agglomeration and economic development import substitution vs. trade liberal is at ion

8/8/2019 Agglomeration and Economic Development Import Substitution vs. Trade Liberal is at Ion

1/21

Agglomeration and Economic Development: Import Substitution vs. Trade Liberalisation

Author(s): Diego Puga and Anthony J. VenablesSource: The Economic Journal, Vol. 109, No. 455 (Apr., 1999), pp. 292-311Published by: Blackwell Publishing for the Royal Economic SocietyStable URL: http://www.jstor.org/stable/2565936

Accessed: 27/08/2009 03:15

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/action/showPublisher?publisherCode=black.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the

scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that

promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

Royal Economic Society andBlackwell Publishing are collaborating with JSTOR to digitize, preserve and

extend access to The Economic Journal.

http://www.jstor.org
http://www.jstor.org/stable/2565936?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/action/showPublisher?publisherCode=blackhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2565936?origin=JSTOR-pdf


2/21

The Economic ournal, 109 (April), 292-311. ?) Royal Economic Society 1999. Published by BlackwellPublishers, 108 Cowley Road, Oxford OX4 1JF,UK and 350 Main Street, Malden, MA 02148, USA.

AGGLOMERATION AND ECONOMIC DEVELOPMENT:IMPORT SUBSTITUTION VS. TRADELIBERALISATION*

Diego Puga and AnthonyJ. VenablesThis paper analyses a model of economic development in which international differences inindustrial structure and income are caused by the agglomeration of industry in a subset ofcountries. Economic development may not be a gradual process of convergence by allcountries, but instead involve countries moving sequentially from the group of poor countriesto the group of rich countries. The role of trade policy in promoting industrialisation isstudied. While both import substitution and unilateral trade liberalisation may be 'successful'in attracting industry, they attract different sectors and welfare levels are higher under tradeliberalisation.

The key determinants of a country's economic development are usually takento be some combination of its factor endowment, technology, institutionalstructure and policy stance. While not denying the importance of theseconsiderations, in this paper we explore a radically different view of economicdevelopment and underdevelopment, based on the idea that economic activitymay agglomerate spatially. In this case it is possible that countries with similar,or even identical, underlying characteristics may nevertheless have differenteconomic structures and income levels. Economic underdevelopment is amanifestation of the spatial pattern of agglomeration, and development occursas this pattern changes, with industry spreading from existing concentrationsto new ones.

Analysis of spatial agglomeration of industry has been formalised in recentwork in economic geography (see for example Krugman and Venables (1995),Puga (1998)), and the goal of the present paper is to draw out the implicationsof this approach for economic development. What forces are conducive to thespatial concentration of industry, and what to its spread from country tocountry? If industrialisation does spread, what form does development take?What is the role of policy-in particular trade policy-in promoting industrial-isation?

The basis of our analysis is a model in which there are forces which maycause industry to concentrate in a few locations. We create these forces fromthree main ingredients. First, there are transport costs or other trade barriers,and these create incentives for firms to locate close to customers and tosuppliers. Second, firms have increasing returns to scale, which play the role offorcing firms to choose where to produce;1 of course, with increasing returns

* This paper was produced as part of the Programme on International Economic Performance atthe Economic and Social Research Council funded Centre for Economic Performance, London Schoolof Economics.1 Without increasing returns, if factors are uniformly distributed, every location can become anautarkic economy producing all goods at an arbitrarilysmall scale (Scotchmer and Thisse (1992)).[ 292 ]


3/21

[APRIL 1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 293we must handle the problem of market structure, and this we do by assumingmonopolistic competition. The third ingredient is the presence of input-output linkages between firms. These linkages create an incentive for firms tolocate close to other firms-their suppliers and customers. Krugman andVenables (1995) showed how this combination of forces creates the possibilitythat industry concentrates in one country, and established the dependence ofthe equilibrium on transport costs. Here we use the framework to study twoissues which we think illuminate the process of economic development.

The first is to consider the spatial implications of growth inn world manu-facturing relative to other tradeable sectors. This growth increases demand forlabour in established manufacturing countries, opening up a larger and largerwage gap between these and other countries. At some point this wage gapbecomes unsustainable, and industry starts to spill over to low wage economies.We analyse this process and establish that it does not lead to steady develop-ment of all low wage economies, but instead to rapid industrialisation ofcountries in turn. The logic of spatial agglomeration implies that developmentcannot proceed simultaneously in all countries. Instead there is a group of richcountries and a group of poor ones, and development takes the form ofcountries being drawn in turn out of the poor group, and taken through aprocess of rapid development into the rich group. We think that this is aninsightful way of thinking about the spread of industry in a number ofcontexts, for example from Japan to its East Asian neighbours.The second issue we address is the role of developing country trade policy inpromoting or hindering industrialisation. While recent papers in economicgeography have focussed on the location effects of reciprocal reductions intrade costs, in this paper we look at the effects of unilateral changes in tradebarriers. We show that either unilateral trade liberalisation or import substitu-tion policies may be used by the low wage economy to attract industry, butthese two policies work through very different mechanisms. Although they areboth superficially 'successful' in attracting industry, they have different effectson economic welfare, with trade liberalisation yielding higher welfare thanimport substitution policies. We analyse this in an aggregate model, and thenin a multi-industry variant of the model calibrated to South Korean input-output and demand data. We use the calibration to show the different sectoralimpacts of trade liberalisation and import substitution, and to confirm thedifferent welfare outcomes generated by the two policies.

Our approach in this paper can be thought of as a formalisation of earlierideas in development economics, in particular the role of forward and back-ward linkages, as emphasised by Hirschman (1958) and others. These linkagesare of no particular economic significance in a perfectly competitive environ-ment, but combined with the other ingredients sketched out above, they createpecuniary externalities between the location decisions of firms, and it is thisthat creates the incentives for agglomeration of industry. To see how thisworks, suppose that there is expansion of a downstream industry. This createsa backward linkage, expanding demand for intermediate goods, raising profitsin an upstream industry, and attracting entry of upstream firms, which in turn?) Royal Economic Society 1999


4/21

294 THE ECONOMIC JOURNAL [APRILmay decrease he price of intermediates. How does this perverse price responseoccur? Entry of firms may make the industry more competitive, squeezingprice cost margins and reducing price, or may lead to entry of more varieties,reducing a price index of industry output as a whole. This perverse priceresponse constitutes a forward linkage-expanding the upstream industryreduces the costs of the downstream. Putting this together, we see a positivefeedback, such that an expansion of the downstream industry makes theindustry more profitable, encouraging further expansion.

The process described above is also suggestive of 'cumulative causation'-the presence of more downstream firms attracts more upstream firms which inturn attract downstream firms and so on. Again this is reminiscent of oldtraditions in development and regional economics (for example in the work ofPerroux (1955), Myrdal (1957), Hirschman (1958), Harris (1954), and Pred(1966)), as well as some newer approaches to development economics.2

1. The ModelWe set out the model for the case of two countries and two sectors-manufacturing and agriculture-relegating a statement of the full multi-country and multi-industry model to the Appendix.

AgricultureEach country is endowed with quantities Li and Ki of labour and arable land(for countries i = 1, 2), both of which are internationally immobile. Agricul-ture is perfectly competitive and produces a homogenous output, which wechoose as nume'raire, nd assume costlessly tradeable.3 Its production function,F, is defined over labour and land, and has constant returns to scale. Ifmanufacturing employment in country i is denoted mi and the labour marketclears, then agricultural output is F(Li - mi, Ki) = Kif[(Li - mi)/Ki]. Thecountry i wage is

wi = f'[ (Li - mi)/Ki]. (1)

Manufacturing IndustryThe industrial sector is monopolistically competitive, producing differentiatedmanufactures under increasing returns to scale. As in Krugman and Venables(1995), we assume that the output of firms in the industry is used both as afinal consumption good and as an intermediate good for use in the same

2 Murphy et al. (1989) model a 'big push' in which increasing modern sector employment raisesaggregate demand, thereby increasing the profitability of modern sector firms. Their model worksthrough aggregate demand, rather than intermediate goods, so has no forward linkages. It also assumesa closed economy.3 Adding a trade cost in agriculture does not change the qualitative results of the model providingthat trade in agriculture occurs in equilibrium. For details of this see Fujita et al. (1999).?) Royal Economic Society 1999


5/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 295industry. The set of firms in each country is endogenously determined by freeentry and exit, and denoted by Ni. The cost function for an industrial firm(firm k) in country i is

Ci(k) =q-w(1_1)[a+ xi(k) (2)xi(k) is the firm's output, and the fixed and marginal input requirements, aand ,B, are the same for all varieties and all countries. The input is a Cobb-Douglas composite of labour, with share (1 - ,), and an aggregate of thedifferentiated industrial goods, with price index qi and share /t.4 This priceindex takes a CES form, so is defined by

qi-{J [pi(k)1(1-f) dk+J [ri pj(k)]( -f)d k} ,(- i, (3)kENi ke Nj

where pj(k) is the producer price of variety k produced in country j. Shipmentof these products is subject to iceberg trade costs: (ri units must be shippedfrom country j in order that one unit arrives in i). Product differentiation ismeasured by the elasticity of substitution between varieties of good, a, andcaptures the idea that firms benefit from access to a wider range of intermedi-ate goods (following Ethier (1982)).

DemandThere is a single representative consumer in each country, who has quasi-homothetic preferences over agriculture (the nume'raire) nd the CES aggregateof industrial goods. Hence there is a 'love for variety' on the consumer side, asin Dixit and Stiglitz (1977). The indirect utility of the consumer in country i is

Vi = q Y1(l-') (yi - eo) (4)where yi is income, and eo is the subsistence level of agricultural consumption.Notice that we use the same price index for varieties of industrial goods inconsumption as in production. This is not necessary for our results, but doesmuch to simplify analysis.

Each product is sold in each country, and the demand for variety k pro-duced in country i, xi(k), can be derived from (3) and (4) as

xi (k) = [pi (k) a[ei qa -1) +eq( -1) l(U1) (where ei is total expenditure on manufactures in country i. Since manufacturesare used both as final consumer goods and as intermediates, ei is given by;

ei = y/ wimi + Kif [(Li-mi)/Ki] - eo} + Ci(k) dk. (6)kENiThe first term in (6) is the value of consumer expenditure on manufactures,

4 In the full multi-industry model given in the Appendix each industry uses inputs from all otherindustries, with value shares given by the input-output matrix of the economy.?) Royal Economic Society 1999


6/21

296 THE ECONOMIC JOURNAL [APRILand the second the value of intermediate demand. In the braces, the first termis wage income in manufacturing, and the second is income generated inagriculture; the consumer devotes the first eo of income to agriculture, andproportion y of income above this level to expenditure on industrial products.The final term in (6) is intermediate demand, generated as firms spendfraction f of their costs on intermediates.

SupplyEach variety is produced by at most one firm, so the firm producing variety kfaces demand curves (5) and cost function (2). Since all products produced inlocation i are symmetric (they have the same technology and demand curves)we drop the label for individual varieties. The profits of a single representativecountry i firm are therefore

Zi = pixi - qi (l-) (a + /3xi). (7)Each firm's perceived elasticity of demand is a, so the equality of marginalrevenue to marginal cost necessary for profit maximisation takes the form

pi(1 - 1/ul) =/3q8wil-. (8)We choose units of measurement for output such that flu = a - 1, giving anequilibrium price of pi = q8wil u) Using this pricing rule in the definition ofprofits, it follows that firms break even if their sales are equal to level x* givenby

x* = a/a. (9)If a firm were to sell less than x* then it would make a loss, and more than x*,a profit. At equilibrium profits are exhausted by free entry and exit. Denotingthe number (mass) of firms in region i by ni #Ni, we therefore have

(xi-x*) ni = 0, xi x*, ni .>? (10)To complete characterisation of equilibrium we need only specify manufac-turing labour demand. The manufacturing wage bill, miwi, is fraction (1 -,u)of costs (equal, in equilibrium, to the value of output), so

Mi i= (1 -u)niCi = (1 -u)nipix*. (11)We have already seen how labour market clearing determines the wage rate(1).

EquilibriaEquilibria of the model are given by (1)- (11). What can be said about them?This question is most easily answered if we assume that the two economies havethe same factor endowments. We therefore set

L1 = L2 = A, K1 = K2 = A, (12)(? Royal Economic Society 1999


7/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 297(where the fact that the quantities of L and K are the same is just a choice ofunits).

There is certainly now a symmetric equilibrium in which industry is equallydivided between countries, although this equilibrium may not be stable. Moreinterestingly, there may also be equilibria in which manufacturing is concen-trated in a single country. To establish whether or not such an equilibriumexists, we assume that all manufacturing is concentrated in one country (saycountry 1) and then see if it is profitable for any firm to start production incountry 2. If not, then the hypothesised concentration of manufacturing incountry 1 is an equilibrium.

Let us assume then that n2 = 0. The price indices of (3) reduce toql = n, pi, q2 = n Plr2 (13)

Sales of each firm in country 1 are,X1 = p-a q(a-l) (el + e2) = X* (14)

where the first equation comes from using (13) in (5), and the second fromthe fact that country 1 industry equilibrium occurs when n1 has adjusted togives zero profits, so each firm sells output level xl = x*.Suppose now that a firm starts producing in country 2. Its sales are (from (5)and (13)),

X2= p-rq(a )[elr(-a) + e2r(a-l)] (15)Relative goods price can be derived from (7) and (13) as

(p2) T(2) (l-8 (16)so taking the ratio of the sales equations (14) and (15) we obtain,

__ _Wl__el e2 (a-1)12~)f\~U(l-1L)[el + .r21(17)X2 (W2 12f [el ('e2e + ee2 2 . (7

This expression provides the criterion that determines whether or not agglom-eration of industry in country 1 is an equilibrium. If the expression has valuegreater than unity then an entrant in 2 can sell more than is required to breakeven (x2 > x*), so agglomeration in 1 is not an equilibrium. Conversely, if theexpression is less than unity then concentration in 1 is sustainable-it is notprofitable for any firm to produce in country 2.

The magnitude of this expression is determined by three forces. First, thefactor market. The larger is wl/w2 the higher is x2 and hence the less likely itis that agglomeration can be sustained.5 Unsurprisingly, the larger are thewage differences associated with agglomeration, the more likely it is thatproduction in country 2 will be profitable.

5 The relativewage term enters with exponent a (1 - ,) because labour accounts for (1 - ,) of costs,and a price increase reduces sales according to elasticity a. Relative wages are endogenous, anddetermined in (18) and (19) below.(? Royal Economic Society 1999


8/21

298 THE ECONOMIC JOURNAL [APRILSecond, forward linkages. The term r-o# captures the fact that a firm setting

up in country 2 would have to import all its intermediate goods, and pay r2more for them than do firms in country 1. Transport costs or other barriers tocountry 2 imports (r2 > 1) make this term less than unity, reducing x2, andmaking it less profitable for a firm to start producing in country 2. Essentially,the term captures the forward linkages foregone by locating in country 2, awayfrom intermediate suppliers.

The third force is backward linkages, and these are captured by the term insquare brackets. To interpret this, suppose that Ti = t2 > 1 and that el + e2 isconstant. Since (r ) e2, making for a relatively small valueof this term. The effect therefore captures the backward linkages foregone bynot being close to industrial consumers.

Equation (17) has endogenous variables on the right hand side, but thesecan be found as follows. By construction, all of country 2's labour force is inagriculture, so m2 = 0. Country l's agricultural labour force must thereforeadjust to equate world supply and demand for agriculture, that is to satisfy,2eo + (1 - y) [ mlf'(1 - ml/A) + Af(1 - ml/A)

+Af (1) - 2eo] = Af (1 - ml/A) + Af (1). (18)The right hand side of this expression is food production and the left handside is demand, coming from the subsistence requirement of the representa-tive consumer in each country, eo, plus proportion (1 - y) of world income inexcess of this subsistence requirement. This equation gives ml as a function ofparameters of the model. We can then find the wage rates from (1)

wl = f '(1 -ml /A), w2 = f '(1), (19)and manufacturing expenditure levels,

el = y[wl ml + Af(1 - ml/A) - eo] + M,(el + e2), e2 = y[Af (1) - eo]. (20)(Derived from (6), noting that manufacturing costs equal total expenditure onmanufactures, and all are incurred in country 1).The dependence of x2/x* on some of the parameters of the model isillustrated by the lines in Fig. 1, in which the vertical axis gives A/eo and thehorizontal the country 2 import barrier, T2.6 Lines correspond to differentvalues of the input-output linkage, ,u, and each line is the locus along whichx2/x* = 1. We call this relationship the sustain curve, because it delimits theregion of parameter space where agglomeration is sustainable. Below thesustain curve x2/x* < 1, so that concentration of manufacturing in country 1 isan equilibrium, while above it we know that concentration is not an equili-

6 The figureis computedusingsolutionsof (18)- (20) in (17). Keyparameters re ,u,and a whichwe set at5, correspondingo a price- marginal ost markup of 25%.Fulldetailsof parameters sedinthisandsubsequent iguresaregiven n theAppendix.?) RoyalEconomicSociety1999


9/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 2991.7

1.6

1.5~ ~ ~ ~ ~ ~~~~~~~~~~~~.1.41.31.2

1.1 \= 04 0.45

1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2't2

Fig. 1. SustainableAgglomeration

brium (it is profitable for a firm to start production in country 2). As theeconomy passes through this line a bifurcation occurs-the qualitative struc-ture of equilibria changes-and in remaining sections of the paper we explorethis transition.Fig. 1 provides a framework for the analysis of the rest of the paper, and insubsequent sections we shall discuss the shape of the x2/x* = 1 curve ingreater detail. In the next section we consider the effects of increasing A/eo,

and show what happens as we move upwards through the bifurcation set. InSection 4 we look at developing country trade policy-that is, at horizontalmovements across the figure. In both these sections we shall at various pointsmove to a more general model, with more countries and more industries.However, much of the intuition for our results comes from (17) and itsillustration on Fig. 1.

3. Growth and the Spread of IndustryWe now turn to investigating the implications of growth in the world economyas a whole for the spatial location of economic activity. Since we do not seek toexplain growth, we simply assume that exogenous technical progress augmentsthe productivity of all primary factors in all countries equally. What are theimplications of such technical progress for the location of manufacturingproduction?

A completely homogenous process of economic growth-raising supply and? Royal Economic Society 1999


10/21

300 THE ECONOMIC JOURNAL [APRILdemand for each sector in each country in the same proportion-will not haveany spatial effects. But if demand for manufactures rises faster than demandfor agriculture, then relative price changes will occur and, as we shall see, thiscan trigger industrial relocation.7 We capture a relatively rapid growth ofdemand for manufactures by using the linear expenditure system with apositive level of subsistence expenditure on agriculture, eo, so that growth ofhousehold income is associated with proportionately faster growth of demandfor manufactures.8

In terms of the model, we assume then that A increase through time, andreinterpret wi and mi as wages and employment of efficiency units of labour.Starting from a situation in which manufacturing is concentrated in country 1,we see from the expression for country 1 manufacturing employment, (18),that such growth causes an equiproportionate increase in ml only if eo = 0,while if eo> 0, then ml increases more than proportionately with A. Turningto the wage equations (19), if ml/A increases, then so too does the country 1wage per efficiency unit, both absolutely and relatively to the country 2 wage.

The implications of this for the location of industry can be seen from (17).The increase in wl/w2 raises x2, working against sustainability of the agglom-eration in country 1. But as the wage in country 1 increases, so does country 1'sshare of world expenditure on manufactures, el / (el + e2), tending to decreasex2 and make 1 a relatively more profitable location. The net effect depends onparameters of the model. If the share of manufactures in consumption, y, isvery small then the effect on wages will be small and the agglomeration willalways be sustainable. However, if y is large enough then the wage effect willcome to dominate, and at a high enough value of A the agglomeration willbecome unsustainable.9 This corresponds to the case illustrated in Fig. 1,where as A increases so the economy crosses the x2/x* = 1 locus.

When x2/x* = 1, entry of a manufacturing firm in country 2 is profitable.What then happens as A increases further? The presence of some manufactur-ing in country 2 creates forward and backward linkages (reduces q2 andincreases e2), but also has the effect of narrowing the wage gap between thecountries (raising w2/WI ). If the linkage effects are very powerful compared tothe wage effects then there may be discontinuous change-production incountry 2 becomes profitable enough that the two economies jump to thesymmetric equilibrium. Discontinuities are avoided if the wage effects arerelatively strong, in which case as A increases further so the two economiesconverge smoothly to the symmetric equilibrium.10To draw out the economics of the process we illustrate it not for our twocountry example, but by numerical simulation of the model for a world of four

7 Between 1960 and 1990 world value added in manufacturing increased fourfold while world GDPincreased threefold.8 Econometrically estimated values of eoare positive, see for example Lluch and Powell (1975).9 In the former case the curve is unbounded for some intermediate values of r2. Puga (1998)investigates further.10 Details of the nature of the bifurcation (whether or not it is discontinuous) are examined by Puga(1998) and Fujita et al. (1999) in similar models, and we do not pursue them here.

?) Royal Economic Society 1999


11/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 301identical countries. Fig. 2 plots real wages per efficiency unit of labour in eacheconomy relative to the average for all economies as A increases. The verticalaxis is this relative real wage per efficiency unit, wi, and the horizontal axis theexogenous technological progress parameter A.

At low A, all industry is in country 1, but growth in A causes an increase indemand for manufactures and hence a divergence of wages-the country 1wage reaching, for our parameter values, nearly one and a half times the levelof wages elsewhere in the world. Despite this wage gap it is not profitable forany firm to move out-the forward and backward linkages received by being incountry 1 compensate for the higher wage. But as pressure builds up incountry 1, so A reaches the bifurcation point at which production in the othercountries becomes profitable. Industrialisation commences in all of them, butas their volume of manufacturing increases so do the associated linkages andpecuniary externalities. There comes a point at which simultaneous industria-lisation in all of them ceases to be a stable equilibrium-if one got slightlyahead of the others then its lead would cumulate.11 This means that just oneof the countries (call it country 2) gains manufacturing; this country's wagepath is illustrated by the first dashed line, and we see very rapid convergence

. (County 1 real wage relative to average for_ countries)1.3 - - - (country realwage relative o average or all countries)_. - - 0()2 (country realwage relative o average or allcountries)(1)3 (country realwage relative o average oral lountries)I.2 - / .. C() (country realwage relative o average or all countries)1.2

(0) 0)2(M1 = 0= )3

............. .......0.9 l-)2 =03 =)4 -

0.8- I1 2 3 5 10

AleoFig. 2. WavesofIndustrialisation

11 Entry and exit of firms occurs in response to instantaneous profits, so the dynamic system isdn1/dt = k7ri,and stability is defined with respect to this system. Analysis of the stability of models ofthis type is undertaken in Fujita et al. (1999), and analytical results on the instability of simultaneousdevelopment by two economies are available on request from the authors.(C Royal Economic Society 1999


12/21

302 THE ECONOMIC JOURNAL [APRILwith country 1. Country 1 suffers both a relative and absolute real wage decline,as it loses a share of its manufacturing to the newly industrialised country.The other countries remain specialised in agriculture following this firsttransition, but continuing growth now starts to raise real wages in 1 and 2relative to these countries. This continues until another critical value isreached at which point industry spreads to a third country, and so on. What wesee then is industrialisation spreading, in a series of waves, from country tocountry. The model predicts that economic development is not a smoothprocess of many countries catching up with the rich. It is instead thecoexistence of a rich and a poor group of nations, but with growth of worlddemand for manufactures causing successive poor countries to join the richclub.

This, we think, provides a useful way of thinking about the spread of industryfrom Japan to several of its East Asian neighbours over the last three decades.Fig. 3 plots the share of manufacturing in total employment in Japan, Taiwan,South Korea, Malaysia and Philippines between 1965 and 1995. Throughoutthis period about 25% of Japanese workers have been employed in manufac-turing; Taiwan, South Korea and Malaysia have gained industrial employmentin that sequence, while the fraction of workers in manufacturing has remainedbelow 12% in Philippines.

0.4g 0.35 . - _0t 0.3 ,

Japan .,tO0.25 }Taiwan,/..* ..

u0.2/-.2 - _ _ zSouthKorea./i

0.15- :~~~~Malaysia,./ - '0.15

0 K -. - -* . Philippines0.05- - .

1965 1970 1975 1980 1985 1990 1995year Sources: eeAppendix

Fig. 3. WavesofIndustrialisation in EastAsia(? Royal Economic Society 1999


13/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 3033. Import Substitution vs. Trade LiberalisationCan the less developed country use trade policy to attract industry? Fig. 1suggests that the answer is affirmative-changing r2 can move the economyout of the region in which country 2 has no industry-and we now investigatethis in more detail.12

Consider first the effects of an import substitution policy, raising tradebarriers. r2 effects x2/ x* in two ways. An increase in r2 makes imported inputsmore expensive, this reducing the term t2 F/ and making it less attractive for afirm to establish production in country 2. But pulling in the opposite direc-tion, an increase in T2 switches country 2 expenditure on manufactures towardsproduction in country 2 (the term e22or-l ).13 Which of these effects is morepowerful? Letting r2 -* 00 we see that x2 - oo providing that there is somemanufacturing expenditure in country 2 (e2 >0) and that (o - 1) / o > Yi.We shall assume that this restriction on parameters is satisfied - without it wehave the curious result that even under autarky it is not profitable to set upproduction to meet local demand.14 In this case then, raising trade barrierscan always be successful in attracting industry (we return in a moment to seeinghow much industry and of what type).

What about trade liberalisation? Can a reduction in r2 cause industrialisationto commence? We see from Fig. 1 that the answer depends on the values ofA/eo and other parameters. We can get some more information on this bylooking at the derivative dx2/dr2. Differentiating (17) this is,

_dX2r2 (-1= 2u 1)1aul~U (21)12I

dT2 X2 =:(1-a) + (a-1) (1The derivative is positive for large enough r2 (providing (a - 1 - ay) e2 > 0),reflecting our earlier discussion about import substitution. The numeratorswitches sign as r2 becomes small enough, and it is this that generates thehump of the sustain curve illustrated on Fig. 1, and suggests that reducing r2will attract industry. However, it is not necessarily the case that this change inthe sign of dx2/dr2 occurs at r2 > 1. From inspection of (21) we see that this ismore likely the stronger are linkages (larger ,u increasing the forward linkagebenefits from trade liberalisation), and the larger is el/ e2 (reducing the valueof protecting the country 2 domestic market).

Although both import substitution and unilateral liberalisation can be effec-tive in attracting industry, they have different welfare implications. This isexplored in Fig. 4, which plots country 2 welfare as a function of its importbarrier, r2. The figure is computed from a two country numerical example,

12 We look only at unilateral changes in trade costs. A multilateral reduction in trade costs can causerelocation of industry to non-industrialised economies, as shown in Krugman and Venables (1995). In athree country set up, the relocation might not be to all countries simultaneously (see Puga andVenables (1997), who also study the effects of regional integration).13 None of the endogeneous variables in (17) depend directly on r2, as may be seen by inspection of

(18)-(20).14 This condition is standard in models of this type, and rules out unbounded agglomeration. Fujitaet al. (1999) have labelled it the 'no-black-hole' condition.(? Royal Economic Society 1999


14/21

304 THE ECONOMIC JOURNAL [APRIL1.05

0.950.9 ~~~~~~~~~~~~withitariffevenue

*0.85O.0I8

0.7510.75 ~~~~~~~~~~~~~~~ttariff revenue0.7 n? 20.65- * * -- i* ' -i' g1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Fig. 4. ImportSubstitutionvs. TradeLiberalisation

details of which are given in the Appendix, but two aspects of which need to beexplained here. First we assume that country 2 is quite small - one third of thesize of country 1 in endowments (and less in income); the reason is that wewant to think in terms of a developing country importing from the rest of theworld. Second, we split the trade barriers into natural and tariff barriers, andset the natural trade barriers at 15%, so define ?* = 1.15. Country 1 has nofurther barriers, so r, = r*, but country 2 has tariff barriers over and above r*.The horizontal axis on Fig. 4 is this extra country 2 barrier, r2 - r*, and thevertical is country 2 real income.15In the interval r- --* to 4+ r* there is no manufacturing in country 2,and these values of r2 are the two solutions of x2/x* = 1 from the sustaincurve. Import substitution draws industry in when r2>4r2,as does trade liberal-isation when r2


15/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 305Looking first at the case in which tariff revenue is included in welfare (the

upper curve), we see four main points. First, there is a range of tariffs withinwhich import substituting industrialisation raises welfare, but this range is oflimited width (between 4' and approximately T2 = 1.73), and higher tariffslead to welfare reduction. This arises from the trade-off between the beneficiallinkages created by industry, and the loss of gains from trade. Second, in theregion in which there is no manufacturing, r2 E (,r-, 4+), reducing the tariffraises welfare. This is simply because it reduces the distortion on importedmanufactures. Third, reducing the tariff further, into the region in whichcountry 2 gains industry (r2 < r-), increases the rate at which welfare increases,because of the linkage benefits that are being achieved. The overall message istherefore clear. While import substitution may be locally welfare raising, ityields lower welfare levels than industrialisation via unilateral trade liberal-isation.

If the tariff does not generate income for domestic consumers, then thelower line applies. The qualitative conclusions of the previous paragraph apply,with two quantitative qualifications. The gains from attracting industry byunilateral liberalisation are now larger-simply because of the real trade costsnow being saved. However, at trade barriers above 4r+ elfare is increasing inr2. And essentially trade barriers are so costly that-within this interval-thebest a country can do is go on raising the trade barriers to create furtherimport substitution and drive imports to zero.Fig. 4 is of course, just from a numerical example. How are things changedas parameters of the model change? The effects of stronger industrial linkages(higher ,u) can be seen from Fig. 1. Higher ,u has the effect of increasing theinterval of trade barriers within which agglomeration occurs (for a given valueof A/eo) raising the point at which import substitution commences, 4+, andreducing the point at which unilateral liberalisation causes industrialisation,'r-. However, it is noteworthy that the upwards shift of 4r+s much larger thanthe downwards shift of 'r-, indicating that strong linkages make inward-lookingindustrialisation more difficult. Reducing the size of country 2 has a similareffect, again with 4+ rising more than 'r- falls; in other words an importsubstitution policy is more difficult to implement the smaller is the economy.Raising developed country import barriers also shifts 4r+ p and 'r- down, butthe relative magnitude of the shift is now reversed; the fall in 'r- is large,reflecting the difficulty that developed country trade barriers create for anoutward looking development strategy.17

4. Trade Policy and Industrial StructureSo far we have assumed a single manufacturing sector. We now disaggregatethis in order to see how import substitution and trade liberalisation policiesaffect the industrial structure of the developing economy.

17 This point suggests that if developed countries reacted to import substituting policies by raisingtheir own tariffs,then the effectiveness of these policies would be reduced.(? Royal Economic Society 1999


16/21

306 THE ECONOMIC JOURNAL [APRILWe base our investigations on a five sector model, with input-output

structure aggregated from a South-Korean input-output matrix. Sector 1 is theaggregate of all primary sectors and is assumed to be perfectly competitive andtradeable. Sectors 2-4 are manufacturing sectors, all of them monopolisticallycompetitive. Sector 2 gathers all labour intensive manufacturing activities(those with an above average labour share). Labour un-intensive manufactur-ing sectors are split by consumer (sector 3) versus industry (sector 4) orienta-tion (consumer oriented being those with an above-average ratio of final tototal demand). Finally, Sector 5 is made up of services and is assumed to bemonopolistically competitive and non-tradeable. (The mapping from the 19-sector transaction table in the 1980 input-output tables for South Korea to ourfive sectors is detailed in the Appendix).

To implement the linear expenditure system we combine the 'subsistence'levels of consumer expenditure calculated for South Korea by Lluch and Powel(1975) and the consumer expenditure shares in the South Korean input-outputtables, as detailed in the Appendix. In all respects other than technical coeffi-cients and demand parameters, we leave the manufacturing sectors identical.Thus, we do not attempt to estimate how product differentiation varies acrossindustries, instead leaving a = 6 in all imperfectly competitive sectors.

We assume that there are three primary factors, arable land, internationallymobile capital and labour. There are three countries, one (on which we shallfocus) having one quarter of the world endowment, and the other two havingthe rest divided equally between them. We continue to abstract from tradi-tional sources of comparative advantage, so relative endowments of each factorare the same in all countries.

Trade barriers and tariffs are the same for all manufacturing sectors, withdeveloped country import barriers for manufactures set at 20%. Our experi-ment is to change the less developed country's trade barriers, but we keep thesebarriers equal across sectors in order to show how the same trade policy affectsdifferent sectors differently, and to reflect the evidence that both Korea andTaiwan shows 'relatively low variances in protection across sectors' (Pack, 1992).Figs 5 and 6 give, for the less developed economy on which we focus, thesimulated shares of the labour force in each manufacturing industry, and theassociated level of welfare.

We see from Fig. 5 that manufacturing employment as a whole is lowest atintermediate levels of the country's trade barriers, although at no point ismanufacturing employment zero. At these intermediate trade barriers thecountry has a small presence in labour intensive manufactures (Sector 2), andalso in labour un-intensive but consumer oriented manufactures (Sector 4).Sector 2 is active because of low wages, and Sector 4 because of consumerdemand. However, the developing country is a net importer of all manufactur-ing products.

Raising import barriers leads to import substituting industrialisation. Thereis expansion of consumer oriented manufactures (Sector 4) in order to meetfinal demand, and beyond some level the labour industry oriented Sector 3becomes active, driven by intermediate demand from the other sectors. As weC Royal Economic Society 1999


17/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 307'n 0.25

0.2 Sector2 (labour-intensivendustries)S 0.15 -

Q O.1 - \ Sector4 (labourun-intensiveconsumerorientedindustries)o

(I 0.05 - 0Sector (labourun-intensiveO ,.' ~ndustry oriented ndustries)ce' O. -" I . ' I I I1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8

Developing economy import barrierFig.5. Sharesof DevelopingEconomyLabourForce n Manufacturing

1.1-1.05 ,_

/

0)950.95 - -_-_-with tariff revenue

X 0.90.85

0.8 without tariffrevenue0.75-

1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8Developing economy import barrier

Fig.6. ImportSubstitutionvs TradeLiberalisationwithSeveralManufacturingSectors? Royal Economic Society 1999


18/21

308 THE ECONOMIC JOURNAL [APRILraise import barriers at no point does the country become a net exporter inany of these sectors.

Trade liberalisation leads to a quite different industrial structure, as wouldbe expected. The low wages of the developing country lead to rapid expansionof labour intensive industries, which become significant net exporters. Thisemployment expansion starts to raise wages, forcing out the other two sectors.

In interpreting these results it is important to bear in mind that we haveassumed that there are no differences in relative factor endowments, and thatthe trade liberalisation is a unilateral import liberalisation by the developingcountry. How then does this import liberalisation lead to such a dramaticexpansion of the labour intensive sector? The reason is that import barriersmake intermediate goods expensive in the developing country, and this is oneof the factors inhibiting industrial development. Reducing import barriersremoves this obstacle. All sectors benefit from this effect but, since wages arelow in the developing country, it is the labour intensive sector that is bestplaced to expand output in response to the change. Interestingly then, evenwithout assuming comparative advantage differences, open developing coun-tries will export labour intensive products.All this conforms with the empirical evidence from the newly industrialisingEast Asian economies which have liberalised trade over the last 20 years. Forexample, Korea reduced its average tariffs from about 32% in 1982 to 22% in1985 and to about 10% in 1992 (World Bank, 1994). These countries haveseen labour intensive products grow most as exports have exploded. Little(1994) looks at the sectoral detail of this process, comparing the patterns ofindustrial development of newly industrialising East Asian economies with thestandard norms for less developed economies calculated by Syrquin andChenery (1989). He finds that textiles, clothing, and metal products andmachinery (all of them included in Sector 2 in our numerical example) havegrown much faster than normal in these economies, while chemicals andprimary metal manufactures (aggregated into Sector 3 in our example)remained well below normal in their shares of GDP.The real income effects in this case confirm those that we saw in theaggregate case. Results are in Fig. 6 and as before the solid line is with realtrade barriers, and the dashed line gives the case in which barriers create tariffrevenue. Concentrating on the dashed line we see that real income is higherwith trade liberalisation than with import substitution. The welfare maximumat low tariffs derives from the optimal tariff argument. This arises primarily forexports of labour intensive products, in which the developing country has asignificant share of world production. (It is absent in Fig. 4 because with asingle manufacturing sector the less developed economy never achieves a largeenough world market share for terms of trade effects to be significant).

5. ConclusionsIn the analysis of this paper we have abstracted from many of the differencesbetween countries which are the focus of traditional development economics-(? Royal Economic Society 1999


19/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 309for example, we have assumed that all countries have the same amounts ofcapital and land per unit labour. Clearly, we do not think that internationaldifferences in these factors are unimportant, but our objective is to see howmuch can be explained without them.

We think the answer is a great deal. Industrial linkages create agglomerationof manufacturing sectors, and this results in substantial real income differ-ences between countries.18 Yet firms do not move to the low wage economy,because if they were to do so they would forego the benefits of proximity tosuppliers of intermediate goods and to their industrial customers.Industrial centres may however become too large to be constricted in theirinitial set of locations. Rising relative demand for manufactures will widen thewage gap between those countries that have industry and those that do not,and at some point industry will spread to other locations. But the logic ofagglomeration dictates that this is not spread evenly over developing countries.Instead industry spreads from one country to the next in a series of steps.Development takes place not as a process of smooth convergence of countries,but instead by countries in turn making the transition from a low level ofdevelopment to the rich country club.The approach also provides insights into the effects of trade policy. Uni-lateral trade policy can be used to attract industry by import substitution. Moresurprisingly, unilateral trade liberalisation can also be successful in attractingindustry, as lower cost intermediate goods remove a barrier to industrialdevelopment. Comparison of import substitution and trade liberalisationindicates that while the former leads to a presence in a wider range of sectors,the latter yields higher levels of welfare. And a simple calibrated example ofthe model indicates a sectoral development pattern that fits well with thatobserved in many newly industrialising countries. Even though we abstractfrom comparative advantage, we see that an open newly industrialising countrywill tend to export labour intensive manufactures.Universityof Torontoand CEPRLondon SchoolofEconomics

AppendixThemulti-industry odel:To the two internationally immobile primary factors of thesingle sector model, labour and arable land, we add capital which is perfectly mobileacross the M countries. There is a perfectly competitive primary sector and a numberof monopolistically competitive tradeable manufacturing and non-tradeable servicesectors. The price index for sector s in the set of industrial sectors I takes the form:

rM 11/(1-C>)qS= ns.(psi) (-a) SE I, (A.1)

18 Of course,the realwagedifferentialsuggestedbyoursimulations renothinglike as largeasrealworld ncomedifferentials, resumablyeflecting he factthatwe haveassumed hesame abourqualityand levelsof socialcapitaln allcountries.? RoyalEconomicSociety1999


20/21

310 THE ECONOMIC JOURNAL [APRILwhere the superscript s denotes the sector. The price index for sector s in the set ofservice sectors S is:

qS = [ns(pS)(lUf)]l/(l),) s S. (A.2)

The cost function of a single firm in sector s at location i is:Cs= (a+ 3xs) rPsw( 'P EkEIUSM 8)ls JI qk)uks, S E I U S, (A.3)

kEIUSwhere the share of primary sector output in the s industry is US, the share of industry kin the s industry is 1kss, r is the rate of return on capital, ps is the share of capital incosts and US s the share of primary sector output, and other parameters and variablesare as in the single-industry version of the model. The cost function in the perfectlycompetitive primary sector P is:

CP = zirP[ tow( U P ,kEIUSU ) 1UP (k)I kP (A.4)kEIUS

where zi is primary sector production in country i, t is the rental price of arable land,and 0 is land share in agriculture. Preferences are given by the following indirect utilityfunction:Vi= l-(Z-SENsYS) fj (qS)(y _ZE es). (A.5)

sEIUS V sE{p}uiusJDemand for primary factors comes from the cost functions in the usual way, andtheir prices are determined by market clearing. Final and derived demands for outputcome from cost functions and indirect utility functions. Given these demand functionsand cost functions (A3) firms maximise profits, and the number of firms in each sectoris determined by free entry and exit.

Simulationparameters.Section 1: or= 5, Tr= 1.1, y = 0.5, and 0 = 0.3, where 0 is theland share in the Cobb-Douglas agricultural production function.Section 2: or= 5, Tl = T2 1.3, y = 0.5, ,u = 0.4, and 0 = 0.1.Section 3: o = 5,Tl =r = 1.15, y = 0.5,,u = 0.5, and 0 = 0.3.Section 4: Sectors are the aggregate of the following sectors from the 19-sector 1980Korean input-output tables (pp. 56-7 in Bank of Korea (1983)):Sector 1 (primary sectors): agriculture, forestry and fishing; and mining.Sector 2 (labour intensive manufactures): textiles and leather; lumber and woodproducts; paper, printing and publishing; non-metal mineral products; metal productsand machinery; and miscellaneous manufactures.Sector 3 (labour un-intensive and industry oriented manufactures): chemicals andchemical products; and primary metal manufacturing.Sector 4 (labour un-intensive and consumer oriented manufactures): food and bev-erages.Sector 5 (services): all service sectors.The input output matrix takes the following form:US 0.085 0.058 0.268 0.498 0.010Y2,s 0.018 0.340 0.041 0.030 0.111Y3, s 0.093 0.201 0.397 0.029 0.116Y4,s - 0.056 0.005 0.004 0.145 0.016Y5, s 0.050 0.135 0.102 0.070 0.2030 0.373 - - - -pS 0.192 0.125 0.129 0.177 0.2811 - US pS - X/trS(-6) 0.132 0.135 0.059 0.050 0.262(? Royal Economic Society 1999


21/21

1999] AGGLOMERATION AND ECONOMIC DEVELOPMENT 311'Subsistence consumer expenditure shares are adapted from those calculated forSouth Korea by Lluch and Powel (1975):es/Iles 0.308 0.198 0.000 0.308 0.186Marginal consumer expenditure shares are calculated from the subsistence levels ofexpenditure and the actual consumer expenditure shares in the South Korean input-output tables:

S 0.060 0.190 0.057 0.215 0.477( = 6, ti = 1.2.Data sources or Fig. 3:International Labour Office. Annual issues. Year Book of Labour Statistics. Geneva:International Labour Office.Republic of China, Directorate General of Budget, Accounting and Statistics. Annualissues. YearBook of Statistics.

ReferencesBank of Korea. (1983). 1980 Input-OutputTablesof Korea.Seoul: The Bank of Korea.Dixit, Avinash K. and Stiglitz, Joseph E. (1977). 'Monopolistic competition and optimum productdiversity'. Amer*can conomicReview, ol. 67, pp. 297-308.Ethier, Wilfred J. (1982). 'National and international returns to scale in the modern theory ofinternational trade.' AmericanEconomicReview, ol. 72, pp. 389-405.Fujita, Masahisa, Krugman, Paul R. and Venables, AnthonyJ. (1999). TheSpatialEconomy:Cities,Regionsand InternationalTrade,Cambridge MA:MIT Press.Harris, C. (1954). 'The market as a factor on the localization of industry in the United States.' Annals oftheAssociationof AmericanGeographers,ol. 64, pp. 640-56.Hirschman, Albert 0. (1958). TheStrategy fEconomicDevelopment. ew Haven: Yale University Press.Krugman, Paul R. and Venables, Anthony J. (1995). 'Globalization and the inequality of nations.'

Quarterlyjournalf Economics,ol. 110, pp. 857-80.Little, Ian. (1994). 'Trade and industrialisation revisited.' Pakistan DevelopmentReview, vol. 33,pp. 359-89.Lluch, Constantino and Powel, Alan (1975). 'International comparisons of expenditure patterns.'EuropeanEconomicReview, ol. 5, pp. 275-303.Murphy, Kevin M., Shleifer, Andrei and Vishny, Robert W. (1989). 'Industrialization and the big push.'Journal ofPoliticalEconomy,ol. 97, pp. 1003-26.Myrdal, Gunnar (1957). EconomicTheory nd Under-developedegions.London: Duckworth.Pack, Howard (1992). 'New perspectives on industrial growth in Taiwan.' In (Gustav Ranis, ed.).Taiwan, romDevelopingo MatureEconomy.Boulder, Colorado: WestviewPress.Perroux, Francois (1955). 'Note sur la notion de p6le de croissance.' EconomiqueAppliquee, ol. 1-2,pp. 307-20.Pred, Alan R. (1966). The Spatial Dynamics of US Urban-IndustrialGrowth,1800-1914: Interpretive ndTheoretical ssays.Cambridge:MITPress.Puga, Diego (1998). 'The rise and fall of regional inequalities.' European Economic Review, (forth-coming).Puga, Diego and Venables, Anthony J. (1997). 'Preferential trading arrangements and industriallocation.' Journal of InternationalEconomics, ol. 43, pp. 347-68.Scotchmer, Susan and Thisse, Jacques-Francois (1992). 'Space and competition: a puzzle.' Annals ofRegionalScience, ol. 26, pp. 269-86.Syrquin, Moshe and Chenery, Hollis B. (1989). 'Patterns of development: 1950-83.' Discussion Paper41, The World Bank.World Bank (1994). East Asia's Trade and Investment:Regional and Global Gains from Liberalization.Washington, DC.

agglomeration and economic development import substitution vs. trade liberal is at ion

Documents