Está en la página 1de 32

Urban Studies, Vol.

36, Nos 5 6, 811 842, 1999

Geographical Econom ics and Urban


Com petitiveness: A Critique

M artin Boddy

[Paper received in nal form, October 1998]

1. Introduction

Since the late 1980s, Paul Krugm an has rap- to national prosperity and the threat to `de-
idly established him self as amongst the most velope d nations posed by cheap labour in
prom inent of econom ists, particularly for his developing countries. He has also addressed
work on `new trade theory and com petitive- questions around the clustering and agglom -
ness. Samuelson, adm ittedly in the preface to eration of econom ic activity at the sub-
Krugm an s 1994 book The Age of Dimin- national scale, linking com petitiveness and
ished Expectations, refers to him as the trade to developm ent at the regional and
rising star of this century and the next . The urban scales (Krugm an, 1991a, 1995). In-
proli c and often controv ersial Krugm an has creasingly, as well, Krugm an s work has
clearly made a major im pact within his own touche d on clustering and the localisation of
discipline. This im pact has been aided by the econom ic activity at the level of the urban
com bination of the polem ical, accessible agglom eration or city scale. There are clearly
and, he would argue, well founded manner also, questions as to the relationship between
of his attacks on som e of his fellow cities and the clustering of activity at the
econom ists and more particularly on policy regiona l scale. He has addressed these issues
advisers (Krugm an, 1997), together with his from a form al econom ic perspective deriving
contributions to form al econom ics. Increas- models of clustering and agglom eration in
ingly, as well, he has attracted interest within highly abstract terms. He has also referred
econom ic geography, regional studies and more generally, and in his terms discursively,
urban studies. In part this is because he has to processes of urbanisation and the growth
explicitly set out to incorporate geography of cities and regions his 1997 book attack-
and space into form al econom ic models of ing P op Internation alism concludes with a
trade and com petitiveness revisiting in the prose essay on `The localisation of the world
process a range of standard `models of econ- econom y com paring the historic develop-
om ic geography and spatial science. In par- ment of Chicago and Los Angeles.
ticular, he has challenged the conventional As within econom ics, his ideas have at-
concept of trade as re ecting prim arily com - tracted both appreciation but also concern
parative advantage, emphasising instead the from a geographical perspective (Knox and
role of increasing returns to scale. He has Agnew, 1994; Martin and Sunley, 1996;
vigoro usly challenged the popula r notion of Dym ski, 1996). Concern has derived in par-
com petitiveness in global markets as the key ticular from his insistence on form al econ-

Martin B oddy is in the School for Policy Studies, University of B ristol, 8 Priory R oad, Bristol, BS8 1TZ, UK. Fax: 1 44(0)117954-6756.
E -mail: martin.boddy@ bristol.ac.uk. This article draws on work supported by the Economic and Social R esearch Council through
the `Cities, Competitiveness and Cohesion P rogramm e, grant num ber L130 30 100 116.

0042-0980/99/05/0811-32 1999 T he E ditors of Urban Studies


812 MA RT IN BODD Y

om ic m odelling as both value free and the external econom ies. This viola tes a key as-
only source of clear understanding (Krug- sum ption of econom ic models which assume
man, 1991a, p. 80) and his whole sale dis- perfect com petition and equilibr ium . In the
missal of the bulk of urban and regional absence of perfect com petition , it is im poss-
studies and, in particular, approa ches ible using what had been conventional, main-
ground ed in `social econom ics . Dym ski stream econom ic models to derive a unique
(1996, p. 439), for example, has suggested equilibrium outcom e. M arket structure de-
that Krugm an is engaged in a form of econ- pends not simply on the interaction of supply
om ic neo-colon ialism, with his work having and dem and but on the decisions, including
becom e: location decisions of producers.
As the mainstream orthodo xy within eco-
less an invitation to geographers to enjoy
nom ics became increasingly focused on for-
mutual gains from trade than an incursion
mal, mathem atically coherent, equilibrium
into underdeveloped theoretical terrain.
models, argues Krugm an, this led to the ne-
An earlier review of Geograph y and Trade glect of both developm ent econom ics and of
(Krugm an 1991a) had referred to Krugm an s spatial aspects of econom ics more generally.
perspective on econom ic geography as This, he argues, explains the lack of attention
`highly simplistic and `patronisin g (John- within mainstream econom ics to urban and
ston, 1992). Both Dym ski, and Martin and regiona l econom ics. Developm ent econom -
Sunley (1996) have provide d a critical ap- ics, for example, declined because:
praisal of Krugm an s `geographical econom -
ics . Martin and Sunley in their more the leading development econom ists failed
extended critique of Krugm an s work, point to turn their intuitiv e insights in to clear-
to the parallels, but also the distance be- cut models that could serve as the core of
tw een, Krugm an s work on agglom eration an enduring discipline (Krugm an, 1995,
and much of the recent work on industrial p. 24).
districts, networks and clustering and the The one basic reason for this, moreover, was:
links betw een innovation and com petitive- the dif culty of reconciling econom ies of
ness ground ed in ideas of post-Fo rdism , scale with a com petitive market structure
social econom ics and institut ional econom ics (Krugm an, 1995, p. 25). M yrdal s work on
of which Krugm an is particularly scathing. circular and cum ulative causation and
Here, therefore, the intention is to examine Hirschm an s work on linkages, stim ulating
Krugm an s work in particular as it relates to and in uential in the 1950s and early 1960s,
cities, urban grow th and com petitiveness. had, by the 1970s, com e to seem not so
Initially, however, it will be useful to look much wrong as meaningless (Krugm an,
brie y at Krugm an s overall approach to 1995, p. 27). L acking the tools to address
econom ic modelling and the process of ab- the increasing returns, external econom ies
straction, as this goes a long way to explain- and im perfect com petition which were
ing the distance between his work and much fundam ental to these areas, the increasingly
recent work within geography and urban and form alised mathem atical modelling of main-
regional studies. stream econom ics simply ignore d them . In
the past few years, argues Krugm an:
2. The New Geograph ical Econom ics
It has becom e apparent that during the
Krugm an s overall approach turns on a cri- 1940s and 1950s a core of ideas emerged
tique of form al econom ic models based on regarding external econom ies, strategic
perfect com petition. Both international trade com plem entarity, and econom ic develop-
and the location of econom ic activity in ment that remains intellectually valid and
space more generally are, in practice, charac- may continue to have practical applica-
terised by increasing returns to scale and tions. This set of ideas anticipated in a
GE OG RAPH ICA L ECO NO MICS 813

num ber of ways the cutting edge of mod- cause they knew they had no way to model
ern trade and growth theory. But these that aspect (Krugm an, 1995, pp. 3637). As
ideas have had to wait to be discov- with developm ent econom ics
ered. Only recently have changes in
econom ics made it possible to reconsider there was a set of core ideas that make
what the developm ent theorists said, and considerable sense in light of recent econ-
to regain the valuable ideas that have been om ic analysis, but that were unacceptable
lost (Krugm an, 1995, p. 7). to mainstream econom ics because they
could not at that tim e be modelled (Krug-
It was a parallel story, according to Krug-
man, 1995, p. 37).
man, in terms of the failure of econom ics to
take account of space and to provide accept-
Krugm an goes on to review a range of `mod-
able explanations for spatial agglom eration
els or approaches from within traditional
or the location of econom ic activity in space.
econom ic geography and location theory.
The problem as Krugm an sees it is, put
Each of these approaches, in one way or
simply,
anothe r, de es form al m odelling due to the
to account for the highly uneven spatial problem s which space poses in terms of in-
distribu tion of econom ic activity we actu- creasing returns and market structure. Hence
ally see, in which most people live on the the failure of mainstream econom ics to em-
small, urbanised fraction of the land and in brace or develop locational models. W ebe-
which most urban areas them selves are rian location theory, for example, aimed to
highly specialised (Krugm an, 1995, p. 35). explain the location of a single factory in
terms of minim ising transport costs to sev-
In contrast, in a world consisting of a hom o- eral suppliers and several markets. In econ-
geneous, featureless plain, with no econom - om ic terms, how ever, it is unclear how actors
ies of scale but subject to transport costs, might make decisions and how the decisions
econom ic activity would itself be evenly of actors might interact to produce a given
spread across the plain in order to avoid any outcom e as required by an econom ic model
transport costs. Uneven distribu tion of natu- as such. It simply reduces the problem to one
ral resources could account, according to of minim ising transport costs given certain
Krugm an, for only a fraction of the observed assumptions including, for example, a sin-
unevenness of econom ic activity across gle production site. The example is less rel-
space he points to mega-cities such as Los evant, in any case, in terms of agglom eration
Angeles or Sa o Paulo as an extrem e case. and the growth of cities.
The answer, as with developm ent econom ics, Central place theory argued that the trade-
relates to market structure in the face of off between econom ies of scale and transport
increasing returns, the need to deal with scale costs leads producers to cluster together into
econom ies and oligopo listic rms: hexagonal market areas (L osch) or a hier-
archy of nested market areas (Christaller).
in order to talk even halfway sensibly
This introdu ces scale econom ies but, as
about econom ic geography it is necessary
Krugm an argues, from an econom ic perspec-
to invoke the role of increasing returns in
tive, it is not clear who is making the deci-
som e form . And that means that even to
sions and what form the market structure
get started on the subject one m ust get into
takes it represents, as Krugm an puts it, a
the issues that did so much to make
schematic way to organise your though ts
high developm ent theory unacceptable to
rather than an econom ic model in which the
mainstream econom ics (Krugm an, 1995,
observed structure can be explained by
p. 36).
deeper (meaning market-based) causes. Prob-
Again, how ever, econom ists avoide d look- lems of modelling im perfect markets given
ing at the spatial aspect of econom ies be- the trade-off between econom ies of scale and
814 MA RT IN BODD Y

transport costs, moreover, prevented devel- Pred (1966) is the im plicit assumption that
opm ent of a form al model (Krugm an, 1995, there are substantial econom ies of scale at
p. 41). the level of the plant. In the absence of
Sim ilar argum ents apply to social physics such scale econom ies, producers would
and the idea of market potential in space have no incentive to concentrate their ac-
which offer plausible stories, som e striking tivities at all: they would simply supply
empirical regularities, and a useful basis for consum ers from many local plants, vitiat-
empirical work (Krugm an, 1995, p. 45). ing Harris s argum ent. And any expansion
Again, im plicit is the notion of increasing of a regional market would not predictably
returns and monopoli stic com petition allow- lead to any increase in the range of goods
ing rms to reap the rewards of locations produced within that region, vitiating
with higher `market potential . M odels of Pred s. Increasing returns, in other words,
cum ulative causation, on the other hand, run were central to the story (Krugm an, 1998a,
into the proble m of circularity and multiple p. 163).
equilibria: rms locate where markets are
More directly relevant to explaining agglom -
large, but markets are large where rms
eration and spatial clustering is, as Krugm an
locate (Krugm an, 1995, p. 46), leading to
observes, the role of local external econom -
the possibility of self-reinforcing grow th (or
ies referring back to Marshall s major contri-
decline). He reiterates this in his more recent
butions to the eld. Essentially, clustering of
paper, referring to Harris (1954) work which
producers in a particular location yields ad-
had dem onstrated that the heavily industri-
vantages and these advantages, in turn, ex-
alised regions of the US were in general
plain clustering. This simple notion of
those with exceptionally high market pote n-
clustering needs conside rable unpacking and
tial (Krugm an, 1998a, p. 162). This similarly
is explored in greater depth below. Krugm an
underlies Pred s `model of regional grow th
him self notes (1995, p. 51) the intrinsic im -
(Pred, 1966) based on replacing im ports of
portanc e of external econom ies as a real-
goods subject to scale econom ies with local
world issue . Krugm an s point in relation to
production. T his draws in workers from else-
mainstream econom ics is its surprising ne-
where, which expands the local market; this
glect of local externalities. Mainstream eco-
may reach a point where a further round of
nom ics simply ignored external econom ies
im port substitution becom es possible :
where market size was a factor and where
The large markets might make it pro table increasing returns to scale and im perfect
to produce goods that had previously been com petition were therefore an issue. This
im ported from other regions; this would would include the ability of larger local mar-
increase the multiplier on the region s ex- kets to suppor t ef cient-scale specialist sup-
port base, leading to a further expansion of pliers of intermediate inputs or the bene ts of
incom e, which would lead to still more larger, pooled labour markets. The main-
local production, and so on (Krugm an, stream allow ed for external econom ies but
1998a, p. 162). only `pure externalities or spillove rs such as
the exchange of inform ation which takes
This is, as Krugm an observes, intuitively place in local clusters of rms but which, it is
reasonable, yet econom ists failed to consider assumed, is not mediated by market pro-
it until around 1990. It remained more a set cesses; hence scale effects do not contravene
of boxes and arrows suggesting relationships the assumption of perfect com petition. Krug-
but defying construction of any form al econ- man goes on to suggest that
om ic model based on an understanding of
market structure. T he reason, according to if you were to ask a mainstream econom ist
Krugm an was that: at any tim e between, say, 1930 and the last
few years why cities exist or why som e
underly ing the work of Harris (1954) and industries are so concentrated in space, he
GE OG RAPH ICA L ECO NO MICS 815

or she would surely answer in terms of just E conom ists understood why econom ic ac-
such local externalities (Krugm an, 1995, tivity spreads out, not why it becom es
p. 51). concentrated and thus the central model
of spatial econom ics became one that
The sources of agglom eration were in effect
deals only with the way that com petition
`made safe for neo-classical econom ics by
for land drives econom ic activities away
assuming them to be pure, technolog ical
from a central market (Krugm an, 1995,
externalities. How ever:
p. 54).
This strategic evasion has been costly in
terms of both credibility and research- Sim ilar criticisms apply to the `new urban
ability the pure-externality assum ption econom ics which aimed to explain the loca-
puts these effects into a black box, where tion of econom ic activities around a central
nothing more can be said you have no core in the sense that the latter was as-
deeper structure to examine, no way to sum ed rather than derived from the model. It
relate agglom eration to m ore micro-level needed, as Krugm an puts it, an appeal to
features of the econom y (Krugm an, 1995, loosely speci ed agglom eration econom ies
p. 52). to com plete the model (Krugm an, 1995,
p. 58). They also shed little light on the
It is worth noting in passing that while this emergence of the polycentric city or `edge
may be true from a form al econom ic per- city growth readily observable in the US.
spective, this assumes that the objective is to This, according to Krugm an, is epitom ised
derive macro-level patterns from micro-level by the eclipse of Chicago, the archetypal
market behaviour. This is Krugm an s com - monocentric city focused around the down-
plaint. As he observes som ewhat disparag- town Loop, by Los Angeles with no single
ingly: centre and a dozen or more of ce districts
which com pete with each other.
Oh, you can try to measure them empiri-
Sim ilarly, regional science and Isard s
cally, and there has been som e im portant
(1956) attempts to develop an overall ap-
work along those lines. But you have
proach in Location and the Space-econom y
no deeper structure (Krugm an, 1995,
failed ow ing to the fact that one cannot view
p. 52).
location simply as anothe r substitutable fac-
It is this which lies at the heart of the gulf tor of production in a general equilibrium
betw een Krugm an s work on externalities model since one has to take account of
and much of the recent work in econom ic increasing returns and hence im perfect com -
geography, urban and regional studies which petition . Krugm an argues that the half
also addresses issues of clustering, agglom er- worked out models of regional science which
ation and externalities and has similar roots made little contribution to econom ic theory
in Marshall s work. This is an issue which were undeniably useful for a variety of
will be returned to below. practical purposes (Krugm an, 1995, p. 56)
Von Thu nen s model of land rent, on the even if they fall well short of an intellectu-
other hand, illustrates, as Krugm an observes, ally satisfying or coherent structure . In al-
many of the features of neo-classical market lowing, more charitably than in other of his
econom ics. Mainstream econom ics could writing, for the practical bene ts of a col-
therefore deal with von Thu nen s model. In lection of tools, som e crude, som e fairly
terms of understanding spatial agglom er- sophisticated (p. 56), Krugm an again illus-
ation, however, it is of no assistance in that it trates his de nition of the intellectually satis-
simply assumes a central urban market fying and coherent model.
around which it explains the distribu tion of According to Krugm an, therefore, the in-
econom ic activity under a range of assump- ability of conventional mainstream econom -
tions. As Krugm an puts it: ics to incorpo rate im perfect com petition
816 MA RT IN BODD Y

arising from situations in which increasing He, him self, set out, therefore, to develop
returns and the role of space were key factors such models both in relation to international
led to blindsp ots. trade and developm ent and as the basis for
understanding agglom eration and the cluster-
Development econom ics, or more
ing of econom ic activity in space. His per-
speci cally the set of ideas that I call `high
spective set out in his lecture on `Models and
developm ent theory had a huge initial
metaphors (Krugm an, 1995) was that the
in uence. But it then faded away, virtually
insistence on models is right, even when it
disappeared from econom ic discourse.
som etimes leads us unfairly to overlook good
Sim ilarly, ideas . Dealing with com plex systems in-
evitably, he argues, requires a process of
to this day the silence of standard econom - abstraction and simpli cation. He offers cri-
ics on such subjects as the location, size or teria for a good model: success in terms of
even existence of cities is startling (Krug- explaining or rationalising som e of what you
man, 1995, p. 67). see in the world in a way that you might not
have expected; predictive success particu-
This was exacerbated by the increasing em-
larly where this is rati ed by the marketabil-
phasis within the discipline of econom ics on
ity of the ideas; im proved insight into why
form al mathematical models as the core of
the vastly more com plex real system behaves
the discipline, models in which macro-
the way it does. Sim pli cation inevitably
behaviour and the resulting patterns could be
entails som e narrowing of vision, hence
derived from micro-m otives in a market con-
blindsp ots lim itations of one s framework
text. It was this which led to the sidelining of
and tools lim itations which, he argues
developm ent econom ics. It was this, as well,
can only be ended de nitively by making
which lay behind the lack of attention to
those tools good enough to transcend those
space and geography within mainstream eco-
lim itations (Krugm an, 1995, p. 73). None of
nom ics and which explains the unsatisfactory
this in itself de nes effective modelling
nature from the perspective of mainstream
uniquel y as con ned to the form al mathemat-
econom ics of a range of models within econ-
ical models of econom ics. Krugm an, goes
om ic geography and regional science. It also
on, however, to assert this to be the case and
explains the partial and unsatisfactory treat-
to de ne all other approaches as essentially
ment of local externalities as the basis for
`metaphors rather than models perhaps pro-
spatial agglom eration and the neglect of
viding, at best, useful insights and raw mate-
much of what Marshall had to say on the
rials for true modelling to draw on.
subject. Unable to pursue these topics with
M ainstream econom ics, he argues, rests on
the `rigour expected of them , econom ists
the two key principles of maxim isation and
largely ignored them .
equilibrium or, as he puts it, obvious oppor-
The solutio n, so far as Krugm an is con-
tunities for gain are rarely left unexplo ited,
cerned, lay not in any dilutio n of the essential
and things add up. It is the success of econ-
tenets of econom ic modelling but in the de-
om ic models which apply these basic princi-
velopm ent of a new set of models which
ples to com plex situations which justi es the
could, successfully in his terms, incorpo rate
approach.
im perfect com petition and hence cope with
increasing returns, the effects of geography T hinking carefully about how self-inter-
and location: ested individ uals would act in a particular
situation, and how these actions would
The reason why space has nally made it
interact, can often produce powerful and
into the econom ic mainstream is therefore
surprising insights (Krugm an, 1995,
obvious: im perfect com petition is no
p. 75).
longer regarded as im possible to model
(Krugm an, 1998a, p. 164). Critics of this approach, the simpli cations
GE OG RAPH ICA L ECO NO MICS 817

and assumptions, he suggests are often well, without having the deeply satisfying
politically motivated driven by values logic of, say, the von Thu nen model
rather than analysis or put off by the appar- (Krugm an, 1995, p. 85).
ent disregard for what people actually think
and feel or by the mathematical com plexity. (which of course Krugm an had him self criti-
The alternatives to mainstream econom ics cised earlier for its need to assume rather
have, he claim s as an empirical prop- than derive the central location around which
osition been notably unsuccessful , citing the model distributed activity).
Galbraith s work on the modern corporation He refers disparagingly as well to what he
as an example im pressive-sounding neo- calls `the antim odel backlash in econom ic
ligism s which make few useful predic- geography of the 1970s, distinc t from the
tions (Krugm an, 1995, p. 77). Krugm an neo-classical tradition itself and drawing on
goes on simply to assert that there is no Marx and, som ewhat curiously, Derrida
alternative to models (Krugm an, 1995, whom he sees as an im porta nt in uence on
p. 79) meaning, again, form al econom ic `post-Fordism and `the regulationist school .
models. Anything else, such as Hirschm ann Backtracking som ewhat, Krugm an allow s
on econom ic developm ent, is at best meta- that Keynsian econom ics, while itself falling
phor, a suggestive heuristic device but, in short of a fully speci ed model, has never-
Krugm an s view, an intellectual dead-end . theless, practical utility. The goal, how ever,
He calls in evidence, as well, the failure of is the `clean model . Ultim ately, he believes,
som e to understand the nature of the trade- there will be a happy ending; in the end,
balance or high-va lue-added industries we will integrate spatial issues into eco-
points which are returned to below . nom ics through clever models (preferably
In terms of developm ent econom ics (refer- but not necessarily mine) that make sense
ring to Murphy et al.) and `geographical of the insights of the geographers in a way
econom ics , he argues that his case is made that meets the standards of the econom ists
by the success of new models of im perfect (Krugm an, 1995, p. 88).
com petition. The new econom ic models are
able, he claims, to capture the essentials Krugm an has, then, a particular disciplinary
without trying to com mitm ent to form al econom ic modelling.
This he vigorou sly asserts as the way for-
capture the richness of reality, either with ward for econom ics as a discipline; he is in
a highly com plex model or with the kind the process dism issive of attempts to under-
of lovely metaphors that seem to evade the stand the world which, in his eyes, fall short
need for a model (Krugm an, 1995, p. 83). of this ideal. Clearly, much of social science
Krugm an him self has developed a series of would dispute the idea of form al econom ic
models in which he has explored central modelling as the holy grail. Many would
place theory, market potential, the localisa- simply dism iss his claims as irrelevant
tion of econom ic activity and the idea of simply the latest version of econom ic im peri-
circular and cum ulative causation (Krugm an, alism while acknowledging that econom ics
1991a, 1991b, 1993a, 1993b, 1995). has its insights and contributions. In this
Econom ic geographers on the other hand, sense, there is much in Krugm an s work
according to Krugm an, when they could not which is interesting and suggestive in rela-
produce models which could derive macro- tion to the agglom eration of econom ic ac-
behaviour from micro-m otives, tivity and the com petitiveness of cities and
regions both in its own right and in
essentially settled for what they could do: counte rpoint to other work within econom ic
schematic descriptions of the data or or- geography and urban and regional studies.
ganising princ iples that made intuitive Much of the latter would clearly fall into his
sense and/or seemed to t the facts fairly category of metaphor and seeks to capture
818 MA RT IN BODD Y

som ething of the richness of reality work com parative advantage. This specialisation is
with which he would appear to be largely in turn re ected at the level of spatial con-
unfam iliar. It should be noted as well that centration within national territories. Search-
Krugm an him self is not averse to discursive ing for a way to explore ideas of increasing
reasoning and speculation with little direct returns, im perfect com petition , multiple
foundation in his touchstone of the form al equilibria and an often decisive role for his-
econom ic model his essay on Chicago and tory and accident, Krugm an turns to the sub-
Los Angeles is a case in point (Krugm an, national level:
1997, pp. 205214). W hen he turns to em-
pirical evidence for corroboration, this can W hen one turns to the location of pro-
also as Martin and Sunley (1996, pp. 278 duction within countries, the evidence for
279) observe in relation to his work on what Nicholas Kaldor called the irrel-
regional specialisation be som ewhat crude. evance of equilibr ium econom ics is far
more com pelling. The long shadow cast by
history and accident over location of pro-
3. Cities and Com petitiveness duction is apparent at all scales, from the
smallest to the largest And this clear
International Trade, Com petitiveness and the
dependence on history is the most con-
Regional Econom y
vincing evidence available that we live in
Krugm an makes a strong case for the im port- an econom y closer to Kaldor s vision of a
ance of linkages betw een international trade dynam ic world driven by cum ulative pro-
and com petitiveness and the sub-national cesses than to the standard constant returns
level: model (Krugm an, 1991a, pp. 910).

One of the best ways to understand how


the international econom y works is to start Increasing Returns and the Localisation of
by looking at what happens inside nations. Econom ic A ctivity
If we want to understand differences in
Krugm an sets out to explore the idea that
national grow th rates, a good place to start
increasing returns represent a key factor in
is by examining differences in regional
international trade and in the localisation of
grow th; if we want to understand inter-
econom ic activity arguing at the outset that:
national specialisation, a good place to
start is with local specialisation (Krugm an, increasing returns affects econom ic
1991a, p. 3). geography at many scales. At the bottom
of the scale, the location of particular
As he observes:
industries clearly often represents the
Step back and ask, what is the most strik- `locking in of transitory advantages. At
ing feature of the geography of econom ic an interm ediate level, the existence of
activity? T he short answer is surely con- cities them selves is evidently an increasing
centration (Krugm an, 1991a, p. 5). returns phenom enon. At the grand
level, the uneven developm ent of whole
This geographical concentration is, more- regions can be driven by cum ulative
over, he argues clear evidence of the per- processes that have increasing returns at
vasive in uence of som e kind of increasing their root (Krugm an, 1991a, p. 10).
returns (Krugm an, 1991a, p. 5). The new
trade theory argued that much trade repre- He poses the question as to persistent con-
sents arbitrary specialisation based on in- centration of manufacturing in the north-east
creasing returns, rather that an effort to take US. This was in essence because each indi-
advantage of exogenous differences in re- vidual manufacturing facility stayed within
sources or productivity (Krugm an, 1991a, the manufacturing belt because of the advan-
p. 7) i.e. increasing returns rather than tages of being near other manufacturers. And
GE OG RAPH ICA L ECO NO MICS 819

the apparent incentive for manufacturers to itself, characterised as it is by high xed


cluster together explains the persistence of costs. T his is im portant in its own right.
the manufacturing belt even after the bulk of Krugm an goes on to argue that:
US prim ary produc tion had shifted to other
M ore im portant than its im mediate
regions. Once the belt had been established,
signi cance, however, is what the history
it was not in the interest of any individu al to
of manufacturing location says about the
move out (Krugm an, 1991a, p. 14).
nature of our econom y in general. And
A range of different explanations might be
what it says is that increasing returns and
plausible. Initially, however, Krugm an ar-
cum ulative processes are pervasive and
gues that:
give an often decisive role to historical
Given suf ciently strong econom ies of accident (Krugm an, 1991a, p. 25).
scale, each manufacturer wants to serve
This has similar im plications, potentially, at
the national market from a single location.
the level of particular urban agglom erations.
To minim ise transportation costs, she
Krugm an goes on to make tw o further
chooses a location with large local
points on the basis of the simple model, both
dem and. But local demand will be large
potentially of wider signi cance. First, the
precisely where the majority of manufac-
circular process whereby the location of de-
turers choose to locate. Thus there is a
mand determines the location of production
circularity that tends to keep a manufactur-
and vice versa can tend to lock into place
ing belt in existence once it is established
existing patterns including centre and periph-
(Krugm an, 1991a, p. 14).
ery relationship s. W hen change does happen,
The forces which tend to make manufactur- how ever, it can in theory at least be quite
ing concentrate emerge from the three-way rapid. Gradual change in the underly ing
interaction of scale econom ies, transport parameters can push the model throug h a
costs and factor mobility . The argum ent is threshold generating quite sudde n change in
thus based simply on the interaction of in- geographical structure. A peripheral market,
creasing returns at the level of individ ual for example, can gradually grow to a point
rms, transport costs and dem and local ex- where it can support ef cient-scale manufac-
ternalities which he discusses at length else- turing which is therefore pulled away from
where are not, here, a factor. He supports his elsewhere. Rational expectations might
argum ent using a simple two-country model, smooth the change but this would depend
dem onstrating that a range of different equi- on the inform ation available in practice. The
librium outcom es are possible dependent on simple model suggests, therefore, that change
the starting-poin t and the parameters. Given in patterns of agglom eration and the shift
suf ciently large econom ies of scale, low from one equilibrium state to another can be
transport costs and suf cient footloo se indus- rapid.
try not tied down by natural resources, a Secondly, again using the simple two-
single location will result. T his, he suggests, country model, Krugm an dem onstrates that
is part of the story behind the emergence of under certain circumstances, expectations
the US manufacturing belt. More generally, and the anticipation of change can becom e
it represents one, highly simpli ed way of self-ful lling. In other words, expectations
deriving the spatial concentration of econ- can determine outcom es and outw eigh the
om ic activity from a simple, form al model. cum ulation of initial advantage. As Krugm an
He goes on to argue that better transport observes, this suggests that local `booster-
links focused on core regions can contribute ism serving simply to create optim ism
to the lock-in of historic al advantage and (rather than confer concrete advantage) may
contribute to cum ulative causation. T his is be justi ed under certain circumstances (par-
because higher traf c volum es will generate ticularly when the speeds of adjustm ent in
econom ies of scale in the transport sector location are rapid): the analysis suggests
820 MA RT IN BODD Y

that in principle, at least, boosterism may producers concentrate (Krugm an, 1991a,
make perfectly good sense (Krugm an, p. 71).
1991a, p. 33). The reverse may also be true.
Again, this is based on a highly simpli ed By `region he is talking roughly at the scale
model. It does, how ever, provide som e sup- of US states or Europe an nation-states. Trade
port from the perspective of form al econom ic in this model is the means of achieving econ-
modelling for policies of boosterism. om ies of scale by localising production in
He goes on in later work to dem onstrate particular industries and exporting to other
num erical models which can, again in a regions (meanwhile im porting goods from
highly simpli ed manner, handle multi- other regions which similarly achieve econ-
regional situations. In one such model, re- om ies of scale in the sectors in which they
gions are assumed to be arranged around a specialise). T he lower transport costs and the
circle, with transport possible only around its larger the potential econom ies of scale, the
circumference the `race-track model. An stronger the tendency towards agglom er-
even spread of manufacturing across the re- ation.
gions represents one equilibrium position . Krugm an develops this argum ent partly to
The circular logic of concentration means, discuss the effects of trade and tariff barriers
however, that any slight unevenness can and the possible effects of moving towards a
cause the spontaneous developm ent of one or single European market. Econom ic inte-
more local concentrations of manufactur- gration by facilitating trade, it is argued, will
ing typically, in simulations generating tend to increase the tendency towards ag-
tw o concentrations on opposite sides of the glom eration. He also makes the point, how-
perim eter of the circle (Krugm an, 1998a, ever, that trade prote ction and/or other
pp. 170171). He also develops a version of policies may under som e circumstances, en-
the original von Thu nen model in which the able a region to secure or protect a localised
central city is not simply assumed; instead, agglom eration of econom ic activity. Using
manufacturing concentrates in the city be- again a simple two-region model, he shows
cause of forwards and backw ards linkages that as transport costs fall (or econom ic inte-
generated by that concentration itself (Fujita gration increases) a point will be reached
and Krugm an, 1995). M ore com plex models where it is more ef cient to produce in one
but still only in one-dimensional space region only and export to the other. In these
spread along a line have been constructed circumstances, a tem porary tariff barrier, for
(Fujita and Mori, 1996a); while Fujita and example, may enable a region to protect the
Mori (1996b) have also modelled the cluster- grow th of industry to the point where
ing of manufacturing around ports and trans- suf cient lead is established and the barrier
port nodes. can later be removed. This is close to what
he acknowledges to be the largely discredited
infant industry im port-substitution policy.
Econom ies of Scale and Spatial Concen- Using the example of Canada from the late
tration 19th century, he argues that an expanding
market can suppor t the growth of industry to
Discussing trade, Krugm an argues that the a scale where it can survive without the need
nature of the externality generating agglom - for protection. In the case of Canada, such
erative tendencies at a broad regiona l level: grow th was fuelled by im migration. It is also
observed that, with econom ies of scale,
com es from market size effects in the face cheapening transport costs will not necess-
of transportation costs from the forward arily encourage econom ic activity to move to
and backward linkages that make produc- more peripheral, cheap labour regions. The
ers want to concentrate near large mar- advantages of such a move may be out-
kets and puts large m arkets where weighed by the additional econom ies of scale
GE OG RAPH ICA L ECO NO MICS 821

achievable in the core region lower trans- 1991a, p. 99). He concludes by emphasising
port costs enabling the peripheral market to the im portance of `path dependence cum u-
be supplied from the centre. If transport costs lative causation xing historically deter-
were to fall further, then the point might be mined patterns QWE RTY econom ics,
reached where the low costs of production in making the analogy with the persistence
the periphery outw eighed the gains from of the original layout of early typew riter
concentrating production in the core area. keyboards.
In essence, Krugm an argues that the com -
bination of econom ies of scale and the fact
External Econom ies and Agglom eration
that there are costs to transactions across
space will tend to generate the spatial con- In Geography and Trade, Krugm an (1991a)
centration of econom ic activity: also develops a parallel argum ent linking the
localisation of econom ic activity to Marshal-
Because of econom ies of scale producers
lian-type external econom ies. This represents
have an incentive to concentrate pro-
an alternative or possibly com plem entary ex-
duction of each good or service in a lim -
planation there is som e suggestion that ag-
ited num ber of locations. Because of the
glom eration based on scale econom ies
costs of transacting across distance, the
operates at a broader national or regiona l
preferred locations for each individual pro-
level while external econom ies apply at a
ducer are those where dem and is large or
more localised level (Krugm an, 1991a,
supply of inputs is particularly con-
pp. 7071). Quite why, and at what scale,
venient which in general are the loca-
each account is applicable is unclear. It is
tions chosen by other producers. Thus
here, how ever, that Krugm an s discussion of
concentrations of industry, once estab-
externalities com es closer to current debates
lished, tend to be self-sustaining; this ap-
within econom ic geography and urban and
plies both to the localisation of individ ual
regiona l studie s. T he starting-point is the
industries and to such grand agglom era-
striking localisation of particular industrial
tions as the BostonW ashington Corridor
sectors which, as Krugm an observes, has
(Krugm an, 1991a, p. 98).
generated a very extensive literature from
The num ber of such agglom erations and any Marshall through Hoover to, more recently,
tendency towards concentration or decentra- Porter. Krugm an goes back to Marshall s
lisation will depend on the balance between original account of the localisation of indus-
econom ies of scale and the transaction costs try based on labour market pooling ,
im posed by space (including, for exam ple, specialised inputs and services, and know -
tariff barriers as well as transport costs). ledge and inform ation.
At one level, as Krugm an acknowledges,
none of this is particularly novel. The differ- Labour market pooling . First, as Marshall
ence now, he suggests, is that these ideas can had argued, the concentration of enterprises
be incorporated into mainstream econom ics in a particular location perm its a pooled
thanks to recent advances. He goes on to labour market for workers with specialist
claim that it is now possible to do geogra- skills, to the bene t of both workers and
phy as rigorou sly as you like . Observing as rms. Using a simple two-region model,
well that: T he geographers them selves prob- Krugm an dem onstrates how clustering
ably won t like this: the econom ics pro- means that rms will at tim es (but not al-
fession s simultaneous love of rigor and ways) be able to recruit the additional labour
contem pt for realism will surely prove infuri- which they require; similarly, workers will,
ating (Krugm an, 1991a, p. 99). Regional at tim es (but again not always), be able to
com parison represents, however, a huge, al- obtain employm ent which otherw ise they
most untapped source of evidence about how would not, by switching employers. Over
our econom y really works (Krugm an, tim e, clustering is m ore ef cient and both
822 MA RT IN BODD Y

employers and workers bene t. This, Krug- logical spillovers. He suggests that the fact
man argues, assumes increasing returns with that much clustering pre-dates Silicon Valley
scale othe rwise employers could split their and is evident in low-tech sectors suggests
production into smaller plants and distribute that other form s of externality are im portant.
them across a dispersed labour force. It is the He attempts to back this up with a simple
com bination of increasing returns and uncer- empirical examination of the concentration
tainty which makes concentration more of different industrial sectors across the US
ef cient. Going beyond Marshall s argum ent which suggests that low -tech industries are in
and reinforcing it, Krugm an also dem on- som e cases more concentrated than high-tech
strates that this is the case even if wages rise ones. This is som ething of a non sequitur
and fall readily with dem and; he also shows since technological spillovers could, in the-
that an open and com petitive labour market ory at least, be the dom inant form of local
is more ef cient than a captive labour force externality, applicable historically and across
(Krugm an, 1991a, pp. 4349). different sectors.
He also argues that technological
Specialist inputs and services. Secondly, the spillovers are assumed true by de nition,
localisation of industry can generate dem and rather than modelled; they also leave no vis-
at a scale which makes possible the ef cient ible trace which can be examined directly.
provision of specialist inputs and services. This sugge sts that investigation should con-
An industrial cluster can suppor t a larger centrate on form s of externality which can be
range of such inputs and allow the producers more directly addressed:
of those inputs to generate additional econ-
Know ledge ows, by contrast are invis-
om ies of scale, making them more ef cient.
ible; they leave no paper trail by which
The producers of these inputs will tend, then,
they may be measured and tracked, and
to locate alongside industry to minim ise
there is nothing to prevent the theorist
transport costs, furthe r reinforcing the clus-
from assum ing anythin g about them that
tering of econom ic activity. Again, as Krug-
she likes. A sociologist might be able to
man demonstrates, the argum ent turns on
help with survey methods; but I would like
econom ies of scale. W ithout scale econom -
to get as far as possible with drab, down-
ies, a cluster of any size could replicate the
to-earth econom ic analysis (Krugm an,
full set of specialist inputs.
1991a, p. 55).
Know ledge and inform ation ow s. T hirdly, Again this is a relatively weak argum ent
the localisation of industry facilitates the and underestim ates the researchability of
ow of know ledge and inform ation between technological spillovers.
rms. These represent in econom ic terms, Krugm an s point that the full range of
`pure or `technologi cal spillov ers or exter- local externalities needs to be considered,
nalities. Given that they do not depend on rehabilitating the elements of M arshall that
scale econom ies, they do not violate the as- mainstream econom ics sidelined is, how ever,
sum ptions of conventional econom ic models very valid. Cum ulative processes are pervas-
and could thus be assumed away. As Krug- ive (Silicon Valley is only the latest glitzy
man argues, unable to handle increasing re- version), he concludes, and labor pooling
turns, econom ists conve ntionally ignored the and the supply of specialised inputs, play a
rst tw o types of externality. His own ap- large role even when pure technological ex-
proach attempts to rehabilitate them . He also ternalities seem unlikely to be im portant
notes that modern high-technolo gy clusters (Krugm an, 1991a, p. 62). Over tim e, he ar-
including Silicon Valley and Route 128, with gues, the bene ts of localisation decline:
the acknowledged im portance of ows of There is a kind of product cycle, in which
know ledge and inform ation, have been used emergent new industries initially ourish in
to argue for the im portance of pure, techno- localised industrial districts, then disperse as
GE OG RAPH ICA L ECO NO MICS 823

they mature (Krugm an, 1991a, p. 63) (Krugm an, 1997, p. 4). The argum ent is po-
althoug h the only argum ent he offers is that tentially signi cant in terms of what it might
autom ation would reduce the im portance of a im ply for the nature of com petition between
pool of skilled labour. cities and urban agglom erations particu-
Krugm an also argues, again, for the im - larly given Krugm an s emphasis on the close
portance of history and of small, accidental relationship between international trade and
events which determine the initial location of the sub-national concentration of econom ic
particular sectors which then becom e locked activity.
in by cum ulative processes, illustrating this According to Krugm an, however,
with anecdotes of the folk-origins of, for
this view of com petitiveness is ` atly
example, carpet-making:
wrong : com petitiveness is a meaningless
An accident led to the establishment of the word when applied to national econom ies.
industry in a particular location, and there- And the obsession with com petitiveness is
after cum ulative processes took over both wrong and dangerous (Krugm an,
(Krugm an, 1991a, p. 61). 1997, p. 22).
And again: Sim ilarly:
small accidental events start a cum ulat- It is simply not the case that the world s
ive process The resulting pattern may leading nations are to any im portant de-
be determined by underlying resources and gree in econom ic com petition with each
technology at som e very aggregative level; other, or that any of their major econom ic
but at ground level there is a striking role problems can be attributed to failure to
for history and accident (Krugm an, 1991a, com pete on world markets (Krugm an,
p. 67). 1997, p. 5).
He also makes the point that while much of
A country , he argues, is not a corporation
service activity follow s the distribution of
and has no nancial bottom -line, making
popula tion (and one would add econom ic
com petitiveness at the outset hard to de ne.
activity in general), other service activities
Growth in real incom e or living standards, he
are among the most highly localised and
argues, essentially equals rates of growth in
given the increasing ef ciency of infor-
dom estic productivity by dom estic factors
mation transfer increasingly so.
rather than by som e form of com petition for
world m arkets. In part this is because trade is
Trade and Com petitiveness relatively small com pared with production
for the dom estic market around 10 per cent
A major them e in Krugm an s work has been
of value added in the US. More fundam en-
his critique of those who maintain that com -
tally, while com panies with similar product
petitiveness in the global econom y is vital to
lines may com pete, trade betw een the major
the continu ed econom ic success of the `de-
industrial countries is essentially bene cial
veloped countrie s and that their position is
providing both export markets and useful
seriously threatened by the transfer of pro-
im ports often at low er prices and of better
duction to low -wage econom ies which would
quality than those produced dom estically:
then ood the dom estic market with cheap
im ports, destroying dom estic industry and International trade is not a zero-sum
employm ent. Trade de cits in manufactured gam e while com petitive problems
goods have pushed up dom estic unem ploy- could arise in principle, as a practical,
ment and forced workers into much low er- empirical matter the major nations of the
paying service-sector jobs. The econom ic world are not to any signi cant degree in
problem, according to this view is essen- econom ic com petition with each other
tially one of com petiting on world markets (Krugm an, 1997, p. 10).
824 MA RT IN BODD Y

The manufacturing trade de cit argum ent, he productivity gain in the now -predom inant
argues, falls on the basis of simple arithm e- service sector.
tic any possible reduction in real wages He develops his line of argum ent and his
falling far short of the observed shift in the attack on various propon ents of com petitive-
US in recent years. He is also critical of the ness in series of articles and lectures col-
com mon association made between com peti- lected together in Pop Internationalis m
tiveness and high-technolo gy industry capi- (Krugm an, 1997). He also shifts or at least
tal and labour need to ow into high clari es his earlier positio n on `strategic
value-added sectors (meaning value added trade theory , based on the new theories of
per worker) which are com monly believed to international trade referred to above whereby
be synonym ous with high technology. Krug- intervention to prom ote exports in particular
man argues however, using US statistics, that sectors could stim ulate growth and raise real
`high value-added sectors are those with a incom es. He argues that subsequent research
large volum e of xed capital relative to has indicated the problem s in practice of
labour. They include, for example, oil identifying relevant sectors and appropriate
re ning and cigarettes. Even steel and car- form s of intervention, while the pay-offs
making are signi cantly ahead of aircraft and even from a highly successful strategy would
electronics where value added is around the be extrem ely modest: the stakes involve d
average for manufacturing as a whole. The are very small, on the order of a few tenths of
obsession with com petition , he argues, one percent of national incom e (Krugm an
re ects a com bination of the exciting im ages 1997, p. 69). He reiterates this in his review
conjured up, the attractions of blam ing other of Tyson (1992) a self-declared cautious
nation-states for dom estic concerns, the rela- propon ent of strategic trade theory as a
tively simple policy agenda such argum ents policy measure. It is the very com plexity
seem to offer, and the political convenience of intervening in im perfect markets for
of such argum ents. T hey re ect as well the example, outcom es depend on the dynam ic
misguide d application of corporate strategy strategic responses of different actors to
concepts, with their emphasis on head- policy interventions which makes such
to-head com petition , to national industrial theory so dif cult to apply in practice.
policy. Krugm an subsequently adds to his argu-
Such argum ents have been prom oted by a ment more speci cally in relation to the sup-
range of policy advisers and supposed ex- posed threat to W estern econom ies posed by
perts who are the target for much of Krug- com petition from low -wage nations. The ar-
man s critique. One propon ent, he argues: gum ent, as sum m arised by Krugm an, is that
with increasingly free trade, many industries
asks us to accept `com petitiveness as a will be tempted into relocating in countries
kind of ineffable essence that cannot be with low labour costs which will attract capi-
either de ned or measured. Data that seem tal and technology. These countries will run
to suggest the im portance of the essence huge trade surpluses creating either large-
are cited as `indicators whatever that scale unem ploym ent or sharply falling wages
means, while those that do not are dis- in the form erly high-w age nations. T his, he
missed as unreliable (Krugm an, 1997, argues, runs com pletely counter to the basic
p. 32). fact of national incom e accounting whereby:

The danger, he argues, is of the misallocation savings 2 investm ent 5 exports 2 im ports
of resources, protectionism and trade wars
com bined with failure to recognise the gains Thus, if investm ent rises in low wage coun-
in terms of international trade from specialis- tries through an in ow of capital, then over-
ation. The key reason for the stagnation in all they will run a trade de cit not a trade
US living standards, he argues, is lagging surplus. The reason for this is that wage rates
GE OG RAPH ICA L ECO NO MICS 825

will rise along with average productivity. argum ent put by Krugm an, moreover, relates
Individ ual modern factories may achieve speci cally to the capital diversion from the
high productivity with relatively little in- productive sector of the US econom y not
crease in their ow n wage rates (since average the im pact of wage levels as such or produc-
productivity rates and wage rates across the tivity in the developing countries, which as
country as a whole will rise only marginally). argued earlier have little im plication.
They may therefore be able to com pete effec-
tively and increase exports. T his, according
Localisation and the W orld Econom y
to Krugm an, simply re ects com parative ad-
vantage and does not contradict the broader Krugm an provide s a further perspective on
conclusion relating to the econom y as a the relationship between international trade
whole and its relation to the high-w age and the localisation of econom ic activity in
countries. The latter bene t, in fact, from his essay com paring the Chicago of a century
im porting goods more cheaply than they can ago with present-day L os Angeles (LA). An
be produced dom estically and from the argum ent which is, he claims, growing in
overall trade de cit run by the low er-wage popular ity among econom ists is: that a par-
countries. Paradoxically at rst sight, it is ticularly good way to unde rstand the Ameri-
rising wages in the low -wage countries that can econom y is by studyin g American cities
may im pact adversely on the higher wage (Krugm an, 1997, p. 206). He makes three
rather than continu ing low wages. As wages main points about LA as a city. First, the
rise throug h productivity gains in som e sec- most striking feature of LA is: what we
tors (for example, cars) goods in other export might call the abstractness of the m odern
sectors (for example, footw ear and clothing) city s econom y the way it seems so discon-
may becom e more expensive re ecting the nected from the physical world (Krugm an,
increase in average wages. International 1997, p. 298). Echoing his argum ents out-
trade, as he puts it elsewhere: is not about lined earlier as to the role of history coupled
com petition, it is about mutually bene cial with cum ulative causation, he suggests
exchange (Krugm an, 1997, p. 120). Krug- (pp. 208209) that:
man s argum ent essentially suppor ts the con-
T ry to understand why any of LA s most
cept of free trade whereas the logic of the
characteristic industries are there now, as
counter argum ent, as he observes, leads in
opposed to how they got started, and you
the other direction. And overall, he maintains
always nd a circular argum ent. The lm
that the predom inant im pact of productivity
studios are there because of the large pool
growth in developing countries is higher
of people with specialised skills, and the
wages in those countries with generally no
skilled people are there because it s where
adverse im pact overall in developed coun-
the jobs are.
tries.
Elsewhere, how ever, he argues the seem- The city s econom y, he argues, seems
ingly contradictory point that, in theory at strangely detached from any sense of place
least, a ow of investm ent from the devel- (p. 210). This is reinforced by the sense in
oped, high-w age countries to low -wage which the working people of LA, the build-
countries (if, for example, manufacturers ings in which they work and whole city look
shifted dom estic produc tion to low -wage very much like anyw here else. Secondly, the
countries), could result in a drop in produc- export base of LA is highly specialised; de-
tivity and real wages in the developed coun- spite the city s im mensity, it is dependent on
tries. Here he argues that, in practice, given a few key industries. It sells a surprisingly
the scale of capital ow s to the low -wage narrow range of goods and services to the
econom ies relative to the overall size of the world at large. Thirdly, most employm ent is
developed world econom y, the effect on real in `non-b ase activities, goods and particu-
wages has been negligible. The theoretical larly services provide d by local workers for
826 MA RT IN BODD Y

local consum ption. Com paring this (p. 211) same things as most people in London,
with Chicago a century ago: Paris and modern Chicago (p. 213).

Although we talk a lot these days about This, he suggests, is also why `deindustriali-
globalization, about a world grown small, sation , the disappearance of m anufacturing
when you look at the econom ies of mod- jobs, is not the concern it is som etimes
ern cities what you see is a process of thought to be.
localisation: a steadily rising share of the
work force produces services that are sold
4. Summ ary
only within that same metropoli tan area.
It is useful at this stage to sum marise the key
Chicago a century ago was, in any case, ideas arising out of Krugm an s work which
strongly tied in with a world-w ide web of are most relevant to the form ation and
markets, trade and nance. grow th of cities, trade and regional develop-
It is this more recent process of localisa- ment, agglom eration and localised externali-
tion, he goes on to argue, which: ties, know ledge-transfer and inform ation.
explains what would otherw ise seem a (1) The spatial concentration of econom ic
paradox for the world econom y: the fact activity at a broad regiona l level (US states,
that international trade is not much bigger Europea n countries) can be derived from for-
now , as a share of world output, than it mal models which com bine scale econom ies
was a century ago (p. 212). related to market size, and transport costs.
Tradeable goods represent a steadily shrink- Econom ies of scale encourage concentration
ing share of the econom y as a whole , even of production. Transaction costs across space
though elements of production seem increas- encourage location where dem and is large
ingly internationalised and transport and and/or the supply of inputs is particularly
com munications so much more effective. convenient. Such models dem onstrate that,
This re ects the greatly increased productiv- depending on the value of the param eters,
ity achieved in the production of goods than either a single agglom eration or m ultiple ag-
in services. Or rather, the fact that while we glom erations may be derived a range of
have made: different equilibrium points being possible.

rapid progress in eld where the infor- (2) The lower the transaction costs across
mation is relatively easy to form alise, to space (including transport costs and tariffs)
embody in a set of instruc tions to a robot and the greater the econom ies of scale, the
or com puter; we have made much less greater the tendency towards agglom eration
progress in activities, from cutting hair to and the smaller the likely num ber of clusters.
medical care, where the inform ation pro- (3) External econom ies play a key role in the
cessing is of the exceedingly subtle and clustering of industries in a particular loca-
extremely com plex kind that we call com - tion. These, however, include not only the
mon sense (p. 213). `pure or `technological externalities of con-
vention al mainstream econom ics (which do
Given this increase in productivity (and re-
not violate assumptions of perfect com pe-
duction in employm ent), the econom y in-
tition) such as ows of know ledge in a
creasingly focuses on the `non-tradeable (by
localised cluster of rms. They also
which he means `non-exportable ) activities
include as Marshall had argued earlier the
that make up the `non-b ase employm ent that
role of market size in support ing ef cient-
occupies most people in modern cities:
scale specialist suppliers of goods and ser-
And that s why most people in Los Ange- vices and labour market pooling . The effects
les produce services for local consum p- of both depend on increasing returns to scale
tion, and therefore do pretty much the and therefore im ply im perfect com petition.
GE OG RAPH ICA L ECO NO MICS 827

(4) The striking feature of econom ic activity world econom y and the shift of production to
is its spatial concentration. In order, there- low-wage countries pop international-
fore, to understand how the international ism is a misleading and dangerous obses-
econom y works or to unde rstand differences sion. Countries are not generally in
in national growth rates, a good place to start com petition a nation is not a corporation.
is by looking at differences in regional
(12) Decline in manufacturing employm ent
growth and at local specialisation.
and output by value in high-w age countries
(5) Increasing returns of one form or another largely re ects increased productivity in
affect econom ic geography at a range of these industries com pared with the growth of
different scales including the location of par- employm ent in a range of service activities
ticular industries, the existence of cities where this has been much slow er. T he effects
them selves and developm ent at the broader of a trade de cit in manufactured goods, of
regional level of US states or European coun- deteriorating terms of trade or a net shift of
tries. investm ent overseas to low-wage countries
have been negligible.
(6) Scale econom ies are at least as im portant
as com parative advantage in explaining trade (13) The levels of national incom e and of
betw een territories. Much of international living standards re ect, above all, dom estic
trade represents arbitrary specialisation based productivity and are largely unaffected by
on increasing returns. Such scale econom ies shifts in produc tivity in other countries. The
re ect both backwards linkage s whereby key reason for the relative decline in US
ef cient-scale suppliers are suppor ted and living standards has been lagging productiv-
forward linkages to markets. ity gain in the service sector which accounts
for an increasing share of consum er expendi-
(7) The agglom eration of econom ic activity
ture.
in space in part re ects processes of cum ulat-
ive causation activity tends to cluster where (14) Strategic trade policy protectionism ,
markets are large and markets tend to be subsidy or suppor t for key industries can in
large where activity clusters. This is rein- theory generate longer-term com petitive ad-
forced by, for example, the concentration of vantage. This will particularly be the case
transport infrastructure to serve such clusters, when resources, including labour and capital,
such infrastructure itself being subject to in- can be drawn in from othe r regions. In prac-
creasing returns. tice, how ever, strategic trade polic y at the
national level is hard to im plem ent and the
(8) Much of the localisation of econom ic
scale of im pact is likely to be very lim ited.
activity within countries re ects historic al
accident and seemingly trivial initial events, (15) Attempts to boost com petitive ness by
coupled with cum ulative causation. support ing high value-added sectors which
com monly associate these with high technol-
(9) Cum ulative processes tend to `lock in
ogy are misguide d. High value-added sectors
patterns of uneven developm ent. Patterns of
are those where capital content is high rela-
econom ic developm ent are subject to `path
tive to labour and in the US, for example,
dependence or QW E RTY econom ics.
include cigarettes and oil re ning. E lectron-
(10) Expectations and the anticipation of ics and aerospace are about average for man-
change can becom e self-ful lling. Local ufacturing as a whole and well below steel
boosterism may therefore be justi ed. and cars.

(11) Trade is generally bene cial, a positiv e- (16) E xternalities as a source of clustering of
sum gam e based on strategic com plem entar- econom ic activity are evident historically and
ity rather than zero-sum . The supposed across different industrial sectors. They are
dangers posed by an increasingly com petitive by no means con ned to modern, high-tech-
828 MA RT IN BODD Y

nology industry as typi ed by Silicon Valley studies then, how im portant is it that the
or Route 128 and where `know ledge effects of one possible (inde ed likely) factor
spillov ers are com monly seen as dom inant. generating agglom eration can be modelled to
the satisfaction of professional econom ists?
(17) The export base of modern cities in the
developed world is narrow and highly spe-
cialised. An increasingly large share of their External Econom ies
popula tions produce `non-b ase goods and This problem of disting uishing between dif-
services for local consum ption. E conom ic ferent factors contributing towards agglom er-
activity is in this sense increasingly localised ation is apparent in Krugm an s own work.
despite the apparently increasing internation- He sees agglom eration as generated by the
alisation of produc tion and im provem ents in interaction of increasing returns and trans-
transport and com munications. Tradeable action costs across space what Henderson,
(export) goods account for a decreasing share (1986, in Glaeser et al., 1992) has termed
of employm ent and the econom y as a whole. `urbanisation externalities . He also argues
This re ects the greatly increased productiv- for the im portance of localised external econ-
ity achieved in these industries com pared om ies to the localisation of econom ic activity
with much of the service sector. in space. T hese rest on the Marshallian trin-
ity of labour pooling , specialised inputs and
5. Discussion and Critique know ledge transfers. How the two relate is
not however made clear. The increasing re-
This section goes on to take a more critical turns/spatial transaction costs model, labour
look at som e of the key ideas arising out of pooling and specialised inputs have in com -
Krugm an s work. The paper then goes on, in mon the im portance of increasing returns
the light of this discussion, to look at a (whereas know ledge transfers, it is assumed,
broader albeit related range of work do not). The source of those increasing re-
within the recent econom ics literature. turns seems, however, to differ among all
three. In relation to the pure increasing re-
turns model in fact, Krugm an does not seem
Agglom eration and Increasing Returns
to be very speci c the assumption seems to
Krugm an dem onstrates that agglom eration be that increasing returns relate to xed start-
can result from the interaction of increasing up costs (and possibly the costs of R&D). It
returns and transaction costs. Using simple is not clear whether labour pooling and
form al models, he can derive single-cluster specialised inputs are to be taken as sub-
and multi-cluster outcom es. From the per- categories of this general case.
spective of the econom ics profession, this T here is also som e suggestion in Krug-
dem onstration of form al derivation may be man s work that the pure increasing returns/
of key im portance. The question is, how ever: spatial transaction costs model with market
how im portant or relevant is this more gener- size a key factor operates at a broader re-
ally? It does suggest that increasing returns gional scale, at the level of US states or
and transport costs may in practice be am ong Europea n countries, generating inter-regiona l
the factors generating agglom eration. It can- centreperiphery patterns. Localised external
not reasonably support any claim that this is econom ies on the other hand operate at a
the only or even the most im portant factor. smaller spatial scale, generating local clus-
The model of necessity assumes away the ters at the level of the single city or small
`com plexity of the real world including the cluster of cities (Krugm an, 1991a, pp. 70
operation of other factors. It does not provide 71). T his suggests, perhaps, that pure in-
any way of com paring the relative im port- creasing returns operate at a broader
ance of other factors. From the perspective of inter-regional scale while Marshallian exter-
econom ic geogra phy or urban and regional nalities are som ehow the glue holding
GE OG RAPH ICA L ECO NO MICS 829

together industrial clustering at a more lo- ledge transfers as such, there is a growing
calised level. T his is not, however, made body of what Krugm an would refer to as
explicit or elaborated. It is not clear whether `sociological research identifying and inves-
both pure increasing returns and local exter- tigating processes of know ledge transfer
nalities operate simultaneously; what might (Jaffe et al., 1993; Scott and Storper, 1992).
determine the relative im porta nce of each; or It is questionable, as well, whether know l-
whether there is a cross-over point in term s edge transfer is in fact unaffected by scale
of their relative im portance for exam ple, econom ies, as conventionally assumed by
related to scale. Krugm an may have set out mainstream econom ics and Krugm an as well.
to incorporate space more effectively into A larger cluster of enterprises would seem to
mainstream econom ics but he is yet to ad- give more scope for useful transfer of know l-
dress the issue of scale. edge and inform ation. Spatial clustering
In discussing the Marshallian trinity of translates into both low er transaction costs
localised externalities, Krugm an discusses and also wider opportu nities for matching
the possible relative im portance of labour needs and capabilities and for the inter-
pooling and specialised inputs versus pure, change of useful know ledge and inform ation
technolog ical spillov ers (in the context, for (Scott and Storper, 1992).
example, of Silicon Valley and Route 128). Again, from the perspective of form al eco-
He argues that the rst two are im portant nom ics, it may be signi cant that the oper-
factors at least, because industries other than ation of labour pooling , for example, can be
high technology have been characterised, derived from a sim ple two-country model
historically, by clustering (Krugm an, 1991a, and captured in a form al, mathematical deri-
pp. 53, 59). This argum ent in itself relies on vation (Krugm an, 1991a, pp. 123128). Iden-
the assumption that the clustering of high tifying the centrality of increasing returns to
technolog y derives from factors different a range of processes generating agglom er-
from those operating in other industrial sec- ation adds to an understanding of how these
tors, in particular, from know ledge transfer operate. The lack of purchase on the relative
an assumption im plicit in Krugm an s im portance of different form s of scale econ-
argum ent (Krugm an, 1991a, p. 53). Later, om ies in relation to agglom eration dem on-
however, he also argues that high technology strates, however, the severe constraints on
clusters relate, in part at least, to pooling of the scope of the form al modelling process.
skilled labour (Krugm an, 1991a, p. 65), un-
derm ining his earlier assumption. The key
point is that the relative im portanc e of the History
three main sources of local externalities Krugm an makes a strong case for history in
clearly cannot be derived from Krugm an s the sense of initial location patterns which
form al models. His argum ent is based, there- then becom e locked in by path dependency
fore, on relatively weak empirical and his- generated by processes of cum ulative
torical evidence and on what is at tim es causation.
som ewhat confused verbal reasoning. He
also argues for highlig hting labour pooling T he long shadow cast by history and acci-
and specialised inputs because they can be dent over the location of production is
modelled, whereas know ledge transfers apparent at all scales, from the smallest to
which, he claims, leave no visible trace, can the largest (Krugm an, 1991a, p. 9).
only be assumed. Here he is guilty of what In this sense, he argues, history matters:
he identi es as a key trait in modern eco-
nom ics, that of sidelining what cannot appar- And this clear dependence on history is the
ently be modelled a shortcom ing which he most convincing evidence available that
would presum ably see as justi ed. And we live in an econom y closer to Kaldor s
whereas it may be dif cult to model know - vision of a dynam ic world driven by
830 MA RT IN BODD Y

cum ulative processes than to the standard the historical grounding of his approach
constant-returns model (Krugm an, 1991a, remains unclear and clouded in ambiguity
pp. 910). (p. 269). Krugm an him self might well argue
that his models simply capture the core of the
This focuses attention on the tendency of processes at work as successfully now as in
clusters or agglom erations of econom ic ac- the late 19th century.
tivity in cities and regions to persist and His emphasis on continu ity in the factors
consolidate over tim e. M artin and Sunley accounting for agglom eration is, as Martin
(1996) highlig ht this as an im portant if not and Sunley point out, in marked contrast to
particularly novel feature of his work. It sug- the recent emphasis in econom ic geography
gests the need to examine, in practice, the on restructuring throug h tim e and the con-
processes by which this happens. From stant reworking of the econom ic landscape.
Krugm an s perspective, path dependence and So too is his emphasis on the random nature
lock-in derive essentially from increasing re- of the initial location of econom ic activity
turns re ecting the com bination of pure in- with all subsequent locational decisions sub-
creasing scale effects togethe r with localised sum ed within `cum ulative causation . This
externalities and, in som e circum stances, ex- again ignores the com plex overlaying of lo-
pectations or active policy. He neglects, cational decisions on inve stment (and disin-
however, the role of local embeddedness and vestm ent) over tim e in response to the
the in uence of local infrastructure, institu- com bined and interacting im pacts of a wide
tional, social and cultural practices the range of factors.
`socio-institut ional externalities emphasised
by the geogra phical literature and social eco-
nom ics of urban and regional studies. Again Local Dependency
this re ects his disparagem ent of social eco- Krugm an, as we have seen, emphasises the
nom ics and his sidelining of factors which role of the pure increasing returns/spatial
cannot readily be modelled. transaction costs argum ent in generating ag-
Martin and Sunley point out, however, that glom eration of econom ic activity in space.
despite an emphasis on history in the sense This in itself should point us in the direction
of path dependence and lock-in, Krugm an of inve stigating the operation of this in prac-
pays too little attention to history in the sense tice. It is likely to be one key factor in
of change and developm ent: explaining the growth of cities and regions
He claim s that the same broad locational or, more accurately, of industries and indus-
forces which explain the growth of nine- trial sectors. He also, as noted, emphasises
teenth-century concentrations also underlie the im portance of local externalities, going
the continued tendency to agglom eration. beyond the conventiona l acceptance by
Indeed, this is one reason why he is reluc- mainstream econom ics of pure technological
tant to emphasise technological spillovers spillovers to rehabilitate and extend M ar-
as a key determinant of contem porary shall s broader range of processes. Recent
clusters (Martin and Sunley, 1996, p. 269). work within econom ic geography and urban
and regional studies has, however, gone con-
Krugm an makes this point, for exam ple, in siderably further in exploring the nature of
discussing Silicon Valley com pared with ear- such `externalities .
lier low -tech clusters. In his form al models, T here are a num ber of related strands to
he allows for the fact that parameters may this work. Initial work based in ideas around
change and different equilibria emerge. He post-Fo rdism and exible specialisation
makes general references to historical change tended to see clusters and agglom eration in
in various of his writings for exam ple, terms of neo-M arshallian externalities. Shifts
when contrasting Chicago and Los Angeles. in technolog y and markets had increasingly
As Martin and Sunle y conclude, however, led to horizontal and vertical disintegration
GE OG RAPH ICA L ECO NO MICS 831

and to the developm ent of analogous internal com ings stem to som e extent from the
organisational structures within larger enter- reliance of much of this work on a lim ited
prises. Spatial proxim ity then served to min- num ber of case studies purposively selected
im ise transaction costs between networks of as representing dynam ic industrial districts.
functionally interlinked enterprises. The rise On the other hand, this work demonstrates
of exible specialisation has led, it is argued, the multiple and com plex form s of externali-
to a renew ed emphasis on agglom eration. ties which contribute to agglom eration
There has been an increasing emphasis in through their im pacts on productivity and
later work on the key role of networking, grow th (although little attention has been
linkages and collaboration as a basis for paid to assessing the relative im portance of
innova tion and grow th. Spatial proxim ity different factors). It emphasises and elabo-
from this perspective is about more than rates, in particular, factors which Krugm an
simply minim ising transaction costs. It fos- dism isses as pure technological spillovers in
ters the transfer and exchange of form al the form of know ledge transfer. It also em-
know ledge and inform ation (pure technologi- phasises the overall regional or sub-regional
cal spillovers). It also, however, facilitates context, blurrin g the distinction between the
the developm ent of trust, shared values and individ ual enterprise and the broader set of
conventions on which successful collabora- enterprises and institutional structures as a
tion is based and it fosters the exchange of spatially concentrated production system:
tacit inform ation embedded in the know ledge econom ic activities are em bedded in and de-
and experience of the workforce and the pendent upon social and cultural contexts.
everyday practices of enterprises included Thus this work emphasises the partial nature
in what Storpe r (1997, p. 80) terms untraded of Krugm an s ow n account of the relevant
dependencies . Individ ual enterprises are factors. Krugm an s account, on the other
`embedded in com plex networks and rela- hand, as Martin and Sunley (1996, p. 273)
tionshi ps both market-based and non-m arket. point out, emphasises the continu ing need to
This is reinforced by the developm ent of consider the im portance of market-based in-
institutional structures and infrastructure sup- teraction alongside the non-m arket, non-hier-
portive of spatially concentrated clusters of archy relationships emphasised by the new
enterprises and processes of learning and regiona l geography. As they argue, his
innova tion.
focus on pecuniary externalities, especially
Krugm an him self is particularly disparag-
market-size effects, and the role that large
ing of such work with its emphasis on social
oligopo listic produc ers can play in indus-
econom ics and lack of rigour. It is not with-
trial agglom eration suggests that industrial
out its own dif culties. As Martin and Sun-
geographers need to adjust their accounts
ley (1996) point out, accounts based in
(p. 285).
exible specialisation im plicitly assum e a
move towards perfect com petition. Empirical It is a useful counter to the emphasis in the
research indicates an overemphasis on the exible specialisation literature and related
local interconnectedness of enterprises. The accounts on local externalities among small
continu ing dom inance of large rms is also com petitive rms which cluster togethe r to
underplayed. Vertical disintegration has been minim ise transaction costs. T his represents a
very lim ited with larger, oligopo listic enter- particular strength of his work.
prises if anythin g of increasing im portance. At the same tim e, a key lim itation of his
As Martin and Sunley argue, internal econ- work is his rejection of non-m arket externali-
om ies of scale may be increasingly im portant ties; this rests prim arily on the fact that they
in generating agglom eration as opposed to cannot be modelled and therefore have to be
external econom ies perhaps closer in som e assumed. W hile this may rule them out from
ways to Krugm an s pure increasing returns/ a form al econom ic perspective, as the new
spatial transaction costs model. T hese short- regiona l geogra phy dem onstrates, they can
832 MA RT IN BODD Y

and have been exam ined from alternative technology. Regionally speci c recession
methodo logical and disciplinary perspec- and crisis are more likely and as Krug-
tives. This leaves him , however, reluctant to man s own models have suggested, crisis
discuss the spatial im pacts of technolog ical when it happens can be relatively rapid.
or know ledge spillove rs or the wider range of Questions of specialisation, growth rates and
local externalities suggested by the new re- vulnera bility to crisis of the specialised `ex-
gional geographical literature. T his is som e- port sector are equally relevant at the sub-
what contrary since, as Martin and Sunley regiona l and city scales possibly more so,
(1996, p. 273) point out, Krugm an s given the tendency towards greater specialis-
own dem onstration of the possible role of ation in smaller areas. As Martin and Sunley
expectations in regional developm ent itself (1996, pp. 278279) again point out, how-
suggests the need to understand their form a- ever, there is no simple relationship between
tion which is itself essentially a social regiona l specialisation and vulnerability to
process. regiona l shock. This will depend in part on
the exibility of industrial structures (ar-
guably greater where exible specialisation
Trade, Competitiveness and Policy
is more developed) and labour market
Krugm an argues, as noted above , that inter- exibility. Krugm an s focus on the im plica-
national trade is generally bene cial, that tions of region-speci c shocks is, how ever,
trade is a positiv e rather than zero-sum gam e as Martin and Sunley observe, an im portant
and that concerns over national com petitive- feature of his work. It is at the same tim e
ness are therefore misplaced. Given his argu- lim ited in the sense that it fails to shed much
ment as to the close relationship between light on why particular econom ies can ex-
international trade and the clustering of econ- perience a rapid reversal in fortune or why
om ic activity at the sub-national scale, his others are better able to adjust, industrial
case for international trade has im plications specialisation and labour market exibility
in turn for econom ic prosperity at the re- being only part of the possible explanation.
gional and possibly sub-regional or city W hile his view s have shifted over tim e,
scale. On the other hand, his argum ent has Krugm an has also argued for the role of
not gone unquestioned. As Martin and `strategic trade policy targeting key sectors
Sunley (1996, p. 275) point out, both dim in- where protection, export subsidy , support for
ishing returns and increasing returns can mit- R&D or other form s of assistance could gen-
igate the bene ts of trade. But, as they also erate increased returns. Sim ilar argum ents
observe (p. 285), one of the key strengths of are advanced in relation to `boosterism at
his work is: the city scale, where expectations may also
play a role. Such advantage may subse-
his linking of external econom ies and
quently be locked in by increasing returns,
regiona l industrial agglom eration with
market size effects or other cum ulative ef-
trade [which] provid es an im portant
fects. Regional and local clusters are central
corrective to the new industrial geography,
to this process for it is at that level, accord-
in which regional industrial developm ent
ing to Krugm an, that advantage can most
is view ed overwhelm ingly as an indige-
effectively be created. To quote M artin and
nous process and the role of trade is typi-
Sunley (1996, p. 282), this im plies that:
cally either subord inated or neglected
altogether.
not only are regional and local industrial
Krugm an also argues that where increasing clusters conside red to provide empirical
returns encourage a greater degree of special- proof of the im portance of external econ-
isation, this may render sub-regional agglom - om ies, such clusters help to de ne what
erations more vulnerable to external shocks industries should be suppor ted. Geographi-
from , for example, change in dem and or cal clustering provid es the justi cation for
GE OG RAPH ICA L ECO NO MICS 833

industrial intervention, and the aim of that tions, handle a range of situations in which
intervention should be to foster local ex- increasing returns and hence im perfect com -
ternalities [emphasis in the original]. petition are of central im portance. As noted
earlier, he claims af nity with Kaldor s be-
W hile this goes som ewhat further than Krug- lief in the irrelevance of equilibrium eco-
man him self, it is close, as Martin and Sun- nom ics . As both Dym ski (1996) and Martin
ley point out, to Porter s view on the role of and Sunley (1996, p. 275) argue, how ever,
local and regiona l econom ic developm ent this does not involve moving very far from
policies in fostering national com petitiveness mainstream rational choice equilibr ium mod-
(see, for example, Porter, 1990). It is also els. It does not, therefore, make much in the
close to those who have argued for the way of inroads into the com plexities of real-
exibly specialised industrial district or com - world situations. As Martin and Sunley
plex securing multiple local externalities as (1996, p. 287) observe:
the model for indige nous local econom ic de-
velopm ent (Cooke, 1990; Storper, 1992a, In his quest for econom ic rigor, Krug-
1992b; Hirst and Zeitlin, 1989). man s mathematical form alization of the
However, the question, again, is whether processes of industrial agglom eration and
this leads to increased specialisation and uneven regional development has taken
therefore vulnerability to external shock, or him away from the richness of Kaldor s
whether increased exibility can ensure the original approach tow ard the lim ited ab-
adaptability of localised industrial com plexes stract landscapes of regional science the
to such events. The extent to which active ghosts of constrained maxim isation and
industrial policy does in fact lead to in- equilibrium solutio ns still haunt much of
creased specialisation or whether a multi- his analysis.
sector approa ch is possible is clearly an
issue. At a more speci c level, Krugm an The breakthrough s of Krugm an and othe rs in
him self has questioned the emphasis in trade recent years may look like a signi cant ex-
and industrial policy on high-technolo gy in- tension to the bounda ries of the discipline as
dustries in particular, arguing that this seen from within the con nes of professional
rests on a misunderstanding of high value- econom ics. From the perspective of econ-
added. This is at one level a useful counter to om ic geography and urban and regional stud-
the often uncritical attachm ent to the pro- ies, on the othe r hand, they do not appear to
motion of high technology and misguided have made much of an inroad into the com -
attempts to recreate Silicon Valley. The plexitie s of real-world processes. Indeed,
case for high technology, how ever, rests not from this perspective, the speci c advances
simply on issues of value-added, but on po- in terms of form al modelling look relatively
tential growth rates, the stim ulus to inno- trivial.
vation and other positiv e externalities, so T his brings us back to the issue of Krug-
Krugm an s case is not as clear-cut as he man s methods and approa ch; as indicated
makes out. earlier, many of these argum ents are based in
form al econom ic models. These focus, in
particular, on situations in which increasing
Rigour M ortis?
returns and therefore im perfect com petition
Krugm an is widely acknow ledged to have are central. Increasing returns represent a
extended the bounda ries of form al econom ic com mon thread to the issues addressed by
modelling. He also makes clear his suppor t Krugm an and to the models developed to
in this respect for the professional standards cope with them . They represent, as well,
of the econom ics profession. Krugm an highly abstract form al models. He frequently
dem onstrates that form al econom ic models uses simple two-country or two-region mod-
can, given appropriate simplifyin g assump- els to develop his case. His argum ent as to
834 MA RT IN BODD Y

the bene ts indeed necessity for this were for example he juxtaposes sim ple form al
noted above. models with narrative accounts or simple
Krugm an s case for form al, mathematical empirical material. He also in places argues
econom ics as the holy grail of social sci- that while, theoretically, there may be a num -
enti c know ledge and his disparagem ent of ber of possible outc om es, empirical material
other form s of social science has been dis- dem onstrates that in practice this has not
cussed at length elsewhere and is not the key been the case. It is im portant to be aware of
issue here. It can simply be noted in passing this and to recognise that his contributions
that his case is true only by de nition. The and claim s rest only in part and som etimes
sociology of `scienti c know ledge tells us remotely on truth-claims ground ed in form al
that different disciplines develop their own models. As M artin and Sunley conclude:
criteria as to what is acceptable and have
different ways of representing the world, It is perhaps less the speci c results of
with different criteria of evidence, proof and Krugm an s analyses that are im portant for
objectivity. That different branches of eco- econom ic geography than the general
nom ics have made major contributions both stim ulus they provide for further enquiry
to our understanding of and practical engage- (Martin and Sunley, 1996, p. 285).
ment with the world clearly goes without Krugm an s own perspective, expressed
saying. There may also be considerable pro- som ewhat more soberly in his most recent
fessional satisfaction to be had from form al com mentaries addressed to the econom ic
mathematical econom ics. Krugm an clearly mainstream (Krugm an, 1998a, 1998b) , ac-
and presum ably deliberately missed the know ledges as much. T he new geographical
point in talking about models and metaphors. econom ics has opened the door to analytical
Social science across the board is engaged in discussion within econom ics of interesting
a process of abstraction from `reality a and im portant issues and the new models are
process of model buildin g. It is simply that novel and engaging but, he asks:
the criteria applied from different perspec-
tives are different. A key issue, however, is are they relevant An unfortu nate feature
whether the sacri ces and simpli cations re- of much of the new theorising since the
quired by the ideals of form al modelling are 1970s is that it has failed to lead to much
simply such that at tim es they negate its validating empirical work (Krugm an,
utility other than as an elegant intellectual 1998a, p. 172).
exercise.
There are clues but the results are so far
Krugm an him self, moreover, is not the
inconc lusive. The new geographic al econom -
consistent purist which might be im plied
ics, then,
from his own discussion of method and his
suppor t for form al econom ic modelling. His has been more successful at raising ques-
approach is in fact more varied and in fact tions than answering them , better at cre-
the broader relevance and value of his work ating a language with which to discuss
derives, in part, from this. His discussion of issues that at creating the tools to resolve
international trade and new trade theory, for those discussions (Krugm an, 1998a,
example, rests in part on these developm ents p. 173).
in modelling to which he has him self con-
tribute d. He also, however, draw s on inter- It is nevertheless better, he argues, to have an
national econom ics, national incom e unresolved discussion about the spatial as-
accounting and `simple arithm etic to make pects of the econom y than to ignore them.
his point. He draws as well, in som e cases,
on empirical inform ation to make or back up
6. City Growth
his case. Here and elsewhere in his work
on the concentration of econom ic activity, This nal section takes the discussion be-
GE OG RAPH ICA L ECO NO MICS 835

yond the work of Krugm an him self to con- higher wages to the same workers in urban
sider a broader range of studie s within the areas re ects the higher productivity of
econom ics literature focusing in particular on workers in such areas. This is so even though
urban agglom eration, externalities, inno- the higher costs of urban living offset the
vation and city grow th. They are generally bene t of higher wages to workers them -
ground ed in much the same area conceptu- selves. W orkers may also bene t from non-
ally, drawing on `new growth theory , `new work-related factors such as a more active
trade theory or the `new econom ic geogra- social environ ment facilitated by proxim ity.
phy . They represent, however, a som ewhat
more empirical approach to the issues around
urban agglom eration and city growth. Agglom erative Effects
Empirical studies drawing on the new geo- Minim ising the cost of moving goods has
graphical econom ics and new grow th theory historic ally been the key factor in the con-
have generally tried to relate growth at na- centration of manufacturing in urban areas,
tional or metropolitan scale to a range of offering market access to residents and more
postulated positiv e and negative in uences widely from transport hubs. Glaeser, like
by means of cross-sectional regression analy- Krugm an, emphasises that:
ses. As Krugm an observes, however, these
have generally failed to offer much in the If there weren t xed setup costs or in-
way of direct testing of the speci cs of the creasing returns of som e sort, factories
models generated by these theoretical devel- would subdivide and locate throug hout
opm ents. The work of Glaeser and associates space to be close to every consum er
com es closest perhaps (Glaeser et al., 1992; (Glaeser, 1998, p. 144).
Ades and Glaeser, 1995; Glaeser, 1998).
Reviewing a wide body of econom ic It is the combinatio n of transport costs and
analysis focusing in particular on the costs increasing returns which generates agglom er-
and bene ts of urban agglom eration and the ation. The geographical concentration of
im plications of this work for the future of the manufacturing has tended to reduce with the
city. Glaeser (1998, p. 140) observes that lowering of transport costs and a lessening of
from an econom ic perspective, simply put, increasing returns with more custom ised pro-
duction and greater inform ation content.
a city is just a dense agglom eration of New manufacturing plants still tend to locate
people and rms. All of the bene ts of near suppliers and markets but the effect is
cities, the contribution of cities to produc- very small (Glaeser et al., 1995). The costs
tivity, com e ultim ately from reduced trans- of urban location for most manufacturing are
port costs for goods, people and ideas. now much higher than the bene ts.
Dense urban areas also m inim ise the fric-
The costs of cities include health costs, pol- tions of distance between people. Access to a
lution, congestion and com muting costs, large m arket allow s individuals to specialise
crime and social problems. T he future of more, generating increasing returns from a
cities depends on the balance betw een the ner divisio n of labour. City size thus allow s
bene ts and costs of urbanisation and how for greater specialisation. They also perm it
these change in the future. labour pooling , as argued by Krugm an. The
The bene ts of urban agglom eration are presence of many alternative employe rs also
dem onstrated by the preference of individ u- increases the bargaining power of workers;
als and rms for this as a location pattern but this in turn encourages them to invest more
also by the wage prem ium paid to workers in in their ow n hum an capital in the know ledge
urban areas. W ages are positively related to that they will be able to reap the return on
city size, other things equal (Glaeser and this. Higher levels of unemploym ent in urban
Mare, 1994). The willingn ess of rms to pay areas might mitigate this but may re ect
836 MA RT IN BODD Y

the fact that cities may attract people who are concentration but with local com petition fos-
prone to unem ploym ent. Dum ais et al. tering innovat ion and rapid adoption. Jacobs
(1996) suggest that labour pooling has been (1969) saw local com petition as im portant
the most im portant determinant of which in- but argued that innova tion was stim ulated by
dustries locate together over the past 20 variety and diversity in geographically con-
years. Gains, moreover, from specialisation centrated industries rather than narrow
and from insuring against rm- or industry- specialisation. These different models all
speci c shocks are probably increasing stress the role of `dynam ic externalities
(Glaeser, 1998). The costs of moving people speci cally, know ledge spillovers for city
are also likely to be increasing with longer grow th.
journey tim es such that reducing the distance
According to these models, cities grow
betw een people is also likely to be increas-
because people in cities interact with other
ingly im portant.
people, either in their own or in other
Glaeser (1998) also argues, starting with
sectors, and learn from them . Because they
the exam ple of Silicon Valley and Stanford,
pick up this know ledge without paying for
that the geographical proxim ity created in
it, these know ledge spillov ers are external-
urban areas allows ideas to travel more read-
ities. T he frequency of interaction with
ily and rapidly both by form al and inform al
other people is ensured by their proxim ity
contact and by workers shifting rms (the
in a city. Because this proxim ity makes
latter, althoug h Glaeser is not speci c on
externalities particularly large in a city, all
this, presum ably because proxim ity between
the models predict that cities grow faster
rms increased the probability of their gain-
than rural areas (Glaeser et al., 1992,
ing useful know ledge through recruitm ent).
p. 1130).
Earlier, it had been argued that:
Exam ining the evidence for these alternatives
If geographical proxim ity facilitates the
by means of an empirical study of US cities,
transm ission of ideas, then we should ex-
Glaeser et al. concluded rst that industries
pect know ledge spillov ers to be particu-
grow more slowly in cities in which they are
larly im portant in cities. After all,
heavily over-represented. Secondly, indus-
intellectual breakthroughs must cross hall-
tries also grow faster in cities in which the
ways and streets more easily than oceans
rms in those industries are smaller than
and continents (Glaeser et al., 1992,
average size. If a smaller rm structure is
p. 1127).
taken to be more com petitive, then this may
Jacobs (1969) and Lucas (1988) had also be stim ulating the spread of know ledge and
linked city grow th and the localised spread of grow th in line with Porter and Jacobs (al-
ideas. Jaffe et al. (1993) show that new though Glaeser et al. are ignorin g other poss-
patents are more likely to cite a prior patent ible connections betw een rm size and
that is close spatially. grow th rates, such as the age and maturity of
Two issues have been raised by alternative rms). Thirdly, city industries grow faster
models of know ledge transfer and inno- where the rest of the city is more diverse.
vation: rst, whether diversity or industrial This is com patible with Jacobs case that
concentration is more conduciv e to growth; diversity stim ulates know -ledge spillov ers,
and, secondly, whether a com petitive, small- innovat ion and grow th.
rm structure fosters innovat ion or whether T he fact that, in practice, many cities are
larger rms with a degree of local monopoly specialised in a few industries, it is sug-
encourage innova tion because the bene ts gested, re ects the fact that externalities
can be internalised and captured. Rom er other than know ledge spillov ers (which
(1986) in line with Marshall s and Arrow s re ect diversity) are more dom inant
earlier work, argued for concentration and Marshall s labour pooling or specialised in-
local monopoly ; Porter (1990) argued for puts and services. These `static spillovers
GE OG RAPH ICA L ECO NO MICS 837

can account for city specialisation as such. lim it on the size of cities (Glaeser, 1998).
Knowledge spillov ers, how ever, are essen- Housing and com muting costs com monly
tially `dynam ic spillovers , linked to inno- represent a trade-off and should therefore be
vation and grow th. Glaeser et al. thus considered toge ther. T he cost of living in
disting uish between static spillovers which general and of housing costs in particular rise
can account for the degree of specialisation with city size, re ecting willingne ss to pay
and dynam ic spillovers which can account for the bene ts of urban living. Average
for growth. As Glaeser observes in a later com muting tim es also increase with city size
paper, however, the evidence is inconc lusive representing a real cost and this cost in-
and: our belief in the intellectual role of creases as the value of tim e rises. The differ-
cities com es mainly from case studies and ence between cities of 100 000 popula tion
anecdotes rather than overwhelm ing hard and 1 million may be in the order of $1000
evidence (Glaeser, 1998, p. 147). Apart annually (Glaeser, 1998, p. 151).
from know ledge spillov ers, urban agglom er- In term s of polluti on, there is little corre-
ation and higher urban density could also lation between city size and, for example,
operate throug h accelerating learning throug h sulphur dioxide and ozone. Particulate levels
interaction and hence hum an capital accumu- do rise with city size. T he cost in econom ic
lation. E vidence suggests that younge r work- terms, how ever, is estimated to be small
ers learn faster in cities and thus experience com pared with com muting costs and falling
faster wage growth (Glaeser, 1998, pp. 148 with changes in technology, emission con-
149). trols and declining urban manufacturing.
It has been argued that inform ation tech- Those in small tow ns suffer almost the sam e
nology will reduce pressures for agglom er- overall level of pollutio n as those in big
ation by making face-to-face contact cities (Glaeser, 1998, pp. 151152).
obsolete. Cities will continu e, however, to Reported crime and victim studies show
facilitate unplann ed inform ation ows and crime rates across a range of offences to be
com binations of ideas, which IT generally positively correlated with city size (Glaeser
does not. Increasing ef ciency and density in et al., 1996). T he urban crime effect, it is
electronic com munication may actually in- argued, seems to re ect the higher returns to
crease the overall num ber of relationships crime in urban areas due for example to
and in fact increase face-to-face contact access to potential victim s, a low er likeli-
overall. Telecommunic ations does not appear hood of arrest and a greater concentration of
to be a substitute for face-to-face contact. individ uals in poverty who are more likely to
People in cities are more likely to use telecom- engage in criminal activities. Again, the in-
munications; those who are close physically creased costs of higher crime rates in urban
are more likely to com municate; business areas and assumed willingne ss to pay to
travel has increased over tim e, even holding avoid crime can be estim ated. T he urban
prices constant. High technology, with cut- effect seem s to be weakening, but the im -
ting-edge com munications technology, is portanc e placed on it to be increasing.
characterised by a high degree of clustering Cities are also characterised by higher
(although this could re ect other factors off- concentrations of poverty . Cities may them -
setting the decentralising potential offered by selves cause poverty through greater segre-
IT). Overall, rising dem and for contact and gation in cities, for example, or the nature of
inform ation may further increase the dem and social networks and interactions. Recent mi-
for urban agglom eration and cities. grants to cities, however, tend to be as poor
as long-term residents. T his suggests that the
concentration of the poor in cities is due to
Congestion Effects
the higher rate of movem ent of the poor into
From an econom ic perspective, it is conges- cities cities attract poor people. Cities it
tion costs which play a key role in placing a is argued are particularly attractive to the
838 MA RT IN BODD Y

poor apparently because of transport costs, handle externalities and redistribu tion which
the public goods bundle and social net- are more ef cient in the longe r run. L and-use
works (Glaeser et al., 1997, p. 154). Public and buildin g regulations are com monly un-
transport, welfare spending and public hous- connected to clear costs and bene ts of exter-
ing differentially attract the poor. Cities also nalities. Finally, they have in som e cases
repel the rich, who are willing to pay to support ed local redistribution al policies and
avoid crime and poore r public schools. Over- high welfare spending; these are inef cient
all, this may tend to reduce urban popula- in terms of redistribution and can distort the
tions. T his, however, seems to be an effect locational choices of both rich and poor
of older cities rather than newer, grow ing attracting the poor and repelling the rich.
cities such that there may be a life-cycle This tendency has declined more recently
effect. And many cities are not affected in re ecting voter preferences. Overall, Glaeser
this way. (It may also re ect structural (1998, p. 157) argues that:
change in historical patterns of industry and
City gove rnm ents should avoid redistribu-
employm ent.)
tion, most quantity controls, and white
Overall, Glaeser concludes that agglom er-
elephant infrastructure proje cts, such as
ation econom ies will continu e to be large;
subw ays. National governments should
inform ation spillovers in particular will re-
follow a policy of spatial neutrality, nei-
main im porta nt; and telecommunic ations
ther arti cially supporting nor taxing cities
may increase rather than eliminate dem and
disprop ortionately.
for face-to-face contact. Considerable costs
remain in urban areas. Place-based policie s such as enterprise zones
create spatial distortions that lim it the
The likeliest possibility is that the future
healthy tendency of the poor to exit declining
will be bright for the relatively hom oge-
areas (p. 157). A city which wishes to
neous and low density agglom erations of
prospe r, should focus on pricing externalities
the western United States, which can offer
correctly, protecting property rights and en-
many of the econom ic advantages of ag-
suring hum an capital develops within its
glom eration while also reducing the costs
borders (p. 158).
of congestion and crime. The poorer, more
heterogeneous, older cities, how ever, will
have more of a struggle. The future may Innovation and Urban A gglom eration
be one where very specialised agglom era-
Krugm an s sidelining of know ledge
tions lled with very particular incom e
spillovers has also been challenged by Au-
groups are the norm (Glaeser et al., 1997,
dretsch (1998) who argues that econom ic
p. 157).
activity based on know ledge has a high
This in itself may reduce innovat ion by re- propensity to cluster in order to maxim ise the
ducing unplann ed contacts betw een diverse production and transm ission of ideas. This,
individ uals, coupled with increased segre- he maintains, explains the apparent paradox
gation along incom e and racial lines. between increasing localisation of econom ic
Arguing from an econom ic perspective, in activity in an era of increasing globalisation
policy term s, local governments have fre- and highly ef cient com munication. Globali-
quently exacerbated the problems of cities. sation and the inform ation and com munica-
They have tended to favour large, prestige tions revolution, he argues, have underm ined
projects such as subway or freeway systems the com parative advantage of high-cost
where the social surplus generated fails to countries in traditional, moderate-technology
cover the cost. Instead, cities need to price industries. T heir com parative advantage is
congestion effectively. T hey have favoured increasingly based on know ledge and inno-
quantity controls (such as zoning or rent vation which can suppor t higher wages,
control) rather than taxes and transfers to re ecting the grow ing dem and for new prod-
GE OG RAPH ICA L ECO NO MICS 839

ucts and services. T his is seen in the rapid ation. By contrast, know ledge is vague,
increase in patents and the rising dem and for dif cult to codify, and often only serendip-
skilled workers. Ef ciency in the trans- itously recognised. W hile the marginal
mission of know ledge and the innovation cost of transm itting inform ation across
process depends on spatial proxim ity. The geographic space has been rendered in-
shift in com parative advantage in high-w age variant by the telecommunications revol-
countries thus helps to explain the grow ing ution, the marginal cost of transm itting
im portance of spatial clustering. know ledge, especially tacit know ledge,
rises with distance (Audretsch, 1998,
High wages are increasingly incom patible
p. 21).
with inform ation-based econom ic activity,
which can easily be transferred to a lower He refers to von Hipple s argum ent that
cost location. By contrast, the creation of highly context-speci c, uncertain know -
new ideas based on tacit know ledge can- ledge what von Hipple calls `sticky know -
not easily be transferred across distance. ledge is best transm itted via face-to-face
Thus, the com parative advantage of the interaction and through repeated and frequent
high-co st countries of North America and contact. Fortune magazine (1993) similarly
W estern Europe is increasingly based observed that business is a social activity,
on know ledge-driven innova tive activity and you have to be where im porta nt work is
(Audretsch, 1998, p. 26). taking place . A survey of top executives had
ranked the Raleigh/Durham area with its
Knowledge spillovers from rms and univer-
research universities as the best city for
sities, it is argued, are essential to innova tive
know ledge workers and innovat ive activity:
activity. Such know ledge spillov ers tend to
be spatially restrictive. Thus: US businesses, especially those whose
success depends on staying at the top of
Even as the relevant geogra phic markets
new technologies and processes, increas-
for most goods and services becom e in-
ingly want to be where hot new ideas are
creasingly global, the increased im port-
percolating. A presence in brain-pow er
ance of innovative activity in the leading
centres like Raleigh/D urham pays off in
developed countries has triggered a re-
new products and new ways of doing busi-
surgence in the im portance of local re-
ness (Fortune, 1993).
gions as a key source of com parative
advantage (Audretsch, 1998, p. 26). Krugm an, as noted earlier, had cited the in-
visibility of know ledge ow s as making
There are close parallels, here, with work on
them im possible to model or measure (and t
know ledge, innovation and the `learning re-
only for sociologists). As Audretsch notes,
gion in econom ic geography, urban and re-
how ever, there is within the econom ics
gional studies which approach the issue from
literature as well, an emerging body of work
a more sociological or social econom ics per-
aimed at measuring the extent of know ledge
spective. Thus Audretsch argues that:
spillovers and linking them to the geography
The resolution to the paradox posed by the of innova tion (Jaffe, 1989; Jaffe et al., 1993;
localisation of know ledge spillov ers in an Feldm an, 1994; Audretsch and Feldm an,
era where the telecommunications revol- 1996). Jaffe et al. (1993) had found that
ution has drastically reduced the cost of patent citations occur more frequently within
com munication lies in a distinction be- the state in which they were patented than
tween know ledge and inform ation. Infor- outside. The propensity of innova tive activity
mation such as the price of gold on the to cluster geographically is also greater in
New York Stock E xchange or the value of industries such as biotechnology where new
the yen in London, can be easily codi ed, know ledge plays a more im portant part
and has a singular meaning and interpret- particularly tacit know ledge (Audretsch and
840 MA RT IN BODD Y

Feldm an, 1996; Audretsch and Stephan, for new ideas prom otes innova tion more than
1996; Prevenzer, 1997). Using city-level data local monopoly . A key conclusion from this
for the US, Audretsch argues that a close and Glaeser et al. s work is that innovative
relationship can be seen betw een the avail- activity requires more than simply factor en-
ability of know ledge resources in a city and dow ment in the form of know ledge inputs:
its innovat ive performance (Audretsch,
T he underlying econom ic and institutional
1998, p. 23). He goes on to suggest that
structure matters, as do the microeconom ic
know ledge spillovers, particularly from uni-
linka ges across agents and rms (Aud-
versities, are a particularly im portant input to
retsch, 1998, p. 26).
innova tion by small rms, com pared with
larger rms where the process is typically This com es much closer to work from a
more internalised. Spillovers, and clustering, sociological, geogra phical or urban and re-
are also likely to be more marked in the early gional studie s perspective than Krugm an s
stages of an industry life-cycle when inno- sidelining of know ledge externalities and
vation is more active (Audretsch and Feld- their study.
man, 1996). Spatial agglom erations may also T he increasing im portance of know ledge
however, be subject to `technological lock- and innova tion to com parative advantage has
in as industries mature, im peding the gener- also been re ected in policy terms. Deregula-
ation and transm ission of new know ledge tion and privatisation have been ac-
and restricting innova tion an additional com panied by less emphasis on com petition
form of `congestion disbene t . policy (anti-trust and monopoly measures).
This work has successfully identi ed the Policy has increasingly focused on enabling
role of know ledge spillovers and spatial clus- the creation and com mercialisation of know l-
tering from an econom ic perspective and has edge including R&D, venture capital and
shifted attention in looking at innovation start-up suppor t. At the same tim e, as central
from the rm to the urban area or geograph- government s role has reduced, Audretsch
ical region. As Audretsch notes, however, suggests, policy initiatives have increasingly
there is little consensus as to exactly how and shifted to the state, regional or more local
why this occurs and the im pact if any of the levels, including suppor t for localised tech-
way in which econom ic activity is organised nology development and innovat ive clusters
in space. He refers to sociological studies (Sternberg, 1996) or, as Audretsch puts it,
for example, Saxenian s work on Silicon the recreation of Silicon Valley.
Valley which focus on the processes of W hile the gulf between much of this work
com munication between individ uals, the role from an econom ic perspective and other
of networks and of culture. Such work, he work on know ledge and innova tion within
acknowledges, suggest the lim itations in- the social sciences is less than with Krug-
herent in attem pts to model the `know ledge man s work, there are nevertheless lim ita-
production function . He notes as well the tions. As the above reference illustra tes, the
econom ist s avoidance of cultural differences im age of the know ledge-based industrial
as explaining grow th or technological devel- cluster often appears to be the iconic and, as
opm ent. There are links, however, as he also Saxenian s own work indicates, essentially
notes, with the work of Glaeser et al. referred atypical examples of high technology in Sili-
to earlier, on the im pacts of specialisation con Valley or Route 128. To cite these as
versus diversity and com petition versus local corroborating evidence for the generalised
monopoly . Forthcoming work by Feldm an role of localised, know ledge-based spillov ers
and Audretsch (referred to in Audretsch, and innovation in econom ic growth is far
1998, p. 26) suggests that diversity across from convincing. Furthermore, while the
com plem entary activities sharing a com mon work of Audretsch, Glaeser et al. and others
science base is m ore conduc ive to innovation argues strongly for the practical role of lo-
than specialisation and that local com petition calised know ledge transfer and innova tion,
GE OG RAPH ICA L ECO NO MICS 841

they ignore a range of othe r empirical work F EL DM AN , M . (1994) Know ledge com plem entar-
ity and innovati on, Small Business Econom ics,
which nds little evidence for such localised
6, pp. 363372.
effects (for example, Sim mie, 1997). In part, Fortune (1993) The best cities for know ledge
this stems from their methodology. T hey w orkers, 15 Novem ber, p. 44.
start by assuming know ledge spillov ers on F UJIT A , M . and K RUG MA N , P. (1995) W hen is the
conceptual ground s. T hey then look for em- econom y m onocentr ic: von Thu nen and Cham -
berlin uni ed, R egional Science and Urban
pirical evidence to corroborate this. E mpiri-
E conom ics, 25(4), pp. 505528.
cal evidence, however, is frequently based on F UJIT A , M. and M ORI , T . (1996a) Structura l stabil-
correlation between supposedly relevant vari- ity and the evolutio n of urban system s, R e-
ables such as innova tion in the form of gional Science and U rban Econom ics, 27(45),
patents cited and city size, rm size or local pp. 399442.
F UJIT A , M . and M ORI , T. (1996b ) T he role of ports
industrial structure. The resulting correla-
in the m aking of m ajor cities: self-aggl om er-
tions may be consistent with or contradict ation and hub-eff ect, Journal of Developm ent
posited in uences on know ledge transfer and E conom ics, 49(1), pp. 93120.
innova tion. They could equally well re ect G L AES ER , E. (1998) Are cities dying?, Journal of
processes other than know ledge externalities E conom ic P erspectiv es, 12, pp. 139160.
G L AES ER , E . and M ARE , D . (1994) Cities and
and the supposed im pacts of these in terms of
skills. W orking Paper E-9411, H oover Insti-
innova tion and, in turn, econom ic grow th. tution.
The com plexity of the possible causal rela- G L AES ER , E ., K A LLA L , H., S CH EINK M AN , J. and
tions is such that it resists unpacking simply S H LE IFE R , A. (1992) Grow th of cities, Journal
by establishing statistical correlations be- of P olitical E conom y, 100, pp. 11261152.
G L AES ER , E ., S AC ER DOT E , J. and S CHE INKM AN , J.
tw een a small num ber of supposedly key
(1996) Crim e and social interact ions, Quarterly
variables. Greater cross-fertilisation between Journal of Econom ics, 111(2), pp. 508548.
econom ic approaches and the rest of social G L AES ER , E ., S C HEIN KM AN , J. and S HLE IF ER , A .
science might prove productive in this (1995) E conom ic grow th in a cross-sec tion of
respect. cities, Journal of M onetary Econom ics, 36(1),
pp. 117143.
H AR RIS , C. D . (1954) The m arket as a factor in the
localisat ion of product ion, A nnals of the A s-
References
sociatio n of A merican Geograph ers, 44,
A DES , A. and G LAE SE R , E . (1995) T rade and pp. 315348.
circuses: explaini ng urban giants, Quarterly H IP PLE , E . VO N (1994) S ticky inform ation and the
Journal of E conom ics, 110(1), pp. 195228. locus of problem solving: im plication s for
A UDRE TS CH , D. (1995) Innovati on and Industry innovat ion, M anagem ent Science , 40, pp. 429
Evolution . Cambridge, M A: M IT Press. 439.
A UDRE TS CH , D. (1998) Agglom eration and the H IR ST , P. and Z EIT LIN , J. (Eds) (1989) R eversing
location of innovat ive activity , Oxford R eview Industri al D ecline ? Oxford: Berg.
of E conom ic P olicy, 14(2), pp. 1829. I SAR D , W . (1956) Location and the Space-
A UDRE TS CH , D . and Feldm an, M . (1996) R&D econom y. Cam bridge, M A: MIT P ress.
spillover s and the geograp hy of innovati on and J ACO BS , J. (1969) The E conom y of Cities. N ew
producti on, A merican E conom ic R eview , 86(4), Y ork: Random House.
pp. 253273. J AFFE , A . (1989) Real effects of academ ic re-
A UDRE TS CH , D . and S T EP HA N , P. (1996) Com - search, A merican Econom ic R eview , 79,
panyscienti st location al links: the case of bio- pp. 957970.
technolo gy, A merican E conom ic R eview , 86(4), J AFFE , A., T RAJT ENB ER G , M. and H E NDE RSON , R.
pp. 641652. (1993) Geograph ical localizat ion of know ledge
C OO KE , P . (1990) M anufactu ring m iracles: the spillove rs as evidence d by patent citations ,
changing nature of the local econom y, in: M . Q uarterly Journal of E conom ics, 63, pp. 577
C A M PBEL L (E d.) L ocal E conom ic P olicy, 598.
pp. 2542. London: Cassell. J OHNS TO N , R. (1992) Review of P. Krugm an,
D UM AIS , G ., E LL IS ON , G. and G L AES ER , E .. (1996) G eography and Trade, Environm ent and P lan-
Geograph ic concentr ation as a dynam ic process ning A , 24, p. 1006.
(m im eograph ). K NO X , P. and A GN EW , J. (1994) The Geograph y of
D YM SK I , G. (1996) O n K rugm an s model of econ- the W orld E conom y, 2nd edn. L ondon: E dw ard
om ic geograp hy, Geoforum , 27, pp. 439452. A rnold.
842 MA RT IN BODD Y

K OZU L -W RIGH T , R. and R OW THO RNE , R. (1998) P RE VE NZE R , M. (1997) T he dynam ics of industria l
Spoilt for choice? Multinati onal corpora tions clusterin g in biotechn ology, Small Business
and the geograp hy of internat ional product ion, E conom ics, 9, pp. 255271.
Oxford Review of E conom ic Policy , 14(2), R OM ER , P. (1986) Increasi ng returns and long
pp. 7492. term grow th, Journal of P olitical E conom y, 94,
K RUGM A N , P . (1991a) Geograph y and Trade . pp. 10021037.
Cam bridge, M A: M IT Press. S AXE NIA N , A. (1990) Regional netw orks and the
K RUGM A N , P. (1991b) Increasin g returns and resurgen ce of Silicon Valley, Californi a M an-
econom ic geograp hy, Journal of P olitical agem ent R eview , 33, pp. 89111.
Econom y, 99, pp. 483499. S AXE NIA N , A. (1994) R egional Advantag e:
K RUGM A N , P. (1993a) On the num ber and location Culture and Com petition in Silicon Valley
of cities, European E conom ic R eview , 37, and R oute 128 . Cambridge, M A: H arvard
pp. 23, 293298. U niversity Press.
K RUGM A N , P. (1993b) F irst nature, second nature S CO TT , A. and S T ORPE R , M . (1992) T he collectiv e
and m etropoli tan location , Journal of Regional order of exible product ion agglom eration: les-
Science , 33, pp. 129144. sons for local econom ic developm ent policy
K RUGM A N , P. (1995) Developm ent, Geograph y and strategic choice, E conom ic Geograph y, 68,
and E conom ic Theory. Cam bridge, MA : MIT pp. 219233.
Press. S IM M IE , J. (1997) Innovati on, Netw orks and
K RUGM A N , P. (1997) P op Internat ionalism . Cam- Learning R egions? London: Regional S tudies
bridge, MA : MIT Press. A ssociation .
K RUGM A N , P. (1998a) Space: the nal frontier , S TE RN BERG , R. (1996) T echnolog y policies and
Journal of E conom ic Perspectiv es, 12, pp. 161 grow th regions, Sm all B usiness Econom ics,
174. 8(2), pp. 7586.
K RUGM A N , P. (1998b) W hat s new about the new S TO RPER , M. (1992a) Regional `worlds of pro-
econom ic geograp hy?, Oxford R eview of Econ- duction : learning and innovat ion in the technol-
om ic Policy, 14(2), pp. 717. ogy districts of F rance, Italy and the USA ,
L UCA S , R. (1988) O n the mechanics of econom ic R egional Studies , 27, pp. 433455.
developm ent, Journal of M onetary E conom ics, S TO RPER , M. (1992b) T he lim its to globalisa tion:
22, pp. 342. technolo gy districts and internati onal trade,
M ART IN , R. and S UNL EY , P. (1996) P aul Krug- E conom ic G eography , 68, pp. 6093.
man s geograp hical econom ics and its implica- S TO RPER , M. (1997) T he Regional W orld . N ew
tions for regiona l develop ment theory: a critical Y ork: Guilford .
assessm ent, Econom ic G eography , 72, pp. 259 T Y SO N , L. (1992) W ho s Bashing W hom ? T rade
292. Con ict in High Technolog y Industri es. Insti-
P O RTE R , M . (1990) The Competitive Advantag e of tute for Internat ional Econom ics, W ashingto n,
Nations. New York: F ree Press. D C.
P R ED , A. (1966) The Spatial Dynam ics of V E NAB LE S , A. (1998) T he assessm ent: trade and
US Urban-In dustrial G row th, 1800 1914 . location , O xford R eview of E conom ic P olicy,
Cam bridge, M A: M IT Press. 14(2), pp. 16.

También podría gustarte