Modern Urban and Regional Economics
Modern Urban and Regional Economics
Modern Urban and Regional Economics
6)~~/r(I'S
3'>tl. oS .r1LC- .~~
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
SSH1Id X-LIS1IHAINil
crnOjXO
uut?))lAJ d!l!4d
S:>!WOUO:>3
leUO!8all pue
ueqJn uJapoll\l
OXFORD
UNIVERSITY PRESS
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
()xford University Press is a department of the University of Oxford.
It furthers the Universitis objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
()xford University Press in the UK and in certain other countries
© Philip McCann 2013
The moral rights of the author have been asserted
First Ed ition copyright 2001
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
add ress above
You must not circulate this work in any other form
and you must in1pose this same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
ISBN 978-0-19-958200-6
Printed in Great Britain by
Ashford Colour Press Ltd, Gosport, Hampshire
To my parents Kath andjoe, and to my family
Philip McCann holds the University of C;roningen Endowed Chair of Economic C;eography in the
Faculty of Spatial Sciences at the University ofC3roningen, 'lhe Netherlands. He is Adjunct Professor
of Economics at the University of Waikato in New Zealand, and a fonner Professor of Urban and
Regional EconoInics at the University of Reading, UK.
During 2010-2013 Philip McCann was Special Adviser to the European Commissioner for
Regional Policy, Johannes Hahn, advising on all aspects of the reform of EU Cohesion Policy. He
also acts as an adviser to C)EC]) and to various governInent departments in several countries, as well
as being a meInber of the steering committees of national and international research
commissions.
Educated at Call1bridge University, Philip McCann has published nUIllerous journal articles,
books, and book chapters over two decades, and won various international awards for his research. He
has previously held long-term visiting professorships at University of Pennsylvania, USA, University
of Tsukuba, Japan, RitsuIneikan University, Japan, and 'lhammasat University, Thailand, as well as
short-term visiting positions at Bocconi University, Italy, Cornell University, USA, and University of
Barcelona, Spain.
Philip McCann is also co-editor of Papers in Regional Science, Spatial Economic Analysis, and
Review of Urban and Regional Development Studies.
Acknowledgements
My thanks go to the editorial team at ()xford University Press who have always been supportive and
professional in everything they do, making the arduous process of writing a book such as this as
straightforward as possible. I am also very grateful for the significant aInount of work undertaken by
five anonymous reviewers whose detailed insights, comments, and criticisms on each chapter greatly
improved the book.
My thanks also go to the numerous scholars and teachers from all over the world who have kindly
provided me with excellent ideas and suggestions for this second book, almost all of which I hope I
have managed to include.
Finally, I am forever grateful for the many friends and family who have supported me in my
work.
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Contents
List of xii
List of tables xiv
Introduction xvii
3.7 Conclusions 96
Appendix 3.1 Spatial monopoly and price discrimination 96
Appendix 3.2 The derivation of Reilly's law 99
Appendix 3,3 The NEG model of the urban-regional economy 100
Appendix 3.4 The Lbschian demand function 104
Appendix 4.1.1 Distance to the edge of the von Thunen area of cultivation 141
Appendix 4.1.2 Distance to a change of land use in the von Thunen model 141
Bibliography 376
Index 398
List of figures
List of tables
lhe front cover of this book provides a wonderful example of why urban and regional economics is
such an exciting and important subject to study. rlhe building featured is the Bank of C:hina Tower
building in Hong Kong, and for me it is one of the most elegant modern buildings in the world. rlhe
building is the home of various international corporations providing knowledge-based financial ser-
vices with enormous investments in China. Yet, as a book-cover image, the building succinctly cap-
tures the urban and regional transformations evident in the modern global context. Hong Kong is one
of the most dynamic cities in the world. Located in the heart of the rapidly expanding global super-
region of South and East Asia, less than five decades ago Hong Kong was a centre for low-value-added
manufacturing. Hong Kong is now one of the major global centres of finance and trade and in com-
mercial terms it is a meeting point and melting pot of east and west, north and south. rlhe wonderful
building therefore is also a sYInbol of the transformations we have witnessed in the global economy
during the last twenty years.
When the first edition of Urban and Regional Economics was published in 2001, we were still little
more than a decade on froIn the fall of the Berlin Wall, the creation of the European Union and its
Single Market, seven years from the founding of the North American Free Trade Agreement (NAFTA),
and only a decade from creation of the world wide web in its modern sense. rlbe global econoInic
changes wrought by these transformations were still only emerging and as such were still relatively
little understood. Today the situation is radically different. Since the Millennium, enormous strides
have been made in terms of urban and regional data provision and urban and regional theoretical
analysis. These empirical and theoretical developlnents have significantly increased our awareness
and understanding of the role played by cities and regions in the global economy, and, in turn, the
impacts of the global economic changes on cities and regions. The debates surrounding cities and
regions have nowadays also broadened to include issues such as creativity, institutions, well-being,
social capital, and sustainability, and cities and regions are seen to be central to all such discussion.
Whereas once urban and regional issues were seen by many economists to be a minor avenue of
research, nowadays many of the most important international institutions including the World Bank,
the C)ECI), the European Commission, and the United Nations are all grappling with the economic
challenges and possibilities associated with regions and cities.
The novelty and originality of the 200 I edition of this textbook was to treat urban and regional
economic issues in an integrated manner, drawing out the links and differences between the various
model approaches. The intention was to make the book accessible to economists, geographers,
regional planners, business and management students, and also public policy-makers. The book was
structured around a series of models presented in a manner which would allow the issues to be taught
and applied irrespective of which part of the world the reader inhabited. The fact that the sales of the
book have been more or less equally distributed across the globe, as well as being translated into
Chinese, Greek, Korean, and Japanese, suggests that it was successful in terms of these objectives.
This new edition of the textbook adopts exactly the same approach as the earlier edition. However,
it has been re-titled as Modern Urban and Regional Economics, and this is in order to Signal the novel
features of this new edition, which is quite different to the first book in several key respects. The new
edition aims to achieve two objectives. First, it seeks to update all the traditional analytical discussions
in the light of the new thinking, the new models, and the new empirics that have emerged over the last
decade. Second, it seeks to position these theoretical and empirical developments in the much broader
and transforming global context in which today's cities and regions find themselves. The book COITI-
prises ten chapters. Each chapter takes a particular theme and discusses the various ways in which we
are able to ask and answer questions related to the topic in question. 'lbe discussions operate at differ-
ent levels, with the first seven chapters discussing the ways in which we analyse the spatial behaviour
and urban and regional impacts of firms, people, land markets, inter-firm linkages, and technology.
The last three chapters take a much broader historical, geographical, and institutional approach to
examine the broader underlying changes that affect cities, regions, and the urban and regional policy
challenges faced in the modern economy. Given the different levels at which these groups of chapters
operate, the book is split into two parts, entitled Urban and regional economic models and methods,
and Globalization: cities, regions, and economic policy. Part I deals with various aspects of urban and
regional economic analysis, while Part II discusses the broader themes relating to the changing con-
text in which regions and cities nowadays find themselves.
Chapter 1 discusses the various theoretical ways we can understand the location behaviour of indi-
vidual firms. Chapter 2 employs these arguments to explain why groups of firms and activities are
often located together in cities, urban agglomerations, and industrial clusters. However, we also know
that firms differ greatly in their location patterns and are often found to be distributed across widely
different spatial scales. Chapter 3 therefore examines these broader patterns of location and disper-
sion, allowing for the fact that different firms face different pressures and advantages from being
clustered or located apart from each other, and uncovers some regularities and systematic features
underpinning the relationships between spatial concentration and dispersion. Where clustering leads
to the growth of a city, once an urban area has arisen at a particular location, Chapter 4 explains how
the urban land market works and how local land allocations are determined. Chapters 5-7 then adopt
a more macro-approach to discussions of various regional economic issues that are generally under-
stood primarily in macroeconomic terms. Chapter 5 discusses multiplier analysis and the ways in
which the linkages between firms and activities in a local area affect the overall output of an area.
Chapter 6 explains the response of spatial labour markets to local demand and supply changes, and
discusses the particular problems associated with local labour market adjustments and interregional
migration flows. Chapter 7 then integrates the arguments in each of the previous chapters in order to
discuss the various approaches we have to analyse regional growth behaviour. As with Chapters 2-4,
Chapters 5-7, which deal with these more macroeconomic issues, are underpinned by explicitly spa-
tial considerations. lhis allows us to identify the differences between analysis of economic phenom-
ena at the urban or regional level and analysis at the national macroeconomic level.
Part II takes a much broader panoramic approach to understanding the impacts of globalization
on cities, regions, and the policy challenges and possibilities associated with these changes. Chapter
8 discusses the historical relationship between urbanization, industrialization, and wealth in the
context of four centuries of globalization processes. The intention here is to explain why studying
cities and regions is so important for understanding national growth, and how these relationships
have changed during different time periods, right up to the present day. What is important here is
to understand that the long-term historical trends linking the growth of cities and densely popu-
lated regions to the growth of the national economy heavily influence the twentieth-century logic
and construction ofthe models discussed in Chapters 1-7. Yet, in terms ofthe relationships between
· cities, regions, and national economies, the twentieth century itself was a period of dramatic twists
and turns, in which many of these relationships were either broken or refashioned in different ways
in different parts of the world. Some of these twists and turns now have profound impacts on the
ways in which we are able to use the ITIodels exan1ined in Chapters 1-7 in the modern economy,
because the underlying context in which such models can be applied may differ significantly in dif-
ferent places. In order to investigate these issues, Chapter 9 examines the major urban and regional
changes that have arisen during the ITIodern era of globalization which began in the late 1980s, and
considers these changes in the light of the long-run trends discussed in Chapter 8. As we will see,
the enormous global changes experienced since the late 1980s have fundamentally altered the long-
run historical relationships between cities, regions, and nations in ways that we are only now begin-
ning to understand better. An awareness of these changes is essential in order to understand how
best to use the models discussed in Chapters 1-7 for analysing conteITIpOrary phenomena. At the
same time, these changes have also led to new debates and controversies regarding regional and
urban policy, debates that are sufficiently profound and influential that they are now taking place at
the highest levels of international econoITIic policy-making. These debates are examined in Chapter
10, which discusses in detail the nature of, and the justification for, urban and regional economic
policy in the context of ITIodern globalization. Chapter 10 explains how we can use the urban and
regional economic ITIodels and ITIethods described in Part I of this book in order to understand,
predict, and target the iITIpacts of various types of urban and regional policies. These discussions
also highlight the limitations of our current understanding, and point out possible directions for
further research.
As we will see in Part II of the book, the impacts of ITIodern globalization on cities and regions have
been profound, while in turn cities and regions have been in the vanguard of modern globalization
processes. The emerging awareness of these relationships over the last decade has shaped entirely new
discourses within politics and policy-making the world over. Many of these discussions are valuable,
but not all are accurate. A genuine understanding of these processes therefore requires a strong
grounding in the theoretical models while at the same time an awareness of the fundamentally chang-
ing context in which these models need to be positioned and interpreted. It is hoped that a good
awareness of both the models and methods, along with the current global context, will help us best
achieve the desired outcomes of public policy initiatives that contain urban and regional elements.
This book aims to provide both this theoretical grounding and also a real awareness of the evolving
global context. No other textbook dealing with urban or regional issues sets these squarely in today's
global context, and no text on globalization deals explicitly with the analytics ofcities and regions. This
new version of the book is therefore genuinely new.
All economic phenomena take place within geographical space. Economic issues invariably involve
either questions concerning the place specificity of particular activities, or alternatively, questions
relating to the overcoming of space and geographical distance. For example, all commodities are
traded at various market locations. However, in order to reach the appropriate market locations,
goods have to be transported and delivered across space. Similarly, service activities take place at par-
ticular locations, and the knowledge and information required to carry out the activity must be trans-
mitted or acquired across geographical space. In each case, the costs incurred in these spatial
transactions will themselves partly determine the price and cost conditions at each market location.
Yet the,reasons why particular markets are located at particular places are also economic questions,
and as we will see in this book, the nature and behaviour of markets depend somewhat on their loca-
tion. Market perfonnance therefore partly depends on geography. At the same time, the economic
performance of a particular area also depends on the nature and performance of the various markets
located within the area. Acknowledging that geography plays a role in determining economic behav-
iour, many discussions about the performance of particular local, urban, or regional economies are, in
fact, fundamentally questions about the relationships between geography and the economy.
Geography and economics are usually interrelated issues.
For Inany years before the 1990s, spatial questions were all too often ignored by economists and
economic policy-makers. 'Ihis is partly a problem of education. In most textbook discussions, the
whole economic system is assumed to take place on a pinhead (Isard 1965). While for a long time there
have been many urban economists, regional scientists, and economic geographers who have been
explicitly concerned with spatial economic phenomena, for many years the majority of geographical
issues were subsumed by Ricardian theories of comparative advantage and international trade. In the
post-war Bretton Woods world of relatively closed economies and currency convertability restric-
tions, such assumptions may have appeared to many economists to be acceptable. However, in the
modern era of free trade areas such as the North American Free Trade Agreement (NAFTA), new
information and communications technologies, currency convertibility, the EU Single Market and
the euro, rapidly increasing capital and labour mobility, and the enormous city-region growth of the
newly emerging BRIICS countries of Brazil, Russia, India, Indonesia, China and South Africa, many
of these traditional assumptions can no longer be justified. These recent developments have high-
lighted the fact that competition between regions is often more important and more complex then
competition between individual countries. Indeed, much international competition is actually domi-
nated by competition between particular regions in different countries, rather than between whole
countries. In each of these cases, the nature of the sub-national and super-national competitive rela-
tionships between various regions depends on the spatial distribution of industrial activities.
Geography is an essential element of the economic system, and the economics of urban and regional
behaviour are just as important as that of national behaviour. The role of geography in the economy
and the importance of the regional economic behaviour provide the motivation and justification for
studying urban and regional economics.
Since the early 1990s there has been an enormous increase in interest is spatial economic questions
within the world of academic research and public policy-making. Testament to this comes from
highly influential work on urban and regional issues undertaken during recent years by the World
Bank, the GECD, the European Commission, and the United Nations, as well as the major changes in
policy-thinking in many countries regarding the critical role in growth played by cities and regions. In
part, this has been because ofthe new institutional and technological developments mentioned above,
which have highlighted the need for explicit considerations of space in economic discussions. The
writings of Paul Krugman and Michael Porter, among others, have also brought the importance of
spatial economic issues to the attention of wider audiences within the international economics, busi-
ness and management fields. The work of these authors has led to significant developments in our
understanding of the relationships between space and the economy. However, there is a long and
broad tradition of spatial economic analysis, the origins of which pre-date both of these authors.
Building on the original seminal works of authors discussed in this book, a huge number of authors
have subsequently provided many fundamental insights into the complex nature of the relationships
between geography and space. A consideration of these insights and the analytical techniques thereby
developed is essential in order to provide a comprehensive understanding of the nature and workings
of the modern spatial economy.
Traditionally spatial economic analysis has broadly been split into two sub-fields, namely urban
economics and regional economics. Ihese are by no means mutually exclusive categories and many
analyses will fall into both categories. 'The distinction between these two categories has arisen as a
result of asking slightly different questions. Urban economics, by definition, is generally concerned
with asking questions about the nature and workings of the econoIny of the city. As such, the models
and techniques developed within this field are primarily designed to analyse phenomena that are
confined within the limits of a single city. Regional economics, on the other hand, tends to ask ques-
tions related to larger spatial areas than single cities, and the models and analytical techniques devel-
oped generally reflect this broader spatial perspective. In essence, urban economics tends to emphasize
issues or relationships operating priInarily at a place, whereas regional economics tends to emphasize
issues or relationships operating primarily between places. Moreover, the the central questions of
regional economics therefore focus on the reasons why individual spatial parts of the same country or
of groups of adjacent countries behave differently to one another, whereas the central questions in
urban economics tend to ignore areas which are primarily rural or primarily a mixture of smaller
urban and rural localities. However, as we will see in this book, there are many issues which can be
analysed within either field, such as questions relating to the locations of cities, the location of finns,
or the migration behaviour of labour. In each urban or regional case, the choice of the appropriate
analytical approach to adopt or the techniques to employ will in part be determined by the particular
real-world context we are considering and the data that are available.
For the purposes of this book, an urban area is defined as a single continuous and contiguous
area of urban developInent and built environment. 'Ihe central questions of urban economics
therefore focus on the workings of the individual city. 'The definition of a region is rather more
complex, because areas can be defined as individual regions in terms of their topography, climate,
economy, culture, or adIninistrative structure. For the purposes of this book we define regions in
terms of spatial units. A region is defined here as a spatial area that is larger than a single urban
area, but that is different the spatial definition of a single nation. In general, we assume that regions
are smaller than individual countries, but in this book we also explore the many cases where
regions are larger than nations. For example, the spatial classifications of urban and regional areas
adopted here are by no means definitive. For example, some individual urban areas such as Los
Angeles and Tokyo can be regarded as major regions in their own right. At the same time, some
regional areas cut across national boundaries. For example, the economy of Detroit and some
parts of Western ()ntario are largely the same regional economy. Similarly, the economy of Seattle
can be considered to be broadly part of the same regional economy as Vancouver, British (:olumbia.
Meanwhile, in Europe, the southern part of Netherlands can be regarded as being largely part of
the same regional economy as parts of northern and eastern Belgium and the Nordrhein-Westfalen
area of Germany. Furthermore, regions can also vary enormously either in geographical or popu-
lation size. For example, the south-west region of the USA is the spatial size of the whole of
Western Europe, while the Tokyo-Kanto regional population is larger than the whole population
of Scandinavia.
Although many spatial economic topics can be analysed within either an urban or a regional
economics framework, this is not to say that the spatial unit of analysis, whether it is a single city
or a multi-city region, is an arbitrary choice. Some economic phenomena primarily affect localized
individual urban areas, whereas the impacts of certain other economic phenomena are generally
felt over much larger regional areas. The appropriate geographical area of analysis will therefore
depend on the nature and spatial extent of the economic phenomena. At the same time, regions
and cities are both valid areas for economic analysis because economic policy is often implemented
at these levels. Individual urban metropolitan governments have a role to play in determining
transportation and land-use policies within the confines of the individual city, and some of the
financing of such policies will be raised by local city taxation. The analysis of the impacts of such
schemes must be made at the level of the individual urban area. Similarly, inter-urban transporta-
tion and land-use policies will have impacts on all the cities within a region. As such, the regions
comprising the groups of cities become the appropriate areas of analysis, as the effects of such
schemes may be rather different between the individual cities. As we will see in this book, the
choice of the area of analysis will determine the models we employ and also how the results we
generate are to be interpreted.
For analytical simplicity, in Chapters 1-7 of this book, which develop the various urban and
regional economic models, we will therefore initially adopt the convention that regions are smaller
than individual countries and larger than individual urban areas, with the additional assumption that
a country is an area with a common currency and free internal capital and labour mobility. We then
relax this assumption in Chapters 8-10 and allow for regions spanning national borders when we
discuss the role of cities and regions in the modern era of globalization.
1.1 Introduction
In this chapter we introduce the basic building blocks of firm location theory. The models
are introduced and explained primarily by means of diagrams, with mathematic treatment
reserved largely for the appendices. We begin by a providing detailed analysis of the Weber
(1909) model, in many ways the fundamental building blocks of microeconomic firm loca-
tional analysis, which is constructed in the context of fixed production factor relationships.
The chapter then extends the Weber approach by examining the insights derived from the
Moses (1958) model, which deals with issues of input factor substitution. Additionalloca-
tion production models are also discussed, including the concept of extension of distance
costs from simply that of transport costs to total logistics costs, which incorporates the costs
associated with time and the frequency of interaction. Taken together, these location-
production models allow us to uncover some general analytical insights regarding the inter-
actions between location and production behaviour. The chapter then moves on to consider
how the geography and spatial delineation of a market influences firm location choices.
A detailed analysis of the Hotelling (1929) model demonstrates that location behaviour and
firm strategy are intertwined, and that the nature of this relationship also depends on the
market pricing structures possible. Price competition and spatial clustering are seen to be
largely incompatible under various situations, due to the risks associated with price wars.
Possible solutions to these problems emerge from both the Hotelling (1929) model and the
more general Salop (1979) model and centre on the role played by non-price competition
behaviour. In particular, product variety and product quality competition are found to offer
reasons why firm clustering is widely observed.
The level of output and activity of an area depends on the total quantities of factor inputs
employed in the area, and the wealth of an area depends on the total payments received by
those factors. Observation suggests that some regions exhibit dense concentrations offactors,
AND
with large numbers of people and investment located in the same area, whereas other
regions exhibit sparse populations and low levels of investment. At the same time, observa-
tion also suggests that people are paid different wages in different areas, while land prices
vary significantly between locations. Therefore, in order to understand the economic per-
formance of a region it is necessary to understand why particular quantities of factors are
employed in that area, and why the factors there earn the particular rewards that they do.
Production factor inputs are usually defined in terms of three broad types, namely capi-
tal, labour, and land, and the factor payments earned by these factors in the production
process are profits, wages, and rents, respectively. In some analyses of the production pro-
cess, additional factor inputs are also identified, such as entrepreneurship and technology.
However, in our initial discussion ofthe causes and reasons for particular types of industrial
location behaviour, we will not initially distinguish these additional factors from the broad
factor groups. We include entrepreneurship in our description of labour, and technology in
our description of capital. Later in our discussion of the causes and reasons for particular
types of industrial location behaviour, we will also investigate the additional issues associ-
ated with entrepreneurship and technology. In this chapter we will concentrate on the
determinants of spatial variations in capital investment, and in later sections of the book we
will focus on spatial variations in labour stocks, and variations in land prices.
We start our analysis by asking the question: what determines the level and type ofcapital
invested in a particular region? When talking about capital, our most basic unit of microe-
conomic analysis is the capital embodied in the firm. In order to understand the level of
capital investment in an area it is necessary to ask why particular firms are located there and
why the particular levels and types of investment in the area are as they are. These are the
questions addressed by industrial location theory. We begin by discussing three classical
and neoclassical models of industrial location behaviour, namely the Weber model, the
Moses model, and the Hotelling model. Each of these models provides us with different
insights into the fundamental reasons for, and the consequences of, industrial location
behaviour. After analysing each of these models in detail, we will discuss two alternative
approaches to analysing industrial location behaviour, namely the behavioural approach
and the evolutionary approach. A broad understanding of these various approaches to
industrial location behaviour will then allow us to discuss the concept of agglomeration
economies.
The model described by Figure 1.1 is often described as a Weber location -production
triangle, in which case the firm consumes two inputs in order to produce a single output.
Notation for use with Figures 1.1 to 1.12.
We assume that the firm consumes material inputs 1 and 2, which are then combined by the
firm in order to produce an output commodity 3. In the Weber location-production model,
we assume that the coefficients of production are fixed, in that there is a fixed relationship
between the quantities ofeach input required in order to produce a single unit ofthe output.
Our production function therefore takes the general form
In the very simplest case k l k 2 == 1, in which case our production function becomes
m) f(ml' m2 ) (1.2)
This represents a situation where the quantity of the output good 3 produced is equal to the
combined weight of the inputs 1 and 2. In other words, for the purposes ofour analysis here,
we can rewrite (1.2) as
The production locations of the input sources of 1 and 2, defined as M I and M 2, are given,
as is the location of the output market M 3, at which output good 3 is sold. The prices per
ton of the inputs 1 and 2 are given as PI and P2' at the points of production M I and M 2,
respectively. The price per ton of the output good 3 at the market location M 3 is given as P3.
As such, the firm is a price taker. Moreover, we assume that the firm is able to sell unlimited
quantities of output 3 at the given price P3' as in perfect competition. The transport rates
are given as t I' t 2 , and t3, and these transport rates represent the costs of transporting one
ton of each commodity 1,2, and 3, respectively, over one mile or one kilometre. Finally,
the distances d l , d 2 , and d 3 represent the distances over which each of the goods 1,2, and 3
are shipped.
We also assume that the input production factors oflabour and capital are freely available
everywhere at factor prices and qualities that do not change with location, and that land is
homogeneous. In other words, the price and quality of labour is assumed to be equal every-
where, as is the cost and quality ofcapital, and the quality and rental price of land. However,
there is no reason to suppose that the prices of labour, capital, and land are equal to each
other. We simply assume that all locations exhibit the same attributes in terms of their pro-
duction factor availability. Space is therefore assumed to be homogeneous.
If the firm is able to locate anywhere, then assuming the firm is rational, the firm will
locate at whichever location it can earn maximum profits. Given that the prices of all the
input and output goods are exogenously set, and the prices of production factors are invari-
ant with respect to space, the only issue which will alter the relative profitability of different
locations is the distance of any particular location from the input source and output market
points. The reason for this is that different locations will incur different costs of transporting
inputs from their production points to the location of the firm, and outputs from the loca-
tion of the firm to the market point.
If the price per unit of output P3 is fixed, the location that ensures maximum profits are
earned by the firm is the location at which the total input plus output transport costs are
minimized, ceteris paribus. This is known as the Weber optimum location. Finding the
Weber optimum location involves comparing the relative total input plus output transport
costs at each location. The Weber optimum location, will be the particular location at which
the sum (TC) of these costs is minimized. The cost condition that determines the Weber
optimum location can be described as
TC = Min L mlA
t=l
(1.4)
where the subscript i refers to the particular weights, transport rates, and distances over
which goods are shipped to and from each location point K. With actual values correspond-
ing to each of the spatial and non-spatial parameters, it is possible to calculate the total
production plus transportation costs incurred by the firm associated with being at any arbi-
trary location K. Given our assumptions that the firm will behave so as to maximize its
profits, the minimum cost location will be the actual chosen location of the firm.
In his original analysis Weber characterized the problem of the optimum location in
terms of a mechanical analogy. I-Ie described a two-dimensional triangular system of pul-
leys with weights called a Varignon Frame. In this system, the locations of the pulleys reflect
the locations of input source and output market points, and the weights attached to each
string passing over each of the pulleys correspond to the transport costs associated with
each shipment. The point at which the strings are all knotted together represents the loca-
tion of the firm. In some case, the knot will settle at a location inside the triangle, whereas in
other cases the knot will settle at one of the corners. This suggests that the optimum location
will sometimes be inside the Weber triangle, whereas in other cases the optimum location
will be at one of the corners. Nowadays, rather than using such mechanical devices, the
optimum location can be calculated using computers. However, although it is always pos-
sible to calculate the optimum location of the firm in each particular case, of interest to us
here is to understand how the location of the Weber optimum will itself be affected by the
levels of, and changes in, any of the parameters described above. In order to explain this, we
adopt a hypothetical example.
Let us imagine that Figure 1.1 represents a firm that produces automobiles from inputs of
steel and plastic. The output good 3 is defined as automobiles and these are sold at the mar-
ket point M 3. We can assume that input 1 is steel and input 2 is plastic, and these are pro-
duced at locations M I and M 2, respectively. If the firm produces a car weighing 2 tons from
1 ton ofsteel and 1 ton of plastic, and the fixed transport rate for steel t 1 is half that for plastic
t2 (given that plastic is much less dense than steel, and transport rates are normally charged
with respect to product bulk), the firm will locate relatively close to the source of the plastic
production. In other words, the firm will locate close to M 2• The reason is that the firm will
wish to reduce the higher total transport costs associated with shipping plastic inputs rela-
tive to steel inputs, ceteris paribus. The firm can do this by reducing the value of d 2 relative
to d i • On the other hand, if the firm has a different production function, such that it pro-
duces a car weighing 2 tons from 1.5 tons of steel and 0.5 tons of plastic, then even with the
same values for the fixed transport rates t 1 and t2 as in the previous case, the firm will now be
incurring higher total transport costs associated with steel shipments, ceteris paribus. The
reason for this is that although plastic is twice as expensive to ship per kilometre as steel, the
total quantity of steel being shipped is three times that of plastic. The result is that the firm
can reduce its total input transport costs by reducing the value of dl relative to d2• The opti-
mum location of the firm will now tend towards the location of production for the steel
input MI.
Within this Weber framework, we can compare the effects of different production func-
tion relationships on the location behaviour of the firm. For example, we can imagine that
the two types of production function relationships described above, one which is relatively
.plastic intensive, and one which is relatively steel intensive, actually refer to the different
production functions exhibited by two different competing automobile producers. Firm A
URBAN AND .AND
M3
"'::-.,
~ ""
/
/
""
/
/
"
/ "" "
/
/
/
""
/ d3A ""
/ ""
/
/
/
"
""
/
/ ""
/d 1B ""
// d~" "
/
/ """
-------------------------------- M2
M1
exhibits the plastic-intensive production function, and firm B exhibits the steel-intensive
production function. As we see in Figure 1.2, from the argument above we know that firm
A will locate relatively close to M 2, the source of plastic, while firm B will locate relatively
close to M I' the source of steel. This is because, if we were to consider the case where steel
and plastic inputs were shipped over identical distances, i.e. diA == d2A , for firm A the total
transport costs associated with plastic transportation would be greater than those associated
with steel transportation. It therefore has an incentive to reduce the higher costs associated
with plastic shipments by reducing d 2A and increasing d iB • Alternatively, for firm B, for
identical input shipment distances, i.e. d iB == d2B , the total transport costs associated with
steel transportation would be greater than those associated with plastic transportation. It
therefore has an incentive to reduce the higher transport costs associated with steel by
reducing diB and increasing d2B •
optimum location of the firm relative to the location of the market and the inputs. Once
again, we can illustrate this point by using our hypothetical example above of two auto-
mobile producer firms, A and B, each consuming inputs of steel and plastic. However, in
this case we can imagine a situation where the input production functions of both firms
are the same. In other words, the relative input combinations for each firm, given as
m}/m 1, are the same. Ifboth firms pay the same respective transport rates t 1 and t2 for each
input shipped, the relative locational pull of each input will be identical for each firm.
However, in this situation we also assume that the firms differ in terms of their technical
efficiency, in that firm A discards 70 per cent of the inputs during the production process,
whereas firm B discards only 40 per cent of the inputs during the production process.
Consequently, the total output weight m 3 of firm B is twice as great as that of firm A, for
any total weight of inputs consumed. This greater output weight will encourage firm B to
move closer towards the market point and further away from the inputs points than firm
A. As seen in Figure 1.3, firm B will therefore be more market oriented than firm A in its
location behaviour.
A more common situation in which similar firms exhibit different location behaviour
with respect to the market is where the density of the product changes through the produc-
tion process at different rates for each of the producers. For example, we can imagine our
two automobile firms A and B producing identical weights of output from identical total
weights of inputs. Here, the production functions of both firms are therefore the same.
However, we can also assume that firm A specializes in the production of small vehicles
suited to urban traffic, while firm B produces large four-wheel-drive vehicles suitable for
rough terrain. As we have already seen, transport rates also depend on the bulk of the prod-
uct, and products which have a high density will exhibit lower unit transport costs than
products with a low density. In this situation firm B produces goods which are very bulky,
whereas firm A produces goods which are relatively dense. Therefore the output of firm B
will be more expensive to transport than that of firm A, and this will encourage firm B to
move closer to the market than firm A. Once again, as seen in Figure 1.3, firm B will be more
mar1)et oriented than firm A.
1.2.4 The location effect of varying factor prices
Our analysis so far has proceeded on the assumption that labour and land prices are identi-
cal across all locations, although in reality we know that factor prices vary significantly over
space. The Weber approach also allows us to consider how factor price variations across
space will affect the location behaviour of the firm. In order to understand this, it is neces-
sary for us to identify the factor price conditions under which a firm will look for alternative
locations.
We assume that the firm is still consuming inputs from M 1 and M 2 and producing an
output for the market at M 3• Under these conditions, we know that the Weber optimum K*
is the minimum transport cost location of the firm, and that if all factor prices are equal
across space this will be the location of the firm. Our starting point is therefore to consider
the factor price variations, relative to the Weber optimum K* which will encourage a firm to
move elsewhere. In order to do this, it is first necessary for us to construct a contour map on
our Weber triangle, as described by Figure 1.3. These contours are known as isodapanes.
M3
" .....
..........
.....
•.....
x w v u ......... s T
M1 :•.••.•••••••••.•.•••.••.••.••.•.••..••.••.•.••.•••.• : ~: .:
••••
w*
-$10
-$20
...................................
-DO I I
-$40 ·······················r·······r···········l
-$50 1······················r·······"I""···········"I""······1
K* Q R s T Distance
~ -$20 ~
: :
···········t························· ····················1············
I -$30 I
....... L·.·.·.·.·.·.·.J.·.~ ·.~·.·.·.·.·.·.·r.~·.·.·.· ~ ~ ~..~:: .~·.~ ~·.·.·.·.·.L.·.·.·.·.~ ~·.J.·.·.·.~·.·.·.~·] ~
x w v u K* Q R S T Distance
investment. For example, if a multinational manufacturing firm is looking for a new pro-
duction site in order to develop its business in a new area, the likelihood of it going to any
particular location will depend on the firm's estimate of the profits it can earn at that loca-
tion. From the isodapane analysis of our Weber location-production model here, we know
that the locations of key input sources such as M 1 and M 2, and market points such as M 3,
will automatically mean that some locations are more profitable than others, with the
Weber optimum being the most profitable location, ceteris paribus. Therefore, in order to
make other locations attractive for investment, local factor prices have to fall relative to the
Weber optimum. The attractiveness of any particular location as a new investment loca-
tion for the firm will depend on the extent to which the local factor price falls can compen-
sate for the increased transport (opportunity) costs associated with any suboptimal
geographical location. If all local factor prices are interregional equilibrium prices, as
described by Figure 1.6, the firm will be indifferent between locations. Under these circum-
stances, the firm will be equally likely to build its new production facility at any location. In
other words, the probability of investment will be equal for all locations. Over large num-
bers of firms with similar input requirements and similar output markets to this particular
firm, the level of investment in any location should be the same as in all locations. On the
other hand, if wages are not in equilibrium over space, certain areas will automatically
appear more attractive as locations for investment, thereby increasing the probability of
investment there.
Geography confers different competitive advantages on different locations, which can
only be compensated for by variations in local factor prices. However, in the above example,
the equilibrium relationship between local factor prices and distance was only applicable to
the particular firm in question here. This is because the interregional factor price gradient
was calculated with respect to the Weber optimum of this particular firm. As we have seen,
different firms will exhibit different Weber optimum locations, and this implies that differ-
ent equilibrium interregional factor price gradients will exist for different types of firms
exhibiting different transport costs, different production functions, and, finally, different
inpl1,t and output locations.
1.2.5 The location effect of new input sources and new markets
Our analysis has so far discussed the locational effect of different transport costs, different
production functions, and the resulting conditions under which a firm will be willing to
move to alternative locations. We will now discuss the question of different input and out-
put locations and the conditions under which a firm will search for alternatives. In the
examples above, it was possible to use isodapane analysis to identify the factor price condi-
tions under which a firm will move from one location to another. However, this process of
movement itself may engender changes in the input sources employed and the output mar-
kets served.
In Figure 1.7 we can consider the situation where the firm relocates from K* to F in
response to the lower factor prices at F, which more than compensate for the additional
input and output transport costs involving in consuming steel and plastic inputs from M 1
and M 2, and serving a market at M 3• Location F has therefore been determined as the new
optimum location with respect to M 1, M 2, and M 3• However, in moving from K* to F, it may
be that alternative suppliers of identical inputs now become available. For example, the
o
$50
c
price gradient must also be continuously changing. However, observation tells us that firms
in reality do not move very frequently, and this raises the question of the extent to which the
Weber model is a useful analytical tool to describe industrial location behaviour.
The reason why firms are not continuously moving is that the relocation process itself
usually incurs very significant costs, such as the dismantling of equipment, the moving of
people, and the hiring of new staff. Part of the transactions costs associated with relocation
are also related to information and uncertainty, which are topics we will deal with later in
the chapter. However, within the above framework we can easily incorporate these reloca-
tion costs by including the annualized cost of these one-off relocation costs into our isodap-
ane model. The existence of these additional costs simply implies that firms will only move
when the factor cost advantages of alternative locations also compensate for these addi-
tional relocation costs as well as the increased transport costs. In other words, the equilib-
rium interregional wage gradient will be even steeper than under the situation where such
costs are negligible. The Weber model therefore still allows us to identify the optimum loca-
tion, and consequently the profit-maximizing behaviour of the firm in space, even in situa-
tions where relocation costs are significant. The observation that firms do not move
frequently does not limit the applicability of the Weber model to real-world phenomena.
The one major location issue which the Weber model does not address is that of the rela-
tionship between input substitution and location behaviour. In order to understand this
relationship, we now turn to a discussion of the Moses location-production model.
The Weber model assumes that the quantities of each input consumed, m i and m 2, are fixed
per unit of output m 3 produced. However, we know from standard microeconomic analysis
that substitution is a characteristic feature of firm behaviour, and that efficiency conditions
mean that firms will substitute in favour of relatively cheaper inputs, ceteris paribus.
Substitution behaviour was first incorporated coherently into the Weber analysis by Moses
(1958), and in order to see how substitution behaviour affects the location behaviour of the
firm, we discuss here the main features and conclusions of the Moses approach.
In Figure 1.8, we construct an arc If in our triangle M I , M 2 , M 3, which is at a constant
distance d3 from the market point M 3• If we constrain our firm to locate along this arc, the
distance from the location of the firm K to the market M 3 will no longer be a variable.
Therefore we can analyse the locational pull on the firm of changes only in the delivered
prices of the inputs produced at M I and M 2•
For example, if the firm is located at I, the delivered price of input 1, given as (PI + t l dI ),
will be a minimum, because the distance dI , from M I to I, will be a minimum. Similarly, the
delivered price of input 2, given as (P2 + t2d2), will be a maximum, because the distance d2,
from M 2to I, will be a maximum. The delivered price ratio, given as (PI + t I dI )/(P2 + t2d2),
will therefore be a minimum at location 1. On the other hand, ifthe firm now moves to f, the
delivered price of input 1 will be a maximum, because the distance dl' from M I to f, will be
a maximum. At the same time, the delivered price ofinput 2 will be a minimum, because the
distance d2 , from M 2 to I, will be a minimum. Therefore the delivered price ratio, (PI + tld l )/
(P2 + t2d2), will be a maximum at location f.
M1
...................................................
d1
...............................
M2
m1
m2
m2
inputs PI and P2' the slope of the budget constraints at each location along If must be differ-
ent. As we move along the arc If from I to f, the delivered price ratio increases, and for every
location along the arc If there is a unique delivered price ratio. This means that the usual
approach to analysing microeconomic efficiency is not applicable to the firm in space, and
must be adapted to incorporate the effects of location on the slope of the budget constraint.
In order to do this we must construct the envelope budget constraint, which just contains all
of the budget constraints associated with each of the locations along the arc If. This is done
by drawing each of the budget constraints for each of the location points on the arc If, as in
Figure 1.10, and the outer limits of this set of individual budget constraints will define the
envelope budget constraint.
The Moses argument is that we can now apply standard efficiency conditions to this
model by finding the point at which the envelope budget constraint is tangent to the highest
isoquant attainable. This is shown in Figure 1.11, where the point of maximum efficiency is
atE*.
At E*, the optimum input combinations are given as m I * and m 2*. However, E* also rep-
resents an optimum location K*. The reason is that the optimum input combination is found
at a particular point on the envelope budget constraint. Yet every point on the budget con-
straint also represents a unique location. Therefore the optimum input mix and the opti-
mum location of the firm are always jointly determined. One is never without the other.
This is a profound insight. Where input substitution is possible, all location problems
become production problems and all production problems become location problems.
We can illustrate the argument with an example. In our Weber-Moses triangle, we can
imagine that a road building programme takes place in the area around location Ml' the
effect ofwhich is generally to reduce the value of t I for all shipments ofgoods from this loca-
tion, relative to all other locations. If all the other parameters remain constant, this will
. imply that the delivered price ratio (PI + t Idl )/(P2 + t2d2), at all locations along If, will fall.
m1
\
Isoquants
Q2
m1
Q1
m2 m2
In other words, the slope ofeach budget constraint becomes steeper, ceteris paribus, and the
envelope budget constraint also becomes steeper and shifts upwards to the left. Strictly
speaking, in accordance with the income effect, the envelope will also shift outwards to the
right, because the price of the input has fallen. However, in this discussion we focus only on
the substitution effect of the change in slope of the envelope. For a given set of production
isoquants, the optimum production combination will change from that represented by E*.
As we see in Figure 1.12, at the new optimum E', the optimum input mix is now m I ' and
m 2'. The reason is that the firm substitutes in favour of input I, which is now relatively
cheaper than before, and away from input 2, which is now relatively more expensive than
before. In doing so, the firm increases the relative quantities of input 1 it consumes and
reduces the relative quantities of input 2 it consumes. However, this also implies that at the
original location K*, the firm now incurs increasing total transport costs (mitId I ) for input
1 relative to the total transport costs (m 2t2d2) for input 2. Therefore the firm will move
towards M I , the source of input I, in order to reduce these costs. The new optimum location
of the firm K' is closer to M I than E*, and so the firm moves towards MI.
The area around M I benefits in two different ways. First, the relative quantity of goods
produced by the area around M I which is purchased by the firm increases. This increases
regional output for the area. Second, the firm itself locates in the vicinity of M I , thereby
increasing the levels of industrial investment in the area.
Exactly the same result would have arisen in the case where, instead of a road building
programme, there had been a fall in the local wages at M I , which reduced the source price
PI' relative to all other locations. Once again, the fall in the delivered price ratio at aliloca-
tions leads to substitution in favour of the cheaper good and also relocation towards MI.
We can contrast this Moses result with that of the Weber model. In the simple Weber
model, if the transport rate t l falls, ceteris paribus, the effect on the location of the firm is to
move the locational optimum away from MI. The reason is that input 2 now becomes rela-
tivel); more expensive to transport, and because the coefficients of production are fixed,
mI
mI '
Isoquant
m1*
m2 ' m2 * m2
such that the relative quantities of m 1 and m 2 consumed remain the same, the firm will move
towards the source of input 2 in order to reduce the total transport costs. The difference
between the location-production results of the two models is that in the Weber model the
fixed coefficients mean that no input substitution is possible, whereas in the Moses model of
variable coefficients, input substitution is possible. In the latter case, the input substitution
behaviour alters the relative total transport costs and consequently the optimum location
behaviour of the firm. In reality, there is a continuum of possible location effects, dependent
on the technical substitution possibilities. In situations where the elasticities of substitution
are zero or very low, the results will tend to mimic those of the Weber model, whereas in
situations where the elasticities of substitution are high, the results will tend towards those
produced by the conclusions of the Moses model.
A second feature of the Moses model is that it allows us to examine the effect of returns to
scale on the location-production behaviour of the firm. In particular, we can ask: how will
the optimum location of the firm be affected by changes in the level of output ofthe firm? In
order to answer this, in Figure 1.13 we construct a series of envelope budget constraints,
represented by the dotted lines, which correspond to different levels of total expenditure on
inputs. Envelope budget constraints further to the right imply greater total expenditure
levels on inputs. An isoquant map, represented by the solid curves, can now be combined
with the envelope map. We can also apply the Moses argument, which states that the opti-
mum point for each level of output and input expenditure is where each particular envelope
is tangent to the highest isoquant, to the case of different output levels. By joining all the
m1
-....
.~ :?2::::'.'"' :' ' ' ' ' ' ' ' ' ' m2
points of tangency we construct a line ABC, which is an output expansion path. Yet this
output expansion path is different from the usual form of an expansion path. Each point on
the expansion path defines a particular optimum input combination. However, each point
on the expansion path also defines an optimum location.
If the expansion path is curved downward, such as in the case of ABC in Figure 1.13, it
implies that, as the output of the firm increases, and the total quantity of inputs consumed
increases, the optimum input mix changes relatively in favour of input 2. The optimum
ratio of mtlm 2 falls and the optimum location of the firm moves towards M 2• Alternatively,
if the expansion path were to curve upwards, this would mean that as the output of the firm
increased, the optimum input combination would change in favour of input 1. As the opti-
mum ratio of m I /m 2 increases, the optimum location of the firm would move towards the
market for mi.
This argument immediately leads to the conclusion that if the expansion path is a straight
line from the origin, such as FGH in Figure 1.14, both the optimum input mix and also the
optimum location of the firm will remain constant as output expands, ceteris paribus. The
actual slope of the expansion path is not important, except that it implies a different opti-
mum location. All that is required to ensure that, once the firm has found its optimum loca-
tion it will always remain at this optimum location as output changes, is that the production
function of the firm exhibits a straight line expansion path from the origin. This is the basic
Moses result.
This basic Moses result holds in the case where the firm is constrained to locate on the arc
If at a fixed distance from the market. However, in the more general case where the distance
from the market is also part of the location problem, the optimum location of the firm will
be independent of the level of output, as long as both the production function of the firm
m1
•.•...•....•.......................•.................. '"
m2
Figure 1.14 The independent of output optimum location solution
and the transportation technology of the firm exhibit constant returns to scale. The Weber
fixed-coefficients production function will satisfy the Moses requirement. However, there
are other more general types of production function allowing for input substitution, which
also satisfy this requirement. These results are detailed in Appendix 1.2.
The Moses result can be viewed somewhat as the spatial equivalent of the firm in perfect
competition. The firm is a price taker, and once it has determined its optimum production
technique and optimum location, the firm will not change its behaviour, ceteris paribus. In
other words, unless there are external changes in technology which alter the production
function relationships, or changes in transportation technology which alter relative trans-
port costs, or externally determined changes in the location of input goods sources and
output market points, the firm will always remain at the same location employing the same
input-output production techniques. It would be wrong, however, to view these spatial
results as implying that the spatial economy is essentially static. From our Weber model
discussion we see that the spatial economy exhibits evolutionary characteristics. These evo-
lutionary features arise because firms continually search for new optimum locations in
response to factor price changes, new input supplier and new market output locations.
Whether firms actually relocate in response to these search processes is an issue discussed
in Chapter 2. However, the key insights ofthe Weber and Moses models are that production
behaviour and location behaviour are completely intertwined issues. Often this point is
overlooked in textbook discussions of industrial economics and the theory of the firm. This
is largely because location adds an extra dimension to the optimization problems, making
the analysis somewhat more complex.
Additional complexity, but also a much richer set of locational insights relating to out-
put prices and the value added during the production process, can be incorporated into
these Weker-Moses types of location-production models when the costs of time and fre-
quency are included in the analysis, as in the logistics-costs location-production discussed
in Box 1.1.
In our analysis so far we have assumed that the market location is simply a point in space.
However, taking geography and space seriously in our models of firm behaviour also
requires that we investigate the explicitly spatial nature of market areas. Market areas fre-
quently differ over space, due to differences in spatial population densities, differences in
AND n~,"M!~~<J~~~~ ~::~~,"-Il~~Vn'&~"", H$~"'..§l<#~_~,J!'
income distributions across space, and differences in consumer demand across space
according to regional variations in consumer tastes. However, even if there were no spatial
variations in population densities, income distributions, and consumer demand patterns,
space would still be an important competitive issue. The reason is that geography and space
can confer monopoly power on firms, which encourages firms to engage in spatial competi-
tion in order to try to acquire monopoly power through location behaviour. In order to see
this we can adopt the approach first used by Palander (1935).
In Figure LIS, we have two firms A and B located at points A and B along a one-dimen-
sional market area defined by O-L. We assume that both firms are producing an identical
product. The production costs pa of firm A at location A can be represented by the vertical
distance a, and the production costs Pb of firm B at location B can be represented by the
vertical distance b. As we see, firm A is more efficient than firm B. The transport costs faced
by each firm as we move in any direction from the location ofthe firm are represented by the
slopes of the transport rate functions. As we see here, the transport rates for the two firms in
this case are identical, i.e. ta tb. For any location at a distance da away from firm A, the
delivered price of the good is given as (Pa + tada), and for any location at a distance db away
from firm B, the delivered price of the good is given as (Pb + tbdb).
Ifwe assume that consumers are evenly distributed along the line 0 L, and we also assume
that consumers, being rational, will buy from the firm which is able to supply at that particu-
1ar location at the lowest delivered price, the total market area will be divided into two sec-
tors, OX and XL. The reason for this is that between 0 and X the delivered price of firm A,
given as (Pa + tada), is always lower than that of firm B. On the other hand, at all locations
between X and L the delivered price of firm B, given as (Pb + tbdb), is always lower than that
Price/Cost
a b
A B
1°
~ ¥
Market of firm A Market of firm B
Figure 1.15 Spatial market areas: a one-dimensional model with equal transport rates
of firm A. Although firm A is more efficient than firm B, and although both firms produce
an identical product, firm A does not gain all of the market. The reason is that location gives
each firm some monopoly power over the area around itself. Firm A cannot capture all of
firm B's market, even though it is more efficient than firm B, because the transport costs
associated with shipping goods to market locations close to firm B increase the delivered
price (pa + tada) to an uncompetitive level in market locations close to firm B. In terms of
selling to consumers in the vicinity of firm B, firm A is unsuccessful simply because it is too
far away. On the other hand, for sales in this area, firm B is successful simply because it is in
the right location, even though it is less efficient in production.
This type of analysis can be extended to allow for differences in transport rates between
firms as well as differences in production costs. In Figure 1.16 (a) and (b), we see that market
areas can be divided up in a variety of ways in situations where the production costs and
transport rates vary between the firms. Generally, the size of a firm's market area will be
larger the lower are the production costs of the firm and the lower are the transport rates
faced by the firm. However, only in the case where transport rates are zero is a lower produc-
tion price sufficient to ensure a firm captures all of the market. The reason is that the exist-
ence of transport costs allows less efficient firms such as firm B to survive by providing each
firm with some monopoly power over particular market areas. In general, the areas over
which firms have some monopoly power are the areas in which the firms are located. For
example, Figure 1.16(b) can be regarded as representing a case such as a local bakery, where
firm B maintains a very small local market area in the face of competition from a national
bakery, firm A, which produces at a much lower unit production costs and transports in
large low-cost shipments.
Monopoly power refers to the ability of the firm to increase the production price of the
good Pa or Pb' and yet maintain some market share. In general, the greater the monopoly
power of the firm, the steeper the firm's downward-sloping demand curve. In many textbook
descriptions of monopoly or monopolistic power, the slope of the firm's downward-sloping
demand curve is viewed as being dependent on brand loyalty, associated with advertising
and marketing. However, location is also an important way in which many firms acquire
monopoly power. The reason is that transport costs are a form oftransactions costs, and from
the theory ofthe firm we know that the existence oftransactions costs such as tariffs and taxes
can provide protection for some inefficient firms. Geography acts in a similar manner,
because the costs of overcoming space in order to carry out market exchanges incur trans-
port-transactions costs. In the context of Figures 1.15 and 1.16, there are two general rules
governing the extent to which distance costs provide a firm with spatial monopoly power:
(i) The greater the values of the transport rates ta and tb, the lower will be the fall in the
market area of the firm, and the greater will be the monopoly power of the firm, for
any marginal increase in the price of either Pa or Pb' ceteris paribus.
(ii) The further apart the firms, the lower will be the fall in the market area of the firm,
and the greater will be the monopoly power of the firm, for any marginal increase in
the price of either Pa or Pb' ceteris paribus.
Therefore firms which are located at a great distance from each other, and which face signifi-
cant transport costs, will consequently exhibit significant local spatial monopoly power.
Price/Cost
A B
1°
(a)
i< Market of A
>1< MarketofB
>1
Price/Cost
A B
1°
(b) r Market of A
* Market
ofB
Market
ofA
~
~
Figure 1.16 Spatial market areas: one-dimensional models with varying transport rates and production costs
;0 ;x C ~ B
7~
L ;
~~ ~
A
Hotelling (1929) model, which describes firms' spatial interdependence within the context
of a locational game.
In Figure 1.17 we adapt Figure 1.15 to the case where both the production costs and
transport rates of firm A and firm B are identical. In other words, Pa = Pb and ta t b, and
we assume that these prices do not change. As before, we assume that consumers are
evenly distributed along OL and we also introduce the assumption that the demand of
consumers is perfectly inelastic, such that all consumers consume a fixed quantity per
time period irrespective of the price. In terms of firm strategy we assume that each firm
makes a competitive decision on the basis of the assumption that its competitor firm will
not change its behaviour. In the game theory literature this particular set of rules describ-
ing the nature of the competitive environment is known as (Cournot conjectures'. Given
that the firms are not competing in terms of their production prices, which are assumed
to be fixed, each firm can only adjust its location in order to acquire greater market share.
If the firms react to each other in sequential time periods, the location result can be pre-
dicted easily.
If we assume that the firms A and B are initially located at one-quarter and three-quarters
of the way along the market, respectively, firm A will have monopoly power over OX and
firm B will have monopoly power over XL. In this case, both firms will have identical market
shares. In time period 1 firm A will therefore move from its original location to a location at
C, just to the left of B. In this way firm A will increase its market share from OX to a new
maximum value of ac. Firm B will still retain market share over BL, although its market
share is now at a minimum.
Firm B will now assume that firm A will maintain its location at C, and so in time period
2, firm B will move just to the left of C. In time period 3, firm A will respond by moving to
the left offirm B, and this process will continue until both firms are located at X, in the mid-
dle of the market. Once both firms are located at X, neither firm has any incentive to change
its location behaviour, because any location change will involve a reduction in market share
relative to their location at X. In game theory, any situation in which neither firm has any
incentive to change its behaviour is known as a (Nash equilibrium'. The locational result in
which both firms are located at the centre of the market is the Nash equilibrium for this
particular locational game. Consequently, once the firms reach this point they no longer
continue to move. This is the Hotelling result. The details of this are given in Appendix 1.4.
At the conclusion ofthe Hotelling game we see that the market share ofboth firms located
at X will be half of the market, exactly the same as at the start of the location game. However,
from Figure 1.18 we see that the Hotelling result leads to a fall in consumer welfare relative
to the original situation. Given that consumers all consume a fixed quantity per time period
of the good produced by firms A and B, there is no substitution effect between the goods
produced by firms A and B and other consumption goods. Therefore the change in the
delivered prices at the each location will accurately reflect the change in welfare of the con-
sumers at each location. The net effect of these welfare gains and losses can be represented
by the areas under the delivered price curves, which are arrived at by comparing the deliv-
ered prices at the respective locations at the start and the end ofthe Hotelling location game.
Price/Cost
m
d
~o A ~x B L ~
r ~
Figure 1.18 The welfare implications of the Hotelling result
l
The consumers who are located in the centre of the market benefit by generally reduced
delivered prices, represented by eghj in Figure 1.18, whereas those located at the edges of the
market lose by generally higher delivered prices, represented by (defc) + (jklm) in Figure
1.18. The gain in lower prices for the central consumers is outweighed by the increase in
prices for the more peripheral consumers. The net effect is therefore a social welfare loss.
In one-dimensional space discussed here, the Hotelling result holds for two firms.
Meanwhile) in the two-dimensional case) the Hotelling result holds for the case of three
firms. However, beyond these numbers, there is no stable equilibrium result as firms keep
changing their location.
Price/Cost
P X L
~( )~( )
Market of A in Time period 1 Market of B in Time period 1
~~ ~
Market of A in Time period 2 with a price cut by Firm A
1.5 Conclusions
The foregoing discussion suggests that we can use classical and neoclassical models of firm
location behaviour to consider the spatial behaviour of firms operating under different cir-
cumstances and also with different objectives. We see that input factor prices, factor substi-
tution, and market prices are all critical aspects of location behaviour. Indeed, all location
problems are also production problems, and vice versa, and these interrelationships begin
to account for why geography is so important for economics. The analysis of market areas,
in which location strategy, output strategy, and pricing strategy are all seen to be interre-
lated, further reinforces this general point. The role of product quality and variety emerges
as an essential aspect of location behaviour and these issues are examined in more detail in
Chapter 2, which deals with the phenomenon of industrial agglomeration and clustering.
Discussion questions
1.1 How does the location of input sources and output markets determine the location
behaviour of the firm?
1.2 To what extent are firm -locational changes dependent on the input substitution
characteristics of the firm's production function?
1.3 How do changes in local land prices and labour prices affect firm location behaviour?
1.4 Explain how firm location changes are linked to changes in the potential input and
output market locations.
1.5 In what ways does space confer monopoly power?
1.6 What role can location play in the competitive strategy of firms, and how are location
and price strategies interrelated?
1.7 What role do logistics costs play in determining firm location behaviour?
1.8 What role does product differentiation play in location behaviour?
Cost
M1 M1
~ ~
d1 d2
and
The first-order conditions for the inner optimum locations are given by:
ml(aTl/ad l ) == m2(aT2 /a d 2)
2
ml(a T l /adr) + m2(a2T2/ad~»O
In other words, the transport costs associated with at least one ofthe inputs must be increas-
ing by more than proportionately with distance, as distance increases, in order for there to
be an optimum location between M 1 and M 2• We can see this in Figure A.1.1.2, where the
interior optimum is at d*.
On the other hand, if transport rates exhibit economies ofdistance, Le. ca
2
T 11ad 2 l ) and
(a2T2/ad~) are negative, or fixed transport rates i.e. (atl/ad l ) and (at2 /a d 2) are zero,
Cost
M1 M2
d*
~ ~
d1 d2
there is no interior solution. As we see in Figures A.I.I.3 and A.I.I.I, in these cases, which
are the usual two situations, the optimal location will always be at an end-point such as M 1
and M 2• In A.I.I.3 the optimum location is at M 1 whereas in A.I.I.I it is at M 2•
The final possibility is where there are trans-shipments costs associated with the loading
and unloading of goods at ports or terminals. In these situations, the 'terminal' costs associ-
ated with these trans-shipments points may alter the transport rates in a variety of ways.
Optimal locations with terminal costs can be either at end-points or at interior locations. As
we see in Figures A.I.I.4 to A.I.I.6, the optimal location will depend on the structure of the
transport costs.
Cost
M1 r "! M2
~ ~
d1 d2
Cost
M1
d1
7 ~
d2
Figure A.l.l.4 One-dimensional location problems with terminal costs and linear transport rates
In Figure A. 1. 1.4 the transport rates are constant, although not equal to each other, and
both transport cost functions exhibit terminal costs. In this case, the optimal location which
minimizes total transport costs is at the end-point M 2•
In Figure A.I.I.S, the transport rates are falling with distance, such that total transport
costs are concave with distance, although they are not equal to each other. Both transport
Cost
(m 1t 1d1) + (m 2 t 2d2 )
m2 t2d2
M1
~ EE
d1 d2
Figure A.l.l.5 One-dimensional location problems with terminal costs and falling transport rates
Cost
m2 t2d2
d'
~ -E
d] d2
Figure A.l.l.6 One-dimensional location problems with terminal costs and increasing and decreasing
transport rates
cost functions exhibit terminal costs. In this case, the optimal location which minimizes
total transport costs is at the end-point MI'
In Figure A.l.l.6, one of the transport rates is falling with distance, whereas the other is
increasing with distance. Both transport cost functions exhibit terminal costs. In this case,
the optimal location which minimizes total transport costs is at the interior location d'. The
classic proof of the one-dimensional location problem is given by Sakashita (1968).
Any profit maximization production-location point will need to satisfy the optimization con-
ditions both with respect to the input combinations, m I and m 2 , and also the locational coor-
dinates. In our Weber-Moses triangle (Figure 1.8) we can define the locational coordinates in
terms of two variables, namely the angle (J and the output shipment distance d3• Any changes
in the input distances d1 and d2 can be defined in terms of changes in these two variables. For
an optimum location-production result, the partial derivatives of the profit function with
respect to the four variables m 1, m 2, (J, and d3 must be equal to zero. Following Miller and
Jensen (1978), by partial differentiation, the first-order conditions for profit maximization are
a(Jr)
am -(p2 + at2 J t d(am
t d)- ~d-·--
( am am2J- md--------~---
------- (amat J( dm
am J 3 3 3
0 (A.I.2.3)
2
2 2
2 2
2 3 3 3 3
3
a(Jrl
ad 3
-mlt 1 (~~:J -mA (~1J( !~: J mt (~4?J
ad - md (_~t2--J(~~?J
ad2 ad 2 2
3
2 2
3
(A.I.2.S)
- m 3t 3 - m3d3 at3 J
(ad 0
3
ami = -(~'!!]-J[t
a(Jr) ami 3d3 + md (--~!~-J]
3 3 am3
-md(~-J
ami - 1 1
(p +
1
td) = 0
1 1
(A.I.2.6)
a(Jr)
ae = -m (-~4lJ[t
1 ae + d (_~-LJ]
ad - m(~E~J[t
ae
1
+ d (-~!.~J]
ae 1 2 2 2
0 (A. 1.2.8)
l
(A. 1.2.9)
- m [t + d{~d~)] = 0
3 3
Equations (A.I.2.6) and (A.I.2.7) together define the production relationships at the
optimum between each of the inputs and the output. Meanwhile equations (A.I.2.8) and
(A.I.2.9) together define the location of the firm at the optimum. To understand the
production function characteristics which will ensure that the optimum location of the
firm is independent ofthe level of output, we need to observe the conditions under which
the marginal rate of substitution between the inputs remains constant for all levels of
output. For an optimum location which is independent of the level of output, equations
(A.1.2.8) and (A.I.2.9) must be satisfied because the firm will not move. Therefore we
need only observe the production relationships. By rearranging equations (A.I.2.6) and
(A.I.2.7) we arrive at
dm 3
-mIdI (~tl)
amI
(PI + tId I )
(A.1.2.10)
dmi
[~:~~:=3~3(:~3 )]
and
am}
-mA(~~2)-(P2 +tA) (A.1.2.1l)
mA(X~J]
dm2
[t d
3 3 +
The term ratio (dm 3Idm l ) is the marginal product of input 1 at the optimum, and the term
ratio (dm 3/dm l ) is the marginal product of input 2 at the optimum. Dividing (A.l.2.ll) by
(A.2.2.l0) gives us an expression for the marginal rate of substitution between the two
inputs at the optimum, thus
The expression (A.l.2.l2) is constant, i.e. the marginal rate of substitution between the
inputs is constant, in the case where there are no economies of scale in transportation. In
this situation, the ratio (dt/dd l ) (dt 2/dd 2 == 0, and the expression reduces to
dm 3 /dm i am 3 /am 2
(A.1.2.13)
+ tIdI ) (P2 + t 2d2)
In other words, the marginal product of input divided by the delivered price of input 1 is equal
to the marginal product ofinput 2 divided by the delivered price ofinput 2. These conditions are
the spatial eqUivalent of standard microeconomic efficiency conditions, and this general
Weber-Moses result has also been shown to hold in the case of multiple inputs and outputs
(Eswaran et al. 1981). Note that economies of distance play no role in the result, because once
the firm is located at the optimum location, the distance relationships are all unchanging.
Productions functions which satisfy the conditions here are functions which are linear or which
are homogeneous ofdegree one. On the other hand, in situations where transport rates exhibit
economies of scale, the results of the Weber-Moses problem become much more restrictive.
where the parameters m, p, t, d are the same as in the above sections, and
The first terms on the right-hand side of(A.I.3.!) represents the ordering and procurement
costs which are incurred each time an input shipment is received, but which are independ-
ent of the shipment size. In manufacturing firms, these costs will also include machinery
set-up costs, and can be shown to very significant. As these costs are independent of the
shipment size but are incurred each time an input shipment is received, the total ordering
costs are a multiple ofthe shipment frequency, i.e. the number ofshipments per time period.
The second term on the right hand side represents the inventory capital holding costs, which
are a function of the average value of inventories held per time period. These costs are the
capital interest plus insurance costs associated with holding inventories. Assuming that we
consume inventories at a constant rate, and that stocks are replenished in a timely manner
such that our inventory levels stay constant, these costs can be seen to be a function of the
delivered price of the goods. In other words, inventory costs are a function of transport
costs. Finally, the third term on the right-hand side of (A.I.3.!) represents the familiar
transport costs term used above.
Using a similar logic we can also define the logistics costs associated with shipments of
output goods, denoted with the subscript (0', as
In this case, the capital costs associated with holding inventories are the opportunity costs
of output revenue which are incurred by not shipping outputs in a continuous manner and
selling them at the market price of Po.
For both input and output goods, the aim of the firm is to determine the optimum ship-
ment size Q* which minimizes the sum of the total logistics costs for any given locational
arrangement ofinput supply points and output markets. However, this is not as straightfor-
ward as it might initially appear, because it can be shown that while the optimal shipment
size Q* is a function of the transport rates, transport rates are also a function of the optimal
shipment size. In order to circumvent these problems, the definition oftransport rates must
be re-specified to allow for discrete shipments. While the details of this problem are beyond
the scope of this book (McCann 1993, 1998, 200 I), it can be demonstrated under very gen-
eral conditions that the optimal shipment size (McCann 2001), and therefore the optimal
average weight of inventories Q* /2 to be held, is a positive function of the distance of the
shipment, and the transport costs associated with the shipment. This also implies that the
ordering costs, which are an inverse multiple of the size of Q, are also a function ofthe trans-
port costs. The combined sum of all the interrelated components of total logistics costs can
be shown to be a concave function of distance transport costs (McCann 1998), with distance
cost curves similar in shape to those in Figure A.1.1.5.
By employing this broader logistics-costs description of the costs associated with trans-
porting goods over geographical distances, we can now re-evaluate the Weber-Moses prob-
lem within a logistics-costs framework. Under very general conditions (McCann 1993,
1998), we can show that there is no solution to the independent ofoutput optimum location
problem. Moreover, as we see in Figure A.1.3.1, as the value-added by a firm increases, the
optimum location of the firm K* moves towards the market.
As the value-added by the firm increases, or the higher up the input-output value chain
is the firm, the steeper is the negatively sloping interregional equilibrium wage gradient,
which would allow a firm to move away from a central market point (McCann 1997, 1998).
We can see this in Figure A.l.3.2, in which the point M represents a location containing
both markets and input source points.
The interregional equilibrium wage gradient associated with total logistics costs, rather
than simply transport costs, becomes steeper, Le. it changes form R} to R2, as the value-
added by the firm increases, or as the firm moves up the value-adding chain. In other words,
in order to encourage a firm to move away from a Weber optimum location at M where
wages are W M' interregional wage differences will need to be greater in order for a high-
value-adding firm to relocate. On the other hand, a low value-adding firm, or a firm lower
down the value-adding input-output chain, will be able to move in response to relatively
minor interregional wage differences (McCann 1997, 1998). Moreover, in Appendix 3.4,
M]
./: K* ......
.
M
1
l·:: \:'~ '\, M
2
WM
R2
M Distance
Figure A.l.3.2 Interregional equilibrium wage gradient associated with logistics costs
knowing that the optimum shipment size Q* is inversely related to the trip frequency f, we
are able to show that the interregional equilibrium wage gradient must be convex with
respect to distance, as in Figure A.1.3.2.
The overall location-production conclusions to come out of the logistics-costs approach
to the Weber-Moses framework is therefore that high-value-adding firms tend to be both
more market-oriented, and much less responsive to regional wage differences, i.e. they are
much less footloose than low-value-adding firms.
Second, a consumer located at point L must always buy from firm B. In other words, the
delivered price of the output of B at L must always be less than the delivered price of the
output of A at L. This can be written as
PB+tBb<PA+tA(d-a) (A.l.4.2)
Price/Cost
a A B
~ ~
I~ a
~ b
~ X'
~
~ c
~
Figure A.l.4.1 The Hotelling spatial framework
At the same time, thirdly, there must also be an indifferent consumer at a distance x' some-
where between A and B. For this indifferent consumer the delivered prices must be the
same. In other words
PA - PB + 2tx' - ta td + tb = 0 (A.l.4.4)
If the transport rates tA and tB for the two firms are the same, and the source prices, PA and
PB, of the two firms are also the same, we have
x' =a+d - b
(A.1.4.5)
2
The value of x' given in equation (A.l.4.5) represents the size of the market captured by firm
A, and the size of the market captured by firm B can thus be represented as
REG mONAl ~::~_1'."~'<ll'>.U&~'g ~"*'.. Ha'il<.,f 1IU"~~8t~~'"
Recalling from Figure A.1.4.1 that c (d b), we can rewrite (A. 1.4.5) as
x' = a + d - (d c) a+c
2 2
As such, if the transport rates are the same and also the product source prices are the same,
the boundary between the two firms is exactly half-way between the two firms, and is inde-
pendent of the transport rates, as we would expect.
For a given source production price PA' known as a (mill' price, the market revenue of
firm A depends on maximizing the value of x'. From (A.l.4.5) and the arguments in section
1.4.1, we see that this is achieved by increasing a and reducing b as much as possible, while
still ensuring that firm A is to the left of firm B. This location change then triggers the leap-
frogging behaviour described in section 1.4.1.
Agglomeration and
clustering
2.1 Introduction
In Chapter 1 we discussed the theoretical issues which affect the location behaviour of the
individual firm. Each of the models presented provides us with a way of analysing the par-
ticular microeconomic effects on firm location behaviour, of various influences such as
transport costs, local factor prices, production and substitution possibilities, market struc-
ture, and competition. In reality the actual location behaviour of the firm is a result of a
complex mix of each of these influences. Therefore this leaves us with the problem of deter-
mining which particular influences are the dominant influences in which situations.
Unfortunately, without information on the individual firm and industry, however, the vari-
ous microeconomic models above do not lead to any systematic conclusion as to whether
optimal firm location behaviour is more likely to result in industrial clustering or in indus-
trial dispersion. Yet, in describing the generally observable features of industrial location
behaviour, two particular features do stand out.
The first generally observed feature of industrial location behaviour is that most indus-
trial activities tend to be clustered together in space, and the reasons for this clustering
behaviour are examined in this chapter. Most productive and commercial activities do take
place in the immediate vicinity of other such activities, and such clusters may take a variety
of forms, including industrial parks, small towns, or major cities. However, as we have dis-
cussed in Chapter 1, firms make different location choices based on their production and
cost functions, the spatial delineation of their markets, and their competitive strategies.
Moreover, firm clustering obviously often leads to congestion costs and higher land prices.
Yet, in spite ofthese different location-production choices and congestion costs, the general
observation that industrial clustering is widespread is indeed valid. As such, this observa-
tion raises the important question of why it is exactly that activities are generally grouped
together geographically, and the reasons why firms are often clustered together in space is
the central issue discussed in this chapter.
The second generally observed feature of industrial location behaviour, which will
be discussed in detail in Chapter 3, is that there appears to be a size distribution of spatial
clusters, with different ranges ofactivities taking place in different clusters. Some activities
AND REGiONAL §::~~~"Jl'§~"<Jn'~§'i&.,.. Ha"."'''.<'"... "....
<l'
are typically dispersed over large areas, with goods being shipped generally over large
distances, while others are more geographically concentrated. These differences in the
degree of dispersion also give rise to regularities in the patterns of clusters. In particular,
within an individual country or market area, there will usually be a single largest city
cluster which exhibits almost all types of activities, followed by larger numbers of other
smaller clusters which increase in number as their individual size falls. The smaller clus-
ters will tend to exhibit a smaller range of activities taking place within them than the
larger clusters. These observations are collectively known as an <urban hierarchy', and the
particular reasons for the development of such a system of cities are discussed in Chapter 3.
Instead, this chapter will focus specifically on the issues explaining why activities cluster
together in space.
In sections 2.2 to 2.5 of this chapter, we will discuss in detail the various arguments and
explanations as to why industrial activities are often observed to be clustered geographi-
cally. The arguments in sections 2.2 and 2.3 centre on the role played by knowledge spillo-
vers, labour markets, and economies of scale in fostering agglomeration. Moving beyond
simply agglomeration) section 2.4 examines the transactions costs relationships underpin-
ning a broader range of different types of clusters, and section 2.5 discusses the role played
by both creativity and consumption in fostering urban agglomeration and clustering, in
particular. Yet these explanations all suggest a level of information availability, awareness)
and rationality on the part of individual firms and people, such that collectively their indi-
vidual decisions give rise to agglomerations and clusters. For this reason section 2.6 raises
the question of how clusters do actually arise in an environment of limited information,
conflicting goals, and uncertainty, and here we see that evolutionary processes become cen-
tral to the discussion. Not surprisingly, these evolutionary processes also reappear as a criti-
cal element in Chapter 7 where we discuss regional growth.
Industrial clustering refers to the observation that all types of commercial activities-
manufacturing, services, resource-based industries-are frequently observed to be grouped
together in space. In attempting to explain this observation, it is necessary to employ the
notion that economies of scale can be place specific. To see this, we can consider the spatial
outcomes of the alternative hypothetical case where all firms achieve constant returns to
scale. If, for whatever reason, a large group of such firms in the same sector or a range of
sectors ends up being located in the same place, the result of this clustering will be a large
level of investment at that particular location. These firms will require land and space for
their activities, and the high demand for land at the particular location will force up its price.
If everything else is unchanged, and if the firms all achieve constant returns to scale, the
increase in the price of land will reduce the profitability of all the firms at that location.
Similarly, the increase in the local land price will mean that the living costs of the labour
employed will also go up. In order to maintain the local labour supply the firms will also
have to increase wages. Once again, this will reduce the profitability of all the firms in the
area. The reduced profits will mean that firms located here will be less competitive than their
competitors located elsewhere and will struggle to survive in the market. Some firms will
move away to alternative locations while others will simply go out of business. Eventually
the cluster will disappear. This hypothetical example is, however, inconsistent with the gen-
eral observation that most activity clusters continue to exist.
On the other hand, let us imagine a situation where each of the firms in the same locality
achieves significant external economies of scale precisely because of the large number of
firms located in the area. In this situation, the high level of investment in the local area will
still imply high local land prices and high labour prices as before. However, the difference
now is that these increased factor prices may be more than compensated for by the
increased efficiency on the part of each firm. The result of this will be even higher profita-
bility for all the local firms, even though the local factor prices may be higher than else-
where. Other firms from other areas may now also consider moving into our area,
contributing to a further growth in the levels of local investment, factor prices, and firm
profitability. This in-migration of new firms will lead to a cumulative process of local
growth.
This hypothetical example is consistent with the observation that most activity clusters
continue to exist. However, in reality, the growth of local clusters tends not to be a process
which is continuously cumulative, as in the latter example. The reason for this is that this
would imply that all activity would end up at one location! Therefore, in order to discuss the
existence of spatial industrial clusters, it is necessary to employ the notion that place-
specific economies of scale do exist, but also to acknowledge that there may be limits to such
effects.
The third source of agglomeration economies is the existence of a specialized local labour
pool. This allows firms to reduce their labour acquisition costs, and there are two aspects to
this. The first is that firms require sufficient quantities of labour to respond to market con-
ditions. Therefore, if market demand conditions improve rapidly, a firm will wish to
expand its labour force quickly, and will need to undertake a search process to acquire the
workers. Secondly, the firm will also need to ensure that the employees are able to carry out
the tasks correctly. In many sectors the costs of training labour and skills acquisition can
be extremely high. This is because workers will need to be provided with specialist courses
and instruction. Also, the opportunity costs involved with the time involved in these train-
ing activities can be extremely high. However, if a firm is located in an area which already
has a large local pool of workers with the specialist skills required by the particular indus-
try, the costs to the firm of expanding its workforce will be relatively low. This is because
the firm will have to undertake little or no retraining activities. For industries in which
skills-acquisition costs are high, or in which the opportunity costs of time are significant
due to rapidly changing market conditions, a local pool of skilled workers will therefore be
of great benefit. The labour acquisition costs on the part of the firms, which include both
the search costs and the retraining costs, will be reduced relative to firms in dispersed
locations.
These three sources of agglomeration economies have been succinctly captured by
Duranton and Puga (2004), who describe them as processes of learning, sharing, and match-
ing. This very neat formulation allows us to see agglomeration not as a static phenomenon,
but as a dynamic phenomenon of simultaneous processes of interaction. Together, these
three sources of agglomeration economies can allow firms within a cluster to experience
economies ofscale which are external to any single firm, but which are internal to the group.
The key feature of each of these sources of agglomeration economies is that spatial cluster-
ing reduces knowledge and information transactions costs. Clustering therefore increases
the likelihood that the appropriate information will be transmitted, that the specialist req-
uisite services will be provided, and the appropriately skilled labour will be available, at that
location, relative to other more dispersed locations.
2.3.2 The types of agglomeration economies
The sources of agglomeration economies described above allow firms within the same
industry which are clustered together in space to achieve localized external economies of
scale. However, in many areas, groups of firms in different industries can be clustered
together geographically. For example, major cities may contain hundreds ofindustrial clus-
ters. The exact nature of the agglomeration economies may therefore be different in differ-
ent locations. In order to describe the particular nature of agglomeration economies in any
particular area, economists often adopt a classification which was first employed by Ohlin
(1933) and Hoover (1937, 1948). This classification splits agglomeration economies into
three types, namely internal returns to scale, localization economies, and urbanization
economies.
is only partially open, in that local rental payments will not guarantee access, although they
will improve the chances ofaccess. The social network model therefore contains elements of
both the Porter model (1990, 1998) and the new industrial areas model (Scott 1988), and has
been employed to describe the characteristics and performance of areas such as Silicon
Valley and the Emilia-Romagna area of Italy. In this model space is once again local, but not
necessarily urban.
When we consider the fundamentally different nature of each of these cluster types, there
are two major issues which immediately arise. Firstly, there is the question of whether there
are any risks associated with locating in a cluster which might deter firms from doing so,
and secondly, there is the question as to how to identify empirically the dominant features
and logic of the cluster.
On the first point regarding the potential risks associated with locating in a cluster, the
major issues here relate to the risks of experiencing what are known as unintended knowl-
edge outflows (Grindley and Teece 1997). From the agglomeration arguments we know
that one of the reasons why firms might wish to co-locate is to benefit from knowledge
spillovers. From the perspective of the individual firms the positive and beneficial aspects
of knowledge spillovers are knowledge inflows from other firms. However, where firms
undertaking R&D have privately developed novel ideas or techniques on which they wish
to build new products or services, in general their wish will be to keep such knowledge
secret until it can be protected by patents, licences, copyrights, and the like. Such firms will
want to avoid unintended knowledge spillovers to other firms, or in other words unin-
tended knowledge outflows. Otherwise the potential competitive advantages afforded by
this proprietary knowledge and proprietary assets will be lost. Tight firm boundaries and
long-term legal contracts are the standard technique for avoiding such unintended knowl-
edge outflows, exactly in accordance with the markets and hierarchies argument of
Williamson (1975) underpinning the organizational logic of the industrial complex. The
industrial complex model therefore allows for large-scale R&D to be undertaken with the
risk of such unintended knowledge outflows occurring. In the case of the social network
model the risks associated with unintended knowledge outflows are controlled by the
shared norms and values of the network, the foundations of which are built on a code of
non-opportunism. In the case of the pure agglomeration model, there are really no means
by which such unintended knowledge outflows can be limited. The point about this argu-
ment is that it is not automatic that firms will wish to cluster together in space (see Urban
and Regional Example 2.2).
What we see is that whether firms are clustered together depends on the knowledge
advantages ofclustering versus the knowledge disadvantages ofclustering, and these issues
are just as pertinent to small companies as they are to large companies. The only proprie-
tary asset that many small companies have is a novel idea, product, or technology. For
these firms the risks of clustering can be very real because if knowledge about the novel
idea, product, or technology unintentionally leaks out to competing firms, the raison d'etre
of the firm disappears. However, these firms are often too small to fully pursue their idea
alone and therefore they must take on board the risks associated with sharing knowledge
and information with potential suppliers, customers, or collaborators, all of whom could
become competitors. As such, the risks of clustering are not just limited to large firms, but
also arise for small firms. What becomes clear is that while there are advantages associated
Urban and Regional Example 2.2 To cluster or not to cluster?
wishes to co-locate with other firms depends on the balance betweenthe positive
I Ut:tICUl.;:' from received knowledge inflows, which are a public good, and the potential losses associated
unintended knowledge outflows. In markets where product life cycles are very short, such as in
in some of the Silicon Valley components of the semiconductor industry
1.·••••• iillSm<entan 1994), for many firms the only way to acquire the relevant market knowledge is by very
1?I~recluent tac€~-to,-talce contact, which can only be afforded by proximity (McCann 2005). Where the
for such high-frequency interactions is less, firms have much more flexibility to move away from
if they wish.
secrecy is absolutely paramount due to the enormous costs of R&D, as is the case in
I pharmaceuticals or in many non-Silicon Valley parts of the semiconductor industry (M.cCann and Arita
a very different location approach is often adopted by firms. In some cases firms will choose to
R&D within a closed environment while developing little or no linkages with the
1•• • >i~sUnroulnding regional economy, or even in effect becoming 'islands' of innovation (Simmie 1988).
I>J~lte!rn(ltlvlelV. firms may even choose to move away from other firms in the same industry. In both
of the firm is to remove the risks of unintended knowledge outflows, as these are
l regaraea as being too costly relative to the benefits of knowledge inflows.
with agglomeration) as outlined above) there are also disadvantages associated with clus-
tering that go well beyond the issues of congestion or rising land prices.
On the second point regarding empirical observation) as we have already noted) all indus-
trial clusters will contain features of one or more of these ideal types. However) determining
the major features of any particular cluster will require us to consider which of these particu-
1ar types is dominant. The reason is that empirically each of these cluster types will have
different manifestations. For example) in the pure model of agglomeration) the dominant
feature will be an appreciation in local real-estate values) with no particular purchasing or
decision-making linkages being evident between local firms. Discussions of the strength of
local purchasing or decision-making linkages) or the types of local alliances undertaken by
firms) will not indicate agglomeration behaviour. On the other hand) in the case of the
industrial complex there will be no real-estate effects) but the dominant feature will be a
stability of both purchasing and decision-making linkages. However) measuring the scale of
such linkages is less important than measuring the duration of the linkages. Finally) in the
network model, although the market environment will be highly competitive) the dominant
feature ofthe firm relations will be a willingness to undertake a variety ofinformal collective
and cooperative activities which cut across organizational boundaries) such as joint lobby..
ing) or inter-firm credit availability. Measuring real-estate appreciation or the strength of
local purchasing linkages will not tell us very much. Industrial clusters are therefore of a
variety of types) the observation and measurement of which is also a complicated topic.
Linking these individual types of clusters to economic growth and development is also
complex. It is clear from the data that both innovation and entrepreneurship are related to
agglomeration (Acs 2002; Van Oort 2004). However) the arguments here also imply that
innovation may just as easily be related to industrial complexes and social networks as to
agglomeration) because each of these three broad types represents a different way of solving
REG ~ 0 N A l. L ">W lV P<S "<J ~~r~ ~...... Hn "_.k 'LV',.. 'L ..l>'
the problems related to the generation, exchange, acquisition, and exploitation of knowl-
edge. In order examine how these cluster types are related to innovation processes,
Iammarino and McCann (2006) develop a cluster-innovation taxonomy which explicitly
links a slightly extended version of the Gordon and McCann (2000) cluster typologies out-
lined here to a famous innovation-system classification scheme known as the Pavitt (1984)
taxonomy. While the detailed issues arising from the Iammarino and McCann (2006)
approach are beyond the scope of this book, the important point to note here is that the dif-
ferent forms of clusters are all related to innovation, but the ways in which we can empiri-
cally observe these relationships are very different. Simple observations of the spatial
clustering of firms cannot be taken at face value as evidence of agglomeration, because their
reasons for locating in the same place may have little or nothing to do with knowledge spillo-
vers (McCann 1995a). In order to identify the relationships between firm location behav-
iour, knowledge, and innovation, it is nearly always necessary to complement the use of
secondary empirical data sources with the use of primary survey and case-study type evi-
dence on the nature of the industrial transactions and interrelationships operating between
the firms, in a manner akin to the Porter (1990) approach.
As well as the various industrial clustering arguments described here, over recent years two
other closely related explanations of the advantages of urban clustering have also emerged.
However, these explanations are somewhat different from those described so far, because
they focus on the advantages of people clustering in urban areas, rather than firm clustering,
and these explanations are the creative class hypothesis (Florida 2002,2005) and the con-
sumer city hypothesis (Glaeser et ale 2001; Glaeser and Gottlieb 2006).
The creative class hypothesis was originally developed by Richard Florida (2002, 2005),
who sought to explain how and why clusters of certain types of people, rather than firms,
were so important for regional growth. Florida's argument in its original form (Florida
2002) examined the performance of places populated with diverse and often rather uncon-
ventional and unorthodox communities, and his initial observations (Florida 2002) centred
on places such as San Francisco, a city with a long history of social innovation and social
tolerance. The central tenet of the creativity hypothesis is that places which are tolerant of
cultural diversity and cultural differences are also environments which are ideally suited for
fostering unconventional approaches to the development of novel ideas, systems, products,
or services. This unconventionality itself is argued to enable and encourage experimenta-
tion' innovation, and entrepreneurial behaviour ofall different forms. Florida (2002) argues
that the natural outcome of this culture of tolerance is an environment of creativity, which
is manifested in the form of various types of creative outputs in art, theatre, and media, as
well as in terms of novel business concepts and new technology and investment ideas. As
such, his original argument is that in these types ofenvironments artistic creativity will also
be associated with commercial dynamism, and the creativity of the community will there-
fore be manifested in different ways.
Florida also argued (Florida 2002, 2005) that people with these requisite creative per-
sonalities and skills will increasingly move to culturally tolerant places, in search of both
creative employment and commercial opportunities but also in search of a particular life-
style, which is tolerant of unconventionality. The in-migration of creative people also
drives further innovation in the tolerant and creative regions, both via local competition
effects and also via innovation-creation effects. In contrast, regions that are intolerant of
cultural diversity will increasingly lose these creative people, and consequently they will
also lose the commercial dynamism associated with these people.
In many ways Florida's arguments echo some of the explanations for the success of the
Dutch Republic during the seventeenth-century <Golden Age', which posit that the flower-
ing burst ofeconomic ingenuity in an environment of religious tolerance allowed for inflows
of Jews, Huguenots, and Catholics seeking refuge from religious intolerance in other coun-
tries. Even explanations for the nineteenth -century growth of some of the US coast cities
contain some of this logic. Moreover, it will become clear that the Florida arguments are
also in some ways related to the bridging and bonding (Putnam 1996) dimensions of social
capital, as discussed in Chapter 7. Indeed, as we will see in Chapter 7, some authors actually
refer to creative capital as a distinct entity in its own right.
A related argument is that of the consumer city hypothesis (Glaeser et al. 2001; Glaeser
and Gottlieb 2006), which posits that high-skilled and high-income people will increasingly
migrate towards cities offering high -quality amenities, such as opera houses, theatres,
museums, culinary outputs, and so on. This argument conceives of the modern city as a
place ofleisure as well as work, where high-income workers consume highly income-elastic
goods and services produced in these particular types of urban environments. This argu-
ment is related to the amenity-migration arguments discussed in Chapter 6, although the
emphasis of the amenity-migration models is on the importance of natural environmental
amenities while the consumer city hypothesis here emphasizes the importance of human-
produced urban amenities. The consumer city hypothesis is also closely related to the urban
gentrification arguments discussed in Chapters 4 and 10.
A point of overlap between these two models comes from the fact that artistic and culi-
nary outputs produced by creative people concentrated in urban areas are exactly the types
of highly income-elastic human-produced goods consumed by high-income workers in
urban areas. Various papers (Comunian et al. 2010; Abreu et al. 2012) have demonstrated
recently that the average wages earned by creative workers working in artistic activities tend
to be lower than those ofequivalently skilled workers working in other activities. In part this
reflects a more skewed wage distribution among creative-artistic occupations than other
activities, but, even allowing for this, it appears that creative workers are willing to accept
lower wages, presumably because some of the rewards to their labour are in the form ofjob
satisfaction. However, this observation suggests that the major beneficiaries of these crea-
tive processes are not the creative workers themselves, but the high-income consumers
working in the city and enjoying the high-quality creative services and amenities on offer in
the city, exactly as argued by the consumer city hypothesis.
Our understanding of these issues is still developing and there are both supporting and
dissenting voices (Caves 2000; Scott 2000; Markusen 2006) on these matters. What is impor-
tant for our purposes, however, is to recognize that urban clustering is not simply about
firms, nor where it relates to labour is it purely about matching. It also appears to be related
to interactions of people, and in particular to social aspects of interactions, which facilitate
the generation and the spreading of new ideas.
2.6 Limited information, uncertainty, and the evolution of
clusters
The models discussed in Chapter 1 and also in the earlier parts of this chapter rely largely on
the assumption that firms and individuals are basically rational, and the behaviour of firms
and individuals predicated on this rationality is sufficient to generate clusters. Yet this argu-
ment itself has three implicit dimensions. First, it is assumed that firms and individuals
either know, or the price mechanism will quickly reveal, the locations where the potential
profitability advantages are greater for their particular activities. Second, it is assumed that
firms and individuals will act on this information and use their location behaviour in order
to maximize their profits. Third, the cumulative effect of such individual choices is that
agglomerations and clusters will naturally emerge.
In reality, however, the information available to firms and individuals is often rather
limited, and different firms will often have different information available to them. For these
reasons, some commentators have argued that firms cannot and do not make rational deci-
sions in order to maximize their profits. Rather, they argue that firms make decisions in
order to achieve goals other than simply profit maximization. Therefore, from the perspec-
tive of both individual and aggregate group location behaviour, this critique might suggest
that the underlying motivation ofour location, clustering, and agglomeration models would
need to be reconsidered. In particular, the ways in which agglomerations and clusters arise
in environments of uncertainty and limited information may need to be reconsidered. This
critique, which is important to understand, has three themes: bounded rationality, conflict-
ing goals, and relocation costs. The first two themes discussed in Box 2.2 can be grouped
~
AND n:. it:. 'U HJ ~.~Jo% L b'i><w 1\..J ~~%..&' ~'lI"~ i! ~ ~'!>'~ 'ilw&' V i!" ~ J>
under the general heading of Behavioural Theories, and these arguments are not fundamen-
tally geographical in nature, nor were they originally directed at location models in particu-
1ar. In contrast, the third theme is essentially an explicitly spatial question.
The arguments concerning <bounded rationality' are most closely associated with Simon
(1952, 1959). This critique concerns the fact that firms in the real world face limited infor-
mation, and this limited information itself limits firms' ability to be <rational' in the sense
assumed in microeconomics textbooks. These arguments are a more general critique of
rationality within microeconomics as a whole. However, they have been argued to be par-
ticularly appropriate to the question of industrial location behaviour. The reason is that
information concerning space and location is very limited, due to the inherent heterogene-
ity ofland, real estate, and local economic environments. Therefore, when considering loca-
tion issues, it would appear that the ability ofthe firm to be <rational' is very much <bounded'
by the limited information available to it. In these circumstances, decisions guided by
straightforward profit maximization behaviour appear to be beyond the ability of the firm.
Therefore location models based on this assumption seem to oversimplify the location issue.
Location behaviour may be determined primarily by other objectives than simply profit
maximization, as discussed in Box 2.2.
The behavioural arguments imply that firms do not necessarily have the ability or the
desire to make decisions which are explicitly aimed at maximizing short-term profits.
Applying this argument to geography means that if we are faced with a set of spatial total
cost and revenue curves, such as those described by Figure 2.2, the firm will make different
location decisions, according to whether it is aiming to maximize profits in the short run or
whether it is aiming to earn satisfactory profits in the short run along with achieving some
other goals. For example, in Figure 2.2, if the firm is aiming to maximize profits in the short
run it will locate at point P. On the other hand, if it is aiming to maximize sales it will locate
at S, and if it is aiming to minimize production costs and to maximize production efficiency
it will locate at C. If the firm had perfect information regarding these different spatial cost
and revenue curves we can argue that the firm would always move to point P. However,
behavioural theories assume that information is imperfect. Given the limited information
Te
TC TR
TR
a p b ( 5 deC f
Figure 2.2 Spatial cost and revenue curves
-
available and the conflicting goals within the organization, the actual location behaviour of
the firm will depend on which is its particular dominant objective.
The third critique of the classical and neo classical location models comes from the ques-
tion of relocation costs. Relocation costs are the costs incurred every time a firm relocates.
The models described above all assume that location is a costless exercise. However, reloca-
tion costs can be very significant, comprising the costs of the real-estate site search and
acquisition, the dismantling, moving, and reconstruction ofexisting facilities, the construc-
tion of new facilities, and the hiring and training of the new labour employed. These signifi-
cant transactions costs, along with imperfect information and conflicting goals, will mean
that firms are unlikely to move in response to small variations in factor prices or market
revenues. In Figure 2.2, the areas in which positive profits are made, i.e. where TR > TC, are
known as (spatial margins ofprofitability' (Rawstron 1958), and are represented by the areas
between locations a and b, c and d, and e and f The relationship between marginal location
change and the profitability of the firm in these areas is given by a(TR - TC)/ad, and this is
represented by the differences in the slopes of the spatial revenue and spatial cost functions
as location changes. In the spatial margins of profitability in which the slopes of the spatial
revenue and spatial cost functions are very shallow, the marginal benefit to the firm of relo-
cation will be very low. Therefore, in the presence of high relocation costs the firm will not
move to a superior location even if it knows which alternative is superior. In conditions of
imperfect information and bounded rationality, conflicting goals, and significant relocation
costs, the behavioural approach would argue that once a firm has chosen a location, it will
tend to maintain its location as long as profits are positive, and not use relocation as a com-
petitive weapon. Rather the firm will attempt to reorganize its factor allocations and activi-
ties among its current set of existing plants. At the same time, the firm will focus primarily
on other price and non-price issues as competitive weapons, and the relocation of a plant,
or the reorganization of multi-plant activities which involves either the closing or opening
of a plant, will only be a last resort strategy. On the other hand, where relocation costs are
insignificant, the firm will take advantage of spatial revenue and spatial cost differences and
will be able to move to superior locations as a competitive strategy.
One obvious weakness of the behavioural critiques is that, unlike the classical and neo-
classical location models discussed in Chapter 1, the behavioural theories do not of them-
selves indicate why a firm chooses a particular location in the first place. In this sense the
behavioural approach is not prescriptive. However, the applicability of the location behav-
iour insights offered by the classical models and theories to real-world situations does need
to be interpreted in the light of the behavioural critique, because uncertainty, bounded
rationality, imperfect information, conflicting goals, and relocation costs are all features
particularly characteristic of the spatial economy. Indeed, as a whole, the application of the
behavioural critique to spatial behaviour suggests that observed spatial patterns are not
necessarily reflective of optimum location behaviour, but rather sub-optimal adjustments
to restricted alternatives. This critique provides an explanation as to why firm location
behaviour may not be so responsive to the available optimization possibilities, and why
aggregate spatial investment patterns in general may be very slow to adjust to the emerging
profitability opportunities. This obviously also puts into question the validity of many of
our models which assume that clusters and agglomerations will arise naturally in response
to price signals. This is important because we implicitly assume that firms and individuals
REGHyNAL ":."'..
,l•• "~,,*,,.§g~'R~"•• """","","" """.: • ,,,,
will voluntarily move and congregate where agglomeration advantages are on offer.
Indeed, if the behavioural arguments provide a powerful critique of the ability of firms in
the real world to act rationally in terms of location behaviour, then it would also appear
prima facie to question our assumptions as to how agglomerations and clusters arise in the
first place.
These difficulties can be overcome by considering the evolutionary argument of Alchian
(1950). Alchian's argument is that the behaviour of firms in conditions of uncertainty can
be understood by discussing the relationship between a firm and its environment, whereby
a firm's environment is understood to encompass all the agents, information, and institu-
tions competing and collaborating in the particular set of markets in which the firm oper-
ates. In Alchian's argument, we can characterize the uncertain economy by two broad types
of environments. One is an 'adoptive' environment and the other is an 'adaptive' environ-
ment. These two classifications are not mutually exclusive, but serve as the two extreme
stylized types, between which the real economy will exist.
In the 'adoptive' environment, all firms are more or less identical in that no firm has any
particular or systematic information advantage over any other firm. However, they will
have differences in their ability to survive, but these differences are often unrevealed or
unknown ex ante. The results of the competitive process will imply ex post that some firms
will be successful while others will not, although ex ante, no firms had any a priori knowl-
edge that their products or techniques would be superior to those of their competitors. This
characterization of the economy is Darwinian, in that the environment 'adopts' the firms
which were better suited to the needs of the economy, even though the firms had no particu-
lar knowledge beforehand that this was the case. In statistical terms, in any given time period
in the 'adoptive' environment, the probability ofa particular single firm making a successful
strategic decision is identical to that of all the other individual firms.
On the other hand, in the 'adaptive' environment, some individual firms are able to
gather and analyse market information, simply by reason oftheir size. Large firms in general
are able to utilize resources in order acquire and process information relating to their mar-
ket environment, and the purpose of these information-gathering activities by the firms is
to subsequently use the information to their own advantage, relative to their competitors. In
statistical terms, therefore, in any given time period in the 'adaptive' environment, the prob-
ability of a particular firm making a profitable strategic decision is increased by reason of
its size.
In the real world of heterogeneous firms and imperfect information, smaller firms will
tend to perceive themselves to be at an information disadvantage relative to larger firms.
Therefore they will tend to make decisions which mimic or dovetail with those of the larger
firms, in matters such as styles, protocols, formats, and technology. In part this is because
they perceive the market leaders to be the best barometers of market conditions, and also
because the behaviour of the market leaders itselfoften contributes significantly to the over-
all economic environment simply by reason of their size. By copying the behaviour of the
larger firms the small firms therefore perceive that they will maximize the likelihood oftheir
own success. The result is that large firms tend to overcome uncertainty by information
gathering and analysis, and small firms tend to overcome uncertainty by imitation.
This type ofleader- follower behaviour is common in models ofoligopoly and uncertainty.
However, this behaviour is particularly pertinent to questions of location. In environments
of uncertainty, larger firms will generally have the information and financial resources to
make more considered location decisions than small firms. Major firms will be able to make
location decisions more akin to those described by the Weber, Moses, and Hotelling models
of Chapter 1, given that they will generally have sufficient resources to evaluate the cost and
revenue implications of their location choice. These large firms will attempt to make rational
and optimal decisions, and the results of their location choices can be analysed by the types
of classical and neo classical models described above. On the other hand, small firms will
generally be located where their founders were initially resident. There will have been no
explicit initial location decision as such when the firm began operating. Often such small
firms are entrepreneurial start-ups, whereby founders who were previously working for large
firms decide to set up a new business in a related field, in many cases selling goods and ser-
vices back to the large firm for which they previously worked, as well as accessing new mar-
kets. The geographical distribution and also the technological profile of such <spin-offs'
therefore tends to closely mirror that of the established firms, and gives rise to the evolution
(Arthur 1991; Boschma and Martin 2010) of localized clusters of small and large firms, often
in related technological fields (Frenken et al. 2007). As we will see in Chapter 7, this techno-
logical relatedness is a fundamental aspect of regional growth.
Over time, increasing competition means that location will eventually become a deci-
sion-making issue even for small firms as they develop and grow. As such, in subsequent
location decisions, many small firms will tend to choose to open new establishments close
to other large firms located in different market areas. Similarly, there is much evidence to
suggest that large multi -plant firms grow by means of the establishment of new ventures, in
particular already well-established spatial concentrations of firms (Delgado et al. 2010).
This itself favours further concentration. Therefore, for small firms which are risk averse,
these clustering strategies are particularly good strategies, because as we see from the
Hotelling model, locating close to competitor firms ensures that an individual firm's market
share is no lower than that ofan equivalent firm. Moreover, the Salop (1979) argument sug-
gests that clustering acts as also partial defence against the instability associated with price
movements. The clustering of small firms around major firms is therefore very commonly
observed. As such, imitation also takes place in terms of the foundation or relocation of new
establishments, as well as the location behaviour of established firms.
2.7 Conclusions
This question of industrial clustering is the topic of Chapter 3, in which we discuss agglom-
eration economies, the growth of cities and urban hierarchies, and centre-periphery
relationships.
Many activities are clustered together in space, giving rise to the formation of cities. This
process of industrial clustering, however, typically leads to an increased demand for local
land and consequently local real-estate prices will tend to increase, as will local labour
prices. These increases in the prices oflocal factor inputs will therefore reduce profits, ceteris
paribus, thereby reducing the attractiveness of the area as a location for the firms, unless
certain countervailing features exist that more than compensate for the increased local fac-
tor costs. These countervailing features are generally understood to be agglomeration economies.
Following on from Marshall's (1890) insights as to the sources of agglomeration and the
Ohlin-Hoover approach to classifying agglomeration effects, various other related reasons
for the growth of cities have been proposed by a range of authors. The common element of
all ofthese approaches is that the generation, acquisition, and transfer ofknowledge is a key
component of all aspects of industrial clustering. However, as our transactions costs
approach has also demonstrated, the actual mechanisms by which these are achieved differ
between different types of cluster. Understanding the types of cluster is critical for identify-
ing the knowledge spillover processes which are economized on by the formation of a clus-
ter, and vice versa. Yet, while these various analyses explain why clusters may arise, they do
not explain exactly how they may arise in an environment of limited information, risk, and
uncertainty. Here, a discussion of the behavioural critique suggests that the leader-follower
behaviour typical of many industries will tend to encourage small firms to cluster together
in space close to larger firms. Such behaviour underpins an evolutionary process of cluster
formation, in which clusters and cities are seen to arise naturally, even in the context ofless-
than-perfect information. That is not to say, however, that agglomeration processes are lin-
ear and indefinite. In contrast, we also observe not only that many activities are geographically
dispersed, but also that many clusters and cities are also dispersed. Indeed, the spatial econ-
omy appears to be characterized by both geographical concentration and geographical dis-
persion, and the balance between these two features produces a system of cities, an urban
hierarchy. It is to these issues that we turn in Chapter 3.
Discussion questions
2.1 What are the three major sources of agglomeration economies and how do they
operate?
2.2 Explain the three major types of agglomeration effects. What difficulties are there in
identifying these different classifications?
2.3 What other descriptions and mechanisms of industrial clustering do we have?
2.4 Using a transactions costs framework, explain the role and contribution of knowledge,
uncertainty, and trust in different types of industrial clusters.
2.5 Which examples of industrial clusters from your country best reflect each of the
different types of clusters you have discussed?
2.6 What role do creativity and consumption play in clustering?
2.7 What insights are provided for industrial location analysis by behavioural theories of
firm behaviour?
2.8 Explain the ways in which evolutionary processes of adaptation and adoption to the
competitive environment are important for firm location behaviour.
-
The spatial distribution
of activities
As such, the firm may employ price discrimination in these situations in order to discriminate
against local customers and in favour of distant customers. Indeed, such practices are com-
monplace in international airline pricing behaviour. The logic ofthis is that the firm will already
have monopoly power over the local customers, and therefore it is able to raise the delivered
prices it charges to these local customers in order to offset the higher prices it charges to distant
consumers. The specific features of this price discrimination argument, which is most closely
associated with the work of Greenhut (1970) and Greenhut and Ohta (1975), are discussed in
detail in Appendix 3.1. In general, however, this type of spatial price discrimination will
increase the spatial market areas of individual firms, increase the average distance over which
goods are shipped, and thereby also increase the tendency towards firm dispersion.
A B (3.1)
a2- (x-a)2
A
J3=r (3.2)
then we have
r =
a2
---~-------~--- (3.3)
(x - a)2
a=
xJ; (3.4)
1+
The actual location of the market boundary between the two retail centres will be depend-
ent on the relative 'pulls' of the markets and the factors inhibiting the overcoming of
distance.
The pull ofthe market is the relative attractiveness of the market as a location for purchas-
ing, and in Reilly's approach, the attractiveness of the retail location, represented by the size
of the retail centre, depends on the variety of goods which can be purchased at the same
location. In other words, a larger centre implies that a greater variety of goods can be pur-
chased on each trip to the centre. Implicit in Reilly's approach here is a notion ofpurchasing
economies of scope, in that a single journey to a single centre will partially or completely
substitute for many individual journeys to purchase a range of products. The greater the
variety of goods available at the retail centre, the greater will be the attractiveness of pur-
chasing from there, even from a significant distance. At the same time, Reilly's approach
also assumes that the marginal costs of overcoming distance become successively greater as
distance increases. This could be represented by, for example, an increasing marginal disu-
tility of travel time. Obviously, if the retail centres are the same size the value of r will be 1,
and a = x/2. As such, in this unique case the market boundary will be equidistant between
the two markets. More generally, however, the different distances between retail centres and
their market area boundaries will be determined by the interaction of the relative attractive-
ness of the centre in terms of the possible economies of scope, and the accessibility of the
centre, which depends on distance.
Although Reilly's market area approach is a simple empirical rule of thumb, it does pro-
vide us with a fundamental insight into the nature of market areas. Goods will be purchased
over greater market areas, and consequently also shipped over greater geographical dis-
tances, from centres which produce a greater variety of goods, ceteris paribus. This observa-
tion therefore suggests that the greater urbanization economies associated with larger cities
will also imply larger hinterland market areas, and larger shipment distances for the goods
produced in the city. In contrast, more local market areas with short-distance goods ship-
ments will tend to be dominated by purchases of a lower variety of goods.
While the Reilly arguments emphasize the consumption economies available at particular
retail centres, the agglomeration and clustering arguments discussed in Chapter 2 empha-
size the production economies available at particular centres. The agglomeration and clus-
tering arguments assume that that the production of high-value, non-standardized, highly
customized and newer goods and services requires greater inputs of knowledge, informa-
tion, skills, and technology inputs than for lower-value goods and services, and that these
inputs tend to be more readily available in major urban areas than in smaller centres. The
knowledge and technology required in producing and marketing high-value non-standardized
products means that high-value products tend to be produced in a smaller number ofloca-
tions and by a smaller number of producers than low-value standardized products, whose
production using less knowledge-intensive inputs is widely dispersed. This smaller number
of locations for the production of high-value goods and services relative to low value prod-
ucts is itself sufficient to ensure that higher-value goods tend to exhibit greater market areas
and longer shipment distances than lower value goods.
Following the Reilly arguments, we can also reinterpret these observations in terms of
the characteristics of the products produced, whereby the relationship between the value
of the product and the distances over which high -value products are captured in terms of
product value-weight ratios or value-bulk ratios. High-value goods are defined as being
high value-weight ratio or high value-bulk ratio goods. In other words the value per
tonne of the good is high or the value per cubic metre is high. This approach assumes that
the unit weight or unit volume of a product can be used as a denominator or a numeraire
for measuring the value of something, and high-value products generally exhibit very
high value-weight ratios. The reason this is important is because transport costs are partly
charged according to the weight or bulk volume of the good being shipped, as well as the
shipment distance. The simple argument here is therefore that the transport costs involved
in moving high value-weight or high value-bulk ratio goods are very low as a percentage
of the price of the good and are likely to remain so even over long distances, whereas for
low value-weight ratio or low value-bulk ratio goods the transport costs are likely to
become a high percentage of the price of the good very quickly, even over short
distances.
This type of thinking therefore suggests that we are likely to see high value-weight ratio
or high value-bulk ratio goods typically shipped over long distances, while low value-
weight ratio or low value-bulk ratio goods will typically be shipped over short distances.
Moreover, to the extent that the major production locations of high-value goods also tend
to be in the larger and dominant urban centres, which are relatively few in number, the
value-weight ratio linkage conclusions outlined above will be even further reinforced. In
general, we would expect to see the average purchasing linkage distance for high value-
weight ratio and high value-bulk ratio goods being higher than for lower value-weight ratio
goods, many of which will simply be purchased locally.
These arguments have traditionally been widely used in the empirical geographical litera-
ture, by adapting them slightly to facilitate measurement into what are known as (linkage'
analyses, which examine the types of commodity flows across space and between firms.
Using this linkage methodology, historically one of the most common findings of tradi-
tionallinkage analysis is that there is a relationship between the delivery distance and the
nature of the products being transported. In particular, the higher the value-weight ratio of
the product, the greater will be the average distance of shipment (Lever 1972, 1974; Hoare
1975; Marshall 1987). The general (rule of thumb' justification for this observation used
primarily by geographers is that high-value products can be transported over large dis-
tances because the high value of the product can absorb long-distance transportation costs.
On the other hand, for low-value products, the transport costs will be very high relative to
the value of the good even over short distances, thereby restricting the distance over which
these goods can be shipped. For example, a very small weight or small volume of micropro-
cessors exhibits a much higher value than the same weight or volume of processed food
products.
While linkage arguments can sometimes be very useful, in general there are, however,
three weaknesses with these simple linkage arguments. The first of these weaknesses is not
too problematic, while the next two represent more fundamental analytical difficulties.
First, large urban centres are not the only production locations for high-value goods,
and major centres are also the production locations for many low value goods as well as
for high-value goods. However, if large urban centres are the source locations of a high
variety of goods, spanning the range from low- to high-value goods, it is clear that the
market areas of major urban centres will tend to be much larger than those for the small
centres, whose product ranges are much smaller. As such, high-value products will tend
to be delivered over long distances, and to the extent that high-values are reflected in high
value-weight ratios, this implies that high value-weigh ratio products will tend to be
shipped over the longer distances on average, given the smaller number of production
locations.
The second, and more fundamental, problem is that many service sector (products' are
without either weight or bulk. As such, value-weight or value-bulk ratios mean little in this
context. As we see in Chapter 9, this is a particularly important issue in the context of the
current phase of globalization with the rise of what is known as the (weightless economy'
(Quah 200 1), in which the provision of services, and in particular services which are primar-
ily delivered electronically, has grown rapidly. Traditional linkage analysis measures were
based on manufacturing and processing industries in a period when such sectors domi-
nated. Today this is no longer the case in many advanced countries, and these traditional
value-weight and value-bulk linkage measures are largely incapable of capturing many
aspects of the emerging weightless activities. This is because the weight or volume denomi-
nator term in all such value-weight or value-bulk ratios when applied to the weightless
economy would be zero, and all such ratios would therefore be infinite, rendering them
useless for capturing the competitive nuances between the different parts of these service
industries. While the traditional linkage approaches may still be of some use in discussions
regarding manufacturing and processing sectors, their more general usefulness may be
increasingly limited.
The third problem, which is more technical and represents a fundamental analytical
problem, is discussed in Box 3.1, and relates to the relationship between the value product
and the required return to investments.
All the issues discussed in this chapter so far provide reasons why many activities are
geographically dispersed, and why many transactions take place over large distances. As
such, these arguments and observations are largely in contrast to those discussed in Chapter
2, all of which emphasizes agglomeration, concentration, and clustering. The spatial econ-
omyas a whole therefore comprizes on the one hand, mechanisms, forces, and issues which
encourage industrial clustering and agglomeration, and on the other hand, mechanisms,
forces, and processes which encourage industrial dispersal. The balance between these two
countervailing tendencies will generate the actual observed spatial patterns of cities and
regions, and these are the issues to which we now turn.
.....
Chapters 1 and 2 and the above sections of this chapter describe the many theoretical reasons
behind both the spatial clustering and the spatial dispersion of industrial activities. However,
in reality the mix of these simultaneous tendencies towards either clustering or dispersion
does not appear to produce a spatial pattern which is entirely random. Observations from
many countries suggest that there is a certain regularity in the spatial patterns of activity, and
this regularity has two aspects. First, the spatial distribution of cities exhibits certain typical
features, and second, the numerical distribution of such cities also exhibits certain typical
features. Nations tend to be dominated by one or two primal cities, generally located in the
centre of the major populated regions of the country. These cities will tend to be the produc-
tion locations ofmost ofthe outputs produced by the economy. Other more peripheral regions
will tend to be focused around successively smaller cities which dominate less populated hin-
terland areas. These smaller cities will also tend to produce a smaller range ofoutputs than the
primal cities. At the same time, as the size of the individual city falls, the number of such cities
generally increases. The result is that both the size and spatial distribution ofthe urban centres
exhibit something of a hierarchical pyramidal pattern, as depicted in Figure 3.1.
~~!bY ~ ~w§ Pill X L R..""" '....", ~'<I'I• ..,.•., ~~$'~ ~ ~• • • '&'18 ".§ 1&.# ~!•• ~•• ,.'" AND
~
Dominant city: Rank 1
/@\ /@\
Third-order cities: Rank 3 @ .@ @
Figure 3.1 The spatial and hierarchical organization of the urban system
The dominant city, which has the largest population, is defined as the city with the highest
rank-ordering. The next group of similar-sized cities is defined as the second level in the
rank ordering of city sizes, and the subsequent group of smaller similar-sized cities as com-
prising the third level in the rank-ordering ofthe urban hierarchy. Exactly why the national-
regional economy should exhibit such a spatial hierarchical pattern is the subject of much
debate, and traditionally has been the focus of an area of research known as (central place'
theory. In the following sections we will discuss three different approaches to central-place
theory. Initially we will discuss the two classical approaches set within a two-dimensional
framework, namely those of Christaller (1933) and Losch (1954), and in section 3.4 we
introduce the more recent (new economic geography' approach to such questions, which
follows the work primarily of Krugman (1991 a, b) and Fujita et al. (1999a).
of the size of the market area of good g == 3, which we denote as rn 3, divided by the size of
the market area of good g == 2, which we denote as rn == 2, is given by k. This rate of increase
ofmarket area is assumed to hold for any successive move up through the hierarchical level
ofgoods produced within the economy. This is obviously a very strong assumption but has
played an important part in earlier eras of planning theory. Christaller also assumed that
there was a direct correspondence between the hierarchical level of the urban centre and
the range of goods it supplied. Any urban centre of level u == (1,2, .... N) is assumed to sup-
ply all of the goods g == (1,2, .... N) up to and including the corresponding level. Therefore
a city of level u 3 is assumed to supply goods g== (1,2,3). The level of the urban centre thus
corresponds to the range of goods produced in the city.
Given these assumptions, Christaller attempted to construct the particular spatial pattern
which ensured that all locations were supplied with all goods from the minimum number of
supply points (Beavon 1977, p. 22). As we see in Figure 3.2, which represents a spatial
economy with three levels of urban centres, markets, and goods, the Christaller system
arrived at a series of overlapping hexagonal market, areas, in which the number of urban
centres is inversely related to the variety of goods produced at each location. The justifica-
tion for these particular spatial market patterns comes in part from observation, as has
\ I
\ I
\ I
\ • I
\ I I
®---~-----®
I
I I
.
\
\
I I \
I tQ\ \ /"
... I ~, \ /'
"', I /", ,\'
I, " ...... / \
I...... / ,/' \
II
I ~ /'
.-----@ \!
: 0 :
I
@---------
I
\\ /@' . . . " /~ . . . II
\
\
A
"
/' '.
' ..... . ,, /
./ ...... I
I
., ......
/' \ 1(y I
"\ ~ I
/ \ . I
. \ I I
\ • I
®---~-----@5
I : \
I I \
I . \
I I \
I \
I \
City of level 3
@ City of level 2 © City of level 1 @
already been mentioned, but also from the principle that the maximum market coverage is
provided for from the minimum number of production points. Allowing for the analytical
limitations involved in the initial assumptions which motivate the system, the major contri-
bution of Christaller's work is to show that a hierarchical urban system can exist automati-
cally with a variety of different sized spatial market areas.
I Eo.b. means (free on board' and c.Lf. means (cost, insurance, freight: These terms are used in the transport
industry and relate to how shipping prices are quoted. With an f.o.b. price quote the product cost is quoted at
the origin and a transport cost mark-up is added to the product cost in order to arrive at the delivered price. The
production and shipment costs are therefore explicitly distinguished. In the case of a c.iJ. quote, the delivered price
of the product is quoted, which includes all transportation and shipment costs as well as the production costs. The
f.o.b. price reflects the price structures used in the various location theory models in Chapter 1. In the Hotelling
model the production cost is known as a (mill price' and transport costs are added to this.
Price/Cost
to
....... p
r I QK
QuantityQ
Quantity demanded
distance distance
.K] .K2
Market
Boundary
is that with hexagonal market areas, the individual production locations are arranged in a
triangular pattern with respect to each other. This ensures that the average distance from any
production location to a market boundary is minimized. Therefore the average delivered
price of the goods is minimized over space, as there is a maximum number of competing
suppliers in the spatial economy. Within the Loschian framework, this hexagonal spatial pat-
tern represents the ideal landscape for a single industry.
The geometrical features of Losch's ideal spatial system initially appear to be very similar
to those of the Christaller system discussed above. However, this is a quite different result
from the Christaller system, which required that the spatial market be supplied from the
minimum number of spatial production points, whereas the Loschian result provides for
the maximum number of supply points. With this observation in mind, however, we must
K6
now ask: what will be the ideal spatial landscape in the case in which the economy is charac-
terized by a range of types of firms?
The demand curves for different firms and the price elasticity of demand of the goods
produced by the various firms will be different according to the different types of goods they
produce. The result of this is that the market areas for different types of goods will also be
different according to the characteristics of the individual demand curves. Allowing for the
caveats discussed in section 3.7.2, we can assume that high-value products tend to be price
inelastic, due to product quality and product heterogeneity, whereas low-value products tend
to be highly price elastic due to product homogeneity. Therefore, in general, firms producing
goods whose demand curves are highly price elastic will tend to exhibit small market areas.
The reason is that the demand for the product will be very sensitive to the transport cost
mark-up on the source price of the good, with demand falling sharply over even small deliv-
ery distances. On the other hand, goods whose demand curves are highly price inelastic will
be relatively insensitive to transport cost mark-ups. The result of this is that demand will fall
only slowly with increasing delivery distance, thereby tending to increase the market area. At
the same time, different firms producing different goods will also exhibit different supply cost
curves, and it is the interrelationship of the demand curves and the supply curves which will
determine the actual sizes of the individual hexagonal market areas. The ideal Loschian hex-
agonal spatial pattern will therefore be different for different types of firms. Firms producing
highly price elastic goods, such as many agricultural commodities, will tend to exhibit small
hexagonal market areas and be located at many points, whereas firms producing low price-
elastic goods will tend to exhibit larger hexagonal spatial areas and be located at fewer points.
In a situation such as this, where the economy is made up of a variety of firms producing a
variety of products, and where the spatial economy exhibits a variety of hexagonal market
area patterns, the Loschian argument is that the most efficient economic landscape is one
where the maximum number of firms is located at the same point. The logic of this argument
is that the maximum number of firms located at the same point will allow agglomeration
economies to take place within each of the sets of firms which are located at the same place.
The mathematics and geometry ofthe Loschian argument and results are beyond the scope of
this book, although the general conclusions of Losch are quite straightforward. As we see in
Figure 3.7, Losch concludes that the economy of any spatial area will tend to be dominated by a
central primal city, the hinterland of which will be characterized by smaller settlements and
alternating areas of industrial concentration and dispersion. The rationale and justification for
Losch's actual conclusions have been the source ofmuch debate (Beavon 1977; Parr 2002). Other
small market area shapes, which allow for all the market spaces to be covered include triangles
and squares, are also possible in the Loschian-type framework, but hexagons are the most effi-
cient for servicing a market in terms ofboth transport costs and delivered prices (Beavon 1977).
Overall, Parr (2002) concludes that the primary contribution of Losch's work is to show
that industrial concentration and urbanization can arise independently of local peculiarity
or particularity.
The classical approaches to urban hierachies and central places represented by the work of
both Christaller and Losch were landmarks in the history ofurban and regional economics, and
both systems ofanalysis represented major breakthroughs at their time. Yet, in comparison to
today's modern microeconomics, and in particular in comparison with modern general
equilibrium frameworks ofanalysis, these classical approaches over time are increasingly seen
URBAN
/
Areas ofindustrial
concentration Areas ofindustrial
~ dispersion
to exhibit major limitations. These limitations relate primarily to the rather restrictive geo-
metrical features imposed on the analysis, and the rather restrictive nature of some of the
assumptions required in order to generate the spatial pattern regularities. Moreover, in these
classical approaches, little room is allowed for the roles played by labour mobility, consump-
tion patterns, or trade creation in driving the evolution of the spatial economy, and few
insights are therefore provided regarding processes of interregional convergence and diver-
gence. These issues are all central to the new generation ofmodels which emerged in the 1990s,
a school of analysis known collectively as (new economic geography', or NEG for short.
The fact that industrial concentration and urbanization can arise independently of local
peculiarity or particularity is a phenomenon which both underpins and also arises out of a
research agenda known as (new economic geography', or NEG for short. The origins ofNEG
are most closely associated with the work ofNobel Laureate Paul Krugman, who adapted and
translated ideas, insights and techniques from international economics to geography
(Krugman 1991a,b). Krugman's landmark 1991 publications sparked offa surge in the devel-
opment of new analytical techniques and empirical approaches to understanding the eco-
nomics of geography. Along with Krugman's key co-authors, most notably Masahisa Fujita
and Anthony Venables, as well as a long list of other researchers, NEG scholars have pro-
vided some major analytical and empirical breakthroughs in urban and regional economics.
The number of papers, books, and analytical contributions made by this research tradition is
so large that it has now become a whole sub-field of spatial economics in its own right.
Fortunately there are now many excellent books (Fujita and Thisse 2002; Baldwin et al. 2003;
Henderson and Thisse 2004; Combes et al. 2008; Brakman et al. 2009) from which the inter-
ested reader can gain a detailed understanding of the major insights generated by the very
large and very technical literature emerging from the NEG tradition. However, rather than
examining a particular individual NEG model, for our purposes here it is important to sketch
out the major features of the NEG tradition and to position and link these with the other
major issues examined in the broad field of urban and regional economics.
New economic geography models, hence forward referred to as NEG models, contain a
small number of basic building blocks. The first is the relationship between size and variety.
NEG models work on the assumption that product diversity increases the welfare of indi-
viduals and households by increasing the range of consumers' choices and consumption
opportunities on offer. At the same time, product variety also increases a firm's production
efficiency by increasing the range of a firm's factor input and market output choices or
opportunities. Importantly, within NEG the setting where size and variety come together is
the urban area, the city, and the justification for this setting is based on the agglomeration
arguments discussed in this chapter.
The second element within these models is factor migration. Labour is able to move to dif-
ferent degrees in different NEG models, some ofwhich allow for zero migration and some for
high levels of mobility depending on whether the labour is working in the production of
agricultural goods or manufactured goods. In many industries producing human-produced
goods rather than land-based agricultural goods we often observe increasing returns to scale
rather than constant returns to scale. Once again, in an NEG setting size comes together with
product variety and also labour inputs in an urban setting, in a city. A greater number oflocal
labour inputs drives down local wages and allows firm's to operate more efficiently. Taken
together, within NEG models cities are seen as the interface where production economies of
scale, welfare in consumption, and efficiency in production all come together.
The third and final element in the basic NEG schema is the particular way that the costs of
overcoming distance are incorporated into the framework. For technical modelling reasons
these are generally defined in a form known as <iceberg' transportation costs. Appendix 3.3
provides a more technical description of the basic elements contained in the original explic-
itly spatial NEG models and a more complete description of how these are understood to
operate in an explicitly spatial setting. However, at this point what is important to note for
our purposes is that in an NEG general equilibrium setting the combination of urban scale-
related production efficiencies, partial labour mobility, and distance costs together gives rise
to both centripetal forces promoting spatial concentration and centrifugal forces promoting
dispersion. The observed spatial distribution of economic activities is seen to depend on the
balance between these two opposing forces, and the outcomes are that in some regions eco-
nomic activity is dispersed and in other regions it is spatially concentrated.
Theoretical NEG models tend to fall into two broad types, namely those which are aspa-
tial and those which are spatial. NEG models which are largely aspatial in nature tend to be
closer to international trade models in that they do not model geographical distance explic-
itly, but rather discuss distance costs within an iceberg transport costs setting as being vari-
ously <high' or <low'. This formulation allows all trade-related costs incurred in crossing
borders, including trade tariffs and customs duties as well as transport costs, to be treated in
an aggregated and consistent manner. On the other hand, the spatial NEG models do explic-
itly model distance using a distance-varying iceberg specification.
There are many different insights generated by NEG models, but for our purposes it is
useful to highlight and summarize two key insights, one ofwhich is associated with an aspa-
tial NEG model and one of which is associated with a spatial NEG model.
URBAN AND n,!!;;'U~~'J§"<!tRL ~:~~l,.jl'~~%§~¥'~~'L. n>"JS""QV~M~'~
In terms of aspatial models, one of the most important NEG insights is provided by the
model of Krugman and Venables (1995). Their model analyses the situation of two neigh-
bouring economies, one of which is large and contains a large agglomeration and one of
which is small and contains only small urban areas. Both countries or both regions are
assumed to produce two sets of outputs, one of which is agricultural goods produced under
constant returns to scale, and the other of which is manufactured goods produced under
increasing returns to scale. If the economies are largely closed to each other because trade
costs are very high, Krugman and Venables (1995) demonstrate that all countries or regions
exhibit similar production patterns. The reason is that high trade costs act as a trade barrier,
thereby encouraging local production and a tendency towards autarky. However, as trade
costs begin to fall an increasing centre-periphery divergence forms in that the agglomeration
advantages ofthe large country or region begin to dominate while the small country or region
declines. Ifproduction factors are mobile, then capital and labour will shift towards the larger
country and away from the small country. The reason is that the large country has a much
stronger 'home-market' effect, in that the larger domestic market provides a platform for
greater domestic levels of competition, product variety, and scale, all of which are linked via
agglomeration processes. The scale-productivity advantages of the large country with a large
home market and large agglomerations only disappear as trade costs fall towards zero. If
trade costs are close to zero, then there is little or no location advantage and the benefits of
agglomeration more or less disappear. In this case investment can just as easily take place
anywhere and convergence between the countries or regions starts to re-occur. The relation-
ship between the level of trade costs and the degree of spatial concentration of economic
activities is represented by an invented-U shape, in which both very high and very low trade
costs are associated with very low spatial concentration of activities, while a large range of
intermediate levels of trade costs is associated with high levels of spatial concentration.
In terms of the explicitly spatial NEG models discussed in detail in Appendix 3.3, the
evolving spatial distribution ofactivities within an NEG framework (Fujita et al. 1999a,b) can
be shown under fairly general conditions to reflect both the types of urban hierarchy patterns
described byChristaller (1933) and Losch (1944), and also the rank-size rule pattern ofurban
distributions discussed below. The cities emerge at discrete points in space interspersed with
areas of no urban development, rather than as a continuum of urban developments across
space. The reason for this is that in the immediate hinterland around a city there are few or
no possibilities for competing via commercial developments because the nearby city domi-
nates the local market area. This is known as a 'shadow effect'. It is only at a significant dis-
tance away from the city that the costs of distance provide enough 'protection' from the city
producers for firms to invest profitably. The role which distance plays in conferring local
monopoly power and allowing less efficient producers to survive has already been discussed
in Chapter 2. The one-dimensional simulations of Fujita et ale (1999a,b) have been extended
to two-dimensional simulations by Stelder (2002, 2005), and these clearly demonstrate that
clustering and dispersion are natural outcomes of spatial competition processes.
The theoretical developments in NEG have been rather more significant than the empiri-
cal developments. The major empirical developments in NEG relate to the ways in which
market potential can be estimated. Market potential is traditionally measured as Harris
(1954) market potential, in which the sizes of the different markets which can be accessed
from a point are weighted according to the distance and then aggregated together. NEG
".
models extend this approach (Redding and Venables 2004) by also including the impacts of
competition, variety, and wages into the system. While the use of these approaches is
becoming more widespread, the results of these models are found to be very sensitive to the
actual specifications employed, and the reasons for this are related to the particular ways in
which transport costs are defined in these models (Bosker and Garretsen 2010).
At this point it is fair to say that in general the contributions of NEG are greater in terms
of theoretical insights than in terms of empirical developments, and whether the balance
changes over time remains to be seen. Certainly, there are increasing numbers of spatial
econometric models incorporating NEG frameworks into their analysis (Fingleton 2005),
but the primary impact of NEG is in terms of analytical insights.
Observations from many countries suggest that there appears to be something of a regular-
ity to the size distribution of the cities within a country, a regularity which NEG model
simulations have also been able to generate (Fujita et al. 1999a,b). This apparent regularity
is often discussed in terms of what is known as the rank-size rule.
As with any economic observation of distributions we can write a very general descrip-
tion of a distribution function as f(x) , whereby f(x) shows the frequency with which a vari-
able X takes a given value x in the sample or population. If X is continuous, the cumulative
distribution which shows the number of observations not greater than x can be written
(Chiang 1984) as
F(x) = J
o
f(x)dx (3.5)
In the case of cities, as we have already mentioned, most cities within a country will tend to
be small, with successively larger cities being progressively smaller in number. The result of
this is that a frequency distribution F(x) of urban areas, ranked according to the size (x) of
the individual urban area, will tend to be skewed to the left. Yet as we shall see in Chapters 8
and 9, very large urban areas playa crucial role in the behaviour and performance of the
overall economy, due primarily to the presence of agglomeration economies. The result of
this is that urban and regional economics tends to place a greater emphasis on the behaviour
and performance of these large urban clusters, which are relatively few in number. For this
reason, city-size distributions are measured from the right-hand side of the distribution. In
other words, if the total number of urban areas is given as T, the city-size distribution func-
tion R(x) is defined as R(x) == T - F(x). The city-size function R(x) therefore describes the
number of urban areas which are greater in size than x. As the actual city-size distribution
is skewed, we employ a non-linear function. Within urban and regional economics, the
usual functional form ofthe city-size distribution is a modified version ofthe Pareto-income
distribution function (Mills 1972; Mills and Hamilton 1994), given as
where M is the population of the dominant metropolitan area. For any country, the size
distribution of the urban areas can be estimated econometrically by taking a natural log
transformation of this function. In other words we estimate the function:
In R(x) = In M - a In x (3.7)
On the basis of many empirical observations, however, a simple common assumption is that
the value of a is close to 1. In other words, our city-size distribution function is given as
where x is the individual city size, and R(x) is the rank order of the particular city within the
urban hierarchy. The situation in which the value of a in equation (3.7) is assumed to be
close to 1 is known as the simple rank-size rule or Zipf's (1949) law.
Zipf's law produces a distribution of urban centres which corresponds to the patterns
observed in Figure. 3.8, and this is a pattern which has been observed in many countries.
Assuming that the slope is equal to -1, the simple rank-size rule can be rewritten as
xR(x) = M (3.9)
which states that the size of the individual city multiplied by its rank order is a constant M,
where M is defined as the population ofthe largest urban centre in the country. For the larg-
est metropolitan area, the population will be M and its rank order will be 1. For the second-
rank urban areas, the population will be approximately half that of the dominant city, and
for the third-rank urban areas the populations will be approximately one-third that of the
dominant city. Obviously, the rank-size rule is neither a rule nor a law, but it is a useful
approximation in many countries.
While advanced countries often closely correspond to Zipf's law across much of their
urban hierarchies, in Box 3.2 we see that in some countries the size of the largest and primal
In R(x)
o
o
0, 0
o~
0,0
In(x)
discussed in Appendix 3.3, suggest that the rank-size rule does indeed approximate to the
long-run spatial distribution of a mature spatial system. This would suggest that the rule
might be the result oflong-run general equilibrium economic processes. In contrast, Gabaix
(1999a,b) demonstrates that the rank-size rule could also be a natural statistical outcome of
a growth process in which random growth shocks are independent of the size of the urban
area. As such, this argument suggests that the rank-size (rule' or Zipf's (law' is in essence a
statistical artefact, and is actually commonly observed in other areas of nature and science.
What does appear to be the case, however, is that the distribution remains robust even
allowing for the complex patterns of growth and decline across cities (Duranton 2007).
The debate as to whether the rank-size rule is primarily a result of economic or statistical
processes has implications which relate in particular to countries undergoing rapid indus-
trial and regional transformations. In Chapter 1 we observed that there are nowadays major
differences between the urban distributions of the industrialized and newly industrializing
economies and those of many of the advanced economies in that many of these countries
are dominated by enormous cities, some of which are many times larger than the next city.
If the rank-size rule is indeed the result of long-run economic processes, this would suggest
that the highly skewed urban distributions within many newly industrializing economies
are something ofa transition phenomenon, and that as their economies continue to develop
and grow, their urban systems will naturally tend to move towards the rank-size spatial
distribution associated with industrialized economies. However, caution needs to be exer-
cised with such interpretations because, as we have already noted, at the moment it is
accepted that the rank-size rule is neither a rule nor a law. Moreover, depending on the
slope coefficients we employ in equation (3.7), a variety of rank-size relationships is possi-
ble. If the observations are primarily statistical rather than economic, then very little can be
inferred regarding the long-run development of the urban system of newly industrializing
economies. Either way, whether there is indeed any long-run adjustment of the urban sys-
tem in developing countries towards to something closer to the patterns evident in advanced
economies is also likely to depend on many other issues will be further examined in detail in
Chapters 8 and 9, on the basis of the analytical arguments developed in Chapters 1 to 7.
As we have seen in this chapter and in Chapter 2, different sizes of cities are associated
with different degrees of specialization and diversity, concentration, and diversification,
and there is a range ofempirical techniques available for capturing these variations, some of
which are discussed in Box 3.3.
URBAN AND ECONOMiC nl'~x,"'IM7~",(Il;~~JI' METHODS
The reason that we wish to measure issues of regional diversity, regional specialization, and
regional concentration is that the relationships between geographical structure and eco-
nomic structure contribute to all the regional economic adjustment and growth processes
captured in the various chapters of this book. As yet, however, there is no overall consensus
as to whether specialization (Henderson et ale 1995) or diversity (Glaeser et ale 1992; Combes
2000) is better for the long-run growth of a region (De Groot et al. 2009), because it appears
to depend on the context. In particular, the degree of a region's diversity or specialization is
often observed to playa different role in different locations and different time periods. We
know that large cities tend to be diversified, and the strong growth performance of many
large cities since the late 1980s would therefore suggest that diversity is an advantage for
many ofthe reasons outlined in this chapter. However, as we will also see in Chapter 8, more
recent evidence from the OECD suggests that many smaller and less densely populated
regions are now exhibiting stronger growth than many large urban areas, evidence which
would appear to point to the advantages of specialization over diversification in many cases.
Following Christaller (1933), Losch (1954), and Fujita et al. (1999), an urban-systems type
of approach would suggest that it is the relationship between different types of places, each
of which plays a different role in the economy, which is critical. However, exactly how we
might demonstrate this empirically is a difficult problem.
The structure and interpretation of spatial data is a major field in its own right, and a set
of topics beyond the scope of this book. However, for those readers interested in exploring
this field there are some excellent books on spatial economics (LeSage and Pace 2009;
Anselin 1988), geographically weighted regression models (Fotheringham et al. 2000, 2002),
and other methods of spatial data and computational analysis (Batty 2005).
Yet the technical issues involved in interpreting spatial data are not trivial. All spatial data
suffer from an underlying statistical problem known as the modifiable areal unit problem-
MAUP (Openshaw and Taylor 1979). The problem arises because even though economic
activities are distributed in space, the empirical analysis of these data is heavily affected by
the cartographical boundaries of the sub-units of the area being analysed. For example,
given the spatial distribution of the activities in the US economy, we can represent the spa-
tial distribution of these activities by reporting data according to the fifty US states.
Alternatively we could report the data at the level of the more than three thousand US coun-
ties. What we observe is that the indices of regional specialization, concentration, or diver-
sity change while the actual geographical distribution of economic activities remains
unchanged. Similarly, we could redraw the cartographical boundaries of the states or coun-
ties so that the average size of the areas remains unchanged while their shapes change. Once
again, the indices of specialization, diversity, or concentration are seen to change for no
actual changes in the spatial distribution of activities (Menon 2012). The reason why the
MAUP exists is because the spatial data generated depend on the definition of the carto-
graphical sub-units into which the overall economy is disaggregated.
A similar, but also quite distinct, problem from MAUP exists in terms of using data classi-
fied at different hierarchical levels (Van Oort 2004) of sectoral decomposition or aggregation
(Mameli et al. 2008) even for exactly the same spatial units. For example, if total economic
output is broken down into ten industries or sectors, the regional indices of specialization,
concentration, or diversity will differ from the case where total economic activity is broken
down into fifty industries or sectors, for exactly the same cartographical units. Moreover, the
difficulties associated with using and interpreting spatial data do not end there. Other indices
have been developed to measure the spatial concentration or dispersion of activities while
allowing and controlling for the observations of clustering which are for statistical reasons
unrelated to actual firm location behaviour or choices. For example, indices have been devel-
oped to control for the fact that some observed patterns of employment clustering and dis-
persion may simply be the result of the distribution of firm sizes with small numbers oflarge
firms and large numbers of small firms (Ellison and Glaeser 1997; Maurel and S'Edillot 1999;
Guimaraes et al. 2007). Other indices have also been developed to control for the effects ofthe
drawing of cartographical boundaries (Duranton and Overman 2005) and the effects of
shared linkages between industries (Ellison et al. 2010). In each of these cases, the aim of
these indices is to account for the degree of spatial clustering which occurs in addition to
what would be expected on the basis of the definitions of firms, areas, or industries.
lHU3AN AND ECONOMU::: AND
3.7 Conclusions
This chapter has discussed the various reasons why the spatial pattern of industrial activity
exhibits both concentration and dispersion. Different industries will exhibit different spa-
tial patterns, according to the extent to which they benefit from spatial proximity. There is a
variety of potential benefits from spatial industrial clustering, the impacts of which will be
different for different firms in different locations. However, at the same time, in other differ-
ent industrial sectors and firms there will be a preference for the dispersal of firms. As we
have seen, the underlying reasons why particular industries tend to benefit from spatial
concentration or dispersion are many and varied, and, at present, there is no full consensus
on these issues. Yet these various patterns of spatial industrial concentration and dispersion
do tend to give rise to a hierarchical pattern of urban centres, the regularity of which can be
captured by a range of empirical measures. Understanding the economics within these
urban centres and clusters and also the economic relationships between these centres and
clusters are central topics within urban and regional economics.
Discussion questions
3.1 Explain how pricing behaviour can lead to firm dispersion.
3.2 Explain why simple product value-weight ratios or value-bulk ratio indices are often
misleading indicators of location and firm linkage patterns.
3.3 Discuss the Christaller and Losch frameworks for explain the structure of urban
hierarchies.
3.4 In what ways do the (new economic geography' models differ from traditional central
place models?
3.5 What is the relationship between urban size, the number of cities, and urban diversity
or specialization?
3.6 How can we measure each of these issues?
3.7 Explain the ideas underlying the rank-size rule and Zipf's law.
3.8 To what extent do the urban patterns in your own country reflect the (rules' or (laws' of
city-size distributions?
which gives
p+(~)(~)p
J(TR)
MR (A.3.l.3)
dQ
dQ/Q
-ap/p- = e (A.3.l.4)
MR = p(l-n (A.3.1.5)
With this general aspatial expression for marginal revenue, we can now consider the dis-
tance costs which will eat into any revenue earned at any location. Following Greenhut
(1970) and Greenhut and Ohta (1975), if we define distance costs as td, where d is the dis-
tance and t is the transport rate, the theory of third-degree price discrimination suggests
that the marginal revenue net of transport costs gained at all locations should be the same.
In other words, the value of (MR td) for all locations should be the same. If we set the net
marginal revenue at any given value k, such that (MR - td) = k, this implies that k is invari-
ant with respect to d. Therefore
MR - td = p( 1- ~) - td k (A.3.1.6)
holds for all locations. If we set t = 1 for simplicity, this can be rewritten as
k =P - Pe- l d (A.3.l.7)
which, by differentiating with respect to d, with the knowledge that k is invariant with
respect to d, allows us observe the behaviour of delivered prices as distance changes, under
a regime of price discrimination, thus:
~~=~~[l-~+~] o (A.3.1.9)
where
dele
a = af;p (A.3.1.10)
which represents the proportionate change in the price elasticity of demand relative to the
proportionate change in price. Our above expression can be rearranged to give
: [~=--~ + a] _ 1= 0 (A.3.1.11)
Therefore
and
dP e
(A.3.1.13)
dd - a-(I-e)
The change in delivered prices with respect to delivery distance therefore depen,ds on the
value of a. If a = 1 there will be no price discrimination in favour of distant customers. The
value of ap/ad will be equal to I, which means that prices will increase linearly with distance.
On the other hand, if a > I, the value of ap/ad will be less than I, which means that prices
will increase less than linearly with distance. In this case, the firm is discriminating delivered
prices in favour of distant customers and against local customers, over whom the firm
already has a monopoly. If a < I, the value of ap/ad will be less than I, which means that the
firm will discriminate against distant customers by increasing transport mark-ups beyond
the actual cost of the freight. However, given that a rational firm will always attempt to
maximize its market area, ceteris paribus, this theoretical possibility makes no sense.
Moreover, as Greenhut and Ohta (1975) point out, as long as demand falls to zero at some
price level, then a cannot be less than unity. Therefore we can rule out this latter
observation.
From the point of view of market areas, this price discrimination argument implies that
market areas can be extended under conditions where firms have some level of spatial
monopoly power. The larger the firm, and the greater the consequent level of firm monop-
oly power, the greater will be the market area of the firm, and the greater will be its ability to
employ price discrimination. Therefore the larger the firm and the greater its level of
monopoly power, the greater will be the average distance over which goods are shipped, and
the greater will be the possibilities for firms to be spatially dispersed.
A B
a 2 - (x a)2 (A.3.2.1)
2
r= a (A.3.2.2)
(x a)2
I; = a (A.3.2.3)
x-a
and
a (A.3.2.4)
x-a=J:
which gives
Therefore we have
xI; (A.3.2.6)
a==l+I;
URBAN
which gives the distance of the market boundary a from retail location A (Richardson
1978). More complete two-dimensional descriptions of Reilly's law can be found in Hoover
and Giarratani (1985) and Parr (1997) in which the shapes ofthe market areas are described.
The two central-place approaches discussed in section 3.7 are set explicitly within a two-
dimensional spatial framework. However, there is a sense in which these two approaches
are rather static, in that the historical evolution of these ideal spatial urban systems is
ignored. In both of the above models it is implicitly assumed that the spatial outcomes of
the competitive market process will automatically converge towards something close to
the ideal landscape. Yet, until recently, these assumptions have not been tested. Many new
insights have been provided by the recent area of research commonly known as (new eco-
nomic geography' (NEG), which follows the work primarily of Krugman (1991a,b; 1993),
Fujita and Krugman (1995), and Fujita et al. (1999a,b). The models developed within this
particular research programme have attempted to generate and simulate Christaller-type
general equilibrium results within a monopolistic competition framework. The analyses
are set in one-dimensional space, and the models are based on a set of simple assumptions
regarding the costs of distance, the utility of consumers, and the productivity of manufac-
turing and agriculture. While a detailed analysis of this particular research field is well
beyond the scope of this book, following Krugman (1991a,b) and Fujita et al. (1999b), the
basic tenets of the original explicitly spatial NEG models are outlined here.
There are three basic assumptions upon which these complex new economic geography
models are built. These assumptions relate to the welfare effects associated with product
variety, the productivity of manufacturing, and finally the costs of transporting goods.
In terms of welfare effects, the NEG models assume that the economy is split into two
sectors, namely agriculture and manufacturing. Manufacturing industry is assumed to pro-
duce a variety of outputs under monopolistically competitive conditions, whereas agricul-
ture is assumed to produce a homogeneous product under conditions ofperfect competition.
All consumers are assumed to have the same tastes, defined by the simple Cobb-Douglas
utility function:
U == MflA1-fl (A.3.3.1)
M= [ t<mY ]
(A.3.3.2)
where m i represents the consumption of each individual variety of manufactured good, and
p is a parameter with a range between zero and one, representing the strength of the con-
sumer preference for product variety among n manufactured goods. If p is close to one, the
different goods are almost perfect substitutes for each other, and the demand curve for each
firm tends towards the horizontal. On the other hand, the closer p is to zero, the greater is
the consumer preference for product variety, and the more price inelastic is the demand
curve for any individual firm. Ifwe set <J 1/(1 - p), and assume that there is a continuum
of n varieties of manufactured goods, we can write the demand function for manufactured
goods as a CES (constant elasticity of substitution) function
]a~l
M =
[
f
tl
m!a-1/aldi (A.3.3.3)
where <J represents the elasticity of substitution between any two varieties of manufactured
good, varying between infinity for perfect substitutes and unity for highly differentiated
products. Fujita et al. (1999, pp. 46-48) show that if the prices of the individual manufac-
tured goods m i are defined as Pi' the general cost ofliving index, which defines the minimum
costs of purchasing a single unit of the composite manufactured good M, can be defined as
1
a
em [I P?-aldir::- (A.3.3.4)
Ifthe number of product varieties available increases, the fact that consumers value product
variety means that the cost of attaining any given level of utility falls. Therefore, the cost of
living falls. We can see this because if all manufactured goods are sold at the same price Pm '
equation (A.3.3.4) reduces to
em = Pmnl/(l-a) (A.3.3.5)
In other words, as the number of varieties n of manufactured products increases, the fall in
the cost of living is greater for lower values of the elasticity of substitution <J. Similarly, for a
given elasticity of substitution a, the cost of living is inversely proportional to the number
of product varieties.
In terms of the productivity of firms, it is assumed that agriculture exhibits constant
returns to scale and is a perfectly competitive economy. On the other hand, manufacturing
is assumed to exhibit increasing returns to scale of the form
Lm = a + bXm (A.3.3.6)
where Lm represents the labour employed by each manufacturing firm, and X m represents
the manufactured output of the firm. This simple specification of increasing returns to scale
implies that, for each firm, the labour required to produce any level of output exhibits both
AN [) *~!&..."'~ *'lSw'r~~!}"''i,~ ~'lMV~,,*V> ~'l?& ~!~" *~I~'ilv.Q 1$,#!&...!l,,>.J'
a fixed overhead component, independent of the level of output, and a variable component
directly related to the level ofoutput. The existence of increasing returns to scale, along with
consumers' preference for product variety, means that each firm will produce a single
unique good. In this monopolistically competitive environment, the number of firms there-
fore will be the same as the number of products produced.
From the perspective of the monopolistic producers, the perceived elasticity of demand
for their own product is (5. Therefore the output price mark-up on their marginal cost can be
expressed as
where W m is the wage for manufacturing labour and Pi is the output price. This gives
Pi = ( 0- ~ 1 }w m (A.3.3.8)
With zero profits, we know that price equals marginal cost. Therefore the ratio of (5/((5- I)
acts also as an index of economies of scale, as well as being a parameter of consumer prefer-
ence for variety. Assuming freedom of entry within the monopolistically competitive envi-
ronment leads to zero profits, then the revenue must equal costs. In notation,
where Pi * is the equilibrium output price. Combining (A.3.3.8) with (A.3.3.9) we have
where X m * is the profit-maximizing equilibrium level of output of the firm, and the equilib-
rium labour demand Lm * for the firm is thus
L*m = a + b a(a - ~
~ = aa (A.3.3.11)
If there are LM workers in a region, the number of manufactured goods produced in the
region will therefore be
n--
L
M - M L L
-M (A.3.3.12)
- L~ - a + bX~ aa
~d = ~e-rd (A.3.3.13)
where Pi is the source f.o.b. price per ton of the good, L is the constant rate ofdecay ofthe good
with respect to the distance d, and Pid is the delivered price of the quantity of good actually
delivered at the distance d. This description of distance costs implies that the costs of distance
are a function of the value of the good produced, because the level of decay is proportional to
the source price of the good. At the same time, we see that the greater the haulage distance, the
greater will be the level of decay, and consequently the smaller will be the quantity of goods
actually delivered. Therefore, in order to ensure that a given quantity ofgood is actually deliv-
ered at any particular location, the total quantity of goods purchased at the source location
must increase as the delivery distance increases, in order to offset the process of goods-decay.
In other words, for any given source value of the good, the total level of goods expenditure
increases with haulage distance at a rate proportional to e'td. The iceberg transport costs associ-
ated with delivering a given weight of product at any given distance can therefore be under-
stood to increase exponentially with distance, as described by Figure A.3.3.1.
distance
Figure A.3.3.1 Iceberg transport costs and haulage distance
If we compare Figure A.3.3.1 with the distance-transport cost functions discussed in the
appendices to Chapter 2) we see that the behaviour of iceberg transport costs employed in
NEG models with respect to distance is very different from the types of transport cost func-
tions typically employed in location theory models (McCann 2005; Fingleton and McCann
2007). As we have seen in Chapter 1) the general form of transport cost functions employed
in location theory models are either linear or concave with distance. Concave transport
costs functions reflect the usual empirical observations of distance costs (McCann 2001)
derived from the distance-frequency optimization problems) either on the basis of the
logistics costs argument discussed in Chapter 1 (McCann 1993) or in terms of the relation-
ship between distance and knowledge-acquisition costs (McCann 2007). The theoretical
and empirical implications of these differences in the transport cost functions between
NEG and classical location theory are discussed elsewhere (McCann 2005) Fingleton and
McCann 2007).
The NEG models integrate these three issues) namely product variety competition) econ-
omies of scale) and transport costs within place-specific considerations) within a framework
of labour mobility. Labour moves between locations according to real wages) which) as
usual, are defined as nominal wages deflated by the local cost of living index. The nominal
wages paid to workers are higher the better the access of (i.e. the closer is) a firm to a market)
the greater the local market income) and the lower the level of local product competition.
Conversely) the cost of living increases according to the geographic distance) and the lack of
local product competition. Large cities with a wide range of manufacturing activities pro-
ducing a high variety of products will be relatively inexpensive to live in) in real terms)
because the high variety of goods locally available will allow any given level of utility to be
achieved at lower real cost. At the same time) these areas will also produce goods at relatively
low cost because of the intense local competition) thereby allowing large market areas to be
captured. However) the point about the exponential form of the iceberg transport cost func-
tion is that the market area tends towards being finite) subject to the source prices of the
goods. Therefore some small cities) or cities which are geographically peripheral with low
product competition and high source prices) will still be able to capture small local market
areas. The role ofdistance-transport costs is therefore to act as a counterbalance to the effect
oflocalized increasing returns to scale in the major cities. Within this integrated framework)
the NEG models show how cities can naturally grow and decline as national and interna-
tional market areas expand. In particular) these models suggest how Christaller-type urban
hierarchies) approximating to the rank-size rule) can be a natural response to economic
development over time.
Q= !(p+tL) (A.3.4.1)
where p is the source price of the good, t is the transport rate per kilometre, and L is
the haulage distance in the Loschian framework. The total market sales of the firm are
therefore given by the sum of all the individual demands at each location. This can be
written as
Q f
= !(p + tL)Ld
o
(A.3.4.2)
Quantity demanded
distanced distanced
Equation (A.3.4.3) defines the total sales of the firm as being the volume of the cone in
Figures 3.4 and A.3.4.1.
The spatial structure of the
urban economy
4.1 Introduction
In Chapters 1-3 we discussed the reasons for the variations in the spatial patterns of indus-
trial investment and activity. As we have seen, firms will locate in different areas for different
reasons, and where this behaviour leads to the co-location of firms we observe spatial con-
centrations of investment. In some cases, the various advantages which are sometimes asso-
ciated with spatial concentrations of such activity give rise to the growth of both cities and
also hierarchical systems of cities. As we know, individual cities can grow to be very large,
and in some cases as large as some individual small countries. There will, however, be a
range of people living within such spatial concentrations, and also there will be a variety of
activities taking place within the city. This consequently brings us to the question of how
such people and activities are distributed within the individual urban economy.
In discussing how people and activities are distributed within the urban economy we
focus on the question of urban land use. In other words, we try to explain why certain
groups of people, or certain industrial activities, occupy land at particular locations within
the city economy. Observation of the behaviour of urban economies suggests that there are
two key features common to all urban areas. These features are that, in general, land prices
tend to fall with increasing distance from the city centre at a diminishing rate, and that the
average land area occupied by each household or business activity tends to increase with
increasing distance from the city centre. Given that there is a market for land in which land
is allocated according to users, in order to understand the allocation of land within the city
and the relationship between location and land prices, we must therefore ask how much
people or firms are willing to pay in order to occupy land at any particular location. In sec-
tions 4.2 to 4.4 of this chapter we will construct a set of models, namely a von Thunen model
and a bid-rent model, which are most commonly used to explain such phenomena. In the
subsequent sections, these models will be contrasted with alternative explanations of the
structure of urban land use and land prices.
We begin by constructing a one-dimensional model of the relationship between location
and land rent, based on the analysis of von Thunen (1826). The von Thunen model is the
simplest model describing the relationship between location and land use, and will act as the
building block upon which our subsequent models of urban land use and land prices are
developed. The simple von Thunen model approach allows us to understand how much land
is employed in productive activities, and how land can be allocated between competing uses.
URBAN
This fixed coefficients von Thunen model will then be extended to a more general variable
coefficients model, known as a <bid-rent' model. The bid-rent model, which allows for sub-
stitution behaviour between land and other production or consumption factors, is the
orthodox model of the urban economy. We will see that applying conventional production
and consumption theory to the bid-rent model provides us with a range of conclusions
regarding the distribution of urban land, the location of urban activities and people, and the
land prices charged at each location.
In order to construct a von Thunen model, we assume that there is a specific market point
located at M, at which all agricultural goods are traded, and we assume that all land is owned
by absentee landlords. We assume that all farmers producing the same agricultural good
exhibit the same production technology and the same fixed production coefficients. We
assume that land is of identical quality at all locations and also that there is freedom ofentry
into the agricultural market. Therefore any production locations which can be shown to be
profitable will result in the agricultural land at that particular location being used for
production.
For example, let us assume that a farmer growing wheat can produce one ton of wheat
from one hectare of land, by combining one hectare of land with one unit of non-land
inputs. Non-land inputs will be a combination of any of the factor inputs employed except
land, such as human labour, animal labour, or human-produced capital inputs such as agri-
cultural machinery. As long as these factor relationships are fixed, it becomes quite straight-
forward for us to consider how much rent the wheat farmer will be willing to pay for a
hectare of land, depending on its location.
In order to see this, we can assume that the price of a ton of wheat at the market location
Mis $100, and that the transport cost t of bringing wheat to the market is $1 per ton-mile or
per ton-kilometre. As we see from Figure 4.1, if the farmer is located immediately adjacent
to M, the haulage distance d from the production location to the market M will be zero. As
such, the farmer will incur no transport costs, and all $100 sales revenue can be spent on
payments to the land and non-land production factor inputs. If the non-land inputs require
payments of$50, the maximum rent the farmer can pay for a hectare of land immediately
adjacent to M will be $50. At a distance of20 kilometres, the maximum the farmer will be
able to pay for a hectare of land is $30, while at a distance of 50 kilometres, the maximum
the farmer will be able to pay for a hectare of land will be zero. Beyond 50 kilometres, there
will be no wheat produced and sold at M. The reason is that the market price of the wheat
will not cover the costs ofproducing plus transporting the wheat to the market from beyond
this distance. As such, the von Thunen model predicts that there will be a negative land-rent
gradient, in which land prices will fall directly with haulage distance, in order to exactly
compensate for higher distance transport costs. At the same time, the von Thunen model
also predicts that there will be a finite limit to the spatial extent over which wheat will be
produced for sale at the market M, beyond which no production will take place.
This basic argument can now be extended to allow for changes in the price of the good, or
changes in the rewards to the factors. For example, in Figure 4.2, if we imagine that the
Rent/Cost
$100
Cost of fixed
non-land inputs
$SO
$30
M! > ~
~
~ 20km
distance d
~ SOkm ~
Figure 4.1 Von Thunen land-rent gradient
Rent/Cost
$lS0
Cost of fixed
non-land inputs
$100
$80
$SO
M~ : 7- distance d
!< ~i
100km
12okm'
Figure 4.2 The effect of increased market prices on the von Thunen land-rent gradient
market price of wheat increases from $100 to $150 per ton, this now implies that the maxi-
mum the farmer will be willing to pay for a hectare of land immediately adjacent to M is
$100. The intercept of the land-rent gradient therefore moves upwards from $50 to $100. At
a distance of 20 kilometres from M, the farmer will be willing to pay $80 rent for the hectare
ofland, and at 50 kilometres from M the farmer will now be willing to pay $50 rent per hec-
tare. Moreover, the maximum land rent will now equal zero at a distance of 100 kilometres,
rather than at 50 kilometres as was previously the case. As such, the distance limit ofthe land
cultivated for wheat production and sale at M will have increased by 50 kilometres from 50
kilometres to 100 kilometres. Within this limit, the maximum possible rents payable to land
at all locations have increased. An increase in the market price therefore brings forth an
increase in the quantity of land brought under cultivation and a consequent increase in the
quantity of output produced and sold, just as we would expect from basic demand and sup-
ply theory.
Exactly the same result as above will also arise if the required payments for the non-land
inputs fall from $50 to zero, with a fixed market output value of $1 00. In this case, the fall in
the payments to non-land inputs will be exactly compensated for by greater payments to the
land inputs.
The effect of changes in the transport rates is slightly different from changes in the output
market prices or changes in the non-land factor payments. This can be explained with the
help of Figure 4.3. For a market price of $100, and non-land input payments of $50, the
maximum the farmer will be able to pay for land immediately adjacent to M will be $50,
irrespective ofthe transport rate. The reason for this is that the total transport costs incurred
by the farmer at a distance d == 0 from M are always zero. As such, the intercept of the land-
rent gradient will remain at $50, irrespective of the transport rate. If, however, the transport
rate t falls from $1 per ton-kilometre to $0.5 per ton-kilometre, the maximum rent the
Rent/Cost
$100
$50
MI ").." :')
distance d
50km
100km
Figure 4.3 The effect of reduced transport rates on the von Thunen land-rent gradient
farmer will now be able to pay at a distance of20 kilometres from M will be $40. Meanwhile)
at a distance of60 kilometres the farmer will be able to pay a maximum of$20) and the maxi-
mum land rent will now equal zero at a distance of 100 kilometres. Once again) the distance
limit of the land which is brought under cultivation to produce wheat for sale at M has
increased from 50 kilometres to 100 kilometres. At the same time) within this limit) the
maximum possible rents payable to land at all locations have increased) except for the land
which is immediately adjacent to the market. The relationship between rental values and the
quantity of land employed is therefore slightly different between the case of changes in
transport rates and the cases of changes in either the output market prices or the non-land
factor payments. Each of the potential changes in the quantity of land used and the maxi-
mum rents payable) described above) will obviously be reversed for equal and opposite
changes in the respective cost parameters.
In the von Thunen model, we treat land as simply a factor input in the production pro-
cess) just like any other production factor) except for the fact that land payments are viewed
as being residual. This assumption is based on the approach of Ricardo (1821) and means
that rental payments to land are distributed only after all other non-land factors and trans-
port costs have been paid. The maximum rents per hectare generated by the von Thunen
model can therefore be described thus:
In the models above) for simplicity we have assumed that a single hectare of land is
employed in the production of wheat. However) if we relax this assumption and allow for
different quantities of land to be employed) with non-land inputs being employed in the
equivalent fixed proportion levels) a more general description of the von Thunen land rent
payable is
Land rent per unit area x Land area = Output revenue - Non-land payments
- Transportcosts
Land rent per unit area) such as per square metre or per hectare) multiplied by the land area
is simply the residual from the total output revenue after all transport costs and non-land
inputs have been paid. Therefore the land rent per unit area is the residual from the total
output revenue after all transport costs and non-land inputs have been paid) divided by the
land area employed) S.
The slope of the negative land-price gradient with respect to distance is given by the
change in the land rent per unit area. This is given by -tiS. This can be understood as that)
for any small increase change in distance i1d) the increase in total transport costs ti1d must
be compensated for by falls in the rent payable to the total land area employed) S. Therefore)
if i1d is approximately zero) the rent per unit area must fall at a rate of -tiS. A formal proof
of this is given in Appendix 4.1. As well as this we can also derive the distance to the outer
limit of the area under cultivation. In Appendix 4.1.1 we show that) as we have seen in the
above example) this is positively related to the market output price) and negatively related to
both the transport rate and the level of non-land payments.
URBAN AND REGHJNAl ECONOM~C *~JS'%>"JS.<$'~~&..~JI'
$150
$100
Rent gradient for barley
M1
distance d
:< >:
~ 33 km ~
leE :~;
I 40km
;(
- 50km
>
Figure 4.4 Competing land uses in the von Thunen model
are not viewed simply as a residual. Rather, in this broader type of model, known as a <bid-
rent' model, we assume that land and non-land production factors can be treated as mutu-
ally substitutable inputs. This allows us to discuss land use within a mainstream
microeconomic framework.
The bid-rent model, associated primarily with the work of Alonso (1964, 1971), and subse-
quently developed by a series ofauthors such as Mills (1969, 1970), Muth (1969), and Evans
(1973), attempts to cast the von Thunen type of framework in a broader setting, which is
more easily related to other areas of microeconomics. In order to do this, the bid-rent model
adopts largely the same basic approach as the von Thunen model, but includes one major
difference. As we have just indicated, the difference is that whereas in the von Thunen model
the land and non-land production factor relationships are fixed, in the bid-rent model land
and non-land production factors are assumed to be mutually substitutable inputs, irrespective
~",~~~~",,~r~'<;lr~1iw ECONOMIC H'~"~#'h7~i....8k.>.Jl'
of whether the firm produces an agricultural or a manufactured good. In terms of the rela-
tionship between the fixed coefficients von Thunen model and the variable coefficients bid-
rent model, there is something of a parallel here between the relationship of the fixed
coefficients Weber model and the variable coefficients Moses model outlined in Chapter 1.
The variable coefficients bid-rent model is much broader than the fixed coefficients von
Thunen model, and provides a wider range of insights.
In order to understand the bid-rent model, we will once again assume that there is a mar-
ket point located at M at which all goods are traded. However, although land is assumed to
be of identical quality at all locations, we will now also assume that land and non-land pro-
duction factors are mutually substitutable. Under these conditions, for a firm producing a
particular good, we can ask the firm what it would be willing to pay per unit area, such as per
square metre, per acre or per hectare, in order to be located at any particular distance away
from M, while still achieving a certain profit level. Assuming that the transportation of
goods to the market M incurs transport costs, we would expect the rents payable by the firm
to fall with increasing distance. As we saw above, for a fixed transport rate per ton-kilometre,
in the von Thunen model the rent gradient is a negatively sloped straight line. However, in
the case of a bid-rent curve, the rents payable by the firm will fall with distance, but at a
decreasing rate. In other words, as we see in Figure 4.5, the bid-rent gradient describes a
rental slope which is both negative and convex to the origin M.
In order to understand the reasons for this observation, we need to reconsider the ques-
tion offactor substitution. In standard microeconomic production theory, in the case where
a firm employs two production factor inputs, such as capital and labour, a firm will equate
the slope of the budget constraint with the slope of the maximum attainable isoquant. If the
price ofone ofthe production factors falls, thereby making it relatively cheap in comparison
to the other factor, the firm will rearrange its consumption of factors by substituting in
favour of the relatively cheap factor and away from the relatively expensive factor. The firm
will continue to substitute its factors until once again the slope of the budget line is equal to
the slope of the highest attainable isoquant.
In the case of a bid-rent curve we construct the rents payable by the firm which will allow
the firm to produce at the same level of profitability, irrespective of the distance from M.
However, we know from our von Thunen model that as we move further away from M, the
Rent/sq. m
M distance d
Figure 4.5 Bid-rent curve for an individual firm
price of land must fall. Assuming that the price of non-land inputs stays constant irrespec-
tive of distance, this means that the price of land must fall relative to the price of non-land
inputs, as the distance from M increases. Production theory suggests the firm will substitute
in favour ofland and away from non-land inputs, as the firm moves away from M. Moreover,
as the distance from M increases, the firm should continue progressively to substitute in
favour of land. Alternatively, given that the price of land increases as we move towards the
market point M, the firm should progressively substitute away from land and in favour of
non-land inputs, as the firm moves closer to the market M. This means that if the firm con-
sumes the optimum amount of factor inputs for each location, given the particular relative
prices of land and non-land factor inputs at each location, it will consume both different
relative and absolute quantities of land and non-land inputs at each location. Close to the
market the firm will consume small parcels of land and large quantities of non-land inputs,
whereas far away from the market the firm will consume large areas of land and small quan-
tities of non-land inputs. Therefore, as the firm moves away from the market, the non-Iand/
land consumption ratio will fall, whereas as the firm moves closer to the market, the non-
landlland consumption ratio will rise.
As with the von Thunen model, the negative slope of the bid-rent curve with respect to
distance is given by the change in the land rent payable per unit area. The slope of the bid-
rent curve is given by -t/5. Although this initially appears to be the same result as in the von
Thunen model, it is fundamentally different in the sense that in the case of the bid-rent
curve, the land area 5 is not fixed, but rather it increases with increasing distance. If the
transport rate t is constant, the negative slope of the bid-rent curve must become shallower
with distance, because the value of 5 will be increasing. The result of this substitution behav-
iour is that the bid-rent curve for the firm with substitutable factor inputs is convex to the
origin, as we see in Figure 4.5. The reason for this is simply that the slopes of the production
isoquants, along which the factor substitution takes place, are also convex. A proof of this is
given in Appendix 4.2.
A second feature of bid-rent analysis is that the higher the position of the bid-rent curve,
the lower the profitability of the individual firm. In other words, in Figure 4.6, the firm prof-
itability n associated with bid-rent curve BR I , which we can write as n(BR I ), is less than that
associated with bid-rent curve BR 2, which in turn is less than that associated with BR 3• In
Rent/sq. m
BR 1
BR2
BR3
M distance d
Figure 4.6 Bid-rent curves for an individual firm
URBAN
Figure 4.6, therefore, n(BR 1) < n(BR 2) < n(BR 3 ). The reason for this is that, given a firm's
budget constraint, the lower the prices ofthe land consumed for any given sales revenue, the
greater the profitability of the firm. In general, however, we adopt the convention that firms
will pay rents such as to ensure that net utility is zero. This is because our assumptions of
freedom of entry into the land market would suggest that if some sectors are systematically
making profits in excess of other sectors, investment flows will move into the most profita-
ble sectors and away from the less profitable sectors, thereby tending to equate profit rates
across sectors with those of normal or zero profits. The result of this is that the bid-rent
curves of firms and industries will tend to reflect the normal or zero profit conditions.
If there are competing producers, some of whom exhibit fixed coefficients of production
in which factor substitution is not possible, as in the von Thunen model, and others for
whom land and non-land inputs are mutually substitutable according to the bid-rent argu-
ment, land will always be allocated to the flexible producer. We can see this in Figure 4.7 if
we compare two producers producing the same output quantity which sells at the same
price per ton at the market M. We can imagine a point at a distance D from M at which a rent
per square metre RD payable by both firms is just sufficient for both firms to earn zero prof-
its. At this point, if the land and non-land inputs employed by both firms are identical, the
rent curves for each firm will coincide. However, as we move towards the market point M,
the rents payable by the flexible firm will increase at a faster rate than those payable by the
inflexible firm. The reason is that the flexible firm will progressively substitute non-land
inputs for land as it moves closer to the market M, thereby reducing the total quantity of
land consumed while increasing the rent per unit area. If there are sufficient numbers of
competing producers of each type of firm, the flexible firms will occupy all the land around
the market. Reversing the argument, a similar conclusion can be arrived at by assuming that
the rent curves for the two types of firms coincide at the intercept M, with the same produc-
tion coefficients. In this case, as we move away from M, the bid-rent curve of the flexible
firms will be shallower than that of the fixed coefficients firms, thereby once again ensuring
that the flexible firms will be able to pay higher land rents at all locations. The result of this
argument is that where fixed and flexible production techniques are competing for land, in
Rent/sq. m
BR
M D distance d
Figure 4.7 Rents payable for fixed and variable coefficients fi rms
general, the land will be allocated to the flexible production techniques, which allow for the
mutual substitution of land and non-land inputs.
Rent/sq. m
I ! . >< ?- <:::::
z
===== . rA
M.
distance d
I<ds>i
~~
>j
I ~ dm
r dr >1
i urban land
the downtown area between the central business district, M, and the outer fringe of the ser-
vice sector at a distance ds from M. The manufacturing activities dominate the area sur-
rounding the city centre between ds and their outer fringe at dm , and the retailing and
distribution sectors dominate the suburban areas between dm and their outer fringe at a
distance dr from M. The actual urban land-rent gradient with respect to distance will be
given by the envelope of the three bid-rent curves, described by the curve WXYZ. In other
words, the actual rental gradient is given by the rental curve which just includes the highest
rent payable by any of the three sectors at any given location, given their individual bid-rent
curves. As we see, the urban land-rent gradient is convex to the point M, which implies that
rents fall at a slower rate as we move further away from the city centre.
As we see in Figure 4.8, the actual urban land-rent gradient is different from the individ-
ual bid-rent curves. The only hypothetical case in which the land-rent gradient and the bid-
rent curves could be the same would be where there is only one type of production activity,
in which case the bid-rent curves for all individual firms will be identical. In all other cases,
in which there is competition for land between different activities, the land-rent gradient
will be the envelope of the individual bid-rent curves.
In this type of model, the distance to the edge of the city is determined by the point at
which it is profitable to convert agricultural land to urban land use. In other words, the
distance to the edge of the city is determined by the point at which the rents payable by
urban activities are just greater than those payable by the agricultural sector rA. Assuming
the profitability ofagricultural land is given, irrespective ofthe distance to the particular city
centre, the agricultural bid-rent curve will be horizontal, as given in Figure 4.7. However,
even ifthe profitability of the agricultural land is dependent on location from the city centre,
as in the von Th unen framework) we can assume that the agricultural bid-rent curve will be
very shallow relative to the other sectors, whose performance is very much more dependent
on accessibility to the particular city centre. As such) within a competitive environment, the
distance to the edge of the city will still be determined by the point at which it is profitable
to convert agricultural land to urban land use. In Figures 4.9 to 4.15 we assume for analytical
and diagrammatic simplicity that r A is given as zero, and therefore we concentrate only on
the urban rent and urban land area.
Rent/sq. m
M; :
I distance d
~
~
s
Id
~( )
r
r
dm ~i
~ 7·
dd
The actual land allocation results outlined in Figure 4.8 depend both on our assumptions
of the relative preference for accessibility to the centre or edge of the city on the part of the
different types of activities, and also on the particular way we have categorized the different
activities. For example, in Figure 4.9 we can split up the retail and distribution sector in
principle into two distinct groups, namely retail and distribution. In this case, we may
hypothesize that retail activities will exhibit a relative preference for accessibility to central
city locations in order to take advantage of any retail agglomeration effects. On the other
hand, the distribution sector may have a relative preference for accessibility to the edge of
the city for the reasons outlined above. As we see in Figure 4.9, under these assumptions the
urban land still dominates the city centre with its outer fringe at ds' but now the retailing
activities are located immediately adjacent to the service activities, with their outer fringe at
dr. Outside these central areas, the manufacturing industry will tend to dominate the land in
immediate proximity to the retailing sector with an outer fringe at dm , and finally the distri-
bution sector will once again be on the edge of the city with its outer fringe at a distance dd
on the city limit.
In principle, we can take this argument even further. For example, we can split up the retail
sector into two groups, namely the traditional type of small- to medium-sized retail stores in
which shops are relatively specialized in certain product ranges, and the large multi-product
sales outlets which favour large-floorspace sales areas. In this case, we may argue that the
former type of retail outlet will exhibit a higher preference for accessibility to the city centre,
whereas the latter type will exhibit a higher preference for edge-of- city locations, in order to
facilitate market access to the hinterland of the city. In this case, we would have the edge of
the city areas dominated both by large-floorspace shopping malls along with distribution
activities, while central areas would exhibit smaller, more specialized shops. Alternatively, we
could split up the service sector into international business services and personal household
services, or we could split the manufacturing sector into large-scale engineering or small-
scale workshop activities. In each ofthese cases, the location preferences of the disaggregated
sectors will tend to be different from those ofthe aggregate sectors discussed above. Therefore
what we see is that our analytical description of a city in part depends on how we classify the
different types of activities which are competing for land in the urban economy. However,
although a city comprises many activities, there will be certain similarities in the preferences
and behaviour oflarge groups of activities. Therefore we can simplify our analysis by treating
groups of different activities as though they were part of a homogeneous individual group.
The justification for this grouping may depend in part on observation and empirical evi-
dence, and our assumptions may therefore be different for different cities in different coun-
tries. Different cities will exhibit different characteristics according to different preferences
on the part of the firms in the city. There is therefore no ideal type of city structure, although
the city structure exhibited by Figure 4.8 is frequently assumed to represent the simplest
description of the most common type of urban land allocation.
Within economics, the question of the allocation of urban land between residential house-
holds is discussed in more or less the same manner as the allocation of urban land between
firms and activities. The assumptions regarding the nature of land and the land market are
the same for the household as for the firm. In other words, we assume that all land is homo-
geneous, land supply is fixed, all land is owned by absentee landlords, and that land is allo-
cated to the person willing to pay the highest rent. We assume that the object of the rational
individual person is to maximize utility given the choices and constraints facing the person.
For the individual person or household we also assume that the individual person gains util-
ity from the consumption of both land and non-land human-produced inputs, and that
these are mutually substitutable.
In order to construct a bid-rent curve for an individual person with these assumptions,
we must once again assume that there is a geographical point M, represented by the central
business district of a city, which is the point on which all household employment activity is
focused. In other words, we assume that all employment takes place at a single point, and
that workers have to commute from their place of residence to the central business district
in order to acquire work. Under these conditions, we can ask an individual person earning
a given wage at the central business district M what they would be willing to pay per unit
area, such as per square metre or per hectare, in order to be located at any particular distance
away from M, while still achieving a certain utility level. Assuming that the cost of commut-
ing to the central business district M incurs transport costs, we would expect that land rents
would fall with increasing distance, thereby altering the relative prices ofland and non-land
inputs at all locations. The individual person will attempt to consume land and non-land
inputs in exactly the proportions which maximize the individual's utility, given the relative
costs ofthe land and non-land inputs at each location. As with the individual firm above, the
bid-rent curve of the individual person can be shown to be convex to the origin.
Assuming the individual person's cost of commuting is given as t per kilometre, the slope
ofthe bid-rent curve ofthe individual person or household can be shown to be given by -tiS,
in a manner similar to that for the bid-rent curve ofthe individual firm. Once again, the land
area S consumed by the household is not fixed, but rather increases with increasing distance
as individual households substitute in favour of land and away from capital as land prices
fall, thereby reducing the non-land/land consumption ratio. Alternatively, as the household
location moves towards the city centre, the individual will substitute non-land inputs in
favour of land, thereby increasing the non-land/land consumption ratio. Therefore, if the
transport rate t is constant, the negative slope of the bid-rent curve must become shallower
with distance, because the value of S will be increasing with distance. The result of this sub-
stitution behaviour is that the bid-rent curve for the individual person, who gains utility
from the consumption of mutually substitutable land and non-land inputs, is convex to the
origin, as we see in Figure 4.10. The reason for this is simply that the slopes of the indiffer-
ence curves along which the factor substitution takes place are also convex. A proof of this
is given in Appendix 4.2.
When we are discussing the question of the utility gained from the consumption of land
and non-land inputs, and the rents payable by an individual for a unit area of land, one
important point to note is that the higher the position of the bid-rent curve, the lower the
utility of the individual. In other words, in Figure 4.10, the personal or household utility U
associated with bid-rent curve BRl' which we can write as U(BR 1), is less than that associated
with bid-rent curve BR 2, which in turn is less than that associated with BR 3• In Figure 4.9,
therefore, U(BR 1) < U(BR 2) < U(BR 3 ). The reason for this is that, given a person's budget
Rent/sq. m
BR 1
BR2
BR]
M distance d
Figure 4.10 Bid-rent curves for an individual household or person
constraint, depending on their employment income, the lower the price of the land con-
sumed, the greater the utility of the consumer. As such, utility can be understood in these
terms as the residual welfare, net of the payments for the consumption of land inputs.
In general, however, we usually adopt the convention that households will pay rents such
as to ensure that net utility is zero. In other words, all income is spent on land, non-land
inputs, and commuting to the city centre, such that there is no surplus. The reason for this
is that we also assume that there is competition for land between homogeneous individuals
within any given income or social group, as well as between different income or social
groups. This will be sufficient to ensure that households' net utility is zero, and the result of
this is that the bid-rent curves of individuals will tend to reflect the zero net utility
conditions.
/
MI~:;~=:J:---t:....:,~--=::::=:..:::::::::::a..-_=:::=====-===·===--~Q~-~') distance d
~ dL ;
: :
~
I d
m
1
~ dh 71
Figure 4.11 Residential urban land allocation for different income groups
bid-rent curves for all three income groups were the same, this would imply that all the
urban land would be occupied by the high-income group, simply because the high-income
group could outbid both of the other lower-income groups at all locations. Therefore, in
order for all income groups to occupy land in a city at the same time, we must assume that
the slopes of the bid-rent curves for each of the income groups are rather different. In Figure
4.11, which represents the urban land allocation of many cities, particularly in North
America, the land occupied by the low-income group is between M and dv the land occu-
pied by the middle-income group extends from dL to dm , and the land occupied by the high-
income group extends from dm to dh• As before, the urban land-rent gradient is given by the
envelope NOPQ of the individual income group bid-rent curves, and is convex to the city
centre M. The edge of the city is given as the point at which the rent from residential land is
just greater than that from agricultural activities. The absolute area ofland occupied by each
individual household will be different according to their distance from M and also accord-
ing to their income, with higher income groups occupying larger areas of land at allloca-
tions. Moreover, as before, the non-Iandlland consumption ratios will tend to fall with
increasing distance.
The land allocation results given in Figure 4.11 are based on strong assumptions relating
to the behaviour and preferences of the different income groups. Low-income people are
assumed to be constrained in terms of their location possibilities, because their low wage
incomes, and therefore their limited budgets, limit their ability to incur the transport costs
AND MODEtS
associated with anything other than short-distance commuting. The bid-rent curve of the
low-income group is therefore very steep, because the transport costs associated with
increasing commuting distance quickly reduce the money they have available to spend on
land and non-land inputs. We assume that both the middle- and high-income groups earn
sufficiently high wage incomes to allow them to incur significant commuting costs, if they
so choose. However, in order for these two groups to co-exist in a city, the slopes of the bid-
rent curves ofthese two income groups must be different from each other, and also different
from that of the low-income group. As we see in Figure 4.11, the slope of the bid-rent curve
of the high-income group is generally assumed to be shallower than that of the middle-
income group, which in turn is shallower than that of the low-income group. The implica-
tions ofthis assumption are quite important, in that we also assume that as incomes increase,
individuals have an increasing preference for land consumption, which is stronger than any
preference for increased accessibility to the city centre.
In order to see this, we can assume that people who earn high wage incomes have a high
opportunity cost of time, in that the opportunity cost to these people of non-wage activity is
high. As wages increase, the increased opportunity costs of time will increase the desire for
proximity to the work location in order to reduce commuting time, the opportunity cost of
which will have risen. On the other hand, however, as incomes increase we assume that peo-
ple have a greater preference for space. Ifwe observe that higher income groups generally live
in the suburban areas ofa city, within a bid-rent model, this implies that the negative slope of
the bid-rent curve falls as income increases. Therefore, in this bid-rent framework we must
assume that the income elasticity of demand for space is higher than the income elasticity of
the demand for reduced travel time. A proof of this is given in Appendix 4.2.1.
As before, our description of the allocation of urban land in part depends on our descrip-
tion ofthe different income groups. For example, we may be able to split up the high-income
group into two quite distinct categories. For example, there may be a high-income group
comprising relatively older people in senior management positions, who have dependent
children and young families. These individuals may exhibit a high preference for space in
order to accommodate for a family unit with more than one or two individuals. As such,
their bid-rent curve will tend to be very shallow. On the other hand, many high wage earners
will also be young single people working in dynamic city-centre industries such as corporate
finance. Many ofthese younger people will not have dependants, and will live in family units
primarily of only one or two income-earning people. This latter group of high earners may
exhibit a relatively low preference for space, instead preferring good accessibility to the
work location. Their bid-rent curves will thus be very steep. If we split up these two groups,
we see that the urban land allocation will be as described by Figure 4.12. This type of resi-
dential pattern broadly represents the urban land allocation in cities with large international
financial activities, such as London, New York, Paris, and Tokyo.
In this case, as we see in Figure 4.12, the land will be occupied by high-income young
people between the centre of the city at M and the outer fringe of residence of the young
high earners at a distance dy from M. The land which is immediately adjacent between dy
and dL will be occupied by the low-income group, the land occupied by the middle-income
group will extend from dL to dm, and the land occupied by the high-income group will
extend from dm to dh• As we see, the area of land occupied by the low-income group in
Figure 4.12 will have fallen relative to the area ofland occupied by this group in Figure 4.11.
Rent/sq. m
t/
distance d
r
~
dm
dh
1
~i
Figure 4.12 Residential urban land allocation with two different high-income groups
Specifically, the width of this area of land will have fallen from dL to (d L - dy )' This implies
that, for a given bid-rent curve, the low-income people will be occupying a smaller total
area. In other words, if the population of this group remains the same, the average area
occupied by individual households within this group will have fallen and the residential
density will have increased. Given that individual utility is in part a function of the quantity
of land consumed, the utility of the low-income group must therefore be reduced by the
presence of the young high-income earners.
Apart from the way in which we categorize different groups, another possible way in
which our model results may change is that our assumptions regarding the relative prefer-
ences for space and accessibility may not always be justified. For example, in some situations
it may be that the income elasticity of demand for accessibility is generally greater than the
income elasticity of demand for space. In this case, as we see in Figure 4.13, the urban land
allocation will be reversed from that which is given in Figure 4.11, in that high-income earn-
ers will live in the city centre, with middle-income earners in immediately adjacent areas,
and low-income groups located on the edge of the city. In Figure 4.13, the high-income
earners will live between the city centre M and the outer fringe of their residence at a dis-
tance dh from the city centre M. The middle-income earners will live immediately adjacent
to the high-income earners between dh and the outer fringe of their residence at a distance
REG~ONAt &;'~\V~"il<>J'n'&~~... $Y"'iS~$U~C;:;~~J>
Rent/sq. m
M ~
~ ! :> =:>n ==- . J distance d
I d
h
;(
~ dm
~
~ dL ~
Figure 4.13 Residential land allocation with high relative preferences for accessibil ity
dm from M. Finally, the low-income earners will occupy the land at the edge of the city
between dm and the urban fringe at a distance dL from the city centre M. In this case, the city
will tend to be very small in area, relative to the city described by Figure 4.11, and the resi-
dential density will be very high. This is because the outer fringe of the city will be defined
with respect to the limited commuting transport costs payable by the lower-income groups.
Cities which exhibit urban land allocations of this type are cities such as Bangkok and
Manila, in which heavy traffic congestion due to insufficient infrastructure limits the ability
of people to commute over anything other than short distances. The opportunity costs of
travel time become very high for all wage earners, but particularly for higher income groups,
which respond by purchasing land in the city centre.
A
Bid-rent curve for low-income group
distance d
rent payable at each location by individuals, because implicitly they will be purchasing dif-
ferent bundles of environmental goods at different locations.
In order to see this, we can consider the example ofthe pollution generated by city-centre
activities. In Figure 4.14 we can assume for simplicity that the city centre is the major source
of urban environmental pollution, due to the generation of exhaust gases caused by local
traffic, plus the presence of smoke from local factories and fumes from city-centre office
ventilation systems. As we see in Figure 4.14, the low-income groups are constrained to
remain close to the city because of their inability to pay high-distance commuting transport
costs. On the other hand, the middle- and high-ncome households may be willing and able
to pay higher rents over a range of locations in order to acquire land further away from the
centre. The reason for this is that the natural environmental quality of land will increase
with distance from the city centre, as it will suffer less from the harmful effects of pollution.
For the middle- and high-income groups, the bid-rent curves will therefore be upward slop-
ing over a large distance, because these groups will be willing to pay higher rents in order to
avoid the pollution damage to their environment. However, beyond a certain distance the
localized effects of the city-centre pollution will be negligible, and the behaviour of rents
with respect to distance will be as predicted by the simple bid-rent model. The shape of the
rent gradient ABCDE in Figure 4.14, which at first rises with distance and subsequently falls
with distance, can be described as concave with distance between Band D, but convex
between A and B and between D and E.
In reality, however, the relationship between environmental quality and the urban rental
gradient may be much more complex than simply the generally concave rent gradient of
Figure 4.14. The reason is that defining exactly what constitutes (environment' is itselfrather
difficult. Urban environmental amenities may be considered to include leisure and enter-
tainment facilities. If these are predominantly located in the city centre, this will tend to
increase city-centre rents relative to those at more distant locations. Alternatively, increas-
ing distance from the city centre may imply that the level of greenery and foliage increases,
thereby improving the local environmental amenities. This will tend to reduce the negative
AN [) H,~~>'« ~~, ~'llllr'''~ ~'iM~:P?"<-#' ?1ii'1'. ~;~ n''18'il<..Q 1&# ""-&..M.'
Rent/sq. m
A'
E'
M' ." '« >:
distance d
~
derelict
land
slope of the rent gradient with distance. Environment may also be considered from the
point of view of social amenities. For example, low-income areas in city centres may be
associated with certain social problems such as crime. In this case, as we see in Figure 4.15,
it may be that the rent gradient A'B' C'D' E' falls to zero in certain areas between B' and C', as
higher-income groups prefer to pay a rent premium in order to isolate themselves from
lower-income groups. The result of this is that there will be a band of derelict space which
remains largely unoccupied by households. Similarly, in such cases, the poor security impli-
cations oflocating in these areas may imply that firms will not wish to invest. The result will
be an inner-city (no-man's land', which is a phenomenon often observed in urban areas.
Once again we can argue that this result is due to the relationship between the location of
land and the qualitative characteristics of the local environment.
The point about all these observations is that the relationship between the rental gradient
and the nature of the environment is not at all clear-cut. If there are environmental changes
associated with location, the urban land-rent gradient may increase, fall or even change
sign, as the distance from the city centre increases. See Appendix 4.2.2 for a formal
discussion.
The bid-rent model is the dominant model of land price-distance convexity, in terms of its
popularity as an analytical approach. However, in reality the problems associated with iso-
lating bid-rent functions from the effects of environmental variations, the weakness of the
assumption of absentee landlords, and the fact that in most urban areas public transporta-
tion infrastructure allows low-income groups to commute over all the urban area, together
may limit the applicability of the bid-rent approach. Therefore there are also alternative
,...
models of rent-gradient convexity and urban land allocation which do not rely on the bid-
rent assumptions of factor substitution between land and non-land inputs. Box 4.1 dis-
cusses two types of models, each of which ascribes rent-gradient convexity to particular
features of the urban land market not fully incorporated in the bid-rent model.
The models discussed in sections 4.2 to 4.5 are based on the assumptions that the city is
monocentric, and that the fixed supply of land available at each location is supplied to the
highest bidder at that location. Moreover, the fact that land is allocated according to its most
profitable use means that the boundary between the edge of the city and its hinterland
-
reflects the optimal size of the city. However) as with any economic model, the results of the
models depend on the assumptions on which they are constructed. Therefore the real-world
applicability of the models for assisting policy decisions must be considered carefully. There
are several issues which need to be raised at this point in order to qualify some of the results
of the standard models described above. These issues relate to the assumption of monocen-
tricity) the questions of land supply and land ownership) the behaviour of the property
developers) and finally the issue of the optimal size of the city.
4.6.1 Monocentricity
The simple models assume that the city is monocentric. In other words) these models assume
that there is a single dominant spatial reference point) with respect to which all location and
land-price decisions are made. In reality) however) large cities have many sub-centres)
which act as local focal points for business and commercial activity. These sub-centres can
often be viewed as local small-scale agglomerations) and the reasons for the existence of
such sub-centres can be attributed to any of the issues discussed in Chapter 2. In cases such
as this) as we see in Figure 4.18) the local rental prices may increase in the immediate vicinity
ofthese sub-centres) such as C I and C2 at a distance del and de2 from M) thereby complicating
the simple downward-sloping distance-land price gradient described above. However) the
existence of such sub-centres does not pose a major problem for our models. The reason is
that we can consider the overall urban land-rent gradient as simply the envelope of two
types of bid-rent gradients) namely those which are determined with respect to the major
urban centre) and those which are determined primarily with respect to the urban sub-
centre. The former are the bid-rent curves discussed above. The latter are the bid-rent curves
of the firms or households whose activities serve specifically local customers) such as retail)
food) and clothing establishments. In addition) ifthe existence ofsuch sub-centres is associated
Rent/sq. m
i (I (2 distance d
r
~
del 71
de2 ~
Figure 4.18 A multicentric city
URBAN REGHJNAl MODELS
with environmental variations, the issues discussed in section 4.4.2 will also become
pertinent.
However, while the calculation ofland prices and location will become more complex the
greater the number of such urban sub-centres, the arguments outlined above do generally
hold. The reason is that all cities have a dominant central business district with respect to
which all urban land prices are at least partially determined. As we see in Figure 4.18, only
the land prices immediately within the vicinity of the sub-centre will be determined primar-
ily by demand for access to the sub-centre. Land prices at all other locations will be deter-
mined by the general urban bid-rent curves with respect to the city centre.
In the case described here it is clear that the polycentric urban structure described above
still represents a single major conurbation. However, as Urban and Regional Example 4.1
makes clear, there are also cases where groups of cities exhibit similar types of polycentric
characteristics.
Whatever the reasons for the emergence of polycentric city patterns, we can still point to
the likely land-price outcomes. From our theoretical arguments, the combination of rental
and land-price gradients in the polycentric system will generate multiple localized peaks
associated with each urban centre. These will each be surrounded by downward-sloping
land-price gradients, which intersect with the other downward-sloping land-price gradi-
ents associated with the adjacent urban centres within the polycentric system, in a manner
largely akin to a standard urban model with many smaller sub-centres, as depicted in Figure
4.18. The overall level of the rent gradients will be determined by the density and economic
performance of the region as a whole, and the three-dimensional topography of the land-
price 'map' of the region will look something like the three-dimensional topography of a
mountain range.
A final issue relates to the fact that in the monocentric city model of residential behaviour
discussed in section 4.4 we assume that households commute to the city centre for work,
whereas in the monocentric city model of firm urban location behaviour discussed in sec-
tion 4.3 we assume that firms are dispersed across a broad area surrounding the central
-
business district. Clearly a more sophisticated approach requires that households are also
able to find work locally in suburbs. However) if we assume that for suburban residents the
transport costs to local suburban employment locations are approximately zero) then the
trade-off for the household is between choosing employment in an adjacent suburban loca-
tion with zero commuting costs and choosing employment at the CBD with its associated
commuting costs. Importantly) the wages earned by the employee in the local suburban
employment location) and therefore the household residential rents payable by this
employee in the same location) will also reflect the profitability and land prices paid by the
firm at that location) as determined by the bid-rent model of firm location described in sec-
tion 4.3. Therefore) as long as the local commuting costs are approximately zero) the model
conclusions still hold. In contrast) the simple argument would start to break down in cases
where a very high proportion ofemployment is in suburban areas with only limited employ-
ment in the CBD. Examples here might be cities such as Los Angeles and Miami) but also
many cities in the Netherlands.
landowners attribute either zero or low attachment value to any marginal parcel of land
because they already have extensive landholdings. The result of this is that any such marginal
parcel of land will be sold at the market price determined by the opportunity cost of land.
On the other hand, according to the law of diminishing marginal utility, small landowners
will tend to ascribe large attachment values to their land because any marginal land sale will
entirely or substantially deplete their current land holding stocks. From the perspective of
the landowners, this argument suggests that the opportunity cost of land sales in a given
area will therefore tend to be inversely related to the size of the individual land holdings in
that area. The result of this is that the more fragmented is the land ownership in any given
area, the higher will tend to be the market price (Dynarski 1986). In an area with heteroge-
neous owners in terms of their landholdings, the price of land may differ between adjacent
properties due to different attachment values on the part of landowners. The result of this is
that land-market development will tend to be piecemeal.
A similar argument may be applied to the question of the length of tenure. Land holders
who have owned a property for a long period may tend to have developed a larger attach-
ment value to the property than land holders who have only owned the property for a short
period. This inflated opportunity cost may inflate land prices above simply the best use
value. Once again, in an area with heterogeneous owners in terms of the length of their cur-
rent ownership tenure, the price of land may differ between adjacent properties due to dif-
ferent attachment values on the part of landowners. As before, the result of this is that
land-market development will tend to be piecemeal.
Piecemeal urban development, particularly on the urban fringe, can also be explained in
terms of information and pecuniary asymmetries between land buyers and sellers, and the
existence of transaction costs. However, the point about all the arguments in this section is
that the simple assumption that land supply is fixed in any location, and that land is simply
supplied to the highest use value, is not always realistic. Other institutional issues sur-
rounding land holding and land tenure must also be explored. These institutional issues
will in addition include questions relating to the behaviour of property development firms.
Such firms engage in land speculation, buying land in advance and often through interme-
diaries, in order to build up land holdings. These firms often make no attempt to supply
land on the basis of the current market price, instead hoping to make greater profits on
future development. The rationale for such behaviour can be understood on the one hand
from the perspective of acquiring a monopoly supply position in a local market. This may
allow the firm to force up the subsequent future sale price in an orthodox monopoly argu-
ment, as described above. On the other hand, in the case of land the determination and
definition of a monopoly position is as much a question of location as it is a question of
land area. Small land holdings in strategically crucial locations can provide monopoly
power. In the case where a seller perceives that a large buyer wishes to buy a large area of
local land in order to undertake a major development, the small seller may attempt to force
up the market price in order to extract as much consumer surplus as possible from the
buyer. However, where the potential buyers are all small, the seller will have little opportu-
nity for such price mark-ups. The result ofall these different types ofinteractions is that the
market prices for land at any location may vary simply because of issues relating to indus-
trial organization. For discussions of the behaviour of the property market see Evans
(1985) and Ball et al. (1998).
-
4.6.3 The optimal size of a city
The arguments outlined in Chapter 2 imply that there is no optimal size for a city, but rather
that there may be an optimal city-size distribution and urban spatial hierarchy. There is,
however, an argument which suggests that the actual size of a city may systematically be
greater than its optimum size (Alonso 1971). This argument is an adaptation of the theory
of the firm and can be understood from Figure 4.19, in which the population of the city is
drawn along the horizontal axis, and the costs and benefits of city are measured along the
vertical axis.
In this argument it is assumed that the costs of city dwelling, which include both pri-
vate and public costs, exhibit economies of scale over a certain range of city size.
For example, such cost efficiencies may include urban agglomeration economies,
plus economies of scale in the provision of public and social infrastructure. Beyond a
certain size, however, it may be that a city begins to experience diseconomies of scale,
associated with increased congestion and pollution. If the benefits of urban dwelling
increase with city size, due to a greater variety of local employment and consumption
opportunities, simple efficiency theory would suggest that the optimal size of the city
should be at the point Q*' where marginal costs equal marginal benefits and the net aver-
age benefits are maximized. However, the argument here is that the city will grow to a
size of Q', at which total costs equal total benefits, and average costs AC equal average
benefits AB. The reason for this is that if city growth is unregulated, the marginal migrant
to the city will perceive the potential net benefits to migration to be positive at all
city sizes below Q', and will ignore his own marginal contribution to the change in
urban dwelling costs. Given that all migrants will ignore their own contribution to
the change in urban costs, the resulting externality problem associated with large num-
bers of marginally erroneous individual calculations will mean that the city grows to Q'
rather than Q *.
Benefi ts/
Costs
AB
4.7 Conclusions
The various institutional and industrial organization issues surrounding the supply of
land discussed in sections 4.6 to 4.6.2 will affect land prices at the very local intra-urban
level. However, along with the environmental issues discussed in section 4.4.2 and indi-
rectly in section 4.6.3, and the possibility of urban sub-centres and polycentric struc-
tures discussed in section 4.6.1, the discussion implies that the actual relationship
between land prices and location will be rather complex over both very small intra-urban
spatial scales and over larger inter-urban and regional scales. As we have seen, over the
large spatial scales of whole metropolitan urban areas or of groups of closely connected
urban areas, the relationship between land prices and location will reflect the theoretical
discussions in sections 4.2 to 4.5.3. In other words, land prices will tend to fall with dis-
tance from the city centre, but at a diminishing rate. From the point of view of this book,
this result is important, because the larger spatial scale of the whole individual metro-
politan area or regions featuring groups of connected urban areas is the spatial scale of
most interest to us. The reason is that in this book we are primarily interested in the
relationship between the urban economy as a whole and the regional and interregional
economy as a whole.
Discussion questions
4.1 ~hat are bid-rent curves? How can an analysis of bid-rent curves help us to
understand the shape of urban land-price gradients?
4.2 How is urban land allocated between different competing income groups?
4.3 How is urban land allocated between competing firms?
4.4 How do urban land allocations differ between different types of cities in different
countries?
4.5 What is the effect of environmental changes on bid-rent curves and urban rental
gradients?
4.6 Apart from bid-rent analysis, what other ways are there of explaining the shape of
urban land-price gradients?
4.7 Under what conditions will the actual size of the individual city differ from its
optimum size?
4.8 Which cities in your country best reflect the theoretical monocentric model of urban
land allocation and which ones appear to differ significantly from the standard model?
Appendix 4.1 The slope of the rent gradient in the von Thunen
model
In the von Thunen framework we can define the total profit accruing to the farmer as
where:
Jr(d) .K S
----- = P -1-- - r--- - td (A.4.1.2)
m m m
If we let K m = Kim and Sm = Slm, whereby Kmand Sm represent the quantities of composite
capital and land inputs required to produce one ton ofoutput, and we also denote Tern = Te( d)lm,
where Tern is the profit per unit of output, we have
In other words, the profit per unit of output is the source price of the good at the point of
production, given as the market price minus the transport costs, minus the total production
factor payments. The maximum rent payable per unit area of land can be calculated by set-
ting Tern = 0 thus:
_ (p - td) - iKm
r- _ - ~~ (A.4.1.6)
Sm
Equation (A.4.1.6) describes the maximum rent payable per unit area of land, as being the
rent payable after all other factors and transport costs have been paid, which ensures that
total profits are zero. The relationship between rents and distance can be found by differen-
tiating (A.4.1.6) with respect to distance d thus:
ar
ad = -~[t
Sm
+ ~d]
ad
(A.4.1.7)
If transport rates are constant, i.e. if (at!ad = 0, then the rent-distance gradient is given by
ar (A.4.1.8)
ad Sm
In order to calculate the distance to the edge of the area of cultivation, we simply set r a,
thus:
0=--------------------------·--·----·····- (A.4.1.9)
Sm
which rearranges to
d= (A.4.1.10)
The conclusions reached in section 4.2 can all be verified from this equation.
By observing that the cross- partial derivative (ad/ ap) of equation (A.4.1.1 0) is posi-
tive' we see that the distance limit of cultivation increases as the market output
price increases. Similarly, the cross-partial derivative (ad/at) is negative. Therefore,
as transport rates increase, the distance limit of cultivation falls. Finally, the cross-
partial derivatives (ad/at) and (ad/ aKm ) are also negative, which implies that as
the payments to non-land production inputs increase, the distance limit of cultivation
falls.
Appendix 4.1.2 Distance to a change of land use in the von Thunen model
In the case of competing land use, in order to calculate the distance at which the land use
changes we simply set the rental prices in each production to be equal. From equation
(A.4.1.6) the land rent in wheat production is given by
(loa 1d) - 50
r = ----------.-------------.-------- (A.4.1.11)
1
(150 - 2.5d) - 50
r = ------------.------------------ (A.4.1.12)
1
Therefore, if the rents are equal between the two uses, we have
(100 1d) - 50
---------,-_._,------~--
(150 - 2.5d) 50 (A.4.1.l3)
1 1
where:
The quantities of both land and non-land inputs consumed are assumed to be functions of
distance d from the central business district.
As we see in Figure 4.10, individual or household utility in a bid-rent model varies accord-
ing to the position of the bid-rent curve, with lower bid-rent curves implying higher utility.
However, this results from the fact that utility is understood here in terms of a residual,
leftover net utility surplus, after all the land and transport payments have been paid. This
notion of net utility, whereby behaviour is determined by the utility surplus which is a resid-
ual, is akin to the understanding of utility in the Salop (1979) model outlined in Chapter 2.
As such, in order to motivate these models we implicitly assume that there is an (outside'
good unrelated to this particular urban location model and which individual or households
in our model are able to consume if they choose. In this case, given that individuals are able
to migrate to other cities or regions, the appropriate outside good is the utility derived from
the difference between the average income and the average living cost of all other location
possibilities in all other cities. The implications of these migration possibilities will be dis-
cussed in Chapter 6. However, for our purposes here, it is sufficient to note that bid-rent
analysis implies that individual urban location behaviour is determined by the surplus or
net utility derived after factor payments are all accounted for.
Given these points, the bid-rent objective ofthe household is to maximize net utility sub-
ject to a given income budget constraint. This can be written as
Max U = U(K(d),S(d))
subject to the budget constraint
Y - iK - rS - T ~ 0 (A.4.2.2)
where:
In a bid-rent model the price of land at each location must fall with distance, because of the
transport-distance costs incurred in commuting to the urban centre. Therefore, in terms of
efficiency analysis, at each location the household must consume land and non-land inputs
in the particular quantities so as to equate the ratio of the marginal utilities from their con-
sumption with their price ratio. The price ratio of non-land and land inputs is given by -ilr,
and the ratio of the marginal utilities of non-land and land inputs is given by MUK/MUs.
However, MUK/ MUs is also equal to the marginal rate of substitution of land and non-land
inputs, given as 11S/11 K, where 11 represents any marginal change in quantity of the inputs
consumed. Therefore we have
I1S (A.4.2.3)
r 11K
For any marginal change in inputs consumed associated with a change in the distance, we
can write
. 11K I1S
1 + r-------- =0 (A.4.2.5)
I1d I1d
However, equation (A.4.2.5) cannot be a complete description of the efficiency conditions a
bid-rent function must fulfil, because the effect of distance changes on the costs of land and
non-land inputs is not symmetrical. For a small change in distance I1d, the price of land r
will fall by a small amount I1r. On the other hand, we assume that the price of non-land
inputs i is independent oflocation. It therefore is necessary to specify each of these relation-
ships as follows:
.aK as
1ad + rad = 0
Therefore
which rearranges to
Jr (t+SJdl
ad ----(S(d) + as)
a;.
(A.4.2.10)
Iftotal transport costs are a function ofdistance, the numerator term will always be positive.
However, even if transport costs are constant with distance, such that if (at/ad) is zero, and
the numerator term reduces simply to t, we still cannot yet determine the sign of (A.4.2.10)
because (as/ar) is negative.
In order to unequivocally establish the sign of (A.4.2.10) is necessary to employ the
Envelope Theorem (Takayama 1993). This theorem is used in situations where we assume
that all variable inputs are employed at their optimal quantities, given the budget con-
straint and the prevailing prices. To employ the Envelope Theorem, it is necessary for us
to distinguish between the variables which are direct and indirect functions of the
distance.
The fall in the price of land will obviously alter the relative prices of the two inputs for
a marginal increase in distance. This will cause a small increase AS in the optimum quan-
tity of land to be purchased, and for a given budget constraint, will consequently also
reduce by a small amount AK the optimum quantity of land to be consumed for any mar-
ginal increase in distance. What we see is that for any budget constraint, the quantity of
1 The residual surplus is the utility from consumption minus the costs of consumption, and this is calculated
with respect to other 'outside good' choice alternatives, in a manner akin to Salop (1979). The relevant 'outside
good' in the case of an urban location choice is that of the best residual surplus achievable in a location in another
urban area.
land consumed is an indirect function of the distance, because the quantity of land con-
sumed is a direct function of the price of land, which itself is a direct function of the dis-
tance. At the same time, given the budget constraint, the quantity of non-land inputs
consumed at any distance is an indirect function of the distance, because the quantity of
non-land inputs consumed is a function of the quantity of land consumed at that dis-
tance, which itself is a function of the distance. This means we can rearrange and rewrite
equation (A.3.2.7) as
iK*(S, Y.d) - td
red) == Y (A.4.2.11 )
s*(Y,d,r)
whereby S* is the optimized quantity of land consumed, given the budget constraint, the
distance and the price of land, and K* is the optimized value of non-land composite capital
inputs, given the budget constraint, the distance and the quantity of land employed.
Applying the Envelope Theorem to (A.4.2.11), we assume that the values of the inputs Sand
K are always at their optimized values, S* and K*, for any given distance. This allows us to
differentiate with respect to only those variables which are directly a function of distance.
From (A.4.2.11) we see that the only such variable is the transport cost. Therefore
dr(d)
.C!!..
+ ad
d)
(A.4.2.12)
ad s*
which, if transport rates are constant, and the land is always consumed in optimum quanti-
ties, gives
dr t (A.4.2.13)
aJ - -·5
In other words, the Envelope Theorem tells us that for a marginal change in distance, the
value of the indirect effect of (as/ar) in equation (A.4.2.10) is approximately zero, and
only the direct effect of distance determines the bid-rent slope. The signs of equations
(A.4.2.10) and (A.4.2.13) are unambiguously negative, and the value of the bid-rent slope
is given by -tiS.
Appendix 4.2.1 The relative income elasticities of the demand for land and accessibility
in the bid-rent model
We can assume that the total expenditure on both transport costs and land is a function of
the income of the household. If transport rates are a constant function of distance, from
(A.4.2.13) the equation of the bid-rent curve is given by
dr t (A.4.2.1.1)
ad- - --5
AND AND ME"fHOOS
To observe the effect of income changes on the bid-rent gradient we take the cross-partial
derivatives of (A.4.2.1.1) with respect to income, thus:
J( ~) 1 Jt t (IS
-JY-- = -S(ad) +S{;;Y) (A.4.2.1.2)
If (A.4.2.1.2) is positive, the bid-rent curve becomes shallower for higher income groups, as
described by Figure 4.11. For the bid-rent curve to become shallower, therefore
(f)(~»(~)(~) (A.4.2.1.5)
which rearranges to
Therefore, from (A.4.2.1.6) the slope of the bid-rent curve will become shallower if the
income elasticity of the demand for space is greater than the income elasticity ofdemand for
reduced travel costs. Alternatively, reversing the inequality (A.4.2.1.6) such that the income
elasticity of the demand for space is less than the income elasticity of demand for reduced
travel costs implies that the bid-rent curve becomes steeper with increasing income, as in
Figure 4.13.
If environmental damage is caused by city-centre pollution, such that the quality ofthe envi-
ronment increases with distance away from the city centre, we can write E = fid), whereby
E represents environmental quality, and fid) describes the functional relationship between
environmental quality and distance from the city centre. We assume that environmental
quality is a location-specific public good, and that E is independent of the quantity of land
consumed at a location. We can regard the effect ofimproved environment as increasing the
utility of the household, for any given level of expenditure on land and non-land inputs.
Therefore we can incorporate environmental quality within our utility function in general
terms as
which, once again differentiating with respect to d using the Envelope Theorem, gives
Jr
d!dJ. aE
(t+ ad ad (A.4.2.2.3 )
ad S + S
which gives
~ = ~[~ - (t + ~d JJ (A.4.2.2.4)
If the bracketed term in (A.4.2.2.4) is positive, the bid-rent curve will be upward sloping. In
other words, if the monetary value of the improvement in the environment with respect to
distance is greater than the increase in total transport costs with respect to distance, the bid-
rent curve will be upward sloping. From the point ofview ofcosts, the monetary value of the
improvement in the environment can be understood in terms of the money that would be
required in order to improve the current environment at the particular location to the
required level. On the other hand, if the environmental improvements with respect to dis-
tance are less significant than the transport costs of distance, the slope of the bid-rent curve
will still be negative, although shallower than would be the case with no environmental
variations.
The argument here follows that of DiPasquale and Wheaton (1996). For any land-based
asset held in perpetuity earning an annual rent of R(t), discounted at a rate of i, the present
value of the property is given as
f
00
PV = R(t)e-itdt (A.4.3.1)
o
URBAN AND '0.". "... "'" "'''l<'' K;, """ 1lY ~~%§ ~"'~lI! 2 'L ."'. """ "'# ".. ~.. c." AND
which is an improper integral (Chiang 1984, ch. 13). If the rent payable at each time period
is fixed, Le. R(t) = R, equation (A.4.3.1.1) can be transformed by taking the limit of a proper
integral, thus:
== · f . R
Y
PV f Re- rt dt = lim Y--7= Re- rt = lim Y--7= .~~i. .~. (I e- iy ) = R
i
(A.4.3.2)
o 0
From equation (4.3) we have an expression for the rent payable for a unit size of property
distributed at an even density around the central business district, given as
R(d)=t(D-d)+rD+k (4.3)
Therefore, from (A.4.3.1.2), the present value of this property asset held in perpetuity pay-
ing the same rent as the current rent defined by equation (4.3) is given as
PV = - +k= C¥ tn +( 7 + ~) (A.4.3.3)
The first bracketed term on the right-hand side reflects the current location value of the
property, in terms of the transport cost savings to the edge of the city, and the second brack-
eted term reflects the agricultural land plus construction value of the property, which we
assume is independent of the location. In the case where a city grows in terms of the wage
incomes payable at the city centre, the urban population and the city radius, we know from
equation (4.6) that the growth in rents is greater for locations further away from the central
business district. If we assume that the growth rate of the urban radius (aD/d) in the long
run takes a constant value of h, the first term in the first bracket on the right-hand side of
equation (A.4.3.1.3) can be rewritten as (tD/i - h). This is because as the city radius grows,
the location value of any location interior to the city grows as the distance to the edge of the
city increases. Therefore, assuming that i > h, the continually increasing transport cost sav-
ing from any location to the edge of the city partially compensates for the depreciating effect
of the discounting on the future value of the location. Therefore we have
PV = (-!E- _t~) + (~ + ~)
1-h 1 1 1
(A.4.3.4)
PV = (r~1 +~)
1
+ (t~ _!~) + (~)
1 1 1(1 - h)
(A.4.3.5)
where
tDh) tD tD (A.4.3.6)
( i(i-h) - i - h - i
In other words, as we see from equation (A.4.3.1.5), in a situation of urban growth,
the present value of the property is equal to the discounted value of the property
given its current location relative to the edge of the city, plus the future growth in its
location value.
Given that the present value of a property is its current market price, in order to under-
stand the relationship between the price of the property P and the rent R of the property we
can divide equation (A.4.3.1.5) by equation (4.3), thus:
P
(A.4.3.7)
R teD
From equation (A.4.3.1.9) we see that the price-rent ratio increases as the rent falls. In other
words, the further the location of the property from the central business district, the greater
will be the price-rent ratio. Following the argument of equation (4.6), the reason for this is
that more peripheral locations experience greater rental gains as a city grows, relative to
central locations. On the other hand, if the city spatial growth h is zero, the price-rent ratio
is given by 1/ i, and is therefore independent of location. Where cities do grow, the result of
equation (A.4.3.1.9) is that even if transport costs are linear, and rents fall linearly with dis-
tance, the market price of properties will fall less than linearly with distance. More specifi-
cally, if the price-rent ratio increases with distance, property prices will be convex with
distance.
Appendix 4.3.1 Property asset appreciation and land price-distance convexity: the role
of income growth in a spatially constrained city
In the situation where a city is constrained in its spatial growth either by physical geo-
graphical restrictions or by land-use planning restrictions, the radius of the city can be
viewed as being held constant. In this case, once the city has expanded to occupy all the
available land, the agricultural rent will no longer be a determining factor in the urban
rents. Therefore all rental values must be calculated with respect to the wage income Y
earned at the city centre. Adopting the notation employed in section 4.5.1 and Appendix
4.3.1, and adapting equation (A.4.1.1.6) such that all measurements relate to a unit area
URBAN REGHJNAl &;'I>...~'>,$~":!i%,.J'n'~!!"""" ."''&.,o",<I'~. . ~•• d'
size, i.e. S = 1, we can write an expression for urban property rents in a city of uniform
density as
R(d) = Y - td k (A.4.3.1.1 )
Differentiating with respect to income Y gives (aRlaY) = 1, and therefore aR aYe Dividing
both sides by R, and multiplying aYby Y/Y, we have
~=(~)(f) (A.4.3.1.2)
From equation (A.4.3.2.2) we see that the rate of rental growth is inversely related to the
share of income accounted for by rent R/Y. In other words, as we move away from the city
centre, the rate of rental growth increases.
These different possibilities for rental appreciation will imply different relationships
between property prices and property rents at each location. Following the approach of
Appendix 4.3.1, we can write the present value of a property at any location as
PV = Y td k (A.4.3.1.3)
i- g
where g here represents the constant long-run rate of growth of city-centre wage incomes
aYI Y. The argument here is that, assuming i>g, the growth in incomes partially offsets the
value-depreciating effects on future income of discounting. Equation (A.4.3.2.3) can be
rearranged to give
Given that the property market price P will be given by present value, we can therefore con-
struct a price-rent ratio thus:
which rearranges to
P _ Yi+{i-g)(-td-k) (A.4.3.1.6)
Ii - i(i - g)(Y - td - k)
Equation (A.4.3.2.6) can be rewritten as
~ _ Yi + (i - g)(R - Y) (A.4.3.1.7)
R - i{i - g)R
which can ~e rearranged to give
P
R
= i(i -Yig)R + 1i - Y
iR
(A.4.3.1.8)
Therefore we have
P 1 (A.4.3.I.9)
R i
+ i(i g)R
From equation (A.4.3.2.9) we see that in the case of a city of uniform density which is spa-
tially constrained, but which experiences incomes growth, the price-rent ratio of property
increases for locations with lower rents. In other words, the price-rent ratio of a property
increases with respect to the distance from the city centre, and the price-distance gradient
is therefore convex. On the other hand, as with equation (A.4.3.1.9), if there is no income
growth the price-rent ratio reduces to Iii, and is therefore independent of location.
The types of arguments discussed in Appendices 4.3.1 and 4.3.2 obviously assume that
growth is correctly factored into the property prices. However, real-estate markets often
exhibit major growth-expectation swings which result in what is known as <over-shooting'
and <under-shooting'. These swings also have major effects on the capital liquidity made
available by banks for housing finance, and the degree to which these monocentricity argu-
ments are affected by shifts in expectations depends on the scale and duration of the over-
shooting and under-shooting. Moreover, this is almost certainly not only a minor issue of
misspecification, in that large movements in the property market associated with changing
expectations are very much the drivers of property markets. Changing expectations heavily
influence the availability of real-estate finance, which in turn influences the levels of con-
struction and supply. However, given the time period involved in construction and devel-
opment activities, real-estate markets tend to lag movements in the rest of the economy by
a period of twelve to eighteen months, and also the real-estate asset prices tend to fluctuate
more markedly than other financial markets (Ball et al. 1998).
Following McCann (1995, 2007), we can set up a trip frequency optimization problem, in
which a firm faces the cost minimization problem:
where:
m, n, (), p, f/J, are positive constants, such that f/Jdp is the total distance costs per journey, and
() is the opportunity cost of less than continuous (i.e. f is less than infinite) face-to-face
contact.
The first term in equation (A.4.4.1) reflects the fact that total transport costs per time
period are a function of trip frequency, while the second term indicates that the opportunity
cost of the lost market revenues of a firm may be negatively related to trip frequency. In
other words, as trip frequency increases the firm will increase its market share up to a maxi-
mum when continuous face-to-face contact is maintained.
In a situation such as this, the firm must decide its optimum trip frequency. In order to
calculate this, we differentiate with respect to f and set equal to zero, thus:
ac
Jj = nqJd P1"-1 - mer m l
- o (A.4.4.2)
The second-order condition can be shown to be positive such that this is the expression for
minimum costs (McCann 1995). Rearranging (A.4.4.2) gives
m()
jm+! = nqJd P jn-I (A.4.4.3)
and thus
Therefore the optimum trip frequency per time period P can be written as
F = C::p)d; (A.4.4.5)
Consequently, what we see is that the optimum number of journeys per time period is
inversely related to the distance ofthe firm from the city centre. In order to calculate the rent
payable at each location, assuming that all trips are undertaken at the optimum frequency
f = P, for each particular location, we can rewrite equation (A.4.4.1) thus:
where c = C when f = P. Differentiating with respect to d, setting to zero and applying the
Envelope Theorem (Takayama 1985, pp. 137-41) gives
de
PlfJdp-1F" +drS 0 (A.4.4.7)
ad ad
such that
dr _PqJd P- 1p n
(A.4.4.8)
ad - S
.me) ;~~~,~
pn
(nqJd P
(A.4.4.9)
Equation (A.4.4.10) is always convex in d as long as p is less than or equal to one. In other
words, as long as total transport costs are less than linear (concave) or linear with distance,
even where the land area of the firm or household is fixed, the rent gradient will still be con-
vex with distance. The standard bid-rent result is achieved here even without substitution
between land and non-land inputs. The point is that where trip frequency is itself a decision
variable, as in the case of all transport, distribution, retail, and consumer shopping activi-
ties, plus all activities where the level of face-to- face contact affects the market share, the
rent-gradient convexity is determined by the optimized trip frequency. This general argu-
ment can subsequently be applied to a range of different real-world examples (McCann
1995, 1998, 2007) with various alternative specifications of costs and factor quantities.
Regional specialization,
trade, and multiplier
analysis
5.1 Introduction
In Chapter 1 we discussed the question of the location behaviour of the individual firm. As
we saw, the reasons for the spatial behaviour of firms depend both on the characteristics of
the firm and on the characteristics of the various regions in which the firm could locate.
Understanding industrial location behaviour then allowed us in Chapter 3 to explain the
economic motivation for the growth of cities and industrial clusters. Our analysis focused
on the issues which determine industrial clustering and industrial dispersion, and, in par-
ticular, on the various types of spillovers and links which take place between firms in the
same area. In Chapter 4, the localized growth of an industrial cluster was then used as the
basis of our comparative analysis of urban land prices and urban land distribution.
One of the issues raised by these arguments which has not yet been dealt with is the local
impact of any local industry changes, such as industry expansion or contraction, or alterna-
tively of the local impact of any local microeconomic changes, such as firm relocation, firm
expansion, or firm closure. Following the arguments in Chapters 2 and 3, any firm reloca-
tion, firm expansion, or firm closure within a local economy must have consequences spe-
cific to the rest of local economy, as well as those consequences for the economy in general.
The reason for this is that such changes will alter the demand for locally supplied factor
inputs, and these demand changes will also engender further changes locally along the lines
discussed in Chapter 3. As we will see in this chapter, the transmission of these effects will
be mediated by the inter-firm linkages which exist in the local economy, plus the linkages
which exist between the local firms and the suppliers of local factor inputs. In general we
would expect the strength of these impacts to be broadly associated with the size of the
change involved. In particular, where the individual firm involved is very large, any changes
in the firm size or organization will have potentially major impacts on the local economy.
Thrrtype of argument can also be extended to the aggregate level where firm relocations,
expansions, or contractions take place at the level of a local industry as a whole. Following
the arguments in Chapters 2, 3, and 4, changes in the output and performance of an indi-
vidual local industrial sector will have implications for other sectors in the local economy.
However, from the arguments of Chapter 3, we would expect that the strength of these
effects will depend on the extent to which the region is specialized in the activities of the
sector in question. If a region is highly specialized in a particular industrial sector, the aggre-
gate effect on the local economy of any changes in the performance of this sector would be
expected to be relatively large. On the other hand, where a region is highly diversified, in the
sense that it contains a wide variety of local industrial activities, we would expect that
changes in the performance of an individual local sector would have a relatively smaller
effect on the local economy. The local impacts of these changes can be analysed in terms of
the aggregate effects of the individual microeconomic changes. This allows us to understand
the relationship between a local region and a local industrial sector. The fortunes of an
industrial sector and a local region therefore become interdependent.
The analysis ofthe impacts ofindustrial change on a host economy, through an assessment
of the various linkages between firms and factor inputs, is known as multiplier analysis. This
involves consideration of the regional trade patterns which exist both within and between
regions. However, unlike countries, by definition, regions do not have the facility to collect
continuous data on the level and patterns of their trade flows with other regions. Therefore,
in the absence of regional trade data, it is necessary for us to employ measures which indi-
rectly impute trade patterns to a region. The process of indirectly estimating regional trade
patterns is done on the basis of observations of the regional industrial structure. Under vari-
ous conditions, these observations allow us to impute regional trade patterns. With these
imputed regional trade data we are then able to consider the regional effects of the expansion
or contraction of a local regional industrial sector.
In this chapter we will initially deal with three different approaches to regional multiplier
analysis. These approaches are economic base models, Keynesian regional income multipli-
ers, and input-output analysis. While the models are somewhat different from each other,
they each throw light on different aspects of the nature of the local impacts associated with
industrial changes, and also on different aspects of the process by which such impacts are
transmitted.
Each of these three model techniques is based on the two fundamental assumptions that,
first, local factor prices are fixed, and second, that there are no local factor supply con-
straints. The first assumption implies that the marginal costs associated with a local output
or employment expansion are constant. In other words, marginal costs and average costs
are both equal to each other and fixed. At the same time, the second assumption implies that
local output and employment can expand without facing any local capacity constraints.
Taken together, these two assumptions imply that regional output or employment can
increase indefinitely in a linear manner, in which average costs stay constant. Although
these assumptions may appear unrealistic at first, they are useful for analytical purposes
here in order to indicate the trade-linkage relationships between different industrial sectors
within the same regional economy. The issues raised by changes in the local availability of
factor inputs, or by changes in the local supply prices of factor inputs, will be discussed in
Chapter 6. Moreover, as we will see in Chapter 8, in conditions oflocal unemployment there
are many cases in which these assumptions, and the models constructed on them, can be
defended as a basis for regional policy.
We will begin by discussing the economic base model, the most general of the three
approaches. The economic base model will then be contrasted with the Keynesian regional
income multiplier model. As we will see, the Keynesian regional multiplier is rather
AND ECONOM~C AND
different from the Keynesian multiplier employed in national income models, but can be
made compatible with the regional economic base model. Third, we will introduce the
input-output approach to regional modelling. Although this is analytically the most
sophisticated of the three multiplier techniques, in that it deals with interregional trade in
a more comprehensive manner, this approach also benefits from the motivation and
insights of the other two approaches. One final point we have not yet mentioned is that the
process of indirectly estimating regional trade patterns, on the basis of observations of the
regional industrial structure, is itself problematic. However, various measures of regional
industrial diversity and specialization can be employed in order to facilitate this process.
Therefore we will conclude the chapter by considering the difficulties associated with some
of the techniques which are available for estimating regional trade patterns on the basis of
observations of regional industrial structures. These issues will also be discussed in detail
in Appendices 5.1-5.3.
The economic base model is conceived at the city-region level of aggregate analysis. Rather
than analysing the impacts of industrial changes at a microeconomic level, the economic
base model focuses on the links between aggregate sectors by characterizing a region as
comprising two broad but distinct industrial sector-groupings. These two sector-groupings
are knows as the basic sector, and the non-basic sector. The definition ofthese two groupings
is that the basic sector is the sector whose performance depends primarily on economic
conditions external to the local economy, while the non-basic sector is made up of the sec-
tors whose performance depends primarily on the economic conditions internal to the local
economy. The definition of external and internal dependence here relates to the location of
the markets for the outputs of the sectors. Industries whose markets are national or global
will tend to sell almost all their output outside the local city-region in which the industry is
based. As such, the market demand for the output of this type of industry will be almost
entirely dependent on market demand conditions outside the local economy. These indus-
tries are classified as basic industries, or sometimes as export-base industries. On the other
hand, there are many sectors whose output tends to be accounted for almost entirely by
local consumers. This is very typical in industries such as retailing, hospitality, and leisure,
and activities such as legal services, real estate and consumer banking, education, health,
and equipment maintenance, all ofwhich tend to cater for households and small businesses.
These industries comprise the non-basic sector, and in economic base terminology are also
known as the service sector, although service here relates to the local orientation of demand,
and is not to be confused with the definition of the tertiary sector.
In situations where regions are dominated by particular major industrial sectors, such as
the automobile industries in Turin and Detroit, the aerospace industries in Everett-Seattle
and Toulouse, and the international financial services industries in London and Frankfurt,
the classification of the basic and non-basic sectors is a relatively easy matter, at least in
principle. The basic sector will comprise the dominant exporting industry in the city-region,
and the non-basic service sector will comprise all other sectors in the local area. On the
other hand, as we will see shortly, for many cities or regions the distinction between the
basic and non-basic sector is not so straightforward. Note here that we are defining exports
in terms of selling outside of the city or region in which the industry is located. In other
words, the definition of regional exports therefore includes both sales to customers in other
parts of the same country as well as sales to customers in other countries.
In order to understand the economic base model, let us imagine we are dealing with an
area where we can easily specify which local industries are basic and non-basic. The most
common form of economic base models treats employment as a proxy for the level of out-
put. In this case, the employment structure in the local economy can be defined by:
T B+N (5.1 )
where:
Equation (5.1) simply says that total employment in an area is the sum of the employment
in the basic industry and the employment in the non-basic sectors. In the economic base
approach, we assume that the output of the non-basic sector is determined by the perfor-
mance of the local economy as a whole, whereas the performance of the basic sector is
determined by factors exogenous to the local economy. As such, we can write N == nT, where
n is a coefficient between zero and one representing the sensitivity of employment genera-
tion in the non-basic sector to the total level of employment generated in the region.
Rewriting equation (5.1) gives
T == B + nT (5.2)
which rearranges to
T 1 (5.3)
B 1- n
The ratio TIB is called the economic base multiplier, and indicates the relationship
between employment in the basic sector and employment in the total economy. The
higher the ratio TIB, the greater the economic base multiplier. The economic base multi-
plier allows us to discuss the overall employment impacts associated with a change in the
basic sector thus:
1
ilT == --------- M (5.4)
1-n
Therefore, for any change ilB in the employment levels in the basic sector, total regional
employment will increase by ilT.
Implicit in this argument is the assumption that the total employment of the region is a
function of the employment generated by the basic sector. The strength of this link between
AND REG~ONAl ll:~lkJl'~"ll%.Jii~';;fllllll~ nH"l..Dil1..""~:"~J
total regional employment and basic sector employment is indicated by 1/(1 - n), where n
represents the strength or sensitivity of the linkage between the local economy and the
locally oriented activities. That this is so can be understood in terms of our assumption that
the performance of the local economy as a whole depends in part on the performance of the
basic sector. At the same time, the employment generated by the basic sector will require
inputs to be provided by the non-basic sector, which itself will generate further employ-
ment. The coefficient n can therefore be perceived as an expenditure-linkage parameter,
reflecting the strength of demand by the basic sector for local non-basic inputs. The higher
the value of n, the smaller the value of (1 n) and the greater the economic base multiplier
TIB = 1/(1 - n).
In areas in which local inter-firm linkages are very strong, the demand linkages between
firms located in the same area will tend to be very high. In some of the industrial clustering
situations described in Chapter 2, and in the cases mentioned above, such as Turin, Detroit,
Toulouse, and Frankfurt, the employment growth in a dominant firm or industry associated
with an output expansion will create enormous additional growth possibilities for local sup-
plier firms in other sectors. These growth possibilities arise from the increased provision of
inputs both to the basic firms themselves and also to the increased number of employees of
those firms. In circumstances such as these, the value of n will tend to be high, and, conse-
quently, the value ofthe economic base multiplier will also be high. On the other hand, there
will be some areas in which the relationship between the basic and non-basic industries will
be rather weak. For example, in many areas dominated by agricultural industries, the falling
demand for labour associated with the increasing use of agricultural mechanization may
mean that local increases in agricultural output will have relatively small impacts on local
employment growth in other sectors. In this case, both the value of n and the value of the
economic base multiplier will tend to be very low. The use of economic base models there-
fore tends to be associated primarily with city-regions dominated by urban concentrations
of both population and production.
Obviously, the economic base relationship may be somewhat more complex that the sim-
pIe description given in equation (5.3). The non-basic sector may not behave in an entirely
linear manner in relation to the basic sector, in that a certain level of non-basic activity may
be somewhat independent of the basic sector. In this case, our economic base model may
look something like this:
where No represents the level of non-basic employment activity which is autonomous of the
basic sector. This rearranges to
T = _ No + B (5.6)
I-nI l-nI
~T=_I_M (5.7)
1- nI
The result in equation (5.7) is the same as in equation (5.4). This implies that even if the non-
basic sector is partly autonomous of the basic sector) in the situation where the marginal
growth of the non-basic sector is constant with respect to the basic sector) the value of the
economic base multiplier will be unaltered.
For practical purposes) models of the form described by equations (5.1) to (5.7) are useful
for empirical purposes. The reason is that they are amenable to simple econometric estima-
tion (Weiss and Gooding 1968) of the form
T = a + fJB + e (5.8)
where the estimated values of a andfi give us values for N o /(l - n) and 1/(1 - n) respectively)
in equation (5.6). In the simplest case outlined by equations (5.1) to (5.4) the value of
No /(l - n) will be zero and 1/(1 - n) will be positive) whereas in the case where the non-basic
sector is partly autonomous of the basic sector) the situation will be best described by equa-
tions (5.5) to (5.7).
It is also possible to conceive of situations in which the marginal relationship between the
total employment change in the basic and non-basic sectors is not constant. For example)
some of the agglomeration arguments discussed in Chapter 3 imply that the sensitivity coef-
ficient n) which defines the strength of the linkage between the local basic and non-basic
sectors) may itself be a function of the size of the local basic sector. Under these circum-
stances) it may be possible to describe the linkage between the sectors as something like
n = no + nIB (5.9)
which rearranges to
T= B(~~n~~) (5.11 )
and
As we see in this case) the value of the economic base multiplier will increase with the size of
the total level of employment in the region. The implication ofthis is that regional growth will
become progressively more sensitive to growth in the basic sector as the size ofthe city-region
increases. This scenario imposes much greater problems of empirical estimation than the pre-
vious models. Moreover) there is no reason to suppose it is indicative of generally continuing
behaviour) as the limits to this explosive growth process will also depend on the arguments
concerning the existence of diseconomies of agglomeration) as discussed in Chapter 3.
_~>W9~~",""'~, '::'.'<,,,l'.,i(~'\:l:%§S""&!If'l-< nH'&,.l'II\§~M&~':§ AND
The argument in section 5.2 is based on the assumption that it is a relatively straightforward
matter for us to decide which industrial activities comprise the regional basic sector, and
which industrial activities comprise the regional non-basic sector. In other words, we
assume it is easy for us to determine which local industrial sectors are primarily regional
exporting sectors, and which local sectors are primarily non-exporting sectors. As we men-
tioned above, in some city-regions which are highly specialized in particular activities, such
as Detroit, Turin, and Toulouse, it is quite easy in principle to identify the major local basic
exporting industries simply from observation. However, even in urban regions like these it
may be that there are many other industries which are basic in nature, but which are not
easily identifiable as being so without additional information on the individual industries.
Similarly, many regions appear not to be dominated by any particular single industry or
group of industries. Areas such as these tend to be very highly diversified in terms of the
industrial activities which are represented locally. Therefore determining what the basic and
non-basic sectors are in these regions is not straightforward. Consequently, without prior
knowledge, in the case of many regions, determining whether an industry is basic or non-
basic is an inexact science.
There are three broad approaches to determining which local industrial sectors are the
regional exporting sectors, namely the assumptions methods, the location quotient method,
and the minimum requirements method. In the following sections we will deal with each of
these three approaches in turn.
where:
In terms of economic base analysis, the logic behind the LQ argument is that if a region has
an employment share in any given sector greater than the national average, the region must
be relatively specialized in the production of the output ofthat particular sector. In this case,
LQir> 1. If we assume for simplicity that, for any given industry sector, all regions have the
same linear production functions, and we also assume that all regional household con-
sumption functions are identical, then a location quotient which is greater than one will
imply that the region must be a net exporter ofthe output ofthe particular sector. Conversely,
a location quotient of less than one implies that a region is a net importer of the good in
question. A location quotient of unity implies zero net regional trade flows.
On the basis of this argument, we would expect that the location quotients for the auto-
motive sectors in cities such as Detroit and Turin will be greater than one, as these areas are
relatively specialized in the automotive sector. At the same time, these areas are net export-
ers of automobiles, simply because the local economies of Detroit and Turin are small rela-
tive to the total markets of the local automotive sectors. Therefore, in these particular cases,
the level ofindustrial specialization in these areas is taken as an indirect indicator ofthe level
of their regional exports. Similarly, if the location quotient of a regional sector is less than
the national average, the region must be a net importer of the goods produced by the sector.
As an example, the level of employment in the whisky industry in England is almost zero, as
almost all such UK employment is in Scotland. The location quotient for the whisky indus-
try in all the regions of England will be close to zero, although whisky is consumed in large
quantities in all regions of England. As such, the English regions are all net importers of
whisky. On the other hand, the location quotient for the whisky industry in Scotland will be
very much greater than one, because almost all of the UK output is produced in Scotland;
Scotland is thus a net exporter of whisky.
This general location quotient argument can be applied to cities and regions in order to
build up a picture of the regional importing and exporting patterns. Table 5.1 indicates the
2012 location quotient values for the aggregate industrial groupings within each of the three
major cities of the Netherlands. These cities are Amsterdam, The Hague, and Rotterdam,
and all three cities are similar in terms of size.
>
2:
0
Table 5.1 City location quotient distributions
A Agriculture, forestry and fishery 0.35 0.48 0.68 0.01 0.30 0.06
B-E Energy and manufacturing 0.49 0.34 0.77 0.40 0.27 0.70
FConstruction 0.50 0.63 1.27 0.39 0.48 0.84
GWholesale and retail 0.90 0.72 0.97 0.78 0.56 0.70
H Transport and storage 1.57 0.68 1.56 0.76 0.77 1.74
I Hotels and restaurants 1.34 0.87 0.86 l.51 0.86 0.85
JInformation and communication 1.71 1.64 0.77 1.93 1.58 0.85
K Financial intermediation 2.01 1.06 0.96 2.67 1.16 1.19
L&N Real estate. rental and other commercial 1.28 1.12 1.11 1.12 1.18 1.19
services
M Knowledge-intensive business services 1.47 1.15 1.16 l.60 0.99 1.29
o Public administration 0.81 2.71 1.02 0.99 3.38 1.33
P Education 0.86 0.88 0.89 1.07 0.87 1.04
Q Health and social work 0.78 0.90 0.95 0.88 0.89 0.99
RCulture, recreation and other services 1.03 1.74 0.98 l.25 1.72 l.15
In economic base models, the choice of technique adopted for determining the relative sizes
of the basic and non-basic sectors will depend in part on the data which are available to us.
The assumptions method is the most basic approach and relies on the least detailed sectoral
employment information. The location quotient approach relies only on data for the region
in question and national data, whereas the minimum requirements approach requires data
on all regions. Unfortunately, these three techniques can give vastly different results.
Therefore, because of the analytical problems with the minimum requirements technique
discussed above and also in AppendiX 5.2, the location quotient method is generally the
most commonly used approach to determining the size of the regional basic and non-basic
sectors.
Even if we are able to overcome the data problems discussed and to successfully identify
the regional basic and non-basic sectors with a reasonable level of accuracy, a second issue
which we have not yet discussed is the question ofdetermining the value of the coefficient n
in equation (5.2), where n represents the strength or sensitivity of the linkage between the
AND ECONOM~C n'li"AQI&#~,,",~,J> METHODS
size of the locally oriented sectors and the externally oriented activities within the local
economy. From our discussions in Chapters 1-4, it is clear that these linkages could arise,
first, due to direct expenditure on local factor inputs by the externallyoriented sectors. The
Weber-Moses arguments of Chapter 1 indicate that firm location behaviour may some-
times be associated with strong local input purchasing linkages. This will be the case where
the optimal location is a corner solution at an end point location. Ifthe firms in question are
primarily export-oriented firms with wide market areas, the result will be that a firm with a
wide market area will exhibit strong local input purchasing. This will imply that the eco-
nomic base linkage between the basic and non-basic sectors will be strong. At the same time,
the agglomeration arguments of Chapter 2 suggest that there will be a second potential
source of linkages between the basic and non-basic sectors. Additional linkages may arise
due to information spillovers, whereby industrial clustering may lead to the improved flow
ofinformation between local firms. The implication here is that such improved information
flows will improve the ability of locally oriented firms to respond to the input market needs
of the local export-oriented firms, relative to more distant firms. This suggests that the
locally oriented firms will maximize their share ofthe market for inputs supplied to the local
URBAN l!"\.~~>J!~%;.)'Pll:p.~~ ECONOMiC MODEtS
basic sector by being better able to customize their outputs to best suit the needs ofthe basic
sector. This will be done in part by investing in the most appropriate technology at the most
appropriate time. The argument is therefore that this process of appropriate investing, in
response to the improved local information flows associated with proximity, will itself
strengthen the linkages between the basic and non-basic sectors.
These arguments provide us with two sources of linkages between the basic and non-
basic sectors, namely the direct expenditure on factor inputs and intermediate goods by
basic firms, and the additional investment linkages possibly associated with agglomeration
arguments. Both of these linkages are monetary linkages. However, distinguishing between
these two sources of linkages between the basic and non-basic sectors can be difficult.
Therefore, evaluating the impact of growth in the basic sector on the local economy, while
taking into account both of these effects, can also be difficult. One way of discussing these
different effects is by constructing an expenditure multiplier in which these monetary link-
ages are explicitly distinguished at the local regional level. The easiest way of doing this is to
construct a Keynesian regional multiplier.
A model which is similar in nature to the economic base multiplier, and which can be made
largely compatible with it, is that of the Keynesian regional multiplier. The Keynesian
regional multiplier is adapted from the standard Keynesian national income-expenditure
multiplier model familiar in introductory and intermediate macroeconomic textbooks.
Assuming that marginal and average input costs remain constant, and assuming that there
are no capacity constraints within the economy, the standard Keynesian national income-
expenditure multiplier model can be described as in Figure 5.1. In Figure 5.1, the change in
income ~ Y associated with any change in aggregate demand L1AD can be described as ~ Y ==
k(L1AD), where k is the value of the multiplier. The multiplier therefore is given by the value
of the ratio of the horizontal shift from Yl to Y2, divided by the vertical shift from AD l to
AD2• In Figure 5.1, the successive rounds of expenditure are indicated by the converging
income-expenditure path abc.
For a change in any of the individual components of aggregate demand, we can therefore
multiply the individual change by the value of the multiplier to provide us with the overall
income change, after all the successive rounds ofexpenditure have taken place. This income-
expenditure process may be dynamic in the sense that successive rounds of expenditure
may take place over several time periods. However, as we see in Appendix 5.1, it is quite
straightforward to reconcile the simple static Keynesian income-expenditure model with a
dynamic model incorporating time into the model, such that the conclusions here hold in
either case.
We can now apply this broad logic to the case of a region. However, from the discussions
in the previous sections, there are particular features of the local regional economy which
are somewhat different from those exhibited by the national economy. The result is that the
Keynesian regional multiplier is also somewhat different from the standard national
Keynesian multiplier. In order to see this we can set up a simple set of expressions out of
which we will construct our multiplier model.
Aggregate
demand AD
MD
AD2
AD 1
i1Y
Y1 Y2 Income Y
Yr == Cr + I r + Gr + X r Mr (5.16)
Cr==C+cYr (5.17)
where:
Our second modification to the simple expression above is adapted from a stand-
ard linear import expenditure function, in which the level of regional imports M r is
partly exogenous of regional income Y r and partly a function of regional income. This is
given as
Mr=M+mYr (5.18)
in which
Tr=tYr (5.19)
where t is the average regional tax rate, such that disposable income after tax is given as Y r
(1 - t).
With equations (5.16) to (5.19) we are now able to provide an income-aggregate
demand expression for regional income. Assuming for the moment, as many national
income-expenditure models do, that regional private sector investment I r, government
regional expenditure Gr " and regional exports X r are exogenous, we can substitute the
information in equations (5.17) to (5.19) in equation (5.16) to give
which rearranges to
- -
C + I r + Gr + Xr - M
Yr =
1- (c - m)(l- t)
(5.22)
The regional income Y r is thus given as the sum of the exogenous components of aggregate
demand multiplied by a regional multiplier kr thus:
Yr = kr(C + I r + Gr + X r - M) (5.23)
where
1
(5.24)
k r == 1 (c - m)(1 - t)
The value of the regional multiplier kr is seen to depend crucially on the value of the brack-
eted term (c - m), which represents the difference between the marginal propensity to con-
sume and the marginal propensity to import from outside the region. The term (c - m)
therefore represents the marginal propensity to consume locally produced goods. As the
value of (c m) increases, the value of the regional multiplier increases, and as the value of
(c m) falls, the value of the regional multiplier falls.
If we consider the effect ofan increase in anyone of the components ofaggregate demand
within the multiplier framework, we can write
~(C + + Gr + M)
~Yr == 1- (c m)(l (5.25)
t)
This· suggests that the greater the regional value of the marginal propensity to consume
locally produced goods, the greater the value of the regional multiplier, and the greater the
increase in regional income. This observation fits well with the discussions in Chapters 1
and 2, in that regions with a strong supply of production factors and intermediate inputs
will benefit greatly from any output increases on the part of the individual components of
demand. The reason for this is that more of the income will be maintained within the region
through successive rounds of expenditure between firms and local suppliers. On the other
hand, regions characterized by firms which have very few local suppliers will tend to exhibit
a high propensity to import m. The result will be that the marginal propensity to consume
locally produced goods will tend to be very low, as will the value of the regional multiplier.
The geography of inter-firm linkages therefore largely determines the value of the regional
multiplier.
The regional multiplier expression given in (5.25) is not substantially different from
simple national multiplier models, except for the geographical definition of imports.
However, there is a second major difference between regional and national income multi-
plier frameworks, and this focuses on the question of investment. In standard national
models investment is treated as being exogenous, as it is regarded as being dependent pri-
marily on issues such as inflation, interest rates and expectations. As such, the level of
investment is not regarded as being primarily dependent on the level of national income.
However, within a regional framework the marginal propensity to invest in the local econ-
omy may be a function of local regional income. The reason for this is that local business
confidence, and also the willingness of banks to provide loans to local businesses, may be
dependent on the existing strength of the local economy, irrespective of national inflation
expectations or interest rates (Dow 1982, 1987; Dow and Rodriguez- Fuentes 1997). The
justification for this type of argument comes from the discussions of location and cluster-
ing in Chapters 1 and 2, in which clustering and proximity may improve flows of informa-
tion between firms and between firms and consumers, and also may improve the efficiency
of the local labour market, thereby increasing local income via agglomeration externalities.
URBAN %..&~>',""*'Ill~&'" &.::''il>..>'i.J§~%.AlS'''''~~~ MODELS AND METHODS
In situations such as this we can assume that local investment levels are partly exogenous
in that they are dependent on national economic conditions, and also partly dependent on
the level of local regional income. We can therefore write regional investment income as
I r == ): + iYr (1 - t) (5.26)
where i is the local regional marginal propensity to invest in the local economy.
The third way in which regional income-expenditure flows are different from national
flows is in terms of government expenditure. In national income-expenditure models gov-
ernment expenditure G is assumed to be exogenous of income, being dependent primarily
on political issues. However, in the case of regions this is not so. Government expenditure in
regions is in part dependent on the level of regional income. In particular, it can be argued
that flows ofgovernment expenditure tend to be inversely related to the level oflocal regional
income. For example, low-income areas often suffer from relatively high unemployment. In
such cases, large flows of welfare benefits will tend to be directed into the regional economy.
Similarly, low-income areas often are eligible for public subsidies from regional policy funds
or urban policy schemes. On the other hand, high-income areas will receive relatively fewer
public expenditure flows ofthe types just mentioned, because there will be less need for such
flows. Government expenditure therefore acts as a partial stabilizer, countering changes in
regional income. Therefore, in terms of regional income-expenditure models, the argument
here suggests that public expenditure is in part an inverse function of local regional income.
We can thus rewrite our government regional income-expenditure function as
Gr == G- g Yr (1 - t) (5.27)
and
Yr == kr (C + 7 + G+ X r - 1\1) (5.30)
where
k r = l-[(c-m)+(i-g)](l-t) (5.31)
This expression is essentially the same as equation (5.22), except for the fact that the denom-
inator term in equation (5.29) is broader than that in (5.22). In equation (5.29) the denomi-
nator term in square brackets contains an additional term (i - g), as well as the term (c - m),
which is the marginal propensity to consume locally produced goods, and (1 - t). This addi-
tional term (i g), which we can call the regional marginal (public plus private) propensity
to invest in the local economy, reflects the total private local investment flows associated
with local income levels, net of the public expenditure withdrawals associated with increas-
ing regional income (Sinclair and Sutcliffe 1983; Black 1981). We assume that i > g, such that
regional income growth is positively associated with overall public plus private sector
investment growth. The type of multiplier model given by equation (5.31), which includes
both the local customer-supplier expenditure linkages (c - m) and the local investment
linkages (i - g), is sometimes known as a (super' multiplier (McCombie and ThirlwallI994).
As before, the justification for this type of model comes from the discussions in
Chapter 2, in that business clustering in many cases is associated with increased regional
income due to improved local information flows between customers, suppliers, and factor
inputs. One of the features of areas characterized by such agglomeration economies is that
there may be an associated increase in the entrepreneurial environment. In part this could
be because of a greater general awareness of potential business investment opportunities,
due to geographical proximity between market agents. At the same time, as we have seen,
this could also imply that credit becomes more readily available for investment opportuni-
ties in already-buoyant economies. For all of these reasons we can assume that the mar-
ginal propensity to invest in the local economy will be positively related to the level of local
regional income.
One final point we must consider concerns the regional propensity for government
investment to be withdrawn from the region as regional income increases, as described by
equation (5.27). As a region grows, the absolute levels of government expenditure in the
region will tend to increase as greater levels ofinvestment are required to provide and main-
tain local public infrastructure such as roads, schools, and hospitals. This suggests that the
marginal propensity for government investment is a positive function of regional income.
However, we can consider this effect as primarily a scale effect on public infrastructure
investment in response to a growth in the local population. On the other hand, the effect
described by equation (5.27) can be related to the income growth of either a stable or grow-
ing population, and relates both to the increasing out-transfer of welfare payment income,
and also public infrastructure investment. In principle, equation (5.27) can be modified to
take both of these effects into account thus:
increase regional income disparities rather than reduce them. Given that such a situation can-
not be sustainable in the long run) we can assume that (i g') is negative. Therefore we can
write g= (i - g') where g represents the net marginal propensity ofgovernment expenditure
to be withdrawn from the local economy as regional income increases. As such) our multiplier
model equations (5.27) to (5.31) do not need to be altered.
As was mentioned in section 5.1) although the economic base multiplier model and the
Keynesian multiplier model are rather different conceptually from each other) they can be
made more or less compatible. In order to see this) we can convert our income-expenditure
multiplier expression (5.29) into an export base multiplier model, which is analogous to the
economic base models discussed in section 5.2. To do this we simply separate the export
income component ofaggregate regional demand from the other domestic regional income
components thus:
Y r = 1- (5.33)
- t)
The first term on the right-hand side of(5.33) which is our export base multiplier) reflects the
economic base multiplier described in terms of income flows rather than employment num-
bers. The second term on the right-hand side of (5.33) describes the income-expenditure
multiplier associated with specifically local activities. If we consider a change in regional
exports AXr ) the corresponding change in regional income Yr can be represented as
AX r
dY r = -1--C-(c---m) + (i - g)J(I- t) (5.34)
Immediately we can see that equation (5.34) is the income-expenditure equivalent of our
economic base equation (5.4) which was given as
dT = _I-dB (5.4)
I-n
The increase in export income AXr in equation (5.34) is the equivalent of the employment
increase in the basic sector dB in equation (5.4). Similarly) the basic/non-basic linkage coef-
ficient n in equation (5.4) is the employment equivalent of the income-expenditure expres-
sion [c - m + (i - g)] (1 - t) in equation (5.34). The same argument can be extended to the
relationship between the income-expenditure multiplier equation (5.33) and the economic
base equation (5.6) which was given as
T=~+~ (5.6)
I-n1 I-n1
The second numerator term on the right-hand side of equation (5.33) is given as
(C + I + G- M)
This term represents the exogenous levels ofregional income which are independent ofregional
exports. By comparing (5.33) with (5.6) we can see that the second numerator term on the
right-hand side of equation (5.33) is the income-expenditure equivalent of No' which in the
economic base model represents the level of regional employment autonomous of exports.
These comparisons therefore allow us to see that the employment linkages between the basic
and non-basic sectors are mediated through the local expenditure linkages generated by the
transactions between the two sectors. These expenditure linkages can be of the form of direct
expenditure on locally produced intermediate inputs and local factor supplies, the strength of
which is given by the marginal propensity to consume locally produced goods (c - m).
Additional induced expenditure can also be generated by increases in local investment, which
are represented by the regional (public plus private) propensity to invest (i - g).
The actual levels of exports and imports depend on a variety of geographical, technologi-
cal, and historical factors, as discussed in Chapter 7. A region's current exports and imports
depend in part on the economic history of the region as well as its economic geography,
because today's products and services produced in the region are also a result of the activi-
ties undertaken in the region in the past. In addition, institutional issues are also important,
as these lead to trade barriers. Gravity equation models, similar to those discussed in
Chapter 6, are often used to estimate interregional trade allowing for such trade barriers
(Anderson and van Wincoop 2003, 2004) in situations where the trade data are limited.
However, such models are never a substitute for detailed input-out and multiplier models.
At the aggregate regional level, the impacts of an increase in regional export activity, associ-
ated with a growth of the basic sector in general, can be analysed in a straightforward man-
ner by observation of equation (5.34). Similarly, using the same approach, we can also
observe the regional multiplier effects of a change in any of the regional domestic compo-
nents of demand given in equation (5.30), simply by multiplying the change in the particu-
lar component of aggregate demand by the regional multiplier, as described in equation
(5.31). At the micro level, however, the situation may be rather more complex.
We can imagine a situation where a firm which serves a wide geographical market area
moves into a particular region. If the market area of the firm is much wider than that of the
host region, it becomes immediately obvious that the exports of the region will increase due
to the local presence of the immigrant firm. In the above expressions this implies that LlXr is
positive. However, following the location theory arguments outlined in Chapter I, a new
firm may initially only have very weak links with the local economy in which it has located.
This is because it may take time for a firm to develop strong linkages with the local economy.
The manifestation of this will be that the firm's marginal propensity to consume locally
produced goods may be much lower than the prevailing average regional marginal propen-
sity to consume locally produced goods (Sinclair and Sutcliffe 1978, 1983). In a situation
such as this, equation (5.33) will need to be adjusted to allow for two different expenditure
URBAN AND AND
effects associated with different types of linkages, and this is known as (impact analysis'. The
type of multiplier framework which accounts for this is known as an (impact multiplier' and
the method of construction of these impact multipliers, as described in Box 5.2, is known as
(impact analysis'.
Linkage expenditure patterns can be defined both in terms of geography and in terms of
time. The geography of linkage patterns, as discussed in Chapters 1 and 2, will determine
the value ofthe regional expenditure coefficients. However, this information will not tell us
anything about the speed of such effects. In order to understand the temporal aspects of
regional impact multipliers, all the models described by equations (5.16) to (5.37) can be
made more sophisticated by splitting up the various rounds of the income-expenditure
process into discrete time periods. This allows us to integrate export base models
with accelerator type models (Hartman and Seckler 1967). There are, however, analytical
weaknesses inherent in all these types of Keynesian income-expenditure multiplier
models, which we must consider. The problems concern the relationship between the
values of the parameters contained in the coefficients e l and e2 , as a structural change of
this sort takes place.
If a firm moves into an area, we know from our Weber analysis in Chapter 1 that the sup-
ply linkages ofthe firm are likely to be very different from those of the host region as a whole.
In particular, the new firm's supply linkages will tend to be less localized than those of the
firms or industries which have been located in the host region for a significant time period.
Over time, the firm's local supply linkages, defined by eI' may converge towards the regional
average values, defined by e2, as the firm seeks to employ more local suppliers. On the other
hand, this may not be the case, either if firms are unable to find suitable local suppliers, or if
corporate organizational concerns militate against the employment of local suppliers,
instead requiring the firm to employ corporation-wide suppliers from other locations. If
either of these latter situations occurs, the firm's regional linkage expenditure patterns,
reflected in its marginal expenditure coefficients, may remain quite different from the
regional linkage expenditure values. Over time, therefore, the immigration of the new firm
may itself change the regional linkage values, and consequently the regional marginal
expenditure values, averaged across all sectors. If the immigrant firm is very large relative to
the size of the local economy, such as in the case of a newly opened automobile production
plant, the regional expenditure coefficients may change very quickly. In some cases these
changes may occur before all of the successive rounds of expenditure have had time to take
place, in which case the value of the regional multiplier will itself have been changed by the
immigration of the new firm.
The immigration ofdifferent types offirms into a region will therefore change the regional
multiplier in different ways according to the particular linkage patterns of the firms in ques-
tion. The reason for this is that firms ofdifferent types will exhibit different demand require-
ments for intermediate inputs and production factors. The demand requirements across the
various sectors within the region will differ according to the expenditure patterns of the
particular firm or sector we are observing. To address these problems coherently, it is there-
fore necessary for us to isolate each of the individual expenditure linkages between each of
the sectors within the region, and between the region sectors and all other regions. Only
then will we be able to identify and accurately estimate, the regional multiplier impact ofany
particular regional structural change. In order to do this, we must undertake what is known
as regional input-output analysis.
The basic principle behind regional input-output analysis is to identify and disaggregate all
the absolute flows of expenditure between different industries, between consumers and
industries, and between industries and factor supplies, in order to reveal the underlying
trading structure of the regional economy. These individual expenditure linkages are then
defined in proportionate terms, and the aggregate pattern of these proportionate relation-
ships is used to construct detailed regional multipliers. It therefore becomes possible to
URBAN l!"i\v.~Mfl!vrl!·~r~:Iw ECONOMH: ~W3'\l..."'I&A"~",IIl...JP
Industry X 70 30 100
Industry Y 20 80 100 200
IndustryZ 20 80 200 300
Regional factor inputs 40 110 140 290
Regional imports 20 10 10 30 70
Total inputs 100 200 300 360 960
identify how the regional economy in general, and how each of the individual regional sec-
tors, is affected by a change in the level of demand of one or more of the individual regional
industrial sectors. As such, regional input-output analysis can avoid many of the analytical
problems described in section 5.6 as long as sufficient updated input-output data are avail-
able, which is a point we will return to at the end of this section.
The first stage of a regional input-output analysis is to construct a regional trade flow
table. In order to understand the logic of this we will use a numerical example employed by
Thorne (1969), in which region R comprises three industrial sectors X, Y, and Z. Table 5.2
indicates the absolute values of the current output and expenditure on inputs by each of the
three sectors X, Y, and Z, plus the expenditure by final consumers in the region on finished
goods. As we see in Table 5.2, at the current levels of regional output, local industry X
spends $20 million on inputs from each of the other local industrial sectors Y and Z,
$20 million on regional imports, and $40 million on local regional inputs. The respective
expenditure flows of each of the sectors Y and Z can also be seen by reading down the
column values in Table 5.2. Similarly, local household consumers are seen to purchase
$30 million worth of finished goods and services from local industry X, $100 million from
local industry Y, $200 million from local industry Z, plus £30 million worth of finished
goods and services produced outside the region. At the same time, we can interpret the pat-
tern of sales by each sector across each of the other sectors and household consumers by
reading across the columns from left to right. For example, local industry Z sells $20 million
worth of outputs to local industry X, $80 million to local industry Y, and $200 million to
local consumers. Meanwhile, local factor inputs supply factor services worth £40 million,
$110 million, and $140 million to each ofthe local industries X, Y, and Z, respectively. These
values represent the combined total wages, rent, and interest earned by the local labour,
land, and capital suppliers employed in the activities ofeach ofthe local industries X, Y, and
Z. The final column on the right-hand side provides the total output sales for each sector.
The second stage in the input-output analysis involves defining the disaggregated expend-
iture flows for each of the three industries and the final consumers as described in Table 5.2
in terms of their proportionate size. In order to do this we simply divide the expenditure
value in each of the cells of Table 5.2 by the respective column total at the foot of each of the
columns. For example, in Table 5.2 we see that local industry Y currently purchases
$80 million worth of inputs from local industry Z. By dividing 80 by the respective column
Table 5.3 Regional expenditure coefficients
total of200, we see that the expenditure by industry Y on inputs from industry Z accounts for
0.4, i.e. 40 per cent, of the total input expenditure of industry Y. The value of 0.4 is defined as
the regional expenditure coefficient of local industry y's purchases of the output of local
industry Z. By repeating this procedure for each ofthe cells in Table 5.2 we arrive at Table 5.3,
which gives each of the individual regional expenditure coefficients for purchases by indus-
tries X, Y, and Z, as well as final household purchases. Table 5.3 can be considered a matrix of
regional expenditure coefficients.
By means of Table 5.3 we can now consider the impact on the regional economy of an
increase in the output demand of one of the local industrial sectors. For example, we can
consider the situation where final consumers increase their consumption of the goods pro-
duced by local industry Z to $1000 million. In order to supply $1000 million worth of goods
and services to final consumers, from Table 5.3 we know that local industry Z must also
purchase 23 per cent of its inputs from local industry X (i.e. 0.23 x 1000 == $230 million),
27 per cent of its inputs from local industry Y(i.e. 0.27 x 1000 $270 million), 47 per cent
of its inputs (i.e. 0.47 x 1000 $470 million) from local factor suppliers, and 3 per cent of
its inputs (i.e. 0.03 x 230 == $6.9 million) as regional imports. Industry X in turn will
then spend 20 per cent of its income on additional inputs from local industry Y (i.e.
0.2 x 230 == $46 million), 20 per cent of its income on additional inputs from local industry
Z (i.e. 0.2 x 230 $46 million), 40 per cent of its income on additional local factor supplies
(i.e. 0.4 x 230 $92 million) and 20 per cent (i.e. 0.2 x 230 == $46 million) on regional
imports. Meanwhile, industry Y will spend 40 per cent of its income on additional inputs
from local industry Z (i.e. 0.4 x 270 == $108 million» 55 per cent of its income on additional
local factor supplies (i.e. 0.55 x 270 == $148.5 million), and 3 per cent of its income (i.e.
0.03 x 270 == $8.1 million) on regional imports. Each of the three sectors in turn will con-
tinue to purchase additional inputs through the successive rounds of expenditure along the
lines described in Appendix 5.1. Assuming that the expenditure coefficients remain con-
stant through the successive rounds of expenditure, it is possible to calculate the total value
of regional output and expenditure associated with local industry Z producing $1000 mil-
lion worth of goods and services for local consumers.
Following Thorne (1969), the absolute output and expenditure values are given in
Table 5.4. A final consumer demand of $1000 million for the output of local industry Z
results in a total regional output of $2868 million, because of the successive rounds of
REG ~ONA l "'.. "".. ~'..$'~~.",.., "'~'''' ~ ~ .,.~ "'.$ ...... ",•. ».. ••"
Table 5.4 Regional output and expenditure flows for consumer purchases of $1 000 million from industry Z
expenditure between each of the local industrial sectors. The value of the regional multiplier
in this particular case is therefore total regional output divided by the particular output
demand in question. In other words, the regional multiplier is given by 2868/1000 == 2.87.
From Table 5.3 we see that each industry has a different pattern of regional expenditure,
as reflected by the different values of the regional expenditure coefficients. However, once
the initial demand stimulus has been transmitted to the supplying sectors, the pattern of
purchasing through the successive rounds of expenditure remains constant. Yet what we
immediately see from this example is that for any given level of final output, the actual value
of the regional multiplier depends on the source of the initial output demand stimulus. For
a fixed pattern of regional sectoral purchase coefficients, given in Table 5.3, the absolute
value of the regional multiplier will depend crucially on the pattern of the first-round local
purchases by the sector in which the initial demand stimulus arose. In the above case, if the
output demand originated in local sector X, the value of the overall regional multiplier
would have been much less, because sector X has a much higher propensity to import than
either sector Y or sector Z. Therefore the first-round expenditure injection into the regional
industrial system would have been much lower than in the above case.
On first inspection, the individual regional expenditure coefficients in each of the cells of
Table 5.3 may appear to be the input-output equivalent of n in the economic base model
(equation 5.3), and (c - m) in the Keynesian multiplier model (equation 5.24). However, in
the input-output case the individual regional expenditure coefficients of Table 5.3 refer
only to a single expenditure linkage between one individual purchasing sector and one indi-
vidual supply sector. On the other hand, the values of n in the economic base model and
(c - m) in the Keynesian multiplier model represent aggregate average regional expenditure
values across all regional purchasing linkages. As we see in Appendix 5.3, the actual input-
output equivalent of the economic base or Keynesian multiplier models is determined by
calculating the inverse of the matrix of all the coefficients given in Table 5.3. It is this tech-
nique that allows us to calculate the values given in Table 5.4.
where updated national input-output tables do exist, regional trade estimates are produced
by adjusting the national input-output coefficients. The details of the techniques employed
are discussed in Appendix 5.2. In situations where it is possible to test the accuracy of
regional input-output tables constructed on the basis of these indirect approaches with
actual regional trade survey data, the results suggest that there is often a high degree oferror.
In particular, as we see in Appendix 5.2.2, regional trade estimates arrived at on the basis of
location quotients generally tend to overestimate local multiplier values.
A final analytical issue relating to input-output analysis, but which also relates to the
economic base and Keynesian regional multiplier models, concerns the realism of the
assumption of the stability of the input-output coefficients through the successive rounds
of expenditure. All three models are based on the assumption that the regional expenditure
coefficients remain stable throughout the successive rounds of local expenditure. In other
words, we assume that the marginal and average costs of production remain equal and con-
stant as output expands. Furthermore, we also assume that all production functions are
linear, in the sense that all input expenditure coefficients remain fixed, and that the relation-
ship between inputs and outputs exhibits constant returns to scale. (Such production func-
tions are referred to as linear input-output functions, or alternatively as Leontief functions,
after the founder of input-output analysis, Wassily Leontief (1953). Taken together, these
assumptions amount to saying that the prices of any factor inputs which are specifically
local, such as labour and land prices, are invariant with the level of output.
In situations in which there are unemployed resources, and in particular in situations in
which there is a pool of unemployed local labour at the prevailing regional wage rate, the
input-output assumptions concerning constant marginal and average costs would appear
to be good approximations of the real world. In these situations, the regional input-output
model can be made compatible with, and provide extensions to, the Keynesian regional
multiplier model (Hewings et ale 1999). Moreover, demographic changes in the composi-
tion of the local labour supply can also be made compatible with such assumptions (Batey
and Madden 1981; Batey et al. 2001). On the other hand, where supply constraints become
evident, the prices of the various local inputs will begin to rise. This may lead to some factor
substitution effects, thereby contravening our multiplier assumptions. The extent to which
the accuracy ofour multiplier model is affected will therefore depend on the extent to which
our linearity assumptions are violated across all regional sectors.
5.8 Conclusions
The question of regional industrial specialization, which was first introduced in Chapter 2,
has implications for the pattern of trade which a region engages in. In particular, spatial
industrial clustering implies that a region is relatively specialized in the production of the
outputs of the locally clustered sector. The result is that the industrial sectors which are
clustered in certain regions tend to provide the major exports for the regions in which they
are located. Conversely, the majority of a region's imports will tend to come from the sec-
tors which have a relatively low level of employment in the region. As we have seen in this
chapter, the regional trade patterns which result from these patterns of regional industrial
specialization have implications for the relationship between the externally oriented
regional export sectors and the locally oriented regional industrial sectors. These relation-
ships are manifested in terms of employment and expenditure linkages, the structure of
which can be understood in terms of regional multipliers. The basic observation to come out
of these multiplier models is that a demand increase in one regional export sector will have
even greater impacts for overall regional demand.
We have discussed the three broad approaches to identifying regional multipliers, namely
the economic base model, the Keynesian regional multiplier model, and regional input-
output analysis. These three approaches to regional multiplier analysis allow us to analyse
the contribution of regional exports to the successive rounds of income and expenditure
which remain in the local economy. Of the three techniques, the economic base model is
seen to be analytically the simplest approach, in which estimates of the basic and non-basic
sectors are most commonly made on the basis of regional industrial employment distribu-
tions. The Keynesian regional multiplier model adopts a different perspective in that it
focuses on the issues which determine the nature of the monetary flows involved in each of
the successive rounds of income and expenditure. As we have seen, however, the economic
base model and the Keynesian regional multiplier model can be made more or less compat-
ible, at least in analytical terms, in that local sectoral employment linkages will be mediated
through local money expenditure linkages. Meanwhile, the input-output approach is much
more comprehensive than the other two approaches, and avoids many of the analytical
problems associated with the two techniques.
On the other hand, all three approaches suffer from the same two basic problems. The
first is the issue of data availability, and the limitations of the various techniques for over-
coming data availability problems. The second problem is the question of the assumption of
constant regional average and marginal costs with excess regional capacity. As we have seen,
in situations where there is excess labour supply, such assumptions can be justified. In par-
ticular, these approaches can be very useful in analysing the impacts of the policies adopted
for rejuvenating under-performing regions. These issues will be dealt with in detail as part
of our regional policy discussions in Chapter 7. At the same time, as we will see in Chapter 6,
these models can also throw some light on the downward adjustment mechanisms of
regions which experience adverse demand shocks. On the other hand, where labour supply
prices are changing, such as in situations in which there are local labour supply constraints,
or alternatively where interregional labour migration is a major feature, regional multiplier
models may exhibit rather greater limitations. These various issues surrounding the regional
demand and supply of local labour are the central topics discussed in Chapter 6.
Discussion questions
5.1 What types of urban and regional economies are most suited to an economic-base
type of analysis?
5.2 What empirical measures can we adopt for estimating basic and non-basic industries?
5.3 Within a Keynesian regional multiplier framework, examine the role played by the
marginal propensity to consume locally produced goods in determining regional
income.
REGiONAL ~;1!<,,~'J~~vn'l1~""" n>~""QM'~M~J
Y=A (A.S.l.I)
and
AE = C+ I (A.S.l.2)
where:
Y income
AE aggregate expenditure
C consumption
I exogenous investment
C = C + cY (A.S.l.3)
where C represents the exogenous component of income, from (A.S.l.3) we can therefore
rewrite (A.S.I.2) as
AE = C + cY + I = Y (A.S.l.4)
which rearranges to
y=C+I=AE (A.S.l.S)
l-c l-c
where the numerator term represent the exogenous components of income. Now, if
for example exogenous investment increases by AI, the multiplier effect can be described as
L\Y = k(L\AE) = k(AI), where the value of the multiplier k is given by
1
k=l_c (A.S.l.6)
Table A.5.l Successive time-period rounds of expenditure
This simple static version of the Keynesian income-expenditure multiplier can also be
defined in simple dynamic terms by assuming that current consumption is a function of the
income in the previous time period thus:
C t == C + eYt - l (A.S.l.?)
If we also assume that the inter-temporal equilibrium is defined by Yt Yt+ 1, we can write
Ct C+eYt (A.S.l.8)
Following Levacic and Rebmann (1982 p. 23), the successive rounds of the income-expend-
iture process can now be described with the help of Table A.S.I.
The total change in income over each of the successive time periods and rounds of
expenditure can thus be given as
As n ~ 00, en ~ o. Therefore, as above, we can write ~ Y == k(AAE) == k (M), where the value of
the multiplier k is given by
I
k==l_e (A.S.l.6)
The successive rounds of the income-expenditure process therefore take place though time,
but the overall effect on income of a change in any of the components of demand can be
described as in Figure 5.1.
Appendix 5.2.1 Estimating regional trade using location quotients where an updated
national input-output table is not available
In cases where updated national input-output tables are not available, the absence ofregional
trade data means that the construction of sub-national regional input-output tables is typi-
cally based on indirect methods of estimating regional trade flows. Once the trade
flows between sectors have been calculated, it is quite straightforward to calculate
the regional input-output expenditure coefficients. The most common method is based on
the observation of regional employment patterns. Comparing regional sectoral employment
shares with those of the national economy allows us to construct an index of net regional
sectoral trade, under various assumptions. As we saw in section 5.3.2, this index is known as
a location quotient (LQ) and from equation (5.13) its simplest form can be represented as
EilEr + Er
LQir (A.5.2.1)
Ei,/E n E in En
where Eir and Ein represent the total employment levels accounted for by sector i in region r
and in the nation n, and Er and En represent the total employment levels of the region and
nation, respectively. In general terms, this approach posits that:
(a) the greater is the LQ above unity, the larger will be the regions net sectoral exports;
(b) the greater is the LQ below unity, the larger will be the regions net sectoral imports;
(c) for an LQ of unity, the region is neither a net exporter nor a net importer.
However, four assumptions are required for this index to give an accurate measure of net
regional sectoral trade (Norcliffe 1983). These are that:
(i) per capita sectoral productivity levels are invariant with respect to location;
(ii) per capita consumption levels and patterns are invariant with respect to location;
(iii) the national economy exhibits no net exports or imports for any sector;
(iv) there is no interregional cross-hauling for any sector, such that for any regional
sector which is an exporter, all local consumption of the output of that sector is
accounted for by the local industry.
Where these conditions are met, for a region which is a net exporter of the output of
sector i, the actual relationship between the LQir and the level of regional sectoral export
X ir employment can be written as (Isserman 1977a, 1980)
which once again can be converted to give net regional sectoral import values for our region
as
Given our assumptions (i)-(iv), the principle behind the simple LQ method is that the
total level of regional production in any particular sector can be described as being pro-
portionate to the relative contribution of regional to national sectoral employment. At
the same time, the total regional consumption of the output of any particular sector is
defined as being in proportion to the size of the region. The net regional sectoral trade
flows are assumed to be defined by the differences in these values. If a region is calculated
as having an LQ which is greater than or equal to unity, we assume that, within a regional
input-output framework, the input expenditure coefficient for that particular activity is
assumed to be one. On the other hand, for regional LQ values of less than one, the region
will be assumed to be a net importer of the goods. In these cases, the regional input-out-
put expenditure coefficients are assumed to be proportional to the regional LQ values of
less than one.
In the simple LQ approach the regional consumption of the output of sector i is assumed
to be a function of the regional population expenditure, defined by the total regional level of
employment. However, even allowing for appropriate regional consumption adjustments it
is arguable that, in many sectors, there is no reason why the size of a region per se should
have any bearing on regional consumption levels. This is true in the case of many intra-
industry transactions not involving final household demand. In these cases, regional secto-
ral demand is more likely to be related to the local level of activity of the various industrial
purchasing sectors within the region rather than to population levels as a whole. In order to
account for the interregional spatial variation in intra-industry sectoral demand we can
construct a cross-industry location quotient (CILQ) which is calculated as the ratio of the
LQ of the supplying sector i over that of the purchasing sector j. Substituting the CILQ for
the simple LQ in equations (A.5.2.2) and (A.S.2.3) gives us
and
In the case where the purchasing sector j is defined as the household consumption sector,
then (A.5.2.6) and (A.5.2.7) will coincide with equations (A.5.2.2) and (A.5.2.4).
There are also several other suggested location quotient formulations which combine the
features of the LQ and CILQ models in a variety of ways (Round 1978; Flegg et al. 1995), but
the general principles underlying all the location quotient formulas are the same.
Both the location quotient approaches described here are based on a comparison of the
regional and national employment structures. However, the minimum requirements
approach (Ullman and Dacey 1960) discussed in section 5.3.3 suggests that there is no theo-
retical economic reason to assume that the national economic structure is the most appropri-
ate benchmark against which regional trade predictions can be generated. For regions of
similar sizes, the smallest share ofsectoral employment in any single region within the appro-
priate size band is taken to represent the local sectoral consumption requirement for regions
of that size, and all relative regional sectoral employment shares greater than this represent
regional export employment. The MR method can thus be represented as
where m is the region with the minimum sectoral employment share, and (E im /Em)E r is local
regional consumption of the output of sector i.
Following Isserman (1980) it is possible to compare this approach with that of the LQ by
noting that equation (A.5.2.2) can be rearranged to give
X ir Eir_Pin)Er (A.5.2.9)
( Er En
Similarly, we can compare the MR approach with that of the CILQ by rearranging equation
(A.5.2.6) to give
X ir = (Ei~
Er
- (E J( E )JEr
jr
Er E;n
in (A.5.2.10)
The key difference between the MR approach and the LQ approaches to determining
regional trade patterns is the question of the appropriate benchmark against which
regional sectoral employment patterns are compared in order to arrive at a measure
of regional sectoral consumption. This is reflected in the differing constructions of the
second bracketed term in each model. The MR approach adopts the sectoral employment
structure of similar sizes areas as the benchmark, whereas the LQ approaches both adopt
the sectoral employment structure of the national economy as the appropriate
benchmark.
The debate as to the accuracy of employment-based regional trade estimates is not a new
one. Assumptions (i) and (ii) above are clearly very difficult to sustain, although, using
regional consumption and output indices, it is possible for the LQ method to be adapted to
some extent to take account ofany regional variations in productivity and consumption due
to technical differences in factor allocations, tastes or transfer payments (Isserman 1977b;
Norcliffe 1983). Similarly, where assumption (iii) is not tenable, these types of models can
be somewhat adjusted to take account of national sectoral trade balances which are non-
zero. Where a regional LQ is greater than unity, Isserman (I977b, 1980) suggests that we can
estimate regional export employment as
X ir (P E
r c gr (1 em
. ))l~Jn
P V LQ > 1 (A.S.2.II)
in rEn r
where Pr is the labour productivity ratio between the region r and the nation n, cr is the
equivalent consumption ratio, and ein is the ratio of national net exports to national output
of sector i. This can be converted to estimate actual export values as before in by substituting
Pin for Ein in equation (A.S.2.II). By similar reasoning, where the LQ value is less than unity
we can write the regional import function as
M ==
ir
(c r En
p,.
Ein ) Pr
VLQir < 1 (A.S.2.I2)
in the case where the national economy is not a net importer. Where the national economy
is net importer of the output of sector i, the appropriate adjustment of the LQ model gives
M. ==
Ir
(c En - pgiL)?_i_l1_
r rEin P
+
r
C
r
(g!)M.
En m
(A.S.2.I3)
where Min is the national level of net imports of sector i, and this can be rearranged to give
This specification allows us to take account of imports into a region both from other regions
as well as from other countries. The most difficult remaining problem, however, arises with
assumption (iv), namely that of the absence of cross-hauling (Harris and Liu 1998). In real-
ity, many products move repeatedly forwards and backwards across the same regional
boundaries during the various stages of the production process. Similarly, the monopolistic
competition models (Fujita et al. 1999) discussed in Chapter 2 predict that many of the
products produced by the same industry will be moved in opposite directions between
regions and locations.
The use of the national employment structure as the appropriate benchmark comparison
against which regional consumption indices are developed is the most common employ-
ment-based method used for estimating regional trade flows. The major reason for this is that
detailed and updated input-output data often exist at the national level. Therefore national
intersectoral expenditure coefficients are available which may be used as a benchmark in the
URBAN AND REGmONAl AND
M ir ~-~ E ir
E r ( p. + M· ) - ~--p. (A.S.2.IS)
En In In E in In
and thus
(A.S.2.I6)
Mir+P ir
~!-(Pin + Min)
En
Assumption (iv) in section A.S.2.I, which rules out the existence of cross-hauling, means
that the regional production of sector i, denoted as P ir, equals the regional consumption of
the regionally produced output of sector i, denoted as Cir. Similarly, national production of
sector i, denoted as Pin' equals the national consumption of the domestically produced out-
put ofsector i, Cin" Therefore the bracketed term on the right-hand side ofequation (A.S.2.I7)
describes the national output of sector i as a proportion of total national consumption of
sector i, and the left-hand-side term describes the net regional imports of sector i as a pro-
portion of total regional consumption of sector i. However, given that
M·
M ir ~r C;r + M v~~
C == 1 can be written as
lr lr
_____ + Pir == I
M ir + P ir M ir + P ir (A.S.2.18)
we have
In other words:
M ir + P ir
Pin
LQ irMi~·+-Pin
(
J (A.5.2.20)
The right-hand bracketed term is the national average propensity to consume the domesti-
cally produced output of sector i. Therefore, by multiplying this by the appropriate LQ
value, we arrive at an expression for the regional average propensity to consume the region-
ally produced output of sector i. Exactly the same result can also be produced if we choose
to employ the CILQ rather than the simple LQ. The only difference in this case is that the
initial regional import function is specified as
M = (E
ir EirJE. + Min (A.5.2.21)
j
1J
Ein In E jn
Xi
Yi = 1 (A.5.2.22)
Ci mi
where Y i is the total domestic sectoral output, Xi is the level ofdomestic sectoral exports, and
ci is the domestic sectoral expenditure coefficient, which equals 1 - m i, where m i is the
domestic import coefficient for the backward input linkages. The total domestic sectoral
import expenditure for the first round will be represented as M i m i Y i, and total first -round
domestic production expenditure in backward linkages will be represented as Pi = Ci Y i. If
there are no other sectoral imports, then Ci accurately measures the first-round domestic
average propensity to consume the domestically produced output of sector i weighted
according to the relative total expenditure on each input as determined by the national
input-output framework. This is because
= ci (A.5.2.23)
Pi+M i CiY i + miY i
*1& ~ ''>ul! ~ '&J nSl_ ll.o l1:: 'l;,.. ~lY &~ %.§ ~~?~ ~ ~ Ha '».Q ~,,, ~AA !I.~ ",,,, A N [)
However, if there are other imports of goods produced by sector i exogenously consumed
by the household sector h, we can represent these additional domestic sectoral imports as
*
M i mi¥h, where i h. Under these conditions, the total domestic propensity to consume
the domestically produced output of sector i can thus be represented as
Under these conditions, observation of the input-output expenditure coefficient alone will
overestimate the domestically produced and consumed output of sector i as a proportion
of the total domestic consumption of sector i, irrespective of whether the region runs a
sectoral balance of payments surplus in which Xi > M i, a balance of payments deficit in
which Xi < M i, or a balance of payments equilibrium, in which Xi = Mi. The result of this is
that the backward linkage input-output expenditure coefficients in the national table will
not accurately reflect overall sectoral net trading balances, and will tend to exceed the over-
all domestic average propensity to consume domestically produced goods. Although the
LQ assumption of the absence of cross-hauling, namely assumption (iv), is not a problem
for input-output models which specifically allow for such behaviour in the first and subse-
quent rounds of expenditure, the LQ adjustment of national input-output coefficients suf-
fers from the problem that cross-hauling can occur at the top level of household demand.
Therefore, if we use national input-output expenditure coefficients as the benchmark
against which regional input-output expenditure coefficients can be produced, this will
also tend to systematically overestimate the regional domestic contribution to sectoral
output, and consequently the regional multiplier (Leven 1956), irrespective of the form of
location quotient employed.
Round (1978) found very little difference in the performance of a variety of LQ specifica-
tions, and although Harrigan et al. (1980) found that the simple LQ approach performed
marginally better than other location quotient specifications, the general accuracy of such
coefficients is open to question. In cases where survey-based regional input-output data do
exist it is possible to compare the survey-based results with those that would have been
predicted on the basis of employment shares. Accepting that the production of survey-
based estimates itself may have required professional judgement based on relative regional
sectoral shares in order to compensate for any missing information, particularly in areas
such as public expenditure, construction, and household consumption, the general picture
we observe is that employment-based estimates of regional trade tend to perform fairly
poorly when compared with survey models (Czamanski and Malizia 1969; Schaffer and Chu
1969; Smith and Morrison 1974) or semi-survey models which employ algorithms to com-
plete the tables (Bacharach 1970; Lahr 1993; Harris 1998).
Chiang (1984 pp. 117-118) and imagine a region where there are industries 1) 2) 3) ... n)
each of which buys from) and sells) inputs to each other) plus an external demand sector
which does not provide inputs to the local production process. The input coefficients for
industries 1) 2) 3) ... n can be arranged into a matrix A == [au] thus:
anI a n2 a n3 .. ·a nn
The coefficients aij represent the requirements of input i needed in the production of one
unit of output j. In the case of industries which do not supply inputs to their own industry)
as in Table 5.3 above) the principal diagonals will all be zero.
If industry 1 produces outputs which are just sufficient to provide for the input require-
ments of each of the other industries 1) 2) 3) ... n) plus the demand requirements of the
external sector) the total output of industry 1 which we denote as Xl' must satisfy the
equation
where:
Ifthe same exercise is repeated for the output ofeach ofthe sectors) we can modify the above
matrix to give
~3
-a31 -a32 (1 - a33) ... -a3n X3 (A.5.3.3)
I
where the matrix on the left-hand side contains the input coefficients) and the vectors on the
left- and right-hand sides contain the outputs of each sector used as inputs by other sectors)
and final external demand for the outputs of each sector) respectively. If we ignore the 1s in
the principal diagonals of the matrix on the left-hand side we see that this matrix is
URBAN AND REGiONAL ECONOM~C AND
simply -A = [aij]. As it is, this matrix is the sum of the identity matrix In' with Is in the prin-
cipal diagonals and zeros elsewhere, and the matrix -A. In other words, we can write (I -
A)x = d, where x represents the variable vector and d the final demand vector. The matrix
(I - A) is known as the <technology matrix' and is usually denoted as T = (I - A) such that
Tx = d. As long as T is non-singular, we can find the inverse of T, denoted as T-l. This now
allows us always to solve the problem
x= T-1d (A.5.3.4)
In other words, for any given level of the external output demand, we can calculate the input
demand requirements through the successive rounds of the input-output expenditure pro-
cess for any ofthe individual production sectors. With this information it is also straightfor-
ward to calculate the total factor earnings, as in Table 5.4, and to calculate the total regional
multiplier impact of any given level of output demand.
Regional and interregional
labour market analysis
6.1 Introduction
In this chapter, we will discuss the question of urban and regional labour markets. Once
again, as with the multiplier models discussed in Chapter 5, we will see that there are some
fundamental differences between the characteristics of the labour market at the regional
and national levels, as well as many similarities between the two. However, these differences
are not simply a question ofscale, but rather an explicit question of the relationship between
market-clearing processes and geography.
In Chapter 4 we discussed the differences between regional and national multiplier mod-
els. The regional multiplier models discussed in Chapter 5 all assume that the marginal costs
of factor inputs are constant. In other words, we assume that the marginal and average costs
of factor inputs are the same as output expands. This allows us to assume that labour, capi-
tal, and land inputs all maintain fixed unit prices independent of the level of output. In situ-
ations where there are unused factor supplies, such as where there is excess capacity in
industrial facilities, or alternatively a pool of unemployed labour, these assumptions may be
justified. However, there are many cases where no such reserve capacity exists. In these situ-
ations factor supplies will be somewhat limited, and the effect of this is that factor supply
prices will not be constant as output expands. The market for factor inputs will therefore
determine factor prices. In the case of geographical labour markets, such factor price
changes may also bring about spatial changes in the allocation of these factors. This is
because such price signals may also encourage factor migration between regions.
In this chapter we will see that the effects of local labour price changes on regional or
urban employment can be rather complicated. Local factor price and income effects can
become somewhat interrelated, with the result that we must consider the spatial problems
discussed in each of the previous chapters in order to come to any coherent analysis of the
issues.
In the next section we will discuss alternative views of the workings of local labour mar-
kets. In section 6.3 we will extend the argument to the question of interregional migration
and factor allocation, and in section 6.4 onwards, we will discuss additional issues which
affect regional labour market and migration behaviour.
AND MODEtS
Labour markets are notoriously complex to analyse, with many labour market outcomes
being the result of complex negotiations between employers and labour representatives
within a bilateral monopoly framework. However, for our purposes here, in order to discuss
the workings of the urban and regional labour market, it is first necessary to return to the
basic microeconomic foundations oflabour market behaviour. These will then be adjusted in
order to allow for the particular characteristics oflocal urban and regional labour markets.
Wage w
D(L) =MPL
the further to the left will be the demand curve for labour. Alternatively, the higher the price
ofthe output good, the further to the right will be the demand curve for labour, for any given
capital stock. Once again, the converse is true. The lower the price of the output good, the
further to the left will be the demand curve for labour, for any given capital stock.
The second basic principle is that the supply oflabour is upward sloping with respect to the
real wage rate. This conclusion is based on an argument which is sometimes known as the
'dual decision hypothesis' (Clower 1965), in which workers use the real wage level in order to
decide simultaneously on the number of hours of labour they wish to supply, the level of
income they wish to earn, and the quantity of human produced goods and services they wish
to consume. The dual decision hypothesis can be explained with the help of Figure 6.2.
In Figure 6.2 we assume that the individual can consume two types of utility-bearing
goods, namely on the one hand, the weekly hours of leisure, and on the other hand, all
human produced goods and services. The vertical axis represents the weekly quantity of
hours of leisure the individual can consume H, with a fixed upper limit F, which represents
a full week. The total number of labour hours supplied per week is thus (F - H). The hori-
zontal axis represents the quantity of human produced goods and services consumed by the
individual 1. We can now employ a standard budget constraint-indifference curve model in
order to understand the supply of labour with respect to the price of labour.
In a standard indifference curve type of framework, assuming the indifference curves are
convex, the object of the individual is to ensure that the price ratio between the two types of
goods is just equal to their. marginal rate of substitution. In Figure 6.2, the slope of the
budget constraint represents the relative prices of leisure and human-produced capital
goods, defined in terms of their opportunity costs with respect to each other. If, for the
moment, we assume that there is a certain element of exogenous consumption even in a
situation oftotal leisure, the origin of the budget constraint will not be on the vertical axis at
c
F
L1
L2
L3
, I'
~
:::J
VI
.Qj
4:-
0
~
:::J
0
:c I
H
12 13
C '1
Quantity of goods consumption'
Figure 6.2 The derivation of the labour supply curve
REGH)NAL
S(L)
~
e
QJ
01
~
""6
&
w* •.............................................................
D(L)
L
Quantity of labour L
F, but will be somewhat shifted to the right of F, at C. As the real wage rate increases from w]
to W 2 to w 3, the slope of the budget constraint becomes shallower, with the result that the
individual consumes less leisure and more human-produced goods. Obviously, there are
both price and income effects operating, in that income itself is the multiple of the wage rate
and the number of hours worked. The optimum combinations of leisure and human-
produced goods consumed, for different budget constraints associated with different wage
levels, can be plotted as an expansion path. Given that the number of labour hours supplied
is represented by L = F - H, we can see that as the real wage rate increases from w] to w 2 to
w 3 , the number of labour hours supplied increases from L] to L 2 to L 3 • The supply of labour
S(L) is therefore assumed to be a positive function of the real wage rate.
The above argument does not rely on the assumption that all labour exhibits the same
preferences. For example, we could assume that the labour market is made up ofheteroge-
nous individuals with different preferences. Some individuals will have a relatively higher
preference for leisure) whereas others will prefer human-produced goods and services.
These different preferences will be represented in Figure 6.2 by different indifference curve
maps. In the former case, the indifference curves will tend to be shifted higher up whereas
in the latter case they will tend to be shifted further down. However, the argument still holds
that as the real wage rate increases, the optimum quantity of labour supplied by each indi-
vidual will increase.
Combining these two basic principles allows us to construct a simple model of a labour
market as in Figure 6.3. The real wage w* is the market-clearing wage at which all labour L*
supplied is demanded. In neoclassical terms the level of employment L * represents full
employment at the current market wage. Under such conditions, there is no involuntary
unemployment, because the labour which is not working, given by the difference between
the total population T and the current employment level L *, is regarded as being voluntar-
ily unemployed.
w
S(L)
w····················..··..
~ 1
e
Q)
~
~
a::::
w* _ .
D(L)
LW1 L* L1
Quantity of labour L
Figure 6.4 Involuntary unemployment: a neoclassical perspective
Given this logic, we can now consider conditions under which unemployment may exist
in such a framework. The first reason why unemployment may exist in such a labour market
is that the real wage being demanded in the labour market is simply too high. We can see
this in Figure 6.4. If the real wage currently offered is WI' this is much higher than the mar-
ket-clearing equilibrium wage of w*. The result of this is that the number of people seeking
work at the current wage is L1, whereas the quantity of labour demanded is Lw1 • The level of
involuntary unemployment is therefore (L 1 - Lw1 ) at the current wage WI. The neoclassical
remedy for such a situation is to allow the real wage to fall from WI to w*, such that the
demand for labour will increase and the supply oflabour will fall until they are brought into
equilibrium. The downward movement in the real wage and the relationship between the
wage fall and the labour demanded is represented by the arrow in Figure 6.4.
In this schema, the only situations in which involuntary unemployment can persist is there-
fore where there is some sort ofimpediment to the free movement of real wages. In particular,
in this case unemployment will persist in situations in which there is some sort of obstacle
which militates against the downward adjustment of the real wage. The question therefore
arises as to what are the possible impediments to the free downward movement of wages.
The first possible impediment is the existence ofa trade union which maintains a monop-
oly over the supply of labour. The role of a trade union is in effect to set up a labour supply
quota. If bargaining between trade unions and corporate management results in a labour
supply quota of LW1 and a union real wage of WI' the current ,market wage for those in
employment, WI' will be higher than the market clearing wage w*. This is what we mean
when we say that the real wage WI is (too high'. Whether involuntary unemployment exists
or not therefore depends on whether the trade unions are able to negotiate real wages for
their members which are higher than the market-clearing wages.
The second possible impediment to the free downward movement of wages is that of
a minimum wage restriction. If a minimum wage policy is instituted by a government,
such that the minimum wage is set at a wage of WI' clearly the effect of this will be to
reduce employment to L WI and to engender involuntary unemployment of (L I L wI ).
Whether involuntary unemployment exists or not therefore depends on whether the
minimum wage is set at a level higher than the market-clearing real wages. Alternatively,
if there is a distribution of wages according to different activities, and a minimum wage
policy raises the lowest wage, it may be that average wages all move upwards as workers
seek to maintain the differentials between different skill occupations. In this case we can
interpret the wage in Figure 6.3 as being the average real wage. Under these conditions
the argument still holds.
In both of these cases, the general neoclassical prescription will be to dismantle the obsta-
cles which militate against the free movement ofwages. This will involve legislation limiting
the power of trade unions, and also the withdrawal of any minimum wage policies.
Apart from the role oftrade unions and minimum wage legislation, there is a third reason
for involuntary unemployment in such a framework, and this is the role of welfare pay-
ments. In order to see this we must return to Figure 6.2. In this diagram we see that there is
an exogenous level of consumption even where no labour is supplied, given by the horizon-
tal distance between F and C. If welfare payments are provided for those without employ-
ment such that the exogenous level of consumption increases, the budget constraint at C
will shift even further to the right. The result of this is that the expansion path which plots
all the efficient consumption points as wages increase and the budget constraint shifts to the
right will be moved further to the right. The effect of this is that, compared with the situation
of little or no welfare payments, in which exogenous income is very low, fewer hours are
worked for any given real wage rate. In terms of our labour market diagram, Figure 6.4, this
implies that the labour supply curve is therefore shifted upwards to the left. The market
wage rate rises and number of people employed therefore falls below the market-clearing
level. Moreover, the greater the level of welfare payments, the further to the left will be the
labour supply curve, and the less will be the total number of people employed.
The neoclassical remedy for the reduced labour demand and supply is, once again, to
dismantle many of these policies. As such, welfare payments will need to be reduced in
absolute terms so as to have a negligible effect on market wages. Alternatively, such pay-
ments will be restricted to a very short time period, after which they will cease to be available
to the individual person.
The simple neoclassical labour market model described above allows for the downward
movement of real wages in order to clear markets. However, an alternative approach to the
labour market question comes from a Keynesian perspective, which argues that wages are
<sticky' downwards. In other words, while wages are able to move upwards over time, down-
ward movements in wages are very difficult to bring about. This is primarily due to the exist-
ence of trade unions and the complex nature of labour bargaining processes. Under these
circumstances, movements down the demand curve for labour, in which wage falls are asso-
ciated with increases in labour demand, are very difficult to effect. In this situation, there is
no guarantee that labour markets exhibiting involuntary unemployment can be expected to
clear. The policy prescription under such conditions is therefore to attempt to expand the
w
/ S(L)
i
j
~
e
<lJ W
c::.n
~ OiL)
]
~
w' _ .
OdL)
LW1 L* L1
Quantity of labour L
demand for labour so as to clear the excess supply oflabour. The argument can be explained
with reference to Figure 6.5, which is constructed on the basis of Figure 6.4.
In the Keynesian argument, a general increase in the demand for labour from D I to D 2
will increase the level of employment from LWI to LI at the current wage WI' thereby clearing
the excess labour without raising the current wage level. If demand increases beyond D 2,
then we will experience wage inflation. However, as long as an expansion in demand can be
limited to a movement from D I to D 2 , the labour market problem can be solved. In macro-
economic terms this implies that the involuntary unemployment can be cleared without
inducing any inflation.
There are many macroeconomic controversies about whether such a costless increase in
demand can actually be effected feasibly or not, and these discussions centre on the questions
of<crowding out', and the relationship between labour market policy, fiscal policy, and mon-
etary policy. It is not our intention here to enter into these debates, as these issues are widely
discussed in detail elsewhere. It is worth noting, however, that until relatively recently it was
fashionable in many circles within the economics profession to talk as though unemploy-
ment were entirely a matter ofeither market imperfections or personal choice, and to assume
that the Keynesian problems associated with deficient demand and deflation no longer
existed (Krugman 1999). The global financial crisis of2008 has changed much of this think-
ing (Krugman 2009), and a result ofthe global financial crisis is that regional unemployment
disparities have reappeared with a vengeance in many countries (OEeD 2011). These devel-
opments have thereby reopened the debate regarding the nature of regional unemployment
and the ability of regions to recover from adverse demand and employment shocks-what
has become known as regional <resilience' (Martin 2011; Fingleton et al. 2011).
It is important to understand the basic analytical principles behind the various approaches
to regional labour market problems and then to apply these principles in the case of an
explicitly spatial local or regional context. This will allow us to discuss the particular features
A t.J D ~'& l!.;.•",,~ ~ ~J ~~R s.... i!: •• '<.• l'.. >! i!:"i!: .....s- i!:""i!: ~ ~ ••
of urban and regional labour markets as distinct from non -spatial analyses. As we will see,
there are certain aspects of urban and regional labour markets which are rather different
from standard textbook models of labour markets, at both the micro and macro levels, and
an understanding of these differences is essential in order to identify the role played by
geography and space in labour market behaviour.
S(L)
w1
~
e
QJ
w3
0)
~ W2 r-·y··········l·······
~
0:::: w*
°l(L)
OiL)
L2 L3 LW1 L* L1
Quantity of labour
Figure 6.6 Local labour markets and downward wage movements
aggregate demand can induce a multiplier effect. Although all the multiplier changes dis-
cussed in Chapter 5 involved positive demand changes, the same types of arguments also
apply to falls in any of the components ofaggregate demand. Negative multiplier effects can
be generated by falls in any of the individual components of aggregate demand, thereby
leading to even greater reductions in income than the original demand fall. In the above
situation of wage falls in the local labour market, local firms may be unwilling to increase
labour demand according to the neoclassical model, and will rather seek to reduce invest-
ment expenditure by running down existing stocks of goods and cancelling future planned
investment. This will also imply that firms will cancel orders from their suppliers. The com-
bined effect of these responses to the local downward wage movements will be a negative
local income multiplier effect given by -LlY = -kr (LlwL w1 ), where kr is the value of the
regional multiplier. In terms of Figure 6.6, this contraction in local expenditure income can
be represented by a downward shift to the left of the demand curve. The vertical distance of
this backwards shift at the employment level of L 1 is represented by a fall in wages of -Llw
from W 1 to w2 , such that -LlY = -kr(LlwL w1 ).
As we see in Figure 6.6, if the demand curve shifts backwards to the left, a range of
wage-employment combinations also becomes possible (McCombie 1988). The actual
employment effect of the wage fall depends on the labour retention policies of the local
firms. If the local firms absorb the negative expenditure-income effects almost entirely
through contractions of their labour stocks, rather than wage reductions to employees,
the wage-employment effects will be represented by the locus b in Figure 6.6 in which we
maintain a wage of W 1 but reduce employment from LW1 to L2• On the other hand, if firms
choose to absorb all local expenditure falls in terms of wage cuts, rather than labour
reductions, the wage-employment locus will be given by c in Figure 6.6 in which we
maintain the employment level at LW1 but reduce wages from W 1 to w2 • The final possibility
is that firms will absorb the fall in local expenditure by cuts in both wages and labour
Output
Q = frlKF,L)
Qc
Qd
Qb
L2 L] LWI Labour
employed, given by the locus d in Figure 6.6, in which wages are cut from WI to W 3• and
employment is cut from L WI to L 3 •
Although the actual wage-employment locus observed in response to a local wage fall
will depend on the labour retention and employment policies of the local firms, it is neces-
sary to consider how it could be possible for the demand curve to be considered to have
actually shifted downwards to the left, as proposed by the Keynesian model. The argument
is that if the type of negative regional income-expenditure effect described above does
indeed operate, not only will local firms cut back future planned investment and input
expenditure, but they will also reduce the current level of local capital utilization. This pos-
sibility is represented by the production function diagram Figure 6.7. As we see in Figure
6.7, changes in capital utilization can be directly associated with changes in both output and
wages. The argument here is that different levels of capital utilization in effect represent dif-
ferent regional production functions. Assuming as usual that capital and labour are comple-
mentary production factors, even temporary cutbacks in regional capital utilization in
response to falling local wage income will move the local firms on to lower capacity produc-
tion functions. This is because less capital is now applied to each unit oflabour employed. In
Figure 6.7, a reduction in the level ofcapital utilization from that of full capital utilization Kp
to a situation of partially unused capital Kp can be represented as a move from a full-capacity
regional production function in which output Q can be defined as Q == !p(Kp,L) to a lower-
capacity regional production function given as Q == !p(Kp,L). The slope of the regional pro-
duction functions represents the marginal product of labour, and consequently the local
regional wage rate. If the local firms choose to adopt the labour retention strategy of b, in
which wages are maintained at their existing levels, and the fall in demand is absorbed
entirely in terms of labour cutbacks, in Figure 6.7 this is represented by a fall in labour
demand from LWI to L2, as in Figure 6.6, and a fall in output from Q 1 to Qb. As we see, the fact
that the regional wages are unchanged at WI means that the slopes of the two regional pro-
duction functions at these two different levels of capacity utilization, employment and out-
put, are the same. The second case is where the labour retention strategies of the regional
firms are represented by locus c is Figure 6.6. In this case, the employment level is main-
tained at L WI but the output level falls from Q I to Qc and the regional wage falls from WI to
w2 . This is represented in Figure 6.7 by the lower slope of the regional production function
at the existing employment of L wI . The final alternative is where firms adopt the labour
retention strategy represented by the locus d in Figure 6.6. In Figure 6.7 this is represented
by a fall in output from Q I to Qd' a fall in employment from LWI to L 3, and a fall in the
regional wage from WI to W 3, a wage level somewhere between WI and W 2•
Given these general observations, it is therefore necessary at this point to consider which of
the possible wage-employment and capital utilization effects described by the loci a, b, c, or
d in Figure 6.6 are likely to take place in a regional labour market in response to local wage
falls. In a Keynesian model of the regional labour market, as we see in Figures 6.6 and 6.7,
downward movements of local wages are not possible without simultaneous backward
shifts in the demand curve for local labour. The reason for this is that the negative income-
expenditure effect on local firms' perceptions of local market demand is regarded as domi-
nating any potential desire on the part ofthese firms to take advantage oflower wages in the
form of increased labour demand. This results in the local firms cutting back the level of
capital employed. In macroeconomic discussions, this particular type of negative income
effect in response to a wage fall, represented by the loci b, c, or d in Figure 6.6, is sometimes
known as a (Keynes effect'. On the other hand, the willingness of firms to increase labour
demand in response to a wage fall, represented by the locus a in Figure 6.6, is sometimes
known in macroeconomic discussions as a (Pigou effect'. In the neoclassical model, the
Pigou effect will generally dominate any possible Keynes effect, whereas in the Keynesian
model the Keynes effect will dominate any potential Pigou effect. The extent to which one
effect dominates the other tends to be both a question of industrial sector and a question
of time.
In the case of local regional or urban labour markets, we can argue that in the short run
at least, the local firms with primarily local markets will tend to interpret local wage falls in
terms of reductions in their potential output market sales revenue. These types of firms are
the firms which we generally classed as (non-basic' in our economic base discussions in
Chapter 5. For the firms of this type, the negative income-expenditure effect will tend to
dominate their labour demand decisions, and will generally lead to cutbacks of the type
represented by the loci b, c, or d in Figure 6.6.
On the other hand, for (basic firms' which rely primarily on regional export markets, falls
in local wages will have little or no effect on their overall market outputs. For these firms,
reduced local wages may mean that the area actually becomes more attractive for expanding
output by employing more labour, and such firms may therefore increase their employment
levels within their current levels ofcapital investment. This will be represented by the wage-
employment locus a in Figure 6.6. As we saw in Chapter I, reduced local labour prices may
W
~
WI
/
e
Q)
0'1
~ DiKR2,L)
"0
Q)
DlKRI,L)
LWI LI
Quantity of labour L
Figure 6.8 Inward investment and the regional labour market
also in the long run encourage these firms to expand their overall local capital stocks.
Similarly, new immigrant firms may move into the region and this process will also increase
the regional capital stock. Both of these capital expansion effects, one which takes place
within existing plants and the other which results in the establishment of new plants, will be
represented by a shift to the right of the labour demand curve, as the regional capital stock
expands from K R1 to K R2 • As we see in Figure 6.8, the long-run result of this regional capital
expansion will be to increase both the local wage and the level of regional employment. The
actual extent to which the demand curves will shift outwards to the right will depend on the
level and the speed of new inward investment flows.
The local effect of regional wage falls will therefore depend on the sectoral balance
between the exporting and domestically oriented firms. Regional economies which are
highly integrated internally, such as those which exhibit strong localization and urbaniza-
tion economies as described in Chapter 3, will tend to suffer from general falls in local
wages, because much of the local demand will be locally generated. On the other hand,
economies which are vertically integrated, in terms of being dominated by strong hierarchi-
cal input-output expenditure linkages between locally based exporting firms and local sup-
plier firms, will tend to benefit from local wage falls. The reason for this is that such
economies will tend to become better places for immigrant mobile investment of the type
discussed in Chapter 2.
There is one exception to the argument that economies which are vertically integrated, in
terms of being dominated by strong hierarchical input-output expenditure linkages
between locally based exporting firms and local supplier firms, will tend to benefit from
local wage falls. This the case ofa local economy which is dominated by strong input-output
linkages between locally based exporting firms and local supplier firms, and where the ini-
tial cause ofthe involuntary unemployment described by Figure 6.4 is actually a contraction
in the local regional export base sector itself.
iN'1rEfU~t:GmONIAl <-.."." ..."'"<-.#»." .... MARKET
~
e
w
1
/ .
S(L)
Q)
c::n
~ °lL)
""6
V
Q)
0::::: w* .
: OiL)
LW1 L* L1
Quantity of labour L
Figure 6.9 Labour market effects of a reduction in export base output demand
In such a situation, we can redraw Figure 6.4 as Figure 6.9 in which the initial level of
regional labour demand is D I , the regional market clearing wage is WI' and the level of
regional employment is L I • Following a fall in regional export demand, the local basic sector
will cut back its output. Therefore the local regional labour demand on the part of both the
basic and non-basic industries will now fall from D I to D 2 • At the existing wage of WI'
the labour employed will fall from Lito LWI and involuntary unemployment will be given as
(L I - L wl )' If wages are unable to fall to w* for the kind of institutional reasons outlined in
Figure 6.4, the involuntary unemployment will tend to be persistent unless there is a com-
pensating change in the position of either the labour demand or supply curves.
In the case where the involuntary regional unemployment has been caused primarily by a
general contraction in regional export demand, this will tend to indicate that the region is cur-
rently not an attractive location for new investment. This implies that in the short or medium
term at least, an expansion of the regional capital stock by new immigrant investment which
is sufficient to compensate for the export base contraction would appear to be unlikely.
Moreover, in such a situation local business confidence will tend to be very low. This will imply
that local business expansion of the type represented by the locus a in Figure 6.6 is unlikely to
take place. Further local wage falls are therefore much more likely to induce further negative
local income-expenditure type effects described by the loci b, c, and d in Figure 6.6.
However, although the regional capital stock can only be expanded relatively slowly, the
regional capital stock can be reduced relatively quickly. If there is a strong local negative
income-expenditure effect such as described by Figure 6.6, firms may withdraw capital
quickly in response to falling local wages by reducing the level of capital utilization, as we
see in Figure 6.7. If such reduced capital utilization does take place, the unused excess
capital (K p K p ) can be withdrawn from production very quickly simply by cutting back
output, although it will not necessarily be scrapped in the short run. The reduction in
investment will mean that the only form of current investment still undertaken is the
depreciation expenditure on existing capital infrastructure, which is necessary to main-
tain it for future use. However, if the demand fall is perceived by firms not to be a short-
term phenomenon, even this depreciation investment may be curtailed. In situations
such as this, the capital will be withdrawn permanently from the productive process, and
the firms will move permanently to a lower-capacity regional production function. As
such, the fact that much of the regional productive capital is both durable and location
specific means that there is something of an asymmetry in regional factor markets.
Regional capital reduction can take place at a much more rapid pace than regional capital
expansion. As we have seen in Figure 6.9, this is particularly a problem in the case where
a region suffers an export demand fall.
In the particular case of urban and regional economies, the withdrawal ofcapital from the
local productive process also has very specific regional implications, and the reason for this
is, once again, the location specificity and durability of regional capital. To see this we must
combine our two key insights, namely, that much regional capital is durable and adjustment
costs are high, and also that regional capital expansion is relatively slow whereas regional
capital contraction can be rapid. The combination of these observations suggests that that a
rapid regional contraction will result in a physical environment characterized by derelict
capital assets. The problem with this is that it can generate a negative externality in the form
of a reduction in the quality of the local environment. Capital dereliction and decay can
therefore playa major role in altering the attractiveness of the regional economy as a loca-
tion for future investment.
An example of this is given in Figure 6.10, which is an extension of the argument
given in Figure 6.9. If regional demand falls from D] to D 2 , capital will be withdrawn
from production as described by Figure 6.7. If the withdrawn capital leads to derelic-
tion in the short to medium term, this may lead to a deterioration in the local eco-
nomic environment and further reduce the attractiveness of the region as a location
for investment. In the subsequent time period, demand may fall even further to D',
with involuntary local unemployment at the existing wage rate of w] increasing from
(L] - L W ]) to (L] - L w').
We can similarly reverse the argument, and imagine a demand expansion from D' to D in
one time period. This could be due to the immigration of a major new immigrant firm. The
consequent expansion in the regional capital stock may engender further growth in local
labour demand from D to D] as other new firms locate in the region and local business con-
fidence increases in general. As such, the investment and labour market decisions in the
current time period depend on the investment and labour market decisions made in previ-
ous time periods in which the economic conditions may have been very different from those
which currently prevail. This particular feature of the labour market, in which there is a
~•. "'·;':~;'I..Z~'SI_&.o l,;;"'.~"'-!l~"Il",.#s~r~~ ...... nH'ILJS,.ff .... ~ ..•." ,AND
wl · · · · · ·~
S(L)
~ ..........~/ .
e
QJ
01
~
"'5
& I DlL)
l. _ .
w·I.............•............
/
: D2(L)
D'(L)
Lw' w7 L* L7
Quantity of labour L
Figure 6.10 Negative interactions between capital withdrawal and labour demand
partial dependence on previous labour market and capital investment behaviour, is often
known as 'hysteresis'. Yet the phenomenon of hysteresis itself in part depends on the dura-
bility and location specificity of the regional capital with which the labour is combined.
Regional capital exhibits a great deal of inertia, in terms of both quantity and quality. (See
Urban and Regional Example 6.1.)
Given that regions tend to exhibit asymmetries in terms of their ability to adjust capital
stocks to increases or decreases in demand, it may be that, in many cases, a reliance on regional
investment and capital changes can be a rather inefficient way of ensuring effective regional
adjustment. The alternative mechanism is therefore to allow the regional supplies oflabour to
adjust. The relative success of these two mechanisms depends on whether the demand for
labour, dependent on the regional capital stock, or the supply oflabour, dependent on migra-
tion behaviour, is more quickly able to adjust to changing regional economic conditions.
As we have seen in the above sections, the individual region has limited internal wage
adjustment capabilities, particularly in response to adverse demand changes. However,
there is a mechanism which operates at the interregional level which can allow the region to
adjust more rapidly to such changes. This mechanism is that of interregional labour migra-
tion. There are three broad types of interregional migration mechanism associated with
wage levels, namely the disequilibrium model, the equilibrium model, and the endogenous
human-capital model. We will discuss each of these models here in turn.
(a) (b)
SA SB
WA
WB
W A1 OBI
DB
DA1 'PA
LA1 LA LB LB1
(c) (d)
SB
SAl
SA SBI
W .
Bl
WA
WB
W A1
OBI
DA1
As we see in Figures 6.11 c and d, this migration behaviour will shift the labour supply
in region A to the left from SA to SAl' and to the right in region B from SB to SBI. As more
people enter region B, the available local labour supply expands, thereby reducing the
marginal productivity of labour and the local wage rate in region B. Similarly, assuming
that the regional capital stocks have remained unchanged, from the law of diminishing
marginal productivity we know that the reduced labour supply remaining in region A
will experience a relative increase in its marginal product. This process of migration will
continue until the falling real wage in region B and the rising real wage in region A are
brought back into equilibrium at the original regional wages of W A == W B • Once this has
been achieved, the interregional migration of labour from region A to region B will
cease. Although real wages will have been brought back into equilibrium between the
regions at their original levels, the total quantity of labour employed in each region will
iNln;~H~E:GION!All&,.F'%.li..8'_\h,8'~" MARKET
W Xl
F
WE t···········································t········£
G ., . W Yl
Oy Ox
i<
l<
LXl
LX
>1<
I
)~(
L Yl
Ly
1
,
Figure 6.12 The welfare gains associated with interregional migration
have changed from the original situation. In the new equilibrium, the total quantity of
labour employed in region A will have fallen in two stages from LA to L AI , and then from
LA 1 to L A2' while regional wages will have fallen and risen in two stages from W A to W Al
and then from W Al to W A' respectively. Meanwhile, in region B, the total quantity of
labour employed will have risen in two stages from L B to L BI , and then from L BI to L B2 ,
while regional wages will have risen and fallen in two stages from W A to W Al and then
from W Al to W A' respectively. In the new equilibrium situation in which there is no inter-
regional migration, region B will now be much larger than region A, whereas initially the
two regions were the same size.
The process of interregional labour migration can be shown to be efficient from the
point of view of the economy as a whole. In order to see this we can employ Figure 6.12,
in which the labour demand curves for two regions X and Y of identical capital stocks are
superimposed on each other by reversing the labour demand curve for region Yhorizon-
tally from left to right. In other words, we can read the labour demand curve of region X
from left to right as normal, and read the labour demand curve of region Y from right to
left. We can begin with a situation in which there is a high marginal product and real wage
in region X of W XI for a low level of employment in region X of LXI' represented in Figure
6.12 by F. Meanwhile there is a low marginal product and real wage in region Y of W Yl for
a high level of employment in region Y of L yI , represented in Figure 6.12 by G. The total
level of employment in the economy is given by L N , where L N = (LXI + L yI ). If one mar-
ginal unit of labour now transfers from region Y to region X, the individual person
achieves an increase both in their marginal product and their real wage of (WXl wYl ).
This marginal transfer of labour between the regions marginally increases the wage in
region Yby L1w Yl and reduces the wage in region X by L1WXl • At the same time, the labour
employed in region Y falls to (L xl - 1) and the labour employed in region X increases to
URSA N !" N 0 ~"'" ~"'%.« ~ 'l.., '; ~~1!"~ ~ &..,-.;;.< ~'-.7 ~~"' ••-? ~.~,~ ~ '&. .• nn: 'ik.,~ 11.# g:;; 1Il•.• :'&
? °V'
W Xl . F
~
WE'
WE
G
Ox'
: Ox
°v
;
)~<
j< LXl
)~( !
LVI
1
j< Lx Lv 7j
r Lx'
r L'
V
~
1<
Figure 6.13 Interregional labour equilibrium with different capital stocks
Lx"
i LV /I
!
(LXI + 1). However, there is still an interregional difference in regional marginal products
given by (W XI LlwXI ) - (w YI - Llw yl ) and for the next person who migrates from region Y
to region X, (W XI - LlwXI ) - (w Y1 - Llwy/) represents the increase in their real wage.
Interregional migration will take place from region Y to region X until real wages are
equalized in both regions at WE' and labour employment in each region is given as
(Lx = Ly), as represented in Figure 6.12 by E. As before, the total national employment is
given as LN (Lx + Ly).
The difference between the marginal products attainable in each region at the existing
regional labour employment levels not only represents the real wage increase available to
the marginal migrant. This difference also represents the foregone national output which is
not produced iflabour migration is not allowed to take place. Therefore the area EFG can be
regarded both as the deadweight loss to society due to a lack of interregional migration, and
the Pareto efficiency gain to society associated with interregional migration.
Although in Figure 6.12 the argument is constructed by assuming each region X and Y
initially exhibits an identical regional capital stock with identical regional demand pat-
terns, there is no reason why this should be the case. If the two regions initially have differ-
ent capital stocks, this would simply imply that the relative positions of the regional labour
demand curves from each of their respective vertical axes would change. We can see this in
Figure 6.13. We assume that the interregional equilibrium was initially the same as that
achieved in Figure 6.12, where regional wages are given as WE in both regions, and labour
employment in each region is given as Lx and Ly, respectively, and where Lx = Ly. If the
stock of capital in region X increases, for example, due to the immigration of a new firm
into the region, the demand for labour will increase in region X from Dx to Dx'. This will
increase the equilibrium interregional wage to w.E', at regional employment levels of Lx' and
L y , and the Pareto efficiency gain associated with migration will increase from EFG to GHJ,
at the initial disequilibrium regional labour supplies of LXl and Lyl •
Ifthe capital increase in region X, on the other hand, has been due to the migration ofa firm
from region Y to region X, this will affect the labour demand curves in the two regions in an
equal and opposite manner. In this latter case, the demand for labour in region Y will have
fallen from D y to D y , such that the equilibrium interregional wage will remain unchanged at
WE' but the new equilibrium labour supply in region X, given as Lx'" will be twice as large as
that in region Y, given as Ly '. As we see here, by comparing this result with the initial result in
Figure 6.12, it is perfectly possible for the sizes of the two regions, defined in terms of their
capital and labour stocks, to be quite different, although the equilibrium interregional wages
are identical. At the same time, the efficiency gains associated with interregional migration are
always available as long as interregional real wages are not in equilibrium.
used in the models, the power of many statistical tests may be weakened by these meas-
urement problems. These data problems may therefore explain why many econometric
models of migration do not appear to find the (correct' results, as suggested by the dis-
equilibrium model.
On the other hand, there is a second and more fundamental critique ofthe disequilibrium
model of migration, known as the (equilibrium' model of migration (Graves 1980,1993).
The equilibrium model of migration argues that there are no (correct' results as such in the
relationship between net migration and real wages, as suggested by the disequilibrium
model. The reason for this is that as well as being a reward for labour services in the produc-
tion process, wages are also perceived to be a partial compensation for amenity differences.
This is because residence in one area or another implies that the bundle of environmental
amenity goods consumed by residents differs by location, and utility is gained from the
consumption of these goods. In areas of high amenity, workers may be willing to accept
lower wages for any given overall level of utility. On the other hand, in areas of poor envi-
ronmental quality, workers may require higher wages to attain any given level of utility. The
overall equilibrium migration argument is based on the consumption models of Roback
(1982) and Tiebout (1956) applied to the case of labour mobility, and these models are dis-
cussed in Appendix 6.1.
If this compensation argument is correct, in a country with heterogeneous regions, com-
paring real wages across regions on the basis of either nominal wage indices or nominal
wages deflated by local cost of living indices will not tell us very much about the relative
utilities of the workers in each of the regions. As such, we cannot assume any particular
migration motives for workers between any two regions unless we can explicitly account for
such amenity differences. This leaves us with enormous empirical problems, in that we
would have to calculate environmental indices for all locations and incorporate these into
our local real wage indices in order to produce appropriate regional real wage data. The
problem the equilibrium migration arguments also then raise is that the construction of
appropriate interregional consumption indices based on a common basket of goods
becomes extremely difficult, for the very reason that different locations mean that different
baskets of environmental goods are consumed. Moreover, the logical limit of this argument
is that we would also have to account for all consumption differences by location, whether
according to natural or human-produced environmental differences.
As we saw in Appendix 3.3 to Chapter 3, the models of Fujita et al. (1999) allow for util-
ity to be related to the local variety of consumption opportunities. Similarly, the Glaeser
et al. (2001) and Gottlieb and Glaeser (2006) arguments suggest that all sorts of urban
consumption amenities are related to urban cultural and leisure services. These in turn
can also be considered as environmental amenity variations, albeit human-produced
ones, and strictly speaking these would also need to be added to the natural environmen-
tal variations to provide a complete local amenity index. Immediately it therefore becomes
clear that the econometric problems involved in these issues are very significant indeed.
Notwithstanding these difficulties, however, in the US case at least, one amenity variable
appears to be significant in determining migration patters over time, and this is the mean
January temperature (Partridge 2010), and as we see in Box 6.1, there is some limited
evidence that similar issues operate within individual European countries (Cheshire and
Magrini 2006).
6.5.3 The endogenous human-capital model of migration
A third approach to analysing the nature of interregional labour migration is based on the
consideration of the microeconomic characteristics of individual migrants themselves.
The basis of this argument is known as the human-capital (HK) model of migration, and
is a development of the standard model of human capital first widely discussed by Becker
(1964). A simple model of human capital is given in Appendix 6.2, applied to the case of
migration (Faggian and McCann 2009a). However, the broad arguments of the model of
human capital, and their relationship to labour migration, can be understood quite
quickly.
The basic human-capital argument is that rational and well-informed individuals will
invest in personal education and training in order to increase their stock of skills, defined
here as human capital, in order to maximize their expected lifetime utility, defined here in
terms of their lifetime income plus job satisfaction. Education and training tend to be
undertaken before employment commences fully, so the costs of such activities are gener-
ally borne at an early stage in the career of an individual, whereas the employment earn-
ings will accrue over the career history of the individual. At the same time, different
lifetime incomes will be earned in different occupations and the cost of training in differ-
ent skills will differ between different occupations. The individual worker therefore has to
consider what is their optimum mode of employment to aspire to, and, consequently,
what is the optimum level of personal education to invest in initially. Given good infor-
mation on expected wages and labour training costs, as we see in Appendix 6.2, such a
calculation is perfectly possible using standard present-value discounting techniques. The
general assumptions are that the higher the human capital of the individual, the relatively
higher will have been the costs of their education in general, due to the extended time
involved in training. At the same time, the higher the human capital of the individual, the
relatively higher will be their expected wage, due to their increased marginal productivity.
REG ~ONl\ L 8~%., '>.",. ~"8"'§, 8~~8 ~ ~~ 0 » . "<-.,4 " ' " ~•. »....0'
However, given that educational investment must generally take place before any long-
term career develops, there is always an element of risk in the educational investment
decision, in that the actual lifetime earnings may differ from those which were initially
expected.
From the perspective of urban and regional labour market behaviour, the problem is to
understand the relationship between migration behaviour and the maximization ofexpected
wages within the human-capital framework. In order to do this, we must combine the
standard human-capital theory outlined above with what is known as <search theory'
(Molho 1986). The basic premise of search theory is that labour will only consider accepting
a job if the wage offered is greater than, or equal to, a particular personal minimum accept-
able wage, known as a <reservation' wage. Individuals will continue with a process of job
search in which job positions are considered sequentially until one offers a wage which at
least matches the individual's reservation wage. From human-capital theory, we know that
greater human capital generally involves greater initial education costs, and also greater
potential wages due to higher skills. Therefore the reservation wage tends to increase for
individuals with greater levels of human capital. This means that the higher the human
capital of the individual, the greater will be the length of the job search process. However, in
terms of regional labour market behaviour, the combination of human-capital theory with
job-search theory also has a direct implication. The implication is that higher human-capital
individuals will tend to search for employment opportunities over a wider geographical
area than those with lower human capital in order to find employment opportunities offer-
ing wages at least equal to their higher reservation wage. In the cases where such employ-
ment opportunities are found and taken up, the result will be that the higher human-capital
individuals will be more likely to have migrated over greater geographical distances than the
lower human-capital individuals. In order to maximize the returns to their human capital,
higher human-capital individuals therefore tend to be more migratory than lower human-
capital individuals, both for reasons of recovering their initial costs of the human-capital
acquisition, and for attaining their expected wages. At the same time, we can also argue that
higher human-capital individuals will also be better informed of alternative employment
opportunities across regions via easier personal access to informal employment networks.
Once again, this will tend to increase the migratory nature of higher human-capital
individuals.
There is now an increasing body of evidence on this issue (Faggian and McCann 2006,
2009a,b,c; Faggian et al. 2006, 2007a,b; McCann et al. 2010) which indeed suggests that
higher human-capital individuals migrate both further and more frequently than lower
human-capital individuals. This is particularly noticeable in the case of university gradu-
ates, whose combination of skills and youth allows for the highest rates of wage enhance-
ment (Mincer 1974), as long as they move to the locations affording the best opportunities.
The argument that higher human-capital individuals will tend to be more migratory than
lower human-capital individuals has profound implications for our understanding of the dis-
equilibrium model of migration. In order to see this we can consider Figures 6.14a, b, c, and d,
which are a modification ofFigures 6.11 a, b, c, and d. In Figures 6.14a and b, we assume initially
that the real wage in region A, denoted as W A' and the real wage in region B, denoted as wB' are
equal, as are the equilibrium employment levels in each region, denoted as LA and LB, respec-
tively. As in Figures 6.11a and b, ifthe demand for labour in region A decreases from DA to DAl
(a) (b)
SA
SB
WA
WB
W Al DBl
DB
DAl "'PA
LAl LA LB LBl
(d)
(c)
SAl
SA
DB
W Bl
WA
WB
W A1
DA
while the demand for labour in region B increases from DB to D BI , the real wage in region A will
fall to W AI , and the real wage in region B will rise to W BI • Similarly, the level of employment in
region A will fall to LAI , and the level of employment in region B will rise to L BI • As before, in
Figures 6.11 c and d, the interregional wage disequilibrium of (WBI - WAI) encourages labour to
migrate from region A to region B, resulting in a labour supply shift in region A to the left from
SA to SAl' and to the right in region B from SB to SBlo However, the human-capital argument
above suggests that the most migratory individuals who most efficiently respond to regional
wage signals will tend to be the most highly educated workers. Ifwe assume, therefore, that high
human-capital individuals migrate relatively quickly, whereas low human-capital individuals
migrate relatively slowly, the effect ofthis selective migration can be understood as altering the
relative regional labour demand in favour ofthe expanding region.
REG~ONAL
To see this we can consider the outward shift in the labour supply SB to SBl in region B
as tending to imply an increase in the supply of human capital within the region. If
we consider the regional capital stock as comprising both physical capital and
also human capital, the average and marginal product of the regional capital stock will
have increased. The effect of this will be to further shift the labour demand curve in
region B from D BI to D B2 • Conversely, in region A, the reduced supply of labour from SA
to SAl will tend to mean a decrease in the supply of human capital within the region.
Once again, if we consider the regional capital stock as comprising both physical capi-
tal and human capital, the average and marginal product of the regional capital stock
can be considered to have decreased. The labour demand curve in region A will shift
further to the left.
In this particular selective migration process, the total quantity of labour employed in
region A will have fallen in two stages from LA to LAJ' and then from L AI to L A3 , while
regional wages will have fallen from W A to W AI. Meanwhile, in region B, the total quantity
of labour employed will have risen in two stages from L B to L Bl , and then from L BI to LIB'
while regional wages will have risen from W A to W AI. Obviously, this process may continue
cumulatively beyond the two stages identified here, and there is no particular unique
interregional wage-employment equilibrium position towards which the regions will
converge. The reason for this is that region B is enjoying agglomeration economies, and,
as we know, any model in which there are economies of scale may have multiple equilib-
rium outcomes. Recent evidence from the USA suggests that if we consider the case of
urban amenities rather than natural amenities, the evidence does indeed tend to support
human-capital explanations of urban growth rather than amenity explanations (Shapiro
2007; Gottlieb and Glaeser 2008). The limits of the cumulative process described are the
same as the limits to agglomeration discussed in Chapter 3. Such a cumulative process of
local growth and decline may eventually reach a point where region B begins to exhibit
agglomeration diseconomies of scale. In this case the cumulative processes, as repre-
sented by Figures 6.14c and d, will begin to be replaced by the equilibrating processes
represented by Figures 6.11 c and d.
a job. If local house prices have fallen significantly, a worker may not be able to cover the
mortgage value of the house from its sale, nor fund the deposit required for a mortgage
in the more expensive region. Therefore, even if alternative superior job offers are avail-
able in more buoyant regions, the worker will be unable to leave the weaker region
(Bover et al. 1989). In such a situation, only intraregional job moves within the same
metropolitan region, which do not involve a change of residence, will be possible for the
worker. As well as the workings of the private housing market, another potential obsta-
cle to interregional migration comes from public sector housing policies. In some cases,
subsidies to publicly provided housing may reduce the likelihood of migration (Hughes
and McCormick 1981; Minford et al. 1988) by artificially increasing the real wages of
those with state housing. In such cases, workers may be much less responsive to inter-
regional wage signals.
The third point concerning migration is the problem of understanding whether the
acquisition of a higher-wage job is the result of, or a cause of, migration behaviour. Many
arguments suggest that, for many people, the availability of employment opportunities
causes migration to take place as a response to job acquisition, rather than as part of the
job-search process itself. The major evidence in support of this is that the levels of inter-
regional migration in many countries tend to be highly pro-cyclical. In other words, as
the national economy expands, the availability of jobs increases, and the levels of inter-
regional migration tend to increase. On the other hand, as the economy contracts and
employment opportunities diminish, the levels of interregional migration tend to fall.
The result of this is that, in many countries, the differences between regional unemploy-
ment rates tend to fall as the economy expands and tend to increase as the economy con-
tracts (Gordon 1985), although the evidence for this is can be rather difficult to interpret
(Hemmings 1991).
An additional alternative approach is to interpret the regional wage arguments outlined
above in terms of the expected wages earned by migrants, rather than the actual wages.
Expected wages are the wages earned on acquiring a job multiplied by the probability of
actually gaining employment in the relevant sector in the respective region. This is the clas-
sic Harris and Todaro (1970) argument which is familiar in the development economics
literature (ThirlwalI1994). From the arguments in Chapter 3, this would imply that there
may be situations where migration will tend to take place from peripheral areas to central
higher-order areas, even though actual real wages in the central urban areas may be lower
than those earned in the peripheral areas. Such cases will arise where central urban areas
are achieving agglomeration economies of scale. The reason for this is that the probability
of finding appropriate employment is much higher in the central urban areas, thereby
allowing real wages earned to be lower. Net migration flows between the regions will con-
tinue until the expected wage is equalized between the regions. In the above scenario, if
economic growth tends to originate in the dominant central areas, for the reasons outlined
in Chapter 3, this will imply that the central areas initially grow faster than the peripheral
regions. Migration will take place from the peripheral regions to the centre, although over
time the net migration levels will fall as the peripheral regions begin to grow and close the
gap with the core regions. In contrast, at the aggregate macroeconomic level, as the econ-
0my as a whole contracts, the migration flows will fall because the overall availability of
jobs in the central core destination regions will fall.
AND
An interesting final aspect of all of these migration models, as discussed in Box 6.2, is that
fact that the vast majority of people do not migrate, and that interregional migration rates
in many countries have been falling, even before the financial crisis of 2008.
6.6 Non-wage-related models of interregional migration
Each of the above models of migration depends primarily on the relationship between
regional wages and employment levels. There are, however, two major models of migration
which are primarily independent of wage levels. These models are known as the gravity
model of migration and the life-cycle model of migration. Both of these models suggest that
migration will take place even though wages or expected wages, or amenity-adjusted wages,
are in equilibrium. The major difference here is that these two models focus on gross migra-
tion and not net migration, as is the case with the models described above in sections 6.5.1
to 6.5.3.
M
AB
= G_!~PB (6.1)
(dAY
whereby PAand PBrepresent the population sizes of the two city-regions, and dAB represents
the distance between the two locations, and the parameters G and a are constants to be
determined. Although the model appears to be a direct analogy from the physical laws of
gravity attraction between any two objects, there is, however, a reasoning behind the model
based on both probability and economics. In order to understand this we must consider the
justifications for the numerator and denominator terms of equation (6.1) separately.
The structure of the numerator term is based on the argument that the expected number
of moves by individuals to or from any region will be directly related to the population sizes
of the regions. In order to see this we can consider the case where the total national popula-
tion is given as PN' and the total number of interregional migration moves per time period is
given as M N. Here, the average number ofinterregional migration moves per person per year
is thus given as MNIPN. In terms of out-migration, ifwe assume that all people in the country
are homogeneous in terms of their propensity to migrate, the expected total number of out-
migrants generated by area A will be given by (MNIPN)PA • Therefore, for any given popula-
tion migration propensity, the total number ofout-migration moves from any area A will be
positively related to the total number of people in the area PA. Meanwhile, if the relative size
of any particular potential destination region B is given by P1/ PN' the expected total number
ofin-migrants per time period to regionBfrom region A will be given as «MNIPN)PA)(PBIPN ) ,
which gives (MNPAPB)/(PNPN)' Similarly, if the relative size of any particular potential desti-
nation region B is given by PBIPN, the expected total number of in-migrants to area B from all
other regions will be given by MN(PBIPN). The contribution of this in-migration to region B
which is accounted for by out-migration from region A will be therefore be given by
(MN(PBIPN ) (PAIPN», which gives (MNPAPB)/(PNPN)' as above. Therefore, in equation (6.1) we
can interpret the migration flows between regions A and B as being a product of PA and PB'
and multiplied by a constant G, where G = MNI(PNPN).
The argument so far has implicitly assumed that migration between any pair of regions is
equally likely as migration between any other pair of regions. However, we can argue that
the spatial transactions costs involved under conditions of uncertainty, as discussed in
Chapter 2, suggests that this will not be so. The agglomeration and spatial information
acquisition arguments in Chapter 2 suggest that migration between contiguous areas will be
much more likely than between distant regions. This argument is sometimes known as (dis-
tance deterrence' (Gordon 1978), and implies that the likelihood of migration between any
two locations will be inversely related to the distance between them, given as dAB' However,
there is no reason to expect that the inverse relationship between the interregional migra-
tion probability and the distance should be linear. Therefore we can specify the distance
function in the denominator in terms of (dAB)a, to allow for any non-linearities. Combining
these two distinct approaches in the construction of the numerator and denominator terms
gives us the general expression (6.1). This can be used to provide indications of migration
flows between regions, even under conditions of real-wage equilibrium between regions.
Gravity models can also be made much more complex than this simple description by
introducing more complex behavioural assumptions (Wilson 1974; Isard et al. 1998). At the
same time, the multiplicative nature of equation (6.1) means that where simple models of
this form are used for estimating interregional flows across all regions, the aggregate flows
do not necessarily sum to the total flows in the system. In order to adjust for this, the models
must be (doubly constrained' so as to ensure the correct total flows into and out of each
region (Isard et al. 1998).
In terms of our regional labour market discussions, however, the general observation to
come out of these gravity models is that interregional migration flows are in part spatially
determined, in the sense that the likelihood of migration is a function of distance. This also
implies that interregional adjustments to labour market shocks may also be in part spa-
tially determined. This is because the efficiency of the migration process as a regional
labour market adjustment mechanism will itself depend on the distance between the local
labour market in question and any other local labour market. More central regions, which
are closer to a larger number of other centres of population, may find it easier to adjust to
local negative demand shocks by means of out-migration flows, whereas geographically
peripheral regions may only adjust much more slowly. In other words, the ability ofregions
to successfully adjust to negative demand shocks may also depend simply on the location
of the regions in question.
As well as labour mobility, the gravity model approach is also used for analysing a range
of different flows between places, including trade flows (Anderson and van Wincoop 2003,
2004) and traffic flows.
H\rrE~lR~:GmO~!Al ii'..N"<:>:ii..Il">.P,..2'&<>;' MARKET
Migration may also exhibit something ofa life-cycle nature. For example, young school and
college graduates may tend to migrate towards large primal cities in order to gain better
access to high-quality employment. This migration takes place because such young job-
seekers assume that their best long-term employment prospects will be served by acquiring
a job in such a central location. The majority of their working life may be spent at such a
location, although eventually the worker will seek to move out of the major city to a smaller,
more geographically peripheral, settlement. This may include migration to regions ofhigher
environmental quality and lower wages (Plane 1983). In the dominant urban centres, such
out-migrants will be continually replaced by new young and generally highly educated in-
migrants. Meanwhile, on the other hand, the peripheral areas will consistently see an out-
migration of such young workers and a continuous in-migration ofolder workers accepting
lower wages than they previously accepted. This has been described as an <escalator' phe-
nomenon (Fielding 1992). As long as the generation of high-quality employment opportu-
nities tends to be dominated by the central higher-order urban areas, such a process will
continue indefinitely. These life-cycle effects on migration will tend to take place over and
above the wage-migration mechanism outlined in sections 6.5 to 6.5.3, and their effect will
be to systematically alter the demographic profile and labour force composition between
particular regions. In addition, the nature of these life-cycle effects is conditioned on the
personal migration history of the individual. People who migrated early, or who have previ-
0usly migrated far or frequently, tend to remain more migratory over their life cycle
(DaVanzo and Morrison 1981). The reason is that breaking the psychological ties with the
home location, the attachment value or the sense of place, allows them to more easily
respond to wage signals with migration. As such, they tend to exhibit enhanced lifetime
earnings. In contrast, those who return home at the earliest possible opportunity (DaVanzo
1983) tend to exhibit lower lifetime earnings, precisely because their social ties become
more embedded and reinforced at an earlier stage in their lives, and these become harder to
break over time, thereby reducing the ability of the individual to realize their wage-earning
potential via migration. This is not to say that local social ties are bad because they inhibit
migration, or that migration is bad because it breaks social ties. It is simply important to
note that the decisions to migrate or not to migrate both involve choices and trade-offs
which depend on personal and household preferences as well as the spatial distribution of
employment opportunities.
6.7 Conclusions
Local urban and regional labour markets can exhibit particular features which are some-
what different from national discussions of the labour market. The hierarchical relation-
ships between the regional export base sectors and the locally oriented sectors of a region
will mediate demand shocks, and the regional responses to such shocks will depend on
the structure of these relationships. Where demand shocks are positive, regions can
respond by expanding their local capital stock, either through the expansion of local
investment or through the immigration of capital from other regions. On the other hand,
in some situations, the fixity and durability of local regional capital, and the hierarchical
demand interactions between the regional export base and non-basic sectors, together
militate against any potential downward adjustment of local wages to market-clearing
levels. Both local market clearing and local involuntary unemployment are possible con-
sequences of this downward wage rigidity. The actual result depends on the interaction
between the employment retention policies of local firms, the expectations of local firms,
and the speed of response of external investors to changes in the local economic
conditions.
All regions exhibit the additional labour market adjustment mechanism of interre-
gional migration. As we have seen, there are various interpretations of the relationship
between regional wages and migration flows. The most common assumption is that of the
disequilibrium model, in which migration will take place as a response to real-wage dif-
ferences between regions. If all regional economies exhibit constant returns to scale, the
process of interregional migration will itself lead to a restoration of the interregional wage
equilibrium. Moreover, we have shown that this process maximizes the welfare to society
by reducing any deadweight loss associated with an inefficient interregional spatial pat-
tern of labour. On the other hand, where differences in human capital exist, the process of
migration itself may cause certain regions to experience agglomeration economies at the
expense of others. In this situation, a process of cumulative growth is possible. This is the
subject of Chapter 7.
Finally, over and above all ofthe equilibrium-disequilibrium issues surrounding regional
labour markets and migration, there are certain characteristics of migration flows depend-
ent on demographic and geographical issues, which occur irrespective of regional wage
levels.
Discussion questions
6.1 Is regional unemployment primarily the result of local labour prices being 'too' high?
What would be the various possible consequences of reducing local wages?
6.2 Explain how interregional migration may solve local unemployment problems.
6.3 In what ways is interregional migration related to national economic efficiency?
6.4 How is human capital related to migration and under what conditions may
interregional migration actually exacerbate local unemployment problems?
6.5 What role do environmental amenities play in determining interregional equilibrium
wages? How does this affect our understanding of whether migration is an
'equilibrium' or a 'disequilibrium' phenomenon?
6.6 What other non-wage-related approaches do we have for analysing interregional
migration patterns?
6.7 What economic reasons are there for explaining why most people choose not to
migrate between regions?
6.8 What other economic reasons are there which limit interregional migration?
!!l''i,L''''];!lU'!l''iJl§",~lb AND ~N~rEfH~f~GmON!Ali:!>.ff'*-l&Al%.<?l<.¥1l~ ANAlVSiS
Wagew
W2
WI
/ °2
~
°1
Ll : : L2 Labour demand/Supply
Wagew
SI
S2
WI 1---- 1---7 :
W2------y----~------
°1
Ll L2
.'
Labour demand/Supply
The model of human capital HK investment can be understood in terms of standard dis-
counting techniques applied to investments in general. If we denote any future annual
income stream at time t which can be earned by an investment as Rt, and we denote the
discount rate as i, and the initial cost of undertaking the investment today as Co' the simplest
present value n calculation of the investment can be defined as
n Rt n
The present value of the investment is the discounted sum of all the future income streams
from time period t == 1 onwards. In the model specification given by (A.6.2.1) we are assum-
ing that the future annual revenues Rt are paid at the end of each year, beginning at the end
of year 1. In other words, the revenue payments which are discounted here are discrete
payments.
In the case where revenues are paid continuously, however, it is necessary for us to con-
vert (A.6.2.1) so as to discount the continuous income stream. In this case the present value
n of the investment can be defined (Chiang 1984) as
n
n= f R(t)e-rtdt - Co (A.6.2.2)
o
If we apply this model to human-capital investments, the initial cost of the investment Co
will be represented by the initial employment training costs. These training costs will
comprise the sum of any tuition fees paid plus the opportunity costs of the current income
foregone during the period of training. The income earned from the human capital will be
represented by the wages earned by working in the occupation for which the individual
trained. If we denote the wage earned on commencing employment as W, we can rewrite
(A.6.2.2) as
n= J
o
W(t)e-rtdt - Co (A.6.2.3)
Over the lifetime of employment, wages tend to increase over time as workers become more
experienced and senior in their chosen occupations. In order to allow for the effect of the
growth in wages over time in the present value model, we note that the current wage at any
time period in the future t can be written as
where ais the rate ofgrowth of wages. Therefore equation (A.6.2.3) can be adjusted to allow
for continuous wage growth thus:
f
n
n= W(t)e(a-r)dt - Co (A.6.2.5)
o
The basic model can be further developed to allow for costs which are incurred in a continu-
0us manner over time, and for wage growth which changes over time.
In migration literature, the fundamental issue raised by the model of human capital is
how the relationship between the costs of human-capital investment, as represented
here by Co' and the future wages earned, W(t), are mediated. In particular, the wages
payable for human-capital investments depend on workers moving to the locations of
the appropriate employment. If the market is perfectly efficient, then workers will be
matched with appropriate jobs at all locations, with expected real wages for each occupa-
tion being equal at all locations. As such, workers will be indifferent between alternative
locations, and variations in workers' spatial patterns will depend only on the distribu-
tion of different types of activities and occupations. Following the logic of Borts and
Stein (1964), as we will see in Chapter 7, spatial equilibrium actually implies that spatial
distributions across activities, firms types, and sectors will tend to become more similar
as they converge, the limit of which is that all regions exhibit the same production func-
tion. In contrast, however, if information transmission improves with human capital,
and constraints to migration also fall with human capital, then we would expect differ-
ential migration propensities and variations in market-clearing mechanisms between
different educational and income groups. Migrants with good information will more
readily move to the locations in which the discounted net income stream of their employ-
ment is the highest, as this will maximize their long-term real earning potential, and
there will develop clear spatial sorting according to skills. The human capital embodied
in this selective migration process points heavily towards processes of cumulative rather
than convergent regional growth.
Appendix 6.3 The Roy (1951) model of migration
The Roy (1951) model of migration based on rates of return to human capital was originally
developed in the context of people moving between different countries. The basic Roy
(1951) insight was that absolute wage-earning-Ievel differences only determine the levels of
migration, and do not determine migration types. The types of migrant are determined by
relative inequality, and Borjas (1987) provided a theoretical framework for how such self-
selection mechanisms work.
In order to help explain this insight, we can first consider high-income, but egalitarian,
countries with high progressive taxes such as many European countries, Australia, or
Canada. The rate of return to human capital in Canada or Europe is relatively lower for
high-skilled workers in comparison with the USA. This is because in more unequal socie-
ties, the rate of return to human capital for high-skilled workers can be very high indeed. As
such, high-skilled workers have an incentive to migrate to, or to remain in, more unequal
societies, where their skills are rewarded relatively more highly, while low-skilled workers
have an incentive to migrate away from more unequal societies to more egalitarian societies
(Borjas 2008). Figures A.6.3.1-A.6.3.4 are all constructed from the perspective of the origin
region.
If we take two countries A and B, and country A is more equal than country B, the returns
to human capital for migrants above the skill level SK are relatively greater in country B than
in country A, and for skills sets below SK the returns to human capital are relatively higher
in country A than in country B. If this is the case, then the migrants from country A (say for
example Canada) to country B (say for example the USA) will tend to be primarily high-skill
Wagesw B
I Skills
o SK
• High-skilled migrants
•
Figure A.6.3.1 Movement of high-skilled migrants
Wagesw A
Skills
o K
• Low-skilled migrants
•
Figure A.6.3.2 Movement of low-skilled migrants
migrants with skills levels above SK. This type of migration is described as 'positive' migra-
tion from country or region A to country or region B.
We can also consider a different situation, described in Figure A.6.3.2, in which the origin
country A (for example Mexico) is a more unequal society than country B (for example the
USA). In this case migration from A to B will comprise primarily 'negative' migration of
low-skill workers.
We can now reframe the Roy (1951) model into an interregional migration model in order
to see how wage-income changes.affect the skills composition of interregional migrants.
If the wage income available in region B increases in general relative to region A, such
that the relationship between wage incomes and skills there is now given by B', the mini-
mum skills levels of 'positive' out-migrants from A to B actually falls from SK to SK'. More
people move away from A to B and these increasing outflows are associated with a falling
average skill composition of the migrant. In contrast, if the wage incomes in region B fall
to B", the minimum skills levels of 'positive' out-migrants from A to B rises from SK to SK'"
such that falling outflows are associated with a rising average skill composition of
out-migrants.
Similarly, if the wage income available in region A increases in general relative to region
B, such that the relationship between wage incomes and skills is now given by A', the mini-
mum skills levels of 'positive' out-migrants from A to B actually increases from SK to SK".
Fewer people move away from A to B and these decreasing outflows are associated with a
rising average skill composition of the migrant. In contrast, if the wage incomes in region A
fall to A", the minimum skills levels of 'positive' out-migrants from A to B falls from SK to
SK" such that the increasing outflows are associated with a falling average skill composition
of the out-migrants.
B'
Wage B
r
income B"
A
1
Skills
o SK' SK SK"
Similar arguments can be constructed for the case where the origin region A is more
unequal than the potential destination region B, as depicted in Figure A.6.3.2. In this case
rising incomes in A (or falling incomes in B) lead to lower outflows of <negative migrants'
and a falling average skills level of the out-migrants, while, in contrast, rising incomes in B
(or falling incomes in A) increase the outflows of <negative' migrants and raise the average
skill sets of the migrants.
We can also apply this Roy-Borjas type of logic to the case of interregional migration
with some additional assumptions, and, in particular, the assumption that clustering and
agglomeration have either a levelling effect and a narrowing of inequality, or a segregation
effect and in increase in inequality.
If we recall the issues we discussed in Chapter 2 on agglomeration, we see that core and
dominant cities tend to be prosperous, buoyant, large, and diverse, and as such they also
offer good environments for entrepreneurship, innovation, job searching, job changing, job
matching, knowledge spillovers, and learning. One of the manifestations of these places is
that many highly skilled people compete with one another in the same place, and this com-
petition drives up living costs and lowers real wages. However, workers will still tend to
remain in these locations for Harris-Todaro reasons, as these types of locations maximize
the long-term employment possibilities and the expected wage for any given nominal wage
paid. However, as well as a wage effect, this type of competitive clustering also often has a
social levelling effect on many skilled people. While superstar workers do indeed emerge
from these urban competitions in the form of prominent business leaders, the employment
trajectories of many highly skilled individuals in large urban areas actually tend to settle
down to something of an equilibrium in which their performance and outcomes are little
different from most of their middle-class and suburban-living peers in the same place. It is
AND
B
Wage A'
income
A
A"
i
1
Skills
a SK' SK SK"
true that their employment status tends to exhibit a high nominal wage in comparison with
the national average and also in comparison with the national average of their skills peers.
Similarly, they also exhibit a high overall expected wage by national standards due to the low
likelihood of facing long-term unemployment. In addition, in the USA these areas are also
associated with higher real wages as well as higher nominal wages (Storper and Kemeny
2012). In contrast, however, the high degree of local skills competition implies that many
such urban residents exhibit little or no wage premium in comparison with a large propor-
tion of the local population, and in particular those living in the same neighbourhoods,
while also exhibiting low wages in comparison with the urban elites.
In contrast, for the small number of highly skilled individuals who do find gainful and
stable employment in smaller towns and cities, these people exhibit very high local standing
in terms of both real wages and social status. Such people are often found in occupations
such as law, dentistry, banking, medicine, and education, as well as in business and govern-
mental activities. These people tend to exhibit lower nominal wages by the national stand-
ards of their skills peers but high real wages due to the relatively low local living costs. At the
same time members of local elites often exhibit very high real wages in comparison with the
local population in general, and this is due to the relatively high local monopoly power
afforded by operating in a small and specialized environment which limits the number of
potential competitors.
Local social status and influence are both likely to be part of the overall utility associated
with work, and local social status and influence are also likely to be correlated with local real
wages. If cities do indeed playa social levelling role as described above, this implies that
high-skilled workers from an outlying non-core region A will not wish to move to the core
city-region B, because their overall real income and real social status will fall with such a
iN·rEJU~t:G~O~~AllABOUR ANAlvsms
movement. This is one possible reason, in addition to local attachment value and sense of
place, why widespread migration ofhigh -skilled workers will often not take place from rural
to urban areas, from peripheral to core areas, from small towns to large cities: these types of
migration imply that a 'someone' with a high local status in a small place becomes an anony-
mous 'no one' in a large place. In marked contrast, a low-skill 'no one' in a small place has a
greater chance of becoming 'someone' in an anonymous place, or at the very least remain-
ing an anonymous 'no one' in a large place, but with a higher real wage.
These types of 'melting-pot' urban social dynamics which operate in large cities reflect
some ofthe insights ofa range ofcommentators including Jane Jacobs (1961) and Peter Hall
(1998), among others, and offer the social reasons why many low-skilled people move to
cities to escape the social and economic progression limitations they face due to the monop-
oly or monopsony power of local elites. They also underpin many of the explanations as to
why people moved from Europe to North America and Australasia in the nineteenth cen-
tury. However, the lack ofsocial dynamism and the existence of many entrenched social and
economic monopoly positions also explain why many high-skilled individuals embedded in
locally elite positions do not move to more prosperous locations. The key to these insights is
derived from the social capital arguments outlined in Chapter 7.
On the other hand, if core cities are characterized by increasing segregation and inequal-
ity above the levels evident in non-core regions and rural areas, this itself may deter the
migration oflower-skilled groups from outer areas to core areas while increasing the mobil-
ity of high-skilled groups to the core cities.
In general, these arguments tend not to apply to new university graduates who are highly
mobile precisely because they are not yet sufficiently established in their careers or com-
munity social hierarchies, and move in order to establish their careers. However, these argu-
ments do apply very much more to workers who are well established in their careers and
also in a particular community.
If increasing inequality is also associated with increasing crime, then the picture becomes
even more complex, with crime increasing the degrees oflocal residential segregation along
the lines discussed in Chapter 4, thereby reducing immigration of lower skills groups.
Furthermore, these arguments also have a cyclical effect. If core regions expand first
through the business cycle with peripheral regions responding with a time-lag, as is often
observed, this is reflected in a shift from B to B' in Figure A.6.3.2 or a shift from A to A" in
Figure A.6.3.3. Increasing core-periphery migration is associated with falling skills levels of
migrants. However, as the positive effects of the business cycle continue to expand and
spread, and peripheral regions begin to catch up, this is reflected in a shift from B to B" in
Figure A.6.3.2 or a shift from A to A' in Figure A.6.3.3. Falling core-periphery migration is
associated with increasing skills levels of migrants.
However, irrespective of the stage of the business cycle, what we observe is that the most
skilled people tend not to migrate once they are established in their jobs and communities.
Regional growth, factor
allocation, and balance
of payments
7.1 Introduction
The object ofthis chapter is to discuss the nature of regional growth and to provide an analy-
sis of the various potential mechanisms by which regional growth takes place. Economic
growth is a complex process, and as with labour markets and multipliers, the analysis of this
issue at the regional level is somewhat different from that at the national level. Various hints
as to the possible causes and consequences of regional growth have been provided in
Chapters 1-6. In Chapter 1 we see that growth may take place via the location behaviour of
firms, as the immigration of firms into a region increases the host region's stock of capital
and employment. In Chapter 2 we see that such industrial location and relocation behav-
iour may also contribute to the development of localized agglomeration economies. In situ-
ations where these agglomeration economies arise, growth becomes possible at particular
locations. In other words, growth is location specific. This may have implications for the size
distribution of urban centres, and as we see from Chapter 3, the extent of the localized
growth will also affect local land and real-estate prices. In Chapter 4 it was argued that the
specifically local impacts of localized growth also depend on the sectoral origin of the
growth, and the strength of the linkages between each of the local industrial sectors. Taken
together, these conclusions suggest that the various regional impacts of growth will depend
on both the sectoral and the spatial industrial structure of the economy.
On the other hand, we may initially perceive that the specifically local effects are relatively
unimportant, in that the national or international economy as a whole will benefit from such
localized growth. This is because any localized efficiency benefits will be spread via private sec-
tor trading relationships and also public sector redistributive fiscal mechanisms to the rest of
the economy. However, the effects of localized growth on individual regions may be quite
diverse, depending on the time required for any localized growth effects to be transmitted to
the rest of the spatial economy. As we see in Chapter 5, differences in migration propensities
between individuals with differing human-capital assets may militate against an even and
rapid dissemination of growth benefits to all regions via labour market adjustment mecha-
nisms. Moreover, the efficiency of labour migration as an equilibrating mechanism itself may
depend on the strength ofthe national economy. Similarly, in periods ofrecession, the negative
environmental effects of dereliction associated with the durability of fixed capital in declining
regions may militate against an even and rapid dissemination ofgrowth benefits to all regions
via capital adjustment mechanisms. Therefore, as well as interregional differences in sectoral
and spatial industrial structures, the spatial dissemination of the benefits of localized growth
may depend on the extent to which the aggregate economy as a whole is buoyant.
In order to discuss the various issues associated with regional growth we will adopt two
broad analytical perspectives. The first perspective, which is broadly neoclassical in nature,
focuses on the questions relating to the spatial allocation of production factors, and the
interrelationships which exist between factor allocation and technological change. This is
the most common approach adopted in analyses of regional growth. The second approach,
which is broadly Keynesian in nature, focuses on questions relating to interregional income
flows, and discusses regional growth behaviour in terms of a balance of payments frame-
work. Conceptually, these two frameworks are fundamentally different from each other,
with the result that they produce somewhat different conclusions as to the nature, causes,
and consequences of regional growth. Each approach can throw some light on different
particular aspects of the nature of the regional growth process. However, there is also a
variety of situations in which the two approaches can be made broadly consistent with each
other, thereby providing a wide-ranging perspective on the nature of the regional growth
process. We will begin by discussing the neoclassical approach to regional growth, factor
allocation, and technological change, and in the subsequent sections we will contrast these
arguments with the Keynesian approach.
The neoclassical approach to macroeconomic growth has developed on the basis of the
original inSights of Solow (1956) and Swan (1956). These arguments have subsequently
been applied to the case of regions, and the neoclassical approach to regional growth has
two major components. The first component concerns the question of the regional alloca-
tion and migration ofproduction factors. The analysis of this issue is based on two analytical
frameworks known as the (one-sector' and (two-sector' models of factor allocation, respec-
tively. The second component concerns the question of the nature of the relationship
between production factors and technological change, and this is generally analysed within
a production function framework. The neoclassical growth models assume that the econ-
omy is competitive, in the sense that factors are paid according to their marginal products,
and also that factors are quickly able to be reallocated so as to be employed in their most
productive use. In sections 7.2.1 to 7.2.3 we will initially discuss and compare the two mod-
els relating to the regional allocation of factors, and then we will use the general conclusions
of these models to motivate our production function approach.
The basic principle underlying this comes from the law of diminishing productivity, which
states that, holding one factor constant, the marginal product of the variable factor falls as a
greater quantity of the variable factor is employed. The assumption here is that the variable
factor is combined with the fixed factor in the production process, and it is the application
of the variable factor to the fixed factor which gives rise to the diminishing marginal pro-
ductivity of the variable factor. We will initially discuss the case where all factors are freely
mobile between regions, and then compare this with the situation where there is a certain
amount of interregional factor immobility.
In the case of capital and labour, for a fixed capital stock, the higher the level of labour
employment, the lower will be the marginal product of labour. In other words, as the quan-
tity of labour increases relative to the quantity of capital employed, the lower will be the
marginal product of labour. Similarly, for a fixed quantity of labour, the higher the level of
capital employed, the lower will be the marginal product of capital. In other words, as the
quantity ofcapital increases relative to the quantity oflabour employed, the lower will be the
marginal product of capital. As we see, in the case of two factors, the law of diminishing
marginal productivity holds for either factor, as long as the other factor is held constant.
Moreover, we can extend the argument to more than two factors. For example, if we hold a
third factor constant, such as land, and add successive quantities of both capital and labour,
the marginal products of both capital and labour will fall.
For reasons of analytical simplicity, however, in the following sections we will assume
that all production activities are the result of the combination of two factors. These two fac-
tors are a composite factor capital, denoted as K, which contains all non-labour inputs, and
all labour inputs, denoted as factor L. We assume that in general, capital K and labour L are
complementary inputs, and the relative quantities of capital and labour employed can be
defined in terms of a capital-labour ratio KIL. Using this notation, the arguments above
concerning the application of the law of diminishing marginal productivity to the comple-
mentary factor inputs can be specified in very general terms. If the quantity of capital is high
relative to the quantity of labour employed, in other words the KIL ratio is high, the mar-
ginal product of capital will be low and the marginal product of labour will be high.
Conversely, if the quantity of capital is low relative to the quantity of labour employed, in
other words the KIL ratio is low, the marginal product of capital will be high and the mar-
ginal product of labour will be low.
These arguments can now be translated into a regional context. We can imagine the case
of a country comprising two regions A and B, in which the capital-labour ratio in region A
is higher than that in region B, such that
KA >---
--- KB (7.1)
LA LB
where:
KA quantity of capital employed in region A
LA quantity of labour employed in region A
KB quantity of capital employed in region B
LB quantity of labour employed in region B
In a situation such as that described by equation (7.1), the marginal product of capital in
region A will be lower than the marginal product of capital in region B. Meanwhile, the
marginal product of labour in region A will be higher than the marginal product of labour
in region B. In other words, assuming that production factors are paid according to their
marginal productivities, marginal profits will be higher in region B while wages will be
higher in region A. If factors are mobile, the different regional capital-labour ratios will
encourage labour to migrate from region B to region A, and capital to migrate from region
A to region B. The difference in the regional capital-labour ratios therefore encourages
the two factors to migrate in opposite directions, in order to earn higher factor rewards.
The two factors will continue to migrate in opposite directions as long as there is still a
difference in the regional capital-labour ratios. This process of interregional factor migra-
tion will therefore only cease when the capital-labour ratios in both regions are the same,
such that
KA (7.2)
14 IE
In this interregional equilibrium situation, wages are the same in both regions and marginal
profits are the same in both regions. The interregional adjustment mechanism, in which
factors migrate in opposite directions until capital-labour ratios are equalized across
regions, is known as the (one-sector' neoclassical model of factor allocation and migration.
The conclusions of the one-sector model can be discussed from the perspective of aggre-
gate national efficiency and welfare. In order to do this we can employ an Edgeworth- Bowley
box in which the factor employment levels and output of both regions are represented.
The Edgeworth-Bowley box in Figure 7.1 represents the two regions A and B.The output
of region A is represented by the isoquants which originate at A, and higher levels of output
are represented by isoquants which are further to the right. Similarly, the output of region B
is represented by the isoquants which originate at B, and higher levels of output are repre-
sented by isoquants which are further to the left. The total level of capital in the economy is
KN , and is represented by the vertical height of the Edgeworth-Bowley box. Assuming that
all factors are employed, KN comprises the sum of the capital employed in both regions. In
other words, KN = (KA + K B ). Meanwhile, the total level of labour in the economy is L N , and
is represented by the horizontal length of the Edgeworth-Bowley box. Assuming that all
factors are employed, LN comprises the sum of the labour employed in both regions. In
other words, LN = (LA + LB ).
If the regional factor allocation is initially at point C, the quantity of capital employed in
region A is given by K AC' and the quantity of labour employed in region A is given by LAC.
Similarly, at point C, the quantity of capital employed in region B is given by KBC' and the
quantity of labour employed in region A is given by LAC. With the palticular interregional
factor allocation at C, the level of output of region A is given by the isoquant QA2 and the
level of output of region B is given by the isoquant QB3. As we see in Figure 7.1, the capital-
labour ratio in region A is much higher than that in region B. Therefore, from the logic of
the one-sector model, a reallocation of factors between the regions can effect a Pareto effi-
ciency gain. The reason for this is that a factor reallocation between regions, in which capital
moves from the high capital-labour ratio region (region A) to the low capital-labour ratio
J\ND
1
r r
LN
LBC ~ B
·..x· _ .
KBC
•••• f .
QA6
KN
KAC
~QA2
,,,,. , , .•.....
A :
LAC
:;;,:
~
Figure 7.1 One-sector interregional Edgeworth-Bowley box
region (region B), will increase the marginal productivity of the mobile capital. Similarly, a
factor reallocation between regions, in which labour moves from the low capital-labour
ratio region (region B) to the high capital-labour ratio region (region A), will increase the
marginal productivity of the mobile labour. This process of factor migration and realloca-
tion, in which the marginal products of both mobile factors are increased, must therefore
necessarily increase aggregate national output.
One-sector factor migration will continue until the capital-labour ratios are equal in
both regions, as represented by equation (7.2) above. Once the capital-labour ratios in each
region are the same, the process of factor reallocation and migration will cease, because
there will be no interregional differences in factor rewards. In other words, when regional
capital-labour ratios are equalized there will be no further potential Pareto efficiency gains
associated with the increasing marginal productivities of migrant factors. In other words, all
the possible Pareto efficient interregional factor allocations must exhibit the same capital-
labour ratios. Within an Edgeworth-Bowley box framework, this argument implies that
where all factors are mobile, the contract curve which links all the Pareto efficient interre-
gional factor allocations must be a straight line, as in Figure 7.1.
For example, iffactors which are initially allocated at C are reallocated to a point D, the total
output of region A will remain the same at QA2' but the output of region B will have increased
from Q B3 to QBS. At point D, the capital-labour ratios of both regions are equal. Alternatively,
if the factors which are initially allocated at C are reallocated to a point E, the total output of
region B will remain the same at QB3' but the output of region A will have increased from QA2
to QA4. At point E, the capital-labour ratios ofboth regions are equal, and are also identical to
AND 'IU'*"""'fw1i"'l>.3'11l",,",iIiw
the capital-labour ratios at D. Finally, if the factors which are initially allocated at C are real-
located to a point G, the total output of region B will increase from QB3 to QB4' and the output
of region A will increase from QA2 to QA3' Each of these three possible interregional factor
reallocations represents a Pareto efficiency gain. More generally, beginning at position C, the
reallocation of factors between regions to any point on the boundary of, or within. the area
defined by CDEF represents a Pareto welfare gain with respect to the factor allocation at point
C. However, only points on the straight-line contract curve within this area ofpotential Pareto
efficiency gains, defined as DGE, represent Pareto efficient factor allocations. The line DGE
represents the (core' of the economy, given the initial allocation at C.
The same logic regarding the Pareto gains associated with interregional factor realloca-
tions can also be applied to any other inefficient initial factor allocation, such as points H
and Jin Figure 7.1. In each case, there will be an area ofpotential efficiency gains which itself
must contain a (core' of Pareto efficient factor allocations. The straight-line contract curve
therefore represents all the possible Pareto efficient core allocations. If the interregional
contract curve is a straight line, this also means that the regional expansion paths are both
linear and identical. In other words, the regional production functions are identical.
Assuming constant returns to scale, this implies that both regional production functions
must be homogeneous of degree one.
If the contract curve is a straight line, the interregional production possibility frontier
must also be a straight line, as in Figure 7.2 (Borts and Stein 1964). Pareto efficient points
such as D and E in Figure 7.1 will be on the production possibility frontier, as shown in
Figure 7.2, whereas inefficient points such as C in Figure 7.1 will be inside the production
possibility frontier.
The slope of the production possibility curve is known as the marginal rate of transfor-
mation, and is given by the ratio of the marginal costs of production. In the case of Figure
7.2 we can write this as MRTAB , = (MCA/MC B), where MRTAB represents the marginal rate
Output of
region B
E
QB3
~C "0
QB5
1
1....·......···••·.....·•.......••....."1""......·...·..···... '\ Output of
r LBT
1 8
······r··
KBU
............... ..•..........-........
KN
KAU
KAT
I I ' ..
A : LAT
~ ~:
employed in region A is now higher than the marginal product of capital employed in region
B. By the same argument, if the price of the output of region A increases relative to region B,
this also implies that the marginal revenue product of labour employed in region A is now
higher than the marginal revenue product oflabour employed in region B. In notation:
(a) (b)
SlA
PA2
PAJ
D2A
D2A
D1A D1A
This would suggest that both capital and labour will migrate from region B to region A in
order to earn the higher factor rewards in region A. The effect of this factor migration, in
which both factors move in the same direction, is to move the supply curve of region A to
the right from SLA to S2A' as in Figure 7.4b. The increased supply consequently leads to a
further increase in the output of region A from Q2A to Q3A. However, this increased output
supply also leads to a fall in the price PA of the output of region A from PA2 to PAI. This fall
in the output price will now reduce the marginal revenue product of both capital and labour
employed in region A.
At the same time, as the output of region B falls from QLB to Q2B' the price of the output
produced by region B will have risen due to the contraction in the supply of the goods pro-
duced by region B, associated with the out-migration of both factors from region B. In
Figure 7.5, this is represented by the backwards shift of region B's supply curve for output
goods. This rise in the output price in region B will increase the marginal revenue product
of both the capital and labour still employed in region B.
Therefore, as region A's output prices fall and region B's output prices rise, the marginal
products of capital in regions A and B converge. Similarly, the marginal products of labour
in regions A and B converge. The process of factor migration, in which both factors move in
the same direction, will continue until the marginal products of both factors are equalized
across the two regions.
The actual point at which the two-sector migration will cease cannot be determined with-
out additional information concerning the regional price elasticities of demand and supply.
However, assuming that the initial interregional factor allocation is Pareto efficient, such as
at point T in Figure 7.3, the long-run effect of the 'two-sector' unidirectional factor migra-
tion can be depicted as a shift from point T to point U. Given that the contract curve is
concave, the relative price of capital with respect to labour in both regions will be lower at
point U than at point T. In Figure 7.3 we can see this change in relative factor prices by
observing the change in slope ofthe marginal rate of substitution, which is perpendicular to
the contract curve. However, for both capital and labour, the marginal factor products and
factor rewards will be equalized across both regions.
S2B
SlB
P2B
P1B
D1B
Q2B Q1B
....................................
Q2B
Q 1B ..................................................................
In the two-sector model of interregional factor allocation, the two regions are assumed to
produce different products with different production functions. The production of one
region, region A, is capital intensive and the production of the other region, region B, is
labour intensive. This implies that the marginal rate of transformation of production
between the two regions is constantly changing according to the level of output in each
region. This is represented in Figure 7.6 by a concave production possibility frontier, in
which a movement along the interregional contract curve from T to U is represented by a
movement down the production possibility frontier.
The fact that the contract curve in the two-sector model is curved implies that the real-
location of factors from one region to another changes the relative rates at which the output
of one region can be expanded while the output of the other region can be contracted, due
to the factor reallocations. In other words, the marginal rate of transformation between the
regions is always changing, and this is why the production possibility frontier is also convex
to origin, as discussed in any intermediate microeconomics textbook.
LN
~ ~
,1
82
.
L1KN
81
.................................................
KBV
KN2
...... KN1
...•...
KAV
............
.•...••••.•.
./ .....
............ A
~ LAV f LBV ~
Figure 7.7 Edgeworth-Bowley box adjustment from a two-sector to a one-sector interregional model
Output of
region B
PPF2
Output of
region A
Figure 7.8 Production possibility frontier adjustment from a two-sector to a one-sector interregional model
effect by an extension to the vertical dimensions of the box. The extent of the capital expan-
sion which occurs with the process ofregional economic integration is represented in Figure
7.7 by i1KN == (KN2 - KN1 ), and the level of capital growth associated with this trade creation
effect is given by (i1KN /KN1 ).
The transition process associated with the regional economic integration therefore has
two main features. The first feature is the interregional reallocation of factors according to
the principles of the one-sector model, which leads to a general flattening of the contract
curve, such that the capital-labour ratios tend towards being equalized across both regions.
The second feature, concomitant with the factor reallocation, is an expansion in the capital
stock of the area of economic integration. The combination of these two integration effects
is assumed to generate regional economic growth. In Figure 7.7 the change in the contract
curve associated with the combination of these effects is represented by the transition in the
contract curve from AB 1 to AB2• The effect of these changes in the contract curve can also be
represented by changes in both the shape and the position of the interregional production
possibility frontier. As we see in Figure 7.8, the process of regional economic integration
encourages the production possibility frontier both to shift outwards from its initial posi-
tion PPF 1, and to become flatter. The long-run result of this process is that the interregional
production possibility frontier will become linear, as represented by PPF2 in Figure 7.8.
Even in the absence of a growth in the labour stock, regional growth therefore comes from
two different sources. These two sources are the interregional reallocation of existing factor
stocks, and the increase in capital stocks associated with any trade creation effect. The long-
run outcome of this one-sector regional integration process is a tendency towards regions
with similar production functions and similar capital-labour ratios, in which regional rates
of return to capital will converge, as will regional wage rates. As such, comparative advan-
tage between regions tends to disappear as an explanation for regional production behav-
iour, and is superseded by explanations based on factor mobility.
The process of one-sector regional economic integration and factor reallocation described
here forms the basis ofmany assumptions about economic growth in areas currently undergo-
ing economic integration. An example of such an area is the European Union. In the case of
the Europe Union, the separate national economies have become progressively more inte-
grated over the last half-century. This integration process has involved the progressive reduc-
tion of border tariffs and the removal of restrictions to trade and factor migration. This
integration process was given an additional spur at the beginning of the 1990s, with the intro-
duction of a common EU passport system, which allows for the free migration of labour
between all EU nations for reasons of employment. Such institutional arrangements ought to
allow for a one-sector reallocation of factors across the EU, as well as some potential regional
trade and capital creation effects. If this one-sector argument is indeed correct, over time we
should observe a tendency towards regional convergence within the EU. Evidence supporting
this one-sector argument was first provided by Barro and Sala-i-Martin (1992, 1995). They
suggested that the level of dispersion across the EU regions of real income per head had fallen
over time, a process which they termed <(J convergence'. Barro and Sala-i-Martin also found
evidence to suggest that there is a negative relationship between the rate of growth in income
per head and the initial level of income per head, a process which they term <p convergence'.
Although there has been much debate as to the appropriateness ofthe data employed by Barro
and Sala-i-Martin and the interpretation of their results (Cheshire and Carbonaro 1995, 1996;
Fingleton and McCombie 1998; Martin and Sunley 1998; Button and Pentecost 1999; Durlauf
and Quah 1999), these tests of (J and p convergence are primarily motivated by the theoretical
conclusions of the one-sector model of interregional factor allocation and growth.
The above sections lead to the general neoclassical conclusion that regional integration pro-
cesses will lead to a one-sector reallocation of factors across regions. The long-run implica-
tions ofthis process will be that all regional production functions will tend to converge, such
that regional capital-labour ratios will converge across regions, as will regional capital
returns and regional wages, and regional expansion paths will also all be linear. Output
growth will increase as factors are allocated more efficiently, and this process itself may
generate additional growth via trade creation effects. Analytically these conclusions are use-
ful because, at least in principle, they allow us to model the sources of regional growth
within a rather straightforward production function framework. To do this we can employ
a Cobb-Douglas production function, which is defined as
Q t = AK a IJ3 (7.3)
where:
The Cobb-Douglas production function has two useful properties. The first property is
that the factor shares are represented by the relative contributions of profits and wages
to the total factor payments in the economy. In the Cobb-Douglas production function
these shares are represented by a and p, respectively. If the factor shares are approxi-
mately constant, this also implies that the capital-labour ratios are approximately con-
stant. The second property of the Cobb-Douglas production function is that if a + PI,
production exhibits constant returns to scale, in terms of the relationship between the
total output produced and the total quantities of input factors employed. A given quan-
tity of capital and labour will produce a given quantity of output, the value of which is
defined as a constant multiple A of the total value of the inputs. Therefore doubling the
quantity of both factor inputs employed will simply double the total level of output pro-
duced, ceteris paribus. The relationship between the level of output and the level of factor
inputs is therefore independent of the total quantity of inputs employed or outputs pro-
duced. If this is so, it implies that the sum of the indices a plus p must equal one. In other
words, p = (1 - a). Our Cobb-Douglas model (7.3) must therefore be modified
accordingly.
A second modification required to equation (7.3) concerns the question of time. Over
time, the relationship between total output and inputs is not static, in the sense that new
production techniques and technologies become available which increase the efficiency of
the production process. The adoption and implementation of these new production tech-
niques and technologies is known as (innovation', and this process of innovation means
that over time the level of output increases for any given stock of factor inputs. For our
purposes, we will define this process of applying new techniques and technologies under
the general heading of (technology'. As such, technology represents the sets of production,
organization, information, and communications blueprints which are available to all
firms, and which mediate the relationship between the input factors employed and the
outputs produced. We denote the level of technology by the technological index 1.
Assuming that the level of technology increases over time, we can incorporate a simple
technological trajectory e4Jt into the Cobb-Douglas function which allows for increases in
technology over time t.
Our modified Cobb-Douglas function, which incorporates both constant returns to scale
and technological change over time, now has the form
Qt = Aet/JtKaL1- a (7.4)
The one-sector interregional factor allocation model discussed in the previous sections
implies that all regions will converge towards the same production function with the
same constant capital-labour ratios. In the Cobb-Douglas function (7.4) the constant
PAYMENTS
capital-labour ratio is given by (all a). Therefore, assuming that aggregate regional
production across markets and industries can be regarded as perfectly competitive, the
process of interregional factor reallocation should lead to all regions exhibiting the same
Cobb-Douglas production function. In other words, if we can model the production
function of one region, we can model the production function of all regions within the
same economic system.
With this production function methodology we are now able to consider how the growth
of regional output is related to changes in the various inputs to the production process. In
order to convert our regional production function into a model of regional growth we can
convert equation (7.4) into natural logarithms and then differentiate with respect to time.
The details of this are given in Appendix 7.1. By these steps we can convert equation (7.4)
into a regional growth accounting expression:
where Qt, Kt, it, represent the rates of growth of output, capital, and labour at time t,
respectively.
This growth accounting expression (7.5) states that the rate of growth of regional output
at time t is the sum of the rates of growth of the input factors (capital and labour), weighted
according to their relative contributions to the economy, plus the level of technology r/J. In
these growth accounting terms, the level of technology represents the contributions to
regional growth which cannot be accounted for simply by changes in the optimally com-
bined stocks of regional capital and labour. As such, the term r/J is sometimes referred to as
the 'Solow residual' or the 'growth of total factor productivity'. As we see in Appendix 7.1,
the growth accounting methodology of equation (7.5) can be shown to predict that, in gen-
eral, wage growth depends on the growth in the capital-labour ratio and also the level of
technology. Moreover, in a long-run steady-state situation in which profit rates are con-
stant, wage growth depends simply on the level of technology.
One of the features that we often observe, however, is that the level of technology is not
ubiqituous, nor does it necessarily spread quickly or evenly. As we see in Box 7.1, technol-
ogy diffusion is often seen to exhibit an 'S-shape' over time. Shortly, we will also apply these
ideas to geography, where similar results are often observed.
The major movement beyond the traditional growth accounting-type framework was
provided by the 'endogenous growth theory' models, initially associated with the work of
Romer (1986, 1987a) and later by Lucas (1988), and involves accounting for growth within
an orthodox neoclassical growth accounting framework, while at the same time dispensing
with the need for an exogenous technology residual (McCombie and ThirlwaIl1994). The
approach assumes that increasing specialization increases output, and as such, output is
defined as a function of the number of units of specialized capital goods, rather than simply
as an aggregate capital stock. Under certain assumptions concerning monopolistic compe-
tition and the role of product variety, in a manner similar to NEG models of Chapter 3,
Romer shows that the production function can be expressed as
Q t = AIJK (7.6)
and transforming equation (7.6) into a logarithmic form and differentiating with respect to
time gives
Ot = fJi t + Kt (7.7)
where p = (I - a) as above. The argument here is that all growth is accounted for in terms
of the growth of inputs, with they key issue being the level of specialized capital inputs and
the associated benefits of labour specialization. As we have seen in Chapter 3, one of the
arguments underlying the agglomeration model is that of increasing location-specific spe-
cialization. If such specialization is indeed place specific, then the endogenous growth
model implies that the benefits of this growth will also tend to be localized.
Romer (1987b) discusses a second potential source of endogenous growth, which is the
stock of knowledge. In order to account for this we can write a production function as
Qt = !(K,L,E)g(N) (7.8)
The increasing returns are external to the firm, and this ensures that a competitive equi-
librium is preserved (McCombie and ThirlwalI1994). In growth rate terms equation (7.9)
becomes
If (a + VJ) is equal to one, growth will be constant. However, if (a + VJ) is greater than one,
growth will be continuously positive and cumulative, and if (a + VJ) is less than one, growth
will decline.
Both of these Romer models conclude that the portion of output growth which would
be considered as a technology residual in the neoclassical model can be attributed
entirely to capital acquisition. In the former case, this is because knowledge growth is
assumed to increase directly in line with the level of specialized capital stock, whereas in
the latter case it is because of the assumption that knowledge increases with the level of
capital inputs.
Lucas (1988) also discusses knowledge inputs, but instead the focus of his model is on the
level of human capital, rather than firm -specific capital. In his approach, we assume that
workers spend a fraction of their time u acquiring human capital H. Following our discus-
sion in Chapter 6 we assume that human capital increases the productivity of the individual
person. However, Lucas assumes that this (internal' effect H also has an (external' effect J,
which benefits all other workers. With these assumptions we can write our production func-
tion as
If we also assume that the external human-capital effect J is equal to the internal human-
capital effect H, equation (7.11) can be rewritten as
Qt = (uH(JL)l-a K a (7.12)
where () = (1 a + y)/(I - a). In order to make growth endogenous, this model requires us
to define the growth of human capital as
d~ = H Pv(1 - u) (7.13)
dt
where p and v are constants, with p being greater than or equal to one, such that there are no
diminishing returns to the generation of human capital. If we take the simplest case where
p is equal to one, the rate ofgrowth ofhuman capital defined by equation (7.13) is a constant
A. This allows us to rewrite equation (7.12) as
Qt = (uLqeAt)l-a K a (7.14)
where L q represents the number of units of labour at a given level of efficiency and quality
and is given by Lq = HOL. As such, a given number of units of labour of increasing human
AND S..il'5I''''Ii!.lw5l'·1l.?·'f .... ",.,
This model concludes that the portion of output growth which would be considered as a
technology residual in the neoclassical model can be attributed entirely to labour through
human-capital acquisition.
These various models of endogenous growth provide different insights into the possible
sources of cumulative growth. For our purposes, we can relate each of these potential
sources of cumulative growth to spatial arguments (Nijkamp and Poot 1998). The Romer
models suggest that endogenous growth can arise due to an increasing variety of specialized
capital goods, or an increasing knowledge base associated with these capital goods and the
associated externality effects of information spillovers. Meanwhile, the Lucas model sug-
gests that endogenous growth can arise due to private investments in human capital, the
benefits of which also spill over to the surrounding environment. However, identifying
empirically whether or not localized growth takes place, and isolating the sources of such
growth, is very difficult because, by definition, such effects are positive externalities. In
order to circumvent these problems, various indirect methods have been employed to
empirically measure systematic growth differences across regions according to location-
specific technology effects. Analyses have attempted to examine if there are any systematic
differences in the spatial extent of information flows by testing for variations in the spatial
distribution of patent citations (Jaffe et al. 1993), variations in the spatial distribution of
research and development activities (Iammarino and Cantwell 2000), or variations in the
spatial distribution of technology-related infrastructure such as universities (Acs et ale
1992). Most of these indirect methods provide support for the arguments relating to
agglomeration externalities, in that there are various localized technological effects which
do not diffuse quickly across space.
A weakness of aspatial endogenous growth theory, however, is that without diminishing
returns to either capital or human-capital accumulation, growth will be implausibly explo-
sive. If we apply this logic to regional development, it would imply that all activities will
converge to one single location. However, as we saw in Chapter 2, the spatial nature of the
economy can provide brakes on any such explosive process. This is because providing for
spatial markets inherently involves the problem of overcoming space. Moreover, conges-
tion effects are always associated with industrial clustering in space, and beyond a certain
point, these negative externalities can work against cumulative clustering. The balance of
these positive and negative externality effects will lead to factor reallocations in which long-
run real returns to factors across space will still tend towards equalization.
Although endogenous growth agglomeration models are not consistent with all develop-
ment becoming localized at a single point, they are, on the other hand, perfectly consistent
with the notion that new innovations may persistently tend to originate at the same loca-
tions' even in an environment in which markets are broadly competitive. As we saw in
Chapter 2, this type of argument comes from the application of product-cycle arguments to
the agglomeration models. The localized positive externality effects in the areas in which the
~ ~'<: %# ~1rK ~~. n'~ ~~,§ R~'" ~'" &,.,.,>' AN [}
innovations originate will generate appreciations in local real-estate prices, such that, in the
long run, real returns to factors will tend to be equalized across regions. On the other hand,
however, a permanent disequilibrium in nominal factor returns is perfectly possible, in
which certain innovative regions always exhibit higher nominal factor prices. These areas
will tend to be the central dominant cities and regions of the spatial economy, while the
more geographically and economically peripheral regions of the economy will tend to
exhibit lower nominal, but much more equal, real factor returns. Short- or medium-term
localized growth effects in space are therefore perfectly consistent with an underlying one-
sector model of interregional factor allocation. However, long-term localized growth is not
consistent with a one-sector model of regional factor allocation, because geographically
localized congestion costs will become evident. At the same time, as we see from the loca-
tion models discussed in Chapters 1 and 3, systematic centre-periphery differences in nom-
inal returns to factors can be perfectly consistent with a one-sector model of factor
allocation.
regional expenditure is constrained by the level of regional export income earned. This is the
basis of the Keynesian approach to regional growth, which posits that regional growth is
constrained by a regional balance ofpayments constraint. While there are some commenta-
tors who would question the validity of a balance of payments discussion applied to regions
(Richardson 1978), the fact that we attempt to discuss interregional trade patterns means
that there must be equivalent interregional income and capital flows (Dow 1982, 1987,1997;
Hess and van Wincoop 2000; Ramos 2007; Crocco et ale 2010). Moreover, from a Keynesian
perspective, the justification for such an analytical approach is that these income flows may
themselves have additional monetary effects, which are in addition to any real-income
effects associated with spatial factor adjustments.
This argument that interregional income flows may themselves have additional monetary
effects, which are in addition to any real-income effects associated with spatial factor adjust-
ments, is well understood by observers of the varying fortunes of the different members
countries of the eurozone in the aftermath of the 2008 global financial crisis. Previously
solvent nations with low public deficits and debts such as Spain, and high-growth and
dynamic countries such as Ireland, were caught in a downward spiral of capital outflows
and rising interest rates. These crises were primarily due to earlier excesses in real-estate
bank lending to developers in the prosperous and rapidly growing regions of their coun-
tries, where price expectations had been overshooting (see section A.4.3.2). The result ofthis
was that profitable business in these previously prosperous regions, as well as firms in other
regions and also in other countries, all started to face increasingly severe credit constraints
which were largely unrelated to the performance of the individual firm. The real-estate asset
sales required by many investors to repay their loans then had further adverse effects on the
capital borrowing ability of other firms in the areas where the sales were concentrated.
Spatial factor adjustments are clearly not the only determinant of regional performance.
Monetary flows are also important.
The first three terms on the right-hand side of equation (7.16) represent the components of
aggregate demand associated with the domestic activity within the regional economy, and
we can group these under the heading of <regional domestic absorption' A r• Meanwhile the
last two components on the right-hand side of equation (7.16) represent the components of
aggregate regional demand associated with the interregional traded sector. In general terms,
from (7.16) we can write
where (Yr - A r) is equal to the net acquisition of assets from other regions.
~JRBAN AND LUI'~,2~"illMll~ ECONOfvHC AND
In order to see why the difference between regional income and regional domestic
absorption is the net acquisition of assets from other regions) we must begin with a discus-
sion of a balance of payments model at a national level, and then translate this argument to
the case of regions. At a national level the simplest balance of payments model can be
defined as
where:
Ifwe rearrange equation (7.18) to give (7.19) we can see that if the left-hand side of(7.19) is
positive) the country is regarded as running a balance of payments surplus) and if the left-
hand side of equation (7.19) is negative) the country is regarded as running a balance of
payments deficit. If a country is running a balance of payments surplus) it must be either
increasing its stock of foreign assets) or alternatively reducing its indebtedness to foreign
citizens. On the other hand) where a country is running a balance of payments deficit) it
must be either reducing its stock of foreign assets) or alternatively increasing its indebted-
ness to foreign citizens. These wealth adjustments are mediated via transactions in the inter-
national currency markets.
AND l'l."'<!"l'. >l.. S',",," l'l'll ""- .• !t:.
In the case of interregional trade, because all transactions within a common currency
regime are denominated in the same currency, there can be no official financing as such.
Moreover, we know that regions do not have customs or trade barriers. However, in
principle, we can still write a balance of payments expression for a region, the key feature
of which is that the right-hand side of equation (7.19) must always be zero when applied
to regions (Ramos 2007). In interregional terms therefore, our balance of payments
expression must be
CAR + KA R = 0 (7.20)
where:
which rearranges to
In other words, the net surplus in a region's trade in goods and services with other regions,
given by (Xr - M r) in equation (7.17), is balanced by the region's net acquisition of assets
from other regions, given by (Yr - A r) in equation (7.17). For example, if a regional export-
ing industry is very successful, this implies that the income generated by the exports can be
used both to import goods and services from other regions and also to buy more assets in
other regions. These asset purchases will include real-estate assets in other regions as well as
share acquisitions in firms located in other regions. Similarly, ifa region is running a balance
of payments deficit, it must be financed by net sales of domestically held assets to buyers
from other regions. If a region is running a balance of payments equilibrium, in means that
the net acquisition of assets from other regions is zero. The problem with balance of pay-
ments surpluses or deficits is that they cannot continue indefinitely. In particular, ifa region
is experiencing a balance of payments deficit, there will only be a finite stock ofdomestically
held assets and properties within the region which can be sold to external buyers in order to
finance the regional deficit. Therefore a region cannot maintain a long-run balance of pay-
ments deficit. This implies that, as well as the level of domestic absorption, the level of
regional income that it is possible to maintain in the long run depends on the level of
regional exports.
In order to see this we can consider a country comprising two regions A and B, whereby
region A exhibits weak investment demand while region B exhibits strong investment
demand. By definition, regions do not have any independent control over monetary issues,
as these are determined by the central banking authorities. Therefore the rate of interest
prevailing in anyone region is equal to the rate of interest in any other region. As such, we
can regard the individual region as a small open economy in which the LM curve is
horizontal.
In Figure 7.10, the rate of interest in both regions is given as r*. At this rate of interest, in
region A the local investment component of regional domestic absorption is only sufficient
Rate of Rate of
interest interest
ISA ISB
r/ _ _ .
ISA ISB
~ FB
YA YFA
Region A Region B
to generate a regional income level of YA. If the investment income required to generate full
local employment is given by Y pA , the local shortfall in labour market demand, expressed in
terms of income, is given by (YPA - YA ). Meanwhile, in region B, the level oflocal investment
is just sufficient to generate a level of regional income which clears the labour market at the
prevailing rate of interest. Therefore a simple comparison of the differences between the
situations in the two regions would suggest that the labour market in region A would clear
if there were a fall in the local interest rate from r* to r/, whereas for region B, if rates fell
below r*, the local labour supply constraints would lead to localized inflation. However,
without an independent currency we know that, by definition, regions do not have the abil-
ity to adjust local interest rates. Therefore, if the monetary authorities presiding over both
regions are charged with maintaining price stability, they will ensure that interest rates are
kept at a level of r* so as to maximize the total employment level in the two regions. This will
mean that the buoyant region maintains full employment whereas the depressed region
exhibits a local labour demand shortfall.
One way of alleviating this labour market imbalance is for unemployed labour to migrate
from region A to region B. However, as we see in Chapter 6, there are many conditions
under which interregional labour market adjustment processes are rather inefficient. In the
absence of local currency and interest rate adjustments, if a region cannot generate internal
investment levels I R *, the only other mechanism a region has for expanding is through an
expansion in its exports. This is explained by Figure 7.11, which represents a fixed-price
model of the regional macroeconomy without currency or interest rate adjustments, the
upper right-hand quadrant contains our familiar income-expenditure diagram.
In this particular case, we are assuming that the level of internally generated regional
domestic investment I R * is constant, as are the externally determined regional interest rates
r*. By reading from right to left in the upper left-hand quadrant of Figure 7.11, we can
observe the inverse relationship between regional domestic investment and interest rates.
IR AD
I *
......................R•••••••••••.•
r* :"If\'1 / ; : I Y
i.v ., y
YI Y2 Y
r r*
' ......
". IS2
••.•••••••... 1 I I 151
.....
....
..•...
.....
.....
The lower left-hand quadrant simply plots interest rates against interest rates, and the lower
right-hand quadrant represents the individual regional investment market as in Figure 7.10.
As we see in the income-expenditure model, as regional exports increase, we can model the
resulting income increase for a given level of domestic investment I R * and interest rates r*
by adjusting the origin and x-axis vertically downwards by exactly the amount of the export
increase AX. The increase in exports provides for an increase in regional income from Y 1 to
Y2 , which at the prevailing rate of interest implies that domestic regional investment
increases from IS 1 to IS2 • At current interest rates, the increase in regional exports therefore
generates increases in regional investment due to a general increase in regional income. This
possibility has already been raised in our discussion of an export base 'super multiplier' in
Chapter 4. In Keynesian regional models, the level of regional exports consequently plays a
key role in determining the level of domestic investment which is sustainable, in situations
where local currency and interest rates adjustments are not possible.
What we have not yet considered, however, is the question of what determines the level
of regional exports, sustainable investment, and income in the long run. In order to discuss
nL'\,..2:'~~iJ$P~~!L E(ONOMIC H"~'%.Qi&.<",,&>.,I6J
this, we can write a simple general long-run regional import demand function as (Thirlwall
1980; McCombie and Thirlwall1994)
M
r
aY1r(~Le]f.1
r P
(7.22)
r
where:
M r regional imports
Yr regional income
J[ regional income elasticity of demand for imports
Pf nominal price of goods produced in other regions
Pr nominal price of goods produced in the domestic region
e exchange rate
f1 price elasticity of demand for imports
Xr bZ E ( fJ r
]T7 (7.23)
ePf
where:
Xr regional exports
Z rest of the world income
£ world income elasticity of demand for exports of region r
y/ price elasticity of demand for the exports of region r by rest of the world
These import and export demand functions simply say that the level of imports and exports
depends on both the price and income elasticities of the goods, as well as on the relative prices
ofdomestic and externally produced goods, subject to the respective exchange rate movements.
Ifwe transform these two functions into natural logarithms and then differentiate with respect
to time, in a manner analogous to sections 7.3 and 7.3.1, we can derive expressions describing
import and export growth rates, respectively. Our import growth rate expression becomes
ti + (1 + 1] + p)[Pr - e]
(7.27)
Yr
TC
When these Keynesian (or more strictly post-Keynesian) types of balance of payments
models are applied to the case of regions, it is generally assumed that the relative price effects
contained in the square numerator bracket of equation (7.27) are relatively unimportant.
There are three major reasons for these assumptions. The first reason is that regions do not
exhibit the capacity to make independent currency adjustments. Second, it is assumed that
most prices are set in oligopolistic industries, which ensure relative price stability between
competing producers, even in the face of cost changes (Lavoie 1992; Davidson 1994). Third,
it is assumed that geographical transactions costs and spatial competition mean that differ-
ences in nominal prices between regions also remain relatively stable over long periods.
Under these assumptions, the long-run regional equilibrium balance of payments expres-
sion reduces to
£Z Xr (7.28)
Yr
TC TC
In other words, the maximum balance of payments constrained long-run growth rate of a
region is equal to the long-run growth in world income, multiplied by the ratio of the world
income elasticity of demand for the exports of the region divided by the regional income
elasticity of demand for imports. This in turn is equal to the long-run rate of growth of
regional exports, divided by the regional income elasticity of demand for imports.
The long-run growth ofregional income is therefore determined by the ratio ofthe income
elasticities of demand for the region's exports and for its imports. This depends on the quali-
tative mix of production sectors in a region. If a region is dominated by the production of
high-value-added, highly income-elastic and low- price elasticity goods, its export growth
will tend to be consistently strong over time, for any given patterns of regional imports.
Similarly, if a region is dominated by industries with strong local linkages, its import growth
will tend to be relatively low over time, for any given pattern of exporting. A combination of
highly income-elastic exports and a low regional income elasticity of demand for imports
will therefore tend to allow a high long-run level ofregional growth, even allowing for the fact
that growth may be constrained by a balance of payments constraint. 1
1 The 'Thirlwall's law' argument assumes that there is hysteresis in terms of the export sales and import demand
functions ofthe regions due to the industrial, technological, and institutional history of the region. A closely related
argument which reverses the causality in these models comes from the '45 degree rule' of Krugman (1989), whereby
population growth influences the variety of output. See McCombie and Thirlwall (1994) for a detailed discussion of
the relationship between the two approaches.
7.4.2 The Verdoorn law and cumulative causation
The final component of Keynesian or post-Keynesian regional growth theory concerns the
question of economies of scale. In this approach, the analysis of economies of scale centres
on the so-called <Verdoorn law', which posits a positive relationship between the rate of
growth of labour productivity and the growth of output. The Verdoorn relationship is
given by
p = a + bQ (7.29)
where p represents the rate of growth of labour productivity, and Q represents the rate of
growth of output. Based on empirical observations, the <Verdoorn law' assumes that the
value of a is approximately 2 per cent, and that the value of b, the Verdoorn coefficient, is 0.5
per cent. These values can be shown to be broadly consistent with a neoclassical production
in which the indices a plus b sum to 1.33 (McCombie and ThirlwallI994).
Ifwe use the notation employed above in sections 7.3 and 7.3.1, we can rewrite equation
(7.29) as
(0. - i) = a + bQ (7.30)
Initial observation of equation (7.30) suggests that econometric estimation of the relation-
ship posited by equation (7.29) will exhibit a simultaneity problem, because the term repre-
senting the rate of growth of output appears on both sides of the equation. While the
treatment of this issue has been the subject of much debate (Kaldor 1975; Rowthorne 1975;
McCombie and Thirlwall 1994; Scott 1989), the general assumption in post-Keynesian
models is that the direction ofcausation is explicitly from right to left (Felipe 1998). In other
words, increasing output growth is regarded as engendering dynamic economies of scale in
production, via both <learning by doing' effects on the part of labour (Arrow 1962), and also
the increased capital accumulation effects associated with easy credit availability in condi-
tions of expanding output. If the assumption of Verdoorn dynamic economies of scale is
now included in our discussion of regional balance of payments constraints, by following
the diagrammatic approach of Dixon and Thirlwall (1975) we can indicate the various
regional growth trajectories which are possible under a regime of cumulative causation.
In Figure 7.12, we observe a set of conditions which gives rise to a constant regional out-
put growth rate. In the upper right-hand quadrant we see that the regional export growth
rate is x, and with an income elasticity of regional demand for imports given by 7[}, this leads
to a balance ofpayments constrained output growth rate of q. Via the Verdoorn effect, in the
upper left-hand quadrant we see that this output growth engenders local labour productiv-
ity growth of h. In the lower left-hand quadrant we see that this itself leads to quality-
adjusted real-price reductions of regional output, which fall at a rate of s. As we discussed
above, in these models we assume that relative prices remain more or less the same across
regions. However, for given output prices, labour productivity gains will be realized in
terms of real quality improvements. Moreover, these regional output quality improvements
will be transmitted in the lower right-hand quadrant to increases in regional export growth
x, the actual extent of which will depend on the income elasticity of demand for the region's
Q
.........
......
•••• 1
.. i
;;::.~.•..••.......
...........
.....1 a .....
h x
p
:: x
............................. _.... ....•.••••..•.....•..•.........•.
,.......................... s .
(-p)
exports £]. In this particular case the export growth will itself lead to a steady-state regional
output growth rate of q. As we see from Figure 7.12, however, there is no particular equilib-
rium steady-state rate of regional growth towards which the region will converge.
In these Keynesian type models, there is no reason why steady-state regional growth
should be automatic. For example, ifa region is characterized by a dense clustering of indus-
tries which exhibit agglomeration economies, the region will tend to produce highly inno-
vative outputs and will also purchase large quantities of its input requirements from within
the local regional economy. This will imply that the income elasticity of demand for the
region's outputs will tend to be greater than under the case of steady-state growth, and also
that the regional income elasticity of demand for imports will be relatively low. In the case
of imports, we can compare this situation with that of the steady-state growth rate of Figure
7.12, by shifting upwards the line in the upper right-hand quadrant of Figure 7.13, which
represents the inverse of the income elasticity of demand for imports, from lin] to I/n2•
Similarly, in the case ofexports, we can shift outwards the line in the lower right-hand quad-
rant of Figure 7.13, which represents the income elasticity of demand for the region's
exports, from £] to £2. As we see, in such a set of circumstances, the combination of highly
income-elastic exports, a low income elasticity of regional imports, and increasing returns
to scale can give rise to cumulative growth. The actual rate of regional growth depends on
the particular values of the regional import and export elasticities.
Similarly, we can envisage the opposite type of situation, in which a region is dominated
by the production of relatively low income elasticity exports, while at the same time being
very dependent on imports. For example, this type of situation could occur in a relatively
low-demand peripheral region which has suffered severe industrial decline and the loss of
Q
~
I
\~I
: .••....
'. I
..... b II ..'
". I .........
.. ..
'
'
.. '
x
:~
p
)
..............
(-p)
many local firms, and which has subsequently experienced employment growth in estab-
lishments which are relatively 'footloose). In this case, the level of expenditure in the local
economy by both the new and old firms will tend to be very small. Moreover, ifboth the new
and old firms in the region are specialized in the production of rather standardized 'mature'
products, the income elasticity of demand for the region's exports will be very low.
In the case of imports, we can compare this situation with that of the steady-state growth
rate of Figure 7.12 by shifting downwards the line in the upper right-hand quadrant of
Figure 7.14, which represents the inverse of the income elasticity of demand for imports,
from l/n j to 1/n3 • Similarly, in the case of exports, we can shift inwards the line in the lower
right-hand quadrant of Figure 7.13, which represents the income elasticity of demand for
the region's exports, from £j to £3. As we see, in such a set ofcircumstances, the combination
oflow income elasticity exports, a high income elasticity ofregional imports, and increasing
returns to scale can give rise to cumulative decline. As above, the actual rate of regional
decline depends on the particular values of the regional import and export elasticities.
These Keynesian and post-Keynesian approaches to regional growth differ fundamen-
tally from neoclassical models in their basic assumptions. In particular, these models do
not require the assumption that factors are paid according to their marginal products. Nor
do they require the assumption that production exhibits constant returns to scale with
respect to input factors. However, in a similar manner to models of endogenous growth,
these models imply that there is no particular long-run rate of growth towards which a
region is expected to converge. The actual regional growth rates will therefore depend on
the extent to which agglomeration economies or diseconomies are operative (McCombie
and Roberts 2007) and also on careful interpretation ofwhat the underlying agglomeration
AND lW1'~""~8'11l'%,,,L
Q
1
1
1
1
•\ ..~1 ~
.. b 1 •••
••••
- : 1
.
••••••
:
_
.
•'. 1 •••••
••••• 1 Q •••••••••• 11K]
.. .........
p
............................ x
•• e.
................ C] •••••••••••••••••••
.
.....................
(-p)
y
' ....
model formulations actually imply (Felipe and McCombie 2012). Moreover, in terms of
regional growth these Keynesian and neoclassical models can be shown to produce largely
equivalent results, such that the interpretation of empirical observations of regional devel-
opment can be shown to be made consistent with either approach (Fingleton and
McCombie 1998).
The models set out in the previous sections provide us with very different views of the
regional economic system. The neoclassical models imply convergence and tendency
towards equilibrium positions, whereas both the endogenous growth and Keynesian-type
mo~els imply processes of cumulative causation. With such an array of diverse arguments
and implications it is necessary to try to draw up an inventory of what we nowadays under-
stand from these various insights.
Within an interregional setting we know from the new economic geography (NEG) mod-
els of Chapter 3 that spatially concentrated growth can be cumulative for long periods.
Obviously localized growth cannot be explosive and infinite, because in that case all indi-
viduals will end up living in the same locality, a point made by Solow (1994). As we saw in
Chapter 1, and as will also be discussed in later chapters, this is indeed a situation towards
which some developing countries are moving. However, most countries exhibit multiple
different agglomerations with varying fortunes; some experience growth and others decline
(OECD 2009a). Similarly, many other regions with few or no concentrations of urban
AND
activity, or regions which are mixed urban-rural regions, exhibit significant growth as well
as decline (OECD 2009b). As we will see in Chapter 8, the most recent evidence appears to
suggest that there is a high degree of heterogeneity between regions in terms of both their
characteristics and their growth experiences, with no simple relationship between the type
of region and the performance of the region.
Over recent years several new arguments and lines of enquiry have emerged to help
explain why it is that different regions should experience different growth rates, and here we
will briefly deal with some of these arguments under the headings of innovation and tech-
nological change, and different notions of capital.
Since the late 1980s in particular we have seen a subtle but fundamental shift in our notions
and understanding of what input production factors are, from primarily a stock concept to
much more ofa flow concept. As we have already seen in many parts ofthis book, production
factors such as labour L, land R, and capital K are traditionally represented in our models as
being defined or measured in terms ofstock values or quantities. However, the stocks ofinput
resources provide flows of production services, and these flows are related to the stocks.
Broadly, larger stocks of factor resources imply larger flows of the services of the resources
into the production process, and this argument underpins the growth accounting-type
framework outlined in Appendices 7.1 and 7.2. However, as we have already seen in Chapter
6 and in this chapter, labour services are nowadays not seen in terms of simple stocks of
labour, defined in terms ofthe number ofworker-hours L, but rather in terms ofhuman capi-
tal HK, a concept which includes a mix ofboth the quality as well as the quantity ofthe labour
services provided. More importantly for our purposes here is that fact that moving from L to
HK implies not only that the concept oflabour has shifted, but the concept of capital has also
shifted. Our notion of capital has shifted from an emphasis on level of stocks and flows of
non-human and non-land input services to an emphasis on the quality as well as the quantity
of the stocks and flows of services which are not purely L or purely K as traditionally under-
stood. The concept of human capital HK clearly incorporates flows related to both our previ-
ous traditional notions of L and K.
This subtle shift in our understanding of the nature of input factors implies that the rela-
tionships between input factors are also rather more complex than has traditionally been
understood. In particular, factors cannot always be easily separated in a manner which is
reflected in the construction of isoquants or bid-rent curves, and then simply added together
in a manner equivalent to stock valuations. In microeconomic language, this means that fac-
tors cannot always be treated as being both additive and separable. The reason is that it is
often the interactions or interrelationships between the factors that are critical for growth
and development. Approaches which emphasize these interrelationships are often known as
systems models, and such approaches are now more or less the mainstream approaches in
much ofthe literature discussing innovation (Porter 1990), creativity (Andersson et ale 2011),
well-being (Stiglitz et ale 2009), and environment (Stern 2007). For our purposes here, what
is important is to recognize that the concept of capital itself has shifted, and this shift has
major implications for how we discuss the various elements of growth and development.
Since the early 1990s much of the recent thinking about economic growth and development
has centred on the role played by institutions and governance systems (Tabellini 2010;
Acemoglu and Johnson 2006). Much of this thinking was spurred by the opening up of the
modern global economy, as described in Chapter 1 and to be discussed further in Chapter
8, in which newly emerging BRICS and former socialist economies entered the capital- and
market-based economic system. Many of these countries exhibited very few of the institu-
tions within which markets in advanced countries operated. In the most minimalist sense
institutions here can be understood as all the formal legal architecture within which eco-
nomic activities operate. This would include issues relating to the clarity and security of
property rights, the ability to enforce contracts, and the separation of the legal and political
arenas. More recently, the argument has been extended to include the extent to which the
legal and political system interferes with commercial activities, including issues such as the
number of days to start a business, the extent of the regulations with which firms must
comply, the ease of capital movements, the restrictions and limitations on asset ownership,
nl&.-.Y~\lw!J~~M~N 8.~",~~wrg'8"M'Hi'~~<>,~ fvl0DElS
the levels of taxation, and the degrees to which 'side-payments' or bribes must be paid
(Heritage Foundation 2009; World Bank 2009). This represents a subtle shift from an
emphasis on the formal architecture of the institutional structure to the behavioural aspects
of the institutional system. The argument here is that the operation and performance of an
institutional system is not merely a question of the design of the system, but also a question
of how all the actors, stakeholders, and interested parties in the system interact with each
other. This structure-system shift in our understanding of the role played by institutions in
growth and development also reflects the more general shift in emphasis from a stock to a
flow way of understanding of the role played by factor inputs in general, as discussed above.
In the case of institutions, one of the most important developments in our understanding
centres on the role of what is known as social capital. The concept of social capital relates to
all the social norms, social rules, and social conventions operating within a society. The
concept was first elaborated many decades ago within the sociological literature, but two
seminal books by Putnam (1993,1996) have brought the arguments into both mainstream
economics and political science. More specifically for our purposes, these arguments have
very strong urban and regional implications (Westlund 2006).
Putnam's (1993) initial argument was based on a historical analysis of the different
regions of Italy, and his contention was that the huge differences in growth and develop-
ment between the north and south of Italy are largely a result of the fact that the north of
Italy has good institutions which people trust, whereas in the south this is not the case. The
reasons for these differences in trust are related to the governance history of the regions, and
Putnam's point is that the historically determined social capital has very long-lasting impli-
cations in terms of the relationships between individuals, households, government, and
commerce. Broadly, the less that individuals trust the formal legal and government institu-
tions, the lower will be the levels of economic development, while the more that they trust
institutions, the more they will invest and take commercial risks. Social trust and economic
activity are argued to be highly correlated over time. The levels of social trust are manifested
in terms of the willingness of people to actively engage in voluntary and social activities
without fear of being undermined by opportunistic behaviour (Williamson 1975) of others.
Public and not-for-profit participation is therefore interpreted as an index of social trust,
and is argued to be correlated with profit-making activities.
The long-term development of a region is therefore seen to be intrinsically related to the
institutional history of the region, but institutions here are understood not just in terms of
formal and legal institutions, but also the informal institutions of social norms, values, and
rules. In particular, it is the interactions between formal and informal institutions which
govern the relationships between individuals, households, firms, and the local state.
The second aspect of the social capital model (Putnam 1996) is that communities mani-
fest social capital in different ways, and the benefits of, and possibilities associated with,
these different social-capital manifestations also have impacts on the economic develop-
ment of different communities. Broadly, there are two types of social capital, namely bond-
ing capital and bridging capital. Bonding capital is the social capital, the social 'glue', which
holds local communities together and helps them to be 'resilient' in times of adversity
(Fingleton et ale 2012). Small isolated communities tend to be very homogeneous in terms
of ethnicity, values, and identity, and these attributes tend to foster community and a sense
of belonging, values closely related to the attachment values discussed in Chapter 4 and the
AND ~J'I'·<!-&..*""'ll.S'll!'?<..,,1!-
7.6 Conclusions
This chapter has discussed the various analytical approaches we have for understanding the
nature of regional growth and development. The most common approaches adopted by ana-
lysts are long-run neoclassical models, which distinguish between growth due to factor allo-
cation processes and growth which is due to technological change. The two broad types of
these models are the one-sector model and the two-sector model, ofwhich it was argued that
the first is regarded as the (true' long-run model of regional factor allocation. The long-run
result of these regional allocation and reallocation processes is a tendency towards interre-
gional convergence in terms of factor proportions and rates of return. The outcomes of
regional factor allocation processes are therefore seen to be somewhat different from the
factor allocation assumptions which underlie models ofinternational trade and comparative
advantage. The assumptions and conclusions of these neoclassical models were then
employed within production function analysis in order to discuss questions of the contribu-
tion of technology to regional growth, as distinct from growth due to factor reallocation.
However, this is a complex problem, and more recent endogenous growth approaches have
focused on the role played by qualitative differences in the stocks ofcapital and labour inputs,
in order to account for variations in regional performance. These various neoclassical models
were subsequently contrasted with Keynesian and post-Keynesian approaches to regional
growth analysis, approaches which centre on the notion of a regional balance of payments
constraint. The assumption here is that regional export levels act as a long-run constraint on
regional growth by limiting the maximum level of inward income-expenditure flows into
the region. Under conditions in which local factor price adjustments are regarded as ofminor
importance, regional growth is perceived as primarily dependent on a region's particular mix
of industries and local linkages. Therefore, where economies of scale are also present, these
models imply that there is no particular reason to expect interregional convergence.
AND
Discussion questions
7.1 What is the relationship between the neoclassical one-sector model of interregional
factor allocation and the nature of regional production functions?
7.2 In what ways does the direction of factors flows in a two-sector neoclassical model of
interregional factor allocation differ from that of a one-sector neoclassical model of
interregional factor allocation?
7.3 How are we to understand the term 'technology' within a neoclassical growth
framework, and how can we model the relationship between 'technology' and regional
factor stocks?
7.4 What are the regional economic implications of endogenous growth models? Are
there any limits to these mechanisms in the regional context?
7.5 To what extent are regions limited in their growth potential by a balance of payments
constraint? How can such a constraint be relaxed?
7.6 What are the implications of the 'Verdoorn law' of regional growth?
7.7 What role does innovation play in regional growth and how is the concept of 'related
variety' connected to innovation advantages?
7.8 What role do institutions and social capital play in local growth and development?
Qt = Ae¢tKaLI-a (A.7.l.I)
FACTOR .11'''1U!",I~~''''~¥'''&lIf AND M'¥"'&~¥"'&?'IIi""',1'lw PAYMENTS
1 dQ
--------- = ¢ + a dK + ----~-------_._--
(1 - a) dL
(A.7.1.3)
Q dt K dt L dt
where Ql' Kt, Lt represent the rates of growth of output, capital, and labour at time t,
respectively.
We know that wages depend on labour productivity, and the growth ofwages will there-
fore be related to the growth of labour productivity. In order to investigate the exact
nature of these relationships, we can take the growth in labour from both sides of equa-
tion (A.7.1.4) in order to arrive at an expression for the growth of labour productivity
thus:
. . .
Qt Kt = ¢+aKt Kt - Lt -aLt (A.7.1.5)
which rearranges to
The left-hand side of equation (A.7.1.6) represents the rate of growth oflabour productivity
at time t, and is given as the sum of the level of technology, plus the growth in the capital-
labour ratio, weighted by the factor share of capital. As we see in Appendix 7.2, the rate of
growth of labour productivity represents the rate of growth of wages. These conclusions are
exactly in agreement with our one-sector model of factor allocation.
We can also adopt a similar approach to investigate the sources ofprofit growth. However,
in order to do this we take the rate of growth of capital from both sides of equation (A.7.1.4)
to arrive at an expression for the rate of growth of capital productivity thus:
The left-hand side of equation (A.7.1.8) represents the rate of growth of capital productiv-
ity at time t, and is given as the sum of the level of technology, plus the growth in the
labour-capital ratio, weighted by the factor share of labour. As we see in Appendix 7.2,
REG~ONAL E(ONOfu1~C
the rate of growth of labour productivity represents the rate of growth of profits.
Once again, these conclusions are exactly in agreement with our one-sector model of
factor allocation.
In order to consider the sources of growth in a steady-state situation in which the rate of
growth of profits is zero, we can set profit growth to zero when the rate of growth of the
output-capital ratio is zero. In other words we have
Qt == K t (A.7.1.9)
o == ¢ + (I - a)(i t - Ot) == ~
(A.7.l.IO)
o == ¢ + (a - 1) (Ot - it ) 1( (A.7.l.II)
wt Qt - L t == I-a (A.7.l.I2)
In other words, in a steady-state situation in which the rate of growth of profits is zero, the
rate of growth of labour productivity and wages depends simply on the level of technology
and the factor share of labour in the economy.
The wage paid to labour w is equal to the marginal product of labour MP v and is given by
the marginal physical product of labour MPP v multiplied by the price of the output Px pro-
duced. Within the Cobb-Douglas production function, the index of labour {J, given as
{J (1 a), is defined as the partial elasticity of output with respect to the input labour:
~QIQ ~Q L ~Q L (A.7.2.1)
(1- a) == ~LIL == Q x ~L ~L
x
Q
Therefore,{J == (I a) == MPLIAP v where MP Lis the marginal product of labour, and APLis
the average product of labour. As such, the wage w, which is given by the marginal product
of labour, is given by w (I - a)APL •
Similarly, the profit rate paid to capital r is equal to the marginal product of capital MPK'
and is given by the marginal physical product ofcapital MPPK , multiplied by the price of the
AND . . "'IC'ii,b'.yi(·'ii,~'Il""',~ PAYMENTS
Therefore a MPK/APK) where MPK is the marginal product of capital, and APK is the aver-
age product of capital. As such) the profit rate r) which is given by the marginal product of
capital) is given by r = aAPK•
AJ!IOd J!WOUOJ3
pue 'SUO!Bali 'sa!I!)
:uO!lez!leq o I9
Economic geography and
economic history
Over recent years we have become increasingly aware of the power of modern globalization
in influencing and shaping almost all aspects of our lives. The central elements of this book,
namely cities and regions, are increasingly understood as playing a key role in driving glo-
balization processes. At the same time, the fortunes of cities and regions are also often
argued to be more than ever subject to the impacts of globalization. Yet are these relation-
ships between the fortunes of cities, regions, and globalization new phenomena? Why are
these issues important for urban and regional economic analysis, models, and methods?
And what is <globalization' and when did it begin?
These are complex questions, but providing some answers is very important in demon-
strating why the study of urban and regional economics is in many ways more important
and timely than ever before. The models and methods discussed in Part I ofthis book can be
applied in many real-world contexts to analyse specific issues and mechanisms. However,
the power and applicability ofthe insights provided also rest on our awareness ofthe broader
backdrop and changing global context in which urban and regional matters take place. This
is the topic dealt with in Chapters 9-11.
How we answer the above questions depends in part on how we view globalization and on
the perspectives that we adopt (McCann 2008). Globalization can be considered from the
perspective of cultural issues, political issues, historical issues, geographical issues, or eco-
nomic issues, and can be understood as either a largely new phenomenon or as an ongoing
process (Steger 2003; MacGillivray 2006). For our purposes in this book we interpret globali-
zation in terms ofeconomics and geography, and this perspective means that we focus on the
relationships between globalization, industrialization, and urbanization. As we will see, the
importance of studying the economics of cities and regions is that they playa crucial role in
long-run economic development. However, the long-run role played by cities and regions in
economic growth is one that has evolved over time. At various stages in history the changing
relationships between trade, growth, and urbanization have both challenged and defined the
notion ofa city, a region, and even the concept ofwhat it is to be a nation. Importantly for our
purposes, these relationships have changed again over the last few decades in response to the
modern era of globalization. An understanding of these recent changes is therefore essential
in order to position the models and methods examined in Part I of this book.
As with all models, urban and regional economic models are by nature always at least
partially abstract, but they are designed to capture aspects of the structure and processes of
change of very concrete phenomena. The aim of Part II of this book is to develop an under-
standing of the impacts of globalization on both the long-run and current development of
cities and regions, and to see how best to use the models discussed in Part I to analyse these
developments. This is best done by building on all the material in Part I and then applying it
within an explicitly historical as well as a geographical context.
While the logic of Part I is built around the theoretical models and methods in which fac-
tors are assumed to move, relocate, or adjust more or less instantly, it is also important to
realize that the structure of many of the regions and cities we observe actually evolves grad-
ually over time. Cities and regions as we perceive them today are in part products ofthe built
environment, as well as being in part constructions of the social, institutional, and natural
environment. These constructions typically change very slowly, but in some cases they can
change quire rapidly, as witnessed by the rapid urban and regional transformations cur-
rently taking place in countries such as China. However, these transformation processes are
not new, and a good knowledge of economic history helps us understand the processes by
which cities and regions emerge and change over time.
In order to provide a solid understanding of the context out of which today's cities and
regions have emerged, we begin here by examining the relationship between industrializa-
tion, urbanization and the economic performance of the countries in which they are located.
In particular, we examine how these relationships have evolved during the various stages of
globalization which have taken place since the late Middle Ages. Box 8.1 examines urbani-
zation in the late Middle Ages, and the subsequent sections then examine the changing
relationships between urbanization and industrialization and globalization which emerged
over the following centuries. As we will see, the interrelationship between industrialization,
urbanization, and economic growth was an ongoing and fairly direct relationship in which
the world's most industrialized states were generally the most urbanized states and also the
wealthiest states in the world. This relationship continued more or less unabated during
each century since the transition from the late Middle Ages to the early Renaissance right up
to the end of the nineteenth century and the beginning of the twentieth century.
As we will see in this chapter, the period when the relationship between urbanization,
industrialization, and economic performance was at its zenith was between the late nine-
teenth century and the early twentieth century, and this was also the period during which
many ofthe seminal authors referred to in this book, such as Marshall (1890), Weber (1909),
Christaller (1933), Ohlin (1933), and Hoover (1937), were making their ground-breaking
observations, developing their analyses, and producing their most famous texts, on which
many of today's models and methods are built.
However, the middle years of the twentieth century also saw something of a hiatus in
this urbanization-industrialization-economic development relationship, in that the mid-
dle years of the twentieth century were a period of anti-globalization, and in some coun-
tries also a period of anti-urbanization. Later on, during the second half of the twentieth
century, the long-run relationship between industrialization, urbanization, and economic
AND H~STORY
development once again emerged, but in a rather different form from the patterns observed
before the early twentieth century. In the second half of the twentieth century, the broadly
positive relationship between urbanization, industrialization, and economic development
was much more focused on developing rather than developed countries.
x .... x .. '~"" X '0>0' X'> .." , A N [)
Moreover, from an urban and regional perspective there do appear to be many aspects of
today's era of globalization which are not only new, but also totally different from anything
experienced in earlier eras. This suggests that while on the one hand the globalization we
experience today can be understood as part of an ongoing and longstanding historical pro-
cess, on the other hand modern globalization can also be understood as a radical departure
from earlier trends. Therefore, in order to understand the urban and regional patterns we
observe today, it is necessary to consider all these features.
This chapter discusses the evolving long-run characteristics of the relationship between
urbanization, industrialization, and globalization. The chapter concludes with an introduc-
tion to the major features of the most recent era of globalization, which began in the late
1980s, a period in which the relationship between industrialization and urbanization again
evolved in different ways. Chapter 9 will deal with these contemporary issues in detail, while
the policy responses will be discussed in Chapter 10, on the basis ofall the analytical insights
provided in earlier chapters.
growth of underwriting and insurance markets in the financial markets of London; the
eighteenth-century industrialization and expansion of Great Britain and the subsequent
industrialization of other parts of the world (Ferguson 2008).
In terms of economic geography, it seems at first sight curious that globalization should
begin in Europe. We know from the various arguments in this book that city size and the
size of the (home) market are all related to economic performance, and this would appear
to favour the Asian empires. Therefore it appears strange that while the world's largest and
richest cities and empires were all originally outside Western Europe, the initial processes of
globalization and industrialization should be primarily a European phenomenon. Yet in
many ways it was the very fragmentation of Europe into different kingdoms and city-states
which encouraged competition between states (Ferguson 2011), whereas the larger and
monolithic governance structures of the Asian empires inhibited this. As such, it is not
simply city size or even the size of the country which drives development, but also the issue
of international competition and governance. For this reason, many other scholars would
perceive globalization in the modern sense as beginning with the competition between the
major trading companies. These trading companies were the early form of multinational
corporations, and in many ways were the paramount agents of globalization (Ferguson
2008), with the country playing a much lesser role.
On this argument, the founding ofthe VOC Dutch East India Company in 1602 (Ferguson
2008) and the establishment of similar types of companies in both England and France
marks a watershed in globalization. The reason is that the founding of the Dutch East India
Company was the first time in which the public raising of investment share capital was also
integrated with processes of trade creation, international migration, foreign direct invest-
ment' and national-colonial expansion. These highly organized, sophisticated and very
powerful trading companies emerged as the first major joint-stock-issuing multinational
corporations, and it was these trading organizations which spearheaded the early globaliz-
ing processes of the major European nations.
The Dutch Republic embarked on its first major wave of colonial expansions during the
latter decades of the sixteenth century, vying with the Portuguese for supremacy in the spice
trade routes to the East around the Cape of Good Hope (Findlay and O'Rourke 2007). The
Dutch set up colonies in South Africa, South Asia, the East Indies, and with Japan at Deshima
Island outside Nagasaki in 1571. By the early seventeenth century the Dutch had overtaken
the Portuguese as the principal power in the eastern trade routes, and throughout the sev-
enteenth century the Dutch themselves were soon followed, first by the British and then by
the French, in competing economically and militarily for access to these trading and colo-
nizing opportunities. At the centre of this emerging global competition were the trading
companies of each country.
In terms of economic geography, the evolution in trade and globalization processes was
also reflected in terms of the evolution in urbanization patterns. As we see in Table 8.1, by
the beginning of the sixteenth century, the Low Countries and Northern Italy were by far
the most urbanized parts of Western Europe. This level of urbanization reflected the
growth of successful city-states and their associated European and Mediterranean trade
networks during the previous two centuries. In these regions, the proportion of the total
population living in cities of over 10,000 was more than twice that of any other part of
Western Europe.
REG~ONS~
1500 urban 1600 urban 1700 urban 1800 urban 1890 urban (Industry
index Index index (indus- index (indus- index (indus- index 1913)
try index in try index in try index in
1750) 1800) 1860)
Urban (urbanization) index: population living in cities of at least 10,000 inhabitants as a percentage of total population
(Maddison 2007a, p. 43)
Industry (industrialization) index: levels of industrialization in the UK in 1900 100 (Findlay and O'Rourke 2007 p. 323)
Source: Maddison (2007a); Findlay and O'Rourke (2007)
Note: Industry indices for the Netherlands are those for Belgium; those for England and Wales, Scotland and Ireland are for
the UK as a whole; those for Scandinavia are for Sweden.
In terms of individual cities, at this time, Paris, the capital of Western Europe's largest
country, also remained Western Europe's largest city with 245,000 inhabitants (Chandler
1987). However, following the transatlantic expansion of the Spanish and Portuguese
empires in the fifteenth century, both Spain and Portugal experienced increasing domestic
urbanization. Within Spanish territory Paris was now closely rivalled by Naples with
224,000 inhabitants (Chandler 1987). More noticeably, a century of Spanish globalizing
activities also meant that by 1600, along with Seville, Milan, and Palermo, the kingdom of
Spain now also contained five of the ten largest cities in Western Europe. All these cities
contained over 100,000 inhabitants, the other five being London, Lisbon, Venice, Prague,
and Rome (Chandler 1987).
By the beginning of the eighteenth century the relationship between industrialization,
urbanization, and globalization had become firmly established in all the growing colonial
powers. The process of rapidly increasing urbanization, which had already been experi-
enced by Spain and Portugal in the previous century, was now also observed on an even
greater scale in the Low Countries during the seventeenth-century Dutch era of colonial
expansion (Ferguson 2008). As we see in Table 8.1 and Table 8.2, by 1700 the century-long
rise of the Dutch Republic and the global growth of Dutch trade had resulted in the Low
Countries becoming by far the most densely urbanized region of Western Europe.
Sources: City population data (Chandler 1987); country population, GOP and GOP per capita data (Maddison 2006)
Note: All GOP and GDP per capita $ values are given in 1990 Geary-Khamis dollars (Maddison 2006).
Amsterdam now emerged as the third-largest Western European city, after London and
Paris. It is no accident that these three largest Western European cities were also the home
locations of the East India and West India trading companies of England, the Netherlands,
and France (Findlay and O'Rourke 2007), all of which were in the vanguard of the globaliz-
ing activities of these countries.
As we see in Table 8.2, the outcome ofthese interrelated processes of urbanization, indus-
trialization, and early globalization is that the largest fifteen cities in the world in 1700
ranged in population from Constantinople (Istanbul) at 700,000 to Lisbon at 188,000. Nine
of the world's fifteen largest cities were in Asia, with five being located in Western Europe,
as well as Constantinople, which is at the crossroads of Europe and Asia. The large and
ancient civilizations of China, India, and Japan were reflected in the size of their major cit-
ies. Only the two largest cities of Western Europe, London and Paris, were of comparable
size to the very largest Asian cities, and these were the capital cities of the two largest
European nations, Great Britain and France.
Table 8.2 also shows that many of the major cities of Europe in 1700 were of the same
order of magnitude as many of the other largest cities in Asia. This may appear surprising
given that the populations and economies of China and India were far larger than for any
other countries. Yet clues as to why the European and Asian cities were of the same order
of magnitude comes from the fact that the national per capita GDP of the dominant
European cities' own countries was already of the order of two to four times that of the
major Asian economies. l As a result of the seventeenth-century long 'Golden Era' of
Dutch global pre-eminence, by 1700 productivity in the Netherlands was almost twice
that of any other country. This suggests that while the size of a major city appears to be
partly related to size of the country in which it is located, city size also appears to be
related to the productivity of the country in which it is located, exactly as agglomeration
arguments imply.
The dominance of the Dutch economy, built as it was around the relationships between
its trading cities and its colonial empire, became increasingly overtaken by other expand-
ing empires. The eighteenth century was the period which witnessed the increasing global
pre-eminence of the empires of Great Britain, France, Russia, and Austria, all of which
challenged the earlier dominance of the Dutch Empire. As with the Dutch, Spanish, and
Portuguese empires in the previous centuries, these newly emerging empires experienced
a similar growing relationship between urbanization, city size, industrialization, and glo-
balization throughout the eighteenth century. As we see in Table 8.3, by 1800 six of the
world's largest fifteen cities were now located in Western Europe, and most notably, these
included Moscow and Vienna. The growth of these cities during the eighteenth century
coincided with the growth and modernization of Russian Empire under Peter the Great
and Catherine the Great, and also that of the Austrian Empire under the Habsburg
Monarchy.
Of these rising empires, whose eighteenth-century globalization processes were all associ-
ated with industrialization and urbanization, the emerging superpower was Great Britain.
1 Pomeranz (2000) argues that the labour productivity gap between the European and Asian economies was very
much lower than the Maddison (2006, 2007a, b) figures would imply. However, these disagreements do not alter
the basic argument here.
/
1800 City population Country population GOP $0005 1820 (% GOP per capita
0005 (% growth 000s 1820 (% growth growth 1700-1820) $1820 (% growth
1700-1800) 1700-1820) 1700-1820)
Beijing (Peking) 1,100 (69.2) 381,000 (276) 228,600 (276) 600 (0)
London 861 (56.5) 21,239 (247) 36,232 (338) 1,706 (36.4)
Guangzhou (Canton) 800 (400) 381,000 (276) 228,600 (276) 600 (0)
Tokyo (Edo) 685 (0) 31,000 (14.8) 20,739 (34.7) 669 (17.3)
Istanbul 570(-18.5) 25,147 (West Asia) 15,269 (West Asia) 607 (West Asia) (0)
(Constanti nople)
Paris 547 (3.2) 31,250 (14.6) 35,468 (182) 1,135 (24.7)
Naples 430 (208) 20,176 (15.2) 22,535 (54) 1,117 (15.4)
Hangzhou (Hangchow) 387 (27.7) 381,000 (276) 228,600 (276) 600(0)
Osaka 383 (0) 31,000 (14.8) 20,739 (34.7) 669 (17.3)
Kyoto 3T7 (108) 31,000 (14.8) 20,739 (34.7) 669 (17.3)
Moscow 248 (217) 54,765 (264) (USSR) 37,678 (232) 688 (12.6)
Soochow 243 (173) 381,000 (276) 228,600 (276) 600 (0)
Lucknow 240 (400) 209,000 (26.6) 111,417 (26.6) 533 (-3.1)
L.isbon 237 (26) 3,297 (64.8) 3,043 (85.7) 923 (12.7)
Vienna 231 (220) 3,369 (34.7) 4,104 (65.2) 1,218 (18.6)
World 1,041,092 (72.5) 694,442 (86.9) 667 (8.4)
This was the period during which Britain underwent the first phase of its industrial revolu-
tion. Rapid mechanization, capitalization, and urban-rural migration, particularly after
1750, all followed on from the widespread adoption of steam power based on large-scale coal
extraction, the introduction ofthe first generation of large-scale factory production systems,
and the spatial transformations enforced by the land enclosure movement. As we see in Table
8.1, by 1800 Britain was the most industrialized country in the world, and as the industrial
transformation accelerated, so did the urban transformation.
At this time, the growth of urbanization in England, Wales, and Scotland was much
greater than in any other areas of Europe, apart from the Low Countries, which are now the
Netherlands and Belgium. As well as generally increasing urbanization, the onset of the
industrial revolution in Britain was also associated with a rapid increase in the size of its
capital city, London. By 1800 London had become the second-largest city in the world, and
some 57 per cent larger than Western Europe's second-largest city, Paris. For almost two
centuries London and Paris had been very similar in size, but during the eighteenth century
France had undergone significantly less technological and industrial change than Britain,
and this was also reflected in much lower levels of urbanization in general and the smaller
size of its capital city.
The period between 1800 and 1850 was a period of increasing urbanization and industri-
alization in both North-Western Europe and the USA, with the advent of railways spear-
heading these changes. As we see in Table 8.1, by the middle of the nineteenth century, the
level of industrialization in Great Britain was more than double that of any other nation. As
we see in Table 8.4, the result ofthis was that in 1850 UK per capita gross domestic product
~(
1850 City popu- Country popula- GOP $OOOs(% GOP per capita
lation OODs tion OOOs (% change change 1820-1850) $ (%change
(%change 1820-1850) 1820-1850)
1800-1850)
St Petersburg 502 (228) 73,l~)0 [USSf\] (34.6) 83,646 [1870] (52.7) 943 [1870] (37.1)
Berlin 446 (259) 33,746 (35.9) 48,178 (79.6) 1,428 (32.6)
Hangchow 434 (12.1) 412,000 (8.1) 247,200 (8.1) 600 (0)
Ulangchow)
Vienna 426 (84.4) 3,950 (17.2) 6,519 (58.8) 1,650 (35.5)
Philadelphia 426 (626) 23,580 (236) 42,583 (426.6) 1,806 (43.6)
Liverpool 422 (555) 27,181 (27,181 (27.9)) 63,342 (74.8) 2,330 (36.5)
Naples 414(3.8) 24,460 (21.2) 33,019 (46.5) 1,350 (20.8)
World [1870] 1,270,014 (21.9) 1,101,369 (58.6) 867 (29.9)
(GDP), a measure of a country's output and wealth per capita, was the highest in the world
and London was by then the world's largest city. The industrialization of France, a process
which started in earnest many decades after Britain, allied with increasing French colonial
expansion, meant that by 1850 Paris was by now the third-largest city in the world.
Significantly, for the first time, cities in North America, most notably New York, appear in
the list of the world's fifteen largest cities, although as we see in Table 8.1 the level of indus-
trialization in the USA at this stage was still far below that of the UK, and more or less
equivalent to those of France (Findlay and O'Rourke 2007).
The overall rate of growth of the Western world's largest cities was increasing during
these years of continuing industrialization and colonial expansion. During the fifty years
between 1800 and 1850, ten out of the world's fifteen largest cities had experienced faster
growth than their equivalently ranked city in 1800 had experienced during the previous
hundred years. Yet, by 1850, only six of the world's fifteen largest cities were in Asia. These
changes in the scale of the major cities therefore also point to a geographical shift in the
nature of urbanization. In particular, the change in global city rankings between 1800 and
1850 reflects the fact that rapid industrialization was taking place primarily in the European
AN [) ~~~>J~'ljj\.J ~~'~ §~
and North American economies, rather than in the major Asian economies, which remained
largely rural.
During the first four centuries of the globalizing processes which took place from the
beginning of the sixteenth century to the turn of the twentieth century, the period with the
fastest growth of industrialization and urbanization was the very end of the era, spanning
the second half of the nineteenth century and continuing right up to the eve of the First
World War. As we see in Table 8.1, during the fifty years leading up to the Second World
War, while the levels of industrialization in the UK and France had doubled, in the USA and
Germany they had increased some six-fold. Between 1820 and 1913 these enormous
increases in the level of industrialization were also associated with rapidly increasing ine-
quality between the different parts of the world, with the rich industrialized North Atlantic
economies along with British off-shoots (Maddison 2006) such as Australia and New
Zealand all rapidly pulling away from the rest of the world (Findlay and O'Rourke 2007).
In terms of cities and urbanization, this period of enormous industrialization also coin-
cided with what until then was the era with the most rapid rates of urbanization the world
had ever known. As we see in Table 8.1, the global urbanization rates had tripled during the
nineteenth century. By comparing Table 8.5 with Table 8.4, we see that the outcome of this
was that during the fifty years between 1850 and 1900, eleven out of the world's fifteen larg-
est cities had experienced faster growth than their equivalent-ranked city in 1850 had expe-
rienced between 1800 and 1850. By now, the world's largest cities reflected the economies
wiQ1 both the highest levels of per capita productivity and also productivity growth, with
twelve out of the world's fifteen largest cities now being either in Western Europe or the
1900 City population 0005 Country popula- GOP $0005 GOP per capita
(% change 1850-1900) tion 0005 (% change (%change $ (%change
1850-1900) 1850-1900) 1850-1900)
The evidence presented so far all points to a clear positive relationship between industriali-
zation) urbanization) and globalization) a relationship which was ongoing from the early
sixteenth century right up to the beginning of the twentieth century. Urbanization was an
unambiguous indicator of industrialization and the largest cities were all in the richest
countries. Not surprisingly therefore) as we see in Table 8.6) by 1925) as the dominant city of
the world)s dominant economy) New Yorkhad emerged as the world)s largest city. Moreover)
by now) fourteen of the world)s fifteen largest cities were located in Europe) the USA) or
Japan. Only Buenos Aires) which was the world)s fastest-growing major city in the early part
of the twentieth century) was outside these regions) and this too was located in what at the
time was a very wealthy country. In 1925) all the world)s largest cities were in the richest and
largest economies.
This longstanding relationship between urbanization) industrialization) and globalization
had been well established for so long that it is not surprising that) as we have already seen in
this book) this was also precisely the period when the first scholars who were seriously inter-
ested in the economics ofcities and regions) including Alfred Marshall (1890)) Alfred Weber
(1909)) Walter Christaller (1933)) Berti! Ohlin (1933)) were making their groundbreaking
observations and writing their seminar works. Cities had emerged as the economic engines
at the heart of the emerging global nation-empire trading systems) and there was nothing to
suggest that these relationships between urbanization) industrialization) and globalization
would change. Yet) after an era of four centuries of largely uninterrupted globalization pro-
cesses) there did indeed begin a period of anti-globalization in many of the richer countries.
This reversal in the long-run historical processes (Ferguson 2008) of urbanization was a
result of changes in the processes of globalization) which themselves were a result of the
Table 8.6 The world's largest cities in 1925
1925 City population Country popula- GOP $0005(% GOP per capita
0005 (% change tion 0005 (% change change 1900-1925) $ (%change
1900-1925) 1900-1925) 1900-1925)
New York 1,1/4 (83.2) 116,284 (52.2) 730,545 (233) 6,282 (53.5)
London 7,142 (19.5) 45,059 (9.48) 231,806 (25.4) 5,144 (14.5)
Tokyo 5,300 (354) 59522 (86.0) 112,209 (216) 1,885 (59.7)
Paris 4,800 (44.1) 40,610 (11.7) 169,197 (44.9) 4,166 (44.8)
Berlin 4,013 (48.2) 63,166 (87.2) 223,082 (37.4) 3,532 (18.3)
Chicago 3,564 (208) 116,284 (52.2) 730,545 (233) 6,282 (53.5)
Ruhr 3,400 (443) 63,166 (87.2) 223,082 (37.4) 3,532 (18.3)
Buenos Ai res 2,410(299) 10,358 (221) 40,597 (233) 3,919 (53.5)
Osaka 2,219 (228) 59,522 (86.0) 112,209 (314) 1,885 (18.3)
Philadelphia 2,085 (47) 116,284 (52.2) 730545 (216) 6,282 (53.5)
Vienna 1,865 (9.8) 6,582 (10.2) 22,161 (233) 3,367 (204)
Boston 1,764 (64.1) 116,284 (52.2) 730,545 (28.7) 6,282 (53.5)
Moscow 1,164 (57.5) 158,983 (27.2)(USSR) 231,886 [1928] (50.5) 1,370 [1928] (10.0)
Manchester 1,125 (20.2) 45,059 (9.48) 231,806 (25.4) 5,144 (14.5)
Birmingham 1,100 (36.2) 45,059 (9.48) 231,806 (25.4) 5,144 (14.5)
global economic disruption engendered by the two world wars, the intervening Depression
of the 1930s, and the reconstruction of the global economic system under the highly regu-
lated international trade architecture of the Bretton-Woods system. In the inter-war years,
all major economies increasingly re-oriented their trade patterns primarily to within the
sphere of their own colonial systems (Findlay and O'Rourke 2007). The outcome of this
period of anti-globalization was that countries became largely closed to each other, and the
importance of international economic linkages and transmission mechanisms fell relatively
in comparison with the importance of domestic mechanisms. In economic terms, the
impacts of these reversals were that the ratio of world trade to global GD P fell during the
period 1929-1950, and barely recovered until the 1970s (Fischer 2003), while the ratio of
foreign assets to global GD P declined from 1914 onwards and was not attained again until
1980 (Crafts 2004).
This period of anti-globalization and the massive global contraction of trade and foreign
investment on the part of the rich countries also coincided with a fall in the importance of
urban growth as an economic engine in many of these same countries. As we have already
seen, urbanization had always been closely associated with industrialization, and as eco-
nomic growth and trade fell, so therefore did the growth of urbanization. In the rich indus-
trialized countries the urbanization growth rates of the largest cities slowed down during
the first half of the twentieth century. As we see in Table 8.7, of the world's largest cities in
1950, only six out of the world's fifteen largest cities had experienced faster growth between
1925 and 1950 than their equivalent-ranked city in 1925 had experienced between 1900 and
1925. Apart from the rapid growth of Los Angeles, in terms of the industrialized countries,
the overall global city rankings remained relatively stable between 1925 and 1950.
x .. '<."'" ... U X ..... ~. i\ N D
1950 City population Country popula- GDP $OOOs GDP per capita
OOOs (% change tion OOOs (% change (%change $ (%change
1925-1950) 1925-1950) 1925-1950) 1925-1950)
New York 12,463 (60.3) 152,271 (30.9) 1,455,916 (99.3) 9,561 (52.2)
London 8,860 (14.4) 50,127 (11.2) 347,850 (50.1 ) 6,939 (34.8)
Tokyo 7,000 (32.1) 83,805 (40.8) 160,966 (43.4) 1,921 (1.9)
Paris 5,900 (22.9) 41,829 (3.0) 220,492 (30.3) 5,271 (26.5)
Shanghai \407 (360) ~)46,81 (13.8) 239,903 (10.0) 439 (21.9)
Moscow 5,100 (289) 179,571 [USSR] (12.9) 510,243 (220) 2,841 (207)
Buenos Aires 5,000 (207) 17,150 (65.6) 85,524 (210) 4,987 (27.2)
Chicago 4,906 (37.6) 1~)2,2 71 (30.9) 1,4:):),916 (99.3) 9,561 (52.2)
Ruhr 4,900 (44.1) 68,375 (8.2) 265,354 (18.9) 3,881 (9.9)
Kolkata (Calcutta) 4,800(345) 3f)9,OOO (12.2) 222,222 (30.3) 619 ( 11.4)
Los Angeles 3,986 (347) 11)2,271 (30.9) 1,4:):),916 (99.3) 9,:)61 (52.2)
Berlin 3,707 (7.7) 68,37:) (8.2) 265,3:)4 (18.9) 3,881 (9.9)
Osaka 3,341 (50.6) 83,80~) (40.8) 160,966 (43.4) 1,921 (1.9)
Philadelphia 2,900 (39.1) 1L)2,2 71 (30.9) 1,455,916 (99.3) 9,561 (52.2)
Mexico City 2,872 (372) 28,485 (53.3) 67,368 (223) 2,365 (73.1)
World (1950) 2,524,324 5,329,719 2,111 (40)
(19131950)
The period characterized by the 1930s Depression and the Second World War can there-
fore in many ways be considered as marking the end of the first major phase of global
urbanization, which from the early seventeenth century to 1950 had been dominated by
Europe and North America. This initial phase of urbanization and industrialization had
led to an increase in the global number of urban dwellers from fifteen million to over 400
million, and an increase in the global urbanization index from 10 per cent to 52 per cent
(UNFPA 2007).
In contrast to the previous four centuries of increasing globalization, industrialization,
and urbanization, the immediate post-Second World War period can be regarded as mark-
ing the start of the second phase of global urbanization (UNFPA 2007, a phase which is
qualitatively quite different in nature to the earlier periods or urbanization. Since 1950, not
only has the urbanization rate increased globally, but this second phase of global urbaniza-
tion has been dominated by the rise of urbanization in developing countries (Satterthwaite
2005). In 1950, only eleven out of the largest thirty cities in the world were from developing
countries (PWC 2009). By 1975, over 75 of the 190 urban agglomerations with over one
million inhabitants were from the developing countries (Chandler 1987). As we see in
Table 8.8, the result of these new forms of urbanization meant that by 1975 seven of the
world's fifteen largest cities, namely Mexico City, Sao Paulo, Buenos Aires, Rio de Janeiro,
Cairo, Shanghai, and Calcutta, were in developing countries, and between 1950 and 1975,
eleven of these mega-cities had experienced faster growth than their equivalent-ranked city
in 1950 had experienced between 1925 and 1950.
Between 1975 and 2000 the global process of urbanization accelerated even more. The
number of cities in the world with a population of more than one million went from 115 in
1960 to 416 in 2000 (Venables 2006) and by the year 2000, there were over 140 cities globally
with populations of over two million inhabitants (Le Gales 2002). The number of cities of
more than 4 million the increased between 1960 and 2000 from 18 to 53 (Venables 2006)
and by 2000 nineteen cities had populations of over ten million (Le Gales 2002). The num-
ber of cities with more than 12 million inhabitants had increased from 1 to 11 during the
four decades between 1960 and 2000 (Venables 2006).
The effect of all of these changes was that, by 2008, at 3.3 billion, the number of people
living in urban areas across the world had for the first time passed 50 per cent of the global
population (OEeD 2007; UNFPA 2007, and this process of increasing urbanization was
common to both the industrialized and the industrializing world. But in the developing
world the rate of urbanization was far more dramatic than in the developed world. By 2005
the developing world's urban population of 2.4 billion accounted for approximately three-
quarters of the global urban population (World Bank 2008; UNFPA 2007. The proportion of
the population in the low- and middle-income countries of the developing world which
lived in urban areas had increased from 37 per cent to 44 per cent between 1990 and 2005,
while during the same period the proportion of the population in developing countries liv-
ing in cities of over one million inhabitants had increased from 14 per cent to 17 per cent
(World Bank 2008).
The result of all of these changes was that, by the year 2000, as we see in Table 8.9, ten of
the world's fifteen largest cities were from the developing world, and this tendency towards
""""""_."_,j(", AND [CONOfvl
2000 City population Country population GOP $0005 (% change GOP per capita
0005 (% change 0005 (% change 1975-2000) $ (%change
1975-2000) 1975-2000) 1975-2000)
mega-cities in the developing world was not specific to one or two countries, in that these
ten cities were located in eight different countries. More recently, PWC (2009) estimated
that 21 out of the 30 largest cities in the world in 2007 were located in developing or newly
industrializing countries.
Moreover, all the evidence suggests that the rate of urbanization in the developing
world will continue to increase even faster relative that that of the developed world. The
recent annual growth of the urban population in low- and middle-income countries in
the developing world between 1990 and 2005 was 2.6 per cent, while that in high-income
countries was only 1.1 per cent (World Bank 2008). United Nations predictions suggest
that the global urban population will increase to 4.9 billion by 2030, of which 3.9 billion
will be in the developing world (UNFPA 2007). As such, the level of urbanization in the
developing world will increase by 60 per cent between 2000 and 2030, which is some
3.75 times greater than the urbanization rate in the developed world over the same
period.
ECONOMiC HiSTORY
This unprecedented urbanization is not simply a result of population growth, in that over
the same period, the global rural population is expected to actually decrease (UNFPA 2007
PWC 2009). The role of urban scale and centrality appears to be becoming ever more critical
as sources of economic growth. Over the last three decades, the increasing importance
played by cities as engines of national, regional, and global economic growth is demon-
strated by the fact that the proportion of people living in urban areas has increased in all
parts of the global economy (Richardson and Bae 2005). Between 1950 and 2030, the total
urbanization index of the developing world is predicted to increase from 18 per cent to 56
per cent (UNFPA 2007, with the majority of this increasing urbanization taking place in
Asia and Africa. By 2030, Asia and Africa will account for 80 per cent of the global urban
population (UNFPA 2007). China alone expects its urban population to double by 2030
(Venables 2006).
Estimates and forecasts of the future growth trends for city sizes, city output levels, and
city productivity performance in different parts ofthe global economy are provided by vari-
ous organizations, including PWC (2009), MGI (2009, 2010, 2011a, b, 2012) and OECD
(2012). Of course cities vary enormously in terms of geographical structures and commut-
ing patterns, and the development of standardized measurement criteria is very important
(0 ECD 2012) in order to ensure comparability. But much ofthe evidence points in the same
direction, namely that the world is becoming more urbanized, and that cities play an
increasingly important role in the global economy. The increasing role of cities is also asso-
ciated with improvements in education and indicators of health and well-being, as it has
been throughout history (Glaeser 2011).
The interrelationships between city size, education, health, and well-being, however, are
by no means straightforward (Burdett et al. 2011), and these relationships are particularly
problematic in the case of very large cities. While on the one hand very large cities contrib-
ute significantly to economic development, the mega-cities of the developing world pose
major challenges in terms of poverty reduction, environmental degradation, health care,
and housing, and in many of these cities the challenges are becoming more acute (Burdett
et al 2011). In contrast, in Europe, well-being and quality of life indicators are generally
higher in smaller and mid-size urban centres, many of which are, by the standards of North
America and the developing countries, actually small cities (European Commission 2010).
However, the relationship between city size and social and health issues are very complex
and the possible policy responses to these types of problems, as they relate to cities and
regions, are discussed in Chapter 10. At this stage we focus on certain key features of mod-
ern cities and modern urbanization processes.
there were globally 67 cities with over one million inhabitants (Chandler 1987) in 1950, as
we have seen, by 1975 there were 190 urban agglomerations with over one million inhabit-
ants (Chandler 1987). The USA alone accounted for 25 of these million-plus agglomera-
tions, Western Europe accounted for 34, and Warsaw Pact Europe accounted for fifteen
(Chandler 1987). Urbanization among the rich countries was therefore dominated by the
USA. The proportion of the US population living in cities increased at a higher rate in the
post-war era than during the inter-war period, and among the other rich countries urbani-
zation and the growth of mega-cities also emerged in both Japan and South Korea during
the post-Second World War era.
In other rich countries, however, the population actually fell in many cities. This was par-
ticularly the case for many traditional industrial cities which declined in population during
this era as manufacturing exhibited an urban-rural drift. In Europe this was particularly
noticeable in the traditional manufacturing heartlands of the UK (Fothergill et al. 1985) and
Germany, while in both the USA (Glaeser 2005) and Europe (Sassen 2006) many of the very
largest cities saw population declines in the core parts of the agglomerations. From the late
1970s similar trends were evident in many of the USA's mid-Western cities, and particularly
those which had been specialized in traditional manufacturing and heavy engineering. While
many of these urban restructuring processes were associated with absolute population
declines within the core city areas, in the outer urban areas it was a different story. Many of
these inner urban population declines reflected processes of urban decentralization in which
residents took advantage of new transportation technologies to relocate to suburban loca-
tions. Large numbers of people moved from inner-city locations to more suburban locations
or to satellite towns and cities, the result of which was that many metropolitan areas actually
increased in terms of their geographical areal coverage, irrespective of whether their total
population levels were falling or increasing. As such, these urban decentralization processes
often tended to be associated with ever-increasing commuting distances into the urban cen-
tres, issues which we now know have contributed to climate change.
As many cities expanded in terms of their geographical areas, there also emerged new
types of commuting patterns in which residents no longer commuted to central-city loca-
tions, but increasingly were employed in suburban or satellite locations. By the 1980s many
urban planners, architectural historians, and sociologists were confidently espousing the
emerging trends towards so-called (post-industrial' residential patterns of living, whereby
changes in technologies and work practices meant that workers would increasingly become
freed from the requirements of proximity and accessibility. It was assumed that workers
would increasingly become free to choose where to live and work on the basis of their life-
style and amenity preferences, rather than according to the traditional requirements oftheir
employers for proximity and density. The results of these new-found freedoms would be
that cities would decline in importance as geography became less important. However, as
we see in Box 8.2, in reality, very different spatial outcomes are also likely, giving rise to
spatial concentration.
Many of the urban decentralization trends evident in the 1970s and 1980s went into
reverse during the 1990s when increasing urbanization was not only observed in developing
countries, but once again was also observed in many advanced and rich countries. On many
indicators, cities appear to have once again re-emerged in economic importance (Glaeser
1998; Glaeser and Gottlieb 2009) at precisely the moment when the arguments ofmany com-
mentators would suggest that they should be dying, or at least significantly diminishing in
importance. Why this should be the case is a result of the shocks associated with the new era
ofglobalization, an era which began at the end ofthe 1980s, and which instilled a complex set
oftechnological and institutional transformations which have changed all aspects ofthe eco-
nomic geography ofthe global economy. Some ofthese changes will be outlined here, and we
will return to these issues in much more detail in Chapters 9 and 10. As we will see, the actual
cn~~
urban and regional economic changes taking place in the modern era of globalization are
very complex, and no simple worldwide urban and regional pattern has been emerging.
The modern era of globalization can be considered on many dimensions to have arisen dur-
ing the period 1989-1994, and as such is approximately still only less than three decades old.
The period 1989-1994 was a period which witnessed enormous institutional transforma-
tions, including: the fall ofthe Berlin Wall in 1989; the accelerated opening up ofthe Chinese
economy after 1989; the reintegration ofSouth Africa into the global system after the release
of Nelson Mandela in 1990; the second industrial reforms which took place in 1991 in both
India and Indonesia; the creation of the European Single Market in 1991/1992; the estab-
lishment of the North American Free Trade Agreement (NAFTA) in 1994; the adoption of
the new constitution of Brazil in 1988 and the flotation of its new currency, the real, in 1994.
All these institutional changes led to a transformation in the global economy.
The opening up of the so-called BRIICS group of countries, namely of Brazil, Russia,
India, Indonesia, China, and South Africa (GECD 2008), meant that these large-population
nations for the first time began to playa very major role in the global labour market. Since
1990 the opening up of the former transition economies has brought some 260 million new
workers into the global labour market, China some 760 million new workers, and India
another 440 million new workers (Venables 2006). As such, a major difference between the
current phase of globalization and all previous eras of globalization is simply in terms of
both the speed and the order of magnitude of the changes (Crafts 2004; MacGillivray 2006).
From the perspective of labour market integration, the period 1989-1994 represented the
largest shift in global labour markets in economic history, with approximately one-third of
the global labour market for the first time emerging into the global economic system within
a very short time period.
The global institutional changes encouraged the rapid integration of these huge countries
into the global economy and provided an enormous range of new possibilities for both the
supply of cheap labour and for new consumption opportunities. These possibilities and
opportunities were magnified and accelerated by the invention in 1991 of the world wide
web (www) interface based on the http protocol which was developed by Tim Berners-Lee.
This technological breakthrough for the first time allowed many different technological
developments to be integrated into a single electronic platform. Many different transporta-
tion and communications technologies had been developing apace over the two or three
decades before the 1990s, most notably technologies such as containerization (Levinson
2006), GPS global positioning systems, mobile phone technologies, electronic file sharing,
but the invention of the world wide web allowed these technologies to be integrated and
exploited in a vast array of new ways which were unimaginable only a few years earlier.
The major beneficiaries of these combined institutional and technological transformations
were those companies which already had the skills and knowledge to exploit these transforma-
tions, namely multinational companies. The competitive advantages of multinational firms
are precisely that they are able to exploit any emerging global market opportunities by devel-
oping highly integrated global supply, marketing, and control systems, and by also controlling
these global systems very tightly so that they are best able to exploit their corporate knowledge.
The modern era of globalization has proved to be an enormous boost to these firms, which
have responded to the transformations by an enormous increase in the extent to which they
outsource and off-shore many oftheir activities. Multinational enterprises (MNEs) are now in
the vanguard of the current phase of globalization, connecting cities and regions to a higher
degree than ever before in terms of all kinds offlows ofknowledge, human capital, investment
finance, goods, and services (McCann and Acs 2011).
In this modern era of globalization the geographical behaviour of these multinational
firms and global corporations exhibits two dominant characteristics. First, multinational
firms, rather than being completely global in outlook, are actually overwhelmingly regional
in nature, in the sense that their sales, investments, and R&D are dominated by the same
'super-regions' or blocs or groupings of adjacent countries in which their parent companies
are located (Rugman 2000, 2005). The spatial patterns of all forms of their cross-border
investments tend to be concentrated in the same super-regions from which the multina-
tional firms emerge. For example, ifwe take the case ofthe three major global super-regional
blocs of countries in the global economy, namely those of NAFTA (USA, Canada and
Mexico), the European Union, and East Asia, we find that the average same-regional sales
share of the world's top 500 companies is over 70 per cent (Rugman 2005). In other words,
a US multinational firm will typically conduct over 70 per cent of its activities within the
NAFTA arena, while a French multinational firm will typically conduct over 70 per cent of
its activities within the European Union. Similar evidence from patent citation data sug-
gests that firms typically gain the vast majority of their knowledge within the local region
and the same country (Peri 2005). At the same time, the geographical patterns of the bilat-
eral investment treaties and double taxation treaties are also very spatially concentrated
(UNCTAD 2003), thereby reinforcing the regionalizing trends of the global companies.
These observations all suggest that modern globalization is more accurately described as
'global regionalism'.
The second geographical aspect of the concentration of multinational activities is that
knowledge-related activities, in particular, tend to be concentrated in certain key 'global
cities' within these same super-regions from which the firm emanates, and there is much
evidence ofthe increasing role ofglobal cities (Sassen 1994,2002; Taylor 2004; Ni and Kresl
2010) in the international economy. The analysis ofglobal cities suggests that in the current
phase of globalization, the links between a city and other parts of the global economy are a
key determinant of the city-region's performance.
In this particular geographical literature, which draws heavily on sociological approaches,
the importance and influence of a city in the global economic system are discussed in terms
ofthe extent ofits global 'connectivity' (Sassen 2002), where 'connectivity' refers not only to
the various aspects of the knowledge and information exchanges which take place between
particular locations, but also to the discretionary decision-making power to act on those
knowledge exchanges (Ni and Kresl 2010). As such, global connectivity may be manifested
via a variety of mechanisms, such as corporate headquarter functions, corporate decision-
making linkages, human-capital mobility patterns, trade linkages, transport linkages, finan-
ciallinkages, and asset management roles (Taylor 2004; Sassen 2006). Network analyses of
trade and knowledge indicators imply that there has emerged a new core-periphery hierar-
chical structure to international trading patterns (Kali and Reyes 2006; Reyes et al. 2009»
and the knowledge reach of technologically leading regions is already far greater than for
other regions (Peri 2005). It is argued that by locating in these centres global firms will
therefore increasingly reap the economic rents associated with their knowledge assets via
the exploitation of the genuinely global production) communication) and financial net-
works (Coleman 1996; Cohen 1998; Zook 2005) which are facilitated by these global city
systems. Estimates suggest that) since 1990) one-third of economic growth in the USA was
accounted for by its multinational companies (MGI 2010b). Understanding the location
behaviour of these firms is therefore also critical for understanding the geography of eco-
nomic growth. We will return to these issues in much more detail in Chapter 9.
In the current era ofglobalization) two ofthe most important features ofthe emerging urban
and regional patterns ofeconomic activity therefore appear to be a trend towards the develop-
ment of cross-border global regions) and also the growing importance of the global cities
embedded within these global regions. This suggests that globalization involves a newly emerg-
ing economic geography of cities and regions in which proximity and accessibility are still
essential features. But the key difference between the economic relationships afforded by
recent technological and institutional changes and earlier eras ofglobalization is that the criti-
cal scale of the relationships which operate nowadays between cities) regions) and nations has
increased dramatically. Yet modern technological and institutional changes do not imply that
the world is getting flatter) and nor do they imply that growing cities are the solution to devel-
opment problems. Rather) these changes imply that the geography ofurban and regional econ-
omies is nowadays evolving in complex and heterogeneous ways) due to ongoing processes of
international economic integration) technological development) and institutional reforms.
For our purposes at this stage it is sufficient to note six features ofthe relationship between
globalization) urbanization) and industrialization:
(i) There has been a clear long-run relationship between increasing urbanization)
industrialization) and globalization since the early sixteenth century) a relationship
which was stalled in many rich countries during the early twentieth century.
(ii) The long-run relationship between globalization and urbanization has re-emerged in
the developed world in the latter decades of the twentieth century) after having been
stalled or even going in reverse for many deacdes in some rich countries.
(iii) The re-emergence or urbanization in rich countries is associated with the
reintegration of the global economy after decades of anti-globalization processes.
(iv) Rising urbanization and the growth of mega-cities in the developing world has been a
major feature of the second phase of global urbanization) a phase which began after
the Second World War and continues to this day.
(v) The modern era of globalization is associated with the super-regional processes of
integration between groups of neighbouring countries) leading to global regionalism
rather than globalization.
(vi) At the core of the newly emerging super-regions are global cities) cities whose trade
linkages are dominated by the presence of multinational companies.
These six features demonstrate why understanding the economic role ofcities and regions
is critical for understanding economic growth and trade. However, simple parallels across
history have to be treated carefully. In many ways, at first sight the features of modern glo-
balization appear reminiscent ofearlier eras of globalization. In particular, modern globali-
zation appears to have much in common with the late nineteenth-century and early
twentieth-century period during which many of the most influential early writers on urban
and regional issues were writing about the importance ofcities and agglomeration. However,
as we will see in Chapter 9, in many ways the modern era ofglobalization is also quite differ-
ent from earlier eras, and an understanding of these differences also changes how we view
the long-run national and international economic role played by cities and regions in
today's global economy.
8.6 Conclusions
The emergence of global firms, global cities, and global regions provides the modern
backdrop and context in which we examine urban and regional economics. The enor-
mous changes wrought by modern globalization since the early 1990s, in which national
economies have again opened up to each other in much more profound ways than ever
before, requires us to consider the role and performance played by cities and regions at
the sub-national level in the light of the behaviour of global cities and global regions
operating at the international level. Cities and regions are again at centre stage in many
discussions of growth and development, and this reinforces the timeliness of studying
urban and regional economics. However, while the focus on cities and regions has re-
emerged, this does not necessarily imply that economic growth is simply about agglom-
eration economies driving the growth of the largest cities and the richest countries, as
it was at the beginning of the twentieth century. While this may be the case nowadays
in developing countries, the fact that so few of the world's largest cities are in advanced
and rich countries suggests that in these countries the picture may be rather more com-
plex. In many ways our economic thinking is heavily shaped by the models and meth-
ods we employ, many of which were originally developed in the light of the early
twentieth-century experience of cities. Yet, as we will see in Chapter 9, for many wealthy
countries urban scale no longer appears to be as good an indicator of growth and devel-
opment as it was in previous eras, whereas in poor and developing countries this rela-
tionship still seems to hold. Understanding these complex emerging patterns therefore
requires us to consider all the analytical issues discussed in Part I of this book in the
light of the long-run global trends discussed in this chapter. These issues are the topics
discussed in Chapter 9. Finally, the changes due to modern globalization regarding the
role played by cities and regions in economic development also require us to rethink
many of the traditional debates regarding urban and regional policy. As we will see in
Chapter 10, the thinking surrounding urban and regional economic policy has evolved
rapidly over recent years as the application of the theoretical tools discussed in this
book to the new global realities is providing new insights, new challenges, and new
opportunities.
Discussion questions
8.1 Discuss the emerging long-run relationships between globalization, urbanization,
and industrialization which were evident in different parts of the world between the
sixteenth century and the early twentieth century.
8.2 Why are the urban size observations of the early twentieth century important for
understanding the nature of today's urban and regional economic theory?
8.3 What were the reasons for the long period of anti-globalization which was evident in
much of the twentieth century? Discuss how this affected the growth of cities.
8.4 Can you describe the long-run patterns of urban development in your own country?
8.5 Discuss the ways in which the global urbanization patterns in the second half of the
twentieth century differ from those in previous eras.
8.6 What are the major characteristics of the world's largest cities today?
8.7 What are the major characteristics of the largest cities in your own country and
how important do you think they are for your own country's national economic
development?
8.8 In what ways has your understanding of the processes driving national and international
economic growth changed by discussing the economics of cities and regions?
Cities and regions in the
modern global economy
9.1 Introduction
As we saw in Chapter 8, there is a great deal of evidence that cities are now once again play-
ing an increasingly important role in the economic development of many countries. In
particular, as we saw in Chapter 2, in the modern era of globalization there is much evi-
dence to suggest that the increasing importance of cities and agglomeration economies is
related to the production and acquisition of knowledge activities. However, in spite of
many popular descriptions, this does not necessarily imply that cities are generally grow-
ing, or indeed that cities are the dominant engines of today's national economic growth. As
we will see in this chapter, the picture is far more complex than this. The role played by
cities and regions is different in different contexts, and is also changing in different ways in
different situations.
In this chapter we will examine why cities and regions have re-emerged in importance
during the modern era ofglobalization, an era which began in the late 1980s. We will exam-
ine the particular features of these newly emerging city-region roles, and discuss the eco-
nomic implications of these for national and regional development. The key here is to
understand the changing role played by the transactions costs of geography, or what we can
call spatial transactions costs. In Chapters 1 to 4 we discussed various aspects of spatial
transactions costs, such as transport costs, logistics costs, knowledge exchanges, and infor-
mation costs, and an understanding of these costs is essential for understanding location
behaviour in cities, region, and clusters. For our purposes here, however, it is also critical for
us to understand that over recent decades the nature and levels of the spatial transactions
costs have fundamentally changed in ways which were previously unimaginable, and these
changes have had profound impacts on all aspects of the relationships between cities,
regions, and countries.
The chapter begins with a description of different types ofspatial transactions costs and an
examination of the changes in these over recent years. A simple model which builds directly
on the material in Chapters 1 to 4 is presented, and this is used to explain how a combination
of rising and falling spatial transactions costs means that the geography of the economy
becomes more uneven, more spiky, and less flat. After examining the empirical evidence on
these matters, we then use the theoretical models and the empirical evidence to capture the
modern emergence of global cities and global regions and their various features.
9.2 Different types of spatial transactions costs
The costs associated with engaging in) and coordinating) economic activities and market
mechanisms across geography can be termed spatial transactions costs. Different types of
spatial transactions costs have been changed by modern globalization in different ways) and
it is important to understand these changes.
A particular subset of spatial transactions costs are those costs which are directly related
to the costs of moving goods or information across space. These costs are dependent on
communications and transportation technologies) and as a combined group we can refer to
these as spatial transmission costs. As we discussed in detail in Chapters 1 to 4) transporta-
tion costs are the simplest expression of spatial transactions costs) but other spatial transac-
tions costs involve the acquisition and transmission of knowledge and information across
space) and these are extremely important in the modern economy. In order to understand
the ways in which spatial transactions costs have changed over recent years we need to split
up spatial transactions costs into three different types.
The first type of spatial transactions costs are the transactions costs associated with mov-
ing goods across geographical space. These are classic transportation costs. The second type
are the transactions costs associated with moving knowledge and information across geo-
graphical space) and for the purposes of this book we will call these knowledge-information
transmission costs. Both the first and second types of spatial transactions costs) namely
transportation costs and knowledge-information transmission costs) are explicitly geo-
graphical in their construction) in that the costs incurred by these transactions always
depend on the geographical distance covered. The third type of spatial transactions costs are
the transactions costs associated with movements across national borders. These are typi-
cally in the form of customs charges or tariffs) which are the institutional costs associated
with a particular border crossing.
These border tariffs and trade barriers are central to models of international economics.
Since the late 1980s) the international institutional changes outlined in Chapter 8 which
have progressively opened up international markets mean that the levels of these tariffs are
generally falling. These falling trade barriers allow for greater degrees of openness and inte-
gration between countries) and increase the possibilities for trade) outsourcing) offshoring
and multinationalism. In some cases the falls in tariffs are quite significant and in some
cases they are very slow; the rate of change depends on progress of the trade negotiations
taking place within international arenas such as the W orId Trade Organization) the
European Union) and NAFTA) along with numerous individual bilateral arrangements
between particular countries regarding specific industrial sectors.
However) these tariff costs are not explicitly defined in geographical terms in the same
way as the first two sets of spatial transactions costs) in the sense that tariffs are not defined
to vary systematically with the distance travelled before or after arriving at a particular
institutional border. As such) from the perspective of urban and regional economics we
can consider border tariff costs as fundamentally aspatial in nature) although explicitly
geographical in terms of their implementation (McCann 2005). The result of these differ-
ences is that falling cross-border trade tariffs are not particularly related to the uneven-
ness of the global spatial economy described in this chapter) whereas the first two types of
THE ful0DERN ECONOMY
spatial transactions costs are essential elements of the explanations for the unevenness
and spikiness that we observe.
In the following sections, which draw heavily on McCann (2007, 2008), we consider
mainly the changes in the first and second types of spatial transactions costs, namely those
transactions costs which are explicitly dependent on distance, while at the same time assum-
ing that tariffs are in general falling slowly worldwide, or at least are not increasing. What
we will observe in the following sections is that spatial transmission costs have fallen dra-
matically over recent years, but that this does not necessarily also imply that spatial transac-
tions costs have fallen. Indeed, much evidence suggests the exact opposite, namely that
many forms of spatial transactions costs have risen. It is the combination of falling spatial
transmission costs and rising spatial transactions costs which over recent years has favoured
certain places over others. This is the issue to which we now turn.
Since the late 1980s we have seen dramatic improvements in the ability of decision-makers
to coordinate activities across space. The primary reason for these improvements has been
the enormous technological developments in information and communications technolo-
gies. Information technologies employing digital, satellite, and fibre-optic technologies
allow for vastly greater quantities of information to be transmitted across space at much
lower costs than was previously imaginable, even just a couple of decades ago. These devel-
opments have both increased the market access for individual firms and meant that com-
plex operations across diverse locations can now be managed both more efficiently and
effectively than was previously possible.
For industries competing on knowledge and trading primarily in information, such as
finance, advertising, marketing, and tourism, modern information technologies provide
entirely new possibilities for the supply of information-based services across a global mar-
ket space. The level of global market access has therefore increased dramatically for huge
numbers of firms trading in knowledge and information-based services, many of which are
in dominant (knowledge hub cities and regions' (OECD 201Ib). At the same time, these
improved technologies also allow decision-makers to undertake the coordination of com-
plex spatial arrangements of activities in ways which were previously not possible. Probably
the most noticeable outcome of these new possibilities and opportunities is the increased
offshoring and outsourcing of many types of activities in both manufacturing and service
industries.
In service industries notable examples of this include international accounting, whereby
New York and London banks transfer their book-keeping requirements overnight to firms
in Ireland or India, in order to have them updated in time for the opening of the money
markets the next day. Other examples include Silicon Valley firms which subcontract soft-
ware development activities to firms in Bangalore, India, while maintaining daily contact
and control of the Indian software development process from California. These observa-
tions all imply that knowledge-information transmission costs must have fallen dramati-
cally over recent decades (McCann 2008).
In manufacturing, advanced communications and control technologies have been widely
applied to the management ofsupply chains, to production and just-in-time (JIT) inventory
L"_~~".$""" -'" AN D
scheduling control systems, and to logistics and distribution operations. The types of firms
which particularly benefit from these technologies are those firms requiring the precise
coordination of complex networks of production and distribution operations across large
geographical distances, and these firms operate in many different sectors. Examples here
include firms employing advanced roll-on, roll-off trucking and containerization technolo-
gies (Levinson 2006), rapid-turnaround shipping, the widespread use of sophisticated
global positioning systems, and the increased speed and efficiency of all forms of air trans-
port technologies.
The various communications and transportation developments allied with increased
outsourcing and offshoring have allowed production systems and corporate organizational
networks to be greatly disaggregated spatially across both regions and countries. These
observations therefore all point to dramatic falls over recent decades in both the costs of
transmitting information across space and of moving goods or people across space. All
these transport and movement costs can collectively be termed spatial transmission costs,
and almost all of the available evidence suggests that they have unambiguously fallen, con-
tributing to a Friedman-type 'flattening' of the spatial economy, as described in Box 8.2.
The theoretical argument that spatial transactions costs have increased while spatial trans-
mission costs have fallen is that improvements in information technologies themselves
increase the quantity, variety, and complexity of the knowledge handled and information
produced. The increased quantity, variety, and complexity of the knowledge handled and
information produced itself increases the costs associated with acquiring and then transact-
ing this knowledge across space. This is because much of the information will originally
have emerged from knowledge which is of a non-standardized tacit nature, as discussed in
Chapter 2, and the acquisition and transmission of the type of information produced from
tacit knowledge increasingly requires greater levels of face-to-face contact in order to main-
tain mutual trust and understanding (Gaspar and Glaeser 1998; Storper and Venables
2004).
Following the arguments in Chapters 2 and 7, one way of thinking about this is to assume
that firms in advanced economies are increasingly competing in terms of high knowledge
content, high value-added, and shorter product life cycles, services, and activities, precisely
so as to avoid the increasing competition from the newly developing countries. The increased
importance of shorter product life cycles is also associated with an increased importance of
face-to-face contact in order to transact the knowledge required for producing and market-
ing these products. As such, for many of these types of knowledge-intensive activities the
required frequency offace-to-face interaction will have increased over recent years (Storper
and Venables 2004; McCann 2007) because the time (opportunity) costs associated with not
having continuous face-to-face contact will also have increased with the quantity, variety,
and complexity of the information produced (McCann 2007). Similar arguments apply to
the speed and frequency of manufacturing and logistics operations (McCann 1993, 1998)
and also to social capital interactions (McCann et al. 2010).
In location theory terms, the outcome is that, in equilibrium, the optimized frequency of
interaction and transactions across space will have increased (McCann 2007; Rietveld and
Vickerman 2004). This in turn increases the spatial transactions costs for any given distance
over which communication and interaction take place (McCann 2007). The features of
these interactions also lead to a natural sorting of firms according to the types of knowledge
exchanges and innovation outcomes they generate (Doloreux and Shearmur 2010). As
advanced economies increasingly shift towards competition- and production-based knowl-
edge-intensive activities where product life cycles are becoming shorter, these time costs
become ever more important, the associated costs of distance increase, and a more pro-
nounced spatial sorting of innovation activities naturally emerges.
Another way of thinking about this is to assume that lower transport costs imply that
firms will increasingly switch to the production of higher-quality customized goods
(Duranton and Storper 2007), for which the costs of providing a given level of service qual-
ity not only becomes more costly with distance, but whose sensitivity to distance is greater
than for standardised products (Duranton and Storper 2008). This is akin to a knowledge-
service quality version of the iceberg model discussed in Appendix 3.3. However, the ques-
tion remains as to why firms would adopt this strategy. Following the Palander-Hotelling
framework in Chapter 1, one argument is that this is a rational strategy for redefining and
using space to bolster a local monopoly position, and avoiding competition from newly
emerging economies which produce more standardized, cheaper and lower-value-added
goods and services. However, there are also counter-arguments which would suggest that
rational firms ought to exploit falling information transmission costs to maximize their
global economies of scale, as many high-technology multinational firms already do.
Whichever theoretical approach we adopt, however, these arguments all suggest that as
advanced industrialized economies increasingly shift into the production of goods and ser-
vices embodying complex knowledge and information, spatial transactions costs as a whole
will have increased in many sectors and activities, even though the spatial transmission costs
of moving information, goods, and people have fallen. As we see in Box 9.1, the reason is
that while the costs of transmitting information, shipping goods, or transporting people
have fallen dramatically, the costs of transacting knowledge via human-capital interactions
(Glaeser and Kohlhase 2004) have actually risen.
&.. '&-.ll&%..&'8:'ll ..1'. AND
MODERN ECONOMY
REGiONS~
The arguments in the above sections suggest that over recent decades many types of spatial
transactions costs have increased at exactly the same time as many types of spatial transmis-
sion costs have decreased. While at first these arguments may appear to be rather contradic-
tory or paradoxical, these apparently conflicting conclusions can be easily reconciled by using
the arguments outlined in Chapters 2 and 3, and applying them to the type of theoretical
framework discussed in Chapter 4. The combination ofthese two opposing forces in different
contexts generates increasing unevenness, increasing spikiness, or more technically increas-
ing convexity, and this can be explained with the help of two diagrams (McCann 2008).
Figure 9.1 depicts a one-dimensional model ofthe economy in which the economic geog-
raphy comprises three city-regions spanning a distance AB. In this particular spatial econ-
omy there are three cities, X, Y, and Z. Cities X and Z are larger than city Y, and all three
cities exhibit two types of production, namely high-value goods and services Hand low-
value goods and services L.
Following the arguments in Chapter 4, the associated bid-rent curves for the production
of each respective good produced by each city are denoted in Figure 9.1 as BRxH for the
AND iN
$ $
BRXH
BRZH
X Y
:~ .; :.-.:
I I
: XH : YH :
XL YL
high-value goods produced in city X, BR xL for the low-value goods produced in city X, BR YH
for the high-value goods produced in city Y and BR YL for the low-value goods produced in
city Y, BRzH for the high-value goods produced in city Z and BR zL for the low-value goods
produced in city Z, respectively. As cities X and Z are larger than city Y, this implies that
there is both a larger local regional hinterland market and also more competition for each
good in cities X and Z than in Y. The result is that at the central-city market locations of X
and Z, land prices are higher for the production of each good Hand L than at Y, and the
bid-rent curves for both goods Hand L produced by X and Z extend further than those
associated with Y.
The urban and regional economic geography of the economy AB is such that the spatial
production area for good L in city Y, denoted as Y L, accounts for the less than one-third of
the total economic geography of production of the low-value good L. Meanwhile, cities X
and Z each account for more than one-third of the economic geography of production of
the low-value good L, with the respective production areas denoted as XL and ZL. In addi-
tion, city Y also has a very small local regional production area YH for good H, while cities X
and Z have much larger local regional production areas, denoted as X H and ZH' for the high-
value good H. Following the arguments in Chapter 4, the equilibrium land prices at each
location in the global economy AB are given by the envelope of the individual local bid-rent
curves, depicted in Figure 9.1 in bold.
Figure 9.2 depicts the situation in response to the types of globalizing and localizing
trends discussed above. As we have seen, falling spatial transmission costs apply primarily
to low-value goods and services which are relatively standardized in nature and whose
product life cycles are not very short. The falling spatial transmission costs for these goods
$ $
AV ; , , ~ Z'" '>JB
x :c y: 0: Z
~ .-: :~ .-:
XH
:.
Low-value goods L .' ZH
mean that the equilibrium bid-rent curves for the production of these goods becomes shal-
lower. In Figure 9.2, the shallower-sloping bid-rent functions for the production of low-
value goods and services L imply that such activities can be profitably undertaken in all
locations. This is the spatial analogy of Friedman's (2007) flat-world argument discussed in
Box 8.2. However, from urban and regional economics we know that this is not the end of
the story. The reason is that the actual distribution of production and activities depends on
the interaction between this flattening effect and the other opposing localization trends
discussed above.
If we consider the localizing tendencies operating with regard to the production of high -
value non-standardized goods and services H, the face-to-face arguments above imply that
the production ofthese goods and services will benefit from location-specific agglomeration
economies. The existence of agglomeration economy advantages as well as increasing spa-
tial transactions costs in the production of high-value goods and services H means that the
land in the major urban areas X and Z is increasingly dominated by the production of these
high-value goods. Cities X and Z expand outwards, the production of high-value goods is
concentrated in just two locations X and Z, and the bid-rent functions associated with the
production of these high-value goods become steeper. The production location Y, which
previously had a local dominance in both the production of low-value L and high-value H
goods for its own hinterland area, now disappears as an independent production centre.
Instead, the whole region between C and D now continues to function simply as a location
for producing low-value goods and services L.
The region CD corresponds to what Leamer (2007) describes as the 'flat terrain' of the
global economy, while X and Z correspond to what he describes as the 'hills and mountains'
of the global economy. As Fujita et al. (1999) demonstrate, the reason for this is that when
agglomeration economies are allied with falling spatial transmission costs, the existence of
agglomeration economies can lead to situations in which intermediate production loca-
tions such as Y disappear altogether. This is partly because the previous economic raison
d' etre of the former intermediate production location now disappears, and also because the
low distance costs now no longer provide any possibility for local producers to be 'pro-
tected) by distance from the producers in the major urban centres. The local and regional
producers lose their limited local monopoly power, as described in Chapter 1) and the result
is that the core central locations dominate the spatial economy.
The equilibrium land prices at each location in the global economy AB are given by the
envelope of the individual bid-rent curves) and once again) as with Figure 9.1, these are
depicted in Figure 9.2 in bold. As we see, the envelope rental curves are now far more curved
than previously. In particular, they are now far more convex) to use the urban economic
terminology. Assuming that land prices and labour prices are highly correlated, because of
the need to maintain local real wages at competitive levels, as described in Chapter 6, we can
assume that the envelope land-price curves also closely resemble the spatial variations in
both local incomes and local productivity levels.
The model predicts that differences between the major centres and the relatively smaller
centres will therefore tend to increase, as both wages and real-estate prices increase in the
core-city regions. This is not to say that all major centres will increase relative to smaller
centres, as it will also depend on the range of technologies and industries evident in particu-
1ar cities, as different industries and technologies rise and fall over time. However, the argu-
ments outlined above do imply that the modern globalization processes will have led to
increasing differences between the fortunes of regions and cities even within the same coun-
try. This is because particular major urban centres will have benefited from the increasing
scale advantages associated with being nodes in global trade networks.
The new economic geography (NEG) frameworks (Krugman and Venables 1995; Fujita et
ale 1999) discussed in Chapter 3 also demonstrate that in the presence ofagglomeration econ-
omies exactly the same type of result can be generated in a situation where the border tariffs
between countries of different sizes are reduced or removed, even where international factor
mobility is limited or non-existent. As we discussed in Chapter 8, in the modern global econ-
omy, the dramatic changes in spatial transactions costs driven by developments in transport
and communications technologies have also taken place at a time ofwidespread falls in trade
and tariff barriers associated with institutional reforms relating to integrated market areas
such as the ED and NAFTA, as well as to taxation and investment. As such, the combination
of both technological changes and cross-border institutional changes reinforces the whole
process towards unevenness, spikiness, and convexity.
The current technological and institutional changes which have taken place since the late
1980s have led to major changes in spatial transactions costs, the effects ofwhich, as we have
already seen, have favoured certain types of places over others. Some regions will be of
increasing economic importance while others will be of declining economic importance.
The emergence and re-emergence of major cities in many parts of the global economy, as
outlined in Chapter 8, is one part of this story. Cities and densely populated regions emerge
in the peaks depicted in Figure 9.2, often at the expense of other intervening regions.
However, this diagrammatic description of the arguments regarding the recent geographi-
cal effects of changes in transactions and transmission costs does not imply that divergence
will continue to increase between all regions as large global cities gain at the expense ofother
types of regions and cities.
In reality, any new rounds of unevenness which are generated by the current phase of
globalization are superimposed on the previous inherited spatial structure (Warf 1995). As
such, the urban and regional legacy of the history of economic development will also alter
the local impacts ofglobalization. This superposition of different effects may lead to increas-
ing interregional divergence in some cases and to increasing interregional convergence in
other cases, depending exactly on which new industries develop where and in which regions
firms choose to locate their new investments. In other words, the actual outcomes of these
changes on any particular region or country will depend on the structure and history of the
region, the experience of the neighbouring regions or countries, the existing spatial struc-
ture, and the actual scale of the local technological and institutional changes. In reality, the
balance of these two opposing effects encouraging convergence and divergence is really a
matter of observation and empirics rather than of theory, and this is the issue to which we
now turn.
As we saw in Chapter 8, we are now witnessing an unprecedented level of urban growth
which is taking place at the global scale (PWC 2009; MGI 2011a). This suggests that in many
parts ofthe world it is becoming increasingly important for firms and people to be clustered
together. In the newly emerging and developing countries in particular, this appears to be
an almost ubiquitous phenomenon (MGI 2009, 2010, 2011 b). In developing countries, both
the increasing levels of overall urbanization and the growth of mega-cities are most marked
(MGI 2011a), and over the next two decades the vast majority of economic growth in devel-
oping countries is expected to come from cities (MGI 2011; PWC 2009). In the developing
world cities, the patterns of clustering relate to both the high-skilled and highly educated
workers, and also to very large numbers of the poorest and lowest-educated groups.
Similar patterns of regional factor reallocation in favour oflarge and dominant cities took
place on both sides of the Atlantic during the 1990s. Within Europe, investment capital and
information became more concentrated in capital cities and large urban centres during the
1990s (Rodriguez-Pose 1998), as it did in the USA (Glaeser 2011). Moreover, empirical
evidence at the US county level (Partridge et al. 2007) and European sub-national regional
level (Caniels and Verspagen 2003) over the same period has also shown that local growth
was directly related to an area's proximity to major urban centres. Either a location in, or
good access to, major urban centres is important for productivity performance on both
sides of the Atlantic.
In advanced countries, urban clustering is most notably associated with the spatial con-
centration of particular types of high-knowledge firms and high-knowledge workers. This
effect appears to be very strong in the case of the USA. Following the human- capital-
migration arguments in Chapters 6 and 7, recent US evidence suggests that not only is there
an increasing share of university-educated human capital living and working in cities (Berry
and Glaeser 2005), but that this proportion of university-educated workers is also correlated
with the existing human-capital stock and both are correlated with the growth of the city
(Glaeser and Shapiro 2003). The result of this is that US cities are actually becoming more
dissimilar in terms of their human-capital composition. In the USA there is no evidence of
the levels ofhigh-school human capital playing any role whatsoever in city growth (Shapiro
2006), whereas the evidence regarding role played by tertiary-educated workers is very
strong. Similar evidence for the UK (Faggian and McCann 2006, 2009) also points to a clear
link between tertiary-educated human-capital migration and regional performance.
The reasons for the productivity premium of today's cities are therefore not just scale
related, but are also related to composition effects. In particular, the increasing interregional
mobility of highly skilled human capital appears to be a key driving force underlying the
reasons why interregional divergence has been reappearing in many countries over the last
two decades, after the previous four decades which had been characterized primarily by
interregional convergence (Berry and Glaeser 2005).
Of the industrialized countries, the USA is the most urbanized large country, followed
by Japan and Korea, and Canada, Australia, and the Netherlands on a much smaller scale
(OECD 2011b). The USA is more urbanized than Western Europe, and US cities on aver-
age are larger than European cities. An estimated over 84 per cent of the US economy is
accounted for by the output of cities with over 150,000 people, whereas the equivalent
share for Western Europe is 64 per cent (MGI 2012). Moreover, US cities exhibit a 35 per
cent GDP per capita premium relative to non-urban areas within the USA, whereas the
equivalent for Western European cities is 30 per cent (MGI 2012; McCann and Acs 2011).
Productivity gains driven by urbanization and large city growth still appear to be ongoing
in the USA (Glaeser and Gottlieb 2009).
The reasons why US cities outperform European cities are complex, and are likely to be a
mixture of scale-related, composition-related, and policy-related factors. US cities on aver-
age are larger than European cities, and the US economy is more urbanized than Europe
(MGI 2012). But city size and urbanization alone cannot account for productivity differ-
THE GLOBAL ~rl§"I&..~,nlilk,l!!Y
ences. On the one hand, in terms of city size, the urban hierarchy arguments of Chapter 3
suggest that the roles and functions of cities also relate to their relative position in the
national urban system. On this criterion, US cities are not particularly large relative to US
national populations, in comparison to many other 0 ECD countries. On the other hand, in
terms of urbanization, much of Europe outperforms Japan and Korea in terms of produc-
tivity, although both Japan and Korea are far more urbanized than almost anywhere in
Europe, and many of the most productive parts of Northern Europe exhibit small cities and
low population densities. As such, differences in urban scale, differences in the levels of
urbanization, and differences in interregional adjustment processes cannot entirely account
for the transatlantic differences in urban productivity.
A quite different literature on the transatlantic productivity gap (Timmer et al. 2010;
Ortega-Argiles 2012) suggests that the productivity advantage of the USA over Europe was
primarily that the US economy was much more efficient in terms of the generation and dif-
fusion of new technologies throughout the economy during the 1990s and 2000s than
Europe. These sectoral and technology-based arguments emphasize the greater integration
of pan-US markets, and the role of US multinational companies in particular (MGI 2010),
in promoting technology transmission across the economy, rather than the more frag-
mented pan-European markets. As such, these arguments emphasize transmission mecha-
nisms associated with firms, sectors, and technologies, rather than cities and regions.
What is still very unclear is exactly how these various industry, sector, and technology
issues relate to urban and regional issues. Are US markets more integrated because US cities
are bigger? Are US cities larger because US markets are more integrated? What is the reia-
tionship between either of these phenomena and the role of US multinational firms? These
issues are still to be resolved. Moreover, other arguments emphasize the growth role played
by the entrepreneurship and innovation effects of small firms, many of which benefit from
being in cities and having connections to multinational firms. How each of these different
firm, sector, and technology issues relate to urban productivity and growth issues is still to
be investigated, but clues as to how they might be connected come from the cluster-
transactions costs arguments in Chapter 2.
In Chapter 2 we saw that there are alternative mechanisms for knowledge exchange and
transmission in which the location priorities for large firms differ from those of small firms.
For small firms the local context, including access to local large firms, is likely to be impor-
tant as a source of both market knowledge and inputs. In contrast, for large firms a wider
national and global context is likely to be more important for market knowledge, while the
local environment serves to improve knowledge of specific inputs. However, large firms will
also seek inputs from a wider spatial extent. Therefore the knowledge spillover, knowledge
exchange, and knowledge transmission effects operating within places, between places,
within sectors, and between sectors will also depend on the organizational make-up of the
firms. It is likely that both urban-regional scale and urban-regional connectivity will become
important, and we will now deal with each of these issues in turn.
have been a crucial element of national wealth generation since the early seventeenth
century. However) the relationship between urban scale and national wealth is no longer
the simple story of urban scale that it was when Marshall (1890) and Weber (1909) were
writing) and this requires us to interpret our theoretical frameworks carefully in the light
of the evidence.
The most economically powerful cities in the world are primarily contained within the
rich countries. According to PWC (2009)) 23 out of the 30 largest GDP cities in the world in
2008 were in the advanced countries) with only seven from the newly industrializing coun-
tries. However) when it comes to labour productivity) defined in terms of GDP per capita)
the advanced and rich countries of the world contain all the world's highest-productivity
cities.
Recent estimates of US urban areas suggest that a doubling of city size is associated with
a productivity increase ofsome 3-8 per cent (Rosenthal and Strange 2004). As such) moving
from a city of 50)000 to one of 5 million would be predicted to increase productivity by more
than 50 per cent (Venables 2006). OECD (2006) data) however) suggest that while US cities
are on average very large and also very productive) there is no general law which states that
urban scale is empirically related to urban productivity. Indeed) outside the USA) the rela-
tionship is actually negative for very large cities.
Many of these world's highest-productivity cities are not mega-cities or even very large
cities (OECD 2006). Of the world's 75 highest-productivity cities (including Singapore)
Hong Kong) and cities in Taiwan and Israel) in 2006) 29 are what the OECD (2006) classifies
as 'small metro areas' of fewer than 3 million inhabitants; 32 are what the OECD (2006)
classifies as 'medium to large metro areas' of between 3 and 6.99 million inhabitants; and
only 14 are mega-cities of at least 7 million inhabitants. Overall) among 0 ECD countries)
the relationship between the size of cities and their GDP per capita productivity appears to
exhibit something of an inverted-U-shaped relationship (OECD 2006).
Very large cities appear to exhibit diseconomies of scale associated with increasing con-
gestion and rising land and labour costs relative to the small to medium-sized cities. Among
the OECD countries there is no simple relationship between city size and city growth. In
both the developed and developing world (MGI 2011)) very large cities have not outper-
formed smaller cities at all since the mid-1990s. In fact) over recent decades) more cities in
the developed world have actually shrunk in size than the number of cities that have grown
(UN-HABITAT 2008; Dijkstra et al. 2012). Intermediate-sized cities and regions are
increasingly seen as drivers of growth. Most of the forecast growth of the US economy over
the coming decades is expected to come from growth in intermediate-sized cities (MGI
2012)) and this is also the case for many emerging parts of the world (MGI 2011).
Following on from our earlier observations) however) there appear to be qualitative dif-
ferences between places. Across the OECD) higher-income cities are actually outgrowing
lower-income cities) irrespective of their population scale (OEeD 2006)) and many of these
high-growth cities are small and medium-sized cities. Among the rich countries) 12 out of
the 15 most entrepreneurial cities are small to medium-sized cities (Acs et al. 2008)) while
11 out of the world)s 15 most competitive cities are small to medium sized (COL 2008).
Nowadays) in the advanced economies) urban productivity is clearly not just a matter of
scale as it was a century ago. In rich countries today) knowledge) creativity) innovation) and
connectivity appear to be far more important for productivity than simply urban scale. As
1II'!>~'Il~~V'8~~ ~N
As Porter (1990) points out, it is not regions which compete, but the firms which are
located in the regions. As such, the clues as to why particular cities are highly productive
must lie in the types offirms which are located there and how their activities relate to wider
global market changes. In terms of the modern era of globalization, the various arguments
above suggest that there have been qualitative changes in the role played by cities in the
industrialized world, changes which favour the competitive advantages associated with
cities as centres ofknowledge. This is particularly the case for cities in the developed world.
However, the scale and pace of modern globalization suggests that the individual city is
much less of a reference point for determining competitive advantages than it would have
been in previous eras when markets were more dominated by domestic and national issues.
For firms which invest heavily in knowledge assets, in order to generate the reqUired
returns to their knowledge investments, many of the knowledge-based firms located in
AND
today's cities must capture markets which extend well beyond the borders of their own
country. Traditionally, the returns to investing in activities related to foreign markets were
generated by exports. However, as we have already seen, one of the key features of the cur-
rent phase of globalization is that there is now an increasing premium associated with face-
to-face contact with foreign customers, suppliers, or collaborators (McCann 2007, 2008,
2011). In global markets this implies that the global engagement facilitated by direct inter-
national investment is becoming relatively far more important than exporting as a means of
global engagement (McCann 2008,2011). Qualitative differences in the role played by mod-
ern cities as centres of communication and engagement are therefore critical for under-
standing the modern economy.
As we saw in Chapter 8, one of the most widely discussed manifestations of the modern
relationship between cities and the global economy comes from the increasingly important
role played by so-called <global cities' (Sassen 1994,2002; Taylor 2004; Ni and Kresl 2010).
The increasing relative dominance of these global cities appears to be associated with the
density of knowledge (Simmie 2004) and information technology assets employed in
the city (Sassen 2002; Button et al. 2006). As such, these global cities are argued to act as the
<knowledge hubs' (OECD 20lla) of the worldwide trade, migration, knowledge, and infor-
mation networks linking all parts of the global economy. Research suggests that many of the
dominant interactions of these cities are with other similar globally oriented cities in other
countries, rather than with other cities in the same country. There is also evidence that the
performance of these dominant global cities has also been playing an increasingly impor-
tant role even with respect to their own hinterland national and continental economies
(HMT-DTI 2001, 2003; BTRE 2004; Col2005a, 2006; 2007c; Glaeser 2005).
The connectivity performance of these global cities is largely related to the scale of the
global engagement of the companies located there, and, in particular, to the globally com-
petitive multinational firms located there (Ni and Kresl 2010; McCann and Acs 2011).
Given that multinationals accounted for almost one-third ofDS economic growth over the
last two decades, global cities are therefore perceived to playa very significant role in the
modern global economy.
It is useful to try to provide a sense of how important these issues of connectivity are in
different settings, and there are various different systems ofglobal city rankings which aim to
rank cities according to their degrees of global connectivity (Taylor 2004; Taylor et al. 2008).
These global city ranking systems include the Mastercard Worldwide Centres of Commerce
rankings (Mastercard 2008), the Global Cities Competiveness rankings (Ni and Kresl 2010),
and the Global Financial Centres rankings (COL 2009; Long Finance 2012). All these differ-
ent ranking systems are based on a wide variety of connectivity data such as corporate head-
quarter functions, corporate decision-making linkages, human-capital mobility patterns,
trade linkages, transport linkages, financial linkages, and asset management roles. These are
analysed using different weighting measures, algorithms. and multivariate methods, and all
three ranking systems provide composite indicators of the level of global connectivity ofdif-
ferent cities viewed from different perspectives. The first two systems cover a range of meas-
ures of global connectivity, whereas the last focuses specifically on financial activities.
Not surprisingly, there is a very close correspondence between the level of global con-
nectivity of the cities via their multinational corporations and their GOP per capita (OECD
2006; MGI 2011; McCann and Acs 2011) and this observation is consistent with the argu-
ment that the cities with the highest levels of global connectivity are also largely the world's
most productive cities. Indeed, global connectivity has been found to be the single most
important feature of a global city competitive position (Ni and Kresl 2010).
That is not to say that small and medium-sized companies are not important for growth;
nor does it imply that agglomeration is not important. On the contrary, as we saw in
Chapters 3 and 7, small entrepreneurial start-ups are critical for innovation and growth,
and agglomeration is often crucial for promoting their success (Acs 2002; Van Oort 2004).
However, the arguments outlined in Chapters 2 and 8, and also in this chapter, all imply that
in the current era of globalization the probability of success for small and medium-sized
firms will be higher in those same city-regions which are also the most globally connected,
because of the potential spillovers from, and global market opportunities associated with,
locally based multinationals. This is exactly what has been found by Aitken et al. (1979);
Andersson (2009); and Johansson and Loof (2009).
In the modern global economy, what becomes very clear is that urban scale is only part of
the story of how regions compete. The global connectivity of cities is now also a critical part
of the story, in a way that has not been the case in previous eras of globalization. As we saw
in Chapter 8, in earlier decades cities were still primarily the economic engines of the
domestic economy, and as global markets developed the major cities also played a role in
servicing these markets. Today, however, the causality is somewhat reversed, with global
cities responding primarily to global changes, and much less so to domestic influences. This
implies that for countries with global cities situated in them, the relationship between the
global city and other more domestically oriented cities will exhibit a relationship similar to
the economic base relationships discussed in Chapter 5. However, in this relationship, city
size is not necessarily the key issue because urban scale is no longer a key indicator of the
performance ofa global city. More important for global city performance is the nature ofthe
activities which are undertaken there and the degree of connectivity provided by the firms
operating these. As we have seen here, infrastructure provision is one part of a broader
story, a story which we saw in Chapter 7 is not simply captured by measures of diversity. As
we discuss in Box 9.2, the modern relationship between urban productivity and globaliza-
tion is related to all the qualitative attributes of cities which foster connectivity and interna-
tional engagement.
9.5 No~-core regions, intermediate regions, and peripheral
regions
All the discussions so far have tended to centre on the role of the global cities and leading
knowledge regions. The cities and regions are defined by the OEeD (2011a) as the large
<knowledge hubs'. These also tend to be the cities and regions which attract the most interest
from researchers, policy analysts, and journalists. However, these places are not typical by
definition. Even in their own countries these types of places tend not to be the norm but
THE MOfDERN ~f>M"',prM:';Ii'>'ro~ f&::~'"ol'~~'>.~~lil3-)ff!,'''''
rather the exception. Most regions do not exhibit these advantageous combinations ofchar-
acteristics of high scale, high diversity, high productivity and high connectivity. Casual
observation tells us that regions vary greatly in size, population, population density, knowl-
edge content, innovation performance, employment and migration patterns, land and real-
estate prices, institutional and governance systems, natural and rural amenities, and climate.
This raises the question as to whether there are any patterns emerging in terms of the more
typical regions which are not the knowledge hubs.
The major evidence on these matters has been generated from analyses of the OECD
regional database (OECD 2006, 2009a,b, 20lla,b, 2012), which provides almost two dec-
ades multi-country regional data disaggregated at different spatial and governance levels.
Analyses of these data show four broad patterns:
(i) In terms of productivity the spiky and uneven nature of the regional economy
implied by Figures 9.1 and 9.2 is confirmed for almost all countries.
(ii) The distribution of growth performance across regions within countries and also
across countries exhibits a regular pattern.
(iii) The share of national and international economic growth accounted for by the big
'knowledge hub' (OECD 20lla) or global city-regions discussed in the previous
sections is between 20 per cent and 30 per cent, depending on the scale of spatial
disaggregation, and that this share remains fairly constant across both countries and
time. As such, between 70 per cent and 80 per cent of economic growth comes from
other types of regions.
(iv) Nowadays, the high-growth regions in most countries are generally not the big urban
hubs (MGI 2010), nor even urban areas in general (OECD 2009a, 20llc), but other
various types of regions (Dijkstra et al. 2012), many of which are even lagging regions
(OECD 2012).
With regard to the second and third points, we can explain these issues by following the
OECD logic applied initially in the case of a country. First, each region's absolute economic
growth contribution x r is calculated as the product of the region's individual GOP growth
rate gr multiplied by the ratio of the region's output GDPr. In other words, we have X r = wrgr
where Wr = GDPr• Once the absolute economic growth share of every region x r is calculated,
the natural logarithm of the individual regional growth x r is plotted for every region along
the horizontal axis, while the vertical axis plots the natural logarithm of the rank order Rr of
the region, defined according to the GOP size of the region GDPr, in a manner similar to the
rank-size-rule diagram of Figure 3.8.
In terms of the OECD growth analysis, the observed shapes for regional growth distribu-
tions which seem to hold in almost all countries are rather more concave in nature than the
-1 slopes of Zipf's law depicted in Figure 3.8. These distributions exhibit a negative down-
ward-sloping shape, the slope of which becomes steeper as we move down and to the right
towards the cases of the very largest regions. Moreover, exactly the same logic can also be
applied to regions across groups of countries, such as across the ED or the OECD (OECD
20llc), and the shapes remain largely unchanged.
Typically, some 20 per cent ofa nation's economic growth is accounted for by less than 5 per
cent ofits regions, depicted in Figure 9.3 as zone A, towards 60 per cent ofa nation's growth is
In rank Rr
----------- ...
... ...
... ...
... ...
,,
,,
":',
: ,
I ,
I ,
I \
: \
: \
Inxr
D c : B : A
~ ~ ~'~ ~'~ ~:
Figure 9.3 Power law relationships for regional growth and rank
accounted for by some 30 per cent of its regions, depicted in Figure 9.3 as zone B, some 20 per
cent ofa nation's growth by 50 per cent ofits regions, depicted as zone C, and negligible contri-
bution to a nation's total economic growth is made by some 15 per cent of regions, depicted in
Figure 9.3 as zone D. Again, these resulting shapes also appear to hold for pooled sets ofregions
across group ofcountries. Note that regions tend to be very densely clustered along the curve in
zones Band C whereas in zones A and D they tend to be much more thinly spread out. Appendix
9.2 provides a detailed explanation of how to interpret such distributions.
Following the logic of the GECD (2011c) regional growth distributions, the reason why
the regional growth patterns are as they are is that many regions which are not predomi-
nantly urban in nature are also currently experiencing strong growth. Regions which are
predominantly rural in nature and also regions which contain a mixture of urban and rural
environments, what the GECD terms predominantly intermediate regions, are also grow-
ing, and often more rapidly than urban regions (GECD 2009a,b,c). GECD economic
growth is not primarily an urban phenomenon, nor is it primarily a phenomenon associ-
ated with global cities (GECD 2011a,b). It is a complex picture, in which the patterns of
regional growth which are greater than, equal to, or less than their respective national
growth rates differ very little between urban, intermediate, or rural regions (GECD 2009a).
In terms ofrural areas, the largest differences are between rural areas which are very remote
from intermediate or urban areas and those which are not so remote from intermediate or
urban areas, with remote regions tending to face problems of depopulation and decline
(Partridge et al. 2008).
The major patterns that we observe among GECD countries are that during the first
decade of the modern era of globalization during the 1990s it was urban areas which
systematically grew faster than intermediate regions, which in turn grew faster than rural
regions. Since the turn of the twenty-first century the ordering has reversed in many
AND ~~ \W'I,,)i~~'~'<S.,,,,, MODERN ECONOMY
places, with rural and intermediate areas in many countries now growing faster than
urban areas. This reversal in regional growth patterns, whereby non-core regions are
growing faster in terms of both total output and labour productivity growth than core
regions, was first observed by Broersma and Van Dijk (2008), and is a pattern now widely
observed (OECD 2009a). The current growth productivity performance of regions does
not present the same spiky picture as depicted in Figures 9.1 and 9.2, which are based on
absolute productivity levels. Many lower-productivity regions are now the fastest-grow-
ing regions such that the productivity and output growth rate spikes are often in different
locations and regions from the spikes in absolute productivity (OECD 2009b).
The long-run effects of these emerging trends remain to be seen. What is important for
our purposes, however, is to note that during the first decade of the current era of globaliza-
tion, namely the 1990s, it was cities, and in particular global city-regions, which seem to
have been the major beneficiaries of the massive international transformations taking place.
Indeed, the role of large cities is still a major feature of development among the BRIICS
countries (McCann 2009). However, the major impacts of these global transformations
increasingly appear to be spreading to other types of regions within the industrialized coun-
tries. Nowadays growth and development are rather more regional issues within OECD
countries than they are urban issues, while the long-run effects ofthese transformations and
developments in the newly emerging countries remain to be seen.
9.6 Conclusions
Modern globalization presents a complex picture of economic geography, in which the cur-
rent roles of cities and regions are evolving in various ways. The evidence regarding the
development of networks of global cities points to the conclusion that in many ways geo-
graphical proximity has become increasingly important since the advent of the modern era
of globalization since the late 1980s, even as transportation and communications technolo-
gies improve. The location behaviour of knowledge-intensive firms is critical, and firms
already embedded in networks connected with these firms will more easily achieve major
returns from globalization. As such, location in knowledge centres will therefore continue
to be critical for high-value activities, and the arguments in Chapters 2 and 7 suggest that
this spatial knowledge concentration will be even stronger for the service industries, because
they are so dependent on trust relations.
While the rate of convergence between advanced economies has also slowed down since
the 1980s (Cappelen et al. 2003; Greunz 2003), the importance ofkey urban areas within the
global economy has increased. The result of all these changes is that over the last few dec-
ades, while between-country inequality has been falling (Crafts 2004), within-country ine-
quality has actually been increasing since the 1980s (Brakman and van Marrewijk 2008).
More recently, however, we have also started to witness the increasingly important role for
intermediate regions, rural regions, and non-core regions in driving economic growth
within many countries. This is a new reality which poses quite different challenges to much
ofwhat the urban-focused literature discusses. Over the coming decades, many ofthe major
new growth opportunities and challenges will be evident in regions which are not the high-
est-productivity centres. This suggests that many of the advantages to core cities which
" •• "-.""~""' .lt'.~
• f\ND
arose out of the advent of the modern era of globalization are waning, and new opportuni-
ties are arising in many non-core regions.
Moreover, the arguments sketched out here are not in any way restricted to advanced
industrialized economies. The importance of large cities in globalization appears to be par-
ticularly pertinent in the case of many developing countries, as witnessed by the rising inter-
national importance ofcentres such as Shanghai, Mumbai, Moscow, Johannesburg, Jakarta,
and Sao Paulo. The increasing urbanization of these countries is associated with rising
incomes, but these large developing BRIICS countries have also been experiencing rapidly
rising interregional inequality to a much greater degree than the rich countries. As with the
advanced economies, over the coming decades much of the economic growth ofthese coun-
tries is also forecast to take place in intermediate centres. However, intermediate-sized cities
in many of these countries are far larger than those in advanced economies.Whether rural
or smaller intermediate regions in these countries also emerge in a similar manner to many
of the GECD countries remains to be seen.
Discussion questions
9.1 Discuss the major institutional, technological, and organization features which have
rapidly transformed the global economy since the late 1980s.
9.2 Many people believe that the world is becoming 'flatter'. In terms of the economics of
cities and regions, do you agree with this view? If so, why? If not, why not?
9.3 Discuss the contemporary relationship between globalization, urbanization, and
industrialization which has emerged in rich countries since the late 1980s.
9.4 What is the current relationship between the wealth of a city, its size, and its global or
regional connectivity?
9.5 How do you think that the current relationship between globalization, urbanization,
and industrialization has reshaped the economic geography of your own country?
9.6 How do you think that the current relationship between globalization, urbanization,
and industrialization has reshaped the economic geography of other countries? Give
examples from different countries.
9.7 Do you think that this relationship is starting to change again in different countries?
9.8 Discuss the economic growth performance of predominantly rural regions and
predominantly intermediate (mixture of urban and rural) regions in your country.
The outcomes of the transition from Figure 9.1 to Figure 9.2 can imply either divergence or
convergence, depending on the spatial units employed. In Figure 9.1, if the areas denoted
as Xv Yv and ZL define the borders of three separate national market areas within the
global economy AB, the transition from the environment depicted in Figure 9.1 to the
environment depicted in Figure 9.2 would imply that the differences between the countries
had increased, resulting in an international income or productivity divergence. On the
other hand, if there were originally only two countries, whereby one country contained
only city-region X while the other country contained both city-regions Yand Z, the transi-
tion from the environment depicted in Figure 9.1 to the environment depicted in Figure
9.2 would imply that the differences between the two countries had decreased, resulting in
international income and productivity convergence. At the same time, irrespective of the
areal definitions employed, the transition from Figure 9.1 to Figure 9.2 implies that the
distribution of city sizes would become more concentrated and the surviving city sizes
more similar, such that the rank-size exponent discussed in Chapter 3 would move above
unity. The reason is that many of the intermediate and intervening locations denoted here
as Y will disappear while the surviving locations X and Z will become more similar.
As we see with this example, the empirical outcomes regarding convergence or diver-
gence depend in part on the definition of the spatial units, as defined by the cartographical
patterns of national or regional borders. In statistical terms this reflects a well-known issue
in spatial statistics and spatial econometrics raised in Chapter 3 known as the modifiable
areal unit problem (MAUP) (Openshaw and Taylor 1979). We therefore note that the rela-
tionship between the cartographical pattern of borders and the spatial distribution of cities
does affect empirical observations, and the interpretation of convergence or divergence
issues must therefore be treated carefully.
gradient of -1. However, we know from Chapter 3 that larger functional (city)-regions often
also exhibit greater population density as well as greater population scales. This means that
local land prices and nominal wages are higher in larger and more densely populated city-
regions, and according to the arguments earlier in the chapter, will also be associated with
higher levels of labour productivity than in smaller centres. This would tend to make the
gradient of the log GDPr curve shallower as the size of the city-region increases, thereby
making the curve rather more convex than linear, in a manner akin to the shape of a bid-
rent curve, with the upper part of the curve exhibiting a gradient of -1, and thereafter
becoming increasingly shallow. The more convex the slope, the more important the agglom-
eration effects.
In contrast, the institutional and governance structure of countries tends to counterbal-
ance this effect. Administrative regions often reflect groupings ofseveral smaller population
centres in a manner which is designed explicitly to counteract the political power of the
dominant city, and to allow for a more even institutional architecture. This grouping of cit-
ies and intervening rural areas into administrative regions will tend to make the slope of a
curve plotting the log of the regional rank size against the log of the regional size rather
more concave, irrespective of whether the regional size is measured in terms of population
or GOP. As such, in terms of log rank-log size distributions these institutional issues will
tend to militate against the convexity issue determined by scale-productivity differences.
We can control for these institutional issues by plotting the natural log of the rank of the
scale against the natural log of the scale, where scale is defined either as regional GDPr or in
terms of regional growth rates gr. We know that TL2 or TL3 regional governance and
administrative levels do not always accurately reflect functional regions, and nor does
labour productivity vary according to population density and population size. Therefore a
plot of the natural logarithm of the administrative regional rank size Rr against the natural
log if its size, where size is defined either according to GDPr == w r or defined according to
regional growth rates gf' will not exactly reflect the -1 slope of Zipf's law, even if Zipf'slaw
does approximately hold in terms of city population sizes.
Using regional data pooled across the European countries and also across time, it becomes
clear that the log rank-log size graphs, whereby size is defined either as regional output
GDPr or as regional GOP growth rates gr' exhibit distributions which are similar in shape to
Zipf's law distributions in that they are downward sloping and approximately linear over a
large extent (Oijsktra et al. 2012). However, the major differences between these distribu-
tions and Zipf's law are that these two distributions both exhibit slopes which are noticeably
steeper than -1 and which also become curved and then flat for small values of size variable.
As such, they are more concave than the Zipf's law distributions.
These distributions therefore allow us to depict the distribution of regional economic
output (in other words the weighting components w r) and also the distribution of regional
growth rates gr' which are a result ofthe interrelationships between the economic geography
of the state and national-regional administrative and governance logic ofthe state. The pat-
terns which become evident tend to hold for almost all countries, and, as we have just seen,
the approach can also be applied to pooled groups of administrative regions across
countries.
If all administrative regions of a country grow at the same rate gr' we know that the curve
which plots the natural logarithm of the regional rank Rr of W r == GDPr against the natural
REGiONS ~'ii'Iil't..ln;""~,$... n,, GLOBAL R~'&M""""~""'!w'?~W~
logarithm of the regional growth contributions x r = wrgr will have exactly the same shape as
the curve which plots the natural logarithm of the regional rank Rr of W r = GDPr against the
natural logarithm of regional output W r = GDPr • Therefore, if the curves do not have the
same shape, we know that growth rates gr differ across administrative regions. More specifi-
cally, if agglomeration effects are Widespread and dominant, the curves plotting the natural
logarithm of the rank Rr of the regional output W r = GDPr against the natural logarithm of
the regional growth contributions x r = wrgr will be more convex and flatter than the curves
plotting the natural logarithm of the rank Rr of the regional output W r = GDPr against the
natural logarithm of the regional output W r = GDPr• In contrast, if the curves plotting the
natural logarithm of the rank Rr of the regional output W r = GDPr against the naturalloga-
rithm of the regional growth contributions x r = wrgr are more concave and steeper than the
curves plotting the natural logarithm ofthe rank Rr of regional output W r = GDPr against the
natural logarithm of the regional output w r = GDPr, it becomes clear that the reason must
be that many intermediate output size regions are growing faster than large output regions.
This is exactly what we do see (Dijsktra et al. 2012). The curve plotting the naturalloga-
rithm of the rank Rr of the regional output W r = GDP r against the natural logarithm of the
regional growth contributions x r = wrgr is very clearly concave, and more so than the curve
plotting the relationship between the rank Rr of regional output W r = GDPr against the
natural logarithm of the regional output W r = GDP r• The reason for this becomes clear if
we plot the log of the rank of GDPr against the log of the regional growth rate gr. Here there
is no real systematic pattern, with all types of regions exhibiting fast, slow, and average
growth rates, except for the fact that many intermediate types of regions are performing
relatively well. Growth rate and growth shares therefore do not systematically favour large
urban areas (OEeD 2009a,b).
Modern urban and
regional economic
policy analysis
Urban and regional economic policy is distinct from other forms of public sector economic
policy, in that it is explicitly related to questions ofgeography. Both the motivation for, and the
implementation of, urban and regional economic policy are specifically spatial in nature, and
decisions as to whether to undertake policy intervention depend on the performance of the
local economic environment. However, our perception of what is (local' will itself determine
the nature ofthe policy, its implementation and its evaluation. This is because the definition of
a (local economic environment' may extend from the spatial scale of an individual suburban
area to that of an urban metropolitan area, and to the even larger spatial scales of a city-plus-
hinterland regional economy, or even a regional economy comprising more than one city.
We know that economic indicators ofaverage employment, unemployment, income lev-
els, house prices, or various other indices of social deprivation will differ according to the
definition of the spatial areas used in order to calculate them. Therefore our assessment of
the performance ofthe (economic environment' will also depend on the spatial scale adopted
to define what is (local'. In other words, the criteria against which any possible policy inter-
vention will be considered or assessed will also depend on the spatial definition of the local
economic environment.
For this reason, urban and regional economic policy is generally broken down into two
distinct groups of policies and initiatives. First, there are various initiatives which are
focused specifically on the urban economic environment and targeted at the urban or sub-
urban spatial scale. We will refer to these policies under the general heading of (urban pol-
icy'. Second, there is a range of initiatives which are targeted at the much broader regional
spatial scale, and we will refer to these policies under the general heading of(regional policy'.
The first major distinction between urban economic policy and regional economic policy is
therefore simply the spatial extent of their focus and implementation, with regional policy
being applied over a much greater spatial scale than urban policy.
The spatial scale, however, is not the only distinction between urban and regional eco-
nomic policies. The second distinction between these two groups of spatial policies is the
nf\\U~v~,n.P<ll URBAN ANALYSiS
nature of the policies which can be implemented. All urban and regional economic poli-
cies are motivated by the desire to improve the local economic environment. However,
the extent to which particular policies can be considered as viable candidates for imple-
mentation will once again depend on the spatial definition of the local environment. For
example, while policies encouraging <gentrification' can be considered as candidates for
economic development at the urban or suburban level, they would clearly be inappropri-
ate as economic policies at the regional scale because their impacts would tend to be too
localized. Similarly, large-scale regional policy, based for example on industrial reloca-
tion incentives, would be inappropriate as economic policies at the suburban scale, and
often largely inappropriate even at the urban scale, because their impacts would tend
to be too widespread. In other words, the justification for the particular spatial economic
polices adopted will in part also depend on the spatial area which is the object of
the policy.
The third distinction between urban and regional policies concerns institutional issues.
The different spatial scales over which these different policies are applied imply that the
governmental and administrative frameworks within which such policies are implemented
must also be different. This is because the geographical areas which are the focus of the spa-
tial policies may cross different administrative boundaries. The implementation of such
policies may therefore require cross-border agreements and cooperative arrangements
between neighbouring local government bodies. This is particularly the case for regional
policy, which usually spans different administrative areas, but generally much less so for
urban policy, which is normally implemented at the level of a single urban municipal level
of government.
The fourth distinction between urban and regional economic policies, and an issue we are
particularly concerned with here, relates to the different analytical approaches which can be
adopted in order to understand and evaluate such policies.
All the various spatial economic models we have discussed in this book are implicitly
constructed at different spatial scales. The Weber-Moses models discussed in Chapter 1 are
primarily discussed in terms of an interregional framework, in which major production
facilities search for optimum production-location arrangements over broad spatial scales.
On the other hand, the models of spatial competition discussed in Chapter 1, such as the
Hotelling framework, can be considered either at an interregional regional scale if we are
discussing large production facilities, or at an urban or suburban scale if we are considering
small retail establishments. In Chapters 2 and 4 the models ofindustrial clustering, agglom-
eration' and urban land use which are discussed are explicitly framed at the level of an indi-
vidual city. This is in contrast, however, to the central-place models of the urban system by
Christaller, Losch, and the more recent work of Krugman (1991) and Fujita et al. (1999), all
of which are discussed in Chapter 3, and which are obviously constructed at the interre-
gional scale. The regional multiplier models discussed in Chapter 5 are all framed at the
city-region and interregional scale, as are the labour employment and migration models
discussed in Chapter 6. Finally, the factor allocation, growth, and balance ofpayments mod-
els discussed in Chapter 7 are all framed at the interregional level.
The fact that each of these models is implicitly constructed at a different spatial scale sug-
gests that different models will be appropriate for different geographical scales of urban and
regional policy analysis. In other words, the appropriate analytical technique, or combina-
tion of analytical techniques, to employ to deal with a particular policy issue will depend on
'%v"'' ' ':l"'~' ~"'k i (
k
the object of the policy and the spatial scale of its implementation. Obviously, all the urban
and regional economic issues discussed in this book have implications for urban and
regional policy. However, in general there are also differences in the analytical underpin-
nings of each policy agenda and differences in the emphasis given to particular approaches.
For example, on the one hand, urban policy issues tend to be closely related to matters of
land use, local labour commuting, and the impacts of localized externalities, both positive
and negative, while on the other hand, regional policy issues tend to be more related to the
issues regarding the interregional mobility oflabour and firms, and also the impacts of insti-
tutional and technological changes.
As we have seen in Part I of this book, over the last two or three decades there have been
many modern developments in urban and regional economic theory which build on and
greatly extend the insights of the earlier seminal theoretical frameworks also discussed in
this book. In particular, our understanding of the mechanics of urban and regional growth
has advanced significantly. At the same time, as we have also seen in Chapters 8 and 9, over
the last two or three decades there have also been fundamental transformations in the global
economy. These changes have enormous implications for the economic role of cities and
regions within national and international economic contexts. In particular, the role of cities
and regions in fostering economic growth appears to be increasing relative to the role played
by countries and nation-states. At the same time, the traditional differences in the analytical
emphasis of the various urban and regional debates have also been somewhat blurred ad
reoriented following the recent impacts of globalization on cities and regions, impacts
which tend to transcend all spatial scales.
This chapter will discuss the policy challenges relating to each of these spatial scales by
examining the policy issues relating first to cities, second to regions, and third to national
policy. The first section of the chapter will examine urban economic policy matters. The
second section will examine traditional regional policy matters, and the third section will
discuss the new debates which have emerged regarding the nature, rationale, and role of
regional policy. As we will see, these new debates have profound impacts in terms of our
understanding of development policy per se, and the outcomes of these debates will shape
regional development policy for many years to come.
Rent per
sq.m
A
Bid-rent curve for retail
c
M~
I do 1
distance
~ ~:
! d
m
r dr
Under these conditions, competition for land will imply that the service sector offices will
be located between the city centre M and a distance do from the city centre. The manufactur-
ing sector factories and workshops will be located in the concentric ring of land around the
service sector, at a distance between do and dm from the city centre. Finally, the retail and
distribution sector will be located in the concentric ring of land around the manufacturing
sector, at a distance between dm and dr from the city centre. The actual urban land-rent gra-
dient is given by the shaded envelope rent gradient ABC which is just tangent to the highest
bid-rent curve at each location. As we see, the urban land-rent gradient will be a smooth
downward-sloping function which is convex to the origin M.
We can now compare this competitive result with a situation in which the urban land is
zoned. For example, we can imagine a situation in which the local metropolitan urban gov-
ernment decides not to allow manufacturing activities to be located as close to the city cen-
tre as they would be under free competition. Such a decision might be taken for reasons of
preserving or enhancing the aesthetic quality of the city centre, or alternatively because of
concerns about the negative effects of local environmental pollution. In this situation, the
local urban planning authorities may decide to permit only retail activities to take place
immediately adjacent to the city centre where the service sector activities take place. As
such, the zoning policy is organized so as to act as a buffer between the city centre and the
manufacturing activities. In Figure 10.2, this retail zone is defined as the area between dis-
tances d] and d2 from the city centre M. At the same time, the local urban planning authori-
ties may also decide that the more peripheral suburban areas which are largely residential,
Rent per
sq.m
M. .
~ distance
~ d1 :
!~
r
d2 •
~.
j
d
r d4 7~
~
~
ds
and in particular those areas which are occupied by middle- and higher-income house-
holds, should not contain manufacturing activities. Manufacturing should only be allowed
to take place in the one particular area which is specifically zoned for such activities. In
order to effect this, the authorities may specify that the manufacturing zone cannot extend
any further than distance d3 from the city centre. Meanwhile, to compensate for the loss of
service sector space available in the city due to the presence of the inner retail zone, the local
authorities may only permit service sector activities to take place in the area between dis-
tances d3 and d4 from the city centre. Beyond this zone the planning system may permit a
mixed use of retail or service sector activities.
In a situation such as this, the actual urban land gradient will be serrated (Evans 1985) as
well as downward sloping, and is given in Figure 10.2 by the bold line. In order to consider
the welfare effects of the planning policy we can compare the area under the rent gradient in
the case where the zoning policy is implemented and in the case where there is no such
policy. Where no zoning policy is in effect and the land market is competitive, the actual
urban land rent gradient is given by the envelope rent gradient in Figure 10.1. As we see in
Figure 10.2, the difference between the area under the competitive rent gradient and the
area under the serrated rent gradient is represented by the sum of the areas abce plus ghjk.
crrH;;S~ ~~\~~~,fr~s,
These two areas together represent the total loss of urban rental revenue which results from
the urban land being occupied at the zoned locations by activities which are unable to pay
the maximum rent attainable under the competitive conditions. Given that Figure 10.2 is a
one-dimensional diagram, the actual total revenue lost to the urban economy due to the
zoning policy can be calculated by rotating the one-dimensional model through 360 degrees.
This total revenue loss represents the opportunity cost of the zoning policy, and as such
reflects the welfare loss to the urban economy of the planning policy.
The key point to come from the above argument is that urban planning policies inevitably
have welfare implications. However, in the analytical assessment of the model described by
Figures 10.1 and 10.2 we assume that the price mechanism is broadly efficient. Yet a true
assessment of the implications of these policies also depends on our perception of the effi-
ciency of the land-price mechanism to correctly price amenity goods. In the situation
described by Figures 10.1 and 10.2, we have assumed that the reason for the implementation
of zoning policies is that government authorities implicitly believe that the land market
exhibits externalities which are not correctly priced by the market mechanism. In particu-
lar, the aesthetic aspects of the central city environment are assumed to be undervalued by
private sector cost considerations, such that the profitability of individual manufacturing
activities will take precedence over the perceived positive benefits associated with an attrac-
tive civic centre. Ifit is perceived that first- or second-best solutions are neither possible nor
appropriate policy responses to the perceived market failure in this context, the government
authorities simply resort to a quota mechanism, in which a quantity constraint is imposed
on the amount of land available at particular locations, and land is allocated by a process of
rationing. Which particular development schemes are undertaken will depend on a prop-
erty developer acquiring planning permission from the local government authority. By
adopting such planning schemes, the government authorities are therefore ruling out direct
comparisons of the form represented by a comparison of Figures 10.1 and 10.2, because it is
assumed that the private sector land prices given in Figure 10.1 do not accurately reflect
marginal social benefits. This is the justification for such interventionist policies.
Rent per
sq.m
I
M~ ~ distance d
~ d]
~ d2
1
Figure 10.3 Derelict urban land
institutional environment in which the local real-estate market works, in order to effect
changes in the physical environment of the local area. The reason for this, as we have seen in
Chapter 3, is that the area of dereliction is due to the perverse upward-sloping behaviour of
the bid-rent curves of high-income groups, due to the externalities associated with social
perceptions. Therefore these urban policies aim to alter the perceptions of the higher-
income groups in order to encourage real-estate investment by this group in the city
centre.
The effects of such policies can be depicted in Figure 10.4 according to the argument in
section 4.4.1 and the description given in Figure 4.12. If the redevelopment of the local city
centre area goes ahead, this area will now become attractive for a certain portion ofthe high-
income group which has a relatively high preference for accessibility to the city, but which
previously was unwilling to pay for this due to the poor local environment. For simplicity
we will characterize this group as a high-income group comprising primarily young people,
but this group may also include dual-income households. The bid-rent curve of this group
will tend to be very steep. On the other hand, the older people within the high-income group
or those with young children will still generally have a higher preference for space in order
to provide for the needs of their dependants. As such, their bid-rent curve will remain very
shallow. The redevelopment of the downtown area will mean that the single perverse bid-
rent curve of the high-income group depicted in Figure 10.3 will now be split into two dif-
ferent downward-sloping convex bid-rent curves for the two distinct high-income groups.
As we see in Figure 10.4, the result of this is that the area within a radius of dy from the
central business district M will be occupied by young high-income earners. Similarly, the
older high-income group will be located beyond a distance dh from the city centre.
Meanwhile, the low-income group will be located in the area between dy and dh from the city
centre.
8'l?~'>vJ?il.JI8-,W&.~'<lI URBAN RE~:lHJ~NJll ECONOM~C POlJCY
Rent per
sq.m
M: distance d
~
1d
~ y
~6' '):
We can compare the area of habitation of the low-income group before and after the
urban regeneration scheme by considering the difference between Figures 10.3 and 10.4. As
we see, for the same bid-rent curve, the low-income group is now constrained to live in a
smaller area of occupation after the redevelopment scheme than before the scheme. The
reason is that both the land immediately adjacent to the city centre and the land on the outer
fringes of the area of low-income occupation are now allocated by the price mechanism to
the two high-income groups. In this situation, the low-income group as a whole will have
suffered a welfare loss with respect to the high-income group. This is because in the situation
where the incomes and preferences of the low-income group, which are embodied in their
bid-rent curve, remain unchanged, many ofthe low-income people will no longer be able to
live in the city and will lose their residences in order to make way for the high-income
group. As such, they will be forced to leave the city.
Alternatively, if the low-income population of the city remains stable, because as we dis-
cussed in Chapter 5 the ability ofthe low-income groups to migrate may be very limited, the
low-income population will now be constrained to a greater residential density of living.
Following the arguments in section 4.4 of Chapter 4, the individual bid-rent curve of the
low-income group, as depicted in Figures 10.3 and 10.4, indicates the rent per square metre
which people are willing to pay at different locations in order to maintain a given level of
utility, where utility is partly a function of the total land area consumed. If at any location
the total land area consumed falls for any given rent per square metre payable, the total
x'> X •• '&.X X %§ X "' ••R. A N [} ?M ~.•, ',...I> ~ ',,* '''-''' ~ ~
'"
utility of the individual household must fall. In terms of bid-rent analysis this utility fall is
represented by an upward shift in the bid-rent curve for the low-income group. As we see in
Figure 10.5, this upward shift in the bid-rent curve for the low-income group after the rede-
velopment scheme will be represented by an increase in the rent per square metre payable
by the low-income group at any location. Competition in the real-estate market implies that
the low-income group will now increase their area of habitation from the area between the
distances dy and dh from the city centre to the area between the distances dr and d[" from the
city centre. As such, the low-income group will increase their area of habitation, relative to
the situation described by Figure 10.4, by encroaching on the area of habitation of both of
the high-income groups. However, this increase in habitation area is paid for by an increase
in dwelling density on the part of the low-income group, the result of which is an upward
shift in the bid-rent function of the low-income group. The low-income group unambigu-
ously suffers a welfare loss due to the urban redevelopment scheme. These complex
welfare-reallocation issues are inherent in most urban gentrification schemes, as discussed
in Box 10.1.
Rent per
sq.m
/
M.
~ distance d
II
=d'
Id
y
~
d/'
Land-use allocations take place within land-use planning systems which vary enormously
across countries (OECD 2010a). In some countries of relatively high spatial population den-
sities, such as the Netherlands, South Korea, and the UK, land use and land allocation at both
the national and local levels are organized primarily within a system of <greenbelts'. A green-
belt is a zone of land surrounding an urban area in which urban development is not permit-
ted under any circumstances. In other words, the greenbelt forms a concentric ring around
the city which clearly defines the outer edge of the urban area. The logic of the greenbelt
policy is to limit the outward expansion of the urban areas in order to preserve and protect
the intervening rural land. The motive for such a restrictive policy is that in densely popu-
1ated countries and regions, the preservation of rural areas is often perceived to be a national
priority, because of the relative scarcity of rural land. From this perspective, the value of the
rural area is defined primarily in terms of aesthetic and archaeological arguments, rather
than simply in economic terms. Therefore, in order to ensure that all people have relatively
easy access to local rural areas, strictly enforced limits are placed on the expansion of urban
areas. Implicitly, such a policy assumes that the private land market mechanism will not
appropriately value the preservation of rural environmental amenities in terms of their social
costs and benefits, and that urban development activities will be priced only with respect to
private costs and benefits. As such, the rationale for greenbelts is an externality problem. On
the other hand, such a policy inevitably also has welfare implications.
In order understand the logic for a greenbelt policy we can consider Figure 10.6, in which
a region comprises two major urban centres with their respective central business districts
rj2 •••••••••.•••••••••••••••••••••••.•••••••••••••••••••••••••
. rK2
rj1 •••••••••••••••••••••••••••••••••••••••••••••••••••••• •-J••
\
I \ I
/ \ I \"····································rK1
/ \ I \
/ \ / \
/
\ / \
/'
/
/
" , /
/ \
\
/'
/' ',.,I' ,
rA .." ".
.,;
,~ ',r
J ~ K
~
d1 d2 d] d4 ds d6
located at J and K. We assume that the nominal wage incomes earned at the central business
location J are higher than the nominal wage incomes earned at K. As such) the competitive
market rental price per square metre rfJ of land at J is higher than the competitive market
rental price per square metre r KI of land at K. Following the arguments in section 4.4.1 of
Chapter 4) the convex envelope rent gradients of each urban area comprise the individual
bid-rent functions of the different income groups employed at each urban centre. If the
agricultural land rent is given as r A at all locations) and the land market is competitive) the
urban area centred on J will be larger than the urban area centred on K. The urban area cen-
tred on J extends from location d2 to d3 ) and the urban area centred on J extends from loca-
tion d4 to ds' The agricultural areas are the areas to the left of d2 ) to the right of ds) and the
area between locations d3 and d4 • In other words) at the prevailing incomes r]I and r KI earned
in urban activities at J and K) respectively) the two urban areas are still separated by an area
of rural activities. The actual land-rent gradient is represented by the bold line in Figure 10.6.
If) for example) due to agglomeration economies) the nominal incomes payable at J and K
increase over time by 50 per cent to r J2 and r K2) respectively) both of the urban areas will
expand. In this case) the new land-rent gradient is given in Figure 10.6 by the perforated
bold line. As we see) the urban area centred on J will now extend to a location d1 to the left
of J) and the urban area centred on J will now extend to a location d6 to the right of K. The
area between Jand K will now all be taken up by urban development such that the interven-
ing urban area will disappear. As such) the new merged urban area will now extend continu-
ously from dl to d6 •
In order to avoid the merging of urban centres over time) for the reasons outlined above)
the land-use planning authorities may enforce a strict greenbelt policy to preserve the exist-
ing urban boundaries. The effects of such a policy can be seen in Figure 10.7. In this case) the
land-use planning system implements a greenbelt around the urban area centred on J) such
that urban development is not permitted beyond locations d2 and d3 ) and a second greenbelt
around K) such that urban development is not permitted beyond locations d4 and ds' In this
situation) if the cities grow over time due to agglomeration economies) the increased num-
ber of locally employed people will be constrained to live in the same urban areas. This will
reduce the average living area of each household. Consequently the bid-rent curve for each
individual urban household will shift upwards to a higher level than would be the case in the
situation where growth is accommodated in a competitive land market without a greenbelt
policy. The result of this is that the envelope-rent gradient will move upwards and the mar-
ket price payable per square metre will increase at all locations within the urban area to
higher levels than would be the case without the greenbelt. In Figure 10.7) the actual rent
gradient under conditions of urban growth after the imposition of the greenbelt policy is
given by the bold line) and the rent gradient under conditions of urban growth without a
greenbelt policy is given by the bold perforated line. On the urban-rural boundaries at
points d2 ) d3 ) d 4 )and ds) there are significant discontinuities in the land rent payable) and
these will continue to exist as long as the greenbelt restrictions remain in place. From the
perspective of living costs) all urban dwellers suffer a welfare loss due to the imposition of
the greenbelt policy.
From the arguments outlined in section 4.4.2 of Chapter 4) it is also possible that the
negative welfare effects of a greenbelt policy may be exacerbated by the expectation that the
environment at the urban fringe will be preserved indefinitely. If environmental amenities
X'OX,,'''.''''''''''''X''' ..''. AND ~(
rJ3 ••••••••••••••••••••••••••••••••••••••••••••••••••••••••
. r k3
rJ2 •••••••••••••••••••••••••••••••••••••••••••••••••••• _••
. rk2
~ K
: d2 d] d4 ds :
are relatively localized and it is perceived that the greenbelt policy will be maintained in the
long term, this implies that the persons who are resident on the urban fringes will always
enjoy superior environmental amenities in comparison with those who are resident closer
to the city centre. Iflow-income groups are constrained to remain close to the urban centre
for the reasons discussed in sections 4.4.1 and 4.4.2 of Chapter 4, this implies that only the
high-income groups will enjoy these environmental benefits at their residential locations.
As we see in Figure 10.8, the effect of this will be to cause the bid-rent curves of the higher-
income groups to become upward sloping as we move towards the urban fringe, such that
the envelope urban rent gradient will now become U-shaped. Under these circumstances,
the discontinuities between the urban and rural land prices at the urban fringes will become
even more marked. Meanwhile, the major beneficiaries of the greenbelt policy will be the
high-income households living on the edges of the urban areas.
The role of, nature of, and rationale for regional economic policy have all recently under-
gone a major rethinking in many parts of the world. The rethinking has been spurred by the
enormous changes associated with modern globalization that were discussed in detail in
''''N''''""'''''',..,~''.''' URBAN AND E(ONOM~C ANAlV5~S
Rentpef Rentpef
sq.m sq.m
fJ3
....................................... f ]
K
~
'I \
/ \
/ \
/
/
/ \
,,
/
/
/ """
fA fA
~
J ~ K
~ d2 d] d4 ds
Chapter 9. Such a process of rethinking has uncovered many complex issues and led to a
fundamental reassessment of many of the assumptions on which traditional regional policy
was based. In addition, this has also led to a redefining of the modern approaches to regional
policy, and a reconsideration of the analytical underpinnings of such policies. Sections
10.3.1 to 10.3.3 will discuss insights from the traditional approaches to regional policy, and
section 10.4 will introduce the new debates in regional policy. As we will see in section 10.4,
many of the traditional insights and justifications for traditional regional policy approaches
are still valid, but they also need to be updated and adapted in the light of the more recent
reconsiderations discussed in section 10.4.
Economic policies implemented at the regional level often involve attempts to improve the
attractiveness of particular regions as locations for investment. In particular, regional poli-
cies attempt to improve the relative attractiveness of investment in less developed regions.
In this sense these regional policies are similar to some of the urban regeneration policies
discussed below. However, a key difference between regional policies and urban policies is
that the industrial sectors towards which regional polices have traditionally been targeted
have tended to be quite different from those targeted at the urban scale. The reason for this
is that, following the arguments in Chapter 1, the types of firms which traditionally have
been generally regarded as being the most sensitive to large-scale spatial cost and price vari-
ations are not the real-estate and property development sectors, but rather the manufacturing
GLOBALmZAT~ON: ~ l\ND ~~.·vn*~. . ; q:1H
and distribution sectors, along with some other types of commercial service sectors carrying
out rather routine standardized activities.
As with urban policy, in order to achieve these goals traditional regional policies involve
relaxations or changes to the institutional or legal framework within which local economic
development takes place. However, the local economic development impacts of regional
policy can differ significantly across both sectors and regions. Therefore it is necessary for us
to consider carefully the size and the spatial pattern of any possible local regional develop-
ment effects. Regional policies tend to be implemented partly or wholly via the provision of
public sector funds, and funding tends to be granted in selected lagging areas which are
chosen as candidates for regional financial aid. However, given that regional policy typically
involves public funds, it is also important to consider the social marginal costs and benefits
of such a regional policy, relative to the situation where no such policy has been initiated.
Counterfactual impact assessment is therefore essential.
The most common types of regional policies are supply-side policies, which attempt
to improve the environment for local investment by upgrading the quality of the local
production factor inputs. In particular, supply-side regional policies tend to focus on
the factor inputs which are location specific. In an interregional economy in which both
capital and labour are mobile, the only production factor inputs which are location
specific are natural raw material inputs, land, and local infrastructure inputs, while
certain segments of the labour market are partly immobile. Given that raw material
locations cannot be affected by policy intervention, the focus of regional policy tends to
be primarily on increasing the quality and variety of local infrastructure inputs or on
the subsidizing of specific labour input costs, such as labour retraining or upskilling
costs. These are generally indirect ways of reducing real local input costs. An alterna-
tive, but rather less commonly used approach nowadays, is for regional policy to focus
on directly reducing the cost of local land inputs. We can consider each of these
approaches individually.
In the case of regional policies which attempt to upgrade the quality and variety of local
inputs, the major focus often traditionally tends to be on the improvement oflocal transpor-
tation infrastructure (Vickerman 1991). The expected effect of these policies is primarily to
reduce the costs of accessibility to the region in question. Therefore improvements are gen-
erally sought in key strategic elements of the local transport infrastructure which connect
the individual region in question to other parts of the interregional economy.
There are two reasons for this overall approach. The first is that transportation inputs are
regarded as essential inputs into almost all industrial and commercial activities, irrespective
of whether it is goods or people that are being moved. On the basis of the production func-
tion discussions in Chapter 7, improvements in transportation infrastructure inputs can be
considered as increases in the level of regional technology. For example, a road-building
programme initiated in a particular lagging region will generally reduce the delivered prices
of all outputs produced in the region at any location. Assuming a broadly competitive mar-
ket, this will increase the overall level of regional outputs sold both within the region and
also to customers in other regions. This output expansion on the part of existing local firms
is a desired effect of regional policy. Therefore improvements in these transportation inputs
should improve the total factor productivity of almost all local regional industrial activities
which are trading interregionally. One ofthe intended effects ofthis type ofinfrastructure-based
AND POLICY
regional policy is therefore to encourage an expansion ofthe existing local industrial base by
enhancing its productivity.
The second desired effect of this type of traditional approach to regional policy is to
encourage the immigration of more firm investment into a region. For industries which are
relatively mobile, spatial variations in transport costs may significantly affect the attractive-
ness of different regions as locations for investment. If the upgrading of transport infra-
structure can alter the relative costs of transportation and accessibility in favour of less
developed locations, it is hoped that this will encourage further immigrant inward invest-
ment. As such, an intended effect of these types of regional policies is to encourage an
expansion of the local industrial base via inflows of additional capital. However, whether or
not such a policy actually has the desired local effects depends on the relationship between
the provision of the transportation infrastructure, changes in transport costs, and the mar-
ginal price and revenue effects of the transport cost changes on both domestic firms and
firms which are geographically mobile.
Following the arguments in Chapter I, one possible effect of transport cost reductions
can be understood as increasing the likelihood of a firm relocation into a particular region
for any given set of interregional labour and land prices. However, competition in factor
markets would imply that any transport cost savings will soon be countered by increases in
local factor prices, such that a permanent advantage cannot be maintained by this policy.
Therefore, whether the transport infrastructure improvements will encourage external
firms at all to invest in that region also depends in part on whether firms will substitute in
favour of inputs produced in that region.
We can analyse the potential effects of such a policy by comparing the conclusions to the
Weber and Moses arguments in Chapter 1. As we see in these sections, the location effect of
reductions in transport costs in particular regions depends on the location-production sub-
stitution possibilities of firms. If firms have zero or only limited input substitution possibili-
ties, localized reductions in transport costs can be efficiently absorbed into firms' cost
schedules by moving away from the area in which transport costs are reduced in order to
reduce relatively higher transport costs associated with other locations. This is a classic
Weber-type result. As such, the regional policy will have exactly the opposite effect from
what was intended. Alternatively, iffirms have a wide range ofMoses-type substitution pos-
sibilities, they will most efficiently absorb the localized transport costs reductions by substi-
tuting in favour of the lower-delivered-price goods of the region in question. This will also
encourage the firm to move towards the area of lower transport costs, thereby having the
desired effect of increasing the immigration of firms into the area. This is a classic Moses-
type result, entirely in keeping with the objectives of the regional policy.
From the foregoing discussion we see that the local regional development effects of tradi-
tional regional policy initiatives based on transport infrastructure provision can actually be
rather hard to predict. If firms are able to substitute between inputs fairly easily, transport
cost reductions should encourage competition in all regions along the lines of a one-sector
model framework. However, ifinput substitution is not so easy, the results can be very com-
plex. Furthermore, improvements in transportation infrastructure can also have additional
regional effects. As we see in Chapters 1 and 3, transport costs over space in part act like a
tariff barrier, protecting less efficient local firms from external competition (Krugman
1991). As transport costs are reduced by regional policy infrastructure improvements, this
GLOBAJJZATHJN: ~~~~.#l:""'..Ji'. AND lk">.v'...... ~~~HH 8"'"
means that some local firms will no longer continue to exist. Krugman and Venables (1990)
have shown that if agglomeration economies operate in some locations, the negative
(shadow' effects on the less developed region can be very significant unless there are major
compensating local wage falls. Therefore the spatial impacts of policies to improve regional
and interregional transportation infrastructure must be evaluated carefully.
Similar lines of reasoning can also be applied to the case of subsidies provided to certain
aspects oflabour inputs. If retraining or upskilling is partly subsidized, the aim is to increase
the efficiency of the firms receiving the subsidy. These efficiency increases ought to be trans-
lated into real falls in the costs ofproduction and increases in profitability for the subsidized
firms. This is the local productivity effect. In addition, the subsidies should also encourage
the immigration of firms into the regions eligible for the subsidies, and this is the substitu-
tion effect. Again, however, as with the transport infrastructure example, the actual impacts
of these labour-training subsidies depend on how sensitive is the overall output perfor-
mance ofthe firm to these subsidies, and also how geographically mobile they are in response
to the subsidy opportunities.
Similar types ofarguments also hold for policies such as (enterprise zones', which encour-
age immigrant investment into a region via indirect land-price reductions. Such policies
operate via the direct or indirect subsidizing of immigrant investment into a region via
rental subsidies, local land tax rebates, or public subsidies for land assembly, land reclama-
tion or conversion, or local infrastructure provision (Swales 1997, 2009). These types of
policies are nowadays often also combined with various financial incentives for local and
regional government authorities to undertake local development activities. Such financial
incentives typically allow the local and regional authorities to borrow additional capital
against the expected incomes earned from the (enterprise zone' via what are known as tax
incremental finance (TIF) schemes, in which the additional revenues raised by the opera-
tion of the enterprise zone are largely tax free over a long period. Examples of these types of
approaches are discussed in Hague et al. (2011), while the types of systems and approaches
used in regional planning in different parts of the world are discussed in Brail (2008), Faludi
(2008), and Seltzer and Carbonell (2011).
more people are able to gain access to such schemes if they are targeted at large firms rather
than at small firms, or alternatively targeted at densely populated central regions rather than
at geographically peripheral regions.
In terms of cost-benefit analysis, on a project-by-project basis the value of an individual
regional policy project depends on both the marginal benefit to the individual benefiting
from the project and also the number of individuals affected by the project. Such valuations
would generally tend to favour densely populated central regions. As such, it would there-
fore appear that if the example in Box 10.2 of a road-building programme in the peripheral
region were indeed initiated, the justification for such a policy could therefore only be pro-
vided primarily on political or social grounds, rather than on economic grounds. However,
there are two major counter-arguments to this.
First, the relative costs of road infrastructure provision may be much lower in peripheral
economies than in congested central regions due to the lower land and labour prices,
although on the other hand lower accessibility may actually increase such costs. Similarly,
congestion effects and high land prices in the densely populated region may limit the poten-
tial benefits of further infrastructure in the central region. These cost differences would all
be taken into account in a cost-benefit analysis, and in either of these situations an evalua-
tion of the long-run social costs and benefits of the scheme may indicate that the net social
welfare gains are greater in the peripheral region than in the central region. In addition, if
localized growth effects are stimulated in the peripheral region by the provision of the pub-
lic infrastructure, the welfare gains associated with the regional policy intervention may be
very significant (Houghwout 1998, 2002). In such situations, the cost-welfare curves
depicted in Figure 10.9 will be reversed, with the cost-welfare curve for the peripheral
region becoming rather shallow whereas that for the central region will be relatively steep.
Second, the cost-benefit calculation of the individual project may be too narrow a refer-
ence point. In the case of regional policy, broader welfare considerations associated with a
set of policy interventions rather than just the return on the moneys invested in an indi-
vidual project may be required. In this case, a project-by-project-based cost-benefit
AND n~·u~'ll>.J1~~JM,,&" i'!.'~" •.fI'~'<I'.!%.#g~fl'l~~
approach to evaluating regional policy will be too narrow, and a broader approach, which
allows for coordination and networks effects and price changes, becomes more appropriate
(Grimes 2013). Under these conditions, a portfolio real-options-based approach which cap-
tures these broader effects is more appropriate for considering these system-wide types of
impacts (Grimes 2013). These issues are discussed in more detail in section 10.4.2.
The welfare evaluation of a regional policy therefore requires not only a cost-benefit
analysis (Sassone and Schaffer 1978; Pearce and Nash 1981; Layard and Glaister 1994) of all
the potential economic and environmental impacts of the policy, but also an explicitly spa-
tial discussion as to the broader distributional and network impacts of the policy (Graham
2007a, b; Venables 2007). Geography plays a role in determining both the absolute size and
the spatial distribution of the economic impacts of public policy initiatives. At the same
time, the impacts of the policy must also be considered against what the objectives of the
policy are. Therefore the success of regional policies must be evaluated carefully with respect
to the intentions of the policy and the explicitly spatial outcomes of the policy.
~
~ ~< )!
LA YA
LAF LA2 YA2 YA1
~
~
15A2
; *I ;*
P' ,
P* YA2 YA1 YA
Region B
. . . . . . . . . . . . . . . .~~~ T. ADB3..T.~~~ !.
~ ~~
~~
15B1 15B3
;* I ;*
unemployment in the depressed region B continues because, at the prevailing interest rate,
local investment levels are unable to generate local demand sufficient to clear the local
labour market. If interregional migration flows are not sufficient to clear all local regional
labour markets quickly, and the interest rate cannot be reduced below i*, the situation is
maintained indefinitely.
At this point, we must consider why the interest rate will be set and maintained at i*. One
of the features of the mutually open and interconnected interregional economy is that price
rises originating in one region can be transmitted very quickly to prices rises in other regions.
In the case ofa national economy in which one buoyant region ofthe economy is consistently
close to facing local supply shortages, the national monetary authorities may decide to set
national interest rates in order to just avoid local labour and land-price inflation in the buoy-
ant region. The reason for this is that the monetary authorities fear that lower interest rates
will engender local inflation in region A, which in turn will immediately be transmitted to the
rest of the national economy. Although interest rate falls will be beneficial for the depressed
regions, in that local investment and employment levels will rise, the land supply constraints
in region A limit the downward movement of interest rates. Moreover, this is the case irre-
spective of the phase the national business cycle at which the national economy finds itself.
The buoyant region therefore consistently acts as a 'bottleneck' region, and constrains the
demand levels in the less buoyant regions. In this particular regional economic system,
unemployment in the less buoyant region is maintained in order to preserve price stability in
region A, and, consequently, price stability in the national economy. In other words, not only
are differences in the individual local regional demand and employment conditions in part a
result of the macroeconomic policy, but at the same time the macroeconomic policy is itself
partly a result of the differences in the regional demand and supply conditions.
In situations where buoyant regions act as regional bottlenecks, regional policy can have
a role to play in encouraging growth and employment in less buoyant regions without
incurring inflationary pressures. In order to understand the logic of this argument we must
compare the local employment demand conditions in the two regions. As we see in Figure
10.10, the prevailing local investment and labour demand conditions in the less buoyant
region are constrained by the factor supply constraints and inflationary pressure in the
buoyant region. However, regional policy can act so as to encourage the diversion ofdemand
from the buoyant region A to the less buoyant region B. Regional policy tools, such as the
provision of infrastructure and real-estate subsidies in the less buoyant region, which in
addition may sometimes be applied in tandem with land-use planning restrictions in the
buoyant region A, can effect a diversion of investment flows away from region A and
towards region B. The manifestations of this will be the migration of new immigrant firms
to region B in preference to region A, and the expansion of existing firms of facilities in
region B in preference to region A. This investment diversion effect can be represented in
Figure 10.10 by the reduction in local investment flows in region A at the prevailing interest
rate i* from ISAJ to ISA2, and an expansion in local investment flows in region B at the pre-
vailing interest rate from IS BJ to IS B2. The reduction in investment in region A will lead to a
fall in local regional income from YAJ to YA2, and the consequent reduction in regional
aggregate demand from ADAJ to ADA2 will lead to a fall in local labour employment from the
full regional employment level of LAP to a lower employment level of L AJ • The local
unemployment shortfall caused by the diversion of investment away from the buoyant
region is therefore given by (L AP - LA2 ). (However, as we will see shortly, this unemployment
does not actually take effect.) In the less buoyant region B, the increase in local investment
associated with the regional policy diversion effect increases the local regional income from
Y BI to YB2 , and the consequent increase in regional aggregate demand from ADBI to ADB2
will lead to an increase in local labour employment from L BI to L B2 • The local unemploy-
ment shortfall caused by the diversion ofinvestment into the buoyant region is now given by
(L BP - L B2 ) and is exactly the same as the apparent local unemployment shortfall (L AF - L A2 )
in region A. In this situation, the current local price level in region A is unaffected either by
the reduction in investment in region A or the expansion in investment in region B.
Moreover, the aggregate demand in both regions can now be allowed to expand without
causing any inflation in region A. In region A aggregate regional demand can be allowed to
expand to its original level of ADAl which was maintained before the application of the
regional policy. At this level of aggregate regional demand region A will exhibit local full
employment income, investment, and labour demand levels of YAI , [SAl' and LAP' respec-
tively, at the prevailing interest rate i* which just ensures price stability at p* in region A.
Similarly, in region B aggregate regional demand can be allowed to expand from ADB2 to
ADB3 • At this level of aggregate regional demand region B will exhibit local full employment
income, investment, and labour demand levels of YB3 , [SB3' and LBF, respectively, at the pre-
vailing interest rate i* which just ensures price stability at p* in region A.
This positive macroeconomic expansion effect across both regions compensates for the
negative diversion effect in the buoyant region A, and is additional to the positive diversion
effect in the less buoyant region B. In other words, the diversion of investment associated
with the regional policy allows regional income and aggregate demand in both regions to be
maintained at levels which ensure full local employment in both regions, without engender-
ing inflation in the bottleneck region. As such, the argument here is that one role of regional
policy is therefore to circumvent many of the regional bottleneck problems associated with
different regional investment levels in a situation where common interregional interest
rates are set primarily with respect to the demand conditions in the buoyant regions.
Therefore, if the policy is implemented successfully, the net result of the regional policy
diversion effect, plus the macroeconomic expansion effect in each of the two regions, is to
allow for full employment aggregate demand to be maintained in both regions under a sta-
ble macroeconomic monetary regime. On this argument regional policy can be imple-
mented without leading to <crowding out'.
As we have seen in this book, the last two or three decades have witnessed many advances in
the theory of urban and regional economics. Moreover, our awareness of the impacts of
modern globalization on cities and regions, and also the role played by cities and regions in
driving modern globalization, have all increased dramatically. In contrast, apart from some
of the financial incentive initiatives alluded to in section 10.3.1, the practice and implemen-
tation of regional development policy has remained largely static until very recently.
In developing countries, policy interventions at the national and regional level have con-
tinued to be based largely on growth thinking derived from the 1950s analyses of Solow
(1956), Swan (1956), Hirschmann (1958), and Rostow (1959), with little serious attention
paid to questions of economic geography, knowledge spillovers, or the complex interac-
tions between these two issues and institutional issues. Consequently, urban and regional
policy in most developing countries has barely moved beyond road-building and infra-
structure provision, with little serious attention paid to the social and environmental conse-
quences of these interventions. Broader distributional and developmental issues first raised
by Myrdal (1957) have been largely ignored at both the regional and urban levels, with the
focus being on encouraging the efficient working of market mechanisms.
Rather differently, in developed countries regional policies tended to be organized in a
top-down hierarchical fashion and at a national level. Such policies were understood as
being largely redistributive in nature, operating primarily by encouraging firms to relocate
to weaker regions via subsidies and tax breaks. These policies have tended to be primarily
sectoral in nature, in that they focused on providing incentives to those particular industries
which were perceived as being most responsive to such relocation incentives. Meanwhile, in
the case of cities, as we have already seen, over the last two or three decades urban policies
in advanced countries have tended to focus either on issues related to gentrification pro-
cesses or on issues relating to zoning and land-use restrictions. In these policies, distribu-
tional issues are almost entirely ignored, and little explicit acknowledgement has been
afforded to the relationship between the changes in the local institutional or governance
issues and the social cohesion or segregation impacts of the urban redevelopment schemes.
In terms of regional and urban policy, this state of affairs has remained largely static for
many decades, in spite ofthe major analytical and political economy transformations which
have been witnessed over the last two decades. However, there are real signs that all this is
about to start changing, and there are two reasons for this, one of which is realpolitik in
nature and one of which is analytical in nature.
First, in terms of realpolitik, the global financial crisis of 2008 and the ongoing economic
disruption caused by this has alerted people in all countries to various weaknesses in many
of the current systems ofeconomic regulation, coordination, and governance. The scale and
complexity of modern globalization processes and their impacts on development have
moved well beyond the accepted conventional (Washington consensus' wisdom regarding
optimal development policies (Stiglitz 2002; Fine 2003; Rodrik 2006), a line of thinking
which had assumed that market deregulation and establishment of well-defined property
rights were more or less sufficient to foster economic growth. Indeed, these weaknesses
were seen to put many processes ofdevelopment in many countries in jeopardy (Spence and
Leipziger 2010) because it has become apparent that the scale ofeconomic shocks in today's
globalized world can undo and often undermine the mechanisms by which many policy
initiatives were assumed to work.
Second, in analytical terms, 2009 and 2010 saw the publication of five very influential
reports on the role of, nature of, and the rationale for regional development policies. These
five reports have reignited many long-dormant debates and forced a rethinking of many of
our assumptions. These five reports are the World Development Report (World Bank 2009),
the Barca Report, An Agendafor a Reformed Cohesion Policy (Barca 2009), two reports pub-
lished by the aECD entitled How Regions Grow (aECD 2009a) and Regions Matter (aECD
2009b), and the 2010 Local Development report of the Corporacion Andina de Fomento
(CAF 2010). All these five reports appear to deal with largely similar analytical and policy
G LOBAlJZA'l'H)N::
issues. However, the policy conclusions that they come to are fundamentally different in
nature.
The highly influential World Development Report (World Bank 2009) adopts what is
known as a (space-blind' approach and comes to what seem to be very different policy con-
clusions from the other four (place-based' reports published by the European Commission,
the GECD, and the CAF. We can also add to this suite of major reports an earlier report
published by the European Commission, namely the influential Agenda for a Growing
Europe (Sapir et al. 2004). In some sense this report can be regarded as being space blind,
inasmuch as it was not focused on regional issues but more on sectoral matters. However,
the arguments and recommendations in the report did potentially have regional implica-
tions, so it is worth discussing it here.
The fact that such high-level reports come to fundamentally different conclusions regard-
ing the case for, and role of, regional development policy suggests that the issues are both
very much alive and very complex. At the very least, the debate is far from being straightfor-
ward and requires a consideration of a range of fundamental and interrelated issues, all of
which have already been referred to in this book.
The reason why there are two broadly different policy camps, with the World Development
Report and the Sapir et al. (2004) report in the (space-blind' camp and the other four reports
in the (place-based' camp, is that they reflect two fundamentally different approaches to
interpreting urban and regional empirical evidence. These differences in empirical interpre-
tation themselves reflect deep-seated differences in analytical positions. The reports derive
their arguments and insights from different elements of the theoretical literature, all of
which are referred to in the various chapters of this book. As such, the reports emphasize
different aspects of the theoretical literature as best reflecting the dominant underlying
urban and regional growth mechanisms, and the result of these different emphases is that
the recommended policy responses are quite different.
While the space-blind and place-based approaches have long intellectual traditions, both
approaches have also recently undergone major changes. From the perspective of regional
development, an awareness of these analytical differences is extremely important. After
years of a development policy consensus which had changed little in five decades, the
debates opened up by these reports will determine the trajectory of regional and urban
development policy in many parts of the world over the coming decades. We now examine
each approach in detail.
argues that policies which focus on places or regions will automatically inhibit migration
and factor adjustment, thereby working against the interests of society as a whole. The
World Bank therefore recommends that the most effective way ofgenerating efficiency is by
adopting a model of development which is 'spatially blind', where 'spatially blind' in this
sense means that the policies 'are designed without explicit consideration to space' (World
Bank 2009, p. 24).
For a basic space-blind model to be optimal from a social welfare perspective it is neces-
sary for factors to be allocated across space in a manner consistent with the underlying
assumptions and conditions of the one-sector model of Borts and Stein (1964). All that is
required to motivate these arguments is that spatial variations in demand are responded to
by factor mobility, so as to equalize rates of return on capital and real wages. As we saw in
Chapter 7, if demand increases in one region relative to another, this localized demand
expansion increases the returns to all factors in that region, relative to other regions. The
factor returns stimulate the in-migration of all factors into the high-demand region and
out-migration from low-demand regions. Assuming that the underlying mechanisms are
not infinitely cumulative, the process continues until factor returns are once again equalized
in all regions. These one-sector factor allocation and migration mechanisms are also
assumed to underpin the more recent welfare-maximizing processes described by the long-
run neoclassical interregional convergence models (Barro and Sala-i-Martin 1992), as dis-
cussed in Chapters 6 and 7.
Early space-blind arguments (Winnick 1966) assume long-run processes of interregional
economic convergence, and as such do not require us to make any assumptions regarding
the long-run role of agglomeration. However, unlike the earlier long-run convergence
models, the contemporary space-blind approach also assumes that agglomeration processes
are fundamental aspect of the economic landscape. As such, if there are localized agglom-
eration processes operating in the destination regions, these are assumed to operate in a
manner which is consistent with a combination of the new economic geography and
agglomeration theories discussed in Chapter 3 and the endogenous growth models of
Romer (1986) and Lucas (1988).
Yet it is important to note here that while these modern space-neutral models usually
tend to refer to the importance of fostering agglomeration, the rationale behind these space-
blind models is not based on arguments about agglomeration. Rather, these space-blind
arguments are actually underpinned simply by the two-sector factor allocation model of
Borts and Stein (1964), as discussed in Chapter 7, and as such do not require agglomeration
arguments per se.
We know that not all locations at all time periods are conducive for fostering the increas-
ing returns mechanisms associated with agglomeration. As such, agglomeration cannot
simply be a matter of geographical clustering. There must be a demand component to the
agglomeration formation process. If the mechanics of agglomeration processes in part
reflect such demand shocks, the two-sector model therefore implies that the agglomeration
processes should be allowed to run their course until interregional factor returns are equal-
ized, and the economy once again settles down to a one-sector model of interregional con-
vergence. In other words, from an aggregate national welfare and efficiency perspective,
encouraging these agglomeration-increasing returns processes is regarded as being inher-
ently good for growth. Therefore, whatever policy regime promotes interregional factor
mobility and the spatial reallocation of firms, people, and activities is seen as the optimal
development policy approach.
Following this approach, in terms ofpolicy recommendations, however, only a very min-
imalist and limited role for public intervention is therefore called for. This is because it is
assumed that policy-makers have no idea what the efficient spatial allocation of factors
would look like (Glaeser and Gottlieb 2008). Therefore a free market factor-mobility mech-
anism, as determined by market signals of regional wages and regional rates of return on
capital, is regarded as being the best way to ensure that the right factors are located in the
right places doing the right things. The invisible hand of the market should be left to decide
of its own accord which places people ought to migrate towards, and this process is likely to
lead to convergence processes operating between lagging areas.
Underlying these arguments, the World Development Report explicitly follows a view of
economic history which reflects the arguments of Rostow (1959) and Williamson (1965),
and which assumes that the urbanization-growth trajectories of today's emerging econo-
mies largely mirror the urbanization-growth processes experienced by today's rich coun-
tries in earlier eras. Applying these Rostow-Williamson arguments within a space-blind
framework implies that if factors follow market signals so as to maximize national growth
and welfare, then economic development will naturally be unbalanced. Economic growth is
automatically assumed to be associated with rising income inequality and spatial concen-
tration, and only at much later stages in development will income equality and spatial equal-
ization processes become evident. The World Development Report therefore argues that any
policy-related attempts aimed at promoting growth in weaker regions by countering these
un -equalizing and concentrating flows will actually undermine national growth and pros-
perity (World Bank 2009) and prolong poverty.
Instead of regional policy approaches which aim to promote equality between regions,
the World Development Report proposes a (development in 3-D' idea as its policy system,
whereby (development in 3-D' is based on the three development dimensions of density,
distance, and division. The solutions that the World Bank proposes to deal with the per-
ceived challenges associated with each of these three dimensions are three (I's, namely insti-
tutions' infrastructure, and interventions (Barca et al. 2012). Each development dimension
is argued to have its own I-solution, and these are mapped according to the following
framework:
The underlying argument of the World Bank approach is that promoting factor mobility to
urban concentrations is the optimal development model for most countries. This is because
it is assumed that these factor adjustment processes will guarantee the widest range of
employment opportunities for all people, thereby improving the lives of the maximum
number of people irrespective of where they come from.
Note at this stage also that these World Bank arguments are assumed to hold in spite of
the inefficiencies and irreversibilities associated with Harris-Todaro-type migration pro-
cesses, as discussed in Chapter 6, and the widespread evidence of squalour, oppression, and
environmental degradation in the barrios, favelas, slums, and shanty towns of the develop-
ing world's mega-cities.
A second major report which can also be considered in many ways space neutral is the
highly influential independent Sapir et al. (2004) report, An Agenda for a Growing Europe.
Although this report was primarily concerned with the reforms of the European Union
budget, and closely linked to debates regarding the so-called Lisbon Agenda for EU eco-
nomic growth policy, the Sapir Report came to conclusions that were not too dissimilar
from those of the World Bank (2009).
The Sapir et al. report recommended that in order to promote intra-EU convergence, the
major focus of EU policy should be that of the poorer new accession countries, and any
European regional policy, or more precisely EU Cohesion Policy, should be targeted at the
level of countries, or EU Member States, as they are known, rather than at the level of the
region. Along with the World Development Report, the Sapir et al. (2004) report also recom-
mended that there should be an agenda for institutional and governance reform in these
new-accession countries, along with additional sector policies focused on the development
and adoption of information and communications technologies (ICTs) across the whole of
Europe and other knowledge-based sectors. As such, the Sapir et al. report shares some
features of the World Development Report in that the authors of both see their reports as
being essentially space neutral, with a primary emphasis on institutional reform (Barca et al.
2012).
However, there are also some fundamental differences between these two reports. First,
there is no explicit urban and regional economics or economic geography in the Sapir et al.
report. The report makes no comments or recommendations regarding optimal urban or
regional growth patterns and the likely mechanisms to achieve these. In contrast, economic
geography and urban and regional economic thinking is fundamental to the World
Development Report and policy recommendations regarding the likely mechanisms to
achieve optimal spatial allocations are very explicit. Second, the Sapir et al. report advocates
funding to less developed parts of Europe, whereas the World Development Report largely
eschews such policies. Finally, the World Development Report takes no position whatsoever
on the sectoral policies related to ICTs, whereas these are central issues in the Sapir et al.
report (Barca et al. 2012).
placed based and highly contingent on context so as to tap any under-utilized local capacity
and to overcome poor or inappropriate institutional configurations. This is because modern
place-based arguments assume that the mobilization of local elements necessary for pro-
moting development requires an explicit consideration of the relationship between institu-
tional and governance issues and economic geography, relationships which are assumed to
differ significantly between places. The major differences between the traditional approaches
to regional policy discussed in section 10.3 and the modern place-based approaches to
regional policy are summarized in Table 10.1.
As we see in Table 10.1, the traditional approaches to regional policy discussed so far
tended to operate under a national architecture in which the policy was dictated by central
government, which decided on the subsidies to be made available and the firms and indus-
tries which were to receive the subsidies. The logic of such policies tended to be built around
influencing the location and geographical investment decisions of particular industrial sec-
tors, encouraging these firms either to relocate or to maintain their locations in particular
regions. Local and regional government generally had little or no major role or influence on
the logic or the design ofthese policies, other than acting as advocates for the particular local
firms, industries, or sectors aiming to attract regional subsidies.
In contrast, as we see in Table 10.1, modern place-based approaches to regional policy
emphasize the importance of designing appropriate institutional systems, whereby regional
policy aims to foster participation between central and local government, and between the
public, private, and civil society sectors. The aim is to design policies with local skills, exper-
tise, and capabilities in mind, in order to foster endogenous local development processes
based on the potential of a particular region. Most importantly, the aim is to design policies
which provide bundles of public goods specifically tailored to the specific challenges of the
region, and to do this in a manner which encourages the engagement ofas many local actors
as possible. The focus is on promoting transparency regarding policy intentions, objectives,
monitoring, and evaluation, and applying the policies in the most appropriate geographical
be unintended effects, and the aims of the policy itself may well be undermined unless the
spatial aspects are explicitly accounted for. In contrast, by taking account ofthese complexi-
ties and specificities, the place-based approach emphasizes the role which policy can play in
the provision of public goods in a manner which is appropriately tailored to the local
requirements of the region. The application of the place-based argument would therefore
appear to be particularly appropriate to the context in which economies are experiencing
major transitions towards new equilibria, a phenomenon very common in the current era
of globalization, and in which the short-run and medium-term transition processes may
heavily influence the long-run outcomes (Thissen and Van Oort 2010).
The publication of these highly influential reports within such a short time period means
that after decades of little or no change in policy thinking, development scholars and prac-
titioners all over the world are proposing, and are also therefore confronted with, two radi-
cally different regional development policy paradigms, namely the space-blind and the
place-based regional development policy approaches (McCann and Rodriguez-Pose 2011;
Barca et al. 2012). However, these differences also provide clues as to the complexity of the
problems which today's regional development policies must confront.
Local and regional development policy is not about drawing lines or zones on a map, but
rather about engaging with the complex interrelationships between institutions, geography,
and economic development (Pike et al. 2010; Barca 2011). Both the space-blind and the
place-based approaches acknowledge the importance of agglomeration effects, network
effects, and other mechanisms leading to spatial spillovers. Both approaches acknowledge
the types of spatial interactions examined in detail throughout this book, and the impor-
tance of factor mobility in allowing economies to adjust to new and emerging conditions
(Barca 2011). The differences between the two approaches therefore relate primarily to how
they perceive the interactions between institutions, factor mobility, and decision-making
processes, and the different emphases that they give to particular issues.
There are four major areas ofcontention, namely: the interactions between geography and
institutions; the nature oflong-term development processes; the nature of national-regional
decision-making processes; and the nature of intentions versus outcomes of a policy.
The place-based approach puts major emphasis on the influence of institutions in shaping
not only local economic geography, but also the ability of regions to adjust to changing
economic circumstances. In particular, the place-based approach assumes that many
regional institutional features, and some of the various dimensions of regional (capital'
including institutional and physical capital, are largely immobile. A region's ability to adjust
to external macroeconomic shocks is therefore regarded as being contingent on how local
institutional systems respond, and in particular to outflows of mobile factors and losses of
human capital. This ability of regions to respond is itself argued to be in part dependent on
the national institutional and governance architecture, and the autonomy given to local
institutions to implement flexible and locally designed policy responses. This is because the
place-based approach assumes that the institutional relationships linking national, regional,
and local levels of governance are central to regional adjustment processes. Increasing
regional inequality is argued to lead primarily to increasing interregional dependency in
which weaker regions increasingly become dependent on fiscal transfers from stronger
regions. This dependency itself can lead to increasing political polarization and paralysis of
many aspects of national-regional governance (Barca et al. 2012). For this reason, the place-
based approach argues that when policies are properly designed on the basis of partnerships
between local and national policy-makers, local policy design and delivery is the best way to
foster appropriate institutional reform which is in the national interest (Barca et al. 2012).
In the place-based approach, interregional inequality is therefore directly linked to national
performance, an argument closely related to the observations of Berg and Ostry (2011) and
Moretti (2012). In contrast, the space-blind approach is largely silent as to the relationship
between regional inequality, institutional reform, and factor adjustment. Rather, the space-
blind approach assumes that factor mobility is the primary adjustment mechanism, and
what it sees as impediments to this process, such as local development and regional policies,
should be removed.
World Bank position has been fundamentally that context does not matter) where context
is defined by place-specific social, cultural, technological, or industrial characteristics) and
that well-functioning institutions and market mechanisms are all that is required for
growth. This has been a longstanding view ofthe World Bank) which has often been referred
to as the 'Washington consensus). However) to say) as the World Development Report (2009)
now does) that national growth in many countries is to be fostered by massive urban growth
via spatial factor adjustment mechanisms) irrespective of the spatial outcomes) ironically
acknowledges precisely the opposite. This is an admission that context really does matter)
because if context did not matter) then the institutional reforms alone would indeed be suf-
ficient for growth) without recourse to making any comments or recommendations regard-
ing cities. Moreover) other publications by the World Bank (World Bank 2003) 2010;
Swinburn et al. 2006; Zoellick 2012) adopt a much more place-based line of argument
emphasizing the need for fostering partnership with the intended beneficiaries) the impor-
tance of engagement of local stakeholders) and the fostering of a broad notion of develop-
ment and growth including sustainability and inclusiveness. This is because it is becoming
increasingly apparent that long-term institutional reform can only be ensured via partner-
ship and collaboration between the external actors (funding agencies) and the local actors
(intended beneficiaries) built around results-based development programmes which
employ monitoring and evaluation (Sachs 2011) 2012). As such) these publications are
rather different from the space-blind approach of the World Development Report (2009))
and suggest that the real relationships between geography and institutions are in reality
much more complex than the simple space-blind arguments imply) and that there is no
simple dichotomy between 'people-based) and 'place-based) approaches.
The reason why large-scale urban growth is recommended by the World Development
Report is that the growth of large cities is seen by the space-blind advocates as being the best
way of solving the problems associated with the generation) acquisition) and sharing of
knowledge and resources (McCann and Rodriguez-Pose 2011). Importantly) if very large
cities are needed) it is not only because the institutions in the country do not work properly)
but also because they cannot reasonably be reformed) due to the entrenched and longstand-
ing self-interests of the major urban-political elites. Major urban expansion may therefore
be the only realistic growth option in nations with little or no realistic possibilities for insti-
tutional reform or development (McCann and Rodriguez-Pose 2011)) particularly where
countries are too small or fragmented to sustain home market effects (Collier 2006; Venables
2010) required for growth. However) the place-based approach argues that this is just one
solution among various possibilities) and that in many other regions and countries other
approaches may be preferable.
The shifting World Bank position in a sense also reflects a more general debate within
both economic geography and development studies regarding the role ofinstitutions versus
geography in economic development (Sachs 2003; Rodrik et al. 2004). While some people
take partisan positions on this) it may well be the case that in reality it is the relationships
between geography and institutions) and how these vary across space) that are the essential
issues. It is easy to think ofplaces where reform has not taken place and where development
has not occurred) and this is the typical case discussed in the literature. However) it is also
possible to think of places such as New Zealand where the institutional reforms have been
excellent but the economic geography outcomes have been adverse) precisely because the
reform of the institutions acted in a manner similar to falling trade costs (McCann 2009).
The point is that the relationships between economic geography, institutions, and develop-
ment are complex, and also they vary across geography. These observations suggest that
context, in the sense of the role of place as determined by social, cultural, and legal issues,
really does matter, and contrary to the World Development Report, other World Bank and
related publications (Zoellick 2012) also reflect thinking which has a great deal in common
with place-based approaches.
se, not just regional policy funding (Partridge and Rickman 2006). The logic of the space-
blind arguments is the removal of all welfare payments, because the provision, monitoring,
and delivery of such payments are place specific, according to the national-regional govern-
ance architecture. The second-best solution is to pay people to move. These arguments are
discussed in detail in Appendix 10.2, and the assessment of these arguments is closely
related to the regional mispricing issues raised by traditional place-based arguments, all of
which are discussed in detail in Appendix 10.1.
In contrast to the space-blind position, as well as both the traditional and modern place-
based responses already discussed in this chapter, the argument against these types of space-
blind conclusions becomes even more powerful when we consider the fact that governance
systems are also place based by definition. The relationship between the national and local
state requires governance and political systems to respond appropriately to the preferences
of the local median voter. This in turn requires national policies to take seriously into con-
sideration the preferences and well-being oflocal residents, and also the tolerance of society
for inequality therefore also becomes a central issue in the discussions (Ferrara 2010). The
modern place-based approach contends that what are ostensibly space-blind decision-
making process are in reality the outcomes of political-bargaining power, and interregional
adjustment mechanisms based on apparently space-blind logic which foster the increasing
dominance of certain places over others may actually contribute to national institutional
sclerosis and in-built political tensions which may undermine donor preferences. The logic
of interregional spatial adjustments may therefore generate other more profound problems
unless the intentions and likely outcomes of the apparently space-blind policies are made
explicit from the start.
10.6 Conclusions
This chapter began by discussing the various major types of specific initiatives traditionally
undertaken under the broad heading of urban policy. As we have seen, a major difference
between urban polices and regional policies is the spatial scale over which the policies are
implemented and assumed to take effect. The intended impacts of urban policies are
expected to take place over a much smaller spatial scale than those of regional policies.
These different spatial scales, however, also mean that the types of policies adopted and the
criteria against which the policies will be evaluated are also different. In the case of urban
policies, the dominant issue which will determine whether or not a policy is implemented is
the nature of the local environment at the suburban level. The notion of 'environment' here
relates to both the physical built environment and the local social and economic environ-
ment. The real-estate market is generally the target of urban policies, and the implementa-
tion ofthese policies generally involves relaxations or changes in the institutional framework
within which the local real-estate market operates. All such urban policies have welfare
impacts, which are realized in terms of changes in the prices of real-estate assets at different
locations. Moreover, the nature and scale of these welfare impacts can be quite different for
different income groups. Urban policies can therefore have welfare distribution effects, and
for urban and regional economists an evaluation of the benefits of urban policies cannot
take place without an assessment of these welfare distribution effects.
Traditional regional policies, on the other hand, simultaneously focus both on encourag-
ing indigenous regional investment growth and on attracting new immigrant investment
into a region from outside. As far as the latter approach is concerned, these polices tend to
operate over a much larger spatial scale than urban policies, and aim to encourage the
migration of firm capital over rather large distances. The focus of regional polices tradition-
ally tends to be on the provision of local regional infrastructure and also, in some cases, the
subsidizing of local real-estate inputs. As with urban policies, however, these regional poli-
cies will obviously have social welfare impacts, the size and spatial distribution of which will
depend on the responsiveness of indigenous and immigrant firms to the regional policy
initiatives.
Traditional regional polices have also been understood as partly playing a macroeconomic
efficiency role in situations where aggregate inflation is very sensitive to the local 'bottleneck'
supply conditions in particular core regions. The argument here is that output can be main-
tained while avoiding some ofthe location-specific real-estate-induced influences on macro-
economic crowding out. The economic justification for using regional policy as an
anti-inflationary device therefore rests on the assumption that underlying land-market
imperfections mean that deregulation ofland and planning regimes is not sufficient to coun-
ter these bottleneck problems. Obviously land-market deregulation may contribute to allevi-
ating such bottlenecks, but, as we saw in Chapter 4, land markets are notoriously complex
and highly efficient, and responsive land-supply adjustments cannot necessarily be assumed.
As we have also seen in this chapter, however, recent debates have thrown a completely
new light on the reasons for regional performance differences and also on the possible pol-
icy solutions. More complex discussions regarding the role of space-blind versus place-
based approaches to regional development policy have been reviewed. As we see, new
insights into the nature of social and institutional capital are argued by the place-based
approach to be key issues in regional development, whereas the space-neutral approach
emphasizes interregional factor mobility and the spatial allocation offactors as being domi-
nant. The latter approach therefore recommends removing anything which distorts the
market or limits factor mobility, and at the urban level this would imply the removal of
greenbelts and many forms of urban zoning. In contrast, the place-based approach would
view these issues as second-order problems, with the interactions between factor mobility
and governance and institutional issues as being dominant. Local development initiatives
underpinned by the provision of specifically tailored local public goods would therefore be
encouraged, as long as these initiatives were also designed so as to correctly align the incen-
tives of all stakeholders, and to avoid the enhancement of local monopoly or monopsony
positions. Whereas traditional regional policy was in part argued to be justified on the basis
of reducing 'crowding-ouf effects associated with local bottlenecks, modern regional policy
is intended to encourage the 'crowding-in' effects (Zoellick 2012) of investment.
These arguments relating to regional development actually relate to even wider policy
issues. As we will see in Appendices 10.1 and 10.2, while the space-blind approach suggests
that no regional assistance funding should be provided in weaker regions, the ultimate
logic of the space-blind approach is actually that no welfare funding should be provided
whatsoever, including unemployment assistance, health, or educational services. This is
because the provision of welfare and the identification of the appropriate recipients are
almost entirely place-specific policies, and as such they are regarded by the space-blind
ANAlYS~S
Discussion questions
10.1 What are the economic motives underlying downtown urban redevelopment
policies? Using standard models) discuss the welfare effects of such schemes on
different local income groups.
10.2 Are urban <greenbelt) policies justified on economic and welfare grounds?
10.3 Is regional policy required during recessions when all regions are suffering
unemployment?
10.4 The best regional policy is simply a freer market in land) labour) and capital. Do you
agree with this statement?
10.5 What issues do we need to consider in order to evaluate the effectiveness of a regional
policy which is based on the provision of transport infrastructure?
10.6 To what extent does regional policy have a role to play in alleviating regional
inflationary <bottlenecks)?
10.7 Discuss the differences between a <space-blind) and a <place-based) approach to
regional economic development policy.
10.8 In what ways has the thinking behind the logic) nature) and role of regional policy
changed over recent years?
The first additional type of value of a place above just the use or consumption value is the
existence value. Existence values associated with a sense of place in a region are the values
that individuals or communities in one region attach to the well-being of individuals or
communities in a second region. In order for this existence value to exist (Bolton 1992), it is
not necessary for the individual members of the first region even to go to the second region,
let alone to move to the second region. All that is required is for the individuals in the first
region to know that the well-being of other communities is intact and is not declining.
The second additional value associated with a sense of place is the option value of a place.
This is the value ascribed by an individual person or firm to a possible future option to
migrate there. The more places there are with vibrant and vital local communities, the
greater will be the number and variety of potentially attractive options available for future
migration for all individuals. Such migration possibilities may occur in response to chang-
ing job opportunities (foreseen or unforeseen), and also changing preferences associated
with, for example, age and income.
As we have seen in Chapters 8 and 9, the changing economic geography of firms and
occupations has been a major feature of the last two decades in all countries. Similarly, as we
saw in Chapter 6, changes in residence associated with highly income-elastic preferences for
amenities are also increasing rapidly. The ability to adjust and adapt to these changes is of
both real and increasing value to individuals in all regions, and the greater the number of
vibrant regions, the greater will be the variety of options available for future migration
choices for all individuals.
While these arguments may at first appear to be rather esoteric, in fact modern survey-
based econometric techniques allow us to rigorously estimate both the existence values and
the option values of places. These existence and option value arguments are also closely
related to the attachment value arguments discussed in Chapter 4. However, they are differ-
ent from the attachment value argument, in that whereas the attachment values are local,
the values here extend well beyond the locality itself to observers in other places.
Donor preferences
A different light can also be shed on these issues by considering the preferences of the
donors, and the easiest way to understand this is by considering international aid (Bolton
1992). We can begin by assuming that the pure equity motive results in individuals support-
ing government assistance to other individuals facing hardship, and there are two ways of
achieving this. First, the moneys could be simply transferred to individuals who are able to
spend the funds where and when they wish. However, we know from studies of donor pref-
erences that even for pure equity reasons, donors rarely prefer open-ended monetary trans-
fers with no conditions attached. The consumption or investment behaviour ofthe recipients
is acknowledged to be of importance to the donors, so that the moneys tied up in the trans-
fers are not squandered but are used constructively for reducing some of the hardships
faced by the recipients. Donor preferences generally prefer that the recipients use the funds
constructively in their own locations, so as to partially correct for the local hardships. At the
REGiONAL ~ ..."1"")!~~'V'*ln~,,,~
same time, the preference by donors is to give the funds to a non-corrupt government, char-
ity, aid organization, or development agency, rather than to individuals. This is both to
ensure that the moneys are spent in an accountable manner according to good governance
principles and to allow for the pooling of resources so as to achieve scale impacts from the
expenditure.
As is clear, the donor preferences argument works against the migration-voucher argu-
ment described above. The reason is that the funding is provided by the donors in the
wealthier regions in order to assist the recipients in their own countries or regions. Donors
would generally not pay for the recipients to move to the countries of the donors. l This is
also true in terms of intra-national mobility, and the argument holds at the level of both
interregional mobility and intra-urban mobility. If we consider regional policy in terms of
income transfers from donors in more prosperous regions to recipients in less prosperous
regions, then by the same logic the residents of prosperous areas will not pay for residents
of less favoured regions to move to their locations. The donor preferences argument works
against the migration-voucher argument, and where redistributive funding is supported on
equity grounds, the donor preferences argument supports funding which is place specific in
the less favoured regions.
What is important here is the fact that donor preferences exist in favour of providing
funding in other people's regions in order to foster development and to help other people to
maintain or enhance their livelihoods in their regions, rather than using the moneys donated
to encourage people to migrate to the donors' regions. This suggests that place does have a
value significantly above use or consumption, a value which is not only defined in terms of
the attachment value and sense-of-place value of the residents-relating to all their various
social, cultural, and economic dimensions ofthe place-but also in terms ofthe value placed
on the well-being of residents of other regions.
In addition to these different valuations of place which go above and beyond use values,
there are three other regional mispricing issues which the place-based approach implies
should be taken into account in valuing regions, but which are largely ignored in the inter-
regional equilibrium arguments.
(i) The traditional place-based argument suggests that the capital costs associated with
public intervention in declining regions are systematically overstated (Bolton 1992).
One aspect of this is a second-best argument regarding downwardly rigid local wages
in declining regions, as discussed in Chapter 6. Such sticky wages overstate the true
opportunity costs of utiliZing the local human capital in these regions, and therefore
underestimate the potential value of such interventions. A second aspect of this
argument is that if the provision of infrastructure capital and other public utilities is
priced according to average costs, then the fact that the marginal costs will be lower
I A rare exception here is the case of the UK, where migrants were subsidized in the post-war decades to move
to countries such as Australia and other former British empire dominions. These schemes were largely funded by
the recipient countries aiming at rapid population growth and were a response to their perceived vulnerability due
to their small populations in the aftermath of the Second World War. At the same time the origin countries, and
by far the most important one was the UK, but also other European nations, all faced post-war food and housing
shortages. The UK families paid only ten pounds per family for a one-way no-return ticket to relocate and were
known as 'ten pound poms'. Such migration subsidies and allocations were primarily restricted to migrants who
were perceived to be ethnically and culturally close to the residents of the receiving countries (Clark 1963).
than the average costs in the declining region means that the costs of providing public
goods in these regions will be once again systematically overstated. Conversely, these
same arguments imply that the opportunity costs of utilizing human capital in
growing regions will be systematically underestimated and the net benefits overstated
(Bolton 1992). Following our discussion in Chapters 5 and 6, if the level of mis-
specification is both systematic and significant, as Laurilla (2004, 2011) implies, then
the interregional distribution of activities may well be fundamentally inefficient.
(ii) Because of the long discounting period over which infrastructure should be priced,
new infrastructure in expanding regions should not be developed until it is clear that
the declining regions will not recover in the long run. Otherwise, infrastructure
capital allocation decisions will be inefficient as they will be made with respect to time
periods which are too short (Bolton 1992). If this is indeed the case, as such decisions
are often taken for shorter-term political rather than long-run economic reasons, the
true opportunity costs of infrastructure expenditure in the declining region will be
understated and the benefits in growing regions will be overstated.
(iii) If we expand our concept of capital to include the intangible forms of capital such as
social capital, cultural capital, institutional capital, creative capital, and
entrepreneurial capital, then valuations based only on physical and human capital will
seriously underestimate the long-run opportunity costs of the processes of decline.
As these various arguments all make clear, the value of a place is therefore systematically
larger than the values ascribed to places by factor adjustment models based only on the use
and consumption values of places. However, the extent to which they alter the orthodox
factor adjustment framework depends on the importance of these other valuations and mis-
pricing issues. If the level of undervaluation or mispricing is small, we could argue that the
neoclassical one-sector factor adjustment model will generate interregional factor distribu-
tions which are close to efficient distributions.
Fujita, M. (1989), Urban Economic Theory, Cambridge Arthur, W. B. (1994), Increasing Returns and Path Depend-
University Press, Cambridge. ence in the Economy, University of Michigan Press,
Ann Arbor, MI.
Greenhut, M. L. (1970), A Theory ofthe Firm in Economic
Space, Appleton Century Crofts, New York. Baumol, W. J. (1959), Business Behaviour, Value and
Growth, Macmillan, New York.
l{oover, E. M. (1948), The Location ofEconomic Activity,
McGraw-Hill, New York. Becker, G. S. (1996), Accountingfor Tastes, Harvard
University Press, Cambridge, MA.
Hotelling, H. (1929), 'Stability in Competition', Economic
Journal, 39, 41-57. Boschma, R., and Martin, R. (2010) (eds), Handbook of
Evolutionary Economic Geography, Edward Elgar,
Isard, W. (1956), Location and the Space Economy, John
Cheltenham.
Wiley, New York.
Carlton, D. W, and Perloff, J. M. (2006), Modern Industrial
Isard, W., Azis, I. J., Drennan, M. P., Miller, R. E., Saltz-
Organization (4th edn) , Addison-Wesley, New York.
man, S., and Thorbecke, E. (1998), Methods ofInterre-
gional and Regional Analysis, Ashgate, Brookfield, VT. Cyert, R. M., and March, J. G. (1963), A Behaviourallheory
ofthe Firm, Prentice-Hall, Englewood Cliff'), NJ.
Krugman, P. (1991), Geography and Trade, MIT Press,
Cambridge, MA. d'Aspremont, C., Gabszewicz, J. ]., and Thisse, J. F. (1979),
'On Hotelling's Stability in Competition', Econometrica,
Losch, A. (1954), 1he Economics ofLocation, Yale Univer-
47.5, 1145-50.
sity Press, New Haven, CT.
Delgado, M., Porter, M. E., and Stern, S. (2010), 'Clusters
Marshall, A. (1890), Principles ofEconomics (1st edn),
and Entrepreneurship' Journal ofEconomic Geography,
Macmillan, London.
10.4,495-518.
Marshall, A. (1920), Principles ofEconomics (8th edn),
Eswaran, M., Kanemoto, Y., and Ryan, D. (1981), 'A Dual
Macmillan, London.
Approach to the Locational Decision of the Firm',
Moses, L. N. (1958), 'Location and the Theory of Produc- Journal ofRegional Science, 21.4, 469-89.
tion', Quarterly Journal ofEconomics, 78, 259-72.
Hotelling, H. (1929), 'Stability in Competition', Economic
Palander, T. (1935), Beitriige zur Standortstheorie, Journal, 39, 41-57.
Almqvist and Wiksells Boktryckeri, Uppsala, Sweden.
Lancaster, K. (1971), Consumer Demand: A New
Perroux, F. (1950), 'Economic Space, Theory and Applica- Approach, Columbia University Press, New York.
tions', Quarterly Journal ofEconomics, 64, 89-104.
Laundhart, W. (1885), Mathematische Begriindung der
Porter, M. E. (1990), The Competitive Advantage of Volkswirtschaftslehre, B. G. Taubner, Leipzig.
Nations, Free Press, New York. McCann, P. (1993), 'The Logistics-Costs Location-
Vernon, R. (1960), Metropolis 1985, Harvard University Production problem', Journal ofRegional Science,
Press, Cambridge, MA. 33.4,503-16.
McCann, P. (1997), 'Logistics-Costs and the Location of Alchian, A. A. (1950), 'Uncertainty, Evolution, and
the Firm: A One- Dimensional Analysis', Location Economic Theory', Journal ofPolitical Economy, 58.3,
Science, 4, 101-16. 211-21.
McCann, ~, (1998), The Economics ofIndustrial Location: Arita, T., and McCann, P. (2000), 'Industrial Alliances and
A Logistics-Costs Approach, Springer, Heidelberg. Firm Location Behaviour: Some Evidence from the US
Semiconductor Industry: Applied Economics, 32,
McCann, ~ (2001), 'A Proof of the Relationship between
1391-403.
Optimal Vehicle Size, Haulage Length and the Struc-
ture of Distance-Transport Costs' Transportation Arthur, W B. (1991), Increasing Returns and Path Depend-
Research A, 35.8, 671-93. ence in the Economy, University of Michigan Press,
Ann Arbor, MI.
McCann, P. (2007), 'Sketching out a Model of Innovation,
Face- to- Face Interaction and Economic Geography', Aydalot, P., and Keeble, D. (1988), Milieux Innovateurs en
Spatial Economic Analysis, 2.2, 117-34. Europe, GREM!, Paris.
Miller, S. M., and Jensen, O. W. (1978), 'Location and the Baumol, W. J. (1959), Business Behaviour, Value and
Theory of Production', Regional Science and Urban Growth, Macmillan, New York.
Economics, 8, 117-28.
Becattini, G., Bellandi, M. and De Propris, 1" (2009),
Miron, J. R. (2010), Ihe Geography ofCompetition: Firms, A Handbook ofIndustrial Districts, Edward Elgar,
Prices and Competition, Springer, Heidelberg. Cheltenham.
Moses, L. N. (1958), 'Location and the Theory of Produc- Boschma, R., and Martin, R. (2010) (eds), Handbook of
tion: Quarterly Journal ofEconomics, 78, 259-72. Evolutionary Economic Geography, Edward Elgar,
Palander, T. (1935), Beitriige zur Standortstheorie, Cheltenham.
Almqvist and Wiksells Boktryckeri, Uppsala, Sweden. Boudeville, J. R. (1966), Problems ofRegional Planning,
Rawstron, E. M. (1958), 'The Principles of Industrial Edinburgh University Press, Edinburgh.
Location', Transactions and Papers ofthe Institute of Breschi, S., and Malerba, F. (2005) (eds), Clusters,
British Geographers, 25, 132-42. Networks, and Innovation, Oxford University Press,
ReVelle, C. (1987), 'Urban Public Facility Location', in Oxford.
Mills, E. S., Handbook ofRegional and Urban Econom- Castells, M., and Hall, P. G. (1994), Technopoles ofthe
ics Volume 2: Urban Economics, North-Holland, World: The Makingof21 S1 Century Industrial Com-
Amsterdam. plexes, Routledge, New York.
Sakashita, N. (1968), 'Production Function, Demand Caves, R. E. (2000), Creative Industries: Contracts between
Function, and Location Tbeory of the Firm', Papers and Art and Commerce, Harvard University Press,
Proceedings ofthe Regional Science Association, 20, Cambridge, MA.
109-22.
Chiang, A. C. (1984), Fundamental Methods ofMathemati-
Salop, S. C. (1979), 'Monopolistic Competition with cal Economics, McGraw-Hill, Singapore.
Outside Goods: Bel/Journal ofEconomics, 10.1, 141-56.
Chinitz, B. (1961), 'Contrasts in Agglomeration: New York
Simon, H. A. (1952), 'A Behavioural Model of Rational and Pittsburgh', American Economic Review, 51,
Choice: Quarterly Journal ofEconomics, 52, 99-118. 279-89.
Simon, H. A. (1959), 'Theories of Decision-Making in Chinitz, B. (1964), 'City and Suburb', in Chinitz, B. (ed.),
Economics and Behavioural Science: American Eco- City and Suburb: The Economics ofMetropolitan
nomic Review, 49,253-83. Growth, Prentice-Hall, Englewood Cliffs, NJ.
Weber, A. (1909), Ober den Standort der Industrien, trans. Comunian, R., Faggian, A., Li, C. Q. (2010), 'Unrewarded
by Friedrich, C. J. (1929), Alfred Webers Theory ofthe Careers in the Creative Class: The Strange Case of
Location ofIndustries, University of Chicago Press, Bohemian Graduates', Papers in Regional Science, 89.2,
Chicago, IL. 389-410.
Cyert, R. M., and March, J. G. (1963), A Behavioural
Chapter 2
Theory ofthe Firm, Prentice- Hall, Englewood
Abreu, M., Faggian, A., Comunian, R., and McCann, P. Cliffs, NJ.
(2012), «'Life is Short, Art is Long": The Persistent
de Groot, H. 1" E, Poot, J., and Smit, M. (2009), 'Agglom-
Wage Gap between Bohemian and non-Bohemian
eration Externalities, Innovation and Regional Growth:
Graduates and Occupations', (2012), Annals ofRegional Theoretical Perspectives and Meta-Analysis', in
Science, 49.2,305-21. Cappello, R., and Nijkamp, P., Handbook ofRegional
Acs, Z. J. (2002), Innovation and the Growth ofCities, Growth and Development Theories, Edward Elgar,
Edward Elgar, Cheltenham. Cheltenham, 256-81.
Delgado, M., Porter, M. E., and Stern, S. (2010), (Clusters Grindley, P. C., and 'reece, D. J. (1997), (Licensing and
and Entrepreneurship', Journal ofEconomic Geography, Cross- Licensing in Semiconductors and Electronics',
10.4,495-518. California Management Review, 39.2, 8-41.
Duranton, G., and Puga, D. (2001), (Nursery Cities: Urban Haug, P. (1986), (US High Technology Multinationals and
Diversity, Process Innovation, and the Life Cycle of Silicon Glen', Regional Studies, 20, 103-16.
Products', American Economic Review 91.5,1454-77. Henderson, J. V, Kuncoro, A., and Turner, M. (1995),
Duranton, G., and Puga, D. (2004), (Micro-Foundations of (Industrial Development in Cities', Journal ofPolitical
Urban Agglomeration Economies', in Henderson, J. V, Economy, 103, 1067-85.
and Thisse, J.-F. (eds), Handbook ofRegional and Urban Hoover, E. M. (1937), Location Iheory and the Shoe and
Economics, Vol IV: Economic Geography, Elsevier, Leather Industries, Harvard University Press,
Amsterdam. Cambridge, MA.
Florida, R. (2002), The Rise ofthe Creative Class, Basic Iloover, E. M. (1948), The Location ofEconomic Activity,
Books, New York. McGraw- Hill, New York.
Florida, R. (2005), Ihe Flight of the Creative Class: The Iammarino, S., and McCann, P. (2006), (The Structure and
New Competition for Global Talent, Harper Business, Evolution of Industrial Clusters: Transactions, Tech-
New York. nology and Knowledge Spillovers', Research Policy, 35,
Frenken, K., Van Oort, F. G, and Verburg, T. (2007), 1018-36.
(Related Variety, Unrelated Variety and Regional Isard, W., and Kuenne, R. E. (1953), (The Impact of Steel
Economic Growth', Regional Studies, 41, 685-97. upon the Greater New York- Philadelphia Industrial
Glaeser, E. L. (2008), Cities, Agglomeration and Spatial Region, Review ofEconomics and Statistics, 35, 289-301.
Equilibrium, Oxford University Press, Oxford. Jacobs, J. (1960), The Economy ofCities, Random tIouse,
Glaeser, E. L. (ed.) (2010), Agglomeration Economics, New York.
University of Chicago Press, Chicago, IL. Keeble, D., and Wilkinson, F. (1999), (Collective Learning
Glaeser, E. L. (2011), Triumph ofthe City: How Our and Knowledge Development in the Evolution of High
Greatest Invention Makes Us Richer, Smarter, Greener, Technology SMEs in Europe', Regional Studies, 33.4,
Healthier, and Happier, The Penguin Press, New York. 295-303.
Glaeser, E. L., Kallal, H., Scheinkman, J. A., and Shleifer, Krugman, P. (1991a), Geography and Trade, MIT Press,
A. (1992), (Growth in Cities', Journal ofPolitical Econ- Cambridge, MA.
omy, 100.6, 1126-51. Mameli, E, Faggian, A., and McCann, P (2008), (Employ-
Glaeser, E. L., Kolko, J., and Saiz, A. (2001), (Consumer ment Growth in Italian Local Labour Systems: Issues of
City', Journal ofEconomic Geography, 1.1,27-50. Model Specification and Sectoral Aggregation', Spatial
Economic Analysis, 3.3, 343-59.
Glaeser, E. L., and Gottlieb, J. D. (2006), (Urban Resur-
gence and the Consumer City', Urban Studies, 43.8, Markusen, A. (1985), Profit Cycles, Oligopoly, and Regional
1275-99. Development, MIT Press, Cambridge, MA.
Gordon, I. R., and McCann, P (2000), (Industrial Clusters: Markusen, A. (2006), (Urban Development and the
Complexes, Agglomeration and /or Social Networks', Politics of a Creative Class: Evidence from a Study of
Urban Studies, 37.3, 513-32. Artists: Environment and Planning A, 38, 1921-40.
Gordon, I. R., and McCann, P. (2005), (Innovation, Marshall, A. (1890), Principles ofEconomics (lst edn),
Macmillan, London.
Agglomeration and Regional Development', Journal of
Economic Geography, 5.5, 523-43. Marshall, A. (1920), Principles ofEconomics (8th edn),
Macmillan, London.
Granovetter, M. (1973), (The Strength of Weak Ties',
American Journal ofSociology, 78, 1360-80. McCann, P. (1995a), (Rethinking the Economics of Loca-
tion and Agglomeration', Urban Studies, 32.3, 563-77.
Granovetter, M. (1985), (Economic Action and Social
Structure, American Journal ofSociology, 91,481-510. McCann, P (1995b), (Journey and Transactions Frequency:
An Alternative Explanation of Rent-Gradient
Granovetter, M. (1991), (The Social Cohesion of Economic
Convexity', Urban Studies, 32.9, 1549-57.
Institutions', in Etzoni, A., and Lawrence, R. (eds),
Socio-Economics: Towards a New Synthesis, Armonk, McCann, P. (1997), (How Deeply Embedded is Silicon
New York. Glen? A Cautionary Note', Regional Studies, 31.7,
695-703.
Granovetter, M. (1992), (Problems of Explanations in
Economic Sociology', in Nohria, N., and Eccles, R. McCann, P. (2005), (Transport Costs and New Economic
(eds), Networks and Organizations: Form and Action, Geography', Journal ofEconomic Geography, 5.3,
Harvard Business School Press, Cambridge, MA. 305-18.
McCann, P., and Arita, T. (2006), 'Clusters and Regional Schumpeter, J. A. (1934), 1he Theory ofEconomic Develop-
Development: Some Cautionary Observations from ment, Harvard University Press, Cambridge, MA.
the Semiconductor Industry: Information Economics
Scott, A. J. (1998), New Industrial Spaces, Pion, London.
and Policy, 18.2, 157-80.
Scott, A. J. (2000), The Cultural Economy ofCities, Sage,
McCann, l~, and Fingleton, B. (1996), "Ihe Regional
London.
Agglomeration Impact of Just-In-Time Input Linkages:
Evidence from the Scottish Electronics Industry', Sinlmie, J. (1988), 'Reasons for the Development of
Scottish Journal ofPolitical Economy, 43.5, 493-518. "Islands of Innovation': Evidence from Hertfordshire',
Urban Studies, 35.8,1261-89.
McCann, P., Poot, J., and Sanderson, I.. (2010), 'Migration,
Relationship Capital and International Travel: Theory Simon, H. A. (1952), 'A Behavioural Model of Rational
and Evidence', Journal ofEconomic Geography, 10.3, Choice', Quarterly Journal ofEconomics, 52, 99-118.
361-87. Simon, H. A. (1959), 'Theories of Decision-Making in
Moses, I.. N. (1958), 'Location and the Theory of Produc- Economics and Behavioural Science', American
tion" Quarterly Journal ofEconomics, 78, 259-72. Economic Review, 49, 253-83.
Ohlin, B. (1933), Interregional and International Trade, Suarez-Villa, I.. and Walrod, W. (1997), 'Operational
Harvard University Press, Cambridge, MA. Strategy, R&D and Intra-metropolitan Clustering in a
Polycentric Structure: The Advanced Electronics
Parr, J. B. (1999a), 'Growth-Pole Strategies in Regional
Industries of the Los Angeles Basin', Urban Studies, 34,
Economic Planning: A Retrospective View. Part 1.
1343-80.
Origins and Advocacy', Urban Studies, 36.7,1195-215.
UN (United Nations) (1992), World Urbanization Pros-
Parr, J. B. (1999b), 'Growth-Pole Strategies in Regional
pects 1992, Department of Economic and Social
Economic Planning: A Retrospective View. Part 2. Imple-
Development, United Nations Secretariat, New York.
mentation and Outcome', Urban Studies, 36.8, 1247-68.
Van Oort, F. G. (2004), Urban Growth and Innovation:
Pavitt, K. (1984), 'Sectoral Patterns of Technical Change:
Spatially Bounded Externalities in the Netherlands,
Towards a Taxonomy and a Theory', Research Policy,
Ashgate, Aldershot.
13,343-73.
Vernon, R. (1960), Metropolis 1985, Harvard University
Perroux, F. (1950), 'Economic Space, Theory and Applica-
Press, Cambridge, MA.
tions', Quarterly Journal ofEconomics, 64, 89-104.
Vernon, R. (1966), 'International Investment and Interna-
Polese, M. (2009), The Wealth and Poverty ofNations: Why
tional Trade in the Product Cycle', Quarterly Journal of
Cities Matter, University of Chicago Press, Chicago, II..
Economics, 80,190-207.
Porter, M. E. (1985), Competitive Advantage: Creating and
Weber, A. (1909), Uber den Standort der Industrien, trans.
Sustaining Superior Performance, Free Press, New York.
by Friedrich, C. J. (1929), Alfred Webers Theory ofthe
Porter, M. E. (1990), The Competitive Advantage of Location ofIndustries, University of Chicago Press,
Nations, Free Press, New York. Chicago, II..
Porter, M. E. (1998a), 'Clusters and the New Economics of Williamson, O. E. (1975), Markets and Hierarchies, Free
Competition', Harvard Business Review, 76.6, 77. Press, New York.
Porter, M. E. (1998b), 'Competing Across Locations', in
Porter, M. E. (ed.), On Competition, Harvard Business Chapter 3
School Press, Cambridge, MA.
Amiti, M. (1998), 'New Trade Theories and Industrial
Putnam, R. (1996), Bowling Alone: The Collapse and Location in the EU: A Survey of Evidence', Oxford
Revival ofAmerican Community, Simon and Schuster, Review ofEconomic Policy, 14,45-53.
New York.
Anselin, I.. (1988), Spatial Econometrics: Methods and
Rawstron, E. M. (1958), 'The Principles of Industrial Models, Kluwer Academic Publishers, Dordrecht.
Location, Transactions and Papers ofthe Institute of
Audretsch, D., and Feldman, M. P. (1996), 'R&D Spillovers
British Geographers, 25, 132-42.
and the Geography of Innovation and Production,
Richardson, H.W (1978), Regional Economics, University American Economic Review, 86, 630-40.
of Illinois Press, Urbana, IL.
Baldwin, R., Forslid, R., Martin, P., Ottaviano, G., and
Salop, S. C. (1979), 'Monopolistic Competition with Robert-Nicoud, F. (2003), Economic Geography and
Outside Goods: Bell Journal ofEconomics, 10.1,141-56. Public Policy, Princeton University Press, Princeton, NJ.
Saxenian, A. (1994), Regional Advantage: Culture and Batty, M. (2005), Cities and Complexity: Understanding
Competition in Silicon Valley and Route 128, Harvard Cities with Cellular Automata, Agent-Based Models,
University Press, Cambridge, MA. and Fractals, MIT Press, Cambridge, MA.
Beavon, K. S. O. (1977), Central Place 1heory: A Reinter- Fingleton, B. (2005), 'Beyond Neoclassical Orthodoxy: A
pretation, Longman, London. View Based on the New Economic Geography and UK
Regional Wage Data', Papers in Regional Science, 84.3,
Black, D., and Henderson, V. (1999), 'Spatial Evolution of
351-75.
Population and Industry in the United States', American
Economic Review: Papers and Proceedings, 89.2, 321-7. Fingleton, B. (2010), "Ibe Empirical Perforn1ance of the
NEG with Reference to Small Areas', Journal of
Blair,1. P. (1995), Local Economic Development: Analysis
Economic Geography, 11.2, 267-79.
and Practice, Sage, Thousand Oaks, CA.
Fingleton, B., and Fischer, M. (2010), 'Neoclassical Theory
Bosker, M., and Garretsen, J. II. (2010), 'Trade Costs in
versus New Economic Geography: Competing Expla-
Empirical New Economic Geography: Papers in
nations of Cross- Regional Variation in Economic
Regional Science, 89.3, 485-511.
Development: Annals ofRegional Science, 44, 467-91.
Brakman, S., Garretsen, II., and Van Marrewijk, C.
Fingleton, B., and McCann, P. (2007), 'Sinking the Iceberg?
(2009), A New Introduction to Geographical Economics,
On the Treatment of Transport Costs in New Economic
Cambridge University Press, Cambridge, MA.
Geography', in Fingleton, B. (ed.), New Directions in
Christaller, W (1933), Die Zentralen Orte in Sud- Economic Geography, Edward Elgar, Cheltenham.
deutschland, Fischer, Jena, trans. by Baskin, C. W.
Fotheringham, A. S., Brunsdon, C., and Charlton, M.
(1966), Central Places in Southern Germany, Prentice-
(2000), Quantitative Geography: Perspectives on Spatial
Hall, Englewood Cliffs, NJ.
Data Analysis, Sage, London.
Combes, P.-P. (2000), 'Economic Structure and Local
Fotheringham, A. S., Brunsdon, C., and Charlton, M.
Growth: France 1984-( 1993)', Journal of Urban
(2002), Geographically Weighted Regression: Jhe
Economics, 47, 329-55.
Analysis ofSpatially Varying Relationships, Wiley,
Combes, P.-P., Mayer, 1'., and Thisse J.-E (2008), Economic London.
Geography: 1he Integration ofRegions and Nations,
Fujita, M., Krugman, P., and Venables, A. J. (1999a),
Princeton University Press, Princeton, NJ.
The Spatial Economy, MIT Press, Cambridge, MA.
De Groot, H. L. E, Poot, 1., and Smit, M. (2009), 'Agglom-
Fujita, M., Krugman, P., and Mori, 1'. (l999b), 'On the
eration Externalities, Innovation and Regional Growth:
Evolution of Hierarchical Urban Systems', European
Theoretical Perspectives and Meta-Analysis', in Cap-
Economic Review, 43, 209-51.
pello, R., and Nijkamp, P. (eds), Handbook ofRegional
Growth and Development Theories, Edward Elgar, Fujita, M., and Krugman, P. (1995), 'When is the Economy
Cheltenham. Monocentric? von 'Ibunen and Chamberlin Unified',
Regional Science and Urban Economics, 18, 87-124.
Dewhurst, J. H. 1,., and McCann, P. (2002), 'A Comparison
of Measures of Industrial Specialisation for Travel-To- Fujita, M., and 'Ibisse, J.-F. (2002), Economics ofAgglom-
Work Areas in Great Britain 1981-97', Regional Studies, eration: Cities, Industrial Location, and Regional
36.5,541-51. Growth, Cambridge University Press, Cambridge.
Dixit, A. K., and Stiglitz, J. E. (1977), 'Monopolistic Compe- Gabaix, X. (1999a), 'Zipf's Law and the Growth of Cities',
tition and Optimum Product Diversity', American American Economic Review: Papers and Proceedings,
Economic Review, 67.3, 297-308. 89.2, 129-32.
Duranton, G. (2007), 'Urban Evolutions: The Fast, the Gabaix, X. (l999b), 'Zipf's Law for Cities: An Explanation',
Slow, and the Still', American Economic Review, 97.1, Quarterly Journal ofEconomics, 114.3, 739-67.
197-221. Glaeser, E. L., Kallal, H. D., Scheinkman, J. A., and Shleifer,
Duranton, G., and Overman, H. G. (2005), 'Testing for A. (1992), 'Growth in Cities', Journal ofPolitical
Localization Using Micro-Geographic Data', Review of Economy, 100, 1126-52.
Economic Studies, 72, 1077-106. Guimaraes, P., Figueiredo, 0., and Woodward, D. (2007),
Duranton, G., and Puga, D. (2000), 'Diversity and Spe- 'Measuring the Localization of Economic Activity: A
cialisation in Cities: Why, Where and When Does it Parametric Approach',Journal ofRegional Science, 47,
Matter?', Urban Studies, 37.3, 533-55. 753-74.
Ellison, G., and Glaeser, E. L. (1997), 'Geographic Concen- Greenhut, M. 1,. (1970), A 1heory ofthe Firm in Economic
tration in US Manufacturing Industries: A Dartboard Space, Appleton Century Crofts, New York.
Approach', Journal ofPolitical Economy, 105,889-927. Greenhut, M. 1,., and Ohta, H. (1975), Theory ofSpatial
Ellison, G., Glaeser, E. 1,., and Kerr, W. R. (2010), 'What
Pricing and Market Areas, Duke University Press,
Durham,NC.
Causes Industry Agglomeration? Evidence from
Coagglomeration Patterns', American Economic Haining, R. (2003), Spatial Data Analysis: Theory and
Review, 100.3, 1195-213. Practice, Cambridge University Press, Cambridge.
Harris, C. D. (1954), 'The Market as a Factor in the Locali- McCann, P. (I995a), 'Rethinking the Economics of
zation ofIndustry in the United States', Annals ofthe Location and Agglomeration', Urban Studies, 32.3,
Association ofAmerican Geographers, 44, 315-48. 563-77.
Henderson, J.v., and Thisse, J.-F (2004), Handbook of McCann, l~ (1995b), 'A Proof of the Relationship between
Regional and Urban Economics Vol. 4: Cities and Optimal Vehicle Size, Haulage Length and the Struc-
Geography, Elsevier, Anlsterdam. ture of Distance-Transport Costs: Transportation
Research A, 35.8, 671-93.
IIenderson, J. v., Kuncoro, A., and Turner, M. (1995),
'Industrial Development in Cities: Journal ofPolitical McCann, P. (2005), 'Transport Costs and New Economic
Economy, 103, 1067-85. Geography: Journal ofEconomic Geography, 5.3, 305-18.
Hoare, A. G. (1975), 'Linkage Flows, Locational Evalua- McCann, P. (2007), 'Sketching out a Model of Innovation,
tion and Industrial Geography', Environment and Face-to- Face Interaction and Economic Geography',
Planning A, 7, 241-58. Spatial Economic Analysis, 2.2,117-34.
Holmes, 1'. J. (1999), 'Scale of Local Production and City Menon, C. (2012), 'The Bright Side ofMAUP: Defining
Size', American Economic Review: Papers and Proceed- New Measures of Industrial Agglomeration', Papers in
ings, 89.2, 317-20. Regional Science, 91.1, 3-28.
Hoover, E. M., and Giarratani, F. (1985), An Introduction to Mills, E. S. (1972), Urban Economics, Scott, Foresman and
Regional Economics (3rd edn), Alfred A. Knopf, New York. Co., Glenview, IL.
Kittiprapas, S., and McCann, P. (1999), 'Industrial Loca- Mills, E. S., and Hamilton, B. W. (1994), Urban Economics
tion Behaviour and Regional Restructuring within the (5th edn), HarperCollins, New York.
Fifth "Tiger" Economy', Applied Economics, 31, 37- 51.
Mundell, R. A. (1957), 'The Geometry of Transport Costs
Krugman, P. (I991a), Geography and Trade, MIT Press, in International Trade Theory: Canadian Journal of
Cambridge, MA. Economics and Political Science, 23, 331-48.
Krugman, P. (1991 b), 'Increasing Returns and Economic Samuelson, P. A. (1952), 'The Transfer Problem and
Geography', Journal ofPolitical Economy, 99, 483-99. Transport Costs: The Terms of Trade when Impedi-
Krugman, P (1993), 'On the Number and Location of ments are Absent', Economic Journal, 62, 278-304.
Cities', European Economic Review, 37, 293-8. OECD (2012a), Redefining 'Urban': A New Way to Measure
Krugman, P, and Venables, A. J. (I995), 'Globalization Metropolitan Areas, Organisation for Economic
and the Inequality of Nations', Quarterly Journal of Cooperation and Development, Paris.
Economics, 110.4,857-80. Openshaw S., and Taylor P. J. (1979), 'A Million or So
LeSage, J., and Pace, R. K. (2009), Introduction to Spatial Correlation Coefficients: Three Experiments on the
Econometrics, Chapman and Hall, Boca Raton, FL. Modifiable Areal Unit Problem' in Wrigley, N. (ed.),
Statistical Applications in the Spatial Sciences, Pion,
Lever, W F. (1972), 'Industrial Movement, Spatial Association
London.
and Functional Linkages: Regional Studies, 6, 371-84.
Parr, J. B. (1997), 'The Law of Retail Gravitation: Insights
Lever, WE (I974), 'Manufacturing Linkages and the
from Another Law', Environment and Planning A, 29,
Search for Suppliers', in Hamilton, E E. (ed.), Spatial
1477-95.
Perspectives on Industrial Organization and Decision-
Making, John Wiley, London. Parr, J. B. (2002), 'The Location of Economic Activity:
Central Place Theory and the Wider Urban System: in
Losch, A. (1944), Die Riiumliche Ordnung der Wirtschaft,
McCann, P. (ed.), Industrial Location Economics,
Fischer, Jena, trans. by Woglom, W. H. (1954), The
Edward Elgar, Cheltenham.
Economics ofLocation, Yale University Press, New
Haven, CT. Quah, D. (2001), 'The weightless economy in economic
development: in Pohoja, M. (ed.), Information Technol-
Mameli, E, Faggian, A., and McCann, P. (2008), 'Employ-
ogy, Productivity, and Economic Growth, Oxford
ment Growth in Italian Local Labour Systems: Issues of
University Press, Oxford.
Model Specification and Sectoral Aggregation', Spatial
Economic Analysis, 3.3, 343-59. Redding, S.J., and Venables, A. J. (2004), 'Economic Geog-
raphy and International Inequality: Journal ofInterna-
Marshall, J. N. (1987), 'Industrial Change, Linkages and
tional Economics, 62, 53-82.
Regional Development', in Lever, WE (ed.), Industrial
Change in the United Kingdom, Longman, Harlow. Reilly, W J. (1929), Methods for the Study ofRetail Rela-
tionships, University of Texas Press, Austin, TX.
Maurel, E and S'Edillot, B. (1999), 'A Measure of the
Geographic Concentration of French Manufacturing Reilly, W. J. (I93l), The Law ofRetail Gravitation, Knicker-
Industries: Regional Science and Urban Economics, bocker Press, New York, republished 1953, Pilsbury
29.5,575-604. Publishers, New York.
Richardson, H. W. (1978), Regional Economics, University Laurila, H. (2011), 'Optimisation of City Size', Urban
of Illinois Press, Urbana, IL. Studies, 48.4, 737-47.
Schabenberger, 0., and Gotway, C. A. (2005), Statistical McCann, P. (1993), 'lbe Logistics-Costs Location-
Methods for Spatial Data Analysis, Boca Raton, FL. Production problem, Journal ofRegional Science,
33.4,503-16.
Stelder, D. (2002), 'Geographic Grids in "New Economic
Geography" Models', in McCann, P. (ed.), Industrial McCann, P. (1995), 'Journey and Transactions Frequency:
Location Economics, Edward Elgar, Cheltenham. An Alternative Explanation of Rent-Gradient Convex-
ity', Urban Studies, 32.9, 1549-57.
Stelder, D. (2005), 'Where do Cities Form? A Geographical
Agglomeration Model for Europe', Journal ofRegional McCann, P., (1998), 1he Economics ofIndustrial Location:
Science, 45, 657-79. A Logistics-Costs Approach, Springer, Heidelberg.
Van Oort, F. G. (2004), Urban Growth and Innovation: McCann, P. (2007), 'Sketching out a Model of Innovation,
Spatially Bounded Externalities in The Netherlands, Face- to- Face Interaction and Economic Geography',
Ashgate, Aldershot. Spatial Economic Analysis, 2.2,117-34.
Venables, A. J. (1996), 'Equilibrium Locations of Vertically Mills, E. S. (1969), "Ibe Value of Urban Land', in Perloff, H.
Linked Industries', International Economic Review, (ed.), 1he Quality ofthe Urban Environment, Resources
37.2,341-59. for the Future, Washington, DC.
Zipf, G. (1949), Human Behavior and the Principle ofLeast Mills, E. S. (1972), Urban Economics, Scott, Foresman and
bjjlJrt, Addison-Wesley, New York. Co., Glenview, IL.
Muth., R. (1969), Cities and Housing: The Spatial Pattern of
Urban Residential Land Use, University of Chicago
Chapter 4
Press, Chicago, IL.
Alonso, W (1964), Location and Land Use, Harvard
Ricardo, D. (1821), On the Principles ofPolitical Economy
University Press, Cambridge, MA.
and Taxation (3rd edn), republished by John Murray,
Alonso, W (1971), 'The Economics of Urban Size', Papers London.
and Proceedings ofthe Regional Science Association,
Riguelle, E, Thomas, I., and Verhetsel, A. (2007), 'Measur-
26,67-83.
ing Urban Polycentrism: a European Case Study and its
Ball, M., Lizieri, C., and MacGregor, B. D. (1998), The Implications', Journal ofEconomic Geography, 7,
Economics ofCommercial Property Markets, Routledge, 193-215.
London.
Salop, S. C. (1979), 'Monopolistic Competition with Out-
Button, K. J. (1993), Transport Economics, Edward Elgar, side Goods', Bell Journal ofEconomics, 10.1, 141-56.
Cheltenham.
Takayama, A. (1993), Mathematical Economics (2nd edn),
Capozza, D. R., and Helsey, R. W. (1989), 'The Fundamen- Cambridge University Press, Cambridge.
tals of Land Prices and Urban Growth', Journal of
von lhunen, 1. H. (1826), Der Isolierte Staat in Beziehung
Urban Economics, 26, 295-306.
aufLandtschaft und Nationalokonomie, Hamburg,
Chiang, A. C. (1984), Fundamental Methods ofMathemati- trans. by Wartenberg, C. M. (1966), Von Thunen's
cal Economics, McGraw-Hill, Singapore. Isolated State, Pergamon Press, Oxford.
DiPasquale, D., and Wheaton, W. C. (1996), Urban
Economics and Real Estate Markets, Prentice- Hall, ChapterS
Englewood Cliffs, NJ.
Amiti, M. (1998), 'New Trade Theories and Industrial
Dynarski, M. (1986), 'Residential Attachment and Hous-
Location in the EU: A Survey of Evidence: Oxford
ing Demand', Urban Studies, 23.1, 11-20.
Review ofEconomic Policy, 14,45-53.
Evans, A. W (1973), The Economics ofResidential Loca-
Anderson, J. E., and van Wincoop, E. (2003), 'Gravity with
tion, Macmillan, London.
Gravitas: A Solution to the Border Puzzle', American
Evans, A. W (1983), 'The Determination of the Price of Economic Review, 93.1, 170-92.
Land', Urban Studies, 10.2, 119-29.
Anderson, J. E., and van Wincoop, E. (2004), 'Trade Costs',
Evans, A. W (1985), Urban Economics, Macmillan, London. Journal ofEconomic Literature, 43.3, 691-751.
Hall, P., and Pain, K. (eds) (2006), 1he Polycentric Metrop- Bacharach, M. (1970), Biproportional Matrices and Input-
olis: Learningfrom Mega-City Regions in Europe, Output Change, Cambridge University Press, Cam-
Earthscan, London. bridge.
Laurila, H. (2004), 'Urban Governance, Competition and Batey, P. W., and Madden, M., (1981), 'Demographic-
Welfare', Urban Studies, 41.3, 683-96. Economic Forecasting within an Activity-Commodity
Framework: Some Theoretical Considerations and Hewings, G. J. D., Sonis, M., Madden, M., and Kimura, Y.
Empirical Results', Environment and Planning A, 13, (eds) (1999), Understanding and Interpreting Economic
1067-83. Structure, Springer, Heidelberg.
Batey, P., Bazzazan, E, and Madden, M. (2001), 'Dynamic l-Ioover, E. M., and Giarratani, E (1985), An Introduction
Extended Input-Output Models: Some Initial to Regional Economics (3rd edn), Alfred A. Knopf, New
lboughts: in Felsenstein, D., McQuaid, R., McCann, P., York.
and Shefer, D. (eds), Public Investment and Regional
Isard, W, Azis, I. J., Drennan, M. P., Miller, R. E., Saltzman,
Development: Essays in Honour ojMass Madden,
S., and 1borbecke, E. (1998), Methods ojInterregional
Edward Elgar, Cheltenham.
and Regional Analysis, Ashgate, Brookfield, VT.
Billings, R. B. (1969), 'The Mathematical Identity of the
Isserman, A. M. (1977a), 'The Location Quotient
Multipliers derived from the Economic Base Model
Approach to Estimating Regional Economic Impacts',
and the Input-Output Model', Journal oJRegional
American Institute ojPlanners Journal, 43, 33-41.
Science, 9.3, 471-3.
Isserman, A. M. (1977b), A Bracketing Approach for
Black, P. A. (1981), 'Injection Leakages, Trade Repercus-
Estimating Regional Economic Impact Multipliers and
sions and the Regional Income Multiplier', Scottish
a Procedure for Assessing their Accuracy', Environment
Journal oJPolitical Economy, 28.3, 227-35.
and Planning A, 9,1003-11.
Blair, J. P. (1995), Local Economic Development: Analysis
Isserman, A. M. (1980), 'EstiInating Export Activity in a
and Practice, Sage, Thousand Oaks, CA.
Regional Economy: A lbeoretical and Empirical
Chiang, A. C. (1984), Fundamental Methods ojMathemati- Analysis of Alternative Methods', International
cal Economics, McGraw- Hill, Singapore. Regional Science Review, 5, 155-84.
Czamanski, S., and Malizia, E. (1969), Applicability and Karaska, G. J. (1968), 'Variation of Input-Output
Limitations in the Use of Input-Output Tables for Coefficients for Different Levels of Aggregation',
Regional Studies: Papers and Proceedings ojthe Journal ojRegional Science, 8.2, 215-27.
Regional Science Association, 23, 66-75.
Lahr, M. L. (1993), A Review of the Literature Supporting
Dow, S. C. (1982), 'The Regional Composition of the the Hybrid Approach to Constructing Regional
Money Multiplier Process', Scottish Journal ojPolitical Input-Output Models', Economic Systems Research, 5,
Economy, 29.1, 22-44. 277-92.
Dow, S. C. (1987), "The Treatment of Money in Regional Lahr, M. L. (2001), Input-Output Analysis: Frontiers and
Economics', Journal ojRegional Science, 27.1, 13-24. Extension, Palgrave, Basingstoke.
Dow, S. C., and Rodriguez- Fuentes, C. J. (I997), 'Regional Leontief, W W. (1953), 'Interregional Theory' in Leontief,
Finance: A Survey: Regional Studies, 31.9, 903-20. W W (ed.), Studies in the Structure ojthe American
Economy, Oxford University Press, New York.
Duranton, G., and Puga, D. (2000), 'Diversity and Spe-
cialisation in Cities: Why, Where and When Does it Levacic, R., and Rebmann, A. (1982), Macroeconomics: An
Matter?: Urban Studies, 37.3, 533-55. Introduction to Keynesian-Neoclassical Controversies,
Macmillan, Basingstoke.
Flegg, A. T., Webber, C. D., and Elliot, M. V. (1995), 'On
the Appropriate Use of Location Quotients in Generat- Leven, C. L. (1956), 'Measuring the Economic Base', Papers
ing Regional Input-Output Tables', Regional Studies, ojthe Regional Science Association, 2.1, 250-58.
29,547-62.
Lichtenberg, R. M. (1960), One- Tenth oja Nation, Harvard
Fujita, M., Krugman, P., and Venables, A. J. (1999), The University Press, Cambridge, MA.
Spatial Economy, MIT Press, Cambridge, MA.
McCombie, J. S. L., and Thirlwall, A. P. (1994), Economic
Harrigan, E H., McGilvray, J. W., and McNicoll, I. H. Growth and the Balance-oj-Payments Constraint,
(I 980), A Comparison of Regional and National Macmillan, Basingstoke.
Technical Structures', Economic Journal, 90, 795-810.
McGregor, P., Partridge, M. D., and Rickman, D. S. (2010),
Harris, R. I. D. (1998), 'The Impact of the University of 'Innovations in Regional Computable General Equilib-
Portsmouth on the Local Economy', Urban Studies, rium (CGE) Modellling: Regional Studies, 44.10, 1307-10.
34.4,605-26.
Miller, R. E. (1998), 'Regional and Interregional Input-
Harris, R. I. D., and Liu, A. (1998), 'Input-Output Model- Output Analysis', in Isard, W, Azis, I. J., Drennan, M.
ling of the Urban and Regional Economy: The Impor- P., Miller, R. E., Saltzman, S. and Thorbecke, E. (eds),
tance of External Trade: Regional Studies, 32.9, 851-62. Methods ojInterregional and Regional Analysis, Ash-
gate, Brookfield, VT.
Hartman, L. M., and Seckler, D. (1967), 'Towards the
Application of Dynamic Growth Theory to Regions', NOMIS, National on-Line Manpower Information
Journal oJRegional Science, 7, 167-73. Service, University of Durham, Durham.
Norcliffe, G. B. (1983), 'Using Location Quotients to Bover, 0., Muellbauer, J., and Murphy, A. (1989), 'Hous-
Estimate the Economic Base and 'rrade Flows: Regional ing, Wages and UK Labour Markets', Oxford Bulletin of
Studies, 17.3,161-8. Economics and Statistics, 51.2, 97-136.
Round, J. I. (1978), ~n Interregional Input-Output Cheshire, P. C., and Magrini, S. (2006), 'Population Growth
Approach to the Evaluation of Nonsurvey Techniques', in European Cities: Weather Matters-but Only
Journal ofRegional Science, 18.2, 179-94. Nationally', Regional Studies, 40.1, 23-37.
Schaffer, W, and Chu, K. (1969), 'Nonsurvey Techniques Cheshire, P C., and Sheppard, S .J. (1995», 'On the Price
for Constructing Regional Interindustry Models', of Land and the Value of Amenities', Economica, 62,
Papers and Proceedings ofthe Regional Science Associa- 247-67.
tion, 23, 35-50.
Cheshire, P C., and Sheppard, S. J. (1998), 'Estimating the
Sinclair, M. T. and Sutcliffe, C. M. S. (1978), 'The First Demand for Housing, Land, and Neighbourhood
Round of the Keynesian Regional Income Multiplier', Characteristics', Oxford Bulletin ofEconomics and
Scottish Journal ofPolitical Economy, 25.2, 177-85. Statistics, 60.3, 357-82.
Sinclair, M. T., and Sutcliffe, C. M. S. (1983), 'Injection Chiang, A. C. (1984), Fundamental Methods ofMathemati-
Leakages, Trade Repercussions and the Regional cal Economics, McGraw- Hill, Singapore.
Income Multiplier: An Extension', Scottish Journal of
Christopherson, S., Michie, J., and Tyler, P. (2010),
Political Economy, 30.3, 275-86.
'Regional Resilience: Theoretical and Empirical
Smith, P, and Morrison, W. I. (1974), Simulating the Perspectives', Cambridge Journal ofRegions, Economy
Urban Economy, Pion, London. and Society, 3.1, 3-10.
Thorne, E. M. F. (1969), 'Regional Input-Output Analysis', Clower, R. W. (1965), 'The Keynesian Counter-Revolution:
in Orr, S. C., and Cullingworth, 1. 8., Regional and A Theoretical Appraisal', in Hahn, F. H., and Brechling,
Urban Studies, George Allen and Unwin, London. F. (eds), Ihe Theory ofInterest Rates, Macmillan,
Ullman, E., and Dacey, M. (1960), 'The Minimum London.
Requirements Approach to the Urban Economic Base', DaVanzo, 1. (1976), 'Differences between Return and
Papers and Proceedings ofthe Regional Science Associa- Nonreturn Migration: An Econometric Analysis: Inter-
tion, 6, 174-94. national Migration Review, 13-27.
Weiss, S. J., and Gooding, E. C. (1968), 'Estimation of DaVanzo, J., and Morrison, P. A. (1981), 'Return and
Differential Employment Multipliers in a Small Other Sequences of Migration in the U.S:, Demography,
Regional Economy', Land Economics, 44, 235-44. 18,85-101.
Wilson, J. H. (1977), 'Impact Analysis and Multiplier DaVanzo, J. (1983), 'Repeat Migration in the United States:
Specification', Growth and Change, 8, 42-6. Who Moves Back and Who Moves On?', Review of
Economics and Statistics, 65, 552-9.
Faggian, A., Corcoran, ]., and McCann, P. (2010), 'Human McCann, P., Poot, ]., and Sanderson, L. (2010), 'Migration,
Capital in Remote and Rural Australia: The Role of Relationship Capital and International Travel: Theory
Graduate Migration: Growth and Change, 41.2, 192-220. and Evidence: Journal ofEconomic Geography, 10.3,
361-87.
Fingleton, B., Garretsen, H., and Martin, R. (2012),
'Recessionary Shocks and Regional Employnlent: McCombie,]. S. L. (1988), A Synoptic View of Regional
Evidence on the Resilience of U.K. Regions', Journal of Growth and Unemployment: I-The Neoclassical
Regional Science, 52.1, 109-33. 'Theory', Urban Studies, 25, 267-81.
Fielding, A. ]. (1992), 'Migration and the Metropolis: Mellander, C., Florida, R., and Stolarick, K. (2011), 'Here to
Recent Research on the Causes and Consequences of Stay-lbe Effects of Community Satisfaction on the
Migration to the Southeast of England', Progress in Decision to Stay', Spatial Economic Analysis, 6.1,5-24.
I-Iuman Geography, 17.2, 195-212. Mincer,]. (1974), Schooling, Experience, and Earnings,
Fujita, M., Krugman, P., and Venables, A. ]. (1999), The Columbia University Press, New York.
Spatial Economy, MIT Press, Cambridge, MA. Minford, P., Ashton, P., and Peel, M. (1988), 'The Effects of
Glaeser, E. L., Kolko, ]., and Saiz, A. (2001), 'Consumer Housing Distortions on Unemployment: Oxford
City', Journal ofEconomic Geography, 1.1,27-50. Economic Papers, 40, 322-45.
Gordon, I. R. (1978), 'Distance Deterrence and Commod- Molho, I. (1986), 'Theories of Migration: A Review:
ity Values', Environment and Planning A, 10,889-900. Scottish Journal ofPolitical Economy, 33, 396-419.
Gordon, I. R. (1985), 'The Cyclical Sensitivity of Regional OECD (2011), Regions at a Glqnce 2011, Organisation for
Employment and Unemployment Differentials: Economic Cooperation and Development, Paris.
Regional Studies, 19.2,95-110. Partridge, M. D. (2010), 'The Duelling Models: NEG vs
Gottlieb, ]. D., and Glaeser, E. L. (2006), 'Urban Resurgence Amenity Migration in Explaining US Engines of
and the Consumer City: Urban Studies, 43.8, 1275-99. Growtl1, Papers in Regional Science, 89.3, 513-36.
Graves, P. E. (1980), 'Migration and Climate', Journal of Partridge, M. D., Rickman, D. S., Olfert, M. R., and Ali, K.
Regional Science, 20.2, 227-37. (2012), 'Dwindling U.S. Internal Migration: Evidence
of Spatial Equilibrium: Regional Science and Urban
Graves, P. E. (1983), 'Migration with a Composite Amenity:
Economics, 42.2, 375-88.
The Role of Rents: Journal ofRegional Science, 23.4, 541-6.
Plane, D. A. (1983), 'Demographic Influences on Migra-
Hall, P. (1998), Cities in Civilisation, Weidenfeld and
tion: Regional Studies, 27.4, 375-83.
Nicolson, London.
Quah, D. (1997), 'Empirics for Growth and Distribution:
Harris, J. R., and Todaro, M. P. (1970), 'Migration, Unem-
Polarization, Stratification, and Convergence Clubs:
ployment and Development: A Two-Sector Analysis:
Journal ofEconomic Growth, 2.1, 27-59.
American Economic Review, 60.1,126-42.
Roback,]. (1982), 'Wages, Rents, and the Quality of Life',
Hemmings, P. J. (1991), 'Regional Earnings Differences in
Journal ofPolitical Economy, 90.6,1257-78.
Great Britain: Evidence from the New Earnings
Survey', Regional Studies, 25.2, 123-33. Roy, A. D. (1951), 'Some Thoughts on the Distribution of
Earnings: Oxford Economic Papers, 3,135-46.
Hughes, G., and McCormick, B. (1981), 'Do Council
House Policies Reduce Migration between Regions?', Shapiro, J. M. (2006), 'Smart Cities: Quality of Life,
Economic Journal, 91, 919-37. Productivity, and the Growth Effects of Human
Capital', Review ofEconomics and Statistics, 88.2, Barro, R. J., and Sala-i-Martin, X. (1995), Economic
324-35. Growth, McGraw- Hill, New York.
Simmie, J., and Martin, R. (2010), 'The Economic Resil- Bathelt, H. (2008), 'Knowledge-Based Clusters: Regional
ience of Regions: Towards an Evolutionary Approach', Multiplier Models and the Role of "Buzz" and "Pipe-
Cambridge Journal ofRegions, Economy and Society, lines»), in Karlsson, C. (ed.), Handbook ofResearch on
3.1,27-43. Cluster Theory, Edward Elgar, Cheltenham and North-
ampton,MA.
Stoneman, P. (1983), The Economic Analysis ofTechnologi-
cal Change, Oxford University Press, Oxford. Bathelt, H., Malmberg, A. and Maskell, P. (2004), 'Clusters
and Knowledge: Local Buzz, Global Pipelines and the
Storper, M., and Kemeny, T. (2012), 'lbe Sources of Urban
Process of Knowledge Creation', Progress in Human
Development: Wages, I-lousing, and Amenity Gaps
Across American Cities', Journal ofRegional Science,
Geography, 28.1,31-56.
52.1,85-108. Berg, A. G., and Ostry, J. D. (2011), 'Inequality and Unsus-
Tiebout, C. M. (1956), 'A Pure Theory of Local Expendi- tainable Growth: Two Sides of the Same Coin, IMF
tures', Journal ofPolitical Economy, 64.5, 416-24. Staff Discussion Note, SDNI11108, International
Monetary Fund, Washington, DC.
Thirlwall, A. P. (1994), Growth and Development (5th edn),
Macmillan, Basingstoke. Black, F. (2010), Exploring General Equilibrium, MIT
Press, Cambridge, MA.
Vanclay, E, Higgins, M., and Blackshaw, A. (2008), Making
Sense ofPlace: Exploring Concepts and Expressions of Boulier, B. L. (1984), 'What Lies behind Verdoorn's Law?',
Place Through Different Senses and Lenses, National Oxford Economic Papers, 36, 259-67.
Museum of Australia Press, Canberra. Borts, G. H., and Stein, J. L. (1964), Economic Growth in a
Wilson, A. G. (1974), Urban and Regional Models in Free Market, Columbia University Press, New York.
Geography and Planning, Wiley, London. Boschma, R. A. (2005), 'Proximity and Innovation: A
Critical Assessment', Regional Studies, 39.1, 61-74.
Durlauf, S. N., and Quah, D. (1999), 'The New Empirics of Krugman, P. (1989), 'Differences in Income Elasticities and
Economic Growth', in Taylor, J. B., and Woodford, M. Trends in Real Exchange Rates', European Economic
(eds), Handbook ofMacroeconomics, North-Holland, Review, 33.5,1031-46.
Amsterdam. Lavoie, M. (1992), Foundations ofPost-Keynesian Economic
Felipe, J. (1998), 'The Role of the Manufacturing Sector in Analysis, Edward Elgar, Cheltenham.
Southeast Asian Development: A Test of Kaldor's First Lucas, R. E. (1988), 'On the Mechanics of Economic
Law: Journal ofPost Keynesian Economics, 20.3, Development: Journal ojMonetary Economics, 22, 3-42.
463-85.
Martin, R., and Sunley, P. (1998), 'Slow Convergence? The
Felipe, J., and McCombie, J. S.L. (2012), 'Problems with
New Endogenous Growth Theory and Economic
Regional Production Functions and Estimates of
Development', Economic Geography, 74, 201-27.
Agglomeration Economies: A Caveat Emptor for Regional
Scientists: Spatial Economic Analysis, 7.4,461-84. McCann, P. (2007), 'Sketching out a Model of Innovation,
Face-to-Face Interaction and Economic Geography',
Fingleton,8. (1997), 'Specification and Testing of Markov
Spatial Economic Analysis, 2.2, 117-34.
Chain Models: An Application to Convergence in the
European Union, Oxford Bulletin ofEconomics and McCann, P., and Acs, Z. J. (2011), 'Globalisation: Coun-
Statistics, 59.3,385-403. tries, Cities and Multinationals', Regional Studies, 45.1,
17-32.
Fingleton, 8., and McCombie, J. S. L. (1998), 'Increasing
Returns and Economic Growth: Some Evidence for McCombie, J. S. L., and Roberts, M. (2007), 'Returns to
Manufacturing from the European Union, Oxford Scale and Regional Growth: The Static-Dynamic
Economic Papers, 51.3, 574-80. Verdoorn Law Paradox Revisited', Journal ojRegional
Science, 47.2, 179-208.
Pingleton, 8., Garretsen, H., and Martin, R. (2012),
'Recessionary Shocks and Regional Employment: McCombie, J. S .L., and Thirlwall, A. P. (1994), Economic
Evidence on the Resilience of U.K. Regions', Journal of Growth and the Balance-oj-Payments Constraint,
Regional Science, 52.1, 109-33. Macmillan, Basingstoke.
Neffke, E, Henning, M., and Boschma, R. (2011), 'The Stern, N. (2009), A BlueprintJor a Safer Planet: IIow to
Impact of Aging and Technological Relatedness on Manage Climate Change and Create a New Era of
Agglomeration Externalities: A Survival Analysis', Progress and Prosperity, Bodley Head, London.
Journal ofEconomic Geography, forthcoming.
Stiglitz, J. E., Sen, A., and Fitoussi, J.-~ (2009), Commission of
Nijkamp, P., and Poot, J. (1998), 'Spatial Perspectives on the Measurement ofEconomic Performance and Social
New Theories of Economic Growth', Annals ofRegional Progress, http://www.stiglitz-sen-fitoussi.fr/en/index.htm
Science, 32.1, 7-38.
Swan, 1'. (1956), 'Economic Growth and Capital Accumu-
OECD (2009a), How Regions Grow: Trends and Analysis, lation', Economic Record, 32, 334-61.
Organisation for Economic Cooperation and Develop-
Tabellini, G. (2010), 'Culture and Institutions: Economic
ment, Paris.
Development in the Regions of Europe', Journal ofthe
OECD (2009b), Regions Matter: Economic Recovery, European Economic Association, 8, 677-716.
Innovation and Sustainable Growth, Organisation for
Thirlwall, A. P. (1980), 'Regional Problems are "Balance of
Economic Cooperation and Development, Paris.
Payments" Problems', Regional Studies, 14,419-25.
Porter, M. E. (1990), 1he Competitive Advantage of
Verdoorn, P. J. (1949), 'Fattori che regolano 10 sviluppo
Nations, Free Press, New York.
della produttivita dellavoro, L'Industria, 1,3-10, trans.
Putnam, R. (1993), Making Democracy Work: Civic by Thirlwall, A. P., and Ironmonger, D., in Perkins, J.,
Traditions in Modern Italy, Princeton University Press, and Hoa, 1'. (eds) (1988), National Income and Eco-
Princeton, NJ. nomic Progress: Essays in Honour ofColin Clark,
Putnam. R. (1996), Bowling Alone: The Collapse and Macmillan, London.
Revival ofAmerican Community, Simon and Schuster, Westlund, II. (2006), Social Capital and the Knowledge
New York. Economy: Theory and Empirics, Springer, Berlin.
Quah, D. (1997), 'Empirics for Growth and Distribution: Williamson, O. E. (1975), Markets and lIierarchies, Free
Stratification, Polarization, and Convergence Clubs', Press, New York.
Journal ofEconomic Growth, 2.1, 27-59.
World Bank (2009), Doing Business 2009: Comparing
Ramos, P. (2007), 'Does the Trade Balance Really Matter Regulation in 181 Countries, Washington, DC.
for Regions?', Annals ofRegional Science, 41.1, 229-43.
World Bank (2010), World Development Report 2010:
Richardson, H. W. (1978), Regional Economics, University Development and Climate Change, World Bank,
of Illinois Press, Urbana, IL. Washington, DC.
Romer, P. M. (1986), 'Increasing Returns and Long-Run
Growt11,Journal ofPolitical Economy, 94,1002-37. ChapterS
Romer, P. M. (1987a), 'Growth Based on Increasing Abrahamson, M. (2004), Global Cities, Oxford University
Returns due to Specialization', American Economic Press, Oxford.
Review, 77.2, 56-62.
Alesina, A., and Spolaore, E. (2003), The Size ofNations,
Romer, P. M. (1987b), 'Crazy Explanations of the Produc- MIT Press, Cambridge, MA.
tivity Slowdown', National Bureau ofEconomic
Research Macroeconomics Annual 1987, MIT Press, Benjamin, 1'. (2009), The Atlantic World: Europeans,
Cambridge, MA. Africans, Indians and Their Shared History, 1400-1900,
Cambridge University Press, Cambridge.
Rowthorn, R. E. (1975), 'What Remains of Kaldor's Law?',
Economic Journal, 85, 10-19. Burdett, R., Taylor, M., and Kaasa, A. (eds) (2011), Cities,
Health and Wellbeing, Proceedings of the Hong Kong
Schumpeter, J. A. (1934), The Theory ofEconomic Develop- Urban Age Conference, 16-17 November, LSE Cities,
ment, Harvard University Press, Cambridge, MA. London.
Scott, M. (1989),A New View ofEconomic Growth, Claren- Cairncross, F. (1997), The Death ofDistance: Ilow The
don Press, Oxford. Communications Revolution will Change our Lives,
Shearmur, R., and Polese, M. (2007), 'Do Local Factors Orion Business Books, London.
Explain Local Employment Growth? Evidence from
Chandler, 1~ (1987), Four Thousand Years of Urban
Canada, 1971-2001', Regional Studies, 41.4, 453-71.
Growth: An Historical Census, The Edwin Mellen Press,
Solow, R. M. (1956), 'A Contribution to the Theory of Lampeter, UK.
Economic Growth', Quarterly Journal ofEconomics,
Christaller, W (1933), Die Zentralen Orte in Sud-
70,65-94.
deutschland, Fischer, Jena, trans. by Baskin, C. W
Solow, R. M. (1994), 'Perspectives on Growth Theory', (1966), Central Places in Southern Germany, Prentice-
Journal ofEconomic Perspectives, 8, 45-54. Hall, Englewood Cliffs, NJ.
Cohen, B. J. (1998), Ihe Geography ofMoney, Cornell Kali, R., and Reyes, J. (2006), 'The Architecture of Globali-
University Press, Ithaca, NY. zation: A Network Approach to International Eco-
nomic Integration', Working Paper: Department of
Coleman, W. D. (1996), Financial Services, Globalization
Economics, University of Arkansas.
and Domestic Policy Change, Macmillan, London.
Le Gales, l? (2002), European Cities, Oxford University
Crafts, N. (2004), 'Globalization and Economic Growth: A
Press, Oxford.
Historical Perspective', 1he World Economy, 27.1, 45-58.
Levinson, M. (2006), 1he Box: How the Shipping Container
European Commission (2010), Survey on Perception of
Quality ofLife in 75 European Cities, Directorate Made the World Smaller and the World Economy Bigger,
Princeton University Press, Princeton, NJ.
General for Regional Policy, Brussels.
MacGillivray, A. (2006), A Briefl-listory ofGlobalization,
Ferguson, N. (2008), The Ascent ofMoney: A Financial
Robinson, London.
History ofthe World, Allen Lane, London.
Ferguson, N. (2011), Civilization: The West and the Rest, Maddison, A. (2006), Ihe World Economy. Volume 1: A
Penguin Press, New York. Millennial Perspective Volume 2: Historical Statistics,
Organisation for Economic Cooperation and Develop-
Findlay, R., and O'Rourke, K. (2007), Power and Plenty: ment, Paris.
Trade, War, and the World Economy in the Second
Millennium, Princeton University Press, Princeton, NJ. Maddison, A. (2007a), Contours ofthe World Economy
1-2030: Essays in Macro-Economic History, Oxford
Fischer, S. (2003), 'Globalization and its Challenges', University Press, Oxford.
American Economic Review: Papers and Proceedings,
93.2,1-30. Maddison, A. (2007b), Chinese Economic Performance in
the Long Run, Second Edition Revised and Updated
Fosher, R. S., Alonso, W, Meyer, J. A., and Kern, R. (1990), 960-2030AD, Organisation for Economic Cooperation
Demographic Change and the American Future, Uni- and Development, Paris.
versity of Pittsburgh Press, Pittsburgh, PA.
Marshall, A. (1890), Principles ofEconomics (1st edn),
Fothergill, S., Kitson, M., and Monk, S. (1985), Urban Macmillan, London.
Industrial Decline: The Causes of Urban/Rural Contrasts
in Manufacturing Employment Change, HMSO, London. McCann, l? (2008), 'Globalization and Economic Geogra-
phy: The World is Curved, Not Flat: Cambridge Journal
Friedman, T.L. (2007), The World is Flat: A BriefHistory of ofRegions, Economy and Society, 1.3,351-70.
the Twenty-First Century (3rd edn), Picador, New York.
McCann, P., and Acs, Z. J. (201l), 'Globalisation: Countries,
Gaspar, J., and Glaeser, E.L. (1998), 'Information Technol-
Cities and Multinationals', Regional Studies, 45.1,17-32.
ogy and the Future of Cities: Journal of Urban Econom-
ics, 43, 136-56. McDonald, J. F. (2008), Urban America: Growth, Crisis and
Rebirth, M.E. Sharpe, New York.
Glaeser, E. L. (1998), l\re Cities Dying?: Journal ofEco-
nomic Perspectives, 12.2, 139-60. MGI (2009), Preparingfor China's Urban Billion, McKinsey
Global Institute.
Glaeser, E. L. (2005), 'Urban Colossus: Why is New York
America's Largest City?: Economic Policy Review: MGI (2010), India's Urban Awakening: Building Inclusive
Urban Dynamics in New York, Federal Reserve Bank of Cities, Sustaining Economic Growth, McKinsey Global
New York, New York. Institute.
Glaeser, E. L. (201l), Triumph ofthe City: How Our MGI (2011a), Urban World: Mapping the Economic Power
Greatest Invention Makes Us Richer, Smarter, Greener, ofCities, McKinsey Global Institute.
Healthier, and Happier, The Penguin Press, New York. MGI (2011 b), Building Globally Competitive Cities: The
Glaeser, E. L. and Gottlieb, J. D. (2009), 'The Wealth of Key to Latin American Growth, McKinsey Global
Cities: Agglomeration Economies and Spatial Equilib- Institute.
rium in the United States', Journal ofEconomic Litera-
Modelski, G. (2003), World Cities - 3000 to 2000, Faros
ture, 47.4, 983-1028. 2000, Washington, DC.
Hall, P. (1966), The World Cities, Weidenfeld and Nicolson,
Ni, P., and Kresl, P. K. (2010), The Global Urban Competi-
London.
tiveness Report 2010, Edward Elgar, Cheltenham.
Hall, P. (1988), Cities ofTomorrow, Blackwell, Oxford.
O'Brien, R. (1992), Global Financial Integration: The End of
Hall, l? (1998), Cities in Civilisation, Weidenfeld and Geography, Council on Foreign Relations Press, New
Nicolson, London. York.
Hoover, E. M. (1937), Location Theory and the Shoe OECD (2006), Competitive Cities in the Global Economy,
and Leather Industries, Harvard University Press, Organisation for Economic Cooperation and Develop-
Cambridge, MA. ment, Paris.
OECD (2007), International Investment Perspectives: UNDP (2009), Iluman Development Report 2009: Over-
Freedom ofInvestment in a Changing World, Organisa- coming Barriers: Human Mobility and Development,
tion for Economic Cooperation and Development, United Nations Development Program, Palgrave
Paris. Macmillan, Basingstoke.
OECD (2009), Globalisation and Emerging Economies: UNFPA (2007), State ofthe World's Population: Unleashing
Brazil, Russia, India, Indonesia, China and South the Potential of Urban Growth, United Nations Popula-
Africa, Organisation for Economic Cooperation and tion Fund, New York.
Development, Paris.
Venables, A. J. (2006), 'Shifts in Economic Geography and
OECD (2012), Redefining Urban: Functional Urban Areas their Causes', Federal Reserve Bank ofKansas City
in OECD Countries, Organisation for Economic Economic Review, 91.4, 61-85.
Cooperation and Development, Paris.
Warf, B. (1995), 'Telecommunications and the Clustering
Ohlin, B. (1933), Interregional and International Trade, Geographies of Knowledge Transmission in the Late
Harvard University Press, Cambridge, MA. 20th Century', Urban Studies, 32.2, 361-78.
Peri, G. (2005), 'Determinants of Knowledge Flows and Weber, A. (1909), Ober den Standort der Industrien, trans.
their Effect on Innovation', Review ofEconomics and by Friedrich, C. J. (1929), Alfred Weber's Theory ofthe
Statistics, 87.2, 308-22. Location ofIndustries, University of Chicago Press,
Chicago, IL.
Polese, M. (2009), Ihe Wealth and Poverty ofNations: Why
Cities Matter, University of Chicago Press, Chicago, IL. World Bank (2008), World Trade Indicators 2000: Bench-
marking Policy and Performance, Washington, DC.
PWC (2009), 'Which are the Largest City Economies in
the World and Might this Change by 2025?', UK Zenou, y. (2009), Urban Labor Economics, Cambridge
Economic Outlook November, Price Waterhouse University Press, Cambridge.
Coopers, London. Zook, M. A., (2005), Ihe Geography ofthe Internet,
Reyes, J. A., Garcia, M., and Lattimore, R. (2009), 'The Blackwell, Oxford.
International Economic Order and Trade Architecture',
Spatial Economic Analysis, 4.1,73-102.
Chapter 9
Richardson, H. W., and Bae, C. H. C. (eds), Globalization
Abreu, M., Comunian, R., Faggian, A., and McCann, P (2011)
and Urban Development; Springer, Heidelberg.
"'Life is Short, Art is Long": 'The Persistent Wage Gap
Rugman, A. (2000), The End ofGlobalization, Random between Bohemian and non-Bohemian Graduates and
House, New York. Occupations: Annals ofRegional Science, forthcoming.
Rugman, A. (2005), The Regional Multinationals, Acs, z. J. (2002), Innovation and the Growth ofCities,
Cambridge University Press, Cambridge. Edward Elgar, Cheltenham.
Sassen, S. (1994), Cities in a World Economy, Pine Forge Acs, Z. J., Bosma, N., and Sterberg, R. (2008), Ihe Entre-
Press, Thousand Oaks, CA. preneurial Advantage of World Cities: Evidence from
Global Entrepreneurship Monitor Data, Working Paper,
Sassen, S. (2001), The Global City: New York, London,
H(2008) 10, SCALES, University of Utrecht and
Tokyo, Princeton University Press, Princeton, NJ.
Netherlands Ministry of Economic Affairs.
Sassen, S. (2002) (ed.), Global Networks: Linked Cities,
Aguilera, A. (2008), 'Business Travel and Mobile Workers',
Routledge, London.
Transportation Research A, 42, 1109-16.
Sassen, S. (2006), Cities in a World Economy (3rd edn),
Aitken, B., Hanson, G., and Harrison, A. E. (1997),
Pine Forge Press, Thousand Oaks, CA.
'Spillovers, Foreign Investment, and Export Behaviour',
Satterthwaite, D. (2005), The Scale of Urban Change Journal ofInternational Economics, 43,103-32.
Worldwide 1950-2000 and its Underpinnings, Interna-
Andersson, M. (2009), 'External Trade and Internal
tional Institute for Environment and Development,
Geography', CESIS Working Paper, Jonkoping Univer-
London.
sity, Jonkoping and Royal Institute of Technology,
Steger, M. B. (2003), Globalization: A Very Short Introduc- Stockholm, Sweden.
tion, Oxford University Press, Oxford.
Anselin, L., Varga, A., and Acs, Z. J. (1997). 'Local Geo-
Taylor, P J. (2004), World City Network: A Global Urban graphic Spillovers between University Research and
Analysis, Routledge, London. High Technology Innovations', Journal of Urban
Economics, 42, 442-8.
UNC'T'AD (2003), World Investment Report: FDI Policies
for Development: National and International Perspec- Ante, S. E. (2009), Creative Capital: Georges Doriot and the
tives, United Nations Conference on Trade and Devel- Birth ofVenture Capital, Harvard Business School
opment, United Nations, New York and Geneva. Press, Cambridge. MA.
Arita, T., and McCann, P. (2000), 'Industrial Alliances and Cappelen, A., Castellacci, E, Fagerberg, J., and Verspagen,
Pirm Location Behaviour: Some Evidence frOlll the US B. (2003), 'Regional Disparities in Income and Unem-
Semiconductor Industry: Applied Economics, 32, ployment in Europe: in Fingleton, B. (ed.), European
1391-403. Regional Growth, Springer, Heidelberg.
Arita, T., and McCann, P. (2006), 'Clusters and Regional Carlino, G. A., Chatterjee, S., and Hunt, R. M. (2007),
Development: Some Cautionary Observations from 'Urban Density and the Rate of Invention', Journal of
the Semiconductor Industry', Information Economics Urban Economics, 61.3, 389-419.
and Policy, 18.2, 157-80.
Chatterji, 1., and Tsai, C.-M. (2006), 'Transport Logistics
Arora, A., and Porman, C. (2007), 'How Local are IT in the Global Economy: Spatial Implications', in
Services Markets? Proximity and IT Outsourcing', Kobayashi, K., Lakshmanan, 1: R., and Anderson, W P.
Journal ofManagement Information Systems, 24.2, (eds), Structural Change in Transportation and Com-
73-102. munications in the Knowledge Society, Edward Elgar,
Cheltenham.
Bel, G., and Fageda, X. (2008), 'Getting there Fast: Globali-
zation, Intercontinental Plights and Location of Head- Chen, N. (2004), 'Intra-National versus International
quarters', Journal ofEconomic Geography, 8, 471-95. Trade in the European Union: Why Do National
Borders Matter?', Journal ofInternational Economics,
Berry and Glaeser, E. 1. (2005), 'The Divergence of
63,93-118.
Human Capital Levels across Cities', Papers in Regional
Science, 84.3, 407-44. Ciccone, A. and Hall, R. E. (1996), 'Productivity and the
Density of Economic Activity', American Economic
Best, M. H. (1990), The New Competition: Institutions of
Review, 86, 54-70.
Industrial Restructuring, Polity Press, Cambridge.
COL (2002), Aviation Services for London, Corporation of
Blum, B. B., and Goldfarb, A. (2006), 'Does the Internet
London, London.
Defy the Law of Gravity?', Journal ofInternational
Economics, 70.2, 384-405. COL (2005a), The Competitive Position ofLondon as a
Global Financial Centre, Corporation of London,
Boulhol, H., and de Serres, A. (2010), 'I-lave Countries
London.
Escaped the Curse of Distance?', Journal ofEconomic
Geography, 10.1, 113-39. COL (2005b), Off-shoring and the City ofLondon, Corpo-
ration of London, London.
Brakman, S., and van Marrewijk, C. (2008), 'It's a Big World
After All: On the Economic Impact of Location and COL (2006), The Importance of Wholesale Financial
Distance', Cambridge Journals ofRegions, Economy and Services to the EU Economy 2006, Corporation of
Society, 1,411-37. London, London.
Broersma, 1., and Van Dijk, J. (2008), 'The Effect of COL (2007), The Impact ofRecent Immigration on the
Congestion and Agglomeration on MFP-growth in London Economy, Corporation of London, London.
Dutch Regions', Journal ofEconomic Geography, 8.2,
COL (2008), The Global Financial Centres Index 3, Corpo-
181-209.
ration of London, London.
BTRE (2004), Focus on Regions 3: Taxable Income, Bureau
COL (2009), The Global Financial Centres Index 6, Corpo-
of Transport and Regional Economics, Canberra,
ration of London, London.
Australia.
Combes, P.-P., and Lafourcade, M. (2005), 'Transport
Burghouwt. G. (2005), Airline Network Development in
Costs: Measures, Determinants, and Regional Policy
Europe and its Implications for Airport Planning,
Implications for France: Journal ofEconomic Geogra-
Utrecht University Press, Netherlands.
phy,319-49.
Button, K., Lall, S., Stough, R., and Trice, M. (1999), 'High-
Crafts, N. (2004), 'Globalization and Economic Growth: A
Technology Employment and Hub Airports', Journal of
Historical Perspective', The World Economy, 27.1, 45-58.
Air Transport Management,S, 53-9.
Deardorff, A. (2003), 'Time and Trade: The Role of Time
Button, K., Stough, R., Bragg, M. and Taylor, S. (2006),
in Determining the Structure and Effects of Interna-
Telecommunications, Transportation and Location,
tional Trade with an Application to Japan, in Stern,
Edward Elgar, Cheltenham.
R. M. (ed.), Analytical Studies in US-Japan Interna-
Caniels, M. C. J. (2000), Knowledge Spillovers and Eco- tional Economic Relations, Edward Elgar, Cheltenham.
nomic Growth: Regional Growth Differences across
Dijkstra, 1., Garcilazo, E., and McCann, P. (2012), 'The
Europe, Edward Elgar, Cheltenham.
Economic Performance of European Cities and City-
Caniels, M., and Verspagen, B. (2003), 'Spatial Distance in a Regions: Myths and Realities', European Planning
Technology Gap Model', in Fingleton, B. (ed.), European Studies, forthcoming 001:10.1080/09654313.
Regional Growth, Springer, Heidelberg. (2012).716245.
DiscHer, A.-C., and l-lead, K. (2007), 'lbe Puzzling Persis- Glaeser, E. L., and Kohhase, J. (2004), 'Cities, Regions and
tence of the Distance Effect on Bilateral Trade', Review the Decline of Transport Costs', Papers in Regional
ofEconomics and Statistics, 90.1, 37-48. Science, 83.1, 197-228.
Doloreux, D., and Shearmur, R. (2012), 'Collaboration, Glaeser, E. L. and Shapiro, J. M. (2003), 'Urban Growth in
Information and the Geography of Innovation in the 1990s: Is City Living Back?', Journal ofRegional
Knowledge Intensive Business Services', Journal of Science, 43.1, 139-65.
Economic Geography, 12.1,79-105.
Glaeser, E. L., Scheinkman, J. A., and Shleifer, A. (1995),
Duranton, G., and Storper, M. (2008), 'Rising Trade Costs? 'Economic Growth in a Cross Section of Cities', Journal
Agglomeration and Trade with Endogenous Transac- ofMonetary Economics, 36, 117-43.
tion Costs', Canadian Journal ofEconomics, 41.1,
Glaeser, E. L., Kolko, J., and Saiz, A. (2001), 'Consumer
291-319.
City: Journal ofEconomic Geography, 1.1,27-50.
Faggian, A., and McCann, P. (2006), 'Human Capital
Gottlieb, J. D., and Glaeser, E. L. (2006), 'Urban Resur-
Flows and Regional Knowledge Assets: A Simultane-
gence and the Consumer City', Urban Studies, 43.8,
ous Equation Approach', Oxford Economic Papers, 58.3,
1275-99.
475-500.
Greunz, L. (2003), 'The Technology Gap and European
Faggian, A., and McCann, P. (2009), 'tIuman Capital,
Regional Growth Dynamics', in Fingleton, B. (ed.),
Graduate Migration and Innovation in British Regions',
European Regional Growth, Springer, lIeideiberg.
Cambridge Journal ofEconomics, 33.2, 317-33.
Grubesic. T. H., Matisziw, 1'. C., and Zook, M. A. (2008),
Fingleton, B. (ed.) (2003a), European Regional Growth,
'Global Airline Networks and Nodal Regions', Geo-
Springer, Heidelberg.
Journal, 71, 53-66.
Fingleton, B. (2003b), 'Increasing Returns: Evidence from
Grubesic,1'. H., Matisziw, 1'. C., and Zook, M. A. (2009),
Local Wage Rates in Great Britain', Oxford Economic
'Spatio-Temporal Fluctuations in the Global Airport
Papers, 55, 716-39.
Hierarchies', Journal of Transport Geography, 17.4,
Florida, R. (2002a), The Rise ofthe Creative Class, Basic 264-75.
Books, New York.
Haynes, K., Lall, S., Stough, R., and Yilmaz, S. (2006),
Florida, R. (2002b), The Flight ofthe Creative Class: 'The 'Network Usage Patterns and the Substitution and
New Competition for Global Talent, Harper Business, Complementarity Effects between Telecommunica-
New York. tions and Transportation: A Demand-Side Approach',
Frankel, J. A., and Romer, D. (1999), 'Does Trade Cause in Kobayashi, K., Lakshmanan, 1'. R., and Anderson,
Growth?', American Economic Review, 89.3, 379-99. W P. (eds), Structural Change in Transportation and
Communications in the Knowledge Society, Edward
Friedman,1'. L. (2007), The World is Flat: A BriefHistory of
Elgar, Cheltenham.
the Twenty-First Century Ord edn), Picador, New York.
Higgins, M. J., Levy, D., and Young, A. 1'. (2006), 'Growth
Fujita, M., Krugman, P., and Venables, A. J. (1999), The
and Convergence across the United States: Evidence
Spatial Economy: Cities Regions and International
from County-Level Data, Review ofEconomics and
Trade, MIT Press, Cambridge, MA.
Statistics, 88.4, 671-81.
Garcilazo, E., and Oliveira Martins, J. (2010), 'The Contri-
Hillberry, R., and Hummels, D. (2003), 'Intra-National
butions of Regions to Aggregate Growth', Paper Pre-
Home Bias: Some Explanations', Review ofEconomics
sented at the ERSA Conference, Jonkoping, Sweden.
and Statistics, 85.4, 1089-92.
Gaspar, J., and Glaeser, E. L. (1998), 'Information Technol-
Hillberry, R., and Hummels, D. (2008), 'Trade Responses
ogy and the Future of Cities', Journal of Urban Econom-
to Geographic Frictions: A Decomposition Using
ics, 43, 136-56.
Micro-Data, European Economics Review, 52, 527-50.
Glaeser, E. L. (2005), 'Urban Colossus: Why is New York
HMT-DTI (2001), Productivity in the UK: The Regional
Americas Largest City?', Economic Policy Review:
Dimension, HM Treasury and Department of Trade
Urban Dynamics in New York, Federal Reserve Bank of
and Industry, London.
New York, New York.
HMT-DTI (2003), Productivity in the UK: The Local
Glaeser, E. L. (2011), Triumph ofthe City: How Our
Dimension, HM Treasury and Department of Trade
Greatest Invention Makes Us Richer, Smarter, Greener,
and Industry, London.
flealthier, and Happier, Penguin Press, New York.
Hummels, D. (1999), 'Toward a Geography of Trade
Glaeser, E. L., and Gottlieb, J. D. (2009), (The Wealth of
Costs', mimeo, University of Chicago, Chicago, IL.
Cities: Agglomeration Economies and Spatial Equilib-
rium in the United States: Journal ofEconomic Litera- Hummels, D. (2001), 'Time as a Trade Barrier', mimeo,
ture, 47.4, 983-1028. Purdue University, Indiana.
Iammarino, S., and McCann, P. (2006), 'The Structure and McCann, P. (2007), 'Sketching out a Model of Innovation,
Evolution of Industrial Clusters: Transactions, Tech- Face- to- Face Interaction and Economic Geography',
nology and Knowledge Spillovers: Research Policy, 35, Spatial Economic Analysis, 2.2, 117-34.
1018-36.
McCann, P. (2008), 'Globalization and Economic Geogra-
Ioannides, Y., Overman, II. G., Rossi-Hansberg, E., and phy: The World is Curved, Not Flat: Cambridge Journal
Schmidheiny, K. (2008), 'lbe Effect of Information and ofRegions, Economy and Society, 1.3,351-70.
Communication Technologies on Urban Structure',
McCann, P. (2011), 'The Role of Industrial Clustering and
Economic Policy 54, 201-42.
Increasing Returns to Scale in Economic Development
Johansson, B., and Loof, H. (2009), 'How are Non-MNEs and Urban Growth: in Brooks, N., Donaghy, K., and
Affected by Globalization?', CESIS Working Paper, Knaap, G. (eds), 1he Oxford Handbook of Urban
Jonkoping University, Jonkoping and Royal Institute of Economics and Planning, Oxford University Press,
Technology, Stockholm, Sweden. Oxford.
Kriitke, S. (2011), The Creative Capital ofCities: Interactive McCann, P., and Fingleton, B. (1996), 'The Regional
Knowledge Creation and the Urbanization Economies of Agglomeration Impact of Just-In-Time Input Linkages:
Innovation, Wiley Blackwell, London. Evidence from the Scottish Electronics Industry',
Krugman, P. (1991), Geography and Trade, MIT Press, Scottish Journal ofPolitical Economy, 43.5, 493-518.
Cambridge, MA. McCann, P., Poot, J., and Sanderson, L. (2010), 'Migration,
Krugman, P., and Venables, A. J. (1995), 'Globalization Relationship Capital and International Travel: Theory
and the Inequality of Nations', Quarterly Journal of and Evidence', Journal ofEconomic Geography, 10.3,
Economics, 110.4, 857-80. 361-87.
Leamer, E. (2007), 'A Flat World, a Level Playing Field, a McCann, P. and Acs, Z. J. (2011), 'Globalisation: Coun-
Small World After All, or None of the Above? A Review tries, Cities and Multinationals', Regional Studies, 45.1,
of Thomas L. Friedman's The World is Flat', Journal of 17-32.
Economic Literature, 83-126. MGI (2009), Preparingfor China's Urban Billion, McKinsey
Leinbach, T. R., and Capineri, C. (2007), Globalized Freight Global Institute.
Transport: Intermodality, E-Commerce, Logistics and MGI (2010), India's Urban Awakening: Building Inclusive
Sustainability, Edward Elgar, Cheltenham. Cities, Sustaining Economic Growth, McKinsey Global
Levinson, M. (2006), 1he Box: How the Shipping Container Institute.
Made the World Smaller and the World Economy Bigger, MGI (2011a), Urban World: Mapping the Economic Power
Princeton University Press, Princeton, NJ. ofCities, McKinsey Global Institute.
Limtanakool, N., Schwanen, '1'., and Dijst, M. (2007), 'Rank- MGI (2011b), Building Globally Competitive Cities: The
ing Functional Urban Regions: A Comparison ofInterac- Key to Latin American Growth, McKinsey Global
tion and Node Attribution Data: Cities, 24.5, 26-42. Institute.
Long Finance (2012), Global Financial Centres Index II, MGI (2012), Urban America: US Cities in the Global
London. Economy, McKinsey Global Institute.
Lucas, R. E. (2000), 'Some Macroeconomics for the Nachum, L., and Zaheer, S. (2005), 'The Persistence of
Twenty-First Century: Journal ofEconomic Perspectives, Distance? The Impact of Technology on MNE Motiva-
14.1, 159-78. tions for Foreign Investment: Strategic Management
Maddison, A. (2006), The World Economy. Volume 1: A Journal, 26, 747-67.
Millennial Perspective Volume 2: Historical Statistics, Ni, P., and Kresl, P. K. (2010), The Global Urban Competi-
Organisation for Economic Cooperation and Develop- tiveness Report 2010, Edward Elgar, Cheltenham.
ment, Paris.
OECD (2006), Competitive Cities in the Global Economy,
Marshall, A. (1890), Principles ofEconomics (1st edn), Organisation for Economic Cooperation and Develop-
Macmillan, London. ment, Paris.
Mastercard (2008), Worldwide Centers ofCommerce Index, OECD (2007a), International Investment Perspectives:
www.mastercardworldwide.com Freedom ofInvestment in a Changing World, Organisa-
McCann, P. (1993), 'The Logistics-Costs Location-Produc- tion for Economic Cooperation and Development, Paris.
tion Problem Journal ofRegional Science, 33.4, 503-16. OECD (2007b), Regions at a Glance 2007, Organisation for
McCann, P. (1998), The Economics ofIndustrial Location: Economic Cooperation and Development, Paris.
A Logistics Costs Approach, Springer, Heidelberg. OECD (2009a), How Regions Growth: Trends and Analysis,
McCann, P. (2005), 'Transport Costs and New Economic Organisation for Economic Cooperation and Develop-
Geography: Journal ofEconomic Geography, 5.3,305-18. ment, Paris.
OECD (2009b» Regions Matter: Economic Recovery) Shapiro, J. M., (2006» (Smart Cities: Quality of Life)
Innovation and Sustainable Growth) Organisation for Productivity) and the Growth Effects of Human
Economic Cooperation and Development) Paris. Capital') Review ofEconomics and Statistics, 88.2)
324-35.
OECD (2009c» Regions at a Glance 2009) Organisation for
Economic Cooperation and Development) Paris. Simmie, J. (2004» 'Innovation and Clustering in the
Globalized International Economy', Urban Studies,
OECD (2011a» Regions and Innovation Policy) Organisation
41.5-6) 1095-112.
for Economic Cooperation and Development, Paris.
Storper) M., and Venables) A .J. (2004» 'Buzz: Face-to-Face
OECD (2011b» Regions at a Glance 2011, Organisation for
Contact and the Urban Economy') Journal ofEconomic
Economic Cooperation and Development) Paris.
Geography) 4) 351-70.
OECD (2012» Redefining Urban: Functional Urban Areas
Taylor, P. J. (2004), World City Network: A Global Urban
in OECD Countries) Organisation for Economic
Analysis) Routledge) London.
Cooperation and Development) Paris.
Taylor) P. J.) Evans) D. M., and Pain) K. (2008» 'Applica-
Openshaw) S. and Taylor) P. J. (1979» A million or so correla-
tion of the Interlocking Network Model to Mega-
tion coefficients: three experiments on the modifiable
City Regions: Measuring Polycentricity Within and
areal unit problem in Wrigley) N. (ed.» Statistical
Beyond City-Regions', Regional Studies) 42.8)
Applications in the Spatial Sciences) Pion) London.
1079-93.
Ortega-Argiles) R. (2012» "The Transatlantic Productivity
UNDP (2009), Human Development Report 2009: Over-
Gap: A Survey of the Main Causes» Journal ofEconomic
coming Barriers: Human Mobility and Development)
Surveys) 26.3) 395-419.
United Nations Development Program, Palgrave
Partridge) M. D.) Rickman) D. S.) Ali) K.) and Olfert) M. R. Macmillan, Basingstoke.
(2008» (Employment Growth in the An1erican Urban
UN-HABITAT (2008» State ofthe World's Cities Report
Hierarchy: Long Live Distance: B.E. Journal ofMacro-
2008109) United Nations-HABITAT Programme)
economics) 8) 1-36.
Nairobi.
Pavitt) K. (1984» (Sectoral Patterns of Technical Change:
Timmer, M. P, Inklaar, R.) O'Mahony, M.) and Van Ark) B.
Towards a Taxonomy and a Theory', Research Policy,
(2010» Economic Growth in Europe) Cambridge
13) 343-73.
University Press) Cambridge.
Piore, M. J.) and Sabel) C. F. (1984» The Second Industrial
Van Oort) F. G. (2004» Urban Growth and Innovation:
Divide: Possibilitiesfor Prosperity, Basic Books) New York.
Spatially Bounded Externalities in the Netherlands)
Porter) M. E. (1990» The Competitive Advantage of Ashgate) Aldershot.
Nations) Free Press) New York.
Venables, A. J. (2006» (Shifts in Economic Geography and
PWC, (2009» (Which are the Largest City Economies in their Causes» Federal Reserve Bank ofKansas City
the World and Might this Change by 2025?') UK Economic Review) 91.4, 61-85.
Economic Outlook November (2009), Price Waterhouse
Ware, B. (1995), (Telecommunications and the Clustering
Coopers) London.
Geographies of Knowledge Transmission in the Late
Rodriguez-Pose, A. (1998), The Dynamics ofRegional 20th Century') Urban Studies) 32.2) 361-78.
Growth in Europe: Social and Political Factors, Claren-
Weber) A. (1909), Ober den Standort der Industrien) trans.
don Press) Oxford.
by Friedrich, C.J. (1929» Alfred Weber's Theory ofthe
Rietveld) p) and Vickerman) R. (2004» (Transport in Location ofIndustries) University of Chicago Press)
Regional Science: lhe "Death of Distance" is Prema- Chicago) IL.
ture') Papers in Regional Science, 83) 229-48.
Westlund) H. (2006), Social Capital in the Knowledge
Rosenthal, S. S.) and Strange) W C. (2004), 'Evidence on Economy: Iheory and Empirics) Springer) Heidelberg.
the Nature and Sources of Agglomeration Economics»
Wickham) J.) and Vecchi) A. (2008» 'Local Firms and
in Henderson) v.) and Thisse, J.-F. (eds» Handbook of
Global Reach: Business Air Travel and the Irish
Urban and Regional Economics Vo.14) North-Holland,
Software Cluster: European Planning Studies) 16.5)
Amsterdam.
693-710.
Sassen) S. (1994» Cities in a World Economy, Pine Forge
Press) Thousand Oaks) CA. World Bank (2008a» World Trade Indicators 2008: Bench-
marking Policy and Performance) World Bank) Wash-
Sassen) S. (ed.) (2002), Global Networks: Linked Cities, ington) DC.
Routledge, London.
World Bank (2009» World Development Report 2009:
Schonberger) R. J. (1996» World Class Manufacturing: Ihe Reshaping Economic Geography) World Bank) Wash-
Next Decade) Free Press, New York. ington) DC.
IUfH.JOGRAPHY
I-lague, C., Hague, E., and Breitbach, C. (2011), Moretti, E. (2012). Ihe New Geography ofJobs, IIoughton
Regional and Local Economic Development, Mifflin Harcourt, New York.
Palgrave, Basingstoke.
Myrdal, G. (1957), Economic Theory and Underdeveloped
Hart, G. (2010), 'Redrawing the Map of the World? Regions, Hutchinson, London.
Reflections on the World Development Report 2009',
OECD (2006), Competitive Cities in the Global Economy,
Economic Geography, 86, 341-50.
Organisation for Economic Cooperation and Develop-
Hirschmann, A. O. (1958), The Strategy ofEconomic ment, Paris.
Development, Yale University Press, New Haven, CT. OECD (2009a). How Regions Grow: Trends and Analysis,
tloughwout, A. (1998), 'Aggregate Production Functions, Organisation for Economic Growth and Development,
Interregional Equilibrium, and the Measurement of Paris.
Infrastructure Productivity; Journal of Urban Econom- OECD (2009b), Regions Matter: Economic Recovery,
ics, 44.2, 216-27. Innovation and Sustainable Growth, Organisation for
lioughwout, A. (2002), 'Public Infrastructure Investments, Economic Growth and Developlnent, Paris.
Productivity and Welfare in Fixed Geographic Areas', OECD (2009c). Regions at a Glance 2009, Organisation for
Journal ofPublic Economics. 83, 405-25. Economic Cooperation and Development, Paris.
Janssens, M., Pinelli, D., Reymen, D. C., and Wallman, S. OECD (2010a), Regional Development Policies in OECD
(2009), Sustainable Cities: Diversity, Economic Growth Countries, Organisation for Economic Cooperation
and Social Cohesion, Edward Elgar, Cheltenham. and Development, Paris.
Kim, S.-J. (2011), 'Non-Market Effects on Agglomeration OECD (2010b), Organising Local Economic Development:
and their Public Policy Responses', in OECD Regional The Role ofDevelopment Agencies and Companies,
Outlook 2011, Organisation for Economic Cooperation Organisation for Economic Cooperation and Develop-
and Development, Paris. ment, Paris.
Krugman, P. (1991), Geography and Trade, MIT Press, OECD (2011a), Regions at a Glance 2011, Organisation for
Cambridge, MA. Economic Cooperation and Development, Paris.
Krugman, P. (2009), The Return ofDepression Economics OECD (2011b), Regional Outlook 2011, Organisation for
and the Crisis of2008, Norton, New York. Economic Cooperation and Development, Paris.
Krugman, P., and Venables, A. J. (1990), 'Integration and Ovennan, H. G. (2002), 'Neighbourhood Effects in Large
Competitiveness of Peripheral Industry', in C. J., and and Small Neighbourhoods: Urban Studies, 39.1,117-30.
J. B. (eds), Unity and Diversity in the European Econ-
Partridge, M. D., and Rickman, D. S. (2006), The Geogra-
omy: The Community's Southern Frontier, Cambridge
phy ofAmerican Poverty: Is there a Need for Place-Based
University Press, Cambridge.
Policies?, Upjohn Institute, Kalamazoo, MI.
Krugman, P., and Venables, A. J. (1995), 'Globalization
Pattachini, E., and Zenou, Y. (2011), 'Intergenerational
and the Inequality of Nations', Quarterly Journal of
Education Transmission: Neighborhood Quality and/
Economics, 110.4,857-80.
or Parents' Involvement?', Journal ofRegional Science,
Laurila, H. (2004), 'Urban Governance, Competition and 51,987-1013.
Welfare', Urban Studies, 41.3, 683-96.
Pearce, D. W and Nash, C. A. (1981), The Social Appraisal
Laurila, H. (2011), 'Optimisation of City Size', Urban ofProjects: A Text in Cost-Benefit Analysis, Macmillan,
Studies, 48.4, 737-47. Basingstoke.
Layard, R., and Glaister, S. (1994), Cost-Benefit Analysis Peck, J, and Sheppard, E. (2010), 'Worlds Apart? Engaging
(2nd edn) Cambridge University Press, Cambridge. with the World Development Report 2009: Reshaping
Lucas, R. E. (1988), 'On the Mechanics of Economic Devel- Economic Geography', Economic Geography, 86, 331-40.
opment: Journal ofMonetary Economics, 22.1, 3-42. Pike, A., Rodriguez-Pose, A., and Tomaney, J. (2006),
Local and Regional Development, Routledge, London.
McCann, P. (2009), 'Economic Geography, Globalisation
and New Zealand's Productivity Paradox', New Zealand Pike, A., Rodriguez-Pose, A., and Tomaney, J. (eds) (2010),
Economic Papers, 43.3, 279-314. llandbook ofLocal and Regional Development,
Routledge, Abingdon.
McCann, P., and Rodriguez-Pose, A. (2011), 'Why and
When Development Policy Should be Place-Based', in Putnam. R. (1993), Making Democracy Work: Civic
OECD Regional Outlook 2011, Organisation for Traditions in Modern Italy, Princeton University Press,
Economic Cooperation and Development, Paris. Princeton, NJ.
McCann, P., and Acs, Z. J. (2011). 'Globalization: Coun- Quigley, J. M., and Raphael, S. (2005), 'Regulation and
tries, Cities and Multinationals', Regional Studies, the High Cost of Housing in California', American
45.1,17-32. Economic Review, 95.2,323-8.
Rodrik, D. (2006), (Goodbye Washington Consensus, lIello Swales, J. K. (2009), 'A Cost-Benefit Approach to Regional
Washington Confusion? A Review of the World Bank's Policy', in Farshchi, M., Janne, O. E. M., and McCann,
Economic Growth in the 1990s: Learning from a Decade P. (eds), Technological Change and Mature Industrial
of Reform: Journal ofEconomic Literature, 44.4, 973-87. Regions, Edward Elgar Cheltenham, 237- 258.
Rodrik, D., Subramanian, A., and Trebbi, F. (2004), Swan, T. W. (1956), (Economic Growth and Capital
(Institutions Rule: the Primacy of Institutions over Accumulation, Economic Record, 32, 334-61.
Geography and Integration in Economic Develop-
Swinburn, G., Goga, S., and Murphy, E (2006), Local
ment',Journal ofEconomic Growth, 9.2,131-65.
Economic Development-A Primer: Developing and
Romer, P. (1986), (Increasing Returns and Long-Run Implementing Local Economic Development Strategies
Growth: Journal ofPolitical Economy, 94,1002-37. and Action Plans, lhe World Bank and Bertelsman
Stiftung, Washington, DC.
Rostow, W W (1959), ('Ihe Stages of Economic Growth',
Economic History Review, 12.1, 1-17. lbissen, M., and Van Oort, E (2010), (European Place-
Based Development Policy and Sustainable Economic
Ruth, M. (ed.) (2006), Smart Growth and Climate Change:
Agglomeration, TESG Journal ofEconomic and Social
Regional Development, Infrastructure and Adaptation,
Geography, 101.4,473-80.
Edward Elgar, Cheltenham.
Vanclay, E, Higgins, M., and Blackshaw, A. (2008), Making
Ruth, M., Donaghy, K., and Kirshen, P. (eds) (2006),
Sense ofPlace: Exploring Concepts and Expressions of
Regional Climate Change and Variability: Impacts and
Place Ihrough Different Senses and Lenses, National
Responses, Edward Elgar, Cheltenham.
Museum of Australia Press, Canberra.
Sachs, J. D. (2003), (Institutions Dont Rule: Direct Effects
Van Oort, F. G. (2004), Urban Growth and Innovation:
of Geography on Per Capita Income', WP 9490, NBER,
Spatially Bounded Externalities in Ihe Netherlands,
Cambridge, MA.
Ashgate, Aldershot.
Sachs, J. D. (2011), The Price ofCivilisation: Reawakening
Venables, A. J. (2007), (Evaluating Urban Transport
American Virtue and Prosperity, Random IIouse,
Improvements: Cost-Benefit Analysis in the Presence
New York.
of Agglomeration and Income Taxation, Journal of
Sachs, J. D. (2012), (Government, Geography and Growth: Transport Economics and Policy, 41.2, 173-88.
The True Drivers of Economic Development', Foreign
Venables, A. J. (2010), (Economic Geography and African
Affairs, 91.5,142-50.
Development', Papers in Regional Science, 89.3, 469-83.
Sapir, A., Aghion, P., Bertola, G., Hellwig, M., Pisani-Ferry,
Vickerman, R. W. (ed.) (1991), Infrastructure and Regional
J., Rosati, D., Viiials, J., and Wallace, H. (2004), An
Development, Pion, London.
Agenda for a Growing Europe: The Sapir Report, Oxford
University Press, Oxford. Williamson, J. G. (1965), (Regional Inequality and the Process
of National Development: A Description of the Patterns:
Sassone, P. G., and Schaffer, W. A. (1978), Cost Benefit
Economic Growth and Cultural Change, 13, 1-84.
Analysis: A Handbook, Academic Press, New York.
Winnick, L. (1966), (Place Prosperity vs People Prosperity:
Seltzer, E., and Carbonell, A. (eds) (2011), Regional
Welfare Considerations in the GeographiC Redistribu-
Planning in America: Practice and Prospect, Lincoln
tion of Economic Activity: in Essays in Urban Land
Institute of Land Policy, Cambridge, MA.
Economics in Honor ofthe Sixty- Fifth Birthday ofLeo
Solow, R. M. (1956), 'A Contribution to the lheory of Grebler, Real Estate Research Program, University of
Economic Growth: Quarterly Journal ofEconomics, California Press, Los Angeles, CA.
70,65-94.
World Bank (2003), World Development Report 2003:
Spence, M., and Liepziger, D. (eds) (2010), Globalization Transforming Institutions, Growth and Quality ofLife,
and Growth: Implications for a Post-Crisis World, Washington, DC.
Corpmission on Growth and Development, Interna-
World Bank (2009), World Development Report 2009:
tional bank for Reconstruction and Development,
Reshaping Economic Geography, World Bank, Wash-
World Bank, Washington, DC.
ington, DC.
Stiglitz, J. (2002), Globalization and its Discontents,
World Bank (2010), (Innovation Policy: A Guide for
Penguin, London.
Developing Countries', World Bank, Washington, DC.
Stiglitz, J. E., Sen, A., and Fitoussi, J.-P. (2009), Commission of
Xu, C. (2011), (The Fundamental Institutions of China's
the Measurement ofEconomic Performance and Sodal
Reforms and Development', Journal ofEconomic
Progress, http://www.stiglitz-sen-fitoussi.fr/en/index.htm
Literature, 49.4,1076-151.
Swales, J. K. (1997), 'A Cost-Benefit Approach to the
Zoellick, R. B. (2012), (Why We Still Need the World Bank:
Evaluation of Regional Selective Assistance: Fiscal
Studies, 18, 73-85. Looking Beyond Aid', Foreign Affairs, 91.2, 66-78.
Index