growth revisited: how big is the puzzle?, Brookings Papers on Economic Activity: Macroeconomics 25(1994-2):273–334 Oliner S, Sichel D (2000) The resurgence of growth in the late 1990s: is information technology the story? J Econ Perspect 14:3–22 R´enyi A (1959) A measure of dependence. Acta Mathematica Academiae Scientiarum Hungaricae, 10:441–451 Roach SS (1991) Services under siege – The restructuring imperative. Harv Bus Rev 69(5):82–91 Siegel D (1994) The impact of technological change on employment and wages: evidence from a firm-level survey of Long Island manufacturers. Mimeo Solow R (1987) We’d better watch out. New York Times Book Review 12:36, July Triplett JE (1999) The Solow productivity paradox: what do computers do to productivity? Can J Econ 32: 309–334
214
M. Alderighi
van Ark B, Inklaar R, McGuckin RH (2002a) Changing gear: productivity, ICT and services industries: Europe and the United States. The Conference Board, Economics Program, EPWP #02–02 van Ark J, Melka BJ, Mulder N, Timmer M, Ypma G (2002b) ICT investments and growth accounts for the European Union, 1980–2000. Groningen Growth and Development Centre, GGDC Research memorandum GD-56 Wolff EN (1999) The productivity paradox: evidence from indirect indicators of service sector productivity growth. Can J Econ 32:281–308
Interconnection-Infrastructure as a Prerequisite for the Development of Territories – The Role of Network Externalities Anna Creti
1 Introduction It has been said that “in a world of global economies and technological change such as information superhighways facilitating transfer of often exponentially growing data, it is likely that the challenge for firms may not always be to be among the first to produce the new information, but may instead be how to recognise, obtain, employ and complement the relevant innovative information” (De Bondt 1996). Telecommunications networks play an important role in information processing: firms “interact with other firms and acquire information on input costs, production technologies and other strategic data”(Tofflemire 1992). Information technologies (computers, mainframes, etc. – henceforth, referred to as IT) and telecommunications (telephone, fax, modem, virtual private networks) increase the speed and accuracy of processing, storing and transmitting information. Furthermore, they also enhance integration among firms by generating informal networks, and fostering the accessibility of large knowledge sources, such as patent publications and databank networks. We believe that these are among the most important forces determining territorial development. By analysing incentives to invest in IT/telecommunication we have a variable that serves as a proxy for measuring the intensity of internal relations of a local production system. As stressed in the Fratesi and Senn (2009, Introduction of this book), networking infrastructure enables interaction of economic actors with regional and external knowledge. Technological change of a given territory based on the use of telecommunications networks and on spillover effects linked to interaction among users is then made “endogenous” as an outcome of agents’ investment decisions.1 Furthermore, we deal with a partial equilibrium model in describing the dynamics of agents belonging to a given territory, but we do not specify the production process of the telecommunications sector. 1
Other models analyse informal networks in a cooperative game theory setting (see for example Calvo-Armengol and Jackson, 2003). U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
217
218
A. Creti
This hypothesis allows us to study the impact of communication technologies on firms’ investment decisions and technological levels. The main aim of this work is to determine factors that enhance or temper the’ private incentives of firms to invest in telecommunication technologies. From this point of view, even though telecommunications can be considered as public services, our analysis differs from the analysis of returns to public infrastructure (Ashauer 1989; Berndt et al. 1992; Garcia-Mila and McGuire 1992; Nadiri and Mamuneas 1994; Morrison-Schartz 1996). Our approach is quite different from that of Romer (1990), who builds a general equilibrium growth model. We maximise the representative firm profit, under the constraint of the demand function derived for monopolistic competition in the output market and consumer utility a` la Dixit-Stiglitz (love for variety), while Romer, setting a general equilibrium model, maximises the constant elasticity of substitution consumer utility under the technological constraint of capital accumulation. By assuming that the usage of network technologies directly affects the evolution of factor productivity, we analyse the impact of the parameters defining the network externality effect – the extent of the network effect and the number of users – on the steady state values of TFP and telecommunications investment. Our model also allows the factors which favour or temper technological change, measured by the steady state intensity of telecommunications investment (the telecommunications expenditures to total revenue ratio) to be studied. In addition, we consider consumer surplus, total profits and welfare in ouranalysis of the impact of communication technologies on the economic system we study. In this chapter, balancedness of connections in a territory imposes some restrictions on the network externality parameter that has to be sufficiently low with respect to the number of users to avoid congestion effects. We will call this restriction the network stability condition. The main conclusion of our study can be summarised as follows. If we look at a given territory, our model predicts that the most favourable condition for a firm to increase its investment in network technologies and obtain the most efficient total factor productivity is: being in a market with low product differentiation (and/or with a large difference between the elasticity of goods-substitution and the inter-industry price), – a large number of users, intensive network externality and considerable technological opportunities linked to the telecommunication network. These factors can be interpreted as determinants of territorial growth with respect to network infrastructure. However, where firms’ decisions involve strategic economic variables exhibiting “externalities”, there is conflict between private incentives to telecommunications usage and their effects on total welfare. In particular, if a large number of users and low differentiation stimulate private investment in telecommunications and usage, the same conditions, together with little network effect, give welfare decreasing with the intensity of competition. Balancedness of network interaction is therefore a necessary, but not sufficient condition to increase welfare. The remainder of the paper is organised as follows. After discussing the hypotheses of our model (Sect. 2), we analyse’ decision making of firms on telecommunication investment and technological level in a dynamic context (Sect. 3). The second part of the paper focuses on the analysis of the steady state levels, together with the welfare analysis (Sects. 4 and 5). Further research directions are suggested in the conclusions (Sect. 6).
Interconnection-Infrastructure as a Prerequisite
219
2 Hypotheses of the Model The major feature of using a telecommunication network is the externality effect. Here, we assume that the number of users and the externality parameter- i.e. the usability of the knowledge obtained by communicating with other users- are the determinants of the spillover effect. The total amount of knowledge obtained by communications can be thus interpreted as pure knowledge spillover. The externality effects are specific for each network configuration. For the purpose of the present model, we choose a star shaped network (Fig. 1), where each firm has bi-directional links with the other n − 1 users. The amount of spillover depends positively on the number of firms belonging to the network and on the externality parameter. We assume that each of the n firms is like firm 1 in Fig. 1, i.e., each firm knows that it is called by n − 1 users, and that it calls them. This hypothesis is more realistic than that of considering a fully connected network, where each firm talks with all of the others and benefits from symmetrical network effects. Firms are less interested than residential customers in network externality (defined simply as the possibility of contacting everybody), but they are more reactive to the effective contacts they make or they receive. Let us consider a representative firm i. In a first step, firm i decides its telecommunications usage Ni and the direct contacts with other j firms, in order to obtain the amount of information or spillover effect NSPi . Each time a firm i contacts other firms, she adds a “piece” of the total amount of their information to her own total amount of information NSPi , gathering ∑ eNSP j . The total amount of information j
obtained from contacting other firms is weighted by the parameter e < 1, which we
2
3
4
5
n
...
1
7 6
Fig. 1 Network interaction: the star-shaped network
220
A. Creti
interpret as the externality effect or the quality/benefit of exchange firm i-firm j. The parameter e is taken as exogenous by each firm. For simplicity, we assume that NSPi is a linear function of Ni and of ∑ eNSP j . j
Moreover, because of indirect links, NSP j will depend on N j and on the contacts between j and the other firms linked to j, since: NSP j = N j + ∑ eNSPk .
(1)
k
Taking into account the direct as well as indirect links, the networking function for firm i becomes: NSPi = Ni + ∑ e N j + ∑ eNSPk j
(2)
k
And so on, if firm k has other indirect links, they will be integrated in the firm i’s networking function. This process will stop when every indirect link of the network has been considered.2 In the case of the star-shaped network, the total amount of information spillover gathered by the communication network is: ⎛ ⎞ NSPi =
n 1 ⎜ ⎟ ⎝Ni + e ∑ N j ⎠ . 2 1 − (n − 1)e j=1
(3)
i= j
Externality effects are positive if e2 (n − 1) < 1: we call this assumption the network stability condition. Note that our formulation significantly differs from that of the R&D spillover model (e.g. Spence 1979; Levin-reiss 1988; De Bondt et al. 1992), which simply model the intra-industry pool of knowledge as follows: n
R&D = Ni + e ∑ N j . NSPi j=1 i= j
The relevant intra-industry pool of knowledge for firm i only consists of the total amount of knowledge created with its own R&D efforts and a part of the knowledge created by the other firms in the industry. Our claim is that, as IT/telecommunications are likely to create a web of links, the above formulation is not adequate to model the multidirectional interactions of a network of users. In fact, (3) indicates that the intra-industry pool of knowledge is a non-linear positive function of the extent of the spillover effect and the number of users; moreover, the sensitivity of the intra-industry pool of knowledge to both the parameter e and the number of users n is greater than that considered by the standard R&D spillover models. However, 2
In a topological approach, this process would correspond to the evaluation of the oriented graph obtained in the star-shaped network, where Ni is the value associated with the node i and eNSP j is the value of the arc originating from the node i toward the node j.
Interconnection-Infrastructure as a Prerequisite
221
as in Levin-reiss (1988), Romer (1990) and Grossman-Helpman (1991), we include a double counting effect: each time that firm i uses the telecommunication network, it contributes to the stock of general knowledge capital. We discuss the case in which a representative firm engages in telecommunications investment in order to increase the efficiency of its production process. We consider a simple Cobb-Douglas production function with constant returns to scale, as in Rebelo (1991): Y = AK (4) where At is the technical change or total factor productivity. The minimum total cost to produce Yt (Sato-Suzawa, 1984), once the optimum allocation of capital has been made, is as follows: TC = (Y cK )/A (5) Producers engage in telecommunications investment Ni , to increase the productivity of their input and to obtain competitive weapons in market rivalry, since a higher A yields lower prices and market expansion. Firms incur cn Ni as variable costs of usage, which represents what managers frequently call the telecommunication budget. We thus consider that the evolution of A, i.e., the dynamic change of the total factor productivity, is a non-linear function of telecommunications investment and spillovers.3 This cost reducing aspect is often used in the literature on technological change and R&D. However, R&D investment and the R&D capital services and spillover are embodied in the production function. The main difference with respect to these models is thus the hypothesis on technological change: we choose Hicks neutral technical change to better analyse the impact of communication technologies on TFP, without considering factor demand bias. For instance, Jones (1995) suggests that perhaps computers and other forms of capital play a complementary role in the discovery of knowledge. We think that, in some way, IT/Telecommunication are more similar to R&D than to other forms of input: like R&D, they allow information useful to the production process to be obtained, processed and stored. However, R&D is an internal source of information, while communication technologies are concerned with information channels external to the firms, and for this reason we believe that the impact of the IT usage is better modelled by a disembodied technological change. We thus assume that the telecommunication equipment is included in the variable capital, so it is not explicitly analysed. In our view, what is new and interesting to model is IT/Telecom usage and network externality effect on the technological decision of firms.
3
For a static model where the total factor productivity level is influenced by the communication technology usage, see Creti (1998). In that model, it is shown that, under perfect competition, the substitution effect between traditional inputs, like capital or labour, and the input information is related to the network effect, or the advantage that a firm obtains by using an input whose costs are shared by other users. Moreover, conditions under which the usage effects (or the factors that increase firms’ individual communications needs) overcome the network effect are investigated.
222
A. Creti
The TFP dynamic is given by: dAt γ δ˜ = At Ntδ NSPt dt
δ + δ˜ < γ < 1
(6)
The term on the right-hand side is analogous to a conventional production function, and, in that respect, it exhibits the standard neo-classical properties of positive marginal product and diminishing returns to communication technologies investment and spillover. Over time, the marginal productivity advantages decrease with respect to IT/Telecommunications expenditures and spillover. This means that later contacts are less significant: there is a crowding of information. Here δ and δ˜ can be interpreted as elasticities of the firm’s own communication investment and the spillovers to the TFP dynamics respectively, and they have a different meaning from the parameter e or the extent of the externality effect. The parameters δ and δ˜ have to be considered as the productivity or the effectiveness of the communication technologies and of their spillovers. We thus assume that a technical improvement of the communication is reflected in an increase of δ and δ˜ . For this reason, we will refer to these parameters in the rest of the paper as proxies of the technological opportunities linked to network usage. Several special cases of the productivity generating function (6) are used in other studies. If γ = 1, δ˜ = 0, we obtain Romer’s (1990) technological progress function; in our specification γ < 1 avoids the so called scale effects, as pointed out by Jones (1995), since it is no longer intuitive to assume that more resources are needed if one possesses a higher productivity level. With γ , δ˜ = 0, we recognise the Dasgupta-Stiglitz (1980) specification. When δ , γ = 0 and δ˜ = 1, we derive the Spence (1979) specification, which is also used by De Bondt et al. (1992). The Levin and Reiss specification can be obtained when γ = 0, and that of Sato-Suzawa (1984) results from setting γ = 1, δ = δ˜ = 0. We assume monopolistic competition in the output market. This allows us to investigate the influence of product differentiation and the number of competitors on technological change and welfare. We consider two kinds of goods: n differentiated products and homogeneous product. The utility is quasi-linear, i.e. it adds two kinds of sub-utility or two sectoral utility levels for the differentiated and the homogeneous goods. As in Grossman-Helpman (1991), the utility of the differentiated goods is a` la Dixit-Stiglitz, which means that there is linear homogeneity among the n number of varieties. The optimal allocation of expenditure across differentiated goods and homogeneous goods yields the usual demand function for each differentiated goods: a−b b yi = p−a X0 (7) i PD where4 a = 1/(1 − ρ ) > 1, b = 1/(1 − ζ ) > 1, with the following price index:
In monopolistic competition, marginal revenue of a firm is equal to ρ p (Grossman-Helpman, 1991). The condition a > 1 is needed so that the price elasticity of demand perceived by a firm will be larger than 1. This is required to avoid negative marginal revenue in a monopolistic situation.
4
Interconnection-Infrastructure as a Prerequisite
PD =
n
∑ p1−a i
223
1 1−a
.
(8)
i=1
Note that a is the elasticity of substitution across n differentiated goods, while b is the overall price elasticity of demand. We then logically assume a > b. The last hypothesis we need is the dynamic of exogenous variables, i.e., scale of demand, and in particular, the costs of N and K. As in Morrison and Berndt (1991), we assume that the exogenous variables increase at given constant rates (X0t = X00 e−σX t , cnt = cn0 e−σnt , ckt = ck0 e−σK t ).
3 Dynamic Profit Maximisation and Balanced Growth Rate The representative firm will maximise the following intertemporal profit:
∞
Max pit ,Nt
e−rt [yit (pit , PD )pit − TCit (yit , Ait ) − cn Nit ] dt,
(9)
t=0
under the constraints of: the demand for differentiated goods, the price index, the total costs, the dynamic of the TFP and the communication externalities, as well as the initial condition Ait = 0 = A0 . The dynamic maximisation is solved by using optimal control techniques.5 We are interested only in the symmetric (Nash) equilibrium, since all firms have the same cost and price conditions, and the same spillover, as firm 1 in Fig. 1. This yields: pit = pt and Nit = Nt ∀i = 1...n. In particular, a pt = (∂ yt /∂ pt )/(pt /yt ), the perceived price elasticity, obtained taking the price and telecommunications investment decisions of other firms as given, becomes aP = a − (a − b)/n. The perceived price elasticity is then always higher than 1. Using the growth rates of the FOC at the symmetric equilibrium (see Appendix I), we are able to calculate the steady state growth rates of the system variables. By combining them, we can look at the relationship between the growth rate of the control variable Ait and the growth rate of the state variable we are interested in, i.e. Nit : (δ + δ˜ )(1 − γ ) [cˆK (1 − b) + δ cˆn − δ bXˆ0 ] = (δ + δ˜ )Nˆ ∗ (10) Aˆ ∗ = g
We have a unique state variable Ait , two control variables, pit and Nit , and the co-state variable µt . Necessary conditions are also sufficient for global maximum if the Mangarasian sufficient theorem is satisfied. This theorem contains two conditions: a) the profit functions and the TFP equations are differentiable and concave in the variables (Nt , At ); and b) in the optimal solution, µt ≥ 0, ∀t ∈ [0,∞), if dAt/dt is non-linear in N¯ t . The model satisfies the first condition; the second condition has to be checked for each optimal solution, because dAt/dt is non-linear in Nt (Chiang, 1992, pp. 215–221). 5
224
A. Creti
where: g=
1 . [1 − γ − (δ + δ˜ )(b − 1)]
It is clear that the autonomous scale demand parameter has a positive impact on the TFP growth rate, while marginal costs of capital and IT/telecommunications have a negative effect.6 Reducing the growth rate of these marginal costs fosters the output growth rate. It could be argued that political measures in favour of deregulating the telecommunications market and fostering competition among IT suppliers, yielding lower marginal costs, would have a direct impact on’ the growth of firms. It is also interesting to look at the growth rate multiplier g. The elasticity of the own knowledge stock (1 − γ ) has a negative effect on g: the more the own knowledge stock is efficient, the less investment is required to increase it.The impact of IT /telecommunications on the TFP growth rate can be assessed looking at the term ˆ both (δ + δ˜ )(b − 1). The communication technologies positively contribute to A, through a direct (δ ) and an indirect effect (δ˜ ), which stimulate further telecommunications investments by increasing the productivity of these efforts. Moreover, the increased productivity induced by the communication technologies decreases total production costs and then the price level (cost reducing effect). This results in the firm’s revenue being increased by (b − 1) times the productivity change. The externality parameter e does not appear in the TFP growth rate. However, it plays an important role in the steady state values of the TFP and the telecommunications investment level, as we will see in the following paragraphs 4–6. Manipulating the FOC, we also acquire the two differential equations dA/dt = f (A, N), dN/dt = g(A, N), which are necessary to study the system dynamics. However, these differential equations depend on time: costs, price and the market scale factor grow in time at an exogenous rate. In order to apply the phase diagram technique, it is a prerequisite that the variable t does not enter into differential equations as a separate argument (the system has to be autonomous); otherwise, each point in a phase space can imply different directions of the system over time. When this is the case, it is not possible to make qualitative statements about the characteristics of a possible equilibrium (Chiang 1984). In order to apply the phase diagram analysis to the inter-temporal maximisation with exponentially growing prices and costs, we have to remove the time component from the dynamic problem. In the Appendix, we use a time elimination method to make the system autonomous: the endogenous variables will be deflated with their steady state growth rates; the appendix also describes the dynamic properties of the system.
6
The marginal cost of capital and IT/telecommunications has a negative influence on the balanced growth rate because it is inversely related to the steady state growth rate of IT/telecommunications investment. This effect may seem counter intuitive, as higher input prices imply that the potential benefits of telecommunications investment per unit of output increase. But in a situation with a price elastic demand curve (a, b > 1), higher input prices also imply that the price level will increase, which decreases demand more than proportionally.
Interconnection-Infrastructure as a Prerequisite
225
The two dynamic constraints in terms of the redefined or deflated variables are: ˜ dAd 1 + e(n − 1) δ γ = Ad Ndδ1 − σa Ad , dt 1 − e2(n − 1)
(11)
dNd = λ1 Nd δ1 (λ2 Nd 1−δ1 − λ3Ad b+γ −2 ) dt
(12)
where: Ad = At e−
(δ +δ˜ )[cˆK (1−γ )+cˆn − bXˆ0 ] t g
= At e−σat ,
(13)
(1 − γ ) [cˆK (1 − b) + bXˆ 0 − cn ] t g = Nt e−σN t , Nd = Nt e −
g=
1 , [1 − γ − (δ + δ˜ )(b − 1)]
δ + δ˜ = δ1 , 1 , λ2 = r + σa (1 − γ ), 1 − δ1 ˜ b b η X00 δ˜ 1 + e(n − 1) δ 1−b a p − 1 c λ3 = δ + n(a−b)/(1−a), cn0 1 − e2(n − 1) 1 + e(n − 1) K0 ap
λ1 =
Note that the subscript d means that the steady state values are obtained after the time elimination method has been considered, and the depreciation factor σa is the steady state growth rate of A. The initial values of the exogenous variables also appear (X00 , cn0 , ck0 ). Equating the dynamic equations to zero, we obtain the steady state value of the total factor productivity Ad and the telecommunication infrastructure Nd 7 :
A∗d
Nd∗
b c1−b X00 K0 = cn0 (r + σa )
b c1−b X00 K0 = r + σa
aP − 1 aP
b n
a−b 1−a
SN SN
1 1−γ +δ1 (b−1)
δ1 −1 1−γ +δ1 (b−1)
σa
,
(14)
1 δ1 (b−γ )−2(1−γ ) 2−γ −b (1−γ +δ1 (b−1))δ1 a−b S aP − 1 1−γ +δ1 (b−1) (1−γ +δ1 (b−1))δ1 N n 1−a σa SN (15) aP cno
where the externalities obtained by using a communication network are as follows:
7
Note that in equilibrium Ad and Nd are constant and positive, but the original variables grow at a constant steady state growth rate, as defined by the FOC of the original non-deflated system.
226
A. Creti
˜ 1 + e(n − 1) δ SN = , 1 − e2(n − 1) δ˜ . SN = δ + 1 + e(n − 1)
(16)
4 Analysis of the Steady State TFP and IT/Communication Levels In this section, we discuss in some depth the impact of the parameters defining the network externality effect (i.e. the number of firms, n, and the extent of the network effect, e) on the steady state levels of TFP and telecommunications investment.8 Several, and sometime contrasting, forces are at work. ˜ The term SN = {[1 + e(n − 1)]/[1 − e2(n − 1)]}δ represents the productivity level of the firm’s telecommunication expenditure or the pure network effect, while the term SN = {δ + δ˜ /[1 + e(n − 1)]} is the elasticity of the telecommunications investment on the TFP dynamics at the symmetrical equilibrium. A higher e and/or an increased number of users n increase the pure network effect: this enhances the productivity of the firm’s own network usage which results in higher marginal benefits of communication technologies. However, a larger amount of spillover as well as a larger number of users means that the appropriability of the firm’s own communications usage declines (or SN decreases).We therefore find two typical effects studied in the models on R&D and spillover. On the one hand, when e and/or n increase, the intra-industry spillover pool is fostered, which results in a benefit for the firm, but on the other hand, less appropriability of knowledge spillovers has a disincentive effect on telecommunications investment, as free riding and exploiting incoming contacts becomes more appealing to the firms. Although these two effects are similar to those analysed by the R&D models, they differ in intensity: as we already noted in paragraph 2, the impact of the parameters e and n on the intra-industry pool of knowledge is stronger in our model than that considered by the standard R&D spillovers models. Indeed, this will have substantial implications in the comparative statics analysis. The parameter n also has a positive impact on the steady state TFP and telecommunication investment level through the perceived price elasticity: a greater number 8
As regards the other parameters, note that the interest rate r, cn has a negative effect on both TFP and telecommunications steady state levels. Since b < 1, cK has a negative influence on both the steady state levels. The autonomous parameter of scale of demand (X) has a positive influence on A∗d and Nd∗ . The depreciation rate σa has a negative impact on A∗d (since δ1 < 1): the faster the depreciation rate, the lower the steady state total factor productivity, while its effect on Nd∗ depends on the “adjusted elasticity of demand”: if b + γ < 2, σa has a positive impact on the steady state telecommunications usage, but if the elasticity of demand is high, the effect could be (b−1) reversed. Remember that 1 − γ + δ1 is positive in order to have a stable dynamic system (see the Appendix).
Interconnection-Infrastructure as a Prerequisite
227
of firms implies a higher perceived price elasticity, which increases the perceived change in demand by a productivity improvement. However, an increase in the number of users also has a negative effect on Ad and Nd , as it increases what we call the the competition effect, i.e. the term n(a−b)/(1−a). This effect is driven by the DixitStiglitz specification of the demand for differentiated goods, which at the symmetric equilibrium is a negative function of the number of available varieties. The DixitStiglitz specification has been mainly used by the endogenous growth models and indeed differs from the linear demand specification, quite common to those R&D models focusing on quantity competition. Both of these opposite effects are stronger when product differentiation is low (the term a − b is high, and/or there is a great difference between the elasticity of substitution among goods a and the inter-industry price elasticity b). As it is analytically quite difficult to disentangle the above-mentioned effects, we mainly use numerical simulations.9 Whenever appropriate, we also compare our results with those of the literature on R&D and spillovers.
4.1 The Extent of the Network ExternalityEffect: Comparative Statics 4.1.1 The Impact of the Extent of Network Externality on the Steady State Telecommunication Infrastructure N∗d The impact of the network extent parameter e on Nd∗ is ambiguous, as the following derivative shows: ∗ ∂ Nd sign ∂e ⎧ ⎪ ⎪ ⎪ ⎨ δ˜ (n − 1)S(δ1(b−γ )−2(1−γ )) N = sign − ⎪ (1 + e(n − 1))2 ⎪ ⎪ ⎩ <0 ⎫ ⎪ ⎪ ⎪ 1 + e(n − 1) ⎬ δ˜ (δ (b−γ )−2(1−γ ))−1 ∂ , + δ˜ SN (δ1 (b − γ ) − 2(1 − γ ))SN SN 1 ∂ e 1 − e2(n − 1) ⎪ ⎪ ⎪ ⎭ >0
Our simulations yield: 9 The figures in this section are obtained from simulations done using “Mathematica 3”. The values of the parameters and exogenous variables are: a = 2.5, b = 1.5, cno = 0.8, cKo = 1.35, X00 = 100, ˜ r = 0.007, σa = 0.005, γ = 0.05,δ = 0.1, δ = 0.3, e = 0.2. Only if different values are used for some parameters do we report them.
228
A. Creti
Simulation result 1: the extent of network externality has a negative impact on telecommunication investment, unless the number of firm is sufficiently high. In particular, we find that when the number of firms is low, an increase of the extent of network effect leads to a decrease of the steady state communication level. However, when the number of firms increases, a higher network effect causes telecommunications investment to decline, and then to rise. In our model, the possibility of a positive impact of network extent on telecommunications investment depends on the double counting method,10 and more specifically on our modelling of the network externality effect. For instance, with a specification similar to that of the research and development model by Spence (1979) – which, in terms of our approach, would imply at the symmetric equilibrium dA/dt = N[1 + e(n − 1)], SN = [1 + e(n − 1)]−1 and SN = 1 – the positive effect of the network spillovers disappears, while the negative one prevails. Another important determinant of the sign of the derivative is the general price elasticity (b). The value of this elasticity is an indicator for the degree of product differentiation with regard to the homogeneous good. In our model, a higher level of b or less differentiated products increases the possibility of a positive influence of e on Nd∗ . This effect also differs from the findings of De Bondt et al. (1992), who conclude that a positive effect is more likely when product differentiation is moderate to high. In their model, moderate to high product differentiation implies that appropriability is larger. In our model, higher product differentiation has an opposite effect: it lowers the perceived price elasticity of demand and therefore it results in less competition, which in turns lowers the incentives to invest in cost-reducing innovation. Therefore, the positive effect of spillovers on IT/telecommunications is more likely to appear under low product differentiation. Our analysis confirms the results of the literature on R&D that high technological opportunities (i.e. high δ and δ˜ ) will cause a positive influence of e on Nd∗ . Summarising the results of this sub-section, we can say that, in our model, a higher number of firms in the product market raises the total volume of network effects and, together with high technological opportunities linked to communication networks and low differentiation, creates the conditions for greater appropriability of telecommunications investment. These three conditions are therefore the prerequisites for a positive effect of network externality on the steady state telecommunications level. 4.1.2 The Impact of the Extent of Network Externality on the Total Factor Productivity A∗d The technological performance of a firm can be measured by its steady state productivity level. In this sub-section, we investigate the relationship between technological performance and the extent of network effects. While the impact 10
It must be stressed that, in our model, the number of firms determines the stability of network effect: to have positive network effect we assume e2 < (n − 1). With many firms, the positive effect is then limited to a low range of the externality parameter.
Interconnection-Infrastructure as a Prerequisite
229
of δ and δ˜ on steady state TFP is always positive, the parameter e has again an ambiguous effect: ⎧ ⎫ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∗ ⎨ ⎬ ˜ ∂ Ad S δ (n − 1) ˜ S−1 S ∂ 1 + e(n − 1) . = sign − N + δ sign N N ⎪ ∂e (1 + e(n − 1))2 ∂ e 1 − e2(n − 1) ⎪ ⎪ ⎪ ⎪ ⎪ ⎭ ⎩ <0
>0
This yields: Simulation result 2: the extent of network externality on the total factor productivity shows a tail of two states. More precisely, when the externality parameter is low, an increase of e lowers A∗d , but this effect is reversed when the externality is higher. Figure 2 shows that the steady state productivity level first decreases then increases, until the “stability condition” is reached. This effect appears regardless of the number of users. To our knowledge, the impact of the extent of network effect on the technological performance of a firm has been little studied not only by the literature on R&D, but also by empirical studies on productivity and usage of information technologies. The rationale for simulation result 2 is that, as long as the extent of the network effect is too low, there is no incentive to invest in IT/Telecommunication usage, which shows a negative or zero impact on TFP. When the quality of the exchanges increases, i.e. when the information gathered by the community of users is more valuable to the firm, then a positive impact of network effect can be predicted. Another interesting explanation of our result should probably take a learning function into account, since it is recognised that extensive experience is needed by firms and organisations to exploit IT gains, as it is for most radically new technologies. Probably, positive learning effects may cause an increase in the extent of interaction among
Ad* 17 16 15 14 13 12 11
0.1 n=5
0.2
0.3 n=10
0.4 n=17
Fig. 2 The impact of extent of network effect on productivity level
0.5
e
230
A. Creti
communication technologies users, thus endogenising the shift from the negative to the positive part of the curve we depicted in Fig. 2. The investigation of this latter point is left for further research.
4.2 The Number of Users: Comparative Statics 4.2.1 The Impact of the Number of Users on Telecommunication Investment N ∗d In this subsection, we discuss the impact of the number of users on the steady state IT/Telecommunication level. When the number of firms increases, the costreduction investment is negatively affected by two effects: the lower appropriability and the competition effect. However, an increase of n has a positive impact on the network effect and on the perceived price elasticity: ∗ ∂ Nd sign ∂n ⎧ ⎪ ⎪ ⎪ ⎨ SN eθ1 1 + e(n − 1) δ1 (b−γ )−3−2γ ∂ ˜ = sign − + θ δ S 2 N ⎪ (1 + e(n − 1))2 ∂ n 1 − e2(n − 1) ⎪ ⎪ ⎩ <0 >0 ⎫ ⎪ ⎪ ⎪ 1 ∂ a − b a−b −1 ⎬ 1−a , 1− n + b θ3 + θ4 ⎪ ∂n ap 1−a ⎪ ⎪ ⎭ >0
δ (b−γ )−2(1−γ )
θ1 = SN1
<0
(1 − 1/aP)n(a−b)/(1−a),
θ2 = SN (1 − 1/aP)n(a−b)/(1−a), δ (b−γ )−2(1−γ )
SN n(a−b)/(1−a),
δ (b−γ )−2(1−γ )
SN (1 − 1/aP).
θ3 = SN1 θ4 = SN1
Our simulations show the key role of the externality parameter in studying the sign of the above derivative. Simulation result 3: a low number of users together with a low externality parameter discourage the IT/telecommunication investment. There is a positive correlation between the extent of the externality effect and the number of users, and this creates two distinct patterns. When the extent of the externality effect is low, an increase of the number of users has a negative impact on Nd∗ ; when e is high, this effect is reversed.
Interconnection-Infrastructure as a Prerequisite
231
The increasing pattern is another element that does not appear in the standard literature on R&D and R&D spillovers. An increasing pattern is more likely to appear when products are less differentiated (higher values of b), the technological oppor˜ tunities linked to the communication network increase (higherδ ) and the difference between the elasticity of substitution (a) and the inter-industry price elasticity (b) is high. 4.2.2 The Impact of the Number of Users on Total Factor Productivity A∗d The impact of an increased number of users on the steady state productivity level has the same determinants as discussed in Sect. 4.2.1: ∗ ∂A sign ∂n ⎧ ⎪ ⎪ ⎪ ⎨ SN aP ˜ SN ∂ 1 + e(n − 1) + ω4 ∂ = sign − ω1 + ω δ 2 ⎪ (1 + e(n − 1))2 ∂ n 1 − e2(n − 1) ∂ n aP − 1 ⎪ ⎪ ⎩ <0 >0 ⎫ ⎪ ⎪ a − b a−b −1 ⎬ 1−a . n + ω3 ⎪ 1−a ⎪ ⎭ <0
ω1 = SN (1 − 1/aP)n(a−b)/(1−a). ω2 = SN (1 − 1/aP)n(a−b)/(1−a). ω3 = SN SN n(a−b)/(1−a). ω4 = SN SN (1 − 1/aP). First, we analyse the impact of elasticity of substitution and inter-industry price elasticity. Simulation result 4: with strong network externality effects, concentrated markets reach the highest productivity level. Our simulations show that when spillovers are low, the productivity level decreases with an increased number of users. This decreasing pattern is also found when the difference between a and b is very low. When spillovers are high, a duopoly always obtains the highest productivity level. The total factor productivity decreases, then increases with an increased number of users, until the stability condition has been reached. The greater the difference between a and b, the faster the decrease. The increasing pattern is an interesting one and, to our knowledge, new to the literature. With higher network effect and less differentiation, entry determines first a decrease then an increase of the steady state productivity level. The intuition behind
232
A. Creti
these results is that the lower market shares due to the higher number of users on the communication network drive the productivity level down. This negative effect decreases when the number of firms increases and at a certain intermediate level of rivalry, the positive influence of more investment in IT/telecommunication and then a larger intra-industry knowledge stock becomes more important. The positive pattern is then driven by the same forces that determine the positive impact of the number of users on the telecommunications steady state level: high network effect and low differentiation. If we combine all the results of these sub-sections, we can expect that the most favourable condition for a firm to increase its investment in network technologies and to obtain more efficient total factor productivity is: being in a market with low product differentiation (and/or with a large difference between the elasticity of substitution among goods and the inter-industry price elasticity), a largenumber of users, a significant/considerable network externality extent and considerable technological opportunities linked to the telecommunication network.
5 Consumer Surplus, Total Profits and Welfare The last section of this paper concentrates on the impact of network externalities and number of users on consumer surplus, total profits and welfare. Using the FOC of the deflated system and the definition of consumer utility, consumer surplus can be written as: aP nTC (17) CS = b − 1 aP − 1 where: a−b
b 1−b a−1 b−1 cK0 n Ad TC = X00
aP aP − 1
−b
.
(18)
Remembering that a, b > 1, a > b, aP > 1, is it easy to see that the consumer surplus depends positively on the number of firms, reflecting love of variety, and on the perceived price elasticity, because if aP increases, the mark-up over costs decreases. The technological level or total factor productivity also has a positive impact on consumer surplus. Using FOC, total profits can be written as follows: ⎤ ⎡ δ˜ σ δ + a 1+e(n−1) 1 ⎦ − nπ = TC ⎣ (19) aP − 1 r + (1 − γ )σa Note that, as for consumer surplus, through TC, the technological level of firms has a positive effect on total profits. In order to have a viable situation in monopolistic competition, the long run total profits have to be equal to zero. This condition defines the no-entry condition and the optimum number of firms on the market (n∗ ).
Interconnection-Infrastructure as a Prerequisite
233
The first term of equation (19) can be interpreted as the gross profit’s ratio, while the second represents the IT fixed investment ratio. The impact of an increase in the number of users on total profit is ambiguous, because there are two opposite effects. On one hand, the entry of new users increases the perceived price elasticity aP , which has a negative effect on profits; on the other, it lowers the fixed cost ratio of investment in communication technology with network effects, which increases profits. We find that the first effect dominates the second one: Simulation result 5: the higher the productivity of the IT/telecommunication network, the stronger the negative impact of entry on aggregate profits. In particular, profit decreases with entry of new users and the appearance of a negative profit is more likely when technological opportunities (summarized by the parameters δ , δ˜ ) are greater. This situation is analysed in the traditional literature on R&D. The long-run equilibrium is attained at different optimun numbers of firms, depending on different values of elasticity of substitution a and the productivity of the network δ . Once the no-entry condition is determined, the stability condition gives us a √ condition for network effect associated with that scenario: e = n∗ − 1. This link between the optimal number of firms and the extent of the network effect ensures the coherence of the model. We find a result similar to that of Dasgupta-Stiglitz (1980): industries with larger technological opportunities tend to have fewer firms. A greater network effect decreases the second term in the brackets (see (19)) and thus reduces the probability of a negative profit rate. If network externality is high, then the number of viable users in a market will increase. A higher degree of product differentiation (lower level of a), decreases the perceived elasticity and hence makes a negative profit less likely. When a decreases, the number of viable users increases. The model therefore predicts that industries with high differentiation have more IT/telecommunication users than the industries with more homogeneous goods. By combining these findings we can conclude that traditional industries with high technological opportunities, littlelittle communication spillover and little product differentiation allow a small equilibrium number of network technologies users. Highly differentiated industries with low technological opportunities and high network effects are characterised by high equilibrium numbers of users. It is worth mentioning that we find an exceptional relationship between profit and entry, which occurs in a very particular scenario. With high differentiation (a = 2.5), very high technological opportunities (high δ = 0.9), and little network effect (e = 0.02), entry first decreases, then increases total profits. Thus, after a certain level, the effect of entry on the perceived price elasticity becomes lower. In other words, entry means smaller market shares and higher expenditure on telecommunications usage, because of a higher appropriability, and technological opportunities. This implies that a very concentrated industry or competitive industries are both viable. Our simulations show that, in this exceptional scenario, when technological opportunities are lower (δ = 0.6), the less competitive equilibrium is obtained for n∗ = 4, and the “more competitive” equilibrium is obtained for n∗ = 9. With respect
234
A. Creti
to the case where δ = 0.9, lower technological opportunities then have the effect of increasing the number of firms that characterises the less competitive equilibrium, and to decrease the optimal number of entrants at the more competitive equilibrium. A low value of a has the opposite effect: it increases the optimal number of firms at the more competitive long-run equilibrium and decreases n∗ at the less competitive solution.
5.1 Welfare We now analyse the total impact of the number of users, and the extent and productivity of the network effect on total welfare, as the sum of consumer surplus and total profits. It can be seen that technological opportunities increase welfare, since both consumer surplus and total profits depends on the productivity level A∗d . The impact of the extent of network effect on welfare is not clear-cut.Firstly, parameter e has a positive impact on total welfare, since it increases total profits. Second, in a previous section (simulation result 2) we found that a positive impact of e on the steady state productivity level A∗d is likely to appear when this parameter is high, regardless of the number of firms. When e is low, its positive effect on the total profit to total costs ratio does not easily counterbalance the negative impact on total costs, which affects both consumer surplus and total profits. The impact of the extent of network effects thus depends on its impact on the steady state technological level. This yields the following result: Simulation result 6: only with a mature network, the externality effect causes welfare to increase. In other words, when the extent of network effect is high, welfare is a positive function of the network externality extent. The influence of an increase in the number of users is again difficult to explain. Firstly, new users increase the perceived price elasticity, which increases consumer surplus and decreases total profits. Second, the influence of new users on the productivity level is ambiguous (see Sect. 10.4.4). The difference between a and b, and the extent of the network effect determines whether the influence is positive or negative. Third, when consumers show love of variety, entry directly increases the total surplus. Our simulations show that: Simulation result 7: with high differentiation, welfare increases with entry of new users. Still, this result can be counterbalanced by low network externality effects, as the following suggests: Simulation result 8: when the extent of network effect is low, total welfare decreases with entry.
Interconnection-Infrastructure as a Prerequisite
235
The degree of differentiation and extent of network externality effect are of crucial importance, since with low differentiation (high inter-industry price elasticity a) and low e, welfare decreases with the entry of new users. This result is consistent with the analysis of the impact of entry on the steady state TFP: in Sect. 4.2, we said that the steady state total factor productivity A∗d decreases with entry if the differentiation and the extent of spillover are low. The decreasing welfare is then driven by this negative effect on A∗d , which decreases consumer surplus and total profits. If we combine welfare characteristics with viability of the market, we findan interesting case. We said that with high a = 2.5 and high technological opportunities, two levels of rivalry are viable. But simulation result 7 indicates that with high differentiation, welfare depends positively on the number of firms. Therefore, if this industry comprises 3 firms, the welfare is lower than in an equilibrium with 12 firms. This industry is “locked into” an inferior industry structure. Even if we change context, we find a result that is often obtained in models that analyse the adoption of products that exhibit network effects (Katz and Shapiro 1986,1990). A “subsidy” to invest in communication technologies could help to bridge un-viable levels of rivalry.
6 Conclusions The main aim of this chapter was to identify factors that enhance or temper firms’ incentives to invest in communication technologies that contribute to the processing of information and that are characterised by the existence of a community of users. In particular, we focused on usage of communications technologies as a special “tool” which allows firms to obtain knowledge and therefore to influence their rate of technological change or total factor productivity in a dynamic context. We solved an inter-temporal profit maximisation, under the constraints of the demand derived from the Dixit-Stiglitz consumer utility function, the price index (as a measure of the intensity of competition) and the dynamic factor productivity. The modelling of the TFP is the most original contribution of this paper because it assumes that network structure matters. Moreover, over time, marginal productivity advantages decrease with respect to IT/Telecommunications expenditures and spillovers. This means that later contacts are less significant than earlier communications: there is a “crowding of information”. We also assumed that, without any cost-reducing investment, the TFP growth rate depreciates due to the technological activities of competitors: this reflects the Schumpeterian idea that innovators have to invest in process innovations (IT/Telecommunications in our model) to keep their market positions. The main conclusion of this study is that the relationship between productivity and the extent of network technologies and TFP or technological change is by no means a simple one. We cannot affirm that higher network effects temper a firm’s incentive to use communications technologies, as is the case of models that consider the similar problem of R&D and R&D spillovers (Spence 1979). These models do
236
A. Creti
not consider that an increased network effect also increases the TFP of a firm and therefore the impact of telecommunication investment on productivity. We found other interesting results: the impact of the extent of network externality effect on the steady state communications level is negative when the number of firms is low, and positive when the both number of users and product differentiation are high. This result is due to the “double counting method” or the effect of more users on the firm’s own TFP. As regards the impact of the network effect on the TFP, when the externality parameter is low, a slightly negative effect appears, but this effect is reversed when the externality is higher, until the “sustainability condition” is reached. This effect appears regardless of the number of firms. The optimal investment in telecom infrastructure, instead, decreases with the number of users when the extent of externality effect is low, and increases when “e” is high. An increasing pattern is more likely to appear when products are less differentiated; the technological opportunities linked to the communication network increase, and the difference between the elasticity of substitution and the inter-industry price elasticity is high. We also found – two patterns when we analysed the influence of the number of users on steady state level of TFP: the first is the traditional decreasing behaviour, and the second shows first a negative, then a positive impact of entry on TFP. The last section of the paper analyses consumer surplus and total profits, which both depend positively on the technological level or total factor productivity. In particular, the impact of the externality parameter is positive, while the effect of entry on total profit is generally negative. The results of our model show that the causality between centralisation/decentralisation and communication technologies is not simply driven by the usage, as has often been suggested by the studies analysing the impact of IT on firms’ productivity and organisation. The “optimal number” of users is determined in a complex way that combines market conditions, productivity of the communication network, and extent of network externality. Finally, we found that all the conditions on the parameters that determine the impact of the number of users on total profits also influence total welfare: welfare, in general, increases when consumers show high love of variety, and decreases if there is low product differentiation or low extent of network effect. Our model could be extended to take more complex network configurations into account. In such cases, we might find other interesting patterns for telecommunications and total factor productivity steady state levels. For example, if we considered the fully connected network, where each firm has symmetric spillovers, the conditions for a positive network effect would limit the value of the externality parameter “e” to discontinuous intervals. We would then expect more “irregular” results than those analysed in the present work. The detailed analysis of different kinds of networks is left for further research.
Interconnection-Infrastructure as a Prerequisite
237
Appendix Optimal Control for the Dynamic profit Maximisation To solve the dynamiuc profit maxmisation, we use the current value Hamiltonian, which allows us to disregard the discount factor (Chiang 1984): γ
˜
δ H c (pit , Nit , µt ) = [pit yit (pit ) − TCit (Ait , yti ) − cn Nit ] + µt (η Ait Nitδ NSPit ).
(20)
Under the symmetic Nash equilibrium conditions, the price index, demand and spillovers can be written as follows: 1 PDt = npt1−a 1−a , a−b
yt = X0b pt−b n 1−a , NSPt = Nt
1 + (n − 1)e . 1 − (n − 1)e2
The first order conditions for profit maximisation are then as follows: aPt T Ct ∂ Hc =0⇔ = pt ∂ pt aPt − 1 yt where
(21)
a−b ∂ yt pt , = a− ∂ pt yt n ∂ Hc ∂ = 0 ⇔ cn = (dAt /dt) ∂ Nt ∂ Nt ⎡ ⎤ ˜ 2 (n − 1) δ ˜ δ 1 + e ˜ ⎦ δ+ , = µt ⎣η Aγ N δ +δ −1 1 − e2(n − 1) 1 + (n − 1)e a pt =
∂ H1c dµ dµ TCt + µr ⇔ − + µr = =− , ∂ At dt dt At δ˜ dAt ∂ Hc γ δ +δ˜ 1 + e(n − 1) = η At Nt = 0⇔ . ∂ µt dt 1 − e2(n − 1) The transversality condition is: lim µt e−rt = 0
t←∞
In order to calculate the steady state growth rate of the system, we first consider the growth rates of the demand function and total costs at the symmetric equilibrium (for simplicity we drop the time subscript): yˆ = b(Xˆ0 − p), ˆ
(22)
238
A. Creti
T Cˆ = yˆ − Aˆ + cˆK . We then differentiate the first order conditions with respect to time to obtain the equilibrium rates of growth, assuming that the growth rate of the system will be constant in the steady state (the changes in growth rates over time are therefore ˆ equal to 0 and d µˆ /dt = d A/dt = 0): pˆ = T Cˆ − y, ˆ ˆ cˆn = µˆ + Aˆ − N,
(23)
ˆ µˆ = T Cˆ − A, ˆ (1 − γ )Aˆ = (δ + δ˜ )N. We are left with six independent equations (22, 23) and six unknown growth rates. We then solve the system, in order to obtain the following steady state growth rates: (δ + δ˜ ) Aˆ ∗ = [cˆK (1 − b) + δ cˆn − δ bXˆ 0], g (1 − γ ) [cˆK (1 − b) + bXˆ 0 − cˆn ], Nˆ ∗ = g b ˆ − γ ) − δ cˆn], y∗ ˆ = [Xˆ0 (1 − γ + δ ) − Z(1 g 1 ˆ − γ ) + δ cˆn − δ bXˆ 0] pˆ∗ = [Z(1 g
where: g=
1 . [1 − γ − (δ + δ˜ )(b − 1)]
Time Elimination Method In order to apply phase diagram analysis to the inter-temporal maximisation with exponentially growing prices and costs, we have to remove the time component from the dynamic problem by a redefinition of the variables. The variables At , Nt , yt , pt grow at constant rates that we denote respectively as σa , σN , σy , σ p . These growth rates are directly linked to the constant growth rates of the exogenous variables and to the various elasticities of the system and are easily obtained by differentiating the FOC of the dynamic maximisation problem with respect to time. When we deflate the endogenous variables with their steady state growth rates, we get an autonomous system of differential equations (Lucas, 1988, Barro-Sala i Martin, 1995). We then define the following new variables:
Interconnection-Infrastructure as a Prerequisite
Ad = At e− Nd = Nt e yd = yt e
239
(δ +δ˜ )[cˆK (1−γ )+cˆn −bXˆ0 ] t g
(1−γ )[cˆK (1−b)+bXˆ0 −cn ] − t g
= At e−σat ,
(24)
= Nt e−σN t ,
(25)
b[Xˆ (1−γ +δ )−cˆK (1−γ )−δ cn ] − 0 t g
pd = pt e
[cˆ (1−γ )+δ cˆn −δ bXˆ0 ] − K t g
= yt e−σyt ,
= pt e−σ pt .
(26) (27)
We redefine the profit maximisation problem in terms of these new variables. In order to get a new dynamic constraint in terms of the TFP, we differentiate equation (20) with respect to time: dAt −oat dAd dAt −oat = e e − σa At e−oat = − σa Ad . dt dt dt
(28)
Differentiating the FCO of the original non-deflated system with respect to time gives σa = γσa + (δ + δ˜ )σN ; we thus have (at the symmetric equilibrium): δ˜ dAd γ δ1 1 + e(n − 1) = Ad Nd − σa Ad . dt 1 − e2(n − 1)
(29)
where, for simplicity, we replaced (δ + δ˜ ) with δ1 . Using the assumption that all the exogenous variables grow at constant rate and the equation (24–29), the profit maximisation problem can be rewritten in terms of the new discounted variables. The starting values of the exogenous variables X00 , cL0 , cK0 , cn0 all enter the current Hamiltonian of this redefined problem. In order to have a meaningful situation, we require that r + σa (1 − γ ) >0. This condition is automatically satisfied when the transversality condition of the redefined problem is met.
Dynamic Properties of the System Looking at the equations (14) and (15) in the text, we see that the dAd /dt = 0 locus is positively sloped and convex. The slope of the dNd /dt = 0 locus is dependent on the value of the elasticity of demand together with the elasticity of the firm’s own TFP with respect to the productivity generation process. We call the term b + γ adjusted elasticity of demand because of the presence of the elasticity of dAd /dt to TFP level. We thus analyse three different situations: low, moderate and high adjusted price elasticity of demand.11 When b + γ < 2, the dNd /dt = 0 locus is negatively sloped, and it is horizontal when b + γ = 2. With moderate price elasticity of demand (2 < b + γ < 3 − δ1 or 3 − δ1 < b + γ < (1 + δ1)/δ1 ) the locus is positively sloped. When the price elasticity is high b + γ ≥ (1 + δ1 )/δ1 the qualitative analysis 11
The detailed analysis of phase diagram and transition to the equilibrium is available from the author upon request.
240
A. Creti
dAd Nd
dt
Nd*
=0
dNd dt
Ad*
=0
A
Fig. 3 Phase diagram with low adjusted price elasticity b + γ = 2
cannot be easily applied. We leave the answer to the analysis of the local stability equilibrium using a linearisation of the non-linear differential system around the steady state. As an example of the transition to the equilibrium path, we analyse the case of the low adjusted price elasticity of demand (b + γ = 2) Fig. 3. b+γ −2 δ1 When dNd /dt = 0, we have A∗d 1− = Ad d . Looking at the definition of λ2 d and λ3 this equality can be interpreted as the equality between the marginal costs of increasing the state variable (left hand side) and the marginal benefits of TFP (right hand side). When b + γ = 2, an increase of the TFP the dd /dt = 0 locus is the saddlepath. On the dAd /dt locus, the depreciation of the TFP level is equal to the telecommunications level that generates TFP. This curve has a positive slope because a higher TFP implies more depreciation, which has to be met with more TFP. This objective can be achieved with more investment in telecommunications level. We assume an initial productivity level of A0 < A∗d . At the steady state equilibrium, the firm will choose a constant level of discounted communications investment level, which is always equal to the constant discounted steady state level Nd∗ . This implies that the change in the productivity level is always the same. Because the initial discounted communications investment level of the firm is lower than the constant discounted steady state level Nd∗ and the change in productivity level is the same, the growth rate of the discounted productivity level of the representative firm will be larger than the constant steady state growth rate. In the next period, our firm will have A1 > A0 . Given A1 , the firm will choose the same discounted communications investment level, which results in a higher growth rate than the constant steady state growth rate. The process will continue but the speed of the adjustment slows down, since the difference between the constant steady state growth rate (A∗d ) and the initially discounted productivity level is smaller. Moreover, the speed of adjustment is lower that in the case of b + γ < 2. Higher adjusted price
Interconnection-Infrastructure as a Prerequisite
241
elasticity decreases the speed of convergence toward the discounted steady state level of communications investment. The qualitative analysis of the system dynamic suggests that, when the adjusted price elasticity of demand gets higher, the dNd /dt = 0 locus rotates counterclockwise. The system stays saddle point stable as long as the slope of the dAd /dt = 0 locus is steeper than the slope of the dNd /dt = 0 locus. This conclusion is also verified by checking the stability of the system by characteristic roots, as follows. The existence of a saddle point can be examined by a linearisation (first order Taylor expansion) of the non-linear differential equation system near the steady state (Chiang, 1984). The Jacobean matrix evaluated at the steady state (A∗d , Nd∗ ) is:
γ −1
γ
γ Ad Ndδ1 SN − σa δ1 Ad Ndδ1 −1 SN . b+ γ −3 b+γ −2 −λ1 λ3 (b + γ − 2)Ndδ1 Ad δ1 λ1 Ndδ1 −1 (λ2 Nd1−δ1 − λ3 Ad ) + λ1 λ2 (1 − δ1 ) N ∗ ,A∗
(30) b+γ −2 ) is 0 at the steady The last term reduces to λ1 λ2 (1 − δ1) since (λ2 N d 1−δ1 − λ3Ad γ state equilibrium. Moreover, at steady state, we also have: Ad Ndδ1 SN = σa Ad . The first term on the left hand side can be rewritten as: −σa (1 − γ ) To check the dynamic stability of the equilibrium, we have to know the signs of the characteristic roots (r1 , r2 ). In order to have a stable saddle point equilibrium, the two characteristic roots must have opposite signs. The Jacobean matrix contains all the relevant information. b+γ −3
det J = r1 r2 = −σa (1 − γ )(1 − δ1)λ1 λ2 + λ1 λ3 δ1 σa (b + γ − 2)Ndδ1 −1 Ad
. (31)
In order to have a locally stable saddle point, detJ has to be negative. The first term of detJ is negative, because γ , δ1 < 1. The second term is positive if (b + γ − 2) > 0, and negative if (b + γ − 2) < 0. In this latter case, we are sure that the dynamic system has a locally stable saddle point. Using the steady state equation: λ2 N d 1−δ1 = λ3 Ad b+γ −2 , the determinant can be rewritten as follows: det J = r1 r2 = −σa λ1 λ2 [1 − γ − δ1 (b − 1)]
(32)
The determinant is positive if 1 < γ − δ1 (b − 1) negative if 1 > γ − δ1 (b − 1) In the latter case, the system has a stable saddle point. When the determinant is positive, we are not able to determine the local stability of the system. In order to calculate this inference, we must calculate the trace of the Jacobean matrix (tr J), which is equal to the sum of the characteristic roots: tr J = r1 + r2 = −σa (1 − γ ) + (1 − δ1)λ1 λ2
(33)
Replacing the parameters λ1 = 1/(1 − δ1 ), λ2 = r + σa (1 − γ ), the trace is rewritten as:
242
A. Creti
tr J = r1 + r2 = r The trace is always positive since r > 0 in order to met the transversality condition of the original non-deflated dynamic system. A positive determinant associated with a positive trace describes an unstable equilibrium. There is an unstable node if: |trJ|2 ≥ 4 |J| ⇔ (r)2 ≥ 4σa
(r + σa (1 − γ )) [1 − γ − δ1 (b − 1)] . (1 − δ1 )
(34)
It then depends on the value of various parameters of the model. If this condition does not hold, there is an unstable f ocus (Chiang 1984, p. 64).
References Ashauer DA (1989) Is public expenditure productive? J Monet Econ 23:177–200 Berndt E, Morrison CJ, Rosemblum L (1992) High-tech capital formation and labor composition in U.S. manufacturing industries: an explanatory analysis. NBER Working Paper No. 4010 Chiang AC (1984) Fundamental methods of mathematical economics, 3rd Edn. Mc Graw Hill, Singapore Chiang AC (1992) Elements of dynamic optimisation. Mc-Graw-Hill, New York Creti A (1998) Telecommunication usage by firms: network externality effect in the production function, Chap.4. Ph.D. Dissertation, Toulouse University Dasgupta O, Stiglitz, JE (1980) Industrial structure and the nature of innovative activity. Econ J 90:266–293 De Bondt R, Slaets P, Cassiman B (1992) The degree of spillovers and the number of rivals for maximum effective R&D. Int J Ind Org 10:35–54 De Bondt J (1996) R&D and R&D spillovers: a survey. Int J Ind Org 3:30–56 Fratesi U, Senn L. (2009) Regional growth, connections and economic modelling: an introduction. In: Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Garcia-Mila T, Mc Guire J (1992) The contribution of publicly provided inputs to states economies. Reg Sci Urban Econ 22:229–241 Grossman G, Helpmann E (1991) Innovation and growth in the global economy. MIT Press, Cambridge Jones C (1995) R&D-based models of economic growth. J Polit Econ 103:759–784 Katz M, Shapiro C (1986) Technology adoption in presence of network externalities. J Polit Econ 94:822–841 Levin RC, Reiss PC (1988) Cost-reducing and demand creating R&D with spillovers. RAND J Econ 19:300–321 Morrison C, Schartz A (1996) State infrastructure and productive performance. Am Econ Rev 86:1095–1251 Morrison C, Berndt E (1991) Short-run productivity in a dynamic model. J Econometrics 16:339– 365 Nadiri I, Mamuneas T (1994) The effects of public infrastructure and R&D capital on the cost structure and performance of US manufacturing industries. Rev Econ Stat 76:22–37 Rebelo S (1991) Long run policy analysis and long run growth. J Polit Econ 99:500–521 Romer P (1990) Endogenous technological change. J Polit Econ 98:71–101
Interconnection-Infrastructure as a Prerequisite
243
Sato R, Suzawa R (1984) Research and productivity: Endogenous technological change. Auburn House, Boston Spence AM (1979) Product selection, fixed costs, and monopolistic competition. Rev Econ Stud 43:217–235 Tofflemire J (1992) Telecommunication external economies, city size and optimal pricing for telecommunications. J Reg Sci 32:77–90
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows Massimiliano R. Riggi and Mario A. Maggioni
1 Introduction Regional performances are traditionally explained in terms of factor endowments, like physical capital accumulation (exogenous growth models), openness to international trade (export-led growth models) or education and innovation dynamics (endogenous growth models). These approaches neglect productive interdependence within or between regions and industries as possible sources of growth. Interdependence may occur in several ways; we have decided to focus on labour flows as the way regions and industries interact. Clusters are defined as the intersection of territorial and industrial units; sectors and territories are the two central dimensions for the evolution of clusters. More precisely, cluster growth depends on two kinds of interactions: those which are industry-specific (intra-industry interactions) and those which are region-specific (inter-industry interactions). Inter-regional dependence ranges from mutualism (in which each region’s growth is positively related to the size of the other region) to competition, in which the growth of one region is obtained at the expenses of the other. In these models, the size of a region is measured by the number of firms and the employment level, but the mechanisms lying beyond these interactions (apart from a generic reference to the existence of spatial spillovers, inter-industry linkages and agglomeration economies), are often left unexplained. One way to make inter- and intra-regional interactions explicit is to take the role played by the mobility of labour into account. Labour flows act as the channel for these inter-industry and inter-regional interdependencies and determine the growth performance of a region. Variation in the growth performance of a group of regions may determine the existence of divergence or convergence dynamics which may be due to the different level of skills embedded in the migrating workers. An important methodological background to the chapter is the Ecological Approach. Ecological models (Dendrinos and Mullally 1985; Maggioni 2002a) define U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
245
246
M.R. Riggi, M.A. Maggioni
regions (or cities) in terms of their economic size and explain their growth as the result of both internal dynamics and the interactions with other regions (or cities). In this chapter, the two parts of the story (the pure ecological growth models and the role of labour flows in determining regional performance) are matched as follows: first the evolution of regional employment is analysed in relation to region-specific and industry-specific features; second, the skill composition of interregional labour flows is taken into account to explain the direct or inverse relationship between employment and wage dynamics. In order to understand the consequences of migration on regional disparities, two effects are identified: • Regional Convergence (mainly due to a “supply-side” effect). The increase in labour supply in the receiving region pushes wages downward, narrowing the interregional wage rate gap (Bencivenga and Smith 1997). • Regional divergence (mainly due to a “demand-side” effect). The inflow of skilled workers boosts economic activities and labour demand in the richer region and cumulative effects take place to the detriment of laggard regions (Ghatak et al. 1996; Dolado et al. 1994). The chapter tests the determinants of regional and industrial dynamics taking into account the effects of the skill composition of workforce (embedded in the migration of skilled workers) and compares these results with a “standard ecological” case in which inter-industry and intra-industry interactions are considered as the sole explanation of regional dynamics. The way territories and sectors affect the growth patterns depends on several factors which are investigated empirically. The empirical analysis conducted on 4233 State/industry couplets over the period 1988–2001, shows the relevance of the following situations: 1. if both inter-industry and intra-industry interdependencies are positive, the cluster experiences an explosive growth path; 2. if inter-industry and intra-industry interdependencies are negative, clusters live in a competitive industrial environment, in which clusters can hurt each other’s growth process. 3. if inter-industry interdependencies are positive and intra-industry interdependencies are negative, territorial synergies are at work (between different sectors in each territory) but there is strong interregional competition. 4. if inter-industry interdependencies are negative and intra-industry interdependencies are positive, the cluster is following a general trend of growth driven by the business cycle; at the same time, a form of competition on local inputs has a negative impact on regional growth paths. 5. if neither type of interdependency is significant, the growth process is rooted in other factors. Looking for a micro-economic foundation of these interactions we focus on the role played by inter-regional labour flows – which, according to the literature, may foster a process of either regional convergence or divergence – and we perform an empirical analysis on the effects of labour flows on both wages and employment level.
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
247
2 Patterns for Cluster Growth in Isolation As already stated above, ecological models are based on the principle of population dynamics. We interpret population here as the cluster employment level, whose regional and industrial composition we investigate. The first dimension allows us to grasp the interactions between firms belonging to different industries within the same region while the second dimension accounts for interactions between firms belonging to the same industry but located in different regions. Both kinds of interdependence act as channels of interconnection, within and between territories. Industries and regions play complementary roles in determining the growth patterns of clusters: Therefore, in order to investigate both aspects, we choose as elementary unit of analysis the concept of ‘cluster’ defined in terms of an industry-region couplet. In order to disentangle the role of interconnections in the growth process, following Maggioni (2002b), we first present the benchmark case of an isolated cluster, as represented in Fig. 1, with three main stages of development:1 • An initial stage in which the development is sparked by an initial, often exogenous, shock (and sustained by the involuntary informational spillovers about the profitability of the location provided by early entrants); • A second stage in which the drivers of Marshall’s (1920) agglomeration economies (labour market pooling, supply of intermediate goods and services, and knowledge spillovers) play a crucial role in sustaining endogenously the growth and the structural transformation of the cluster through start-ups and spin-offs;
Number of incumbents maximum m dimension (K) Maturity/ Decline
golden age discovery time
Fig. 1 The development pattern of a cluster 1
When interactions are considered, the shape of the curve can significantly change at any stage, depending on the strength and type of interaction.
248
M.R. Riggi, M.A. Maggioni
• A third (twofold) stage in which either the cluster achieves national/international leadership in a given sector/technology and becomes resilient, (i.e. able to withstand technological shocks and economic recessions2 ) or the cluster declines (both socially and economically) generating – within different institutional frameworks – huge migration outflows or mass unemployment.3 In basic ecological models, cluster size is interchangeably measured in terms of number of firms or employment level, but the relationships between stock and flows; between cluster size and cluster growth, are often left unexplained. This intuition – which corresponds to a well known recognised stylised fact of the development of industrial clusters – may be derived from alternative or complementary “explanations” such as the following (Maggioni 2006): 1. Spin-off and imitation: new firms within a cluster are often started by former employees of existing firms or originated by local people imitating successful entrepreneurs (through a sort of “contagion” process). Both phenomena are proportional to the incumbent mass; however, while the spin-off story alone will generate an exponential “explosive” development (if not balanced by some counteracting force or controlled by a variable “birth” rate), the imitation story – in a population of a given size – will generate an S-shaped development process since the imitation process is proportional to the product of the number of potential entrepreneurs and of actual ones (Anton and Yao 1995; Klepper and Sleeper 2002; Dahl et al. 2003; Maggioni and Roncari 2009, in this volume, and Garavaglia and Breschi 2009, in this volume). 2. Signalling (a): in an uncertain environment, with strong information asymmetries between insiders and outsiders, the number of firms (belonging to the same industry) already located in the cluster signals the profitability of the location (due to the quality of the workforce, the availability of intermediate inputs, the general “business climate”) to potential entrants (Pascal and McCall 1980). This informational forward spillover mechanism works even in absence of agglomeration economies and may generate “informational cascades” (Bikhchandani et al. 1992; Hirshleifer 1993), “herd behaviour” (Banerjee 1992) and, with strong relocation costs”, lock-in phenomena. An interesting point is made by Choi (1997) who shows that the presence of informational externalities and spillovers may also work backward, i.e. the herd behaviour of subsequent entrants influences the initial location decision, so that a bias against new locations can be created by the “fear of being stranded” (Choi 1997, p. 408).4 2
Becoming what Markusen (1996) calls a ‘sticky place’. A region starts to decline when it gets large enough to suffer from congestion, when the indigenous industries start to decline and do not attract a new generation of entrants into new industries and firms (Swann et al., 1998). 4 Similar behaviour, but in the context of complete information, has been studied by Farrell and Saloner (1987) under the name of “penguin effect” from the behaviour of a flock of penguins gathered “on the edges of ice floes, each trying to jostle the other in first, because although all are hungry for fish, each fears there may be a predator lurking nearby” (Farrel and Saloner 1987, p. 14). 3
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
249
3. Signalling (b): by choosing to locate into a well established (i.e. larger) cluster, a firm signals its quality to potential customers by showing its ability to survive arm’s length competition in the inputs (i.e. skilled labour, venture capital/bank funding, land, etc.) and in the output (especially if sold to other local firms as intermediate input) markets. This point is highlighted by Vettas (1999) who shows that – in an Hotelling-type model with both vertical and horizontal product differentiation – spatial agglomeration may be used as a high-quality signal by firms acting in an incomplete information environment. Thus choosing an established cluster is a quality-signalling and reputation-building strategy which, with imperfect ex-ante information on its own ability and risk-prone entrepreneurs, may even generate excessive clustering.5 Once the cluster becomes sufficiently large,6 the fear of excessive competition reduces the entry rate, thus stabilising the size of the cluster (Nocke 2003). 4. Information diffusion7 : information (news or rumours) about a new profitable location for a given type of firm may be diffused in a given population of potential entrants and entrepreneurs through an epidemic model (Grilliches 1957; Bass 1969). If one assumes that information diffuses through contacts, and that these contacts are random, then the rate of diffusion of an innovation at any moment is proportional to both the fraction of actual users and the fraction of potential users. Alternative interpretations assume that, at any moment, there is perfect information on the existence and nature of the new cluster. However each potential entrepreneur/existing firm, before deciding whether or not to locate (or re-locate), must compare the benefits and the costs of location. In “rank effects” models (David 1969; Ireland and Stoneman 1986) it is assumed that the heterogeneity of potential entrants causes different returns from entry and, indirectly, different dates of location. In the “stock effects” models (Reinganum 1981; Quirmbach 1986) the benefits from location depend on the existence of agglomeration economies and dis-economies. In the “order effects” models (Ireland and Stoneman 1986) it is argued that location benefits to a firm depend on its position on the order of entry (on the basis of a “first come, better served” criterion). 5. Anchor tenant: originally conceived in the real estate economics literature, this label has been imported into the high-tech clusters literature by Feldman (2002) and refers to the fact that the existence of a large established industrial firm creates externalities that “contribute to benefits of agglomeration” (Feldman 2002, p. 14). Thus the number of new start-up firms (and their internal growth) is therefore positively related to the number of anchor tenants in the cluster (due to knowledge spillovers, specialised input procurements and user innovation 5
A similar result, applied to the clustering of scientists on a minority of “hot topics”, is obtained by Rocco (2003). 6 The toughness of price competition is positively related to the size of the cluster, hence in larger markets opportunities are greater (more consumers, more suppliers) but price-cost margins are narrower. More talented (i.e. efficient) entrepreneurs benefit relatively more from larger markets (Nocke 2003, p. 4). 7 All the original references quoted in this explanation deal with technological diffusion. The interpretation within the location analysis framework has been proposed in Maggioni (2002a).
250
M.R. Riggi, M.A. Maggioni
networks). This process is empirically confirmed and theoretically modelled by Rauch (1993) with specific reference to “artificial” clusters where developers play an active role in building the membership of an industrial park through a carefully designed strategy of discriminatory land pricing.8 The location of large firms (either spontaneous or sponsored) may therefore act as catalyst of the clustering process in the early stages of an industry, when uncertainty is strong and no obvious location has still emerged. 6. Leader–suppliers relationship: Originally conceived in the Italian literature on industrial districts (Belussi 1988; Garofoli 1991; Bramanti and Maggioni 1997; Paniccia 1998), this explanation focuses on the composite (both synergetic and competitive) relationship existing between a small number of large leading and innovative firms – acting as organizers and coordinators of the activity of the clusters – and a large number of imitative small firms which act mainly, but not exclusively, as sub-contractors. The relationship between the development of these two populations of firms within the same industrial cluster/district is an example of complex co-evolution in which pecuniary externalities and competitive dynamics play different roles at different times (Folloni and Maggioni 1994; Folloni 2009, in this volume; Bramanti and Fratesi 2009, in this volume). Suppliers like to be in a district with a sufficient number of leaders because of the higher price they can get for their product and because of the “insurance” they derive from the plurality of buyers but they suffer from being in a district with many leaders because they fear their competition on inputs (land, labour, credit). Leaders like to be in a district with a sufficient number of suppliers because of the lower price they can pay for their intermediate inputs but they suffer from being in a district with too many sub-contractors because of the limited knowledge spillovers and the reduction in the appropriability of their innovation. 7. Institutional processes and social legitimacy: originally conceived in the organisational ecology literature (Carrol 1988; Hannan and Freeman 1989; Staber 1997) this explanation refers to the fact that density affects founding rates of “organisational population” (for our purposes: a given type of firm) through institutional processes. “when numbers are small, those who attempt to create a form must fight for legitimacy. (. . . ) Once a sufficient number of instances of the form exist, the need for justification (and thus the cost of organizing) declines. Other things being equal, legitimation of a form increases the founding rate of population using the form. If legitimacy increases with the prevalence of the form in the society, then legitimation processes produce positive density dependence in founding rates” Hannan and Freeman (1988, p. 21). If knowledge is assumed to be local, then the natural consequence of such a process is spatial clustering. The same process does not produce unbound growth because it is counterbalanced by competition: “the main source of negative density dependence is competition within and between populations. The more abundant are competitors, the smaller 8
This strategy is based on the subsidisation of early (or “seed”) tenants (usually nationally prominent firms) while “later tenants are paying for the privilege of benefiting from economies of agglomeration as firms accumulate in the park, allowing the developers to recoup the cost they incurred in subsidising early tenants” (Rauch 1993, p. 858).
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
251
the potential gains from founding an organization at a given level of demand for product and services” (ibidem). 8. Agglomeration economies and diseconomies: originally conceived by Marshall (1920) and later rediscovered – firstly by the regional and urban economics literature (Isard 1956; Henderson 1977) and secondly within the “new economic geography framework by Krugman (1991) – this explanation highlights that each new entrant increases the locational benefits to incumbents (because of the existence of a labour market pool, intermediate inputs pool, technological externalities and knowledge spillovers) only up to a point, then it reduces them when competition and congestion prevail. The previous factors are sufficient for explaining the growth of clusters per se, so the following paragraph which accounts for an extension is necessary to consider the interaction between different clusters.
3 Patterns for Cluster Growth in Interaction The foregoing is enough to explain the growth of an isolated cluster lacking any external interaction and therefore based only on its own economic size. Now consider the case in which two clusters exist. Define r a generic region while the ‘rest of the world’ is indicated by −r. Within each region, the economic activity is grouped in various industries (which may be divided into sector i and sector −i all other industries). By defining an industry-region couplet as ‘cluster’, it may be said that the evolution of a cluster population depends on its own economic mass (nir ), the consistence of other industries (n−ir ) within the same region (inter-industry dependencies) and the relevance of the same industry in other regions (ni−r ) (inter regional dependencies): dnir = f (nir , n−ir , ni−r ), (1) dt where dndtir expresses the variation of population (or its growth rate if the variable is taken in logarithms). Inter and intra-regional dependencies sum up any form of relational interconnection among clusters, and depending on the (beneficial versus detrimental) reciprocal influence, different forms of interactions arise. Say that a12 represents the effect of interaction that cluster 1 has on cluster 2 and a21 represents the effect of interaction that cluster 2 has on cluster 1. Different relationships can be modelled simply by changing the signs (or values) of a12 and a21 , so that the forms of interaction summed up in Table 1 arise. The fact that the unit of analysis is the “cluster” (i.e. a combination of a specific territory and a given sector) stresses the focus on the different interactions that can take place at regional level, allowing us to disentangle both internal and external interdependencies (Maggioni and Riggi 2008). Here we decided to focus our attention on labour market dynamics as preferred channels of inter- and intra-regional interactions, as worker flows represent,
252
M.R. Riggi, M.A. Maggioni
Table 1 Different types of bilateral interactions Types of interactions Competition Mutualism Commensalism Predation Neutralism Amensalism
Sign and values of interaction coefficients a12 and a21 −− ++ +0 +− 00 0−
especially within the US economy (which constitutes the focus of the empirical analysis) the most important connection between industries (within regions) and regions (within industries) as described in model I. The empirical analysis is then extended to consider the effects of population dynamics on income and growth. To this extent, we assume that a crucial role is played by the skills embedded in the workers moving between regions and industries. We measured the skill content of the labour force by proxying it with a suitable sectoral detail of the regional wage and the inclusion of appropriate control variables. Thus, in model II, we study the relationship between employment and wage variations. Labour mobility has become crucial in the recent economic debate. In particular, it is often recognized as a main driver for regional development. Emphasis has been given to either the role of flexibility of wage adjustment in matching labour demand and supply at macro level (Blanchard and Katz 1992), or the importance of wage determinants at a micro level nested in personal and firm features such as the relevant industry and skills (Bentivogli and Pagano 1999). Consistently with these conclusions, Ghatak et al. (1996) and Shields and Shields (1989) discuss the long term impact of migration, arguing that it is positive through all forms of capital deepening. In a regional framework, Fratesi and Riggi (2007) argue that inter-regional divergences may deepen if migration flows have a high share of skilled people. This turns to the role of human capital in the process of growth, as recently stressed in the endogenous growth literature. Dolado et al. (1994) use a Solow (1956) growth model, augmented by migration, and assess the contribution of human capital to economic growth. They found empirical evidence, at country level, of a positive impact of the human capital embedded in the immigrant population. An important role in the cumulative process is played by the level of technology in the advanced regions, where the inflow of high skilled immigrants workers can be easier due to skill specialisation and localised externalities. The introduction of a new technology may either switch around or reinforce the existing development pattern: a region’s attractiveness in fact depends on both wage levels and the productivity of the existing skills applied to the new technology (Desmet 2000). Notice that labour mobility can take place not only across regions but also across industries. In both cases, two opposite effects can be disentangled, depending on the size and importance of the skills embedded in migrant workers. The net effects
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
253
D
W
L1
W
S
D L1
L0
S
L0
D
L0
D
L0
S
L1
W0
W1
W1
W0
E0
E1
E
E0
E1
E
Fig. 2 Labour migration: complementary and substitution effects
depend on the balance between substitution effects (related to the supply effect), and cooperative effects (related to the demand effect). In Fig. 2, we show how inmigration flows may or may not boost cumulative effects on labour demand. More generally, according to the elasticity of labour supply and labour demand functions, two different outcomes can arise. Figure 2 (left side) illustrates the case in which the increase in labour demand is not able to compensate migration inflows, because skills brought by movers are not enough to boost growth and, in that way, to significantly shift upward the labour demand schedule (substitution effects prevailing). The final equilibrium exhibits an increase in employment (to E1 ), at lower wages (from w0 to w1 ). In Fig. 2 (right side) in-migration is able to boost growth within the host region and the final equilibrium is at higher employment and wages levels (complementary effects prevailing). The final equilibrium exhibits an increase in employment (to E1 ), at higher wages (from w0 to w1 ). In section 3 we test the existence and relevance of these substitution and complementary effects by measuring: • Intra-industry dependencies, as proxy for industrial interactions; a positive (negative) coefficient on this variable shows cooperative (competitive) effects operating across regions, with positive (negative) implications on the employment dynamics. • Inter-industry dependencies, as proxy for regional interactions; a positive (negative) coefficient on this variable shows cooperative (competitive) effects working within the region, with positive (negative) implications on the employment dynamics.
254
M.R. Riggi, M.A. Maggioni
4 The Empirical Analysis Two dimensions of analysis drive the chapter: the region and the industry. Regions are largely heterogeneous in terms of population, economic size, geographical dimension and the choice of a particular level of analysis may lead to very different results and divergent policy implications (Br¨ulhart and Traeger 2002). The problem is referred to as the Modifiable Area Unit Problem (MAUP), and concerns the arbitrariness of the geographical partition used in the analysis; it implies that results from statistical data analysis for geographical dimensions can be different if geographical boundaries are varied. The problem is twofold. On the one hand, the best aggregation scale has to be chosen (scale problem), on the other hand, units have to be correctly assigned to the right area (aggregation problem). To address this point, we refer to the data-set on which the analysis is implemented (County Business Pattern, by the US Census Bureau). County Business Patterns provides data for US counties and some thousand sub-sectors. The spatial aggregation level is not independent of the industrial aggregation level. Following Swann et al. (1998) and Maggioni (2002a), ‘States’ within the US were considered as the basic regional units of the analysis combined with digit3 NAICS sectors (see Table 2). Eventually, the work was implemented on 4233 clusters (State/industry couplets) across the time span 1988–2001. In order to assess the inter-regional interaction mechanisms, we centred our analysis on labour which shows a high degree of mobility between firms and regions, especially in the US. Labour mobility is thus considered as the way regions and sectors interplay in the current setting. The choice of labour as the focus of our empirical exercise may be explained both in terms of data availability and theoretical motivation. Labour flows may be considered as a rough proxy of human capital movements and entrepreneurship dynamics which are central to the process of cluster and regional development. Here we address the original question about the determinants of regional dynamics in two steps: first, we focus on the original idea of pure ecological models where the evolution of clusters is related to spatial and industrial interdependencies (Model I) and then, other variables, such as wages, human capital and immigration rates, are included (Model II), to assess the effect of regional growth on convergence or divergence at aggregate level.
4.1 The Basic Ecological Model (Model I) In this section, we analyse the occupational mass variation of clusters and its possible explanations. In terms of inter-industry and inter-regional dependencies (model I), the basic version of the model estimated is: dNtis = α0 + α1 esttis + α2 (esttis )2 dt + α3 empsizetis + α4 Intrat + α5 Intert + α6 igrti− + εit ,
(2)
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
255
Table 2 Industrial classification by 3-digits NAICS NAICS Sector code
NAICS Sector code
113
Forestry and logging
451
114 115
Fishing, hunting and trapping Agriculture and forestry support activities Oil and gas extraction Mining (except oil and gas) Mining support activities Utilities
452 453
486
234 235 311 312 313 314
Building, developing and general contracting Heavy construction Special trade contractors Food mfg Beverage and tobacco product mfg Textile mills Textile product mills
487 488 492 493 511 512
315 316
Apparel manufacturing Leather and allied product mfg
513 514
321
Wood product mfg
522
322
Paper mfg
523
323
Printing and related support activities
524
324
Petroleum and coal products mfg
525
325 326 327
Chemical mfg Plastics and rubber products mfg Nonmetallic mineral product mfg
531 532 533
331
Primary metal mfg
541
332
Fabricated metal product mfg
551
333 334
Machinery mfg Computer and electronic product mfg
561 562
335
Electrical equip, appliance and component mfg Transportation equipment mfg Furniture and related product mfg Miscellaneous mfg Wholesale trade
611
Scenic and sightseeing transportation Transportation support activities Couriers and messengers Warehousing and storage Publishing industries Motion picture and sound recording industries Broadcasting and telecommunications Information and data processing services Credit intermediation and related activities Security, commodity contracts and like activity Insurance carriers and related activities Funds, trusts, and other financial vehicles (part) Real estate Rental and leasing services Lessors of other nonfinancial intangible asset Professional, scientific and technical services Management of companies and enterprises Administrative and support services Waste management and remediation services Educational services
621 622 623 624
Ambulatory health care services Hospitals Nursing and residential care facilities Social assistance
211 212 213 221 233
336 337 339 420
454 483 484 485
Sporting goods, hobby, book and music stores General merchandise stores Miscellaneous store retailers Nonstore retailers Water transportation Truck transportation Transit and ground passenger transportation Pipeline transportation
(continued)
256
M.R. Riggi, M.A. Maggioni
Table 2 (Continued). NAICS Sector code
NAICS Sector code
441
Motor vehicle and parts dealers
711
442
Furniture and home furnishing stores
712
443
Electronics and appliance stores
713
444
Bldg material and garden equip and supp dealers Food and beverage stores Health and personal care stores Gasoline stations Clothing and clothing accessories stores
721
445 446 447 448
722 811 812 813
Performing arts, spectator sports, and related industries Museums, historical sites and like institutions Amusement, gambling and recreation industries Accommodation Food services and drinking places Repair and maintenance Personal and laundry services Religious, grantmaking, civic, prof and like organizations
dN is
where dtt is the time variation of employment, esttis represents the number of establishments, empsizetis the employment size of establishments, igrti− is the nationwide industrial rate of growth and εit is an error component. In order to account for cluster inter-dependencies, we set up two interaction variables: Intra and Inter. The rest of the country is supposed to influence one State development path, if it belongs to its “competitive arena” as defined through a Linda series;9 in this way, only the most important region/industry couplets are considered to be able to influence the ‘rest of the world’ (all other clusters) development path. The ecological model is implemented on 80digit-3 NAICS sectors. The results are shown in Table 3, where both inter-industry and intra-industry dependencies are reported if statistically significant at 10% l.o.s or below. Interestingly, competitive (synergic) intra-industry dependencies are generally associated with synergic (competitive) inter-industry dependencies.10 Two groups of sectors can be thus identified. 9
Maggioni (2002a) applied this method to territorial analysis to gauge regional interactions within sectors; on this basis, we can define interactions between regions. For each sector, it is possible to say whether or not a region plays a crucial role in the structure of some given sectors and to what extent sectors are concentrated or dispersed. When Linda indices are ranked, their graphical representation (the Linda structural curve) shows a decreasing monotonic path at the beginning; the first discontinuity corresponds to the individuation of the competitive arena within the upper part of the series. As a general finding, apart from some traditional sectors and natural resources, the number of States within each sector and each oligopolistic arena, shows that the most important sectoral shares of economic activities are not heavily concentrated (results omitted). In particular, sectors defining inter- and intra-industry dependencies (that are significant in the inferential analysis carried out on 80 sectors on the ‘ecological model’) show pretty dispersed sector shares. 10 In many cases the coefficients of either intra- or inter-industry dependencies (sometimes both) are positive or negative and significant to least at 10% of the significance interval, but the focus here is on the case in which both are significant. For a detailed presentation of the empirical analysis, see Riggi 2004).
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
257
Table 3 Ecological model (model I): estimation by 3-digits NAICS Positive intra-industry dependencies and Negative inter-industry dependencies
Negative intra-industry dependencies and positive inter-industry dependencies
Heavy construction Food manufacturing Fabricated metal product manufacturing Miscellaneous manufacturing Real estate Nursing care facilities Recreation industries Food and drinking services
Motor vehicle dealers Bldg mat. and garden dealers Health and personal care stores Gasoline stations Clothing and accessories stores Sporting and hobby stores General merchandise Stores Publishing industries Ambulatory health care services Ambulatory health care services Personal and laundry services Religious, grant-making and like organizations
The first (Agriculture support activities, Heavy construction, Food manufacturing, Fabricated metal product manufacturing, Miscellaneous manufacturing, Publishing industries, Real estate, Nursing care facilities, Recreation industries, Food and drinking services) shows similar growth patterns across States suggesting the relevance of a country-based industrial dynamics together with the existence a certain level of competition for inputs and resources with other industries within the same State. The second group (Motor vehicle dealers, Building material and garden dealers, Health and personal care stores, Gasoline stations, Clothing and accessories stores, Sporting and hobby stores, General merchandise stores, Credit and related Activities, Ambulance health care services, Hospitals, Personal and laundry services, Religious grant making and like organizations) collects sectors with strong vocation to serve local markets (non-base industries). For this reason their growth is positively associated with the size of other industries in the same State and negatively with the size of the same industry in other States (suggesting the existence of a moderate effect of spatial competition induced by cross-border trade).
4.2 Taking Skills into Account (Model II) This section addresses the implication of inter-industry and interregional dependence at the aggregate level, to assess the role of workers’ skills in affecting regional performance and convergence/divergence dynamics (model II). Since the skill composition of the employment workforce is not available in the CBP dataset, we used industries as proxies of the skill level. This hypothesis is very strong considering that several skills levels are used within each industry, but the argument still holds
258
M.R. Riggi, M.A. Maggioni
on average: each sector has a particular skill content which is most intensively used. The drawbacks deriving from such a hypothesis are mitigated by the industrial classification which is adopted in the analysis. The NAICS ‘North Atlantic Industrial Classification System’ groups activities instead of products, which makes industries better suited to describe the skill content necessary within a process. Consistently with the theoretical model, the dependent variable to be investigated is defined as the temporal variation of the employment mass in each cluster (or its growth rate, if variables are taken in logarithms). The skill effect component can be included by exploiting the large variety in sector classification available in CBP; it is embedded in the variable wage, computed as a difference between the wage prevailing in the cluster (i.e. in the couplets: region/industry) and average US industry wage. A cluster may enjoy a higher wage than the rest of the industry located elsewhere for a number of reasons. Leaving aside institutional and legal reasons (which in the US should account for a proportion of such a difference) higher wages would, in principle, signal a higher productivity which, in turn, may be explained in terms of different average size of firms, better organizational and management practices, better technologies, and above all labour force skills. Agglomeration economies are also important: when the effects of human capital formation and labour movement are taken into account, clustering behaviour can become crucial; in fact, conditions can be provided under which human capital is created and used more efficiently within specialised regions containing similar firms (Almazan et al. 2003). Furthermore, in order to assess the impact of population on income growth, the focus is then centred on the role of workers skills in affecting the evolution of clusters. For this purpose, the framework is enriched with the State average educational attainment (HK −s ) and controlled for the rate of foreign immigration on employment (Imm − rate−s). dNtis = β0 + β1dwtis + β2esttis + β3(esttis )2 + β4 Intrat dt + β5Intert + β6gi− + β7 HKt −s + β8 Imm − ratet −s + ζit
(3)
where symbols have the same meaning as above. Equations 2 and 3 are formally built to consider a ‘two cluster’ case. In reality, the sample encompasses some 4000 State/industry couplets; such a wide variety would complicate the empirical analysis and retain too many degrees of freedom if all possible combinations were taken into account. Sector peculiarities are thus tackled by applying the Frisch-Waugh-Lovell theorem (Frisch and Waugh 1933) within macro-sectors and implementing estimates for each, as listed in Table 2. Employment and wages are thus expressed in deviation from the relevant macro-sector average. The industrial classification is applied to 80 sub-sectors, and the role of skills is taken into account (model II). A first distinction is made between manufacturingrelated activities and services. After controlling for other variables (see (3)), the attention is focused on the relationship between variations in wages and variations
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
259
Table 4 Wage and employment variations by sector Manufacturing
Services
Negative ratio
Building contracting Beverage, tobacco prod. mfg Leather prod. mfg Nonmet. mineral prod. mfg Electr. equip, appl. stores Bldg mat. and garden dealers Food and beverage stores Pipeline transp. Publish. ind.
Inform. and data process. serv. Security and like act. Real estate Mgmnt of comp. and enterpr. Amb. health care serv. Hospitals
Positive ratio
Fishing and hunting Mining (exc. oil and gas) Utilities Textile mills Textile prod. Mills Apparel mfg Petroleum and coal prod. Mfg Chemical mfg Primary metal mfg Fabricated metal prod. Mfg Transp. equip. Mfg Miscellaneous mfg
Motor vehicle dealers Furniture stores Agr. And forestry supp act. Health and pers. care stores Miscellaneous stores Motion pict. and sound rec. ind. Broadcast. and telecom. Admin. and supp. serv. Nursing care facilities Social assistance Museums and like instit. Recreation industries Accommodation Food and drinking serv.
in employment, and we discuss the coefficients which are significant at 10% level of the confidence interval or less. Both manufacturing and services show interesting results (Table 4): within some sectors there is a positive ratio between variations in wage and variations in employment, within others an increase in wages is associated with a reduction in employment. The explanation of these results lies in the very nature of industries. In manufacturing industries, capital intensive sectors present a positive relationship, whereas in labour intensive sectors the relationship works the other way round. This is considered as evidence of the adjustment mechanism acting in the labour market when the sector is labour intensive – an increase in the number of workers is absorbed through a reduction in wages – and of cumulative effects in capital intensive sectors, where skills are more likely to matter within production (due to strong complementarities between labour and capital). In services, the boundary line between positive and negative influence of wage variations on employment variations divides services to production from services to persons. In the first case employment levels shrink as wages increase, with wages acting as the classical market adjustment mechanism; in the second case (mainly social services and entertainment), higher wages and higher employment are positively related (meaning that in those sectors it is likely that skills accumulated by
260
M.R. Riggi, M.A. Maggioni
Table 5 Wage and employment variations: a taxonomy
Services Manufacturing
Positive ratio
Negative ratio
Services to persons K-intensive
Services to production L-intensive
learning-by-doing procedures play a crucial role in the dynamics of employment and wage). These results are summarised in Table 5 which shows the different effects recorded for different kinds of manufacturing and service industries, while Table 4 lists the names of all individual industries to allow a more detailed analysis. Summing up the empirical results, it can be inferred that: the pure ecological model (model I) identifies, among others, two sets of industries: the first group of industries (mainly basic services) showing negative intra-industry dependencies and positive inter-industry dependencies; the second group of industries with positive intra-industry dependencies and negative inter-industry dependencies. The first group identifies clusters for which sector-specific competition goes hand in hand with territorial synergies. In these cases, clusters grow with diversity and compete within the same sector across States. The second group corresponds to clusters that, to some extent, benefit from vertical integration or compete on region-specific demand or supply factors with other sectors in the same State. In these cases, specialisation seems to boost growth. This echoes the analysis of Glaeser et al. (1992) who tested the empirical relevance of specialization as predicted by Marshall (1920), Arrow (1962), Romer (1986) and Porter (1990) versus diversification as defined by Jacobs (1969), – in explaining the growth of local economic systems. When the analysis is performed at the macroeconomic level (model II) a more sophisticated taxonomy within services and manufacturing activities arises. Services to persons and capital intensive industries show a positive correlation between wages and employment variation while the opposite holds for services to production and labour intensive industries, suggesting a more crucial role played by informal skills (not described by educational attainment), tacit knowledge and know-how in the first group of activities.
5 Conclusions The paper suggests that regional development paths can be better understood by considering interactions between and within regions. Two dimensions of interactions have been identified: inter-regional interactions, when considering spatial differences within the same industry and inter-industry interactions, when considering sector peculiarities within the same region. Starting from a basic ecological approach, where employees are the units of the population investigated, the role of labour flows is introduced as an explicit channel
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
261
of regional interactions when looking at the effect of skills content on the dynamics of regional unbalances. In theory, two different outcomes may arise following the inflow of labour. When an increase in demand for labour is not able to compensate for the increased labour supply (which is the case if the skills brought by in-movers are not strong enough to boost labour demand), the ecological setting forecasts the “competition effect”. When in-migration is able to boost labour demand within the host region, we expect a prevailing complementary effect and the final equilibrium is reached athigher employment and wage levels. The empirical analysis has shown that the sign of the ratio between employment dynamics and wages is in fact ambiguous and, crucially, industry-dependent.11 Building a tentative (and very rough) taxonomy based on the industry average skills level, we find that in general a complementary effect arises in high skilled sectors, whereas a competition effect arises in low skill sectors. The analysis performed in the chapter has, therefore, some relevant policy implications: regional disparities are more likely to persist (because of the attraction of skilled workers to more advanced regions)12 rather than to narrow (because of the existence of automatic adjustment mechanisms within the sole labour market). Counteracting this tendency would imply trading off efficiency for equity. Optimal policies can be based on education and training provision as escapes to cumulative effects acting in favour of more advanced regions. These policies can be successful only within a systemic framework encompassing, on the one hand, targeted skill-building policies and, on the other hand, the diffusion of new technologies into lagging regions in order to reduce the interregional productivity gap (Crescenzi and Rodr`ıguez-Pose 2009, in this volume). At the same time, a necessary condition for the growth process to be triggered is the setting up of targeted infrastructure policies. Scholars (see, among others Baldwin et al. 2003; Martin 1999) usually support the idea that important market failures come from some factors being immobile. As a consequence, it is not agglomeration per se that is harmful, but the fact that the welfare of immobile factors is not taken into account. If this were the whole story, the reduction of transport costs would be a sufficient response. In the framework of this paper it would be desirable to displace workers from clusters with negative inter/intra dependencies to clusters with positive values. Such intervention, however, would imply social costs for people moving and require targeted labour force re-training programs. All the reasoning above is crucially dependent on the rate and direction of technological change. In a world without innovation, agglomeration dynamics – only partially counteracted by congestion mechanisms on immobile factors in the advanced regions – may determine the desertification of peripheral and laggard areas as theoretically demonstrated in the seminal paper of the “new economic geog11
This results may be seen as (indirectly) consistent with the analysis performed by Arbia et al. in this volume, which shows very different locational patterns and dynamics of different industries at the NUTS2 level in the EU. 12 Through different forms of “brain drain” dynamics (Commander et al., 2003; Kanbur and Rapoport 2005).
262
M.R. Riggi, M.A. Maggioni
raphy” literature (Krugman 1991). If, on the contrary, the speed of technological change is high, the economic leadership of a region may become the source of its own downfall when a radical innovation is introduced. Brezis et al. (1993) identify the following conditions for leap-frogging to happen, thus resulting in catch-up opportunities for laggard regions with respect to richer ones: • Past experience with the old technology is irrelevant for the new technology (i.e. cumulative output produced by using the old technology has no effect on the new technology learning curve). • The new technology is potentially superior (i.e. for a given amount of cumulative output the new technology is more productive than the older or, in other words, learning effects are greater for the new technology). • Despite a potential advantage, the new technology is initially inferior to the old in an established cluster – given the higher wages prevailing there and the fact that no cumulative output exists for the new technology – but superior in a newly established cluster (with no cumulative output for any technology). These conditions lead to higher productivity and higher locational benefits in the new clusters. In this sense in a world dominated by fast technological change and radical innovation, only integrated policies (targeting worker mobility and skills endowment) would give an opportunity to laggard regions. In particular, pursuing a proper mix of technologically oriented intervention and human capital formation programs would offer the laggards a chance to catch up (if not to leapfrog) leading regions (Brezis et al. 1993; Leach 1996; Desmet 2000; Suedekum 2002).
Appendix: Data Issues The County Business Patterns, produced by the US Census Bureau, provides data on establishments, employment and annual payroll by detailed industry for all counties in the United States since 1946. We decided to use data since 1988, because of earlier classification breaks raised problems of data comparability. Most recent data are available in the NAICS classification but it is possible to bridge them to earlier data (1988 to 1997), to SIC (Standard Industry Classification) to the current NAICS (North American Industrial Classification System) classification. This bridging is non fully possible, but at the degree of detail chosen, it has been proven to be safe from significant time series breaks due to errors in measurement. Other data are used in the empirical analysis, from various sources, mainly derived from the Census Bureau. In particular, from the U.S. Citizenship and Immigration Services (USCIS) data have been made available about Immigration Statistics. From the U.S. Department of Education, National Center for Education, the following proxy for human capital has been used: Population age 25 and over, percent of population age 25 and over with bachelor’s degree are provided by
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
263
Table 6 Sector classification for macrosectors in the CBP Macrosectors in the County Business Patterns NAICS Classification 11 21 22 23 31 420 44 48 51 52 53 54 55 56 61 62 71 72 81
Agriculture, forestry, fishing and hunting Mining Utilities Construction Manufacturing Wholesale trade Retail trade Transportation and warehousing Information Finance and insurance Real estate and rental and leasing Professional, scientific, and technical services Management of companies and enterprises Administrative and support and waste mgt and remediation services Educational services Health care and social assistance Arts, entertainment, and recreation Accommodation and food services Other services (except public administration)
“School Enrolment-Social and Economic Characteristics of Students” (P-20 Current Population Reports). In the empirical analysis 4233 clusters (i.e. State – industry couplets) across the time span 1988–2001 are considered. The choice is consistent with the target pursued here of investigating sector specific patterns of occupational flows. Table 2 reports the highest level of industry detail adopted. Macro-sectors were built by grouping industries according to the first two figures of the relevant NAICS codes13 and are illustrated in Table 6. From the data set, we have computed the following list of variables: Variation in Employment (dempw). The variation in employment over time is the dependent variable to be explained, representing the regional population dynamics. It is defined as the time differences of the employment mass by each industry-State couplet in logarithms (ln – ln). Number of Establishments and squared term (estD and estD2). An establishment is a single physical location at which business is conducted and/or services are provided. It is not necessarily identical with a company or enterprise, which may consist of one or more establishments. Sector heterogeneity is taken into account by adjusting variables by their within-industry mean. It mainly represents scale effects within
13 The only exception is the Wholesale trade, which is both a 2-digit and a 3-digit sector. It has been artificially created by grouping the durable and non durable goods distinction, because bridging from SIC to NAICS classification was not possible.
264
M.R. Riggi, M.A. Maggioni
industry-State and the squared term allows possible non-linearities to be controlled for. Employment Size (empsize). From the theoretical insights, employment is the main independent variable, but it cannot be included as a regressor for endogeneity issues. The solution used in this chapter consisted of instrumenting it, through the’ employment size of firms. Furthermore, these variables allow us to control for some dimension-related variables, preventing the results from being affected by different behaviour across different industry-State couplets. Interdependencies (Intra and Inter). To set up the variables Intra and Inter, two kinds of industry-State couplets are considered: the one examined and all the others. The rest of the world (USA nationwide) is supposed to affect one’s development path, if it belongs to the “competitive arena” defined by a Linda series. In this way, only the most important industries in the most important States14 are considered as being able to affect the ‘rest of the world’ development path. After this first selection, a distinction is made between industries belonging to the same State as the one under analysis, but operating in a different sector (inter-industry dependencies) and industries operating in different States (intra-industry dependencies). Intra and Inter variables, provide cooperation (with a positive sign) and competition (with a negative sign) effects of regional and sector interactions. Wages. This variable is included as the deviation in growth rate of wages in each industry-State couplet from the State average(dwageD) and accounts for the ill-effect component of the wage growth specific to each industry. Wages are included in the regression as deviations to render the performances of industry-State couplets relative to all others within the same State. How much does a sector in one State pay relative to other industries within the same State? Let be the prevailing wage in the couplet identified by State s and industry i, and be the mean of wages across States. This allows the variable, to take into account for differences in living costs across States and the level of different purchasing power. The analysis aimed at assessing the contribution to the employment mass, identifies the specific contribution of wages to growth. This factor has to be taken into account in order to correctly assess regional population dynamics, which otherwise could be blurred because of sector-wide trends and structural differences across States. furthermore, this variable is taken as a proxy for the skill component of employment growth. National Industrial Trend. How much does a sector in a region grow compared to the nationwide industry it belongs to? This effect is controlled for through this variable which allows a source of unobserved heterogeneity stemming from the industry specific skilled content of employment to be controlled for in part. This is theoretically relevant to the framework outlined but cannot be directly tested because the CBP database does not contain such information. Educational Attainment. As a proxy of human capital, the hypothesis is made that the degree of education of the State population can control for State specific human capital endowment, which is able to explain one part of the variation in wages.15 14
In terms of market share held. Educational Attainment of the Population 25 Years and Over, By State, Bachelor’s degree or more. Percent of population. Data for 1992 not being available, a computational trick was adopted. 15
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
265
International Immigration. The Immigration from outside the US to each State is assumed to control for occupational effects on wages for reasons not strictly related to workforce skill composition.16
References Almazan A, de Motta A, Titman S (2003) Firm location and the creation and utilization of human capital. NBER Working Paper Series, (WP N. 10106), November Anton JJ, Yao DA (1995) Start-ups, spin-offs, and internal projects. J Law Econ Org 11:362–378 Arrow KJ (1962) The economic implications of learning by doing. Rev Econ Stud 29(3):155–173 Baldwin RE, Forslid RP, Martin X, Ottaviano G, Nicoud FR (2003) Economic geography and public policy. Princeton University, Princeton, NJ Banerjee AV (1992) A simple model of herd behavior. Q J Econ 107(3):797–817 Bass F (1969) A new product growth model for consumer durables. Manage Sci 15:215–227 Belussi F (ed) (1988) Nuovi modelli di impresa, gerarchie organizzative ed imprese rete, FrancoAngeli, Milano Bencivenga V, Smith B (1997) Unemployment, migration, and growth. J Polit Econ 105(3):582– 608 Bentivogli C, Pagano P (1999) Regional disparities and labour mobility: the Euro-11 versus the USA. Labour 13(3):737–760 Bikhchandani S, Hirshleifer D, Welch I, (1992) A theory of fads, fashion, custom and cultural change as informational cascades. J Polit Econ 5:992–1026 Blanchard OJ, Katz LF (1992) Regional evolutions, brookings papers on economic activity, 1, 1–75.Borjas G. J. The economics of immigration. J Econ Lit 32(4): 1667–1717, December 1994 Bramanti A, Fratesi U, (2009) The dynamics of an innovation driven territorial system. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Bramanti A, Maggioni MA (eds.) (1997) La dinamica dei sistemi produttivi territoriali: teorie, tecniche, politiche. Angeli, Milano Brezis ES, Krugman PR, Tsiddon D (1993) Leapfrogging in international competition: A theory of cycles in national technological leadership. Am Econ Rev 83(5):1211–1219 Br¨ulhart M, Traeger R (2002) An account of geographic concentration patterns in Europe. Mimeo, University of Lausanne Carrol GC (eds.) (1988) Ecological models of organizations. Ballinger, Cambridge, MA Choi JP (1997) Herd behaviour, the “penguin effect”, and the suppression of informational diffusion: an analysis of informational externalities and payoff interdependency. Rand J Econ 28(3):407–425 Commander S, Kangasniemi M, Winters LA (2003) The brain drain: curse or boon? IZA Discuss Pap (809):1–34 Crescenzi R, Rodr`ıguez-Pose A (2009) Systems of innovation and regional growth in the EU: endogenous vs external innovative efforts and socioeconomic conditions. In: Fratesi U, Senn For 1992, data is available for percentage of high school attenders. In this case, Bachelor’s degree or more is proxied by the high school attenders, corrected by the difference between Bachelor’s degree or more and high school attenders for 1993, for which both data are available. Data for District of Columbia are missing. Data for 1990 are provided by the U.S. Department of Commerce, Bureau of the Census, Current Population Reports. Data for 1988 are missing. 16 Immigrants Admitted by State of intended residence, fiscal years 1989–2001. For the empirical analysis, the relative ratio to employment is used.
266
M.R. Riggi, M.A. Maggioni
L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Dahl MS, Pedersen COR, Dalum B (2003) Entry by spinoff in a high-tech cluster. DRUID Working Paper No. 11–3 David PA (1969) A contribution to the theory of diffusion, centre for research in economic growth research memorandum, 71. Stanford University, Stanford Dendrinos DS, Mullally H (1985) Urban evolution. Studies in the mathematical ecology of cities. Oxford University Press, Oxford Desmet K (2000) A perfect foresight model of regional development and skill specialization. Reg Sci Urban Econ 30:221–242 Dolado JJ, Goria A, Ichino A (1994) Immigration, human capital and growth in the host country: evidence from pooled country data. J Popul Econo 2(7):193–215 Farrell J, Saloner G (1987) Competition, compatibility and Standards: The Economics of Horses, Penguins, and Lemmings. In: Gabel HL (ed.) Product standardization and competitive strategy. Elsevier, Amsterdam, pp 1–21 Feldman M (2002) The locational dynamics of the u.s. biotech industry. knowledge externalities and the anchor hypothesis. Paper prepared for the DRUID conference Folloni G, (2009) A model of local development. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer Berlin Folloni G, Maggioni MA (1994) Un modello dinamico di crescita regionale: leader e attivit`a indotte, Quaderni della ricerca di base “Modelli di sviluppo e regional competition”, 3, Universit`a Bocconi, Milano Fratesi U, Riggi M (2007) Does migration reduce regional disparities? The role of skill-selective flows. Rev Urban Region Dev Stud 19:78–102 Frisch R, Waugh F (1933) Partial time regressions as compared with individual trends. Econometrica (1):387–401 Garavaglia C, Breschi S (2009) The co-evolution of entrepreneurship and clusters. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Garofoli G (1991) Modelli locali di sviluppo. FrancoAngeli, Milano Ghatak S, Levin P, Wheatley Price S (1996) Migration Theories and Evidence: An Assessment. J Econ Surv 10(2):159–198 Glaeser EL, Kallal HD, Scheinkman JA, Shleifer A (1992) Growth in cities. J Polit Econ 100(6):1126–1152 Grilliches Z (1957) Hybrid corn: an exploration in the economics of technological change. Econometrica 25(4):501–522 Hannan MT, Freeman J (1989) Organizational ecology. Harvard University Press, Cambridge, MA Hannan MT, Freeman J (1988) Density dependence in the growth of organizational populations.In: Carroll GR (ed.) Ecological models of organizations. Ballinger, Cambridge, pp 7–32 Henderson JV (1977) Economic theory and the cities. National Academic Press, New York Hirshleifer D (1993) The blind leading the blind: social influences, fads, and informational cascades. Working Paper of the Anderson Graduate School of Management No. 24, UCLA, Los Angeles Ireland N, Stoneman P (1986) Technological diffusion, expectations and welfare. Oxf Econ Pap 38:283–304 Isard W (1956) Location and space-economy. MIT Press, Cambridge, MA Jacobs J (1969) The economy of cities. Random House, New York Kanbur R, Rapoport H (2005) Migration selectivity and the evolution of spatial inequality. J Econ Geogr 5(1):43–57 Klepper S, Sleeper S (2002) Entry by spinoffs. Mimeo, Carnegie Mellon University, Pitsbourg Krugman PR (1991) Geography and trade. MIT, Cambridge Krugman PR (1991) Increasing returns and economic geography. J Polit Econ 99(3):483–499 Leach J, (1996) Training, migration and regional income disparities, Journal of Public Economics, 61, 429–443
Regional Growth and the Co-Evolution of Clusters: The Role of Labour Flows
267
Maggioni MA (2002a) Clustering dynamics and the location of high-tech firms. Springer-Verlag, Heidelberg, New York Maggioni MA (2002b) The development of high-tech clusters: theoretical insights and policy implications. In: Feldman M, Massard N (eds.) Institutions and systems in the geography of innovation. Kluwer, Dordrecht, pp 309–340 Maggioni MA (2006) Mors tua, vita mea? the rise and fall of innovative industrial clusters. In: Braunerhjelm P, Feldman M (a cura di) Cluster Genesis The origins and emergence of technology. Oxford University Press, Oxford, pp 219–242 Maggioni MA, Riggi MR (2008) High-tech firms and the dynamics of innovative industrial clusters. In: Karlsson C (ed.) Handbook of research on clusters: theories, policies and case studies. Edward Elgar, Aldershot Maggioni MA, Roncari S (2009) Learning, innovation and growth within interconnected clusters: an agent based approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Markusen A (1996) Sticky places in slippery space: a typology of industrial districts. Econ Geogr 72(3):293–313 Marshall A (1920) Principles of economics. Macmillan, London Martin Ph (1999) Public policies, regional inequalities and growth. J Public Econ 73(1):85–105 Nocke V (2003) A gap for me: entrepreneurs and entry. Penn Institute for Economic Research, Working Paper No. 03–019. Philadelphia, (mimeo) Paniccia I (1998) One, a hundred, thousand of industrial districts. organizational variety in local network of small and medium-sized enterprises. Org Stud 19(4):667–699 Pascal AH, McCall JJ (1980) Agglomeration economies, search costs and industrial location. J Urban Econ 8:383–388 Porter ME (1990) The competitive advantage of nations. The Free Press, New York Quirmbach HC (1986) The diffusion of new technology and the market for an innovation. Rand J Econ 17:33–47 Rauch JE (1993) Does history matter only when it matters little? the case of city-industry location. Q J Econ 108:843–867 Reinganum JF (1981) On the diffusion of new technology: a game theoretic approach. Rev Econ Stud 48:395–405 Riggi M, (2004) Labour market dynamics and the evolution of industrial clusters: towards a microfoundation of the ecological approach. PhD thesis, Universit`a di Bologna Rocco L (2003) Externalities in the economics of science. mimeo, University of Milan Bicocca, Milan Romer P (1986) Increasing returns and long-run growth. J Polit Econ 94:1002–1037 Shields GM, Shields MP (1989) The emergence of migration theory and a suggested new direction. J Econ Surv 3(4):277–304 Solow RM (1956) A contribution to the theory of economic growth. Q J Econ 70(1):65–94 Staber U (1997) An ecological perspective on entrepreneurship in industrial districts. Entrepreneurship Reg Dev 9:45–64 Suedekum J (2002) Subsidizing education in the economic periphery: another pitfall of regional policies? University of Goettingen, Mimeo Swann GMP, Prevezer M, Stout D (1998) The dynamics of industrial clustering. international comparisons in computing and biotechnology. Oxford University Press, New York, USA Vettas N (1999) Location as a signal of quality. CEPR Working Paper No. 2165 Centre for Economic Policy Research, London
Intra-National Disparities, Regional Interactions and the Growth of Countries Marco Alderighi and Marco Percoco
1 Introduction The concept of spatial interaction is certainly one of the most important research topics in the regional science field. As stated by Olsson (1970; p.223 reported by Fotheringham and O’Kelly 1989): The concept of spatial interaction is central for everyone concerned with theoretical geographical and regional science . . . Under the umbrella of spatial interaction and distance decay, it has been possible to accommodate most model work in transportation, migration, commuting, and the diffusion, as well as significant aspects of location theory.
Additionally, it is clear that spatial interaction among agents or regional systems is the key to economic development through a series of phenomena. Nijkamp and Poot (1998) point out that the neoclassical growth model can easily be extended to take account of several issues considered to be strategic in the process of regional development, such as the degree of openness, R&D activities and the related innovation capacity, human capital, inter-regional and inter-industry linkages. However, those elements can be considered as belonging to different theories and paths of research. In what follows we will consider three lines of research: 1. Inter-regional trade 2. Strategic interdependence among governments 3. Spatial dependency International trade. In the 1960s, the attention of scholars and policy makers was directed towards the issue of the role of feedback effects and the nature and extent of extra-regional influences on an economy, that returned to provide an additional stimulus. Miyazawa (1966) firstly considered the interregional trade issue as complementary to spillover effects between income groups. This insight constituted the basis of the pioneering works by Miller (1966, 1969, 1986), although a clear connection between these two approaches was not made until the paper by Sonis and Hewings (1998) was published. U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
269
270
M. Alderighi, M. Percoco
According to the traditional international trade theory, based on the HecksherOhlin model, different factor endowment and regional specialization in the most abundant local factor in conjunction with free factor mobility build regional comparative advantages, which are at the origin of trade. In the European case, greater integration has favored inter-regional dependence at the expense of intra-regional dependence. In addition, the structure of the economies is becoming similar, which is reflected by a growing dominance of intra-industry trade (indicating diversification) as opposed to inter-industry trade (specialization) (Percoco et al. 2006). In the EU, the Single Market has not led to strong specialization of European economies, but rather to specialization of regional economies within countries, depending on the geographic position of the region on the European scale, and their level of investment in technology and human capital. Fatas (1997) demonstrates that specialization in technology and quality is more obvious between EU regions than between EU countries. In addition, it seems that agglomeration forces are limited to the country where they take place: lower transaction costs and higher factor mobility within countries rather than between countries (due to cultural, language differences) can maintain regional dynamics in the form of increasing polarization/specialization. As regions within a country become less similar over time, we could expect region-specific fluctuations to increase within countries, whereas the smaller specialization of the national economies makes them less sensitive to specific shock. As a result, disparities increase within countries, but decrease between countries (Martin 1999). This is particularly relevant to our case study: in Italy the homogeneous Northern regions have returns and advantages that far outweigh those of the declining Mezzogiorno. Therefore, our primary belief is that competitive relationships of the regional economies would tend to dominate their complementary relationships. Strategic interdependence. Following the current path of research in political economics (Persson and Tabellini 2000), strategic interaction among local governments has recently become an interesting line of theoretical and empirical analysis. In this field, interaction between regions occur because of many factors: (a) spillovers,1 (b) tax competition (Wilson 1999); (c) welfare competition through migration flows from poor regions (Brueckner 2000). Most of the empirical works in this field aim at estimating reaction functions, which show how a jurisdiction responds to the choices of neighboring jurisdictions in setting the level of its own decision variable (Brueckner 2006). Spatial dependency. Somehow related to the government strategic interaction literature is the concept of spatial externalities as defined by Anselin (1988) in terms of spatial autoregressive lag and error models. In this framework, one can consider the production function of region i to be in the form (Ertur and Koch 2005; Quah 1993): yi = f (α , X, ρ , W, X−i ) 1
See Wilson (1996) for a survey on pollution abatement policies.
(1)
Intra-National Disparities, Regional Interactions and the Growth of Countries
271
where yi is production in region i, α is a vector of technological parameters, X is the matrix of inputs located in the region, ρ is a parameter governing the spatial diffusion of spillovers, with the spatial weights matrix W, of the inputs of neighboring regions, X−i that are meant to produce externalities on i. Regional and the spatial dimensions of development are at the heart of European Union policies for regional convergence. After the road leading to a common market was established, the need for a fair balance of regional economies has lead to the building of Structural Funds for interventions aimed at bringing lagging regions to prosperous development. There is now a wide consensus that European regional policies should be considered, from a theoretical economic viewpoint, as redistributive public intervention (de la Fuente 2004; Percoco 2005). On this point, while there is extensive literature on and evidence of the inefficiency of regional redistribution (see, among others, de la Fuente 2004 and Rossell`o 2003), few studies have considered how regional disparities in a given country affect national growth (Quah 1993, 1997; and, more recently, Arbia et al. 2005). A growing body of economic literature is considering the impact of inequality among individuals on the growth path of nations. In particular, Alesina and Rodrik (1994) and Persson and Tabellini (1994), while arguing partially against the Kuznets curve, find that inequality negatively affects growth rates, whilst Forbes (2000), by using panel instead of cross-sectional data, finds a positive relationship between inequality and growth. Lack of strong theoretical findings make the direction of the impact of regional disparities on growth unpredictable. In effect, whether similarities or differences foster the economic growth of countries depends on the linkage between regions and whether this linkage is positive or negative. At the moment, the literature is too fragmented and there are many competing theories that could be used to justify different theoretical results. We provide some interpretations of “classical works” to sustain the opposing conclusions theoretically. The aim of this paper is to study the impact of regional differences in per capita income on the growth rates of national economies. Our framework is based on a bottom-up approach: national growth is the sum of the regional growths. Multi-regional models are usually classified into two main groups: bottom-up or topdown.2 Contrary to a top-down approach, which assumes that redistribution among regions is a zero-sum game, we presume that regional interaction may change the result of the game (i.e., it is not necessary that one region grows when another loses).3 The key question we analyze is whether higher growth rates are more likely in countries characterized by strong regional duality (i.e., large differences in the regional per capita income). Our analysis is, first of all, an empirical investigation, as we observe that this promising issue is not sufficiently analyzed. However, in the 2
A top-down approach decomposes the national growth into regional growth without allowing for a feedback influence. Hence the impact of a regional economy on other regions was not considered. Adams and Glickman (1980), Bolton (1985), Glickman (1977), Nijkamp et al. (1986), Rey and Dev (1997) provide interesting surveys on the issue. 3 See, e.g. Alderighi (2003).
272
M. Alderighi, M. Percoco
first part of the paper we try to present a simple theoretical multi-regional model to justify the econometric exercise. From an empirical point of view, we find some evidence from EU data that regional heterogeneity fosters growth, suggesting that an unbalanced growth path may have a positive impact. The remainder of the work is organized as follows. In Sect. 2, we introduce a theoretical model which accounts for regional linkage and we provide some interpretations in the light of theoretical findings. Section 3 provides some empirical evidence from the EU countries. Finally, Sect. 4 concludes the paper.
2 A Theoretical Model We think of a country as a collection of I regions. National per capita growth rate y˙ is the weighted average of regional per capita growth, i.e., y˙ = ∑i wi y˙i where y˙i is the regional growth and wi = yi /y. We suppose that internal and external factors affecting regional growth can be grouped into three classes. The first class is composed of national and international factors (e.g., the international exchange rate, level of debt, etc.). The second class includes regional factors (infrastructure, distance from markets, etc.). Finally, the third class includes multi-regional and inter-regional feedback and spillover effects, i.e., positive and negative linkage between regions. The following equation summarizes the per capita economic growth of regioni: y˙i = β0 + β1 xN + β2xRi + β3xIi ,
(2)
where β = (β0 , β1 , β2 , β3 ) is the parameter vector, xN is a vector of national/ international factors, xRi is a vector of regional factors and xIi is a vector capturing regional interdependencies. The analysis of the impact of regional interactions on regional growth is quite complex. The literature identifies many channels which affect regional performance and may have a nationwide effects (e.g., inter-regional and international trade, labour and capital mobility, human and physical capital, availability of energy and infrastructure resources, environmental pollution levels, financial and commercial services, the location of multi-regional and multi-national firms, technology transfer and knowledge spillovers, agglomeration economies and diseconomies and demand-supply feedbacks). Some of these channels may help to understand the impact of regional interactions on national growth. The literature on trade and growth (see for example: Kelly 1997) suggests that growth may be driven by increased specialization caused by geographical expansion of the markets: “when markets move from isolation to fusion into one large, economy-wide market, it causes growth to accelerate.”4 Under this assumption, 4 Numerous empirical studies have confirmed a positive correlation between trade and income (Lewer and Van den Berg 2003). In addition to the Smithian interpretation of the relationship between trade and growth acceleration, the literature has emphasized additional factors: the exploitation of increasing returns from large markets (Balassa 1961), the transmission of international technology spillovers (Coe and Helpman 1995), the exchange of ideas through travel and communication, as well as the pro-competitive effect of international trade (Bhagwati 1965).
Intra-National Disparities, Regional Interactions and the Growth of Countries
273
opposite results emerge according to whether we expect trade to be higher between similar or different regions. In favor of the first thesis there are the “classical” theories of commerce (as Heckscher-Ohlin) which hold that trade raises when there are differences in factor endowment. On the contrary, according to intra-industry trade literature, trade between regions with similar levels of technology and income should be higher. Endogenous growth theory emphasizes that knowledge spillovers and multiregional and inter-regional linkage are crucial for the growth of regional and national economies. In fact, spillovers lead to dynamic externalities and agglomeration effects5 which may provoke faster growth. Like the previous example, however, it is possible to have two competing theses. On one hand, the “homogeneity” argument suggests that regional spillovers are more intensive and growth is faster when regions are similar. In fact, similar regions produce and use similar technologies, which lead to intensive inter-regional crossfertilization (Acemoglu and Zilibotti 1999; Quah 1999). On the other hand, strong differences between regions may favor spillovers and growth through technological and knowledge transfer. In the latter thesis, it is argued that regional spillovers are strong when regions differ because of imitation and leap-frogging investment. In the same vein, catching-up models (for a review, see Badinger and Tondl 2005) suggest that initial differences can foster growth; or in innovation literature, Giovannetti (2000) shows that innovation necessary produces regional disparities. Hence, regional disparities can both foster or hinder regional and national growth. Finally, a standard two-region Solow model can be used to explain the link between regional differences and national growth. Denote the capital endowment and total output of region i = 1, 2 by Ki and Yi , respectively, and denote per unit of output capital endowment by ki = Ki /Yi . As previously mentioned, variables without indices refers to the country. Let the weight of the region in the total economy be wi = 1 − w j = Yi /Y . In addition, assume that there is no capital depreciation or population growth. Standard results (Barro and Sala-i-Martin 2003) state that, when production function f is in a Cobb-Douglas form, the regional growth rate is given by: y˙i = s f (ki ), where s is the saving rate, f is the marginal productivity of capital and, by definition, y˙ = wi y˙i . Now, we ' (notice that y˙ = s (w1 f (k1 ) + (1 − w1) f (k2 )) > s f (w1 k1 + (1 − w1 ) k2 ) = s f k¯ where k¯ is the average capital endowment (Fig. 1). Previous inequality comes from the standard assumptions on the concavity of the production function, meaning that we expect higher growth rates when regions have different capital endowments (and different output levels) with respect to a case in which the same total capital is equally distributed. Summarizing previous results we observe that there are two competing theses and at the moment it is only an empirical matter whether one be preferred to the other. We will investigate this later in the paper.
5
The importance of knowledge spillovers has been emphasized by Glaeser et al. (1992), Griliches (1992), Smolny (2000), Fritsch and Franke (2000).
274
M. Alderighi, M. Percoco
. y s . f ⬘(k )
s . f ⬘( k1)
s
2
Σw f ⬘( k ) i= 1
i
i
sf ⬘( k )
s . f ⬘( k2)
k1
k
k2
k
Fig. 1 Growth rates in regions with different capital endowment
Now, we assume that regional interactions depend upon regional differences: ) ) xIi = f (y1 , ..yI ) = ϕ (I) ∑ )yi − y j ) , (3) j
where ϕ (I) is function of the number of regions. Note that xIi is a “measure of our ignorance” in the sense that, because of the lack of data at regional level, we do not include the channels previously sketched but we assume only that these channels are affected by differences or similarities between regions. Replacing (2) in (1) and averaging regional growth into country growth, we obtain: ) ) (4) y˙ = β0 + β1xN + β2 xR + β3 · ϕ (I) ∑ wi )yi − y j ) i, j
) ) where xR = ∑i wi xRi and ∆ = ∑i, j wi )yi − y j ) is the average difference. Gini (1914) showed that the so called Gini Index G is given by ∆/ max ∆. If we choose ϕ (I) = 1/ max∆, it follows that the national growth equation can be written as: y˙ = β0 + β1 xN + β2 xR + β3G
(5)
Equation (4) suggests two important considerations. First, when we focus on national growth rate, national and regional variables are indistinguishable as of
Intra-National Disparities, Regional Interactions and the Growth of Countries
275
linearity in equation (1). Thus, regional determinants of growth can be approximated (prima facie) by the average national value without distorting the estimates. This result is quite important from a regional point of view as we can derive national growth equation as a sum of regional growth. Second, when determinants of regional growth include regional interactions measured by average differences in levels, it can be summarized by a Gini index on regional disparities.
3 Some Empirical Evidence In this section we report some empirical evidence on European Union growth for a cross section of countries over the period 1980–1999. Countries in our data set are Austria, Belgium, France, Finland, Germany, Greece, Ireland, Italy, the Netherlands, Portugal, Spain, Sweden, UK. We have excluded Denmark and Luxembourg because of the impossibility of computing a Gini coefficient as those countries do not have regional repartition of their territories. Data are from the Euregio Dataset provided by EUROSTAT and from the World Development Indicators collected by the World Bank, while data on infrastructure stocks are taken from Calderon and Serv´em (2004). The variables are described in Table 1, while Table 2 reports some summary statistics for our variables. Our empirical strategy consists in estimating the following regression: y˙it = const. + α Ginii,t−1 + γ Xi + ε
(6)
where y˙ is the growth rate of country i between time t and t − 1, Ginii,t−1 is the Gini index for country i at time t − 1, aiming at providing information on the GDP distribution among regions. It should be noted that the use of the Gini index is particularly suitable in our case as a measure of production dispersion as it is highly
Table 1 Description of variables Acronym
Variable description
Source
Gini
The index has been calculated at national level on regional per capita GDP Per capita GDP Private investment rate in fixed capital Public expenditure in education in percentage of GDP Government consumption as a share of GDP Kilometers of paved roads Total population Number of telephone mainlines Growth of exports
The source of regional production is EUROSTAT, REGIO
GDPpc Investment EduExp GovCons Road Pop Tcom ExportGrowth
EUROSTAT World development indicators World development indicators World development indicators Calderon and Serv´em (2004) World development indicators Calderon and Serv´em (2004) World development indicators
276
M. Alderighi, M. Percoco
Table 2 Summary statistics
Mean Median Maximum Minimum Std. Dev.
EduSpend
GrowthPC
ExpGrowth
5.289208 5.255000 8.630000 1.990000 1.427381
2.120689 2.147985 9.835993 −6.779953 2.227978
5.60609 5.25847 21.03666 −16.46386 4.88258
GovCons
Investment
2.172565 2.97558 2.128689 3.33282 9.646518 37.13331 −5.499539 −19.58401 2.104858 7.23425
GDP
Road
TCOM
GINI
383786 182913 2064742 364 429396
221821.6 128001.0 893500.0 5000.5 222978.7
10268787 5036612 48210000 157112 10856954
0.101557 0.106071 0.172237 0.039426 0.033921
Table 3 Estimation results Model: Constant Ginit−1 Log(GDPpc)t−1 Investment EduExp∗ GovCons
1
2
3
4
0.009 (0.078) 0.0557 (2.208)a −0.00073 (−0.259) 0.0024 (15.794)c 0.00023 (2.330)a
0.0012 (0.080) 0.062 (2.241)a −0.008 (−1.404) 0.0024 (15.425)c 0.00017 (1.6698)a 0.0065 (2.042)a 0.00092 (0.164)
−0.0133 (−1.243) 0.0347 (1.603)b −0.002 (−1.088) 0.002 (16.997)c 0.000346 (4.030)c
0.038 (0.424) 0.400 (1.887)b 0.027 (0.739) 0.003 (2.570)c 0.011 (1.781)b 0.027 (0.739)
Log(Road/Pop)t−1 Log(Tcom/Pop)t−1 Export Growth R2 N. Obs. OTHER TESTS
0.655 156
0.646 156
0.0015 (8.764)c 0.776 156
0.386 124
Notes: t-statistics are in brackets; a 95% significance level; b 90% significance level; c 99% significance level. Models 1–3 estimated by GLS with random effect. Model 4 was estimated by GMM-IV. Instrumental variables are: Log(Road/Pop)t−2 , Log(Tcom/Pop)t−1 , Eduspend t−1 , Log(GDPpc)t−2 , Ginit−2
robust to extreme values (Cowell and Flachaire 2007). Finally, Xi is a set of control variables at time t or t − 1. To estimate equation (6) we use standard least squares estimator with random effects.6 In Table 3 we report some empirical evidence from our regression analysis. In our regressions, the estimated coefficient for the per capita GDP at time t − 1 is slightly negative, although not statistically significant, implying the presence of a weak convergence process across countries.
6
We have run similar regressions with fixed effects but results were very unsatisfactory. This could be due to the fact that any significant nation-specific effect does affect growth pattern.
Intra-National Disparities, Regional Interactions and the Growth of Countries
277
As expected, private investment as well as government expenditure on education (as interacted with total public consumption) are positive and highly significant. Lagged values of infrastructure stocks a have positive impact on national growth rates, although telecommunication infrastructure is not significant. This could be surprising if we compare this result with an actual situation, however it is worth noting that our time period spans 1970–1999, thus largely before the “ICT Revolution” (Alderighi et al. 2005). Finally, in Model 3 we consider the export growth and find that it is significantly related with growth. Notice that this variable becomes insignificant when we add the infrastructure stocks into the regression. This could be due to there being some uncertain correlation among those variables leading to unclear results from the viewpoint of competitive advantage trade theory. The parameter associated with the Gini index is positive and significant across all models. This, in turn, means that the higher the polarization of production, the higher the growth rate of the national economy. This result seems to confirm the predictions of the multi-regional Solow model, as sketched in the previous section. Additionally, we could not find any evidence of a negative correlation between intra-national disparities and country growth, hence the arguments in Alesina and Rodrik (1994) and Persson and Tabellini (1994) cannot be extended into the regional context. However, in our analysis we have not considered any interaction between regions (which might change some of the predictions of our theoretical framework if it were present). In what follows, we run a robustness analysis of our main result in the case that the path of national growth is influenced by regional interaction. In particular, our strategy consists in showing that our analysis of the correlation between the Gini index and growth is robust to the presence of spatial interactions. Let us now turn to discuss our results in the light of regional interaction. According to what was stated in the Introduction and in equation (1), we should define a sort of measure of economic distance (Conley and Lingon 2002) to explain the way regions interact. This, in turn, would mean that the Gini coefficient should be multidimensional, given the different inputs in the production function. Put it in a different way, we could consider a Gini coefficient calculated over the fitted values of (1l) to take into account interaction between regions, although we are making the strong assumption that all inputs are mutual substitutes. In other terms, we assume that regional interactions are fully captured by the predicted value of production. We can rewrite equation (1) as: yi = y*i + ei , where y*i is the fitted value and ei is the residual. Residuals distribute according to N(0, σe2i ). The Gini coefficient can thus be written as: ) ) yi + ei ) − (* y j + e j )) ∑i ∑ j )(* (7) Gini = 2n2 µ rearranging terms and by simplifying the notation we have: Gini = Gy* + δ
(8)
where Gy* is the Gini index computed over the fitted values of the production functions and δ is an error term.
278
M. Alderighi, M. Percoco
Given equation (8), when considering regional interactions, we need to estimate the following regression: ' ( y˙it = α + α Gy* i,t−1 + γ Xi + ε However, Gy* = Gini − δ , hence we can rewrite the previous equation as: y˙it = const. + α (Gini − δ )i,t−1 + γ Xi + ε
(9)
By grouping the Gini coefficient and the error term ε , we have: y˙it = const. + α Ginii,t−1 + γ Xi + ϖ
(10)
where ϖ = ε − αδ is assumed to be distributed according to a Normal distribution. It is clear that in equation (10), the covariance between the Gini coefficient and the error term is different from zero, E [Gini, ϖ ] = 0, because of the presence of δ in the structure of errors. Thus standard least squares estimator is no longer consistent and, mostly, it is biased. To overcome this problem, we make use of GMM-IV regression, as reported in Model 4 of Table 3. We find that most of the results are preserved also in the case of regional interaction, although road variable is no longer significant.7 The parameter for the Gini coefficient is still largely positive, implying that, also in the case of incorrect specification of that variable, due perhaps to spatial spillovers, the main result still holds.
4 Conclusions As argued by Giovannetti (2000), the temporal diversification in the adoption of new technologies causes production polarization among regions and the formation of clusters. Once regional asymmetries are in place they can be considered as an engine for national growth. However, the reduction of regional disparities is the main objective of European Union policies through Structural Fund intervention. In this paper we have directly addressed this issue and found that there is a positive link between regional disparities and national performance. Future research can be directed towards a more correct characterization of the structural relationship by exploiting other forms of regional interactions, such as interregional migration and trade, location factors and resources. Shortages. Our analysis does not necessarily imply the elimination of European Union spending for regional policies, rather it indicates that expenditure should be directed towards sectoral policies in order to sustain specific regional growth path, i.e., it should rely on the comparative advantage of regions. 7
It should be noted that parameters associated with variables other than Gini are biased.
Intra-National Disparities, Regional Interactions and the Growth of Countries
279
References Acemoglu D, Zilibotti F (1999) Information accumulation in development countries. J Econ Growth 4:5–38 Adams FG, Glickman NJ (eds) (1980) Modeling the multiregional economic system. D.C. Heath and Company, Lexington, MA Alderighi M (2003) Investment trajectories in domestic and international infrastructure, WP Certet, 1–03 Alderighi M, Biagi F, Riggi M (2005) Growth, information and communication technologies and regional disparities: a review of theory and evidence, mimeo Alesina A, Rodrik D (1994) Distributive politics and economic growth. Q J Econ 109(2):465–490 Anselin L (1988) Spatial econometrics. Kluwer, Amsterdam Arbia G, Basile R, Piras G (2005) Analyzing intra-distribution dynamics: a reappraisal. University of Illinois at Urbana-Champaign, REAL Technical paper 05-T-11 Badinger H, Tondl G (2005). The factors behind European regional growth: trade, human capital and innovation. Rev Reg Res 25:67–89 Barro RJ, Sala-i-Martin X (2003). Economic growth, 2nd edn. MIT Press, Cambridge, MA Bolton R (1985) “Regional econometric models”. J Reg Sci 25(4):495–520 Bhagwati JN (1965) On the equivalence of tariffs and quotas. In: Baldwin RE, et al. (eds) Trade, growth, and the balance of payments: essays in honor of gottfried haberler. North Holland, Amsterdam, pp. 53–67 Brueckner JK (2000) Welfare reform and the race to the bottom: theory and evidence. South Econ J 66(1):505–525 Brueckner J (2006) Strategic interaction among governments. In: Arnott R, McMillen D (eds) Companion to urban economics. Blackwell, Oxford Calderon C, Serv´em L (2004) The effect of infrastructure development on growth and income distribution, The World Bank, mimeo Coe DT, Helpman E (1995) International R&D spillovers. Eur Econ Rev 39(5):859–887 Conley TG, Lingon E (2002) Economic distance and cross-country spillovers. J Econ Growth 7(2):157–187 Cowell, F, Flachaire E (2007) Income distribution and inequality measurement: the problem of extreme values. J Econom forthcoming de la Fuente A (2004) Second-best redistribution through public investment: a characterization, an empirical test and an application to the case of Spain. Reg Sci Urban Econ 34(5):489–503 Ertur C, Koch W (2005) Growth, technological interdependence and spatial externalities: theory and evidence. University of Illinois at Urbana-Champaign, REAL Technical Paper 05-T-10 Fatas A (1997) EMU: countries or regions? Lessons from the EMS experience. Eur Econ Rev 41(3–5):743–751 Forbes KJ (2000) A reassessment of the relationship between inequality and growth. Am Econ Rev 90(4):869–887 Fotheringham AS, O’Kelly ME (1989) Spatial interaction models: formulations and applications. Kluwer, Amsterdam Fritsch M, Franke G (2000) Innovation, regional knowledge spillovers and R&D cooperation. Universit¨at Freiberg, Working Paper, Freiberg Giovannetti E (2000) Technology adoption and the emergence of regional asymmetries. J Ind Econ 68:71–102 Glaeser EL, Kallal HD, Scheinkman JA, Shleifer A (1992) Growth in cities. J Polit Econ 100:1126– 1152 Griliches Z (1992) The search for R&D spillovers. Scand J Econ 94:29–47 Glickman N (1977) Econometric analysis of regional systems. Academic Press, New York Kelly M (1997) The dynamics of Smithian growth. Q J Econ 112(3):939–964 Lewer JJ, Van den Berg H (2003) How large is international trade’s effect on economic growth. J Econ Surv 17(3):363–396
280
M. Alderighi, M. Percoco
Martin P (1999) Public policies, regional inequalities and growth. J Public Econ 73(1):85–105 Miller RE (1966) Interregional feedbacks in input-output models: some preliminary results. Pap Reg Sci Assoc 17:105–125 Miller RE (1969) Interregional feedbacks in input-output models: some experimental results. West Econ J 7:57–70 Miller RE (1986). Upper bounds on the sizes of interregional feedbacks in multiregional inputoutput models. J Reg Sci 26:285–306 Miyazawa K (1966) “Internal and external matrix multipliers in the input-output model.” Hitotsubashi J Econ 7:38–55 Nijkamp P, Poot J (1998) Spatial perspectives on new theories of economic growth. Ann Reg Sci 32(1):7–37 Nijkamp P, Rietveld P, Snickars K (1986) Regional and multiregional economic models: a survey. In: Nijkamp P (ed) Handbook of regional and urban economics. North-Holland, Amsterdam, pp. 257–294 Olsson G (1970) Explanation, prediction, and meaning variance: an assessment of distance interaction models. Econ Geogr 46:223–233 Percoco M (2005) The impact of structural funds on the Italian Mezzogiorno, 1994–1999, Region and Development, n. 21 Percoco M, Dall’erba SG, Hewings, JHD (2006) Structural convergence of the national economies of Europe. In: Van Geenhuizen M, Reggiani A, et Rietveld P (eds) Policy analysis of networks and regions. Ashgate, Aldershot, UK Persson T, Tabellini G (1994) Is inequality harmful for growth? Am Econ Rev 84:600–621 Persson T, Tabellini G (2000) Political economics. MIT Press, Cambridge, MA Quah D (1993) “Galton’s fallacy and tests of the convergence hypothesis”. Scand J Econ 95:427– 443 Quah D (1997) Empirics for growth and distribution: stratification, polarization, and convergence clubs. J Econ Growth 2:27–59 Rossell`o J (2003) Regional redistribution and growth. Investigaciones Economicas 27(2):369–392 Smolny W (2000) Endogenous innovations and knowledge spillovers. Springer, Heidelberg/ New York Sonis M, Hewings GJHD (1998) Economic complexity as network complication: multiregional input-output structural path analysis. Ann Reg Sci 32(3):407–436 Rey SJ, Dev B (1997) Integrating econometric and input-output models in a multiregional context. Growth Change 28:222–243 Wilson JD (1996) Capital mobility and environmental standards: is there a theoretical basis for the race to the bottom? In: Bhagwati J, Hundee R (eds) Fair trade and harmonization: prerequisites for free trade? vol 1. MIT Press, Cambridge, MA Wilson JD (1999) Theories of tax competition. Natl Tax J 52(2):269–304
Creativity, Cultural Investment and Local Development: A New Theoretical Framework for Endogenous Growth Pier Luigi Sacco and Giovanna Segre
1 Introduction Creativity may be seen as one, albeit crucial, instance of a deeper process of reorientation of production processes toward intangible forms of added value, a point that has been made several times in the recent literature. In industrialised countries, an increasing number of goods and services incorporate an essential, intangible added value deriving from design, aesthetics, and symbolic and identity values: the key elements of state-of-the-art competition. When competition cannot take place through costs cutting, product innovation represents the distinctive successful factor which can be obtained though a massive employment of applied creativity, craftsmanship and technological transfer. Within this framework, by taking a closer look at the organizational features of economic activities, we provide a theoretical account of the new epoch of local development processes. In particular, we focus our analysis on the emergence of new forms of horizontally-, rather than vertically-integrated clusters of economic activities, where an important role is played by the strategic complementarity between artistic production and firm production. This is the basis of a new vision of the functioning of cultural districts, where the creative value chain, starting from the cultural and artistic dimension, drives economic systems in the field of applied research and creative production. Within this context, the pure cultural artistic dimension of the district, and the creativity diffusion process which arises from it, represent the key explaining factors of culture-led economic development. Recent literature which explicitly considers the role of culture in fostering economic development, lacks a deep analysis of the causal links between all the important factors involved. All - and sometimes not even all – the important factors have generally been grouped together in a “black box”, without any attempt to identify the causal relationship between them.
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
281
282
P.L. Sacco, G. Segre
2 Core Vs. Non-Core Creativity in Culture-Led Local Development In the past few years there has been a stream of new literature on the issues of creative cities and culture-led local development (Miles and Paddison 2005 for a critical review). In particular, books by Richard Florida (2002, 2005) have sparked a considerable amount of interest in the potential of local policies aimed at attracting talented creative workers, high-tech firms, creative minorities, and so on. However, the strong focus that such literature has developed toward some specific dimensions (as summarized in Florida’s creativity index) carries the risk of rendering such development processes too mechanically (Nathan 2005; Markusen 2006, for a detailed critique of Florida’s approach). In particular, the weakness of Florida’s theory is twofold. On the one hand, the lack of a fully-fledged theoretical analysis of the deep causal links that make creativity so important in the current economic scenario, making it difficult to assess what the critical conditions are that determine whether or not a given local system is successful (Comunian and Sacco 2006). On the other hand, an excessively broad definition of the creative class, in that it includes almost every professional profile that is characterized by a relatively high level of human capital. Critical discussion of Florida’s theory therefore points to the need for narrowing the focus on creativity, concentrating on its core dimensions, and discussing its relationship with cultural activities. To this end, it is important to discuss definitions of culture, as well as the issue of how to locate the borders of the cultural sectors, broadly interpreted. One definition of “culture” given by Throsby (2001) refers to the set of attitudes, practices and beliefs that are fundamental to the functioning of different societies and groups defined in geographical, political, religious, or ethnical terms. Culture thus finds its expression in a particular society’s values and customs, which evolve over time as they are transmitted from one generation to the next. Accordingly, culture is both tangible and intangible. The stock of tangible cultural capital assets consists of buildings, structures, sites and locations endowed with cultural significance (called “cultural heritage”) and artworks and artifacts existing as private goods, such as paintings, sculptures, and other objects. Intangible cultural capital includes the set of ideas, practices, beliefs, traditions and values which serve to identify and bind a given group of people together, however the group may be determined, together with the stock of artwork existing as public goods in the public domain, such as certain instances of literature and music. But then, who works in the creative sector? What do creative workers do? It is certainly a misconception that the creative sector workforce comprises only artists and culture-based occupations. If one takes as the real criterion for discrimination the amount of innovative problem solving that is normally undertaken by carrying out one’s professional tasks, a definition as broadly inclusive as Florida’s is relatively justified. One can easily argue that, whereas members of the non-creative goods and service fields are paid mainly to work according to pre-existing plans, members of the creative fields are paid mainly to create such plans, are engaged
Creativity, Cultural Investment and Local Development
283
in complex problem-solving that is characterised by a great deal of independent judgement and requires high levels of education or intellectual capital. Therefore, anyone who makes use of sophisticated intellectual abilities is fully entitled to be part of the creative field. Accordingly, one must think of an extensive definition of the creative sectors according to the “TAPES” formulation, based on the “TAPE” scheme of Stolarick et al. (2005): T = Technology and Innovation A = Arts and Culture P = Professional and Managerial E = Education and Training S = Science and Basic Research Each of these sectors is subject to the “3T” (talent, technology, tolerance) characterization of Richard Florida. But nevertheless, cultural and scientific activities, more strictly defined, play a crucial role in this picture. Whereas in the other sectors creativity is instrumental to other ends (i.e. involves extrinsic motivation and is seen not as an end in itself but as an asset that may be useful in solving functional problems), proper artistic or scientific activities attract people who have intrinsic motivation for creative thinking and apply it to non-functional problems. In this sense, artistic production and pure scientific research make up the deep core of the creative sector. The more the balance between functional and non-functional thinking shifts toward the former, the closer we are to the outer layers. We can therefore envision a spectrum of activities moving from the core to the periphery, that range from pure arts/science through applied arts/science, to aesthetic/technological transfer activities, in which the functional element becomes dominant although a relatively high level of creative thinking may still be involved. Whereas the technological transfer domain is very well known and studied nowadays, there is not yet enough recognition of the fact that all those fields where aesthetic ideas conceived in basic artistic production are creatively digested and transformed, are instances of aesthetic transfer processes: from fashion to communication, design to entertainment, to give just a few examples. By putting both artistic and scientific activities under the common ‘culture’ label, we can therefore define the cultural value chain as a passage from the ‘basic/pure’ through the ‘applied’ to the ‘transfer’ step, with an implicit parallel between the artistic and the scientific dimension. To clarify the process we have in mind, the following figure can be drawn (Fig. 1).
Super-core creativity activities
Core creativity activities
Fig. 1 The cultural value chain
Creative products
Functional consumption products
decreasing degree of creative content
284
P.L. Sacco, G. Segre
The notion of super-core creativity denotes the activities directly deriving from the pursuit of non-instrumental cultural purposes, such as new creative expression, cultural experiments, ground-breaking artistic creation. It can be thought of as the laboratory from which new cultural paradigms emerge. Given this perspective, basic scientific research is included in the notion of super-core creativity. At a lower cultural innovation potential, core creativity defines all the activities with substantial cultural content, such as those deriving from the operation of cultural industries on the art side, and, those connected with applied scientific research on the science side. Super-core creativity can then be regarded as the fuel that feeds not only the development of the cultural sector but also the progress of applied research; it is the intangible asset which ultimately enables the expansion of the stock of cultural and research goods. Within this context, the idea of creativity supercedes the Romantic view of creative genius, once regarded as a gift from the gods, i.e. something extraneous to the determination of the surrounding social context. Conversely, creativity is now regarded as a collective process, which evolves through the systematic modification of tastes, habits of mind and perceptions, whereas major creative accomplishments by individuals and teams are regarded as emergent phenomena that provide a more effective synthesis of ongoing tendencies, ideas and frames of mind. Creative products originate directly from this humus. Most goods and services that are produced and consumed in advanced economies carry an essential, intangible added value derived from design, aesthetic, symbolic and identity values. These commodities are the endpoint of the creative value chain, and (as already emphasised) they are generated though massive employment of applied art, craft and technological transfer activities. A wide range of commodities is thus obtained, from the quasi-artistic dimensions of gourmet food and fashion, which incorporate a considerable amount of (applied) creative added value, through down-to-earth goods and services in which the creative dimension is not outstanding but which are characterised by indicators that derive from the innovation and creativity orientation of the production environment, as is the case for several ‘made in Italy’ products, to products which are entirely functional consumer goods with no creative dimension whatsoever. This classification may not be unanimously accepted or may only partially overlap existing ones. For instance, Forte and Mantovani (2005), identify the following classes of “core” cultural activities: visual arts; cultural knowledge and testimony (archaeology, heritage, history, scientific, technical and humanist thought); civil, political, social information (media of all sorts); artistic and cultural entertainment (music, cinema, performing arts, literature); pure entertainment. In this case, for instance, we find, within the same category, activities that are characterised by widely varying degrees of functional orientation. Our definition is, by the way, more restrictive than Florida’s definition of ‘core’ creativity, which is much closer in principle to that of Forte and Mantovani. The distinction between the core and non-core components of the creative class is crucial to understanding their differential impact on and implications for the process of culture-led local development.
Creativity, Cultural Investment and Local Development
285
3 Creativity, Knowledge Spillovers and Local Clusters of Activities Throughout much of the twentieth century, creativity was totally ignored by economists and cultural aspects were largely ignored. However, by analysing the usual key factors of economic theories which attempted to explain economic growth, we can identify a specific role played by cultural and creative factors. Starting from the seminal work of Josef Schumpeter (1911) on innovation, the economic literature focus on the importance of new knowledge and technological change arising from innovation and knowledge spillovers immediately directed attention to the fundamental role played by information and its diffusion. Moreover, the studies on endogenous growth, initiated by Romer (1986) and Lucas (1988), introduced a new perspective, which explicitly recognized the role of human capital, made up of education and skills, and knowledge capital. The role of intangible assets was then accepted in economics. Nevertheless, the human act of producing creative thoughts was always considered an exogenous variable. Within this context, the importance of the proximity of individuals emerges. By allowing the knowledge of one individual to spill over onto others, the productivity of the others is improved in a virtuous circle.1 Furthermore, the widespread diffusion of knowl edge derived from knowledge spillovers, enhanced productivity not only among individuals working within the same sector, but also across different and sometimes apparently very distant sectors, creating a process of cross-fertilization. The value of industrial diversity in generating innovation in the economy has been well documented (Maignan et al. 2003). However, these results must be extended to include more broadly defined creative diversity, since connections at all levels in a rich, creativity-based economy can generate innovation. A relatively high density of creative-class employment generates the conditions under which positive interaction between individuals is more likely to occur spontaneously. These connections take place among artistic, design, technology and business sectors. Our contention is that technology alone cannot lead to sustained innovation in the long run: the role of ‘super-core’ and ‘core’ cultural activities is needed if innovation is to be socially sustainable. A particular role for cities and clusters of activities then arises, as acknowledged by the recent upsurge in new literature on the issues of creative cities and cultureled local development (Landry 2000; Scott 2000; Caves 2001; Florida 2002; Florida and Tinagli 2004). The main focus of Landry’s (2000) analysis, for instance, is the idea that people are the crucial resource of cities. Human cleverness, desire, motivation, imaginations and creativity are the driving forces, replacing location, natural resources and market access. Creativity and innovation - or the lack of them - are critical, especially when cities face a period of transition. This is the case in today’s world economy, mainly as a consequence of the increasing globalization, which produces growth of cities in some areas e.g. Asia, while in others e.g. Europe, a shortage 1
The existence of “localized knowledge spillovers” is discussed by Garavaglia and Breschi in this book.
286
P.L. Sacco, G. Segre
of opportunities for cities in the field of old industries and an expansion of the possibility of the application of intellectual capital to product, process and services. However, little is known about the conditions necessary for creativity and innovation to emerge. What is evident in modern economic systems, according to this view, is that place, culture and economy are symbiotic, and this symbiosis is significantly influencing the expansion of certain important cities (as documented by the cases of cities such as Austin (Texas), Denver (Colorado) or Linz (Austria); see Sacco and Pedrini 2003). Accordingly, Scott (2000) argues that the presence of skilled and socialized cultural workers is not a sufficient condition to obtain efficacious patterns of productive employment. It is not only the usual concept of agglomerations of technologically dynamic firms that generate developments in a regional system, but also the existence of qualities such as cultural insight, imagination, and originality created within the local system of production. Creative and innovative energies should be considered endogenous properties of a production system. The production system and the geographic milieu are therefore just two aspects of a single economic and cultural situation represented by dense agglomerated structures of employment and social life. In an effort to prepare themselves for the twenty first century, many communities focused on adjusting to the needs of an age in which information is the most valuable commodity. Their strategy was to encourage investment in human and financial resources to prepare individuals to meet the challenges of the rapidly evolving post-industrial, knowledge-based economy and society. At the heart of this effort is the identification of the vital linkage between art, culture and economic systems: the diffusion of knowledge is greatly influenced by cultural production, which originates in socially and economically embedded creative processes. Local development arises in a new urban landscape made by powerful regional economies based on the city, where creativity and cultural production play an essential role in sustaining economic growth. Many examples confirm such an opinion. According to Stolarick et al. (2005), the recent growth of the Montr´eal region is partially driven by the convergence of art- and technology-based entrepreneurs from new fields such as entertainment technology, voice recognition, genomics, novel materials, and nanotechnology. In Montr´eal the number of cultural events per square kilometer is remarkably high and the bilingual and multicultural setting offers a rich and original set of interconnections. Artists work directly with manufacturers to develop new materials and techniques for their artistic purposes. By, reinvesting 40% of its profits in a “creative think tank” the Cirque du Soleil with its R&D division contributes to the development of new technology as much as any traditional high-tech company. Ex-Centris produces screens to provide a high definition digital movie theatre experience at affordable prices, thus considerably improving the access that independent film-makers have to movie theatres. A new sound of a “Techno” band can become part of a video game, as well as a new color combination presented in an art gallery can make its way into a graphic designer’s product. In Montr´eal, industrial designers are widely employed, and toys designed for Canadian, U.S. and European markets are exported from the city. A positive impact of design on corporate financial performance was empirically identified by Hertenstein et al. (2005) for the period 1995–2001. The relationship
Creativity, Cultural Investment and Local Development
287
between effective industrial design and’ financial performance - investigated by measuring traditional financial ratios, such as return on sales, return on assets, and total stock market returns - highlights a significant difference between firms which are more effective at using good industrial design and firms that are less so. Firms with more effective industrial design had higher returns on sales and higher returns on assets. Moreover, trends in total stock market returns measured by the Standard&Poors 500 Index are not only higher for the highly effective industrial design group of firms, but the difference between the two groups’ returns also increased over the sample period. The mainstream of world urban development further highlights how culture is a source of prosperity and cosmopolitanism through international events and centers of excellence, and drives high-growth business sectors such as creative industries, commercial leisure and tourism, and increases the profile and name recognition of an urban centre. A high concentration of higher education and research institutes, moreover, as well as a renewed tendency to city centre living, accessible business-incubator and start-up facilities, and sophisticated consumer/ producer inter-relationships within a context of permeable networks and major cultural institutions are, when present collectively, are factors which enhance a powerful dynamic. Cities such as London, Los Angeles, New York, Paris and Tokyo are certainly global cultural players; however, smaller cities can increasingly earn visibility and economic development if they recognise the potential of cultural spillovers (Comedia 2002). The development of cities can be dramatically influenced by the organization of major events such as sports competitions2 (The Olympic Games, World Championships and World Cups) and cultural gatherings (Expo, Universal Exhibitions, International Forums, Culture Capital Programs). This is confirmed by a comparative study of cities such as Berlin, Paris, Seoul, Sydney, Toronto, Barcelona, Melbourne, Rio de Janeiro, Seville and Shenyang, hosting different events3 (Metropolis 2002). Following Sharpe et al. (2004) an important distinction should be made between impacts arising from non-cultural-producers accessing a cultural facility and impacts arising from cultural producers themselves using a cultural facility to do their work. This distinction is particularly significant when public policy is under dis2
Note that usually the sport competitions also have an essentially cultural dimension and there are a host of artistic events such as parades, exhibitions, concerts, etc. to go with them. 3 Three - Berlin, Paris, Seoul - are the political capitals of their countries: Berlin is considered because of the bid for the 2000 Olympic Games and the 2006 Football World Cup; Paris is considered because of 1998 Football World Cup; Seoul is considered because of 1988 Olympic Games and 2002 Football World Cup. Two - Sydney and Toronto - are the economic capitals of federal states: Sydney is considered because of the 2000 Olympic Games and Toronto is considered because of the World Youth Days in 2000. Three - Barcelona, Melbourne, Rio de Janeiro - are cities challenging and rivaling the main cities in their country: Barcelona hosted the 1992 Olympic Games and The Universal Forum of Cultures in 2004; Melbourne is considered because of the 2006 Commonwealth Games; Rio de Janeiro for the annual Carnival. Finally, two of them - Seville and Shenyang - are rather “outsiders”, but represent cities that occupy an important place inside their countries: Seville is considered because of the 1992 Universal Exhibition and for the 1999 World Athletics Championship; Shenyang (China) is considered because of the International Amities Event Month in 1999.
288
P.L. Sacco, G. Segre
cussion, since the direction and intensity of the spillover effects of investment in cultural facilities can be different. In general, however, all levels of government are involved in all the dimensions of the impacts of cultural facilities: social, cultural, and economic. In conclusion, the arts strongly influence technological development and innovation, and technology and innovation change the arts in a perpetual virtuous circle.
4 Genius Loci Place and culture are persistently intertwined. Any given place is always a locus of dense human interrelationships - out of which culture partly grows - and culture is a phenomenon that tends to have intensely local characteristics which help to differentiate places from one another (Scott 2000). However, a deepening tension between local and global is evident: culture is both narrowly place-bound and source of non-place globalised events and experiences. The production of culture tends to become more and more concentrated in a privileged set of localized clusters of firms and workers, while final output are channelled into ever more spatially extended networks of consumption. A question remains open. Which forces in the on-going globalization process tend to preserve cultural diversity, and which forces tend to induce cultural convergence? Several mechanisms may be at work here. Increased trade integration with other countries may affect the evolution over time of preferences and cultural traits of individuals, and lead to convergence of preferences in the participating countries. In the case of cluster formation, as pointed out by Garavaglia and Breschi (2009, in this book), the cultural homogeneity arising in the district can limit the exploration of new ideas, new technology and new investment. On the other hand, individuals who are geographically far apart are becoming more exposed to new ideas, new ways of thinking and behaving. Transport costs for people have been reduced and migration has increased. With enhanced facility to move across regions and countries, individuals are more likely to come into contact with other individuals from different cultures. This is going to affect intra-regional as well as inter-regional cultural dynamics. However, trade conflicts may arise in the cultural industries and feed back into cultural dynamics. For example, the expansion of US film, music and TV production into Europe may cause a reaction in European industries at the EU level, and trigger a revival in European activities and cultures. In the same way, the cultural perception of immigrants by natives may trigger a renaissance of local (as well as migrant) culture and identities. Within this context, UNESCO Universal Declaration on Cultural Diversity, says: “As a source of exchange, innovation and creativity, cultural diversity is as necessary for humankind as biodiversity is for nature”.
Creativity, Cultural Investment and Local Development
289
The key element in global competition is no longer the trade in goods and services or capital flows, but the competition for people. In the past, firms generally selected locations for new factories or offices based on where doing business was cheapest. Looking for the cheapest worker is now an old way of thinking: what firms really want is a talented labor force, not the least expensive. According to this view, “when Prashanth Boccasam founded a small software company in the fall of 2001, he could have remained in Houston, where he lived at the time; but he chose to relocate the company to the Tyson’s Corner section of Washington D.C., which had much more to offer culturally than Houston. It was clear to him that the young, talented people he was trying to hire all wanted to reside in a place4 with a healthy and vibrant cultural life” (Eger 2003, p. 13). Successful cities recognize that art and culture are vital not only to a region’s life quality, but also to the preparedness of its workforce.
5 Intrinsic Motivations: Arts and Capabilities Throughout history, science, mathematics and technology have flourished only where and when all the arts have flourished. However, no clear evidence exists about the causal relationship between the two groups of activities. Several recent studies reveal that exposure to and participation in the arts strengthens children’s educational performance. Following Eger (2003), a study of 150 biographies of great inventors - from Pasteur to Einstein - demonstrated that nearly all of the eminent scientists were also musicians, artists, writers or poets. A ninemonth study of three-year-old children demonstrated that early training in singing and playing musical instruments stimulates the brain in pre-school children and enhances learning. Moreover, college student test scores were higher when students listened to ten minutes of Mozart’s piano music immediately prior to taking an IQ test. The assumption that the arts can play a very positive role in the success of the post-industrial information economy and society, and ensure a creative and competitive workforce to meet the economic opportunities of both the present and future, is achieving increasing consensus. According to this view, for instance, in 2002 the Los Angeles County Board of Supervisors decided that sequential instruction in the multiple arts disciplines would be scheduled into each school day and accounted for in the budget of every county school district. Within this context, the Lisbon objectives, adopted by the Lisbon European Council in 2000, that Europe should become the most competitive and dynamic knowledge-based economy in the world by 2010, capable of sustainable economic growth with more and better jobs and greater social cohesion, calls for a precise educational strategy. At the Lisbon meeting, the European Council invited the Education Council to undertake a general reflection on the concrete future objectives of education systems in order to allow European citizens to develop their own skills and 4
Note, however, that suburban quarters – those replete with strip malls that creative young people tend to shun – have grown more rapidly than most original urban environments (Eger, 2003).
290
P.L. Sacco, G. Segre
competences and realise their potential. The main steps, opening up education and training systems to the wider world of non EU-countries, adapting to family-friendly timetables, and making learning more attractive, include the following key variables: increasing literacy and mathematical skills, updating the definition of basic skills for the knowledge society, maintaining the ability to learn, sustaining curiosity and interest in new development and skills, equipping schools and learning centers in order to ensure access to information and communication technologies, increasing the recruitment to scientific and technical studies and improving education and training for teachers and trainers. Finally, since education and training are considered instruments at the disposal of society, a general principle is that they should be used to develop the sort of society we want, where racism, intolerance or discrimination on any grounds are unacceptable and social cohesion is essential. The framework developed by the EU Education Council is highly compatible with the idea that education should involve all aspects of modern life, in the effort to create competitive knowledge-based economies and forward-looking societies. It lacks, however, any explicit discussion of the role of the arts in this important and strategic process. In addition, the introduction of a role for the arts in a strategic plan for local and global development should also recognize that the traditional way of considering arts in education systems must evolve. Students must go beyond the mere analysis of the products of creative understanding, such as poems, choreographies, paintings, songs, experiments and theories. They must be able to deconstruct them to gain an understanding of the real, highly complex process that led to their invention. This view is very close to the main traits of the “capability approach” developed by the Nobel prize-winner Amartya Sen (Sen 1999). Capability represents the freedom to achieve valuable “being and doing” (what Sen calls “functioning”), which reflect the person’s freedom to lead one type of life or another. The aim of society should be, indeed, to give people genuine choices, and not to force them into a particular life, however rich it might be in other respects. Future patterns of local economic development depend on opportunities for “capability building”: only those communities that can maintain a social attitude to having access to cultural experience on a widely diffused basis can be truly competitive in the global contest for innovativeness. The crucial issue in this respect is enabling individuals to access the competences that are needed to appreciate and value a given experience or creative goods. For instance, although everybody can drink a sophisticated wine, not everybody is able to reconstruct all of the sensory complexity that is potentially associated with the act of drinking that particular wine. This requires a specific experience base, training and development of individual sensitivity that results from a specific form of investment in building specific skills. The more wide-ranging the set of experience-specific skills one is able to develop, the wider is, in principle, one’s experience space, i.e. the set of experiences which one is able to value and appreciate. We define this process as the acquisition of competences, where the notion of competence identifies the effect of the stimulus of cultural, symbolic and identity capital, as will be explained in the following section. The acquisition of competences can be seen as a way of substantiating Sen’s “capability” approach. In particular, one can see Sen’s capabilities as a basic competence level that is crucial
Creativity, Cultural Investment and Local Development
291
in determining the capacity of individuals to exploit the cultural opportunity potential of experiences and to rely on it to develop self-fulfilling patterns of demand for commodities.
6 Toward a New Generation of Endogenous Growth Models The foregoing discussion is useful in laying the foundations for a novel approach to endogenous growth that takes account of the role of super-core and core cultural activities (and their impact on the creative and production sectors) in fostering local development processes. The acquisition of competence is the key factor that regulates the process of growth, and in particular, is responsible for its onset. In its essence, it runs as follows. Assume that the level of competence and capability of consumers is large enough to guarantee that they will be willing to pay for the creative component of a given quality commodity.5 If a part of this sophisticated consumers base is made up of creative workers, firms hiring them will take advantage of such skills to create better creative goods and services, and will have an incentive to invest in increasing its creative assets, and will expecting a relatively high return from this investment on the basis of the workers’ skill pool and of the consumers’ willingness to pay for quality. This will cause an increase in the menu of cultural opportunities, represented by the increase in the stock of cultural, symbolic and identity capital, which is only partly appropriable by the firm. The consequent increase in the quality and dimension of local cultural supply will foster a higher cultural demand by non-core creative workers (opportunity effect). If the size of the stock of social capital is also large enough, the higher exposure to creative experiences of the latter will increase the acquisition of competence of the non-core creative workers by means of social awareness and pressure (sociability effect). Therefore, they will increase their personal investment in their own competence and capabilities. Provided that there is enough complementarity between the creative experience and their job tasks, this will encourage their intrinsic motivation component, spurring their innovativeness, and thus organizational performance. At this point, if part of the added value thus generated is devoted to the financing of super-core and core creativity activities, both as an initiative of the creative firms themselves and of the public sector, a virtuous circle is created. In particular, the virtuous circle consists of a relationship between the demand for cultural and creative goods and the corresponding supply, which, by increasing creative choice opportunities, further boosts creative demand. A virtuous circle that can be seen as part of the story we want to tell is also discussed by Garavaglia and Breschi (2009, in this book), where the mechanism of positive externalities arising from concentration of production is described. A greater number of producers and of product varieties in a local area, raises the demand for labor and wages, attracts workers, who are also consumers. This 5
The framework is therefore more informative in the context of developed rather than developing countries.
292
P.L. Sacco, G. Segre
competence
social awareness
creative demand Social capital Cultural capital Symbolic capital Identitarian capital
increased choices possibility
willingness to pay
creative supply
Fig. 2 The competence virtuous circle
generates a larger market in the region, where demand is sustained. This further attracts entrepreneurs, both in the core activities of the area and in the supportive and complementary activities. Within this framework, the entrepreneurship activity is the engine of cluster formation and persistence, reinforced by the idea that, where the share of entrepreneurial self-employment is higher, the supply of entrepreneurship is stronger. Accordingly, the endogenous growth mechanism, in our view, is sustained by cultural investment which increases the stock of cultural, symbolic and identity capital, as well as by the possibility of involving entrepreneurs in the process of product innovation. We can illustrate the theoretical working of this mechanism with the following simple scheme (Fig. 2). The scheme is entirely compatible with the logic usually described by policy makers when strategic plans aimed at revitalising old industrial cities are undertaken. This is, for instance, the case of several cities such as Austin, Pittsburgh or St. Louis in the “fly over states” of the US (see Sacco and Pedrini 2003, for more details), as well as, with varying degrees of success, Homebush Bay in Sydney and Bicocca in Milan (Sacco and Tavano Blessi 2005). However, the main point we want to make is that the functioning of culture-led economic development (often invoked by policy makers in recent years and increasingly taken into account by economists) has always been described essentially as a “black box” where all - and sometimes not even all - important factors were considered together, without any attempt to identify the causal relationships between them.
Creativity, Cultural Investment and Local Development
293
In this book, the capacity to transform advances in the technological field into new products is analyzed by Bramanti and Riggi (2009, in this book) within the framework of a neoclassical Solowian model where total output growth depends on innovation. Innovation is described as the output of knowledge circuits - where competence and creativity are considered, but not explicitly analyzed - based on accumulation of endogenous human capital, originating from local connections and tacit knowledge, and of exogenous human capital, originating from exchange of information. In Bramanti and Fratesi (2009, in this book), the process of knowledge creation, accumulation and exchange is further discussed and the effect of innovative leader firms on product innovation is identified.
7 Conclusions The present chapter is a first attempt at spelling out the mechanics of endogenous growth that can be generated by a culture-led local development model in which culture basically works as a systemic “activator of innovation” through its impact on people’s willingness to invest in the development of their cognitive competences. This conceptual core has to be developed into a fully fledged formal growth model in order to proceed to a comparative assessment of alternative models and, ideally, to an empirical test. There is clearly much more work yet to be done, and several open questions to be addressed, such as providing a detailed account of the mechanics of strategic complementarity between access to culture and innovativeness in noncultural sectors, or defining the cognitive mechanisms that lie behind competence building (see however Sacco and Zarrri 2004, for a preliminary account). However, we believe this to be a very promising field for further research and will pursue these and other open issues in our forthcoming research.
References Bramanti A, Fratesi U (2009) The dynamics of an innovation driven territorial system. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Bramanti A, Riggi MR (2009) Sustainable interrelated growth: a phenomenal approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Caves RE (2001) Creative industries. Harvard University Press, Cambridge, Massachusetts, USA Comedia (2002) Releasing the cultural potential of our core cities. Core Cities Group, London, UK Comunian R, Sacco PL (2006) NewcastleGateshead: riqualificazione urbana e limiti della citt`a creativa, Archivio di Studi Urbani e Regionali 37:5–34 Eger JM (2003) The creative community. The California Institute for Smart Communities, San Diego, California Florida R (2002) The rise of the creative class. Basic Books, New York, USA
294
P.L. Sacco, G. Segre
Florida R (2005) The flight of the creative class: The new global competition for talent. Harper Collins, London, UK Florida R, Tinagli I (2004) Europe in the creative age. Carnegie Mellon University, mimeo Forte F, Mantovani M (2005) Manuale di Economia e Politica dei Beni Culturali. Rubettino, Catanzaro: Italy Garavaglia C, Breschi S (2009) The co-evolution of entrepreneurship and clusters. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Hertenstein JH, Platt MB, Veryzer RW (2005), Impact of design effectiveness on corporate financial performance. J Prod Innov Manage 22:3–21 Landry C (2000) The creative city. Earthscan Publications, London, UK Lucas RE Jr (1988), On the mechanics of economic development. J Monet Econ 22:3–42 Maignan C, Ottaviano G, Pinelli D (2003) Economic growth, innovation, cultural diversity, Nota di Lavoro 12, Fondazione Eni Enrico Mattei Markusen A (2006), Urban development and the politics of a creative class: evidence from the study of artists. Forthcoming, Environment and Planning Miles S, Paddison R (2005) Introduction: the rise and rise of culture-led urban regeneration. Urban Stud 42(5–6):833–839 Metropolis (2002) The impact of major events on the development of large cities. Commission 1, World Association of the Major Metropolises, Barcelona, Spain Nathan M (2005) The wrong stuff: creative class theory, diversity and city performance, discussion paper no.1, September 2005, Centreforcities Romer PM (1986) Increasing returns and long-run growth. J Polit Econ 94:1002–1037 Sacco PL, Pedrini S (2003) Il distretto culturale: mito o opportunit`a? Il Risparmio 51:101–155 Sacco PL, Tavano BG (2005), Distretto culturale e aree urbane. Economia della Cultura 2:153–165 Sacco PL, Zarrri L (2004) Cultura, promozione della libert`a positiva e integrazione sociale. Economia della Cultura 13:499–507 Scott AJ (2000) The cultural economy of cities. Sage, London, UK Sen A (1999) Development as freedom. Oxford University Press, Oxford, UK Sharpe D et al (2004) Critique and consolidation of research on the spillover effects of investments in cultural facilities. Ryerson University Board Room, Toronto, Canada Schumpeter J (1911) The theory of economic development. Harvard University Press, Cambridge Solow RM (1956) A contribution to the theory of economic growth. Q J Econ 70:65–94 Stolarick K, Florida R, Musante L (2005) Montr´eal’s capacity for creative connectivity: Outlook and opportunities. Catalytix mimeo Throsby D (2001) Economics and culture. Cambridge University Press, Cambridge, UK
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration Giuseppe Arbia, Massimiliano Copetti, and Peter Diggle
1 Introduction The aim of this chapter is to present a class of statistical models to study the location of economic agents and their geographical concentration and explain their spatial interacting behaviour. Traditionally, the problem of the spatial location of economic activities has been approached in three different ways. According to the first way of thinking, spatial location of economic agents is of no interest. This is the view of traditional macro-economists who treat an economy as if it is collapsed into a single, dimensionless, point. The second traditional approach is that of treating spatial location by looking at the distribution of activities within geographical partitions such as administrative units, regions, municipalities etc. We will refer to this approach as to the meso-economic approach. While this approach deserves some attention, it is often undermined by the problem that any conclusion reached depends on the geographical partition chosen (e.g. administrative regions) and could be different if referred to a different partition (e.g. provinces) or to the single economic agent (see Arbia 1989). The third way in which spatial location has been approached refers to the study of the spatial equilibrium reached under some pre-defined theoretical conditions. This approach was very fashionable in the past (Hotelling 1929; Christaller 1933; Palander 1935) and underwent various re-appraisals in the fifties (L¨osch 1940,1954; Isard 1956; Greenhut 1956) and in the seventies (Paelink and Nijkamp 1975; Paelink and Klaassen 1979). However, it has the drawback that it does not provide statistically testable models. In contrast, in this chapter we will present a class of spatial micro-econometric models that are amenable to statistical estimation.1 In particular, we will review a set of spatial statistical models derived from the so called point-pattern analysis 1
A similar “agent-based approach” is also employed by Maggioni and Roncari (2009) in this book.
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
297
298
G. Arbia et al.
(Cressie 1991; Rathbun and Cressie 1994; Diggle 1983) that have been recently proposed in the economic literature by Arbia (1996, 2001a,b), Arbia and Espa (1996), Quah (2002, 2003) and Arbia et al. (2008). The layout of the present chapter is the following. In Sect. 2 we discuss more thoroughly the need for a sound statistical theory underling the procedures commonly used to measure the level and the dynamics of the spatial concentration of economic activities. In Sect. 3 we will consider the study of spatial concentration measured on a discrete space (e.g. observations within regional boundaries) by stressing some of the deficiencies of the statistical tools commonly employed and proposing some alternative approaches. Starting from the weakness of the approaches based on discrete-space observations, in Sect. 4 we will review some methods for the study of spatial concentration on a continuous space. Specifically, in Sect. 4.1 we introduce a kernel regression approach whereas Sect. 4.2 is devoted to a fully parametric modelling framework. Finally, Sect. 5 contains some concluding remarks.
2 Measuring the Spatial Concentration of Economic Activities One of the most remarkable findings in empirical analysis is the different regional concentration of economic activities. Economists have long been aware of this problem and statisticians have provided tools for measuring it. In this field we have seen a recent acceleration of interest and a growing number of studies in Europe especially due to many issues raised in the EU by the unification process. In fact, the narrowing of regional disparities is a fundamental objective of EU policy and the persistence of large differentials in per-capita GDP across regions is regarded as an impediment (link between growth and agglomeration of economic activities due to spill-overs and increasing returns to scales, see e.g. Barro 2000). An important question linked to this issues is whetherin the long run the recent European economic integration will imply a higher (as suggested by Krugman and Venables 1996—among others) or a lower concentration of economic activities (as is desired to achieve the aim of a more even distribution of income and wealth). Having stressed the growing importance of the issue of spatial concentration and the massive literature produced in this field in the recent decades, it is extremely surprising to find that the literature on the empirical measurement of spatial concentration has not kept up with the increased demand and that most of the empirical work is still based on some basic (a-spatial) statistical measurements. For instance Krugman’s (1991a, b) model is originally motivated by the study of the concentration pattern of US industries based on Gini locational coefficients (Gini 1912; Dalton 1920). Similarly Ellison and Glaeser (1997) (and subsequently Maurel and Sedillot 1999) suggest a measure based on the simple correlation coefficient. A third important example concerns the study of inequality and economic convergence (Barro and Sala-i-Martin 1992) where the conclusions are still based on statistical measures like the variance or the linear correlation coefficient.
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
299
In all the examples quoted, no attention is paid to the spatial nature of data. However, neglecting the spatial features of data produces serious biases when quantifying concepts like regional concentration and industrial agglomeration. In fact (as is clearly motivated by Arbia 2001a) spatial concentration consists of two different features that are very rarely kept separate: (1) an a-spatial concept of variability which is invariant to permutations, and (2) the concept of geographical pattern (agglomeration or polarisation). The measures currently in use capture only the first and disregard the second, important, effect. In the literature there appear to be two different approaches to taking space and spatial relationship into consideration when studying the concentration of economic activities. The first approach consists of using traditional measures based on discrete regional data (e.g. state, regions or counties), but introducing further measures to capture the feature of agglomeration. The second considers the individual economic agents distributed on a continuous space as points in a plane and consequently derives parametric and non-parametric modelling frameworks. The two approaches will now be discussed in turn in the next two sections.
3 The Analysis of the Concentration on a Discrete Space 3.1 Basic Concepts The fundamental problem with the statistical measures used in the empirical literature to characterise spatial concentration of economic activities is that they do not take into account anything which is truly spatial. In fact any statistical measure of variation or concentration satisfies the condition of anonymity with respect to the individuals (May 1952; Sen 1972), that is the property of being insensitive to any permutation of individual orderings. However, is this a desirable property for a spatial inequality measure? Of course, the answer is no. To illustrate this point, let us consider the Gini coefficient of location used by Krugman (1991a, b; p. 75). In a unitary square in the same graph the author reports the cumulative share of manufacturing activity (e.g. value added or employment) in the various regions of a study area on the horizontal axis, and the share of the widget activity in the same regions on the vertical axis. Following Hoover (1936), he then derives a concentration index as the area between the resulting Lorentz-type curve (Lorentz (1905) and the 45-degree line. In the particular example reported in the quoted work he considers three regions. “Region 1 has 20% of the total manufacturing employment, but 50% of employment in the widget industry. Region 2 has 40% of total manufacturing employment, and 40% of widget employment. Region 3 has the remaining 40% of manufacturing employment, and the remaining 10% of widget employment” (Krugman 1991a; pp. 55–56). The resulting curve is reported in Fig. 1 and the index is equal to the surface area A. It is often convenient to divide the index by its maximum (that is 0.5) to obtain an index ranging in the interval (0,1).
300
G. Arbia et al.
Fig. 1 Gini-type concentration curve. After Krugman (1991a, b, p. 56)
40%
40%
Region 1
Region 2
20%
Region 3
a
40%
Region 1
20%
Region 3
40%
Region 2
b Fig. 2 Two hypothetical situations where the locational Gini coefficient is the same
It is clear that the index used is totally insensitive to the relative position of the three regions. For instance, Fig. 2 reports two different hypothetical situations referring to the data used in Fig. 1. Apart from the different dimension of the three regions (a problem treated in Arbia 1996) the two spatial distributions differ in terms of the position of the regions in the area. As a consequence, the distribution of employment is very different in the two cases with stronger evidences of geographical concentration in the first case (see Fig. 2a). However from the viewpoint of the concentration index used they are indistinguishable. Now consider a second hypothetical example where we have 12 firms located in a study area exhaustively partitioned into 16 squared cells (subregions) arranged in a 4-by-4 regular lattice grid. Figure 3 reports three very different locational situations. It is clear that the level of polarisation is much higher in case 3a than in case 3c. However the Gini concentration ratio remains unchanged in the three cases. This is because the index is essentially an a-spatial measure which fails in distinguishing between inequality of the distribution and polarisation.
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
301
3
3
0
0
3
0
0
0
0
0
0
0
3
3
0
0
3
0
0
0
0
3
0
3
0
0
0
0
3
0
0
0
0
0
0
0
0
0
0
0
3
0
0
0
0
3
0
3
a
c
b
Fig. 3 Three extreme cases of polarisation of 12 firms with constant concentration ratio. In the three cases Gini’s Concentration Ratio is 0.75, but polarisation is higher in case (a) than in case (c)
The problem is therefore that of characterising the three situations reported in Fig. 3 so as to be able to rank them in descending order from case 3a to 3c. An exploratory tool that can help in distinguishing the three cases of polarisation is the spatial correlation coefficient introduced by Moran (1950) and Whittle (1954) and much studied by Cliff and Ord (1973) and criticised by Arbia (1989). Spatial correlation measures derive directly from the concept of serial correlation in time series analysis and from the definition of spatial lag (for the analytical derivation see Arbia and Espa 1996 amongst others). One popular measure is that proposed by Moran (1950): n
I=
n n
n
∑ ∑ wd,ij
i=1 i=1
n
∑ ∑ wd,ij (xi − x¯ )(xj − x¯ )
i=1 j=1
n
(1)
∑ (xi − x¯ )2
i=1
Index (1) can be used to characterise different polarisation cases. For instance, going back to the cases displayed in Fig. 3 Moran’s I assumes the value I = 0.277 in the case 3a (high polarisation); I = 0.222 in the intermediate case 2b; and a negative value (I = −0.183) in the case of high dispersion 83c). The case of positive spatial correlation therefore is associated with a high degree of polarisation, whereas the case of negative spatial correlation is associated with low degree of polarisation, where data are spatially dispersed in the study area. Finally, the case of zero spatial correlation is associated with no particular geographical regularity when data are randomly distributed over space. It is important to remark, however, that a spatial correlation coefficient alone is not a measure of spatial concentration in the meaning we attach to the word in the present context. Let us go back to the example of the 12 firms distributed in a 4-by-4 regular lattice grid and consider the two hypothetical situations reported in Fig. 4. It is clear from the example that the level of dispersion is higher in the case 4b than in the case 4a as described by a Gini’s ratio which is three times greater in the first case. However, the Moran spatial correlation coefficient remains unchanged in the two cases because it measures the level of polarisation and it is insensitive to the general variability of the phenomenon.
302
G. Arbia et al.
3 3 3 3
0 0 0 0
0 0 0 0
0 0 0 0
a Moran’s I = 0.222 Gini’s Concentration Ratio = 0.75
0 0 0 0
1 1 1 1
1 1 1 1
1 1 1 1
b Moran’s I = 0.222 Gini’s Concentration Ratio = 0.25
Fig. 4 Same spatial correlation with different concentration ratio. Two possible spatial distributions of 12 firms in 16 regular cells. Case (a) refers to high concentration and case (b) to low concentration, but the spatial correlation index is the same in the two cases. (a) Moran’s I = 0.222, Gini’s concentration ratio = 0.75 and (b) Moran’s I = 0.222, Gini’s concentration ratio = 0.25
The general conclusion deriving from these examples is that if we want to characterise a spatial distribution in terms of both concentration in an a-spatial sense and polarisation, it is necessary to consider both features simultaneously. An exploratory analysis that aims at identifying both characteristics could start by inspecting the descriptive plot termed GI-plot (Arbia 2001a). In Fig. 5 we report the plot and four extreme cases of spatial concentration. The top-right box of the GI-plot refers to cases of high spatial concentration in that it contains geographical situations characterised by both high concentration (G > 0.5) in a non-spatial sense and high polarisation evidenced by a positive spatial correlation. Conversely the bottom-left box of the GI-plot refers to cases where spatial concentration is low in that it includes situations characterised by low aspatial concentration (G < 0.5) associated with negative spatial correlation. The GI-plot does not produce a complete ranking, but only a quasi-ordering (Sen 1972) of the various geographical situations. In particular, there is no obvious way of ranking vis-`a-vis points falling within the top-left box and points falling within the bottom-right box. In the next section we illustrate the use of the GI-plot by referring to a European dataset on the distribution of manufacturing industries in Europe reported by Arbia (2006).
3.2 Some Empirical Evidence for the Spatial Concentration of Economic Activity in Europe (1980–2000) 3.2.1 The Dataset The analysis that follows illustrates the use of the GI-plot with reference to the spatial concentration of value added in Europe in the years between 1980 and 2000 (see Arbia 2006). Firstly, the differences between sectors in the level of regional concentration and polarisation in 2000 are analysed. Secondly, we analyse how concentration and polarisation of the different sectors change over time. In particular,
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
303
G 0 0 0 0
0 3 0
0 0 0
0 3 0
3 3 0
3
0
3
0
Region A
3 3 0
0 0 0
0 0 0
0
0
0
Region B
.5
1 1 0 1
1 0
1 1
1 0
0 0
1 0 1 1 1 1 Region C
1 1
1 1
1 1
0 1 1 1 0 1 1 1 Region D 0
I
Fig. 5 The GI-plot. The horizontal axis is the I-Moran spatial correlation coefficient, the vertical axis is the G-Gini concentration ratio. The graph reports four extreme cases of spatial concentration. In region a I = −0.183 and G = 0.75; in region b I = 0.277 and G = 0.75; in region c I = −0.222 and G = 0.25; finally in region d I = 0.222 and G = 0.25
the dynamic analysis is performed by dividing and comparing regional concentration/polarisation in two different periods (1980–1990 and 1990–2000) in order to properly investigate the impact of economic integration. It was during the nineties, indeed, that the ongoing process of EU integration accelerated, due to the Single Market Program. We use the Cambridge Econometrics’ European Regional Databank, which is based on the EUROSTAT series. The data on Gross Value Added (GVA) were transformed in terms of PPP. Lacking estimates on the PPP at a regional level, the correction was based on national price levels.2 The sample used in the empirical analysis covers the group of regions belonging to the EU-15 countries, thus including regions of Austria, Finland and Sweden, 2 Most of the previous studies on regional concentration have examined employment as the central activity variable. As suggested by Aiginger and Leitner (2003), the use of value added has definite advantages, since it eliminates differences in productivity levels, as well as differences in the productivity gradient between the core and the periphery. Moreover, one must consider that recent employment trends have been blurred by the increasing shares of part-time work.
304
G. Arbia et al.
which entered the Union only in 1995. The NUTS-1 level was used for the United Kingdom, while Denmark, Ireland and Luxembourg were considered one-region countries (level NUTS-0). Germany does not include the former GDR Eastern Lander. The outermost regions of France (Departement d’Outre Mer, Guadeloupe, Martinique, Guyane, Reunion), Spain (Ceuta and Melilla, the Balearic and Canary Islands) and Portugal (Ac¸ores and Madeira) were excluded from the sample.Two other regions were excluded for different reasons: Groningen (Netherlands) and Vastsverige (Sweden). The list of 164 regions included in the sample is reported in the Appendix 1. Regional data are classified in 14 branches namely: Manufacturing (further divided into “Mining and energy supply”, “Food, beverages and Tobacco”, “Fuel, chemical, rubber and plastic products”, “Textiles and clothing”, “Electronics”, “Transport equipment”, and “Other Manufacturing”), Constructions; Total Market Services (further divided into “Wholesales and retail”, “Transport and Communication”, “Hotel and restaurants”, “Financial services”, and “Other market services”) and Non Market services.
3.2.2 Static Analysis For each sector we compute a Gini-type location index (Gini 1912; Dalton 1920) using the procedure described in Aiginger and Leitner (2002) (see Arbia et al. 2003 for details). As for the I-Moran computation we apply (1) to the regional share of value added in the various sectors. By working with ratios rather than absolute values the effect of the different dimension of each locational unit is eliminated. The results of the computation of the Gini location index and the Moran’s I in 2000 are reported in Fig. 6. As we already said, the GI-plot induces a quasi-ordering in the sectors: those falling in the top-right box dominate the sectors falling in the bottom-left box for both spatial polarisation and concentration. However, the other two boxes are not exactly rankable. In one case polarisation prevails on a-spatial concentration, in the other a high a-spatial concentration is present, but data are geographically dispersed. Notice that the two main sectoral aggregates, Manufacturing & Energy and Market Services, lie respectively in the bottom right and in the top left boxes. Although not exactly rankable, their relative position indicates a stronger polarisation for Manufacturing & Energy: the I-Moran coefficient value is almost double that of Market Services. This sector, on the other hand, shows a higher level of concentration, due substantially to the performance of Financial services, Transport & Communication and Other Market Services. In 2000, Non-market Services and Construction are less concentrated and less polarised than the other aggregates. This result seems, however, quite obvious since each country/region has its own Public Service sector and its own building sector with which to serve the domestic market. Within Manufacturing & Energy, we can distinguish between two situations. Three out of seven sectors (Electronics, Textile & Clothing and Transport equipment) are spatially polarised: they lie in the top right box, thus indicating a high level
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
305
Static GI - plot (2000) .62 Trans
.60 Fin .58
Ele
Tex
OthS
Gini
.56 .54 .52 .50 .48 .46
MktS
T&C
H&R
OthM Whol
NMktS
MAN Food
Fuel
TOT
Min Cons
.44 .1
.2
.3
.4
I- M oran
Fig. 6 Static GI-plot of the spatial concentration of the various sectors in 164 European regions in the year 2000. (TOT Total; MAN Total manufacturing; Cons Constructions; Mkts Total Market Services; Nmkts Non-market services; Min Mining and energy supply; Food Food, beverages and tobacco; Fuel Fuel, chemical, rubber and plastic products; Tex Textiles and clothing; Ele Electronics; Trans Transport equipment; othm Other manufacturing; Whol Wholesales and retail; T&C Transport and Communication; H&R Hotel and restaurants; Fin Financial services; oths Other market services)
of both I-Moran and Gini. This is probably due to the existence of strong agglomeration economies and spill-over effects in these fields. In contrast, Mining & Energy and Food sectors seem to be quite spatially dispersed. A similar pattern is seen in the two remaining service sectors, i.e. Hotel & Restaurant and Wholesale & Retail.
3.2.3 Dynamic Analysis On the basis of the data examined, it is also possible to construct a dynamic GI-plot with reference to different moments of time. Figures 7 and 8 report the 14 sectors on the GI-plot for the years 1980, 1990 and 2000. In particular, Fig. 7 shows the path of movements of Gini and I-Moran coefficients in the three years considered for the main sectoral aggregates (Manufacturing & Energy, Market Services, Construction and Non-market Services) and for the Total value added, while Fig. 8 illustrates the global dynamics of polarisation and concentration for the various manufacturing and service sectors in the same period. In both graphs, the origin of the arrows refers to the first year, and the arrow-head refers to the final year. Arrows that are (approximately) parallel to the vertical axis indicate changes in the polarisation with a constant level of a-spatial concentration whereas arrows parallel
306
G. Arbia et al. Dynamic GI-plot (1980, 1990, 2000) .56 .54 E&M MktS
Gini
.52 .50
NMktS
TOT
.48 .46 .44 .1
Cons
.2
.3
I Moran
Fig. 7 Dynamic GI-plot of the spatial concentration of the various sectors in 164 European regions in the three years 1980, 1990 and 2000 (TOT Total; MAN Total manufacturing; Cons Constructions; Mkts Total Market Services; nmkts Non-market services; Min Mining and energy supply; Food Food, beverages and Tobacco; Fuel Fuel, chemical, rubber and plastic products; Tex Textiles and clothing; Ele Electronics; Trans Transport equipment; othm Other Manufacturing; Whol Wholesales and retail; T&C Transport and Communication; H&R Hotel and restaurants; Fin Financial services; oths other market services)
to the horizontal axis represent cases where the concentration changes, but without producing changes in the spatial distribution. From Fig. 7 we learn that the polarisation of the Total value added declined over time (the relative change is −7.6% between 1980 and 1990 and −4.1% between 1990 and 2000; see Tables 1 and 2), while changes in concentration levels appear always close to zero. Interestingly enough, this performance seems to be the result of two opposed paths followed by the two main aggregates, Manufacturing & Energy and Market Services. In line with the existing empirical literature, our data give evidence of a declining concentration of Manufacturing & Energy value added (−3.5% from 1980 to 2000) along with a reduction in polarisation (−3.9% the relative change). It is quite interesting to note that the reduction of the de-concentration process was particularly strong in the nineties, when the process of economic integration accelerated in Europe. Over this period the decline in a-spatial concentration was also accompanied by a reduction in spatial polarisation, so that the sector moved toward the bottom-left box. It is however important to remember that over this period, the European Union also increased its public support (through Structural and Cohesion Funds) in favour of backward regions in order to oppose the trend towards productive localization in core regions. Actually, Midelfart-Knarvik et al. (2000) show that European Structural Fund expenditure influenced the location of industry in Europe, thus mitigating the economic forces at work.
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
307
Dynamic GI-plot (1980-2000) .7 Trans Ele Fin
Tex
Gini
.6
OthS Fuel
T&C
Food
H&R Whol
.5
OthM
Min
.4 0.0
.1
.2
.3
.4
I Moran
Fig. 8 Dynamic GI-plot of the spatial concentration of the various sectors in 164 European regions in the time interval 1980–2000. (TOT Total, MAN Total manufacturing; Cons Constructions; Mkts Total Market Services; nmkts Non-market services; Min Mining and energy supply; Food Food, beverages and Tobacco; Fuel Fuel, chemical, rubber and plastic products; Tex Textiles and clothing; Ele Electronics; Trans Transport equipment; othm Other Manufacturing; Whol Wholesales and retail; T&C Transport and Communication; H&R Hotel and restaurants; Fin Financial services; oths other market services)
In the case of Market Services, instead, the concentration level increased slightly all through the period (+1.8%) and the polarisation level decreased in the ‘80s and increased in the ‘90s. In spite of this dynamics, the level of spatial polarisation of Manufacturing & Energy remains much higher than that of Market Services. In terms of a-spatial concentration, instead, we assess that in the final year the services sector became more concentrated than manufacturing. Figure 8 illustrates the dynamic GI plots for the various manufacturing and service sectors in the different period 1980–2000. It emerges that in most industrial sectors the Gini concentration ratio declined along with the I-Moran coefficient, especially in the nineties. Particularly strong is the reduction of concentration in Mining & Energy (the relative change 1980– 2000 is −11.2%), Transport Equipment (−5.7%), Electronics (−4.4%) and Other Manufacturing (−5.5%). The only sector showing a slight increase in the level of a-spatial concentration and in the level of polarisation in the nineties was Textile & Clothing. Moreover, for Transport Equipment we observed an increase in polarisation along with the decrease in concentration. These trends appear clearer in the figures in Appendix 2. Generally speaking, from observing the entire dynamics of the Gini concentration ratio and of the I-Moran index, a positive time correlation emerges between the two indices (see Table 3), thus suggesting that the de-concentration process of manufacturing added value in Europe observed by different recent empirical studies has been accompanied by a polarisation process.
308
G. Arbia et al.
A different situation emerges in the case of Market Services. Transport & Communication and Other Market Services show an increase in the concentration level; the other sectors do not reveal any relevant change in concentration. A huge change in the polarisation level occurred in the case of Financial Services. Unlike manufacturing sectors, service sectors do not reveal a strong correlation between concentration and polarisation levels. When the Spearman rank correlation turns out to be significantly different from zero, the sign is always negative.
3.2.4 Conclusions In this section we stressed some of the weaknesses and the inadequacies of the standard statistical measures of inequality when used to characterize spatial distributions of economic activities on a discrete geographical space (e.g. regions). We have also shown how some of these inadequacies could be eliminated by making joint use of two distinct statistical indicators: one for the a-spatial concentration (the Gini coefficient) and one for polarization (the Moran’s I coefficient). This leads to the exploratory tool termed GI-plot (Arbia 2001a; Arbia 2006). The method could be criticized because with the GI-plot we only obtain a quasi-ordering of the various situations and in some instances it is necessary to make a full ranking of the various territorial units (e.g. when targeting re-equilibrating resources). A naive way of producing a summary index that measures simultaneously a-spatial concentration and polarisation would be to combine the two proposed measure (G-Gini and I-Moran) multiplicatively or additively and, rescale the measure thus obtained so as to range between 0 and 1. However, this way of proceeding raises the question whether both features have to be weighted equally in the summary measure and whether the index preserves the properties of the two parent measures. Furthermore the sampling distribution of the summary statistic is not straightforward even in the simplest case of Gaussian distribution. A further possibility is to explore the use of normative, rather than positive, measures of concentration where some hypothesis on the social welfare function is explicitly assumed (Dalton 1920; Atkinson 1970). Normative measures of polarisation could equally be proposed. Finally, we could approach the problem in a completely different way by studying the spatial characteristics of concentration them on a continuous space. This is the approach followed in the next section.
4 The Analysis of the Concentration on a Continuous Space 4.1 Introduction The path-breaking lectures of Paul Krugman at Leuven University (Krugman 1991a) stimulated interest in the so-called “new economic geography” by revealing unexpected links between the two previously unrelated fields of international economics
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
309
and economic geography. Statistical evidence of the geographical concentration pattern of many industries in the US, based on the computation of Gini locational coefficients, is the basis of the Krugman model. In Sect. 2, we discussed the problems arising with an approach based on discrete space observations. Since these problems arise from the arbitrariness of geographical partitions, why not simply remove the boundaries, and analyse the economy on a continuous space? Economists often see that economic activities locate on a continuous space and that “there is no particular reason to think that national boundaries define a relevant region” (Krugman 1991a, b). So why should a regional boundary define a relevant region? We are not saying here that boundaries should be altogether ignored, but only that we need to distinguish between the meanings of boundaries in different situations. In some instances boundaries can be classified as significant borders, that is, as places where the economic conditions change abruptly because of some change e.g. in the tax system or in transport costs. In other instances we can speak of irrelevant borders, where nothing actually happens from an economic standpoint. In such instances it becomes reasonable to incorporate information on significant boundaries into an explanatory model (as we shall see later). But in the meantime at this early stage of an exploratory analysis we can simply ignore both kinds of boundaries. Starting from these considerations, Arbia (2001b) proposed shifting the emphasis from a meso to a micro-level of analysis. Krugman (1991a) remarked that “If we want to understand differences in national growth rate, a good place to start is by examining differences in regional growth”. Arbia (2001b) similarly asserts that “a good way to understand regional economics is to begin by examining the micro behaviour of economic agents in the space economy”. In fact, phenomena in nature are encountered on a continuous space and are developed over continuous time; it is only our limitations that force us to isolate phenomena in some way (and subsequently distort them by reducing the quantity of information). The same general idea has been adopted in time series analysis with the development of continuous time econometrics, and is providing significant contributions to many branches of economics (see Gandolfo 1990; Bergstrom 1990). The idea of continuous space modelling is not new in economic geography and spatial economics; it was already referred to by Weber in theoretical studies on industrial location at the beginning of the twentieth century (Weber 1909). More recently Beckmann (1970) and Beckmann and Puu (1985) analysed equilibrium conditions of models defined on a continuous space. Griffith (1986) discussed a spatial demand curve based on a central place economic landscape defined on a continuous surface. Kaashoek and Paelinck (1994, 1996, 1998) derived the properties of a non-equilibrium dynamic path of continuous space economic variables based on partial differential equation theory (John 1978; Toda 1989). However, these studies are all concerned with the theoretical properties of models and not with their statistical estimation.
310
G. Arbia et al.
There are many reasons why such an approach has not been adopted thus far; the most obvious are: lack of an appropriate statistical methodology, lack of accurate data (often not available for reasons of confidentiality), and lack of appropriate computer technology. However, the methods for analysing spatial data on a continuous space now form a well-consolidated methodological body (Cressie 1991; Diggle 1983; Arbia and Espa 1996), even though not all the inconsistencies with the discrete counter-part have yet been removed (see Arbia 1992; Rozanov 1993; Cressie 1996). The availability of statistical data at the individual agent level has also increased considerably in recent times, due to the diffusion of spatially-referenced administrative records, and the development of methods of concealing confidential data without seriously distorting the statistical information (see Cox 1980; Duncan and Lambert 1986; de Waal and Willenborg 1994; Willenborg and de Waal 1996). Finally, the possibility of an automated statistical analysis at surprisingly numerous levels of disaggregation has also increased dramatically due to the introduction of modern GIS technologies. Therefore, there no longer appear to be any technical obstacles to a microeconomic approach to regional problems In order to introduce the continuous-space approach to the problem of spatial concentration, let us characterize each economic agent with its the geographical coordinates. If we can measure the economic space at this level of resolution then we can work on a continuous geographical space and this will eliminate the problems connected with the arbitrariness of geographical boundaries. On the other hand, obviously, the statistical structures in such environments usually become, more complex and sophisticated. In this new framework we see the single economic agent as a point in a continuous space and observation related to it as one single realization of a spatial point random field: a stochastic mechanism that generates points in the space. For the study of dynamic phenomena, (e.g. the birth of new firms) we can characterize the agent with three coordinates, two for space and one for time, thus allowing a spatio-temporal point process analysis. From a purely statistical point of view we are interested in detecting the main features of the point process in order to give them some economically significant interpretation. The hypothesis of Poisson Process for the underlying point process is usually well justified in geographical studies and represents a point of departure for more sophisticated studies. To introduce this approach, we need to model two main features: the number of the events in any planar region and their position. For a general Poisson Process (Diggle 2003; Arbia and Espa 1995), let λ (x) be a non-negative valued function over the plane, x ∈ ℜ2 , called the intensity function, defined as: • For any +planar region A, the number of events, N(A) ∼ Poisson(µ (A)) where µ (A) = A λ (x)dx. • Given N(A) = n, the locations of the n events form an independent random sample from the distribution on A with probability density function proportional to λ (x).
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
311
Analytically, the intensity function is defined as the mean number of points per unit area, that is: , E[N(dx)] (2) λ (x) = lim |dx| |dx|→0 where dx is the infinitesimal region containing the point x and |dx| is the area of that region. If λ (x) = λ ∀x (i.e. if the intensity function is constant over space) the process is a homogenous Poisson process and the different regions within the same area have the same probability to host a point. This is obviously not a realistic assumption for most economic phenomena. Conversely, if the points tend to form local concentrations we encounter two different paradigmatic situations: a) inhibition (i.e. firms are more dispersed than in the random case); and (b) concentration (i.e. firms are more concentrated than in the random case. See Diggle (1983) and Arbia and Espa (1996), for examples). If λ (x) varies within the regions the process is said to be a non-homogenous Poisson process: this is a typical situation in most economic studies. The well-known discrete Poisson distribution provides the expected number of points in each location in accord with a pre-defined intensity function. The statistical literature suggests several tools to estimate the intensity function in a parametric, semi-parametric or non-parametric way (Diggle 2003, Arbia and Espa 1995). The main aims of these inferential tools is to design a very intuitive map to help identify regularities, clusters or agglomerations displayed by the point pattern. In the next two sections we will introduce two modelling frameworks along these lines: the first is based on a kernel regression approach and the second on a space–time regression approach.
4.2 A Kernel-Regression Approach 4.2.1 Multi-Type Point Pattern and Spatial Segregation In this section we will introduce the concept of spatial segregation and we will propose a new methodology to quantify such a concept. We will discuss the idea of spatial segregation with reference to a specific case-study that will allow us to simplify the discussion and present the main characteristics of the method. In recent years a lot of attention has been devoted to the geographical dimension of the mechanisms of creation and diffusion of technology. Consistently with the economic literature on the subject (Jaffe et al. 1993, 2000; Thompson and Fox-Kean 2002; Breschi and Lissoni 2004), we will represent “innovation” as “patent production” and we will assign the inventor’s residence as the location of the patent. In such a way we can represent patents as points of a spatial pattern. The temporal dimension of the phenomenon can be represented by taking into account the application dates of the patents. Our empirical analysis focuses on the use of the EPO-Dataset containing all patent applications made to the European Patent Office (established by the
312
G. Arbia et al.
Munich Convention) starting from 1978. In particular we will use the version developed by CESPRI, Bocconi University, Milan. The dataset provides us with information about applicants, inventors, date of application (day-level), International Patent Classification code (distinguishing the industrial sector), and citations among patents. The use of the inventors’s residence allows us to locate each patent precisely. In the present context we consider the case of Italy in 1990. In our database we omitted Sicily, Sardinia and the other Italian islands for lack of spatial continuity which produces serious biases in our procedure based on Euclidean distance. The resulting dataset consists of 44078 inventors, but only 25312 patents due to multi-inventor patents. To avoid any subjectivity in the assignment of a unique location to each patent, we consider only patents with one single inventor numbering 14632 in our database. The patents are classified into six industrial sectors, namely: Electricity Electronics, Instruments, Chemical Pharmaceutical, Process Engineering, Mechanical Engineering Machinery and Consumer goods Civil Engineering. In order to introduce temporal information we avail of the record of two dates related to the registration of each patent: (1) the publication date and (2) the priority date, that is the date of the earliest filling of an application made in one of the patent offices adhering to the convention. The choice of one date or the other is crucial because the time lag between the priority date and the publication date may range from 1.5 to 2.5 years. We chose the latter because it is the date that gets closer to the actual revelation of the patented invention. Distinguishing each patent in terms of its specific industrial sector allows us to interpret our data as a single realization of a multivariate spatial point process. The map reported in Fig. 9 displays the overall spatial distribution of the 14632 cases. The map reveals an evident agglomeration of points within innovative regions, namely the Milan area and the north-eastern part of the country. Indeed, all northern regions are very innovative and contain more than 83% of all patents. Conversely, the central part of Italy contains only about 14%, of the patents with a heavy concentration around Rome and Florence, and only 3% located in the south, mostly in the neighbourhood of Naples. This empirical evidence is not surprising considering the industrial gap which exists between Italian regions and, specifically the north-south dualism. The estimation of the intensity function for the whole country will further corroborate this visual impression. It is helpful to think of the previous map as a mixture of as many point processes as there are industrial sectors, each characterised by a certain intensity. Figure 10 reports an individual point map for each sector. The multivariate inhomogeneous Poisson point process hypothesis appears to be particularly well motivated in Fig. 10. We will henceforth assume, for each of the component processes, an independent Poisson process distribution with intensity functions λk (x), k = 1, . . . , 6 k indicating the industrial sector. Even though the six sectors have a different number of patents, they display similar spatial behaviour. For this reason a “comparative” analysis is more interesting than a separate study of each individual sector. If, a certain industrial sector is present more in one region than in the other regions, we will detect “spatial
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
313
Fig. 9 Location of 14632 patents in Italy in 1990. Source: European Patent Office data processed at CESPRI, Bocconi University, Milan
segregation” or “agglomeration”. More precisely, if an industrial sector is predominant in one particular sub-region with respect to the other sectors, we say that a phenomenon of segregation occurs. Formally a multivariate point pattern exhibits spatial segregation if for at least some j = i the conditional intensity of a type- j
314
G. Arbia et al.
Fig. 10 Location of 14632 patents in Italy in 1990 distinguished by sector. Source: European Patent Office data processed at CESPRI, Bocconi University, Milan
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
315
point at location x conditional upon type-i points at the same location is less than the marginal intensity of type- j point in that location (Diggle et al. 2005). Using the symbolism introduced by Diggle et al. (2005), we denote the type-specific probability by the symbol pk (x) i.e. the conditional probability that a case occurring at location x be of type k: pk (x) =
λk (x) 6 ∑ j=1 λ j (x)
(3)
We claim this tool to be a measure of spatial segregation. In fact, a constant typespecific probability, pk (x) = pk implies a completely un-segregated Poisson process, in that the different types of points show no propensity to occur in one particular sub-region more than in another. In contrast, a case of complete spatial segregation occurs if, for each location x, pk (x) = 1 for some sectors and pk (x) = 0 for the others. In fact in this case at any location only one single type of point occurs. In We can monitor various forms of spatial segregation between these two extreme situations by analysing the spatial distribution of the estimated functions pˆ k (x), k = 1, . . . , 6. The statistical framework for the estimation of the type-specific probabilities was proposed by Diggle et al. (2005). The procedure is based on a kernel regression estimator for probabilities surfaces pk (x). To start with, data are represented as multi-nomial outcomes Yi : i = 1, . . . , 6 where Yi = k denotes a patent belonging to k industrial sector, so that: n
pˆk (x) = ∑ wik (x)I (Yi = k);
(4)
i=1
where, for each industrial sector k: wik =
wk (x − xi ) ∑nj=1 wk (x − x j )
(5)
with wk () kernel function with band-width hk > 0, wk =
w0 (x/hk ) , h2k
where w0 () is the
standardised form of the function. In this section, we use the Gaussian kernel kernel
x 2 such that w0 (x) = exp − 2 . We also use a common band-width for all sectors chosen in order to maximize the cross-validated log-likelihood: n
6
(i)
Lc (h) = ∑ ∑ I(Yi = k) log pˆk (xi )
(6)
i=1 k=1
(i)
In (6), pˆk (xi ) denotes the kernel estimator based on all data except (xi ,Yi ). In such a way we can employ the pˆk (x)-surfaces to assess spatial segregation. Based on this methodological framework, Diggle et al. (2005) proposed to formally test the visual impression of segregation provided by the type-specific probability maps with a Monte Carlo test for the null hypothesis of non-segregation. In this setting, the null hypothesis is represented by H0 : pk (x) = αk ∀x, i.e. constant
316
G. Arbia et al.
type-probabilities and the test statistic is given by: n
6
T = ∑ ∑ [ pˆk (xi ) − αk ]2
(7)
i=1 k=1
with αˆ k = nnk and nk the number of cases of type k and n is the total number of cases. The value of T for the data is t1 and it is compared to the value of T,t2 ,t3 , . . . ,ts , for new datasets obtained via MonteCarlo simulation under the null hypothesis by relabelling the data at random whilst preserving the observed number of cases of each type. The p-value for this test is p = q+1 s , where q is the number of t j > t1 . We can also provide the statistical significance of the type-specific probabilities estimates through a Monte Carlo test. We will call this test the local spatial test. Once the estimates are obtained, we randomly re-label the sectors while preserving the amounts. We then estimate, as usual, the type-specific probabilities and we repeat this Monte Carlo simulation 99 times. At the end of the simulations, for each point of the grid where we estimated the probabilities we have the p-value of our estimates in the same way as above.3 We will now use the new statistical tools described above to characterize the spatial distribution of patents reported in Figs. 9 and 10. To start with, we choose the optimal band-width for the kernel estimator by maximizing the cross-validated log-likelihood and in (5) we set hopt = 16.5 km. Secondly, we estimate the typespecific probability surfaces. The evidence of segregation is now strong for all sectors. Figure 11 reports the map obtained for the sector “Instruments” in which the hypothesis of constant type-specific probability equal to 1481/14632 = 0.1 is rejected. White areas indicate no statistically significant estimates. The Monte Carlo test proposed in (7) (which considers all sectors at the same time) rejects the null hypothesis of no segregation with a p-value of 0.01 as was to be expected from looking at the estimated type-specific probabilities maps. The map shows small sub-regions where the number of patents for “Instruments” dominates the other sectors. This happens in particular in southern Abruzzo, in southern Lazio and in Calabria, (see white areas). This effect could be due to the low number of patents in those regions. In contrast, the same feature in Trentino, Friuli-Venezia Giulia and Liguria is due to a true dominance of the sector.
4.2.2 Temporal Segregation The methodological framework presented in Sect. 4.2.1 can be easily extended to encompass the time dimension. If we look at the temporal pattern of production rates in the nineties we see a marked increase for all sectors except Chemicals & pharmaceuticals which, conversely,, experienced a decline after an expansion in the mid-nineties. In a temporal context, segregation occurs when one or more types of points are predominant in one certain time-period and not another. The 3
The ad-hoc routines for the computations described were provided by Pingping Zheng of Lancaster University. We gratefully acknowledge his contribution here.
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
317
Fig. 11 Kernel regression estimation of the intensity of 14632 patents in Italy in 1990. Coloured areas represent the estimates that are statistical significant
318
G. Arbia et al.
main hypothesis on which the model is based closely resembles that considered in the static version presented in Sect. 4.2.1. A multivariate inhomogeneous Poisson point process is considered in which each component process has an intensity function which varies over time. In this new framework the type-specific probabilities become the conditional probabilities that a case occurring at time t is of type k. The statistical estimation procedure for temporal type-specific probabilities is also similar to that used for static analogues, but now x ∈ (0, +∞) and the cross-validated log-likelihood give a different optimal band-width. In our dataset this was equal to 135 days. In this new setting we want to estimate the type-specific temporal probabilities to understand the differences between sectors. In addition to the overall test for temporal segregation, we can test, at each time, whether the current typespecific probability pk (t) is significantly different from the time-average pk . Times at which this test gives a significant result, 95% level of significance, are indicated in Fig. 12 by bold line segments. In each sector we contrast the time-averagepk non-segregation - (number of patents in one sector divided by the total number of sectors) with the estimated type-specific probabilities. In Fig. 12 some peaks and depressions are particularly evident and indicate temporal segregation. Two features
Fig. 12 Time evolution of type-specific probabilities of the patents in Italy in 1990–2000. Grey scale represents the temporal pattern. The horizontal straight line represents the average. Bold lines represents the estimates that are statistical significant
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
319
are particularly evident by inspecting the graph: the growth at the end of the 90’s of the “Mechanical engineering machinery” and “Consumer goods civil engineering” sectors.
4.2.3 Conclusions In this section we proposed a new approach to the study of spatial concentration of economic activities based on point pattern analysis. In particular, we considered a kernel regression approach that enables us to produce maps of sectoral type-specific probabilities and formal statistical testing procedures that quantify the idea of static and dynamic segregation. In the next section we will consider a second way of exploiting the point pattern characteristics of data to model spatial concentration by making reference to a continuous space version of Krugman’s concentration model proposed in Arbia (2001b).
4.3 A Dynamic Model for Space–Time Birth–Survival-Growth of Firms 4.3.1 A Continuous Space Version of Krugman’s Concentration Model Generalities In this section we will introduce a class of models to explain the concentration and diffusion of firms in a continuous space. The formalism is taken from Arbia (1996; 2001b) and derives from a model proposed by Rathbun and Cressie (1994) for the spatial diffusion of vegetation. Generally speaking, there are two different reasons (which we shall keep separate in the present context) for spatial concentration of economic activities. We can have concentration either because of a large number of firms in a definite region of space, or alternatively, have a small number of firms that are very large in terms of production or employment. From a dynamic point of view, a higher concentration is recorded either because new firms locate in the same area, or because the existing firms grow. The distinction obviously becomes unnecessary when we analyse data at a regional level, but it is indeed essential in our continuous space context. We will therefore model the birth process for the new firms and the growth process for existing firms, separately. A Birth Process In order to introduce a birth process for the location of firms in a continuous space, let us start by defining A as the study area, x as the location of the firm, N as the number of firms located in A, Di(i = 1, . . ., N) as the dimension of ith firm (e.g. as measured by number of employees, gross product, or other), and δ (x) as a region in the neighbourhood of the point x. Let us further use the definition of intensity given in (2)
320
G. Arbia et al.
A good way to model a situation of non-homogeneity is to build up a birth process for the firms where the spatial intensity is non-stationary across the economy and varies according to the location. In specifying a model for the birth process we can assume that the location of the firms at some time in the past is taken exogenously (thus incorporating Krugman’s idea of “historical initial conditions”. See Krugman 1991a), and the location of new firms is the result of a non-stationary point process (Diggle 1983). More explicitly, we can express the intensity of the process in the point x as:
λ (x) = exp {β0 + β1d(x) + β2W (x) + β 3 X + β 4 R + Φ(x)}
(8)
with β = {β0 , β1 , β2 , β 3 , β 4 } a vector of parameters to be estimated. High λ (x) indicate a concentration of economic activity in the infinitesimal area centred in x. If we associate each point in the area with its corresponding estimated density, we monitor spatial concentration on a continuous space. In (7), d(x) indicates the distance of x from main roads, communication networks, and other points of interest. This term incorporates the cost of shipping the raw materials from the source to the production site and the final goods from the production site to the market. W (x) is a term measuring the sign and the intensity of the interaction between the firm located in x and the other existing firms and incorporates the idea of non-constant spatial returns. Furthermore, in (7), X represents a vector of independent variables assumed to be spatially heterogeneous (such as demand or unitary transport costs), and R represents a vector of exogenous regional policy instruments (such as local taxation, incentives) that can stimulate (or depress) the concentration of economic activities in the long run. Finally, Φ(x) is the error term of the model assumed to be spatially stationary, Gaussian and zero mean, but non-zero spatial correlations. Due to the nature of the error term, the estimation of (7) presents some problems that will be discussed more thoroughly in Sect. 4 while presenting a numerical application of the model. Equation (7) can be seen as a continuous space version of Krugman’s concentration model that avoids the problems connected with the arbitrariness of geographical partitions. The birth process can be supplemented by a death/survival process that accounts for dynamics. In order to introduce dynamics into the model, let us define the survival indicator at time t as: . 1 if the i-th firm survives at timet Mi (t) = (9) 0 otherwise Let us also define the survival conditional probability at time t + 1 as: Pi (t + 1, θ ) = Prob {Mi (t + 1, θ ) = 1 |Mi (t, θ ) = 1 }
(10)
We will assume that the survival at time t + 1 is a function of the dimension of the firm at time t, of the competitive influence (positive or negative) with the
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
321
neighbouring firms, of the global growth of the sector (nation-wide or region-wide), and of a set of explanatory variables and regional policy instruments. Applying the logistic transform to the probability Pi, we can model it by utilizing a spatial auto-logit model (Heagerty and Lele 1998) specified as: Logit[Pi (t + 1, θ )] = φi0 + φi1 Di,t + φi2Y ∗ + ϕ i3 X + ϕ i4 R + φ i5 Wt (xi )
(12)
where, in addition to the previously introduced notation, ϕ = {φi0 , φi1 , φ i2 , ϕ i3 , ϕ i4 , φi5 } is a set of parameters to be estimated, Di,t the dimension of the ith firm at time t, Y ∗ the global growth of the sector, and Wt(xi) a measure of the intensity of interaction at time t of the ith firm with its geographical neighbours. For a specification of Wt(xi) see Arbia (2001b). For a meso-level analogue of (12) see Van Wissen L (2000). Equation (12) could be estimated using the standard Maximum Likelihood procedure. A Growth Process The birth and the death/survival models account for the mere absence/presence of firms in the space, but, as we said, there is a second source of spatial concentration in the process of growth of existing firms. In order to describe a growth process, we can Ntas the number of firms existing at time t in K dimensional classes. Define further xik as the location of the ith firm belonging to the dimensional class k (k = 1, . . . , K). We further consider the first mk (i = 1, . . . , mk) firms of dimensional class k as surviving at time (t + 1) and the remaining nk − mk firms (i = mk + 1, . . ., .nk) as ceasing their activity in the interval (t,t + 1). Furthermore let Dik,t be the dimension of the ith firm belonging to the dimensional class k at time t, and let Yik,t be a measure of the growth of the ith firm belonging to the dimensional class k between time t and time t + 1. For instance we can assume Yik,t = (Dik,t + 1 − Dik,t)/Dik,t, but other definitions are obviously possible. The proposed model describes the spatial growth Yik,t as a function of the stage of development of the firm, its exposure to external sectoral shocks, the non-constant returns (positive or negative) deriving from being close to other existing firms, and a set of explanatory variables. More formally we have: K
Yik,t = αi0 + αi1 Dik,t + αi2Y ∗ + α i3 X + α i4 R + ∑ γi jWt j (xik ) + εik
(12)
j=1
where, in addition to the previous notation, α = [αi0 , αi1 , αi2 , α i3 , α i4 , γi1 , . . . γik ] is a set of parameters to be estimated, Wjt (xik) is a measure of the intensity of interaction of firm i belonging to the dimensional class k with its geographical neighbours belonging to the dimensional class j, and ε ik is independently and normally distributed noise with zero mean and finite variance. Given these assumptions, the estimation of (12) presents no statistical problem and can be performed by making use of the Ordinary Least Squares estimators.
322
G. Arbia et al.
Fig. 13 Location of 298 manufacturing industries in the San Marino Republic on 1 January 1996 (indicated by dots), of the 18 firms set up during 1995 (indicated by plus), and the main roads. The map is oriented with the top to the north. The markers in the horizontal and vertical axes are approximately every 830 m
An Empirical Example of Birth Process Modelling For illustrative purposes, we present an empirical estimation of the birth model (8) reported by Arbia (2001b). The data refer to the concentration of manufacturing industries in the Republic of San Marino. Even thoughthe data refer to a very small region of the world (approximately 61.19 square kilometres), they fully describe an area that constitutes an autonomous region with its own peculiar characteristics in terms of laws, tax regulations and political institutions. Figure 13 shows the location of 298 manufacturing industries on 1 January 1996 (Statistical Office 1996). Eighteen of these firms, set up in 1995, are displayed with a “+” in the same graph. Figure 13 also shows the main roads that cross San Marino. There is clear evidence of agglomeration between firms, in the neighbourhood of the main roads, and in proximity of the border with Italy (the top-right corner of the map). For the peculiarity of the San Marino Republic, it is reasonable to neglect the spatial variability of transport costs, the demand for goods, and regional policy
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
323
measures. Hence we can express the spatial intensity of new firms, λ (x), as:
λ (x) = exp {β0 + β1 d(x) + β2W (x) + Φ(x)} where d(x) indicates the distance in meters of each point of coordinates x from the closest road. For the interaction term W (x) we consider the particular specification suggested in Arbia (1996; 2001b). In particular it is assumed: ⎧ ⎨∑ exp[−ri (x)] if ri (x) ≤ 3 kilometers (13) W (x) = i ⎩0 otherwise where the index i = 1, . . . , 280 and ri (x) represents the distance between the firm located at x set up during 1995 and the ith firm already existing at the beginning of the same year. Interaction is restricted to firms falling within a 3-km radius. The theoretical maximum distance in the area (the diagonal of the rectangular region in Fig. 13), is about 8 km. The error term Φ(x) is assumed to be continuous-space stationary, Gaussian and with a zero mean subsequently discretised onto a regular grid of 200-by-200 squares (say Φ(xi j ), i = 1, . . . , 200; j = 1, . . . , 200) and it is assumed conditionally Gaussian with E(Φ(xi j|neighbours)) = γ ∑i ∑ j wi j Φ(xi j) (with wi j = 1 if the ith and jth squares are neighbours in the grid and zero otherwise), and Var(Φxi j|neighbours)) = τ 2 . Since the error term is unobservable, we use the modification of the EM algorithm for our estimation (Dempster et al. 1977) – proposed by Rathbun and Cressie (1994) – using the Gibbs sampler in the E step to compute the expected value of the conditional distributions, and the Newton–Raphson algorithm to find the maximum in the M step. Both the E and M steps were run using ad hoc routines. In Arbia (2001b) we obtained the following estimated model: , λ (x) = exp 12.02 − 2.602 d(x) + 0.127 W (x) (0.744)
(0.002)
(0.012)
where the figures in brackets refer to the standard errors of estimates. The negative value displayed by the parameter β1 indicates a negative relationship between the distance from the main roads and the probability of a new location of firms as it was to be expected due to the ease of access and shipment of raw materials and manufacturing goods. Conversely, the positive value of β2 indicates a positive interaction between new firm location and the location of previously existing firms, thus reflecting the action of positive spatial externalities typical of many economic geographic situations (Griffith 1999). The accuracy of the fit of the model was assessed with Monte Carlo procedures. The visual inspection of the empirical Kfunction (Ripley 1977) of the observed points contrasted with the 99% bands derived from 500 simulations of the estimated model (see Arbia and Espa 1996) led to the acceptance of the estimated model as a good description of reality.
324
G. Arbia et al.
4.4 Conclusions In this section of the chapter we have presented two alternative approaches to the study of spatial concentration of economic activities on a continuous space. In both cases the starting point was the computation of the spatial intensity of the process on a continuous set of co-ordinates. In Sect. 4.2 we summarized the main locational characteristics of economic activities as a map intensity and, on the basis of empirical data, we estimated the intensity surface using a kernel regression procedure separately for each economic sector. Peaks and depressions in this surface reveal situations of spatial concentration of each sector with respect to the others. In Sect. 4.3 we followed a regression approach to explain the spatial intensity of sets of point-data as a function of a series of explanatory variables including geographical features such as distances from points of interest. By using both approaches, we have shown that it is possible to extend the framework so as to consider the dynamic aspects of concentration, growth and diffusion of firms in the economic space.
5 Summary of the Chapter and Concluding Remarks Spatial concentration of economic agents is a complex phenomenon and its statistical measurement is a problematic issue. This chapter had two major aims. First of all, we stressed the weaknesses of approaches based on regional data and we argued strongly in favour of an approach based on micro-economic data modelled on a continuous space. Secondly, we presented two examples of how spatial concentration could be studied by making use of point data on a continuous space framework. In Sect. 4.2, we presented an approach that quantifies the concept of spatial segregation. In Sect. 4.3, we reported a prototype spatial micro-economic model describing the birth, survival and growth of firms on a continuous surface. This model can be thought as the continuous time version of Krugman’s model. The two examples should clarify the advantages of a modelling framework based on a continuous space over the traditional approach based on regional observations of economic phenomena.
References Aiginger K, Leitner W (2002) Regional concentration in the United States and Europe: Who follows whom? Review of World Economics (Weltwirtschaftliches Archiv), Springer, 138(4), 652–679, December Amrhein C (1995) Searching for the elusive aggregation effect: evidence from statistical simulations. Environ Plan A 27: 105–119 Anselin L (1988) Spatial econometrics. Kluwer, Dordrecht
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
325
Anselin L, Bera AK (1998) Spatial dependence in linear regression models with an introduction to spatial econometrics. In: Ullah A, Giles DEA (eds) Handbook of applied economic statistics. Marcel Dekker, New York, pp 237–290 Anselin L, Florax R (1995) New directions in spatial econometrics. Springer, Berlin Arbia G (1989) Spatial data configuration in statistical analysis of regional economic and related problems. Kluwer, Dordrecht Arbia G (1992) Aggregation over time space and individuals in economic modelling: a generating mechanism approach. In: Gandolfo G (ed) Continuous time econometrics. Chapman and Hall, London Arbia G (1993) The use of GIS in spatial statistical surveys. Int Stat Rev 61(2): 339–359 Arbia G (1996) Modelling individual behaviour of firms in the spatial economy. Quaderni Universit`a Commerciale Bocconi Milano 1(96): 1–11 Arbia G (2001a) The role of spatial effects in the empirical analysis of regional concentration. Geogr Syst 3: 271–281 Arbia G (2001b) Modelling the geography of economic activities on a continuous space. Pap Reg sci 80: 411–424 Arbia G (2006) Statistical foundations of spatial econometrics. Springer-Verlag, Berlin Arbia G, Espa G (1996) Statistica economica territoriale. CEDAM, Padova Arbia G, Benedetti B, Espa G (1996) Effects of the MAUP on image classification. Geogr Syst 3(2–3): 123–141 Arbia G, Espa G, Quah Q (2008) A class of spatial econometric methods in the empirical analysis of clusters of firms in the space. Empirical Economics – springer-Verlag – Vol. 34, Number 1/February Arbia G, Haining R, Griffith D (2003) Error propagation computing vegetation indices based on Landsat imagery. Environmental and ecological statistic 10(2): 375–396 Atkinson AB (1970) On the measurement of inequality. J Econ Theory 2: 244 263 Bade F-J, Nerlinger EA (2000) The spatial distribution of new technology-based firms: empirical results from West Germany. Pap Reg Sci 79(2): 155–176 Barro RJ (2000) Inequality and Growth in a Panel of Countries. J Econ Growth, Springer 5(1): 5–32, March Barro RJ, Sala-i-Martin X (1992) Convergence. J Political Econ 100(2): 223–251 Barro RJ, Sala-i-Martin X (1997) Technological diffusion, convergence and growth. J Econ Growth 2: 1–26 Bartlett MS (1964) The spectral analysis of two-dimensional point processes. Biometrika 51: 299–311 Beckmann MJ (1970) The analysis of spatial diffusion processes. Pap Reg Sci Assoc 25: 109–117 Beckmann MJ, Puu T (1985) Spatial economics: density, potential and flow. North Holland, Amsterdam Beckman MJ, Thisse JF (1987) The location of production activities. In: Nijkamp P (ed) Handbook of regional and urban economics, Vol. I Bergstrom AR (1990) Continuous time econometric modelling. Oxford University Press, Oxford Breschi S, Lissoni F (2004) Knowledge networks from patent data: methodological issues and research targets. CESPRI WP N◦150 Christaller W (1933) Die Zentralen orte in S¨uddeutschland: Eine o¨ konomisch-geographische Untersuchung u¨ ber die Gesetzm¨assigkeit der Verbreitung und Entwicklung der Siedlungen mit st¨adtischen Funktionen. Fisher, Jena Cliff AD, Ord JK (1973) Spatial autocorrelation. Pion, London Combes PP, Overman HG (2003) The spatial distribution of economic activities in the European union. In: Henderson V, Thisse JF (eds) Handbook of Urban and Regional Economics. Elsevier, North-Holland Conley T (1999) GMM estimation with cross-sectional dependence. J Econom 92(1): 1–45 Cox LH (1970) Analysis of binary data. Chapman & Hall, London Cox LH (1980) Suppression methodology and statistical disclosure control. J Am Stat Assoc 75(370): 377–385
326
G. Arbia et al.
Cressie N (1991) Statistics for spatial data.Wiley, New York Cressie N (1996) Change of support and the modifiable areal unit problem. Geogr Syst 3(2–3): 159–180 Dalton H (1920) The measurement of the inequality of incomes. Econ J 30: 348-361 Decressu J, Fatas A (1995) Regional labour market dynamics in Europe. Eur Econ Rev 39(9): 1627–1655 Dempster AP, Laird NM, Rubin DB (1977) Maximum likelihood for incomplete data via the EM algorithm. J R Stat Soc B 39: 1–22 de Waal AG, Willenborg LCRJ (1994) Minimizing the number of local suppressions in a Microdata set. Technical Report, Department of Statistical Methods. CBS, BPA no. 5997-94-M1. Diggle PJ (1983) Statistical analysis of spatial point patterns. Academic Press, New York Diggle PJ (2003) Statistical analysis of spatial point patterns (2nd edition). Edward Arnold, London Diggle PJ, Zheng P, Durr P (2005) Nonparametric estimation of spatial segregation in a multivariate point process: bovine tuberculosis in Cornwall, UK. J R Stat Soc Ser C (Appl Stat) 54(3): 645–658 Duncan GT, Lambert D (1986) Disclosure-limited data dissemination (with discussion). J Official Stat 81(393): 10–28 Ellison G, Glaeser EL (1997) Geographic Concentration in U.S. Manufacturing Industries: A Dartboard Approach. J Political Econ 105(5): 889–927 Fujita M, Krugman P, Venables AJ (1999) The spatial economy: cities, regions, and international trade. MIT Press, Cambridge Gandolfo G (1990) Continuous time econometrics. Chapman &Hall, London Gehlke CE, Biehl K (1934) Certain effects of grouping upon the size of the correlation coefficient in census tract material. J Am Stat Assoc 29(Supplement ): 169–170 Gini C (1912) Variabilit`a e mutabilit`a. Reprinted in Memorie di metodologica statistica (Ed. Pizetti E, Salvemini, T). Rome: Libreria Eredi Virgilio Veschi (1955) Greenhut ML (1956) Plant location in theory and practice. University of North Carolina Press, Chapel Hill Griffith DA (1986) Central place structures using constant elasticity of consumption demand cones. Part I: the infinite plane. Econ Geogr 62: 74–84 Griffith DA (1999) Statistical and mathematical sources of regional science theory: map pattern analysis as an example. Pap Reg Sci 78: 21–45 Haining RP (1990) Spatial data analysis in the social and environmental sciences. Cambridge University Press, Cambridge Haining RP (1994) Testing spatial interacting market hypothesis. Rev Econ Stat 66: 576–583 Heagerty PJ, Lele SR (1998) A composite likelihood approach to binary spatial data. J Am Stat Assoc 93(443): 1099–1111 Hoover EM (1936) The measurement of industrial localization. Rev Econ Stat 18: 162–171 Hotelling H (1929) Stability in competition. Econ J 39: 41–57 Isard W (1956) Location and space economy. MIT Press, Cambridge, MA Jaffe AB, Trajtenberg M, Henderson R (1993) Geographic localization of knowledge spillovers as evidenced by patent citations. Q J Econ 108: 929–952 Jaffe AB, Trajtenberg M, Fogarty MS (2000) Knowledge spillovers and patent citations: evidence from a survey of inventors. Am Econ Rev 90: 215–218 John F (1978) Partial differential equations. Springer, Berlin Kaashoek JF, Paelinck JHP (1994) On potentialized partial differential equations in theoretical spatial economics. Chaos solitons fractals 4(4): 585–594 Kaashoek JF, Paelinck JHP (1996) Studying the dynamics of pre-geographical space by means of space and time-potential partial differential equations. Geogr syst 3: 259–277 Kaashoek JF, Paelinck JHP (1998) Potentialized partial differential equations in economic geography and spatial economics: multiple dimension and control. Acta Appl Math 51: 1–23 Krugman P (1991a) Geography and trade. MIT Press, Cambridge, MA Krugman P (1991b) Increasing returns and economic geography. J Polit Econ 99(3): 183–199
Modelling Individual Behaviour of Firms in the Study of Spatial Concentration
327
Krugman P, Venables AJ (1996) Integration, specialization and adjustment. Eur Econ Rev 40: 959–967 Lorentz MO (1905) Methods for measuring concentration of wealth. J Am Stat Assoc 9(70): 209–219 L¨osch A (1940) Die r¨aumliche Ordnung der Wirtscaft. Gustav Fischer Verlag, Jena L¨osch A (1954) The economics of location. Wiley, London Maggioni MA, Roncari S (2009) Learning, innovation and growth within interconnected clusters: an agent based approach. In: Fratesi U, Senn L (eds) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin May KO (1952) A set of independent, necessary and sufficient conditions for simple majority decisions. Econometrica 20: 680–684 Maurel F, Sedillot B (1999) A measure of the Geographic Concentration in French Manufacturing Industires. Regional Science and Urban Economics 29(5): 575–604 Midelfart-Knarvik KH, Overman HG, Redding SJ, Venables AJ (2000) The location of European industry. Eur Commission Econ Pap Number 142 Moran PAP (1950) Notes on continuous stochastic phenomena. Biometrika 37: 17–23 Paelink JHP, Klaassen LH (1979) Spatial econometrics. Saxon House, Farnborough Paelink JHP, Nijkamp P (1975) Operational theory and methods in regional analysis. Saxon House, Lexington Books Palander T (1935) Beitrage zur standortstheorie. Almquist and Wiksell, Uppsala Quah D (May, 2002) Spatial agglomeration dynamics. Am Econ Rev (Pap Proceed), 92(2): 247–252 Quah D (2003) Spatial cluster empirics, paper presented at the EEA/NBER/ISOM ConferenceBarcelona June 2003. Available at http://econ.lse.ac.uk/staff/dquah/index own.html Rathbun SL, Cressie N (1994) A space time survival point process for a longleaf pine forest in Southern Georgia. J Am Stat Assoc 89(428): 1164–1174 Ripley BD (1977) Modelling spatial patterns. J R Stat Soc B 39: 172–212 Rozanov IA (1993) Markov random fields. Springer-Verlag, Berlin Sala-i-Martin X (1990) On growth and states. Ph.D. Dissertation, Harvard University, Cambridge, MA Sala-i-Martin X (1994) Cross-sectional regressions and the empirics of economic growth. Eur Econ Rev 38: 739–747 Sen A (1972) On economic inequality. Claredon, Oxford Statistical Office (1996) Operatori economici, liberi professionisti, enti vari. Centro elaborazione dati e statistica, Repubblica di San Marino Thompson P, Fox-Kean M (2002) Patent citations and the geography of knowledge spillovers: a reassessment. Am Econ Rev 95(1): 450–460 Toda M (1989) Nonlinear waves and solitons. Kluwer, Dordrecht Van Wissen L (2000) A micro-simulation model of firms: application of concepts of the demography of the firm. Pap Reg Sci 79(2): 111–134 Weber A (1909) Uber den Standort der Industrien.JCB Mohr,Tubingen Whittle P (1954) On stationary processes in the plane. Biometrika 4: 434–449 Willenborg LCRJ, de Waal T (1996) Statistical disclosure control in practice. Springer, Berlin
The New Approach to Regional Economics Dynamics: Path Dependence and Spatial Self-Reinforcing Mechanisms Domenico Marino and Raffaele Trapasso
1 Introduction Globalisation increased the level of competition between regions all over the world. Although the country effect is still significant, the (competitive) advantage of regions has dramatically changed and some areas – even in some industrialised countries – are suffering a general worsening of their economic performance (i.e. GDP trends), while some others are enjoying astonishing development. The ongoing situation confirms part of the theoretical conclusions of the New Economic Geography, and, at the same time, creates a huge number of opportunities for further research on regional development. An innovative theoretical approach extends to drawing the economic dynamic as the evolution of complex systems. Complexity can be introduced in economic formalization in many different shapes and patterns. A crisis of traditional economic models and (accordingly) of related policies is often a first result. The “Agent Based” models are sophisticated formalisations for studying complexity within regional economy and they also will be the main background for the analysis presented in this section.1 Specifically, by using sophisticated mathematical instruments it is possible to assess ongoing dynamics by combining three main issues. First of all, the presence of multiple specialisations in regions and their effect on consumer utility function (monopolistic competition a` la Dixit and Stiglitz 1977). Secondly, the effect that territorial contiguity of actors has on local development (shipping charges in transportation costs as in the ice-berg model of Samuelson, 1962). Lastly, the source of better performance in those regions which host haphazard interactions among firms of different branches and industries (Aoki 2002; Storper and Venables 2003). The first two points are embedded in the New Economic Geography (especially in the Krugman formalisation), which is the starting point of modern regional economics. The third (the evolutionary one) 1
For other approaches using the agent-based formalization, see the contribution of Maggioni and Roncari, 2009, in this volume.
U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
329
330
D. Marino, R. Trapasso
characterizes this contribution, which aims at giving a new interpretation of the concept of endogenous development, here considered as the dynamic development of a complex economic system. This paper will assess the economic dynamic as a self-reinforcing mechanism: positive (or negative) feedback that characterizes the evolution of a dynamic system. The concept of a self-reinforcing mechanism can be expressed as a dynamic system, with path dependence and positive feedback, which tends to produce a large variety of asymptotic states. Every evolutionary step of the system influences the next and thus the evolution of the entire system, so generating path dependence (David, 1985). Such a system has a high number of asymptotic states, and the initial state (Time zero), unforeseen shocks, or other kinds of fluctuations, can lead the system into any of the different domains of the asymptotic states (Arthur 1988). Furthermore, the system selects the state in which it places itself. Such dynamics are well known in physics, chemistry and biology and the final asymptotic state is called the emergent structure. The concept of positive feedback is relatively new in economics. The latter generally deals with problems of optimal allocation of scarce/insufficient resources, thus the feedback is usually considered to be negative (decreasing utility and decreasing productivity). Sf-reinforcing mechanism dynamics can be used to assess many different economic problems with different origins, from those related to the international dimension to those typical of the industrial economy, and, last but not least, problems related to regional economics. Many scholars have assessed multiple equilibria and their inefficiency (Marshall 1890; Arrow and Hahn 1971; Brown and Heal 1979; Scarf 1981). Multiple equilibria depend on the existence of increasing returns to scale. If the self-reinforcing mechanism is not counterbalanced by some opposite force, the output is local positive feedback. The latter, in turn, will amplify deviation from some states. Since these states derive from a local positive feedback, they are unstable by definition, so multiple equilibria exist and are efficient. If the vector field related to a given dynamic system is regular and its critical points follow some particular rules, then the existence of other critical points or of stable cycles (also called attractors) is a result (Marino 1998).2 The multi-attractor systems have some particular properties that are very useful to our research (Marino 1998). Strict path dependence is therefore manifested, and the final state of the system will depend on the particular path it has covered during its dynamic evolution from one (unstable) equilibrium to another (unstable) equilibrium. Accordingly, the system’s dynamic is a non-ergodic one. Three are the points where the research can be focussed. First of all, the identification of forces that act as attractors for the system; secondly, if these forces exist, assessing the possibility that the system will move from a lower to a higher equilibrium (and if so, in which way and how); finally, whether this transition from one level to another is spontaneous or needs some particular policy (effectiveness of policies). A first remarkable result is that different mathematical instruments give the same result. Accordingly, patterns of evolution can be numerous and different from each other, because of the existence of many stable multiple equilibria, and 2
For instance, this issue justifies the efficiency of the lower technology pattern of production within the market.
The New Approach to Regional Economics Dynamics
331
convergence paths (or phase transitions between the states). The stylized facts confirm that the process of regional development is discontinuous and unexpected: as in the case of new territorial agglomeration (clusters) created by a collective reorganization of the local productive framework.
2 Self-Reinforcing Mechanism and Complexity in Regional Economics For many years, regional economics has not been considered as the economic mainstream. The main reasons for such a situation are mainly related to two orders of factors. First of all, the perfect competition approach required a world in which all agents were equal (or divided into well-defined categories such as households or firms), without any difference between them. Secondly, the economic system as a whole was trying to reach a stable equilibrium and then to maintain it for as long as possible. In other words, the steady state was considered as a locus in which the system had no more incentives to move toward any other state. The result of this kind of formalization was weak and counterfactual, too weak to be benchmarked with the empirical evidence of many regions. The first attempt to give a theoretical (even though qualitative) basis to the empirical evidence for agglomeration dynamics dates back to 1890, when Alfred Marshall defined as “external economies” those economies which are external to a single firm but are internal to a specific area which is characterised by an “industrial atmosphere” (the latter being a form of public good). According to his definition, there are three main pillars that underpin the individual location choice of firms and workers: 1. The existence of a pooled labour market that enhances the probability of finding a job for workers, and, on the other hand, lowers the probability of labour shortages for firms 2. The localized production of non-tradable specialized inputs 3. The possibility for firms to gain a better production function thanks to the existence of informational spillover Marshall didn’t leave a formalized model of his insight. He avoided facing a theoretical “Gordian Knot” since the existence of a source of competitive advantage for firms localized in a specific area was a sort of “shock” for orthodox economic theory: the presence of “unexhausted economies of scale at the level of firms undermine[d] perfect competition” (Krugman 1998). The aim of preserving the coherence and elegance of the “perfect competition” formalization led many scholars to bypass the problem of the competitive advantage of firms by using the concept of “central city” in their static models considering the territory in a passive form.3 This clearly appears, for example, in the Christaller (1933) assumption that larger cities can support a wider range of activities, and in the hexagonal market formalized by 3 Territory, in those pioneer formalizations, was homogeneous and isotropous (i.e. the same in every direction). In other words, the basic concept of land space was that of the endless plains of the central USA.
332
D. Marino, R. Trapasso
L¨osch (1954), where some specialized economic activities can be undertaken only at a limited number of sites.4 Both the models of Christaller and L¨osch considered a manufacturing sector which sells its products to an agricultural sector. Accordingly, this kind of approach was not able to describe the circular feature of production in which some of the demand for manufacturing commodities comes from the manufacturing sector itself (commodities produced using other commodities). Empirical evidence shows that the presence of a well developed, strongly localized, manufacturing sector is attractive for other firms of the same sector or production chain.5 This dynamic can be summarised with the expression “circular causation” utilized by Myrdal (1957) to describe a self-fulfilling process in which a given location starts attracting firms from a certain dimension of its manufacturing sector. The circularity of the process is due to the “backward and forward linkages” (Hirschman 1958), that link firms to each other.6 Furthermore, the physical proximity to suppliers and seller makes for lower transactional costs (Coase 1937). The next step in the theory was to recognise the evolutionary nature of external economies. Vernon (1962), having analysed the New York productive framework during the 1950s, stressed the “rise and spread of external economies”: new sectors are localised in central areas because they need a high concentration of positive externalities. The standardisation of the production reduces the need for a specialised external economy and thus firms leave the expensive urban centre and locate in the periphery of metropolitan areas. The last issue was to discover the way in which a territory was able to achieve the right concentration of (manufacturing) firms to start a self-sustaining process of circular causation. Only in the early 1990s did economists find a sound theoretical basis for the empirical evidence by modelling a system of “monopolistic competition” (`a la Dixit-Stiglitz) and, so, consider the “increasing returns of scale” which firms gain by choosing (or by being in) a particular region.7 4
It is important to note that neitherthe formalisation of Christaller or Losh gave any explanation for the development of the central city, which existed “by default”. 5 The existence of strong relationships between clusters of firms in a well-defined territory was first discovered during the 1920s, as a consortium of economists of Columbia University analysed the collocation of firms and industries in New York. They discovered that standardisation of output played a remarkable role in location decision of agents. Firms with a low level of standardisation operating, for example, in the fashion sector, were located in the centre, closely related to their suppliers or sellers by wide use of face-to-face relationships. On the contrary, firms with a high level of standardisation and vertical integration (cooperage is the original example), were located in the outskirts of the city. 6 “The economies are external in the sense that the firm obtains them from outsiders, and they are economies in the sense that the firm can satisfy its variable or part-time needs in this manner more cheaply than it could satisfy them from within. The outsiders, in turn, can afford to cater to the firm’s fractional needs because they also cater to many other firms” Hall (1959). This kind of inter-firm relationship, under some particular conditions (high level of environmental trustiness, strong meso-institutions, etc.), can be so strong that firms start to externalize their “Value Chain”, forming what some scholars call a “Value Constellation” (Lorenzoni 1990). 7 We are referring to the contributions of Fujita, Krugman, and Venables, among others, in the creation of the so called “New Economic Geography”.
The New Approach to Regional Economics Dynamics
333
Specifically, the three fundamental conditions are: 1. The manufacturing sector has to employ a large proportion of the local population in order to generate large local demand 2. The sector has to be characterised by large economies of scale 3. Low transportation costs When these conditions are satisfied, a region (or an urban area) with a large local market and large availability of goods and services will attract population from regions whose economic frameworks don’t have such as characteristics (or have then in a less intensive form). In other words, territories start competing against each other to attract manufacturing activities. The approach to agglomeration seen above (New Economic Geography) can be useful to assess some long run dynamics. Indeed, when a broad temporal horizon is considered (i.e. starting from the Industrial Revolution) the importance of cheaper transportation costs in the development path of agglomeration is clearly appearant. However, “circular causation” seems to reduce dramatically when a shorter period (e.g. from the 1970s) is considered. Given that transportation costs were in a constant decreasing trend, empirical evidence seems to suggest a U-shaped relationship between the level of agglomeration and the cost of transportation, as shown in the figure below.
dispersion
dispersion
max agglomeration Very low
Very high Transportation costs
Fig. 1 Impact of trasportation costs level on urban agglomeration
334
D. Marino, R. Trapasso
This dynamic can be explained theoretically by considering a system in which firms produce both for other firms and for the agricultural sector: when transportation costs are very high firms disperse to meet the demand of farmers in every region, on the other hand, if the cost of transportation is very low firms disperse, because of easy access to other firms and consumers. However, this formalisation assumes the intra-city transportation cost to be zero and the inter-city transportation cost to be positive. In other words, it is useful only to understand the conditions in which agglomeration arises in a given large region.
3 Heterogeneity of Agents Regions are often the location of a complex structure of heterogeneous agents acting in different ways. Agents do not actually optimize a common utility function and they do not share a common endowment of perfect information. Conversely, agents are part of a complex system and every agent (or group of agents) evolves toward unstable equilibria in which they adjust their strategies and their expectations continuously. Strategies and expectations together change the environment itself. Accordingly, the path toward the equilibrium point (or the linear dynamic of growth, as in the neo-classical Solow formalization) is only one of an infinite number of patterns in which the system may evolve. In this situation, even small changes in some variables are able to change the system from one pattern to another (an emergent structure). As Arthur recently stressed (Arthur, 2005) a dynamic like that has three main features: – Perpetual novelty: there is a constant incentive to evolve (while according to static economics, agents should not have any incentive to move from the equilibrium once it is achieved). – Equilibrium indeterminacy and a selection process that means the evolutionary path of the system is not given and even small variations can change the intensity or the direction of the vector field. – Expectational indeterminacyand inductive behaviour. In static economics, agents try to form their expectations about an outcome that is a function of their very expectations: a self-referential situation. With rational expectation the problem remains; indeed to avoid the onset of multiple equilibria, all the agents should adopt the same base theory (i.e. based on the same assumptions), which is at least a very special event. Accordingly, complexity theory can be regarded as an emerging paradigm for understanding the complex dynamics underlying processes in regional economics, as, according to our definition above, regions are complex systems made up of many interacting parts. Complex systems can be described as a graph with nodes (elements) and edges (interactions). The number of interactions that exist between elements can define complexity. Accordingly, it is a function of the number of elements (N) acting in the defined domain.
The New Approach to Regional Economics Dynamics
335
Complexity ranges thus from a maximum level of N elements or agents generating N(N-1) interactions (assuming that interactions are not necessarily mutual) to a minimum of complexity in which there is only one agent (or a group of agents – firms and households) without any direct relationship (or with direct and linear relationships). However, empty graphs cannot really be considered systems because the elements have no relations with other elements. Agent interactions can also have differing degrees of intensity, they can be weak or strong, and usually intensity of interaction is a function of proximity to different agents. In this way it is possible to describe a pattern of interactions between elements along a continuum (instead of using a dichotomy approach). For instance, it is possible to use a range in which 0 represents the absence of interactions, and 1 represents a point of the system that is fully connected to the others. This possibility is particularly useful when a geographical area is taken in account, since geographic proximity is an important generator of mutual interaction. Nonetheless, it is also possible that some interactions are strong and effective over a long distance.8 This methodology allows the use of a single parameter for studying complexity. Hence, the latter should not be mistaken for complicated models with many parameters and multiple behaviour patterns (Axelrod 1997). There are three main approaches to model complexity that satisfy the conditions imposed above: Fitness landscapes or Adaptive landscapes, Complex networks, and Percolation.
3.1 Fitness Landscape Models In evolutionary biology, fitness landscapes or adaptive landscapes (Wright 1932) are used to visualize the relationship between genotypes (or phenotypes) and replicatory success. It is assumed that every genotype has a well defined replication rate (often called fitness). This fitness is the “height” of the landscape. Genotypes which are very similar are said to be “close” to each other, while those that are very different are “far” from each other. The two concepts of height and distance are sufficient to form the concept of a “landscape”. The set of all possible genotypes, their degree of similarity, and their related fitness values is then called a fitness landscape. A typical formalization is the NK-model. Every component of the system has an “epistatic” relationship with the other components or elements.9 In other word, each agent affects all other elements through a particular property.
8
Storper and Venables (2003) developed a model in which the diffusion of information (intellectual spillovers) depends on face-to-face interactions of agents. Accordingly, geographical contiguity plays a fundamental role in developing some particular sectors in which knowledge evolves quickly. For a deeper assessment of the role of face-to-face interaction in spreading innovation, see Maggioni M.A. Roncari S.N. in this publication. 9 In biology epistatic relationships refer to the case in which the action of one gene is modified by one or several genes that are classed independently. The two genes may be quite tightly linked, but their effects must reside at different loci in the genome. The gene whose phenotype is expressed is said to be epistatic, while the phenotype altered or suppressed is said to be hypostatic.
336
D. Marino, R. Trapasso
In the formalization of Kauffman (1993) each element of the system (where N is the total number of elements) is affected by K other elements. Through this model it is possible to simulate the effects of epistasis by constructing a fitness landscape. The original model deals with technology, and fitness landscapes are used to refer to efficiency or quality (for production process, and for products respectively). The fitness value W of a certain strategy s is calculated as the mean of the fitness values wi of each element i. 1 N W (s) = ∗ ∑ wi (s). N i=1 This model analyses mutation in the system due to epistatic relationships between the elements. If K = 0 there are no epistatic relationship and wi has only two random values 0 or 1. When the epistatic relationships are at their maximum level (K = N − 1), any mutation in a single element will produce new fitness values for each element within the system. It is important to note that in the case of clusters of epistatic relationships, the system tends to develop a variety of local equilibria at different heights. If the information is moderately complex, the level of equilibrium reached through a local search (within the epistatic cluster) will be quite efficient, and the level of local equilibria (on average) could be quite high. On the contrary, if the information is complex, the local search carried out by the cluster could be insufficient to generate a high equilibrium and the local search (or research) will be inefficient.
3.2 Complex Network Models Complex networks are related to the idea of many agents connected in different patterns and with different intensities. The proprieties of networks are measured by using two fundamental dimensions: the “cliquishness” or local density of the network X ( j, l) 1 , C= ∑ ∑ N i j,l∈Γi Γ ( Γ −1)/2 (where Γi is the set of neighbours of agent I and Γi is the size of neighbourhood, while X can be either 0 absent – or 1 – present); and average path length between any two agents: d(i, j) 1 , L = ∑∑ N i j=i N − 1 (where d(i,j) is the shortest path between I and j). According to these two properties, the formation of a cluster of the closer (or less distant) elements is highly probable in complex networks.
The New Approach to Regional Economics Dynamics
337
3.3 Percolation Models Percolation Models refer to the movement and filtering of fluids through porous materials. In others words, they concern a stochastic dynamic of a phenomenon that can evolve in an environment that is able, in turn, to influence the dynamic. In economics percolation has been used to model the transmission of information in a given environment. It is mostly based on the concept of phase transition: a change of a given condition in the agent, or in the system, causes the agent to “jump” from one state into another. Broadly speaking, every step in the evolution of the system is influenced by the previous one, generating path dependence.Such a system has a huge number of asymptotic states, and the initial state (Time zero), unforeseen shocks, or other kinds of fluctuations, can conduct the system in any of the different domains of the asymptotic states (Arthur 1988). Accordingly, the concept of a self-reinforcing mechanism can be expressed as a dynamic system, with path dependence and positive feedback, which tends to lead to a large variety of asymptotic states. Furthermore, the system selects the state in which it places itself. Such dynamics are well known in physics, chemistry and biology and the final asymptotic state it is called the emergent structure. The concept of positive feedback is relatively new in economics. Indeed, economics generally deals with problems of optimal allocation of scarce/insufficient resources, thus feedback is usually considered to be negative (decreasing utility and decreasing productivity). Path dependence, in turn, is the main characteristic of sf-reinforcing mechanisms (the other being multiple equilibria in the system, possible inefficiency of the equilibrium, and lock-in).The next section focuses on this approach and shows two different applications of it.
3.3.1 Path Dependence as an Allocation Process It is not possible to define precisely the dynamic occurring in a system which has the tendency to lock-in in a specific equilibrium, given the existence of multiple equilibria and a sf-reinforcing mechanism. Nonetheless, it is possible to define a system which has some characteristics that allow broad classes of analytical systems to be designed that encompass large number of examples. First of all, to avoid excessive complexity, the system should follow the linear sequence in which choices are undertaken. Second, the proportion of groups of feasible alternatives influences the choice itself (a concentration of alternatives in a particular group at a particular time influence the choice of the system). Finally, a self-reinforcing mechanism usually begins from a “balanced” but unstable position, thus the end-state can be determined by both the initial conditions of the system and by small events outside the model. In this case, a small variation in a given exogenous variable could cause a catastrophic effect on the entire system. Therefore, the actual state of the system cannot determine the next position of the system, but rather the probability of the next action and then of the next position. Considering a general class of dynamic systems, it is possible to assess the dynamic of the allocation process. One of the possible applications of the allocation process concerns, for instance, the
338
D. Marino, R. Trapasso Probability Next agent chooses A
pı(x)
pıı(x)
Xı Proportion of A in the system
Fig. 2 Two illustrative Allocation functions for dimension K = 2 (Arthur 1988)
distribution of firms in K locations at a certain “event time”. The probability that the next firm will join category i is pi (x) where x is the vector of current proportion or firm location.10 That formalization allows us to determine p, at least implicitly. By taking only two territories (K = 2) into account, it is possible to show (Fig. 2) all the possible dynamics of the system graphically. In the graph, it can be seen the quantity of agents concentrated in the A region is influenced by the number of agents that are already there. Specifically, if the number of agents in A is larger than a given proportion xi , the probability that the next agent will decide to locate in region A will be higher. Therefore, the region A will attract more agents. On the contrary, if the number of agents in A is lower than the proportion xi , the probability that agents will choose A as their next location will decrease over time. It is worth noting that it is impossible to use the Strong Law of Large Numbers in this stochastic distribution of elements, since past distributions influence the dynamic of the system, while in the Strong Law increments are independent. In this dynamic process, each choice of the system is irreversible and the process must converge to one of the points p of the feasible allocations. System att + 1 = System att + the choice with the highest probability + a random exogenous dynamic Without the random exogenous variable the expected value of System at time +1 will be equal to the actual state at time +1 : (E(Xt+1 |Xt ) = Xt+1 ) , which is the “The vector of probabilities p = (p1 (x), p2 (x), . . ., pK (x)) is the allocation function that maps the unit simplex SK of proportions into the unit simplex of probabilities” (Arthur, 1988). 10
The New Approach to Regional Economics Dynamics
339
equivalent deterministic solution. The formalization assessed above is the pillar of many studies on the location of firms by a spin off process.11 In these models, new firms are added by “spinning off” from parent firms one at time. Accordingly, firms are added incrementally to regions with probabilities equal to the proportion of firms in each region at that time. Empirical evidence underpins this process especially in the high-tech/knowledge-intensive sectors. Every point of the unit simplex (the total of regions) may become an attractor point, so the system can converge to any point. In other words, “chance” dominates the dynamic completely.
3.3.2 Path Dependence with Recontracting Processes In the allocation process assessed above, choices made by the system are irreversible. But what happens if every time the system can “change its mind”, it decides to re-contract previous choices? To model this dynamic it is necessary to consider a Markov-transition in which the concentration of firms in region A influences the location choice of firms in region B which can change their location every time by “jumping” into the other region. The region that attracts more firms increases its probability of attracting the “next one” at time t + 1; hence, a self-reinforcing mechanism is still possible. To give a formalization, let’s imagine a case in which there are only two regions K(K = (A, B) = 2) and total population is T = 2N, with a state variable m. Accordingly, N +m firms will prefer region A, and N–m firms prefer region B. Since pAB (m) is the probability that a firm will change its location from A to B, and pBA (m) the probability that a firm willchange its location from B to A (at every unit of time), the probability P(m,t) of finding the system at state m at time t will evolves as: P(m,t + 1) = P(m,t)((1 − pAB(m) − pBA(m)) +P(m + 1,t)pBA(m + 1) + P(m − 1,t)pAB(m − 1). From which we derive the Master Equation: dP(m,t) = [P(m + 1,t)pBA(m + 1) − P(m,t)pBA(m)] dt + [P(m − 1,t)pAB(m − 1) − P(m,t)pAB(m)] (∗ ) which normalized to the variable x in the continuous interval (−1, 1), m , N 1 ε= , N P(x,t) = NP(m,t),
x=
11
See Cohen, 1976 or Klepper, 2004.
340
D. Marino, R. Trapasso
R(x) =
[pAB (m) − pBA(m)] , N
Q(x) =
[pAB (m) + pBA (m)] , N
yields the possibility of rewriting (*) in the form of a one-dimensional Fokker-Plank diffusion equation
∂ P(x,t) ∂ ε ∂2 = − R(x)P(x,t) + Q(x)P(x,t). ∂t ∂x 2 ∂ x2 By substituting diffusion functions R and Q to describe some specific transition mechanism, it is possible to study the evolution of P over time and its distribution. It is worth noting that in recontracting process dynamics, transitions remain constant over time, while transition magnitude decrease over time in the allocation process formalization To give another example, we can show a model that refers to this kind of dynamic in the labour market (Aoki 2003). By adopting the mathematical instrument of the master equation (also called Chapman-Kormogorov equation), it is possible to assess a stochastic dynamic in which heterogeneous agents face the same limitations in their mobility or in their possibility of being hired by some sectors of the economy.12 One of the first results that this kind of formalization gives is a stationary distribution of equilibria instead of a single stable equilibrium. Another feature of this approach is the possibility of considering workers with differences in work experience, human capital stock, geographical location, and the sector in which they work. The economy has Ksectors, and sector i employs a certain number ni , i = 1, . . ., K of workers. There are two “states” in which a sector could be: the first is the “normal time”: y i = ci n i . In this situation the sector produces an output that is equal to the demand expressed by the market for the sector’s commodities. In the second case the demand is higher than the level of supply, and the sector goes into overtime capacity; with the same number of workers producing a higher output than before: yi = ci (ni + 1). Demand for goods i is given by siY , with K
Y = ∑ yi i=1
and si is a positive share of the total output Y referri g to goods produced by sector i with ∑i si = 1. Every sector has the excess demand defined by: 12
The model refers to the entire dynamics in the macroeconomic environment but here we refer to the part of labour market.
The New Approach to Regional Economics Dynamics
341
fi = siY − yi with i = 1, 2, . . ., K. Sets of sectors with positive and negative excess demand are denoted by I+ = {i : fi ≥ 0} ; I− = {i : fi < 0} . (∗∗ ) Changes in Y due to changes in any one of sectors affect the excess demand of all sectors. The model uses (∗∗ ) as proxy to indicate which group of sectors is profitable (and thus whose production it wants to expand), and, conversely, which one is unprofitable (and whose production it tries to reduce). According to the model, only one sector can adjust its production up or down by one unit at any given time. The sector with the shortest sojourn time will be the one to jump first (because of path dependence). And so dynamics are only determined by the transition rates in continuous-time Markov chains. Distance between different sectors is defined by using ultrametric distance. Therefore, the economic environment is structured as a tree diagram in which every sector is a “leaf” which is connected to the rest of the tree through “nodes”. Transmission of economic shocks in the environment depends on distance between leaves and branches. The distance is measured between “nodes”.
2
3
2
1
Ultrametric distance d(i, j) has the following properties: a. It is positive unless i = j (in which case it is zero) b. It is symmetric d(i, j) = d( j, i) c. It satisfies d(i, j) ≤ maxk {d(i, k), d( j, k)}
342
D. Marino, R. Trapasso
Every sector in overtime fills its vacancies (if there are no vacancies the overtime condition creates them) with workers laid off by itself or by the other sectors of the economy. Obviously, workers belonging to the hiring sector have more possibilityof being hired than workers belonging to more distant sectors.The distribution of the stochastic probability that a certain worker of a certain sector will be hired by a sector can be assessed by using the master equation. Ultrametrics can also be introduced as dummies for institutions and other kind of “special agents” whose actions can influence the system as a whole.13 Accordingly, the analysis can be used not only to forecast the evolution of the system sic rebus stantibus, but it can also show which are the main attractors in the system. Another important result of this approach is that it may be helpful to design policies taking into account other variables characterising the contemporary economy such as natural and environmental resources, human resources, and technology. Furthermore, incorporating these factors into the model does not increase the complexity of the mathematical instrument. This specific issue is broadly analysed in the next section.
4 Economic Policies in Spatially Extended Systems: New Paradigms Description of the evolution of spatialised economies emphasizes the role of new rather than classical paradigms. New factors seem to have replaced land, work and physical capital. Natural and environmental resources, human resources and technology are beginning to get the upper hand following the “technological revolution”. Co-operation within businesses and between businesses and business systems takes place on a vertical and horizontal scale in which the local dimension and the territorial variables constitute the catalyst for processes of development. Technological expertise and social capabilities (Latella and Marino 1996) are the basic elements capable of explaining the different levels of development seen in different territorial contexts. Territorial variables, in other words, are decisive factors in explaining development differentials, especially when they are associated with the idea of the market conceived as a social construction. This new market requires rules that will guarantee its smooth running given that access rights, exchange mechanisms and opportunities for distribution of the wealth generated not only do not re-assemble uniformly and autonomously in time and space (Sen 1984, 1985), but almost always require outside intervention to achieve the objectives set for development policies. Re-equilibrium policies thus appear necessary to guarantee a more equitable development process. Within the market it is necessary to define collective rules ensuring that positive dynamics (increasing returns) can develop through the interaction of the agents operating in it. The territorial dimension and the systemic nature of
13
The role of institutions in regional agglomeration dynamics is assessed below in section 5.
The New Approach to Regional Economics Dynamics
343
the production process are fundamental elements to understanding and governing development processes. Public intervention in such a scenario cannot simply be thought of as a mechanism for allocating resources within the economy but must assume the role of guide and director of processes. It must taking the shape, on the one hand, of a set of actions aimed at defining and guaranteeing individual access rights and, on the other, of interventions aimed at developing the exchange capacities of markets and business systems (Bianchi 1995). An explanation may be sought in the fact that local communities increasingly interact with the rest of the world in a continuous process of integration and globalization without necessarily responding to stimuli from the central state. This obliges us to re-examine the composition of the economic policy maker’s “tool box” and, at the same time, forces us to radically rethink the very meaning of government policies, given that the central public authority is no longer able to guarantee the development of the local community in the presence of particular actions enforced by the central authorities (Bianchi 1995). Traditional economic policies lose their capacity to produce the expected results when enforced in the context of an open market or of a market characterized by strong interrelations between agents, because the mechanism of response to the policy maker’s input has to deal with a system characterized by high levels of interrelations between individual decisions and which therefore displays collective response characteristics which are different from individual response mechanisms. The consolidated logic of public intervention in economics assumes that the government authority will identify objectives for which the instruments most likely to achieve results (which can be verified and therefore simulated) are chosen. Traditional macroeconomic policies only work if acting on a closed system for which it is possible to order objectives and priorities with certainty. In this case the policy maker can govern the system of underlying relations by assuming lineartype response mechanisms. If these assumptions are not verified, the complexity of the system makes traditional policies pointless; therefore, to govern a complex system policy-makers must equip themselves with a set of objective instruments and programming actions able to cope with non-linearity and the consequences of complexity.
4.1 Planning Actions in Spatially Extended Systems: Old and New Approach From the idea that an economy is a “complex evolving system” in which single individuals are linked to each other by strong relationships, it follows that dynamic characteristics cannot be represented by individual approaches but rather by collective properties subjected to subsequent non-reversible scansions (Arthur 1988). It is thus conceivable that each economic system, in its evolution, might manifest both a multiplicity of equilibria, each dependent on previous historical interrelations, and the presence of inefficiencies and lock-in which can be selected during
344
D. Marino, R. Trapasso
the evolutionary course of the system to the detriment of possible efficient solutions. Government of an economy seen as a complex evolving system therefore excludes the possibility that commands might be expressed with a prescriptive-type mechanism in mind, as would happen if the system being analysed were essentially closed and characterized by low levels of interactions between agents. To this must be added the considerable incidence of variables of a territorial nature. Territory cannot be thought of simply as a physical support for business activities but must itself become an active factor conditioning the exploitation of local resources and the capacities of single businesses to cope with international competition. Therefore, the general objective of regional policy becomes that of structural adjustment with a view to greater economic and social territorial integration. So new regional policy must firstly contemplate a “transactive” rather than a “prescriptive” type of approach and the basis for any action must consider not just “what must be done” but “in what manner, by what procedures and with whom”. This means making systematic and widespread use at all levels of the principle of subsidiarity which implies that decisions should be taken as near as possible to the problem and be appropriate to its solution, and individual responsibilities should also be identified using the same criterion. Thus the main task of decision-makers in each Spatial Extended System is to aim at reassembling the rules and re-establishing the access rights which are the basis of any subsequent action designed to re-appropriate local culture and raise the threshold of contextual knowledge. On these premises it is possible to imagine the transfer of outside knowledge and the creation of networks which build up the basis for the realization of a self-sustained model of development. To achieve these aims the Spatially Extended System (SES) needs to equip itself with instruments capable of identifying moments of participation and complementarity among all the actors that make up the local system. To do this opportunities must be created to allow the human resources to increase the know-how and acquired cognition that will qualify them to introduce innovative codes and routines within the productive system. If such cognitive improvement occurs, there will be an increase in flexibility and specialization and a greater capacity to understand and govern change and innovation and ultimately an improvement in the overall efficiency of the productive system. The government of a local system which is complex because of the continuous, strong interrelationships between the individuals operating within it cannot be of a deterministic kind unless part of it is isolated from the rest of the relationships. The government of a complex system demands a series of deliberations over interventions, which by their intrinsic nature are irreversible, i.e. they produce permanent changes in the state of the system. To return to the now extensively examined concept of SES, multiplicity of equilibria, co-operation, proximity, resilience and freedom of access can be pointed to as some important categories in the description and government of a complex system. The conceptual field within which the local system has to move is, in fact, of a bottom-up kind and provides the archetype for programming actions capable of leading the evolutionary paths of the SES towards states of greater growth.
The New Approach to Regional Economics Dynamics
345
Bianchi’s (1995) taxonomy of interventions identifies the following three procedures: 1. Programming according to exogenous concepts 2. Programming according to critical situations 3. Programming according to integration contexts Programming according to exogenous concepts is nothing more than the traditional concept of programming, achieved by means of the exogenous definition of objectives by the policy maker in conjunction with the identification of the instruments necessary to achieve the pre-established goals. If complexity and environmental turbulence are low, this method of programming is effective. This type of programming enters a crisis when the system enters those critical areas characterised by high levels of turbulence or uncertainty. In such circumstances it is necessary to programme according to critical situations, i.e. to devise programming capable of self-regulation in the presence of criticality and of varying parameters in order to overcome any lock-in or bottle-neck situations. As long as the critical areas are small in size, this approach is sufficient. If, however, levels of turbulence and complexity are so high that criticality can occur at any moment, then it is necessary to programme according to integration contexts, i.e. considering the system as a whole as an organism capable of adapting continuously to the outside environment. In this case policies have to take into account the changes they induce in the system itself, i.e. the way the system metabolises them. The need for programming according to integration contexts therefore justifies, as fundamental elements for regional policy, forms of structural adjustment whose objective is to lower the costs of transaction and which concern: • The social dimension, linked to the quality of life and culture • The ecological aspect, closely connected to the urban habitat, the landscape and the ecosystem • Public institutions and productive sectors, with special reference to the organizational aspect and the quest for efficiency Public-private co-operation, improved social standards, the construction of R&S networks and appropriate territorial policies designed to provide the basis for integration are irreplaceable instruments for governing the economy and for leading it to the highest levels of development.
5 An Outline of the Transmission Mechanism of Economic Policy in the Presence of Complexity The collective properties of a territorial economic system in relation to the link existing between productivity growth and information could be represented in terms of response function. We would like, at this point, to generalize the previous relationship by constructing an interpretative model which describes the propagation
346
D. Marino, R. Trapasso
mechanism of economic policy in a situation of complexity. The description of the transmission mechanism logically completes the previous observations regarding objectives and instruments. Single economic policy decisions, aimed at achieving the j-th objective through the use of the i-th instrument, can be represented as an outside stimulus which superimposes itself on interactions between agents. Agents in this approach are thought of as being spatially distributed and linked to each other by local mutual interactions (of a nearest neighbours type). We use H to indicate the effect of the economic policy. We can thus define an effective Heff stimulus which includes both outside stimulus and agent interaction.14 Obviously, without agent interaction H and Heff are equal. Heff therefore assumes the form: Heff = H +
drc(r − r )δ γ (r )
Where c(r-r’) is a function of correlation between agents which can constitute an acceptable means of modelling the concept of proximity, δ γ (r ) is a variation in the behaviour of agents induced by the policy applied, the integral can be linked to the concept of resilience. This type of behaviour arises in the area of a linear response model for systems with collective properties. The effect of an economic policy on a complex system made up of many agents interacting with each other can therefore be described in this way and modelled, as seen in the previous chapter, by means of the response properties of the system itself. Therefore, in the area of linear response theory we have a cause-effect relationship of the type: E(X) = G(X) ⊗ H(X) where E (X) represents the generalized effect, G(X) the response function, and H(X) the generalized cause. Therefore it is possible to study the generalised transmission mechanism of economic policy by describing the response function as a sort of susceptivity which comes to depend on the distribution of agents within the market. Obviously the type of response depends not only on distribution, but also on the type of interaction between agents.
6 Some Concluding Considerations The debate in economics between those who maintain that complexity and its causes play a decisive role in the construction of models with high levels of realism and those who think that a complete and exhaustive description of economic phenomena can be achieved by using linear and equilibrium-type models regardless of the complexity of the behaviour of agents and markets is relatively recent. In this work we analysed the relationship between complexity and economic policies from the 14
Heff represents the actual output of the implemented policy.
The New Approach to Regional Economics Dynamics
347
point of view of regional and territorial economics. The economy as a complex evolving system (Arthur 1988) therefore implies that: • Individuals are bound to each other by strong relationships • Dynamic characteristics cannot be represented by means of individual approaches but only by collective properties • Evolution manifests itself by means of multiple equilibria; • Each equilibrium depends on previous historical interrelations through possible inefficiencies and/or lock-in From a conceptual point of view, the main characteristics of the effects that emerge in the dynamic evolution of a system with complex behaviour can be explained by: • The difficulty prescriptive-type regional and territorial policies have had in promoting and sustaining economic development • The loss of importance of the national dimension: the local dimension clashes with the global dimension • The faltering view of economic policy and its propagation mechanism as being based on principles of command and control • The inability of a central planner to govern all the underlying relationships between economic agents at any given time according to linear-type response procedures
References Aoki M (2002) Modelling aggregate behaviour and fluctuations. In: Economics stocastic views of interacting agents. Cambridge University Press, Cambridge Aoki M (2003) A new model of labour dynamics: Ultrametrics, Okun’s law, and transient dynamics. working paper 2003 Arrow K, Hahn F (1971) General competitive analysis. Edinburgh, Oliver and Boyd Arthur WB (1988) Self-reinforcing mechanisms in economics, In: Anderson PW, Arrow KJ, Pines D (eds.) The economy as an evolving complex system. Addison-Wesley, Reading, MA Arthur WB (1989) Competing technologies, increasing returns and lock in by historical events. Econ J 99:116–131 Arthur WB (2005) Out-of-equilibrium and Agent-based Modelling. In: Hand-book for computational economics, vol. 2. Elsevier, Amsterdam Axelrod R (1997) The complexity of cooperation. Princeton University Press, Princeton, NJ Bianchi P (1995) Le politiche industriali dell’Unione Europea. Il Mulino, Bologna Brown DJ, Heal GM (1979) Equity, efficiency and increasing returns. Rev Econ Stud 46:571–585 Coase R (1937) The nature of the firm. Economica 4:386–405 Cohen AC (1976) Progressively censored sampling in the three parameter log-normal distribution. Technometrics 18:99–103 Christaller W (1933) Die zentralen Orte in S¨uddeutschland. Gustav Fischer, Jena David P (1985), Clio and the economics of QWERTY. Am Econ Rev (Pap Proc), 75:332–337 Dixit AK, Stiglitz E (1977) Monopolistic competition and optimum product diversity. Am Econ Rev 67:297–308 Fujita M, Krugman P, Venables AJ (1999) The spatial economy cities, regions and international trade. MIT Press, Cambridge, MA Hall M (ed.) (1959) Made in New York. Harvard University Press, Cambridge, MA
348
D. Marino, R. Trapasso
Hirschman AO (1958) The strategy of economic development. Yale University Press, New Haven, CT Kauffman SA (1993) The origin of order. Self-organisation and selection in evolution. Oxford University Press, Oxford Klepper S (2004) The geography of organizational knowledge. Presented at the 4th EMAEE Conference, Utrecht Krugman P (1998) Two Cheers for Formalism. The Economic Journal 106(451):1829–1836 Latella F, Marino D (1996) Diffusione della conoscenza ed innovazione territoriale: verso la costruzione di un modello, Quaderni di Ricerca di Base dell’ Universit`a Bocconi n. 2, Febbraio Lorenzoni G (1990) L’architettura di sviluppo delle imprese minori. Costellazioni e piccoli gruppi di imprese. Il Mulino, Bologna L¨osch A (1954) The economics of location. Newhaven, Connecticut Maggioni MA, Roncari S (2009) Learning, innovation and growth within interconnected clusters: an agent based approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Marino D (1998) Territorial economic systems and artificial interacting agents: models based on neural networks. Int J Chaos Theory Appl 3:23–28 Marshall A (1890) Principles of economics: an introductory volume, 1990 reprint of 1920 edition. Porcupine, Philadelphia Myrdal G (1957) Economic theory and underdeveloped regions. Gerald Duckworth, London Samuleson PA (1952) The transfer problem and the transport costs: analysis of effects of trade impediments. Econ J 62:278–304 Scarf H (1981) Production sets with indivisibilities. Econometrica 49:1–32 Sen AK (1984) Resources, values and development. Harvard University Press, Cambridge, MA Sen AK (1985) Commodities and capabilities. North-Holland, Amsterdam Storper M, Venables AJ (2003) The Buzz city The economic force of the city. J Econ Geogr 4:351–370 Vernon R (1962) Anatomy of a Metropolis. Garden City, New York Wright S ([1932] 1986), “The Roles of Mutation, Inbreeding, Crossbreeding and Selection in Evolution”, Proceedings of the Sixth Annual Congress of Genetics 1:356–366. Reprinted in Sewall Wright, Evolution: Selected Papers, William B. Provine (ed.). Chicago: University of Chicago Press, 161–177
What Policy for Interconnected Territories? Conclusions and Openings Lanfranco Senn and Ugo Fratesi
1 Aim of the Chapter The main argument of the whole book is that regions, however their functional boundaries are defined, grow with a virtuous process if the public and private agents of their territory maintain an equilibrated set of connections both inside and outside the region, and can hence be defined as interconnected territories. Some recent literature has defined the process by which the growth of regional territories is shaped not only by global forces but by their internal structure and governance, as ‘glocalism’(e.g. Swyngedouw 2000). Virtuous regions are in fact characterized by strong interconnections within the territory, with other territories, and also on the global scale. Only equilibrated combinations of local (internal) and global (external) interconnections can guarantee a virtuous cumulative growth process, inasmuch as strong internal interconnections (networks) not coupled with adequate external interconnections risk leading the territory into an implosive process; and strong external interconnections not coupled with adequate internal networking risk bringing about explosive growth (see Fig. 2 of the chapter by Fratesi and Senn 2009, in this book). Some attempts have been made to derive this intuition from more general models of regional growth. Most of these attempts adopted a macroeconomic and/or ‘phenomenal’ approach, whereas a full micro-founded model has not been developed yet. One model is the Endogenous Interrelated Growth model (Bramanti and Miglierina 1995), where the innovative capability of a region is made to depend on external connections, growth is made to depend on regional innovation, the generation of local economic and social connections is made to depend on growth and, finally, the equilibrium between internal and external connections within a region determines whether the region is able or not to maintain a sustainable growth rate (Fig. 1). Other attempts to show theoretically the need to balance internal and external connections are published in this book, for instance in the chapters by Bramanti and Riggi (2009), by Folloni (2009) and by Bramanti and Fratesi (2009). U. Fratesi and L. Senn (eds.) Growth and Innovation of Competitive Regions – The Role of Internal and External Connections. c Springer-Verlag Berlin Heidelberg 2009
351
352
L. Senn, U. Fratesi
External connections
Explosive path: Disintegration of the local system
Implosive path: Death by asphyxia
Local connections
Innovation
Growth
Fig. 1 Explosive and implosive paths in the Endogenous Interrelated Growth model. Notice that the relations need not be linear but, since we have no demonstrated indication of their actual shape, we chose to represent them as such
This theoretical intuition lacks sufficient empirical evidence at present. Tests may be carried out for the above mentioned or other models to demonstrate the existence of each functional relation contained in them, but the real objective for testing ought to be assessing empirically if the theoretical intuition of an equilibrium between external and internal connections being necessary for sustainable regional growth is corroborated. In fact, this would lead to the design of appropriate policies for equilibrated interconnections within and between territories, as summarized in Fig. 2 of the Introduction (page 16). In this short concluding chapter an attempt is hence made to suggest some of the possible fields of application rather than to provide empirical evidence. Seven are exemplified, related to as many important areas of policy (among the many possible) in order to suggest a number of cases where proving the intuition would have important policy implications. The issues (corresponding to the next sections of the chapter) are the following: • Promotion of interregional and international trade. • Attraction of foreign direct investments (FDI).
What Policy for Interconnected Territories? Conclusions and Openings
353
• • • •
Development of R&D and innovation. Tourism attraction and marketing. Management and governing the impact of large infrastructural projects. Promotion and consolidation of clusters, industrial districts and local production systems. • Regional and urban strategic planning.
2 Promotion of Interregional and International Trade Interregional trade (both of final goods and intermediate products) is important for the competitiveness of local economies. Regions are in fact too small to be selfcontained markets, and it is important for firms to be able to chose the most dynamic markets for their products. Moreover, to manufacture internationally competitive products, it is essential to select the best suppliers on an international basis. Re-positioning is hence essential for many sectors, especially for those most internationalized, dynamic, qualified and least protected. These same sectors are those that grow most and those in which openness is highest and increasing fastest. To allow the regional sectors to compete internationally, it is essential to have adequate local services, so that the local fabric can act as a stepping stone to allow local firms to participate effectively in the global competition. Policies of trade promotion should hence enhance the market information available to local firms. This can be achieved, for example, by sectoral economic analyses; by providing support to the marketing of firms by giving them a recognizable “local brand” by which the firms can be immediately recognized as coming from a reliable area and hence be perceived favourably by potential customers and suppliers. Export consortia are one of the tools by which this can be achieved. Also, services are helpful that support local specializations and, in particular, their integration into production chains. Of course, all business services directed towards quality (even though not directly concerned with trade promotion) are of paramount importance. If policies are only directed towards the internationalization of local firms, without targeting the local productive fabric at the same time, there is a risk that the most active firms will choose to re-locate outside the region, hence relegating the less active firms to the domestic market which is marginal at an international level but perceived as more protected and easy to understand. However, in a world with continually increasing market integration, even the protection of domestic markets is weak and decreasing, and the domestic market is insufficient in an international context: the growth potential of regions which are sheltered from the international markets is weak (Rodr´ıguez-Pose and Fratesi 2007). In this case, the regional economy can even die by asphyxia.
354
L. Senn, U. Fratesi
3 Attraction of Foreign Direct Investments (FDI) Foreign direct investments (FDI) increasingly choose between alternative possible destinations on the basis of a comparison between opportunities (e.g. incentives, labour costs, effective exchange rates) and the efficiency of territories and productive structures. Hence, the evaluation of localization factors takes place on a comparative and not an absolute basis, so that regions are in competition with other regions often located in a completely different economic and geographic context. Since attracting FDI means not only attracting capital and production, but also technology, best practices and knowledge of international markets and establishing new linkages (Markusen and Venables 1998; Lipsey 2002), a policy of openness to FDI best brings positive effects if it is able to actively select firms and sectors through policies of information and promotion of the opportunities of the region, essentially abroad. This is most effective if the network of local entrepreneurial association extends globally. Target sectors should not be selected on a purely external, exogenous basis – by choosing the internationally most performing sectors. Internal performance should take into consideration the characteristics and capacities of the local production system, so that incentives and facilitations go towards those sectors and firms which are best able to integrate locally and produce most development for the whole region. Policies of external attraction need to be complemented by internal policies such as: de-bureaucratization and shortening of document processing times; offering integrated, systemic services (enhancing the productive environment); i.e. liberalization instead of protectionism. Networks assume paramount importance since incoming FDI flows more easily when other firms from the investing country are already present, due to the perceived risk being lowered because of previous experiences (Buckley and Prescott 1989). Firm networks are also important because they allow fruitful exchanges with complementary firms located abroad. Without this complementarity (and despite all efforts by the policy-maker) it is probable that the incoming investors will be opportunist, since they are non-selected, only coming to plunder the public incentives or the short-term opportunities which are often market-related. These investments will impoverish the local fabric instead of strengthening it, so that when these investments are withdrawn, the region is weaker than before the arrival of the investors. Active policies are in any case essential, since without them the local characteristics are not enough to win the competition for FDI and hence the attraction rate diminishes in favour of the more aggressive areas/countries. This would be unfortunate for the region, since selected FDI can generate cumulative effects which go beyond directly generated employment to encompass employment in subcontractor activities at a first stage and an enhancement of the local technological and innovative capabilities at a second stage.
What Policy for Interconnected Territories? Conclusions and Openings
355
4 Development of R&D and Innovation The openness of a region to interregional and international connections favours the learning of best practices and the acquisition of technology developed elsewhere through imitation processes. Regions are in fact too small to develop internally all the knowledge needed to be competitive, but acquiring external knowledge is possible if and only if there are people and firms within the region able to receive, interpret and use the new knowledge (Bilbao-Osorio and Rodr´ıguez-Pose 2004; Malmberg and Maskell 2006). Hence, innovation policies should target many aspects at the same time: first it is essential to target the participation of regional firms and research institutions in international research networks, to enable local actors to access innovation developed elsewhere and to participate in frontier research. It is important to create real regional innovation systems, i.e. to make all the research and innovative components of the region act systemically in favour of regional growth and in particular of the strongest regional specializations, in a context which sees the region playing its specific role in the interregional and international distribution of poles of excellence. Innovativeness in itself is in fact not necessarily able to bring economic growth (Crescenzi and Rodr´ıguez-Pose 2009, in this book) and a cooperative attitude of the regional scientific and industrial worlds is a strategic pre-condition for an effective local knowledge-society (Camagni and Capello 2009, in this book). At the same time, it is also important to target the local “champion” sectors and try to attract selected “advanced” sectors. To do this, it is necessary to put in place internal actions which promote the research intensity of the region. Actions targeting regional R&D capabilities may involve: private and public training; universities; business-university connections; technology transfer centres; centres for the technological integration of production chains; policies for youth training abroad with successive repatriation; specialized technical conferences; participation in research events abroad. The most important factor in any regional R&D policy is the attraction, training and retaining of specialized human capital, including the most entrepreneurial and/or creative people (for the importance of these aspects for regional growth see Bramanti and Fratesi 2009; Garavaglia and Breschi 2009; Riggi and Maggioni 2009 and Sacco and Segre 2009, all in this volume). These people are currently the most mobile inter-regionally and internationally, so policies should facilitate the creation of contexts attractive to such people. Although the targets are clear, policies in this field are not easy to design, since thresholds exist below which research facilities cannot be centres of excellence. Moreover, research has to be specialized and the research specialization of a region needs to coincide with the regional productive specialization if it is to bring economic development. Finally, there are two risks: first, that research produced locally could remain within the research centre without spilling over to firms or, second, the outcome of research could flow too far and generate its growth effects outside the region. The latter case is more likely when the regional productive and entrepreneurial structure is weak and not able to take advantage of the locally
356
L. Senn, U. Fratesi
generated innovations. Hence, the need for a holistic approach to regional innovation policies.
5 Tourism Attraction and Marketing Incoming tourism has an exceptional multiplier effect on the development of the territories that host it. In fact, tourism is a service export activity (a territory which hosts tourists sells its services to a demand which is external–unlike most other service activities). This makes it acceptable to analyse the impact of tourism with export-led models; for example, tourism would belong to the base in an economic base model. The Keynesian multiplier of the economic base and Input-Output multiplier are high because the direct demand for services is very diverse: accomodation, transport, food consumption, cultural and environmental goods, and manufactured or crafted products such as clothing and souvenirs. If we look at the indirect demand of tourism services, sectors positively involved by tourism are even more diverse: agriculture, energy (such as petrol-derived products for transport, or electricity for accommodation), banking and insurance services (which are increasingly included in travel packages), industrial semi-manufactured components (such as wood for furniture in the hotels). A large part of these indirect effects are likely to remain in the region if the local fabric is strong enough to complement the external tourism attractiveness of the region and has the capacity to supply locally what the visitor requires. Hence it is clear that tourism, an external activity by definition, brings development to a region when it is coupled with a strong and diversified internal economic structure. In the case of policies, it is therefore important to complement external with internal actions. A marketing policy which aims at tourism development by attracting visitors (either for leisure or business) from other regions or abroad, is hence a policy of external linkage which is essential and should be made effective. As far as the public sector is concerned, this policy will involve the creation of means of increasing the visibility of the region and the information available to the potential visitor, and of means of coordinating all tourism reservations. The private sector ought to retain control of such activities as tour operators and airlines. However important it is for policies to attract tourism, these cannot be effective in generating regional development without internal policies that target the organization of the activities of tourists in the area once they have arrived. This should be done in an interesting and “creative” way: for example, by proposing that even less well-known localities host cultural attractions. It also means that the economy of the host region has to be ready to meet the demand not only of the front-office (direct) but also of the back-office (indirect and induced). A good hotel or a good restaurant needs to be integrated into an efficient supply chain. There should not be shortages of input to make up the tourism product on offer, nor of intermediate products, otherwise there will be bottlenecks and a lack
What Policy for Interconnected Territories? Conclusions and Openings
357
of prompt reaction. But it is also possible and necessary to organize “horizontal production chains”: for example, tourism itineraries integrated with a number of associated localities; hotel chains; organized inter-modal transport with single ticketing; cycles of events such as cultural festivals, or professional exhibitions and fairs. This coordinated “systemic” offer highlights the importance of a high degree of internal networking consistent with the capacity for external visibility which is instead dependent on networking with the localities of origin of the tourists. On the other hand, if the territory is very attractive, with high levels of equipment and interest because the local system has been activated effectively, but is unable to read and interpret the external demand, there is a risk of over-investing in some infrastructure or initiatives for which there is insufficient market demand. Thus, it is important to know the external market and to segment the tourist market by country of origin but also, increasingly, by age, personal income and social status. In this case, too, there is a need for internal/external equilibrium.
6 Management and Governing of the Impact of Large Infrastructural Projects All infrastructure, but in particular all complex infrastructure such as that for mobility or of telecommunications, effectively provide their function as enhancer of potential accessibility if they are well connected in networks. When a person has to move, or goods have to be delivered from one place to another, this inevitably involves some long-distance traits and some short-distance traits, i.e. long-range networks and short-range networks. For example, a businessman travelling from New York (Manhattan) to Tokyo (Ginza), will necessarily use a flight between the two airports (long extent network) but he will have to add a number of short-range local stages, such as the trip from home to the airport and from the airport to the city centre, involving various means of transport. The fluidity, the comfort and also the total duration of a trip will hence depend on the effective integration of long-range and short range networks. Should the long-range networks be efficient but, once arrived at the destination airport should there be difficulties in getting a taxi or a train to the city centre, the global accessibility would be strongly penalized. On the other hand, if a territory had a capillary local transport network, but, due to the insufficiency of long-range connections, these would not allow rich external markets to be reached, the territory would implode due to the impossibility of maintaining external relationships. The equilibrated construction of long-range and short-range networks avoids bottlenecks and allows a virtuous cycle of accessibility and development to be set up. In the opposite case, the accessibility would be limited to the gateways of cities and regions, the “hubs” - including airports, ports, train stations or multi-modal exchange facilities - but without an articulated system of origins and alternative destinations. In this case, the accessibility of the territory would be lacking despite high expensive investments in infrastructure.
358
L. Senn, U. Fratesi
To be effective, the integration of long-range and short-range infrastructural networks ought not only to concern homogeneous means of transport (e.g. roads with motorways or trains with metropolitan underground trains) but extend to different means of transport (intermodality) and to different operators, public or private. Physical infrastructure networks need the development and the integration of all transport (or telecommunication) that is offered in the network. Inefficient metropolitan services make it difficult to access airport flights, for example, being unable to be sure of catching a flight without leaving inordinately early. The same happens when the links and the integration of the means of communication or transport are lacking, which may cause bottlenecks, delays and queues in some air or rail services. All this can harm the development of a territory and hence we can say that the endowment of infrastructure is not enough if it is not complemented by effective management of services. Finally, territorial development can be distorted by long-range and short-range infrastructure that is not in line with the needs of the areas served. This occurs both if the infrastructure of the inter-modal channels are undersized (congestion) and if they are oversized (“cathedrals in the desert” phenomenon or stable under-use of infrastructure, which implies very high management costs of the infrastructure itself).
7 Promotion and Consolidation of Clusters, Industrial Districts and Local Production Systems Firms, especially those which are small or medium sized, find it increasingly difficult to compete alone in the global market. They therefore need to delay their foray into the integrated international markets until they have developed some forms of cooperation with other firms, which often happens with other firms in the same territory because belonging to the same area makes forming economic and personal relationships easier. On a macro scale, a correlation is often observed in the life of territories between density and sectoral integration on the one hand, and competitive success on the other, whereas when density and integration are lacking, the territorial system tends to decline. Co-operation with other firms can either occur horizontally, with other firms of the same sector, in order to generate external economies of scale or scope, such as in industrial districts, or vertically, with firms operating in the same production chain in order to have increasing efficiency and flexibility due to their links with subcontractors. Finally, collaboration and/or integration can also take place deliberately with a multiplicity of firms, especially of technological, productive and commercial services, in order to increase the effects of urbanization economies. At the same time, if firms unbalanced their collaborations and connections locally, this would be negative for two reasons: first, on the demand side, they could become too focused on the local market without a reasonable probability that their production could be absorbed locally, and lose their ability to interpret and predict
What Policy for Interconnected Territories? Conclusions and Openings
359
the trends in the international markets. Second, on the supply side, it is not always true that every input can be best obtained locally even if local connections remain important because they allow greater flexibility and interaction, to firms. Also in the case of connections between firms, therefore, local networks have to be combined in an equilibrated way with relationships with external firms. Local networks are important because they foster the competitiveness of local firms, because they ensure that a consistent part of the induct is local and because they tie firms to their territory, retaining them when other opportunities are available elsewhere. External connections, at the same time, are very important because they enable the arrival of market and technology information which is essential for the viability of local firms. The implications that this discussion has for policies is that policies should foster internal and external networking, without focusing on just one of the two, since it is only the complement of the two types of network that really enhances regional development.
8 Regional and Urban Strategic Planning With the openness of global competition, the self-sufficiency and the selfcontainment of urban and regional strategic planning are undoubtedly condemned to failure. Urban and regional competitiveness is increasingly characterized on the one hand by the identification and implementation of specializations in functions or sectors that are unique or rare in the world landscape, and, on the other hand, by the maintenance and development of a series of relationships between the region (or the city) and the rest of the world. These relationships are to allow exchanges, imitation and finally, to increase attractiveness. It is significant that the most recent rankings of world cities put their accent on relational indicators (flows, interactive) rather than on the traditional indicators of localized activities (stocks, static) as the most effective measurement of urban hierarchy (Taylor 2003). Moreover, administrative boundaries of cities and municipalities are increasingly outdated, since they rarely coincide with the boundaries of economic, social and cultural relationships that link the areas to the external world. The same also increasingly applies to the identification of functional boundaries of regions and countries, which is more and more problematic. For example: where do global cities such as New York, Tokyo or London end today? This is almost clear, or at least conventionally defined, from an administrative point of view, but not from a functional point of view. The firms from these global cities which have localized their branches abroad certainly still contribute to the competitiveness and attractiveness of their mother cities even though they are mainly localized around the world. This is true also for second order cities on a global scale: for example, the fashion sector of Milan makes it a “world” city for the commercial relationships of the firms linked in production
360
L. Senn, U. Fratesi
chains with the fashion industry of Milan, even though they are localized (especially with regard to production) throughout the world. From the point of view of regional and urban planning, this has made it necessary to conceive and implement development essentially strategic policies which on the one hand make the most of local strengths, but on the other also target international relationships. Planning an effective development strategy for a region or a city which aims at improving its position in the respective international rankings often involves attracting international events (sports, fairs, culture, etc.) and/or specializing strongly with respect to the principal competitors. The planning of infrastructure and international accessibility is also part of a winning strategic plan, as are policies for territorial marketing and international missions of local representatives. To sum up, the negative, implosive, implications of a planning self-contained in the internal boundaries are evident as are those, as much negative, explosive, of a policy only devoted to external relations, which ends up exporting its own talents without re-generating the internal network at the same time. For example, it is always possible to host a fair in a large city, but it is necessary that the internal networks work systemicly: all the local institutions in charge of all the relevant services, the environmental quality, the communication sector and the hotel hospitality need to collaborate with each other to prevent the single entities from gaining credit for themselves abroad without activating a systemic process of accumulation and growth for the whole city or region.
References Bilbao-Osorio B, Rod´ıguez-Pose A (2004) From R&D to Innovation and Economic Growth in the EU. Growth Change 35(4):434–455 Bramanti A, Fratesi U (2009) The dynamics of an innovation driven territorial system. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Bramanti A, Miglierina C (1995) Alle radici della crescita regionale: fattori, fenimeni, agenti. L’Industria 16(1):5–31 Bramanti A, Riggi MR (2009) Sustainable interrelated growth: a phenomenal approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Buckley PJ, Prescott K (1989) The structure of british industry’s sales in foreign markets. Manage Decision Econ 10(3):189–208 Camagni R, Capello R (2009) Knowledge-based economy and knowledge creation: the role of space. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Crescenzi R, Rodr´ıguez-Pose A (2009) Systems of innovation and regional growth in the EU: endogenous vs external innovative efforts and socioeconomic conditions, in Fratesi, U. and Senn, L. (eds.) Growth and innovation of competitive regions: the role of internal and external connections, Springer, Berlin Folloni G (2009) A model of local development. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin
What Policy for Interconnected Territories? Conclusions and Openings
361
Fratesi U, Senn L (2009) Regional growth, connections and economic modelling: an introduction. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Garavaglia C, Breschi S (2009) The co-evolution of entrepreneurship and clusters. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Lipsey RE (2002) Home and host country effects of FDI. NBER Working Paper 9293 Malmberg A, Maskell P (2006) Localized Learning Revisited. Growth Change 37(1):1–18 Maggioni MA, Roncari S (2009) Learning, innovation and growth within interconnected clusters: an agent based approach. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Markusen JR, Venables AJ (1998) Foreign direct investment as a catalyst for industrial development. Eur Econ Rev 43:335–356 Riggi MR, Maggioni M (2009) Regional growth and the co-evolution of clusters: the role of labour flows. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Rodr´ıguez-Pose A, Fratesi U (2007) Regional business cycles and the emergence of sheltered economies in the southern periphery of Europe. Growth Change 38(4):621–648 Sacco PL, Segre A (2009) Creativity, cultural investment and local development: a new theoretical framework for endogenous growth. In: Fratesi U, Senn L (eds.) Growth and innovation of competitive regions: the role of internal and external connections. Springer, Berlin Swyngedouw E (2000) Elite power, global forces and the political economy of ‘glocal’ development. In: Clark G, Feldman M, Gertler M (eds.) The oxford handbook of economic geography. Oxford University Press, Oxford, pp. 541–558 Taylor P (2003) World city networks: A global urban analysis. Routledge, London
Index
accessibility, 7, 11, 150–152, 170, 171, 177, 180, 181, 183, 186, 217, 357, 360 accessibility index, 181 adaptation, 65, 127, 290, 345 adaptive landscapes, 22, 335 agent-based approaches, 18, 20, 117, 118, 131, 141, 199, 297, 329 agglomeration, 3, 4, 8, 11, 13–15, 18, 22, 48, 61, 95, 101–105, 109–113, 129, 131, 133, 138, 140, 147, 150, 208, 245, 247–251, 258, 261, 270, 272, 273, 286, 298, 299, 305, 311–313, 315, 322, 331, 333, 334, 342 anchor tenant, 249 arts, 22, 281, 283–290 attractiveness, 32, 252, 356, 359 of regions, 4, 356 brain drain, 261 business-university relations, 11, 15, 355 capital, 7, 9, 11, 13, 15, 19, 31, 33, 34, 37–42, 45, 48, 55, 62, 64, 70, 100, 108, 109, 111, 145, 150, 151, 156, 159–161, 193, 195, 218, 221, 224, 245, 252, 258–260, 272, 273, 282, 283, 285–287, 289–293, 342, 354 accumulation of, 31, 34, 38, 218, 245 carrying capacity, 38, 70, 72, 76, 87 circular cumulative causation, 18, 19, 45, 104, 332, 333 clusters, 6, 7, 11, 12, 20–23, 55, 95, 96, 99–112, 117, 118, 120, 128–132, 136–141, 147, 149, 150, 156–159, 176, 245–251, 254, 256, 258, 260–262, 278, 285, 288, 292, 311, 331, 332, 336, 353, 358
differentiated, 20, 140 dynamics of, 18, 21, 118 emergence of, 95, 102, 103, 105, 107, 110, 112, 139 vertically-integrated, 281 co-operative sub-contractors, 60, 61, 66–72, 75–78, 80, 82–84, 86, 87 cognitive capabilities, 20, 147, 149 collective behaviour, 29 collective learning, 9, 14, 15, 39, 43, 62, 65, 145–147, 149, 150, 152 collective processes, 60, 151, 284 competences, 31, 33, 37, 111, 119, 138, 145, 153, 163, 290–293 competition, 3, 4, 6, 10, 14, 29, 46, 55, 84, 87, 88, 105, 123, 133, 137, 138, 141, 218, 221, 224, 227, 228, 230, 235, 245, 246, 249–251, 257, 260, 261, 270, 281, 289, 329, 331, 344, 354 for people, 289 on prices, 138, 249 competitiveness, 3, 9–13, 16, 19, 21, 22, 29, 32, 45–51, 55, 56, 61, 63, 70, 74, 85, 86, 95, 103, 106, 145, 153, 161, 187, 353, 359 of regions, 4, 7, 10, 12, 19–21, 60, 145, 161, 359 complex networks, 22, 335, 336 complexity, 18, 22, 113, 131, 290, 329, 331, 334, 335, 337, 342–346 congestion, 13, 182, 218, 248, 251, 261, 358 connections external, 10, 14–17, 20, 21, 31, 32, 245, 257, 270, 351, 359 internal, 16, 32, 39, 46, 49, 352 consumer surplus, 218, 232, 234–236
363
364 continuous space, 22, 298, 299, 308–310, 319, 320, 324 contractible investment, 199, 200, 205, 210 convergence, 6, 30, 40, 64, 183, 245, 246, 254, 257, 271, 276, 286, 288, 298, 331, 338, 339 cooperation, 8, 29, 33, 45, 47, 48, 55, 60, 66–72, 75–80, 82–84, 86, 87, 105, 107, 131, 149–153, 160, 163, 164, 217, 253, 342, 344, 345, 355, 358 creative class, 22, 33, 282, 284 creativity, 9, 12, 13, 21, 31, 33, 37, 64, 100, 149, 152, 281–286, 288, 291, 293 culture, 6, 12, 22, 29–31, 46, 59, 63, 100, 101, 106, 107, 109, 111, 145, 146, 149, 150, 154, 163, 270, 281–293, 344, 345, 356, 357, 359, 360 cumulative growth, 17, 351 customer-supplier relations, 66, 150 death of distance, 9 decreasing returns to scale, 35, 51, 52, 56, 72 delocalisation, 32, 37, 80, 84, 87 desertification, 261 development, 3–6, 12, 14, 18, 19, 22, 29, 30, 34, 36, 37, 42–51, 53, 55–57, 61, 62, 67, 70, 74, 78, 84, 95, 96, 98–100, 102, 104–120, 123–127, 137, 140, 146–156, 164, 167, 168, 170, 172, 173, 177, 180, 187, 188, 194, 217, 247–256, 260, 269, 271, 275, 281–293, 309, 310, 321, 329, 331–333, 342–347, 353–360 discrete space, 298, 299, 309 disintegration, 246, 351 divergence, 183, 245, 246, 254, 257 diversification, 6, 13, 123, 131, 139, 140, 155, 260, 270, 278, 356 ecological approach, 18, 21, 65, 245–248, 254, 256, 257, 260, 261, 345 econometric models, 18, 297 economic agents, 4, 6, 8, 10, 12, 14, 15, 18, 22, 29, 30, 46, 60–66, 68, 82, 85, 96, 98, 102, 109, 121, 128, 131, 136, 149, 160, 163, 170, 200, 201, 205, 210, 217, 269, 297, 299, 309, 324, 331, 332, 334–336, 338, 342–344, 346, 347, 351 economic base model, 7, 17, 356 economic geography, 3, 12, 103, 111, 177, 309 economic growth, 3, 14, 37, 59, 64, 95, 96, 145, 167–169, 172, 183, 186, 187, 194, 252, 271, 272, 285, 286, 289, 355 economies of scale, 8, 103, 104, 331, 333
Index efficiency, 3, 39, 118, 120, 122, 129, 140, 141, 148, 155, 157, 158, 221, 261, 330, 336, 344, 345, 354, 358 embeddedness, 3, 8, 14–17, 29, 31, 34, 37, 39, 77, 84, 85, 100, 107, 145, 150, 153, 162, 169, 180, 245, 246, 252, 258, 286, 329 endogenous growth, 8, 22, 30, 169, 180, 227, 245, 252, 273, 281, 285, 291–293 entrepreneurial alertness, 98–100, 107, 112 entrepreneurship, 4, 12, 15, 17, 19, 20, 33, 37, 45, 48, 55, 66, 95–103, 105–112, 119, 133, 152–156, 162, 164, 248, 249, 254, 286, 292, 354, 355 equilibrium indeterminacy, 334 evolutionary economics, 5, 8, 64, 80, 96, 103, 112, 170, 329, 330, 332, 334, 335, 344 excludability, 8 expectational indeterminacy, 334 exposure effects, 108, 112 external economies, 8, 13, 29, 30, 95, 102, 103, 110–112, 131, 138, 331, 332, 358 external pecuniary economies, 103, 104, 111, 112 external resources, 3 externalities, 13, 21, 102, 104, 110, 125, 145, 149, 150, 172, 208, 218–220, 222–225, 228–230, 234, 236, 248, 249, 251, 252, 270, 271, 273, 291, 323, 332 pecuniary, 102, 104, 106, 250 factors of regional growth, 9–11, 21 feed-backs, 18, 33, 43, 62, 63, 65, 69, 74, 99, 102, 269, 271, 272, 330, 337 mechanisms, 71 firms re-location, 17, 20, 55, 71, 84, 86, 108, 118, 125–128, 248, 249, 289, 353 fitness landscape models, 335 follower firms, 47, 60, 61, 66–68, 70, 72, 75–77, 79–84, 86 foreign direct investments, 14, 22, 352, 354 formalization of theories, 17 function of production, 33, 34, 72, 193, 221, 222, 270, 273, 277, 331 genius loci, 22, 32, 33, 288 geographic concentration, 6, 102, 145, 297, 300, 309 Gini index, 21, 274, 275, 277, 278, 298–305, 307–309 global cities, 359 global competition, 31, 289, 353, 359 globalization, 3, 4, 285, 288, 343 government, 6, 8, 63, 64, 108, 269, 270, 277, 288, 343, 344
Index growth differentials, 3 growth triggers, 4, 7 heterogeneous agents, 22, 117, 129, 139, 334, 340 human capital, 8, 11, 19, 31, 33–43, 46, 48, 62, 100, 102, 109, 111, 146, 149, 150, 153, 176, 252, 254, 258, 262, 269, 270, 282, 285, 293, 340, 355 externally-trained, 19, 33, 34 internally-trained, 19, 33, 34, 36 hypothesis-testing, 18 ICTs, 11, 21, 59, 62, 105, 108, 171, 193–201, 203–208, 211, 212, 217, 218, 221, 222, 224, 226, 229, 233, 235, 236, 277, 290 divide, 21 investment in, 21, 193–197, 202–204, 206–208, 210, 211 paradox, 196, 211 training in, 197 implosive processes, 351 increasing returns to scale, 12, 59, 72, 102, 104, 111, 272, 298, 330, 332, 342 incubators, 13, 108, 287 incumbents, 71, 105, 108, 109, 111, 248, 251 inductive behaviours, 334 industrial districts, 6, 8, 14, 16, 23, 42, 45, 48, 55, 56, 59, 96, 101, 102, 105–107, 149, 160, 169, 250, 353, 358 industrial structure of regions, 11 information, 9, 13, 15, 29–32, 37, 46, 48, 59, 62–65, 98, 100, 102, 105, 109, 112, 121, 122, 124, 128, 129, 137, 146, 149–151, 155, 171–173, 175, 193, 195–200, 204–206, 210, 211, 217, 219–222, 229, 235, 248, 249, 275, 284–286, 289, 290, 293, 309, 310, 312, 334, 336, 337, 345, 353, 354, 356, 359 diffusion of, 249, 335 information asymmetry, 98, 100, 150, 199, 248 infrastructure, 11, 13, 14, 61, 70, 75, 76, 102, 110, 170, 171, 175, 186, 194, 207, 217, 218, 225, 227, 236, 261, 272, 275, 277, 357, 358, 360 investment, 357 large projects, 22, 353, 357 research, 11, 75, 170 innovation, 4–16, 19, 20, 22, 29–34, 36, 37, 39, 42, 47, 59–71, 73, 75, 77–79, 81–87, 89, 91, 98, 100, 102, 104, 118, 123, 124, 137–141, 146–150, 154–160, 162–164, 167–173, 175, 177, 179–183, 186–188, 228, 245, 249, 250, 261, 269, 273,
365 283–285, 288, 293, 311, 335, 344, 351, 353, 355, 356 adaptive, 10 and local competitiveness, 59 averse societies, 167 incremental, 10, 31 prone societies, 167, 168, 170, 172, 175, 177, 179, 182, 186 radical, 11, 31, 147, 262 innovative effort, 66, 78, 168, 169, 171, 172, 178–180, 187 innovative performance, 11, 13, 87, 137, 171, 172 input-output relations, 11, 14, 15, 76 institutional economics, 12, 194, 258 institutions, 8, 12, 15, 16, 29, 30, 33, 37, 46, 63, 64, 69, 86, 96, 107, 108, 110, 112, 147, 152–154, 163, 164, 169, 170, 287, 322, 332, 342, 355, 360 public, 4, 39, 107, 108, 345 intangible investments, 37 inter-industry relations, 118, 141, 246 interactions, 3–5, 8, 9, 11, 12, 15, 20, 21, 29, 30, 32, 38, 39, 41, 42, 47, 59–67, 69, 88, 99, 100, 102, 103, 105, 109, 110, 112, 117, 118, 121–123, 126, 129–131, 145, 146, 149–154, 159, 162, 168–170, 178, 180, 182, 183, 187, 217–220, 229, 245–247, 251, 253, 256, 260, 269, 270, 272, 277, 285, 297, 320, 321, 323, 329, 334, 335, 342–344, 346, 359 interactive learning, 63, 121, 128, 140, 149, 152 interconnected territories, 16, 21, 168, 171, 172, 186, 217, 247, 251, 286, 351, 352 interconnection infrastructure, 21, 217 internal resources, 3, 82 international trade, 272 interregional trade, 14, 22, 245, 269, 272, 352, 353 intra-industry relations, 246 invention, 10, 69, 105, 119, 124, 125, 131, 137, 138, 140, 147, 149, 152, 290, 312 investment, 11, 21, 30, 31, 34–37, 49, 60, 61, 66, 69, 80, 105, 107, 147, 153, 154, 162, 164, 176, 179, 180, 186, 193, 195–207, 210, 211, 217–233, 235, 236, 270, 273, 277, 281, 286–292, 354, 357 incentives, 21, 217, 228, 229, 235, 291 non-contractible, 199, 205, 210 Kaldor, 18 kernel-regression approach, 298, 311, 315, 317, 319, 324
366 Keynes, 7, 17, 45, 356 Knight, 97, 99 knowledge, 3, 4, 7–11, 13–17, 20–22, 29–34, 37–39, 46–48, 55, 56, 60–64, 66–70, 72, 75, 77–86, 95, 97, 98, 100, 102, 103, 105–109, 111, 112, 117–133, 136–141, 145–159, 161–165, 167–173, 176, 180–183, 186, 187, 217, 219–221, 224, 226, 229, 231, 235, 247–251, 272, 273, 284–286, 289, 293, 335, 339, 344, 354, 355 barter, 20, 118, 121, 122, 128 codified, 11, 31, 32, 38, 39, 62, 63, 151, 152, 171, 180 creation of, 7, 20, 67, 69, 105, 138, 145, 146, 148, 150, 152, 153, 156, 163, 293 depletion of, 127 diffusion of, 61, 63, 102, 145, 150, 285, 286 external, 11, 16, 17, 47, 60, 61, 63, 66, 69, 151, 155, 168, 217, 355 flows, 14, 150, 168, 171 obsolescence of, 123, 138 tacit, 34, 38, 39, 46, 62, 102, 105, 111, 126, 151, 152, 171, 260, 293 transcoding of, 20, 152–160, 162–164 transfer of, 15, 63, 118, 121, 122, 130, 131, 136, 137, 140, 141, 273 knowledge-based economy, 95, 145, 146, 148, 161, 162, 167, 176, 286, 289 labour, 7, 9, 11, 13, 15, 21, 32, 45, 46, 48, 50, 55, 62, 67, 70, 72, 82, 83, 86, 102, 104, 106, 108–110, 125, 145–153, 164, 173, 174, 186, 187, 194, 196, 197, 204, 206, 208, 221, 245–254, 258–261, 272, 289, 291, 331, 340, 354 labour market mobility, 15, 21, 70, 176, 207, 208, 245, 246, 248, 252–254, 260, 261, 269, 270, 278, 288 poaching, 15, 61, 82 pooling, 13, 15, 247, 331 lagging regions, 4, 21, 66, 172, 179, 195, 208, 246, 261, 262, 271 land, 7, 70, 104, 249, 250, 331, 342 leader firms, 19, 30, 45–51, 54–56, 60, 66–72, 74–80, 82–87, 112, 179, 250, 293 learning, 4, 9, 14, 30–33, 38, 39, 41, 42, 59, 62–64, 69, 79, 80, 86, 87, 98, 110, 117–123, 126–131, 136, 138–141, 147–153, 164, 167, 169, 170, 173, 174, 194, 207, 208, 229, 260, 262, 289, 290, 355
Index learning by doing, 31, 38, 41, 117, 122, 126, 128, 207, 208, 260 learning by interacting, 38, 41 learning regions, 9, 14, 147, 167, 169, 170 Lisbon process, 147, 148, 167, 194, 289 local growth, 12, 19 local identity, 19, 30, 32, 47, 48 local innovation systems, 63, 102, 177 local production system, 7, 10, 23, 47, 54–56, 217, 353, 354, 358 local tastes, 108, 284 localism, 17, 151, 353 localization, 13, 18, 60, 102, 103, 112, 125, 131, 306, 354 economies of, 112 location, 3, 7, 8, 11, 13, 15, 22, 54, 67, 72, 75, 87, 102, 104–106, 109–111, 125–130, 146, 147, 150, 154, 157, 160, 168, 186, 207, 247–250, 258, 269, 271, 272, 278, 282, 285, 289, 297, 299, 300, 304, 306, 309–315, 319–323, 331, 332, 334, 338–340, 354 theory, 269 lock-in, 22, 103, 107, 235, 248, 337, 343, 345, 347 loops, 18, 33, 65, 74 macroeconomic approach, 17, 18 manufacturing sectors, 193, 308, 332, 333 market integration, 3, 15, 353 Marshall, 14, 55, 101, 106, 125, 247, 251, 260, 330, 331 meso-economic approach, 4, 6, 59, 170, 297, 309, 321, 332 methodological approaches, 17 microeconomic approaches, 18, 21, 131, 310, 351 microeconomic factors, 10, 11 milieux innovateurs, 6, 8, 13, 14, 16, 29, 32, 42, 48, 59–61, 64, 69, 70, 72, 79, 86, 87, 101, 102, 147, 149–152, 161, 169, 286 effect, 79 model calibration, 61, 68, 72, 80, 87 models, 3, 5, 7–9, 11, 13, 15, 17–19, 21–23, 25, 27, 45, 46, 65, 66, 72, 87, 97, 100, 104, 122, 129, 131, 141, 149, 151, 161, 169, 210, 217, 220, 221, 226–228, 235, 245–249, 254, 270, 271, 273, 277, 291, 293, 297–299, 309, 311, 319, 321, 322, 324, 329, 331, 332, 335, 337, 339, 346, 351, 352, 356 monopolistic competition, 218, 222, 232, 329, 332 multiple equilibria, 22, 68, 330, 334, 337, 347
Index national systems of innovation, 8, 63, 170 neoclassic economics, 222, 334 network externalities, 217–219, 221, 226–229, 231–236 networking, 4, 11, 14, 17, 21, 38, 64, 95, 195, 217–220, 227, 228, 233–235, 351, 357, 359 new economic geography, 8, 102–104, 251, 262, 308, 329, 332, 333 new industrial spaces, 9 non-material resources, 145 non-pecuniary factors, 96, 104, 106, 112 openness, 14, 16, 19, 20, 29–37, 39, 42, 43, 60, 61, 69, 70, 72, 75–77, 86, 87, 155, 245, 269, 353–355, 359 path-dependency, 10, 22, 103, 208, 329, 330, 337, 339, 341 pecuniary factors, 96 percolation models, 22, 335, 337 perpetual novelty models, 334 personal interactions, 38, 62 phenomena, 4, 9, 10, 15, 18, 22, 29–33, 35, 37–39, 41, 43, 48, 69, 96, 98, 100, 101, 110, 112, 117, 120, 123, 131, 136, 137, 141, 154, 156, 158, 194, 204, 207, 248, 269, 284, 288, 301, 309–311, 313, 324, 337, 346, 351, 358 polarization, 21, 270, 277, 278, 299–302, 304–308 policy, 3, 4, 6, 7, 11, 17, 20–22, 30, 31, 36, 43, 55, 56, 61, 86, 87, 110, 113, 137, 138, 141, 145, 146, 151–153, 164, 167, 168, 170, 172, 186–188, 195–197, 206, 207, 254, 261, 262, 269–271, 278, 282, 287, 292, 298, 329, 330, 342–347, 351–356, 359, 360 effects of, 17 experiments, 80 principal-agent models, 21, 195, 197 prior knowledge, 98, 100, 107, 108, 112 process innovation, 147, 154, 235 product differentiation, 218, 222, 227, 228, 232, 233, 236, 249 product innovation, 70, 80, 85, 147, 155–157, 159, 160, 162, 281, 292, 293 production factors, 7, 33, 35, 37, 39, 118, 125, 126, 128, 151 productivity, 10, 13, 21, 32, 37, 41, 48, 61, 147–149, 176, 193–196, 206, 209–212, 218, 221, 222, 224, 226–229, 231, 233–236, 252, 258, 261, 262, 273, 285, 303, 330, 337, 345
367 public/private partnerships, 151 reciprocity, 31, 46, 63 reduction of dynamic uncertainty, 8 region administrative, 6, 7, 297 definition of, 6 economic, 6, 7 regional disparities, 21, 187, 194, 195, 204, 206–208, 246, 261, 271, 273, 275, 278, 298 regional government, 15 regional growth, 3–5, 7, 9, 11, 13–15, 17–23, 25, 27, 45, 145, 167–170, 180, 188, 246, 254, 271–275, 278, 309, 351, 355 regional interactions, 21, 245, 251, 253, 254, 256, 260, 261, 271, 272, 274, 275, 277, 278 regional performance, 3, 7, 12, 13, 21, 138, 196, 245, 246, 257, 272 regional policy, 36, 170, 187, 271, 278, 320–322, 344, 345 regional science, 3, 7, 269 regional systems of innovation, 12, 14, 20, 85, 87, 101, 141, 167, 168, 170, 180, 187, 188, 355 relational capital, 147, 151, 152, 154–157, 159–162 relational development, 33, 36, 37 research and development, 10, 13, 20, 22, 47, 60–64, 66, 69, 70, 76, 77, 80, 82, 84–87, 117, 141, 146–153, 155, 156, 163, 164, 167–170, 173, 178–183, 186, 187, 220, 221, 226–229, 231, 233, 235, 269, 286, 353, 355 infrastructure, 75 into productive activity, 61 investment in, 13, 20, 47, 60, 61, 85, 148, 151–153, 164, 167, 168, 179, 180, 186, 221 resilience, 7, 20, 82, 141, 248, 344, 346 of the System, 80 robustness, 14, 19, 29–37, 42, 43, 54, 60, 86, 87, 277 routines, 10, 15, 33, 34, 37, 39, 41, 63, 111, 316, 323, 344 scale effects, 62, 150, 222 schooling, 11 self-reinforcing mechanisms, 22, 77, 96, 102, 103, 329–331, 337, 339 signalling, 140, 248, 249 skills, 4, 15, 21, 33, 37, 38, 62, 66, 70–72, 76–79, 82–84, 86, 100–102, 106–108,
368 110, 150, 152, 157, 160–162, 173, 187, 195–197, 201–204, 206–208, 245, 246, 249, 252, 253, 257–262, 285, 286, 289–291 social capital, 14, 15, 29, 64, 111, 291 social fabric, 10, 29 social filter, 20, 61, 63, 170–179, 181–183, 186–188 social networks, 14, 17, 64, 86, 95, 98, 100, 109, 111, 151 socio-economic factors, 10, 168, 178, 188 Solow, 19, 33, 34, 36, 42, 193–195, 252, 273, 277, 293, 334 spatial concentration measures of, 22, 298, 301 spatial econometrics, 22, 129, 182, 183, 269, 270, 301–303, 320 spatial interaction, 269, 277 spatial organization factors, 12 spatial segregation, 311, 313, 315, 324 specialization, 6, 30, 42, 45–48, 55, 102, 108, 118, 125, 131, 139–141, 147, 154, 157, 160–162, 249, 252, 258, 260, 270, 272, 331, 332, 344, 355 spillovers, 12, 15, 20, 46, 61, 62, 69, 75–80, 82, 86, 95, 101, 103, 106, 109, 122, 126, 129, 136, 139, 140, 147, 150, 168, 172, 180–183, 186, 187, 217, 219–223, 226–236, 245, 247, 248, 269–273, 278, 285, 287, 288, 298, 305, 331, 335 of knowledge, 13, 15, 17, 22, 56, 63, 68, 86, 95, 102, 103, 105, 106, 111, 112, 118, 121, 122, 125, 138, 145, 149, 150, 169, 180, 183, 186, 219, 226, 247, 249–251, 272, 273, 285 unintentional, 20 spin offs, 9, 96, 108, 110–112, 119, 120, 129, 130, 140, 141, 147, 149, 247, 248, 339 steady state, 36, 50, 52, 53, 218, 223–226, 228–231, 234–236, 331 strategic planning, 23, 353, 359 strong ties, 31, 34, 39 structural adjustment, 9, 10, 344, 345 sub-contractors, 47, 67, 70, 72, 74, 75, 77–79, 81, 82, 84, 86, 105, 108, 250, 354, 358
Index sustainability, 6, 7, 10, 19, 29–33, 35–37, 39, 41–43, 68, 70, 84, 167, 236, 285, 289, 351, 352 sustainable growth, 7, 10, 19, 29, 30, 32, 36, 39, 42, 351, 352 synergy, 29, 31, 42, 60, 151, 162, 164, 169, 187, 250 system dynamics, 60, 61, 65, 74, 224 tacitness, 11, 29, 34, 38, 39, 46, 62, 63, 102, 105, 107, 111, 126, 151, 152, 171, 260, 293 telecommunications, 14, 21, 217–224, 226, 228, 232, 233, 235, 236, 357 usage of, 21, 218, 219, 226, 229, 233 territorial systems of production and innovation (TSPI), 19, 29–33, 35–37, 39, 40, 42, 43, 59–72, 75–80, 82, 84, 86–88, 169 endowed with different R&D, 75 fitness and policy, 80 showing different levels of milieu, 79 with different degrees of openness, 76 total factor productivity, 21, 62, 194, 218, 221–226, 229, 231, 232, 235, 236 tourism, 22, 287, 353, 356 transmission mechanisms, 346 trust, 8, 14, 15, 29, 31, 46, 62–64, 67, 82, 87, 102, 109, 150, 171 universities, 11, 15, 29, 30, 59, 64, 69, 75, 95, 108, 153, 155, 170, 308, 312–314, 316, 332, 355 untraded interdependencies, 13, 46 urban structure, 13 urbanization economies of, 358 usage of, 219 venture capital, 11, 110, 249 wages, 21, 33, 66, 71, 72, 76–84, 86, 101, 103, 104, 108, 195, 208, 246, 252–254, 258–262, 291 welfare, 21, 117, 137, 218, 222, 232, 234–236, 261, 270, 308 world integration, 4, 5