thirlwall - the nature of economic growth

The Nature of Economic Growth

Other Books by the Author

Growth and Development: with Special Reference toDeveloping Economies (Sixth Edition)

Inflation, Saving and Growth in Developing EconomiesRegional Growth and Unemployment in the United

Kingdom (with R. Dixon)Financing Economic DevelopmentBalance-of-Payments Theory and the United Kingdom

Experience (Fourth Edition with H. Gibson)Nicholas KaldorUK Industrialisation and Deindustrialisation (Third Edition

with S. Bazen)The Performance and Prospects of the Pacific Island

Economies in the World EconomyEconomic Growth and the Balance-of-Payments Constraint

(with J. McCombie)The Economics of Growth and Development: Selected

Essays, Vol. 1Macroeconomic Issues from a Keynesian Perspective:

Selected Essays, Vol. 2

Edited Works

Keynes and International Monetary RelationsKeynes and Laissez-FaireKeynes and the Bloomsbury Group (with D. Crabtree)Keynes as a Policy AdviserKeynes and Economic DevelopmentKeynes and the Role of the State (with D. Crabtree)European Factor Mobility: Trends and Consequences

(with I. Gordon)The Essential Kaldor (with F. Targetti)Further Essays in Economic Theory and Policy, Volume 9,

Collected Economic Papers of N. Kaldor (with F. Targetti)Causes of Growth and Stagnation in the World Economy (the

Mattioli Lectures of N. Kaldor with F. Targetti)Economic Dynamics, Trade and Growth: Essays on

Harrodian Themes (with G. Rampa and L. Stella)

The Nature ofEconomic GrowthAn Alternative Framework forUnderstanding the Performance of Nations

A.P. ThirlwallProfessor of Applied Economics, University of Kent at Canterbury, UK

Edward ElgarCheltenham, UK • Northampton, MA, USA

© A.P. Thirlwall 2002

All rights reserved. No part of this publication may be reproduced,stored in a retrieval system or transmitted in any form or by anymeans, electronic, mechanical, photocopying, recording, or otherwisewithout the prior permission of the publisher.

Published byEdward Elgar Publishing LimitedGlensanda HouseMontpellier ParadeCheltenhamGlos GL50 1UAUK

Edward Elgar Publishing, Inc.136 West StreetSuite 202NorthamptonMassachusetts 01060USA

A catalogue record for this book is available from the British Library

Library of Congress Cataloguing in Publication DataThirlwall, A. P.

The nature of economic growth : an alternative framework forunderstanding the performance of nations / A.P. Thirlwall.

p. cm.Includes index.1. Economic development. I. Title.

HD82 .T483 2002338.9—dc21

2001051074

ISBN 1 84064 864 3 (cased)Printed and bound in Great Britain by Biddles Ltd, www.biddles.co.uk

Contents

Preface vii

1 Growth theory in the history of thought 12 Neoclassical and ‘new’ growth theory:

a critique 203 Manufacturing industry as the engine of

growth 404 A demand-oriented approach to economic

growth: export-led growth models 525 Balance of payments constrained growth:

theory and evidence 666 The endogeneity of the natural rate of

growth 79

Bibliography 97Index 107

To my Grandchildren,Lorenzo and Sienna

Preface

This short book has arisen out of a series of lecturesand seminars that I gave at the National Universityof Mexico in September 2000. They, in turn, werebased on a selection of lectures I have been givingfor a long time at the University of Kent to studentsstudying for a Master’s degree in developmenteconomics. The fact that the lectures were given tograduate students, however, does not mean that thebook will not be intelligible to others, includingundergraduates and practitioners in the develop-ment field. First of all, the basic principles of growthand development theory are not that difficult tograsp by anyone with a willingness and interest tolearn, and secondly, following the dictum of AlfredMarshall (the great 19th-century Cambridgeeconomist), I have tried to translate theoreticalmodels expressed in mathematics into words.

The matter of why some countries are rich andothers are poor, and why some countries growfaster than others over long periods of time(although not necessarily continuously), hasalways fascinated me as an economist, and in thechapters to follow I try to present the conventionalwisdom, as it has evolved historically, but with acritical eye, from Adam Smith, the author of AnInquiry into the Nature and Causes of the Wealth of

vii

Nations (1776) to ‘new’ or endogenous growththeory. I am critical of the latter, and its predeces-sor, neoclassical growth theory, and my owncontribution is to try and put (back) demand intogrowth theory as a driving force. In my view, neo-classical and ‘new’ growth theory is far toosupply-oriented in its approach, not recognizingsufficiently the various constraints on demandlong before supply constraints bite. In an open,developing economy one of the major constraintsis the availability of foreign exchange to pay forimports, so that export growth which relaxes abalance of payments constraint on demandbecomes a crucial determinant of overall growthperformance. This is entirely missing from ‘new’growth theory, but is a central feature of my ownthinking and research. There are not manydeveloping countries in the world that could notutilize resources more fully, and grow faster, givengreater availability of foreign exchange. Within thisframework, the main factors of production – labourand capital – are considered to be elastic todemand, and so too is productivity growth basedon static and dynamic returns to scale, captured byVerdoorn’s Law. Demand creating its own supply(within limits) in a growth context (as well as in astatic context), rather than the pre-Keynesian viewof supply creating its own demand, provides analternative framework to the neoclassical one forunderstanding the differential growth perfor-mance of nations.

viii The nature of economic growth

I am grateful to Dr Roberto Escalante (Chairmanof the Department), and his colleagues at theNational University of Mexico, for the opportunityto give these lectures, and to write this short book,and for the feedback on them.

I am also grateful to the editor of Banca Nazionaledel Lavoro Quarterly Review for allowing me toreproduce in Chapter 6 certain material publishedin the December 2000 issue of the journal.

A.P.T.

Preface ix

1. Growth theory in the historyof thought

Growth and development theory is at least as oldas Adam Smith’s famous book published in 1776entitled An Inquiry into the Nature and Causes of theWealth of Nations. The macro issues of growth, andthe distribution of income between wages andprofits, were the major preoccupation of all thegreat classical economists including Adam Smith,Thomas Malthus, John Stuart Mill, David Ricardoand Karl Marx.

One of Smith’s most important contributionswas to introduce into economics the notion ofincreasing returns – a concept that ‘new’ growththeory (or endogenous growth theory) has recentlyrediscovered (see chapter 2). In Smith, increasingreturns is based on the division of labour. He sawthe division of labour, or gains from specialization,as the very basis of a social economy, otherwiseeverybody might as well be their own RobinsonCrusoe doing everything for themselves. And itis the notion of increasing returns, based on thedivision of labour, that lay at the heart of Smith’soptimistic vision of economic progress as a self-generating process, in contrast to the later classical

1

economists, such as Ricardo and Mill, whobelieved that economies would end up in astationary state due to diminishing returns in agri-culture; and also in contrast to Marx who believedthat capitalism would collapse through its own‘inner contradictions’ (competition between capi-talists reducing the rate of profit; a failure ofeffective demand as capital is substituted forlabour, and the alienation of workers).

The notion of increasing returns may sound atrivial one but it is of profound significance for theway we view economic processes. It is not possibleto understand divisions in the world economy, andso-called ‘centre–periphery’ models of growth anddevelopment (between ‘north’ and ‘south’ and richand poor countries), without distinguishingbetween activities subject to increasing returns onthe one hand and diminishing returns on the other.Increasing returns means rising labour productiv-ity and per capita income, and no limits to theemployment of labour set by the (subsistence)wage, whereas diminishing returns implies theopposite. Industry is, by and large, an increasingreturns activity, while land-based activities, suchas agriculture and mining, are diminishing returnsactivities. Rich, developed countries tend tospecialize in increasing returns activities, whilepoor developing countries tend to specialize indiminishing returns activities. It is almost as simpleas that, but not quite!

2 The nature of economic growth

Adam Smith

If we go back to Adam Smith, he recognized threeways in which the productivity of labour isincreased through specialization: first, theincreased dexterity or skill of labour through whatwe now call ‘learning by doing’; second, the savingof time which is otherwise lost through switchingfrom one job to another, and third, the greaterscope for capital accumulation, that is, the abilityto break up complex processes into simplerprocesses permitting the use of machinery, whichraises productivity still further. But the division oflabour, or the ability to specialize, depends on theextent of the market. Smith used the example ofproducing pins. There is no point in installingsophisticated machinery to work on the differentprocesses involved in producing a pin if only a fewpins are demanded. Workers may as well produceeach pin individually. But if the market is large,there is great scope for economies of scale. Theextent of the market, however, depends in turn onthe division of labour because this determines thelevel of productivity, per capita income andpurchasing power. We have here an interdepen-dent and circular process. The division of labourdepends on the extent of the market, but the extentof the market depends on the division of labour.

Smith recognizes, however, that the process hedescribed was much more a feature of industrythan agriculture. He says explicitly:

Growth theory in the history of thought 3

the nature of agriculture, indeed, does not admit of somany subdivisions of labour, nor of so complete aseparation of one business from another, as manufactures.It is impossible to separate so entirely the business of thegrazier from that of the corn farmer, as the trade of thecarpenter is commonly separated from that of the smith.(p. 16)

There is not the scope for increasing returns in agri-culture. Indeed, if land is a fixed factor ofproduction, there will be diminishing returns tolabour – one of the few incontrovertible laws ofeconomics, as Keynes once said.

As far as the extent of the market is concerned,Smith also recognized the importance of exports,as we do today particularly for small countries.Exports provide a ‘vent for surplus’; that is, anoutlet for surplus commodities that otherwisewould go unsold. There is a limit to whichindigenous populations can consume fish, bananasand coconuts, or can use copper, diamonds and oil:

without an extensive foreign market, [manufacturers]could not well flourish, either in countries so moderatelyextensive as to afford but a narrow home market; or incountries where the communication between one provinceand another [is] so difficult as to render it impossible forthe goods of any particular place to enjoy the whole of thathome market which the country can afford (p. 680)

This vision of Smith of growth and developmentas a cumulative interactive process based on thedivision of labour and increasing returns inindustry lay effectively dormant until the


American economist, Allyn Young, based at theLondon School of Economics, revived it in aneglected but profound paper in 1928 entitled‘Increasing Returns and Economic Progress’(another paper rediscovered by ‘new’ growththeory). As Young observed:

Adam Smith’s famous theorem amounts to saying that thedivision of labour depends in large part on the divisionof labour. [But] this is more than mere tautology. It meansthat the counter forces which are continually defeating theforces which make for equilibrium are more pervasiveand more deeply rooted than we commonly realise –change becomes progressive and propagates itself in acumulative way.

In Young, increasing returns are not simplyconfined to factors which raise productivity withinindividual industries, but are related to the outputof all industries which he argues must be seen asan interrelated whole. For example, a larger marketfor one good may make it profitable to use moremachinery in its production, which reduces thecost of the good and the cost of machinery whichthen makes the use of machinery profitable inother industries, and so on. In other words, a largermarket for one good confers a positive externalityon others. Under certain conditions, change willbecome progressive and propagate itself in acumulative way: the precise conditions beingincreasing returns and an elastic demand forproducts so that, as their exchange value falls, pro-portionately more is bought. Let us consider a


simple example of Young’s vision of increasingreturns as a macro phenomenon. Take the steel andtextile industries, both subject to increasing returnsand producing price-elastic products. As thesupply of steel increases, its relative price falls. Ifdemand is elastic textile producers demand pro-portionately more steel. Textile productionincreases and its relative price then falls. If demandis elastic steel producers demand proportionatelymore textiles, and so on. As Young says: ‘undercertain circumstances there are no limits to theprocess of expansion except the limits beyondwhich demand is not elastic and returns do notincrease’.

This process could not happen with diminishingreturns activities, such as primary products, withdemand price inelastic. No wonder levels of devel-opment, both historically and today, seem to beassociated with the process of industrialization.There is, indeed, a strong association acrosscountries between the level of per capita incomeand the share of industry in GDP, and also a strongassociation across countries between industrialgrowth and the growth of GDP (see Chapter 3).

Allyn Young’s 1928 vision also got lost untileconomists such as Gunnar Myrdal (SwedishNobel Prize winner in economics), AlbertHirschman and Nicholas Kaldor (a pupil of Youngat the LSE, and later joint-architect of theCambridge post-Keynesian school of economists)started to develop non-equilibrium models of thedevelopment process in such books as Economic


Theory and Underdeveloped Regions (Myrdal, 1957);Strategy of Economic Development (Hirschman,1958), and Economics without Equilibrium (Kaldor,1985). Kaldor used to joke that economics wentwrong after Chapter 4 of Book I of the Wealth ofNations (1776) when Adam Smith abandoned theassumption of increasing returns in favour ofconstant returns, and the foundations for generalequilibrium theory were laid: but foundationstotally inappropriate for analysing the dynamicsof growth and change.

The Classical Pessimists

The prevailing classical view after Smith was verypessimistic about the process of economic devel-opment, which led the historian, Thomas Carlyle,to describe economics as the dismal science – nota view shared by present readers, I hope! The firstof the pessimists was Thomas Malthus, who wrotehis famous Essay on the Principle of Population in1798, in which he claimed that there is a ‘tendencyin all animated life to increase beyond the nour-ishment prepared for it’. According to Malthus,‘population, when unchecked, goes on doublingitself every 25 years, or increases in a geometricratio [whereas] it may be fairly said – that themeans of subsistence increases in an arithmeticalratio’. Taking the world as a whole, therefore,Malthus concludes that ‘the human species wouldincrease (if unchecked) as the numbers 1, 2, 4, 8, 16,32, 64, 128, 256 and subsistence as 1, 2, 3, 4, 5, 6, 7,


8, 9’. This implies, of course, a diminishing pro-portional rate of increase of food production, ordiminishing returns to agriculture. The result ofthis imbalance between food supply andpopulation will be that living standards oscillatearound a subsistence level, with rising livingstandards leading to more children, which thenreduces living standards again.

This Malthusian vision forms the basis in thedevelopment literature of models of the low-levelequilibrium trap associated originally with Nelson(1956) and Leibenstein (1957), and models of thebig push to escape from it. The ghost of Malthusdoes, indeed, still haunt many Third Worldcountries, although it has to be said that, for theworld as a whole, food production has grownmuch faster than population for at least the lastcentury. The reason is that technical progress,always underestimated by the classical pessimists,has offset diminishing returns, leading to sub-stantial increases in productivity, particularly inEurope and North America, but also in developingcountries that experienced a ‘green revolution’.

Another of the great classical pessimists wasDavid Ricardo. In 1817 he published his Principlesof Political Economy and Taxation, in which hepredicted that capitalist economies would end upin a stationary state with no capital accumulationand therefore no growth, also due to diminishingreturns in agriculture. In Ricardo’s model, capitalaccumulation is determined by profits, but profitsget squeezed between subsistence wages and the


payment of rent to landowners which increases asthe price of food increases owing to diminishingreturns to land and rising marginal cost. As theprofit rate in agriculture falls, capital shifts toindustry, causing the profit rate to decline theretoo. In industry, profits also get squeezed becausethe subsistence wage rises in terms of food. Asprofits fall to zero, capital accumulation ceases,heralding the stationary state. Ricardo recognizedthat the cheap import of food could delay thestationary state, and as an industrialist andpolitician, as well as an economist, he campaignedvigorously for the repeal of the Corn Laws inEngland which protected British farmers. ArthurLewis’s famous model economic developmentwith unlimited supplies of labour (Lewis, 1954) isa classical Ricardian model, but one where theindustrial wage stays the same as long as surpluslabour exists. Ricardo’s pessimism has also beenconfounded by technical progress, and thestationary state has never appeared on the horizon,except, perhaps, in Africa in recent times, but thecauses there are different and complex, related topolitical failure.

Karl Marx in his famous book, Das Kapital (1867),also predicted crisis due to falling profits, butthrough a different mechanism related to compe-tition between capitalists, overproduction andsocial upheaval. The wages of labour aredetermined institutionally, and profit (or surplusvalue, which only labour can create) is thedifference between output per man and the wage


rate. The rate of profit is given by s/(v+c) or(s/v)/(1+c/v), where s is surplus value, c is‘constant’ capital, v is ‘variable’ capital (the wagebill), and c/v is defined as the organic compositionof capital. The latter is assumed to rise throughtime, and as it does so, the rate of profit will fallunless the rate of surplus value rises. As long assurplus labour (or what Marx called a ‘reservearmy of unemployed’) exists there is no problem,but Marx predicted that, as capital accumulationtakes place, the reserve army will disappear,driving wages up and profits down. The capital-ists’ response is either to attempt to keep wagesdown (the immiseration of workers) leading tosocial conflict, or to substitute more capital forlabour, which raises the organic composition ofcapital and worsens the problem of a falling profitrate. Moreover, as labour is replaced, it cannotconsume all the goods produced, and there is afailure of effective demand, or a ‘realization crisis’,as Marx called it. Capitalism collapses through itsown ‘inner contradictions’, and power passes tothe working classes.

Classical models of growth and distribution stillform an integral part of growth and developmenttheory, particularly the emphasis on the capitalistsurplus for investment, but the gloomy prognosti-cations of the classical economists have notmaterialized, at least for the capitalist world as awhole. As said before, what is wrong with Malthusand Ricardo is that they both underestimated thestrength of technical progress in agriculture as an


offset to diminishing returns. What is wrong withMarx is that he first of all confused money and realwages, and secondly underestimated the effect oftechnical progress in industry on the productivityof labour. A rise in money wages as labourbecomes scarcer does not necessarily mean a risein real wages; and a rise in real wages could beoffset by a rise in productivity, leaving the rate ofprofit unchanged. In other words, in a growingeconomy, there is no necessary conflict betweenwages and the rate of profit.

For nearly 60 years after Marx’s death in 1883,growth and development theory lay virtuallydormant, until it was revived by the Britisheconomist (Sir) Roy Harrod in 1939 in a classicarticle, ‘An Essay in Dynamic Theory’. In the late19th and early 20th centuries, economics wasdominated by neoclassical value theory under theinfluence of Jevons, Walras and particularly AlfredMarshall’s Principles of Economics, published in1890. Growth and development was regarded asan evolutionary natural process akin to biologicaldevelopments in the natural world. All thischanged in 1939 with Harrod’s article, which ledto the development of what came to be called theHarrod–Domar growth model (named afterEvesey Domar as well, who derived independentlyHarrod’s fundamental result in 1947 but in adifferent way (Domar, 1947)). The model hasplayed a major part in thinking about developmentissues ever since, and is still widely used as aplanning framework in developing countries.


Neoclassical growth theory was born as a reactionto the Harrod–Domar model, and ‘new’ growththeory developed as a reaction to neoclassicalgrowth theory.

Harrod–Domar Growth Model

Harrod was one of the most original and versatileeconomists of the 20th century. He was theinventor of the marginal revenue product curve inmicro theory; the life cycle hypothesis of savingand the absorption approach to the balance ofpayments in macro theory; the biographer ofKeynes; the author of a book on inductive logic; aswell as the originator of modern growth theory.

Harrod’s 1939 model was an extension ofKeynes’s static equilibrium analysis of The GeneralTheory. The question Harrod asked was: if thecondition for a static equilibrium is that plans toinvest must equal plans to save, what must be therate of growth of income for this equilibriumcondition to hold in a growing economy throughtime? Moreover, is there any guarantee that thisrequired rate of growth will prevail?

Harrod introduced three different growthconcepts: the actual growth rate (ga); the warrantedgrowth rate (gw) and the natural growth rate (gn).The actual growth rate is defined as ga = s/c, wheres is the savings ratio and c is the actual incremen-tal capital–output ratio (that is, the amount of extracapital accumulation or investment associated witha unit increase in output). This expression is defi-


nitionally true because in the national incomeaccounts, savings and investment are equal. Thuss/c = (S/Y)/(I/ΔY) = (ΔY/Y), where S is saving, Iis investment, Y is output, and ΔY/Y is the growthrate (ga).

This rate of growth, however, does not neces-sarily guarantee a moving equilibrium throughtime in the sense that it induces just enoughinvestment to match planned saving. Harrod calledthis rate the warranted growth rate. Formally, it isthe rate that keeps capital fully employed, so thatthere is no overproduction or underproduction,and manufacturers are therefore willing to carryon investment in the future at the same rate as inthe past. How is this rate determined? The demandfor investment is given by an acceleratormechanism (or what Harrod called ‘the relation’)with planned investment (Ip) a function of thechange in output, so that Ip = crΔY, where cr is therequired incremental capital–output ratio at agiven rate of interest, determined by technologicalconditions. Planned saving (Sp) is a function ofincome so that Sp = sY where s is the propensityto save. Setting planned investment equal toplanned saving gives crΔY = sY or ΔY/Y = s/cr,which equals the warranted growth rate (gw). Fordynamic equilibrium, output must grow at the rates/cr. If not, the economic system will be cumula-tively unstable. If actual growth exceeds thewarranted growth rate, plans to invest will exceedplans to save; and the actual growth rate is pushedeven further above the warranted rate. Contrari-


wise, if actual growth is less than the warrantedrate, plans to invest will be less than plans to saveand growth will fall further below the warrantedrate. This is the Harrod instability problem.Economies appeared to be poised on a ‘knife-edge’.Any departure from equilibrium, instead of beingself-righting, will be self-aggravating.

The American economist, Evesey Domar,working independently of Harrod, also arrived atHarrod’s central conclusion by a different route –hence the linking of their two names. What Domarrealized was that investment both increasesdemand via the Keynesian multiplier and alsoincreases supply by expanding capacity. So thequestion he posed was: what is the rate of growthof investment that will guarantee that demandmatches supply? The crucial rate of growth ofinvestment can be derived in the following way. Achange in the level of investment increases demandby ΔYd = ΔI/s, and investment itself increasessupply by ΔYs = Iσ, where σ is the productivity ofcapital (ΔY/I). Therefore, for ΔYd = ΔYs we musthave ΔI/s = Iσ, or ΔI/I = sσ. That is to say,investment must grow at a rate equal to theproduct of the savings ratio and the productivityof investment. With a constant savings–investmentratio, this implies output growth at the rate sσ.Since σ = 1/cr (at full employment), the Harrod andDomar result for equilibrium growth is the same.

Even if the actual and warranted growth ratesare equal, however, guaranteeing the full utiliza-tion of capital, this does not guarantee the full


utilization of labour which depends on the naturalrate of growth (gn) made up of two components:the growth of the labour force (l) and the growthof labour productivity (t), both exogenously given.The sum of the two gives the growth of the labourforce in efficiency units. If all labour is to beemployed, the actual growth rate must match thenatural rate. If the actual growth rate falls belowthe natural rate there will be growing unemploy-ment of the structural variety.

It should be clear that the full employment ofboth capital and labour requires that ga = gw = gn;a happy coincidental state of affairs that JoanRobinson once coined ‘the golden age’ toemphazise its mythical nature.

Where do the developing countries fit into thisstory? The short-run (trade cycle) problem is therelation between ga and gw, and we will not saymore about this here. The long-run problem is therelation between gw and gn, or the relation betweenthe growth of capital and the growth of the labourforce in efficiency units. Almost certainly, in mostdeveloping countries, gn exceeds gw. Labour forcegrowth (determined by population growth) maybe 2 per cent per annum, and productivity growth3 per cent per annum, giving a natural growth rateof 5 per cent. If the net savings ratio is 9 per centand the required incremental capital–output ratiois 3, the warranted growth rate is only 3 per cent.Therefore gn > gw. This has two main conse-quences. Firstly, it means that the effective labourforce is growing faster than capital accumulation,


so that with fixed coefficients of production therewill be unemployment of the structural variety.Secondly, it means that plans to invest will exceedplans to save, because if the economy could growat 5 per cent there are profitable investment oppor-tunities for more than 9 per cent saving, and therewill be inflationary pressure. Hence the simulta-neous existence of unemployment and inflation indeveloping countries is not a paradox; it is theoutcome of an inequality between the natural andwarranted growth rates.

A good deal of development policy can beunderstood and considered within this Harrodframework. The task is to bring gn and gw closertogether; to reduce gn and to increase gw. The onlyfeasible way to reduce the growth of the labourforce is to reduce population growth. The Harrodmodel provides a rationale for population control.A second way to reduce gn is to reduce the rate oflabour-saving technical progress, but this has theserious drawback of reducing the growth of livingstandards. A rise in gw could be brought about byincreases in the savings ratio. This is whatmonetary and fiscal policy programmes aredesigned to do, with emphasis on tax reform andpolicies of financial liberalization. A rise in gwcould also come about if the capital–output ratiowas reduced by countries using more labour-intensive techniques of production. There is acontinuing debate on the choice of appropriatetechniques in developing countries, and whether


more labour-intensive techniques could beemployed without the sacrifice of output or saving.

The Harrod (and Domar) model provided thestarting point for the great debates in growtheconomics that preoccupied large sections of theeconomics profession for at least three decadesbetween the mid-1950s and the 1980s. The battle-lines were drawn up between the neoclassicalgrowth school on the one hand, based inCambridge, Massachusetts, USA, with the majorprotagonists being Robert Solow, Paul Samuelsonand Franco Modigliani, and the Keynesian growthschool on the other, based in Cambridge, England,with the major protagonists being Nicholas Kaldor,Joan Robinson, Richard Kahn and Luigi Pasinetti.What was immediately apparent to both campswas that, if the Harrod–Domar model was a rep-resentation of the real world, all economies, richand poor, capitalist and communist, would be infor a bumpy ride. The variables and parametersdetermining gn and gw were all independentlygiven, and there were apparently no automaticmechanisms for bringing the two rates of growthinto line to provide the basis for steady long-rungrowth at the natural rate. The task that both con-flicting camps set themselves was to developmechanisms to reconcile divergences between gnand gw.

The Cambridge, England camp focused on thesavings ratio, making it a function of the distribu-tion of income between wages and profits whichin turn was assumed to be related to whether the


economy was in boom or slump. Specifically, intheir model the propensity to save out of profitsis assumed to be higher than out of wages, and theshare of profits in national income is assumed torise during booms and fall during slumps.Therefore, if gn exceeds gw, generating a boom, theshare of profits rises and the savings ratio will rise,raising gw towards gn. The only constraint mightbe an ‘inflation barrier’ caused by workers notbeing willing to see the share of wages fall belowa certain minimum. Conversely, if gn is less thangw, generating a slump, the share of profits fallsand the savings ratio falls, lowering gw towards gn.The only limit here might be a minimum rate ofprofit acceptable to entrepreneurs which sets alimit to the fall in the share of profits.

The Cambridge, Massachusetts camp focused onthe capital–output ratio, arguing that, if the labourforce grows faster than capital, the pricemechanism will operate in such a way as to inducethe use of more labour intensive techniques, andvice versa. Thus, if gn exceeds gw, thecapital–output ratio will fall, raising gw to gn. If gnis less than gw, the capital–output ratio will rise,lowering gw to gn. This neoclassical adjustmentmechanism, however, presupposes two things:firstly, that the relative price of labour and capitalare flexible enough, and secondly that there is aspectrum of techniques to choose from so thateconomies can move easily and smoothly along acontinuous production function relating output tothe factor inputs, capital and labour. If this is true,


economies can achieve a growth equilibrium at thenatural rate (see Chapter 2).

Out of the neoclassical model, however, camethe extraordinary counterintuitive conclusion thatinvestment does not matter for long-run growthbecause the natural rate depends on the growth ofthe labour force and labour productivity(determined by technical progress) and both areexogenously determined. Any increase in acountry’s saving or investment ratio would beoffset by an increase in the capital–output ratio,leaving the long-run growth rate unchanged. Theargument depends crucially, however, on the pro-ductivity of capital falling as the capital to labourratio rises. In other words, it depends on theassumption of diminishing returns to capital. Thisis the neoclassical story that ‘new’ endogenousgrowth theory objects to. If there are mechanismswhich keep the productivity of capital from fallingas more investment takes place, then theinvestment ratio will matter for long-run growth,and growth is endogenous in this sense; that is,growth is not simply determined by the exogenousgrowth of the labour force in efficiency units.

In the next chapter we consider in more detailthe assumptions and predictions of the neoclassi-cal model, and the criticisms made of it. We thenlook at the challenge of ‘new’ growth theory, andwe are critical of this too.


2. Neoclassical and ‘new’growth theory: a critique

Our task in this chapter is to outline formally theassumptions and predictions of neoclassicalgrowth theory as a background to showing, firstly,how the neoclassical production function is usedfor analysing growth rate differences betweencountries, and its weaknesses; and secondly, howneoclassical growth theory forms the basis for‘new’ endogenous growth theory – the only majordifference being that the assumption of diminish-ing returns to capital is relaxed, so that ‘new’growth theory is subject to the same majorcriticisms as conventional neoclassical theory as faras analysing and understanding growth rate dif-ferences between countries is concerned.

The Neoclassical Model

The neoclassical growth model is based on threekey assumptions. The first is that the labour force(l) and labour-saving technical progress (t) grow ata constant exogenous rate. The second assumptionis that all saving is invested: S = I = sY. There is noindependent investment function. The third

20

assumption is that output is a function of capitaland labour, where the production function exhibitsconstant returns to scale, and diminishing returnsto individual factors of production. The mostcommonly used neoclassical production function,with constant returns to scale, is the so-calledCobb–Douglas production function, named afterCharles Cobb, a mathematician, and Paul Douglas,a well-known Chicago economist before WorldWar II (who later became a US senator). Thefunction takes the form:

Y = TKαL1–α, (2.1)

where Y is output, K is capital, L is labour, T is thelevel of technology, α is the elasticity of outputwith respect to capital and 1–α is the elasticity ofoutput with respect to labour. Obviously α + (1–α)= 1 (the assumption of constant returns to scale),so that a 1 per cent increase in capital and labourleads to a 1 per cent increase in output.

To consider the predictions of the model, it isconvenient to transform equation (2.1) into its‘labour-intensive’ form by dividing both sides byL, so that the dependent variable is output perhead, and the independent variables are the levelof technology and capital per head.

Y/L = (TKαL1–α)/L = T(K/L)α

or

q = T(k)α (2.2)

Neoclassical and ‘new’ growth theory: a critique 21

where q is output per head and k is capital perhead.

The basic predictions of the neoclassical model,which can be shown diagrammatically (see below),are as follows:

1. in the steady state, the level of output per head(q) is positively related to the savings–investment ratio and negatively related to thegrowth of population (or labour force);

2. the growth of output is independent of thesavings–investment ratio and is determined bythe exogenously given rate of growth of thelabour force in efficiency units (l + t). This isbecause a higher savings–investment ratio isoffset by a higher capital–output ratio (or alower productivity of capital) owing to theassumption of diminishing returns to capital;

3. given identical tastes and preferences (that is,the same savings ratio) and technology (that is,production function), there will be an inverserelation across countries between thecapital–labour ratio and the productivity ofcapital, so that poor countries should growfaster than rich countries, leading to the conver-gence of per capita incomes across the world.

Figure 2.1 illustrates the first two predictions.The production function, q = f(k), with dimin-

ishing returns to capital, comes from equation (2.2).The ray from the origin with slope (l + t)/s givespoints of equality between the rate of growth of


capital and labour measured in efficiency units.1

Only at k* is the level of output per head such as togive a rate of growth of capital equal to the rate ofgrowth of the labour force. To the left of k*(k1), thegrowth of capital is greater than the growth oflabour, and economies are assumed to move alongtheir smooth production function towards k* usingmore capital-intensive methods of production. Tothe right of k*(k2), the growth of capital is less thanthe growth of labour, and economies are assumedto use more labour-intensive techniques ofproduction. At k*, where the capital to labour ratio


q = f (k)

q = [(l + t)/s]kq

q*

l + ts

k* kk1

k2

Figure 2.1

is in equilibrium, output per head will also be inequilibrium at q*. It can be seen from the figure thata rise in the savings ratio (s) pivots downwards theray from the origin and raises the equilibrium kand raises the level of q, but does not affect thegrowth rate of the economy. It can also be seen thatthe level of q will be inversely related to the rate ofgrowth of the labour force because a rise in l pivotsupwards the ray from the origin.

The explanation for convergence of per capitaincome across countries can be seen from theformula for the capital–output ratio:

K/Y = (K/L) (L/Y). (2.3)

If there is diminishing returns to capital, a higherK/L will not be offset by a higher Y/L ratio, andtherefore K/Y will be higher. Thus, if thesavings–investment ratio is the same acrosscountries, rich countries with a higher K/L ratioshould grow more slowly than poor countries witha lower K/L because the productivity of capital islower in the former case than in the latter.

What major criticisms can be made of thismodel, apart from the empirical fact that across theworld we do not observe the convergence of livingstandards? The fundamental point to be made atthis stage is that the neoclassical model is a supply-oriented model par excellence. First, demand neverenters the picture. Saving leads to investment, sothat supply creates its own demand. The neoclas-sical model of growth takes us back to a


pre-Keynesian world where demand does notmatter for an understanding of the determinationof the level of output (and, by implication, thegrowth of output). Secondly, factors of productionand technical progress are treated as exogenouslydetermined, unresponsive to demand. But, by andlarge, the demand for factors of production is aderived demand, derived from the growth ofoutput itself. Much technical progress and labourproductivity growth is also induced by the growthof output itself (see later).

The assumption of exogeneity of factor suppliesis no more apparent than in the studies that use theaggregate production function for analysinggrowth rate differences between countries; anapproach pioneered by Abramovitz (1956) andSolow (1957) and still widely utilized. Let usconsider this approach and comment on its limi-tations.

Using the Production Function for AnalysingGrowth Differences

If we go back to the Cobb–Douglas productionfunction in equation (2.1), it is easy to see how itcan be used for analysing the sources of growth;that is, decomposing a country’s growth rate intothe contribution of capital, labour and technicalprogress. The question is, how useful is it for aproper understanding of the growth performance ofcountries if the main inputs into the growthprocess are not exogenous but endogenous?


The function in equation (2.1) is made opera-tional by taking logarithms of the variables anddifferentiating with respect to time, which gives:

y = t + α(k) + (1 – α)l, (2.4)

or in labour-intensive form:

y – l = t + α (k – l), (2.5)

where lower-case letters represent rates of growthof the variables.

Given estimates of α and (1 – α), the contribu-tion of capital growth and labour force growth toany measured growth rate can be estimated,leaving the contribution of technical progress as aresidual. For example, suppose y = 5%, k = 5%,l = 2%, α = 0.3 and (1 – α) = 0.7. The contributionof capital to growth is then (0.3) (5%) = 1.5percentage points or 30 per cent; the contributionof labour is (0.7) (2%) = 1.4 percentage points or 28per cent, leaving the contribution of technicalprogress as 5% – 2.9% = 2.1% or 42 per cent.

Solow (1957) was the first to use the labour-intensive form of the Cobb–Douglas productionfunction in analysing the growth performance ofthe US economy over the previous 50 years, andconcluded that only 10 per cent of the growth ofoutput per man could be ‘explained’ by the growthof capital per man, leaving 90 per cent of growthto be ‘explained’ by various forms of technicalprogress. Denison (1962, 1967) used the same


production function approach, or growthaccounting framework, to study growth perfor-mance in the USA and between the countries ofEurope, disaggregating the technical progress term(or residual) into various component parts.Maddison (1970) used the approach to studygrowth rate differences between developingcountries. Since this early research, there has beena mass of other studies too extensive to survey here(however, see Felipe, 1999), but two recent studiesmay be mentioned as illustrative. The World Bank(1991) did a study of 68 countries showing capitalaccumulation to be of prime importance, withtechnical progress minimal. This seems to be thecentral conclusion for developing countries incontrast to developed countries. Secondly, thereis the controversial study by Alwyn Young (1995)of the four East Asian ‘dragons’ of Hong Kong,Singapore, South Korea and Taiwan which alsoshows that most of the growth in these countriescan be explained by the growth of factor inputsand not technical progress, so that, according toYoung, there has been no growth miracle in thesecountries – contrary to the conventional wisdom.

Before accepting this conclusion, however, theobserver still has to explain why there was such arapid growth of factor inputs, and it is this pointwhich exposes the fundamental weakness of theproduction function approach to the analysis ofgrowth performance. Inputs are not manna fromheaven dropped by God. Something ‘miraculous’must have been driving these economies, to which


input growth responded. On closer inspection,what distinguishes these countries is their outwardorientation and relentless search for exportmarkets, and their remarkable growth of exportswhich confers benefits on an economy from boththe demand and the supply side (see Chapter 4).This exposes another weakness of neoclassicalgrowth theory and that is that the models areclosed. There is no trade in these simple models,and no balance of payments to worry about. Theyare supply-oriented, supply-driven, closedeconomy models unsuitable for the analysis ofopen economies in which foreign exchange isinvariably a scarce resource acting to constrain thegrowth process. We return to this topic in Chapters4 and 5, but first we must look at the challenge of‘new’ growth theory.

‘New’ Endogenous Growth Theory

Since the mid-1980s there has been an outpouringof literature and research on the applied economicsof growth, attempting to understand and explaindifferences in output growth and living standardsacross countries of the world – most inspired byso-called ‘new’ growth theory or endogenousgrowth theory. This spate of studies seems to havebeen prompted by a number of factors: firstly, bythe increased concern with the economic perfor-mance of poorer parts of the world, andparticularly major differences between continentsand between countries, with South East Asia


forging ahead, Africa left behind and SouthAmerica somewhere in the middle; secondly, bythe increased availability of standardized data onwhich to do research (Summers and Heston, 1991);and thirdly, by studies showing no convergence ofper capita incomes in the world economy (forexample, Baumol, 1986), contrary to the predictionof neoclassical growth theory based on theassumption of diminishing returns to capital.

If there are not diminishing returns to capital –but, say, constant returns – a higher capital–labourratio will be exactly offset by a higher output perhead,2 and the capital–output ratio will not behigher in capital-rich countries than in capital-poorcountries, and the savings–investment ratio willtherefore matter for long-run growth. Growth isendogenously determined in this sense and notsimply determined by the exogenous rate ofgrowth of the labour force and technical progress.This is the starting point for ‘new’, endogenousgrowth theory which seeks an explanation of whythere has not been a convergence of livingstandards in the world economy.

The explanation of ‘new’ growth theory is thatthere are forces at work which prevent themarginal product of capital from falling (and thecapital–output ratio from rising) as moreinvestment takes place as countries get richer. PaulRomer (1986) first suggested externalities toresearch and development (R&D) expenditure.Robert Lucas (1988) focuses on externalities tohuman capital formation (education). Grossman


and Helpman (1991) concentrate on technologicalspillovers from trade and foreign direct investment(FDI). Other economists have stressed the role ofinfrastructure investment and its complementar-ity with other types of investment. In fact, it can beseen from the formula for the capital–output ratiothat increasing returns to labour for all sorts ofreasons could keep the capital–output ratio fromrising.

So now let us turn to ‘new’ growth theory, seewhat it has to say, see whether it is saying anythingnew, and consider some of the problems of inter-preting the empirical results from testing newgrowth theory.

The first crude test of new growth theory is toobserve whether or not there is an inverse relationacross countries between the growth of output perhead and the initial level of per capita income ofcountries. If there is, this would be supportive ofthe neoclassical prediction of convergence. If not,it would be supportive of ‘new’ growth theory thatthe marginal product of capital does not decline.This is referred to as the test for beta (β) conver-gence. It can be said straight away that no globalstudies find evidence of unconditional beta conver-gence. Virtually all studies find evidence ofdivergence. The coefficient linking the growth ofoutput per head to the initial level of per capitaincome is positive, not negative.

Before jumping to the conclusion that this isunequivocal support for ‘new’ growth theory,however, it must be remembered that the neoclas-


sical prediction of convergence assumes all otherthings the same across countries: populationgrowth; tastes and preferences (for example, thesavings ratio); technology and so on. Since theseassumptions are manifestly false, there can neverbe the presumption of unconditional convergence –only conditional convergence controlling for differ-ences in all other factors that affect the growth ofliving standards, including differences in the ratioof investment to GDP and variables that affect theproductivity of capital and labour such aseducation and training, R&D expenditure, trade,macroeconomic performance and political stability.The question is, what happens to the sign on theinitial per capita income variable when thesecontrol variables are introduced into the equation?If the sign on initial per capita income turnsnegative, this is supposed to represent a rehabili-tation of the neoclassical model. In other words,living standards would converge if only levels ofinvestment, education, R&D expenditure and soon were the same in poor countries as richcountries, but they are not! The argument is remi-niscent of the way neoclassical economistscontinue to work with fictitious models of com-petitive equilibrium in the presence of increasingreturns, by treating the latter as externalities (thedevice originally adopted by Alfred Marshall in1890). Indeed, most ‘new’ growth theorists, andparticularly Robert Barro (1991), are clearly neo-classical economists in disguise. We will look at thework of Barro and others later, but first let us


consider the ‘newness’ of ‘new’ growth theory andthe interpretation of results.

First, I find it amusing that it seems to havecome as a surprise to many members of theeconomics profession that living standards in theworld have not been converging according to theprediction of neoclassical growth theory. Longbefore the advent of ‘new’ growth theory, many‘non-orthodox’ economists had been pointing towidening divisions in the world economy, anddeveloped models to explain divergence. That iswhat the centre–periphery models of Prebisch(1950), Myrdal (1957), Hirschman (1958), Seers(1962) and the neo-Marxist school (for example,Emmanuel, 1972; Frank, 1967) were all about,many based on a combination of internationaltrade and increasing returns.

Secondly, it has to be said that many of the ideasof ‘new’ growth theory are not new at all. Who,apart from strict adherents to the neoclassicalmodel, ever believed that investment did notmatter for long-run growth? Kaldor (1957), withhis technical progress function, precisely antici-pated new growth theory by arguing that technicalprogress requires capital accumulation and capitalaccumulation requires technical progress (it isimpossible to have one without the other), and hismodel of growth gives an explanation of why thecapital–output ratio stays constant through timedespite a rising ratio of capital to labour (see later).On the origins of increasing returns, we couldmention Adam Smith and the division of labour


(see Chapter 1), Allyn Young and the idea ofincreasing returns as a macroeconomicphenomenon related to the interaction betweenactivities (see Chapter 1), Kenneth Arrow’s modelof learning by doing (Arrow, 1962), the work ofSchultz (1961) and Denison (1962) on the socialreturns to education, and the work of Griliches(1958) on the social returns to R&D. We have anendearing tendency in economics to reinvent thewheel.

Thirdly, when it comes to interpreting theempirical results from testing models of newgrowth theory and convergence, some care needsto be taken. In particular, great care needs to beexercised in interpreting the negative sign on theinitial level of per capita income as necessarilyrehabilitating the neoclassical model of growth, asfor example, Barro (1991) does, because there areother conceptually distinct reasons for expecting anegative sign. Firstly, outside the neoclassicalparadigm, there is a whole body of literature thatargues that economic growth should be inverselyrelated to the initial level of per capita incomebecause, the more backward a country, the greaterthe scope for catch-up; that is, for absorbing abacklog of technology, which represents a shift inthe whole production function. Is conditional con-vergence picking up diminishing returns to capitalin the neoclassical sense, or catch-up? The twoconcepts are conceptually distinct, but not easy todisentangle empirically. Secondly, the negativeterm could simply be picking up structural change,


with poor countries growing faster than richcountries (controlling for other variables) becauseof a more rapid shift of resources from low pro-ductivity to high productivity sectors (for example,from agriculture to industry). How do we dis-criminate between these hypotheses?

A fourth point concerns the specification of‘new’ growth theory in its simplest form as the so-called AK model:

Y = AK, (2.6)

where A is a constant, which implies a constantproportional relation between output (Y) andcapital (K), or constant returns to capital. On closeinspection, this specification is none other than theHarrod growth equation g = s/c (see Chapter 1).This can be seen by taking changes in Y and K anddividing by Y, which gives:

ΔY/Y = A ΔK/Y = A (I/Y), (2.7)

where ΔY/Y is the growth rate (g); I/Y is thesavings–investment ratio (s), and A is the produc-tivity of investment, ΔY/I = 1/c or the reciprocalof the incremental capital–output ratio. What thismeans is that, if the productivity of investment (A)was the same across all countries, there would bea perfect correlation between growth and theinvestment ratio. If there is not a perfect correla-tion, then definitionally there must be differencesacross countries in the productivity of capital. All


that empirical studies of ‘new’ growth theory arereally doing is trying to explain differences in theproductivity of capital across countries (providedthe investment ratio is in the equation) in terms ofdifferences in education, R&D expenditure, tradeand so on, and initial endowments (see Husseinand Thirlwall, 2000, for further elaboration of thispoint).

As far as the constancy of the capital–outputratio is concerned, it was pointed out by Kaldor(1957) many years ago, as one of his six stylizedfacts of economic growth, that, despite capitalaccumulation and increases in capital per headthrough time, the capital–output ratio hasremained broadly unchanged, implying some formof externalities or increasing returns. It is worthquoting Kaldor in full:

As regards the process of economic change and develop-ment in capitalist societies, I suggest the following‘stylised facts’ as a starting point for the construction oftheoretical models – (4) steady capital–output ratios overlong periods; at least there are no clear long-term trends,either rising or falling, if differences in the degree of capitalutilization are allowed for. This implies, or reflects, thenear identity in the percentage growth of production andof the capital stock i.e. for the economy as a whole, andover long periods, income and capital tend to grow at thesame rate.

Kaldor’s explanation lay in his innovation of thetechnical progress function (TPF) relating thegrowth of output per man (q

.) to the growth of

capital per man (k.), as in Figure 2.2.


The position of the (linear) TPF drawn in Figure2.2 depends on the exogenous rate of technicalprogress, and the slope of the function depends onthe extent to which technical progress is embodiedin capital. Along the 45° line, the capital–outputratio is constant, and the equilibrium growth ofoutput per head is q

.1*. An upward shift of the

function associated with new discoveries, techno-logical breakthroughs and so on will cause thegrowth of output to exceed the growth of capital,raising the rate of profit and inducing moreinvestment, to give a new equilibrium growth of


q. = f (k

.)

q.

q.*

1

q.*

2

k.*1

k.*2

k.

x

x1

45°

Figure 2.2

output per head at q.2* (follow the arrows). An

increase in capital accumulation not accompaniedby technical progress will simply cause thecapital–output ratio to rise. If the capital–outputratio is observed to be constant there must be tech-nological forces at work shifting the functionupwards. ‘New’ growth theory is precisely antici-pated.

What applies to countries through time appliespari passu to different countries at a point in time,with differences in country growth rates at thesame capital–output ratio associated with differenttechnical progress functions. To quote Kaldoragain:

A lower capital–labour ratio does not necessarily implya lower capital–output ratio – indeed, the reverse is oftenthe case. The countries with the most highly mechanisedindustries, such as the United States, do not require ahigher ratio of capital to output. The capital–output ratioin the United States has been falling over the past 50 yearswhilst the capital–labour ratio has been steadily rising;and it is lower in the United States today than in the man-ufacturing industries of many underdeveloped countries.(Kaldor, 1972)

In other words, rich and poor countries are simplynot on the same production function.

A final point concerns the way that new growththeory models trade. First of all, some of themodels and empirical studies do not consider therole of trade at all, as if economies are completelyclosed. It is hard to imagine how it is possible to


explain growth rate differences between countrieswithout reference to trade, and particularlywithout reference to the balance of payments ofcountries which constitutes for many developingcountries the major constraint on the growth ofdemand and output (which will reduce the pro-ductivity of capital). When a trade variable isincluded in the model, it is invariably insignificant,or loses its significance when combined with othervariables. On the surface, this is a puzzle. It wouldconflict with the rich historical literature that existson the relation between trade and growth(Thirlwall, 2000). It would conflict with thevoluminous work of the World Bank and otherorganizations showing the beneficial effects oftrade liberalization, and it would undermine thewhole thrust of international policy making sinceWorld War II, which has been to free up marketsand to promote trade in the interests of economicdevelopment.

There may be several explanations for the weakresults, but I believe the major one is that the tradevariable normally taken is the share of exports inGDP as a measure of ‘openness’ which may pickup the static gains from trade and technologicalspillovers, but not the dynamic effects of tradewhich can only be properly captured by the growthof exports which affects demand, both directly andindirectly (by relaxing a balance of paymentsconstraint on demand), and also the supply side ofthe economy by permitting a faster growth ofimports. This point relates to my general criticism


of ‘new’ growth theory that it neglects demand-side variables. When an export growth variable isincluded in a ‘new’ growth theory equation, it ishighly significant (see Thirlwall and Sanna,1996).

When it comes to evaluating the empiricalevidence, only four variables in ‘new’ growththeory equations appear to be robust (see Levineand Renelt, 1992): the initial level of per capitaincome, the savings–investment ratio, investmentin human capital, and population growth(usually). All other variables are fragile in the sensethat, when they are combined with other variables,they lose their significance. The robust variablesare ones that growth analysts have stressed formany years, long before the advent of ‘new’growth theory. Plus ça change, plus c’est la mêmechose.

Notes

1. This can be seen by rearranging the equation q = [(l + t)/s]k to qs/k = l + t, where q = Y/L; s = S/Y = ΔK/Y (since allsaving leads to capital accumulation) and k = K/L.Therefore (Y/L) (ΔK/Y) (L/K) = ΔK/K = l + t.

2. Remember K/Y = (K/L)/(Y/L).


3. Manufacturing industry as the engine of growth

The neoclassical approach to economic growth,and its offspring ‘new’ growth theory, are not onlyvery supply-oriented, treating factor supplies asexogenously given, but are also very aggregative.They treat all sectors of the economy as if they arealike. They do not explicitly pick out any one sectoras more important than another. In practice,however, aggregate growth will naturally berelated to the rate of expansion of the sector withthe most favourable growth characteristics.

There is a lot of historical, empirical evidence tosuggest that there is something special aboutindustrial activity, and particularly manufacturing.There seems to be a close association acrosscountries between the level of per capita incomeand the degree of industrialization, and there alsoseems to be a close association across countriesbetween the growth of GDP and the growth ofmanufacturing industry. Countries which aregrowing fast tend to be those where the share ofindustry in GDP is rising most rapidly: the so-called ‘newly industrializing countries’ (the NICs).Is this an accident?

40

One of the first economists to have seriouslyaddressed this issue is the late Nicholas Kaldor,who argued in many of his writings (see Targettiand Thirlwall, 1989) that it is impossible tounderstand the growth and development processwithout taking a sectoral approach, distinguishingbetween increasing returns activities on the onehand (which he associated with industry) anddiminishing returns activities on the other (whichhe associated with the land-based activities of agri-culture and mining). Kaldor first articulated histheory about why growth rates differ in twolectures: one in Cambridge in 1966 entitled Causesof the Slow Rate of Economic Growth of the UnitedKingdom (Kaldor, 1966); the other at CornellUniversity in the same year entitled StrategicFactors in Economic Development (Kaldor, 1967). Inthese lectures he presented a series of ‘laws’ orempirical generalizations which attempted toaccount for growth rate differences betweenadvanced capitalist countries, but which also haveapplicability to developing countries as well.

There are three laws to focus on, plus a numberof subsidiary propositions. The first law is thatthere exists a strong causal relation between thegrowth of manufacturing output and the growthof GDP. The second law states that there exists astrong positive causal relation between the growthof manufacturing output and the growth of pro-ductivity in manufacturing as a result of static anddynamic returns to scale. This is also known asVerdoorn’s Law (see Chapter 1 and later). The third

Manufacturing industry as the engine of growth 41

law states that there exists a strong positive causalrelation between the rate at which the manufac-turing sector expands and the growth ofproductivity outside the manufacturing sectorbecause of diminishing returns in agriculture andmany petty service activities which supply labourto the industrial sector. If the marginal product oflabour is below the average product in thesesectors, the average product (productivity) will riseas employment is depleted. For this reason, overallGDP growth will tend to slow up as the scope forabsorbing labour from diminishing returnsactivities dries up.

Given these ‘laws’, the question remains of whatdetermines the growth of the manufacturing sectorin the first place. Kaldor’s answer is demandcoming from agriculture in the early stages ofdevelopment and export growth in the later stages.These are the two fundamental sources ofautonomous demand to match the leakages ofincome from the industrial sector of food importsfrom agriculture on the one hand and imports fromabroad on the other. A fast growth of exports andoutput may then set up a virtuous circle of growthwith rapid export growth leading to rapid outputgrowth, and rapid output growth leading to fastexport growth through the favourable impact ofoutput growth on competitiveness. Other countriesfind it difficult to break into such virtuous circles,and this is why the polarization between countriesoccurs. The present north–south divide in theworld economy has its origins in the fact that the


‘north’ contains the first set of countries to indus-trialize, and only a handful of countries since havemanaged to challenge their industrial supremacyand to match their living standards.

Kaldor’s growth laws can be tested acrosscountries, across regions within countries, acrossregions and countries using panel data (forexample across the regions of the European Union)and for individual countries using time series data(although care has to be taken with the second lawnot to confuse Verdoorn’s Law with Okun’s Law,which relates to pro-cyclical variations in produc-tivity over the trade cycle). (See McCombie andThirlwall, 1994.)

The first test of the first law is to run a regressionof the rate of growth of GDP against the rate ofgrowth of manufacturing output and to test for sta-tistical significance. When this is done acrosscountries or regions, the relation is invariablyhighly significant, but this could be a spuriousrelation due to the fact that manufacturing outputconstitutes a sizeable fraction of total output. Sidetests therefore need to be undertaken. One is toregress the growth of GDP on the excess of thegrowth of manufacturing output over the growthof non-manufacturing output; another is to regressthe growth of non-manufacturing output on thegrowth of manufacturing output. When these sidetests are performed, the first law is generallyconfirmed. A recent interesting study across theregions of China strongly supports Kaldor’s firstlaw (Hansen and Zhang, 1996). For manufacturing


to be regarded as special, however, it needs to beshown that GDP growth is not closely related tothe growth of other sectors such as agriculture,mining or services. It is hard to find any significantcross-section relation between the growth of GDPand the growth of the agricultural sector. Therelation between the growth of GDP and thegrowth of services is stronger but there is reason tobelieve that the direction of causation may be theother way round from GDP growth to servicegrowth since the demand for many services isderived from the demand for manufacturingoutput itself. The question is to what extent serviceactivities have an ‘autonomous’ existence, andwhether they have the production characteristics(for example, static and dynamic scale economies)to induce fast growth. This is still an open question,ripe for further research.

If the first law is accepted, what accounts for thefact that, the faster manufacturing output growsrelatively to GDP, the faster GDP seems to grow?Since differences in growth rates are largelyaccounted for by differences in labour productiv-ity growth (rather than the growth of the labourforce), there must be some relationship betweenthe growth of the manufacturing sector and pro-ductivity growth in the economy as a whole. Thisis to be expected for two main reasons. The first isthat, wherever industrial production andemployment expand, labour resources are drawnfrom sectors which have open or disguised unem-ployment (that is, where there is no relation


between employment and output), so that labourtransference to manufacturing will not cause adiminution in the output of these sectors, and pro-ductivity growth increases outside manufacturing(the third law – see below).

The second reason is the existence of increasingreturns within industry, both static and dynamic.Static returns relate to the size and scale ofproduction units and are a characteristic largely ofmanufacturing where, for example, in the processof doubling the linear dimensions of equipment,the surface increases by the square and the volumeby the cube (the so-called ‘cube rule’). Dynamiceconomies refer to increasing returns broughtabout by ‘induced’ technical progress, learning bydoing, external economies in production and soon. Kaldor draws inspiration here from AllynYoung’s pioneering paper of 1928, ‘IncreasingReturns and Economic Progress’ with its emphasison increasing returns as a macroeconomicphenomenon resulting from the interactionbetween activities in the process of generalindustrial expansion; ideas now taken up by ‘new’growth theory (see Chapter 2). For those interestedin the history of economic thought and the inter-generational transmission of ideas (whichsometimes take a long time to resurface!), Kaldorwas a pupil of Allyn Young at the London Schoolof Economics in 1928 and took a full set of lecturenotes from him, including his thoughts onincreasing returns (see Thirlwall, 1987a;Sandilands, 1990).


The empirical relation between productivitygrowth and output growth in manufacturing isknown as Verdoorn’s Law, following Verdoorn’s(1949) paper published in Italian, entitled ‘Fattoriche Regolano lo Sviluppo della Produttivita delLavoro’. Interestingly, at the time of publication,Verdoorn was working for Kaldor in the Researchand Planning Division of the EconomicCommission for Europe in Geneva, of which Kaldorwas Director. It was Kaldor who revived Verdoorn’sLaw in 1966, and it is also known as Kaldor’ssecond law; that is, there is a strong positive causalrelation between the growth of manufacturingoutput and the growth of productivity in manu-facturing. In recent years, the relation has beenextensively tested across countries (Kaldor, 1966;Michl, 1985); across regions within countries forboth developed and developing countries(McCombie and de Ridder, 1983; Fingleton andMcCombie, 1998; Leon-Ledesma, 2000a; Hansenand Zhang, 1996) and across industries (McCombie,1985a). Typically, the estimated Verdoorn coefficientis 0.5, which means that manufacturing outputgrowth is split evenly between induced productiv-ity growth on the one hand and employmentgrowth on the other. The relation is always robustfor manufacturing and industry more broadly. Theprimary sector of agriculture and mining reveals nosuch relationship, but some studies (for example,Leon-Ledesma, 2000) find evidence of a Verdoornrelation also operating in service activities, althoughnot so strongly.


There are a number of ways in which theVerdoorn relation can be generated. Verdoornhimself derived it from a static Cobb–Douglasproduction function where the coefficient linkingoutput growth and productivity growth dependson the parameters of the production function, theexogenous rate of technical progress and the rateat which capital is growing relative to the labourforce. The Verdoorn coefficient can also be thoughtof, however, as a much more dynamic relationlinked to Kaldor’s technical progress function (seeChapter 2) where the coefficient depends on therate at which capital accumulation is induced byoutput growth (the accelerator effect), the extent towhich technical progress is embodied in capital(reflected in the slope of the technical progressfunction) and the rate of disembodied technicalprogress induced by growth (learning by doing).

The estimation of the Verdoorn relation, byregressing productivity growth on output growth,is not without its critics, however, because thequestion has been raised, quite rightly, of what iscause and what is effect. Some argue that thedirection of causation could be from fast produc-tivity growth to fast output growth because fastproductivity growth causes demand to expandfaster through improved competitiveness. In this(opposite) view, all productivity growth would beautonomous; none induced by output growthitself. Also, for the mechanism to work, the priceelasticities of demand would have to be relativelyhigh and wage growth would have to lag behind


productivity growth for relative prices to fall.Kaldor did not deny the reverse causationargument – indeed it is part of his export-ledgrowth model (see Chapter 4) – but his argumentwas always that it would be very difficult toexplain such large differences in productivitygrowth in the same industry over the same timeperiod in different countries without reference tothe growth of output itself. To assume that all pro-ductivity growth is autonomous would be a denialof the existence of dynamic scale economies andincreasing returns. The two-way relation betweenoutput growth and productivity growth doesmean, however, that the Verdoorn relation shouldbe estimated using simultaneous equationmethods to avoid biased estimates of the Verdoorncoefficient.

Whether or not Verdoorn’s Law holds, it is not,contrary to the popular view, an indispensableelement of the complete Kaldor model. Even in theabsence of induced productivity growth in themanufacturing sector (which is difficult to believe)the growth of industry would still be the governingfactor determining overall output growth as longas resources used by industry represent a netaddition to output either because they wouldotherwise have been unused or because of dimin-ishing returns elsewhere, or because industrygenerates its own resources in a way that othersectors do not by the reinvestment of profits. Thisleads on to Kaldor’s third law which states that,the faster the growth of manufacturing output, the


faster the rate of labour transference from non-manufacturing, so that productivity growth innon-manufacturing is negatively associated withthe growth of employment outside manufacturing.In practice, it is difficult to measure productivitygrowth in many non-manufacturing activitiesbecause output can only be measured by inputs.But it is possible to relate the overall rate of pro-ductivity growth in the economy as a whole toemployment growth in non-manufacturing, con-trolling for differences in the growth ofmanufacturing employment or output. When thisis done, Kaldor’s third law is generally supported.The study referred to earlier across the regions ofChina by Hansen and Zhang estimates thefollowing equation:

p = 0.02 + 0.49 (gm) – 0.82 (enm) (3.1)(16.4) (5.4)

where p is overall productivity growth; gm is thegrowth of manufacturing output and enm is thegrowth of employment in non-manufacturing. Thesign on enm is negative and significant, ashypothesized, and the sign on gm is positive andsignificant (bracketed terms are t values).

There are a number of subsidiary propositionswhich complete Kaldor’s wide vision of thegrowth and development process. Following onfrom the third law, as surplus labour becomesexhausted in the non-manufacturing sector, andproductivity levels tend to equalize across sectors,


the degree of overall productivity growth inducedby manufacturing output growth is likely todiminish. This is why country growth rates tend tobe fastest in the take-off stage of development anddecelerate in maturity (to use Rostow’s terminol-ogy). It is in this sense that countries at a high levelof development may suffer from a ‘labourshortage’, not in the sense that manufacturingoutput growth itself is constrained by a shortageof labour because labour is a very elastic factor ofproduction, as we shall argue in Chapter 6. Themanufacturing sector can always get the labour itwants, although it may have to pay a higher realwage which eats into profits and investment (à laLewis and Marx). What may constrain manufac-turing output growth is not a shortage of labourbut demand from agriculture in the early stages ofdevelopment and exports in the later stages. Anascent industrial sector needs a market to sell to.In the pre-take-off stage of development, agricul-ture is by far the largest ‘external’ sector; hence theimportance of rising agricultural productivity toprovide the purchasing power and growingmarket for industrial goods.

Kaldor’s two-sector model of agriculture andindustry (Kaldor, 1996; Thirlwall, 1986) shows theimportance of establishing an equilibrium terms oftrade between the two sectors if the growth of theeconomy is to be maximized, so that industrialgrowth is neither supply-constrained because agri-cultural prices are too high relative to industrialprices, or demand constrained because they are too


low. Through time, however, the importance ofagriculture as an autonomous market for industrialgoods will diminish and exports will take over, anda fast growth of exports and industrial output willtend to set up a virtuous circle of growth workingthrough Verdoorn’s Law and other feedback, rein-forcing mechanisms. Fast export growth leads tofast output growth; fast export growth depends oncompetitiveness and the growth of world income;competitiveness depends on the relationshipbetween wage growth and productivity growth;and fast productivity growth depends on fastoutput growth. The circle is complete.

I shall outline this model more fully in the nextchapter. Suffice it to say, at this point, that acountry ignores the performance of its manufac-turing sector at its peril, but the foundations mustfirst be laid for the manufacturing sector toprosper. Balanced growth is required betweenindustry and agriculture, and between internalgrowth and the traded goods sector if balance ofpayments problems are to be avoided. It is to therole of exports and the balance of payments thatwe now turn.


4. A demand-oriented approachto economic growth: export-led growth models

In Chapter 2 it was argued that ‘new’ growththeory is an improvement on old (neoclassical)growth theory in the sense that it can explain whywe do not observe convergence in the worldeconomy, but ‘new’ growth theory is still open tothe same criticism as old growth theory; that it issupply-oriented. Moreover, ‘new’ growth theory isnot the only model in town to explain divergenttrends in the world economy. In neoclassicaltheory, output growth is a function of factor inputsand factor productivity with no recognition thatfactor inputs are endogenous, and that factor pro-ductivity growth may also be a function of thepressure of demand in an economy. In practice,labour is a derived demand, derived from thedemand for output itself. Capital is a producedmeans of production, and is therefore as much aconsequence of the growth of output as its cause.Factor productivity growth will be endogenous ifthere are static and dynamic returns to scale.

As a starting point for the analysis of growth,therefore, it would seem just as sensible, if not

52

more so, to take a (Keynesian) demand-orientedapproach to growth and ask what are the majorconstraints on demand, and assume that demandconstraints generally bite long before supply con-straints become operative. In static macro theory,students are taught that national income (oroutput) is the sum of consumption expenditure,investment and exports, minus imports. In growthanalysis, why not teach that national incomegrowth is the weighted sum of the growth of con-sumption, investment and the balance betweenexports and imports, and proceed from there? Ifwe take this approach, the role of exports is imme-diately apparent. Exports differ from othercomponents of demand in three importantrespects. Firstly, exports are the only truecomponent of autonomous demand in aneconomic system, in the sense of demandemanating from outside the system. This is veryimportant to bear in mind. The major part of con-sumption and investment demand is dependenton the growth of income itself. Secondly, exportsare the only component of demand that can payfor the import requirements for growth. It may bepossible to initiate consumption-led growth,investment-led growth or government expendi-ture-led growth for a short time, but each of thesecomponents of demand has an import content (thatis why imports are subtracted in the nationalincome equation). If there are no export earningsto pay for the import content of other componentsof expenditure, demand will have to be con-

A demand-oriented approach to economic growth 53

strained. In this respect, exports are of great sig-nificance if balance of payments equilibrium oncurrent account is a long-run requirement. What itmeans is that exports have not only a direct effecton demand, but also an indirect effect by allowingall other components of demand to rise faster thanotherwise would be the case. This is the idea of theHicks supermultiplier (Hicks, 1950; McCombie,1985b) in which the rate of growth of an economybecomes attuned to the rate of growth of thedominant component of autonomous demand,which in the case of the open economy is exports.The third important aspect of exports is thatimports (permitted by exports) may be moreproductive than domestic resources becausecertain crucial goods necessary for development(such as capital goods) are not produced domesti-cally. This is the supply-side argument forexport-led growth.

It can then be shown that, if there are increasingreturns and induced productivity growth, exportgrowth can set up a virtuous circle of growthwhich leads into centre–periphery models ofgrowth and development which, on certainconditions, predict divergence between regionsand countries in the world economy. In this chapterI develop this demand-oriented export-led growthmodel and consider the conditions under whichdivergence is likely to take place, but withoutimposing a balance of payments constraint. (Thatis done in Chapter 5.)


The Model

As mentioned already, the main idea behind themodel is that export demand is the most importantcomponent of autonomous demand in an openeconomy, so that the growth of exports will governthe long-run growth of output to which othercomponents of demand adapt. Thus we may write:

gt = γ (xt), (4.1)

where gt is the growth of output at time t and xt isthe growth of exports. But what determines thegrowth of exports? We can use a conventional mul-tiplicative (constant elasticity) export demandfunction which makes export demand a functionof relative prices measured in a common currency(competitiveness), and income outside the country:

Xt = A (Pdt/Pft)η Zt

ε, (4.2)

so that taking rates of change (lower-case letters):

xt = η (pdt – pft) + ε (zt), (4.3)

where Pd is domestic prices; Pf is competitors’prices measured in a common currency; Z isincome outside the country; η (< 0) is the priceelasticity of demand for exports; and ε (> 0) is theincome elasticity of demand for exports.

The growth of income outside the economy andforeign prices may be taken as exogenous, but the


growth of domestic prices is assumed to beendogenous, derived from a mark-up pricingequation in which prices are based on labour costsper unit of output plus a percentage mark-up:

Pdt = (Wt/Rt) (Τt), (4.4)

where W is the money wage rate, R is the averageproduct of labour, and Τ is 1 + % mark-up on unitlabour costs. Taking rates of change gives:

pdt = wt – rt + τt. (4.5)

Productivity growth, however, is partlydependent on the growth of output itself throughstatic and dynamic returns to scale: Verdoorn’sLaw (see Chapter 3):

rt = rat + λ (gt), (4.6)

where rat is autonomous productivity growth, andλ is the Verdoorn coefficient.

The Verdoorn relation opens up the possibilityof a virtuous circle of export-led growth. Themodel becomes circular because the faster thegrowth of output the faster the growth of produc-tivity; and the faster the growth of productivity theslower the growth of unit labour costs, and hencethe faster the growth of exports and output. Themodel also implies that, once a country obtains agrowth advantage, it will tend to sustain it.Suppose, for example, that an economy acquires


an advantage in the production of goods with ahigh income elasticity of demand in world markets(technology-based activities) which raises itsgrowth rate above that of other economies. Owingto the Verdoorn effect, productivity growth will behigher and the competitive advantage of theeconomy in these goods will be reinforced, makingit difficult for other economies to produce the samecommodities except through protection or excep-tional industrial enterprise. In centre–peripherymodels of growth and development, it is differ-ences between the income elasticity characteristicsof exports and imports which lie at the core of theproblem for the periphery and at the heart of thesuccess of the centre (Thirlwall, 1983).

The equilibrium solution of the model isobtained by successive substitution of (4.6) into(4.5), the result into (4.3) and this into (4.1) whichgives:

(4.7)

Remembering that η < 0, the equilibrium growthrate is shown to vary positively with autonomousproductivity growth, the rate of growth of foreignprices and the growth of world income, andnegatively with domestic wage growth and anincrease in the mark-up. The Verdoorn coefficient(λ) serves to exaggerate growth rate differences

g

w r p zt

t at t ft t=

− + −( ) + ( )[ ]+

γη τ ε

γηλ1.


between economies arising from differences inother parameters and variables (that is, the higherλ, the smaller the denominator, since η < 0). If λ = 0,there is no exaggeration of differences.

Now it is an interesting question whethercountry growth rates will tend to diverge throughtime. This depends on the behaviour of the modelout of equilibrium. In a two-country model, anecessary condition for divergence is that thegrowth rate of one of the countries diverges fromits own equilibrium rate. One way to consider amodel in disequilibrium, and to examine itsdynamics, is to put lags into the equation. If we puta one-period lag into the export growth equation(4.2), we obtain a first-order difference equation,the solution to which is:

gt = A (– γηλ)t + particular (equilibrium) solution.(4.8)

Since η < 0, (– γηλ) > 0, so there are no cycles. If⎢γηλ| > 1, there will be explosive growth as tincreases. If ⎜γηλ⎜ < 1, there will be convergence toequilibrium. If, for the moment, it is assumed thatγ = 1, this would mean there would be cumulativedivergence away from equilibrium if ⎜– ηλ⎜ > 1.Given a Verdoorn coefficient of 0.5, this wouldimply a price elasticity of demand for exportsgreater than 2. This is possible.

In practice, however, it is not usual to observegrowth rates between countries diverging throughtime. Levels of per capita income diverge, but not


the growth of output. Growth rates betweencountries differ not because we observe countriesin the process of divergence but because the equi-librium growth rates differ, associated mainly withdifferences in the income elasticity of demand forexports (ε). What keeps growth on its equilibriumpath is likely to be a balance of payments equilib-rium requirement. Typically, imports grow fasterthan output. This means that exports must alsogrow faster than output. This implies that γ inequation (4.1) will be substantially less than unity.If relative price changes are ruled out as a balanceof payments adjustment mechanism, γ will be thereciprocal of the income elasticity of demand forimports. For example, if the import elasticity is 2,then γ = 0.5. This means (from equation 4.8) thatthe price elasticity of demand for exports wouldhave to be greater than 4 for divergence from equi-librium to occur. Such a high elasticity foraggregate exports is highly unusual.1

If the above model is simply treated as anexport-led growth model with no feedbackmechanism through the Verdoorn effect, andrelative prices are held constant, equation (4.7)reduces to:

gt = γ ε (zt). (4.9)

If a balance of payments constraint is imposed, γ =1/π, where π is the income elasticity of demand forimports. Therefore:


gt = ε (zt)/π (4.10)

or

gt/zt = ε/π. (4.11)

This says that one country’s growth rate relative toall others (z) is equiproportional to the ratio of theincome elasticity of demand for exports andimports. I discovered this rule in 1979 (Thirlwall,1979), which turns out to be the dynamic analogueof the static Harrod trade multiplier (Harrod, 1933)and I elaborate it and its implications more fully inthe next chapter. Paul Krugman (1989) discoveredit ten years later and for obvious reasons called itthe 45-degree rule (relative growth rates areequiproportional to relative income elasticities).However, he reverses the direction of causation,which makes him an orthodox neoclassicaleconomist as far as growth theory is concerned. Inhis model, the growth of the labour forcedetermines output growth, and fast output growthleads to fast export growth – hence an apparentlyhigher income elasticity of demand for exports.The direction of causation is therefore from growthto export elasticities, not from elasticities to growth.It is tautologically true, of course, that if faster-growing countries manage to sell more exports,they will be observed to have a higher elasticity,but the model does not explain how fast growtharises in the first place (except by the assumptionof a faster growth of the labour force), or why a


faster-growing country will necessarily exportmore independent of the characteristics of thegoods it produces. Greater supply availabilityand/or variety is not sufficient if demand isrelatively lacking.

In the final analysis, it is a question of to whatextent income elasticities can be considered asexogenously determined and to what extent theyare endogenously determined by the growth ofoutput itself. In this respect, it should not beforgotten that, in many instances, countries’income elasticities are largely determined bynatural resource endowments and the characteris-tics of the goods produced which are the productof history and independent of the growth ofoutput. An obvious example is the contrastbetween primary product production andindustrial production, where primary productstend to have an income elasticity of demand lessthan unity (Engel’s Law) while most industrialproducts have an income elasticity greater thanunity. In my model, where the direction ofcausation is from elasticities to growth, the elas-ticities reflect the structure of production. This isthe basic assumption of all the classiccentre–periphery models including those ofPrebisch, Myrdal and Seers, and also Kaldor (1970).Even between industrial countries (with whichKrugman is primarily concerned), feedbackmechanisms of the type already described(associated with Verdoorn’s Law) will tend toperpetuate initial differences in income elasticities


associated with ‘inferior’ industrial structures onthe one hand and ‘superior’ industrial structureson the other.

Empirical Evidence on Exports and Growth

Let us now consider the interpretation of theempirical evidence that exists on the relationshipbetween export growth and GDP growth. Therehas been a massive amount of research in recentyears showing a link between export and outputgrowth (see Thirlwall, 2000 for a survey). In fact,output growth probably correlates more closelywith export growth than any other variableintroduced into growth equations. There is likelyto be bi-directional causality through mechanismsdescribed earlier. The causal mechanism by whichexport growth affects output growth, however, isoften not specified and, where it is, it is normallya neoclassical supply-side argument. It is assumedthat the export sector has a higher level of pro-ductivity than the non-export sector, and that,because of exposure to foreign competition, theexport sector confers externalities on the non-export sector. Therefore both the share of exportsin GDP and the growth of exports matters foroverall growth performance.

Feder (1983) was the first to develop a formalmodel on these lines which fits neatly intomainstream neoclassical growth theory, where theconventional production function is augmented bythree terms: the growth of exports, the share of


exports in GDP, and a coefficient combining thedifferential productivity and externality effects.The equation derived is:

g = a(I/Y) + b(dL/L) + [δ (1 + δ) + Fx] (X/Y) (dX/X),(4.12)

where I/Y is the investment ratio as a proxy forcapital accumulation; dL/L is the growth of thelabour force; X/Y is the share of exports in GDP;dX/X is the growth of exports; δ(1 + δ) measuresthe differential productivity effect between the twosectors and Fx measures the externality effect.Feder tests the model across 31 countries over theperiod 1964–73, first without export growth andthen with. The inclusion of dX/X improves con-siderably the explanatory power. Then the exportshare term is excluded to isolate the externalityeffect. The difference between the total export effecton growth and the externality effect is the differ-ential productivity effect. There is evidence of bothexternality and differential productivity effects.

The model certainly has plausibility, but it is apure supply-side argument, and not the onlypossible one. There are other possible supply-sidearguments, and also demand-side arguments,consistent with a strong correlation between exportand output growth. For example, as we argued atthe beginning, fast export growth permits fastimport growth. Imports, particularly of capitalgoods and intermediate inputs, are a vehicle forthe transfer of technology which can have spillover


effects on output (Grossman and Helpman, 1991).Also, if countries are short of foreign exchange, anddomestic and foreign resources are not fully sub-stitutable, more imports permit a fuller use ofdomestic resources. Esfahani (1991) recognizes thislatter point and re-estimates Feder’s equation for31 countries, including import growth as well asexport growth. The export growth variable nowloses its significance, while the import growthvariable is significant. The regression equation isalso run without export growth, and it is foundthat, once the import supply-side effect of exportsis taken into account, there is apparently no sig-nificant externality effect of exports left to explain.Esfahani concludes, ‘even though exports do notappear to have had much direct externality effecton GDP – export promotion policies in thesecountries can be quite valuable in supplyingforeign exchange which relieves import shortagesand permits output expansion’.

But even the Esfahani argument does not go farenough because, as argued at the beginning, thereare equally (if not more) important demand-sideconsiderations to take into account which wouldalso be consistent with finding a positive relationbetween export growth and GDP growth, but theseconsiderations are rarely articulated in themainstream trade and growth literature. Specifi-cally, export growth is a major component ofaggregate demand, and may set up a virtuouscircle of growth. But even more important, in mostdeveloping countries at least, the major constraint


on the growth of demand is the current balance ofpayments and a shortage of foreign exchange.Export growth relaxes this constraint and is uniquein allowing all other components of demand togrow faster without balance of payments difficul-ties arising. This is the simplest of all explanationsof the relationship between export and outputgrowth, and leads on to the modelling of balanceof payments constrained growth models.

Note

1. Recently, more sophisticated models of cumulativecausation have been developed (for example, Leon-Ledesma, 2000b) containing elements of both divergenceand convergence. Specifically, the export growth equationis augmented to include a technology variable whichdepends on cumulative output, education and the pro-ductivity gap between a country and the technologicalleader. The productivity growth equation also dependson the level of technology and the technological gap.Whether there is divergence or convergence becomes anempirical matter depending on the parameter values ofthe model.


5. Balance of payments constrained growth: theory and evidence

It has been a central feature of most of my ownwork on growth to try and put demand back intogrowth theory, and to argue that for most countriesdemand constraints bite long before supply con-straints operate, and that, to understand growthrate differences between countries over the longrun, the analysis and understanding of demandconstraints cannot be ignored. In an open economy,the major constraint on the growth of demand (andtherefore growth performance) is likely to be itsbalance of payments. At a theoretical level, it canbe stated as a fundamental proposition that nocountry can grow faster than that rate consistentwith balance of payments equilibrium on currentaccount unless it can finance ever-growing deficits,which, in general, it cannot. There is a limit to thedeficit to GDP ratio (Moreno Brid, 1998), and alimit to the debt to GDP ratio beyond which thefinancial markets become nervous and a countryis unable to borrow more. If capital flows areincluded in the model, every country must have agrowth rate consistent with its overall balance of

66

payments because, by definition, the total balanceof payments must balance.

At the empirical (observational) level, theevidence for the proposition I am making is thatmany countries find themselves in balance ofpayments difficulties, and have to constraingrowth, while the economy still has surpluscapacity and surplus labour. There are certainly notmany developing countries in the world that couldnot grow faster given the greater availability offoreign exchange. Obviously, not every country inthe world can be balance of payments constrainedsimultaneously, since the world itself is a closedtrading system, but it only requires one country orgroup of countries not to be constrained for all therest to be so. Constrainers in the past have beencountries such as Japan, Germany, Switzerland andmany of the oil-producing countries of the MiddleEast. Keynes at Bretton Woods recognized thedeflationary consequences of persistent balance ofpayments surpluses, and would have penalizedsurplus countries in the same way that deficitcountries are penalized, but his proposal wasrejected (see Thirlwall, 1987b, where the KeynesPlan ‘Proposals for an International ClearingUnion’ is reprinted).

Below, I develop a model of balance of paymentsconstrained growth, first without capital flows,and then including capital flows. I will then discusstests of the model, and examine some of the morerecent empirical evidence. The inspiration fordeveloping this class of model came in the 1970s

Balance of payments constrained growth 67

when I was working with a PhD student (R.J.Dixon) on regional export-led growth models (ofthe type outlined in Chapter 4), but where balanceof payments problems are not apparent in thenormal sense of regions within a country havingto defend an exchange rate, since they are part ofa common currency area. It occurred to me,however, that in these regional export-led growthmodels, when applied to countries, it cannot beassumed that there are no demand constraints atall. Suppose, for example, that the growth ratedetermined by the parameters of the model (as inequation (4.7)) leads to a faster growth of importsthan exports. The growth rate would not be sus-tainable. In other words, imports need modellingin export-led growth models, and the obviousapproach is to model starting with the condition ofcurrent account equilibrium.

The Model

The structure of the model is very simple. We startwith the balance of payments equilibriumcondition. We then specify export and importdemand functions (as in Chapter 4). Since importgrowth is a function of income growth, we can thensolve for the growth of income consistent withbalance of payments equilibrium.

Current account equilibrium is given by:

PdX = Pf ME (5.1)


where X is the quantity of exports; Pd is the priceof exports in domestic currency; M is the quantityof imports; Pf is the price of imports in foreigncurrency, and E is the exchange rate measured asthe domestic price of foreign currency. Taking ratesof growth gives:

pd + x = pf + m + e. (5.2)

The growth of exports (as in Chapter 4, but withthe exchange rate included) is given as:

x = η (pd – pf – e) + ε (z). (5.3)

But now imports need modelling. The importdemand function may be specified in the sameway as the export demand function: as a multi-plicative (constant elasticity) function in whichimports are related to competitiveness and todomestic income as a proxy for expenditure. Thus:

M = B (Pf E/Pd)ψ Yπ, (5.4)

where ψ (< 0) is the price elasticity of demand forimports; Y is domestic income, and π (> 0) is theincome elasticity of demand for imports. Takingrates of change gives:

m = ψ (pf + e – pd) + π (y). (5.5)


Substituting equations (5.3) and (5.5) into (5.2)gives the rate of growth of income consistent withbalance of payments equilibrium (yB):

yB = [(1 + η + ψ) (pd – pf – e) + εz]/π. (5.6)

Equation (5.6) expresses a number of interestingand familiar economic propositions:

1. an improvement in the real terms of trade, (pd – pf – e) > 0, will improve a country’s growthrate consistent with balance of payments equi-librium. This is the pure terms of trade effect onreal income growth;

2. one country’s prices rising faster than anothermeasured in a common currency will lower acountry’s balance of payments equilibriumgrowth rate if the sum of the (negative) priceelasticities is greater than unity: that is (1 + η +ψ) < 0;

3. currency depreciation (e > 0) will raise thebalance of payments equilibrium growth rateif the sum of the price elasticities is greater thanunity. This is the dynamic analogue of the staticMarshall–Lerner condition for an improvementin the balance of payments following currencydepreciation. Note, however, that a once-for-alldepreciation or devaluation of the currencycannot put a country on a permanently highergrowth path consistent with balance ofpayments equilibrium since in the period afterthe devaluation e = 0, and the growth rate would


revert to its former level. Using constantelasticity demand functions, the currency depre-ciation would have to be continuous, but thiswould soon feed through to domestic prices,nullifying the exchange rate advantage;

4. the equation shows the mutual interdependenceof countries because one country’s growth per-formance (y) is linked to all others (z). But howfast one country can grow relative to all others,while preserving its balance of payments,depends crucially on ε, the income elasticity ofdemand for exports. For some countries, ε isvery high (in the range 3 to 4); in other countriesit is very low (less than unity);

5. the balance of payments equilibrium growthrate is inversely related to its appetite forimports, measured by π.

If it is now assumed that relative prices measuredin a common currency remain unchanged,equation (5.6) reduces to:

yB = ε (z)/π = x/π. (5.7)

This is the dynamic analogue of the static Harrodtrade multiplier result Y = X/m (where Y is the levelof income, X is the level of exports and m is themarginal propensity to import), which Harrodderived in his book International Economics in 1933on the same assumptions as above, namely balanceof payments equilibrium and no change in the realterms of trade. I had not read Harrod before I


derived the result in equation (5.7), but when I didI realized I had reinvented the wheel, although itshould be emphasized that Harrod never derivedthe growth implications of his result. The Harrodtrade multiplier of 1/m was eclipsed by the closedeconomy Keynesian multiplier of 1/s (where s isthe propensity to save), but in the open economyit is probably more difficult to plug animport–export gap than it is to bridge asavings–investment gap and therefore the foreigntrade multiplier has more relevance for under-standing the macroeconomic performance ofcountries. If relative prices do not adjust in inter-national trade, or trade flows are relativelyinsensitive to price changes, it is output andgrowth that adjust to bring imports and exportsinto line.

The test of this model is to see how close the longrun growth of countries approximates to thepredicted growth rate x/π. If it is equal, or slightlyabove with countries running deficits, and thereexist unemployed domestic resources, this is prettyconvincing evidence (to me at least) that growth isbalance of payments constrained. When I firstapplied this simple model in 1979 to a series ofdeveloped countries, I did not perform any formalparametric tests. I simply observed how close theactual growth rate was to the predicted rate andestimated rank correlations. The actual andpredicted rates were remarkably close, and I ratherpompously remarked, ‘it might almost be stated asa fundamental law that – the rate of growth of a


country will approximate to the ratio of its rate ofgrowth of exports and its income elasticity ofdemand for imports. The approximation itselfvindicates the assumptions used to arrive at thesimple rule’. Ever since, this result has come to beknown in the literature as Thirlwall’s Law: not aspowerful as e = mc2(!), but a powerful predictor,nonetheless, of inter-country growth performance(for surveys, see Journal of Post Keynesian Economics,1997, and McCombie and Thirlwall, 1997).

There are parametric tests of the model. The twomain ones are as follows. The first is to run aregression of actual growth (y) on yB for a seriesof countries and test whether the constant term iszero and the regression coefficient is equal to unity.If so, yB will be a good predictor of y. This test hasbeen performed by some investigators with mixedresults, but there are at least two major problemswith the test. The first is that there will be bias if asample of countries is taken in which the balanceof payments deficits and surpluses do not cancelout: that is, if there is a systematic tendency for y >yB or y < yB. Secondly, there may well be outlierswhere y ≠ yB (for example, Japan) which gives aregression coefficient significantly different fromunity, leading to a rejection of the theory for allother countries.

A second (alternative) test which avoids theabove problems is to take each country separatelyand to estimate the income elasticity of demand forimports (say π') that would make y = yB, and thento compare this with the estimated π from time


series regression analysis of the import demandfunction. If π' is not significantly different from π,y and yB will not differ significantly either. Whenthis test is done, the model is supported in the vastmajority of cases. The cases where it is not aretypically countries that have run either largebalance of payments surpluses for a long period oftime, or large deficits financed by capital inflows.This leads on to the extension of the basic modelto include the capital account of the balance ofpayments.

The Model with Capital Flows

With capital flows, equation (5.1) becomes:

PdX + C = Pf ME (5.8)

where C > 0 is capital inflows measured indomestic currency. This is an identity because thebalance of payments must balance in total. Takingrates of change of (5.8), and substituting (5.3) and(5.5), gives the rate of growth of income consistentwith the total balance of payments:

yBT = [(pd – pf – e) + (θη + ψ) (pd – pf – e) + θ ε z+ (1 – θ) (c – pd)]/π,

(5.9)

where c is the growth of nominal capital inflows;θ is the share of exports in total receipts to pay forimports, and (1 – θ) is the share of capital inflows


in total receipts. The first term in equation (5.9)gives the pure terms of trade effect on real incomegrowth. The second term gives the volume effectof relative price changes. The third term gives theeffect of exogenous changes in income growthabroad, and the fourth term gives the effect of thegrowth of real capital inflows which ‘finance’growth in excess of the rate of growth consistentwith equilibrium on current account.

Since equation (5.9) is derived from an identity,it is possible to disaggregate any country’s growthrate into the above four components, and tocompare countries or groups of countries. One ofthe latest studies to do this is by Nureldin-Hussain(1999). He takes a sample of 29 African countrieswhich grew on average at 3.66 per cent per annumand 11 Asian countries which grew on average at6.60 per cent, and analyses differences betweenthem in terms of equation (5.9). There is not muchdifference between the two sets of countries as faras terms of trade movements are concerned. Theeffect of capital inflows on growth is slightly lowerin Asia than in Africa. The big difference comesthrough the growth in the volume of exportswhich in Asia gives a growth rate of 5.91 per centwhile in Africa it produces a growth rate of only2.45 per cent – less than half. This highlights onceagain the importance of differences in the structureof production and income elasticities of demandfor exports in contributing to differences in growthperformance between countries. Africa is still


dominated by the export of primary products,while Asia has diversified into manufactures.

Policy Implications

The simple policy implication for most countries isthat, if they wish to grow faster, they must first raisethe balance of payments constraint on demand.The challenge for economic policy-making is howto do this effectively. The IMF prescription isnormally liberalization and currency depreciation.While trade liberalization may improve export per-formance, it may also lead to a faster growth ofimports which worsens the balance of payments(see Santos-Paulino and Thirlwall, 2001). Amonginternational organizations, only UNCTAD (1999)seems to recognize this possibility. Liberalizationof the capital account of the balance of payments isalso fraught with problems without internal macro-economic stability. Domestic interest rates whichare too high will lead to capital inflows andovervalued currencies which damage the tradeablegoods sector. Equally, domestic crisis may lead torapid capital outflows, depreciating the currencyexcessively, leading to inflation.

As far as devaluation is concerned, we haveshown that currency depreciation cannot raise acountry’s growth rate on a permanent basis unlessit is continuous, or it changes favourably otherparameters of the model. The exchange rate,however, is not an efficient instrument forstructural change because it simply makes


countries more competitive (temporarily) in thegoods that cause the balance of paymentsproblems in the first place. Countries can try andmake their goods more price-competitive by othermeans, but many of the goods developingcountries produce (at least collectively) are priceinelastic (for example, primary commodities). Itis the non-price characteristics of goods such astheir quality, technical sophistication andmarketing which seem to be the most importantfactor in determining trade performance.

Countries can impose import controls to reducethe income elasticity of demand for imports (π) butthis can breed serious inefficiency. It is true,however, and worth remembering in debates overprotection, that no country in the world, apart fromthe United Kingdom, has ever industrializedwithout protection of one form or another. Exportpromotion and import substitution are not incom-patible strategies, as Japan and South Korea havedemonstrated in the post-war years. The distin-guished development economist Ajit Singh tellshow, when he first went to Cambridge to studyeconomics, Nicholas Kaldor taught him threethings: first, the only way for a country to developis to industrialize; second, the only way for acountry to industrialize is to protect itself; andthird, anyone who says otherwise is beingdishonest! The developed economies do preachdouble standards. They preach free trade fordeveloping countries, yet protect their ownmarkets. There is an economic case for protection


to alter the structure of production and to improvethe balance of payments, but it needs to be imple-mented with prudence and skill to avoid theprotection of high-cost inefficient industries andthe pursuit of rent seeking.

Countries can encourage greater capital inflowsto finance import growth in excess of exportgrowth, but care needs to be taken with the type ofcapital inflow. Long-term direct investment isprobably the most stable and beneficial, but foreigninvestment can also cause problems relating to thenature of the goods produced, the techniques ofproduction employed and the outflow of profits.Most other types of inflow, apart from pure aid,involve debt-service repayments, and debtproblems can arise if the inflows are not translatedinto improved export performance which earn theforeign exchange to pay interest and amortization.Even if the borrowing is invested in the tradeablegoods sector, foreign exchange is not guaranteedbecause the growth of exports is outside the controlof the countries concerned. The export growth ofdeveloping countries depends largely on the healthof the world economy, which became dramaticallyapparent during the debt crisis of the early 1980s.

The only sure and long-term solution to raisinga country’s growth rate consistent with balance ofpayments equilibrium on current account isstructural change to raise ε and to reduce π. We areback to the ideas of Raul Prebisch and the questionof the most appropriate industrial policy forcountries, and the role of protection.


6. The endogeneity of thenatural rate of growth

It was Harrod who first formally introduced theconcept of the natural rate of growth into economictheory in his paper, ‘An Essay in Dynamic Theory’(Harrod, 1939), which was discussed in Chapter 1.The natural rate of growth refers to the rate ofgrowth of productive potential of an economy, orthe ‘social optimum’ rate of growth, as Harrodcalled it. In all of mainstream growth theory, thenatural rate of growth (composed of labour forcegrowth and labour productivity growth) is treatedas exogenously determined, unresponsive to theactual rate of growth or the pressure of demand inan economy. It is exogenous in Harrod’s originalmodel, which is why Harrod’s growth model is notreally a growth model at all, but a trade cyclemodel, because it does not explain growth. It istreated as exogenous in the neoclassical responseto Harrod, as in the original model of Solow (1956),for example, as discussed in Chapter 2. It is treatedas exogenous (by and large) in the post-Keynesianresponse to the neoclassicals, as in the originalmodels of Kaldor (1957) and Joan Robinson (1956).Paradoxically, it is even treated as exogenous in

79

‘new’ endogenous growth theory. In ‘new’ growththeory, growth is endogenous in the sense thatinvestment matters for growth, because theassumption of diminishing returns to capital isrelaxed, not in the sense that labour force growthand productivity growth respond to demand andthe growth of output itself. Demand is entirelymissing from ‘new’ endogenous growth theory.

Then, when it comes to empirical studies ofgrowth rate differences between countries, we findthat exogeneity of factor supplies and productiv-ity growth permeates the whole of the mainstreamliterature on the sources of growth, as in thepioneer studies of Abramovitz (1956), Solow(1957), Denison (1967) and Maddison (1970)among others, and the recent work of AlwynYoung (1995) on South East Asia and Hu and Khan(1997) on China, as discussed in Chapter 2.

But the question arises, suppose the natural rateof growth, or a country’s growth of productivepotential, is not exogenous, but endogenous todemand or the actual rate of growth? What impli-cations does this have? It has two majorimplications. First, at the theoretical level, it hasimplications for the efficiency and speed of theadjustment process between the warranted andnatural rates of growth in the Harrod model.Secondly, and more important, it has implicationsfor the way we view the growth process, and whygrowth rates differ between countries: whether weview growth as supply-determined, or whether weview growth as demand-determined or


determined by constraints on demand beforesupply constraints become operative. The viewtaken here is that it is a mistake to think of thenatural rate of growth as exogenously determined.In other words, there is nothing natural about thenatural rate of growth, just as there is nothingnatural about the natural rate of unemployment(but that is another story). Both the growth of thelabour force and productivity growth arepositively related to demand or the actual rate ofgrowth.

The view that growth is primarily demand-driven, to which supply responds, does not mean,of course, that demand growth determines supplygrowth without limit; rather, that aggregatedemand determines aggregate supply over a rangeof full employment growth rates, and that in mostcountries demand constraints tend to bite longbefore supply constraints are ever reached.

Later, I will suggest a simple technique fortesting the endogeneity of the natural rate ofgrowth and give some empirical results for asample of 15 OECD countries over the period 1961to 1995. First, however, let us discuss the theoreti-cal consequences of the natural rate beingendogenous.

Although it was Harrod in 1939 who first coinedthe term ‘the natural rate of growth’, as a matter ofhistorical interest, Keynes had effectively antici-pated Harrod’s idea two years earlier in his GaltonLecture to the Eugenics Society in 1937 on ‘SomeEconomic Consequences of a Declining

The endogeneity of the natural rate of growth 81

Population’ (Keynes, 1937), where he expressed theworry that, because of a falling population, therewould not be enough demand to absorb fullemployment saving. Consider, he says, aneconomy with a savings ratio of 8–15 per cent ofnational income, and a capital–output ratio of 4,giving a rate of capital accumulation which willabsorb saving of approximately 2–4 per cent. Witha constant capital–output ratio, this is the requiredgrowth of output, but can this rate be guaranteed?Historically, it appeared to Keynes that one-half ofthe increase in capital accumulation (or demandfor investment) could be attributed to increasedpopulation; the other half to increased livingstandards (productivity growth). Now supposepopulation growth falls to zero. Since the standardof life cannot be expected to grow by more than 1per cent per annum, this means that the demandfor capital will only grow at 1 per cent while thesupply grows at between 2 and 4 per cent – a clearand worrying imbalance which would have to berectified either by reducing saving or by reducingthe rate of interest to lengthen the average periodof production (that is, to raise the capital–outputratio). This discussion is exactly analogous toHarrod’s discussion of divergence between thewarranted and natural rates of growth. Therequired rate of growth to absorb saving is thewarranted rate of growth, and the long-run growthrate determined by population (labour force)growth and rising living standards (productivitygrowth through technical progress) is the natural


rate of growth. Harrod’s dynamic theory isprecisely anticipated by Keynes; and Keynes, likeHarrod, treats the natural growth rate asexogenous.

Given the definition of the natural rate of growthas the sum of the rate of growth of the labour forceand the rate of growth of labour productivity, itfollows that the measured natural rate must be thatrate of growth that keeps the unemployment rateconstant. Otherwise, if the actual growth rate isabove the natural rate, the unemployment rate willfall; and if the actual growth rate is below thenatural rate, the unemployment rate will rise. Forthe purposes here, I define and measure the naturalgrowth rate of countries as the rate which keepsthe rate of unemployment constant.

The natural growth rate fulfils two functions inthe Harrod model. Firstly, it sets the ceiling to thedivergence between the actual and warrantedgrowth rates and turns cyclical booms into slumps.Secondly, as implied earlier, it gives the long-runpotential growth rate to which economies mightgravitate given the right conditions. But there wasno mechanism in the original Harrod model tobring the warranted and natural rates of growth inline with one another, with the consequence thateconomies might experience perpetual secularstagnation (if the warranted rate exceeds thenatural rate) or permanent inflation and structuralunemployment (if the natural rate exceeds thewarranted rate, as in most developing countrieswhere population growth is high and savings low).


Mechanisms that might achieve equilibrium,however, were soon invented. The Cambridge,Massachusetts school, represented by RobertSolow, Paul Samuelson and Franco Modigliani,used the neoclassical production function andvariations in the capital–output ratio to show thatthe warranted growth rate would adjust to thenatural rate (assuming, of course, appropriatefactor price adjustment and a spectrum ofproduction techniques to choose from). TheCambridge, England school, represented byNicholas Kaldor, Joan Robinson, Richard Kahn andLuigi Pasinetti, used variations in the savings ratiobrought about by changes in the functional distri-bution of income between wages and profits as themechanism to bring about equilibrium. But bothschools have equilibrium growth proceeding at theexogenously given natural rate.

What happens, however, if the natural rate ofgrowth is not exogenous? This has interesting con-sequences both for the short-run trade cycle modelof Harrod and for the long-run equilibrium growthmodel. Recall that, in the trade cycle model (seeChapter 1), if the actual growth rate diverges fromthe warranted growth rate in either direction,forces come into play which widen the divergence– but divergence is bounded by ceilings and floors.The ceiling is the natural rate of growth becausethe level of output cannot exceed the fullemployment ceiling. But suppose the natural rateincreases with the actual rate of growth (becauselabour force growth and productivity growth are


induced); this will perpetuate the cyclical upturn.One interesting conjecture is that this increases thepossibility that the cyclical upturn is not broughtto an end by an absolute ceiling, but by demandconstraints associated with inflation and balanceof payments problems due to bottlenecks in thesystem. This may explain why cyclical peaks areoften accompanied by excess capacity. In any case,the endogeneity of the natural rate will surelylengthen the cycle.

In the long-period model of divergence betweenthe warranted and natural growth rate, the endo-geneity of the natural rate will impede adjustmentto equilibrium. If the warranted rate exceeds thenatural rate, it means that the growth of capitalexceeds the growth of the labour force in efficiencyunits and the warranted rate must fall for equilib-rium. In conditions of recession, however, thenatural rate is also likely to fall as workers leavethe labour force and productivity growth slows,impeding adjustment. Similarly, if the natural rateexceeds the warranted rate, this implies that thegrowth of the effective labour force exceeds thegrowth of capital and the warranted rate must risefor equilibrium. In booms, however, the naturalrate is also likely to rise as workers are attractedinto the labour force and productivity growthaccelerates, also impeding adjustment.

In general, the endogeneity of the natural rate ofgrowth has serious implications for the notion of agiven full employment production frontier whicheconomies will gravitate towards. In practice, the


frontier will continually shift with the actualgrowth rate.

In What Ways is the Natural Rate Endogenous?

There are many mechanisms through which thenatural rate of growth is likely to be endogenousto the actual rate of growth. Consider first thegrowth of the labour force or labour supply.Labour supply is extremely elastic to demand.When the demand for labour is strong, labourinput responds in a number of ways. Firstly, par-ticipation rates rise. Workers previously out of thelabour force decide to join the labour force. Theparticipation rates of the young, the old, andmarried women are particularly flexible. Secondly,hours worked increase. Part-time workers becomefull-time workers, and overtime work increases.Thirdly, and significantly for many countries acrossthe world, labour migration takes place in responseto booming labour markets. If countries are shortof labour, they import it. Cornwall (1977) andKindleberger (1967) document the important rolethat immigrant labour played in Europe during the‘golden age’ of economic growth between 1950 and1973. The migration of labour from Portugal,Spain, Greece and Turkey into Germany, France,Switzerland and northern Italy was not anexogenous movement but was fuelled by an excessdemand for labour in the receiving countriesbecause the growth of demand for output was so


high (largely due to rapid export growth). Similarstories could be told for other parts of the world.

Now consider the growth of labour productiv-ity. There are several mechanisms through whichlabour productivity growth is endogenous todemand, and well documented. First, there arestatic and dynamic returns to scale associated withincreases in the volume of output and the technicalprogress incorporated in capital accumulation.Some technical progress is autonomous, but a greatdeal is demand-driven, particularly processinnovation. Necessity is the mother of invention!Secondly, there are macro increasing returns in theAllyn Young (1928) sense associated with the inter-related expansion of all activities. If the market fora good expands, it makes it profitable to use moresophisticated machinery, which cuts costs. This notonly reduces the price of the good (leading tofurther expansion of demand) but will also reducethe price of machinery if there are economies ofscale in its production which makes it profitable touse machinery in other activities. The initialdemand expansion leads to a series of changeswhich propagate themselves in a cumulative way,causing labour productivity to rise. Thirdly, thereis the well known phenomenon of learning bydoing whereby the efficiency or productivity oflabour is an increasing function of a learningprocess related to cumulative output. The moreoutput produced, the more adept labour becomesat producing it. Clearly, the impact of learning willgradually diminish with successive amounts of the


same output, but as long as product ranges changeover time, the effect of learning on productivitygrowth will be a continuous process related to theexpansion of output. All the phenomenamentioned above are captured by the Verdoornrelation, or Verdoorn’s Law, discussed in Chapter3. Given this relationship between the growth ofoutput and induced productivity growth, it is noaccident that, when growth slows down, produc-tivity growth also slows down. The productivitygrowth slowdown after the shocks to the worldeconomy in the 1970s was regarded as a puzzle bysome economists, but can be readily understood inthe context of models in which productivitygrowth is endogenous.

Estimating the Natural Rate and Testing itsEndogeneity

Let us now turn to the question of how the naturalrate of growth of a country may be estimated, andto test whether it is endogenous. Many years ago(Thirlwall, 1969), I suggested a simple techniquefor estimating the natural rate of growth based ona modification of the equations used for testingOkun’s Law (Okun, 1962) relating to the relationbetween changes in unemployment and the gapbetween actual and potential output. We sawearlier that, by definition, the natural rate must bethe growth rate which keeps the rate of unem-ployment constant. If we therefore relate changesin unemployment in a country to its growth rate,


we can solve for the growth of output that keepsunemployment constant. In other words, let

Δ% U = a – b (g), (6.1)

where % U is the percentage rate of unemploymentand g is the growth rate. Solving for g when Δ% U = 0gives an expression for the natural rate of growth ofgn = a/b. The technique is simple, but there arecertain problems. The estimate of the coefficient ‘b’may be biased downwards because of labourhoarding which would exaggerate the estimate ofgn. Equally, however, the constant term ‘a’ may bebiased downwards through workers leaving thelabour force where g is low, depressing the estimateof gn. It is difficult to know a priori what the relativestrengths of the (offsetting) biases are likely to be.

An alternative procedure is to reverse thevariables in equation (6.1) to give:

g = a1 – b1 (Δ% U). (6.2)

Solving for g when Δ% U = 0 now gives an estimatefor the natural rate of growth of gn = a1. This alsohas statistical problems since the change in unem-ployment is an endogenous variable, although ittranspires empirically that this does not affect theresults obtained from fitting (6.2). Originally, Itested both ways for the United States and theUnited Kingdom using (6.1) and (6.2) and obtainedthe same results for the period 1950 to 1967: anatural rate for the UK of 2.9 per cent and for the


US of 3.3 per cent, which seemed eminentlyreasonable estimates.

To my knowledge, the technique has not beenused subsequently, but if this simple way ofestimating the natural rate of growth is accepted,the obvious way to test for endogeneity is toinclude a dummy variable into (say) equation (6.2)in periods when the actual growth rate is above theestimated natural rate and test for its significance:

g = a2 + b2D – c2 (Δ% U), (6.3)

where D takes the value of 1 when actual growthis greater than the natural rate and zero otherwise.If the dummy is significant, this must mean thatthe rate of growth in periods of boom to keepunemployment constant has risen. The actualgrowth rate must have been pulling more workersinto the labour force and inducing productivitygrowth. The constant term ‘a2’ plus b2 gives thenatural rate of growth in boom periods. The inter-esting question is then how this estimate of thenatural rate in boom periods compares with theestimate of the natural rate which does not distin-guish between boom and slump. What is theelasticity of the natural rate in periods of boom?

The procedures described above can be illus-trated by means of a simple diagram (Figure 6.1).The growth of output is measured on the verticalaxis and the change in the percentage level ofunemployment on the horizontal axis. The scatterpoints relate to the time series relation between


the two variables. Since the natural rate of growthis defined as that rate which keeps unemploy-ment constant, a1 is the estimated natural rateover the whole sample period not distinguishingbetween periods of boom and slump. If we thendistinguish between periods when g > gn and g< gn, the question is whether the intercept differs,or do they share the common intercept a1? Notethat in practice not all observations will lie wherethey should theoretically in the top left andbottom right quadrants of the diagram (with Δ%U > 0 when g < gn and Δ% U < 0 when


–

g

a2 + b

2

Δ%u

a1

a2

gn

+0

Figure 6.1

g > gn) because the relation between g and Δ% Uis stochastic. Some observations may lie in the topright and bottom left quadrants, which could biasestimates of the intercepts between the tworegimes. This is tested for, and it is found that‘rogue’ observations make no statistical differenceto the results to be presented below.

Empirical Results

To test the model, a sample of 15 OECD countriesis taken over the period 1961 to 1995. Bothequations (6.1) and (6.2) were fitted to estimate thenatural rate of growth over the whole period. Ingeneral, equation (6.2) gave the best results interms of goodness of fit of the equations and thereasonableness of the results. In equation (6.2), theestimate of the natural rate of growth is given bythe constant term (a1), and this is reported for allcountries in the first column of Table 6.1. Theconstant term was estimated as statistically signif-icant in all 15 countries. The estimates of thenatural rate all look reasonable for the countriesconcerned, and range from 2.5 per cent in the UK(the lowest) to 4.6 per cent in Japan (the highest).The average natural growth rate for the 15 OECDcountries as a whole is 3.5 per cent.

When a dummy variable was added to equation(6.2) for years when the actual growth rateexceeded the estimated natural rate (equation 6.3),it was found to be significant in all 15 countries.


The sum of the dummy plus the new constant (a2)gives the natural rate in boom periods, and isshown in column 2 of Table 6.1. The natural rateis seen to increase considerably in all countries, butin some countries by more than others. Taking thecountries as a whole, the average increase is 1.8percentage points, which is to say that the actualrate of growth in boom periods has induced labourforce growth and productivity growth by thatamount. The countries where the sensitivity of thenatural rate seems to be greatest are those wherethe reserves of labour are known to be highest,such as Greece and Italy (due to surplus labour inthe south), and where output growth has inducedimpressive technical progress through learningand sectoral rationalization, such as Japan. Ingeneral, the results show substantial elasticity ofthe labour force and productivity growth withrespect to the pressure of demand in the economy,and it is important to stress that these results arenot measuring simply the cyclical effect of demandon output growth because this is captured by thecoefficient, c2, in equation (6.3). The results arecapturing the longer-lasting effects that sustaineddemand expansion has had on the growth ofproductive potential over the cycle.

Conclusion

The conclusion to this chapter can provide a fittingconclusion to the book as a whole. In mainstreamgrowth theory, including ‘new’ growth theory, the


natural rate of growth is treated as exogenouslydetermined, unrelated to demand or the actual rateof growth. If supply or output potential respondsto demand, however, this raises the crucialquestion, what does it mean to say that outputgrowth is supply-determined, or constrained by


Table 6.1 Sensitivity of the natural rate of growth tothe actual rate of growth

Country Increase in natural rate

Natural rate in boom in boom periods

Natural rate periods Absolute(%) (%) difference Increase (1) (2) (2)–(1) (%)

Australia 3.9985 5.7131 1.7146 42.9Austria 3.1358 4.9563 1.8205 58.1Belgium 3.5239 4.9102 1.3863 39.3Canada 3.8352 5.2613 1.4261 37.2Denmark 2.9424 4.7826 1.8402 62.5France 2.8270 3.9343 1.1073 39.2Germany 3.5054 4.7091 1.2037 34.3Greece 4.5089 7.6711 3.1622 70.1Italy 3.3439 5.9104 2.5665 76.8Japan 4.5671 8.7199 4.1528 90.9Netherlands 3.2817 5.3151 2.0334 62.0Norway 3.9722 5.0094 1.0372 26.1Spain 4.0623 6.0928 2.0305 50.0UK 2.5438 3.8022 1.2584 49.5USA 2.9911 3.6642 0.6731 22.5

Average 3.5359 5.3634 1.8275 51.7

supply? Of course, it is true in a trivial sense thatcapital and labour are required to produce output,and how much output is produced will alsodepend on the level of technical efficiency, but thereally important question is, why does the growthof capital, and technical progress, differ so muchbetween countries? The supply-oriented, neoclas-sical production function approach to the analysisof growth cannot answer this question, and for themost part never asks it!

What has been shown in this last chapter is thatit is a mistake to regard the natural rate of growthas exogenously given. The rate of growth necessaryto keep the percentage level of unemploymentconstant rises in boom periods and falls inrecession because the labour force and productiv-ity growth are elastic to demand and outputgrowth. This is also confirmed using causality testsbetween input and output growth (not reportedhere, but see Leon-Ledesma and Thirlwall, 2002).The orthodox and ‘new’ growth theories thatassume that it is input growth that unidirectionallycauses output growth finds no support from theevidence. The implication for growth theory andpolicy is that it makes little economic sense to thinkof growth as supply-constrained if demand, withinlimits, creates its own supply. If factor inputs(including productivity growth) react endoge-nously, the process of growth, and growth ratedifference between countries, can only be properlyunderstood in terms of differences in the strengthof demand, and constraints on demand. For most


countries, and particularly developing countries,demand constraints operate long before capacityis reached. Demand constraints are likely to berelated to supply bottlenecks which cause inflationand balance of payments difficulties for countries.It is this aspect of supply, and not the growth ofinputs in a production function, that should be themain focus of enquiry in any supply-orientedtheory of economic growth.


Bibliography

M. Abramovitz (1956), Resource and OutputTrends in the United States since 1870, AmericanEconomic Review Papers and Proceedings, May.

K. Arrow (1962), The Economic Implications ofLearning by Doing, Review of Economic Studies,June.

R. Barro (1991), Economic Growth in a CrossSection of Countries, Quarterly Journal ofEconomics, May.

W. Baumol (1986), Productivity Growth, Conver-gence and Welfare, American Economic Review,December.

J. Cornwall (1977), Modern Capitalism: Its Growthand Transformation (London: Martin Robertson).

E. Denison (1962), The Sources of Economic Growthin the US and the Alternatives before Us (New York:Committee for Economic Development, Libraryof Congress).

E. Denison (1967), Why Growth Rates Differ: Post-war Experience in Nine Western Countries(Washington, DC: Brookings Institution).

E. Domar (1947), Expansion and Employment,American Economic Review, March.

A. Emmanuel (1972), Unequal Exchange: A Study ofthe Imperialism of Trade (New York: MonthlyReview Press).

97

H. Esfahani (1991), Exports, Imports and EconomicGrowth in Semi-Industrialised Countries, Journalof Development Economics, January.

G. Feder (1983), On Exports and EconomicGrowth, Journal of Development Economics,February/April.

J. Felipe (1999), Total Factor Productivity Growthin East Asia: A Critical Survey, Journal of Devel-opment Studies, April.

B. Fingleton and J. McCombie (1998), IncreasingReturns and Economic Growth: Some Evidencefrom the European Union Regions, OxfordEconomic Papers, January

G. Frank (1967), Capitalism and Underdevelopment inLatin America (New York: Monthly ReviewPress).

Z. Griliches (1958), Research Costs and SocialReturns: Hybrid Corn and Related Innovations,Journal of Political Economy, October.

G. Grossman and E. Helpman (1991), Innovationand Growth in the Global Economy (Cambridge,Mass.: MIT Press).

J.D. Hansen and J. Zhang (1996), A KaldorianApproach to Regional Economic Growth inChina, Applied Economics, June.

R. Harrod (1933), International Economics(Cambridge: Cambridge University Press).

R. Harrod (1939), An Essay in Dynamic Theory,Economic Journal, March.

J. Hicks (1950), The Trade Cycle (Oxford: ClarendonPress).


A. Hirschman (1958), Strategy of Economic Develop-ment (New Haven: Yale University Press).

Z.F. Hu and M.S. Khan (1997), Why is ChinaGrowing so Fast?, IMF Staff Papers, March.

K. Hussein and A.P. Thirlwall (2000), The AKModel of ‘New’ Growth Theory is theHarrod–Domar Growth Equation: Investmentand Growth Revisited, Journal of Post KeynesianEconomics, Spring.

Journal of Post Keynesian Economics (1997),Symposium on Thirlwall’s Law, Spring.

N. Kaldor (1957), A Model of Economic Growth,Economic Journal, December.

N. Kaldor (1966), Causes of the Slow Rate of EconomicGrowth of the United Kingdom (Cambridge:Cambridge University Press).

N. Kaldor (1967), Strategic Factors in EconomicDevelopment (New York, Ithaca: New York StateSchool of Industrial and Labour Relations,Cornell University).

N. Kaldor (1970), The Case for Regional Policies,Scottish Journal of Political Economy, November.

N. Kaldor (1972), ‘Advanced Technology in aStrategy for Development: Some Lessons fromBritain’s Experience’, in Automation andDeveloping Countries (Geneva: ILO).

N. Kaldor (1985), Economics without Equilibrium(Cardiff: University College Cardiff Press).

N. Kaldor (1996), Causes of Growth and Stagnationin the World Economy (The Raffaele MattioliLectures), (Cambridge: Cambridge UniversityPress).

Bibliography 99

J.M. Keynes (1937), Some Economic Consequencesof a Declining Population, Eugenics Review,April.

C. Kindleberger (1967), Europe’s Postwar Growth: theRole of the Labour Supply (Cambridge, Mass.:Harvard University Press).

P. Krugman (1989), Differences in Income Elastic-ities and Trends in Real Exchange Rates,European Economic Review, May.

H. Leibenstein (1957), Economic Backwardness andEconomic Growth (New York: Wiley).

M. Leon-Ledesma (2000a), Economic Growth andVerdoorn’s Law in the Spanish Regions 1962–91,International Review of Applied Economics, January.

M. Leon-Ledesma (2000b), Accumulation,Innovation and Catching-Up: an ExtendedCumulative Growth Model, Cambridge Journal ofEconomics , May.

M. Leon-Ledesma and A.P. Thirlwall (2002), TheEndogeneity of the Natural Rate of Growth,Cambridge Journal of Economics, forthcoming.

R. Levine and D. Renelt (1992), A SensitivityAnalysis of Cross-Country Growth Regressions,American Economic Review, September.

A. Lewis (1954), Economic Development withUnlimited Supplies of Labour, Manchester School,May.

J. Lopez and A. Cruz (2000), ‘Thirlwall’s Law’ andBeyond: the Latin American Experience, Journalof Post Keynesian Economics, Spring

R. Lucas (1988), On the Mechanics of EconomicDevelopment, Journal of Monetary Economics,vol. 22.


J. McCombie (1985a), Increasing Returns and theManufacturing Industries: Some EmpiricalIssues, Manchester School, March.

J. McCombie (1985b), Economic Growth, theHarrod Foreign Trade Multiplier and the HicksSuper Multiplier, Applied Economics, February.

J. McCombie and J.R. de Ridder (1983), IncreasingReturns, Productivity and Output Growth: TheCase of the United States, Journal of PostKeynesian Economics, Spring.

J. McCombie and A.P. Thirlwall (1994), EconomicGrowth and the Balance of Payments Constraint(London: Macmillan).

J. McCombie and A.P. Thirlwall (1997), TheDynamic Harrod Foreign Trade Multiplier andthe Demand Oriented Approach to EconomicGrowth: An Evaluation, International Review ofApplied Economics, January.

A. Maddison (1970), Economic Progress and Policyin Developing Countries (London: Allen &Unwin).

T. Malthus (1798), Essay on the Principle of Population(London: Penguin, 1983).

K. Marx (1867), Capital: A Critique of PoliticalEconomy, Vol. 1 (reprinted New York: Interna-tional Publishers, 1967).

T.R. Michl (1985), International Comparisons ofProductivity Growth: Verdoorn’s Law Revisited,Journal of Post Keynesian Economics, Summer

J.C. Moreno-Brid (1998), Balance of Payments Con-strained Economic Growth: The Case of Mexico,

Bibliography 101

Banca Nazionale del Lavoro Quarterly Review,December.

J.C. Moreno-Brid (1999), Mexico’s EconomicGrowth and the Balance of Payments Constraint:A Cointegration Analysis, International Review ofApplied Economics vol. 13, no. 2.

G. Myrdal (1957), Economic Theory and Underdevel-oped Regions (London: Duckworth).

R. Nelson (1956), A Theory of the Low Level Equi-librium Trap in Underdeveloped Countries,American Economic Review, December.

M. Nureldin-Hussain (1999), The Balance ofPayments Constraint and Growth Rate Differ-ences among African and East Asian Economies,African Development Review, June.

A. Okun (1962), Potential GNP: Its Measurementand Significance, Proceedings of the Business andFinance Statistics Section of the American Statisti-cal Association.

R. Prebisch (1950), The Economic Development ofLatin America and its Principal Problems (NewYork: ECLA, UN Dept. of Economic Affairs).

D. Ricardo (1817), Principles of Political Economy andTaxation (reprinted London: Everyman, 1992).

J. Robinson (1956), The Accumulation of Capital(London: Macmillan).

P. Romer (1986), Increasing Returns and Long RunGrowth, Journal of Political Economy, October.

R. Sandilands (1990), Nicholas Kaldor’s Notes onAllyn Young’s LSE Lectures, 1927–29, Journal ofEconomic Studies, vol. 17, no. 3/4.


A. Santos-Paulino and A.P. Thirlwall (2001), TheImpact of Trade Liberalisation on ExportGrowth, Import Growth, the Balance of Tradeand the Balance of Payments of DevelopingCountries, University of Kent, mimeo.

T.W. Schultz (1961), Investment in Human Capital,American Economic Review, March.

D. Seers (1962), A Model of Comparative Rates ofGrowth of the World Economy, Economic Journal,March.

A. Smith (1776), An Inquiry into the Nature andCauses of the Wealth of Nations (London: Strahamand Caddell).

R. Solow (1956), A Contribution to the Theory ofEconomic Growth, Quarterly Journal of Economics,February.

R. Solow (1957), Technical Change and theAggregate Production Function, Review ofEconomics and Statistics, August.

R. Summers and A. Heston (1991), The Penn WorldTable (Mark 5): An Expanded Set of Interna-tional Comparisons 1950–1966, Quarterly Journalof Economics, May.

F. Targetti and A.P. Thirlwall (1989), The EssentialKaldor (London: Duckworth).

A.P. Thirlwall (1969), Okun’s Law and the NaturalRate of Growth, Southern Economic Journal, July.

A.P. Thirlwall (1979), The Balance of PaymentsConstraint as an Explanation of InternationalGrowth Rate Differences, Banca Nazionale delLavoro Quarterly Review, March.

Bibliography 103

A.P. Thirlwall (1983), Foreign Trade Elasticities inCentre–Periphery Models of Growth and Devel-opment, Banca Nazionale del Lavoro QuarterlyReview, September.

A.P. Thirlwall (1986), A General Model of Growthand Development on Kaldorian Lines, OxfordEconomic Papers, July.

A.P. Thirlwall (1987a), Nicholas Kaldor (Brighton:Harvester Wheatsheaf).

A.P. Thirlwall (1987b), Keynes and Economic Devel-opment (London: Macmillan).

A.P. Thirlwall (1999), Growth and Development: withSpecial Reference to Developing Economies, 6th edn(London: Macmillan).

A.P. Thirlwall (2000), Trade Agreements, Trade Lib-eralisation and Economic Growth: A SelectiveSurvey, African Development Review, December.

A.P. Thirlwall and G. Sanna (1996), ‘“New”Growth Theory and the Macrodeterminants ofGrowth: An Evaluation and Further Evidence’,in P. Arestis (ed.), Employment, Economic Growthand the Tyranny of the Market: Essays in Honour ofPaul Davidson, Vol. 2 (Cheltenham, UK andBrookfield, US: Edward Elgar).

UNCTAD (1999), Trade and Development Report(Geneva: UNCTAD).

P.J. Verdoorn (1949), Fattori che Regolano loSviluppo della Produttivita del Lavoro,L’Industria, no. 1. English translation by A.P.Thirlwall in L. Pasinetti (ed.), Italian EconomicPapers Vol. II (Oxford: Oxford University Press,1993).


World Bank (1991), World Development Report(Washington: World Bank).

Allyn Young (1928), Increasing Returns andEconomic Progress, Economic Journal, December.

Alwyn Young (1995), The Tyranny of Numbers:Confronting the Statistical Realities of the EastAsian Growth Experience, Quarterly Journal ofEconomics, August.

Bibliography 105

Abramovitz, M. 25, 80Africa 9, 75aggregate demand, and

export growth 64–5agriculture 2, 3–4, 8, 10–11,

42, 46, 50, 51diminishing returns in 4

agriculture and industrymodel 50–51

AK model 34Arrow, Kenneth 33Asia 75–6

balance of payments 66–78balance of payments

constrained growthmodel 67, 68–74

with capital flows 74–6policy implications 76–8

Barro, Robert 31, 33Baumol, W. 29

capital, diminishing returns29

capital flows 66, 67, 74–6, 78capital-output ratio 18, 30, 84capitalism 10Carlyle, Thomas 7catch-up 33Causes of the Slow Rate of

Economic Growth of theUnited Kingdom 41

centre-periphery growthmodels 2, 32, 57, 61

China 43–4, 49, 80classical pessimists 7–12Cobb, Charles 21Cobb-Douglas production

function 21–2, 25labour-intensive form 26

competitiveness 51convergence 30, 31Corn Laws 9Cornwall, J. 86country growth rates,

divergence 58–9‘cube rule’ 45currency depreciation 70,

76–7current account equilibrium

68–9

Das Kapital 9debt-service repayments 78demand 24–5

and new endogenous growth theory 80

demand constraints 95–6demand driven growth 81demand oriented export-led

growth model 55–62equilibrium solution 57–8

demand-oriented approach52–65

Index

107

Denison, E. 26, 33, 80developing countries 15–16development policy 16diminishing returns 2

to capital 29direct investment 78divergence

in country growth rates 58–9

in the world economy 32division of labour 1, 3, 5Dixon, R.J. 68Domar, Evesey 11, 14Douglas, Paul 21

East Asia 27Economic Theory and Under-

developed Regions 7Economics without Equilib-

rium 7education 29, 33elasticities, and growth

60–61Emmanuel, A. 32Engle’s Law 61equilibrium, mechanisms to

achieve 84equilibrium growth model

84Esfahani, H. 64Essay on the Principle of

Population 7Europe

growth performance 27immigrant labour 86

export growth 51, 69, 78and aggregate demand

64–5and GDP growth 62–5

export promotion 77export-led growth model 48

exports 4, 28, 38, 39, 42,53–4, 55, 62

factors of production 25Feder, G. 62–3Felipe, J. 27Fingleton, B. 46foreign direct investment

(FDI) 3045-degree rule 60Frank, G. 32full employment production

frontier 85–6

GDP 6, 38, 66and manufacturing 40

GDP growth 44and export growth 62–5and manufacturing

output growth 41The General Theory 12‘green revolution’ 8Griliches, Z. 33Grossman, G. 29, 64growth 14–15, 40

demand driven 81Harrod’s concepts 12–14

growth economics 17growth rates, differences in

25–8, 37, 41

Hansen, J.D. 43, 46, 49Harrod growth equation 34Harrod instability problem

14Harrod, R. 11, 12, 60, 81, 83

‘An Essay in Dynamic Theory’ 79

International Economics 71Harrod trade multiplier 60,

71, 72


Harrod-Domar growthmodel 11, 12–19

Helpman, E. 29, 64Heston, A. 29Hicks, J. 54Hicks supermultiplier 54Hirschman, Albert 6, 32

Strategy of Economic Development 7

Hu, Z.F. 80Hussein, K. 35

IMF 76import controls 77import substitution 77imports 53–4, 69increasing returns 1–2, 5, 6,

45in industry 4

industrialization 77input growth 95inputs 27–8An Inquiry into the Nature

and Causes of the Wealthof Nations 1, 7

inter-country growth per-formance 73

International Economics 71investment 19, 20, 24, 30, 82

and growth 32

Japan 77Journal of Post Keynesian

Economics 73

Kahn, Richard 84Kaldor, Nicholas 6, 32, 35–6,

37, 41, 45, 46, 50, 61, 77,79, 84

Causes of the Slow Rate of

Economic Growth ofthe United Kingdom 41

Economics without Equilibrium 7

Strategic Factors in Economic Development41

Kaldor’s first law 41, 43–4Kaldor’s second law

(Verdoorn’s Law) 41,46–8

Kaldor’s third law 42, 48–9Keynes, J.M. 4, 83

The General Theory 12‘Proposals for an

InternationalClearing Union’ 67

‘Some Economic Conse-quencesofaDecliningPopulation’ 81–2

Keynesian growth school 17

Khan, M.S. 80Kindleberger, C. 86Krugman, Paul 60

labour 10, 19, 20migration 86shortage 50

labour force 81growth 86

labour productivity 3, 44growth 87

learning by doing 3, 87–8Leibenstein, H. 8Leon-Ledesma, M. 46, 65, 95Levine, R. 39Lewis, Arthur 9, 50living standards, conver-

gence 29, 30, 31, 32Lucas, Robert 29

Index 109

McCombie, J. 43, 46, 54, 73Maddison, A. 80Malthus, Thomas, Essay on

the Principle ofPopulation 7

manufacturing 40–52and GDP 40

manufacturing growth 42manufacturing output

growth, and GDPgrowth 41

market, extent of 3–4, 5Marshall, Alfred 31

Principles of Economics 11Marshall-Lerner condition

70Marx, Karl 2, 50

Das Kapital 9Michl, T.R. 46Mill, J.S. 2Modigliani, Franco 84Moreno-Brid, J.C. 66Myrdal, Gunnar 6, 32, 61

Economic Theory and UnderdevelopedRegions 7

natural rate of growth 79–96definition 83empirical results 92–3endogeneity 80–81, 84–5,

86–8testing 88–92

estimating 88–92as exogenous 79–80, 83,

94–5in the Harrod model 83in OECD countries 92–3,

94Nelson, R. 8neo-Marxist school 32

neoclassical growth model20–25

predictions 22–4neoclassical growth theory

12, 17neoclassical production

function 84neoclassical value theory 11new endogenous growth

theory 1, 28–39and demand 80

newly industrializingcountries (NICs) 40

non-equilibrium models ofdevelopment 6

Nureldin-Hussain, M. 75

OECD countries, naturalrate of growth 92–3, 94

Okun, A. 88Okun’s Law 43, 88output 21, 22

and unemployment 88–91

output per head 30

Pasinetti, Luigi 84policy implications, balance

of payments con-strained growth 76–8

population 7–8, 16, 82Prebisch, R. 32, 61, 78primary product

production 61Principles of Economics 11Principles of Political

Economy and Taxation 8production function, use for

analysis of growth dif-ferences 25–8


productivity growth 56, 82,88

profit 8–9, 9–10protection 77–8

regional export-led growthmodels 68

Renelt, D. 39research and development

(R&D) 29Ricardo, David 2

Principles of Political Economy and Taxation8

Ridder, J.R. de 46Robinson, Joan 15, 79, 84Romer, Paul 27

Samuelson, Paul 84Sandilands, R. 45Sanna, G. 39savings ratio 17–18Schultz, T.W. 33Seers, D. 32Singh, Ajit 77Smith, Adam 3–7, 32

An Inquiry into the Nature and Causes of theWealth of Nations 1, 7

‘social optimum rate ofgrowth’ 79

Solow, R. 25, 26, 79, 80, 84South Korea 77specialization 3Strategic Factors in Economic

Development 41Strategy of Economic

Development 7Summers, R. 29

supply 24

Targetti, F. 41technical progress 87terms of trade 70Thirlwall, A.P. 35, 38, 39, 41,

43, 45, 50, 57, 60, 62, 67,73, 88, 95

Thirlwall’s Law 73trade 37–8trade cycle model 79, 84trade liberalization 76

UNCTAD 76unemployment 15, 44–5, 81,

83, 95and output 88–91

USA, growth performance27

Verdoorn, P.J. 46Verdoorn’s Law (Kaldor’s

second law) 41, 43,46–8, 56–7, 88

wages 9–10, 11warranted rate of growth

84, 85Wealth of Nations 1, 7World Bank 27

Young, Allyn 5–6, 33, 87‘Increasing Returns and

Economic Progress’45

Young, Alwyn 27, 80

Zhang, J. 43, 46, 49

Index 111

thirlwall - the nature of economic growth

Economy & Finance

monthly review

gn exceeds

american economic

oxford economic

world war

warranted

cobbdouglas

natural rate