David F. Batten, Cristoforo S. Bertuglia, Dino Martellato, Sylvie Occelli Auth., David F. Batten, Cristoforo S. Bertuglia, Dino Martellato, Sylvie Occelli Eds. Learning, Innovation and Urban Evolution

LEARNING, INNOVATION
AND URBAN EVOLUTION

LEARNING,INNOVATION
ANO URBAN EVOLUTION
edited by
David F. Batten
Cristoforo S. Bertuglia
Dino Martellato
Sylvie Occelli
"
~.
Springer Science+Business Media, LLC

Library of Congress Cataloging-in-Publication Data
Learning, innovation, and urban evolution / edited by David F. Batten ... [et al].
p. efi.
Includes bibliographieal references and index.
ISBN 978-1-4613-7083-3 ISBN 978-1-4615-4609-2 (eBook)
DOI 10.1007/978-1-4615-4609-2
1. Cities and towns--Effect of technological innovations. 2. City planning. 3.
Information society. 4. Technology--Social aspects. 5. Technology-Economic
aspects. 1. Batten, David F.
HTI66.L4262000
307.76--dc21
00-040546
Copyright 2000 by Springer Science+Business Media New York

Origioally published by Kluwer Academic Publishers in 2000
Softcover reprint ofthe hardcover Ist edition 2000
AII rights reserved. No part of this publication may be reproduced, stored in a

retrieval system or transmitted in any form or by any means, mechanical, photo-
copying, recording, or otherwise, without the prior written permission of the
publisher, Springer Science+Business Media, LLC.
Printed on acid-free paper.

Contents
Preface VB
Acknowledgement ix
I. Learning, Innovation and Urban Evolution: An Introduction

David F. Batten, Cristoforo S. Bertuglia, Dino Martellato, Sylvie Occelli
PART I: LEARNING AND KNOWLEDGE EXCHANGE

Section A: Theories of Learning and Complex Dynamics
2. Innovation and Patterns of Learning: A Survey of Evolutionary

Theories 11
Mario Cimoli and Marina della Giusta
3. Emergence and Co-Evolutionary Learning in Self-Organised

Urban Development 45
David F. Batten
4. Nonlinear Dynamics, Innovation and Metropolitan Development 75

Dimitrios S. Dendrinos
Section B: Path-Dependent Processes of Knowledge Exchange
5. The Knowledge-Intensive Nature of Japan's Urban Development 109

Kiyoshi Kobayashi, Sotaro Kunihisa and Kei Fukuyama
6. Interurban Knowledge Networks 127

Martin J Beckmann
7. Innovation and Urban Planning 137

Britton Harris
VI
PART II: INNOVATION AND ITS SPATIAL IMPACTS

Section C: Economic and Technological Changes
8. Product Vintages and Specialisation Dynamics in a Hierarchical

Urban System 165
Ul/a Forslund and Barje Johansson
9. The Impact of Research Activities on the European Urban System 197

Stefano Magrini
10. Is the Real Exchange Rate Relevant for Determining the

Innovation Potential of European Regions? 221
Dino Martel/ato
Section D: Impacts of Innovations in Telecommunications

and Transport
11. Impact of the New Information Technologies on Economic-

Spatial Systems: Towards an Agenda for Future Research 237
Cristoforo S. Bertuglia and Sylvie Occelli
12. Telematics Innovation and the Development of Non-Metropolitan

Areas: Lessons from Policy Experience 255
Andrew Gillespie and Ranald Richardson
13. The Impact of Innovations in Telecommunications and

Information Systems on Urban Mobility 271
Ennio Cascetta and Bruno Montella
Index 299
List of Contributors 309

Preface
In the global information society, innovation does not only constitute a

technological issue. It is a highly pervasive process which influences all
facets of human life, whether cultural, economic, political or institutional.
Awareness of the impact of innovation on so many areas of urban life
prompted the research project which led to the production of this volume.
In order to understand change in our cities, classical notions of
innovation must be broadened to encompass a wide-ranging set of
interactions between the various constituents of urban life. A key element is
the learning process: at the core of urban evolution is the complex interplay
between the local dimension of learning and the city or region's collective
propensity for innovative change. Aspects of the nexus: learning-
innovation-urban evolution, have been examined here by a number of
experts from different parts ofthe world, all actively involved in research in
the urban field. Their willingness to participate in our joint investigation of
this complex theme has been much appreciated.
The editors would in addition like to express their thanks to Angela
Spence, not only for co-ordinating the various stages of the books
preparation, but also for her excellent technical suggestions and careful
linguistic editing, in which she was ably assisted by Jennifer Wundersitz.
Many thanks are also due to Franco Vaio for his expertise in dealing with the
formatting and the production of the final copy.
The Editors
The editors wish to acknowledge the generous support of the National
Research Council of Italy (CNR) for providing the financial backing
(contracts nos. 97.00185.PF74 and 97.001 86.PF74) which made the
production of this volume possible.
PART I:
LEARNING AND KNOWLEDGE EXCHANGE
Section A:
Theories of Learning and Complex Dynamics

Chapter 1
Learning, Innovation and Urban Evolution:

An Introduction
David F. Batten 1, Cristoforo S. Bertuglia2, Dino Martellat03 and Sylvie

Occelli4
J The Temaplan Group, Melbourne, Australia
2Turin Polytechnic, Turin, Italy

3 University Ca ' Foscari, Venice, Italy
4/RES-Piemonte, Turin, Italy
1.1 PREAMBLE
An important component of the current debate on the impacts of the new

information technologies is that their adoption does not simply constitute a
technological issue. Technological innovations are key factors in the
evolution of a great many aspects of human development: cultural,
economic, political and institutional, for example. Because of the pervasive
nature of these impacts, there is an urgent need for the classical notions of
innovation and technological analysis to be broadened to encompass a more
diverse set of interactions between the various evolving parts of a modem
society. Innovation is a product of human knowledge and learning, and
leaming is an interactive, dynamic process. In order to come to grips with
the pervasive nature of technological change, there is a need to adopt a
socio-dynamic systems approach.
Some of the pertinent impacts are well known. For example, the time and
space-shrinking potential of new information technologies are having a
profound impact on many kinds of urban interdependencies. Interpersonal
contacts, economic linkages, and many other functional and spatial
relationships in our cities are undergoing major transformations in response
to the new modes of communication and movement. Even our traditional
desire for face-to-face transactions is being challenged by a plethora of
distance-insensitive technologies. As a result of their adoption, distant
2 Batten D.F., Bertuglia C.S., Martellato D., Occelli S.
locations seem to draw closer. Geographical space is distorted as the ease of

communication between places grows. This compression of space is
spawning new classes of urban products, virtual marketplaces and new ways
of transacting sales.
But many of the impacts of socio-economic change at an urban scale are
less predictable. Human evolution does not follow a smooth path into the
future. Neither does urban evolution. The collective advances of society as a
whole show a definite alternation between long periods of consolidation,
during which change is rather gradual and predictable, punctuated by
relatively brief bursts of unexpected change. These eruptions are catalysed
by innovative ideas which often result in the onset of an entirely new regime.
Arthur Koestler has put the biological metaphor quite forcefully:
"Mental evolution is a continuation of biological evolution, and in
various respects resembles its crooked ways. Evolution is known to be a
wasteful, fumbling process characterised by sudden mutations of
unknown cause, by the slow grinding of selection, and by the dead-ends
of over-specialisation and loss of adaptability ... Moreover, there occur in
biological evolution periods of crisis and transition when there is a rapid,
almost explosive, branching out in all directions, often resulting in a
radical change in the dominant trend of development."l
There is plenty of evidence to suggest that we are presently in the midst
of such an era of rapid and explosive change. The powerful motor driving
this latest transformation of the post-industrial society is generally thought to
be technological. But such an explanation seems too narrow. Broadly
speaking, the age in which we find ourselves could be more aptly described
as a knowledge-intensive age on an international scale. 2 Knowledge-
intensive activities, such as research and development (R&D) programmes,
are being conducted with relative ease on a global scale. Scientific inputs
have a growing impact on processes of innovation, yet R&D itself has
become far more complex. The explanation may lie in the growing
sophistication of learning-by-doing among creative economic agents.
We live in an era of global knowledge exchange, facilitated by major
advances in our infrastructure networks. In this frictionless kind of world,
many innovations are by-products of collective exchanges between cities far
apart, simplified by the ease of transport and communication. Thus there is a
need for us to look more closely at various collective processes of learning,
knowledge exchange and innovation in a spatial setting. This is the primary
purpose of the present book.
See Koestler (1964, p.226).

2 Today's society has also been described as a 'cosmo-creative' one (see, for example,
Andersson et at 1993).
Learning, Innovation and Urban Evolution: An Introduction 3
Learning and innovation have a dual role in urban evolution. On the one
hand, they possess a cosmopolitan dimension in the sense that many creative
exchanges permeate cultural borders and rapidly take on a global
significance. This highly interactive part of the learning process is a
collective one, as mentioned above. It is often regarded as a public good
because it is available to all who can engage in it. On the other hand,
creativity has a private dimension, sometimes even a confidential dimension,
associated with the skills of an individual who produces outstanding kinds of
work - such as music, poetry, scientific ideas or technical innovations. This
private aspect introduces a personal dimension to creativity and innovation,
one that is intimately associated with the local environment.
The complex interplay between the local (or private) dimensions of
learning or innovation, and a city's collective (or public) propensity for
change, is at the core of urban evolution. It also highlights an emerging
paradox. As the literature on innovation and innovation diffusion
accumulates, our individual knowledge about its implications for urban
evolution appears increasingly fragmented, often irrelevant. The complexity
of the issues involved demands a special kind of analysis: a collective,
participatory, dynamic approach. Yet this introduces greater uncertainty
owing to its sensitivity to initial conditions, the importance of path-
dependence and the existence of multiple outcomes.
In a series of earlier books, closely-related notions such as cosmo-
creativity, technogenesis and communication networks have been explored. 3
The point of departure for the collection of papers in this volume is that they
conceive of innovation in an urban context as a broader set of interactions
between the various evolving parts of a modem society. Innovation is seen
as a product of human knowledge and learning, both collective and
individual, and learning is viewed as an interactive, dynamic process. In
other words, the chapters herein adopt a socio-dynamic systems approach.
In order to adopt such an approach, it is helpful to make a distinction
between those contributions which examine the dynamic processes of
learning and knowledge exchange (Part I), and those which examine the
nature of innovation as well as its economic and technological impacts on
urban evolution (Part II). This break-up emphasises the fact that learning and
knowledge acquisition are precursors to any kind of innovation. In short,
invention precedes innovation.
Also implied in this order is a progression from a set of theoretical
papers, through to an exploration of a set of dynamic processes, then finally
to an examination of a variety of impacts that follow from these dynamic
processes. This natural sequence: theory ~ processes ~ impacts provides a
See, for example, Andersson et at (1993), Bertuglia et at (1995), Batten et at (1995),

Graham and Marvin (1996), Bertuglia et at (1997).
4 Batten D.F., Bertuglia C.S., Martellato D., OcceIli S.
cohesive organisational structure for the book. It also recognIses that

innovation is a path-dependent process.
1.2 AN OUTLINE OF THE VOLUME
1.2.1 Theories of Learning and Complex Dynamics
The opening chapter of Part I, by Mario Cimoli and Marina Della Giusta,
provides a survey of the various evolutionary theories of innovation and
economic change. The authors follow Dosi (1996) in demanding a
consideration of dynamics, the presence of microfounded theories, the
assumption of bounded rationality and heterogeneity among agents, as well
as the recognition of the continuous appearance of novelty, the view of
collective interactions as selection mechanisms, and the consideration of
aggregate properties as emergent phenomena. In so doing, they develop an
alternative view of behaviour and learning on the part of firms. Their ideas
turn out to be embedded in broader national and regional 'systems of
innovation' which account for the persistent differences in technological
capacity. The importance of the institutional dimension for evolutionary
theories of innovation is thereby demonstrated.
The next chapter, written by David Batten, explores the theme of
emergence and co-evolutionary learning in urban development. It is argued
that cities and brains, as living organisms whose patterns of development
display traits typical of complex dynamic systems, have much in common.
For example, because they are both subject to co-evolutionary learning, each
can produce spontaneous unexpected outcomes referred to as 'emergent
behaviour'. The author examines these collective processes of emergence
and co-evolution, looking in particular at the interaction patterns within and
between individual groups of economic agents. One of his key conclusions is
that even very small changes in the behavioural ecology of a city's
inhabitants can have profound effects on its economy and its urban
landscape.
The aim of the chapter by Dimitrios Dendrinos is to state the effects of
technological innovation on metropolitan development as an example of the
application of the mathematics of nonlinear dynamics. The chapter shows
that the process of innovation can be regarded as an unexpected,
exogenously induced perturbation of paths, which can lead to phase
transitions or divergences in the dynamic trajectories. A substantive message
of the chapter is that technological innovation is likely to lead to further
spatial or sectoral dualism, due to the inherent instability found in such

highly interactive complex systems.
1.2.2 Path-Dependent Processes of Knowledge Exchange
In their chapter, Kiyoshi Kobayashi, Sotari Kunihisa and Kei Fukuyama

discuss the knowledge intensive nature of Japan's urban development. The
hypothesis they investigate is that location decisions are not simply a
function of the geographical endowment of places, but are path-dependent.
Thus the main driving force consists of agglomeration economies, which
operate through positive feedback and cumulative, self-reinforcing
processes. Here, their ideas support those of Brian Arthur (1994). As an
empirical example, the authors examine the factors which have influenced
location behaviour in Japan's Science City, Tsukuba. In particular, they
examine the dynamics of industrial laboratory location using a simple logit
model that permits a description of the lock-in effects. After reviewing
sources of increasing returns in regional knowledge economies, they manage
to identify the major factors determining laboratory location.
Martin Beckmann's chapter focuses on the production and diffusion of
pure basic knowledge as opposed to technological knowledge. Although
knowledge production is often seen to be a 'footloose' activity, Beckmann
points out that successful research organisations are found almost
exclusively in or near major metropolitan areas. This highlights the
importance of face-to-face contacts and the notions of centre and periphery
in knowledge exchange networks. The question of location is studied in
connection with the interaction between scholars and scientists, and suggests
that distance still matters, even in this seemingly friction-free age of
advanced telecommunications.
The final chapter in Part I of the book presents a picture of the complex
process of planning, its susceptibility to invention and innovation, and its
need for innovative content. Britton Harris reviews the last fifty years,
illustrating the development of our analytical understanding of the city, and
claiming that the current 'impasse' confirms the historical view that design is
an indispensable component of planning. His modern view of design entails
greater reliance on scientific support, as well as more complete engagement
in organised collective activity. It is necessary to redefine the public process,
going beyond public conflict of interest to public participation. Harris argues
that this will require major changes in the respective roles played by
analysts, practitioners and the public, as well as a daunting new stock of
knowledge and skills.
1.2.3 Economic and Technological Changes
In their chapter on product vintages and specialisation dynamics, Ulla

Forslund and Bo1je Johansson develop three complementary, but partly
competing, theories of how and why location intensities differ between
urban regions. These are used to analyse both static equilibrium patterns and
the dynamics of location intensities, thereby helping to explain the observed
regional leads and lags in the process of economic change. Their lead-lag
model is then applied to an empirical analysis of patterns of change in
Sweden's urban regions, which focuses on the spatial concentration of
activity vintages. The assumptions of the model are examined separately for
medium-sized urban regions or metropolitan areas and small peripheral
municipalities.
Stefano Magrini examines the role played by various factors, especially
R&D, in the recent economic growth of the European urban system. The
author formalises a theoretical model of endogenous growth which takes into
account the way technological knowledge diffuses across space. In so doing,
he distinguishes between tacit knowledge and codified general knowledge.
The resulting model is used as a basis for an empirical study of the 122
largest functional urban regions in Europe, as defined by Hall and Hay
(1980). The study attempts to identify the sphere of socio-economic
influence of its urban core. It is shown that the use of conventional
administrative regions (such as NUTS regions) can distort, perhaps even
invalidate, any conclusions concerning growth potential.
In a chapter which adopts a monetary perspective, Dino Martel/alo asks
whether the real exchange rate is relevant for determining the innovation
potential of European regions. The hypothesis explored is that certain types
of foreign investment are particularly effective for transferring the
innovative technology, skills and know-how incorporated in goods from
advanced to less advanced regions. It is demonstrated that a country or
region can increase its export share only by attracting more direct
investment, because this increases the flow of innovation. The most
fundamental problem for a region in a currency area such as the EMU is to
be able to manage its competitive position. In the case of the EMU, a
competitive real exchange rate with a unit nominal exchange rate requires a
sufficient degree of price flexibility to bring about the desired amount of net
exports and resulting flow of foreign resources.
1.2.4 Impacts of Innovations in Telecommunications and

Transport
The final section of the volume explores some of the impacts of advances
in telecommunications and transport on the evolutionary path of
metropolitan development. It opens with a chapter by Cristofaro Bertuglia
and Sylvie Dccelli who begin by reflecting on the concept of innovation,
including the question "How new is new?" Various taxonomies of
innovation are examined, and distinctions made between the systemic and
the substantive perspectives. The authors argue that one weakness of such
taxonomies is their preoccupation with the 'hard' (or tangible) components
of innovation at the expense of the 'soft' (or intangible) components. The
latter are associated with changes in the pattern of interactions and
restructuring of relationships, and have strong links to dynamic processes of
knowledge exchange, learning and experience. Their chapter concludes with
an agenda for future research which concentrates more heavily on the softer
components of innovation.
Andrew Gillespie and Ranald Richardson look critically at the potential
of telematics to overcome the problems of peripheral areas. A review is
made of policy approaches to telematics in the UK, designed to stimulate
economic development in rural areas. The authors point out that whereas
early policies emphasised infrastructure provision, more recent policy has
focused on market access. This was the main constraint on smaller firms
wishing to innovate through telematics. Non-metropolitan areas can
doubtless benefit from telecommunications designed to overcome distance,
but rural telematic policies thus far have, by and large, failed to deliver the
expected economic benefits. In order to compete successfully in a global
knowledge society, it would appear that rural firms require people-oriented
skills and market knowledge which normally can be found only in larger
urban areas.
As many existing studies have been keen to show, our use of
telecommunications can serve as a substitute for transport. However, it is
also recognised that telecommunications have an indirect effect in
stimulating the demand for mobility. The final chapter, written jointly by
Ennio Cascetta and Bruno Montella, examines the impact of new types of
communication systems (including telecommuting, teleconferencing, and
teleshopping) on urban transport. They pay particular attention to the
competitive and complementary effects on urban mobility and transport
demand. This is a highly topical research area where, despite the significant
amount of work already carried out4, findings continue to be inconclusive.

The strength of the paper by Cascetta and Montella lies in their review of the
models and methods adopted to estimate the pertinent impacts of the more
recent innovations in this area.
REFERENCES
Andersson, A.E., Batten, D.F., Kobayashi, K., Yoshikawa, K., eds. (1993) The Cosmo-
Creative Society. Springer, Berlin
Arthur, B. (1994) Increasing Returns and Path Dependence in the Economy. The University
of Michigan Press, Ann Arbor
Batten, D., Thord, R. (1989) Transportationfor the Future. Springer, Berlin
Batten, D., Casti, J., Thord, R., eds. (1995) Networks in Action. Springer, Berlin
Bertuglia, C.S., Fischer, M.M., Preto, G., eds. (1995) Technological Change. Economic
Development and Space. Springer, Berlin
Bertuglia C.S., Lombardo S., Nijkamp P., eds (1997) Innovative Behaviour in Space and
Time. Springer, Berlin
Dosi, G. (1996) Opportunities, Incentives and the Collective Patterns of Technological
Change. Economic Journal
Graham, S., Marvin, S. (1996) Telecommunications and the City. Routledge, London
Hall, P., Hay, D.G. (1980) Growth Centres in the European Urban System. Heinemann,
London
Hepworth, M., Ducatel, K. (1992) Transport in the Information Age: Wheels and Wires,
Belhaven Press, London, New York
Koestler, A. (1964) The Act of Creation. Pan Books, New York
Nijkamp, P., Pepping, G, Banister, D. (1996) Telematics and Transport Behaviour, Springer,
Berlin
Young, W. (1987) Transport. Communication and Urban Form: Issues and Policies (2
volumes), Monash University, Melbourne
4 In addition to the many references cited in the Cascetta-Montella chapter, further studies
are reported in Young (1987), Batten and Thord (1989), Hepworth and Ducatel (1992),
and Nijkamp et al (1996).
Chapter 2
Innovation and Patterns of Learning: A Survey of

Evolutionary Theories
Mario Cimoli\ Marina della Giusta2

IECLAC-United Nations, Santiago, Chile
2Department ofEconomics, University ofReading, UK
2.1 INTRODUCTION
Some of the main international organisations concerned with

development issues, like the ECLAC-UN, WB and the OECD, are becoming
increasingly interested in studying systems of innovation known as National
Innovation Systems. We believe that this interest needs to be complemented
by a thorough understanding of the micro-foundations of innovative change,
namely evolutionary theory. Such an evolutionary perspective is needed in
order to appreciate the crucial role of organisations and institutions in
development processes (Cimoli and Dosi 1995). Without such deeper
insights, it may be impossible to trace the impact of learning and innovation
on possible trajectories of metropolitan development.
This chapter aims to provide a selective survey of evolutionary
approaches to innovation and learning. By 'selective', we imply a focus on
the role of technological paradigms and trajectories. By 'evolutionary', we
mean all those theoretical contributions that possess the following
methodological building blocks: consideration of dynamics; the presence of
micro-founded assumptions; bounded rationality and heterogeneity among
agents; explicit recognition of the continuous appearance of novelty; a view
of collective interactions as selection mechanisms; and, finally, the
consideration of aggregate phenomena as emergent properties of an unstable
system (Dosi 1997).
By adopting a systematic approach, we hope to accomplish two
important tasks. The first is the identification of a collective structure within
which the main threads that link technologies, institutions, capabilities and
12 CimoJi M., Della Giusta M.
economic performance can be identified and described. The second is an

attempt to identify the policy implications of this view, and to relate them to
an analysis of the actual mechanisms responsible for technical change and
innovation.
We begin our survey by explaining the micro-foundations of an
evolutionary view, including the notions of paradigms and trajectories. We
then proceed to describe the implications of this definition in terms of a
theory of production. Section 2.4 provides a brief introduction to the
behavioural assumptions that describe individuals, organisations and
institutions. In Section 2.5, some of the main models that describe the
evolution of industries are examined, while Section 2.6 describes the role of
technological capabilities and production capacity in the development
process. Section 2.7 discusses National Systems of Innovation. Innovative
activity is deemed to include all the processes by which firms master and put
into practice new product designs and manufacturing processes (Nelson
1993). In the final section, we draw some conclusions about the implications
of our survey findings for metropolitan development.
2.2 THE MICRO-FOUNDATIONS OF TECHNICAL

CHANGE
The model of technical change proposed in Schumpeter's early work has
been construed as being linear in nature, since it implied a set of simple
relationships running from invention through innovation to diffusion.
Nevertheless, Schumpeter saw inventions as happening discontinuously and
exogenously, with entrepreneurs exploiting them by transforming them into
innovations at a profit. In his later work, however, he recognised and
emphasised the role of corporate R&D, spotting the possibility of feedback
from successful innovation to increased R&D. Moreover, he also
acknowledged that large corporations could influence market demand
(Freeman 1982).
An opposite line of causation may be found in Schmookler's (1962)
demand-pull model. In his empirical analysis of patent data in railroading,
petroleum refining and the building sector, inventive effort was found to
vary directly with output, but lag slightly behind it. He therefore argued that
expected profits from invention, the ability to finance it, the number of
potential inventors and the dissatisfaction which stimulated them, were all
positively associated with sales. As a consequence, variations in inventions
were seen to be caused by those economic conditions with which output was
also positively correlated. Thus a causal relationship from economic growth
to innovation was derived.
Innovation and Patterns of Learning 13
The demand-pull model spawned a number of empirical studies. Mowery

and Rosenberg (1979) demonstrated that most empirical analyses did reveal
the importance of demand in supporting successful innovation, but failed to
find any causal relationship between the two factors. Later studies did not
contain sufficient evidence to conclude that innovation was stimulated by a
shift in demand, rather than by technology. Demand-based theories of
innovation can also be criticised on different grounds. Perhaps the most
serious criticism is that they reduce the innovative process to a simple
deterministic phenomenon strictly connected to market conditions. This
implicitly undervalues the complexity of the scientific and technological
processes that are conducive to innovation. What the review by Mowery and
Rosenberg demonstrated was that the perception of a potential market is a
necessary condition for innovation, but not a sufficient one (Dosi 1984) 1.
Linear models of technical change ignore what happens inside firms,
because they are treated as black boxes (Rosenberg 1982). It was claimed
that the 'early' Schumpeter model described a process running from the
science base through firms to markets. On the other hand, the demand-pull
model ran the opposite way. By taking into account the feedback
mechanisms proposed in Schumpeter's later work, and the notions of
leaming-by-producing inherent in Rosenberg's work, the model of
technological change became much more complex. R&D labs were now
seen as providers of inputs for learning in production and receivers of inputs
from the science base and from problems arising in production. Furthermore,
it was recognised that science often spills out of technology, rather than
impacts being exclusively the other way round.
To begin to understand the complex nature of innovative activity, it is
useful to summarise some of its main features. Firstly, scientific inputs have
become increasingly important in the innovative process. At the same time,
R&D activities have become more complex. Thus it is necessary to adopt a
long-run perspective in planning such activities within firms. A growing
body of empirical literature has emerged, correlating R&D efforts with
innovative output for various industrial sectors. In these studies, market and
demand changes did not exhibit significant correlation with innovation.
Secondly, innovation generated by leaming-by-doing, and the issues related
to knowledge embodied in people and organisations, have also come to the
fore. Regarding the innovation process as intrinsically uncertain by no means
implies that technical change occurs randomly. Evolutionary theories
suggest that the directions of technical change are determined jointly by
state-of-the-art technologies and technologies that individual firms possess.
For an exhaustive list of studies on this theme, see Rothwell and Walsh (1979) and
Combs, Saviotti and Walsh (1987).
14 Cimoli M., Della Giusta M.
Indeed, it is possible to identify patterns of change which are defined in

terms of technological and economic characteristics of products and
processes (Dosi 1984; Dosi, Freeman, Nelson and Soete 1988; Freeman
1994).
Clearly, it has not been possible to formulate a general theory of
technical change based exclusively on technology-push or demand-pull
factors. Furthermore, certain components of technology prevent us from
adopting definitions that would apply to all production sectors, industries
and firms. The evolutionary nature of technical change at a macroeconomic
level is based on the core notion of a technological paradigm. By adapting
the notion of paradigms formulated by Kuhn in the philosophical sciences,
Dosi (1988, p. 1127) defines a technological paradigm as:
"a pattern of solution of selected techno-economic problems based on
highly selected principles derived from the natural sciences, jointly with
specific rules aimed to acquire new knowledge and safeguard it,
whenever possible, against rapid diffusion to the competitors"
The notion of technological paradigms is grounded on three fundamental
ideas2 Firstly, it suggests that .any satisfactory description of 'what is
technology' and how it changes must also embody the representation of the
specific forms of knowledge on which a particular activity is based. This
means that technology cannot be reduced to the standard view of a set of
well-defined blueprints. Technology primarily concerns problem-solving
activities. It also involves those tacit forms of knowledge embodied in
individuals and organisational procedures (Cimoli and Dosi 1995). It is very
important to analyse the specific elements that are common to the various
A variety of concepts have recently been put forward to define the nature of innovative
activities: technological regimes, paradigms, trajectories, salients, guideposts, dominants
designs and so on. More crucially, there is a considerable overlap in these concepts, in
that they try to capture a few common features of the procedures and direction of
technical change (for a discussion and references, see Dosi 1988). The rates and direction
of technical change are shaped by the dominant paradigm and their disruption is
correlated with radical changes in paradigms. Freeman and Perez (1988) propose the
notion of techno-economic paradigm, and changes in this are caused by a combination of
interrelated product and process, technical, organisational and managerial innovations
involving an increase in potential productivity for all or most of the economy. In their
view a new paradigm emerges only gradually, since the world continues to be dominated
by an old paradigm. The new paradigm therefore begins to demonstrate its comparative
advantage initially only in few sectors. The new key factors have to satisfy three criteria:
being in rapidly growth, having pervasive applications and falling costs. The current
dominating information technology paradigm clearly possesses all these features, as
previously did the electrical equipment and chemical technology-based paradigm in the
interwar period and the mechanical paradigm associated with the industrial revolution.
evolutionary definitions, on which our rudimentary understanding of the

nature of technology is grounded.
In the work of Nelson and Winter (1982), problem-solving activity has
been characterised as a process of an irreversible, contingent, dependent and
uncertain nature. It generates both technical advance and technological
competence among the actors performing it. Regarding uncertainty, Dosi
(1988, p. 1126) suggests that:
"an innovative solution to a certain problem involves discovery and
creation, since no general algorithm can be derived from the information
about the problem that generates its solution automatically".
Dependency and contingency derive from the solution of technological
problems, involving the use of information drawn from previous experience
and formal knowledge, as well as specific and uncodified capabilities on the
part of the inventors. Thus the outcome of the search process will be
determined by the history of the inventor, the formal knowledge available
and the inventor's capabilities. Luck may also playa part.
The knowledge base of inventors draws upon two different elements: the
potentially public and the tacit. Public knowledge consists of the formal
knowledge available to us all, access to which may be restricted owing to
different ways of conceptualising and codifying knowledge. Following
Polany (1967) and Dosi (1988), tacit knowledge relates to those elements of
knowledge, insight, and so on that individuals have, which are ill defined,
uncodified, and unpublished, and which they themselves cannot fully
express. They differ from person to person, but may to some significant
degree be shared by collaborators and colleagues who have a common
experience.
Secondly, the paradigms entail specific heuristics of how to do things and
how to improve them. These heuristics may be shared by the community of
practitioners in each particular activity. Sometimes they are called
'collectively shared cognitive frames' (Constant 1985). Following Cantwell
(1991) at the level of individual firms, they are routines which incorporate
the skilful behaviour required for the generation and application of
technology and consist of an interlinked sequence of steps which require
knowledge on the part of those who perform them, and which cannot be
fully communicated to them unless they join the firm's team and undergo the
same learning process.
In terms of evolutionary theory, it is argued that the technological
capabilities that define the competence of firms are best understood in terms
of the tacit element of technology. Strategic assets are those assets which are
non-tradable, non-imitable and non-substitutable (Dierick and Cool 1989).
16 Cirnoli M., Della Giusta M.
These assets are accumulated and refined over time, making them
impossible to copy. Such characteristics confirm that industrial R&D reflects
a private and a public dimension of knowledge and technology (Nelson
1981). They also facilitate a distinction between technology and
information3 . The latter spreads easily among firms, whereas the former
includes
"tacit and specific knowledge that cannot be written down in blueprint
form and cannot, therefore, be entirely diffused either in the form of
public or proprietary information" (Dosi 1988, p. 1131).
The fact that tacit knowledge is embodied in individuals is particularly
important, and plays a major role in explaining the nature of the impact of
science on technology (see Pavitt 1991; Batten, Kobayashi and Andersson
1989).
Technological paradigms also define basic artefacts which are
progressively modified and improved over time. These artefacts can be
described in terms of some fundamental technological and economic
characteristics. In the case of an aircraft, for example, the basic attributes can
be described not only in terms of the inputs and production costs, but also in
terms of technological features such as wing-load, take-off weight, speed,
refuelling distance, etc. Of interest is the fact that technical progress often
displays patterns of order and invariance in terms of these product
characteristics. Technological invariance has been be found in
semiconductors, agricultural equipment, and cars, as well as in micro-
technological studies.
Technological trajectories may be associated with the progressive
realisation of innovative opportunities associated with each new paradigm.
In principle, this can be measured in terms of the changes in the fundamental
techno-economic characteristics of artefacts and production process4 . Nelson
and Winter (1982) define 'natural trajectories of technical progress' as those
paths which shape the direction in which problem-solving activities move
and which posses a momentum of their own. In this sense, a trajectory
represents the normal problem-solving activity determined by a paradigm
(Dosi 1988).
The core ideas involved in the notion of technological trajectories are:
See also the importance of co-specialised assets described in Teece, Pisano and Shuen
(1990) and Teece (1998), and to which we return later.
4 The interpretation of technical change and a number of historical examples can be found
in pioneering works on the economics of technical change by Chris Freeman, Nathan
Rosenberg, Richard Nelson, Sidney Winter, Thomas Hughes; Paul David, Joel Mokyr,
Paolo Saviotti and others; for a partial survey see Dosi (1988).
1. that each particular body of knowledge (i.e. each paradigm) shapes and
constrains the rates and direction of technological change irrespective of
market inducements;
2. that one can observe regularities and invariance in the pattern of technical
change that hold under different market conditions (e.g. different relative
prices) and whose disruption is correlated with radical changes in the
knowledge base (in paradigms);
3. that technical change is partly driven by repeated attempts to cope with
the technological imbalances which it creates.
Such technological imbalances are described by Rosenberg (1976) as

'bottlenecks'. They serve as focusing devices, since efforts to overcome
them are themselves an important source of technical change. The
cumulative impact of small increments is important. Adopting Gilfillian's
view, improvements in shipbuilding embodied:
"the gradual and piecemeal nature of technological change, drawing
heavily on small refinements based on experience and gradually
incorporating a succession of improved components or materials
developed in other industries" (Rosenberg 1982, p. 17).
2.3 IMPLICATIONS FOR PRODUCTION THEORY
The elements of technical change discussed so far, especially the

implications of localised technical change, have been investigated recently
by Atkinson and Stiglitz (1969). Their hypothesis is that technical advances
in one sector can improve production techniques in that sector, but with very
few or no spillover effects upon techniques in neighbouring sectors. In terms
of a neo-classical production function, technical change would imply the
outward movement of one point of the function, rather than the whole
function. If the effects of learning-by-producing, and the costs of switching
from one technique to another, are also taken into account, then the effect of
technical change on production differs greatly from that described by
standard neo-classical theory.
A general property, now widely acknowledged in the innovation
literature, is that learning is both local and cumulative. Local implies that the
exploration and development of new techniques is likely to occur in the
vicinity of the techniques already in use. Cumulative means that current
technological development builds upon past experience of production and
innovation. Thus it proceeds via sequences of specific problem-solving
junctures (Vincenti 1992). This fits together with the ideas of paradigmatic
knowledge and the ensuing trajectories. An important implication is that, at

any point in time, there is little scope for substitution between techniques
because of the limited availability of blueprints that differ from those in use.
Because the notion of technological paradigms contains elements of
production and innovation theory, it is an evolutionary theory. In order to
summarise what we have learnt so far, we list below a few key observations
which are derivable from our evolutionary approach (Cimoli and Dosi
1995):
(a) in general, one or a few best practice techniques dominate the others at
any point in time (irrespective of relative prices);
(b) different firms may be characterised by their adoption of better or worse
techniques;
(c) the observed aggregate dynamics of technical coefficients in each
particular activity is the joint result of the process of imitation/diffusion
of existing best-practice techniques, the search for new ones, and market
selection amongst heterogeneous agents;
(d) changes of best-practice techniques over time trace out reasonably regular
trajectories, both in the space of input coefficients and in the space of the
core technical characteristics of outputs.
Observation (a) follows from processes of diffusion and competItIOn

among technologies, whereas observation (b) is a consequence of the
differences in tacit forms of knowledge among firms. The nature of learning
processes is responsible for observation (c), whereas (d) follows from the
fact that the prevailing paradigm determines the actual direction that the
learning trajectory pursues. A distribution of micro coefficients in the space
of unit inputs is presented in Fig. 2.1, under the simplifying assumption of a
homogeneous good being produced under constant returns to scale with two
inputs.
'\
XI
Figure 2.1 The distribution of technological coefficients
This graph depicts the distribution of coefficients (Cj) observed at time t,

where we denotes the various firms (or techniques) in decreasing order of
efficiency. The scatter plot of coefficients reveals the degree of asymmetry,
or heterogeneity of firms, within the sector. The dispersion follows from the
above-mentioned fact that some firms adopt a technique which is inferior to
the best practice technique, because they do not know what the best practice
technique is or how to adopt it. The situation at time 1+ 1 is also portrayed.
Here the distribution of micro-coefficients has changed. Precipitants of such
change include attempts by below-best-practice firms to imitate the
technological leader, innovative efforts to generate new techniques (which
may be superior to those available), and the exit of existing firms and entry
of new ones. All those processes governing the diffusion of innovation are
largely responsible for the dynamics by which such changes in the
distribution of technical coefficients take place.
In this framework, changes in relative prices have an influence on the
direction of imitation and innovative search pursued by agents (Atkinson and
Stiglitz 1969). However, prices remain constrained by the nature of the
underlying knowledge base, the principles it exploits, and the existing
technological paradigm. Persistent shocks to relative prices tend to be
reflected in the diffusion of alternative paradigms, rather than in substitution
between techniques (Verspagen 1990)5. Of course, a much grander
theoretical story has been written about development, diffusion and
competition among alternative paradigms. Fragments of it can be found in
explicit evolutionary models (e.g. Nelson and Winter 1982 or Silverberg,
Dosi and Orsenigo 1988), in path-dependent stochastic models (e.g. Arthur
1989; Arthur, Ermoliev and Kaniovski 1987; Dosi and Kaniovski 1994; and
David 1989) and in sociological models of network development (e.g.
Calion 1991). These may be worthy of study when attempting to unravel
spatial and locational trajectories.
Metcalfe's (1981, 1988) work provides a useful set of links between
micro studies on the diffusion of innovation and the macrodynarnics of
industrial growth. He does this by viewing the dynamic triggers as impulses
from innovation which are transmitted across the economy via incentives
provided by profitable rewards. Metcalfe is critical of the standard diffusion
5 He stresses that the type of technical progress taken into consideration in his own, and in
Atkinson and Stiglitz's analysis is Hicks' neutral technological progress, i.e. purely
labour-saving or capital-saving technical progress is not considered. The effects of
unexpected price shocks on productivity are described at the aggregate level, and a
possible interpretation of the productivity slowdown could be the presence of a
continuous adaptation to a fast-changing environment. The latter would compel firms to
produce with techniques they have not yet learnt to exploit efficiently, or old techniques
that they master efficiently but are inferior to the new ones.
model for neglecting the supply side, i.e. the profitability perceived by
producers of innovation. His own model depicts the pace of diffusion of an
innovation as being determined by both supply and adoption constraints.
This ensures a balanced diffusion path which is determined by an adjustment
gap (i.e. the difference between saturation output level and the initial rate of
demand) and dynamic elements in demand and capacity growth. Thus the
diffusion process becomes the force determining the pace and direction of
technical change. A similar line of reasoning lies at the heart of the model by
Silverberg et al (1988), in which innovation diffusion, diversity of
technological capabilities, business strategies and expectations, are formally
incorporated into a theory of the evolutionary patterns of industries and
nations.
The dynamics in all the above models are founded on learning
mechanisms within firms. This learning is associated with the ways in which
new production factors are introduced and how firms learn to use them. In
other words, the dynamics depends on Arrow's process of learning-by-
doing. Adoption and learning substantially modify the cost functions faced
by individual firms, and thus their productivity dynamics. An interesting
example of this learning-by-doing approach is the work by Gurisatti et al
(1997). They discuss the patterns of diffusion of microelectronics-based
technical change in machine tools employed in some Italian metal working
firms. By interviewing mechanical engineers, they obtained a description of
the process of innovation which consisted of radical improvements (the
installation of new machines) followed by long phases of endogenous
improvements that substantially improved the whole process. The gains in
productivity resulting from the latter often exceeded those from the
installation of the new machines. Moreover, the authors found that the
diffusion of new machines among firms takes a considerable period of time,
and that large variations exist between firms due to differences in their
technological and organisational capabilities.
A graphical description of this process can be integrated with another
described by Dosi (1984). Because unit costs decrease along a
technologically determined learning curve, human skills and abilities clearly
display a cumulative character (see Fig. 2.2). This also helps to explain how
the set of points representing an industry recomposes itself (as we depicted
in Fig. 2.1). Furthermore, it is interesting to note that the existence of
increasing returns to adoption can be explained purely by the cumulative
nature of the learning-by-doing process.
Thus far, the question of paradigms, trajectories or equivalent concepts
has been discussed at a micro-technological level. We have argued that a
paradigm-based theory of innovation and production seems to be highly
consistent with observed patterns of technical change, microeconomic
heterogeneity and technological gaps. Also, it may be associated with the

theories of production which allow for increasing returns - from those put
forward by Young and Kaldor to the recent, more rigorous formalisations of
path-dependent models of innovation diffusion. In the latter, interactions
between micro decisions and some form of adaptive learning (or
externalities) produce unique technological paths which may be reinforced
by lock-in effects with respect to technologies. Although some technological
choices may well be inferior to others, they still prosper and grow to a
dominant position because of the advantages of being first and the quirks of
their history. These kinds of models have been proposed by Brian Arthur and
Paul David.
WorkinEf-_ _...,
hours
Adoption Adoption Adoption

Learning Learning
and and
adaptation adaptation
Time
Figure 2.2 A learning curve (adapted from Dosi 1984 and Gurisatti
et a11997)
The evolutionary nature of technical change may be responsible for what

Arthur has called 'lock-in by historical events' (Arthur 1989). This suggests
that the process of selection and adoption is strongly path-dependent,
unpredictable, irreversible, and liable (on some occasions) to select inferior
technologies. The lock-in approach has several implications which are
especially important when it comes to the location of industry and the
development of cities. These include the fact that the specialisation history
of a particular location (in terms of the techniques already adopted) is crucial
in determining its current choice of technique. Furthermore, it may attract
more firms to that location purely by boasting a significant number of firms

already. In other words, agglomeration economies can be enjoyed there.
Referring again to Fig. 2.1, the two distributions may be representative of
two cities or regions separated by a technological gap. Once again, the
evolutionary explanation resides in the relative learning processes of each
location. In keeping with other theories that stress the importance of
imperfect learning (ahead of the optimal allocation of resources) as the
engine of development, evolutionary theory predicts persistent asymmetries
between regions in their capacity to master production processes. This has
two main consequences:
1. it is possible to rank different regions or countries by the efficiency of

their average production techniques and the performance characteristics
of their outputs;
2. these asymmetries are not related to differences in capital/labour ratios.
Differences in technological capability not only account for such

asymmetries, but also for the differing abilities and time lags in developing
new products. The specific capability of a region determines its ability to
borrow and adapt the more advanced technologies developed elsewhere.
This lies at the very roots of its industrialisation process. The next section is
devoted to investigating in more detail the behavioural assumptions which
lie at the core of this approach, as well as the implications of the behaviour
of individuals and organisations for the dynamics of firms, industries,
regions and countries. When firms are seen as repositories of knowledge,
forming networks of linkages with other firms and institutions, it becomes
possible to apply the same line of reasoning to national systems of
innovation. The technological gaps between countries can then be seen as
the outcome of different national technological and institutional capabilities.
2.4 FROM INDIVIDUALS TO INSTITUTIONS
In order to investigate the behavioural assumptions used to describe

economic agents in evolutionary models, it is useful to consider approaches
that abandon the assumption of rationality made in traditional orthodox
economic theory. As early as 1936, Hayek asserted that agents would not be
capable of fully rational decisions, if the unrealistic assumptions regarding
their unlimited capacity of acquiring and processing knowledge were
removed. Thus knowledge would be diffused heterogeneously and
asymmetrically amongst agents. According to Egidi (1996), this intuition lies
behind the bounded rationality approach formulated by Simon. The
existence of limits to the possibility of individuals taking fully rational

decisions explains the formation of institutions. Institutions exist in order to
gather knowledge and information. According to Hayek, they are:
"the historical and unintended product of the consolidation of inter-
individual relationships" (Egidi 1996, p. 23).
Hayek, and later Schum peter, placed the creation of knowledge at the
core of the process of co-ordination among individuals and consequently of
economic change. The microeconomic foundations of this approach can be
found in the work undertaken in the fifties by Cyert, March, Simon and
Trow, who analysed the role of learning activity in human decision-making.
Within organisations, individuals learn to solve problems through stable
behavioural patterns of action. Thus their behaviour becomes routinised.
Routines are defined as "procedures which solve sets of problems internal to
the organisation", where a procedure is "a set of instructions determining the
actions to be taken when dealing with a particular circumstance". This
replication of procedures enables individuals to reduce the complexity of
individual decisions, so that routines become automatic and partly tacit. By
using a theoretical framework in which co-ordination between individuals
and their activities is central, it is possible to classify organisations as
'devices with which to co-ordinate economic activities'. These can range
over a continuum extending between pure markets and pure hierarchies.
In work by Dosi and Lovallo (1995), the presence and consequences of
'decision biases' in organisations are discussed in the context of corporate
entry and the evolution of industrial structures. Decision biases, such as
overconfidence in the future, are a result of the process by which firms build
their competence. This, in turn, is shaped by the characteristics within which
technical change takes place. The behaviour of individuals and organisations
is shaped by the features of the knowledge bases they can draw upon. The
emergence of cognitive frames and decision routines is a result of the
presence of "ever-changing and potentially surprising environments". Action
rules often take the form of relatively event-invariant routines which are
nonetheless robust, in the sense that they apply to entire classes of seemingly
analogous problems. Adaptive learning, tends to lead to lock-in phenomena
(Dosi and Lovallo 1995). Again, the characteristics of knowledge shape the
behaviour of individuals and organisations.
In institutional theory, organisations are at the heart of change (North
1990). This core position is due to their demand for investment in
knowledge, the interactions which they engender between economic activity,
scientific knowledge and institutional stru~ture, and finally the gradual
change in informal rules which they give birth to in the course of their
activities. According to North, institutions define the set of opportunities of a
society, whereas organisations exist in order to exploit such opportunities. In

doing so, however, they develop and gradually alter institutions, so the
characteristics of institutional change are depicted as intrinsically
evolutionary. Indeed, the institutional dimension has central importance in
evolutionary theories of production and innovation. A bi-directional relation
between market structures and patterns of technological learning is
acknowledged. The dependence of firms' performance and industrial
structures on learning characteristics is a direction which has already been
illustrated earlier in this chapter. Here, we shall draw implications at the firm
level from the relationship between institutions and organisations.
According to the above interpretation, heterogeneity will manifest itself
not only at the level of technical efficiency, but also at the level of
profitability. Different rates of learning influence the ability of firms to
survive and expand, and thus affect industrial structures. Firms are crucial
(but not exclusive) repositories of knowledge which is, to a large extent,
embodied in their operational routines and modified over time by their rules
of behaviour and chosen strategies. This idea is central in the
characterisation of firms' technological capabilities, as proposed in Nelson
and Winter (1982), and in the idea of competence, as proposed by Dosi,
Teece and Winter (1992). According to the latter, a firm's competence is a
set of differentiated technological skills, complementary assets, and
organisational routines and capacities that provide the basis for a firm's
competitive capacities in a particular business and, in essence, competence is
a measure of a firm's ability to solve both technical and organisational
problems.
When the role of firms as actors in the process of technical advance is
recognised, it is easy to understand how the nature of technological change is
fundamentally shaped by the learning processes of firms. Learning has been
described as local and cumulative. One classical perspective on the complex
interrelations between technological learning and organisational change is
Alfred Chandler's reconstruction of the origins of the modern multi-
divisional (M-form) corporation and its effects on American competitive
leadership over several decades (Chandler 1990~ 1992a and 1993). As
Chandler argued, there are close links between business history and
evolutionary theories (Chandler, 1992b). It is useful to recall one of his main
messages:
" ... it was the institutionalising of the learning involved in product and
process development that gave established managerial firms advantages
over start-ups in the commercialisation of technological innovations.
Development remained a simple process involving a wide variety of
usually highly product-specific skills, experience and information. It
required a close interaction between functional specialists, such as

designers, engineers, production managers, marketers and managers ...
Such individuals had to co-ordinate their activities, particularly during
the scale-up processes and the initial introduction of the new products on
the market ... Existing firms with established core lines had retained
earnings as a source of inexpensive capital and often had specialised
organisational and technical competence not available to new
entrepreneurial firms" (Chandler 1993, p. 37).
Organisational dynamics can be interpreted as an evolutionary story of
competence accumulation and development of specific organisational
routines (Chandler 1992b). The model has been refined by incorporating the
importance of the co-specialised assets of firms. These assets are
complementary to production and lie downstream from product-process
development in the value-added chain. They also play an important role in
stimulating technical change. Did seemingly superior organisational forms
spread evenly throughout the world? The Chandlerian enterprise diffused,
albeit rather slowly, across other OECD countries (Chandler 1990; Kogut
1992). However, the development of organisational forms, strategies and
control methods have differed from nation to nation, because of their cultural
differences (Chandler 1992a). Moreover, the diffusion of the archetypal M-
form corporation has been limited to around half a dozen developed
countries.
A growing literature identifies some of the specific features of German,
Japanese or Italian systems of production in their early corporate histories,
and shows how these have influenced the contemporary form of organisation
and learning (see Chandler 1990; Coriat 1990; Kogl,lt 1993, Dursleifer and
Kocka 1993; Dosi, Giannetti and Toninelli 1993). There is also a wealth of
information about the New Industrial Economies. In Korea, for example, it
seems that the major actors in technological learning have been large
business groups known as the the chaebols. At a very early stage of
development, they were able to internalise the skills needed for the efficient
selection and adaption of technologies from abroad. In a relatively brief
time, they managed to develop impressive engineering capabilities. This
issue is discussed in greater depth in Amsden (1989), Amsden and Hikino
(1993, 1994), Enos and Park (1988), Bell and Pavitt (1993), Cimoli and Dosi
(1988, 1990), Dahlman and Westphal, (1982), Lall (1992), Katz (1984, 1987),
Kim and Dahlman (1992), Teitel (1981, 1987), Teubal (1984, 1987).
Conversely, Taiwanese organisational learning has rested more on large
networks of small and medium firms which are open to the international
markets and often develop production capabilities complementing those of
first world companies (Dahlman and Sananikone 1990, Ernest and O'Connor
1989). It is precisely these diverse learning patterns that constitute our

primary interest.
The different organisational patterns of learning simply serve to illustrate
the multiplicity of evolutionary outcomes that exist. The fundamental point
is that the rates and directions of learning are dependent on the ways
corporate organisations emerge, change, and develop particular problem-
solving capabilities. In other words, they depend on a firm's individual
history. This is the co-evolutionary view emphasised by Nelson (1996),
where market structures and patterns of technological learning are
interdependent. Different rates of learning influence the ability of firms to
survive and expand. In tum, this affects industrial structures. Conversely,
any particular structure - with its associated distribution of corporate
features - influences and constrains the know-what and know-how of firms,
i.e. their propensity to learn. Formal applications of this positive feedback
loop approach may be found in Nelson and Winter (1982), Winter (1984),
Dosi, Marsili, Orsenigo and Salvatore (1993).
2.5 EVOLUTIONARY INDUSTRIAL MODELS
A recent assessment of evolutionary theories by Dosi and Nelson (1993)

interprets the continuous turbulence in industry dynamics, and the
considerable variety in the patterns they follow, as a direct consequence of
differences in firms' behaviour. The literature on the role of innovation in
the evolution of an industry, however, has embraced a different perspective:
the industry life cycle model as described by Utterbach and Abernathy
(1975). This model relates both process and product innovation to stages of
development of an industry. Innovations are firstly stimulated by market
needs. Here, product and process development are still uncoordinated. In the
next stage, innovations are stimulated by technological opportunity, product
development aims to maximise sales, and process development is segmental.
The third and final stage is characterised by innovations stimulated by
production factors, with cost minimising product development and systemic
process development. In this model, therefore, the locus of innovation, its
type and the barriers to it change according to the stage of development of
the industry. Since the type of industry is not influential, technical change is
considered to have a uniform effect on all industrial activities.
In the evolutionary models of Nelson and Winter (1982), firms are seen
as the central actors, and their essential characteristics are given by their
capital stocks and prevailing routines. The relative superiority of a
technology is determined by its profitability, insofar as it is able to generate
profits and lead to capital formation and growth of the firm (Dosi and Nelson
1993). Through imitation by other firms, such technology spreads and

replaces less profitable forms. More recent models (Dosi et al 1993)
explicitly describe the existing regularities (in terms of rates of entry and exit
and variations in market shares) in industrial structures as 'emergent
properties', ansmg from non-equilibrium interactions amongst
technologically heterogeneous firms. In particular, the selection criteria
among firms are endogenous to the model, which is capable of generating,
the aggregate, empirically-observed dynamics through simulations in which
the system parameters describe learning processes and market selection.
Nelson and Winter (1982) also recognised two different technological
regimes, corresponding to the two phases of Schumpeter's work: the
entrepreneurial regime (favourable to innovative entry) and the routinised
regime (where established firms perform the bulk of innovative activities).
The former was also associated with highly innovative industries in which
large firms are dominant; whereas the latter was characteristic of capital
intensive, advertising intensive, concentrated and highly unionised
industries. On the basis of this framework, Audretsch (1996) observed that
entry rates should be relatively high in industries belonging to the
entrepreneurial regime, and lower under the routinised regime.
As we have repeatedly underlined, one of the building blocks of
evolutionary thinking is the recognition of the specificities of technical
change. According to Dosi (1988), it is simply not possible to find one all-
encompassing model of technical change which can describe the
characteristics of all industries or sectors. The particular characteristics of
innovative processes, historically observed in empirical studies of different
sectors, have led Pavitt (1984) to formulate a taxonomy describing industry-
specific models of technical change. This taxonomy includes five sectoral
patterns, allowing the derivation of industry-specific models of technological
change: supplier dominated (agriculture, services, and traditional
manufacture), scale intensive (consumer durables, automobiles, civil
engineering, and bulk materials), information intensive (finance, retailing,
publishing, and travel), science based (electronics and chemicals), and
specialised suppliers (machinery, instruments, and software).
In the supplier dominated and information intensive sectors, the main
sources of technical knowledge are situated outside the firm. In the science-
based sectors, on the other hand, the main sources of technical advance are
in-house R&D and basic research. In terms of the discussion presented
above, this sector can be characterised as being of the late Schumpeter-type.
The scale intensive sector, characterised by continuous processes, finds its
main sources of technology in production engineering, production learning,
suppliers and design offices, whereas design and advanced users are the
sources for specialised suppliers. Both these sectors are characterised by
conservative and incremental processes, and can thus be described as

Schmookler-type (demand drivent
The peculiar features of each sector in terms of its technological
characteristics can be combined with the issue of the influence that changes
in relative prices possess on innovative activities. As already discussed in the
presentation of the localised technical change model, and the view of the
behaviour of individuals and organisations which lies at the heart of
evolutionary theorising, changes in relative prices do have an influence on
the directions of innovative efforts. These, however, remain constrained by
the nature of the knowledge base of the particular activity, the physical and
chemical principles it exploits, and the technological system in which the
activity is embodied. We wish to push the argument further, and suggest that
it is in fact possible to express a relationship between the level of localisation
of technical progress and the influence of prices on the substitutability
among techniques. We therefore draw a representation of such relationship
derived for different sectors, for example in Pavitt's taxonomy, and associate
it with a graphic representation (upper part of Fig. 2.3) of the varying types
of technical progress (non-localised, weak-localised, and strong-localised)
that correspond to increasingly localised technical progress shown in the
lower part of the figure .
non-localised t.c. weak-localised tc. stro ng-localised tc.
from non-localised to strong-localised techniad change

high
_A_ .................. ___ ....................... . . . . . . . . . . . .
Influence of TradilionaVsupplier dominated
relative prices =tor.; "
Science based sectors
.. ....... ...... . . ...... . . .. ....... ,................
.........
. .&.
low
non-localised Lc. weak-localised Lc. strong-localised Lc.
Figure 2.3 Relative prices and technical change
See on these points: Rosenberg (1990, 1991).

This further step is based on the theory of production derived in Section

2.2 (see in particular the representation of technical coefficients in Fig. 2.1
and the discussion presented there), and the problem of the influence of
relative prices on substitutability among techniques (Atkinson and Stiglitz
1969; Verspagen 1990). It aims to connect these micro-foundations with the
sectoral characteristics oftechnical change that are now being presented.
Even if it were possible to think about the start of a particular production
activity facing non-localised technical change, given the behavioural
assumption based on bounded rationality and the characteristics of the
process of technical change, the situation would eventually be similar to the
right hand side of the figure. There we find ourselves in that part of the
continuum of degrees of localisation that varies between weakly and
strongly localised technical change. In particular, non-localised technical
change would be, in our view, the exceptional case, whereas the 'real'
situation would be represented over the second half of the continuum. If one
then relates the influence of relative prices on substitutability among
techniques with the situation described above, we find that different
behaviour emerges depending on the sector one is looking at. The science-
based sector is an example of a situation in which the influence of changes in
relative prices over the choice of technique starts to decline in importance
very early. The curve for the traditional and supplier-dominated sectors, on
the other end, depicts a situation in which the influence of prices remains
important for production processes belonging to this group.
Walsh's (1984) analysis of innovation during the development stages of
two subsectors of the chemical industry (plastics and dyestuffs) confirms
both the general sectoral characteristics included in Pavitt's taxonomy, and
the specificities of the evolution of subsectors. At first glance, plastics seem
to have followed an early-Schumpeterian pattern. The early plastics were
primarily developed through the entrepreneurial activity of inventors. Later
on, however, science and anticipated demand in large corporations played a
major role, satisfying the late-Schumpeterian model. The analysis of patents
in dyestuffs, on the other hand, produces contrasting results depending on
whether a solely quantitative or qualitative analysis is performed. When a
quantitative analysis is made, a demand-pull model seems to emerge. Using
a qualitative analysis, however, produces an early Schumpeterian pattern.
Recent evolutionary studies have tried to account for several of the
peculiarities present in the evolution of industries. For example, Malerba and
Orsenigo (1996, 1997) have identified persistent diversity among firms, the
presence of a high degree of turbulence in terms of both entry and exit rates
and changes in market share, the persistence of certain sectoral specificities,
and the presence of regularities in the relationship between sectoral
dynamics and rates and modes of technical change. This last observation
allowed Malerba and Orsenigo to produce a more restricted taxonomy based

on the two stages in Schumpeter's work: (a) a 'Schumpeter Mark l' group,
characterised by the technological ease of entry in an industry, the major role
played by new firms in innovative activities and the presence of a continuous
erosion in competitive and technological advantages of the established firms;
and (b) a 'Schumpeter Mark II' group, in which new innovators face
significant barriers to entry - because large established firms are prevalent in
innovative activities and/or a few firms dominate because they are
continuously innovative, thanks to lock-in and the accumulation of
technological leadership over time.
These empirical regularities may be seen at a more abstract level in the
model presented in Dosi, Marsili, Orsenigo, and Salvatore (1995, 1997).
This is based on evolutionary micro-foundations and describes the behaviour
of firms in the two learning regimes (Marks I and II). It emphasises the
process of the accumulation of competence by firms. The model contains a
stylised industry structure composed of three levels in which firms can be
aggregated by micro-sectors, sector, and industry. They are measured
according to three dimensions: their age, size, and competitiveness. All the
endogenous variables depend on learning, market selection and dynamics of
their markets. The size of the micro-sectors is determined by an exogenously
given demand cycle, but their dynamics are the product of the endogenous
firms' variables.
By distinguishing between Mark I (no learning) and Mark II (cumulative
learning), aggregate statistics can be produced to describe the market
structure in ways that are systematically different according to the learning
regime hypothesised. There are more firms in Mark I than in Mark II, the
number increasing with the technological opportunities for entrants.
However, the concentration turns out to be directly proportional to
opportunities for incumbents and inversely proportional to opportunities for
entry. At the level of micro-sectors, life cycles emerge, whereas at the
industry level, as stated above, the results are compatible with the empirical
evidence observed for the Mark I and Mark II regimes described in Malerba
and Orsenigo (1996). Dosi and Orsenigo (1994) claim that their evolutionary
models possess the ability to show how an economic system self-organises
and generates aggregate dynamics similar to those observed in the
marketplace. With economically meaningful values of system parameters,
the macro-dynamics produced are consistent with the micro-level stylised
facts available.
In order to describe behaviour under conditions of environmental
complexity and non-stationarity, it is necessary to have a micro-level
description of agents (individuals and firms) in terms of institution-
embedded routines, knowledge specificity and belief-conditioned learning.
In this way, evolutionary representations can be imbued with multiple

equilibria and imperfect information models in order to highlight the
intrinsic richness of representations based on the continuous micro-level
endogenous production of aggregate environmental nonstationarity.
2.6 TECHNOLOGICAL LEARNING AND

PRODUCTION CAPACITIES IN THE PROCESS
OF DEVELOPMENT
The evolutionary path of technological learning is related to the capacity

to acquire suitable technologies (e.g. capital goods, know-how) and the
ability to absorb these technologies and adapt them to local conditions.
During the last three decades, significant technological progress has
occurred in the Newly Industrialised Countries. A number of empirical
studies have examined the increased technological capabilities of developing
countries, revealing that some of them have even become exporters of
technology (see Lall 1982; Fransman 1986; Fransman and King 1984; Teitel
1984; and Teubal 1984). These analyses of technological capacity have
revealed the crucial role played by certain 'core technologies' associated
with network infrastructure (such as energy, transportation and
communication technologies) as sources of technological skills, problem-
solving opportunities and productivity improvements.
Core technologies determine the relative advantages or disadvantages of
any location, since they provide infrastructures for a wide range of activities.
Moreover, it is possible to identify a pattern of industrialisation that evolves
through the emergence of the sectors classified in Pavitt's taxonomy. The
initial stage involves the development of the supplier-dominated and
specialised supplier sectors (various forms of incremental learning take
place, e.g. in use of equipment, development of engineering skills in the
adaptation and transformation of machines and products, etc.). The second
stage involves the emergence of scale intensive industries. These focus on
creating a technological synergy between production and sets of innovations
(which gives rise to horizontal and vertical integration), the adoption of
technologies associated with the exploitation of static and dynamic
economies of scale, and finally the development of formal R&D
complementary to informal learning. In the final stage, a science-based
sector is created, in which the knowledge base is exploited economically
through formalised research efforts and R&D is the typical learning
mechanism.
Among these sectors tnere exist input-output linkages which give rise to
a wide set of externalities and interdependencies based upon the
communality of knowledge bases, complementarities, and technological
spillovers. Such untraded technological flows are essential not only for the
technological development of the firms involved, but also for the
development of the whole industry. Fig. 2.4 portrays some sources of
technological linkages among the sectors described in the dynamic
taxonomy.
I Specialised supp6ers
~uipment
~~- Traditional
Science based Equipment Materials

~aturalresources
intensive sectors
~
Components ~
Materials
r----------,I
Scale intensive _
~ "A_~
Materials
Figure 2.4 Technological flows and sectoral specificities (Source:

Guerrieri 1993)
Specialised suppliers provide product innovation and capital inputs for

other sectors whereas, through the production of components and materials,
the science-based sectors generate positive effects which propagate to the
whole system. All these linkages are fundamental for the process of
industrialisation, in particular those connections established between the
most innovative and the traditional natural resource-based sectors. An
application of this type of taxonomic dynamic analysis to some Latin
American and Southeast Asian countries can be found in Cimoli (1988) and
Bell and Pavitt (1993).
Sectoral learning patterns, however, are nested within broader structures
which exist at the regional and national level, like those defining the
educational system. For example, in 'supplier-dominated' and 'specialised
supplier' sectors, a significant role is played by the levels of literacy and
skills of the workforce. In scale-intensive sectors, the existence of managers
Innovation and Patterns of Leaming 33
capable of efficiently running complex organisations is also important. In

science-based sectors, the quality of higher education and research
capabilities is obviously relevant. The role of technology transfer as a source
of development of local capabilities has been extensively investigated:
increasing technology flows towards developing economies have taken
place, especially towards Asian countries.
A significant body of literature stresses the importance of institutions and
their role in economic and industrial development. Regarding the Pacific
Rim nations, studies by Amsden (1989), Wade (1990), Cantwell (1991) and
others help to explain the reasons for institutional success and institutional
failure. Bardhan (1996) analyses issues such as co-ordination, which is seen
in terms of the interaction between institutions, or state governance structure,
and industry. He suggests that the success of institutions in some NIEs (e.g.
South Korea and Taiwan) can be put down to their capacity to establish and
enforce performance criteria. For example, credit allocation by the state was
tightly bound with export performance; in this way, international
competition was used to foster internal learning. The following section will
analyse institutional success and failure in this context, and what its
implications might be in terms ofthe technological capability framework.
2.7 AN EVOLUTIONARY VIEW OF NATIONAL

SYSTEMS OF INNOVATION
Despite the recent turmoil in Asian markets, government intervention has

played an essential role in their successful industrialisation. Most of the
Southeast Asian economies have focused on export orientation and
specialisation in a few manufactured commodities. Particular emphasis was
given to the promotion of linkages between enterprises, with the aim of
promoting stable access to technology transfer and a fruitful mode of
diffusion across the whole economy. Another key aspect in the industrial
development of these countries has been human capital formation. The role
of the scientific and educational system in industrial development has been
repeatedly underlined in the literature on Asian nations. It is often cited as a
fundamental precondition to their success. On the whole, the general pattern
of incentives defined by the institutions has accounted for the type of
response to internal and external stimuli, and determined the successes and
failures of the these young tigers. This pattern of development provides an
example of the functioning of what is referred to as the National System of
Innovation (Freeman 1987; Lundvalll993; Nelson 1993; Edquist 1997).
The specific features of national systems of production and innovation

are seen as the combined outcome of the three levels of analysis presented in
this section:
1. the micro level - where individual firms may be seen as repositories of

knowledge, embodied in their operational routines and modified over
time by their own strategies and higher level rules of behaviour;
ii. the meso level - where networks of linkages between firms and other
organisations enhance the firms' opportunities for improving their
problem-solving abilities; and
iii. the macrolevel- where the firms' microeconomic behaviour is embedded
in a complex set of social relationships, rules and political constraints
(Cimoli and Dosi 1995).
To assemble the components of our evolutionary account of economic

structure, and enhance our understanding of the process of technological
change at these three levels, we advocate the notion of vectors of
technological capabilities. These vectors evolve in time and space, and are
defined by competence and performance. In between these two, and shaping
their interaction (and hence the span that exists between them), lies the
national innovation system (NIS). Because this kind of innovation system
acts at both the national and regional levels, it also possesses an inherently
local nature.
Fig. 2.5 depicts the vector of technological capabilities at a given point in
time and space. It has permeable borders between the three zones, which are
the micro, meso and macro levels. The performance interface provides
feedback to and from all the other subsystems. Such a structure attempts to
assimilate the dynamics of each agent with the inherently systemic
properties of innovation mechanisms. The interesting feature from an urban
perspective is that the 'state' of a location at a specific point in time mimics
the mathematical state of a dynamic system along its chosen trajectory.
Because such systems are inherently nonlinear, a simplified study of the
characteristics of a point along one trajectory requires linearisation of the
system in the neighbourhood of such a point. We shall return to this issue
shortly.
At the chosen location in space, the microlevel capabilities reflect
educational skills, R&D capabilities, and technology transfer propensities.
At the macro level, policy indicators should be viewed as conditioning
elements and as aggregate measures of the system's performance. Such
macrovariables include GDP per capita, population growth, exports as a
percentage of GDP, the average inflation rate, interest rates, and real
exchange rates. Key features of this system are the interface between
capabilities and performance, and the role played by the NIS as the
representative 'arm' of (public and private) institutions. Knowledge flows
are triggered by individuals and their organisations, and thus we find that a
collection of institutions lies at the heart of the whole system.

.
Macro-system

Pei4form ance

...

NIS
Competence Performance

i

. .

NIS

Figure 2.5 A point in the technological capabilities vector defining
the state of an economic system
Can we measure the 'goodness' or 'badness' of a specific NIS? Given the

evolutionary micro-foundations introduced above, it may be possible to
explain why technological gaps arise among different locations (or nations)
over time. Such gaps may open up because individual responses produce
different (sometimes suboptimal) collective outcomes. The existence of
diverse institutions and organisations, with specific modes of interaction,
determine unique national systems of innovation which, over time, exhibit
certain invariant characteristics. It is these invariances which account for
phases of relative 'technological success and failure' (Cimoli and Dosi
1995). Thus we can see that well-organised systems of innovation serve as a
powerful motor of progress, whereas poorly organised systems can seriously
inhibit the whole process (Metcalfe 1995).
2.8 SOME CONCLUDING THOUGHTS ABOUT

INNOVATION AT THE URBAN LEVEL
Clearly, most of the observations at the national level are too aggregate to
allow a deeper understanding of the effect of the innovative process on
specific locations. To understand metropolitan development, we must focus
on the appropriate locational units of analysis. Key factors will be those
systems that are geographically and institutionally localised. One might ask,
for example, whether those metropolitan institutions and capabilities that are
supporting technical change are sufficiently responsive and adaptive to
adjust to whatever underlying economic changes that emerge from market
interactions. Conversely, we may ask whether they are sufficiently creative
to shape innovation and economic performance. In other words, local
institutions and networks of individuals need to be both proactive and
reactive as the principal agents of innovation.
All of the above is clearly an oversimplification of reality. The vector of
capabilities and the matrix of the NIS must be of compatible dimensions, and
there still exists the problem of defining and measuring such components.
Measurement is further complicated by the presence of many informal
relationships between organisations, institutions and sets of actors 7. In the
classical literature, such relationships are treated as 'externalities'. Thus
further investigation is needed to provide a solid foundation of support for
our chosen structural representation. Nonetheless, we believe that the
structure can help to understand those mechanisms by which a NIS may
determine the success and failure of technological progress, and thus the
positioning of the vector of technological capabilities (see Fig 2.6).
The NIS matrix is closely related to the "institutional matrix which
supports and sustains the activities of innovating firms" (proposed by
Metcalfe, (995). Although it has been depicted as a linear representation, it
should be borne in mind that this is only for the purpose of illustration. The
evolutionary foundations which account for the characteristics of national
systems of production and innovation dictate that firms are repositories of
knowledge, nested in various networks of linkages with other firms and
other non-profit organisations. Such networks enhance the opportunities for
each firm to improve their problem-solving capabilities. An even broader
notion (at a wider level of aggregation) tends to map microeconomic
behaviours into a set of social relationships, rules and political constraints
(Granovetter 1985).
Recalling North's analysis of institutional change, institutions are characterised by their

slow adjustment to changes induced by organisations (here, basically changes in
competence of business firms). They also, however, possess a competence of their own.
All of this implies that at the empirical level there is bound to be an overlap in the
variables used to measure firms' competence and NIS characteristics, but those used to
proxy for the NIS will be modelled as lagged variables to reflect the inertial adjustment.
Firms' competence should be understood in terms of effort, and analogously system's
competence as a representation of the aggregate potential ofthe system.
l\tK." IICIIIIIIdc Hiaba" fllb:alim

tdtbIc: in8a1ion inlDllllicm (HEls). rescan:h
~e, exc:hanjle rile, &: tecInoIogjcal LiIII'Ky rua (1ICOIIdDy
IIIC8I pdicy deYdClplllld Q1IIIIIisaIims
IIId tsIiIry lewia
........,.~ (RTOII),. iDcUIIIiaI ~
tIIIdma 1lli0l),. tird
hie policy, aedit IIIxnmes (IRLs),
.
1ewI ...... (ill IIIIIbt,
lei..,md...-.
1)'tIaII,~ .,~..-dI
ICitIxle IIId II!F-iIJl),
riP system, inIIituIcI (0RlI), rn1iIIIy iD
nwbt~ r-m inmIIhs (MRIs),.
mcdwipno lpIICiestirdaiiDlland
IDdateriIlIlld niIina 1II1J- bee. R.ctD. %GNP.
Indr poIIcia: (AETI), c:enillc:aud
Dilday ecpmcil\ll'e in
IDizcxuJ vs taeIIdl and tcdlnoIogy
R.ctD.lllio of pi~ vs
YCIIical poUcies, aaaaAlims(CRTOs),.
pubIie R.ctD. iIMcnaII ill
iotcrwnion II lIDdIloIosic:aI infmnetim ICi.all:lCJaIb, FDI
seaaiaI and Dian
level,
c:aIIn:S (TIQ),
apmSlliOlllIUppCl1ioa
.a. iqIOIU ofcaptal
pdI
~OIID1l scico;e I*b (0SSPs).
IecIIDoIccY and FDI ICdJodo&y lI1DSfcr.
IIOWI imlitucicm (11'11)
NIS (Institutional vector of ..t !ii~-:

.'
...
.... :.._~;.~.
.~
''''
matril:) :. - ,~~
~ . .
Figure 2.6 A simple representation of the interaction between

competence and NIS
The momentum associated with single technological trajectories is itself

a largely social concept. As Misa has argued:
"it points to the organisations and people committed by various interests
to the system, to manufacturing corporations, research and development
laboratories, investment banking houses, educational institutions and
regulatory bodies" (Misa 1991, pIS).
In tum, these interests and institutions are sustained by the positive
feedbacks (or increasing returns) associated with most learning activities.
Markets cannot exist or operate outside the set of rules and institutions that
established them in the first place. The institutional structure of the economy
creates a distinct pattern of constraints and incentives, which defines the
interests of the actors as well as shaping and channelling their behaviours
(Zysman 1994).
Cities and nations are characterised by particular modes of institutional
governance which make them self-reproducing entities. Moreover, an
element of nationality follows from the shared language and culture, and
38 Cimoli M., D!')lla Giusta M.
environment (Metcalfe 1995). Together, they shape the organisational and

technological context within which each economic activity takes place. In a
sense, they establish the set of opportunities and constraints facing each
individual process of production and innovation. Thus institutional and
technological diversities may be seen as the true determinants of
development. Because the processes described here are inherently co-
evolutionary (see Dosi and Nelson 1993; Nelson 1995; Dosi 1997), they are
characterised by positive feedback mechanisms. Such feedbacks link
performance and capability, driven on by the adaptive nature of learning-by-
doing. The role of institutional strategies is a fundamental part of this co-
evolution.
The basic approach is applicable at different levels of analysis: national,
regional, urban or local. For example, an urban system can be defined in
terms of a set of innovative efforts (and technological activities) within the
city, with which it is possible to associate a vector of economic performance
and approximate the interplay between such efforts and performance.
Economic performance is identified for each specific system. Clearly, this
fits well with the identification of different levels of analysis. At the national
level, there is the macro-economy containing industry and institutions whose
analysis is mainly based on the industrial technological specificity associated
with the traditional NIS concept (Freeman 1987; Nelson 1993). At the
regional and sectoral levels, boundaries come into play. Regional boundaries
aim at the identification of an area where a specific institutional matrix, a
vector of capabilities, and their interaction with industry, can be seen to
affect local economic performances. A Sectoral Innovation System (SIS)
could be defined as a "system (group) of firms active in developing and
making a sector's products and in generating and utilising specific sector
technologies. Performance at the regional or sectoral level can be viewed as
an effort dominated by local institutions and local capabilities" (Breschi and
Malerba 1997). However, performance may be affected by the overlap
between these two levels8 .
An interdependent system could result from the interaction of the
regional and sectoral levels. In this perspective, the empirical and theoretical
domain to which this system can be applied is related to the interaction, co-
operation and competition of firm's activities developed in a specific region.
In this context, the system itself could be characterised by different firms
localised at different stages in the 'value added chain'. For example, think of
the mechanical and textile industries and the interactions between them.
8 According to Carlsson and Stankiewicz (1995) and Edquist (1997) the systems of
innovation "may be supranational, national or subnational (regional and local), and at the
same time they may be sectoral within any of these geographical demarcations".
Another example is the close spatial relationship established between the

software and modern microelectronics industry. Successful regional systems
of innovation, like those located along Route 128 in Massachusetts or in
Silicon Valley, rely on the knowledge-oriented milieu associated with their
local areas to produce the capabilities and performances for which they have
become renowned.
By building on the micro-foundations laid by the evolutionary theory of
technical change, we have introduced the reader to the concept of national,
regional, urban and local systems of innovation and learning. Our
suggestions serve mostly as an introductory roadmap linking the micro-
foundations of technical change to system-wide performance conjectures.
Another aim of the chapter has been to demonstrate that the NIS concept is
based on a substantial body of theory (see e.g. Nelson 1993). Thus it should
not be seen as something new, but more as a useful means of broadening our
understanding of the developmental potential ofa given economic system.
A general thesis, widely acknowledged in the literature, is that learning is
both local and cumulative. Localised learning implies a clustering of
individuals and institutions at much the same location in geographical space.
Cumulative learning means that these same agents build upon past
experiences through sequences of problem-solving exercises. To learn in this
way requires interaction with others.
It came as no surprise to learn that the institutional dimension plays a
fundamental role in evolutionary theories of innovation. After all, at the
regional and urban levels, technology is largely incorporated in particular
institutions (mostly firms) whose characteristics, decision rules, capabilities
and behaviours are the key factors shaping technological progress.
Furthermore, the firms themselves are embedded in complex relational
networks linking them to other firms and institutional actors. Finally, the
emphasis on adaptive, local learning, together with rationally bounded
decision-making, is in keeping with the views of those political economists
and sociologists who stress that a major ingredient of development is the
process of change in social norms, expectations, and forms of collective
organisation. We would expect such ingredients to also lie at the heart of
urban development.
Acknowledgements
A number of people made helpful comments on this essay. The authors particularly
wish to thank the following, without implicating them in any way in the opinions
expressed: Giovanni Dosi, Paolo Guerrieri, Sanjaya Lall, Francesco Luna, and
Franco Malerba. The research on which this chapter is based benefited from the
support of a 40% grant from MURST, Crescita Endogena nelle Aree Arretrate
(Italian Ministry for University and Scientific and Technological Research).
REFERENCES
Amsden, A (1989) Asia's Next Giant: South Korea and the Last Industrialisation, Oxford
University Press, New York, Oxford
Amsden, A, Hikino, T. (1993) "Staying Behind, Stumbling Back, Sneaking Up, Soaring
Ahead: Late Industrialisation in Historical Perspective". In International Convergence of
Productivity with Some Evidence from History, Baumol, W., Nelson, R., Wolff, E., eds,
Oxford University Press, Oxford
Amsden, A, Hikino, T. (1994) Project Execution Capability, Organisational Know-How, and
Conglomerate Corporate Growth in Late Industrialisation. Industrial and Corporate
Change 3:111-47
Arthur, W.B. (1989) Competing Technologies, Increasing Returns and Lock-In by Historical
Events. Economic Journal 99( I): 116-31
Arthur, W.8., Ermoliev, Y., Kaniovski, Y. (l987) Path Dependent Processes and the
Emergence of Macro Structures. European Journal ofOperational Research 30: 294-303
Atkinson, AB., Stiglitz, I.E. (1969) A New View of Technological Change. Economic
Journal 79 (3): 573-78
Audretsch, D.B. (1996) Technological Regimes, Industrial Demography and the Evolution of
Industrial Structures. IIASA Working Paper 42, April 1996
Bardhan, P. (1996) The Political Economy of Development Policy: An Asian Perspective.
Development Thinking and Practice Conference, Sept 3-5, Washington DC
Batten, D.F., Kobayashi, K., Andersson, A. E. (1989) "Knowledge, Nodes and Networks: An
Analytical Perspective." In Knowledge and Industrial Organisation. Andersson, A. E.,
Batten D.F., Karlsson, C., eds, Springer, Berlin, 34-46
Bell, M., Pavitt, K.L.R. (1993) Technological Accumulation and Industrial Growth: Contrasts
between Developed and Developing Countries. SPRU, University of Sussex, Sussex
(mimeo)
Breschi, S., Malerba, F. (1997) "Sectoral Innovation Systems: Technological Regimes,
Shumpeterian Dynamics and Spatial Boundaries." In Systems of Innovation, Edquist, C.,
ed, Pinter, London Washington
Calion, M. (1991) Technological Development and Adoption Networks, Conference on
Constructive Technology Assessment, Twente, The Netherlands
Cantwell, AI. (\ 991) "The Theory of Technological Competence and its Application to
International Production." In Foreign Investment, Technology and Growth, McFetridge,
D.G., ed, University of Calgary Press, Calgary
Carlsson, 8., Stankiewicz, R. (1995) "On the Nature, Function and Composition of
Technological Systems." In Technological Systems and Economic Performance: The Case
ofFactory Automation, Carlsson, 8., ed, Kluwer, Dordrecht
Chandler, AD. (1990) Scale and Scope: The Dynamics of Industrial Capitalism. Harvard
University Press, Cambridge Mass
Chandler, AD. (1992a) Corporate Strategy, Structure and Control Methods in the United
States during the 20th Century. Industrial and Corporate Change I: 263-84
Chandler, AD. (1992b) Organisational Capabilities and the Economic History of the
Industrial Enterprise. Journal of Economic Perspectives, 79-100
Chandler, AD. (1993) "Learning and Technological Change: the Perspective from Business
History." In Learning and Technical Change, Thomson, ed, MacMillan, London
Cimoli, M. (1988) Industrial Structures, Technical Change and the Modes of Regulation in
the Labour Market. DRC Discussion Paper 60, SPRU, University of Sussex, Sussex
Cimoli, M., Dosi, G. (1988) "Technology and Development: Some Implications of Recent
Advances in the Economics of Innovation for the Process of Development." In Science,
Technology and Development: A Critical Perspective. Wad, A., ed, Westview Press,
Boulder Colorado
Cimoli, M, Dosi, G. (1990) "The Characteristics of Technology and the Development
Process: Some Introductory Notes." In Technology Transfer in the Less Developed
Countries, Chatterji, M., ed, Macmillan, London
Cimoli, M., Dosi, G. (1995) "Technological Paradigms, Patterns of Leaming and
Development. An Introductory Roadmap." Journal of Evolutionary Economics 5 (3) (also
Working Paper 83, IIASA, Laxenburg, Austria)
Constant, E. (1985), The Turbojet Revolution, Johns Hopkins University Press, Baltimore.
Coriat, B. (1990) Penser l'envers. Bourgeois, Paris
Dahlman, C., Sananikone, O. (1990) Technological Strategy in Taiwan: Exploiting Foreign
Linkages and Developing Local Capabilities, mimeo
Dahlman, C., Westphal, L. (1982) "Technological Effort in Industrial Development: A
Survey." In The Economics of New Technology in Developing Countries. Stewart, F.,
James, 1. eds, Frances Pinter, London
David, P.A. (1989) Path-Dependence: Putting the Past into the Future of Economics.
Discussion Paper, Stanford University, Stanford
Della Giusta, M. (1996) Education and Technological Capabilities in Industrialisation.
Master's Thesis, University of Reading, Reading
Dierickx, I., Cool, K. (1989) Asset Stock Accumulation and Sustainability of Competitive
Advantage. Management Science 35 (3), September
Dosi, G. (1984) Technical Change and Industrial Transformation. MacMillan, London New
York
Dosi, G. (1988) Sources, Procedures, and Microeconomic Effects of Innovation." Journal of
Economic Literature 26: 1120-71
Dosi, G. (1997) Opportunities, Incentives and the Collective Patterns of Technological
Change. Economic Journal 107: 1530-57
Dosi, G., Freeman, c., Nelson, R., Soete, L. (1988) Technical Change and Economic Theory,
Pinter, London
Dosi, G., Giannetti, R., Toninelli, P.A. (1993) Technology and Enterprise in a Historical
Perspective. The Clarendon Press, Oxford
Dosi, G., Kaniovski, Y. (1994) On Badly Behaved Dynamics. Some Applications of
Generalised Urn Schemes to Technological and Economic Change. Journal of
Evolutionary Economics 4: 93-123
Dosi, G., Lovallo, D. (1995) Rational Entrepreneurs or Optimistic Martyrs? Some
Considerations on Technological Regimes, Corporate Entries, and the Evolutionary Role
of Decision Biases. IIASA Working Paper 77, August
Dosi, G., Marsili, 0., Orsenigo, L, Salvatore, R. (1993) Learning, Market Selection and the
Evolution oflndustrial Structures. IIASA Working Paper
Dosi, G., Marsili, 0., Orsenigo, L., Salvatore, R. (1995) Leaming, Market Selection and the
Evolution oflndustrial Structures. Small Business Economics 7:411-36
Dosi, G., Nelson, R. (1993) "Evolutionary Theories in Economics: Assessment and
Prospects." In Market and Organisation: The Competitive Firm and its Environment,
EECffempus Programme, LATAPSES, Nice, France
Dosi, G., Orsenigo, L. (1994) "Macrodynamics and Microfoundations: an Evolutionary
Perspective." In Economics of Technology. Granstrand, 0., ed, Elsevier Science
Dosi, G., Pavitt, K., Soete, L. (1990) The Economics of Technical Change and International
Trade, Harvester Wheatsheaf Press, London, New York
Dosi, G, Teece, D. J., Winter, S. (1992) "Toward a Theory of Corporate Coherence:

Preliminary Remarks." In Technology and Enterprise in a Historical Perspective. Dosi,
G., Giannetti, R., Toninelli, P.A., The Clarendon Press, Oxford
Dursleifer, Kocka (1993) The Impact of the Pre-Industrial Heritage. Reconsiderations on the
German Patterns of Corporate Development in the Late 19th and Early 20th Century.
Industrial and Corporate Change 2: \04-18
Edquist, C., ed (1997) Systems ofInnovation, Pinter, London, Washington
Egidi, M. (1996) Virtuous and Adverse Selection within Economic Organisations. llASA
Working Paper 16, February
Enos, J., Park, W. (1988) The Adoption and Diffusion of Imported Technology: the Case of
Korea. Croom Helm, London
Ernest, D., O'Connor, D. (1989) Technology and Global Competition, OECD, Paris
Fransman, M. (1986) Technology and Economic Development. Wheatsheaf, Brighton
Fransman, M., King, K. (1984) Technological Capability in the Third World. Macmillan,
London
Freeman, C. (1982) The Economics ofIndustrial Innovation. Pinter, London
Freeman, C. (1987) Technology Policy and Economic Performance: Lessons From Japan.
Pinter, London
Freeman, C. (1994) The Economics of Technical Change. Cambridge Journal of Economics
18: 463-514
Freeman, C., Clark, H., Soete, L. (1982) Unemployment and Technical Innovation. A Study of
Long Waves and Economic Development. Pinter, London
Freeman, C.,. Perez, C. (1988) "Structural Crises of Adjustment: Business Cycles and
Investment Behaviour." In Technical Change and Economic Theory. Dosi, G. et aI, eds,
Pinter, London
Granovetter, M. (1985) Economic Action and Social Structure: the Problem of
Embeddedness. American Journal ofSociology 91: 481-5 \0
Guerrieri, P. (1993) International Competitiveness, Trade Integration and Technological
Interdependence in Major LA Countries. Paper presented at 9 th Scientific Convention, 30
Sept- IOct, Milan
Gurisatti, P., Soli, V., Tattara, G. (1997) Patterns of Diffusion of New Technologies in Metal
Working Small Firms. The Case of an Italian Region. Industrial and Corporate Change 6:
275-312
Hirschman, A.O. (1992) "Linkages in Economic Development." In Rival Views of Market
Society and Other Recent Essays. Harvard University Press, Harvard
Katz, 1. (1984) Domestic Technological Innovations and Dynamic Comparative Advantage.
Journal of Development Economics 16: 13-38
Katz, 1. (1987) Technology Generation in Latin American Manufacturing Industries.
Macmillan, London
Kim, L., Dahlman, C. (1992) Technology Policy for Industrialization. An Integrative
Framework and Korea's Experience. Research Policy 21: 437-52
Kogut, B. (1992) National Organizing Principles of Work and the Dominance of the
American Multinational Corporation. Industrial and Corporate Change 1: 285-326
Kogut, B. (1993) Country Competitiveness, Oxford University Press, Oxford
Lall, S. (1982) Developing Countries as Exporters of Technology, MacMillan, London
Lall, S. (1984) "India's Technological Capacity: Effects of Trade, Industrial Science and
Technology Policies." In Technological Capability in the Third World, Fransman, M.,
King, K., eds, Macmillan, London
Lall, S. (1992) Technological Capabilities and Industrialisation, World Development 20
(2): 165-86
Lundvall, B. (1993) National Systems of Innovation: Towards a Theory of Innovation and

Interactive Learning. Pinter, London
Malerba, F., Orsenigo, L. (1996) The Dynamics and Evolution ofIndustries. Industrial and
Corporate Change 5: 51-87
Malerba, F., Orsenigo, L. (1997) Technological Regimes and Sectoral Patterns ofInnovative
Activities. Industrial and Corporate Change 6: 83-118
Metcalfe, S. (1981) Impulse and Diffusion in the Study of Technical Change. Futures B: 347-
59
Metcalfe, S. (1988) "The Diffusion of Innovation: An Interpretative Survey." In Technical
Change and Economic Theory. Dosi, G., Freeman, C., Nelson, R., Soete, L., eds, Pinter,
London
Metcalfe, S. (1995) "The Economic Foundations of Technology Policy." In Handbook of the
Economics of Innovation and Technical Change. Stoneman, P., ed, Blackwell, Oxford
Misa, T.J. (1991) Constructive Technology Assessment: Cases, Concepts, Conceptualization.
Paper presented at the Conference on Constructive Technology Assessment, Twente, The
Netherlands, September
Mowery, D., Rosenberg, N. (1979) The Influence of Market Demand Upon Innovation: A
Critical Review of Some Recent Empirical Studies. Research Policy 8: 103-53
Nelson, R (1981) Research on Productivity Growth and Productivity Differences: Dead Ends
and Departures. Journal of Economic Literature 19: 1029-64
Nelson, R., ed (1993) National Systems ofInnovation, Oxford University Press, Oxford
Nelson, R. (1995) Recent Evolutionary Theorizing about Economic Change. Journal of
Nelson, R (1996) The Concept of Institution as an Attractor, Snare and Challenge. IIASA,
Laxenberg, (mimeo)
Nelson, RR, Winter, S.G. (1982) An Evolutionary Theory of Economic Change. Harvard
University Press, Cambridge Mass
Noble, D. (1984) Forces of Production, Knopf, New York
North, D. (1990), Institutions, Institutional Change and Economic Performance, Cambridge
University Press, Cambridge
Orsenigo, L. (1995) Technological Regimes, Patterns of Innovative Activities and Industrial
Dynamics. Cahiers d'economie et sociologie rurales 37: 26-69.
Pavitt, K. (1984) Sectoral Patterns of Technical Change: Towards a Taxonomy and a Theory.
Research Policy \3: 343-75
Pavitt, K. (1991) What Makes Basic Research Economically Useful? Research Policy 20 (2):
109-20
Pol any, M. (1967) The Tacit Dimension. Doubleday Anchor, New York
Rosenberg, N. (1976) Perspectives on Technology. Cambridge University Press, Cambridge
Rosenberg, N. (1982) Inside the Black Box: Technology and Economics. Cambridge
University Press, Cambridge
Rosenberg, N. (1990) Why Do Firms Do Basic Research (With Their Own Money)? Research
Policy 19: 165-74
Rosenberg, N. (1991) Critical Issues in Science Policy Research. Science and Public Policy
18: 12-18
Rothwell, R., Walsh, V. (1979) Regulation and Innovation in the Chemical Industry. OECD,
Paris (mimeo)
Schmookler, J. (1962) "Economic Sources of Inventive Activity." In The Economics of
Technological Change, Rosenberg, N., ed, 1971
Schumpeter, J.A. (1934) The Theory of Economic Development. Harvard University Press,
Cambridge Mass
Schumpeter, J.A. (1943) Capitalism. Socialism and Democracy, Harper and Row, New York
Silverberg, u.; Dosi, G., Orsenigo. L. (1988) Innovation, Diversity and Diffusion: A Self-
Organisation Model. The Economic Journal 98: 1032-54
Simon, H.A. (1955) A Behavioural Model of Rational Choice. Quarterly Journal 0/
&onomics
Teece, D. (1998) Strategic Technology and Public Policy. Selected Papers, Edward Elgar,
London
Teece, D.J., Pisano, G., Shuen, A. (1990) Firm Capabilities, Resources, and the Concept of
Strategy. University of California at Berkeley (mimeo)
Teitel, S. (1981) Towards an Understanding of Technical Change in Semi-Industrial
Countries. Research Policy 10: 127-47
Teitel, S. (1984) Technology Creation in Semi-Industrial Economies. Journal 0/ Development
Economics 16,: 39-61
Teitel, S. (1987) "Towards Conceptualisation of Technological Development as An
Evolutionary Process." In Structural Change. Economic Interdependence and World
Development. Dunning, J., Mikoto, U., eds, Macmillan, London
Teubal, M. (1984) "The Role of Technological Learning in the Exports of Manufactured
Goods: the Case of Selected Capital Goods in Brazil", World Development, 12: 349-865.
Teubal, M. (1987) "Innovation and Development: A Review of Some Work at the
IDBIECLNUNDP Programme." In Technology Generation in Latin American
Manu/acturing Industries. Katz, 1., ed, Macmillan, London
Thomson, R. (1993) Learning and Technical Change. Macmillan, London
Utterback, 1., Abernathy, W. (1975) A Dynamic Model of Process and Product Innovation".
Omega Vol.3
Verspagen, B. (1990) "Localised Technological Change, Factor Substitution and the
Productivity Slowdown." In New Explorations in the Economics o/Technological Change.
Freeman, c., Soete, L., eds, Pinter, London
Vincenti (1992) What Do Engineers Do and How Do They Know It? Johns Hopkins
University Press, Baltimore
Wade, R. (1990) Governing the Market: Economic Theory and the Role o/Government in
East Asian Industrialisation. Princeton University Press, Princeton, New Jersey
Walsh, V. (1984) Invention and Innovation in the Chemical Industry: Demand Pull or
Discovery Push? Research Policy 13: 211-34
Winter, S. (1982) "An Essay on the Theory of Production." In &onomics and the World
Around It. Hymans, S.H., ed, University of Michigan Press, Ann Arbor
Winter, S. (1984) Shumpeterian Competition Under Alternative Technological Regimes.
Journal 0/Economic Behaviour o/Organizations 5: 287-320
Zysman, 1. (1994) How Institutions Create Historically Rooted Tr*ctories of Growth.
Industrial and Corporate Change 3: 243-83
Chapter 3
Emergence and Co-Evolutionary Learning in

Self-Organised Urban Development
David F. Batten
The Temaplan Group, Melbourne, Australia
3.1 INTRODUCTION
The celebrated astronomer, Carl Sagan, reminded us all some time ago
that:
"the pattern of evolutionary causality is a web of astonishing complexity;
the incompleteness of our understanding humbles us," (Sagan 1980)
Almost forty years earlier, the noted economist Joseph Schumpeter
delivered a similar cautionary message:
"The process of social life is a function of so many variables, many of
which are not amenable to anything like measurement, that even mere
diagnosis of a given state of things becomes a doubtful matter, quite apart
from the formidable sources of error that open up as soon as we attempt
prognosis." (Schumpeter 1942)
Causality remains a mystery in the case of urban development. Like most
problems in evolutionary causality, the difficulty is that too many variables
are involved. With the recent hindsight of nonlinear dynamics, we are
beginning to realise that a city's future path of development cannot be
predicted. If it cannot be predicted, it certainly cannot be optimised. Most of
the world's oldest cities have grown haphazardly, bit by bit in response to
the needs and opportunities of the moment. No compact, lawlike description
of their evolution can be found. Some continue to grow, others have fallen
from grace. The causes of one's survival and another's demise are difficult
46 Batten D.F.
to discern and impossible to generalise. Perhaps the best we can do is simply

stand back and watch the pageant unfold.
A city's development has certain features in common with the evolution
of the human brain. The brain's neurochemistry is astonishingly busy, its
circuitry being more incredible than any machine devised by humans. It
develops from a small centre, grows slowly and changes discontinuously.
Innovative replacement can only be selective, since the system must
continue to function during any renovation. The functions of many old parts
are too vital for them to be replaced altogether. They possess vintage
properties, in the sense that some bits are older and less reliable than others.
Nevertheless these parts struggle on, mostly out-of-date and sometimes even
counter-productive, an inevitable consequence of evolution.
In this chapter, I argue that cities and brains actually have a great deal
more in common. I shall argue that both are living organisms whose patterns
of development display the traits of complex dynamic systems. Perhaps the
most important feature they have in common is that cities and brains are co-
evolving complex systems. An important discovery in biology is that
organisms do not merely evolve, rather they co-evolve, both with other
organisms and with a changing environment. Co-evolution can produce
surprising outcomes, often referred to as emergent behaviour. To further our
understanding of urban change, we must strive to understand both these
processes: emergence and co-evolution.
Although we can never hope to predict the exact branchings of cities or
brains, we can hope to uncover laws that govern their general behaviour. We
must seek explanation rather than prediction. The aim of this paper is to
describe the nature of emergence and co-evolutionary learning and to
demonstrate their powerful explanatory potential when it comes to various
facets of urban development.
3.2 SELF-ORGANISING URBAN NETWORKS
3.2.1 Ziprs Law and City-Size Distributions
What do we really know about any nation's system of human

settlements? One thing we know is that small towns are far more common
than large cities. The same is also true for earthquakes, storms, avalanches,
wars and stock-market price gyrations (see Bak 1996). In 1949, George
Kingsley Zipf observed some striking regularities in systems of human
origin (see Zipf 1949). One of his observations linked city size to frequency
of occurrence. Fig. 3.1 shows how the population of each American city in
Emergence and Co-Evolutionary Learning 47
1950 occupied a unique position in that nation's urban hierarchy. The size
distribution of cities corresponded closely to a rank-size relationship. If we
plot population size against rank for every urban centre in the United States,
then the resulting downward-sloping curve is roughly a straight line on a
logarithmic plot. This simply means that the size of each centre is inversely
proportional to its population rank.
~
~ 250
CI'j~
~i /00
~~s
so
Q..
C~
~ 10
I
/ 5 10 2S /00 ZOO 500 ~OOO 2pOO
R.ANK
Figure 3.1 Rank-size distribution of urban centres in the United States,
1950 (Source: Vining 1955)
48 Batten D.F.
Zipf produced similar plots for many geographical regions in the 1920s
and found the same kind of regularity. Subsequent research has confirmed
that this rank-size rule relationship also seems to hold outside the United
States. In a large comparative study of thirty-eight countries at varying levels
of economic development, Berry (1961) showed that thirteen of these
countries had city-size distributions that conformed to the rank-size rule (see
Table 3.1). Later studies have also confirmed that rank-size distributions
exist in some of the nations of South-East Asia (e.g. Sendut 1966). The
regularity expressed by the straight lines in many of these logarithmic plots
of size versus rank, with slope near unity, is known as Zipfs Law.
Table 3. J City-size distributions in selected countries
Countries with Countries with Countries with

rank-size pattern primate pattern intermediate pattern
Belgium Austria Australia
Brazil Ceylon Canada
China Denmark Malaya
EI Salvador Dominican Republic New Zealand
Finland Greece Nicaragua
India Guatemala Norway
Italy Japan Pakistan
Korea Mexico England and Wales
Poland Netherlands
South Africa Peru
Switzerland Portugal
United States Spain
Sweden
Thailand
Uruguay
Source: Based on Berry (1961)
But what about the twenty-five countries in Table 3.1 whose distributions
did not follow the rank-size rule? Fifteen exhibited a primate distribution, in
which one or two very large cities dominated and there was an absence of
medium-sized cities. The rest revealed patterns that could be seen as
intermediate between rank-size and primate. In these cases, one or two cities
were rather dominant, but the overall pattern was generally closer to rank-
size than to primate.
Is there a simple explanation for the differences? The table itself does not
suggest any obvious explanation. While many of the lesser developed
nations (like Guatemala, Mexico, Peru and Thailand) do have primate
patterns, others (like EI Salvador and India) do not. Conversely, a number of
advanced European nations (like Austria, Sweden and the Netherlands) seem
to display a primate pattern rather than a rank-size distribution. On the basis
of Berry's table alone, we feel obliged to conclude that numerous factors
probably account for the differences - factors such as stage of economic
development, changes in political status, length oftime urbanised, and so on.
Before we accept such a conclusion, however, two observations warrant
our attention. First, there is no convincing argument to support the arbitrary
choice of national frontiers as the most appropriate way of defining a system
of cities. Quite different results can be achieved after taking cultural and
linguistic borders into account. More importantly, an historical analysis of
flows and interaction patterns between all candidate cities is a more reliable
way of defining a truly interactive system of cities. This method would
suggest that the towns and cities of Malaysia and Singapore should be
regarded as part of a single distribution of cities which, when treated in this
way, is approximately rank-size (see Sendut 1966). Furthermore, in Japan
one finds a rank-size distribution of cities once Tokyo's chief rival is seen to
be the multi-centred Kansai or Keihanshin conurbation (see Ginsburg 1988;
Batten 1995).
Second, there is no reason to expect that very different systems of cities
will converge to the same kind of city-size distribution at any single point in
time. After all, we are comparing the outcomes of complex evolutionary
processes which have commenced from different initial conditions and are
evolving in different directions at different speeds. Some settlement systems
are already well-developed and strongly interactive. Others remain largely
immature and only weakly interactive. The very fact that one system
displays a rank-size distribution, while another remains primate, merely
confirms that we are dealing with different classes of settlement systems.
Some of these are dynamically stable. Others may be relatively unstable.
3.2.2 Macro-Stability and Power Laws
In some countries, the rank-size relationship has been remarkably stable

over long periods of time. Take the case of the United States. Fig. 3.2 shows
that the slope of the rank-size relationship has remained more or less the
same since 1790, despite the fact that the American urban system has grown
enormously since that time. The rank-size structure of the Malaysia-
Singapore system of cities has also remained fairly stable since the 1950s.
50 Batten D.F.
25'000
I ~ ______L - - L__- L____ ~~-L--~--~--~~
I 5' 10 25 /00 200 sao 1.000 2,tXJO ~G\70
RANK
Figure 3.2 Stability of the rank-size distribution of urban centres in the
United States, 1790-1990
Why should the size distribution (or macrostructure) of some nation's

system of cities display such stability over time? Zipf (1949) attributed this
underlying regularity of city-size structures to the individual agent trying to
minimise his or her effort. But he gave no clues about how to get from this
individual level to his aggregate statistical observations. Berry (1961)
suggested that as the economic, political and social life of a country becomes
more complex, its city-size distribution will tend to develop towards a rank-
size pattern, which represents the steady-state of an urban system.
Berry's suggestion raises two intriguing questions:
1. Under what conditions can a nation's economic, political and social life
be regarded as complex?
2. Could the rank-size pattern be a sufficiently strong attractor (in the case
of the United States) that the individual towns and cities self-organise in
ways which ensure that the system as a whole preserves its rank-size
structure over time?
Let us deal with the second question first. There is evidence to support
the idea that the rank-size pattern is an attractor. Despite the high degree of
macro-stability shown in Fig. 3.2, the relative position of individual cities in
the U.S urban system has varied considerably. Between 1870 and 1950, for
example, Los Angeles leapt ahead 314 places to reach the top five. Among
those cities with a population of more than 100,000 in 1950, 14 enjoyed a
relatively stable ranking but another 10 experienced changes of more than
100 places. This is something like a forest whose contours never seem to
change, but whose individual trees do. Perhaps the towns and cities self-
organise into an equilibrium pattern or rank-size 'ecology'.
Similar macro-stability can also be found in several city-size distributions
in Eastern Asia. Indonesia, Japan, Malaysia-Singapore, South Korea and
Taiwan have more-or-Iess preserved their rank-size distributions over the
last fifty years, despite the fact that many individual towns and cities have
changed their rankings markedly. Although Zipf's law implies this stability
in the number of cities within a given range of population sizes, it cannot tell
us why a particular city has a certain number of inhabitants. All the
observations are of a collective or statistical nature.
What is so magical about the rank-size distribution that it can command
such rigid adherence? The rank-size distribution is emergent in the sense that
it is not an obvious consequence of the underlying dynamic rules governing
the behaviour of the individmll elements (i.e. towns and cities).
Mathematically speaking, however, Zipf's Law is nothing more than a
specific kind of power law. Power laws express a quantity N as some power
of another quantity s:
N(s) = s-t (3.1)
By taking logarithms of both sides of the above equation, we find that:
log N(s) =-t log s (3.2)

The exponent t is the slope of the straight line formed when log N(s) is
plotted against log s. In Zipf's law, the number N corresponds to the number
of cities with more than s inhabitants, and is given by N(s) = lis = S-I. It is
52 Batten D.F.
simply a power law with exponent -1. Thus the problem of explaining the
observed statistical features of complex systems such as rank-size
distributions of cities can be seen as the problem of explaining the
underlying power-law relationship.
3.2.3 Sand piles and Self-Organised Criticality
City size distributions are not the only phenomena which can be
expressed as straight lines on a double logarithmic plot. Various catastrophic
events, fractals and noise also exhibit this pattern (see Bak 1996). Such
coupled dissipative phenomena cannot be understood by studying them
within a time frame which is short compared with the evolutionary process
which created them. For example, the laws governing earthquakes cannot be
understood just by studying earthquakes occurring in a single human
lifetime.
A special feature of coupled dissipative systems is that they evolve
naturally towards a self-organised critical state. To illustrate the basic idea of
self-organised criticality in an interactive system, consider a simple pile of
sand. Suppose we start from scratch and build our pile by randomly adding
sand, one grain at a time. The pile will grow and the slope will increase until
eventually it reaches a critical value. If more sand is then added, it will slide
off. Alternatively, if we had started off from a situation where the pile had
been too steep, it would have collapsed until it reached the same critical
state. By 'self-organised', we mean that the system evolves naturally to this
critical state, without detailed specification of the initial conditions. The
critical state is an attractor for the dynamics. It is just barely stable with
respect to further perturbations.
As the pile is built up, the characteristic size of the largest sandslide (or
avalanche) grows, until at the critical point there are avalanches of all sizes
up to size of the whole system. Once this critical point is reached, the system
stays there. This distribution of avalanches in a sandpile also happens to
obey Zipf's law. Perhaps we should be thinking of cities as being formed by
avalanches of human migrations (see Fujita 1996)? More importantly, we
might ask whether certain nation's city-size distributions evolve
spontaneously towards a critical state and later return to it, even if perturbed
at some later stage.
Self-organised systems start from disordered initial conditions and tend
to move to more highly ordered behaviour. This highly ordered behaviour
exhibits surprisingly simple regularities, such as rank-size distributions. One
can imagine with city-size distributions that there might be many potential
outcomes depending on initial conditions. But a handful of these potential
outcomes - namely those self-organised ones in which city sizes roughly
obey the rank-size rule - might be more or less independent of historical

events in anyone location. After all, a rank-size distribution does not depend
on which particular cities belong to the various size classes. It simply defines
the overall shape of the distribution. It might therefore be a favoured
outcome under certain rather common conditions.
3.3 LOCK-IN, INTERACTIONS AND EMERGENCE
3.3.1 Lock-In Based on Interaction Patterns
Some urban networks persist for many centuries. In New York City, for
example, the arrangement of many of the major streets dates back to the
seventeenth century, the stock exchange to the eighteenth century, the
waterworks to the nineteenth, and the electrical power system to the
twentieth. In the seventeenth century, you travelled between Brooklyn and
Manhattan across the East River by ferry. In the nineteenth century, new
technology allowed the construction of a suspension bridge across the river.
It was built precisely at the site of the ferry terminal. When newer
technology permitted the construction of a tunnel under the river, it too was
built in exactly the same place. This reinforcement of existing circulatory
systems by upgrading the technology of critical links is very much like the
pattern of biological evolution. It also provides some insights into
technological choices which permit the city to function more or less
continuously over many centuries.
Given the vagaries of political decision-making and the costs of altering
large infrastructure networks, it is rather easy for a city to 'lock-in' to a
particular arterial system which was perceived at the time to possess a socio-
economic advantage (see Arthur 1994). Lock-in can occur as sequential
decisions carve out an advantage which a city (or more correctly a city
government) finds it impossible to escape from at a later date. In the
nineteenth century, for example, Philadelphia was a more important port
than New York. The opening of the Erie Canal then tipped the scales in
favour of New York. Although the Erie has been more of a tourist attraction
than a serious transportation route for more than a century, New York still
dominates the USA's urban hierarchy. We shall never know whether
Philadelphia might have achieved that status if the Erie Canal had never
materialised. New York moved ahead and has never looked back. In cases
such as these, history really matters.
The growth of New York versus Philadelphia highlights the importance
of circulation and interaction patterns within and between cities. Circulation
54 Batten D.F.
and interaction are the lifeblood of human activity. Although urban

infrastructure networks evolve relatively slowly, the patterns of human
interaction which they facilitate can change quickly and unexpectedly (see
Batten and Johansson 1985). This is the signature of co-evolution. To
identify the co-evolutionary potentials of various cities, we need a deeper
understanding of what Karl Marx called the sphere of circulation. In modern
jargon, we must study the games of circulation and interaction which are
played out on those logistical networks within and between cities.
Why are interaction patterns so important? The quick answer is because
they provide fundamental clues to how a city really functions and changes.
Again there are striking similarities between the functioning of a city and the
workings of the brain. We know precious little about how nerve cells work
and we lack rudimentary circuit diagrams for the brain. Even an expert
electronic engineer would have trouble understanding how a circuit worked
if he didn't know what its components did or how they were linked together
(Cohen and Stewart 1994). But this is precisely the situation in our cities.
Each person knows very little about what every other person or urban
organisation really does, and has limited knowledge of the circulatory
patterns which a city generates. Yet the city somehow manages to survive
despite all this uncertainty!
The sheer complexity of each system poses fundamental problems. For
example, the brain has ten million neurons, each connected to (roughly) a
hundred others. An accurate circuit diagram would feature one trillion
connecting wires! This requires a very large computer system for its storage.
Similar problems arise when we try to study the interactions between or
within urban populations. Even when city dwellers seem to be doing very
simple things, their interactions can quickly add up to more complexity than
we can handle.
In an urban population, for example, the maximum number of ways for
pairs of people to interact is very close to half the square of the number of
inhabitants. As the city's population increases, the number of potential
interactions increases much faster (see Table 3.2):
Table 3.2 Interactions between city inhabitants
Number of inhabitants Possible maximum number of interactions

100 4,950
1,000 499,500
10,000 49,995,000
100,000 4,999,950,000
1,000,000 499,999,950,000
10,000,000 49,999,999,950,000
Obviously each inhabitant cannot interact with all the others, so the
population's interactive potential falls well short of the figures in the right
hand column. But the key message to be learnt from this is more subtle.
Once the ratio of interactions per inhabitant passes a critical threshold, then
the qualitative impact of that population's interaction patterns can change
unexpectedly. The following simple experiment helps to illustrate the
unusual nature of these critical thresholds of interaction.
3.3.2 Buttons, Threads and Phase Transitions
It is often helpful to look at a simpler toy problem first when trying to

understand a more complex one. The toy problem to be discussed here was
devised by Stuart Kauffman (1995) and involves random graphs. A random
graph (or random network) consists of a set of nodes connected at random by
a set of links. On paper it looks like a set of dots connected at random by a
set of lines (see Fig. 3.3). To visualise this particular toy problem in an
everyday context, we can think of the dots as 'buttons' and the lines as
'threads'.
Imagine 100 buttons scattered on a wooden floor. Now randomly choose
two buttons and connect them with a thread. After putting this pair down,
randomly choose two more buttons, pick them up, and connect them with a
thread. As you continue to do this, at first you will almost certainly pick up
buttons that you have not picked up before. Sooner or later, however, you
are more likely to pick at random a pair of buttons and find that you have
already chosen one of the pair. So when you tie a thread between these two
buttons, you will find that you have linked together three buttons. As you go
on choosing pairs of buttons randomly to link together with a thread, you
will find that some of the buttons soon become interconnected into larger
clusters. This is shown in Fig. 3.3. Each connected cluster is known as a
component in our random graph.
The thing to notice is that random graphs show very regular statistical
behaviour as we increase the ratio of threads to buttons. Once the ratio of
threads to buttons passes the 0.5 mark, something seemingly magical occurs.
All of a sudden most of the clusters become cross-connected into one giant
structure! When this giant component forms, most of the buttons are directly
or indirectly connected. As the ratio of threads to buttons approaches one,
virtually all of the remaining isolated buttons and small clusters become
cross-connected into the giant component.
This rather sudden and unexpected change in the size of the largest
cluster of buttons, as the ratio of threads to buttons passes 0.5, is the
signature of something resembling a phase transition (see Fig. 3.4).
56 Batten D.F.
a) Edges I Nodes = 5/20 b) Edges I Nodes = 10/20
@
0
CD 0

@
@
@
/
17
c) Edges I Nodes = 15/20 d) Edges I Nodes = 20/20
Figure 3.3 The crystallisation of connected webs. (As the ratio of edges
to nodes exceeds 0.5, most points become connected; once it
exceeds 1.0, closed pathways begin to emerge)
e) Edges 1Nodes = 25/20
Figure 3.3 The crystallisation of connected webs (continued)
~
~
~ 400~----------------------------------~
~
h..
t1 300
~
'~"
-..../
200
100
to
I\j
(;j
o
o 0.5 1.0 1.5
RATIO OF TJlREAJ)S/BUTTONS
Figure 3.4 A phase transition (Source: Kauffman 1995)

58 Batten D.F.
Note that the curve is S-shaped or sigmoidal. The size of the largest
cluster of buttons increases slowly at first, then rapidly, then slows again as
the ratio of threads to buttons increases further. Were there an infinite
number of buttons, then as the ratio of threads to buttons passed 0.5, the size
of the largest component would jump discontinuously from tiny to
enormous. The steep part of the curve would become more vertical than it is
in Fig. 3.4. This is typical of a phase transition, just like when separate water
molecules freeze into a block of ice.
3.3.3 Emergence and Urban Complexity
Now let's return to our urban population. Consider what might happen if
some residents start to interact more frequently with certain other residents.
As the number of interactions increases, clusters of 'like-minded' residents
begin to emerge spontaneously. Like-minded residents do not know in
advance who their closest friends and allies may be. These kinships can only
emerge after-the-fact. Furthermore, such like-minded clusters may do more
than simply interact among themselves. In order to pursue their common
interests more widely, eventually they may link up with other like-minded
clusters; thereby creating even larger clusters. Sounds familiar, doesn't it?
People behaving like buttons and threads!
In fact, the socialisation process whereby common interest groups
emerge and develop is analogous to the toy example discussed in the
previous section. People form clusters (e.g. political parties, unions, clubs,
teams) in order to pursue their common interests. These clusters represent
decision-making units which can play a powerful role in the shaping of
urban development. But because such clusters must take their decisions
collectively (e.g. by preference voting), sometimes the outcomes can be
quite different from those intended or expected by some of their members.
We shall refer to these unpredictable outcomes as emergence.
For the purposes of urban research, it is helpful to think of a city as a
complex social system which results from a large number of interactions
between its inhabitants. The defining characteristic of this complex urban
system is that some, perhaps even many, of its collective properties cannot
be predicted simply from our individual knowledge of these underlying
interactions. The whole urban system is greater than all of its parts! Some
collective outcomes can be predicted, but others cannot. They only emerge
as byproducts of the collective experience.
Complex systems breed emergence. For example, a car is a complex
system because its ability to move cannot be predicted simply by knowing
the individual rules which govern its parts (axles, gears, carburettors, etc.). A
car's motion is an emergent property - a process that it can carry out only by
virtue of its unique collective organisation. Your mind appears to be an

emergent property of your brain. Mind is a process, not a thing, and it
emerges from the collective interactions of appropriately organised pieces of
ordinary matter. Life itself seems to have emerged from chemistry by way of
DNA.
In the urban context, voting outcomes at public meetings can be swayed
spontaneously by the interactive debate and discussion of the key
protagonists. Many meeting decisions are unpredictable in advance. As we
shall see later, traffic jams on city highways emerge from the collective
interactions of a critical density of drivers on a transportation network.
Similar kinds of emergent phenomena can arise in many other social
situations. Their common feature is that the population of interacting
individuals can 'spontaneously' develop collective properties that may not
have been intended or expected by each individual a priori. Each of these
spontaneous outcomes is an example of emergence.
In this paper, our modest aim is to begin to examine the role which
emergence plays in urban evolution. We think of emergent phenomena as
any large scale, group behaviour which does not seem to have any clear
explanation in terms of the system's constituent parts (Darley 1995). Only
some of the collective phenomena observed in cities will be emergent. Our
aim is to explore a few of these in depth and to form some tentative
conclusions about their importance in the overall process of urban
development.
An important point to recognise is that urban systems may exhibit non-
emergent behaviour under some conditions, but generate emergent behaviour
under others. Emergence may simply be the result of a phase change in the
amount of computation necessary for the optimal prediction of certain
phenomena (Darley 1995). As the size and rule complexity of a city
increases, more and more of the useful predictive knowledge about it will be
contained in the accumulation of interactions.
Surprising outcomes that can alter the course of urban evolution may
occur when certain people alter their own interactive behaviour unexpectedly
in response to changes in perceived opportunities. For example, the
emergence of a completely new urban hierarchy may be set in motion by
little more than the opportunity for travellers to move about more quickly or
easily between places. Remember how the Erie Canal favoured New York
ahead of Philadelphia. Seemingly small changes to transport or
communication networks can generate an entirely different set of
entrepreneurial opportunities. Such changes can facilitate learning-by-
circulating, from which novel forms of urban and interurban
entrepreneurship may emerge. This is the subject of Section 3.4.
60 Batten D.F.
Because learning processes are co-evolutionary, small changes in the

behavioural ecology of urban residents can also have profound effects on the
urban landscape as a whole. For example, transportation and communication
networks provide the urban arena on which an ever-increasing volume of
human interactions accumulates. In response to growing congestion, the
more innovative residents search for novel ways to exploit these networks.
As their numbers grow, the resulting temporal and spatial innovations
increase the network's throughput. Nevertheless, more conservative
residents are unwilling to alter their behaviour until the network is modified.
Co-evolutionary learning is discussed more fully in Section 3.5.
3.4 LEARNING-BY-CIRCULATING
3.4.1 The Pirenne-Mees Hypothesis
According to the Belgian historian, Henri Pirenne (1925), the sudden

growth of population and the revival of towns and cities in Medieval Europe
during the eleventh century resulted from the expansion of trade over longer
distances. Such long distance trading was possible only after key
transportation routes were opened up again and the safety of travelling
merchants could be guaranteed. Of course the Mediterranean was never
closed off completely to enterprising Oriental merchants, who travelled in
convoys to trade precious Eastern goods with Christian, Jew and Moslem
alike. But Pirenne's hypothesis aimed to explain the sudden emergence of a
European merchant class. Although controversial and still disputed by some
historians, his views have found support in Fernand Braudel's great trilogy
(1979) charting the history of Western society.
Medieval Europe in the Carolingian period was a rather sparsely
populated continent. Villages, or groups of villages, were mostly self-
contained, forming small oases of cultivated land in a largely uncultivated
continent. The dangers of sea travel together with the difficulties of overland
travel forced it into an isolationist state in which many small, self-sufficient
settlements (demesnes) refrained from exchanging goods over longer
distances. Most bulky items were produced and exchanged at local markets.
Manors sought self-sufficiency and petty control of trade. Exchanges over
long distances were mainly restricted to luxury goods peddled by Orientals.
In economic terms, there was a strong tendency towards autarky.
Yet some popular goods like salt, metals and wine were not produced
locally and had to be found. High quality wool was also popular. If such
things could not be obtained by war and plunder, the last resort was to
engage in trade over longer distances. Although some long-distance trade

was carried on by the Oriental merchants, the total amount was insignificant
for Europe. Then, quite unexpectedly, the circulation of goods intensified.
During the latter part of the tenth century, all forms of trade grew
significantly, but is was that over medium and longer distances which
expanded most of all. Curiously, this sudden expansion of trade occurred at
the same time as large increases in urban populations.
The Cambridge mathematician, Alistair Mees (1975) analysed the effect
of increasing trading opportunities on urban and rural populations in terms of
catastrophe theory. His dynamic analysis is interesting because it shows us
how a stable, self-sufficient society could change dramatically when exposed
to the vagaries of circulation and trade - in other words, an accumulation of
interactions. It may also be an important example of emergent behaviour at
the urban level, i.e. an urban phase transition. A summary of his analysis
follows.
During the Medieval period, regional populations consisted mostly of
farmers (group f) and city merchants (group c). Of course there were a few
land owners and Oriental merchants, but they were vastly outnumbered by
farmers and city merchants. Furthermore, the self-sufficient nature of most
towns and villages meant that each region's total population (farmers plus
city merchants) remained roughly constant.
The farmers and city merchants depended only on each other for their
needs. Thus the attractiveness of becoming a farmer or a city merchant could
be assessed in simple terms, namely a comparison of each group's
production compared with the demand for their products. Mees' basic idea
was that if the attractiveness of belonging to one group (f or c) changes
suddenly, people will try to move from the less to the more attractive
situation. The dynamics of such a situation can be understood with the aid of
Fig. 3.5.
Em is a stable equilibrium point, corresponding to a mixed region
employing both farmers and city merchants. It is stable because, on either
side of this point, small changes in each group's working population will not
alter markedly the relative attractiveness of either group. The points Ec and
Ej are unstable. Because there is demand for both products, complete
specialisation in farming (the point Ef ) or city merchandising (the point
Ec)will always be inferior to a balanced mix of farmers and city merchants
(the point En,). Thus we can see how a stable and self-sufficient Medieval
economy could be attained; with a balanced mix of farmers and city
merchants.
However, this balanced mix of farmers and city merchants could easily
disappear if trading opportunities with distant towns became easier.
Pirenne's idea was that safer transportation routes did in fact make trade
62 Batten D.F.
easier. If some enterprising city merchants felt that the risks and costs of
travel over longer distances were acceptable, they would visit more distant
places to sell their goods more profitably. Such a quest for more distant
markets would be more compelling once local demand had been satisfied. If
these early merchants reported profitable trading ventures, then other city
merchants would be more likely to follow their lead and eventually engage
in trade.
RATE OF CHANGE
IN FA.RMEIlS
..RATE OF CHANGE
IN Clry MERCHANTS
Figure 3.5 Dynamics with no trade: stable and unstable equilibria

(Source: Mees 1975)
By raising the relative attractiveness of becoming a merchant instead of a

farmer, this new class of European merchants (willing to trade over longer
distances) served to flatten out the curve between Ec and Em (see Fig. 3.6).
As this new breed of merchants continued to expand, eventually a stage was
reached where the curve between Ec and Er dropped below the axis. The
stable equilibrium point Em disappeared and the previously unstable Ec
stabilised. Everyone switched from farming to trading, farming died out and
the region then specialised in mercantile activities.
RA TE OF CflANGE
IN FARMERS
~c'-----------------------~~--------~Er
RATE OF CHANGE
IN CITY MERCHANTS
Figure 3.6 Catastrophic change as trade costs decrease (Source: Mees 1975)
This is Mees' explanation for the simultaneity of selective urban growth

and the escalation of trade during the Medieval period. Gradual
improvements to the transport system led to more travelling merchants,
growing trade and greater specialisation. Some regions specialised in city
goods, others in agricultural commodities. If each region's total population is
also assumed to vary with its attractiveness, overall growth can also be
explained. Additional arguments based on increasing returns to scale and
agglomeration could also be drawn upon to show why some cities expanded
much more rapidly than others or than their rural surroundings (Fig. 3.7).
SPECIALISATION
INCREASING
CIRCULATION AND ,. AND INCREASING
"-
RETURNS TO
TRADE
SCALE
II'
,
IMPROVEMENTS TO
SELECTIVE
TRANSPORT AND ~---~
GROWTH IN URBAN
COMMUNICATION
POPULATIONS
NETWORKS
Figure 3.7 Cumulative causation in urban growth and development

64 Batten D.F.
3.4.2 The Innovative Merchant's Decision Problems
The travelling merchant was basically a broker whose primary capacity

was to observe and exploit price differences (Andersson 1995). He sought to
profit from buying goods in excess supply, then selling them at a higher
price in places of excess demand. Whereas city merchants confined their
trade to local markets, travelling merchants realised that geographical space
(in the form of more distant trading centres) provided a richer set of profit-
making opportunities. These merchants were willing to accept the risks of
goods perishing or being stolen on long journeys in return for the promise of
larger profits if their journey was successful. In other words, they were risk-
takers.
In order to decide on the viability of such a venture, a merchant needed to
(1) compare the purchase price with the price prevailing at the point of sale,
(2) calculate the additional costs of organising the transportation and
protection of the goods between their purchase point and their point of sale,
and (3) estimate the risk of losing some or all of his cargo on route. His
simple rule-of-thumb for the viability of a safe journey could have been that:
(3.3)
where PI = the unit purchase price in city 1, P2 = the unit sale price in city 2,
t\2 = the unit costs of transporting and protecting the goods between city 1
and city 2, and r\2 = the unit profit accruing from the sale. For any journey
deemed reasonably safe, the merchant's problem thus reduces to that of
estimating the profitability, r\2, of trading between candidate pairs of cities.
From the viewpoint of today's information-intensive world, this may
look like a relatively simple task (although, because of segmented markets
and the complicated dynamics of price movements, the pairwise comparison
of prices is no easy task even nowadays). But when the Medieval merchants
faced it almost one thousand years ago, networks were unreliable and
communication was no easy matter. At least three decision problems posed
great difficulties:
1. pairwise price comparisons required specialised knowledge, which could

only be gained by visiting (or at least monitoring) the places involved on
a regular basis;
2. risk assessment could only be accomplished with the benefit of
experience accumulated over a history of journeys;
3. estimates of the costs of transportation could only be approximate if more
than one good was to be transported together.
The knowledge and experience needed to solve the first two decision
problems could only be generated by a series of trading journeys. In other
words, the entrepreneurial skills of a successful travelling merchant could
only be acquired through learning-by-circulating. The idea of circulating
over longer distances clearly appealed to the Italians, especially to the
Venetians. Given the modest size of local demand, long-distance trading was
seen as a natural way to combine trading for profit with the Venetians'
acknowledged skills as seafaring navigators. The incentive to exploit their
maritime prowess was stronger than elsewhere, because farming was not a
realistic alternative.
The early fish-eating, marsh-dwelling Veneti gathered and processed salt,
then sought other markets where it might be sold for profit. One such market
was found in Constantinople. As the final decade of the tenth century began,
Venice prospered under the strong hand of the statesman, warrior and
diplomatic genius, Pietro Orseolo II. In less than a year, he negotiated tariffs
with Basil II in Constantinople that were more favourable than any that
Venice had previously enjoyed (Norwich 1977). By effectively reducing the
total transaction costs (tn) for Venetian goods sold in Constantinople, the
Doge established a major competitive advantage for Venetians trading in the
Greek city.
Imagine the thoughts of an enterprising Venetian merchant who knew
that salt was highly regarded by both Muslim and Christian alike. He also
knew that Constantinople supplied many luxury goods such as silk cloth,
gold and silver plate, carved ivory, jewellery, and semi-precious stones. It
also produced more pedestrian commodities such as linen, cotton cloth, and
armaments. Sicilian grain could be bought in Palermo and sold for profit in
many other ports. This merchant had the following bright idea: (1) to buy
salt in Venice for sale in Palermo and Constantinople, (2) to buy a few
luxury goods in Constantinople for sale in Venice or Palermo, and (3) to buy
grain in Palermo for sale in Venice. Thus the more enterprising Venetian
merchants added to salt exports the business of general trade (Jacobs 1969).
In addition to the three decision problems mentioned earlier, this merchant
also needed to charter a suitable vessel, chart the most judicious route and
decide on the quantities to be carried (see Fig. 3.8).
The Venetian merchant's willingness to search for new markets for some
luxury goods bought in Constantinople was not just an act of great courage.
It also revealed a natural talent for sniffing out entrepreneurial profit. An
unfamiliar product is valued by purchasers much as gifts of nature or
pictures by old masters (Schumpeter 1934). Often its price may be
determined without regard to the actual cost. Despite the innumerable
difficulties of such trading ventures, clearly the rewards could be exceptional
(although unpredictable).
66 Batten D.F.
()
J
Salt @ P1
t1.2
~ ~:()
Silk@ P2
t2.3
Grain @ P3
Figure 3.8 The network economy of Venice, Palermo and Constantinople

Granting such merchants the foresight to charter a suitable vessel, chart a

reasonable route and estimate all transaction costs was a generous but
plausible assumption. But would such a merchant have been sufficiently
knowledgeable to have made pairwise price comparisons and then decided
upon the amounts to be carried in such a way that the venture's profitability
could be guaranteed (if not maximised)? Even with the benefit of price
comparisons recorded during earlier voyages, his complete logistical
problem is far from simple because of the risks associated with each journey.
Because each merchant's beliefs were personal and shaped mostly by his
own experiences, it is also unclear how the trading system would have
behaved collectively. There is no certainty that it would have ever converged
to a classical general equilibrium like that discussed in Andersson (1995).
Even if all merchants had expected to profit from every trading transaction,
they could never have achieved the same degree of mercantile proficiency
and profitability. A handful of experienced merchants might have known
that they had chosen optimally, while others would surely have believed that
they had done the same. Most would probably have had no idea whether
their choices had been optimal. The more enterprising a Medieval merchant
was, the more complicated and ill-defined his decision problem would have
been.
3.4.3 Mental Models and Learning-by-Circulating
How might our enterprising Venetian merchant have come to terms with
such a complicated situation? According to modern psychology, human
beings have always resorted to both deductive and inductive reasoning. The
left side of our brain presides over rational, analytical and critical thinking.
But the scope of our deductive powers is limited. The right side is mainly
responsible for pattern recognition, intuition, sensitivity and creative insights
(see e.g. Bower and Hilgard 1981). When situations become too complicated
or ill-defined, human nature dictates that we tum more to our inductive
abilities. We resort to intuition or search for patterns which, once recognised,
can help us to simplify the problem. We use our intuition or recognisable
patterns to build temporary mental models to work with, carry out localised
deductions based on these mental models, and then act on such models.
When he first set sail for Constantinople, our Venetian merchant's profit
expectations may have been a little unrealistic. He may have enjoyed mixed
success on the first few trips owing to the many uncertainties associated with
such a novel and risky undertaking. But once he had collected a reasonable
diary of information about the prices prevailing in all three cities, his ability
to make the necessary pairwise price comparisons would have improved
greatly. More accurate risk assessment would have come with the benefit of
68 Batten D.F.
hindsight and experience accumulated over a series of journeys. No doubt he

would have exhibited adaptive behaviour by modifying his chosen route in
order to minimise the costs and risks of transportation.
Eventually he might have been able to extend his simple rule-of-thumb
hypothesis for the profitability of pairwise trades (rl2) into a crude mental
model of the profitability and risk associated with various combinations of
goods, places and chosen routes. Relying more on intuition (rather than
mathematics), he would have chosen his strategy and then put his plan into
practice.
He would have gained considerable feedback from his own experiences
during each journey, seeing how dangerous his chosen course was, how
judicious his choice of goods for trade were, and how profitable his pricing
strategy. Such feedback would have strengthened or weakened his belief in
his crude mental model of the whole trading venture. For example, he might
have discarded some previous routes if they proved to be too risky or if he
had heard about a superior route from other merchants. He might have
changed his bundle of goods for trade if they failed to achieve his profit
expectations, sometimes even replacing them with new ones as needed.
Wherever he lacked full definition of his problem and could not be sure
about the best way to reach a good decision, he would have been obliged to
'paper over' the gaps in his understanding. In the words of Tom Sargent
(1993), he would have acted as an economic statistician, using and testing
and discarding simple expectational models to fill these gaps. As logic, such
behaviour is not deductive but inductive. Inductive reasoning is from a part
to a whole, from the particular to the general, or from the local to the
universal. Far from being the antithesis of 'reason' or science, it is precisely
the way in which science itself operates and progresses (Arthur 1994).
Tremendous competitive advantages would have accrued to the
merchants who could select the swiftest vessel, chart the safest route and
carry the most profitable combination of goods. Small wonder that the status
of the most successful merchants grew remarkably quickly. Entrepreneurial
merchants became the economic leaders of the Medieval period. They were
upwardly mobile, became aristocrats, and formed powerful patriciates to
govern the great trading cities. Eventually even the higher nobility 'turned
merchant' as the great trading cities prospered to an unprecedented extent.
Learning-by-circulating turned out to be a very rewarding activity for the
enterprising ones. It also contributed to the emergence of a completely new
urban hierarchy in Europe.
3.5 CO-EVOLUTION
3.5.1 Co-Evolutionary Learning Under Uncertainty
Small changes in the behavioural ecology of urban residents can have

profound effects on the urban landscape as a whole. For example, today's
transportation networks provide the urban arena on which an ever-increasing
volume of human interactions accumulates. In response to the growing
incidence of congestion, some of the more innovative commuters search for
novel ways to exploit these networks more efficiently. As their numbers
grow, the resulting temporal and spatial innovations increase the network's
throughput. Such adaptive learning processes are co-evolutionary in
character; the decisions of the individual commuters combine to shape the
collective outcome for the traffic as a whole, which in turn affects the
decisions of each individual commuter. The interested reader is referred to a
separate paper (Batten 1998) for further details of co-evolutionary learning
on congested traffic networks.
There are many urban situations where the collective outcome (of a
group of individuals coming together under certain external constraints) is
difficult to predict in advance. Instead ofthe collective risk manifesting itself
in the form of congestion or loss of time (as on congested roads), it may
involve loss of money, power, respect or urban quality. The unifying aspect
is that many of these situations are complicated or ill-defined.
In each instance, the complete set of hunches or hypotheses entertained
by a group of individuals in the face of such uncertainties forms a kind of
ecology (or knowledge base). A key question of interest is how this ecology
co-evolves over time. Does it ever converge to some standard equilibrium of
beliefs (as we often assume in equilibrium theory) or does it remain open-
ended, perpetually discovering new hunches and hypotheses?
Because the set of potential hypotheses is usually open-ended, this seems
to be a difficult question to answer analytically. One might generate a kind
of 'alphabet soup' of hypotheses and then proceed by computer experiments.
In this way, Brian Arthur showed that a set of hypotheses may self-organise
into an equilibrium pattern which is "almost organic in nature ... something
like a forest whose contours do not change but whose individual trees do"
(Arthur 1994). However, even the contours of the emergent ecology can be
expected to change in many urban situations, because the urban population is
always growing in number and the city's network structure may also change.
This emergent inductive world is rather like an electoral world, where
each voter's beliefs are highly individual, largely subjective and mostly
private. When called upon collectively, however, each individual's
70 Batten D.F.
preferences can only manifest themselves in crudely simple forms (e.g. a

single vote, a single trip or a single location). Moreover, it is a dynamic
world. All those beliefs or hypotheses that residents or economic agents
form are constantly being tested in a world that forms from their and others'
actions and subjective beliefs. This vast collection of beliefs or hypotheses is
incessantly being formulated, acted upon, changed and discarded; all are
interacting and competing and evolving and co-evolving; forming an ocean
of ever-changing, predictive models (Arthur 1994).
Because these beliefs are constantly evolving and co-evolving, there is no
evidence to suggest that urban decision-making behaviour ever settles down
into any stable predictable pattern. Instead, the emerging behavioural
ecology becomes more complex and contains an even richer population of
active hypotheses. Reasoning is both deductive and inductive. This uneven
pattern of co-evolutionary learning marches forever onwards, occasionally
into but mostly out of equilibrium states.
3.5.2 Complexity, Timescales and Stages of Development
The complexity associated with urban evolution arises partly from the
interdependencies which exist between individual learning, collective
behaviour and the design of physical systems to facilitate behaviour. It is
also related to the stage (or epoch) of development which a city or system of
cities has reached. For example, the innovative behaviour of circulating
merchants was a very positive developmental factor for the system of
Medieval cities. It promoted product and knowledge exchanges, urban
specialisation and productivity improvements in many places at a time when
Europe was ripe for such catalytic changes. Although some negative
externalities were doubtless generated, their cumulative impact would not
have been excessive. On balance, Europe was ready to enter a new epoch of
communication, growth and prosperity.
The innovative behaviour of drivers in today's cities may be seem as an
opportunistic response to a system which has attained quite a different stage
of development. Jamming transitions are symptomatic of a heavily
congested urban system where negative externalities have reached
unacceptable levels. Those urban arteries which have served cities so well
over so many centuries are finally showing signs of distress. Under these
conditions, driver behaviour is often motivated more by desperation than
inspiration! Because levels of pollution generated are at their highest under
stop-start conditions, there is a need for some major qualitative
improvements to the system. On balance, such conditions correspond to a
mature system of cities nearing the end of a prosperous epoch of
development. Their future is uncertain.
In the latter case, each explorer's decision is flow-inducing; it helps to

alleviate the congested state arising from collective human decision-making
under uncertainty. It is motivated, however, by the more frequent experience
of a flow-inhibiting state. Although the natural conclusion might be that
links should be added to the network, there is clearly a limit to this kind of
innovation. When networks are large and sophisticated, it is often difficult to
predict the impact of additional new links. Patterns of collective behaviour
are very sensitive to the connectivity of a network, so much so that counter-
intuitive responses may occur over time.
Because individual beliefs, collective behaviour and physical systems all
evolve and co-evolve on different timescales, a proper understanding of
evolutionary complexity requires us to recognise the emergent behaviour
which may arise unexpectedly from within. Certain forms of emergent
behaviour arise principally because of the different speeds of these co-
evolutionary processes (Batten and Johansson 1985). Some of this behaviour
will be beneficial for a city, some of it may be detrimental. Few would argue
that the rise of a European merchant class was detrimental to urban Europe.
It engendered increasing prosperity and the emergence of an innovative
network of great trading cities. On the other hand, few would disagree with
the view that congested networks are an undesirable outcome of
contemporary urban evolution and traffic management.
Many attempts have been made to formulate models of urban evolution.
Some of these models are formally stated in terms of a relatively slow-
moving set of parameters and a relatively fast-moving set of state variables.
For modelling purposes, the relatively static parts of an urban system may be
regarded as parameters and the relatively dynamic parts may be treated as
state variables. An interesting attempt to broaden the Pirenne-Mees
hypothesis along these lines was made by Andersson (1985). He asserted
that a series of revolutionary changes to the world economy (from 1000 AD
until 2000 AD) were triggered by slow but steady changes to its logistical
networks 1 Using the same third-order differential equation system each
time, he portrayed the qualitative development of interregional economic
relations throughout the world in terms of four logistical revolutions. The
relatively slow-moving variable was network infrastructure capacity and the
relatively fast-moving state variable was production capacity.
The basic problem with dynamic models of urban evolution is that they
are obliged to be partial. A complete metropolis possesses such an
exceedingly complex system of interdependent parts that its evolutionary
potential becomes difficult to discern. Because of this uncertainty, we know
now that the likelihood of reaching a stable and unique equilibrium state is
Logistical networks are those systems in space which are used for the movement of
commodities, information, money and people.
72 Batten D.F.
relatively low. In the face of multiple candidate states, the question of how a
particular state mayor may not be reached is of particular importance. This
says nothing about how systems of cities might evolve interdependently.
3.6 CONCLUDING REMARKS
A city's development certainly shares some features with the evolution of

the human brain. The brain's neurochemistry is astonishingly complex, its
circuitry being more incredible than any machine devised by humans. It
develops from a small centre, grows slowly and changes discontinuously.
Innovative replacement can only be selective since the system must continue
to function during any renovation. The functions of many old parts are too
vital for them to be replaced altogether. They possess vintage properties, in
the sense that some bits are older and less reliable than others. Nevertheless
these parts struggle on, mostly out-of-date and sometimes even counter-
productive, an inevitable consequence of evolution.
In this chapter, I have hinted at the idea that cities and human brains may
have a great deal more in common. Both are living organisms whose patterns
of development display the traits of complex dynamic systems. Perhaps the
most important feature they have in common is that cities and brains are co-
evolving complex systems. Cities co-evolve with other cities, with other
living organisms (including brains) and with an ever-changing environment.
This kind of co-evolution can produce surprising macroscopic outcomes,
such as the rank-size distribution of cities discussed in this paper. Such
collective outcomes are sometimes referred to as emergent behaviour.
Although we can never predict the exact branchings of cities (or brains),
we can hope to uncover laws that govern their general behaviour. We must
seek explanation rather than prediction. To further our understanding of
urban change, we must strive to understand various processes of self-
organisation and their emergent properties. This chapter has focused on such
processes because they offer the urban analyst a powerful new lens for the
recognition of certain comparative aspects of urban development.
In urban decision situations which are complicated or ill-defined,
individuals are forced to resort to inductive rather than deductive reasoning.
Orderly deduction works well when an inhabitant feels confident about the
collective performance of (components of) a city. His confidence remains
high because his expectations are more-or-Iess satisfied. If all urban
inhabitants felt similarly satisfied, an equilibrium state may even prevail. But
as soon as any decision situation becomes complicated or ill-defined, the
confidence of some inhabitants is easily undermined. Each individual is
forced to draw on his experience of past situations, to resort to half-held
hopes, or even to draw on analogies. When the information at hand is

inadequate, he must 'paper over' any gaps in his understanding as best he
can. He employs inductive reasoning and hypothesises about the global on
the basis of the local. Learning and adaptation are central to this problem.
The uneven nature of co-evolutionary learning suggests that the
collective preferences of urban decision makers rarely coincide, least of all
converge to a rational expectations equilibrium. Spontaneous emergence of
inductive explorers from a seemingly homogeneous population of deductive
'sheep' is a nonlinear perturbation which tends to be repeated over and over
again. The incentive for repetition is strong. In the true Schumpeterian spirit,
each time a new group of explorers emerges and evolves, their innovation
enhances the evolutionary potential of the city.
As a small step towards a behavioural formalism for the study of urban
co-evolution, we must look more closely at the subjective expectations and
multiple hypotheses held by the different individuals and cultures involved.
This is a rich and complex world, in which co-evolutionary learning is
incessant and sometimes surprising. Beliefs can be mutually reinforcing or
mutually competing. Like vintages of technology and products in the
marketplace, beliefs are invented, establish a small niche, grow in
importance, begin to dominate, mature, fall back, and finally decay. In so
much as they resemble an ocean of interacting, competing, arising and
decaying entities, occasionally they may simplify into a simple,
homogeneous equilibrium pattern. But more often than not they produce
complex, ever-changing patterns in which non-equilibrium beliefs are
unavoidable. Because they form the DNA of our cities, such beliefs should
command our closest attention.
Acknowledgments
An unfinished draft version of this chapter was prepared while the author was
visiting the Politecnico di Torino. For providing generous support to facilitate these
visits, sincere thanks are. due to Professor Cristoforo S. Bertuglia and the National
Research Council (CNR) of Italy. The author is also grateful for the comments
provided by the audiences attending his lectures at the Politecnico di Torino, the
University of Bologna and the University of Venice. Specific comments from
Dimitrios Dendrinos, Britton Harris, Dino Martellato, Sylvie Occelli, Aura Reggiani
and Angela Spence are also acknowledged.
REFERENCES
Andersson, A.E. (1985) Presidential Address: The Four Logistical Revolutions. Papers of the
Regional Science Association, 59: 1-12
74 Batten D.F.
Andersson, A.E. (1995) "Economic Network Synergetics." In Networks in Action. Batten,

D.F., Casti, J.L., R. Thord, eds, Springer Verlag, Berlin, 309-18
Arthur, B. (1994) Inductive Behaviour and Bounded Rationality. American Economic Review,
84: 406-11
Bak, P. (1996) How Nature Works: The Science of Self-Organized Criticality. Springer-
Verlag, New York
Batten, D.F. (1995) Network Cities: Creative Urban Agglomerations for the 21 st Century.
Urban Studies, 32: 313-27
Batten, D.F. (1998) "Coevolutionary Learning on Networks." In Knowledge and Networks in
a Dynamic Economy, Beckmann, M.l, Johansson, B., Snickars, F., Thord, R., eds,
Springer Verlag, Berlin, 311-32
Batten, D.F., Johansson, B. (1985) The Dynamics of Metropolitan Change. Geographical
Analysis 19: 189-99
Berry, BJ.L. (1961) City Size Distributions and Economic Development. Economic
Development and Cultural Change 9: 573-88
Bower, G.H., Hilgard, E.R. (1981) Theories ofLearning, Prentice Hall, Englewood Cliffs
Braudel, F. (1979) The Wheels of Commerce, William Collins and Sons, London
Cohen, J., Stewart, I. (1994) The Col/apse of Chaos: Discovering Simplicity in a Complex
World. Penguin Books, New York
Darley, V. (1995) Emergent Phenomena and Complexity. Artificial Lifo IV Proceedings, 411-
16
Fujita, M. (1996) On the Self-Organization and Evolution of Economic Geography, The
Japanese Economic Review, 47: 34-61
Ginsburg, N. (1988) "Reflections on Primacy: Cases from Asia." In Asian Urbanization:
Problems and Processes, Tietze, W., ed, Gebrueder Bomtraeger, Berlin
Jacobs, 1. (1969) The Economy of Cities. Random House, New York
Kauffman, S. (1995) At Home in the Universe: The Search for Laws of Complexity. Penguin,
London
Madden, C.H. (1956) Some Indicators of Stability in the Growth of Cities in the United
States. Economic Development and Cultural Change 4: 236-52
Mees, A. (1975) The Revival of Cities in Medieval Europe. Regional Science and Urban
Economics 5: 403-25
Norwich, lJ. (1977) A History of Venice. Penguin, London
Pirenne, H. (1925) Medieval Cities: their Origins and the Revival of Trade (English
translation by Halsey, F.D., 1952, Princeton University Press, Princeton)
Sagan, C. (1980) Cosmos. Ballantine Books, New York
Sargent, TJ. (1993) Bounded Rationality in Macroeconomics. Oxford University Press, New
York
Schumpeter, 1. (1934) The Theory of Economic Development. Harvard University Press,
Cambridge Mass
Schumpeter, 1. (1942) Capitalism, Socialism and Democracy. Harvard University Press,
Cambridge Mass
Sendut, H. (1966) City-Size Distributions of South-East Asia. Asian Studies 4: 165-72
Vining, R. (1955) A Description of Certain Spatial Aspects of an Economic System.
Economic Development and Cultural Change 3: 147-95
Zipf, G.K. (1949) Human Behaviour and the Principle of Least Effort. Addison-Wesley,
Reading Mass
Chapter 4
Nonlinear Dynamics, Innovation and Metropolitan
Development
Dimitrios S. Dendrinos.
Urban and Transportation Dynamics Laboratory, University of Kansas, USA
4.1 INTRODUCTION
This chapter describes the effects of innovation on metropolitan

development in terms of the mathematics of nonlinear dynamics. Its main
methodological message is that the analysis of these effects can be
conducted productively as an unexpected, exogenously induced, perturbation
of paths. Such perturbation could lead either to phase transitions and a
change of state for the metropolitan area concerned, or to divergencies in
dynamic trajectories. Path disturbance, imposed either as an outside shock or
as an endogenous event in metropolitan development, could be depicted
either by changes in model parameters or the altering of initial conditions.
Whereas parameter change may be attributed to basic innovation (which is
always exogenous), disturbances of initial conditions can be due at times to
endogenous marginal innovation. Both can bring about phase transitions,
whereas marginal innovations may be responsible for divergencies. The
focus of this chapter is on the marginal type, as they seem to be much more
likely to occur than the basic ones, although they have rather confined
impacts.
A substantive message of the chapter is that innovation could lead to
further spatial or sectoral dualism, i.e. increased spatial or sectoral disparities
within or among metropolitan settings. Dualism is due to an inherent
dynamic instability found in highly interactive complex systems. To the
extent that the perturbation of paths could induce instability, it is argued that
innovation might be the force behind this shock. Furthermore, it should be
remembered that relative nonlinear dynamics allow automatically for the
76 Dendrinos D.S.
consideration of an environment, i.e. the context within which particular

metropolitan (spatial) dynamics unfold. Consequently, they inherently link
local dynamics to global behaviour. Globalisation, to the extent that it
increases interdependencies (and thus interaction) among localities, is
viewed in abstract within a dynamic framework as an inevitably
destabilising force. All these hypotheses have already been tentatively
explored by the author (Dendrinos 1992), but here certain themes are further
elaborated. It should be noted that a number of contributions have been made
over the past decade to the spatio-temporal social science literature on the
nonlinearity involved in the spatio-temporal dispersion of innovation, see for
instance Bertuglia et al (1995). The difference between the contents of the
present work and studies associated with the geography and economics of
innovation diffusion is stressed.
The chapter is organised into two main parts. In the first, Section 4.2, the
methodology of nonlinear dynamics is very briefly reviewed and references
are made to how it can account for innovation in the form of outside
disturbances of dynamic paths close to borders separating different domains
of attraction in phase portraits depicting abstract metropolitan dynamics. The
second part, Section 4.3, addresses issues of speculative behaviour by
individual agents within a metropolitan context, stating that when some of
these individual actions might trigger disturbances in initial states for such
dynamic paths. Certain speculative instruments for action, specifically the
derivative instruments of options and futures, are discussed. We also briefly
address the question of how nonlinear dynamics may be used to describe the
temporal behaviour of these instruments as causes for perturbing
metropolitan developmental trajectories. An innovative element is the
combination of the two themes.
4.2 NONLINEAR DYNAMICS AND INSIGHTS INTO

METROPOLITAN DEVELOPMENT
4.2.1 A Very Brief Reference to Nonlinear Dynamics
An abbreviated reference to the mathematics of nonlinear dynamics is

presented below. It is confined to some central ingredients of nonlinear
dynamics as they relate to the subject, i.e. to an understanding of how
innovation affects metropolitan development, and is not claimed to be self-
contained, indeed it is far from it. This introduction neither covers the full
menu of tools available to date in the field of nonlinear dynamics (an
impossible task to undertake here), nor the full gamut of dynamic behaviours
Nonlinear Dynamics, Innovation and Metropolitan Development 77
which could be encountered in nonlinear dynamics (unlikely ever to be

accomplished). Consequently, it is not directed to the mathematician or the
social scientist with an already active and advanced involvement in
nonlinear dynamics, but to the uninitiated social scientist and the practitioner
who wish to acquire a rudimentary introduction into these rather advanced
mathematical notions. It does however go beyond the material encountered
in the popular press or even the book by Gleick (1987). In spite of its
simplicity, the reader of this section must nonetheless have some knowledge
of differential calculus and elements of topology.
Much more can be obtained by way of introduction in Devaney (1989)
and Thompson and Stewart (1986), where fundamental and additional
material can be found. An experimental type of approach to the subject is
given in Tufillaro et al (1992).
a) Linear vs. nonlinear dynamics
A non-randomly behaving dynamic system (more precisely, any fictional

model of a real system as drawn by an observer of it) contains two central
features: first, the underlying rule (law, behaviour, statement or code) which
depicts (describes) the continuous or discrete development or evolution I of
the system in time and/or space in a deterministic or stochastic
(probabilistic) mode; and second, the initial condition (starting state or initial
perturbation) of the system.
A model's rule is formally (mathematically) stated in terms of one or
more relatively slow moving parameter(s) and one or more relatively fast
moving, endogenously changing state variable(s), within a set of
simultaneous equations. One already detects that any outside (exogenous)
change in the dynamic system involves either the model's rule, or one (or
more) of its component parameters, or its initial state. Changes in parameters
and initial conditions will preoccupy our attention here for two reasons.
Firstly, because something systematic can be said about them, whereas
changes in the form of the difference or differential equations of a model
(i.e. its rule) is largely outside the realm of systematic analysis. Secondly,
because effects of innovation can be well accommodated within these less
severe interventions without having to resort to exogenous changes in a
model's rule, which is an efficiency principle in modelling.
The notions of 'development' and 'evolution' are not equivalent, although they are used
interchangeably in this chapter. In Dendrinos D.S. and Mullally H. (1985), the definition
of 'development' is given as: a dynamic path of a dynamic system given an initial
condition and a fixed set of parameter values in phase space, whereas 'evolution' is
defined as a phase transition in phase space.
78 Dendrinos D.S.
Continuous or discrete dynamics are stated depending on whether time,

space (or both) are thought of as continuous or discrete in reference to the
specific observed system. Consequently, either differential (in case of
continuous time) or difference (in case of discrete time) equations are
employed as the system's model. A basic question one can ask under these
conditions is whether the dynamic system's model in fact has a solution. A
solution of a dynamic system is the condition which allows one to come up
with predictions regarding future states of the system. An externality of such
conditions is that the solution affords the observer complementary
understanding of the system's underlying rule(s) and the trajectory (orbit) of
the system in time (space, or both) on the basis on which predictions
(forecasts or projections) can be based.
In certain systems studied by natural scientists (like, for example, the
orbits of celestial objects like planets, comets, etc.) the statement of a
system's rule, its initial conditions and solution can be derived and recorded
with satisfactory accuracy. Consequently, these systems' behaviour is well
understood and very accurate predictions of the future states of the systems
can be obtained, even for relatively long time periods into the future.
However, this is still not the case for a large class of natural science systems
(for example; the classical three-body problem, hydrodynamic systems or the
atmosphere), and it is certainly not the case for the vast majority of social
science systems.
Usually, the dynamic systems which have been studied extensively with
solutions which are well known to natural scientists are linear, i.e., they
contain simple (first or second order) uncoupled (isolated) linear differential
equations, like for example the following isolated second order linear
differential equation:
(4.1)
given in Tufillaro et al (1992, pJ). The above differential equation, one of

the most widely studied differential equations in physics (depicting the
continuous motion of a frictionless pendulum) is referred to as 'linear'
because the second derivative of the state variable x is a linear function of x,
i.e. the right hand side in equation (4.1) is an equation of first degree in x.
Two of the solutions to this simple differential equation are: Xl(t) = sin(t),
and X2(t) = cos(t). But these two solutions are by no means the only solutions
of this differential equation, as their sum x(1} = Xl (t) + xdl} is also a solution
to the simple linear differential equation as is any linear combination of the
two solutions Xl (I) and xdl}
This is the principle of linear superposition, a principle lying at the heart
of the theory of linear dynamics. Such a principle, however, which allows
for a full understanding of these linear dynamics, their rules and their
solutions, is absent from nonlinear dynamics.
Consider for example the case of the differential equation, also given in
Tufillaro et al:
(4.2)
which is a nonlinear equation, as the right hand side of (2) is of second

degree in the state variable x. Very simple calculations can show that the
sum of any two solutions Xt(t), X2(t) is not a solution to (4.2), and that neither
is a linear combination of them.
Efforts towards obtaining either the underlying rules or solutions to
complicated (i.e., nonlinear) dynamic equations have so far been
unsuccessful. Due to this failure, reliable predictions regarding the futures of
such systems have eluded even the most ingenious of scientific efforts over
the past few centuries. Basically, these two shortcomings have been
attributed either to the fact that the solutions to these, so far unbreakable,
codes correspond to irregular (random) instead of regular (or even periodic)
motions, or to the realisation that some of these systems are very
'complicated' or complex, i.e. they consist of an extremely large number of
interactive variables, and would need to be modelled by an involved system
of dynamic equations, far beyond our (or any computer's) capacity to derive
analytically (i.e. explicitly) derive, let alone solve. Such complex systems
and their dynamic equations have been thought to be beyond our knowledge
or even understanding. Indeed it may be true that such complex dynamic
systems will always be beyond our reach, but chaotic dynamics renders this
aspect of dynamics irrelevant!
Faced with such complications, scientists have developed strategies for
dealing with them. One has been to construct caricatures (i.e. abstract
models) and linear approximations to such nonlinear multi-dimensional
dynamic systems, making it possible to treat these (often overly and at times
erroneously) simplified systems analytically. Another has been to seek the
assistance of (mostly ad hoc) computer executed algorithms, and
approximations to their true solutions have been derived by way of
numerical simulations, i.e. computationally or experimentally. The belief,
however, has persisted that even a nonlinear dynamic system always has a
solution, if only one were ingenious enough to find it.
This was the commonly accepted scientific wisdom, until Poincare
(1957) pointed out that this belief was wrong. Initial conditions, a
component of differential equations which had been neglected up till then,
playa fundamental role in the derivation of solutions. When slight changes
in the initial state of a dynamic system result in drastically different solutions
80 Dendrinos D.S.
(orbits in time), then the system's solutions are impossible to derive

explicitly, i.e. the dynamic system has no solution, and is in fact chaotic.
Almost one century had to pass before it was realised just how profound
this realisation was for natural scientists. Its importance is still to be
recognised in the social sciences. Since then, this situation has been
christened as extreme 'sensitivity to initial conditions', and it is the
cornerstone of modern nonlinear (specifically, chaotic) dynamics. Sensitivity
to initial conditions is also the central focus of this chapter, which
emphasises the perturbation of metropolitan dynamic orbits (trajectories or
paths) in phase space as the result of marginal innovation, manifested
through the use of certain speculative instruments.
The reason for the difficulty in obtaining solutions to dynamic systems,
and consequently predictions, was thus shifted by Poincare from lack of
insight or cleverness on behalf of the mathematician or the observer, to the
impossibility at times of determining exactly the initial perturbation of a
dynamic system. By extension, since any state of the system can be
construed as an initial state (for any future time period), it follows that this
difficulty might apply in specifying any state, not only those arbitrarily
chosen as an initial state.
In effect, there are systems (such as those which are extremely sensitive
to initial conditions, stock markets for instance) whose state can never be
determined exactly by an observer. This is so because these dynamic systems
are indeed chaotic. This important finding is at the heart of speculative
action. Speculative action basically attempts to predict the behaviour of real
systems, only to find out that prediction is impossible for relatively lengthy
time horizons. Fllrther, the nonlinear dynamic model one uses cannot in any
accurate way replicate the real system it intends to emulate (that is, it cannot
derive the solution to the underlying nonlinear dynamics). By extension,
Poincare alerts us to the fact that this impossibility is not due to a failure in
intelligence, or the incapacity to come up with a smart model and its
solution, but worse. Not only can the range of nonlinear dynamic models
currently used not accurately simulate reality, but in effect none can or ever
will. This subject will be further pursued in Section 4.3.
Approximations to initial conditions can be fatally wrong in the case of a
dynamic system exhibiting extreme sensitivity to original perturbations (i.e.
starting values), as small differences at any point in time can be blown up
considerably even within very short time intervals. Under chaotic conditions,
predictions are virtually impossible, since they are by necessity inaccurate
and thus unreliable for any length of time beyond a very brief one.
This considerable difficulty in arriving at solutions to nonlinear dynamic
systems is an additional problem, as we already have the difficulties
associated with large dimensionality and interactivity among variables
(degree of nonlinearity). Which one is more decisive in impeding our ability

to derive solutions (and thus able to predict) is still an open question,
although largely irrelevant. The benefit of Poincare's finding (and the legacy
chaotic dynamics have bestowed upon us) is that it proved such irregularly
behaving systems cannot have solutions, and by extension they cannot afford
predictions.
Further, a discovery by Poincare of extreme importance in relation to
dynamic systems which behave in a chaotic fashion is that the study of
individual trajectories (orbits) of these specific systems is nonsensical. Put
differently, to place any confidence on the behaviour of a particular solution
corresponding to a specific initial state of a specific dynamic system which
exhibits irregular motion, is totally unwarranted. Instead, one ought to
examine the collective statistical properties of many solutions, not only of
this particular system but also of topologically neighbouring (i.e. somewhat
similar) dynamic systems. Only then, might one be able to say something
qualitative about such collections of systems with any confidence or
relevance.
A study of the statistical properties of solutions constitutes an ergodic
theory of dynamic systems, and it is associated with the idea behind the so-
called Poincare maps, both subjects being central pieces of nonlinear
dynamics. The importance of these statements for social systems cannot be
understated, and will be further addressed later in Section 4.3.
Another important feature of nonlinear dynamics is the fact that the study
of any such system cannot be referred to or derived from a general theory of
nonlinear dynamics, as is the case with linear dynamics. A general theory of
nonlinear dynamics is simply non existent and perhaps impossible to derive.
In effect, one cannot have a global view of nonlinear dynamics, but only a
local and partial one. Every nonlinear dynamic system is a different animal,
as its nonlinearity is different from another system's.
A nonlinear dynamic system might share a few common and basic
characteristics with some neighbouring systems (as, for example, in the case
of the universal numbers, found by Feigenbaum 1980, which govern the
period-doubling cascades of certain nonlinearities), nonetheless it remains a
'local' event in the universe of nonlinear dynamics, a point to be further
discussed later. Consequently, by studying just some of them, one can
understand only a few elements of specific nonlinear dynamic systems.
Thus, in the field of social and spatial dynamics, this issue is which
nonlinear dynamic system should be used as a basis for an exposition that
will be inherently limited. As numerous such systems have been suggested
in the literature, we shall confine ourselves to one very simple but
pedagogically rich example.
82 Dendrinos D.S.
b) An example: the simple logistic prototype
An influential model in the mathematical literature of nonlinear

dynamics, and also in the socio-spatial literature over the past quarter
century, has been the simple discrete logistic prototype model (quadratic
map) first employed in the field of population dynamics in mathematical
ecology by May (1976).
x(t+l) = ax(t)[l-x(t)] (4.3)
x(O) =x 0< x(t) < 1 O<a<4
This is a discrete time model with relative population stock dynamics and
one dimension, as it contains a single state variable, x. The only parameter
[a] employed by this model, captures all relevant factors from the
environment within which the state variable x has been normalised. As a
resurt, the logistic prototype is not unlike the Malthusian model of
exponential growth, where all complicated environmental factors have been
included (discounted) in the exponential growth rate parameter. In (4.3), a
replication (or reproduction) function is presented, where a future time
period's relative (normalised) stock size is a deterministic function of the
current relative stock size. Time-delay effects can be easily incorporated in
these discrete dynamics, but this is not necessary for our purposes.
Whether this is simply an interesting mathematical tool, or in any way
reproduces in model form some real dynamic event of relative population
growth and decline is a question far deeper than academic speculation can
probe. We shall revisit the issue later, after examining certain key qualitative
ingredients of this simple mathematical formulation. This prototype presents
us with only three distinct qualitative dynamics, well analysed in the
mathematical literature, as one scans the linear parameter space within a
bifurcation diagram:
I. in the area of the parameter space O<a<3 a time independent fixed point
solution x* occurs (with extinction being the attractor outcome in the
range O<a<l; the solution x* = 0 is always a solution to the quadratic
map, albeit unstable beyond a = 1).
11. precisely the moment when the environmental parameter [a] becomes
equal to three, the fixed point is transformed (through a flip bifurcation)
into an unstable equilibrium (repeller) and a stable period-2 limit cycle.
In effect, as the scanning of the linear parameter space continues in the
bifurcation diagram and when a= 3, a cascade of stable period-doubling
cycles sets in at geometrically decreasing intervals given by one of the
universal Feigenbaum numbers. After all even-period cycles have been

exhausted, odd-period cycles commence with decreasing periodicity,
until the last ofthem (i.e., the period-3 cycle) appears.
111. exactly at this point, a slight increase in the parameter produces non-
periodic (in effect chaotic) motion in x, and the simple logistic
prototype has no solution up till the parameter reaches its limit for this
case, a = 4.
The simple logistic prototype offers us all the qualitative types of

dynamic behaviour we know in low dimensional problems, except one:
quasi-periodicity. To produce this type of motion, which involves movement
over invariant tori, one needs at least two dimensions in state variable space
under discrete dynamics. In continuous dynamics, quasi-periodicity and
chaotic motion (which might involve strange attractors and possibly fractal)
in phase space require at least three dimensions to be recorded. This
realisation inexorably links particular dynamics to a mInimUm
dimensionality, a subject of some interest to modellers of socio-spatial
dynamics.
Two additional points can be made. Firstly, multi (i.e. more than three)
dimensional (discrete or continuous) nonlinear dynamics involve qualities in
motion (paths or trajectories in phase space) and areas in phase space with
special separatrices largely unexplored until now. There are numerous types
of chaotic motion, and only a few of them have been fully analysed.
Evidence even seems to suggest that chaotic motion might be a local event in
higher dimensional dynamic systems. In other words, certain variables might
exhibit turbulent motion while others obey calm dynamics, Dendrinos and
Sonis (1990) and Dendrinos (1997).
Secondly, no empirical study in the social sciences has so far
conclusively demonstrated that either quasi-periodic or chaotic motion (of
the types we know from the mathematical models available to date) actually
occur in social systems. But the lack of sufficiently extensive time series on
any socio-economic variable does not seem to be the only problem. More
fundamentally, the real issue seems to be whether model chaos and reality's
chaos can ever be matched. Challenges involving these two problems lie
ahead, and certainly constitute a potentially productive research agenda.
4.2.2 Innovation and Nonlinear Dynamics
We now take a closer look at what can be extracted in general and in

principle from the abstract notions surrounding nonlinear dynamics
presented above with reference to innovation and spatial development. No
particular socio-spatial mathematical model of innovation will be discussed,
84 Dendrinos D.S.
as there has been a multiplicity of such models already proposed in the

socio-spatial literature, each with a different substantive focus. For a
relatively recent exposition to this variety of models the reader is directed,
among other references, to Bertuglia et al (1995), which focuses on the
specific subject of innovation, and also to the Special Issue of the journal
Chaos, Solitons and Fractals on Nonlinear Dynamics in Urban and
Transportation Analysis (Volume 4, Number 4, April 1994), which
investigates the broader subject of urban dynamics.
By definition, the notion of 'innovation' implies something new and,
when first occurring, unexpected. Both concepts of 'novelty' and
'randomness' are involved in a definition of innovation. Mathematical
modelling of human as well as of non human (i.e. broadly biological)
systems has not had an easy time coping with the emergence of new species
(whatever these new species might be). The challenge of deriving speciation
models or models of mutations is one which still has to be met effectively. A
reason for such difficulties is that for innovation to be modelled, one must
account for it beforehand in some fashion in mathematical terms. But by
doing so one strips from it the novelty component, let alone its
unexpectedness (randomness). Further, mathematical modelling is still trying
to solve the problem of deriving truly random numbers, let alone arriving at
a satisfactory definition of what random events really are (Penrose 1989). In
fact, there might well not be a logically consistent way to derive a definition
of randomness, nor may it be possible for novelty.
Compounding these difficulties is the question of origin, i.e. whether
innovation in general, technological innovation in particular or any other
specific innovation, is endogenously derived by a socio-economic system (in
this case by agents within a given metropolitan area) or whether it is
exogenously imposed (or imported, though still humanly induced) by its
relevant environment. The question is a long standing issue in the socio-
spatial sciences and we shall attempt here to add a few items in the debate.
Innovation can be either basic (as for example, the invention of tool-
making, fire, agriculture, the wheel, writing, printing, the steam engine,
electricity, or the microchip), or marginal (a scientific paper on traffic
management, for instance, which, to be published, must conform to some
'innovation' criterion). No distinction is made here between concepts such as
'discovery', 'invention', 'scientific or technological change', the generation
of 'basic knowledge' or 'understanding', the 'initial application' of a new or
old technology, etc. For the purpose of this chapter, these concepts are
considered to be interchangeable and identical to the term 'innovation'. This
is not because they all share an identical definition, or because they imply
the same social process(es), but, as explained in Section 4.2, because they
have the same dynamic effects, i.e. they result in a change in a model's
parameter(s) or its initial state.
We make a distinction between basic and marginal innovation, in order to
be able to say something about them, their possible origins and their likely
effects. Both types of innovation can be assumed to first appear randomly at
a particular point in time and space (a metropolitan area), and then to diffuse.
It should be noted, however, that this is a controversial assumption.
Simultaneous or independent appearance of a basic innovation at different
~ocations is also possible to imagine. In any case, they are both construed as
being predominantly exogenous to such setting's on-going dynamics. Put
differently, they are not the likely outcome of internal in-place metropolitan
dynamics, but rather an initial shock to these dynamics. They generate what
Jacobs (1969) called 'new work' and, in the case of basic innovations, what
Schumpeter (1934) described as 'creative gales of destruction'.
a) Basic innovations
Basic innovation is randomly produced at some point in space, I.e. 10

some metropolitan setting(s), at some time period. Then, gradually or
abruptly, it diffuses in space over time. The diffusion process might occur
quickly or slowly, but its impact lasts over an extended time horizon, having
irreversibly altered almost all the metropolitan area's dynamics. Adoption of
a basic innovation is logistically spread over very extended spatial horizons.
Such geographical processes have been well modelled by nonlinear
dynamics (see, for example, Sonis 1994). The effects of a basic innovation
are thus of a long term and large scale nature, although occasionally the
innovation is only recognised as such post jacto. Obviously, there could be
basic innovations which are more and less fundamental, but both can be well
and effectively accommodated within a nonlinear dynamics framework, as
will be seen below.
Basic innovations, having highly significant effects, tend not to appear
very often. They also occur in a non random sequence, since one innovation
generally cannot have happened unless a previous one has appeared. Human
societies would not be able to cope with frequent and severe shocks. When
the frequency of basic innovations increases, as for example during the
1920s, then severe socio-economic shocks can ensue (such as the Great
Depression and the Second World War, which were accompanied by a
restructuring of the world economies). One should expect basic innovations
to be a few and far between. One might call this the 'requisite comfort'
principle, picturing a shock-averse society which opts for environmental
conditions that have settled into stable enough patterns allowing for it ample
86 Dendrinos D.S.
time to adjust and develop. It is hard to imagine viable and sustainable

human societies under a frequently and dramatically changing environment.
A productive way to incorporate such shocks into an analysis of spatial
settings is to consider that at times the parameters (or slow moving
variables) of dynamic models are altered. Parameter values determine the
model's fast dynamics, i.e. the dynamics of its state variable(s), given any
initial condition. The importance of basic innovation is captured in that it
reshapes phase space. Any initial condition positions and locks the
metropolitan dynamics onto a particular locality in a phase space
corresponding to a particular set of parameter values in the bifurcation
diagram (which plots the phase space against the parameter space).
Reshaping of phase space, i.e. movement in the bifurcation diagram,
requires a considerable shock to the dynamic system. The source of such
shocks can be thought to lie outside the metropolitan area in question, in that
it is exogenous to it. Not being the result of internal metropolitan dynamics,
it must be due to a considerable change originating from the environment
within which the metropolitan setting in question operates.
If it happens that a change in model parameter(s) catapults the dynamics
beyond bifurcation points (lines or solids) in a bifurcation diagram which
separate areas of different fast dynamics, then a basic technological
innovation causes a phase transition through a bifurcation in a metropolitan
area's dynamics. In the example of the quadratic map discussed earlier, if
innovation is picked up by a change in the single environmental parameter
[a] (say from 1.8 to 3.2), then the fast dynamics are transformed from a fixed
point (at 1.8) to a stable period-2 cycle at 3.2. In effect, the innovation
assumed to correspond to this jump in [a] has pushed the dynamics of the
system beyond the critical point where a=3.
Still using the above example, if a change from 1.8 to 3.2 were to occur,
we must assume that there has been a considerable shock, absorbed by the
underlying phase transition. The fixed point x*:
x* = [1-1/a]=0.44,..) when a = 1.8
is quite some way away from the two stable periodic solutions Xl * and X2*
given respectively by:
Xl* = (1+a+A.[a2-2a-3])/2a = 0.79945549 ...

X2* = (1+a-A.[a2-2a-3])/2a = 0.513044509,.. when a = 3.2.
Assume now a shock of lesser magnitude, but one which still results in a
bifurcation. Whereas the qualitative effect would be the same if the fast
dynamics are perturbed by changing the environmental parameter [a] from a
value of, say, 2.99 to 3.01, the quantitative effects are relatively minor in
difference: the fixed point x*= 0.665 ... when a = 2.99 is quite close to both
stable periodic solutions x\*= 0.7005 ... and X2*= 0.6339 ... when a =3.0l.
Thus, proximity to a bifurcation point is substituted by severity of the
exogenous shock in the ensuing phase transition: close proximity implies
less severity as far as the simple logistic prototype is concerned. What this
implies is that a basic innovation associated with a large size jump in a
model's parameter might have the same effect as one with a smaller size
change, depending upon the location of the system's dynamics in the
bifurcation diagram in reference to critical (bifurcation) points.
A large change in parameter value might be due to a fundamental basic
innovation. For example, the advent of electricity might have been more
fundamental than (and must come before) the invention of the microchip,
and thus the former might be depicted by a larger magnitude change in the
environmental parameter than the latter. Under some sequence in parameter
jumps, including timing and magnitude, one might structure the underlying
causes of socio-spatial evolution. The other cause might be attributed to a
series of perturbations in initial conditions to be discussed in the next
subsection. However, both conclusions regarding proximity to a bifurcation
point and impacts of small or large changes in parameter values upon
solutions (end-states) are by no means general conclusions applicable to the
whole gamut of nonlinear dynamics. This is another strIking example of the
locality involved in the analysis of these dynamics.
But beyond differences in end-state, phase transitions due to bifurcations
also imply that the path to the end-states is very different: the type of
oscillatory movement towards the final equilibrium value under fixed point
dynamics is quite different from that involved in a motion converging to
either a stable 2"-period cycle or to quasi-periodic and non periodic
solutions.
Here, a noteworthy technical point should be made. Insofar as the
number of iterations is concerned, another interesting feature emerges from
the above quadratic map related example. Given some starting condition
x(O), the number of iterations (i.e. discrete time periods) necessary to
converge to any arbitrarily chosen neighbourhood of the fixed point
(solution) is relatively small at parameter values further away from
bifurcation points. The number of iterations increases almost exponentially
fast as one draws closer to critical borders in parameter space.
Put differently, the closer one is to a bifurcation point in the quadratic
map, the more difficult it becomes to recognise a solution (for example, to
distinguish fixed point behaviour from a stable two-period motion, a stable
two-period from a stable four-period motion, etc). A period of uncertainty
prevails at the borders of phase transitions as to the final outcome of a basic
88 Dendrinos D.S.
innovation, the more so the closer to the border one is. Clean breaks, i.e.
innovations which involve great changes in parameter values may be more
easily susceptible to detection (since they leave more distinct fingerprints),
than innovations causing small shocks in parameter values in the bifurcation
diagram.
As parameter change implies a considerable shock to metropolitan
dynamics, it must be the result of an expenditure of resources. As different
parameter changes may correspond to different speeds, they will necessitate
different levels of resource expenditure. The two could well be correlated,
and constitute a problem of potential research interest.
b) Marginal innovations
Marginal innovations are far more numerous and of a much higher

frequency than basic ones. But they, too, can be assumed to occur randomly
in space and time. They also diffuse once they appear, although their
diffusion rates might be less dramatic compared to those of basic
innovations. The adopters may be more densely located and closer to the
innovations' origin than for basic innovations. The time which elapses from
the appearance of a marginal innovation to its terminal time horizon (the
lifespan of its effects) is likely to be far shorter than for a basic one. Origins
of marginal innovation are specific: there is much less likelihood of a
marginal innovation appearing simultaneously at very distant points in
space. Also, there does not seem to be a necessary sequence for marginal
innovations.
Innovations are likely to be recognised as marginal as soon as they
appear, as a result of their expected market area and effects, although ex post
facto some of them could be reclassified and elevated to the status of basic
innovations. Those innovations which have had major global effects and
market areas for a limited time period, but have soon afterwards retreated
and been surpassed by subsequent ones, might also be classify as minor.
This phenomenon, whereby an innovation has a brief lifespan with effects
which acquire a major spatial extent for only a short time period, may
constitute something equivalent to the butterfly effect in nonlinear dynamics.
Many examples can be cited, for instance innovations in refrigeration,
telephony and information storage. We shall discuss this further later in the
chapter.
Marginal innovations, with their much higher frequency - in sharp
contrast to basic innovations - are the engine of the everyday life of social
systems. Human societies are structured around these lesser innovations,
which occur in all areas of social activity: in political, social and economic
arenas, changes and novelty are constantly present and usually surrounded
by an element of mild surprise.
Art and architecture, fashion and law are fields where these marginal
innovations have been vividly abundant at all points in space and in all time
periods throughout human history. Almost by necessity, socio-spatial
systems feed from these abundant, albeit minor, innovations. Examples
include the novelty in sculptures and buildings, clothing and legal statutes.
Plans and policies for metropolitan growth are further cases in point. Some
of them may be endogenously produced within a metropolitan setting, but
others are likely to be subject to outside influence. Exogenous creativity is
unquestionable in the case of art and architecture, fashion and design, but it
is doubtful for law and planning, these areas of human activity being very
much the product of endogenous metropolitan dynamics.
Within a nonlinear dynamic context, the most appropriate way of
accounting for these marginal innovations (in either products or processes) is
by exercising the other freedom one obtains through the use of dynamic
modelling: i.e., by altering initial conditions in a setting's dynamics. The
source of changes in the initial perturbation could be attributed to either
actions by agents within the metropolitan setting in question, i.e. these
changes are endogenous to a particular metropolitan area as they originate
within a spatial setting, or they could be exogenously imposed.
Actions by individual agents almost always contain an element of
innovation. As a consequence of the individual actions constantly affecting a
metropolitan area, many of the plethora of marginal innovations will cancel
each other out, but a few will accentuate the perturbation. Obviously, the
vast majority of individual human activities are so minor as to be
inconsequential, not even disturbing the initial conditions. But some will
transcend the confines of routine activity. Since individual or collective
agents' action is what fuels these lesser innovations, the motives behind
them can be found in speculative agent behaviour, a subject to be further
explored in Section 4.3. .
Dynamic analysis generally considers initial conditions to be of
paramount importance when in the phase space there are separatrices
delineating areas of different dynamic behaviour (possibly involving areas of
distinct chaotic attractors and fractals). Nonlinear dynamics clearly indicate
that sensitivity to initial conditions matters particularly when a system is in
either quasi-periodic or non periodic (chaotic) motion, and vice versa. This is
the reason why, as already stated, chaotic regimes in nonlinear dynamics are
characterised, among other requirements, by extreme sensitivity to the initial
perturbation.
Phase transitions follow parameter changes, i.e. basic innovations occur
when bifurcation points are present. But phase transitions are also possible
90 Dendrinos D.S.
when starting conditions cross over the separatrices in phase space due to the
push of marginal innovations push. In addition, alteration of initial
conditions (i.e. marginal innovation) could have significant effects when the
dynamics are quasi-periodic or chaotic. As a result, within this framework
exists the potential for two types of marginal innovations: those more
important ones, which guide dynamics in crossing over separatrices, and
those less important ones which, if the dynamic system finds itself in nOn
periodic regimes, might cause it to divert to different end-states.
Effects due to diversion can be shown quite easily. Continuing with the
example from the quadratic map, using a pocket calculator with only 11
significant digits approximation and assuming that a = 3.9 when x(O) = 0.1,
the 5-time period sequence for the state variable is: 0.351, 0.8884161,
0.38661843972, 0.924864025, 0.27101318509. By slightly altering the
initial state to x(O) = 0.1001, the same time horizon sequence is:
0.351311961, 0.8887782815, 0.38552064646, 0.92388846269,
0.2742424277, and at only the ninth iteration have the corresponding values
become 0.5379486459 and 0.4911126596.
Such divergency effects, due to extreme sensitivity to starting values,
brings about the problems of truncation, approximation and estimation
(calibration) of either initial conditions or parameter values in view of an
error accumulation. Is it possible to record in a reliable manner dynamic
paths at any level of approximation and for a given time horizon and
accurately reproduce them? Is it possible to statistically estimate parameters
when confronted with extreme sensitivity conditions (i.e. in regimes of quasi
or non periodicity), independently of how extensive the time series are?
One could speculate that social systems might in effect have built-in
processes to correct errors due to approximations in dynamic series of social
behaviour! A high frequency in political processes (frequent elections) might
be viewed in such a light. Beyond this, we do not intend to add further here
to these potentially important questions, except to point out that innovation
might be feeding from the inability to record, estimate and reproduce such
processes accurately (if at all).
Besides differences in end-state configurations, phase transitions imply
that the way towards an end-state is significantly altered. Otherwise, changes
in initial conditions become irrelevant as far as the end-state is concerned. It
may, however, have a mild effect upon the number of iterations and hence
the time which elapses from the initial perturbation to the terminal time
period. This has been referred to as the' butterfly effect', and was first stated
by Lorenz (1963), who cited the case where the flapping of the wings of a
butterfly in Brazil causes the weather to change a week later in New York
City.
Butterfly effects should in reality be very infrequent, unless one seriously

considers the future of New York City to be left hanging on the wings of the
butterflies of the world! Only under the most opportune of circumstances can
one expect a butterfly effect to occur, namely when one is in a domain of
quasi or non periodicity, which is very rare in the parameter space. As the
simple logistic prototype indicates, only in the very narrow range of the
parameter 3.83 ... <[a]<4.00 is a change (even a minute one) in an initial
condition x(O) of any consequence. Changes in the initial state in any other
range of [a] between 0 and 4 has only the effect of delaying or prolonging
the final solution. That is, a marginal innovation will impact the temporal
trajectory followed by the dynamics of the setting, but the equilibrium (fixed
point or periodic) to which the nonlinear dynamics will eventually lead the
metropolitan area in question, will be unaffected by the marginal innovation.
One would expect, according to the above analysis, that only an action by
very few lucky individuals will gain the distinction of having had a 'butterfly
effect'. Many more can be granted the distinction of simply having
accelerated or decelerated the development of a metropolitan setting. The
vast majority of individual actions are cancelled out in the soup of collective
action, as they have neither a path nor a time-independent final state effect.
And an extremely small number of all those people who have ever lived can
be credited with the acknowledgement that their actions have had a basic
effect on human history.
Much has been said over the past decade, in qualitative terms, about 'path
dependence' and evolution, and whether history matters along evolutionary
paths. Some of these issues have been popularised by discussions on the
subject of complexity held at the Santa Fe Institute. Along these lines, a
rigorous attachment to nonlinear dynamics seems to imply that:
I. if development occurs along consistent deterministic dynamic paths in

phase space (orbits, trajectories), then in order to replicate these, the only
information necessary is the initial conditions and location within phase
space. History therefore does not matter or, put differently, it can be
deterministically derived and known;
11. if, on the other hand, the evolution involves shifts in paths due to changes
in parameters and/or initial positions, then history does matter and cannot
be deterministically derived. To be able replicate the dynamic
trajectories, in addition to the initial conditions and position in phase
space, one would need to know the exact size and direction of changes in
parameter(s) and perturbations.
Here, I should like to make a final point about innovations and nonlinear
dynamics. Innovations, whether basic or marginal, are in effect both
92 Dendrinos D.S.
hierarchical and cumulative. Some are of a 'higher level' (the fundamental

basic innovations), others are of lesser importance (the small and relatively
inconsequential marginal ones). However, some principle of determinism or
'slaving' must be at work, manifested in the various time constants
associated with the different parameter changes. While some innovations
emerge and are adopted, others fail to take off, but nevertheless a few
marginal innovations will occasionally result in a basic innovation. This
latter is a point made, among others, by Cimoli and Oosi (1995).
4.2.3 Some Implications for Metropolitan Dynamics
Relative dynamics, along the lines of the quadratic map, express the size
of the state variables in normalised terms and with reference to a broader
environment. Normalisation thus automatically ties up the local
(metropolitan level) dynamics with the global (environmental) ones by
expressly modelling the local level's share of the total (be they shares of
population, income, wealth, etc). The broader the environment over which a
locality's variables have been normalised, the more global is the source of
impacts affecting the locality.
A number of principles seem to underlie relative dynamics, particularly
in relation to inter and intra-metropolitan dynamics. A few of these
principles are pertinent here. Smaller shares identify the larger environment
within which a locality's state variables are normalised. An environment
which is more extended in space in turn implies a locality more open to
outside influences. The smaller a locality with regards to its effective
environment, the more likely it is that the locality's dynamics may be
turbulent. Larger settings are less likely to be subject to chaotic dynamics
since the more open a setting is to outside influences, i.e. the more
interactions there are with its environment or the larger its environment, the
more likely it is that the setting's aggregate as well as disaggregate
metropolitan dynamics are unstable.
Dynamic instability implies that a setting (as a whole or in parts) is
driven either to extinction or to a far larger size. Instability is manifested,
among other things, in the rank size distribution of settings, where a very
few very large (and wealthy) ones coexist with a very large number of very
small (and poorer) settings. It is also found in the incessant movement of
metropolitan settings up and down the urban hierarchies, as well as in the
continuous change at the top of urban hierarchies.
These principles, together with a few others connected with the
magnitude of interaction, are tied to some basic findings associated with
nonlinear dynamic systems, Oendrinos (1992). A number of these are based
on conclusions reported initially by May (1974).
In view of these principles, one might expect the smallest and the largest
metropolitan areas to be far more vulnerable to exogenously imposed basic
innovations as they are exposed to more severe impacts than are the settings
occupying the middle ranges of urban hierarchies. Although larger settings
(due to scale effects) are more likely to develop buffers to soften the impact
of exogenous shocks, giant metropolitan agglomerations may be simply too
open to global forces for their size. Medium size settings tend to be more
diversified in their employment base relative to their size than either the
smaller ones or the giant ones, although there may be always exceptions.
Medium size metropolitan areas may principally consist of those moving up
or down the hierarchy. According to these hypotheses, globalisation (i.e.
world-wide interactions) must clearly be viewed as a destabilising factor in
metropolitan dynamics, whether of an inter or intra-metropolitan character.
Given this likely reaction to an outside environment by both small and
large metropolitan settings, the question is: does environmentally instigated
innovation increase the chances for instability? If the answer is yes, then one
should also expect that innovation (basic or marginal) fuels further increases
in spatial (and possibly sectoral) disparities, i.e. dualism, since dynamic
instability in metropolitan development leads to disparities.
To address such a question, it is necessary to have a broader dynamic
framework is than that given by the quadratic map or the mathematical
ecology of May. Dendrinos and Sonis (1990) have formulated the dynamics
of a 'universal map', where the term universal is used to point out that its
dynamics could potentially be inclusive of a vast variety of qualitative
dynamic effects, far richer than the ones discussed earlier and found in the
simple logistic prototype (which emerges as a special case of these universal
dynamics).
Analytical as well as numerical experimentation with this universal map
through computer simulation seems to indicate that dynamic stability is the
overwhelming outcome in the universe of parameter space in low
dimensional cases (where the number of state variables is no greater than
four). But small islands of instability do exist in that sea of calm dynamics,
and there are also islets of turbulent motion. The greater the system's
dimension (and the greater the number of parameters), the more likely it is to
randomly experience stable dynamics. In fact, turbulent dynamics are
eccentric events.
These findings stand in sharp contrast to those reported by May, and
demonstrate the fact that, because no general theory of nonlinear dynamics
exists, the analytical framework one decides to adopt more or less dictates
the qualitative findings one obtains - and maybe vice versa.
One might conclude that this is a cynical view of science, particularly
social science. However, a more careful look at the implications of this
94 Dendrinos D.S.
dictum leads to some important and deep conclusions regarding social reality
and social action. Social reality is perceived by social actors who act on the
basis of these (possibly coexisting and multiple) perceptions, thus affecting
it. Although some of these perceptions may have been way off the mark, this
is irrelevant. What matters is that actors act on such perceptions, thus
disturbing dynamic paths. Innovation is triggered by such actions - the more
unusual and unexpected the actions and perceptions, the greater the impact.
In effect, as evolution is tied here to disturbance of paths, one may conclude
that the existence of multiple and severely varying perceptions is a necessary
condition for socio-spatial evolution!
Creative actions by individual social actors (agents for change) are
beyond the norm and the repetitive routine of human activity. New work
requires the realisation of actions which are at great variance from the
average; in other words, outliers are at the centre of innovation. Creativity in
innovation, novelty and surprise, are the sine qua non not only for
metropolitan development but, more broadly, human development in
general.
A final point to this section. Perturbation of the initial conditions of
metropolitan dynamics, which is the effect of marginal innovation, requires
far fewer resources than those required by a basic innovation, because it is
simply much easier to perturb trajectories than change them.
How one deals with innovative action of the marginal type in
metropolitan development is the subject of the next section.
4.3 SPECULATIVE BEHAVIOUR AND

METROPOLITAN DEVELOPMENT
4.3.1 Instruments for Speculative Action
In this section, we present some notions potentially treatable by nonlinear

dynamics. The applications, however, are still to be worked out in detail. As
such, they are hypotheses to be regarded as a research agenda.
Since speculative behaviour is at the centre of the ensuing discussion, a
definition of 'speculative action' must be supplied, together with a brief
outline of a number of instruments employed by individuals for dealing with
their speculative actions in a dynamic framework. A speculative action is a
current action undertaken by an agent (individual or collective) in
anticipation of a current expectation about a future event. Speculative
behaviour is a set of rules which determine speculative actions by individual
agents. Since no individual has perfect knowledge about any future state of a
social system, only perceptions about it which in turn create expectations,

such actions invoke uncertainty and carry risk. Individuals can be risk-
averse, risk-prone or risk-neutral, depending on their attitude towards
uncertainty. However, before a speculative action is taken, no individual
knows the exact risk involved; individuals possess only perceptions
regarding risks. A number of hedging (insurance) strategies are available for
individuals to limit their exposure to perceived risk. Metropolitan
development, due to its future oriented outlook, involves by necessity
speculative behaviour by numerous agents.
Speculative actions are ways of coping with the uncertainties of the
future. They are also ways of taking advantage of opportunities that
expectations and uncertainties provide. The time horizon, i.e. the period
between the current point in time and the future point in time to which the
expectation relates, carries a premium in speculative action. This premium is
the currently perceived price of the time horizon within which the expected
event is to occur, in effect it is the price of time. Changes in it, along with
changes in perceptions about the future (primarily due to changes in and
caused by the environment), determine the future course of events, i.e.
whether an action is undertaken or not.
Through speculative actions, the future affects the present. The central
concept in speculative action is that of expectations. As explained above,
individuals almost inevitably have different expectations about the future,
and it is this difference (variation) in expectations which leads to the use of
instruments as a basis for action in the various speculative markets. If all
agents were to have identical expectations, there would be no action markets
and no need for speculative action.
Variety of expectations is a necessary (but not sufficient) condition for
action. The availability of resources for action is the sufficient condition.
The amount of resources available for action determines the degree by which
one can disturb developmental paths.
Expectations born out by social events vouch well for their holders.
Those that guessed correctly as a result of informed judgement, or simply by
chance, win in the speculation game. But even a casual look at primary or
derivatives markets reveals that no one guesses correctly always, or even for
a prolonged period. This seems to be so, no matter how much and how
accurate the information that individuals may have, nor how close to or far
from the markets and trading the individuals are, nor how elaborate the
models constructed by the individuals to guide their actions. Somehow,
reality (in this instance the markets) tends to react to individual and
collective actions so as to render itself unreadable.
The evolutionary principles which derive from the above statements (i.e.
from the assumptions regarding the future and from associated expectations)
96 Dendrinos D.S.
could lay down a foundation for analysing metropolitan (and broader socio-
spatial) development. Using traditional evolutionary arguments of
competition, one would expect the winner of a speculative (guessing) game
to amass more resources than others and thus have more to spend on
speculative actions in the future. This would imply that, in an evolutionary
context, competition would favour those who tend to guess right, either
because they possess the right information or due to favourable chance (good
luck). Consequently, one would expect expectations markets to eliminate
those who tend to guess wrong (either because they did not make informed
choices or due to adverse chance), as they eventually exhaust the resources
allowing them to speculate. Under this evolutionary argument of
competition, markets would only comprise of winners!
Obviously, this is not possible, as there are at least two fundamental
flaws to this evolutionary argument:
l. winning requires that there be losers, and in fact there are always far
more losers than winners at any point in time. The argument would
require then that, although there must be winners, those who win are not
necessarily the same all the time, or even most of the time, as life cannot
afford to have only a few that guess consistently correctly. Thus, an
evolutionary argument concerning the competition for right guesses
requires that a co-operative association exists between a few winners and
many more losers, and that the roles be constantly shifting among
individuals, commuting between these two groups;
ii. the term 'informed' can never be granted to a judgement ex ante but ex
post Jacto. It is only the results of speculative action which indicate
whether a judgement was informed or not.
This reasoning stands in sharp contrast to that of 'rational expectations'

where only agents who consistently (although not necessarily always) guess
correctly win. The rational expectations model does not lend itself to any
evolutionary argument. But the evolutionary paradigm of competition has
shortcomings too. Almost everything can be argued on evolutionary grounds
ex post Jacto: one can supply an explanation (and at times even more than
one), for all that has already happened, but this is hardly a scientific base for
analysis and the testing of scientific hypotheses.
a) Instruments Jor speculative action
Instruments for speculative action include, but are not limited to, the
availability and purchase of commodities, stocks and bonds, as well as
derivatives of these tangible products (commodities, resources) and financial
instruments (stocks, bonds, etc.). Urban development implies the acquisition

of commodities or resources (such as land, and primary production factors,
including labour). Financing of development in the private sector is quite
often carried out through the issue of stocks or the floating of bonds. The
undertaking of large capital investment projects by government agencies too
is achieved in particular through the use of bonds.
Commodities, resources, stocks and bonds are primary instruments for
speculative behaviour. Trading in derivative instruments, referred to as
derivatives because they derive their value from the underlying primary
instrument, is also possible (and indeed necessary, as it will be argued later)
for development. The reader is assumed to be familiar with the basics of
these instruments.
On the introduction of such issues for investment, markets in which they
can be continuously traded or exchanged are created. These markets allow
the continuous updating of the prospects (expectations) in relation to these
issues, and thus ensure their eventual viability (economic necessity) and
importance (severity of perturbing development paths). Signals about the
issues are produced through the speculative actions markets for the actors
involved. The efficiency of the signals depends on the proper functioning of
the markets. This issue will be further addressed later.
4.3.2 Derivatives and Nonlinear Dynamics
Two derivative instruments, options and futures contracts, are now

addressed within a nonlinear dynamic framework. Options and futures
occupy a prominent position in today's investment activity in advanced
capitalist economies and are of paramount importance in private sector
activity, Derivative instruments are not without their critics, as their sheer
volume tends to considerably increase or decrease the value of the
underlying stocks. Their total value at any point in time can well be a many
times the value of the underlying primary instruments.
Through the close and highly nonlinear interaction between the
underlying stock prices and the value of derivatives (one obviously affecting
the other in a feedback loop), the volatility in stock prices increases as a
result of the mere existence of derivatives. When the value of the
outstanding derivatives is relatively high compared with that of the
underlying stocks, the impact of derivatives markets can significantly affect
the markets of the stocks themselves.
Another criticism which has been labelled against the application of
derivative instruments, and of far more consequence to us here, is that of the
considerable market imperfections present in their trading forum(s). Because
of the relative thinness of these derivatives markets, (innovative) actions by
98 Dendrinos D.S.
a few actors can have a significant effect on the performance of the markets.
Vice versa, a market action may trigger individual responses which, if
coming from influential actors, can have catastrophic effects (in the
mathematical sense of the word) on the market.
A few of these imperfections will be discussed at greater length in the
context of metropolitan markets. In spite of these and other criticisms, the
use of derivative instruments offers considerable advantages which we shall
also touch upon later. Firstly, we shall provide a few basic introductory
definitions for two instruments, options and futures, for the reader who is
unfamiliar with them.
a) Definitions offutures and options
Options and futures are legal contracts offering their holders certain
purchasing rights under strict expiry conditions. Those rights are acquired in
appropriate derivatives' markets and at prices determined by free and
continuous trading. Markets for options and futures may be formal or
informal, i.e. explicit and regulated, or implicit and unregulated. Futures are
contracts which offer their holder the right (but not the obligation) to
purchase a prespecified quantity of a stock (commodity or resource) in
futures markets, at a prespecified future point in time at a price which is
determined currently (at the time when the contract was purchased). In
effect, the market (presumably a large number of individuals) speculates on
the future price of the underlying issue by discounting for all perceived
factors which could possibly affect the price at that time in the future.
At all points in time between the point of purchasing the contract and the
its expiry or selling date, the holder maintains a current balance (in the form
of either credit or debit) on the value of the contract. Thus, the exposure to
risk could be considerable, as relatively large quantities of the underlying
issue (stock, bond, commodity or resource) is contracted.
Options are contracts giving the holder the right (but again not the
obligation) to purchase (or sell) a prespecified quantity of an issue (stock,
bond, commodity or resource) at a prespecified point in time in the future
and at a preset price (referred to as the striking price). The price of the
contract is determined currently on the appropriate option's market. There
are numerous types of option (call vs. put options, U.S. vs. European
options, etc), but for the purpose of this exposition, these differences are not
relevant. The sellers of these contracts are governed by equivalent
conditions. How and what transaction costs enter into these markets is
largely irrelevant here, although it is recognised that the magnitude and type
of transaction costs involved could potentially affect the performance of
these markets.
Since the most one might lose by participating in an options market is the
price of the acquired options contract(s) (usually a fraction of the value of
the contracted issue), the exposure to risk is much less than for futures
contracts. On the other hand, the time horizons over which futures are
contracted are far longer than those for options. So options, generally,
simply pick up the value (premium) of the remaining time before the
option's expiry date.
Introduction of innovation within derivatives markets may have profound
effects upon these markets. For instance, a decrease in the reaction times to
information diffusion among actors or to trading may offer opportunities for
greater efficiency, but also traps for exaggerated responses.
b) The need for and criticism of derivatives
Derivatives result from the need for risk management. Firstly, because of
the durability of large scale public capital investments, the viability of
alternatives (courses of action) can be sorted out in the derivatives markets
before costly public projects are undertaken. Projects are often built before
obtaining signals concerning potential failure. The planning literature is full
of accounts of unfinished projects or schemes which ended in disaster,
giving testimony to failed planning.
A second reason for the existence of derivatives is that they present
opportunities for those who hold large quantities of an underlying
commodity, resource or any primary financial instrument (stocks, bonds) to
hedge against undesired price movements. The presence of options and
futures markets is a form of insurance. In a metropolitan scene, this concern
for insurance mostly affects land holdings (although by no means not
exclusively). In brief, the existence of options and futures contracts markets
for public plans could offer the opportunity for greater flexibility in deriving
more socially optimum plans or policies more formally and effectively.
On the other hand, criticisms of derivatives' markets abound. For one
thing, these markets tend to be dominated by a few actors-participants, and
monopolistic or oligopolistic conditions are commonplace. Bandwagon
feedback dynamics are widespread, as minor actions or erroneous
information can result in major price swings. As options (or futures) contract
prices increase, the volume of contracts demanded tends to increase too, in a
pattern of trading where time is at a premium. This interdependency between
volume and price can lead to a mass phenomenon of panic buying. Similarly,
as contract prices decrease, the same frenzy can occur, causing prices to
decline yet further as contracts holders decide to exit from the markets as
fast as they entered it.
100 Dendrinos D.S.
Another criticism is that, whereas these derivative markets are supposed

to currently discount future events, they tend to be rather myopic. Over and
under reactions are abundant. Insider trading is also very likely to occur, but
is quite difficult to detect. Highly nonlinear interactions among a number of
agents, influencing their actions in an oligopolistic (or oligopsonistic)
context, are central in the functioning of these markets. Abundant
externalities, high transaction costs, and lack of homogeneity in the
underlying primary instrument augment the criticisms of derivative markets
in metropolitan development.
One might well reject these markets for anyone of the above reasons (or
others too). Nonetheless, it is clear that their working provides considerable
insights into the functioning of social systems, especially their nonlinearities
and dynamics. And from this angle at least, they are useful in understanding
social system dynamics, including the role of chance in metropolitan
development. They also give an idea of how metropolitan evolution might
result from the myriad of individuals and collectives exercising a variety of
options in multiple options markets.
Principles of (human, and possibly non human) evolution can be drawn
from these markets, as already mentioned. Central to such a theory is the
realisation that in a soup of over and under reaction by most actors-
participants in such markets, only a few actors win and then for only short
time periods and never consistently. Chance clearly has the upper hand as to
who will win and when. By focussing on individual actions, as well as
planning and policy making in the context of an options or futures contracts
markets, it becomes more evident why socio-spatial (and broadly, biological)
evolution is largely unpredictable.
Markets (whether secondary, such as options and futures, or primary)
provide us with ample evidence for unpredictability in social dynamics. This
realisation follows numerous routes. One sees the way of approaching the
final collective market outcome as the unpredictable result of a soup of
activity involving the vast heterogeneity of actions performed by individuals
in a game context and following some individual learning process. Another
arrives at this unpredictability through way that the vast differences in
individuals' expectations about the market's final outcome affect individual
behaviour and consequently the performance of the markets.
It could be argued that markets also learn, i.e. collective behaviour
acquires properties of collective action which can be interpreted as the
outcome of some learning. If so, then a central part of such collective
learning may be an effort to act in a manner which would not allow any
specific individual to learn how the collective behaves for a prolonged
enough time period!
The analysis of multiple individual actions within a primary or

derivatives markets framework, using the nonlinear dynamics lens, lends
itself neatly to issues of perturbation. Most individual actions do not matter,
as they are not significant enough to disturb initial conditions of a system's
dynamic path. Only a few actors, mostly by chance, if not by the sheer size
of their intervention, can materially affect paths, as most cancel themselves
out. Imperfect markets require them to do so. And, to the extent that a
system is in a state of chaotic dynamics, individually perturbed trajectories
studied in isolation reveal nothing. A statistical (ergodic) theory of multiple
trajectories must be considered if one is to say anything with any degree of
confidence. This is another legacy of nonlinear dynamics.
4.3.3 Options, Plans and Metropolitan Development
In the final part of this section, we address the role of plans for
metropolitan development. It is argued that there is a greater need to employ
options on metropolitan resources (including land, water, air, the physical
urban environment, and even its historical landmarks) as well as on
metropolitan infrastructure (housing, utilities, transportation network, public
services etc.). Above all, there is a need to derive an options markets for
formal plans.
Options on metropolitan and regional resources is an instrument for
investment already in place in the US, as well as in other Western countries.
They are usually employed in the form of "rights", although in a strict sense
rights and options (or futures) are not identical. However, the differences
between them, which are mostly of an accounting nature, are of no interest
here. Land transactions represent the most widely used forum for their use.
Four examples of rights already widespread in the US are the transfer of
'development rights', 'water rights', 'pollution rights' and 'congestion
rights'. Although these rights behave in an option-like manner, because of
their thin markets they are not traded like options in formal options markets.
In this difference lies the potential for significant innovation, but also the
risk of pitfalls.
In such formal markets, the effects of a large set of metropolitan
development policies could be fully and effectively evaluated. Policies might
be traded in a more formal public forum than the largely informal ones
currently used for metropolitan policy markets. A number of ways of
promoting greater and more systematic public participation and optimisation
(or sub optimisation) in the planning process, as advocated by Harris in this
book, might be more fully addressed in this context. The question of
innovation of this kind in the approach to drawing up metropolitan plans and
policy-making would require much deeper consideration than is possible in
102 Dendrinos D.S.
the confines of this chapter. It constitutes in fact a call for further research as
it is clear that there is a need to address the economic, financial, legal,
institutional, social and other problems associated with setting up these
formal options markets. Such issues, although extremely interesting, are not
of course the main subject of this chapter or book.
a) Plans as Options
Options afford a significant vantage point for interpreting the nature of

plans. One might look at a plan along the lines of a derivative from an
underlying issue. In the case of metropolitan development, a metropolitan
plan could be a special or a general purpose plan, in other words, either a
sectoral or a comprehensive plan. The plan could also be seen as the
derivative of the physical or social structure(s) envisioned by it for the
metropolitan area in question. The plan becomes then, in effect, a contract
offering the planner (or the holder of the contract, the city commission, the
state, or a multi-jurisdictional political entity for example) the right, but not
the obligation to execute the plan in question at some preset date in the
future.
All plans contain a prespecified list of projects, which in this light
represent the preset quantity of the underlying commodities, resources, etc,
in accordance with the formal definition of an option given earlier. Further,
all plans contain an evaluation of the intended action(s), involving the
currently expected future benefits and costs from the plan, i.e. a striking
price. Finally, every plan carries a current composite (political, social,
economic) price at the time it is adopted.
The currently perceived price of a plan changes over time, as new
information comes to bear upon it, or the environmental conditions are
altered. Plans can be abandoned at any time between their adoption date and
their planned implementation (expiration) date, when and if the current price
of that option-plan has collapsed.
Here lies squarely the utility of options markets for plans: before a plan,
which at some time in the past might have looked promising, is implemented
the market for it (and other plan-options) might signal that currently it is no
longer so. The plan is then abandoned, without the social system having to
go through the trauma of undertaking it, only to see it fail once constructed.
Signals regarding the viability of options are of paramount importance in
planning. The more efficient they are, the better the plan-option markets
work. But the markets for obtaining such signals are highly imperfect, and
sometimes do not exist at all. Government monopolies, corrupt political
decision-making processes and the like are widespread in metropolitan
development, under both capitalist and centrally planned economies.
Here lies the responsibility of planners and policy makers: instead of

focusing on producing plans, they would act as facilitators and efficient
regulators of these policy-options markets. This new function would
constitute a fundamental shift in planning and policy-making: from actually
planning or being involved with decision-making, planners and policy
makers could switch to managing (setting up, monitoring, regulating) the
necessary institutional infrastructure which would eventually produce the
most efficient plans and public policies. This might be an effective
instrument for dealing with the unpredictability of reality, as well as the
current malaise in and about planning.
It also points to a moral, ethical and legal responsibility on the part of
metropolitan planners and policy makers, whose role would be to safeguard
the efficient and equitable functioning of the plan and policy options
markets, ensuring high ethical standards in regUlating and even formulating
formal policy and the plan options markets.
b) Metropolitan areas seen as options for development
Economic and social (or, more broadly, cultural) activity settles, i.e.
locates at specific points in space and time, and thereafter develops as a
result of the spatio-temporal choices exercised by the human population.
Metropolitan areas themselves are the outcome of concrete choices exercised
by a myriad of individuals and collectives when confronted with alternatives
in space, in a much broader context than that envisioned by Tiebout (1956).
When the population of a metropolitan area grows, or its economic
output and wealth increases, or its resources appreciate in value relative to
other competing metropolitan areas, this implies that the city's fortunes in
options markets looks brighter and its option's price is on the move. On the
other hand, when a metropolis declines (in terms of population, employment
or wealth) or is abandoned by its population, this simply means that its
option price has collapsed in metropolitan growth-options markets within a
national or global context.
Marginal innovations, occurring through individual actions, which either
set in motion the construction of cities or disturb their dynamic paths, for the
better or the worse, are central in metropolitan development and evolution.
Individual actions manifest themselves in highly interactive, albeit
imperfect, options and futures markets. It therefore seems of some
importance to investigate the specifics involved (qualitatively and
quantitatively) in derivatives markets if we are to understand better the
process of metropolitan development.
104 Dendrinos D.S.
4.5 CONCLUSIONS
A rather ambitious effort has been undertaken in this chapter. Firstly, we

have explored the linkages between three topics: the notion of defining
innovation, the examination of its effects upon metropolitan development in
abstract form, seen in terms of nonlinear dynamics, and an attempt to tie
some marginal innovations (still within a nonlinear dynamics framework) to
individual speculative actions guided by expectations. In the second part,
speculative actions have been approached as the exercise of options in local
markets.
The first set of issues, developed in Section 4.2, is clearly more concrete
and can be utilised in formal mathematical models of metropolitan
development (whatever these models might be). The second set of issues,
presented in Section 4.3, is still in preliminary stages of exploration, and
needs quite a bit of further elaboration. The two sets of issues are, however,
very closely linked. ,
It was demonstrated that approaching innovation from a nonlinear
dynamics standpoint affords not only a useful, but also a comprehensive
classification of innovations into basic and marginal ones. They differ as to
the way in which they are reflected in changes in either parameters or initial
conditions, and as to their effects upon urban development paths potentially
involving phase transitions (metropolitan evolution). Whereas basic
innovations are generally exogenous to a metropolitan setting, marginal ones
could also be endogenous. Certain marginal innovations were seen to be tied
to individual actions taking place within a metropolitan area, and it was
claimed that some later prove themselves to be of consequence.
In a butterfly effect, these individual actions were perceived as altering
initial conditions for metropolitan dynamic paths in phase space. How these
broad principles could be made much more concrete in the context of
existing (and new) dynamic metropolitan models could be the subject of
further research.
Section 4.3 investigated an innovative way of regarding metropolitan
plan formulation and was clearly far more speculative. It was argued that it
could be fruitful to approach individual (and collective) actions and plans
from the standpoint of futures in a formal or informal options markets.
Furthermore, it was advocated that plans and policies could be more
effectively and optimally evaluated in this way. If they were allowed to
operate within the setting of a more formal metropolitan plan and policy
options market, this would permit a much higher degree of flexibility.
Structural advantages and disadvantages to such markets were discussed.
Potentially, the approach proposed could be highly innovative not only
for research, but also planning practice and education. Clearly, the associated
legal, institutional, economic, financial, social and other aspects of such plan
options markets would need to be addressed. These represent another set of
issues requiring further research, especially the question of how the
hypotheses presented could be translated into the form of a mathematical
model.
Acknowledgements
This chapter was prepared while the author was a Visiting Scholar at the Politecnico
di Torino under a grant from the Italian National Research Council (CNR). An
earlier draft of this chapter was used as the basis for Class Lectures at the
Department of Regional Development at Turin Polytechnic. Support from Professor
C.S. Bertuglia was instrumental for the realisation of this project and is gratefully
acknowledged. Helpful comments were received from Professors Britton Harris,
David Batten and Dino Martellato, as well as from Sylvie Occelli from IRES and
Angela Spence. The usual caveats, however, apply.
REFERENCES
Bertuglia, C.S., Fischer, M.M., Preto, G., eds (1995) Technological Change, Economic
Development and Space. Springer-Verlag, Berlin
Cimoli, M., Dosi, G. (1995) "Technological Paradigms, Patterns of Learning and
Development: An Introductory Road Map". Evolutionary Economics 5: 243-68
Dendrinos, D.S. (1992) The Dynamics of Cities: Ecological Determinism, Dualism and
Chaos. Routledge, London
Dendrinos, D.S. (1996) Chaos: Challenges from and to Socio-Spatial Form and Policy,
Discrete Dynamics in Nature and Society
Dendrinos, D.S., Mullally, H. (1985) Urban Evolution: Studies in the Mathematical Ecology
of Cities. Oxford University Press, Oxford
Dendrinos, D.S., Sonis, M.(1990) Chaos and Socio-Spatial Dynamics. Springer, New York
Devaney, R.L. (1989) An Introduction to Chaotic Dynamic Systems. Second Edition,
Addison-Wesley, New York
Feigenbaum, M.J. (1980) Universal Behavior in Nonlinear Systems. Los Alamos Science 1: 4-
27
G1eick, J. (1987) Chaos: Making a New Science. Viking, New York
Jacobs, J. (1969) The Economy of Cities. Random House, New York
Lorenz, E.N. (1963) Deterministic Nonperiodic Flows. Journal of Atmospheric Sciences 20:
130-41
May, R. M. (1976) Simple Mathematical Models With Very Complicated Dynamics. Nature
261: 459-67
Penrose, R. (1989) The Emperor's New Mind. Oxford University Press, Oxford
Poincare, H. (1957) New Methods of Celestial Mechanics. (Reprinted from 1899 French
Edition), Dover
May, R. (1974) Complexity and Stability in Model Ecosystems. Princeton University Press,
Princeton
Schumpeter, J.A. (1934) The Theory of Economic Development, Harvard University Press,
Cambridge
106 Dendrinos D.S.
Sonis, M. (1994) Spatio-Temporal Spread of Competitive Innovations: An Ecological

Approach. Papers ofthe Regional Science Association 52: 159-74
Thompson, J., Stewart, H. (1986) Nonlinear Dynamics and Chaos. John Wiley, New York
Tiebout, C. (1956) A Pure Theory of Local Expenditures. Journal of Political Economy 64,
October
Tufillaro, T., Abbott, Reilly, 1. (1992) An Experimental Approach to Nonlinear Dynamics and
Chaos. Addison-Wesley, New York
Section B:
Path-Dependent Processes of Knowledge Exchange

Chapter 5
The Knowledge-Intensive Nature of Japan's Urban

Development
Kiyoshi Kobayashi I , Sotaro Kunihisa2 and Kei Fukuyama3

IGraduate School of Civil Engineering. Kyoto University. Japan
2The Institute ofBehavioral Science. Tokyo. Japan
3Department ofSocial Systems Engineering. Tottori University. Japan
5.1 INTRODUCTION
Small variations in events in history could have created a pattern of urban

centres very different from that which exists today. In particular, the
locational patterns of knowledge-intensive agents, such as research
laboratories, seem to follow paths that depend upon history (Kobayashi et al
1991). When agents and firms wish to congregate in places where others are
already located, one or a few locations may end up with the entire industry.
Agglomeration economies therefore introduce an indeterminacy. If we
bypass this indeterminacy by arguing that historical accident is responsible
for the dominant locations, we must then define historical. accident and
explain how each winning location is selected.
Tsukuba is a unique example of such a winning location for industrial
research laboratories. It is a Japanese science city which is a new settlement
planned and built by central government, with the aim of generating
scientific excellence and synergetic research activities, by concentrating a
critical mass of research organisations and scientists within a high-quality
urban space. However, especially in the beginning, the progress of Tsukuba
was slower than expected. The response of private industry was particularly
slow in the 1970s, but a major change occurred after 1985, when the mass of
industrial research laboratories in Tsukuba became large enough to outpace
the existing agglomerations in Tokyo.
The spatial economic literature tends to see the spatial ordering of
industry as an economic response to geographical endowments, transport
110 Kobayashi K., Kunihisa S. and Fukuyama K.
possibilities and firms' needs. The locational pattern is considered to be an

equilibrium outcome of individual agents' decisions. In this view, locational
history has no importance, to the extent that the equilibrium outcome is
unique. The locational system is therefore deterministic and predictable.
Recently, an alternative view has emerged whereby agents' locations are
seen as being path-dependent, i.e. an organic process in which new agents
are influenced by inherited locational patterns already in place. Geographical
differences and transport possibilities are still important initially, but the
main driving forces are agglomeration economies, i.e. the benefits of being
close to other agents. Latecomers are likely to be attracted to these same
places by the presence of the early locators, rather than by geography.
What Tsukuba's development story tells us is that once the positive
feedback processes are 'in gear', the agglomeration of research activities is
cumulative and self-reinforcing. In other words, the emergence of a
particular site as a major agglomeration does not depend at this point upon
the intrinsic nature of the site. Historical factors, including political
initiatives, appear to be essential in the selection of a particular equilibrium.
It is well known that minor changes in the socio-economic environment that
occur at critical periods may well result in a very different geographical
configuration (Batten et al 1989; Kobayashi et al 1991; Krugman 1991;
Krugman 1993; Arthur 1994; Fujita and Krugman 1995; Matsuyama 1995).
This chapter attempts to provide a sound empirical example of the
'historical accident plus agglomeration' viewpoint by relating Tsukuba's
development story. In particular, we examine the dynamics of industrial
laboratory location and the selection of an asymptotic locational pattern,
using a simple logit model that permits a description of the lock-in effects
associated with existing agglomerations. By assuming that firms differ
slightly in their locational tastes and that they make their locational choice
one by one - in an order that is random as far as an observer is concerned -
we show that industrial laboratories will indeed cluster in one dominant
location; which one, however, depends both on the geographical
attractiveness of the site and the historical order of the firms' choice. We
also show that the establishment of new research facilities like universities
does not always guarantee the emergence of research cores. The next section
reviews the source of increasing returns in regional knowledge economies.
In Section 5.3, the major driving forces behind the evolutionary locations of
industrial research laboratories are identified. In Section 5.4, we provide an
overview of the geographic landscape of research laboratories in Japan.
Finally, the locational behaviour of industrial research laboratories is
formulated, in order to trace the evolutionary development path ofTsukuba.
The Knowledge-Intensive Nature of Japan's Urban Development 111
5.2 KNOWLEDGE PRODUCTION AND REGIONAL

EXTERNALITIES
Path-dependency is omnipresent in the evolutionary locational patterns of

knowledge-intensive agents such as industrial research laboratories. More
and more laboratories want to locate close to each other because of various
factors that allow for a greater diversity and higher specialisation in
knowledge production, as well as the broader body of knowledge available
for consumption. Setting up a new university in such regions gives rise to
new incentives for researchers to migrate there, because they can expect an
environment which suits their research requirements. In tum, this makes the
place more attractive to research laboratories, because they can expect to
find the types of agents and services they need, as well as new outlets for
their products. Hence, both types of agents benefit from being together. This
idea has already been suggested by Marshall (1920).
In general, Marshallian externalities arise because of (1) internal
economies, (2) the formation of a highly specialised labour force based upon
the accumulation of human capital and face-to-face communications, (3) the
availability of specialised input services and (4) the existence of modem
infrastructure. Apart from the production side, there is an important
advantage of proximity for consumption too. For example, cities are
typically associated with a wide range of products and a large spectrum of
public services. This means that consumers can reach higher utility levels
and therefore have stronger incentives to migrate to cities. The propensity to
interact with others is a fundamental human trait and, as distance is an
impediment to interaction, cities are the ideal institution for the development
of social contacts. Economic life is creative in the same way as are the arts
and sciences, and - as was pointed out by Romer (1986) and Lucas (1988) -
personal communication within groups of individuals sharing common
interests can be a vital input to creativity. In this respect, it is well known
that face-to-face communications are the most effective. Given that different
people have different skills, the size of such groups also gives rise to
significant scale effects (Fujita and Thisse 1996).
There is a circular causation associated with the agglomeration of
research laboratories and knowledge-handling workers through both forward
linkages (an increased number of laboratory locations enhances researchers'
amenity) and backward linkages (a greater number of researchers attracts
more laboratories). Through these linkage effects, scale economies at the
individual laboratory level are transformed into increasing returns at the
level of the region as a whole. The spatial ordering is not unique when these
types of scale economies function in loeational fields. The first firms set up
by historical accident in one or two locations; others are then attracted by
their presence. A different set of events at the outset could have steered the
locational pattern to a different outcome, so settlement history is crucial.
Because of the existence of multiple equilibria, minor changes in the values
of the critical parameters may generate dramatic changes in the equilibrium
spatial configuration. This suggests that historical events can account for
current industrial patterns, and that circular causation generates a snowball
effect that leads manufacturing firms to be locked-in to the same region for
long periods of time (Arthur 1994).
5.3 THE MAJOR FACTORS FOR LABORATORY

LOCATIONS
Thanks to the advent of advanced telecommunication devices, firms

typically conduct some of their activities - such as general business
activities and negotiations with other firms - at headquarters located in the
CBD. Conversely, their plants and factories, in-house laboratories, planning
and R&D activities are located in subsidiary offices out in the suburbs. This
spatial separation of a firm's activities stems from the development of
telecommunication technologies, a factor which is likely to play an even
more decisive role in future patterns of spatial organisation as
communication costs continue to decrease (Fujita and Ogawa 1982).
The most typical intra-firm division of activities involves a head office in
the CBD, plants and factories in the suburbs, and research laboratories
located in scientific environments with good accessibility to public
knowledge stocks, like universities and public research institutes. The
increased intensity of knowledge production is also a significant factor in the
separation of R&D sectors from the firm's other activities. Availability of
skilled labour and the proximity to academic institutions are among the most
significant determinants of the location of industrial research laboratories
(Premus 1982).
Markusen et al (1986) did not find any evidence that research spending is
a significant factor in explaining the distribution of high tech firms in the
US, although educational options at certain educational institutions are an
important determinant. Oakey (1984) indicates that high tech firms seldom
have strong research links with universities. The impacts of the bodies of
knowledge generated by universities (hereafter referred to as 'knowledge
impacts') on firms' activities can result not only from research activities, but
also the accumulation of human capital or derive from the university's
services to the community. The places in which universities are located are
likely to take the lead in generating truly new scientific knowledge.
Consequently, it may be advantageous for firms with R&D departments to
locate in the vicinity of these knowledge generating nodes. The transfer of

knowledge via consultancy activities obviously requires face-to-face
contacts (Florax 1992).
Scale effects refer to critical thresholds that may be reached for the
provision of particular goods and services, while agglomeration effects
denote the amenity that a university might constitute in terms of the social or
economic attractiveness of a region. Knowledge impacts may result in a
dynamic, self reinforcing process. Due to the expenditure impacts of a
university, the resulting increase in regional income may lead to a growing
demand for higher education, as the income elasticity of the demand for
higher education is known to be relatively high. The subsequent increase in
enrolments may lead to a greater knowledge impact and eventually
(assuming a human capital type of effect on earnings) lead to an additional
increase in expenditure. This kind of positive feedback can be seen in
Tsukuba's development process, as shown in Section 5.5.
5.4 THE GEOGRAPHICAL LANDSCAPE OF

INDUSTRIAL LABORATORIES
Very little statistical data are available in Japan for the locational
behaviour of research laboratories, especially those belonging to private
companies. While data on the total number of research laboratories in each
Prefecture are available, they are highly aggregated and therefore unsuitable
for micro behavioural analyses. Moreover, the 'research laboratories'
category may include various other research-related centres and laboratories,
along with the research divisions of companies in the manufacturing
industry. Thus, in this study, the database was constructed by extracting
research laboratory data from the Directory of National Experimental and
Research Organisations, published by the Science and Technology Agency.
In this way, the relationships between these research laboratories and their
parent companies are transparent. In addition, details about the relationships
between the locations of research laboratories, the functions of the parent
companies, and the locations of their factories were ascertained from the
Survey of National Industrial Factories (Nikkan Industrial Newspaper). Due
to the restriction of our focus to laboratories that perform research functions,
we have examined those industries with high R&D investments, such as the
electrical and electronic, mechanical, chemical, medical, and automobile
industries. From these industries, we have chosen the one hundred and forty
two research laboratories of companies with listed stocks, on the basis that
they furnished sufficient data for the intended observation.
114 Kobayashi K, Kunihisa S. and Fukuyama K
Table 5.1 shows the number of research laboratories in the sample which
have located in each area since 1950. The number in brackets shows the
percentage of the total sample number. It can been observed that there is a
strong tendency for research laboratories to concentrate in large cities,
especially the Tokyo Metropolitan Area, which includes the Ibaragi,
Saitama, Chiba, and Kanagawa Prefectures. The total number of new
laboratories set up reached its peak during the period 1961-1970.
Table 5.1 Locations of research laboratories by date and region (sample

results only)
Location Year of establishment

Pre 1950 1951-60 1961-70 1971-80 After 1980 Total
Tokyo Metrop. Area 22 12 21 12 I3 80

% of total sample (15.5) (8.5) (14.8) (8.5) (9.2) (56.3)
Kansai Metrop. Area 8 5 5 4 3 25
(Osaka-Kobe-Kyoto)
% of total sample (5.6) (3.5) (3.5) (2.8) (2.1) (17.6)
National total 40 20 37 25 20 142
% oftotal sample (28.2) (14.1) (26.1) (17.6) (14.1) (100)
The research laboratories that were selected can be categorised according

to their relationships with their parent companies: (1) located at the head
office, (2) located at the factory, or (3) located separately (stand-alone).
Table 5.2 indicates the distribution of these categories by industry. Among
the one hundred and forty two sample research laboratories with complete
data about the characteristics of their associated companies, it emerged that
twenty seven were located at their head offices (19.0%), sixty three at their
factories (44.4%), and seventy one (36.6%) separately. (Note that because
the companies with head offices and factories in the same location were
allocated to both categories 1 and 2, the total sum of research laboratories
counted exceeds the total sample of one hundred and forty two.) The
absolute number of research laboratories located at the site of head offices is
relatively low (under 20%), most of these being in and around the Tokyo
Metropolitan Area. While many of research laboratories located at the
factory site can be found in the Tokyo and Kansai Metropolitan Areas, the
overall number of research laboratories located in these areas has been
declining significantly. This is partly due to the restrictions introduced by the
enforcement of the 'Law on Restriction of Locations of Factories and Other

Facilities in Existing Urban Areas' (1959). Recently, there has been a strong
tendency for research laboratories located at factories to decentralise to
suburban areas. Other types of research laboratories have tended to
concentrate in Tsukuba (Ibaragi Prefecture).
Table 5.2 Locational classes of research laboratories in sample

(by industry)
Locational Class Type of Industry
Chemical Medical General Electrical Auto Others Total

mechanical and
electronic
Stand-alone 21 10 7 17 o 16 71
% total sample (14.8) (7.0) (4.9) (12.0) (0.0) (11.3) (50.0)
With head office 2 4 6 10 I 4 27
% total sample (1.4) (2.8) (4.2) (7.0) (0.7) (2.8) (19.0)
With factory 11 II 10 15 3 13 63
% total sample (7.7) (7.7) (7.0) (10.6) (2.1) (9.2) (44.4)
With head office o 4 4 7 3 19
and factory
% total sample (0.0) (2.8) (2.8) (7.0) (0.7) (2.1) (13.4)
Total 34 21 19 35 3 30 142
% total sample (23.9) (14.8) (13.4) (24.6) (2.1) (21.1) (100)
Table 5.2 shows the locational classes of research laboratories by

industry. Here we can see that the electrical and electronics industry has the
largest in number (24.6%), followed by the chemical (23.9%) and medical
(14.8%) industries. In the chemical industry, the stand-alone laboratory is the
largest class with 21 research laboratories. Following the chemical industry
in the number of stand-alone research laboratories are the electrical and
electronic, medical, and mechanical industries. Thus, it would seem that the
industries which depend relatively intensively on R&D and fundamental
research, tend to have stand-alone research laboratories. On the other hand,
laboratories that require large-scale assembly machinery or involve the
development of production processes may need to be located close to the
manufacturing plant. In fact, all three samples associated with the
automobile industry have their research facilities located at the factory site
and no stand-alone laboratories.
Data on the private companies that located their stand-alone research

laboratories in the Tokyo Metropolitan Area after 1970 were extracted from
the Directory of National Experimental and Research Organisations (Science
and Technology Agency). The targeted industries were those with more
stand-alone research laboratories, namely the chemical, electrical and
electronic, medical, and mechanical industries. Moreover, a telephone
survey was carried out in order to enrich the information for those
companies where data in the Directory was insufficient. The largest
percentage of the sample (a total of one hundred and fifteen laboratories)
belonged to the chemical industry (approximately 30%) and to electrical and
electronic industry (around 20%).
Fig. 5.1 shows the changing location pattern of these laboratories. Before
1980 the main concentration was in Kanagawa and Tokyo. In that period,
there were few laboratory locations in Ibaragi, but since 1985 the number
has increased dramatically. The overall number of laboratories increased
most rapidly during the period from 1985 to 1989. Fig. 5.2 shows the spatial
location of laboratories by industry. All four industries illustrated in this
figure, and particularly the chemical industry, are concentrated in the Ibaragi
Prefecture (where Tsukuba is located).
120
100
80
60
40
20
Figure 5.1 Changing patterns of research laboratory locations in the Tokyo

Metropolitan Area
120
100
80
60
40
20
Total
o General Mechanical
Industry
~~ '. Automobile Industry
ro
.Q +
'n
.c \
U \
Chemical Industry
Figure 5.2 Spatial location of research laboratories by industry
Fig. 5.3 indicates the spatial configuration of research laboratory

locations throughout the Tokyo Metropolitan Area. It can be seen that the
laboratories tend to be clustered in specific parts of the metropolitan area.
The stand-alone research laboratories have evolved not in the Tokyo CBD,
but in the suburban towns with a population of one hundred thousand to two
hundred thousand people and with a land price of one hundred thousand to
three hundred thousand yen/m 2.
The construction of a planned academic city, called Tsukuba Research
and Academic City (hereafter Tsukuba), commenced in 1970, after the
Tsukuba Science City Construction Law was passed by the government.
Undoubtedly, the emergence of a new research agglomeration affected the
locational behaviour of industrial research laboratories. However, at first
Tsukuba developed more slowly than expected. In its first decade the
response of the private sector in particular was very slow and few
laboratories were established there. A dramatic change occurred in 1985,
when the International Science and Technology Exposition was held in
Tsukuba, bringing improvements in infrastructure, including a new
expressway. Since 1985 the number of private research laboratories has

increased dramatically and the city is now the dominant location for research
laboratories. In the next section, we trace the evolutionary development path
of Tsukuba to elucidate how it was geared to take off as a self-sustaining
research agglomeration.
Tohoku Shif:j,kansen; {
/ Ibaragi Pref.
I
Tochigi Pref. I.
I
I
I
I
o more than 20
o 10-20
o 5-10
less than 5
ml\ior expressways
Shinkansen railways
Figure 5.3 Spatial configuration of research laboratories

5.5 A LOCATION MODEL OF IN-HOUSE

RESEARCH LABORATORIES
5.5.1 The Structure of the Model
The locational behaviour of industrial research laboratories has been

described here by a binary discrete choice model which identifies whether
they locate 'in Tsukuba' or 'not in Tsukuba', i.e. elsewhere within the Tokyo
Metropolitan Area. Given the particular characteristics of our choice
contexts, the following points should be taken into account. The choice
option denoted 'not in Tsukuba' is a highly aggregate one, as it includes
many bundles of potential location sites. In fact, the metropolitan area boasts
agglomeration magnets caused by the existing milieu of knowledge
resources. The locational attributes are extremely heterogeneous, a feature
which reflects the power of inertia of the Tokyo Metropolitan Area. In order
to grow autonomously, Tsukuba needed to be able to conquer the lock-in
effects of the existing spatial structure. This implies that there was a critical
threshold beyond which Tsukuba could be achieve a self-reinforcing
evolutionary process. The potency of the Tokyo Metropolitan Area can be
modelled using a disaggregate choice approach (McFadden 1978).
5.5.2 Model Formulation
When the number of alternatives is large, corresponding probability

models are data hungry and need a great deal of processing time. The special
structure of the multi-nomial logit model permits a reduction in the scale of
the problem by either aggregating alternatives or by collapsing a sample of
the full alternatives into a smaller number of primary types. In what follows,
primary choices are introduced to aggregate a number of potential
alternatives.
Suppose that firm n (n = 1,., N) is faced with the binary choice
of Ai (i = 1,2), where AI denotes 'in Tsukuba' and A2 'not in Tsukuba'. The
primary choice Ai can be regarded as a set of Mi possible alternative
location points Bij (j = 1,, Mi). The desirability (utility) of firm n locating
at point Bij (j = 1,, Mi) E Ai' denoted by U;, is fully characterised by its
attributes xij = {x!, ... ,xij )1 . To avoid the heavy burden of data collection, let
us suppose that a firm's utility varies stochastically over alternative

locations. Let the partial utility that the firm can obtain by locating its
research laboratory at Bi} E Ai , denoted by V;', be expressed as the following

linear sum:
(5.1)
where vi} is the deterministic part of the utility function which is common to
all laboratories, &; is the stochastic part of utility that fluctuates
stochastically due to unobservable factors, and ()'= (Ot, ,Om! are
parameters to be estimated. Now, the probability Pi)' that point / is chosen
from amongst the possible location points in Ai can be given as:
p~.y = Prob{V.n.y~ ,
V;~ for all j in Ai} (5.2)
where the stochastic variables &; are identically and independently

distributed according to a Gumbel distribution. The probability that
alternative )" is chosen is given by the following equation:
(5.3)
The expected utility is then expressed as:
I exp{ ()' Xi}}

M
E[max j {V'i }]= In (5.4)

j
Defining the mean of the locational attributes across Mi location points

of type i as:
Xi= LXi}IMi (5.5)

j
let the attributes vector of point Bij E Ai' (}xij' be expressed as follows (by
utilising the vector ~ij that represents the estrangement from the average
attribute of the arbitrary possible location point in choice set Ai, Xi:
(5.6)
Then equation (5.4) can be rewritten as follows:
(5.7)
where OJ; (B) is a correction factor for heterogeneity within alternative A;

which satisfies:
(5.8)
The research laboratories locate either 'in Tsukuba' or 'elsewhere, within

the Tokyo Metropolitan Area'. Let this locational behaviour be expressed as
a choice between 'in Tsukuba (AI)' and 'not in Tsukuba (A 2 )'. Now the
utility that firm n obtains by alternative A; (i =1,2), U;n , can be formulated
as follows:
(5.9)
where t' are unknown parameters, y; is the characteristic vector peculiar to

the aggregated alternative A;, and '1:' is the observation error inherent to
firm n. The first and second terms of equation (5.9) represent the average
utility level of alternative A; , the third term the feasibility of location, and
the fourth the degree of distinction among possible location points (i.e. the
diversity of choice). The third and fourth terms represent the potency of the
existing agglomeration. If these terms are sufficiently large, the spatial
structure can exhibit lock-in effects. The option 'not in Tsukuba (A 2 )'
strongly overwhelms the 'in Tsukuba' (AI) option in the initial stage. On the
other hand, the effect of the first and the second terms in (5.9) can be
controlled to some extent by policy initiatives. Tsukuba becomes self-
reinforcing when the magnitude of these two terms passes the threshold
represented by the third and fourth terms. By assuming again that the error
term 'lin is distributed according to a Gumbel distribution, we obtain the
following binary logit model:
p = exp{t' y; + O'x; + InM; +w;(O)}

(5.10)
I Lkexp{t'Yk+O'xk+lnMk+wk(O)}
A useful approximation to OJ; (B) can be obtained if the deviations qij

within A; can be treated as independent random variables drawn from a
multivariate distribution which has a cumulative generating function WO. If

the number of alternatives M is sufficiently large, then the law of large
numbers implies that Wi certainly almost converges to Wi = W(B). For
example, if qij is multivariate normal with covariance matrix Ii , then:
(5.11)
A practical method of estimation is to use the normal approximation to

Wi' but with some restrictions imposed on i when summing over i (1:;).
Then B can be estimated using maximum likelihood estimation of equation
(5.10). The procedure can be iterated using intermediate estimates of B III
the formula for Wi (McFadden 1978).
5.5.3 Model Estimation
The data set used for model estimation was gathered from eighty research
laboratories located throughout the Tokyo Metropolitan Area (including
Tsukuba). All data were pooled for parameter estimation. The attributes of
the chosen location were described using data drawn from the specific site in
that year when the laboratory concerned was actually located. Those
locations which were not chosen were represented by the average values of
the corresponding alternatives in each location. In calculating the average
value of the attributes of the Tokyo Metropolitan Area (option A2 ), the
whole area was divided into six zones, and the average attributes of these
areas was assumed to represent the average locational attributes of each
location. Alternative models were estimated according to different
combinations of explanatory variables. The desired estimations are
summarised in Table 5.3, which has the largest likelihood ratio among the
estimates satisfying the desired conditions. The likelihood ratio and hit ratio
are 0.806 and 95.4%, respectively, and therefore the model fits the data well.
The selected explanatory variables are: 'the number of researchers', 'land
price', 'travel time to Tokyo', and 'population density'. In calculating the
number of researchers, the researchers in the public and private sectors have
been included. 'Population density' also includes those working in both
public and private sectors. Hence, the variables 'travel time to Tokyo',
'number of researchers' and 'population density' can be partially controlled
by policy initiatives. However, the basic feature of the estimated results is
that the attractiveness of Tsukuba is regulated by endogenous variables
which serve as proxies for agglomeration economies. In other words, it can
be seen from Table 5.3 that the agglomeration itself furnishes the conditions
for further agglomeration.
Table 5.3 Results of parameter estimation
Variables Parameter values t -values

Travel time to Tokyo (minutes) -0.1359 -1.43
Population density (no. o/personslkm2) 0.5775.10-3 1.31
Land price (l 04 yenlm2) -0.2202.10-5 -1.54
Number of researchers (no. o/persons) 0.1245.10-2 2.84
The likelihood ratio 0.806

The hit ratio 95.4%
5.5.4 Self-Reinforced Evolutionary Development in

Tsukuba
As mentioned earlier, Tsukuba is a science city founded with the aim of

generating scientific excellence and synergetic research activities by
concentrating a critical mass of research organisations and scientists within a
high-quality urban space. Tsukuba is located about 60 kilometers north-east
of Tokyo and 40 kilometers northwest of Tokyo International Airport. It is a
national research centre totally funded by central government. Laboratories
engaged in basic research are also government funded. Tsukuba covers
28,559 hectares, encompassing 48 national research and educational
laboratories. Some relocated here, others were newly established. At the end
of the 1980s, 30 percent of all national research agencies and 40 percent of
their researchers were located in Tsukuba, and about 50 percent of the total
budget for all laboratories was invested there.
Fig. 5.4 shows how Tsukuba has evolved over the last two decades. The
central government continued to invest in Tsukuba until 1985, when
Tsukuba became geared to self-reinforcing evolution. The establishment of
hotels went hand in hand with the birth of research laboratories. Service
firms followed soon thereafter. The most prominent feature of Fig. 5.4 is the
changing probability that research laboratories will choose to locate in
Tsukuba. Before 1981, the probability of choosing Tsukuba was almost nil,
but by the late 1980s it had approached unity. This shift was not smooth, but
abruptly nonlinear.
From Tsukuba's experience, we have learned that it is difficult to change
the existing spatial structure because of the lock-in effects associated with
existing agglomerations. This can be partly explained by the magnetic power
of Tokyo and the formation of self-fulfilling prophecies about the
development of Tsukuba. It seems reasonable to view existing cities as focal
points that help agents co-ordinate their spatial decisions. Reshaping the
urban landscape therefore requires major changes in agents' expectations
(Krugman 1991). Although there is a great deal of flexibility in the choice of
location a priori, there is also a strong rigidity of spatial structure once the
process of urbanisation has started. These factors could explain why many
planned research cities have failed to develop successfully. Governments
have often withdrawn their support from key urban activities before some
critical mass could be reached (Fujita and Thisse 1996).
1 r-------------------------------~----~~
Number of national
09 r---------~~~~~~~~----~~----~~~~----~
research institutions (2)
08 r---------------------~_=--~~~----------- r----~
01 r-------------------------~~~--------.----------~
06 r-----------------------~~~---------------------~
os r-------~~------~--~~--~----~~----------~
03 r-------------__~-:--~--~----------------------~
02
Ol ~~------~~----------L-------------------------~
o - A --"-'-_
1975 1980 1985 1990
(I), (2), (3) are nonnalised such that their 1990 values = 1.0
Figure 5.4 The evolutionary development path ofTsukuba
5.6 CONCLUSIONS
This paper has attempted to provide a sound empirical example of self-

reinforcing evolutionary trends by describing Japanese experiences in
Tsukuba. As we have shown, the location of knowledge-intensive agents
such as research laboratories is path-dependent; in the sense that latecomers'
locations are very much influenced by inherited locational patterns already
in place. The agglomeration of research activities displays all the features of
positive feedback processes (Arthur 1994). The mechanism underlying the
agglomeration processes is nonlinear in character. Because the system has
multiple equilibria, the location of a new agglomeration is difficult to
predict. This implies that historical factors like political initiatives appear to
be essential in the emergence of research cores. It is also true, however, that
once an agglomeration of knowledge-intensive agents has formed, it

becomes very difficult to change the existing spatial structure by any policy
means because of the lock-in effect exercised by existing agglomerations.
REFERENCES
Arthur, W. B. (1994) Increasing Returns and Path Dependence in the Economy. Ann Arbor,
The University of Michigan Press, Michigan
Batten, D. F., Kobayashi, K., Andersson, A.E. (1989) "Knowledge, Nodes, and Networks: An
Analytical Perspective." In Knowledge and Industrial Organization, A. E. Andersson et al.
(eds. ), Springer-Verlag, Heidelberg Berlin
Florax, R. (1992) The University, A Regional Booster? Avebury, Aldershot
Fujita, M., Krugman, P. (1995) When is the Economy Monocentric? Von Thtinen and
Chamberlin Unified. Regional Science and Urban Economics 25: 505-28
Fujita, M., Ogawa, H. (1982) Multiple Equilibria and Structural Transition of Non-
Monocentric Urban Configurations. Regional Science and Urban Economics 12: 161-96
Fujita, M., Thisse, J.-F. (1996) Economic Agglomeration. Discussion Paper 430, Institute of
Economic Research, Kyoto University, Kyoto
Kobayashi, K., Batten, D.F., Andersson, A. E. (1991) The Sequential Location of Knowledge-
Oriented Firms Over Time and Space. Papers in Regional Science 70: 381-97
Krugman, P. (1991) History Versus Expectations. Quarterly Journal of Economics 106: 651-
67
Krugman, P. (1993) First Nature, Second Nature, and Metropolitan Location. Journal of
Regional Science 33: 129-44
Lucas, R. E. (1988) On the Mechanics of Economic Development. Journal of Monetary
Economics 22: 3-22
McFadden, D. (1978) "Modelling the Choice of Residential Location." In Spatial Interaction
Theory and Residential Location, Karlqvist, A. et ai, eds, North-Holland, Amsterdam, 75-
96
Markusen, A., Happ, P., Glasmeier (1986) High Tech America, The What, Where. and Why of
the Sunrise Industries. Allen & Unwin, Boston
Marshall, A. (1920) Principles of Economics. 8th edition, Macmillan. London
Matsuyama, K. (1995) Complementarities and Cumulative Process in Models of Monopolistic
Competition. Journal of Economic Literature 33: 701-29
Oakey, R. P. (1984) High Technology Small Firms, Innovation. Frances Pinter, London
Premus, R. (1982) Location of High Technology Firms and Regional Economic Development.
US Government Planning Office, Washington DC
Romer, P. M. (1986) Increasing Returns and Long-Run Growth. Journal of Political
Economy, 94: 1002-37
Chapter 6
Interurban Knowledge Networks
Martin J. Beckmann
Technical University. Munich. Germany
6.1 INTRODUCTION
In the examination of the processes through which innovative techniques

and ideas are produced and diffused, considerable interest has been raised by
the concept of knowledge networks. By knowledge networks we mean the
nodes and links that map the interactions of knowledge producers in
scientific, technical and academic fields. The nature of these interactions and
the strength of the links differ substantially between pure basic research and
applied research and development, which in many respects represent the two
extremes.
In the case of applied research, knowledge becomes a marketable good,
whose exclusive possession confers significant economic advantages. These
advantages can be appropriated by profit-seeking enterprises, but this puts a
premium on secrecy and results in the exchange of knowledge only on a cash
basis or against other knowledge of equal market value.
In this chapter, we focus on the case of pure basic research which, unlike
applied research, produces a public good of no market value. In fact, to
achieve maximum benefit it needs to be communicated as cheaply and
rapidly as possible. The exchange which operates here is between knowledge
and the recognition of 'prestige', i.e. the gain in scientific reputation that
adheres to new discoveries. Not only would secrecy bring no advantage, but
it would also be in violation of 'the ethos of science', which stresses the
importance of unrestricted disclosure of new findings and their sources,
including data, historical or experimental records.
In small fields, where the creative scholars or scientists are so few as to
be personally acquainted, this communication can be personal and informal,
128 Beckmann M.J.
occurring through oral communication or correspondence. Even then, the

claim to scientific fame must eventually be established through publication,
i.e. documentation of the innovative findings in regular outlets such as
scholarly journals or conference proceedings.
In larger fields, too, scientific communication before definitive
publication in journals or proceedings plays an active and ever growing part
in the knowledge production process. This is illustrated by the typical story
of how a scientific paper, as the embodiment of new knowledge, is brought
into existence. (Note that we refer here to the case of non-proprietary, i.e.
pure basic research).
The new ideas themselves are most likely to come to light not through
lecturing or discussion, but in the isolation of individual thought. Their
validity and novelty, however, need to be tested through informal talks with
close colleagues. If they pass this initial 'test', the originator will then
present them to other staff of his or her own institute and for this purpose
will draft a discussion paper. Following the presentation and the attendant
discussion, the head of the institute may advise presentation to a wider
professional audience at a regional or national meeting of a professional
association. This will normally require the approval of one or more referees
invited by the conference organisers.
Publication, either in the form of proceedings or in a scientific journal,
will once more be subject to a refereeing process plus editorial approval. All
of these steps imply interactions of scientific personnel, which after the
initial phase need not, and in general will not be, in the same location I. The
audience, too, at a professional meeting, will be brought together from a
more or less extensive 'market area' for knowledge in this particular field.
Publication in book form may also be sought, but usually takes too long
for the purpose of dissemination of new knowledge. Books are however an
important aid in the later distribution of knowledge through teaching and the
consultation of reference works.
6.2 LOCATION FACTORS
Publications are equally accessible in all locations: the cost and time
difference in the mailing of journals are not significant. Moreover, research
is essentially a footloose industry, tied neither to 'resource deposits' for raw
material inputs, nor to markets of a large consuming public. It would be
In applied research, the only exchange of ideas without prohibitions or restructions is

possible locally, i.e. with colleagues in the same organisation.
Interurban Knowledge Networks 129
erroneous, however, to conclude from this that location does not matter and
that all places are equally suitable for knowledge production.
As a first observation, we note that the researcher invariably requires an
organisation to provide support, since only organisations can afford the
infrastructure of libraries, advanced computers, data banks and
communication links, not to mention the secretarial and research assistance.
Even Karl Marx, denied a professorial chair, had to immerse himself in the
British Museum to write Das Kapital! Organisations also provide the
opportunity for constant interaction with fellow researchers, who serve as
sources of both reference information and critical opinion. Moreover, the
synergy of teaching and research can be achieved only in a teaching
institution, such as a university.
Secondly, people of scholarly and scientific talent, especially at the
higher levels, show distinct preferences in their choice of residential
location. As their 'consumption' tends to include cultural services, such as
good schools for their children, musical and theatrical events, exhibitions,
and so on, they will opt where possible, to live in places where these are
available, which almost invariably implies a metropolitan area. Students, too,
will be more attracted to metropolitan areas with varied cultural facilities
than to remote places in the hinterland. It is not easy for organisations
located in inaccessible places to draw the specialised personnel needed to
build up a scientific community of renown. The knowledge network of
scholars/scientists is thus imbedded in a network of institutions located, for
the most part, in a network of metropolitan cities.
6.3 MEASURES OF DISTANCE
It is an axiom of regional science that 'interaction decreases with

distance'. Is this true also in the arcane world of learning, where ideas might
move freely and be not tied to any particular place? Whereas journals and
books may be equally accessible in any location, this is not the case for kind
of the personal contacts whose importance in the formulation and
development of new scientific ideas was stressed above. Even in an age of
easy communication and rapid transportation, such contacts appear
hampered by distance. This can be tested through the analysis of two types
of interaction and flow of ideas which are sufficiently manifest to be easily
observed and measured - citation and joint publication. Table 6.1 shows the
location of works cited in the Administrative Science Quarterly 1981-1992
and indicates that citation decreases with the distance between institutions.
The majority of citations refer to research undertaken in the same
institution or neighbouring ones, suggesting that, even in the researchers'
w
Table 6.1 Citations in papers from US universities by location o
IJ:l
(I)
<)
o
c o:l ;0;-
00 o:l .v; tI) E
;>, <) '0 <)
<) :0 0 ::J tI) o:l tI) ...; S
"E tI) ."0
.: c d if] til o:l _
C <) ~ ~
"0 0 .D E o .5 "0 <) >-
<8 ..:.: ;:: u E d ..:.: -.o o C U u o:l
tI) N
E ... C cIi
c... :.c:u
~ ...
<) :. o:l "0 0... c<) o o ~ ~ 0
~ ~ :.a
cZi co u u ~ ::i 0... if] ~ ~ ~ U Z i ~ if] U f- ~
- -i >- .....
Stanford 35 16 10 10 8 7 6 6 5 6 4 4 4 4 3 3 3 3 3 3 142
Berkeley 11 II 5 2 7 6 11 o o 2 3 9 o \3 4 5 3 o 94
Illinois 4 o 9 4 2 3 o 7 2 o 1 2 3 2 o 6 4 o 51
Cambridge 2 o o o 4 2 2 o o o o o o 2 2 o o 17
Columbia 4 3 0 1 10 2 5 o 2 o 3 o o o 4 1 o o 37
Wisconsin 2 1 002 2 o 0 2 0 o o 1 2 2 o o o 16
U.Penn. 3 3 209 3 o 0 3 o o o 4 2 1 2 o 35
Penn.SU 0 o 0 0 2 o o 6 o 0 o o o 3 o o o o \3
York 0 2 002 o 0 o 0 I 0 o o o 0 I o o o o 10
San Jose 3 o 0 0 2 o 0 o 0 o 3 o 1 o o o o o o 10
Arizona o 0 o 0 2 0 o 0 o o o o o o o 7
Minnesota o I 2 0 o 3 o 0 o o o 1 1 2 o o 14
Cornell 3 o 0 I 14 4 2 o 8 o 3 o o 8 o I I o o 46
Michigan 2 o 000 2 2 o 4 o 2 o o o 0 o o o o o 12
SUNY 2 o 1 0 1 o 2 o 0 o o o I o 2 4 o 16
Total 73 37 28 21 59 31 40 10 32 16 22 14 7 9 46 20 15 16 16 4
Source: Dr. OIle Persson, Dept. of Sociology, University ofUmea
field of specification, there is limited knowledge of work carried out

elsewhere. In some cases, contact is maintained with students, friends or
former colleagues who have moved away to more distant places. But
sustained personal contact tends to decrease with distance, even when
communication costs are uniform. What the citation statistics imply is a
rather serious phenomenon, i.e. that even scientific results travel with some
difficulty in space, even where no language barriers intervene. This would
appear to be in contrast with scholarly standards and ethics, which require
that reference be made to the sources most appropriate to the enquiry,
regardless of where they are located.
How should distance in knowledge networks be defined or measured?
This clearly depends on how distance enters into scientific interaction.
Sociologists of science have argued (Crane 1972) that scientific interaction
is a social phenomenon, based on personal and often face to face contacts.
The natural measure of distance in knowledge networks would seem
therefore to be travel time. This was the variable used in the studies of
collaboration at a distance whose results are shown below.
The overpowering deterrence of language differences reveals itself for
instance in the citation matrix for 'Invisible Colleges' (Persson and
Beckmann 1995) in which three French-speaking authors were discovered to
form an independent subset receiving no citations from authors not in the
subset.
In the case of scientific collaboration, the distance effect is both stronger
and the evidence more weighty. In an unpublished paper, Persson (1991) has
documented the effect of distance on collaboration for the Nordic countries.
Ake E. Andersson and Olle Persson (1993) have fitted a gravity model to
articles from authors in the so-called Creative Regions of Europe 1988-1990.
Distance was measured as cost of contact based on travel time and general
transportation cost, producing statistically significant results.
l.S.Katz (1994) has studied geographical proximity and scientific
collaboration for authors from Canada, Australia and the United Kingdom.
Beckmann and Persson (1996) have studied collaboration between authors in
Swedish universities and estimated a gravity model with statistically
significant results.
A theoretical model can in fact be constructed, showing that the
probability with which authors, acting as rational agents, will choose
collaborators in their field decreases with distance. In a random utility model
containing a normally distributed productivity factor bij for collaboration
between persons i and j, the distance effect is approximately exponential,
putting scientific collaboration squarely into the company of most other
types of human interaction (Beckmann 1994).
132 Beckmann M.J.
6.4 INTERURBAN KNOWLEDGE NETWORKS
What does this imply for interurban knowledge networks? In an

interurban network, some cities are located closer to the main grouping,
while others are relatively isolated. The notion of centre and periphery is
thus relevant also in knowledge networks. Even in medireval Europe, centres
of learning such as Bologna, Padua, Paris, Oxford, Heidelberg, Vienna and
Prague, fared better than more peripheral places like St. Andrews, Uppsala
or Salamanca. Although the citizens of Gottingen in the 18th century are said
to have expressed disappointment about receiving a university rather than
the state prison (which went to Celie), universities and research facilities are
generally considered assets to their urban community. They attract scholars
aT!A "cientists whose education and income set them above the average and
attract students who, although sometimes troublesome, may nevertheless
contribute substantially to a city's income.
On the other hand, what are the factors that induce a university or
research centre to locate in a city? Most important would seem to be its
cultural amenities, in particular good schools, theatres and a flourishing
musical culture. Decent restaurants and attractive 'watering holes' (drinking
places) are also contributing factors. A benign climate, not too diminished by
air and water pollution, would seem to be an added attraction. Once
established, a university will itself in no small way augment a city's cultural
life. An elevated position in the knowledge network thus tends to be
associated with a city's prominence in the cultural network.
Today's most distinguished research universities are invariably close to
or even in the centres of great metropolitan areas rather than in the
periphery. This is quite clearly seen in North America, itself in a wider sense
a centre for the world economy. Are 'think tanks' an exception to this? By
think tanks are meant communities of scholars brought together from their
home institutions for interdisciplinary interaction for limited periods of time.
To discourage distractions from the outside world, the locations chosen are
usually quiet environments away from busy centres, and often beyond the
fringes of major metropolitan areas. There is therefore greater freedom in
choosing locations for knowledge activities over shorter periods than longer
ones. In principle, the same also applies to professional meetings, except that
the need for ample hotel facilities imposes limitations that. brings them back
to major metropolitan centres!
We should like to mention in passing that Applied Research and
Technical Development are not footloose activities, but clearly need to be
market oriented. Contacts with clients and buyers, and therefore
geographical proximity, is of paramount importance. Laboratories and
research centres are typically located near corporate head quarters (General
Motors once moved a research facility from Santa Barbara, California, to

Warren, Michigan, for better contact and control). Other examples of
successful research facilities near corporate headquarters are Bell
laboratories and IBM's main research centre in Yorktown, New York.
Commercial research enterprises seek locations in centrally located
metropolitan areas.
In the scientific world, 'periphery' is frequently used to refer to areas
outside the large knowledge-producing countries of Europe and North
America. In this sense, even economically important countries like Japan
and Australia are 'peripheral'. Relative isolation may also result from
differences of language. As English has become the language of
international communication in science, publications in languages other than
English have suffered from neglect. As mentioned above, this phenomenon
shows up in citation behaviour. The citation matrix reveals independent
language-specific subsets of authors not cited by authors outside their subset
(Persson and Beckmann 1995). Authors in countries outside the English-
speaking world are thus handicapped unless they publish in English, and in
any case are more limited in the informal types of verbal exchange that play
such an important part in scientific knowledge production. It is to be
expected, however, that improvements in communication and transportation
technology, which have stimulated a considerable growth in scientific
collaboration (Fig. 6.1), will further reduce the disadvantage of the periphery .
..
90
80
70
60
~ ----------------
40
2D
10
o~--~~--~--~--~--~~
1900 1910 1920 1930 l!I4n 1!l5(J 1960 I!I7O
Figure 6. J Incidence of multiple authorship as a function of date

(Source: Price 1963)
134 Beckmann M.J.
MATHEMATICAL APPENDIX
Utilisation of outside knowledge accessible at a distance.

To demonstrate the importance of the accessibility of other knowledge
centres for knowledge production in a location i, let knowledge be produced
from labour Lj (scientific effort), local knowledge Kj and outside knowledge
*
Kj available in a location J i. Assume a Cobb Douglas production function
and opportunity costs (in terms of time spent) Wj for labour, Cj for utilising
local knowledge and Cjj for accessing knowledge through contacts with
location J (J should be thought of as an aggregate of various locations away
from i).
Scientists in the efficient pursuit of new knowledge to benefit themselves

and the scientific community seek:
B1 =L;,K
max ,Kj j
b.L~Kf1K~
1 1 1 J
-wL.
1 1
-cK
1 1
-c IJ.. K.J (6.1)
Alternatively, the opportunity costs may be replaced by a time budget

constraint. (Beckmann 1994). Comparing the use of local and outside
knowledge Kjand Kj respectively.
(6.2)
aBj- --yL'
0 -- - TaKPKY-1
. J -c .. (6.3)
aK. 1 1 IJ
J
Dividing:
K = ---.!...
y c
_I (6.4)
KJ fic IJ..
Thus the relative utilisation of outside knowledge decreases with the cost
ratio, cij/cj which in tum means a decrease with distance rjj (say), since cjlrij)
is an increasing function. This must result in a loss of output, i.e. of research
achievement.
REFERENCES
Andersson A., Persson O. (1993) Networking Scientists. Annals ofRegional Science 27:11-21
Beckmann, M. (1994) On Knowledge Networks in Science: Collaboration Among Equals.
Annals of Regional Science. 28:233-42
Beckmann, M., Persson O. (1996) Scientific Collaboration as Spatial Interaction. Institute for
Futures Studies, Stockholm
Crane, D. (1972) Invisible Colleges. University of Chicago Press, Chicago
Katz, 1.S. (1994) Geographical Proximity and Scientific Collaboration. Scientometrics, Vol.
31(1): 31-43
Persson, O. (1991) Regional Collaboration in Science. Data on Nordic Coauthorships 1988-
1990. CERUM Working Paper 16
Persson, 0., Beckmann O. (1995) Locating the Network oflnteracting Authors in Scientific
Specialties. Scientometrics 33(3): 35\-66
Price, D. de Solla (1963) Little Science. Big SCience, Columbia University Press, New York
Stewart, 1.Q. (1950) The Development of Social Physics. American Journal of Physics, 18(5):
239-53
Chapter 7
Innovation and Urban Planning
Britton Harris
Department of City and Regional Planning, University ofPennsylvania, USA
7.1 INTRODUCTION
The coming twenty-first century will be one of great challenge and also
of danger for humanity. The population of the world may more than double
its present level of over five billion. The pressure of population on resources
will greatly increase the danger of starvation and pestilence. Such dangers
will interact with social unrest and political instability, and all of these
together will add to the dangers of war and of environmental catastrophe.
The scene in which most of this drama will be played out will differ in
many respects from the present world. While now far less than half of the
world population lives in cities, by 2050 far more than half will do so. Urban
populations will increase by a factor of more than five, and it will be difficult
to provide resources for infrastructure, transport, and daily living, with the
result that overcrowding and pollution may continue and increase.
Some hope may however be seen in the intrinsic nature of present
technological growth and change. The two most important sectors are the
biological sciences and the information and computing technologies. Both
sectors are relatively environmentally benign in providing a technology
which does not consume vast amounts of power, nor materials which deplete
the environment, or produce residues which poison it. These technologies
can redirect consumption in environmentally less harmful directions, and
provide at the same time other benefits.
Up until now, the world hope for controlling population growth and
increasing the level of world citizenship has rested mainly on a slow
transitional process of social, economic, demographic, and urban-rural
change. In largely urban societies, birth rates have fallen and levels of
literacy have risen. Exceptional rural cases like the Indian state of Kerala and
13 8 Harris 8.
the nation of Sri Lanka emphasise the importance in this process of

education and literacy, not least for women. It seems possible that the
information revolution may accelerate the increase of literacy everywhere,
and that biological science will not only provide means to improve health,
nutrition, and the quality of life, but will facilitate a rapid reduction in
population growth by finding cheap, safe, and effective means of birth
control which are not susceptible to prohibition by the church, the state, or a
dominant male hierarchy. Such hopes are offset by other dangers, known and
unknown, which we shall not discuss here.
In this rapidly changing and dangerous situation, action based on the
analysis of urbanism may play a decisive role in conserving resources, in
ameliorating conflicts, and in speeding a transition to a more stable and
beneficent organisation of society. The context for this action may be viewed
from two related perspectives. First, we may consider the future of
technology and social organisation and speculatively analyse their impacts
on the future of cities. These impacts will depend not only on the inherent
nature of the technologies, but also on the way in which planning is carried
out to channel and harness their potential. At the same time, however, we
should consider the changing nature of the planning process itself, and
consider reinforcing or redirecting the innovative aspects of those changes.
My intention is to focus on the second of these viewpoints, and to
develop a picture of the complex process of planning, its susceptibility to
invention and innovation, and its need for innovative content. Planning will
itself depend on the progress of technology, and will help to direct its
impacts on cities.
In this context, I would describe urban planning as consisting of three
major components: an analytic or scientific portion; an innovative or
inventive portion; and a public or participatory portion. In my view, these
three aspects of planning are intertwined and inseparable, but due respect
must be accorded to the professionalism of practitioners and researchers who
focus on as little as one or another of them, or on their parts. We will thus
pursue these three ideas separately and together, each being elaborated in
one of the three following sections, with a concluding review of the creative
aspects of invention and innovation.
First, I discuss the innovative nature of the scientific laws and analytic
understanding of cities which have arisen during the past fifty years or so.
The development and general success of these ideas has, however, come to
an impasse based on so-called 'market failure', or more deeply, on
difficulties with optimisation, and this has confirmed the historical view that
'design' is an indispensable component of planning.
Second, I examine the nature of design, in its most innovative aspects,
and suggest in a largely speculative vein the possibility of innovation in
Innovation and Urban Planning 139
methods of innovation. This more modem view of design entails greater

reliance than before on the support of science, and more complete
engagement in the public process and in organised collective activity than is
suggested by the solitary nature of much creative work.
Third the public process itself is redefined in terms of going beyond
public conflicts of interest and their resolution, to public participation in both
the analytic and design aspects of planning. Such a shift poses many
problems in the roles played by analysts, practitioners, and the public alike.
Finally, we consider the nature of innovation and invention in all of these
three areas.
7.2 THE REVOLUTION IN URBAN ANALYSIS
Planning is the premeditation of action. A planner, like any professional,

needs to provide some qualified guarantee of the results of his
recommendations. This thus calls for conditional predictions of the impacts
of plans, and these are made difficult by the extent, diversity, and
interlocking complexity of urban systems.
The general systems nature of most public problems (and many private
ones as well) is amply illustrated by the most pervasive difficulty with
premeditated action: the proliferation of unintended consequences.
Significant actions in social and economic affairs produce a cascade of
effects which are propagated over space, time, and function. When, as in
planning, it is desired to predict or project the consequences of action, then
such actions cannot be considered in isolation, both because any single one
has effects beyond its targets, and because it has synergistic and antagonistic
relations with other actions.
All this was well understood by the most broadly capable of planners and
policy-makers in the first half of this century, and before. But this
understanding resided in the thinking of individuals and small groups, and
was not an established component either of scientific understanding or of
practical policy-making. There have been three major obstacles to acquiring
and organising the necessary knowledge to broaden the scientific and
political understanding of metropolitan problems. First, knowledge is
organised by disciplines, which are inimical to holistic views of life's
problems. Second, action is organised by function and jurisdiction, and this
fragmentation is inimical to the integration and collaboration demanded by
the problems of large systems. Third, an atomistic and competitive view of
life has paralysed a great deal of public action, and in espousing the power of
free markets has derided the necessity for scientific understanding.
140 Harris B.
Fifty years ago, at mid-century, the world was recovering from war, and
there was a widespread atmosphere of optimism and growth, particularly in
the USA, with a neo-enlightenment view that rational action could overcome
many social problems. These and other factors conspired to speed the
development of a new understanding of metropolitan function and its
changes. We will try to review these developments in a way which may
serve at least three purposes: we will show concretely how urban analysis
changed and what potentialities grew out of this change; we will present this
in a way which emphasises the innovative nature of scientific progress itself;
and we will show in the end how the science of urban analysis has defined
some of its own limits.
The presentation will take the form of a roughly chronological review of
the more important episodes in this development.
The computer was probably the most important single factor in assisting
this scientific revolution. The first general-purpose digital computer was
invented during the war and used for ballistic computations. The first
commercially available computer, the UNIV AC, was produced in 1950. The
first major use of computers in urban analysis was in network analysis by the
Chicago Area Transportation Study in about 1958 (CATS 1959).
Auxiliary disciplines included in particular Operations Research, with its
emphasis on man-machine systems, and Computer Science, dealing with
algorithmic methods and optimisation along with other tools. Disciplines
like statistics and survey research were revitalised.
Transportation planning introduced the first large scale metropolitan
analysis. Its new methods were developed in the course of a struggle by the
US Bureau of Public Roads to achieve a methodology with which to plan the
already burgeoning system of Interstate highways and urban expressways in
the US. Some elements of this set of methods (many originating in CATS)
were the following:
The' gravity model', which had been in development and use for about
ten years, became a central feature of many urban analyses; contradicting
conventional economic theory, it proposed and showed that similar
people in similar circumstances behaved dissimilarly (for example, not
everyone travelled to the closest available job). Much earlier, the
sociologist Stouffer (1940) had developed an explanation in the idea of
'intervening opportunities' , which was applied in a new form by
Schneider at CATS.
The 'shortest path through a maze' was an algorithm of utmost
importance to all spatial analysis, solving the problem of realistic
measurement of spatial separation between all pairs of localities in a
metropolis. The method was found by Moore (1957) and based on the
dynamic programming principles of Bellman. Schneider retrieved

Moore's original paper and applied it in Chicago from 1956, using an
IBM 709 computer.
The start of network planning at CATS was based on a little-known
analysis of 'The Optimal Spacing of Expressways', still perhaps a covert
classic. Land uses were modelled using a detailed approach to
decentralisation, on the basis of a model by Hamburg and Creighton
(1959), with no changes for different networks.
Virtually all of the large US urban transportation studies, including
CATS, have been based on a large sample of actual travel behaviour, and
subsequent detailed analysis of the behaviour of households in a given or
variable transport environment. Much of this work was carried out under
the influence of an important book "Traffic - A Function of Land-Use"
by Mitchell and Rapkin (1954).
The high visibility of transportation planning, and the vast scale of its
research efforts energised the whole field of metropolitan locational analysis
in several ways. It showed by example how large-scale urban research could
be mounted. It provided data and methods which facilitated new research.
And by virtue of its shortcomings, real or imagined, it led practitioners and
academicians to undertake novel investigations.
Rough empirical generalisations, like the observed negative exponential
decline of density with distance from the Central Business District (CBD),
inspired statisticians like Colin Clark to prodigies of ex post generalisation,
and others to look for more specific causal explanations. This led to a
prolonged effort to improve understanding of location in general and
housing markets in particular.
Wingo (1961) and Alonso (1964) elaborated a model of the urban

residential land market derived from the earlier agricultural land model of
von Thunen. This model described the behaviour of households in a
monocentric city, trading off travel for space under a budget constraint, and
thus reproducing in theory many familiar observed phenomena. (Alonso'S
model was circulated prior to publication as early as 1959.)
Herbert and Stevens (1960) designed an application of Alonso's model
intended for practical use, i.e. for the Penn Jersey Transportation Study. The
Herbert-Stevens Model used linear programming (a post-war development,
largely computer-based) and tried to overcome the unreality of
monocentricity. Analyses by Harris (1963) and Wheaton (1974) sharpened
the interpretation of this model, but it was never deployed in practice.
By the end of the 1960s, the underlying concepts of Wingo, Alonso, Herbert
and Stevens, and others had been elaborated in various linear programming
142 Harris B.
models by Mills (1972) and by the staff of the NBER modelling project,
among others. These models all tended to suffer from the difficulty that
linear programming does not 'mix people together', but populates each
subarea with a single household type, or usually at the most two types.
The connection between transportation planning and land-use planning went
beyond the transfer of computer methods and data, with an implicit
recognition of interaction, as in Mitchell and Rapkin, and more explicit
connections showing how accessibility shapes land-use (Hansen 1959).
Simultaneously, another modelling trend was developing at some speed.

This trend undertook to reproduce observed patterns of development by
various means, but without assuming any well-defined causal relationships.
Some early examples of this type include EMPIRIC (see Hill 1965), which is
still being applied in policy analysis in a few sites. These did not become the
progenitors of later families of models - perhaps because their structures
were unique and not readily transferable. An exception to this can be found
in the diverse and widespread application of the gravity model.
One somewhat isolated instance is significant. Huff (1964), partly
following Carroll (1955), showed that shopping trips were not always to the
closest centre, but followed a gravity model distribution. This finding dealt a
death-blow to the principle underlying geography's well-established Central
Place Theory which nevertheless, while dead in principle, lingers on in the
practice of geographic analysis. Meanwhile better instruments for analysing
autonomous service location (not enforced by laws or regulations) have
continued to develop. Many types of service location (mainly concerned
with public health and safety) have been analysed in an optimising
framework, which is directly connected with policy-making. In the private
sector, Harris (1964), simultaneously with Lakshmanan and Hansen (1965),
and later in Harris and Wilson (1979) provided a model of retail trade
location under a gravity model of consumer choice, which reproduced
centres of various sizes and frequency by introducing a simple rule of
entrepreneurial behaviour.
The most important and seminal locational study using principally the
gravity model was the famous Lowry Model (1964). Lowry demonstrated
convincingly that a large-scale metropolis could be simulated in a computer,
mimicking the locational choices of households and the growth of retail
centres in 650 different mile-square tracts of the Pittsburgh region. The
simulation took as given a locational pattern for 'basic' industry - including
'export' industry and perhaps producer services. The separations of homes
from work places and shops were not represented by actual travel times or
costs but by airline distances. Trips from work to home and from home to
shop were distributed by various gravity formulations. The capacity of areas
to receive would-be residents or shopping facilities was loosely and

arbitrarily constrained, but there was no representation of improvements and
structures as such.
The Lowry model's simplicity and verisimilitude started a chain of
imitations and attempted improvements. (See Goldner, 1971, for a review of
applications in North America.)
In the US models called PLUM (Projective Land Use Model) and BASS
(Bay Area Simulation Study) adopted Lowry's methods, sometimes using
actual transportation times to measure separation. The location of 'basic
industry' was made at least partially endogenous. In the UK, similarly
modified Lowry Models became an important implement in the analysis
needed for structure planning.
Putman (1976, 1982) in the US, along with Echenique (see Echenique and
Owers 1994) in the UK, recognised and emphasised the reciprocal
interrelationship between transportation conditions and land use change.
They showed how different transport plans affected land-use development
differentially, thus indirectly affecting transport demand as well, and hence
how transport-dependent land-use models were necessary for transportation
planning. Putman's models - PLUM, DRAM, ITLUP, and EMPAL -
follow the modified Lowry model tradition, and are widely used in US
transport planning agencies. Echenique extended the Lowry tradition in
ways to be discussed, and enjoyed a different type of success in Europe and
Latin America.
In the seventies there was a new round of creative activity, followed in the
eighties by a process of mostly incremental improvements and fine tuning.
The seventies started with a strong attack on modelling (Lee 1973), which
based itself on experiments preceding many of those discussed here. Lee
disregarded these later experiments and the self-critical attitudes of the
model developers, thus utterly miscalculating the future of modelling. The
main effect of his paper was to weaken the connection between theory and
practice and delay the professional use of models outside of transport
planning.
Wilson (1970) found a new derivation of the gravity model and initiated a
more complete new style of locational modelling through the use of so-
called 'entropy maximisation', an idea closely linked with thermodynamics,
but yielding new insights into optimisation in location theory. These new
models were also related to the measurement and interpretation of consumer
and producer surpluses, and hence of welfare. In many cases they depended
on a more careful use of constraints than the pure Lowry model, producing
144 Harris B.
(like several other types of mathematical programming) shadow prices or

monetary measurements of locational advantage and disadvantage.
In economics, McFadden (1973) developed a theory of discrete choice
which while addressing a wide variety of situations, also proved to generate
a family of specific types of gravity model. Maximum entropy, gravity, and
discrete choice all depend on our ignorance of many of the influences on
choice. This is explicit in McFadden's development.
Up to this point, the economic theory of Wingo-Alonso-Mills leads to
models distinct from those motivated by the gravity theory of Lowry.
Echenique overcame this gap in his modelling by introducing land and
building rents to enforce constraints in the allocation of scarce and desirable
facilities. Anas (1975, 1987) similarly showed that such constraints and
their shadow prices would lead to an equilibrium pattern of location. A
hybrid model, using linear programming and a limited amount of discrete
choice was developed by the National Bureau of Economic Research
(Ingram et al 1972). Models of these types have produced powerful results
which ultimately merge the best features of the two earlier traditions.
At about this time, the current phase of the expanded practical development
of GIS began. Together with the improvement of the US Census, and its
orientation toward cartographic locational measurement, this trend has
opened up a new prospect for the expanded use of data in analysis. GIS and
urban models are cohabiting, but are not yet married.
An important ultimate bound to the effectiveness of modelling was found,
when a paper in computer science by Cook (1971) identified the existence
of a class of intractable problems related to optimisation which cannot be
solved in practical times. He called these problems NP-complete, showing
how to identify them by direct or indirect reference to a basic problem and
establishing that if one of these problems could be solved in practical terms
(that is, with respect to their size, in polynomial time rather than exponential
time), then all could be so solved, and if not, they couldn't. We must, at the
conclusion of this topical review, examine some of the implications of this
finding.
The idea of NP-completeness was developed in an environment of

optimising (which may be interpreted as finding optimal plans), but so far in
this discussion we have been emphasising the importance of finding
reasonable models which embodied an understanding of process and
behaviours in metropolitan location. We have described a variety of kinds of
choices in making locational decisions, which take into account the costs of
interaction by transportation, and the competition for land or other scarce
locational resources. In models which follow the economic tradition of
Wingo, Alonso, Mills, and others, it may be seen that the locational rules
tend to imply clearing markets for land, labour, goods, and services; also,
network equilibrium is a pseudo-market in congestion avoidance. Under
these market conditions, and in the absence of externalities, the competition
amongst locators and users of facilities tends to produce an optimal
distribution of activity within any subsector, subject to the locational
decisions of all other locators or sectors and subsectors.
As may be seen from the diversity of criteria for location, separating
activities by sector or function has been practical and attractive to the
disciplines of analysis. At the same time, it removes some of the most
important externalities from the individual models and puts them in the
realm of sector interaction. This means however that we cannot
automatically be sure that the joint location of all groups will result in a
system optimum. For obvious reasons, optimising all of location involves
taking account of externalities, economies of scale, and indivisibilities.
These can, more unusually, impeach the optimality of each sectoral model in
its own sphere, since these factors favour the appearance of local optima.
It is exactly problems of this type which generate so-called market
failures, meaning that purely economic rules may lead to solution which may
be good but not necessarily the best possible. These failures arise in cases of
indivisibilities, externalities, economies of scale or agglomeration, public
goods, and monopolies. Planners have realised the existence of exactly these
difficulties for years, and mathematical programmers had been frustrated by
the difficulty of solving intransigent problems like that of the travelling
salesman. The discovery of NP-completeness thus in a sense unified the
common difficulties of economics, planning, and computer science.
The reader should be struck by the fact that nearly every paragraph of the
discussion in this section has contained a capsule description of one or more
creative acts of invention in theory or its application to analysis. It is not
surprising, however, that these discoveries may have led to (or resolved) a
situation in which a major new jump would be required to overcome a
newly-identified major obstacle. In facing the unknown, we are mostly
confined to speculation until the exploration has gained momentum, and it is
this spirit that we will examine the territory which now seems beyond a fully
rational limit to engaging in rational action.
7.3 CROSSING THE BOUNDARY BETWEEN

DESIGN AND ANALYSIS IN PLANNING
Whatever limitations are found to exist for scientific approaches to

achieving perfect plans, there remains a public imperative to improve their
performance - that is, to better them. Even though such improvement may
146 Harris B.
fail to achieve a genuine optimum, we must continue to regard the planning

process as optimum-seeking, and we may examine the barriers to success
and the probable best methods by which they may be overcome. This
examination will lead us into issues of invention and creativity which are
very hard to explore.
In the resulting discussion we do not surrender our attention to the
importance of analysis. A well-ordered process of optimum-seeking requires
an analytic understanding of the impacts of plans so far considered. This will
guide the directions of improvement, inform the planner when a new start is
indicated, and determine the possible limits to further efforts.
However, no matter how complete our understanding of any system, and
our agreement on the goals being pursued, there is frequently an intrinsic
limitation on the possibility of finding a globally optimal solution for the
problems of that system. This occurs when the problem formulation
unavoidably contains conditions which permit the existence of local optima.
These conditions imply non-convexity, a condition which renders simple
optimisation unlikely, and which is very important in the mathematical
formulation and analysis of problems.
For example, suppose that residential locators have strong social
preferences, and that other things being equal they greatly prefer housing
near people of their own class, occupation, or ethnic group. In trying to find
an optimal locational scheme, an initial pattern might be generated using
preferences which are not self-referential, but this pattern would have to be
explored and adjusted taking social preferences into account. Then a new
pattern would arise, with physical as well as social aspects, such as density
and dwelling-unit costs. Once such a pattern was established it would tend to
be self-perpetuating, if social preferences are sufficiently strong. This indeed
seems to be the case for some residential groups not only in American cities
but throughout the world, although alternative explanations are available and
may prevail.
The tendency of like to seek like thus leads to a kind of community
indivisibility, which is further intensified with suburban self-rule, leading to
the Thiebout phenomenon of local standards being imposed in the manner of
the formation of clubs. This indivisibility appears in a different guise in the
quadratic assignment model of Koopmans and Beckmann (1957), where the
great difficulty of finding a guaranteed optimal layout of factories or cities is
implicitly demonstrated. It is shown that in general finding a globally
optimal pattern for a large number of locators whose interaction costs
dominate their site costs is extremely intractable. Even with the simplest
possible objective function which captures the costs and possibly some
benefits of interaction induced by separation, small problems (with, say, fifty
activities to be located) can require hours of computation for the complete
enumeration of possibilities which would be necessary to find a global

optimum. Approximate methods can be used, making step-wise
improvements from random starts, but the costs of computation remain
large, and there is no guarantee of their complete effectiveness.
Superficially, the difficulty of the quadratic assignment problem seems to
depend on the definition of indivisible activities which compete for sites, but
in actuality the same problem arises with atomistic activities which desire to
avoid externalities or have strong economies of agglomeration. The idea of
indivisibility obviously applies for example to planned shopping centres, but
it may be seen to persist in many urban configurations which grew naturally.
Either case has a certain durability and persistence which results from its
coherence and, as a result, it also follows that changes, if they occur at all,
can involve the catastrophic collapse of at least one centre.
In brief, the intractable problems in optimising location arise from the
presence of externalities, indivisibilities, economies of scale and
agglomeration, decreasing costs, and the general interdependence of
decisions. These as we have seen are the difficulties which lead to problems
characterised in economics as cases of market failure, and in computer
science as NP-complete. A precise statement of this difficulty is not needed
this discussion, but we can identify at least three major problems of urban
planning which have been shown to fall into this class, and which partake of
some or all of its observed difficulties.
The practice of zoning and of locating significant large facilities is
intended to facilitate (and hopefully to optimise) interaction while preserving
the benefits of agglomeration and preventing the adverse effects of
externalities or spillovers among diverse activities. This is strictly analogous
with an enlarged problem in quadratic assignment, which even in its simplest
form has been proved NP-complete.
In my view, this is the most essential activity in planning, and its
successful execution by exact optimising methods is possible only on the
margins of urban development, if then. This limitation undermines the
significance of optimising and emphasises the importance of planning and
large scale design in providing good development guidelines over long
periods of time. These methods must overcome the failure of market
solutions to planning problems, which are beyond both the horizons of time
and scale, and the means, of private investors.
Two other major problems are planning the provision of transportation
facilities, or network design, and the efficient distribution of supply or
service centres. Both of these problems are NP-complete, although of
slightly less difficulty than those above. Every one of the three depends on
the solution to the others, that is, on the outcomes of all locational decisions,
both public and private. The solution of one problem can influence the
148 Harris B.
outcome of the solution of another, and may then invalidate the first result.
While a rejection of optimising can be fully supported on the grounds that it
is unthinkable to try to find truly optimal solutions to these three problems
jointly, there are still other very important considerations which militate
against formal optimisation ..
The first of these depends on the nature of the planning process itself.
Optimising methods require a full knowledge of the possible decisions: this
is the policy space. Characteristically planners (and designers in general)
tend, in the very process of planning, to invent new measures or actions
which might be taken. Any such invention, if effective, would require a
complete reconsideration of all optimising work done previously. If on the
other hand, optimising commitments were permitted to rule out the invention
of actions, and to enforce their banishment to an earlier stage, much of the
possible gains from invention in planning would be given up.
A second, and overriding, consideration is the fact that most optimising
processes have limited sets of criteria, and that these do not match the
breadth of the criteria which will emerge in discussions in the process of
public choice. It is widely recognised that there are mUltiple bases of choice
which are hard to combine in a single model, and this gives rise to methods
of multi-criterion decision-making, but this is beside the point. We contend
that if several local optima are collected from an optimising process, the one
which is formally the best will often not be judged the best by the users of
the process, because they will use a larger and partly inaccessible set of
criteria and change the ordinal ranking of the cases based on the formal
criteria.
This line of thought underscores the fundamental reason for preserving
less than optimal solutions to a problem and providing both planning staffs
and the decision-making public (or its representatives) with alternatives. The
means by which alternatives can be encountered are not immediately
obvious, and the locus of creative actions in the planning process is obscure,
so we need to discuss innovative planning in the light of both existing
practice and optimising procedures. We will see that in a qualified sense,
much professional practice contains the same elements as formal optimising.
Generally speaking, both creativity and optimising are combinatorial
processes. They require the discovery and selection of good or superior
combinations of decisions from a very large (and in a practical sense,
unbounded) universe of possible decision sets. Since most of these sets are
vastly inferior to a thin layer of those with superior performance, the
identification and selection of good sets is important and may be non-trivial.
We start with the somewhat arbitrary distinction between inventing
elements of a plan, and finding or inventing good combinations of elements.
The elements of a plan may be physical constructs or social measures, and in
some basic sense they are themselves combinations of even more elementary
components. Examples which come readily to mind are: the skyscraper
(which in tum depended on the invention of structural forms and of the
elevator or lift); the limited access, divided, multilane highway - and its
extension into a system of such highways; and Federal mortgage guarantees
in the US, which incubated the post-war suburbs. I consider a particularly
charming example to be the decision taken in New York, San Francisco,
Prince Edward Island, and a few other places with toll water crossings of
suitable configurations, to collect tolls in only one direction but at a higher
level. This decision has saved many millions of dollars in collection costs
and in eliminating user delays - yet its discovery has no precedents, no
history that I can discover, and no reason why it should not have been
invented much earlier.
Obviously engineering, architecture, planning, and policy abound with
such examples of inventing elements in the process of planning. Indeed, the
same thing occurs in the course of building scientific theories and
experiments, and in writing, painting, and other creative arts. This process
needs to be pursued in planning, but will be done at the expense of orderly
predefined optimising procedures. Most particularly, new measures may
create new difficulties in predicting the impact of plans which embody them,
and this is a problem for planning analysis. Because the invention of new
measures is so obscure, and its optimum-seeking characteristics so ill-
defined, we reluctantly forego further discussion of it at this point.
There is a handful of standard optimum-seeking procedures which when
fully developed can become actual optimising methods. We discuss the most
important of these in relation to planning practice.
The most fundamental optimising procedure, and in many ways the least
creative, is the application of successive improvements to a complete
configuration up to the point where no further improvements are available.
This is called hill-climbing and is a method of finding a peak in
multidimensional space. In discrete space, methods of the calculus (dear to
engineers) may not be available, and there is no analogy to finding Mount
Everest on a clear day by triangulation, because the peaks and their ranges
are not visible.
Successive improvements are often tested by swapping elements (or
locators), and swapping is regarded as a fundamental procedure for discrete
problems. It is the method which, for example, permits linear programming
to proceed inexorably toward a guaranteed optimum. This guarantee depends
on convexity, and planners should know that most of their problems are non-
convex. Thus there is always a danger that effort is being applied to
improving the wrong plan. This calls our attention to the fact that most
improvement methods must start with a trial plan (formally called a feasible
150 Harris B.
solution), and in the case of non-convexity, many such starting plans might
be used. The selection of these starting points is crucial, since a blind and
random selection may waste resources by testing many starts whose
outcomes are all alike, while a better-informed set of starts, designed by a
knowledgeable planner, may suffer from some bias which eliminates
unusual but possibly strong contenders.
Two points are at issue here. The first relates more to the previous
discussion of science in planning: the use of market clearing models to
approach or achieve equilibrium models in locating many populations,
determining densities, and specifying interaction patterns implicitly makes
many planning decisions, and can provide guiding indicators such as land
prices and congestion for changing them or making new ones. The original
starting plans and their improvement can thus be conducted with a focus on a
smaller number of higher level decisions.
The second issue is the multiplicity of plans based on different starting
points. In the extreme case, where there is good reason to have confidence in
a single start, the process is greatly simplified. The Chicago Area
Transportation Study began with an analysis of "The Optimal Spacing of
Expressways" which guided an initial plan, and some forty additional
modified plans were generated in succession, largely without any of the
backtracking which would indicate difficulties in the improvement process.
A process like this does not generate alternatives for later examination, since
the problem is treated as if it were convex and only a single start were
required.
Such certainty is often encountered among planners, and may from time
to time be subject to severe criticism. In the CATS case, for example, plans
with alternative levels of mass transportation should probably have been
considered. Their omission reflects the bias of the sponsors of the study,
which was transmitted to the staff and which generated some conceptual
errors in the analytic procedures. In general, methods which generate
alternatives, or require their introduction, also generate cognitive dissonance
amongst planners and politicians. Despite the repeated mention of
alternatives in planning theory discussions, the issue of their sources is
largely unexamined, as is the startling absence of them in planning practice.
We feel that the creative exploration of alternatives in planning, in a
fashion which avoids the prevalence of failures and promotes the discovery
of configurations which are very different in structure but very similar in
value - is worthy of much study and encouragement.
We finally turn to a cluster of other optimising methods which show how
planners and many others try to avoid alternatives. We shall try briefly to
exam me how these methods might be used to advantage in a creative
process.
These methods might be broadly defined as constructive. In

contradistinction to methods which work with a complete plan, or a feasible
solution, attempting to improve it, constructive methods consider starting
with a tabula rasa, with an existing pattern of development, or with some
combination of these (as in urban redevelopment), and then proceed to take
one decision at a time, until a complete plan has been generated. This
approach can result at one extreme in examining only a single plan (not
necessarily optimal), and at another extreme in the implicit enumeration of
all possible plans, with a certain kind of guarantee of optimality.
The so-called greedy algorithm provides a straight path to the final plan.
This method takes one decision at a time, each one chosen so as to provide
the maximum beneficial effect as measured at the point where the decision is
taken. This is a procedure which is congenial to those, like businessmen and
politicians, with short horizons and a willingness to disregard the interaction
of decisions, and also those with a strong confidence in the prescience of
their step-by-step judgements or with an impatience for alternatives, like
some planners. In rare cases this procedure can produce optimal results: for
instance, in graph theory, the shortest 'spanning tree' connecting a set of
nodes can be found by selecting at each step the shortest unused link which,
when added to the potential tree, does not create a circuit. On the other hand,
adding one firehouse at a time to the system at the most beneficial current
location is certain to produce suboptimal results if the earlier choices cannot
be relocated.
Various versions of dynamic programming provide examples of a greedy
approach in a carefully structured decision space, which do provide optimal
results. A special instance is the shortest path algorithm discussed above,
which rapidly finds optimal solutions to a large number of related problems,
but here the greed is modified by a process of correcting initial decisions if
they are proved incorrect. Some other applications of dynamic programming
are more difficult, and less directly related to greedy, because they depend
on carrying in memory a large number of alternative combinations, which
are gradually reduced in number.
The extreme case of the constructive approach, which is not greedy, is
the procedure known as branch and bound. This undertakes to make one
decision at a time, but evaluates it taking into account some function of the
decisions yet to be made as they are influenced by those already taken.
Combinations whose later consequences are found to be sufficiently adverse
are ruled out, and no further dependent decisions are made. This is done by
finding a 'bound' on the minimum potential cost of a incomplete solution,
and ruling it out if it exceeds a value already found to be possible for a
complete solution. The 'branch' part of the procedure examines branches of
the decision process which have at this point not yet been ruled out, and if
152 Harris B.
none is available, backtracks toward the root of the decision tree until a node
is found with options still available.
We note first that this procedure requires at least one complete plan to
establish a comparative value or cost. This could be the existing plan for a
city similar to the one under study. Later complete but improved plans can
arise successively in the branch and bound procedure, tightening the bounds
at each step. Complete solutions which are later discarded could be
preserved as alternatives, but usually are not. Second, we observe that the
great difficulty with this method is finding ways to estimate the
consequences of incomplete plans. Doing so may be very costly or uncertain
under available estimating methods, while failing to do so leads to an
unconscionable expansion of the number of combinations explored, and the
time and resources expended.
Nevertheless, planners often implicitly follow this procedure in planning.
Its efficiency is improved by placing the decisions with the most
consequences, or the highest net costs, early in the decision process. As
before, all of the decisions should be strategic, with the ultimate aim of
exploring the details through the use of equilibrium models to optimise
subsystems by developing their performance in detail. In ordinary planning,
the decision whether to explore further from a given combination is taken
judgementally, on the basis of intuition rather than complete analysis. This is
the point at which planning procedures can vary most widely when operating
in this mode. Very tight and self-assured methods of judgement will leave
only one option leading out of each node, and the procedure reduces to the
greedy algorithm. Different bounding criteria will lead to different solutions,
even in the greedy case: this is then a new means of generating alternatives.
As we repeatedly notice, strong bases of judgement introduce bias, but weak
ones grossly extend the planning process.
This sketch of the broad relationship between creative but professional
planning and optimising has, I think, demonstrated the importance of further
articulating planning method. Complete optimisation is practically, if not
conceptually, impossible. I have for instance not dealt in detail with the
problem of different interests and value systems, since I feel that this is a
problem not in optimising, but in civil government and conflict resolution. It
affects not the outcomes or consequences of decisions and thus of plans, but
their evaluation and acceptability; accommodating these differences is
another, but subsidiary, reason for preserving alternatives. Given the
practical obstacles to optimisation, we must fall back on a professional but
judgmental process, where the greatest difficulties are presented by bias and
single-mindedness. But there is an additional problem, which is the lack of
imagination and the inability to innovate, which must also be overcome.
Some of the problems in all of these areas will be related to the scope and
style of public participation in the next section, and will be approached more
generally in the last section.
7.4 HOW PLANNING DEPENDS ON THE PUBLIC
It is a truism that while the client of planning may be a commission or a

legislative body, this activity is indirectly in the service of the public. There
is first the public of the citizenry taken as a whole, and the possible
responsibility of the planner to represent the 'public interest', if that can be
identified and if it is not already represented in the planning process and in
the judgements of the client. In addition, there are various publics, consisting
of classes and organised entities which differentiate the broad citizenry, and
give rise to particular interests, which may be overlapping, mutually
reinforcing, or conflicting. The satisfaction, amelioration, and resolution of
these interests and their conflicts is a problem for both planning and
governance.
It is not my intention in this section to deal in any detail with most of
these aspects of planning, but to review some special features of public
participation in the planning process taken as an effort at a rational collective
exploration of spatial problems, and at creative forward-looking problem
solving.
Limited experience has shown that quite aside from observing the
principle of democratic participation, the direct participation of the public(s)
in planning has the potential of reducing conflict, increasing consensus, and
improving the quality of plans. The structure of conventional planning does
not always facilitate this participation. Governing bodies (one or more of
which may be viewed as the client) usually reserve the right of final
decision, which they feel selected to exercise, and this throws the results of
planning into a political process, which may foster the influence of special
interests and new but hidden agendas.
Planners and planning agencies, on the other hand, may regard
themselves as technical experts, whose planning judgement is superior to
that ofthe client and the publics. They may also assume the responsibility of
representing a limited set of otherwise neglected public interests. In this
case, the public's admission into the planning process, other than through its
elected officials, may be limited to giving voice to private interests and
perhaps to providing some local knowledge.
Such a process is therefore participatory only in a minimal sense. In a
more developed process, it has been found that public groups meeting
together to establish criteria and examine the process by which evaluations
are made of proposed plans reaches new levels of understanding and
154 Harris B.
surprisingly wide agreement among diverse interests. One study of an

enormous irrigation and water supply project which would impact the San
Francisco Bay Region illustrates these findings. After study, a representative
local group decided unanimously to adopt the very sophisticated criterion of
maintaining biodiversity. It then followed and criticised constructively a
complicated evaluative model using hydrological and biological simulation.
In the final analysis the project was abandoned, partly as a result of this high
level of participation.
These and many other findings lend support to the sketch which follows
of an expanded public role in planning. I will develop this in terms of a
paradigm for planning based on optimisation, in order to show how different
demands of the process lead to different kinds of participation.
The kernel, but not the whole, of an optimisation scheme lies in the
means which are used to determine the outcomes of planned decisions - a
topic which occupied most of an earlier section of this chapter. The
predictive part of this exploratory process is related to a set of inputs and
outputs. The inputs include the measures which may potentially be taken, the
resources available, and the initial state of the system. These are connected
to the outputs by the behaviour of the system as it is modelled. The outputs
are an array of physical, social, economic, and environmental happenings,
which must provide all the information needed for the publics to be able to
evaluate these outcomes - that is to calculate their impacts on the goals and
objectives held by different groups and members of the public. The active
part of the planning process involves varying the inputs so as to produce
different outcomes, thus seeking an optimum. The inputs are changed by
inventing new measures, and by postulating new combinations of measures
in order to find the most successful plan to satisfy the anticipated needs of
the public and its parts.
Any discussion confined to inputs and outputs, however, neglects
consideration of the intervening model which translates the first into the
second. It has been suggested by various students of participatory planning
that it is essential for the participants to 'buy into' the analytical basis on
which proposals are tested - that is, to understand and accept the models
which are used. This is difficult with complex models, and involves breaking
down the barriers between scientists or professionals and the public. The
scientists must give up their claims to complete expertise, and the public
participants must surrender the belief that a vigorous expression of their
interests is an adequate substitute for wider understanding and informed
exchange of views. All this was evident, for example, in the Bay Area case,
where the science was partly removed from the daily experience of the
participants. This provided a cloak of neutrality for the scientists which is
more difficult to establish in the case of urban planning.
I think we must recognise in the case of urban models, that various

publics ought to have the right to practice what might be called 'public
science' - that is, to urge the application of their knowledge of the actual
facts and the potential relationships and interactions of urban life in building
the theories and models that are used in the study of plans. This view can in
fact be extended in a general way to many of the findings of social science,
where academics may fail to appreciate the texture and feel of real life, and
could benefit by being more 'anthropological' both in studying the
experience of people in their societal environment, and in seeing societies
whole rather than chopped into disciplinary categories like economics,
sociology, and political science.
All of these ideas may be extended to the other aspects of planning, and
this extension will be sketched in a moment. At this juncture, it is opportune
however to interject the comment that all of these changes will be stressful to
all of the participants - to the scientists, the planners, and the publics.
'Openness' will have a new definition, which will include access to
scientific and planning data and tools, the use of public computational
facilities, together with (on both sides) a willingness to learn, and a decent
respect for the opinions and attitudes of others. These are not characteristic
of bureaucracies or of the contending parties in public disputes.
To sum up briefly, we list activities which establish a framework for
planning, in which an organised effort can elicit public assistance and
participation:
The statement of public values or goals, which is often placed first in

describing a planning effort, is not a static phenomenon; it changes with the
elaboration of plans and their implications, with experience in the planning
process, and in exchanges with other people and groups with other
experiences, goals, and interests.
The assessment of the initial state of the system for which planning is being
done is often deficient owing to limited data collection, and even more
owing to limited perspectives - which the public can correct.
The specification of constraints and resources is also apt to be deficient in
more than one direction, and can be revised from the knowledge of the
public.
The relationships and causal chains in models also involve costs, benefits,
and constraints, and the exploration of all of these by planners and social
scientists, in collaboration with the public, provides a great deal of
potentially useful and important knowledge.
This knowledge must, in practice, be organised into models; public
participation in the design of models and their implementation in computing
is a topic of special importance, and will be taken up below.
156 Harris B.
Very powerful contributions by the public to planning can be made in its

operational aspects. The public can contribute greatly through:
Inventing new measures, actions, and devices to improve the functioning of

the metropolitan system. Often these inventions must be accompanied by an
original view of how they will function, and what their effects will be in a
larger context.
Proposing alternative sets of decisions, and alternative actions on them, in
the context of the optimum-seeking processes described in the last section.
In the context of planning, these last two operating functions require the
most creative capacity and lead to the most intensive computing. I like to
think of widespread public planning with citizen participation as a kind of
parallel creativity, with people and organisations providing millions of
parallel processors by analogy with parallel computation. Here again, the
real problem down the road is the programming and co-ordination of the
processors, so as to make their contributions mutually reinforcing and
effective, and not just a jumble of diverse but undirected activity.
In looking toward this future there are many aspects, attitudes, and
practices which need to be reviewed, rethought, and reorganised before the
whole process can be made coherent and effective.
The analytic community needs to make its models more open and
understandable to the users and the public. (I avoid the use of the term
'transparent', which actually means 'invisible when looked through'. The
surface of a model should be transparent, but the interior should not; this
reverses the usual practice, in which the surface hides the working parts.)
Such understanding is usually quite possible at the level of concepts,
provided the terminology is demystified - as in translating 'negative price
elasticity' into 'increasing reluctance to pay higher prices for a given
commodity'. However, not all these terms survive their translation into an
operating model, and the validity of the implementation is a matter of quite
proper concern to both professional and lay members of the using
community. There is some urgency in pursuing the development of a set of
protocols for such validation of models in a non-superficial way - that is, by
content rather than by immediate results.
The analytic process, once agreed upon, needs to be greatly opened and
speeded up, and to some degree simplified. If we imagine a member of the
public or groups of the public and clients at a planning discussion, we ought
to be able also to imagine that a new proposal could be analysed with the use
of a computer model in a few minutes, and surely in less than an hour.
Unfortunately, what tends to happen is that either proposals of alternatives
are analysed by planners off the top of the head, or the process of interaction
is drawn out interminably, and interest in it falls toward zero or is turned into
antagonism. Instant expertise is not a characteristic of planning
bureaucracies, and failed negotiations are a not always desired end of the
public process!
In fact, the availability of rapid means of analysis could be beneficial to
the planning process even within agencies. Currently, the tedium of
elaborating and evaluating alternatives leads to an unseemly rush to
judgement, and to the premature rejection of novel ideas. An imaginative
and agile planning process, with active staff participation, requires adequate
and neutral means of exploring new suggestions, rather than authoritarian or
uninformed ends to discussion.
We have already mentioned the need for public access to data, models,
and computing facilities. With the development of processes of the kind
envisaged, it is apparent that public advocacy will often require expert
technical support. Such support has already been used in civil suits against
certain practices in transportation planning in the US. Over the long haul,
such analytic support for public advocacy will become a fixture of the
profession, will be established internally in large public organisations, and
will become an acquired skill among some members of these organisations.
It is hardly necessary to add that major accommodations will be required
by both the analytic and the planning professions - within their practices and
home agencies, in their professional contacts with each other, and in their
relations with the public. More respect will have to be offered to others in
place of professional assurances and disciplinary hauteur. And new activities
will have to be invented and undertaken, such as the public explanation and
justification of modelling, the preservation of planning alternatives, and the
activities of public education and advocacy in general.
It may by now be apparent that there are many different types and levels
of innovation in planning - in the analytic process, in the development of
plans, and in securing the participation, contribution and trust of the public.
In our concluding section, therefore, we will take a very brief look at the
attitudes and processes which tend to support these changes.
7.5 INNOVATION AND INVENTION IN URBAN

AFFAIRS
Invention is generally conceived to involve the origination of new things,

while innovation is the activity of adopting such things in practice. We will
not try to keep too strictly to this distinction, because in many cases it is hard
to make, and in others the requirements and motives of the innovator have
158 Harris B.
some of the nature of invention. We will, rather, look at the components and
sources of invention.
It is sometimes said of invention, and especially in well-trodden fields
like urbanism, that there is 'nothing new under the sun'. We will avoid
debate by conceding that in general there are no new things, but that there
are new combinations of old things. (A safety-pin is no more than a bent
piece of wire.) Finding new combinations, subject to constraints, is not easy;
as we have seen above, a major aspect of design is its combinatorial nature,
and th~s design is a part, or form, of invention. Science and technology
constitute another part, and without analysing them we will note that they
may impinge on urban innovation from outside of planning, but not without
the invention of new modes of organisation and action in the city, without
which they would not become integrated into its life.
In principle, we could invent new combinations of enormous variety, but
the constraints on innovation tend to curb the exuberance of such
inventiveness. These constraints are of the general nature of workability,
usefulness, affordability, a benign nature, moral acceptability, and beauty (or
more aptly and practically, non-ugliness). Some constraints, like legality,
may be susceptible to change outside the process of planning and design.
These overlapping and sometimes conflicting constraints are indicative only;
the process of planning takes them, and others, into account directly and
indirectly. Insofar as planning and design are purposeful (and this is most
often the case - but that is a different paper!), they accord with the maxim
that necessity is the mother of invention. We may be looking for the father.
In complex systems like cities, the problem does not invent the solution,
any more than in medicine the disease invents the cure, or even in difficult
cases, the diagnosis. The necessity which is thought to provoke invention at
the most makes only a suggestion, and most often this is - at least in its
simplest form - wrong. This is because all inventions and innovations have
indirect effects which may defeat the original purpose. This is particularly
true in the reverse case, where invention is the mother of necessity - that is,
where something is desired and adopted because it is new and fulfils a need
only just now revealed, and with effects only partially explored. This
imperfection of untried novelty is a principal source of conservatism in daily
life, and of the distrust of major innovations and technological fixes.
Borrowing often takes the place of invention, in whole or in part. Cities
can borrow technology from other fields, and can do so creatively and
imaginatively. Borrowing from the experience of other nations and other
cities is particularly apt, but is subject to the need for adaptation to variations
in history, resources, terrain, climate, culture, and so on. This kind of
borrowing can be especially useful because only a few cities are at the
cutting edge of urban development, and hence most other cities can still
learn from them. This option is closed if development is too rapid, or if the
leading cities are taken as an object lesson in an undesired future.
We have already seen how analysis and design can borrow extensively
from the social sciences, and from the 'managerial sciences' such as
operations research and computing. Here the transfer of knowledge and
practice is two-edged. Some of these disciplines are privileged in the realm
of invention because they are a step removed from practice and not
constrained in the ways we have discussed. Their inventions may therefore
be more advanced than others which are so to speak home-cooked, but they
may not be suitable for innovation in urban analysis, planning, or function.
Finally, invention which is not purely inspirational (and such invention
may not exist) probably depends most extensively on an extended form of
borrowing - namely metaphor. A more pedestrian form of metaphor is
analogy, and an even more pedestrian form of analogy is simulation. All
three of these might be thought of as models, either of the urban system or of
the planning process, but there is a difference of major importance. Models
are judged on the basis of their realism or accuracy, while metaphors are
judged by their suggestive power; that is, metaphors have enough realism to
sustain a connection, and enough novelty to suggest that our view of reality
might be improved.
The most powerful metaphors are those which inject the most novelty
into our views of a phenomenon, without losing a sense of realistic fitness.
The best inventors probably have the widest knowledge of distant things of
all kinds, and a distinctive ability to make the connections, see the relevance,
and resolve the dissonance by redefining the nearer objective of the
metaphor. Metaphors have to do with bringing together disparate forms, or
discovering hidden but congruent forms, in the widest sense of the word
'form'. There are underlying metaphors in writing, painting, architectural
design, and scientific and technological innovation. Often, the borrowing by
one field from another is largely metaphorical.
An important source of metaphor in any activity which involves
systematic patterns, or their denial, is mathematics. 'Pure' mathematics is a
cornucopia of systematic expressions of form; many of these are not
invented for any practical purpose, and may lie unused for many decades.
But most of these forms ultimately find some application in the real world,
starting as a metaphor and evolving into a theory which has to be tested.
Conversely, many real world findings stimulate mathematical invention,
because they are uncomfortable without a metaphor supporting their
systematic understanding.
Dealing in urban form, and in the formal statement of urban relations,
metaphors from mathematics must be treated with both enthusiasm and
caution. They have an intrinsic regularity and beauty which supports their
160 Harris B.
potential power, but they may not fit the relational purposes of a metaphor.
Since no systematic representation of complex system can be mounted
without computer support and mathematical statements of its inner
relationships, this dilemma is important to planning analysts and planners in
everything they undertake.
This gallop through the nature of creative analysis and design has left us
with a daunting sketch of the knowledge and skills which must be brought to
these activities. Let us conclude by listing them:
Knowledge of the science and technology presently available and in the

wings, together with some understanding of its impacts on urban relations.
Knowledge of the history and status of many different urban complexes
throughout the world, and of the planning measures, economic and
geographic forces, and social relations and cultural preferences which have
shaped their development.
An ability to grasp the nature of urban problems in the context of available
resources, desired ends, and political will - without foreclosing or
neglecting alternative solutions in favour of previously tried or
institutionally favoured ones.
A deep understanding of the data and computational means available now,
or in the foreseeable future, for the analysis of urban regions and of the
effects of planning actions on them.
An understanding of the strengths and limitations of optimum-seeking
methods, in practical detail, and their application or replacement in the
planning process.
An experimental and inventive cast of mind, and an ability to deploy it in
practice and in organising a planning effort, and using the knowledge of a
situation, of goals, and of means of plann ing and acting, in a creative way.
This list is indicative and not exhaustive. Its scope is beyond the
complete grasp of most if not all individuals, and compels us to recognise
once again that planning and analysis are collaborative activities - as well as
admitting above all that the public belongs in that collaboration as a full
partner.
REFERENCES
Alonso, W. (1964) Location and Land Use. Harvard University Press, Cambridge Mass
Anas, A. (1975) The Empirical Calibration and Testing ofa Simulation Model of Residential
Location. Environment and Planning A 7: 899-920
Anas, A. (1987) Modeling in Urban and Regional Economics. Harwood Academic
Publishers, Chur, Switzerland
Carroll, J.D. (1955) Defining Urban Trade Areas. Traffic Quarterly 9: 149-61
CATS (1959) Final Report Volume 1. Chicago Area Transportation Study, Chicago III
Cook, S.A. (1971) The Complexity of Theorem Proving Procedures. Proceedings of3rd ACM
Symposium on the Theory of Computing, ACM, 151-58
Echenique, M., Owers, J., eds. (1994) Research into Practice: The Work of the Martin Center
in Urban and Regional Modeling. Environment and Planning B (Special Issue) 21:5
Goldner, W. (1971) The Lowry Model Heritage. Journal of the American Institute of
Planners 37: 100-10
Hamburg, J.R., Creighton, J.R. (1959) Predicting Chicago's Land Use Pattern. Journal of the
American Institute of Planners 25: 67-72
Hansen, W.G. (1959) How Accessibility Shapes Land Use. Journal of the American Institute
ofPlanners 25: 73-6
Harris, B. (1963) Linear Programming and the Projection of Land Uses. Penn-Jersey Paper
No. 20, Philadelphia
Harris, B. (1964) A Model of Locational Equilibrium for Retail Trade. Institute for Urban
Studies (mimeo), University of Pennsylvania, Philadelphia PA
Harris, B., Wilson, A.G. (1979) Equilibrium Values and Dynamics of Attractiveness Terms in
a Production-Constrained Spatial-Interaction Model. Environment and Planning A, 10:
371-88
Herbert, J. S., Stevens, B. H. (1960) A Model of the Distribution of Residential Activity in
Urban Areas. Journal of Regional Science 2: 21-36
Hill, D.M. (1965) A Growth Model for the Boston Region. Journal of the American Institute
ofPlanners 31: 111-19
Huff, D.L. (1964) Defining and Estimating a Trade Area. Journal ofMarketing 28: 4-38
Ingram, G.K., Kain, J.F., Ginn, J.R. (1972) The Detroit Prototype of the NBER Urban
Simulation Model, National Bureau of Economic Research, New York
Koopmans, T.C., Beckmann, M.J. (1957) Assignment Problems in the Location of Economic
Activities. Econometrica 25: 53-76
Lakshmanan, T.R., Hansen, W.G. (1965) A Retail Market Potential Model. Journal of the
American Institute of Planners, 31: 134-43
Lee, D.B. (1973) Requiem for Large-Scale Models. Journal of the American Institute of
Planners 39: 163-78
Lowry, I.S. (1964) A Model of a Metropolis, RM-4035-RC, The Rand Corporation, Santa
MonicaCA
McFadden, D. (1973) "Conditional Logit Analysis and Quantitative Choice Behavior." In
Frontiers in Econometrics, Zarembka, P., ed, Academic Press, New York, 105-42
Mills, E.S. (1972) Urban Economics, Scott Foresman, Glenview III
Mitchell, R.B., Rapkin, C. (1954) Urban Traffic: a Function of Land Use. Columbia
University Press, New York
Moore, E.F. (1957) The Shortest Path Through a Maze. International Symposium on the
Theory of Switching, Harvard University, Cambridge Mass
Putman, S.H. (1976) The Interrelationships of Transportation Development and Land
Development. U.S. Department of Transportation, Federal Highway Administration,
Washington DC
Putman, S.H. (1982) Integrated Urban Models: Policy Analysis of Transportation and Land
Use. Pion, London
Stouffer, S.H. (1940) Intervening Opportunities: a Theory Relating Mobility and Distance.
American Sociological Review 5: 845-57
Wheaton, W.C. (1974) Linear Programming and Locational Equilibrium: The
Herbert-Stevens Model Revisited. Journal of Urban Economics 1: 278-87
162 Harris B.
Wilson, A.G. (1970) Entropy in Urban and Regional Modeling. Pion, London
Wingo, L. Jr. (1961) Transportation and Urban Land Use. The Johns Hopkins Press,
Baltimore MD
PART II:
INNOVATION AND ITS SPATIAL IMPACTS
Section C:
Economic and Technological Changes

Chapter 8
Product Vintages and Specialisation Dynamics in a

Hierarchical Urban System
Ulla M. Forslund) and Borje Johansson2

IThe Royal Institute o/Technology. Stockholm. Sweden
2Jonkoping International Business School. Jonkoping. Sweden
8.1 INTRODUCTION
8.1.1 Studies in Urban Development and Industrial

Dynamics
This chapter follows on from a series of studies of urban development

carried out in various European and other OECD countries (Johansson
1986a, 1986b; Nijkamp 1990; Cheshire and Gordon 1995a). It has its recent
background in empirical investigations of industrial dynamics undertaken in
a Nordic context (Forslund 1996). These studies examined the systematic
diffusion and relocation of economic activities across a system urban regions
at a fine sector specification. They identified a distinct spatial hierarchy as
regards the reduction or disappearance of those economic activities with
declining markets and shrinking employment. Some of the results may be
compared with earlier contributions in product cycle analysis or filtering-
down models (Erickson 1976; Johansson and Karlsson 1987; Cheshire and
Gordon 1995b).
The findings relating to Sweden and Norway are examined here in a
European perspective. One objective is to provide a framework for analysing
leads and lags in the location dynamics of economic activities with reference
to Europe as whole. Within this framework, models are specified to explain
(i) how the frequency of product innovations (new economic activities) vary
166 Forslund U.M. and Johansson B.
across different types of urban region, and (ii) how these novelties are
diffused and relocated in a system of urban regions. The study also analyses
how an urban region may remain an initiator or early follower, or may lose
its leading role over time.
Three alternative approaches to the explanation of hierarchical
specialisation dynamics are outlined and compared. First, a model of
location dynamics is formulated on the basis of the theory of central place
systems (CPS-model). Second, specialisation theories are examined with a
focus on dynamic location advantages (DLA-model). Third, a type of
product cycle model is introduced with special emphasis on product vintages
(PV -model). Each of these three theoretical models helps to explain specific
aspects of observed location dynamics. The analysis shows under what
conditions predictions from the CPS-model become inconsistent with
predictions from the DLA-model and PV-model.
8.1.2 A Model of Observed Leads and Lags
The model outlined in Forslund and Johansson (1995) and Forslund

(1996) was formulated with the specific objective of generating testable
hypotheses and empirical estimations. This model attempted to classify
economic activities in such way that they referred to product groups and
clusters of products with synchronised location dynamics. For a given
system of urban regions, the model specifies for each type of economic
activity (each product), i, its average share, Si, of all activities in the urban
system (measured in terms of employment or value added). The activity
share of a particular region, r, is denoted by Sir. The model identifies a
specific leading urban region for a given multiregional context. For this
leading region, denoted by K, the relative shares, Pi = SiK / Si, are used as
predictive indicators. Two basic hypotheses are associated with the
innovation leadership of region K. The first states that economic activities
with a high p-value should be expected to grow in other regions. The second
states that sectors with a low concentration in the leading region (with a low
p-value) should be expected to decline in other regions. The first assumption
implies that new activities frequently start their growth in the leading
region(s). The second assumption implies that leading regions lose
employment in mature activities before other regions. Hence, they also lead
in the decline phase.
Certain economic activities are based on natural resources which have to
be extracted or harvested in the region where they are located. The above
lead-lag model obviously does not apply to such activities. However, their
location requirements can be characterised by standard DLA-models, which
are able to predict where, among the potential locations (regions), the
Product Vintages and Specialisation Dynamics 167
resources will be harvested (Moroney and Walker 1966; Smith 1975). For
production whose location is not based on proximity to natural resources, the
DLA-model can easily be transformed into the PV class of models. These
models assume that a high proportion of new products are initiated or
imitated (at an early stage) in a leading region. As production expands, the
products become standardised and techniques are routinised, then relocation
begins to takes place. However, the number of followers is expected to be
limited, since growing economies of scale prevent full decentralisation to
other regions. According to the PV-theory, the location is assumed to be
supply-based also at this follower stage of a product cycle Vernon 1960,
1966; Andersson and Johansson 1984).
The dynamic version of the CPS-model is based on a similar logic to that
of the PV-model, with standardisation and routinisation being the driving
forces. However, in the CPS-theory, density of demand is the governing
principle (Beckmann 1958; Camagni, Diappi and Leonardi 1986). Hence, as
production is started at lower levels in the hierarchy, all regions with a
sufficient demand density are likely to become hosts of the activities which
'filter down'. In the long term, all existing activities may be challenged and
replaced by a new substitute. The decline of the obsolete activity is also
assumed to initiate in the leading region. In this sense, the CPS and PV-
models have the same overall description of long-run substitution.
As a consequence of the assumptions for the dynamic CPS and PV-
models outlined above, a set of sub-hypotheses can be formulated. For
example, activities with both a high ,u-value and fast growth are assumed to
be non-routinised and refer to non-standardised products. They are also
assumed to be R&D and knowledge intensive. Activities with a low ,u-value
are assumed to be routinised and price competing, with a strong tendency to
reduce the labour input coefficient.
8.1.2 Outline of the Chapter
Three complementary (but partly competing) theories of how and why

location intensities differ between urban regions in a multiregional setting
are outlined in Section 8.2. The theories - represented by the CPS-model,
the DLA-model and the PV-model - are used to analyse both static
equilibrium patterns and dynamics of location intensities. According to this
last model, standardised products may have a different location and delivery
pattern from that predicted by the CPS-mode\. However, all three models
help explain the observed regional leads and lags in the economic change
process. In the CPS-model the dynamic process is generated by the declining
importance of 'demand density' over time. In the PV -model the importance
of 'supply density' changes over time and, as a consequence, new location
options arise. Both the CPS-model and PV-model make similar predictions
about where new activities should be expected to appear.
In this section a description is also given of the differences between the
models. Basically, the CPS-model refers to activities which gradually
become more standardised and routinised by reducing average input costs as
well as the set-up costs. The PV-model refers to activities and products for
which routinisation means that knowledge-oriented input coefficients
decrease much faster than other input coefficients.
Section 8.3 shifts the perspective from interregional conditions to
processes and structural features within each urban region. By focusing on
the economic milieu and the associated location attributes of a region, an
analysis is made of the specialisation pattern. In subsection 8.3.2 a model of
sectoral change is introduced to clarify under what conditions the region can
succeed in renewing its economy and under what conditions its activity
pattern ages and leads to decline. There is also a discussion about the
dynamics of a region's location attributes.
In Section 8.4 the leader-follower assumption is introduced and applied
to an empirical analysis of changes in a system of urban regions in Sweden
and Norway consisting of municipalities and groups of municipalities. The
empirical investigation focuses on the spatial concentration of declining
product vintages. In Section 8.5 the perspective is reversed. Here the
knowledge intensity of expanding product groups is examined. In particular,
the location attributes are characterised with regard to initiating regions, as
well as early and late followers. In Section 8.6 a set of conclusions and
conjectures is presented about the leader-follower model in a Europe-wide
context.
8.2 MULTIREGIONAL THEORIES OF LOCATION

INTENSITIES
8.2.1 Location Theory and Central Place Systems
This subsection introduces a hierarchical urban system based on the

central place theory as formulated in Beckmann (1958) and Beckmann and
McPherson (1970). In this setting, each city has a rank representing its
market size: the rank determines which types of commodity will be supplied
by the city itself and which commodities the city has to import. For some
countries (like Germany or Italy) it may be a non-trivial task to determine
which region should be attributed the highest rank. However, in a multi-
country setting such problems tend to resolve themselves.
A central-place systems model (CPS-model) classifies all urban regions

(cities) such that each city belongs to a rank level k. The market size of cities
belonging to category k is denoted by M k The system is characterised by its
ordering of market sizes such that MK > ... > Mk > M\, where MK refers to the
city with the highest rank and MI all regions of the lowest rank.
Commodities (goods and services) of the highest rank are supplied only
from the city of category K and distributed over the urban system.
Commodities of category k+ 1 are supplied from several centres of rank k+ 1
to subsystems of centres on lower levels. Some of these features have been
studied in detail elsewhere. Here we shall focus on how commodities (or
product groups) may shift position over time, thereby reshaping the
specialisation pattern (see Camagni, Diappi and Leonardi 1985).
Beckmann (1996) presents a CPS-model with equilibrium properties,
such that the market size of a city attracts location of activities which in turn
attracts population and determines the city size. Moreover, the market size
for all ranks of the central place system is determined under complete
interdependency (see also Bos 1965 and Tinbergen 1967). Following
Beckmann, a central place of rank k+ 1 serves its internal market of size,
Vk+I-Vk+Mk, where v i denotes the population size of a city with rank i = k,
k+ 1. In addition, the same centre delivers commodities of category k+ J to
mk+1 central places of rank k. Thus, for a city of rank k+ 1, the market size
can be specified as:
(8.1)
where d k+1 signifies the distance from a city of rank k+ 1 to its demand-
associated cities of rank k, A the friction parameter referring to deliveries
from supply sites of rank k+ 1 to markets of rank k, and where A may have a
specific value for each commodity.
Given the urban system described above, a feasible location for a market-
oriented firm (which supplies a given type of commodity) depends on the
cost function which applies at a given point in time for the production of this
commodity. To describe this, let F denote the fixed costs, including the non-
variable part of development costs. Consider also the variable costs which
may here be approximated by a proportional cost component cQ, where c is
a fixed coefficient and Q denotes output or supply. This yields:
C=F+cQ (8.2)
Given this cost function a firm supplying the commodity in question can
locate in a central place k, if it can break even there. For a given price p this
is equivalent to the following condition:
(p-c)f(p)M k -F~O (8.3)
Next, let p* denote the price which maximises the aggregate profit
function:
(8.4)
which describes the maximum profit a single supplier can obtain. With n
suppliers, the function changes to trk (p) = (p - c)f(p)Mk / n - F. Given
the profit condition in (8.4) one can specify a set of industries or supply
activities which locate in a central place of rank k. This set consists of all
sectors (commodity groups) j such that:
This type of set will form a hierarchy which reflects the city-size
hierarchy.
We shall study the dynamics of this hierarchy by introducing a technical
change process. The change of production techniques in an industry is
described by the technique vintage index r, which is close to 0 for a new
process and approaches 1 as the technique matures. Because of this, the cost
function in (8.2) has to be respecified to read:
C(r) = F(r) + c(r)Q (8.2')
We assume that c( r) decreases and that F( r) does not increase as r

increases. This implies that an industry which supplies a commodity may
initially have a feasible location in central places with k+ 1 as the lowest
rank. As the technique vintage gradually increases, feasible locations can be
found in places of rank k, k-l, k-2, etc. Hence, the location of all supply
activities with the described technical change process will filter down (or
diffuse through) the hierarchy of the urban system.
Remark i: Let us consider the CPS-model as given by (8.1)-(8.4) and
(8.2'). Suppose that rank k is unfeasible with technique r and becomes
feasible with technique r+u. Such a switch requires that
[p - c(r)]f(p)M k < F(r) and [p * -c(r + u)]f(p*)Mk ~ F(r + u), where
c( r + u) < c( r). This may involve a price reduction, p* < p, a
corresponding demand increase,j(p*) > j(p), and a reduction of set-up costs
such that F( r) > F( r+u). Moreover, suppose that the vintage development
also makes distribution costs (or the distance friction) diminish. We should
then expect MkC r+u) > MkC r), which will also stimulate filtering down.
8.2.2 Specialisation Theory and Location Advantages
The CPS-model combined with the dynamic process in the preceding

subsection is one of the cornerstones of the dynamic theory we shall now
outline. But let us put this model aside for a moment and concentrate on
additional location attributes which may influence the specialisation of urban
regions in a hierarchical urban system.
The CPS-model is designed to reflect the conditions of market-oriented
industries. In Beckmann (1996) the CPS-model is assumed to apply to all
production (or supply) activities, except resource-oriented industries such as
mining and the processing of raw materials. Here, we claim that it is possible
to identify other resources, not just natural resources, which remain
'trapped'. In European regions, the composition of the labour force (as
regards competence and knowledge intensity) adjusts on a slow time scale.
In other words, the spatial relocation (migration) of employment categories
can be assumed to respond very slowly to regional wage differentials and to
excess demand and excess supply of specific labour force categories. There
are also other location attributes which are fairly invariant even for extended
time periods. This means that the size of the market is only one of the
fundamental characteristics of an urban region.
In the pure CPS-model the rank size of an urban region determines the
size of market a supplier in this region may expect. However, the size and
economic density of an urban region should also be expected to generate
localisation and agglomeration economies (e.g. Ciccone and Hall 1996).
This is a supply side effect which is reflected by cost or productivity
differentials between regions. Part of this effect is included in the CPS-
model, in which the fixed cost, F / Mk in the production of a specific
commodity falls as Mk increases. In the following, we stress that certain
regions have location advantages as regards knowledge-oriented economic
activities.
Specialisation theories teach us that differences in the economic milieu of
regions provide opportunities for specialisation in accordance with the
particular advantages of each region. This type of analysis was initiated by
Ricardo and was later developed by Heckscher and Ohlin. In the standard
version of this model, an equilibrium implies a pattern such that each region
makes use of its relative advantages - given by the fixed (trapped) resources
which constitute the region's economic milieu. The central-place rank of a
region may be thought of as just one aspect of the economic milieu. Hence,
we cannot be content with just an identification of each region's rank, we
have to consider the entire urban system, which we denote U. Moreover, by
Uk. we signify the set of urban regions belonging to rank k = 1, ... , K. Any
region r which belongs to Uk is denoted by r(k) and the complete set of

regions is defined by U = {U Uk: k = 1, ... , K} .
To start with, we consider the following three categories of specialisation
phenomena: (i) agglomeration economies, (ii) localisation economies, and
(iii) other specific resource advantages. Agglomeration economies can be
described as a supply side version of the CPS-model (Vernon 1960). Four
specific assumptions can be made about the location advantages of an
agglomeration. First, an agglomeration offers a robust labour market with a
diversified set of specialised skills and competence profiles. Second, the
same urban region can provide a diversity of non-traded (public) inputs
which affect the productivity or cost level of firms in the region. Third, the
agglomeration has a comparatively large and differentiated supply' of
producer services. Fourth, the agglomeration is characterised by
intraregional information flows about new production techniques, products,
suppliers and customers (see Johansson and Wigren 1996).
The above arguments refer back to Marshall (1920). He used them to
explain localisation economies, but one can observe the same type of effects
as with agglomeration economies. However, localisation economies are
effects which are specific to a particular industry, i.e. to economic activities
centred around a given product group.
Two urban regions with the same rank may differ considerably with
regard to the factors giving rise to agglomeration and localisation
economies. We begin by looking at two conditions which are not directly
determined by the size of the region and are thus partly independent of the
rank:
infrastructure capacity and quality, which provides a prerequisite for

accessibility within a region and also interregional accessibility;
knowledge-intensity of the labour force in the region and R&D-intensity
measured as the capacity of the region to carry out R&D activities. In
regions characterised by localisation economies both these intensities
may be industry-specific.
Let us consider the often discussed city-sequence Amsterdam-Dortrnund-

Koln-Bonn-Mannheim-Karlsruhe-Stuttgart. In this corridor we can identify
subareas with high knowledge intensities as well as R&D intensities, in spite
of the fact that each functional region is of limited size. In northern Italy we
may again observe that in relative terms Bologna and Venice have more of
certain knowledge resources than the larger neighbours Turin and Milan.
However, the following three components of a region's economic milieu
are definitely strongly correlated with the size ofthe individual region:
the customer-intensity, i.e. the number and diversity of customers which

are accessible to a supplier in the region. In the case of localisation
economies, a supplier has accessibility to many but not diversified
customers (consisting of the localised industries);
the supplier-intensity, i.e. a firm's accessibility to other firms which can
offer a variety of specialised input deliveries - often under competition;
the import-intensity, i.e. the richness of product variants, technical
solutions and ideas that influence an urban region in the form of import
flows from the rest of the world.
In small countries like Austria, Finland, Norway, Denmark or Sweden, it

is easy to conclude that their capital cities will have both the highest rank in
the CPS-hierarchy and the highest intensities with regard to the above list.
For very large metropolitan regions like London or Paris, the interest is
instead directed towards the location pattern within the region, including
satellite regions like Reading outside London (Cheshire and Gordon 1995b).
In order to assess the importance of the above factors, the following
average cost function is introduced:
, = F, / x, + c,
(8.5)
c, = p,a + m,b
where for a certain product, , denotes total cost per unit output, c, the
associated variable cost, F, the fixed cost which may vary between locations,
x, the production or supply of the product in region r. Moreover, (8.5)
specifies two input coefficients where a signifies the necessary input of
routine resources and b the necessary input of development and knowledge
activities. The two price variables p, and m, are location specific.
Let us first consider a case where the product is standardised and
produced with a routinised technique. Region r will then have an advantage
if routine resources, including labour with standard skills, are available at
low prices, i.e. when p, is low. Moreover, the supply of land and suitable
infrastructure may be such that F, is smaller in region r than in alternative
locations.
We consider next a case where the product is non-standardised and still
in a development phase. It may be a new product or a product for which each
delivery is customised. Then F, is assumed to decrease as a region acquires
higher customer, supplier and import intensity. In particular, the average
price of knowledge resources is assumed to be lower in regions with a high
R&D and knowledge intensity as well as accessibility to specialised,
development-oriented suppliers. In order to remain development-oriented, a
region must be able to keep a balance between the diversity of the economic
milieu and its specialisation on few activities. The European urban regions
which in the period 1960-1980 were specialised in shipbuilding, all seem to
have lost the quality of balance in their economic milieu during this period,
although some may have found new specialisations later. Another strong
impression is that automobile specialisation, as observed in Wolfsburg,
Goteborg or Turin, tends to bring about a similar impoverishment of the
economic milieu.
The above examples may serve to illustrate location advantages which
are based on the economic milieu and which may add to or subtract from the
location properties which follow from the size of an urban region. In
particular, the economic milieu is influenced by the composition of activities
in the region. Hence, we should expect the dynamics of location advantages
to be non-linear, with both increasing returns to scale and increasing returns
to scope. Because of this, a region with a lower rank sometimes manages to
develop into a system-wide supplier in a specialised field.
8.2.3 Product Vintages and Location Dynamics
Let a product be identified by its product group i and its vintage index
0~~~1, where a low B-value indicates either a young product or an early
vintage, and a high B-value indicates an late vintage. The former refers to
non-standardised products which require large shares of knowledge or
development resources in the production process. High T-values therefore
refer to routinised (and usually automated) production and interaction
activities, whereas low T-values represent non-routinised activities. Late
product vintages are assumed to use smaller shares of knowledge resources.
With these input categories the cost function in (8.2) can be extended as in
(8.6), which is an alternative specification of the cost function in (8.5). When
focusing on a particular product group, we can suppress the associated
index, i, and consider the following cost function which refers to product
e,
vintage which is based on technique T, and which is applied in a rank-k
region r = r(k):
, (T, e) = F, (T, e)/ x, (T) + c, (T, e)

(8.6)
C r (T, e) = Pra( T, e) + OJrb( T, e)
where ,( T,B) denotes total cost per unit output, C,( T,B) is the associated
variable cost, F,(T,B) refers to fixed cost associated with technique T, a(T,B)
represents the input coefficient (delivered output per unit) referring to
routine input resources, and b( T,B) represents the input coefficient referring
to development resources or, in other words, input resources for knowledge
activities or knowledge production. The input coefficients are assumed to be
determined by the selected technique and are not dependent on the location.
On the other hand, the two price variables, Pr and OJr, are location specific,
where Pn denotes the price of routine inputs and OJ,. denotes the price of
development, knowledge-intensive resources.
In this subsection we study how the knowledge intensity develops as the
technique-vintage and product-vintage indexes increase. For these analyses
we construct the variable f3( .,8) = b( .,8) f a( .,8) or simply f3 = bfa, which is
called the knowledge-routine ratio.
In order to bring out the essential aspect of ageing product vintages, we
that.=
assume that the process of technical change is driven by the change in () so
1(8). Following the product cycle model in Johansson and
Andersson (1996), we assume that the routinisation of production increases
as the product becomes more standardised, i.e. as the ()..value gets larger. We
consider two stages of the product cycle development represented by two
pairs of input coefficients (b O, an) and (b I , a I ), with b g =b('f(Og and
d =a( 'f(~, where g = 0,1. Let the initial technique (b O, an) refer to a
young product and the second to a more mature one so that:
(8.7)
In Remark 2 we consider two regions, where the first has rank k and the
second rank h, and k> h. We examine the variable cost, cqg =p q a g + OJ q b g ,
for q = {k, h} and g E {O,l}. The region with the higher rank, k, is assumed to
have lower prices (costs) of knowledge resources and higher prices of
routine resources - in comparison with the level h region. The following
remark is self-evident from the definitions given in (8.6)-(8.7).
Remark 2: Let k > h, Pk > Ph and OJk < OJ h Consider a particular product
which is initially located in a region with rank k with < c2 c2 .
Moreover,
assume that the technique develops as specified in (8.7). Then a shift in
variable cost advantage occurs iff the knowledge-routine ratio reduces in
such a way that:
The described variable cost advantage of rank h regions does not imply
that such regions offer a feasible location. The size of demand must also be
large enough at this level of the urban hierarchy. In order to be a feasible
location, a rank h region, r = r(h), must satisfy the condition.
1r; =(p - c; )f(p) - FfM r > O. This observation leads to Remark 3.
Remark 3: Consider the assumptions in Remark 2 and assume also that the
price shifts to pi < pO when the knowledge-routine ratio shifts from pO to
pl. A rank h region becomes a feasible location if (pI - c! )f(pl) ~ F I M h
Such a change in technique contradicts the rank-size order of the CPS-model
ifinaddition (pl-c!)f(pl)<FIM k fork>h.
The described contradiction may well to occur. It implies that the
commodity is delivered from a region at a lower level to regions at higher
levels in the hierarchy. Observe that the effect of the new technique is (i)
ck >c! and (ii)i</.
The contradiction occurs if ;r! = (pI - c! )f(pl) - F I M k < O. When
such a contradiction occurs it becomes necessary to shift model framework.
In particular, the distribution system cannot be restricted to the hierarchy
given by the CPS-model, we have to consider the entire urban system U.
Because of this, we also have to refer to a network of deliveries from region
r to all other regions, where drs denotes the distance between rand s in this
network. Reformulating (8.1 ???) so that it refers to the introduced network
system, the potential demand for a particular product delivered from region r
can be defined as follows:
Mr =Ls exp{- Adrs ~s (8.8)
where v s represents the size of the market in region s E U.

Now, let ()o refer to a young product and let ()I > ()o refer to the same
product when it has matured with a routinised production. By r = r(k) we
-
r -
refer to a region with a high rank k > k, and by = r( k) a region with a
low rank. We assume that the ranking is such that Pk > PiC and OJ k < OJiC .
The following two conditions show (i) that r is a feasible location for a
young product and not for a mature one, (ii) that r has a reversed feasibility
pattern:
(pO -c~)<Fr()o)/Mr()o)~Fr()o)/Mr()o)po -c~)

(8.9)
(pI _ c~) > Fr()I)1 Mr()I) ~ Fr()I)1 Mr()I) > (pI - c~)
With this formulation there may be a large set of regions at rank k which
are feasible locations for the technique indexed by 1. In order to obtain
predictions about likely location patterns, a probability distribution therefore
has to be superimposed.
8.3 AN INTRA-URBAN MODEL OF LOCATION

DYNAMICS
8.3.1 Location Attributes and Economic Milieu
The previous section examined how the economic milieu in each region
affects the interregional location dynamics. The ambition in the present
section is to show how the economic milieu itself is influenced by the
location process. In this way, Section 8.3 investigates the same process as in
Section 8.2, but with a specification of zones within each individual region.
In this subsection, the five aspects of a region's economic milieu (from
Section 8.2.2) are considered: (i) infrastructure, (ii) knowledge and R&D
intensity, (iii) customer intensity, (iv) supplier intensity and (v) import
intensity. The infrastructure of the urban region as a whole as well as of
individual zones may facilitate the improvement of (ii)-(v). However, the
critical factor is the dynamic pattern of the location of activities within the
region. If the capacity of the built environment is taken up by routinised
activities related to late product vintages, then there is limited space for new
activities. Moreover, in such a situation the customer and supplier intensity
will be oriented towards late product vintages and not early ones. Over time,
such a situation will make the economic milieu gradually less suitable for
innovation activities.
Let us consider the zones in an urban region. The capacity and quality of
a zone is denoted by Z and the value added of the activities in the zone is
denoted by Y. A frequent problem is that the activities which have in the past
had a prosperous period tend to remain in the same zone even when the
value added or profit they can generate continue to fall (due to routinised
techniques and standardised products). In order to assess this type of
dynamic problem the following simple model is introduced:
Y(O)=G(Z,O) (8.10)
where G is a production possibility function describing the size of value

added, Y, which can be generated in a zone with infrastructure value Z and
an average product vintage denoted bye. Consider then the following
assumption.
e
It is assumed that for a given -value, G has a sigmoid form with
iG / 8Z > O. For a given Z-value, it is assumed that late product vintages
generate a smaller value added than early ones. The major reason for this is
that the output price must fall as e increases. Hence iG / t30 < 0, for given
Z. It is also assumed that the G-function is different for different types of

zones.
We next consider two product vintage indices OB < OA , where the first
refers to young products and the second to old (standardised) products with
routinised techniques. Moreover, we consider two types of zones, a central
B-zone and a peripheral A-zone. These are associated with the functions GA
and GB . Let the B-zone represent areas which are suitable locations for
contact intensive activities, including the development and production of
new products, and let the A-zone refer to zones which provide a location
advantage to routine activities. The potential value added may then be
related to the average product vintage indices, OB and OA, and to the
infrastructure index of the two types of zones, as illustrated in Fig. 8.1.
Assume then the each zone has a fixed infrastructure level, signified by z!3
and z4 and consider the following surplus or profit function specified for
each given zone q = A, B:
1C q(0) =Gq(zq ,0)[1- Cq(zq )], for q = A,B (8.11 )
where Cq(Z) signifies the overall production cost per unit value added in a
zone of type q, which is endowed with the infrastructure level Z. Given this
specification, a set of conclusions are specified below. They are based on the
vintage theory assumptions introduced in Section 8.2, together with the
assumptions introduced above in this subsection. They also refer to the
structures depicted in Fig. 8.1. We should note the following:
(i) the B-zone has by assumption an advantage in production with low ()-
values, in the sense that GB(ZB,OBGB(ZB,OA), which implies that
1C B(OB) > 1C B(OA);
(ii) the B-zone is by assumption superior to the A-zone for early product
vintages (low ()) in the sense that GB(ZB,(}BGA(ZA,(}B). If
CB(ZB,OB) is not too high relative to CA(ZA,OB), this yields
1C B(OB1C A(OB) ;
(iii) for certain routine activities one may expect that the A-zone is superior
to the B-zone, in the sense that GA(ZA ,OA) > GB(ZA ,OA). However,
irrespective of this productivity condition, the cost level in the A-zone
means that 1C A(0 A) > 1C B(0 A) ;
(iv) by assumption, the A-zone has an advantage as a site for routine
activities, in the sense that G A (Z A, 0 A ) > G A(Z A ,0 B), which yields
1fA((}A) >1fA((}B).
In Fig. 8.1, z! is optimal for low O-values and zA is optimal for high 0-
values. Hence, for non-routine activities and young products with a low B-
value, the B-zone is superior to the A-zone. The reversed situation obtains
for routine activities and old products with a high B-value. In particular, one
can observe from the figure:
1r B(OB) > 0, and 1r B((}A)::; 0

(8.12)
1r A((}A) > 0, and 1r A((}B)::; 0
8-zone with a G
A-zone with a
lowe lowe
y y
z z
y A-zone with a
B-zone with
high e c y high e
J.-_--G
G
z z
Figure 8.1 Value added and cost in A -zones and B-zones for high
and low B-values
This implies that when the B-value increases for a given production,
profits will fall and become negative if it remains located in a B-zone. A
relocation to an A-zone is then a rewarding alternative. Another option is to
relocate these activities to an A-zone outside the urban region. The major
lesson is that exit of routinised activities from B-zones is necessary in order
to provide new space for the introduction of new and development-oriented
activities (with low B-value).
A basic argument in the above analysis is that the infrastructure of each

zone changes very slowly. Hence the dynamic vitality of each zone requires
that the composition of activities must change continuously. The schematic
description of zones also provides guidelines about infrastructure
investments and formation of the economic milieu. Urban policy can
therefore be seen as the orchestration of change processes in a system of
interlinked zones. In such a self-organised process, one can observe the
development of new B-zones in an urban region. An typical example is
Reading outside London. During the period 1970-1990 there was a growing
concentration of new economic activities with an international orientation, as
reported by Cheshire and Gordon (1995b).
In Johansson (1993) the above observations are included in a model
showing that zones should be expected to follow cyclic paths of
infrastructure renewal and decline. The need for renewal follows from the
fact that zones frequently have a specialised design and an economic milieu
with associated localisation economies. Hence, when more and more of old
activities are replaced by new ones, there is a demand for adjusted location
attributes. In an urban region, the demand for such adjustment may become
strong for many zones simultaneously.
8.3.2 Location Dynamics of Aggregate Clusters
This subsection continues the approach in subsection 8.3.1, but now all
A -resources and all B-resources in an urban region are aggregated.
Moreover, all activities with a low B-values are grouped together and called
J-activities. In the same way, all activities with a high B-value are grouped
together under the label I-activities. Hence, this subsection transforms the
micro approach of Section 8.2 by analysing broad categories of products
which are identified as clusters or groups of products. We assume that
individual but associated product cycles can be expected to develop in
parallel clusters, with a base in the same urban region or in a small set of
interconnected regions. When such a specialisation pattern obtains, we shall
use the term product group specialisation of a region. Such a group may
then be characterised by its average vintage index.
Let I denote routine-oriented, standardised activities, and let J denote
non-routinised, knowledge-oriented activities. The output from activities of
type I is called a type I product, and the output from the second type of
activities a type J product. Moreover, let A and B represent the routine-
oriented resources and the knowledge resources, respectively, in a given
region and let these resources comprise both market resources and other
parts of the economic milieu. We also need two pairs of technical
coefficients associated with resources of type A. These coefficients are

denoted by all' au, and a JI , a JJ , where:
QijAx i shows how the A-resources available for activity} shrink and
aul1x i shows how the A-resources available for activity i shrink as the
activity level Xi increases by Lhj; i = I, J and} = I, J (8.l3)
With an alternative formulation, A - aijxi denotes the amount of A-

resources available to activity} when activities of type i operate at level xi'
Moreover, aij and ajj may be functions whose value vary with x j and Xi
As regards B-resources, analogous coefficients bl/' ~, and hJ' bJJ are

introduced:
byLlx, shows how the B-resources available for activity} shrink and
biiLlxi shows how the B-resources available for activity i shrink as

the activity level x i increases by l1X i ' i = I, J and} = I, J (8.14)
From the specifications in (13)-(14) one may conclude that:

Remark 4: Type J activities can grow only if both [A - au X/ - aJJxJ ] >0
and [B - hu x I - bJJ x J ] > O. The corresponding condition for activities of
type lis [A - aI/xI - aJlx./] > 0 and [B -hI/xI - b./lxJ ] > O.
Let us call the region we have focused on the 'export region'. The
demand in region s for products of type J from the export region is denoted
by D Js , and the corresponding demand for products of type I is denoted by
DIs' Observe that D./s and DIs represent the amount of demand which is
directed towards the export region. Excess demand in the form
[Is Dis - Xi] > 0 is a necessary condition for growth of i = I, J.
In order to make the above model specification complete, one additional
dynamic feature has to be considered. First, we assume that there is a
technique renewal process such that non-routine activities gradually are
transformed to routine activities. In order to illustrate this idea, it is assumed
that there is an ongoing transformation of techniques of type J given by
q./x./, where 0 < qJ < 1 is a coefficient expressing the rate of
transformation. Thus, during the time interval dt, the amount qJx./dt is
assumed to leave the existing type J capacity and to join the type I category,
thereby adding to the size of XI' Using the terminology of Section 8.2,
activities of type J refer to low .-values, while activities of type I refers to

high .-values. Hence, ~JxJ may be thought of as a shift of technique
vintage. Assembling our assumptions, the change process of a given urban
region can be formulated as follows:
(8.15)
where the functions f(x[) and f(x J ) specify the speed of adjustment, given
the size of each type of activity at each point in time. These adjustment
factors will express self-reinforcing adjustments if df / dx > O. In many
modelsfisexpressedas f(x[)=y[x[ and f(xJ)=YJxJ' where Y[ andYJ
are two positive constants (e.g. Fisher and Pry 1971; Batten and Johansson
1989).
The model in (8.15) is designed to be compatible with product cycle
assumptions as expressed in Remarks 1-3. Hence, the condition
b ll / all > b l / ai' must be satisfied, which implies that the knowledge-
routine ratio is higher in non-routine than in routine activities. Moreover, if
equal activities disturb each other less than unequal activities, one obtains
aij ~aii and bij ~bii.
The change process in (8.15) is characterised by competitive exclusion,
and the time profile of both growth and decline has a sigmoid form. It is also
evident from the equations in (8.15) that if x I is not declining, or if the
process is very slow, then the entry of non-routine activities will be delayed
or completely obstructed.
Thus, given that the resources A and B remain unchanged over time, the
output from type J activities can, with sufficient demand, continue to grow if
activities of type I continue to reduce their input consumption of the same
resources. The latter is in this case equivalent with a gradually declining
output of type 1. It should be observed that if such a process is constrained
by the available supply of B-resources, it will generate an excess supply of
A-resources. This follows from the specification of the knowledge-routine
ratios above. Hence, when studying long time periods, expansion and
contraction of the two categories of resources must be contemplated. This
may include investment processes, education of the labour force and

migration of different employment categories into and out ofthe region.
The technique renewal coefficient ~J describes how non-routine
activities change technique and are transformed into routine activities. These
represent product vintages which are more mature than the average vintage
level of type I products. Hence, if ~JxJ > 0, it should be expected to
increase the competitiveness of type I activities. This may be assumed to
affect the export delivery so that it increases or remains on a high level. If
the competitiveness of type I activities improve in this way over a long time
period, type J activities may be forced to decline. It can also be observed
from (8.15) that exit from the region is essential to allow the possibility entry
of novel activities. In particular, XI ~ 0 is compatible with exit of activities
of type I, since there is a gradual transformation of J-activities into 1-
activities, as expressed by ~JxJ > O. This change process is illustrated in
Fig. 8.2.
In summary, the dynamic model provides a framework for an analysis of
the complex dynamics of an urban region's internal structure and its
interplay with the external networks. The complex dynamics may generate a
sequence of economic episodes, in which we find time intervals where type
J activities dominate, and other intervals where type I activities dominate.
Moreover, the model may generate waves of aggregate growth and decline
episodes (e.g. Johansson and Nijkamp 1987).
Aspects of the dynamic process outlined in Fig. 8.2 have been
investigated for the Milan region, the London region, the Randstadt region,
the ZUrich region and the Stockholm region (Cheshire and Gordon 1995a).
In some of these cases, the delayed exit of 'obsolete' activities can be shown
to obstruct the entry of new and expanding activities which make full use of
a rich urban economic milieu. However, there are also cases of delayed
adaptation of the economic milieu (in particular the built environment,
including transportation facilities).
European regions with a history of harbour and shipbuilding can be
considered to belong to a special class. In all of them, these activities have
been drastically reduced over a 10-20 year period. In most cases this has
liberated a large amount of centrally located land which could have been
used for new activities. Often, however, the regions have not managed to
launch sufficient renovation of the urban infrastructure. This corresponds to
an exit process which has not been balanced by a complementary entry
process, at least in the short or medium term. Glasgow, Goteborg, Malmo
and Rotterdam are typical examples of unbalanced development of this kind.
However, they have all made efforts to rejuvenate the regional milieu to
stimulate the introduction of new activities (see Cheshire and Hay 1989).
lS4 Forslund U.M. and Johansson B.
Entry of 'new' activities
,
ECONOMIC MILIEU OF THE
Evolution of:
(1) A-resources
and
..- URBAN REGION
(1) Expansion or decline of
J-activities
(2) B-resources
(2) Transformation of J-activities
to I-activities
(3) Expansion or decline of
I-activities
,
Exit of 'old' activities
Figure 8.2 Illustration of the urban dynamics outlined in formula (S.15)
8.4 IDENTIFYING THE TIME-SPACE HIERARCHY
8.4.1 A Leader-Follower Dichotomy
In the previous subsection, the economy of a region was schematically

divided into sectors I and J. The analysis in the following section is based on
a finer classification of economic sectors, which are divided into nine
different groups. It is assumed that each sector is sufficiently homogenous to
correspond to our concept of a product group, and that each of the nine
categories can represent an aggregate cluster of product groups as specified
above.
The basic distinction is between product groups and associated activities
which expand and decline as a share of total production. From this, one can
identify a leader region as discussed in Section S.l. This region is
characterised by having an earlier start and, initially, faster growth than the
follower regions in expanding activities. The follower regions, on the other
hand, are expected to achieve a higher expansion rate in a later phase.
A dichotomy of (i) leaders and (ii) early and late followers should be
interpreted in terms of frequency. Innovations and early imitations are
assumed to occur much more often in the leader regions than elsewhere.
Moreover, they serve as examples to other regions and demonstrate that the
new activities are both feasible and profitable. According to the theory
introduced earlier, novel activities and still-developing, non-standardised
product vintages are frequently initiated and promoted in leader regions. In
order to leave space for new activities, standardised and routinised activities
must gradually be removed from the leading region.
The approach described above was applied to the set of municipalities in
Sweden and Norway. In many cases, the functional region was composed of
three or more municipalities. The results presented below have been
aggregated into the following categories: (i) the set of cities in the Malardal
region (including Stockholm), (ii) Oslo-Akershus (Norway's leading region),
(iii) all Swedish regions outside the Malardal region, (iv) all Norwegian
regions including Oslo-Akershus. The result of the empirical analysis is that
the Malardal region (M-region) has a leading role in the Sweden-Norway
context. (In this study, the M-region comprises the county of Stockholm and
the municipalities of Uppsala, Enkoping, Habo, Vasteras, Koping, Arboga,
Eskilstuna and Strangnas.)
The first step was to measure for each municipality the relative share of
the employment in each activity. Next, the average share of each sector was
calculated for the set of all Swedish regions. For sector j, this share is
denoted by S j The corresponding value for region k is denoted by S jk By
comparing such shares in the M-region, S jM , with the corresponding share
in Sweden as a whole, one may classify sectors into sets with high, medium
and low concentration in the M-region. Let P j = S jM / S j and define:
P j ~ 1.4 sectors with high concentration
1.4 > P j ~ 0.7 sectors with medium concentration (8.16)
0.7> Pj sectors with low concentration.
In addition, the growth of the sectors during the period 1980-1990, g j , is

classified as follows:
gj ~ 0.3 is fast, 0.3> g j ~ 0 is medium, 0 > g j is negative (8.17)
The class limits between high, medium and low concentrations, as well
as between fast and medium growth, have been selected to generate a sharp
division between the corresponding clusters. In the empirical analysis,
sectors are cross-classified according to their concentration in the M-region
and their growth rate in the rest of Sweden (all municipalities outside the M-
region). In a similar way, the M-region concentration is classified against
sectoral growth in the Oslo-Akershus region, and against Norway as a
whole.
8.4.2 Location Dynamics and the Size-Hierarchy of

Functional Regions
Two alternative hierarchies of urban regions can be distinguished. The

first is based on Central Place Theory and is referred to as the CPS-ranking
of the regions. The second is an ordering which depends on each region's
advantages as a birth-place for product cycles and has been called a PV-
ranking. According to the discussion of the model and the assessment made
in Sections 8.2 and 8.3, these two hierarchies will to a large extent coincide.
With this background, a set of empirical hypotheses can be formulated. The
first may be expressed as follows.
Sectors with a high ,u-value will have a fast growth in the rest of Sweden.
These sectors will have a high concentration also in the Oslo-Akershus
region, since it is a leading region in Norway. However, the concentration
value for most of these sectors is higher in the M-region than in the Oslo-
Akershus region. This follows from the assumption that the M-region is a
leading region for both Sweden and Norway.
Sectors with a low ,u-value are expected decline in the rest of Sweden, in
Norway, and in Oslo-Akershus.
These hypotheses were examined on the basis of 3x3 contingency

matrices, in which the concentration of sectors in the M-region was cross-
classified against the pattern of change in the same sectors in (i) the rest of
Sweden, (ii) Norway and (iii) the Oslo-Akershus region. With regard to the
rest of Sweden, we obtained the matrix described in Table 8.1, with an
acronym for each cell group. The major result from the examination is that
frequencies are clustered along the diagonal, indicated by bold type. In the
following, some of the results are presented in truncated versions of the 3 x3
matrix.
Some sectors with medium concentration in the M-region may still be
growing in other regions, while other sectors with a medium ,u-value have
already started to decline. Sectors which decline in the rest of Sweden may
be concentrated in one or a few regions in which specialisation based on
economies of scale takes place. However, those declining sectors experience
reduced employment in most regions.
Table 8.1 Summary of empirical observations .related to the model of

specialisation dynamics in an urban hierarchy
Growth in the rest High p-value Medium p-value Low p-value

of Sweden
Fast H++ M++ L++
Medium (~O) H+ M+ L+
Negative H- M- L-
Sources: Forslund (1996); Forslund and Johansson (1995); National Bureau of Statistics in Sweden and
Norway.
Table 8.2 Contingency matrix of sectoral concentration and growth rates

for the rest of Sweden 1980-1990 (number of sectors)
Growth in the High and medium Low p-value Total

rest of Sweden p-value
Positive 46 15 61
Negative 13 31 44
Total 59 46 105
Sources: Forslund (1996), Forslund and Johansson (1995); National Bureau of Statistics in Sweden and
Norway.
The strong diagonal in Table 8.2 provides support to the empirical

hypotheses. In other words, we cannot reject the hypothesis that a high and
medium ,u-value is associated with fast growth in the rest of Sweden. More
than 77 percent of all sectors with such a ,u-value are characterised by
growth. Reciprocally, 67 percent of all sectors with a low ,u-value decline
during the period examined.
Table 8.3 Contingency matrix of sectoral concentration and growth rates

for Norway 1980-1990
Growth in Norway High and medium Low p-value Total

p-value
Positive 27 7 34
Negative 18 38 56
Total 45 44 90
Sources: Forslund (1996); Forslund and Johansson (1995); National Bureau of Statistics in Sweden and
Norway.
Table 8.3 reveals that during the 1980s Norway had a larger number of
sectors with declining employment than Sweden. However, the table still
supports the hypothesis that a low concentration in the M-region predicts
decline among Norwegian regions. Moreover, of the sectors with a positive
growth in Norway, 79 percent have a high ,u-value and only 21 percent a low
,u-value. In summary, we find that 72 percent of all sectors in Norway are
placed on the diagonal of the matrix. Standard i-measures for Tables 8.2
and 8.3 show that we cannot reject the hypotheses stated above. Also, when
the growth pattern in the Oslo-Akershus regions is examined in the same
way as in Table 8.3, it proves to be a follower of the M-region.
8.4.3 Spatial Concentration of Declining Product Vintages
In this subsection we examine sectors producing declining product

vintages. Table 8.4 shows that only 9 percent of the declining sectors in both
Norway and Sweden had a high concentration in the M-region. Hence, a
high ,u-value of a sector predicts correctly with few exceptions that the
sector is not declining in the rest of Sweden and Norway.
The core assumption in this study is that products and product groups
tend to mature over time, in the sense that they develop more standardised
attributes (higher 8-value) and select more routinised techniques (higher T-
values). According to the relevant theory, this implies that declining product
groups should have a lower knowledge-routine ratio than expanding groups.
The share of employees with formal education of varying length is used here
as a proxy for the knowledge-routine ratio. Table 8.5 shows that around 90
percent of the labour force in sectors with a low ,u-value, irrespective of
whether these sectors were declining or not in the rest of Sweden, had no
university level education. From Table 8.3 we know that most of the sectors
in Table 8.5 are declining.
Table 8.4 Sectors with a declining number of employees in the rest of

Sweden and in Norway 1980-1990
Sectors with a Sectors with a Sectors with a Total

high II-value medium II-value low II-value
Number of sectors in 4 9 31 44
the rest of Sweden
Number of sectors in 5 \3 38 56
Norway
Table 8.5 Average education shares (as %) in 1992 in Sweden for sectors
with a low fl value
Short university Long university Total university
Fast growth 6.6 3.1 9.8

Medium growth 6.1 2.8 9.0
Negative growth 6.9 3.5 10.6
Note: University refers to three or more years at college. Total university includes all university
education (short and long), including post graduate studies.
Are the university shares reported for 1992 in Table 8.5 particularly low?
Table 8.6 provides an answer. It shows that sectors with a high concentration
in the M-region and fast expansion in the rest of the country had a much
higher university share in 1992 than all sectors in Table 8.5. Moreover, even
for sectors with a medium concentration in the M-region, the university
share is almost three times higher than in sectors with a low concentration.
All this suggests that the knowledge intensity is relatively low in sectors
with a low ,u-value.
Table 8.6 Comparison of education shares for different groups of sectors

(national averages 1990)
Share of employees with any university education in the group of Percentage

sectors with a low .u-value divided by:
(I) Share of university education in (H++) sectors 30
(2) Share of university education in (M++) sectors 37
(3) Share of university education in (M+) sectors 36
Note: (H++) = a high .u-value and fast expansion in the rest of Sweden
(M++) = a medium .u-value and fast expansion in the rest of Sweden
(M+) = a medium .u-value and medium expansion in the rest of Sweden
8.5 LOCATION ATTRIBUTES ATTRACTING

EXPANDING PRODUCT GROUPS
8.5.1 Location Intensity as a Function of Location

Attributes
We shall begin by describing the activities which have a high

concentration in the M-region (high ,u-value) and then examine the
characteristics ofthe expanding activities in terms of knowledge intensity.
The relevant sectors are listed in Table 8.7, where we can observe that
the group (H++) basically contains advanced services together with
pharmaceuticals.
Table 8.7 Expanding sectors in the rest of Sweden
(H++) Sectors with high (H+) Sectors with high fl- (M++) Sectors with medium
fl-value and fast growth value and medium growth fl-value and fast growth
Financial services Urban transport Hotel services
Computer services Wine & liquor Cleaning services
R&D services Instruments Accounting services
Advertisement & marketing Photo & optical equipment Rental of machinery
Air transport Printing Engineering consulting
Technology trade Misc. transport &

communications
Pharmaceuticals Other financial institutions
Restaurants Insurance services
Special consulting Telecommunication products
Legal services Computer & office

equipment
Electrical motors, generators

and equipment
In which municipalities and functional regions do the above sectors have

a high concentration? What are the location attributes of these regions? By
inspecting Fig. 8.3 an answer is provided. This figure depicts a dynamic
ordering of all 70 functional 'labour market regions' in Sweden (see Batten
1985). The highest concentrations are found for the Stockholm, Uppsala,
Goteborg, Malmo and Vastenis regions. The chain Stockholm-Goteborg-
Malmo represent the upper triple of the rank size distribution. Moreover,
Uppsala and Vasteras are characterised by proximity to Stockholm and
belong to the M-region. When comparing the leading regions (in the figure)
with the average Swedish regions, they can be classified as knowledge and
R&D intensive. Moreover, they have a high customer, supplier and import
intensity, and a diversified labour market. University regions outside the
metropolitan agglomerations, like Umea and Linkoping, are placed in the
middle of the development curve.
There is one more essential observation which relates to Table 8.3. The
expanding sectors which are concentrated in the leading region and early
followers represent to a large extent business and producer services and
characterise the economic milieu in these regions. Moreover, this
concentration also demonstrates that leading regions are service (knowledge)
exporters to other urban regions at lower levels in the urban hierarchy.
s/(l-s) t
I
The Malardal region
Stockholm
Figure 8.3 Dynamic hierarchy of functional regions (A-regions) in Sweden

1990 (Batten 1985)
8.5.2 Knowledge Intensity of Expanding Product Groups
One of the basic assumptions in the theory of specialisation dynamics

presented in this chapter is that economic activities with a high knowledge
intensity have a high concentration in the leading region, and that sector with
a low knowledge intensity have a low concentration in the leading region.
The latter feature has been examined in the preceding section. This
subsection shows that sectors with fast growth and a high ,u-value have a
higher share of employees with further education (hence higher knowledge
intensity) than other sectors. Moreover, a high ,u-value is always associated
with a much higher knowledge intensity than one finds in sectors with a low
,u-value, and this holds for both growing and declining sectors. One may also
observe that both (M++)-sectors and (M+)-sectors are associated with a high
share of employees with further education (see Table 8.8 for this percentage
in Sweden).
Table 8. 8 Average shares of further education in Sweden in 1992 of sectors

with a high and medium ,u-value (as %)
Short university Long university Total university

H++ 15.6 15.7 33.5
H+ 11.2 9.4 21.0
M++ 13.1 13.1 26.8
M+ 14.6 12.9 28.1
M- 6.7 3.0 9.8

Note: University refers to three or more years at college. Total university includes all
university education courses (short and long), including post graduate studies.
It can be seen that all classes of sectors with a high ,u-value have twice as
large a share of employees with pre-university education than sectors with a
low ,u-value. They have around 2-3 times the share of employees with some
form of university education. Sectors with a high ,u-value have as much as
10-15 times the share of employees with post graduate studies than sectors
with a low ,u-value. Equation (8.18) describes a regression of a sector's
concentration (,u-value) and its share of employees with some kind of
university education, denoted by u. The association between fl and u is quite
strong, and indicated by a high t-value in brackets (8.18).
fl=0.32+3.19u; R2 =0.39 (8.18)

8.6 CONCLUSIONS: THE EUROPEAN URBAN

SYSTEM
The analysis in this chapter has shown how the CPS-model can function
as a core element in the analysis of specialisation dynamics in a CPS-
structured urban system. This model generates precise predictions as regards
those economic activities for which the density and size of demand is the
prime factor of the economic milieu. The CPS-model is complemented by
the PV-model, which focus on the distribution over the urban system of (i)
purchasing power, (ii) the supply of knowledge and development resources,
(iii) other interaction features related to agglomeration and localisation
forces. Hence, the PV-model combines the supply and demand aspects of the
economic milieu. For a smaller number of activities the pure DLA-approach
applies, with its concentration on activity-specific supply factors.
The CPS approach predicts that urban regions with a high rank are likely
to be initiators of new economic activities, new commodities and new
consumption patterns. By combining the static CPS-model with vintage
production theory, this extended CPS-model predicts a filtering down to
lower levels in the urban hierarchy. Hence, an increasing number of urban
regions become hosts of a given activity. Leading regions in the PV -model
are characterised by a favourable economic milieu - in this case size and
purchasing power is only one factor. The model applies to economic
activities for which the gradual change of the product vintage generates
changing milieu requirements and location options. The location change
follows a pattern where initially certain regions are specialised in a given
activity. At later stages, as the product vintage increases, these regions lose
their advantage, as other regions become specialised in the same field.
Hence, in the PV case there is no filtering down. Instead one observes a
discrete relocation to a few sites at lower levels in the urban hierarchy.
According to the above discussion, it is possible schematically to
distinguish between CPS-products and PV -products, though it should be
stressed that there is no sharp borderline. The empirical analysis presented in
Sections 8.4 and 8.5 was designed to capture predictions for both CPS-
products and PV-products. These two groups seem to have the same leading
region, which is matched against the rest of Sweden, as well as the whole of
Norway. Hence, for a CPS-product, a growing share in the rest of Sweden
implies growth in almost all urban regions. On the other hand, with a PV-
product, it should be observed that a growing share in the rest of Sweden
reflects a growing share in a small or limited set of urban regions. Naturally,
we can identify products which have an intermediate position on a scale with
CPS and PV as extreme points. An analysis of these aspects requires the
application of a multi-regional framework, and this could represent an
important second stage in the empirical analysis presented here. Such a step
forward could also be combined with a specification of attributes of the
economic milieu.
Another extension of the empirical investigation presented would be to
include a larger set of European countries. Since the data requirements are
fairly modest (employment figures with a fine sector specification), such an
extension is quite feasible. An interesting challenge would be to identify a
set of leading regions in Europe, examining to what extent they lead in
different fields, and investigating the relationship between dynamic
specialisation and regional distribution of economic milieu attributes. This
effort could also shed new light on existing pictures of the economic
geography of Europe. Would the leading region be located in the famous
'corridor' from the London region, curving down through Randstad,
following the Rhein, passing the Swiss domain and ending in the Milan
region? Other major central places may include Paris, Munich, Berlin and
Brussels. Secondary leading regions will include Rome, Florence, Vienna,
Helsinki, Stockholm, Oslo-Akerhus, Copenhagen, several British
metropolitan regions and still others. Moreover, could one apply the same
model to describe the technology diffusion and the location dynamics for
lower levels in the urban hierarchy of Europe?
Finally, it should be stressed that the PV-model has both an interregional
and an intra-regional message. The interregional aspect comprises the
relations between technical change, product development and location
dynamics in a multi-regional setting, where regions are characterised by
their economic milieu. But the intra-regional aspect is equally important. In
this case, the economic milieu is influenced by composition of economic
activities in the region, and hence the entry and exit of activities is part of the
ongoing formation of a region's economic milieu. If obsolete activities exit at
slow pace, the milieu characteristics deteriorate and there is not enough
space for new activities to enter. Moreover, a significant policy aspect is that
the economic milieu can evolve as the result of development and design of a
region's infrastructure.
REFERENCES
Andersson, A.E., Johansson, B. (1984) Knowledge Intensity and Product Cycles in

Metropolitan Regions. WP-84-13, IIASA, Laxenburg, Austria
Batten, D.F. (1985) The Changing Economic Structure of Metropolitan Regions.
Scandinavian Housing and Planning Research 2: 207-23
Batten, D.F., Johansson, B (1989) "Dynamics of Product Substitution" In Advances in Spatial
Theory and Dynamics. Andersson, A.E., Batten, D.F., Johansson, 8., Nijkamp, P., eds,
North-Holland, Amsterdam
Beckmann, M.J. (1958) City Hierarchies and the Distribution of City Sizes. Economic
Development and Cultural Change, VI(3): 243-48
Beckmann, M.J., McPherson, l (1970) City Size Distribution in a Central Place Hierarchy:
An Alternative Approach. Journal ofRegional Science 10: 25-33
Beckmann, M.J. (1996) The Location of Market Oriented Industries in a Growing Economy,
Paper presented at the 5th World Congress of the RSAI in Tokyo, May
Bos, H.C. (1965) Spatial Dispersion of Economic Activity. Rotterdam
Camagni, R., Diappi, L., Leonardi, G. (1986) Urban Growth and Decline in a Hierarchical
System - A Supply-Oriented Dynamic Approach, Regional Science and Urban Economics
16: 145-60
Cheshire, P.C., Gordon, I., eds (1995a) Territorial Competition in an Integrating Europe.
Avebury, Aldershot
Cheshire, P.C., Gordon, I. (1995b) "Change and Spatial Specialisation within the South East's
Economy." In Territorial Competition in an Integrating Europe, Cheshire, P., Gordon, I.,
eds, A vebury, Aldershot, 65-97
Cheshire, P.e., Hay, D.G. (1989) Urban Problems in Western Europe. An Economic Analysis,
Unwin Hyman, London
Ciccone, A., Hall, R.E. (1996) Productivity and the Density of Economic Activity, American
Economic Review 86: 54-70
Erickson, R.A, (1976) The Filtering-Down Process: Industrial Location in a Non-
Metropolitan Area. Professional Geographer, XXVIII
Fisher, lC., Pry, R.H. (1971) A Simple Substitution Model of Technological Change.
Technological Forecasting and Social Change 3: 75-88
Forslund, U.M. (1996) Industrial Location - Interregional Leads and Lags in Sweden and
Norway, Working Paper No 1, Swedish Institute for Regional Research, Ostersund
Forslund, U.M., Johansson, B. (1995) "The Mlilardalen: A Leading Region in Scandinavia
and Europe?" In Territorial Competition in an Integrating Europe, Cheshire, P.C.,
Gordon, I., eds, Gower, Aldershot, 3-27
Johansson, 8., ed (1996a) Dynamics in Metropolitan Processes and Policies, RR-86-8, IIASA,
Laxenburg, Austria
Johansson, B, ed (1996b) Spatial Dynamics and Metropolitan Change, RR-86-9, IIASA,
Laxenburg, Austria
Johansson, 8. (1987) "Technological Vintages and Substitution Processes." In Economic
Evolution and Structural Adjustment Batten, D.F., Casti, J., Johansson, 8., eds, Springer-
Verlag, Berlin, 145-65
Johansson, 8. (1993) "Economic Evolution and Urban Infrastructure Dynamics." In The
Cosmo-Creative SOCiety - Logistical Networks in a Dynamic Economy, Andersson, A.E.,
Batten, D.F., Kobayashi, K., Yoshikawa, K., eds, Springer-Verlag, Berlin, 151-75
Johansson, B., Karlsson, C. (1987) "Processes ofIndustrial Change: Scale, Location and Type
of Job." In Urban and Regional Labour Markets, Nijkamp. P., Fischer, M.M., eds, North-
Holland, Amsterdam, d 133-79
Johansson, B., Nijkamp, P. (1987) "An Analysis of Episodes in Urban Event Histories." In
Spatial Cycles. van den Berg, L., Klaassen, L.H., eds, Gower, Aldershot, 43-66
Johansson, B., Andersson, A.E. (1996) A Schloss Laxenburg Model of Product Cycle
Dynamics, mimeo, JOnkoping International Business School
Johansson, 8., Wigren, R. (1996) "Production Milieu and Competitive Advantages." In
Infrastructure and the Complexity of Economic Development, Batten, D.F., Karlsson, C.,
eds, Springer-Verlag, Berlin, 187-212
Marshall, A. (1920) Principles of Economics. Macmillan, London
Moroney, J.R., Walker, J.M. (1966) A Regional Test of the Heckscher-Ohlin Theorem.
Journal ofPolitical Economy
Nijkamp, P., ed (1990) Sustainability of Urban Systems - A Cross-National Evolutionary
Analysis of Urban Innovation, Avebury, Aldershot
Smith, B. (1975) Regional Specialisation and Trade in the UK. Scottish Journal of Political
Economy
Tinbergen, J. (1967) The Hierarchy Model of the Size Distribution of Centres. Papers of the
Regional Science Association, 20: 65-80
Vernon, R. (1960) Metropolis 1985, Harvard University Press, Cambridge Mass
Vernon, R. (1966) International Investment and International Trade in the Product Cycle.
Quarterly Journal ofEconomics, 80: 190-207
Chapter 9
The Impact of Research Activities on the European

Urban System
Stefano Magrini
University Ca' Foscari, Venice, Italy
9.1 INTRODUCTION
The renewed attention over the past decade given to the theoretical issues
of economic growth has been accompanied by a growing body of empirical
analyses aimed, in particular, at confirming the existence of a process of
convergence across national and regional economies. A substantial part of
this empirical literature is made up of cross-sectional and panel data
regression analyses that focus on the behaviour of a representative economy,
making the implicit assumption that it is moving along a steady path of
growth.
Although these studies are developed within the framework of the
traditional neo-classical model of growth, many authors do not consider
cross-sectional and panel data regression analyses of convergence a valid
test for this theory (Romer 1993, 1994; Fagerberg 1994; Cheshire and
Carbonaro 1995; Barro and Sala-i-Martin 1995; Sala-i-Martin 1996). Indeed,
the typical finding that economic systems are converging at a stable annual
rate of2 per cent (among others Barro 1991; Barro and Sala-i-Martin 1991,
1992, 1995; Holtz-Eakin 1992; Armstrong 1995a, 1995b; Sala-i-Martin
1996) is consistent with many other explanations of the growth process. It
could, for example, be equally consistent with both the evolutionary and
endogenous growth theories. Other sources of empirical evidence on the
determinants of economic growth need to be utilised therefore in evaluating
the relative merits of the different approaches. Several findings (Lucas 1988;
Barro and Sala-i-Martin 1995; Pavitt and Soete 1982; Fagerberg 1987, 1988;
Acs et a/1992, 1994; Feldman 1994; Audretsch and Feldman 1996; Coe et
198 MagriniS.
al 1997 among others) suggest that the study of economic growth cannot be
separated from the study of technological change and its determinants. For
this purpose, cross-sectional regressions can still represent an useful tool,
provided that they are used to shed light on the role played by different
factors in the growth process, rather than to analyse whether the poor
economies are converging towards the richer ones.
Within the traditional neo-classical model, technological change is
interpreted as a purely exogenous phenomenon and thus no economic
explanation of its evolution is given. By contrast, the evolutionary approach
has developed a framework in which technological change is explained by
the action of economic agents. Much of the theoretical work within the
evolutionary tradition has relied primarily on appreciative theory, i.e. on less
abstract, more descriptive modelling. In response, recent 'endogenous
growth' contributions have tried to codify some of the fundamental elements
of the evolutionary view within the formal modelling tradition of
mainstream economics. As a result, although differing profoundly in many
ways, both frameworks interpret technological progress either as the by-
product of other economic activities or as the intentional result of research
efforts carried out by profit-seeking agents. They therefore consider human
capital and innovation as fundamental elements in the explanation of the
process of economic growth.
One of the aspects that has received insufficient attention from
endogenous growth theorists is the relationship between technological
progress, knowledge spillovers and space. Fagerberg emphasises that:
"appreciative theorising often describes technology as organizationally
embedded, tacit and cumulative in nature, influenced by the interaction
between firms and their environments, and geographically localised"
(Fagerberg 1994, p. 1170).
Tacit knowledge, in particular, being the non-written personal heritage of
individuals or groups is naturally concentrated in space. Moreover, because
of its personal nature, tacit knowledge spills over space essentially through
direct, face-to-face contacts. It is important therefore to explicitly analyse the
geographical dimension of these spillovers since it is a feature of knowledge
creation and transmission that has been entirely neglected in formal theories
of endogenous growth.
The present empirical analysis of the role of research and development
(R&D) activities in regional growth is based on a theoretical model already
developed elsewhere (Magrini 1997, 1998). The main elements will be
briefly recalled in the following section. Section 9.3 presents the empirical
The Impact of Research Activities on the European Urban System 199
results for a data set of 122 major European Functional Urban Regions
(FURs) over the period 1979-1990, and in Section 9.4 we draw some
conclusions.
9.2 A MODEL OF REGIONAL GROWTH: BASIC

FEATURES AND FINDINGS
The main aim of the theoretical model developed in Magrini (1997,

1998) is to describe the role of formal research organizations - firms' R&D
laboratories, government laboratories, universities, etc. - in shaping the
spatial distribution of wealth within a two-region economic system. The
model, which builds on the existing literature on endogenous growth and, in
particular, on the work of Romer (1990a, 1990b), Rivera-Batiz and Romer
(1991a, 1991b) and Rivera-Batiz and Xie (1993), has three main features.
Firstly, economic growth is assumed to be endogenous and driven by the
research activity of profit-seeking agents. Secondly, an explicit role in
regional production structures is assigned to human capital, which is
considered to be the crucial input in the research sector. Thirdly, knowledge
spillovers across space are an essential feature of research activity aimed at
designing and developing new products.
The model presents a stable equilibrium characterised by permanent
differences in per capita income levels. By resorting to a definition of
research activities that recognises the important role played by spillovers of
both tacit and abstract knowledge, it is suggested that income disparities owe
their existence to a process of regional specialisation between 'knowledge
creating' and 'knowledge applying' regions. The ability to innovate within a
regional economy depends on the interaction between the macro-economic
system and the various factors shaping its Regional Innovation System
(RIS). The result is the development of a location-specific ability to innovate
which is referred to as the regional technological competence in research.
Those regions which are better able to innovate through the development of
superior technological competence in research will be characterised by a
relative specialisation in research activities and thus become 'knowledge
The present analysis adopts the set of functional regions derived by Hall and Hay (1980)
and described in the Data Appendix. As argued by Cheshire and Hay (1989), Cheshire
and Carbonaro (1996), Cheshire et at (1996) and Magrini (1999), the very nature of
regional economic disparities means that any empirical study on the subject must take
space into consideration and opt for a definition of the region which centres on the spatial
sphere of socio-economic influence. Since the functional links between spatial units are
limited by space, functional regions seem to be a suitable choice, as they take account
explicitly ofthe distance factor.
200 Magrini s.
creating' regions. Since research activities tend to make a more intensive use
of human capital than manufacturing activities, the relative concentration of
research in one location leads to a parallel concentration of human capital.
Moreover, since wages for human capital tend to be higher than wages for
unskilled labour, the relative concentration of human capital in one region
implies that the average level of per capita income in these 'knowledge
creating' regions will be higher than in 'manufacturing' regions.
The model also offers a possible interpretation of the effects of European
integration on the disparities in per capita income. Indeed, the process of
integration which has characterised recent European history seems able to
foster the long term growth in the economic system, by determining a
reduction in the cost of physical distance. At the same time, however, given
that the reduction in the cost of distance has been achieved primarily through
a reduction in the travel time between locations (there have been few
improvements in cultural and institutional homogeneity, so a high degree of
heterogeneity still characterises the European system), the price to pay could
be an increase in regional differentials. In such a situation, even though
integration may reduce existing gaps in regional levels of technological
competence in research, disparities in per capita income are likely to widen.
In other words, the integration process might determine the emergence of a
new steady-state equilibrium characterised by a further concentration of
research activities in the regions which were already relatively specialised in
research. During the transition towards this new equilibrium, per capita
income growth rates differ across regions. While the adjustment takes place
through the reallocation of unskilled labour and human capital, the average
per capita income in the more innovative, relatively research-intensive
region grows at a faster rate than in the other region.
The fundamental equation of the model describes the activity of each
regional research sector. Let us consider an economic system made up of
two regions, i and j, in which the cost of moving from region one region to
the other is equal to dij. The flow of new knowledge, i.e. the number of new
designs created in region i at any point in time is given by:
(9.1)
where Hr; is the level of human capital employed in the research sector of
region i, and 0; is the level of technological competence characteristic of the
research sector located in region i. A is the number of intermediate inputs
existing in the system and the overall level of abstract knowledge created so
far (and available to all researcher due to aspatial knowledge spillovers). As
far as the spatial spillovers of technological tacit knowledge are concerned,
Hr/ reflects the size of the intra-regional spillovers, whilst the term
Hr} d;ll Pij represents the extent of the inter-regional spillovers of tacit
knowledge that benefit the research effort in region i and originate from the
interaction with the research sector located in region j. These are functions of
the level of human capital existing in the other region, Hrj, weighted by the
cost of physical distance, dij' and a measure of the potential technological
benefit from interaction, pij. As these spillovers result primarily from the
physical interaction between researchers, their size is inversely related to
(the cost of) physical distance. However, the 'catch-up' argument
(Gerschenkron 1962; Abramovitz 1986) emphasises the potential benefit that
can be enjoyed by technologically less advanced economies from the
interaction with economies closer to the technological frontier, due to the
possibility of imitating technologies already developed elsewhere. It is
assumed that the potential technological benefit accruing to researchers
located in one region from the interaction with the researchers of the
technologically more advanced region is an increasing function of the
relative local technological competencies in research ~echnological leader / 80ther
region and a decreasing function of the cost of the physical distance. The
measure of the potential technological benefit accruing to region i from the
interaction with region j can therefore be represented by:
(9.2)
where
8 if 8; > Ii}
{
Ii technological leader = ~ if 8; < Ii}
A crucial element of this description of the research effort is the

parameter t5, the level of' local technological competence in research', which
is a concept introduced to account for location-specific tacit knowledge. It is
defined as the ability of the RlS to perform research and allows for spatially
bounded spillovers of knowledge arising from interaction between
researchers. In turn, the RlS is the local network of public and private
institutions supporting the initiation, modification and diffusion of new
technologies (Freeman 1987; Nelson and Rosenberg 1993; Patel and Pavitt
1994). Among the factors that constitute the RlS, it is possible to emphasise
the role played by: the size and quality of the education system, the
availability of technical, financial and networking services, the quantity and
202 Magrini S.
quality of space available for innovative activities, the structure of the local
industrial sector, and both the system-wide and local macro-economic
setting.
9.3 THE VARIABLES OF THE EMPIRICAL MODEL
The aim of the present section is to provide a simple test for the model's
predictions. In particular, we study the fundamental determinants of
economic growth in the 122 major European FURs. The period chosen, from
1979 to 1990, conforms to two fundament requirements. Firstly, it is long
enough to allow for cyclical movements to occur around the growth trend.
Secondly, this is a period in which the European system underwent
important steps in its process of economic integration.
The dependent variable is the growth rate of per capita GDP in each
FUR. The formula for the growth rate is the traditional logarithmic
transformation of the ratio of regional per capita GDP at the beginning and
end of the study period:
GROWTH. =~ln(Yi'J990) (9.3)

I 11 Y;,1979
The fundamental independent variables of the empirical model relate to

the activity of research performed in the regions. Indeed, on the basis of the
theoretical analysis summarised above, the regions which are relatively
specialised in research activities are expected to grow faster than regions
specialised in manufacturing activities. However, according to equation
(9.1), it is not only the level of research activity carried out within a region
that matters, but also the level of knowledge spillovers. Whilst aspatial
spillovers are not of interest as far as the relative performance of regional
economies is concerned (as they accrue to all regions indifferently), other
forms of spatially asymmetric spillovers must be taken into account. An
attempt is therefore made here to estimate the total effect of research activity
on the growth performance of the region by considering both intra- and
inter-regional spillovers oftacit knowledge. Ideally, this requires data for the
level of employment in research activities in all the regions at the beginning
of the study period, as well as data for technological and physical distances.
Unfortunately, however, such data is not easily available and it has been
necessary to resort to proxies. The level of regional research activity has
therefore been measured by counting the number of R&D laboratories
located in the region at the beginning of the period. (More precisely, the
laboratories considered were those belonging to corporations which appear

in the Fortune 'Top 500' lists for the United States and also outside the
United States). The data on the laboratories and their location has been
obtained from the Directory of the European Research Centres published in
1982. To represent the relative concentration of research activities within the
region, the number of R&D laboratories has been expressed per unit of
population. As pointed out by Cheshire and Carbonaro (1995, 1996), this is
only a crude measure of the theoretically appropriate variable. It does
however seem able to provide a general indication of the relative
specialisation in research and the extent of the spatial spillovers of
knowledge.
To obtain an estimate of the parameter measuring the strength of the
intra-regional spillovers, the R&D variable has been divided by the area of
the region. As for inter-regional spillovers, the initial step was the
calculation of two matrices of time distances (expressed in minutes) between
each pair of FURs. The first matrix takes the time distance by road, whilst
the second matrix takes the shorter time distance, when a choice between air
travel and road is available. Given the importance of air transport
infrastructure for regions with a strong commitment to research-intensive
activities (emphasised by many empirical studies of European urban regions,
for instance, Andersson et al 1990; Batten 1995), this second matrix was
then used in the estimation of inter-regional spillovers of knowledge. A
proxy for the technological distance between pairs of FURs, flij, was then
calculated on the basis of equation (9.2), where the relative levels of
technological competence in research were estimated using data on regional
technological creativity for the early 1980s derived by Ake Andersson and
reported by Batten (1995).
To sum up, a first set of variables is used in the empirical analysis to
account for the role of research activity and the spatial spillovers of tacit
knowledge. More details on these variables are reported in Table 9.1. The
first of these variables, labelled R&D, simply reflects the relative
concentration of research without allowing for spatial spillovers. The other
variables, R&DS I to R&DS7, consider both types of spatial effects and
allow for different sizes of the distance range over which inter-regional
spillovers are calculated. The a priori expectation is that, whilst all variables
should be positively related to per capita GDP growth, those allowing for
spatial spillovers should be statistically more significant and improve the
overall performance of the model. The statistical significance of these
variables, together with the measure of fit of the resulting models, will then
be used to identify the distance range over which the inter-regional spillovers
appear to be the strongest.
204 Magrini s.
A second set of variables has been introduced in the empirical analysis in
order to reflect the local factors shaping the RIS. These in tum determine the
regional level of technological competence in research, 8, and, most
importantly, its likely evolution. Although not all these factors have been
explicitly considered here due to data availability problems, it seems
nonetheless possible to take into account some of the most relevant ones.
Table 9.1 Research and development variables
Variable Spatial Distance threshold

spillovers
R&D No
R&DSI Yes 90 minutes
R&DS2 Yes 110 minutes
Universities are producers of education and therefore influence the

quality of the human capital available to firms. They also engage in research
activities and produce knowledge. Universities are therefore an essential
feature of the RIS and their influence on the evolution of regional
technological competence in research must be accounted for. This has been
done by considering the number of academic staff employed in universities
and other further education institutions in the academic year 1976-1977.2
Clearly, this variable is expected to playa positive role in the economic
performance of the region.
The level of regional innovative activity is also influenced by the
structure of the local industrial sector. The theoretical model indicates that a
concentration of manufacturing activities is detrimental for economic
growth: it can hamper the ability of the existing regional research sector both
to develop a superior technological competence in research and to attract
other researchers. Clearly, not all manufacturing activities have the same
negative role. The variables COAL and PORT are intended to account for
those industries which are likely to playa particularly negative influence on
the growth prospects of local economies or, in other words, identify "old
industrial regions suffering from industrial decline and employment loss"
(Objective 2 regions). As pointed out by Cheshire and Carbonaro (1995), the
2 See the Data Appendix for details on the sources.

presence of coal mining can adversely affect local growth prospects for a
considerable period of time, even after this industry has ceased to account
for a substantial share of employment. Consequently, the influence of the
coal industry is taken into account through a dummy variable related to the
coincidence of the area of the FUR with a coal field as defined in the Oxford
Regional Economic Atlas (1971).
The second variable reflects the extent of port activities, measured in
terms of the amount of sea-going freight handled in 1978. Dramatic new
developments in transport technology and, in particular, the introduction of
containerisation and roll-on roll-off ferries, have greatly reduced the
attractiveness of ports as locations for processing activities. During the study
period, this is likely to have negatively affected all ports in relation to their
size. However, the transformation in the industry is likely to have led to a re-
organisation of traffic flows and, therefore, to an increase in the degree of
competition among existing ports. Large ports, thanks to their economies of
scale, may have taken advantage of the process of re-organisation and
increased their share of traffic at the expense of smaller ports. As a result,
the relationship between port size and growth of per capita income in the
region could be quadratic rather than linear.
Another relevant feature of the local industrial structure concerns the
relative importance of service sector due to the role played by the variety of
business services in providing firms with market, financial and commercial
knowledge. This factor is measured as the percentage share of employment
in service activities over employment in services and manufacturing in 1980,
and is expected to be positively related to per capita GDP growth.
An interesting question related to the local industrial structure is whether
local ability to innovate is promoted by industrial specialisation, due to intra-
industry spillovers, rather than by industrial diversity and inter-industry
spillovers. The underlying theoretical model adopted here does not provide
any indication of the relative importance of these two possibilities, but
leaves the question open to empirical investigation. The degree of
specialisation of the regional economies is therefore measured on the basis
of employment data for nine industrial sectors. 3 The employment in each
regional sector is expressed as a percentage of the total industrial
employment in the region. After ranking the sectors by size, the index of
regional specialisation is calculated as the ratio between the average
percentage of employment for the four smallest regional sectors over the
average percentages of employment for the four largest ones. The index,
therefore, ranges between 0 and 1, these two extremes indicating
respectively specialisation and diversity in the regional industrial structures.
A breakdown of the sectors is provided in the Data Appendix.

206 Magrini S.
The variable AGR is the share of employment in agriculture in 1975

within the wider NUTS2 region. This variable therefore focuses on
Objective 5b regions and, at least partially, Objective 1 regions. We argue,
as do Cheshire and Carbonaro (1995, 1996), that the relationship between
growth in per capita GDP and specialisation in agriculture in the NUTS2
region should be quadratic. Indeed, in FURs surrounded by regions with a
relative specialisation in agriculture, we can expect economic growth to be
relatively slow because these FURs are unable to attract research activities or
other human capital rich activities. At the same time, the migration of
unskilled workers from the countryside to urban areas is likely to lead to
population increasing faster than output and hence falling average levels of
human capital. On the other hand, FURs located in densely urbanised
regions will probably suffer from congestion and other environment-related
problems and could find it difficult to attract human capital.
A second variable that considers more directly the quantity and quality of
the local supply of space suitable for research activities is represented by the
density of the population in the FUR area in 1981. The level of density can
be considered as a proxy for the land rent. At the same time, urban areas
have witnessed a rapid increase in traffic levels that in many cases has led to
acute congestion problems. In both cases, population density, measured in
terms of the number of inhabitants per square kilometre, is expected to be
negatively associated with growth.
The variable labelled SDG represents the sum of the difference between
the growth rate of a given FUR and the growth rates in other FURs within a
150 minutes radius. In particular, the variable is calculated as
Yi _Yj
Yi Yj
(9.4)
dij
where dij represents the road distance between regions. Moreover, to avoid
problems of definitional correlation with the dependent variable, the growth
rates are calculated over the period 1979-1985. This variable is introduced
into the analysis in order to take into account the spatial adjustment between
neighbouring FURs. As explained by Cheshire (1979), adjacent local labour
markets tend to interact primarily through the adjustment of commuting
patterns (see also Evans and Richardson 1981; Burridge and Gordon 1981;
Gordon and Lamont 1982; Gordon 1985). Relatively rapid growth of per
capita GDP in one FUR will tend to attract additional in-commuters from
surrounding FURs. The impact of this mechanism is twofold. The first effect
is essentially statistical, as the increase in the number of in-commuters will

lead to an apparent rise in per capita GDP (since output, which is measured
at workplaces, will increase, while the resident population remains stable).
The second effect concerns the level of human capital and productivity of
the workers employed in the FUR. As the additional in-commuters induced
by the differential in growth rates will be relatively long distance
commuters, who tend to have higher human capital and productivity than
short distance commuters, this is likely to increase the average level of
human capital in the recipient FUR. This, in tum, could have dynamic
implications due to the intra-regional spillovers of knowledge of the research
sector. The combination of both effects could well result in a positive
relationship between the growth rate of a FUR and the sum ofthe differential
growth with adjacent FURs.
The variable NFGROWTH reflects the influence of the macro-economic
system. National macro-economic policies, education policies, national
culture, as well as legal and social institutions all help to shape the local
technological competence in research and its evolution over time. This
variable has been calculated as the growth rate of per capita GDP in the part
of the nation that remains outside the major FURs. At a sub-national level,
empirical analyses have often stressed the specificity of the southern regions
of Spain.4 For instance, in their analysis of the Spanish Provinces, Mas et al
(1995) find that growth prospects for the southern agricultural Provinces of
Spain are significantly worse than for northern and eastern part of the
country. As a consequence, a dummy variable for the south of Spain has
been introduced in the model. s
9.3 THE RESULTS

The empirical model can therefore be summarised as follows:
-1 1n (Yi'1990
11 YiI979
J
- - =ao+a E i1979+Si1990
"
(9.5)
where E1979 is the vector of explanatory variables just described. The results
of the OLS cross-sectional estimation of these equations have been given in
Table 9.2.
4 A dummy variable for the Italian Mezzogiorno was also introduced. However, the
inclusion of this variable proved to add no explanatory power to the model.
S The FURs included in this variable are Alicante, Cordoba, Granada, Malaga, Murcia, and
Seville.
N
0
00
Table 9.2 The determinants of per capita GDP growth in the FURs
~
~
(JQ
2 3 4 S 6 7 8 :
Constant 0.00718 0.00692 0.0070S 0.00743 O.OO72S 0.00690 0.00713 0.00736 ~
(0.52) (O.SO) (0.51 ) (0.S4) (0.S3) (0.50) (0.52) (0.S3)
R&D 0.00011
(2.74)
R&DSI 6.6ge-S
(2.81 )
R&DS2 6.02e-S
(2.90)
R&DS3 S.92e-S
(3.01 )
R&DS4 S.6Se-S
(3.00)
R&DSS S.1ge-5
(3.00)
R&DS6 4.86e-S
(2.98)
R&DS7 4.34e-S
(2.9S)
University 3.81e-7 4.24e-7 4.23e-7 4.23e-7 4.24e-7 4.17e-7 4.0ge-7 4.15e-7
(2.22) (2.47) (2.47) (2.48) (2.48) (2.44) (2.40) (2.43)
Coal -0.00296 -0.00331 -0.00339 -0.00336 -0.00336 -0.00336 -0.00339 -0.00338
(-2.31) (-2.57) (-2.63) (-2.62) (-2.62) (-2.62) (-2.64) (-2.63)
(continues)
Table 9.2 The determinants of per capita GDP growth in the FURs (continued)
Port -9.74e-5 -9.35e-5 -9.54e-5 -9.54e-5 -9.64e-5 -9.75e-5 -9.96e-5 -9.6ge-5

(-2.45) (-2.37) (-2.42) (-2.43) (-2.45) (-2.47) (-2.52) (-2.46) ...,
Port2 3.94e-7 3.80e-7 3.8ge-7 3.91e-7 3.94e-7 3.97e-7 4.04e-7 3.93e-7 ::r
(l)
(2.20) (2.13) (2.18) (2.20) (2.22) (2.23) (2.26) (2.21) 3

-
'"d
Service 0.01742 O.oI726 0.01684 0.01630 0.01655 0.01710 0.01717 0.01682 s:>l
n
....
(2.21) (2.19) (2.14) (2.07) (2.11) (2.18) (2.19) (2.14) 0
....,
IQ
Specialisation -0.03957 -0.04063 -0.03917 -0.03907 -0.03942 -0.03960 -0.04015 -0.04026 (l)
(l)
(-2.53) (-2.62) (-2.52) (-2.52) (-2.54) (-2.56) (-2.60) (-2.60) '"
!:l
n
Agriculture 8.9ge-4 9.26e-4 9.46e-4 9.55e-4 9.58e-4 9.54e-4 9.50e-4 9.4ge-4 ::r
(4.01) (4.09) (4.16) (4.21 ) (4.21) (4.20) (4.19) (4.18) :>
n
....
Agriculture 2 -3.35e-5 -3.43e-5 -3.48e-5 -3.50e-5 -3.51e-5 -3.50e-5 -3.50e-5 -3.50e-5 ~:
....
(-5.28) (-5.38) (-5.44) (-5.48) (-5.49) (-5.48) (-5.47) (-5.47) G
'"0
Density -0.00936 -0.00952 -0.00957 -0.00954 -0.00957 -0.00960 -0.00963 -0.00959 ::t
....
::r
(-3.99) (-4.06) (-4.09) (-4.09) (-4.10) (-4.12) (-4.12) (-4.11) (l)
tTl
SDG 0.19645 0.20122 0.20366 0.20424 0.20538 0.20554 0.20437 0.20251 ~
...
0
(4.93) (5.04) (5.10) (5.13) (5.15) (5.15) (5.13) (5.09) '"d
(l)
NFGrowth 1.02707 1.03278 1.03051 1.02832 1.02922 1.03037 1.02888 1.02855

(6.92) (6.98) (6.98) (6.98) (6.99) (6.99) (6.98) (6.97) c:::::
...
0-
DSE -0.00936 -0.00952 -0.00957 -0.00954 -0.00957 -0.00960 -0.00963 -0.00959
r/:l
(-3.99) (-4.06) (-4.09) (-4.09) (-4.10) (-4.12) (-4.12) (-4.11) '<!
'"~
3
R2 0.5812 0.5826 0.5844 0.5866 0.5865 0.5865 0.5860 0.5855 N
0
Note: t-ratios are shown in brackets \0
210 Magrini S.
Whereas the first version of the model (shown in the first column) makes
use of the variable on research activity without considering spatial spillover
effects, these effects are allowed for in all the other estimated versions. The
results appear quite robust in all versions. The R 2 values range between 0.58
and 0.59, a satisfactory level for a large cross sectional data set. All the
expectations on the signs of the coefficients are met, and all coefficients are
generally highly significant.
The first important result is that not only do all the coefficients for the
variables reflecting the role of R&D activities on per capita GDP growth
have the expected positive sign, but they are also highly statistically
significant. Despite all the caveats concerning the measurement of this
activity expressed in the previous section, this is nonetheless an encouraging
result.
The comparison between the results on the R&D variables for the
different models gives some insight into the role of spatial knowledge
spillovers. Indeed, we should note that the inclusion of the spillovers in the
R&D variable determines a generalised improvement in the regression
results. Both the R 2 values and the level of statistical significance of the
coefficients for research activities are generally enhanced when these spatial
effects are accounted for. At the same time, the statistical significance of the
other variable closely related to innovation activity, the number of university
staff, is also substantially improved by the inclusion of these effects. All
these results could therefore be interpreted as supporting the view that spatial
spillovers of knowledge are an important feature of innovation activities.
Concentrating on the versions of the model that allow for these effects
(columns 2-8), it is possible to analyse the effect of space on the strength of
interaction between research sectors of neighbouring FURs. As explained in
the previous section, these variable are calculated by considering different
distance ranges for the inter-regional spillovers. The distance ranges
considered here vary from a minimum of 90 minutes to a maximum of 150
minutes. The best version of the model, both in terms of the regression
R 2 and of the t-ratios for the 'research activity' and the 'university'
variables, corresponds to the R&DS3 variable (column 4), which allows for
interaction between regional research sectors located within a range of 115
minutes. In other words, the strength of the inter-regional interaction
between researchers appears to reach its strongest level when the researchers
are on average less than 115 minutes apart. Moreover, the strength of the
inter-regional spillovers declines slowly with increases in the distance range
up to 125 minutes, whilst it appears to decline more rapidly for larger
distances.
Given these results, we now concentrate on the fourth version of the
model. The coefficient for the index of specialisation of the local industrial
structure is negative and significant at the 1% level. This suggests that, other
things being equal, those regions that were characterised by a higher degree
of specialisation in their industrial structures at the beginning of the study
period have grown faster. With regard to the nature of the intra-regional
spillovers, this suggests that, in aggregate terms, intra-industry dynamic
externalities have been more effective than inter-regional dynamic
externalities in stimulating per capita GDP growth.
The role of port activities on regional growth according to the estimated
coefficients is described in Fig. 9.1. The regression results seem to support
the view, expressed in the previous section, that the relationship between
port size and regional growth is quadratic. However, a closer look shows that
the minimum point of the curve is reached for an amount of trade just
exceeding 120 million tons. The only port which at that time was handling
more than 100 million tons was the port of Rotterdam (with 259 million
tons), whilst the second port in terms of goods handled was Marseilles with
93 million tons.
0.002
50 100 150 200
-0.004
-0.006
Figure 9.1 The role of port activities
It is therefore likely that the functional form of the influence of port

activities on regional growth is heavily influenced by the observation for
Rotterdam. To check for this possibility, two further regressions were run in
which the port activity variable excludes Rotterdam. In the first of these
regressions, the functional form for the influence of port activity on regional
growth was linear; in the second regression a quadratic form was considered.
The results, together with the results of the best version of the previous set of
regressions (version 4), are shown in Table 9.3.
They confirm the impression that the quadratic form is in fact due to the
very high leverage on the observation for Rotterdam. The coefficients for
212 Magrini S.
'Port2' and 'Port22 " the variables on port activity which exclude the
observation for Rotterdam, are both statistically non significant, thus
rejecting the hypothesis of a quadratic form for the influence of the other
European ports on regional growth. On the contrary, when the relation
between port activity and growth is assumed to be linear (version 10), the
coefficient is negative and highly significant.
Table 9.3 The influence of port activities
4 9 \0
Constant 0.00743 0.00851 0.00823
(0.54) (0.62) (0.60)
R&DS3 5.92e-5 5.72e-5 5.76e-5
(3.01) (2.91) (2.97)
University 4.23e-7 4.16e-7 4.16e-7
(2.48) (2.46) (2.47)
Coal -0.00336 -0.00340 -0.00340
(-2.62) (-2.66) (-2.67)
Port -9.54e-5
(-2.43)
Porr 3.9Ie-7
(2.20)
Port2 -6.15e-4 -7.30e-4
(-0.82) (-2.47)
Port2 2 -1.78e-5
(-0.17)
Service 0.01630 0.01649 0.01658
(2.07) (2.12) (2.15)
Specialisation -0.03907 -0.03979 -0.03967
(-2.52) (-2.59) (-2.59)
Agriculture 9.55e-4 9.43e-4 9.44e-4
(4.21) (4.19) (4.22)
Agriculture 2 -3.50e-5 -3.48e-5 -3.48e-5
(-5.48) (-5.49) (-5.51)
Density -0.00954 -0.00942 -0.00947
(-4.09) (-4.02) (-4.09)
SDG 0.20424 0.20356 0.20343
(5.13) (5.14) (5.16)
NFGrowth 1.02832 1.01503 1.01784
(6.98) (6.90) (6.99)
DSE -0.00954 -0.00942 -0.00947
(-4.09) (-4.02) (-4.09)
R2 0.5866 0.5873 0.59\0

In other words, it seems possible to conclude that, generally speaking,

port activity has a negative influence on the growth prospects of a region.
The most noticeable exception is represented by the port of Rotterdam
which, thanks to a process of profound - and successful - restructuring, has
positively contributed to the growth performance of the region. Because of
the amount of the goods traffic handled by this port in 1978 (almost one
fourth of the total handled by the ports included in this database passed
through the port of Rotterdam), the relationship between port activity and
regional per capita GOP growth appears to be quadratic.
As far as the role of agriculture is concerned, the results of the regression
confirm the expectation of a quadratic relationship with regional growth.
This relationship, which is represented in Fig. 9.2, is stable in the face of
changes in the port variable. Both for version 4 and version 10, the curve
representing the influence of agricultural employment reaches its maximum
at a value of 13.6. Unlike the previous case, almost one sixth of the
observations in the database have a value exceeding the maximum.
0.006
10 20 % share
-0.004
Figure 9.2 The role of the share of employment in agriculture
Finally, a set of diagnostics was performed on versions 4 and 9 of the

model. The first test was the Kiefer-Salmon test for the normality of the
residuals. Secondly, the heteroscedasticity was tested with one of two
different diagnostics: either the Breusch-Pagan (BP) Lagrange Multiplier test
or the Koenker-Bassett (KB) test. Following the testing procedure in
SpaceStat (see Anselin 1994), the choice depended on the results of the
normality test. When the errors were non-normal (for a probability level of
0.01) the KB test was preferred. The Ramsey's RESET test was then used to
check the functional form. Finally, four separate diagnostic statistics for
spatial dependence were produced: Moran's I statistic, Burridge's Lagrange
214 Magrini S.
multiplier test, Kelejian and Robinson's test for spatial error, and Anselin's
test for spatial lag (for details see Anselin 1988, 1994). In each case, the
tests were based on both distance matrices used in the derivation of the
variables of the model. All diagnostics excluded the presence of
specification problems with either of the two preferred versions of the
model.
9.4 CONCLUSIONS
The results of the regression analyses lend some support to the main
predictions of the theoretical model outlined in Section 2. These results can
be summarised as follows. Firstly, research activities appear to play an
important role in the process of regional growth. Indeed, the coefficients for
the variables measuring regional research efforts were always positive and
highly significant. Secondly, by considering different specifications of the
spatial interaction between researchers, it has been possible to find evidence
supporting the existence of spatial spillovers of knowledge. Inter-regional
spillovers of knowledge appear to reach their maximum effect within a
radius of about 115 minutes. Thirdly, it has been possible to identify several
factors which affect the regional growth rate of per capita GDP by shaping
the local level of technological competence in research. One of these appears
to be the existence of universities. Indeed, universities contribute to the
regional research effort both directly, in their role as research centres, and
indirectly, as part of the regional infrastructure providing new human capital.
Data limitations did not allow us to analyse these effects separately.
Nonetheless, according to the results of the empirical analysis, the
combination of these effects has a significant positive impact on regional
growth. Finally, another interesting outcome concerns the controversy on the
relative importance of intra-industry and inter-industry dynamic spillovers in
promoting growth. An index of the degree of sectoral specialisation of
regional industrial specialisation has been used to shed light on this widely
debated issue. The results indicate that, during the period 1979-1990,
European regions characterised by a higher degree of sectoral specialisation
grew faster than regions with a more diverse industrial structure. In other
words, intra-regional dynamic spillovers appear to have been more
successful than inter-regional dynamic spillovers in fostering regional
economic growth.
Finally, it is necessary to sound a note of caution in connection with the
problem of data limitations since these have, in some instances, led to the
use of rather rough measures for the variables required by the theoretical
model. This is the case, in particular, for the variable relating to research
activity. The lack of spatially disaggregate data on research employment

made it necessary to adopt a crude measure of research activity, but this
appears nonetheless to have been adequate in providing an initial indication
of the influence of research activities on regional growth.
DATA APPENDIX
9.Al Definition of the FURs
The present analysis adopts the set of functional regions proposed by

Hall and Hay (1980) and adopted by Cheshire and Hay (1989) in their
analysis of urban problems in Europe between 1951 and 1981. Each of these
regions, referred to as Functional Urban Regions (FURs), was derived from
a two-step procedure. Firstly, a core was defined by identifying an urban
centre with 20,000 jobs or more, then adding all those contiguous
surrounding areas - at the lowest level of disaggregation available - with a
density of 12.35 jobs or more per hectare. Secondly, to each core were added
all those contiguous administrative areas from which more workers
commuted to the core in question than to any other core. The calculation of
the GDP series was based on the data provided by the Community Statistical
Office for NUTS 3 regions, which made it possible to derive a consistent
time series for the 122 largest FURs6 from 1979 to 1991.
The FURs are:
Belgium Antwerpen, Bruxelles-Brussel, Chaleroi, Liege;

Denmark Arhus, K0benhavns;
Germany Aachen, Augsburg, Berlin, Bielefeld, Bochum, Bonn,
Braunschweig, Bremen, Dortmund, Dusseldorf, Duisburg,
Essen, Frankfurt, Hamburg, Hannover, Karlsruhe, Kassel, KOln,
Krefeld, Mannheim, Monchengladbach, Munchen, Munster,
Numberg, Saarbrucken, Stuttgart, Wiesbaden, Wuppertal;
Greece Athinai, Thessaloniki;
Spain Alicante, Barcelona, Bilbao, Cordoba, GijoniAviles, Granada,
La Coruna, Madrid, Malaga, Murcia, Palma de Mallorca,
Sevilla, Valencia, Valladolid, Vigo, Zaragoza;
6 Defined as those with a core city greater than 200,000 inhabitants and a total population
greater than a third of a million inhabitants.
216 Magrini S.
France Bordeaux, Clermont-Ferrand, Dijon, Grenoble, Le Havre, Lille,

Lyon, Marseille, Montpellier, Mulhouse, Nancy, Nantes, Nice,
Orleans, Paris, Rennes, Rouen, St. Etienne, Strasbourg, Toulon,
Toulouse, Valenciennes;
Ireland Dublin;
Italy Bari, Bologna, Brescia, Cagliari, Catania, Firenze, Genova,
Messina, Milano, Napoli, Padova, Palermo, Roma, Taranto,
Torino, Venezia, Verona;
The Netherlands Amsterdam, Rotterdam, S-gravenhage, Utrecht;
Portugal Lisboa, Porto;
UK Belfast, Birmingham, Brighton, Bristol, Cardiff, Coventry,
Derby, Edinburgh, Glasgow, Hull, Leeds, Leicester, Liverpool,
London, Manchester, Newcastle, Nottingham, Plymouth,
Portsmouth, Sheffield, Southampton, Stoke, Sunderland,
Teeside.
9.A2 University Staff
This variable reflects the number of academic staff (full-time plus part-
time) employed in universities and other further education institutions during
the 1976-1977 academic year. The sources of data on employment are:
The International Association of Universities (1978) International

Handbook of Universities, (7th edition), The Macmillan Press, London;
Association of Commonwealth Universities (1978) Commonwealth
University Yearbook 1979, (55 th edition) The Association of
Commonwealth Universities, London;
The World of Learning 1978-1979 (29 th edition), Europa Publications,
London.
Due to the different national education systems, it has been necessary to

identify comparable institutions on the basis of:
The British Council (1990) International Guide to Qualifications in

Education, (2nd edition), Mansell Publishing, London.
9.A3 Degree of Industrial Specialisation
The industrial specialisation index was calculated on the basis of data on

employment for nine industrial NACE classes. In particular, the employment
in each regional sector is expressed as a percentage of the total industrial

employment in the region. After having ranked the sectors by size, the index
of regional specialisation was calculated as the ratio between the average
percentage of employment for the smallest four regional sectors over the
average percentage of employment in the largest four ones. The index,
therefore ranges between 0 and 1, these two extremes indicating respectively
specialisation and diversity in the regional industrial structures. The data on
employment 1980 were derived from the REGIO Databank (Eurostat). In the
case of Greece and Portugal this source has been complemented by the
respective national statistical offices. The 9 sectors considered in the analysis
are:
Table 9.A.l Index of industrial specialisation: NACE classes
Sector denomination NACE classes

Energy and Water I
2 Extraction and Preparation of Metalliferous Ores 21
Production and Preliminary Processing of Metals 22
Extraction of Minerals other than Metalliferous and Energy- 23
producing minerals, Peat Extraction 24
Manufacture of Non-Metallic Mineral Products
3 Chemicals 25
Man-Made Fibres 26
4 Metals Manufacture: Mechanical, Electrical and Instrument 3
Engineering
5 Food, Drink and Tobacco 41,42
6 Textiles 43
Leather and Leather Goods 44
Footwear and Clothing 45
Processing of Rubber and Plastics 48
Other Manufacturing Industries 49
7 Timber and Wooden Furniture 46
8 Manufacture of Paper and Paper Products, Printing and Publishing 47
9 Building and Civil Engineering 5
REFERENCES
Abramovitz, M. (1986) Catching Up, Forging Ahead, and Falling Behind. Journal of
Economic History 46: 385-406
Acs, Z.1., Audretsch, 0.8., Feldman, M.P. (1992) Real Effects of Academic Research:
Comment. American Economic Review 82: 363-67
Acs, Z.1.,Audretsch, 0.8., Feldman, M.P. (1994) R&D Spillovers and Recipient Firm Size.
Review ofEconomics and Statistics 100: 336-40
Anselin L. (1988) Spatial Econometrics: Methods and Models. Kluwer Academic Publishers,
London
218 Magrini S.
Anselin L. (1994) Spacestat Tutorial. Morgantown Regional Research Institute, West

Virginia University
Audretsch, D.B., Feldman M.P. (1996) R&D Spillovers and the Geography ofInnovation and
Production. American Economic Review 86: 630-40
Andersson A.E., Anderstig, C., Harsman B. (1990) Knowledge and Communications
Infrastructure and Regional Economic Change. Regional Science and Urban Economics,
20: 359-76
Armstrong, H.W. (1995a) Convergence among Regions of the European Union, 1950-1990.
Papers in Regional Science 74: 143-52
Armstrong, H.W. (1995b) "An Appraisal of the Evidence from Cross-Sectional Analysis of
the Regional Growth Process within the European Union." In Convergence and
Divergence Among European Regions, Vickerman, R.W., Armstrong, H.W., eds, Pion,
London
Barro, R.J. (1991) Economic Growth in a Cross Section of Countries. Quarterly Journal of
Economics 106: 407-501
Barro, R.J., Sala-i-Martin, X.x. (1991) Convergence across States and Regions. Brookings
Papers on Economic Activity 1: 107-82
Barro, R.J., Sala-i-Martin, X.X (1992) Convergence. Journal of Political Economy 100: 223-
51
Barro, R.J., Sala-i-Martin, X.x. (1995) Economic Growth. McGraw-Hili, London
Batten D.F. (1995) Network Cities: Creative Urban Agglomerations for the 21 51 Century.
Urban Studies 32: 313-327
Burridge, P., Gordon, 1. (1981) Unemployment in the British Metropolitan Labour Areas.
Oxford Economic Papers 33: 274-97
Cheshire, P.e. (1979) Inner Areas as Spatial Labour Market: a Critique of the Inner Area
Studies. Urban Studies 16: 29-43
Cheshire, P.e., Carbonaro, G. (1995) "Convergence-Divergence in Regional Growth Rates:
An Empty Black Box?" In Convergence and Divergence Among European Regions.
Vickerman, R.W., Armstrong, H.W., eds, Pion, London
Cheshire, P.e., Carbonaro, G. (1996) Urban Economic Growth in Europe: Testing Theory and
Policy Prescriptions. Urban Studies 33: 1111-28
Cheshire, P.e., Furtado, AM., Magrini, S. (1996) "Analysis of European Cities and Regions:
Problems of Quantitative Comparison." In Cross-National Research Methods in the Social
Sciences. Hantrais, L., Mangen, S., Pinter, London
Cheshire, P.e., Hay, D.G. (1989) Urban Problems in Western Europe: an Economic Analysis.
Unwin Hyman, London
Coe, D.T., Helpman, E., Hoffmaister, AW. (1997) North-South R&D Spillovers. The
Economic Journal 107: 134-49
Evans, AW., Richardson, R. (1981) Urban Employment: Interpretation and Additional
Evidence. Scottish Journal of Political Economy 282: 107-24
Feldman, M. (1994) The Geography ofInnovation, Kluwer Academic Publishers, Boston
Fagerberg, 1. (1987) A Technology Gap Approach to Why Growth Rates Differ. Research
Policy 16: 87-99
Fagerberg, 1. (1988) "Why Growth Rates Differ." In Technical Change and Economic
Theory. Dosi, G., Freeman, C., Nelson, R., Silverberg, G., Soete, L., eds., Pinter, London
Fagerberg, 1. (1994) Technology and International Differences in Growth Rates. Journal of
Freeman, C. (1987) Technology Policy and Economic Performance: Lessons From Japan.
Frances Pinter, London
Gerschenkron, A. (1962) Economic Backwardness in Historical Perspective. Belknap,

Cambridge Mass
Gordon, I., Lamont, D. (1982) A model of Labour-market Interdependencies in the London
Region. Environment and Planning A 14: 238-64
Gordon, I. (1985) The Cyclical Sensitivity of Regional Employment and Unemployment
Differentials. Regional Studies 19: 95-109
Hall, P., Hay, D.G. (1980) Growth Centres in the European Urban System. Heinemann,
London
Holtz-Eakin, D. (1992) Solow and the States: Capital Accumulation, Productivity, and
Economic Growth. NBER Working Paper No. 4144
Lucas, R.E. (1988) On the Mechanics of Economic Growth. Journal of Monetary Economics,
22: 3-42
Magrini, S. (1997) Spatial Concentration in Research and Regional Income Disparities in a
Decentralised Model of Endogenous Growth. Research Paper in Environmental and
Spatial Analysis No.43, London School of Economics
Magrini, S. (1998) Modelling Regional Economic Growth: The Role of Human Capital and
Innovation. London School of Economics, Unpublished Ph.D. Thesis
Magrini, S. (1999) The Evolution ofIncome Disparities among the Regions of the European
Union. Regional Science and Urban Economics 29: 257-281
Mas, M., Perez, F., Uriel, E., Maudos, I. (1995) "Growth and Convergence in the Spanish
Provinces." In Convergence and Divergence Among European Regions, Vickerman, R.W.,
Armstrong, H.W. eds, Pion, London
Nelson, R.R., Rosenberg, N. (1993) "Technical Innovation and National Systems." In
National Innovation Systems: A Comparative Analysis, Nelson, R.R., ed, Oxford
University Press, Oxford
Patel, P., Pavitt, K.L.R. (1994) National Innovation Systems: Why They Are Important, and
How They May be Measured and Compared. Economics of Innovation and New
Technology 3: 77-95
Pavitt, K., Soete, L.G. (1982) "International Differences in Economic Growth and the
International Location of Innovation." In Emerging Technologies: Consequences for
Economic Growth, Structural Change, and Employment, Giersch, H., ed, I.C.B. Mohr,
Tubingen
Rivera-Batiz, L.A., Romer, P. (1991 a) Economic Integration and Endogenous Growth.
Quarterly Journal of Economics 106: 531-56
Rivera-Batiz, L.A., Romer, P. (l99Ib) International Trade with Endogenous Technological
Change. European Economic Review 35: 971-1004
Rivera-Batiz, L.A., Xie, D. (1993) Integration among Unequals. Regional Science and Urban
Economics 23: 337-54
Romer, P.M. (1990a) Capital, Labour and Productivity. Brookings Papers: Microeconomics,
337-67
Romer, P.M. (1990b) Endogenous Technological Change. Journal of Political Economy 98:
71-102
Romer, P.M. (1993) Idea Gaps and Object Gaps in Economic Development. Journal of
Monetary Economics 32: 543-73
Romer, P.M. (1994) The Origins of Endogenous Growth. Journal of Economic Perspectives
8: 3-22
Sala-i-Martin, X.x. (1996) Regional Cohesion: Evidence and Theories of Regional Growth
and Convergenc.e. European Economic Review 40: 1325-52
Chapter 10
Is the Real Exchange Rate Relevant for Determining

the Innovation Potential of European Regions?
Dino Martellato
University Ca' Foscari, Venice, Italy
10.1 INTRODUCTION
In a well known paper on foreign investment, Richard Caves (1982) gave

a definition of Foreign Direct Investment (FDI) and a classification of the
different forms which this can take. Both are still relevant today. According
to Caves:
"A firm's expansion into a new, geographically segregated production
facility can take any of three forms: horizontal extension (producing the
same goods as elsewhere), vertical extension (adding a stage in the
production process that comes earlier or later than the firm's principal
processing activity), or conglomerate diversification."
This definition focuses on an aspect of location which is implicit in any
FDI decision. An equally important point to note, however, is that FDI
inevitably involves not only a transfer of capital, but also a transfer to the
foreign subsidiary of technology, know-how, entrepreneurship and
managerial capacities. It is possible then that the transferred knowledge will
spill over from the foreign subsidiary into the surrounding region. Needless
to say, in this instance, the knowledge spillover may directly impinge in a
positive way on the economic performance of the receiving area, but may
well erode the technological lead of the source.
For an economist, the fundamental questions which come to mind in this
context are: why and when? That is, why do certain firms and enterprises at
a certain point in time decide to set up a foreign subsidiary and become
multinational? And when should a growing firm set up a foreign subsidiary?
222 Martellato D.
The answers to these questions bear essentially on the same arguments and
depart considerably from the early analyses of FDI.
It is recognised that a large number of economic factors are involved in
the decision to undertake FDI. In the explanations, a distinction can usually
be made between those based on theories assuming perfect markets and
theories which assume some degree of market imperfection. But whatever
the nature of the market, it would seem fair to say that the most common
factor behind FDI is the search for a cheaper - if not the cheapest - location
or way of serving the market (whether the foreign market, the home market
or the general market).
Here, I intend to focus on one single factor, namely the real exchange
rate, which was already considered by Aliber (1970, 1971) to be a powerful
factor behind a great many FDI decisions. However, the risk connected with
exchange rate fluctuations can be covered by hedging, so why should we
bother with this aspect? The point is that the exchange rate risk can be
hedged only partially. Real exchange rate deviations from the purchasing
parity equilibrium are considerable and appear to be one of the most
important factors influencing the expected revenue in many FDI initiatives.
Indeed, the role of real exchange rates has been revised more recently by
Caves (1988), Froot and Stein (1991) and others.
More specifically, I wish to consider the role of exchange rate for the
following reasons:
it is a fact that real exchange rates are not as stable as generally thought;
by managing nominal exchange rates, a country can manage its real
exchange rate.
a higher level of real exchange rate may be very positive for some
exporting sectors.
the current and expected level of the exchange rate is relevant for many
types of capital transfer and thus relevant for technology transfer.
The divergence between the current level of the real exchange rate of a
currency and its level in equilibrium is directly related to the strength of the
currency.
The nominal exchange rate is a strategic factor in the determination of the
competitiveness of a region (directly, through its effect on the real exchange
rate and indirectly, through FDI and innovation, the exchange rate drives the
competitive position of any region).
While the first three points do not need any special comment here, the
remaining three deserve some explanations. They wiIl be dealt with in the
following sections, but only after having considered the relationship between
FDI and the possibility of technology spillovers.
Is the Real Exchange Rate Relevant for Determining Innovation Potential? 223
10.2 FDI AND TECHNOLOGY SPILLOVERS
Technology spillovers can be horizontal or vertical. The former occurs

when the technological novelty is copied by a firm that is in competition
with the one responsible for the innovation. Vertical spillovers, on the other
hand, occur when a firm transfers technology to other firms that sell
production inputs, thus enabling its customers to lower their production costs
or to increase the variety of their products.
There is considerable evidence to show that technology spillovers tend
generally to be local. Silicon Valley, Route 128 in the Boston area and other
industrial districts provide ample illustrations of how knowledge capital,
being to some extent incorporated in trained people, tends to be location
specific. However, it has also been demonstrated that trade and capital
transfers cannot easily be separated from the transmission of knowledge.
Grossman and Helpman (1991) have shown how trade makes the diffusion
of knowledge possible. The combination of location specific spillovers and
long distance capital transfer can generate a leap-frog pattern of growth.
Some earlier studies! have recognised that the relationship between trade
and innovation goes both ways. On one hand, the technological lag between
advanced and less advanced economies gives rise to trade2, on the other,
wage differentials induce a continuous transfer of technology from advanced
to less advanced economies. The transfer which occurs when mature
products are left to less advanced economies is often backed by the
appearance of new products in the advanced economies. The advantage of
these latter economies thus appear to be in their capacity to offer a variety of
new products - which implies a quasi-rent and higher wages - rather than in
the higher efficiency of its labour force. The advanced economies however
are obliged to constantly innovate, because their monopoly power is
progressively eroded by the technology transfer.
Paul Krugman suggests two conclusions. In the first instance, he
identifies a clear link between technological change and capital movement:
"The region experiencing the most rapid technological advance will also
experience capital inflow" (Krugman 1979, p. 264)
The causation however runs from technological change to capital
movement, not vice versa. Technological change raises profits, thus
providing an incentive for FDI. Advanced economies therefore can maintain
their edge and higher wage levels by developing new industries and selling
new products, rather than by seeking to increase the productivity of labour in
Most notably that of Krugman (1979).

A fact which is by no means in contradiction with the strong trade relations between
countries at the same growth stage.
224 Martellato D.
old and mature products. From a certain point of view, Krugman's model is
a development of that presented in a celebrated article written by Dornbusch,
et al (1977) twenty years ago. Using a Ricardian model, in which goods can
be produced anywhere by labour only, the authors show how tastes,
technology and size jointly determine the equilibrium wage and the efficient
production specialisation of different countries. Their model takes into
account a technology transfer, but no new products. In additional, it is
assumed that income, but not capital, may be transferred from one country to
another. Krugman's contribution therefore provides those elements of
analysis which are lacking in the Dornbusch-Fisher-Samuelson model, i.e.
new products, relocation, capital transfer and technological diffusion.
In a recent research report, it was pointed out by the World Bank (1997)
that in the 1990s FDI has been used not so much a way for multinational
corporations to indulge in 'tariff jumping', as a way of competing. Currently,
FDI benefits in particular from factors such as:
the lifting of restrictions

falling transport costs
falling communications costs
cross-border mergers and acquisitions
privatisation
strong macro-economic fundamentals
a sound enabling environment (divergence between private and social
returns).
While it has been demonstrated that FDI has positive effects

(productivity, product composition, human capital, additional domestic
investments) all of which boost economic growth, it has not been
conclusively proven that technological spillovers always occur. However,
FDI has very evident effects on trade flows. First of all, trade is occurring
increasingly in the nature of intrafirm exchange. Secondly, FDI flows are
highly correlated with trade flows. Thirdly, it has been shown3 that
developing countries that have strong links with DECD countries tend to
have higher productivity levels. In conclusion, although it is clear that FDI,
trade flows and productivity are highly correlated, it is difficult to establish
whether the productivity effect and technology spillover are more affected
by FDI or by trade.
Later in this chapter, in Section 10.4, I propose a model where the link
between trade and FDI is explicit and very strong: region cannot have a
stable inflow ofFDI without a comparable trade surplus.
3 I refer to Coe et at (1995).

10.3 FDI AND THE STRENGTH OF CURRENCIES
There is no doubt that FDI is a phenomenon which defies the pedagogic

distinction between old and new products which has been used so often in
relation to the product life cycle theory. Both new and old products can
indeed be moved as exports from one country to another, entirely or
partially, sometimes as a substitute, sometimes not.
To give at least a partial assessment of the general impact of real
exchange rate movements on FDI and technology transfer, it is useful to bear
in mind the basic distinction between portfolio and foreign direct
investments. While portfolio investments pursue a purely financial return
and are devoid of any interest for corporate control, FDI is undertaken for
the purpose of better corporate performance, and involves the birth of a
foreign subsidiary as a production unit which is distinct from the parent firm.
The exchange rate exerts a different impact on the two types of capital
transfer. Let us consider, for instance, an expected nominal appreciation of
the currency40f the home country. This could attract portfolio investments
hoping for an appreciation of the currency in which the claims are
denominated and which could therefore add to the interest payments. But, at
the same time, if the nominal appreciation is not followed by a proportional
relative price contraction, the real exchange rate decreases proportionally5. If
the higher valuation in real terms lasts for a sufficient period of time, this
could encourage a capital outflow, which translates into direct investment
abroad and particularly where an opposite situation is found. It is obvious
that other factors may well be in operation - any microeconomic analysis of
the FDI would not fail to take them into consideration - but here I wish to
focus only on the exchange effect and the macroeconomic fundamentals.
Why then should an enterprise decide to undertake FDI on the basis of
the expected exchange rate change only? Probably the first scholar to give an
answer to this question was Aliber (1970,1971). With the following basic
interpretation. Firms located in countries with a strong currency6 have an
advantage in making FDI in countries with a weak currency. However, this
advantage is provided by a supposed bias in the capital market. According to
Aliber, the market is more likely to capitalises the income stream generated
in the weak currency country if it is acquired by a firm located in a strong
currency country, where the exchange rate is not expected to revert rapidly
to its equilibrium level. Hence the conclusion that strong currency countries
export their capital towards weak currency countries.
4 The nominal exchange rate of course decreases.

The currency appreciates in real terms.
Strong, in this instance, means an overvalued currency.
226 Martellato D.
But if we take into consideration the existence of cross-haulings, which

clearly cannot be explained in terms of exchange rates, this means that
countries with a highly valued currency tend to be the origin of net capital
transfers, while countries with a low valued currency are the destination of
net transfers. Aliber's proposition yields the conclusion that the
overvaluation of a currency, other things being equal, implies outward FOI,
while undervaluation is associated with inward FOL It would therefore seem
helpful to briefly focus on the concept of over and under-evaluation, and the
related concept of currency strength.
When a currency is overvalued, and its real exchange rate is below the
equilibrium level, the nominal exchange rate does not correctly reflect the
relative prices of exportable goods. In common parlance, this currency is
strong. However things could be very differene. If an increase in the
nominal rate of the currency is expected, then the same currency is seen as
weak and there is a structural tendency to devaluate it. Mutatis mutandis, a
currency is strong when it is undervalued in terms of relative prices and there
are expectations of a revaluation, not vice versa.
It is also helpful to understand how the real exchange rate could interfere
with capital transfer in general, and FOI in particular. In attempting to
explain this, I adopt the approach suggested by Caves (1971, 1982), since it
seems to me the most appropriate. His approach is essentially based on a
classification of investment types which allows us to formulate some
reasonable assumptions not only about the composition of the subsidiary's
cash flow, and its exposure in terms of exchange rate risk, but also the
effects of a change in the exchange rate.
Taking up once again the definition given by Caves at the beginning of
the chapter, FOI involves an extension of production abroad which can take
three different forms: horizontal extension, vertical extension and
conglomerate diversification.
A. Horizontal extension may be the outcome of two different conditions
(i) In the first case, the creation of a foreign subsidiary implies export
substitution, which may occur in two situations:
because the parent firm wants the production of its foreign subsidiary to
replace current domestic production. If the outside production is for the
outside market, FOI leads to a reduction in exports, i.e. a trade
contraction. However, a revaluation of the home currency would also
imply a reduction in exports, thus FOI is a way of keeping the market
share;
Fitoussi (1995) recently adopted another view.

if the outside production is both for the domestic market and the general
market, there is a switch from exports to imports. This leads, at least
partially, to a case of trade diversion. An increase in the real exchange
rate in the country where the foreign subsidiary has been located (or a
revaluation of the home currency, which is equivalent) yields a
reduction in production costs (imports) and hence higher profits.
(ii)ln the second case, the creation of a foreign subsidiary does not imply
that there is export substitution. This case of horizontal extension is
explained by the pursuit of profits by direct penetration in a foreign and,
arguably, growing market. In this situation, an increase of the real
exchange rate of the country where the foreign subsidiary has been
located could imply a reduction in profits, as the market could be best
served with imports from outside.
B. Vertical extension may also be of two types
(i) A backward extension occurs when the early stages of the production
process are decentralised. In this case imports of intermediate products
increase and the outflow of foreign currency increases, but a devaluation
of the foreign currency, as before, implies lower costs and thus higher
profits for the parent firm.
(ii)A forward extension occurs when the end stages of the production
process are decentralised. In this case, the parent firm sells intermediate
products at a lower price, thus increasing profits in the foreign subsidiary.
A devaluation of the currency would imply higher profits for the parent
firm, but lower ones for the subsidiary.
c. The case of conglomerate diversification IS similar, in terms of its

implications, to the last case cited above.
It is therefore evident that a permanent gap between the real exchange

rate of the foreign currency and its equilibrium level is able to influence FDI
in almost all the cases contemplated here. This suggests that it would be
worthwhile further investigating the role of exchange rate in relation to FDI
and net exports.
10.4 THE TRADE BALANCE AND CAPITAL

TRANSFER
In the preceding analysis it has been pointed out that there are many
factors which impinge on FDI, but it has also been argued that the current
228 Martellato D.
and expected levels of the real exchange rate are able to determine the rate of
return on investments and thus influence many types of FOI. It has also been
argued that capital transfer has an impact on a region's innovation potential,
although it is not clear if technology spillovers are necessarily an outcome of
FOI. Furthermore, it is clear that trade flows and FDI are correlated. In this
section, I wish to show that a sustainable condition for a capital importing
region implies a trade surplus. Accordingly, we investigate the macro-
economic conditions which tend to favour both a trade surplus and a foreign
capital inflow.
As the region's trade balance can be explained, other things being
constant, by its competitive position, i.e. by the real rate of exchange and the
innovative content of its products, we need to focus on the relationship
between the trade balance and the net debt position of the region, as it is
obvious that this will be modified by FOI. Particular attention is therefore
given to the relationship between trade balance and capital mobility. To do
so, we adopt a very simple model that has also been used to highlight the
regional impact of the EMUs and which directly involves the relationship
between trade and debt. The resulting equation has some interesting
properties. First of all, it illustrates the role of the rate of return and that of
the real exchange rate in determining the long run or sustainable ratio
between the trade balance and the net financial position. Furthermore, it
shows quite clearly that the trade balance and the region's debt position are
cyclical in nature.
There is an obvious link between the investment/savings gap and the
change in the stock of net foreign assets owned by an economic agent. When
personal savings are insufficient, the agent must raise extra funds. This
constraint holds true for whatever set of agents (whether these are found in a
region or in a country). It reads:
(10.1)
where (t) is the year of reference. If the savings gap is permanent, the
regional net debt will be piling up. Let us assume that this condition holds
for a while and that the region's net debt position grows at an exponential
rate equal to (g):
(10.2)
See: Martellato (1999).

A second relationship which must be considered is the balance of outside

payments according to which the change in the net foreign position is equal
to the current account:
(10.3)
where r is the average rate of interest and the expression in brackets is the
trade balance. By combining the last two equations, we obtain the most basic
relationship of this section, that is:
(10.4)
This relationship makes the growth rate (g), the interest rate (r) and the
trade balance (M - X)I mutually consistent for a given net stock of external
debt (DI). From this equation it is easy to derive a relationship in which the
rate of change (g) or the stock (D1+1) is related to the real exchange rate and
the rate of interest.
An equivalent equation says that the net stock of external debt is equal to
the current one, conveniently capitalised, plus the trade balance:
(10.5)
To proceed, we need a suitable specification of factors which may be

able to influence the trade balance. If the region were a nation with its own
currency, it would have its nominal exchange rate (a). We could then
assume that nominal regional exports and imports were linked to the real
exchange rate:
e = aP' / P (10.6)
where the external and internal price indices of exportable goods are
respectively ( P * ) and (P). These also depend on external demand Y * and
internal demand Y in a standard way:
with a > 0, It > 0 (10.7)
with b < 0, J1 > 0 (10.8)

230 Martellato D.
The combined price elasticity is assumed to be positive in order to make

sure that a devaluation has a positive effect on the trade balance, i.e.:
e=a-b-l>O (10.9)
Simple substitution and rearrangement of terms gives:
g=r+ D (M) [1- (a pp*)f: Y*A]

t y,u (10.10)
This equation gives the feasible current growth rate of the net debt, given
the average rate of return (r), the stock (D) and the current trade balance. It
can be used to describe a regional cycle of trade and capital transfer. The
stock of net external debt (credit) may be at any of the points of the straight
line in Fig. 10.1, but not all these points are equivalent from an economic
point of view. In the figure there are three pairs of symmetric conditions
which we shall now comment upon.
87
f6\.
Sustainable
conditions

Unsustainable
conditions
Figure J O. J Sustainable and unsustainable debt and trade conditions

In CASE 1, a net external debt is associated with a trade deficit (in the
symmetrical CASE 6, a net external surplus is associated with a trade
surplus). Clearly, both these conditions cannot be sustained for long.
In CASE 2, the net external debt is increasing because a trade surplus
only partially offsets negative interest payments. In CASE 5, a net external
credit position is paired by a trade deficit and positive interest payments
again with only partial compensation.
CASE 3 represents that of a region where the net debt decreases as the
trade surplus outperforms the negative interest outlays. Case 4 is
symmetrical, in that the stock of net credit decreases over time as the trade
deficit outperforms the positive interest payments.
The dynamics of the stock of debt is given by equation (10.11) slightly
reformulated as:
Dt+1 P*J& y*,A.]

= (1+r)Dt +Mt [1- (a p y.u (10.11)
A more complete way of looking at the same cycle is shown in Fig. 10.2
where the trade deficit and the stock of net debt are simply plotted against
time. The figure indicates all six cases, but some of them (where the curves
cross) have to be considered non sustainable since they may be maintained
for very short time periods only. The system is able to move from one stage
to the next through a suitable change in the rate of return on capital and the
real exchange rate.
The first conclusion that can be drawn from the preceding analysis is that
a region can modify its competitive position in many ways, but - in the
present instance - the main ways would be a) by increasing the rate of return
of capital or b) by increasing the real exchange rate (here I refer to the
expression within the inner brackets of the last two equations). This latter
can be done in two ways: by undertaking a competitive devaluation
(increasing a) or through competitive deflation (i.e. by controlling
production costs and increasing export prices, but by a lesser amount).
It is interesting to note that the first alternative, a), is by no means in
conflict with either of the last two alternatives, but that these last two may be
in conflict with each other. There is however another possibility. This
implies a squeeze on domestic demand and employment, but - at least in
theory - no government would be keen to deliberately pursue it.
More in general, the conclusion is that a region has to improve its macro-
economic fundamentals as well as the internal micro-economic conditions
(e.g. infrastructure and human capital supply, social environment, etc.).
However, since R&D activities are capital intensive and probably have
232 Martellato D.
strong scale economies, small countries and regions may not be well suited
to competing in research activities and would possibly do better limiting
themselves to importing new technologies from outside. In this case,
innovation may not be the result of domestic R&D investments, but the
consequence of strong macro-economic conditions, if these favour an inflow
ofFDI.
The argument we tried to develop in the first part of the chapter was that
by encouraging inward FDI, and by using a suitable combination of the three
alternatives mentioned above 9. a region can benefit from technology
spillovers, faster innovation and increased capacity in the industries with
exportable goods. A further striking conclusion is that by entering EMU, a
region is obliged to exclude exchange manoeuvres and to rely on the other
two policy options a) or b), as well as internal demand restraint.
eo
,l 60
I
I
" l
f I
jl .~~~~[JJIIIII~JI.II~III'1 I;])
20
~
11"1 o ~
20

-40
Peiods
- Trade balance ~ Stock ri net det:t
Figure 10.2 Trade balance and net debt
9 Alternative a) is by no means in conflict with alternative b). A high exchange rate may
sustain the investment return and encourage the inflow of capital.
REFERENCES
Aliber, R.Z. (1970) "A Theory of Direct Foreign Investment." In The International
Corporation: A Symposium. Kindleberger, C.P., ed, The MIT Press, Cambridge Mass
Aliber, R.Z. (1971) The Multinational Enterprise in a Multiple Currency World. In The
Multinational Enterprise, Dunning, 1.H., ed, The MIT Press, Cambridge Mass
Caves, R.E. (1971) International Corporations: The Industrial Economics of Foreign
Investment. Economica 38: 1-27
Caves, R.E. (1982) Multinational Enterprise and Economic Analysis. Cambridge University
Press, Cambridge
Caves, R.E. (1988) Exchange Rate Movements and Foreign Direct Investment in the United
States; Harvard Institute of Economic Research, Discussion Paper 1383, Harvard
University
Coe, D., Helpman, E., Hoffmaister, A. (1995) North-South R&D Spillovers. Discussion Paper
1133, Centre for Economic Policy Research, London
Dornbusch, R., Fischer, S., Samuelson, P.A. (1977) Comparative Advantage, Trade, and
Payments in a Ricardian Model with a Continuum of Goods. American Economic Review
67 (5): 823-39
Fitoussi, J.-P. (1995) Le Debat Interdit, Monnaie, Europe, Pauvrete. Editions Arlea, Paris
Froot, K.A., Stein, 1., Stein, C. (1991) Exchange Rates and Foreign Direct Investments: An
Imperfect Capital Markets Approach. Quarterly Journal of Economics 106: 1191-217
Grossman, G.M., Helpman, E. (1991) Innovation and Growth in the Global Economy. The
MIT Press, Cambridge Mass
Krugman, P. (1979) A Model oflnnovation, Technology Transfer, and the World Distribution
of Income. Journal of Political Economy 87 (2): 253-66
Martellato, D. (1999) "Trade and Investments in the EMU Regions." In Spatial Dynamics of
European Integration. Regional Policy Issues at the Turn of the Century. Fischer, M.M.
and Nijkamp, P., eds, Springer, Berlin
World Bank (1997) Global Development Finance, Washington DC
Section D:
Impacts of Innovations in Telecommunications and

Transport
Chapter 11
Impact of the New Information Technologies on

Economic-Spatial Systems: Towards an Agenda for
Future Research
Cristoforo S. Bertuglia 1 and Sylvie Occelli 2

I Turin Polytechnic, Turin, Italy
2IRES-Piemonte, Turin, Italy
11.1 INTRODUCTION
This chapter takes up some of the themes presented in the introductory

chapter, and aims to put forward an 'agenda' for a research programme to
investigate the relationship between innovation and urban development.
Besides contributing to a better understanding of these relationships, an
agenda of this kind could help focus attention on those questions which
should be given priority in addressing the issue.
It is a challenging topic of research which has in fact been much debated
in the literature for several decades, attracting the interest of scientists,
practitioners and politicians (among the numerous books published in the
nineties, we should like to mention Graham and Dominy 1991; Brotchie et al
1991; Robins 1992; Andersson et al1993a; Batten et al1995; Bertuglia et al
1996; IRES 1996; Bertuglia et al 1997). In investigations of the spatial
impact of New Information Technologies (NIT) on cities and regions, i.e.
studies taking a specifically geographical point of view, three fundamental
questions have come to the fore (see, among many others, Giaoutzi 1990;
Robins and Gillespie 1992):
1. what kind of 'restructuring' can be expected to result from the adoption

of NIT? In other words, what role is NIT likely to play within the current
social, cultural, political and economic changes which urban systems are
currently undergoing?;
238 Bertuglia C.S., Occelli S.
2. what outcome is likely for the different types of area (i.e. different kinds
of region) affected by the above processes?;
3. what kind of phenomena should be observed and studied in a 'spatial'
analysis of NIT?
However, in the present chapter, rather than look in detail at these

questions, we take a step back and focus on a preliminary issue, the concept
of innovation. Our discussion is divided into two main parts. Firstly, we
attempt to explain why some of the major aspects of novelty nowadays
concern the very concept of innovation. Secondly, on the basis of these
arguments, we suggest a possible extension of the conventional notion of
innovation. Some implications of this extension for the formulation of the
research agenda will also be mentioned.
11.2 THE CONCEPT OF INNOVATION: ITS

TWOFOLD NATURE
There is no single or commonly agreed definition of innovation.

Historically, it has been defined by means of various formulations which
have usually made reference to other more familiar concepts, such as
discovery, invention or creativity, which were felt to bear a resemblance to
it, though not overlapping precisely (Brown 1981). Whether referring to an
artefact, an idea or a situation, the concept of innovation has always raised
questions about the definition of 'how new is new?'
The investigation of this question entails the observation and
interpretation of the innovations themselves and also their impacts on the
various components of society. Two major topics of interest are: a) the
assessment of the role played by innovation in the evolution of urban and
regional systems and b) the nature of its manifestations in the spatial
evolution process, i.e. those processes connected with the fact that an
innovation often does not reveal itself immediately, but seems to be highly
selective in terms of the timing of its appearance, the place where it occurs
and the people involved, showing changing characteristics as it becomes
diffused in space and time (Brown 1981).
The approach in the literature to the fundamental questions associated
with the concept of innovation has tended to adopt two distinct, although
overlapping, perspectives:
a. a systemic perspective, in which the main focus is the outline of a

coherent framework against which innovation is analysed and assessed.
This usually entails making reference to some analytical dimensions,
Impact of the New Information Technologies 239
consisting of either temporal dimensions or spatial scales or both (i.e.

push-pull based explanations). On a more theoretical level, such a
framework can also lead to more general questions concerning the micro-
macro relationships which rule the system;
b. a substantive perspective, whose main purpose is to investigate in more
depth both the nature and components of innovation. This typically
involves a distinction between the 'result' of innovation, be it a social
outcome or an institutional change, a new artefact or market product, and
the 'process' by which it was achieved, i.e. how, with what resources and
in what time period the result is obtained.
Despite being implicit in the mainstream of innovation studies, the above

perspectives have seldom been acknowledged. They are nevertheless at the
basis of the various categories of theoretical thinking about innovation
diffusion and innovative processes. The systemic perspective, for example, is
clearly embedded in the examination of general types of innovation which
can be assessed through their impacts on the socio-economic environment.
According to this view we can distinguish (Brown 1981):
a. continuous innovations, defined as innovations which have little

influence on the established consumption pattern;
b. dynamically continuous innovations, which are more influential than the
preceding category and involve the creation of new products or the
modification of existing ones;
c. discontinuous innovations, which involve the establishment of new
consumption patterns and new products.
Kondratiev related the various stages or waves of development of modern

societies over historical time to the appearance of new technologies (see
Table 11.1). According to this interpretation, the path of evolution of modern
societies can be subdivided into periods identified by the appearance of a
given set of innovations whose introduction occurs in a wave-like fashion.
During each period, a specific bundle of innovations develops (the upswing
part of the diffusion curve) and is then spread more and more widely until
replaced by a new bundle (the downswing part of the curve). This bundle
then becomes predominant, marking the beginning of a new stage of
development. The process is accompanied by (and interacts with) other
social and economic changes which jointly determine the distinctive
descriptive profile for the period.
Table 11.1 Main characteristics of technological innovation according to

long wave analysis (after Freeman and Perez 1988)
I 1/ III III IV
Period 178011850 185011890 189011940 194011980 Since 1980
Main features Mechanisation Steam engines Electrical and Fordist mass Information
heavy production and tele-
engineering communica-
tions
Main sectors Textiles, Steam Electrical Cars, Computers,
iron-working, engines, machinery, armaments, capital goods,
water power steamships steel ships consumer telecomm-
durables, unications,
petro- optical fibres
chemicals
Infrastructure Canals, roads Railways, Electrical Highways, Digital
world cables airlines, communica-
shipping airports tions,
networks,
satellites
The substantive perspective has prevailed, on the other hand, in studies

whose main aim has been to assess the innovative performance of economic
sectors. From consideration of 'market' determinants, both internal and
external to the firm, characterising the innovative behaviour of firms (e.g. the
relative weights of product and process innovation, the size of the firm, the
intensity and type of technological diversification, and the source of
innovation), a typology of innovative behaviour can be identified and
economic sectors classified accordingly. Gaffard (1990), for example,
related the various categories of firms with a definition of the kind of
technological trajectory they were likely to follow. The typology consisted
of the following groups:
a. firms which are supplier dependent (i.e. dependent on other firms

producing equipment and intermediate goods). Innovation mostly
involves the process rather than the product and is achieved by means of
new machinery and processing techniques. It originates outside the sector
to which the firm belongs and the associated technological trajectory is
mainly oriented towards cost-cutting;
b. firms which are specialised suppliers. Innovation mainly concerns the
product, which then enters the other sectors as fixed capital inputs (i.e.
intermediate goods and specialised equipment). Innovation is thus
implemented by user firms (and in particular those producing durable
goods). The associated technological trajectory is mostly product design

oriented, aimed at refining the conception of the product;
c. firms which are mass producers (and usually large). Innovation is can
involve both product and process. It is often the outcome of a complex
research process which mainly takes place within the firm and uses
specialised suppliers. Innovation is essentially a technical artefact, and its
diffusion depends on the existence of dynamic learning economies and
patents. The associated technological trajectory is both cost cutting and
product design oriented;
d. firms which are science-based (and usually large). Innovation is linked
with the appearance of new scientific knowledge, conceived in R&D
research laboratories, and forms the basis of new technological
paradigms. It is mainly product innovation (a technical artefact) which
enters almost all sectors of the economy. Its diffusion requires patents,
learning economies and know-how. The associated technological
trajectory is mixed, including cost cutting and product design strategies.
Freeman and Perez (1988), posit a more detailed approach in which

elements of both the systemic and substantive perspectives are introduced.
They distinguish between: 1) incremental innovation, 2) radical innovation,
3) new technology systems and 4) changes in the techno-economic
paradigm. As shown in Table 11.2, the classification takes into account the
following aspects:
a. the type of innovation, which is defined according to the pervasiveness

and persistency of its impact;
b. the kind of economic sectors involved;
d. the factors characterising the environment which may foster or favour the
birth of innovation;
e. a general assessment of the expected outcome of the innovation;
d. the time necessary for an innovation to display itself.
The systemic perspective is perhaps most fully developed in the regional

taxonomy proposed by Steiner (1990) which using concepts derived from
the evolutionary approach. Here, the area types reflect the different abilities
(i.e. the performance) of regions in adjusting to change. These result from
the economic and cultural environments shaping the behaviour of firms, but
also depend on the stage of development of the firms themselves. Innovation
can happen anywhere, but the probability of its survival is not the same
everywhere. It is claimed that in socio-economic systems the variations and
mutations, which are key concepts in the biological interpretation of
evolution, are not completely chance events, since:
Table 11.2 A taxonomy of innovations (after Freeman and Perez 1988)
Type of events Sectors involved Enablingfactors Main effocts Time ofreali=ation
Efficiency
Demand pressures,
Continuousl y improvements,
Incremental socio-cultural factors,
distributed over ! All productivity gains, medium run
innovation I technological
ti me and sectors quality
opportunities
imDrovements
R&D activities
Structural changes,
Discontinuously in firms,
Radical limited and
distributed over universities, ns long-run
innovation localised economic
time and sectors government
impacts
laboratories -- ~-
Combination of Several branches

Changes of Organisational and Formation of
radical and of the economy,
'technology managerial clusters of long-run
incremental formation of
system' innovations innovations
innovations new sectors
Changes in Combination of Pervasively
Formation of new
'techno- radical and affecting many Deep social and
technological long-run
economic incremental branches of the institutional changes
regimes
oaradirzm' innovations economv
choice and imitation are conscious actions

skill and experience can be passed on, and anticipation of future
developments is possible.
For an area, then, the overall outcome of adjustments is determined by

the simultaneous survival of efficient and less efficient firms (not only the
fittest survive). Three types of region can be identified, as shown in Fig.
11.1. These are:
1. areas which can be considered as adaptable, as they possess the potential

resources to exploit the' creative' behaviour of firms;
2. areas which are regarded as adaptive, as they try to use external situations
in an optimal way, the allocative function being the main factors steering
behaviour;
3. areas which are non adapted, either because of an 'over reaction' or too
rapid change in the environment.
Although they provide valuable insights into the many facets of

innovation, the above taxonomies prove far from satisfactory when
confronted with the current dynamics of change in spatial systems (whether
concerning metropolitan or peripheral areas).
Legend
Adaptable areas Areas in which there is a high resource potential
and firms with creative (entrepreneurial) behaviour
Adaptive areas Areas dominated by standardised production in which
firms are sensitive to locational costs and factor prices
Non-adapted areas Areas with an insufficient adjustment performance
and stagnating firms
Relevant indicators Adaptable area Adaptive areas Non adaptable

area
Qualifications ofthe workforce high low
Percentage of blue collar workers low
Possibility of communications low
Diversification of industry high low
Rates of firm entry high
Wage levels low high
Female participation high
Turnover of firms high high
Dominance of large firms low high
Figure 11.1 Regional types in relation to their adjustment capacity

(after Steiner 1990)
But even the most recent literature, where the two analytical perspectives,
the systemic and the substantive, are being brought together to build sounder
approaches to the interpretation of the relationships between innovation and
urban development, i.e. what some authors call "the social construction of
technology" (Graham and Marvin 1996), there are difficulties with the very
concept of innovation as far as explanations of urban evolution are
concerned.
244 Bertuglia c.s., Occelli S.
We argue that these difficulties are due to the fact that the concept of
innovation - commonly understood in general terms as a 'thing' (i.e. a
product, an artefact, a technique, an idea or even a behaviour) which is new
because it is qualitatively different from the existing form - needs to be
extended and refined.
First of all, we need to recognise that an innovation is made up of two
main components, the 'hard' and the 'soft' part, which although closely
interrelated are nonetheless distinct:
a. the hard component is the tangible part of an innovation, and commonly

perceived as the 'thing' mentioned above. This is also the component of
innovation most usually referred to in the literature;
b. the soft component is the intangible part that is brought into existence by
the use of the innovation itself. The latter should not be confused with
what are generally termed the 'effects' produced by the introduction of
innovation, and which are extensively examined in the analysis of
innovation diffusion. This intangible part can be broadly defined as the
change in the interaction pattern, i.e. the man/machine interactions,
interpersonal contacts and man/environment interactions (Salomon 1986)
that, for an individual agent, is produced by using an innovation. In this
respect it bears a resemblance to a number of interpretations of
innovation already proposed in the past. From a cultural point of view,
the essence of such change lies in "the restructuring of the parts so that a
new pattern results, the distinctness of which cannot be characterised
simply as a variation of a number of its elements" (Barnett 1953). This
intangible part can also be linked with the notions of knowledge,
experience and learning, which some economic literature has recognised
as being incorporated in an innovation (i.e. necessary for creating an
innovation) (Metcalfe 1988). It certainly shares similar features with the
sociological view according to which innovation can occur, be adopted
and influence social systems, only when it involves the activation and
modification of circuits of communicative interaction (Mel a 1991).
Lastly, it can be considered to be embedded in what has been called the
'software network' of logistical networks, such as education, the arts and
science, which enhance human infrastructure (Andersson et al 1993b).
The above concept of innovation is very general and can be seen to hold
for the different kinds of innovation (including products, artefacts,
techniques, or simply ideas). It is, of course, particularly relevant for the so-
called New Information Technologies (NIT), i.e. those technologies derived
from the convergence of telecommunications and computing, which
characterise the 'new informational society' (Castells 1989). The point worth
emphasising here is that the soft component represents the most outstanding
aspect of innovation today. Although still largely unnoticed, it constitutes the
'live wire' connecting most of the wide-ranging issues in the current debate
on innovation.
The two components of innovation also have deep implications for the
way in which we interpret the nature of innovation and identify the agents
likely to trigger and/or diffuse innovation. As far as the nature of innovation
is concerned, it seems to be increasingly difficult, as noted by some authors
(see Barnett 1953), to put forward any explanation which depends
exclusively on the properties of the innovation itself, since these are context
dependent. In this respect, for example, the distinction between basic and
derived innovations becomes blurred and meaningless. Only when seen in
terms of its hard component can innovation be 'referenced' to its
characteristics, its spatial location and the timing of its appearance.
When the soft component is taken into account, the interaction patterns
come to the fore. As innovation implies re-organisation and reshaping, the
interaction patterns are changed through a whole range of behavioural (i.e.
goal seeking, explorative, imitative, hyper-selective) processes. These can
act at both the individual and collective level and affect spatial systems in a
number of different ways which all feed its evolution. In this respect, it
appears that innovation is a permanent feature (a process) of system
evolution. Moreover, everyone is likely to be an agent of innovation. By
identifying the ordinary agent, whether an individual, a household, an
organisation or any collective society, as the main repository of innovation, a
less elitist view is posited.
Finally, the awareness of the twofold components of innovation with
their corollary implications will also significantly affect our ways of
conceiving the relationships between innovation and urban development. It
is the ability to continuously adjust, modify and create new interpretations of
these relationships (i.e. the mental models of the future course of the world,
see Silverberg 1988; Andersson et at 1993b) which moves the agents'
actions and ultimately shapes the direction of urban development.
11.3 TOWARDS AN AGENDA FOR FUTURE

RESEARCH
Notwithstanding the growing literature on innovation (and in particular

the studies on NIT), the implications for urban evolution remain largely
unknown, mostly because empirical evidence is scarce and theoretical
thinking is still at a speculative stage. Any interpretation purporting to be
exhaustive would therefore be meaningless and naive. However, as that same
'soft component' of innovation also holds in interpreting the relationships

between innovation and urban development, an effort is made in this final
section to formulate some key questions which we consider are worth
addressing in future studies. The fundamental questions are:
How can the 'enabling' role of innovation (that endogenously determined

by individual actions, i.e. by firms, households or local institutions), be
facilitated in order to make the system opt for a more favourable
innovative path?
Can we identify particular properties of urban systems likely to promote
innovative behaviour?
To what extent will innovation, and in particular NIT, affect what cities
look like and influence ways of living, producing, moving and consuming
in spatial systems?
How can we tackle any undesirable effects which may result from the
adoption of innovation (unemployment, increase in regional disparities,
etc.)?
What are the implications of innovation adoption for urban policy and
planning (including how the planning activity and decision-making is
carried out.
In the following, we shall limit ourselves to commenting on the different

approaches to the spatial analysis of innovation diffusion, and then
summarise the main considerations deriving from the above discussion.
The approaches to the spatial analysis of innovation diffusion can be
brought together under the four main complementary perspectives
summarised in Fig. 11.2. Although these perspectives are all equally
relevant, we believe that in analysing the relationships between innovation
and urban development priority should be given to the so-called
'development perspective'. As indicated in Fig. 11.2, this perspective
focuses on the consequences of innovation on the socio-economic, spatial
and cultural environment of a region. It represents the final stage of the
diffusion process, that is the one in which the effects on the urban
environment manifest themselves and their outcomes become fully
observable. Within the on-going process of urban innovation, this final stage
also sets the conditions for a new bundle of innovation to occur. Essentially,
this perspective purports to assess the overall outcome of the multiplicity of
changes induced in the individual interaction patterns associated with the
soft component of innovation. The relevance of this perspective lies in the
fact that it entails two further dimensions which are increasingly important in
the current changes in urban societies (Brotchie et at 1986; Amin 1994):
evaluation and planning.
Adoption perspective
Consumer innovation
Demand aspect of diffusion
Processes bv IVhich it occurs
Market and infrastructure

perspective
Economic history Establishment of diffusion agencies
perspective Implementation of an adoption strategy 1------"
Where, when and for whom ~--..
Supply side of diffusion
innovation occurs Conditions for adoDtion
Innovation improvement and
IadaDtation t
Development perspective
Economic disparities among
regions and social classes
~
Impact of diffusion on
individual welfare and regional
IdeyeloDment
Figure 11.2 The main perspectives in the analysis of innovation

diffusion (after Brown 1981)
The implications of the above discussion can be summarised as follows:
a. The thesis advanced in this chapter is that major potential for urban
development is to be found in the 'soft component' associated with the
use of innovation. This consists of the continuously modified interaction
patterns, which define what is usually understood as the enabling role of
innovation. The soft component makes a wide range of opportunities
accessible and beneficial to individual urban agents as well as collective
agents. How to create and make use of these opportunities represents the
challenge of planning the city of tomorrow. A corollary of this thesis is
that the soft component is primarily an 'individual experience' on the part
of ordinary agents (whether an individual, a firm, a household, an
institution or a planning department) encountered in performing their
daily activities which occur in built places located in geographical space.
b. Whereas innovation has to be considered an intrinsic feature of system
evolution, whose appearance can be casual and unexpected, the
environmental conditions (i.e. the economic, social, spatial, cultural and
institutional context) are largely responsible for its viability. This holds at
the micro (individual) level as well as at the macro (collective) level. The
existing spatial structure as well as the built environment and cultural
milieu, therefore, are likely to play an important role in this respect.
248 Bertuglia c.s., Occelli S.
c. The effects of innovation are not neutral. We must recognise that

innovation does not represent a panacea to the solution of the urban
problems. There is a need to assess its outcomes, some of which may
even be undesirable. The evaluation of innovation performance, then,
represents a substantial issue in the relationships between innovation and
urban development.
d. Innovation effectiveness is strictly associated with the awareness (in
terms of perception, information and knowledge) of innovation potential.
The explanations - that is, broadly speaking, the individual and collective
representations - of both innovation and its likely impacts are therefore
likely to have no lesser role in the adoption of innovation than the
appearance of the innovation itself.
Within the above framework, we shall now briefly discuss some aspects
which could become major topics of a future research agenda dealing with
the relationships between innovation and urban development. These can be
listed under the headings: a) accessibility, b) people and places, c)
representations of innovation.
a) Accessibility
Accessibility is a well-established concept in urban and regional analysis

(see Burns 1979; IRES 1995, for a review). While providing the conceptual
links between the functional component of an urban system, i.e. the range
and temporal organisation of activities which are available to individuals in a
geographical space, and the spatial component, i.e. the transportation
allowing individuals to overcome space and participate in specific activities,
it also gives a sound methodological basis from which operational
measurement tools can be derived (Occelli and Gallino 1992). In particular,
it provides a meaningful analytical perspective through which we can
investigate the enabling role of innovation relative to the functional and
spatial pattern of urban environment.
From a theoretical viewpoint, such an investigation has already been
undertaken. For example, Kobayashi et al (1993) have developed a
representation of accessibility to describe the interdependencies created by
knowledge exchanges. Martellato (1993) has considered transport and
telematic accessibilities as distinct inputs in the production function of firms,
and analyses the impact of change in their price on the firm's location;
Bertuglia et al (1995) developed a model for simulating urban scenarios in
which the location choice by firms depends on both transport and telematic
accessibility. Finally, in this book, Beckmann addresses the role of
accessibility in the establishment of knowledge networks.
Focusing more specifically on accessibility allows us to extend the

conventional view which tends to see the beneficial role of innovation (in
particular of NIT) in the increased connectedness and greater interaction
possibilities without taking into account:
the kind and quality of opportunities in terms ofthe activities likely to be

undertaken by individuals as well as the new kind of services which
might be provided by the NIT;
the costs of NIT provision (i.e. installation and running costs of the
information network, tariffs and user costs) and variations between areas
determined by the different infrastructure facilities and supply conditions
offered by NIT providers;
the socio-economic and spatial barriers to NIT, which result not only
from those intrinsic geographical, environmental, cultural and
institutional factors which can hamper the spread of NIT, but also from
the 'forms of exclusion' which some regions risk as a consequence of
their low information and educational level.
The increased relevance of accessibility fostered by NIT's enabling role

thus raises both old and new issues which will have to be addressed in future
research. These include:
the extent to which the development of new forms of

telecommunications-based accessibility in competition with more
traditional transportation-based forms will affect the mobility levels of
people and freight, as well as their spatial and temporal distribution (see
Bertuglia and Occelli 1995);
11 the need to assess accessibility according to the different spatial (i.e.
local, regional, international) and temporal (i.e. short, medium and long
run) levels at which spatial and functional interactions take place within
the system;
iii the awareness that any accessibility assessment should also be
distinguished according to the types of agent desiring to partake in the
various urban activities and, more generally, the spatial resources which
would need to be accessed;
Ivan examination of the relationships between mobility, spatial patterns and
the use of environmental resources.
b) People and places
The pervasiveness of the impact of NIT in remodelling and re-organising

urban places has already been acknowledged by several authors (see, for
example, Castells 1989; Bertuglia and Occelli 1995; Batty 1995; Capello and
Nijkamp 1996; Graham and Marvin 1996). A major feature emphasised in
these studies is that the spatial and temporal re-organisation induced by NIT
allows socio-economic processes to be 'co-evolutionary', thus increasing the
degree of freedom in the evolution paths of urban systems. This also
enhances the planning possibilities, requiring the formulation of 'creative
visions' of likely futures, as well as careful assessments of the social benefits
and costs involved.
An urban system, i.e. a region, is however - no matter what kind of
description (iconographic, linguistic or formalised) we use to define it (Preto
1995) - an entity in which the space of flows (i.e. the tangible and intangible
relationships determining the system interactions) coexists with a space
consisting of places (i.e. the human settlements located in a built, natural and
cultural environment). What ultimately matters is the path of changes made
possible by the co-evolutionary role of NIT. We therefore need to make an
assessment of the outcome of lock-in and/or bifurcation processes which are
liable to occur, and an assessment of the capacity ofthe various agents of the
system to cope with changes.
As system innovative ness has its roots in the ability of individual agents
to exploit the soft component of an innovation and seize the resulting
opportunities (Salomon 1996), a major dimension to be looked at in a future
research agenda is what can be called 'the local dimension of the urban
action space'. By that we simply mean that improvements achieved through
innovation should be an endogenous. and permanent driving force in the
behaviour of the various urban actors and the functioning of the system.
To give an example, we mention three domains where such a drive would
seem likely to produce relevant transformations. The first is transportation
where, besides the complementary/substitution issue between physical
movement and virtual interaction, also the logistical organisation of
transportation modes and user services are deeply affected by innovation
(see, Hepworth and Ducatel 1992). A second domain concerns the
management of local development as the outcome of a multitude of
corporate and institutional decisions (Bennett and McCoshan 1993). A major
area in which innovation can have a significant role is the enhancement of
the capacity of local administrations and 'governance' to respond to the
global socio-economic changes as well as to deal with everyday
commitments. The third domain concerns the conservation of the built
environment as having a value of its own. This is a multi-faceted topic which
is attracting growing interest, especially in the most developed countries, as
a result of increasing sensitivity to the historical heritage consolidated in the
built landscape of human settlements (Coccossis and Nijkamp 1995). To be
Impact of the New Infonnation Technologies 251
able to pass on this heritage to others is a fundamental component of

innovative behaviour.
c) Representations of innovation
The concept of innovation posited here, and referred to in other chapters

in this book, also has significant implications for our interpretations of
innovative activities, as well as the kind of knowledge likely to be associated
with them. A fundamental point we would like to emphasise is that we also
need to be innovative in our representations of innovation. That does not
mean that we need to put forward new interpretations just for the sake of
being innovative. The most salient feature of new interpretations should, we
believe, be their capacity for mutual interaction and learning. A fundamental
requisite is of course that they should be understandable and communicable
not only to the decision-makers, but also to the general public. Creating
'knowledge' about innovations is therefore an essential determinant if
innovation is to be an ongoing process of urban development and societal
change.
Innovation - and we refer here in particular to the hard component of
innovation - is also deeply affecting the ways in which interpretations are
developed. In the field of spatial and regional analysis, for example, the
following consequences of the impact of the introduction of NIT have been
identified (Rabino and Occelli 1997):
a. a shift in the role of computers. Whereas previously the computer was

simply a computing tool and the level of technological sophistication set
the benchmark for the application potential, the computer is now an
integral part of spatial analysis (think for example of GIS, simulation
experiments, and interactive visualisations). Most of the current empirical
studies could not even have been conceived of without a computing
device;
b. the increasing efficiency of operational procedures. Thanks to the
advance in information technology, it is now possible to deal with
problems of a kinds that could not have been tackled earlier or only at
prohibitive costs. In addition, by making possible alternative procedures
of implementation (i.e. parallel computing, computer vision) an updating
of existing approaches is possible;
c. enhancement of communication possibilities. This results from the
improvement of both user and method interfaces as well as from the
diffusion of computing potentialities among a wider and more diversified
public. The possibility of running experiments locally broadens the scope
of spatial analysis. In this respect, the application of models provides the
means for sharing the experience gained elsewhere, and offers support in
learning about 'how to learn about' urban processes.
REFERENCES
Amin, A., ed (1994) Post-Fordism. Blackwell, Oxford
Andersson, A.E., Batten, D.F., Kobayashi, K.,Yoshikawa, K., eds (1993a) The Cosmo-
Creative Society. Springer, Berlin
Andersson, A.E., Batten, D.F., Kobayashi, K., Yoshikawa, K. (1993b) "Logistical Dynamics,
Creativity and Infrastructure." In The Cosmo-Creative Society. Andersson, A.E., Batten,
D.F., Kobayashi, K., Yoshikawa, K., eds, Springer, Berlin, 1-16
Barnett, H.G. (1953) Innovation: The Basis of Cultural Change. McGraw-Hili, New York
Batten, D., Casti, 1., Thord, R., eds (1995) Networks in Action. Springer, Berlin
Batty, M. (1995) The Computable City. Keynote Address at the Fourth International
Conference on "Computers in Urban Planning and Urban Management", Melbourne, July
11-14
Bennett, R.1., McCoshan, A(1993) Enterprise and Human Resource Development: Local
Capacity Building. Chapman, London
Bertuglia, C.S., Fischer, M.M., Preto, G. (eds.) (1995) Technological Change, Economic
Development and Space, Springer, Berlin
Bertuglia, C.S., Lombardo, S., Nijkamp, P., eds (1997) Innovative Behaviour in Space and
Time, Springer, Berlin
Bertuglia, C.S., Lombardo, S., Occelli, S. (1995) "The Interacting Choice Processes of
Innovation, Location and Mobility: A Compartmental Approach." In Technological
Change, Economic Development and Space. Bertuglia, C.S., Fischer, M.M., Preto, G., eds,
Springer, Berlin, 118-44
Bertuglia, C.S., Occelli, S. (1995) Transportation, Communications and Patterns of Location,
in Technological Change, Economic Development and Space, Bertuglia C.S., Fischer
M.M., Preto G., eds, Springer, Berlin, 92-117
Brotchie, 1., Batty, M., Hall, P., Newton, P., eds, (1991) Cities of the 21st Century: New
Technologies and Spatial Systems. Longman, London
Brotchie, J, Hall, P., Newton, P. (1986) The Transition to an Information Society. In Brotchie
1., Hall P., Newton P., eds, The Spatial Impact of Technological Change. Croom Helm,
London, 435-51
Brown, L.A. (1981) Innovation Diffusion: A New Perspective. Methuen, London
Burns, L.D. (1979) Transportation, Temporal and Spatial Components of Accessibility.
Lexington Books, Lexington Mass
Capello, R., Nijkamp, P. (1996) Telecommunications Technologies and Regional
Development. The Annals of Regional Science 30: 7-30
Castells, M. (1989) The Informational City. Blackwell, Cambridge
Coccosis, H., Nijkamp, P. (1995) Planning for Our Cultural Heritage. Avebury, Aldershot
Freeman, C., Perez, C. (1988) "Structural Crises of Adjustment: Business Cycles and
Investment Behaviour." In Technical Change and Economic Theory. Dosi, G., Freeman,
c., Nelson, R., Silverberg, G., Soete, L., eds, Pinter, London, 38-66
Gaffard, J.L. (1990) Innovations et changements structurels, Revue d'Economie Politique 3,
325-82
Giaoutzi, M.(I990) "Telecommunications Infrastructure and Regional Development." In
Infrastructure and the Space-Economy, Peschel, K., ed, Springer, Berlin, 116-30
Graham, S.D.N., Dominy, G.R. (1991) Planning for the Information City: The U.K Case,
Progress in Planning 35, 169-248
Graham, S., Marvin, S. (1996) Telecommunications and the City. Routledge, London
Hepworth, M., Ducatel, K (1992) Transport in the Information Age. Belhaven Press, London
IRES (1995) Un 'analisi dell 'accessibifita in Piemonte. Studio di supporto alia valutazione
delle politiche del piano regionale dei trasporti. Occelli, S., Gallino, T., eds, Quaderni di
Ricerca IRES 74, Turin
IRES (1996) Telecomunicazioni e imprese: if casa del Piemonte. Lanzetti, R., Antonelli, C.,
Rizzello, S. (eds.) Quaderni di Ricerca IRES 80, Turin
Kobayashi, K., Sunao, S., Yoshikawa, K.(I993) "Spatial Equilibria of Knowledge Production
with Meeting-Facilities." In The Cosmo-Creative Society. Andersson, A.E., Batten, D.F.,
Kobayashi, K, Yoshikawa, K., eds, Springer, Berlin, 219-44
Martellato, D.(1993) "Reti di interazione e tecnologia della informazione. Un'analisi dei loro
effetti sulla localizzazione." In Innovazione e trasformazioni della citta: Teorie, metodi e
programmi per if mutamento. Lombardo, S., Preto, G., eds, Collana AiSRre, Franco
Angeli, Milan, 130-47
Mela, A.(1991) Innovation, Communication and Social Structures. Sistemi Urbani 13: 97-108
Metcalfe, 1.(1988) 'The Diffusion of Innovation: an Interpretative Survey." In Technical
Change and Economic Theory. Dosi, G., Freeman, C., Nelson, R., Silverberg, G., Soete L.,
eds, Pinter, London, 560-89
Miles, I., Robins, K (1992) "Making Sense of Information." In Understanding Information,
Business, Technology and Geography. Robins, K (ed.) Belhaven, London, 1-26
Occelli, S., Gallino, T. (1992) Un'analisi dell'accessibilita in Piemonte: note metododologiche
ed applicazione empirica di alcuni indicatori, Atti della XlII Conferenza Italiana di Scienze
Regionali, Ancona, 337-68
Preto, G.(I995) "The Region as an Evolutive System." In Technological Change, Economic
Development and Space Bertuglia, C.S., Fischer, M.M., Preto, G. (eds.), Springer, Berlin,
257-75
Rabino, G.A., Occelli, S. (1997) Re-thinking Urban System Modelling: New Features and a
Proposal. Paper presented at the ICCS Conference, Cortona, 22-25 September
Robins, K. (ed.) (1992) Understanding Information, Business, Technology and Geography.
Belhaven, London
Robins, K., Gillespie, A. (1992) "Communication, Organisation and Territory." In
Understanding Information, Business, Technology and Geography. Robins, K, ed,
Belhaven, London, 157-64
Salomon, I. (1986) Telecommunications and Travel Relationships: A Review. Transportation
Research A, 20A: 223-38
Salomon, I. (1996) Telecommunications, Cities and Technological Opportunism. The Annals
of Regional Science 30, 75-90
Steiner, M.(1990) "How Different Are Regions? An Evolutionary Approach to Regional
Inequality." In Infrastructure and the Space-Economy. Peschel, K., ed, Springer, Berlin,
294-316
Chapter 12
Telematics Innovation and the Development of Non-

Metropolitan Areas: Lessons From Policy
Experience
Andrew Gillespie and Ranald Richardson

Centre for Urban and Regional Development Studies, Newcastle upon Tyne, UK
12.1 INTRODUCTION
In a century of unprecedented economic and social change, technological

progress has usually been associated with the growth of large cities and of a
particularly metropolitan form of modernity. It was Marshall McLuhan in
the 1960s, however, who first argued that new electronic media made
concentration in cities unnecessary:
"With instant electric technology, the globe itself can never again be
more than a village, and the very nature of (the) city as a form of major
dimensions must invariably dissolve like a fading shot in a movie"
(McLuhan 1964, p.306).
The potential of this development to stimulate a re-birth of rural
communities, thereby halting the seemingly inexorable process of rural
decline brought about by the subjection of the country to the city under the
conditions of industrial capitalism (Williams 1973), has attracted various
anti-urban utopianists. Peter Goldmark (1972), for example, suggested that
this could lead to the development of a 'new rural society' brought about by
the diffusion of the urban population into low-density settlements in rural
areas interconnected by telecommunications networks. The most influential
of such perspectives came from Alvin Toffler (1980), who envisioned a
'third wave' society coming into being (the first and second waves triggered
by the agricultural and industrial revolutions respectively), in which work
would return to the home - the 'electronic cottage' - under the stimulus
256 Gillespie A. and Richardson R.
provided by new technology, which would supply the precondition and

dynamo for social transformation (Gold 1991).
12.2 THE RURAL TELECOMMUNICATIONS

OPPORTUNITY
12.2.1 The Role of Telematics Services
In recent years, there has been increasing recognition of the role that
telecommunications could play in facilitating rural development. In part, this
recognition is based on technical advances within telecommunications
networks, particularly their digitalisation, which is opening up a range of
possibilities for combining communications with computing to create new
'telematics' services and applications. Beyond this technological impetus,
there is also recognition that profound structural changes in the advanced
economies - embodied in sectoral shifts from industrial to post-industrial or
informational actIvities, and in shifts from centralised 'Fordist'
organisational structures towards more decentralised and flexible 'post-
Fordist' forms - are reducing or eliminating many of the distance-related
constraints upon processes of economic development in remote or rural areas
(Gillespie et aI1994).
In the North American context, for example, Hudson and Parker (1990)
see the combination of telecommunications advances and structural
economic changes as having the potential to remove the 'distance penalty'
under which rural America has long laboured:
"In the provision of physical goods and services, rural areas could
compete across barriers of distance and geography if they had a natural
resource advantage. In the provision of information goods and services,
reliable telecommunications infrastructure can make geography and
distance irrelevant" (p.195).
There are a number of ways in which telematics innovations could,
potentially at least, contribute to overcoming the rural 'distance penalty'.
These include:
Access to markets: through telematics, rural enterprises could gain access

to markets in core regions and metropolitan areas;
Access to business services: electronic delivery mechanisms could help
rural enterprises to gain access to higher quality and/or lower cost
business services;
Telematics Innovation and the Development of Non-Metropolitan Areas 257
Decentralisation of iriformation work: teleservice firms and teleworkers

could be attracted to locate in rural areas, through cost advantages or
perceived 'quality of life' benefits;
Access to public services: rural citizens could gain improved access to the
public services - such as health and education services - available in
metropolitan areas, through telematics innovations such as distance-
learning or tele-medicine.
Despite this potential, some commentators are nevertheless sceptical of

the likelihood of rural re-birth though new technology. The US Congress
Office of Technology Assessment (OTA) (1995), for example, argues that:
"Visions of life spent conducting business through the Internet, hooked
up by video phone, and receiving and sending faxes, all the while living
in bucolic and isolated bliss, are likely to be a dream that only a few can
fulfil. OTA concludes that the new wave of information technologies will
not prove to be the salvation of a rural US economy that has undergone
decades of population job loss as its natural resource-based economy has
shrunk" (p.6).
12.2.2 Opportunities Foregone?
Why might such scepticism be justified? There appear to be both supply

side and demand side constraints that might prevent the positive scenario
from being realised. On the supply side, there is a risk that under liberalised
telecommunications environments, non-metropolitan areas will find it
increasingly difficult to demonstrate sufficient demand to justify investment
in new infrastructure (Salomon 1988; Parker et al 1989; Gillespie and
Robins 1991). In the UK context, for example, Gillespie and Cornford
(1996) have suggested that rural areas represent the 'cold shadows' on the
map of telecommunications investment, remaining a de facto monopoly
some 15 years after liberalisation, in stark contrast to the 'hot spots' of
competitive provision in the big cities.
It is however on the demand side that lie the most significant barriers to
rural areas grasping the potential offered by telematics. Empirical evidence
points firmly to the fact that non-metropolitan areas make consistently less
use of distance-shrinking technologies than do metropolitan areas. The EU's
SARBA project, for example, surveyed 2000 businesses across a number of
rural regions in Europe, and concluded:
"There is a considerable underuse of equipment and telematics services.
The failure to exploit potential suggests that the opportunities offered by
telematics... are far from being realised. The principal reason is user-
resistance"(llbery et al1995, p.66).
The suggestion that user-resistance lies behind the observed phenomenon
of rural under-utilisation is undoubtedly part of the explanation. The
adoption of telematics is a form of innovation likely to be suppressed in
traditional or conservative approaches to management, which we might
reasonably expect to characterise many firms in rural areas. Further,
however, the expectation that firms in rural areas should be making more use
of telecommunications services because of their remote location is based on
a misconception of the role of telecommunications; telecommunications is
not primarily a tool for overcoming distance, but rather for managing
complexity. Firms with simple transactional structures, with relatively stable
markets and which operate from single sites, will in consequence have
relatively limited 'need' for telecommunications, regardless of the location
from which they are operating. By contrast, firms with complex transactional
structures, with rapidly changing markets or which operate from multiple
sites, will display high levels of 'need' to telecommunicate, again regardless
of the particular locations from which they operate. The propensity to
telecommunicate is thus dependent primarily on the characteristics of the
firm, not on its location or its geographical 'distance' from centres of
economic activity. We would thus contend that the observation that firms in
rural areas make less use of telecommunications than expected is, at least in
part, explained by the nature of the firms concerned; simply, they are making
as much use of telecommunications as they need to, given the requirements
stemming from their transactional structures and forms of organisation.
For such reasons, rural areas with broadly similar geographical
characteristics in terms of remoteness and distance from cities can display
markedly different telecommunications usage profiles, due to variations in
the types of firms operating from them. A study of telecommunications
usage in rural areas in Scotland, for example, revealed a marked contrast
between the relatively high use of data communications in firms in Dumfries
and Galloway, a remote rural region with few large towns, and the much
lower use evident in the Borders region, a less remote rural region with an
urbanised core (Gillespie et al 1994). The explanation for the difference in
usage lay in the nature of the firms present in each area. In Dumfries and
Galloway, a polarised enterprise structure was apparent, with some large
branch plants making intensive use of data communications to connect them
into global corporate networks; in contrast, in the Borders region the firms
were mainly medium-sized, locally-owned and operating from single or a
few local sites, and data communications was little used. Significantly,
however, their limited use oftelematics cannot be taken to imply a constraint
on their activities, or a barrier to accessing remote markets. Indeed, the firms
in the Borders region, which has an industrial specialisation in high quality

woollen textiles, have international markets, necessitating intensive use of
international telephony which is regarded as vital to the competitiveness of
the firms concerned. The differences between the telecommunications usage
profiles in the two areas is thus explained by the detailed characteristics of
their enterprise profiles, in terms of markets, organisational structures, etc.
The degree of 'rurality' of the two areas is entirely irrelevant, except in as
much as it may have influenced the characteristics of the enterprises located
within them.
It follows from this that some of the concerns with the low levels of
usage of telematics in rural areas, which has led to a range of policy
interventions designed to increase rural usage, may be misplaced. The
'barriers' which policy initiatives have been designed to address may
amount to little more than a simple reflection of the characteristics of rural
enterprises. With this possibility in mind, we turn in the remainder of the
paper to a review of the range of policies which have been applied to
encouraging telematics usage in rural areas. It is our contention that many of
these policies, while worthy in their objectives and well-intentioned, are
misguided in that they address wrongly specified problems.
12.3 RURAL TELEMATICS POLICIES
Given the allure of the scenario of 'rural renaissance through

technology', as well as the widespread belief in a set of barriers preventing
the opportunities in rural areas from being realised, it is unsurprising that
those with concern for rural development have attempted to stimulate the up-
take of telematics in rural areas through policy interventions. We are
interested here in investigating primarily the interventions designed to
contribute to rural economic development, and hence targeted at existing
firms and new enterprises. These interventions can be categorised according
to the nature of the barrier they are designed to address, and include, inter
alia:
the provision of advanced infrastructure;

the stimulation of services and applications appropriate to rural SMEs;
the provision of shared facility service centres, or 'telecottages';
the stimulation of rural teleworking.
This is not of course an exhaustive list of the range of policy options

relevant to rural areas. In the particular regulatory context of the USA, for
example, where telecommunications services have already been provided for
a long period by private companies, as liberalisation occurred much earlier

than in Europe, regulatory policies have been of great importance in
ensuring that rural areas got a 'fair deal' in the provision of
telecommunications (see Parker et a11989, and Parker and Hudson 1995, for
excellent accounts of federal and state policies with respect to rural
telecommunications in the US). In the European context, with which we are
primarily concerned in this chapter, policies to stimulate rural economic
development through telecommunications have mainly concentrated on the
forms of intervention listed above. In the sections below, we review
examples of each of these forms of intervention in turn, with the aim of
establishing the extent to which the desired economic development outcomes
have in fact materialised.
12.4 THE PROVISION OF ADVANCED

INFRASTRUCTURE
One of the most persistent misconceptions associated with regional

development policy is that by investing in certain types of infrastructure it is
possible, necessarily and inevitably, to bring about the development of
lagging regions. Although many researchers have identified a statistical
correlation at the regional level between levels of investment in
infrastructure and levels of economic development (e.g. Biehl 1986), it has
proved a formidable methodological challenge to tease out the direction of
causality (Vickerman 1989). Given the perceived significance of the
information society for economic development, it is not surprising that the
notion of investing in telecommunications infrastructure in order to stimulate
rural or less-favoured region economic development has proved attractive to
policy-makers. Research-based evidence to support the validity of such
policy intervention can be found from econometric analysis undertaken in
the US (Parker et a11989; Cronin et aI1993).
In the European context, the limited empirical evidence is derived from
the ex post evaluation of policies which have attempted to provide
telecommunications 'ahead of demand' in order to stimulate regional
development. The most significant of these was the European Union's STAR
programme (Special Telecommunications Action for Regional
Development), which between 1988 and 1991 provided 780 million ECUs of
funding from the European Regional Development Fund (ERDF) in order to
accelerate the rate of advanced telecommunications infrastructure investment
in the less favoured regions in seven member states. According to the
Community level evaluation of STAR (Ewbank Preece 1993), in the case of
network digitalisation in Greece and the launch of cellular radio in Portugal,
the programme was successful in bringing forward the investment plans of

the Public Telecommunications Operators (PTOs) by two years. The
evaluators were not able to demonstrate, however, that the accelerated
infrastructure provision contributed directly and unequivocally to regional
development, concluding only that in their opinion the investments
"represented a good use of regional development funds" (p. S-17).
Another interesting example of an attempt to stimulate rural development
through telecommunications infrastructure investment is provided by the
Highlands and Islands Initiative in Scotland (Richardson and Gillespie
1996). The issue of advanced communications became prominent in the mid-
1980s, with the recognition by the region's development agency, Highlands
and Islands Enterprise, that an advanced communications infrastructure was
becoming an increasingly important factor in economic development. They
realised that although existing infrastructure was satisfactory for voice
telephony, it was not sufficiently advanced to carry high levels of data
traffic. Further, they were concerned that BT, the de facto monopoly supplier
of telecommunications services, would regard levels of expressed demand in
this sparsely-populated region as insufficient to justify expenditure on up-
grading the infrastructure. They therefore decided to contribute 5 million of
public money to a BT-Ied 16 million programme of investment designed to
bring advanced communications technologies to the region ahead of
demand. The three year programme, which was completed in 1992, covered
three main areas of infrastructure up-grading:
improving data transfer services within the area, and making them
available at local call rates;
up-grading 65 exchanges to support ISDN services;
establishing a Network Services Agency to provide value-added services
and access to national databases.
The rationale for the initiative was to defend existing jobs, to allow new
data-intensive businesses to be set up, to encourage existing firms to expand
into new ventures, and to improve the area's attractiveness for inward
investment. Of these objectives, only the latter has really borne fruit, with a
number of back offices and telephone call centres having decided to locate in
the region. The take-up of advanced communications services by existing
firms has been limited, and the infrastructure remains massively under-
utilised (Richardson and Gillespie 1996).
Although the Highlands and Islands Initiative was not based on the belief
that inadequate infrastructure was holding back the region's development,
the decision to invest in infrastructure 'ahead of demand' was clearly
intended both to prevent future bottlenecks developing and stimulate new
economic activities. Whilst incoming firms have managed to harness the

capabilities of the new infrastructures - utilising, for example, ISDN
services - few existing firms in the region have had the need, the inclination
or the skills and competencies required to take advantage of the new
networks and services. Providing an advanced infrastructure is thus no
guarantee that its existence will stimulate economic development. It may in
other words be necessary, but it is certainly not sufficient. For some rural
areas, however, an advanced infrastructure may not even be necessary, in
that the existing infrastructure is more than adequate to cope with current or
likely future demand.
12.5 THE STIMULATION OF SERVICES AND

APPLICATIONS APPROPRIATE TO RURAL
SMES
In recognising that advanced telecommunications infrastructures are

insufficient to stimulate rural economic development, our attention shifts to
the conditions for the effective use of telecommunications networks by firms
(which in the context of most rural areas, means SMEs). Various policy
initiatives have attempted to bridge the gap between infrastructure provision
and effective use, for example by stimulating new value-added services
which are targeted specifically at SMEs. The above-mentioned ERDF-
funded STAR Programme, in addition to providing infrastructure also placed
much emphasis on stimulating the development of database and information
services for SMEs. Although the evaluation of the STAR Programme
reported several relatively successful projects of this sort, the vast majority
failed to achieve commercial viability - in fact, when support funding
expired, the services failed. On the other hand, services designed to support
the transactional activities of the firm, notably EDI services, or those that
were integrated into other core processes, such as CAD-CAM were far more
successful (Ewbank Preece 1993).
The latest transactional service, which appears to offer much hope to
those promoting the growth of rural SMEs, is the Internet, and particularly
the World Wide Web. This is because it appears to offer the possibility of
'breaking free' of the market-access limitations imposed by peripheral
location and remoteness. Richardson (1996, pI 0) thus argues:
"Rural food producers in north America now use the Internet to sell a
wide variety of products ... Rural craft and manufactured goods
producers sell everything from clothing to furniture over the Internet...
The Internet represents a global storefront for such rural and remote
businesses, providing them with access to customers never before

possible."
Nevertheless, the validity of these claims for the Internet cannot be taken
for granted. The EU-funded WOLF project, for example, worked closely
with 150 SMEs drawn from ten of Europe's less favoured regions, many of
them rural, with the aim of assisting the firms to take advantage of the
WWW for business benefit (WOLF Consortium 1997). Even though the
SMEs were hand-picked according to their likelihood of benefiting from
electronic commerce opportunities, and despite the considerable assistance
they received from external consultants during the course of the project,
there were few instances of increases in sales reported as a result of the
firms' use of the Internet. This could be due to the relatively short duration
of the project, or to the current limitations of the Internet for trading which
will soon be overcome by the establishment of secure trading standards.
However, the explanation could also be that, although the Internet will
undoubtedly become a widely-used business tool, it is not likely, in any
generalisable sense, to transform the market access opportunities of rural
SMEs, most of whom are selling to local markets and do not have the
managerial or market-serving capacity to efficiently serve remote markets.
The two mis-specifications here are, firstly, that the main barrier to rural
SMEs gaining access to core markets is geographical distance and, secondly,
that this distance barrier can be overcome by telematics. We would contend
that geographical distance per se is not the main barrier facing rural SMEs
wishing to gain access to core markets. It is more likely to be their inability
to compete in these markets, due either to problems of price (by definition,
small firms will often not be able to reap the economies of scale which larger
firms can exploit) or to problems of product/service quality or design.
Evidence suggests that high rates of innovation and achieving high quality
are generally associated with having particularly demanding customers,
which most rural areas are unlikely to possess. (A good example is provided
by food, one of the specialised products of rural areas. The most exacting
demands for specialised products or quality standards tend to come not from
rural consumers, but from specialised suppliers in the largest cities).
For the relatively few rural SMEs which do have competitive products or
services in metropolitan or even global terms, the problem of access to
geographically distant markets is not likely to be overcome solely by
telematics solutions, such as the Internet. The ability to serve a remote
market will usually require direct market presence, due to the requirements
for negotiation in the selling process and for after-sales support. Telematics
can usefully supplement the type of direct market presence normally
required, but only in a limited range of products/services can it substitute
entirely for this presence. It follows that policies to stimulate the up-take by
rural SMEs of such telematics applications and services, with the specific
aim of helping them gain access to remote markets, will often have
disappointing or limited results: Many of the firms concerned will have
products or services which are not competitive outside the local market, and
others will need to put in place mechanisms for providing them with direct
presence in remote markets if they wish to compete effectively.
12.6 THE PROVISION OF SHARED FACILITY

CENTRES OR 'TELECOTTAGES'
Another barrier, or assumed barrier, which rural telematics policies have

attempted to address, concerns the difficulty of access to IT and
telecommunication facilities and associated training and support services.
This has been tackled by setting up shared facility centres or 'telecottages'.
Originating in Scandinavia, telecottages have social and community
development goals, as well as providing services to local small businesses
and facilities for undertaking teleworking (Qvortrup 1989). The
establishment of telecottages has been one of the main rural leT-related
policies applied in Europe, with particularly widespread application in
Scandinavia (where there were around 100 telecottages by the early 1990s;
Hillman 1993), Britain and Ireland. By mid-1997, the Telecottage
Association, which covers the latter two countries, had 160 members
(Telecottage Association 1997). Telecottages are predominantly, though not
exclusively or necessarily, a rural phenomenon. In Britain, a quarter of all
telecottages are located in remote rural settings, another quarter in small
villages, and half in towns, mainly market towns situated in rural areas
(Gillespie et aI1995).
However, there is little evidence that telecottages have contributed
appreciably to rural economic development. In Norway, for example, of the
five telecottages established in the rural province of Finnmark, which were
intended to become self-financing, only one remains, playing a mainly
socio-cultural role (Qvortrup 1994). In Britain, the limited survey evidence
suggests that there are few instances of teleworking being undertaken from
telecottages (Gillespie et al 1995), due not only to the lack of necessary
skills but, particularly, to the difficulties of identifying potential customers
beyond the local market and of marketing capabilities to those potential
customers. Some nascent attempts, for example schemes underway in Wales
and as part of the EU's TeleMart project, are tackling this problem by
forming networks of telecottages to integrate their competencies and market
them in metropolitan areas. Again, then, it is not inadequate access to IT
facilities in rural areas which is preventing such areas from taking advantage
of new telematics-enabled opportunities, the real constraints are the lack of

necessary competencies, of market knowledge, and the contacts that would
facilitate market access.
12.7 THE STIMULATION OF RURAL

TELEWORKING
Ever since Toffler (1980) put the electronic cottage at the centre of his
'third wave' society, those concerned with the development of rural areas
have attached considerable importance to the prospects for rural teleworking.
The limited and relatively narrow range of employment opportunities in
rural areas could, it is suggested, be transformed by the availability of new
teleworkingjobs which are not tied to any particular location.
There are, however, a number of problems with this vision. Firstly,
teleworking does not really provide new jobs, but rather a new tele-mediated
way of undertaking existing jobs, or types of jobs. A pre-requisite for
participation in teleworking is, therefore, the possession of the types of skills
and competencies which can be amenable to teleworking. Many rural
residents will not possess such skills. Secondly, many teleworking jobs are
not as locationally footloose as is often supposed. For employed teleworkers
(as opposed to those who are self-employed), survey evidence from Britain
suggests that they spend on average a quarter of their time on the employer's
premises and a further quarter elsewhere, for example on clients' premises
(Huws 1993). As a consequence, they are most likely to be found in
suburban and urban areas, where they are close to their employers and
clients. A recent study on the prospects for partially home-based teleworking
in Britain concluded that
"while a minority of rural areas offer considerable potential for the
development of this form ofteleworking, the majority, including all those
categorised as 'deep rural', do not" (Huws et a11996, p. 33).
More locational flexibility is likely for those who are self-employed or
free-lance teleworkers. Though there are still constraints due to the need for
access to clients, Huws et al (1996) conclude that this type of teleworking
has the highest potential for rural areas, as result of life style choices which
are favouring attractive rural areas.
For such reasons, teleworking in rural areas is likely to be primarily a
feature of in-migrating professionals seeking a particular quality of life. This
is certainly the pattern in the Highlands of Scotland, which have proved
successful in attracting teleworkers to locate in often deeply rural
environments (Richardson and Gillespie 1996). Significantly, these
teleworkers tend to be in-comers or people returning to the area rather than

residents who have been based in the area most of their lives. In order to be
commercially successful, teleworkers need to have marketable skills and a
product or service which can be offered by means of telecommunications -
attributes unlikely to have been established in rural areas. Furthermore, the
well-established market contacts required will, almost by definition, have
been developed in non-rural areas and then imported into the rural area
through life-style-based migration.
In connection with policies aiming to increase the incidence of
teleworking in rural areas, it needs to recognised that the scope for
encouraging teleworking amongst existing rural residents is likely to be
limited. For those rural areas, such as the Highlands of Scotland, perceived
by many metropolitan professionals to have an attractive quality of life, there
is likely to be considerable scope for attracting mobile professionals to come
and live in the area. While such policies, if successful, will undoubtedly
broaden the base of the rural economy, they are unlikely to create significant
new job opportunities for existing rural inhabitants, and may even have some
undesirable side-effects, such as, for example, in pushing up the price of
residential property.
12.8 CONCLUSIONS
Our contention in this paper is that rural telematics policies have, by and
large, failed to deliver the expected benefits to rural economic development.
This is largely because of the mis-specification of the barriers or obstacles to
development. The two main types of barrier which the policies were
designed to overcome are: a) barriers of access to telecommunications
infrastructure, IT facilities and appropriate telematics services, and b)
barriers of access to markets and customers. .
The first barrier has proved to be largely illusory. Policies designed to
tackle the various impedances to access (individually or in combination),
have made it increasingly clear that the real barrier lies not in access, but in
the limited use which rural SMEs make of infrastructures, facilities and
services. While policy-makers have correctly identified the existence of a
'usage gap', the diagnosis of access problems has led, too often, to policies
likely to exacerbate the real problem. Once the supposed supply-side access
problems have been 'solved' through the provision of infrastructure ahead of
demand, the provision of telecottages with IT facilities, or the design of
telematics services targeted at rural SMEs, it becomes even more starkly
apparent that the real barrier is not one of availability, but the lack of usage
of these infrastructures.
In part, there are real barriers to be overcome, concerning the lack of

awareness, lack of appropriate skills, lack of marketing expertise and lack of
managerial competence in rural (and indeed urban) SMEs, which result in
them making less than optimal use of available computing and
telecommunications technologies and tools. These deficiencies can and
should be addressed by policies of enterprise and managerial development in
which the effective use of available telematics tools need to be integrated. It
needs to be remembered, however, that even if these managerial
shortcomings are overcome, it is still likely that rural SMEs would make less
use of telematic tools than other firms, simply because of their
characteristics - single site firms, serving primarily local markets, will
display relatively low levels of need (or, to put it another way, will derive
relatively little benefit from usage), regardless of where they are located.
The second type of barrier towards which rural telematics policies have
been directed concerns the barrier of distance in limiting access to remote
markets. Again, we have argued that this barrier is mis-specified. What most
rural SMEs (or indeed rural teleworkers) lack is not geographical proximity
to metropolitan markets, but the lack of understanding of these markets and
the appropriate mechanisms for serving these markets, or the products and
services that would be competitive. The issue to be addressed is then not one
of geography or distance per se, but rather of market credibility and
competitiveness. The appropriate policy responses again concern the
development of skills, management competence, marketing expertise,
product/service innovation capabilities, etc.
For both sets of barriers, it is thus clear that telematics policies can play
no more than a subsidiary role in tackling the identified weaknesses or
shortcomings of rural enterprises. Such enterprises are not, with few
exceptions, being constrained by inadequate access to leTs, nor does their
primary shortcoming with respect to market access concern geographical
distance. The appropriate policy responses would seem to be an emphasis on
improving the fundamentals of rural enterprise performance (business
efficiency, marketing, innovation, etc), and perhaps also the stimulation of
in-migration of individuals, such as professional teleworkers, and firms
which already possess well-developed skills, products/services and market
access mechanisms. As we have argued elsewhere (Richardson and Gillespie
1996), the most significant contribution of information and communications
technologies to rural economic development may well lie in facilitating the
movement of mobile people, firms and activities from metropolitan areas to
rural areas.
REFERENCES
Biehl, D., ed, (1986) The Contribution of Infrastructure to Regional Development. (2 vols),
Office for Official Publications of the European Communities, Luxembourg
Cronin, FJ., Parker, E.B., Colleran, E.K., Gold, M.A. (1993) Telecommunications
Infrastructure Investment and Economic Development. Telecommunications Policy 17(6):
415-30
Ewbank Preece (1993) STAR - Special Telecommunications Action for Regional
Development: Community-level Evaluation, Executive Summary. Ewbank Preece,
Brighton
Gillespie, A, Coombes, M., Raybould, S. (1994) Contribution of Telecommunications To
Rural Economic Development: Variations on a Theme? Entrepreneurship and Regional
Development 6: 201-217
Gillespie, A, Cornford, 1. (1996) "Telecommunications Infrastructures and Regional
Development". In Information and Communication Technologies: Visions and Realities.
Dutton, W., ed, Oxford University Press, Oxford
Gillespie, A, Richardson, R., Cornford, 1. (1995) Review ofTelework in Britain: Implications
for Public Policy. Report prepared for the UK Parliamentary Office of Science and
Technology, CURDS, University of Newcastle upon Tyne
Gillespie, A, Robins, K. (1991) "Non-universal Service? Political Economy and
Communications Geography." In Cities of the 21'1 Century: New Technologies and Spatial
Systems. Brotchie, J., Batty, M., Hall, P., Newton. P., eds, Longmans, Harlow, 159-70
Gold, 1.R. (1991) "Fishing in Muddy Waters: Communications Media, Homeworking and the
Electronic Cottage." In Collapsing Space and Time: Geographic Aspects of
Communication and Information. Brunn, S.D., Leinbach, T.R., eds, Harper Collins
Academic, London, 327-41
Goldmark, P.G. (1972) Tomorrow we will Telecommute to our Jobs. The Futurist 6: 55-8
Hillman, J. (1993) Telelifestyles and the Flexicity: A European Study. European Foundation
for Living and Working Conditions, Dublin
Hudson, H.E., Parker, E.B. (1990) Information Gaps in Rural America: Telecommunications
Policies for Rural Development. Telecommunications Policy, 14: 193-205
Huws, U. (1993) Teleworking in Britain, Research Series No.18, Employment Department,
Sheffield
Huws, U., Honey, S., Morris, S. (1996) Teleworking and Rural Development. Rural Research
Report No.27, Rural Development Commission. Rural Development Commission, London
I1bery, 8., Clark, D., Berkeley, N., Goldman, I. (1995) "Telematics and Rural Development:
Evidence for a Survey of Small Businesses in The European Union." European Urban and
Regional Studies 2 (1): 55-68
McLuhan, H.M. (1964) Understanding Media: The Extension ofMan. Sphere, London
Parker, E.B., Hudson, H.E., Dillman, D.A, Roscoe, AD. (1989) Rural America in the
Information Age: Telecommunications Policy for Rural Development. The Aspen Institute,
Washington DC and University Press of America, Boston
Parker, E.B., Hudson, H.E. (with Dillman, D.A., Strover, S., Williams, F.) (1995) Electronic
Byways: State Policies for Rural Development Through Telecommunications. The Aspen
Institute, Washington DC
Qvortrup, L. (1989) The Nordic Telecottages: Community Teleservice Centres for Rural
Regions, Telecommunications Policy 13: 59-68
Qvortrup, L. (1994) Advanced Communications and Regional Development: the Case of
Finnmark. Report for the CEC ACCORDE Project
Richardson, D. (1996) The Internet and Rural Development: Recommendations for Strategy
and Activity: Final Report and Executive Summary. Report for the Food and Agriculture
Organization of the United Nations (http://www.fao.orglwaic... irect/CDDO/)
Richardson, R., Gillespie, A. (1996) Advanced Communications and Employment Creation in
Rural and Peripheral Regions: A Case Study of the Highlands and Islands of Scotland.
Annals of Regional Science, 30: 91-110
Salomon, I. (1988) Geographical Variations in Telecommunications Systems: the
Implications for Location of Activities. Transportation 14: 311-27
Telecottage Association (1997) Telecottage Map: British and Irish Telecottages. Teleworker 4
(5): 20-21
Toffier, A. (1980) The Third Wave. Pan, London
US Congress, Office of Technology Assessment (1995) The Technological Shaping of
Metropolitan America. OTA-ETI-643. us Government Printing Office, Washington DC
Vickerman, R. (1989) "Measuring Changes in Regional Competitiveness: the Effects of
International Infrastructure Investments." Annals ofRegional Science 23: 275-286
Williams, R. (1973) The Country and the City. Chatto and Windus, London
WOLF Consortium (1997) WWW and Internet Opportunities in the Less Favoured Regions.
Project Final Report, WOLF Consortium, Onyx Internet, Middlesborough
Chapter 13
The Impact of Innovations in Telecommunications

and Information Systems on Urban Mobility
Ennio Cascetta and Bruno Montella

University "Federico II". Naples. Italy
13.1 INTRODUCTION
During the last few decades the considerable increase in mobility, evident
almost everywhere, has derived from a number of socio-economic factors.
Among these, the growth of income and increased involvement in various
urban activities (changes in lifestyle) have played a major role. In the larger
urban areas this has led to widespread congestion, which appears to be
growing steadily in all cities of the Western world. If this current trend in
mobility continues, within a few years the increase in congestion and
pollution in large cities is predicted to exceed the alarm threshold.
The growth in mobility demand which has induced this phenomenon has
been generated by the increasing need, or desire, to participate in a wide
range of activities. However, many of these activities could be carried out
either by means of physical proximity or telematic links. The supply (or
network) of the opportunity for 'exchange', in its various forms, may thus be
seen as consisting of two parallel networks: a physical communications
network and a telecommunications network. This suggests that the
development of telecommunications (TLC) could have a significant effect
upon both urban mobility demand and transportation supply. The issue has
already been raised within the EU, and in May 1994 the Council of Ministers
identified five salient objectives relating to the development of interactions
between transportation and telecommunications. These can be summarised
as follows:
272 Cascetta E. and Montella B.
the need to promote telecommuting both at home and in satellite centres

so as to reduce commuter traffic. The possible benefits of such a policy
consist of increased productivity, greater working flexibility and reduced
production costs as well as less traffic, pollution and lower energy
consumption. The main disadvantage consists of the reduction of
opportunities for personal contact between individuals. Nevertheless, by
the year 2000 10 million telecommuting jobs are expected to be created;
the setting up of advanced TLC systems to manage road traffic, including
functions ranging from the provision of information for motorists and
public transport users to the real-time management of traffic flows and
parking spaces. The aim is to cover 300 metropolitan areas with such
systems by the beginning of the year 2000;
a system of air traffic management should be created in Europe between
all the ground control centres (LATC);
private homes should be equipped with an on-line system of multimedia
services, ranging from entertainment to information and e-commerce;
public administration networks in Europe should to be interconnected in
order to guarantee that all EU citizens have rapid and easy access to
information.
The large-scale application ofTLC is likely to result in:
a substitution effect: this can occur where the use of telecommunications

substitutes certain physical trips (telecommuting, teleshopping, etc.);
a complementarity effect: this can affect both demand and supply. An
increase in demand may result from the time-saving connected with the
use of TLC, since the free time gained leads to new trips being
undertaken. Supply, on the other hand, may be affected through the
provision of better information to users, leading to a reduction in the
generalised trip cost (Le. an increase in efficiency of the transportation
supply).
As will be seen, the more serious the congestion and the higher the
relative transportation costs, the more intense these effects are likely to be.
Substitution effects are generally connected with changes in lifestyle, which
are reflected in mobility needs and in the performance of the transport
system, with effects in both the short and medium-long term. We are already
seeing the need for some physical trips being eliminated as a result of the
performance of certain activities 'at a distance', thanks to the possibilities
offered by telecommuting, teleconferencing, teleshopping and other tele-
assisted activities (telebanking, tele-entertainment, etc.). On the other hand,
the development of telecommunications can also create complementarity
effects, since the increased efficiency of the transportation system made
The Impact ofInnovations in Telecommunications and Information Systems 273
possible by the use of TLC systems tends to generates new mobility demand.
This increase in demand results from the reduction of time previously
consumed in physical trips. It is generally presumed that the 'extra' time
made available will be used to make trips for leisure or cultural purposes and
will therefore modify the whole pattern of mobility, since the trips are likely
to be oriented towards different parts of the city and transport modes.
The application of TLC systems can help to achieve greater efficiency in
both private and public transportation. As regards private transportation,
TLC functions enable the user to adapt his or her departure time and route
choice in accordance with real-time information on the current and predicted
state of the transport system. As far as public transportation is concerned, the
functions can be used by transport authorities to increase the efficiency of
their service and also affect users by improving trip quality. In the former
case, the information channelled through centralised vehicle control systems
has important spin-offs in optimising a series of corporate functions. It
allows more efficient management of operations, as well as of depots and
vehicle maintenance. As regards the users, information systems may be
static or dynamic, and be 'pre-trip', i.e. consulted prior to beginning the trip
(and hence possibly affecting the mode chosen as well as the itinerary) as
well as 'en route'. Increasingly, the information provided relates to all
possible transportation modes available for the proposed trip (including
public and private transportation systems, with possible interchange and
parking facilities). Thus the user is provided with a complete picture of the
alternatives and assisted in the choice of the most appropriate one.
A recent survey of research carried out in this field has been made by
Mokhtarian and Salomon (1997), who show that there have been numerous
investigations of the possible positive and negative effects of TLC
applications, examined at both national and international level. The authors
have drawn up a classification of the studies, dividing them into four
categories based on the scale (macro and micro) and the scope of the
coverage. It emerges however that although some effects have received
particular attention, especially user information systems for road and public
transport systems in Europe and the USA (probably due to their implications
for industry), there are still no experimental results concerning large scale
applications. Some suggestions about the possibility of integrating TLC
options into spatial choice structure have been presented by Ben-Akiva
(Ben-Akiva et aI1996).
The first part of this chapter describes the results of research on the main
substitution effects on mobility demand and the possible complementarity
effects on transport demand and supply. The second part provides a brief
update on the development of mathematical models and calculation
methodologies for evaluating the effects ofTLC on transport systems.
13.2 SUBSTITUTION EFFECTS ON MOBILITY

DEMAND
To understand the extent of the substitution effect of TLC technologies

with regard to physical trips, it is first necessary to analyse the
interrelationships between the need or desire to participate in an urban
activity and the choice of the mode, i.e. whether to use TLC or physical
mobility (transportation). Figs. 13.1 and 13.2 (Moore 1988) represent,
respectively, the relations between the supply and demand of TLC services
and the choice model for the telecommunications mode.
(A) Supply of communication (8) Demandfor communica-

facilities: tion facilities:
- equipment and facilities - formal structure of the
....
...
r----1..~ provided by organisation
- spatial office structure
organisation
- informal structure of the
organisation
(C) Quantity and type of

communication activity
....
(0) Use transportation
- r-
or
telecommunications?
(E) Communication (F) Transportation

media choice choice
/
(G) Evaluation of outcome
of communication choice
.. .......
Figure J3. J An integrated framework for communication/transportation
interactions: the supply-demand model
The Impact ofInnovations in Telecommunications and Infonnation Systems 275
I. Communication activity 2. Characteristics ofthe

characteristics initiator:
- recipient relationship
3. Individual characteristics 5. Communication

media characteristics
4. Organisational
characteristics
~,
6. Affective perceptions 7. Cognitive media

of media perceptions
" " " ~,
8. Determination ofthe relative 9. Perceptions and

importance of communication feelings about media
attributes and personal needs
+ +
10. Formation of media preferences
in specific context
11. Situational
constraints
~r
12. Choice of media in ........

specific context
Figure 13.2 Integrated framework for communication/transportation

interactions
13.3 THE MEDIA CHOICE MODEL
As may be inferred from Fig. 13.1, the demand for communication

interacts with the supply, jointly determining the quantity and type of
exchange required, which in turn determines the choice between the
telecommunications system and the transport system. The choice of system
results from an evaluation of the two alternatives and ends in an iterative
process with modifications to telecommunications supply and demand. Fig.
13.2 shows a sequential logic block-diagram which takes us from the stage
where the decision-maker becomes aware of having a need for
telecommunications to the choice model resulting from the specific
constraints.
Substitution effects can affect the physical movement of both passengers
and freight, involving trip generation (i.e. the number of trips undertaken for
each purpose or activity) and trip distribution (the physical destinations).
Below, we examine in particular the situation regarding passenger trips,
highlighting the possible changes in terms of quantity and the effects on
transport. To test the substitution effects, Selvanathan and Selvanathan
(1994) examined the demand system at the consumer level in the city of
Rotterdam, and found pair-wise substitutes with elasticities with respect to
costs of -0.8 for private transport and -0.03 for public transport.
The main TLC services which can currently be considered to have
significant effects on the structure of mobility demand are the following:
telecommuting (Salomon 1985; Nilles 1988; Garrison and Deakin 1988;

Kitamura et a11990; Mokhatarian 1991; Pendayala et a11991; Hamer et
aI1991,1992; Hu and Young 1992; DOT 1993; Divieti 1995; Henderson
and Mokhatarian 1996; Koenig et aI1996);
teleconferencing and distance learning (Bennison 1988; Mokhtarian
1988; Salomon et a11991; Button and Maggi 1994);
teleshopping (Salomon 1988; Batty 1997);
teleservices (telebanking, tele-entertainment, etc.).
Telecommuting is undoubtedly the TLC system likely to produce the

greatest impact on mobility demand. Telework activities may completely
substitute office work, but can also be part-time and not necessarily
undertaken at home, but in special local centres (satellite centres). This latter
case would obviously entail a change in trip destination and a reduction in
the average length of the trip, rather than its total elimination.
Although work opportunities at home (both as a complete and a partial
substitute for office working) have increased enormously with the expansion
of TLC, they are generally considered to be underexploited. According to
certain estimates, it has been calculated that the widespread adoption of

telecommuting could theoretically make it possible to avoid over 20% of
journeys to work in urban areas (and, in particular, about 50% of trips for
managerial staff). Nevertheless, in current circumstances, only 5-10% of
such trips appear to be realistically avoidable. The cause of this divergence
lies both in the nature of corporate structures and in worker behaviour. On
the one hand, firms do not always view positively the loss of control over
work undertaken and of the direct contact between employer and employee.
Furthermore, employees themselves are often not particularly enthusiastic
about working at home, because of possible conflict with the normal family
routine, the problem of using domestic space for working activities and the
elimination of social relationships with work colleagues. This suggests that
future development will possibly lie in the direction of working only a few
weekdays at home. Tables 13.1 and 13.2 show future telecommuting
projections and their consequent impacts upon the transportation system
forecast in the USA (Divieti 1995). As the tables indicate, the number of
telecommuters is forecast to double over the next few years, with satellite
telecommuting centres developing considerably, in order to overcome the
difficulties mentioned above, connected with worker isolation and the use of
private homes for working purposes.
Table 13.1 Future projections for teleworking in the US
Future Projections For Teleworking 1997 2002

Number ofteleworkers (in millions) 3.1- 6.2 7.5 - 15.0
Percentage of work form (in %) 2.3 - 4.6 5.2 - 10.4
Home teleworkers (in %) 74.3 49.7
Teleworkers in satellite centres (in %) 25.7 50.3
A verage days per week 2-3 3-4
Other indirect impacts on trip structure may occur due to changes in the
trip chains (i.e. sequences of trips starting and ending at home ). At present,
these chains are chiefly conditioned by the journey to work as the primary
purpose (i.e. work + shopping and/or accompanying another person). By
reducing work trips, these chains may be reduced in favour of simpler or
circular trips (home - activity - home). There could also be impacts on
modal choice, with cars being used more for the dispersed destinations and
less for trips to the central zones better served by public transportation.
Lastly, there may be effects on the urban structure itself, due for example to
a reduction in urban centrality. However, an in-depth assessment of this type
of effect, which may in some cases be undesirable, lies beyond the scope of
this chapter.
Table 13.2 Impacts of telecommuting on the transportation system
Impacts on Transportation System 1997 2002

Saving in vehicle miles (VM) (L.x 10 12 ) 10.0 - 12.9 17.6-35.1
Saving in VM - total passengers (in %) 0.49 - 0.63 0.7 - 1.4
Saving in VM - commuting (in %) 1.6 - 2.0 2.3 - 4.5
Saving in oil (in millions of gallons) 475.9 - 619 840 - 1.679
Saving in oil (as %) 0.6 - 0.8 1.I - 2.1
Value of oil saved (in million $) 543 -706 958 - 1.914
Saving in emissions (in %) NOx 0.6 - 0.8 1.I - 2. I
HC 0.8 - 1.I 1.4-2.7
CO 1.0 - I.3 1.7-3.4
Hours saved per year for average teleworker 93 110.3
Total hours saved per year (in millions) 444 - 577 826 - 1.652
Another effect of the potential of telecommunications for improving

transport efficiency is the generation of 'latent demand'. This phenomenon
consists of the attraction of new vehicle trips due to an increase in the
capacity of the transportation system or the reduction of costs. Studies of
latent demand in a general context (Hansen and Huang 1997) shows that it
could amount to anywhere between 30% and 90% of the newly available
capacity.
Teleconferencing and distance learning represent another possible field of
development in the use of telecommunications. These substitute face-to-face
meetings and are chiefly used to avoid long or difficult trips. It is possible to
'teleconnect' several appropriate locations so as to substitute a traditional
conference, meeting or lecture with video and audio communication. Two
studies in the field of videoconferencing - those carried out by SCAG (the
Southern California Association of Governments) (Mokhtarian 1988) and by
British Telecom (Bennison 1988) would appear to have produced significant
findings. The former showed that the use of videoconferencing allowed a
time saving of 18% and a saving of 24% in distance travelled compared with
the trips needed for face-to-face meetings. Furthermore, a 30% higher
participation rate was reported compared with conventional meetings. In
relation to the second study, it is worth highlighting the results of a survey of
reasons for not using videoconferencing (Bennison 1988) in order to grasp
the perceived disadvantages (see Table 13.3). The main causes would appear
to lie in the current underpromotion of such systems and a certain reluctance
by many people to use the technology.
A number of disadvantages of teleconferencing have been demonstrated

in other studies (Button and Maggi 1994). Apart from technical problems
and (the present) unfamiliarity with the technology, a more serious limitation
on the future potential of the teleconferencing would seem to be the
elimination of the external benefits of face-to-face interaction. We can
perhaps conclude that the effect is likely to be only partial substitution.
Table 13.3 The perceived disadvantages of videoconferencing
Mean score * Rating as % Rating as

important ...... unimportant+
Too many people in average meeting 4.1 44 40
Average meeting too long 4.2 51 38
Average meeting needs in-person contact 4.7 55 27
System was unreliable 3.6 28 47
Room access difficult 3.9 39 42
People like to get out of the office 3.3 31 51
Limited contact between sites 4.0 45 39
People felt uncomfortable 4.0 47 38
Not strongly promoted 5.7 78 10
No colour picture 3.0 21 61
Terminal too small 3.9 37 39
* Mean score on a seven point scale from 1-7 (I = not at all important; 7 = extremely important)
*. Those rating the factor as 5, 6 or 7
+ Those rating the factor as I, 2 or 3
Teleshopping consists of the possibility of making purchases without the

customer needing to undertake any physical trips, since he or she can search
for the target object (the product) on the Internet, receive all relevant
information and place an order by 'remote' means, then have the purchase
delivered at home. It is now even possible to go 'surfing' inside stores,
viewing goods and maybe deciding on purchases not foreseen prior to the
'visit', just as in a conventional shopping trip. In the last few years, such
applications of information technology have extended the potential of
teleshopping and are overcoming the reticence of many retailers about its
use. Several of the barriers to its adoption have been removed by refinements
in the technology, which may also include an 'audio' element, enabling the
customer to be guided within the virtual store, told about special offers and
led to a specific department.
A paper by Batty (Batty 1997) explores many interesting issues
concerning teleshopping, especially as regards the physical interaction
between sales staff and customers. If we assume that the use of teleshopping
services will depend on the quality and user-friendliness of product
information supplied, then its use seem likely to increase.
The large scale development of teleshopping would presumably lead to a
net decrease in the number of physical shopping trips made, especially the
longer ones, as these could be substituted in both the market search and the
purchase phase. It would also be possible to eliminate many trips involving
the distribution of goods to sales outlets, as it would no longer be necessary
to display goods nor hold stocks at local wholesale or retail outlets. Indeed, it
is possible to imagine central warehouses where goods would be stocked for
direct delivery to purchasers.
There appear to be very good prospects for the development of such e-
commerce: with the increasing diffusion of the Internet, sound foundations
have been laid for the spread of teleshopping on a worldwide basis. It
therefore seems likely that a substantial change in the level of trip demand
for shopping (above all for trips outside the city) can be expected in the near
future. This would also affect their distribution, as sales would no longer be
tied to physical structures, but to their virtual counterparts. A simple shop of
a few square meters may well sell quantities of goods comparable to those
sold in a large department store.
As mentioned above, a highly significant implication for urban mobility
would be the transformation in the whole pattern of freight distribution, due
to the modifications in the distribution chain. There would be new nodes,
and different delivery routes used as goods could be transported directly
from the producer to the consumer.
Teleservices cover a wide range of services which can be performed by
means of TLC, including telebanking, tele-entertainment, telemedicine, etc.
Also in this sector, the development potential and consequent impact on
mobility are high. Undoubtedly, many services, especially those which are
quite tedious, are eminently suitable for substitution with teleservices. For
some such services, such as many banking and postal transactions,
telediagnostics, access to data banks and libraries, information and
certification services, etc. facilities already exist and are spreading rapidly.
Finally, as regards freight transportation, TLC may in some cases
completely substitute physical transport. Already a large amount of written
matter, whether single letters or whole documents, is now 'dispatched' bye-
mail rather than sent by post. In other cases, as explained in the example
above, the physical transport of goods may be considerably modified as an
The Impact of Innovations in Telecommunications and Information Systems 281
indirect consequence of the replacement of other activities by tele-

communications.
Various studies have attempted to quantify and formulate the possible
effects of TLC substitution with the use of mathematical models. In a study
of preferences relating to communication modes for information exchange,
Moore and Jovanis (1988) used a Logit model to examine the attributes and
relative weights of various communication systems. Their survey of stated
preferences suggested that the preferred system was the telephone, followed
bye-mail and ordinary mail. Another study carried out in Italy (La Bella and
Silvestrelli 1991) set up the following substitution model:
f4 = 69.86 - 0.755f) - 0.851f2
The model relates the number (f4) of physical trips (in a year) to number
of the telephone transactions (TLC trips) (f)) and contacts through TLC
terminals (f2)' This relationship, calibrated with a survey conducted in 1987,
produces substitution coefficients close to one for both types of distance
interaction. Efforts to arrive at a rigorous quantification of substitution
effects and the formulation of appropriate models are still being undertaken.
13.4 COMPLEMENTARITY EFFECTS
13.4.1 Complementarity Effects on Demand
As already discussed above, the advent of TLC may entail a

complementarity effect as well as a substitution effect - the former resulting
from the increased efficiency of the transportation system. This improved
performance leads to a reduction in the time 'consumed' by the individual in
existing trips and thus the possibility of undertaking further activities, which
seem likely to be chiefly recreational trips. Mobility demand therefore
becomes modified in space and time as a result of new trip requirements.
Detailed studies of the actual increases in demand have not yet been
carried out, given the relatively limited diffusion of TLC systems and the
methodological difficulties involved. Systematic and reliable estimates are
therefore not yet available, but many researchers maintain that such effects
do exist and are significant.
In a study carried out in Italy in the 1980s (La Bella and Silvestrelli
1991), the relative weight of substitution and complementarity effects were
estimated for a number of components of demand mobility (see Table 13.4).
The table shows that substitution effects have the greatest relative weight
or impact for the work-work component, closely followed by services. It
seems probable that, with time, the latter will increase until it has the same
relative weight as the former. There are also substantial complementarity
effects in relaton to the work and service activities.
Table 13. 4 Relative weights of substitution and complementary effects
Activity type Substitution Complementarity

Work-work 0.36 0.34
Services 0.29 0.25
Shopping 0.14 0.14
Home - work 0.12 0.15
Leisure 0.09 0.12
Total 100.00 100.00
Average 0.42 0.58
In the same study, a survey was performed to determine the sensitivity of

the various components of the mobility and traffic system to the introduction
of TLC-based innovations (Table 13.5). The survey shows that TLC could
have considerable impact in the transportation supply sector, as will be
explained in the next section, especially as regards traffic control, affecting
both public and private transportation.
13.4.2 Complementarity Effects on Supply
By complementarity effects upon supply, we refer to the impact of TLC

technologies in enhancing the efficiency of physical trips, increasing the
comfort and safety of the trip (Amott et al 1991) and often reducing the
travel time. There are two types of TLC-based technological systems for
increasing the efficiency of transport supply: ATIS (Advanced Transport
Information Systems and ATCS (Advanced Transport Control Systems). The
former supplies information to the user about the system state, possibly
providing a short-term forecast, while the latter supports a transport control
system so as to ensure maximum efficiency in the operation of the system
itself.
When an individual considers undertaking a trip, after deciding whether

and where to go, he or she combines the information available about the
probable state of the transportation system relating to the origin, path and
destination, then decides the transportation mode to use, the departure time
and detailed path (roads or set of public transportation lines) to follow.
Table 13.5 Sensitivity of mobility system components to

introduction ofTLC
Components Effect (%)
Demand structure 100

frequency 0.52
distribution in time 0.18
duration 0.16
length 0.14
transport supply 100
public transportation 0.73
private transportation 0.27
public transportation 100
traffic control 0.43
on-board TLC 0.27
organisation 0.19
planning 0.11
private transportation 100
traffic control 0.56
on-board TLC 0.28
planning 016
environmental effects 100
congestion reduction 0.56
energy saving 0.28
pollution reduction 0.16
system effects 100
overall efficiency 0.52
demand structure 0.25
demand volume 0.23
User information systems consist in general of three subsystems: one for

monitoring the system state, one for data transmission to a control centre and
one for elaborating and transmitting information to various users (firms,
individuals, external systems). There are several types of ATIS system,
which may be classified:
a) according to the time horizon of the information;

b) according to the trip phase in which information is supplied;
c) according to information type.
As regards the time horizon, information for the user can be of two types
(Ben Akiva et aI1991):
historic information: this consists of all the information the user

possesses about the probable system state (during the period in which he
intends to travel, both for the outward and return route) based on his or
her own experience and that of others. There is also information supplied
by the system itself. As we shall see below, both outward and return
journeys should be considered since both condition the user's choice of
mode, departure time and return time, and routes taken;
current information (real time): this consists of all information based on
the traffic situation (transport system attributes) at the moment of
interrogation. Such information, provided electronically or by radio, may
be descriptive (reactive control) or predictive (anticipatory control).
Predictive information concerns the traffic situation (possible transport
system attributes) during the time periods in which the outward and
return trips are to be undertaken. Such information is tied to TLC
technology development and predictive models of traffic conditions. The
latter, in particular, are difficult to construct because there is a mutual
interrelation between information and user behaviour, i.e. there are
problems concerning feedback cycles between information and
description of the system state.
As regards the second criterion, information may be supplied to the user

during two travel phases:
pre-trip communication by telephone, radio, computer, etc.;

en route communication by telephone, radio or on-board computer in the
case of private transportation, or at stops and stations on appropriate
display boards for public transportation.
The Impact ofinnovations in Telecommunications and Infonnation Systems 285
As regards the third criterion, the information may be descriptive or

indicative. The user is assumed to choose whether or not to follow the
indication contained in the information. There are currently few
experimental results for large scale applications of ATIS to real systems.
Current knowledge of possible effects is primarily based on theoretical
studies, some of which will be reviewed in Section 13.4.
Advanced transport control systems involve the real-time management of
operational features designed to optimise system performance, i.e. its
efficiency and effectiveness. The main types of ATCS systems can be
classified as follows:
a) for private transportation

b) for roads sections
c) road pricing systems
d) speed lane control systems
e) for junctions
t) network access control systems (ramp metering)
g) traffic light network control systems
h) for public transportation
i) systems for real-time location and emergency management of bus fleets
A systematic review is outside the scope of this chapter and the interested
reader is referred to Papageorgiou (Papageorgiou 1991) and the Proceedings
of the First World Congress on Applications of Transport Telematics and
Intelligent Vehicle-Highway Systems (Ertico 1995), which describe a
number of theoretical and empirical studies of such systems.
13.5 MATHEMATICAL MODELS FOR EVALUATING

THE EFFECTS OF USER INFORMATION
SYSTEMS
As noted in the previous section, the introduction of TLC can have

significant impacts on mobility demand and on the operating conditions of
the urban transport system. In the last few years considerable efforts have
been devoted to developing mathematical models capable of predicting such
impacts.
Several studies of the implications of TLC applications in mobility
demand have been made with the use of stated preference (SP) techniques,
(Bernardino et al 1994; Mahamassani and Jayarishnan 1991). Some
examples were reported in the previous sections. However, the aspects
relating to the impact of user information systems are generally highly
286 Case etta E. and Montella B.
complex and the results less consolidated. This does not hold for models
simulating the complementarity effects of ATIS and ATCS systems. These
have been the subject of various research programs in Europe (DG XIII
framework programs such as DRIVE 1,2 and 3), the USA (IHVS) and Japan
(interest being motivated at least in part by the industrial interests involved).
It is worth pointing out (Shofer et at 1993) that decisions regarding the
implementation of ATIS systems should be based on the social as well as the
private benefits expected from the technology. But such benefits will depend
largely on how users respond to such sources of information.
Given the complexity of demand and supply interactions involved, ATIS
systems need to be designed, evaluated and operated on the basis of
sophisticated mathematical models able to simulate the performance of the
transport system. To adequately capture the possible impacts and to evaluate
the potential benefits of an ATIS in specific area, such models must have a
series of specific requisites and, in particular, be able to represent:
the uncertainty of the control system (concerning information acquisition

and transmission, and any delays) in real time and with a flexible
stochastic model;
the dynamics of transportation supply and demand in relation to the
schedule ofthe outward and return trips, as well as day-to-day variations;
user heterogeneity, especially in the reactions to information provided. A
particular segmentation of mobility demand must thus be considered not
only with classical static prediction models, it must also represent the
different behaviour of different types of users.
The effects on users of real-time TLC information about the state of the
transport system may be simulated on the basis of behavioural theory by
means of random utility models, which are now widely used in transport
systems modelling. Random utility models simulate the choice behaviour of
the transport system user, who is considered to be a 'rational' decision-
maker, i.e. an individual who follows three fundamental behavioural
hypotheses (Ben Akiva and Lerman 1985; Cascetta, 1990):
in making the choice, the user considers a set of available alternatives and
associates a perceived utility with each alternative;
the alternative chosen is that which has the greatest perceived utility;
the utility associated with each alternative can be described as a function
of attributes relating to the alternative (e.g. the level of service and the
decision-maker's socio-economic characteristics).
The Impact of Innovations in Telecommunications and Infonnation Systems 287
Moreover, the utility associated with each alternative is not known a

priori by the analyst and is therefore a random variable which owes its
randomness to variations in individual taste, to the incompleteness of the
model attributes, to evaluation errors and the intrinsic randomness of the
choice process. In A TIS systems, in addition to the possible 'intrinsic'
sources of randomness relative to the information acquisition process, it is
necessary to take into account the user's capacity to understand, evaluate and
process the information.
The information supplied to the user may, as stated above, be pre-trip or
en route and entail different decisional levels, or relate to 'choice
dimensions'. As regards pre-trip information, the user is able to choose the
transportation mode, departure time and best route or even the possibility of
cancelling the trip altogether. On the other hand, en-route information made
available during the trip enables the user to choose (or vary) only the route.
In the following sections we briefly describe the main characteristics of
models which may be used to simulate behaviour regarding the choice of
mode, time and route (Sections 13.6 and 13.7) and demand-supply
interactions (Section 13.8). Finally, Section 13.9 describes some systems of
models used for overall ATIS policy evaluation, and presents the results of
the simulations.
13.6 EFFECTS ON MODAL CHOICE
The modal choice model supplies the fraction pC (mlodsh) of trips of user
category c who use mode m to reach zone d from zone 0 for purpose s in
time slice h. The modal choice models used almost always have a
behavioural interpretation in practice, as their name would suggest. As
regards the functional form of modal choice models, the Multinomial Logit
is the one most frequently used:
(13.1)
where the systemic utilities for each mode m, Vm , are calculated by using the
level of service attributes of the trip between 0 and d in time slice h and the
socio-economic attributes of the market segment, or individual, c.
Increasing use is being made of Nested Logit modal choice models,
which can introduce correlation structures between perceived utilities of
various mode groups, such as individual modes and public modes, and/or
between various services within the same mode (e.g. slow and fast lines).
As stated above, the attributes which appear in the utility function of a

modal choice model are generally level of service attributes and socio-
economic attributes. Level of service attributes concern the characteristics of
the transport service, such as travel time, service regularity, number of
transfers and so forth. In fact these attributes are usually assessed by the user
on the basis of his or her historic knowledge of the system in question.
However, a user who is informed in real time by an ATIS system about the
actual situation, may significantly change behaviour and achieve a reduction
in the value of the modal constant, as this incorporates aspects not explicitly
predicted by the model. These include the expected variance of attribute
values and a lower error variance of attribute perception.
Information also plays a significant role in the correct perception of the
availability of modes not used regularly. This holds particularly for public
transport and non-habitual trips.
13.7 EFFECTS ON DEPARTURE TIME AND ROUTE

CHOICE
The route choice model supplies the number, p(k/mods), of trips using
route k, by mode m, to go from 0 to d for purpose s. The route choice models
most commonly used are all behavioural models, although hypotheses on
user behaviour differ for road route choice and trips using the public
transportation network. In the case of road networks, it is normally assumed
that the user chooses the route before starting the trip (preventive choice). In
this case, it is assumed that the variables affecting route choice are the
service attributes, known either historically or through real-time information.
However, the level of service information obtained is generally
'negative', or in other words represents a 'cost' for the driver (travel time,
monetary cost). Hence, reference will be made below to the perceived cost
1\
C k relative to route k. This perceived cost may be expressed as:
(13.2)
where Ck is the average perceived cost of route k, & is the random residual
and fo d m is the set of 'feasible' paths linking the origin/destination (o/d) pair
with mode m.
Various types of behavioural model may be used to obtain route choice
probabilities. The most commonly used random utility models are the Logit
and Probit. The former is specified as follows:
exp[-aC ]
p(k I mods) = k (13.3)
"L..,heiodm exp[- aCh ]
However, the hypothesis of identically and independently distributed

random residuals, which underlies the Logit model, is barely acceptable
theoretically when alternative routes have several links in common. As an
alternative, it is possible to use a Probit model, where the variance in
perception error is assumed to be proportional to route cost, and the
covariance of the error for two routes is proportional, according to the same
parameter 0, to the cost of the links belonging to both routes (Ch k) (Sheffi
1985; Cascetta 1990).
The computational problems entailed by the Probit model can be avoided
by using a modified form of the Logit model, known as the C-Logit model,
recently proposed by Cascetta et aT 1996. The latter requires explicit path
enumeration, which is unavoidable for simulating within-day dynamics.
When additional information is received during the trip (e.g. through on
board route guidance or variable message signs), further choice of path may
occur en route. In this case, the same class of models can be used for
descriptive information, but with additional compliance problems in the case
of prescriptive information.
Where public transport networks are concerned, the picture is quite
different. Changes need to be made both to the definition of the single choice
alternative and the expression of average perceived cost of an alternative.
The single choice alternative is generally no longer the elementary route, i.e.
the sequence of links connecting an old pair, but a set of links which is
called a hyperpath. It is assumed that a user who undertakes a trip in a high
frequency public transportation system does not initially have all the
information necessary to be able to take a complete decision. For example,
the user is not able to predict his exact arrival time at the various stops
andlor the vehicle schedule on the various lines (trains, buses, etc.) serving
~ach stop. Under such conditions, the user does not choose a predetermined
route, but a travel strategy which is defined by a set of preventive choices
fixed prior to starting the trip, and adaptive choices, i.e. decisions that the
user takes during the trip in adapting to random or non-predictable events
(Cascetta and Nuzzolo 1986).
Adaptive choices are assumed to occur most commonly at stops: at stop i
the user is presumed to follow the rule of boarding the first passing vehicle
in the set of alternative or 'attractive lines' L iH. Hence, the probability Pij,H of
boarding line j from stop i is equal to the ratio between the average service
frequency of line j (fPj) and the cumulative frequency of all the lines
belonging to the set LiH of attractive lines in i:
(13.4)
Under these adaptive behaviour assumptions, a strategy, i.e. a preventive

choice alternative, may be represented on a network graph of public
transport services, as a sub-graph or 'hyperpath'. The average cost GH of the
hyperpath H may be calculated as a weighted average of the travel times
associated with on-board links relative to all the lines belonging to the
strategy and waiting time depending on cumulative frequencies at stops.
As regards the hyperpath choice model, model (4) may be extended by
substituting the average hyperpath cost GH with the path cost Ck. For further
details on hyperpath models and calculation algorithms, see the specialised
references (Nguyen and Pallottino 1986, 1988; Florian and Spiess 1989).
Both in the case of road systems (continuous service) and public
transport systems (discontinuous service), the path choice model may be
extended to include the departure time, using the classic Logit model
(Cascetta et al 1992). The model specification usually adopted takes into
account route characteristics, travel time and schedule delay variables (Small
1982). The choice set includes every available pair (h,k), where h is a time
interval, for example 15 minutes, and k is one of the available routes. An
indifference time band on the desired arrival time is typically considered in
the model specification to take into account a non linear perception of small
'early' or 'late' arrival discrepancies with respect to desired or target time
arrival.
From what has been said, it is evident that both pre-trip and en-route
information provided by ATIS can influence route choice and that pre-trip
information can significantly influence departure time choices. Various
models have been proposed to capture these phenomena.
13.8 DEMAND-SUPPLY INTERACTION MODELS

(ASSIGNMENT)
Assignment models used to simulate ATIS systems must allow

representation of within-period dynamics, i.e. allow simulation of within-day
variations of the system state (expressed in terms of flow characteristics such
as link flows, queues, etc.) and any temporary oversaturation of capacity, the
forming of queues or spill-backs. Such phenomena can only be represented
very approximately with traditional 'within-day static' assignment models.
The increase in urban road network congestion and the potential offered by
ATIS and ATCS justify the considerable interest in dynamic assignment

models shown within the scientific community. Within-period dynamic
assignment models explicitly consider variations within the reference period
(whole day, rush-hour, etc.) in demand (due to departure time choice
behaviour) and/or supply (due to the presence of real-time control systems,
accidents, etc.) (Ben Akiva and Lerman 1985; Boyce 1988; Cascetta and
Cantarella 1993). The use of such models also allows us to take into account
any modifications to the initially chosen route occurring in response to
traffic conditions encountered and/or to information or route guidance
provided en-route.
Adapting static demand and supply models or assignment models to
within-period dynamics is not straightforward. In general, a discrete time
representation is adopted in which the reference period is assumed to be
subdivided into intervals. A continuous time representation is possible, at
least in theory, but its solution requires some form of discretisation.
Dynamic assignment models may be classified into two groups according
to the approach used in simulating decision-maker behaviour and generating
network flows. In the aggregate approach (macrosimulation), the overall
effects of user behaviour are simulated as in the static case; in the non-
segregate approach (microsimulation) the behaviour of each user is explicitly
simulated. Non-segregate models may be more sophisticated and accurate,
but are not in general suited to large-scale applications because of the
computational power required. Meso-simulation models are thus
increasingly numerous. These use intermediate aggregation levels for the
network and performance functions of groups of individuals (packets),
representing explicitly the movements of such groups on the network.
It is is particularly difficult to specify supply models for within-period
dynamics since, unlike the static case, link flow characteristics cannot be
uniquely represented, as averages in space and time do not coincide. Some
of the proposed dynamic assignment (dynamic process) models explicitly
simulate the system evolution from one period to another (day-to-day
dynamics). They may be viewed as a generalisation of the tradition
equilibrium paradigm, insofar as they explicitly simulate user learning
mechanisms and knowledge review, including possible habitual behaviour.
They therefore permit analysis of a broader class of problems, amongst
which is the effectiveness of ATIS and ATCS dynamic systems based on
real time information.
Between-period dynamic models adopt two different approaches.
Deterministic process models, based on non-linear dynamic systems theory,
simulate system evolution with a system of recursive equations. This
explicitly defines the predicted link cost vector in the generic reference
period, based on the costs incurred in the previous periods, and thus takes
into account user behaviour, defined by the link flow vector. Stochastic
process models, on the other hand, model the evolution of the
demand/supply system as a random process. Conditions may be identified
for which there is a stationary probability distribution, allowing the averages
and dispersion of the main variables to be obtained, such as link flows
(Cascetta and Cantarella 1993).
13.9 MODEL APPLICATIONS FOR EVALUATING

ATIS SYSTEM BENEFITS
Applications of the models described above should provide a response to

a number of issues regarding the suitability of the various ATIS systems for
individual users and for society as a whole. The kind of questions we need to
answer may be expressed as follows:
1) how many users should be informed?

2) how should the information be communicated?
3) how should the information be used?
4) how extensive should the relative information be?
Attempts to provide responses to these questions have usually been based

on simulation exercises or small scale empirical tests, which make it
impossible to draw any general conclusions quantifying the effectiveness of
pre-trip or en route information.
As regards public transport, some simulation experiments have been
performed to evaluate the benefits of real-time information (Montella 1996).
In these experiments, user time saving was evaluated by formalising a model
to estimate average user disutility due to lost time in a scenario with no
information available on the transport system characteristics. This was
compared with a scenario where information allowed users to choose the
best route in terms of travel time, comfort etc. (Anderson 1993). In the
former scenario, the user does not know real traffic conditions, nor actual
frequencies, nor schedules at the stops of public transport lines, and adopts
preventive behaviour by comparing perceived utilities in the choice of stop,
and a completely indifferent adaptive behaviour in choosing which vehicle
to board from the set of attractive lines. When there are several possible
alternative lines, the user will choose according to their relative frequencies
at the stops, i.e. by boarding the first which is compatible with the
destination.
User behaviour is very different in a scenario where there is a 'pre-trip'
information system capable of providing details on the actual conditions of
The Impact of Innovations in Telecommunications and Infonnation Systems 293
the public transport system before the trip takes place. This knowledge
ensures that all the combinations 'stop at origin - run - stop at destination'
are seen by the user as primary alternatives. To evaluate the user's average
time saving with a pre-trip information system, generalised path costs were
evaluated. The results obtained show an average potential generalised cost
saving of approximately 25% in small-scale cases.
A different procedure has been defined for evaluating saving in user time
deriving from full real time knowledge of the transport system obtained the
trip, usually at the stop (Biggiero et al 1996). The time saving, in terms of
generalised costs, was calculated through a procedure for obtaining the
service level provided to users in two different scenarios with different levels
of information available. The attributes known to affect choice are total
travel time, waiting time, degree of crowding, and the number of transfers
made. When this information is provided by the ATIS, the user chooses the
set of service which minimise his or her generalised cost (Balogh and Smith
1992; Baneres 1990; Murnaghan 1993). Once the system state is known, two
mechanisms affecting choice come into play: the first involves a reduction in
the weight that the user attributes to waiting time, insofar as it no longer
constitutes an uncertainty (in fact, if known a priori, it may be used
differently); the second involves a trade-off between waiting time and travel
time, i.e. a longer wait at the stop may be preferred in order to travel in a
faster vehicle. Overall, generalised cost reductions of 10-12% were
achieved.
In private transport (car networks), the time-saving effects due to the
introduction of TLC technologies providing drivers with real time
information have been studied since the mid 1980s (Cascetta and Cantarella
1991). An example of an evaluation model which includes driver response to
information, network flow variations and new used paths, is that developed
by Jayakrishnan et al (1994) and known by the acronym DYNASMART
(Dynamic Network Assignment Simulation Model for Advanced Road
Telematics) (Peeta and Mahamassani 1995; Hu and Mahamassani 1995).
This model has the following features:
the ability to model the route choice of drivers with and without access to
information;
responsiveness to dynamic O-D information available to the controller;
the ability to track the locations of all drivers;
the ability to predict the travel time based on the assignment results;
the ability to model real traffic control strategies in the network;
the ability to comprehensively model both freeway and surface street
traffic.
The DYNASMART simulation model was a meso-simulation model and

estimated the generalised cost reduction for drivers exploiting the received
information for average trips in urban areas in standard conditions to be
around 8-10%, whereas for states of emergency (accidents) it was as much
as 20-30%.
13.10 CONCLUSIONS
The application of TLC is playing an important role in the mobility of

people and of goods at all scales, from the urban to the international. The
impacts range from the total substitution of physical trips with virtual ones
(telecommuting, teleshopping, and so on) to various complementary effects
which lead to new demands for mobility. The substitution effects, even
though not free from social problems (i.e. the reduced opportunity for
personal contact), are spreading rapidly especially in the field of secondary
trips - from shopping to social services - and seem likely to expand very
quickly in the near future. Many complementary effects of TLC applications
are already evident, but rather more difficult to examine systematically and
extremely hard to predict. Studies of the use of TLC and information
systems within the transport field itself, both theoretical and experimental,
are demonstrating the validity of such applications and their potential for
achieving greater efficiency in traffic management. Greater attention should
however be paid to the integration of these innovative technologies with
mobility, transportation, urban planning, and the environment on the part of
operators and policy-makers in all these fields, as well as the industries
producing vehicles and TLC systems.
REFERENCES
Anderson, T. (1993) Real-Time Passenger Information for Transit Systems: Proposal for the
Jubilee Line Extension. PTRC Summer Annual Meeting, 21
Arnott R. et al (1991) Does Providing Information to Drivers Reduce Traffic Congestion?
Transportation Research 25A (5): 309-18
Balogh, S., Smith, R. (1992) Real-time Bus Information - The London Transport Route 18
Demonstration. PTRC Summer Annual Meeting, 20
Baneres,1. (1990) Automatic Vehicle Monitoring System for the Bus Network of Barcelona.
URBANI A Congress, Padova Italy
Batty, M. (1997) The Retail Revolution. Environment and Planning B, Planning and Design
24: 1-2
Ben Akiva, M.E., Lerman, S.R. (1985) Discrete Choice Analysis. MIT Press, Cambridge
Mass
Ben-Akiva, M., de Palma, A., Kaysi, I. (\991) Dynamic Network Models and Driver
Information Systems. Transportation Research 25A (5) 251-66
Ben-Akiva, M., Bowman, J., Dinesh, G., (\996) Travel Demand Model System for the
Information Era. Transportation 23: 241-66
Bennison, DJ. (1988) Transport/Telecommunication Interactions: Empirical Evidence from a
Videoconferencing Field Trial in the United Kingdom. Transportation Research 22A (4)
291-300
Bernardino, A., Ben Akiva, M., Salomon, I. (1993) Stated Preference Approach to Modeling
the Adoption of Telecommuting. Transportation Research Record 1413: 22-30
Biggiero, L., Di Gangi, M., Montella, B. (1996) A Methodology for Estimating Effects on
Users of Introducing TLC Into Urban Public Road Transport. Applications of Advanced
Technologies in Transportation Engineering, ASCE New York
Boyce, D.E. (1988) Route Guidance Systems for Improving Urban Travel and Location
Choice. Transportation Research 22A (4) 275-81
Button, K., Maggi, R. (1994) Videoconferencing and its Implications for Transport: An
Anglo-Swiss Perspective. Transport Reviews 15(1): 59-75
Cantarella, G.E., Cascetta, E. (1996) Stochastic Assignment to Transportation Networks:
Models and Algorithms. CTR XXV, Montreal 10-11 October
Cascetta, E. (1990) Metodi quantitativi per la pianificazione dei sistemi di trasporto CEDAM,
Padova Italy
Cascetta, E., Cantarella, G.E. (1991) A Day To Day and Within Dynamic Stochastic
Assignment Model. Transportation Research 25A (5): 277-91
Cascetta, E., Cantarella, G.E. (\ 993) Modelling Dynamics in Transportation Networks: State
of the Art and Future Developments. Journal ofSimulation Practice and Theory I: 65-91
Cascetta, E., Nuzzolo, A. (\986) Uno schema comportamentale per la modellizzazione delle
scelte di percorso nelle reti di trasporto pubblico urbano. Proceedings of the 4th PFT CNR
National Congress, Turin, Italy
Cascetta, E., Nuzzolo, A., Biggiero, L. (1992) Analysis and Modelling of Commuters'
Departure Time and Route Choice in Urban Networks. Proceedings of 2nd International
Seminar on Urban Traffic Networks, Capri 5-8 July
Divieti L. (1995) "Trasporti e telecomunicazioni: potenziali interazioni tra spostamenti fisici e
spostamenti immateriali." In Sviluppi della ricerca sui sistemi di trasporto, SlOT, Franco
Angeli, Milan Italy, 185-201
ERTICO (1995) Proceedings of the First World Congress on Applications of Transport
Telematics and Intelligent Vehicle-Highway Systems
Florian, M., Spiess, H. (1989) Optimal Strategies: a New Assigment Model for Transit
Network. Transportation Research 5: 152-74
Garrison, W.L., Deakin, E. (1988) Travel, Work, and Telecommunications: a View of
Electronics Revolution and its Potential Impacts. Transportation Research 22A (4): 239-
45
Hamer, R, Kroes, E., van Ooststroom (1991) Teleworking in the Netherlands: an Evaluation
of Changes in Travel Behaviour. Transportation 18: 365-82
Hamer, R, Kroes, E., van Ooststroom, Kockelkoren, M. (1992) Teleworking in the
Netherlands: Evaluation of Changes in Travel Behaviour - Further Results. Transportation
Research Record 1357: 82-9
Hansen, M., Huong, Y. (1997) Road Supply and Traffic in California Urban Areas.
Transportation Research A 31 (3): 215-18
Henderson, D.K., Mokhtarian, P.L. (1996) Impacts of Center-Based Telecommuting on
Travel and Emissions: Analysis ofthe Puget Sound Demonstration Project. Transportation
Research 01(1): 25-49
Hu, P.S., Young, 1. (1992) Nationwide Personal Transportation Survey: Summary of Travel
Trends. Report FHWA-PL-92-027. Federal Highway Administration Office of Highway
Information Management, Washington DC, March
Hu, T., Mahamassani, H. (1995) Evolution of Network Flows Under Real Time Information:
A Day To Day Dynamic Simulation-Assignment Framework. Transportation Research
Record, 24-31
Jayakrishnan, R., Hu, T., Mahamassani, H. (1994) An Evaluation Tool for Advanced Traffic
Information and Management Systems in Urban Networks. Transportation Research 2C
(3): 129-47
Kitamura, R., Nilles, 1., Fleming, D., Conroy, P. (1990) Telecommuting as a Transportation
Planning Measure: Initial Results of State of California Pilot Project. Transportation
Research Record 1285: 98-104
Koenig, B. E., Henderson, D. K., Mokhtarian, P.L. (1996) The Travel and Emissions Impacts
of Telecommuting for the State of California Telecommuting Pilot Project. Transportation
Research 4C (1): 13-32
La Bella, A. Silvestrelli, M.A. (1991) Telecomunicazioni e trasporti nella citta del futuro.
Proceedings of the 5th PFT CNR National Congress, Rome
Mahamassani, H. S., Jayakrishnan, R. (1991) System Performance and User Response Under
Real-Time Information in a Congested Traffic Corridor. Transportation Research 25A (5):
293-307
Mahamassani, H., Yen, 1., Herman, R., Sullivan, M. (1994) Employee Attitudes and Stated
Preference Toward Telecommuting: An Exploratory Analysis. Transportation Research
Record 1413
Mokhtarian, P.L. (1991) Telecommuting and Travel: State of the Practice, State of the Art.
Transportation, 18(4): 319-42
Mokhtarian, P.L. (1988) An Empirical Evaluation of the Travel Impacts of Teleconferencing.
Transportation Research, 22A (4): 283-88
Mokhtarian P.L. (1997) Emerging Travel Patterns: Do Telecommunications Make a
Difference? 8th Meeting of the International Association for Travel Behavior Research,
Austin Texas, September
Mohtarian, P.L., Salomon, I. (1997) Modelling the Desire to Telecommute. Transportation
Research 31A: 35-50
Montella, B. (1996) A Methodology for Estimating Effects on Users ofIntroducing TLC into
Urban Public Road Transport. Applications of Advanced Technologies in Transportation
Engineering, ASCE, New York
Moore, A., Jovanis, P.P. (1988) Modelling Media Choice in Business Organizations:
Implications for Analyzing Telecommunications - Transportation Interactions.
Transportation Research 22A (4): 257-73
Murnaghan, P. (1993) Real Time Information at Bus Stops Romance Project Southampton.
PTRC Summer Annual Meeting, 21
Nguyen, S., Pallottino, S. (1986) Assegnamento dei passeggeri ad un sistema di linee urbane:
determinazione degli ipercammini minimi. Ricerca Operativa 38: 207-30
Nilles, 1. M. (1988) Traffic Reduction by Telecommuting: A Status Review and Selected
Bibliography. Transportation Research 22A( 4): 301-17
Papageorgiou, M. (1991) Traffic & Transportation Systems. Pergamon Press, New York
Peeta, S. Mahamassani, H. (1995) Multiple User Classes Real Time Traffic Assignment for
Online Operations: A Rolling Horizon Solution Framework. Transportation Research 3C
(2): 83-98
Salomon, I. (1985) Telecommunications and Travel: Substitution or Modified Mobility?
Journal a/Transport Economics and Policy 19(3): 219-35
Salomon, I., Koppelman, F. (1988) A Framework for Studying Teleshopping Versus Store
Shopping. Transportation Research 22A(4): 247-55
Salomon, I., Schneider, H.N., Schofer, J.L. (1991) Is Telecommunicating Cheaper than
Travel? An Examination of Interaction Costs in a Business Setting. Transportation 18:
291-318
Schofer, J. Khattak, A., Koppelman, F. (1993) Behavioral Issues in the Design and Evaluation
of Advanced Traveller Information Systems. Transportation Research 1C (2): 107-17
Selvanathan, E.A., Selvanathan, S. (1994) The Demand for Transport and Communication in
the United Kindom and Australia. Transportation Research 28B (1): 1-9
Small, K.A. (1982) The Scheduling of Consumer Activities: Work Trips. The American
Economic Review 73 (3): 467-79
US Department of Transportation (1993) Transportation Implications of Telecommuting US
Government Printing Office
Subject Index
abstract knowledge, 199 Arrow's learning-by-doing. See learning-by-

accessibility: doing
to business and public services, 257 assignment models:
and knowledge networks, 248 between-period dynamic, 291
to markets, 256 within-period dynamic, 291
transport and telematic, 248 ATCS. See Advanced Transport Control
adaptive learning, 21, 23, 39,69 Systems
Advanced Transport Control Systems ATIS. See Advanced Transport Information
(ATCS). 282, 286 Systems
types ot~ 285 attractors, fractal and strange, 83

Advanced Transport Information Systems
(ATIS), 282, 286-87 barriers to development, mis-specification of,
evaluating, 292-294 266
types of, 284 beliefs, mutually reinforcing or mutually
agents: competitive, 73
knowledge-intensive, 109. See also best-practice techniques, 18-19
scientists bifurcation diagram, 82-83, 86
nonlinear interactions between, 100 bifurcation processes, 250
R&D by profit-seeking, 198 bounded rationality, 22
speculative behaviour of, 76, 89, 95 brain, 54
agglomeration economies, 5, 22, 109-10, as a complex dynamic system, 46, 72
193. See also increasing returns; mind as an emergent property of the, 59
positive feedback neurochemistry of the human, 46, 72
and NP-completeness, 147 butterfly effect, 88, 90-91
proxies for, 122

supply-side generation of, 171-72 causality, challenges of evolutionary, 45-46
algorithm, greedy, 151 catastrophe theory, 61
analysis, creative, 160 central place systems (CPS):
location theory and, 168-70
300
theory of, 166 connectivity, network, 71

chaebols, 25 core technologies associated with network
chance events, 241 infrastructure, 3 I
role in urban development, 100 CPS. See central place systems
chaotic dynamics, 79-81 CPS-model, 166-67:
citations, influence of distance on, 129, description of, 169-70
Table 6.1 extended, 193
cities, rank-size relationship of, 47 and specialisation dynamics, 17 I -72, 193
city: CPS-products, 193
as a complex dynamic system, 4, 58 CPS-ranking of regions, 186
evolution of a, 45-46 cultural amenities, importance of, 132
city-size distributions in selected countries, cumulative causation, Fig. 3.7
Table 3.1 currency strength, definition of, 226
codified knowledge, tacit and, 6
co-evolutionary learning, 4, 26, 38, 60. See deductive reasoning, 67, 72
also adaptive learning demand-based theories, 13, 28
on congested traffic networks, 69 demand-pull model, 12-13,29
uneven pattern of, 70, 73 departure time, modelling the effects of, 288-
co-evolutionary processes, 46, 54, 250 90
between cities, 72 dependence. See path-dependence
collective action: derivatives, 96-97
as an outcome oflearning, 100 need for and criticism of, 99-101
self-cancelling nature of, 91 distance:
competitiveness, exchange rate influence on learning, 276. See a/so teleconferencing
regional, 222 penalty of, 256, 263
complementarity of transport and DLA. See dynamic location advantages
telecommunications, 272, 281-85 DLA-model, 166-67, 193
effects on: DNA of cities, 73
demand, 281-82 dualism, spatial or sectoral, 75
supply, 282-85 dynamic location advantages (DLA),
relative weights of substitution and, nonlinearity of, 174
Table 13.4 dynamic programming, 141, 151
complex dynamics, 4-5. See also nonlinear dynamics:
dynamics continuous or discrete, 78
ofa region's internal and external of industrial laboratory location, 110
relationships, 183 nested (slow and fast), 71, 86
complex dynamic systems: nonlinear. See nonlinear dynamics
cities and brains as, 4, 54, 58 dynamic stability, 93
ergodic theory of, 81 DYNASMART (Dynamic Network
highly interactive, 75 Assignment Simulation Model for
congestion, 69, 206 Advanced Road Telematics), 293-94
and transportation costs, 272
301
e-commerce, 280 effect of distance on frequency of, 131

economic geography of Europe, 194 importance of, 5
economic growth: knowledge spillovers via, 198
determinants of, 202 FDI. See foreign direct investment
evolutionary model of, 198 feedback, positive. See positive feedback
and innovation, 12 financial derivatives. See derivatives
neoclassical model of, 197-98 follower regions, 184
economic milieu, 171 foreign direct investment (FDI), 221-27, 232
adaptation of an urban, 183, Fig. 8.2 correlation with trade, 224
components ofa region's, 172-73, 177- effect of real exchange rates on, 222,
80, 194 225-27
diversity of a region's, 173-74 forms of, 221, 226
\
education: foreign investment and technology transfer, 6
organised by disciplines, 139 fractals, 52
importance of university, 188-89, 192, freight transportation, 280-81
Table 8.8 functional urban regions (FUR), 199-209
electronic cottage, 255, 265. See also definition of, 215-16
telecottage FUR. See functional urban regions
emergence, 4 futures, 76, 97-98
definition of, 59 definition of, 98
unpredictability of, 58
urban change portrayed as co-evolution globalisation, 76
and, 46 gravity model, 131
emergent behaviour. See emergence as central tool of urban analysis, 140
emergent properties, 27, 72 entropy-maximising, 143-44
of complex systems, 58
European urban system, 6, 197-221 Hayek's theory of knowledge diffusion, 22-
evolution: 23
description of, 2 hierarchies:
micro-foundations of technological, 12- time-space, 184-89
17 of urban regions, 186. See a/so urban
role of innovation in urban, 3 hierarchies
self-reinforcing, 123 human capital:
theories of, 4, 11 difficulties of attracting, 206
evolutionary causality, difficulties of, 45-46 and innovation, 198
evolutionary models, 19
exchange rate, 6, 222 imitation, 19, 27, 158-59
expectations, central role of, 95 early, 184-85
exponential growth, Malthusian model of, increasing returns, 21, 63
82 in urban growth and development, Fig.
3.7
face-to-face contacts, 111,278 definition of, 68
302
micro-foundations of technological, innovation systems, national. See national

12-17 innovation systems (NIS)
at the regional level, 110-11
to scale and scope, 174 institutional change, evolutionary character
inductive reasoning, 67, 72-73 of, 24
definition of, 68 interaction:
industry Iifecycle model, 26 importance of, 245
information technology. See new learning and mutual, 251
information technologies (NIT) models, 290-92. See a/so assignment
infrastructure: models
difficulty of justifying investment in intervening opportunities, model of, 140
new, 257 intuition, 67
facilitating role of, 177 invention, role in planning, 148-49, 157-60
importance of air transport, 203 invisible colleges, 131
lifecycle patterns of renewal and
decline, 180 Japan's urban development:
provision of advanced, 260-62 knowledge-intensive nature of, 5, 109-
initial conditions: 27
altering, 89
sensitivity to, 77, 79-80 know-how, 26, 31, 241
innovation: knowledge exchange:
basic and marginal, 84-92, 245 path-dependent processes of, 5
concept of, 3, 238-39, 251 and trade, 223
cumulative and hierarchical, 91 knowledge-intensive milieu, 39, 192
difficulty of modelling, 84 knowledge networks, 132-33
effects of customer sensitivity on, 263 definition of, 127
evolutionary theories of learning and, scholarly, 129
11-39 knowledge spillovers, 198-99, 202, 214
hard and soft, 244-48 diminishing with distance, 210
in leader and follower regions, 185-85 from foreign subsidiaries, 221
and nonlinear dynamics, 83-92 know-what, 26
as product of knowledge and learning, I Kondratiev waves of development, 239
taxonomies and typologies of, 7, 240-
42, Table 11.2 language, isolating effects of, 133
and trade, 223 leader regions, 184-86
in urban planning, 157-60 in Europe, 194
innovation diffusion, 12 frequency of innovations in, 184-85
path-dependent models of, 21 knowledge-oriented, 191-92
perspectives in the analysis of, 247 (Fig. leads and lags:
11.2) model of, 166
stages of, 246 regional, 6, 167
303
learning: media choice model, 276-81

belief-conditioned, 30 mental models, 67-68, 245
co-evolutionary. See co-evolutionary merchants:
learning decision problem of travelling, 64-67
collective processes of, 2 Venetian, 65
and innovation: mesosimulation, 291
evolutionary theories of, 11-39 metaphor, 159
local and cumulative. 17 metropolitan development. See urban
and mutual interaction, 251 evolution
technological, 25 mobility:
and development, 31-33 socio-economic factors enhancing, 271-
role of chaebols in, 25 72
theories of, 4-5 substitution effects affecting, 249, 274-75
learning-by-circulating, 59-68 modal choice models, 287-88
learning-by-doing, 13.20,38 modelling, 140-45, 156, 159
learning-by-producing, 13, 17 models:
learning curve, 20 discrete choice, 144
learning processes: modal choice, 287-88
of each location, 22 residential land market, 141
simulation of knowledge and, 291 retail trade location, 142
within firms, 20 route choice, 288-90
linear dynamics: M-region. See Malardal region
nonlinear and, 77-81 mUltiple equilibria, 31, 72
theory of, 78-79 prevalence of, 112, 124
localization economies, 171-72 Malardal region (M-region), 185-91
location:
attributes of, 190-91 national innovation systems (NIS), 11-12,33-
intensities, 6 39
patterns of, 109-24, 141, 144-45 network economy, Fig. 3.8
lock-in, 30, 250 new information technologies (NIT):
to an existing spatial structure, 53, 119, barriers to, 249
123 co-evolutionary role of, 250
by historical events, 21 contributions to rural development, 267
Lowry model, 142-43 costs of providing, 249
spatial impacts of, 237,244-51
macrodynamics: time and space-shrinking capacity of, 1
of industrial growth, 19 New York versus Philadelphia, growth of,
and micromotives, 30 53-54
macrosimulation, 291 NIS. See national innovation systems
Malthusian model of exponential growth, NIT. See new information technologies
82 non-convexity, 146, 150
Marshallian externalities, 111
304
nonlinear dynamics: transportation, 140-42

butterfly effect in, 88-91 Poincare maps, 81
and financial derivatives, 97-101 positive feedback, 5, 26, 37-38
and innovation, 83-92 in financial markets, 99
linear versus, 77-81 between research agglomerations, 110,
mathematics of, 75 200
and technological innovation, 4 power law, 51-52
theories of, 291 production theory, 29
partial nature of, 81, 93 effects of technical change on, 17-22
urban development in terms of, 45-46, product cycle analysis, 165, 175,225
76-83 product cycles, 180, 186
Norway, leader and follower regions in, product innovations, frequency of, 165
185-88 product vintages, 6
NP-completeness, 144-45, 147 and location dynamics, 174-76
maturity of, 183
Office of Technology Assessment (OTA), spatial concentration of declining, 188-89
257 public knowledge, definition of, 15
optimising methods, 146-53 public participation:
limitations of, 148, 152 in planning, 153-57
options, 76, 97-100 paralysis by economic rationality, 139
definition of, 98 public problems, general systems nature of,
urban plans as, 102-3 139
punctuated equilibria, 2
parameters, effects of changes in, 77 PV. See product vintages
path dependence, 21, 75, 91 PV -model, 166-67
and innovation, 3-4 and CPS-model, 193
of agents' locational choices, 110-11, PV -products, 193
124 PV -ranking, 186
pattern recognition, 67
Pavitt's taxonomy of industry-specific quadratic assignment problem, 147
models quadratic map, 87
phase transitions, 55, 58
due to bifurcations, 86-90 random graph, definition of a, 55
in urban systems, 61, 75, 86, 104 rank-size distribution, 47
Pirenne-Mees hypothesis, 60-63, 71 among Asian nations, 48-49
planning: as an attractor, 51
complexity of, 5, 138 as an emergent property of urban
holistic view of, 155 dynamics, 51
major components of, 138 stability of, 49, Fig. 3.2, 92
public participation in, 153-57 Swedish,191
role of invention and innovation in, 149, in the United States, Fig. 3.1
157-60
305
rational expectations model, weaknesses of, Schumpeterian models of technical change,

96 13, 27, 29-30
regional growth: sectoral innovation system (SIS), definition
influence of port activity on, 213 of, 38
self-organisation, 72
modelling the role of R&D in, 198-207 economic, 30
regional innovation systems (RIS), 199-201 among mental models, 69
local factors shaping, 204 in systems of cities, 51, 180
regions: self-organised criticality, 52
corridor, 172 self-reinforcing processes. See positive
Europe's functional urban, 6, 183 feedback
knowledge-creating and knowledge- sensitivity to initial conditions, 80, 90
applying, 199 simulation experiments, to assess benefits of
leader and follower, 184-86, 191, 193- real-time information, 292
94 sociological ~ode\s, 19
research and development (R&D), 34 spatial economics, 110
characteristics of, 13, 128,231-32 spatial hierarchy, 165
differences between pure and applied, specialisation, 174, 199
127 factors contributing to, 172
importance of, 2, 5 index of industrial, 217
intensity of, 172-73, 177, 203, 232 product group, 180
location of laboratories, 112-18, Table in research, 203
5.1, Fig. 5.2 specialisation dynamics:
modeling, 119-22 hierarchical, 166
and regional economic growth, 6, 198- theories of, 6, 171-73
207,210,214 speciation models, 84
role of corporate, 12 speculative behaviour, 94-97
scale economies of, 231-32 spillovers. See knowledge spillovers
as a source of learning, 13, 31 stochastic processes, models of, 292
and technical progress, 27 stock markets, unpredictability of, 80
Ricardian model, 224 strategic assets, 15
rights, 101. See also options substitution:
RIS. See regional innovation systems between transport and communication,
risk-takers, 64-65, 95 250, 272, 276-81
Rotterdam, dominance as Europe's leading models of, 281
port, 211-13 influence of relative prices on, 29
route choice, modelling the effects of, 288- superposition, principle of linear, 78
90 Sweden, leader and follower regions in, 185-
routines, automatic and tacit, 23 92
systems approach, socio-dynamic, \-3
Schmookler's demand-pull model, 12-13,
28
306
tacit knowledge, 14, 18, 198 knowledge spillovers; technology transfer

abstract and, 199 and foreign direct investment (FOI), 223-
codified and, 6 24
definition of, 15 horizontal and vertical, 223
importance of, 16 technology transfer, 33
spatial spillovers of, 202-3 and foreign investment, 6, 224
tariff jumping, 224 telecommunications, 112
technical advance. See technical progress digitalisation of, 256
technical change: impacts on urban development, 7
and capital movement, 223 as a substitute for transport, 7, 250
micro-foundations of, 12-17 as a tool for managing complexity, 258
models of: telecommuting, 7, 272, 276-78. See also
industry-specific, 27-29 telecommunications; teleworking
Schumpeterian, 113,27,29-30 impacts on transportation, Table 13.2
relative prices and, Fig. 2.3 teleconferencing, 7, 272, 276,278. See also
theories of, 14,39 videoconferencing
technical progress, 15, 17. See also disadvantages of, 279
technical change Telecottage Association, 264
natural trajectories of, 16 telecottages:
stimulation of rural communities by, definition of, 264
255 rural location of, 264
types of, 28 stimulation of rural, 265-66
technological capabilities, vectors of, 34 telematics. See also telecommunications
technological competence, 201, 204, 214 barriers to the use of, 259
technological imbalances, 17 definition of, 256
technological innovations: policies, 267
characteristics of, Table 11.1 rural areas:
and nonlinear dynamics, 4 levels of usage in, 259
pervasive nature of, 1 role in, 7, 256
technological paradigms, 12-18 uneven spatial demand for, 257
as technical artefacts, 16, 241 teleservices, 276, 280
definition of, 14 teleshopping, 7, 272, 276, 279-80. See a/so
technological progress. See technical e-commerce
progress teleworking, 264-65
technological trajectories, 12,37 future projections in the U.S., Table 13.1
definition of, 16 Thlinen's model of agricultural location, 141
technology: Tokyo metropolitan area, 114-17,119,122
core, 31 trade, two-way relationship between
new developments in transport, 205 innovation and, 223
vintages of, 73 trade balance:
technology spillovers, 222. See also and capital mobility, 228
and debt, Fig. 10.2
307
traffic: nested (slow and fast) dynamic models

information systems, 272 of, 71, 86
management of air, 272 role oflearning and innovation in, 3, 11,
transaction costs, of derivatives, 98 238,245
transportation costs, 272 urban hierarchies, 92, 168-70. See also
central place systems (CPS)
transportation, impacts on urban specialisation dynamics in, 165-94
development, 7 urban planning. See planning
transportation models, 287-92 urban populations, controlling growth of,
transportation planning and metropolitan 137-38
analysis, 140-42 user information systems, 284. See also
Tsukuba, 5: traffic information systems
evolutionary development path of, 109- models for evaluating the effects of, 285-
10,117-24 287
R&D concentration in, 115, 121-24
vector of technological capabilities, 34
universities: videoconferencing, perceived disadvantages
direct and indirect effects of, 214 of, Table 13.3
importance of education at, 188-89, vintages, product, 6, 174-76
192, Table 8.8 vintage theory, 178
as knowledge producers, 204 volatility, stock market prices, 97
urban analysis, revolution in, 139-45
urban complexity, 58-60, 70 Zipfs law, 46-49. See also rank-size
urban development. See urban evolution distributions
urban evolution, 4: as a power law, 51-52
economic and technological impacts, 3
309
List of Contributors
David F. Batten Dimitrios Dendrinos

The Temaplan Group, Urban and Transportation Laboratory,
PO Box 3026, University of Kansas, 302 Marvin Hall,
Dendy Brighton, 3186 Australia Lawrence, Kansas 66045, USA
Cristoforo S. Bertuglia Ulla M. Forslund

DINSE, Politecnico di Torino, Department of Infrastructure and Planning,
viale Mattioli 39, The Royal Institute of Technology,
10125 Torino, Italy 10044 Stockholm, Sweden
Martin Beckmann Kei Fukujama

Lehrstuhl flir Angewandte Mathematik und Department of Social Systems Engineering,
Statistik, Technische Universitat MUnchen, Tottori University,
Archisstrasse 21, Tottori 680, Japan
80333 MUnchen, Germany
Andrew Gillespie
Ennio Cascetta Centre for Urban and Regional Development
Dipartimento di Ingegneria dei Trasporti, Studies,
Facolta di Ingegneria, University of Newcastle upon Tyne,
Universita di Napoli 'Federico II', NEI 7RU, United Kingdom
Napoli, Italy
Marina Della Giusta
Mario Cimoli Economics Department,
Division of Production and Management, University of Reading, PO Box 218,
ECLAC - United Nations, Casilla 179D, Whiteknights,
Santiago, Chile Reading RG6 6AA, United Kingdom
310
Britton Harris Dino Martellato

Department of City and Regional Planning, Dipartimento di Scienze Economiche,
127 Meyerson Hall, Fondamenta San Giobbe 873,
University of Pennsylvania, 30121 Cannaregio, Venezia, Italy
Philadelphia, PA 19104-6311, USA
Bruno Montella
Blirje Johansson Dipartimento di Ingegneria dei Trasporti,
Jlinkliping International Business School, Facolta di Ingegneria,
Box 1026, 55111 Jlinkliping, Sweden Universita di Napoli 'Federico II',
Napoli, Italy
Kiyoshi Kobayashi
Graduate School of Civil Engineering, Sylvie Occelli
Kyoto University, Kyoto 606, Japan IRES-Piemonte,
via Nizza 12,
Sotaro Kunihisa 10125 Torino, Italy
The Institute of Behavioral Science,
Honmura-cho 2-9, Ichigaya, Ranald Richardson
Shinzyukuku, Tokyo 162, Japan Centre for Urban and Regional Development
Studies,
Stefano Magrini University of Newcastle upon Tyne,
Dipartimento di Scienze Economiche, NEI 7RU, United Kingdom
Fondamenta San Giobbe 873,
30121 Cannaregio, Venezia, Italy

David F. Batten, Cristoforo S. Bertuglia, Dino Martellato, Sylvie Occelli Auth., David F. Batten, Cristoforo S. Bertuglia, Dino Martellato, Sylvie Occelli Eds. Learning, Innovation and Urban Evolution

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

David F. Batten, Cristoforo S. Bertuglia, Dino Martellato, Sylvie Occelli Auth., David F. Batten, Cristoforo S. Bertuglia, Dino Martellato, Sylvie Occelli Eds. Learning, Innovation and Urban Evolution

Uploaded by

Copyright:

Available Formats

LEARNING, INNOVATION

AND URBAN EVOLUTION

Springer Science+Business Media, LLC

Copyright 2000 by Springer Science+Business Media New York

AII rights reserved. No part of this publication may be reproduced, stored in a

Printed on acid-free paper.

I. Learning, Innovation and Urban Evolution: An Introduction

PART I: LEARNING AND KNOWLEDGE EXCHANGE

2. Innovation and Patterns of Learning: A Survey of Evolutionary

3. Emergence and Co-Evolutionary Learning in Self-Organised

4. Nonlinear Dynamics, Innovation and Metropolitan Development 75

Section B: Path-Dependent Processes of Knowledge Exchange

5. The Knowledge-Intensive Nature of Japan's Urban Development 109

6. Interurban Knowledge Networks 127

7. Innovation and Urban Planning 137

PART II: INNOVATION AND ITS SPATIAL IMPACTS

8. Product Vintages and Specialisation Dynamics in a Hierarchical

9. The Impact of Research Activities on the European Urban System 197

10. Is the Real Exchange Rate Relevant for Determining the

Section D: Impacts of Innovations in Telecommunications

11. Impact of the New Information Technologies on Economic-

12. Telematics Innovation and the Development of Non-Metropolitan

13. The Impact of Innovations in Telecommunications and

List of Contributors 309

In the global information society, innovation does not only constitute a

Theories of Learning and Complex Dynamics

Learning, Innovation and Urban Evolution:

David F. Batten 1, Cristoforo S. Bertuglia2, Dino Martellat03 and Sylvie

2Turin Polytechnic, Turin, Italy

An important component of the current debate on the impacts of the new

locations seem to draw closer. Geographical space is distorted as the ease of

See Koestler (1964, p.226).

See, for example, Andersson et at (1993), Bertuglia et at (1995), Batten et at (1995),

cohesive organisational structure for the book. It also recognIses that

1.2 AN OUTLINE OF THE VOLUME

1.2.1 Theories of Learning and Complex Dynamics

spatial or sectoral dualism, due to the inherent instability found in such

1.2.2 Path-Dependent Processes of Knowledge Exchange

In their chapter, Kiyoshi Kobayashi, Sotari Kunihisa and Kei Fukuyama

1.2.3 Economic and Technological Changes

In their chapter on product vintages and specialisation dynamics, Ulla

1.2.4 Impacts of Innovations in Telecommunications and

amount of work already carried out4, findings continue to be inconclusive.

Innovation and Patterns of Learning: A Survey of

Mario Cimoli\ Marina della Giusta2

Some of the main international organisations concerned with

economic performance can be identified and described. The second is an

2.2 THE MICRO-FOUNDATIONS OF TECHNICAL

The demand-pull model spawned a number of empirical studies. Mowery

Indeed, it is possible to identify patterns of change which are defined in

evolutionary definitions, on which our rudimentary understanding of the

Such technological imbalances are described by Rosenberg (1976) as

2.3 IMPLICATIONS FOR PRODUCTION THEORY

The elements of technical change discussed so far, especially the

knowledge and the ensuing trajectories. An important implication is that, at

Observation (a) follows from processes of diffusion and competItIOn

This graph depicts the distribution of coefficients (Cj) observed at time t,

heterogeneity and technological gaps. Also, it may be associated with the

Adoption Adoption Adoption

The evolutionary nature of technical change may be responsible for what

more firms to that location purely by boasting a significant number of firms

1. it is possible to rank different regions or countries by the efficiency of

Differences in technological capability not only account for such

2.4 FROM INDIVIDUALS TO INSTITUTIONS

I ~ L - - L- L ~~-L--~--~--~~