The Theory of Economic Growth - A 'Classical' Perspective PDF

The Theory of
Economic Growth: a
‘Classical’ Perspective
Edited by
Neri Salvadori
University of Pisa, Italy
Edward Elgar
Cheltenham, UK • Northampton, MA, USA
vi Contents
10. The evolutionary perspective on growth 205

Grazia D. Santangelo
11. Competition, rent seeking and growth: Smith versus the

Contents endogenous growth theory
Antonio D’Agata
222
12. R&D models of economic growth and the long-term

Introduction by Neri Salvadori xi
evolution of productivity and innovation 236
Mauro Caminati
1. Theories of economic growth: old and new 1
Heinz D. Kurz and Neri Salvadori
13. Competition and technical change in Aghion & Howitt:
a formalisation of Marx’s ideas? 260
2. The structure of growth models: a comparative survey 23
Maria Daniela Giammanco
Antonio D’Agata and Giuseppe Freni
14. Division of labour and economic growth: Paul Romer’s
3. Endogenous growth theory as a lakatosian case study 42
contribution in an historical perspective 272
Mario Pomini
Andrea Mario Lavezzi
4. Endogenous growth in a multi-sector economy 61
15. The interaction between growth and cycle in macrodynamic
Giuseppe Freni, Fausto Gozzi and Neri Salvadori
models of the economy 285
Serena Sordi
5. Income distribution and consumption patterns in a
‘classical’ growth model 82
16. Real business cycle models, endogenous growth models and
Davide Fiaschi and Rodolfo Signorino
cyclical growth: a critical survey 306
Davide Fiaschi and Serena Sordi
6. Keynesian theories of growth 104
Pasquale Commendatore, Salvatore D’Acunto, Carlo Panico
17. Growth theory and the environment: how to include matter
and Antonio Pinto
without making it really matter 330
Tommaso Luzzati
7. Should the theory of endogenous growth be based on
Say’s law and the full employment of resources? 139
18. Modelling growth and financial intermediation through
Fabio Petri
information frictions: a critical survey 342
Salvatore Capasso
8. The demographic transition and neo-classical models
of balanced growth 161
Piero Manfredi and Luciano Fanti
9. Human capital formation in the new growth theory:

the role of ‘social factors’ 186
Maria Rosaria Carillo
v
List of contributors Introduction
Mauro Caminati (University of Siena, Italy)
Neri Salvadori
Salvatore Capasso (University of Napoli “Federico II”, Italy)
Interest in the study of economic growth has experienced remarkable ups and
Maria Rosaria Carillo (University of Napoli “Parthenope”, Italy)
downs in the history of economics. It was central in Classical political
Pasquale Commendatore (University of Napoli “Federico II”, Italy) economy from Adam Smith to David Ricardo, and then in its ‘critique’ by
Karl Marx, but moved to the periphery during the so-called ‘marginal
Salvatore D'Acunto (University of Napoli “Federico II”, Italy)
revolution’. John von Neumann’s growth model and Roy Harrod’s attempt to
Antonio D’Agata (University of Catania, Italy) generalise Keynes’s principle of effective demand to the long run re-ignited
Luciano Fanti (University of Pisa, Italy) interest in growth theory. Following the publication of papers by Robert
Solow and Nicholas Kaldor in the mid 1950s, growth theory became one of
Davide Fiaschi (University of Pisa, Italy) the central topics of the economics profession until the early 1970s. After a
Giuseppe Freni (University of Napoli “Parthenope”, Italy) decade of dormancy, since the mid 1980s, economic growth has once again
become a central topic in economic theorising. The recent theory is called
Maria Daniela Giammanco (University of Catania, Italy) ‘endogenous growth theory’, since according to it the growth rate is
Fausto Gozzi (University of Rome “La Sapienza”, Italy) determined from within the model and is not given as an exogenous variable.
This book is the main product of a research group on the theory of growth
Heinz D. Kurz (University of Graz, Autriche) and the relation between modern growth theory and ‘Classical’ growth
Andrea Lavezzi (University of Pisa, Italy) theory. The scholars involved were motivated to this task not only by the
emergence at the end of the 1980s and the rapid development of the literature
Tommaso Luzzati (University of Pisa, Italy) on economic growth, but also by the contributions of Kurz and Salvadori
Piero Manfredi (University of Pisa, Italy) (1998b, 1999) who have shown that the logical structure underlying most of
the early models of endogenous growth is very similar to the logical structure
Carlo Panico (University of Napoli “Federico II”, Italy) of ‘Classical’ growth models. Put schematically, in the latter a given real
Fabio Petri (University of Siena, Italy) wage rate determines (together with the technological data) the rate of profits
and thus, through the saving-investment mechanism, the rate of growth; in
Antonio Pinto (University of Napoli “Federico II”, Italy) the modern literature, ‘human capital’ or ‘knowledge’ works in the same way
Mario Pomini (University of Verona, Italy) since there is a ‘technology’ producing them, exactly like the real wage rate
‘produced’ labour in the analyses of the Classical economists. The scholars
Neri Salvadori (University of Pisa) involved have also investigated the connection between the Classical
Grazia D. Santangelo (University of Catania, Italy) economists and the modern theories of growth in the analysis of competition,
technical change, economic cycles, and financial intermediation.
Rodolfo Signorino (University of Napoli “Federico II”) The readers may ask themselves whether classifying economic ideas in
Serena Sordi (University of Siena, Italy) distinct analytical approaches to certain economic problems and even in
different schools of economic thought is a futile enterprise. The title of this
ix xi
xii The Theory of Economic Growth: a ‘Classical’ Perspective Introduction xiii
book implies that its authors think that it is not. We rather hold the view that of growth. Pasquale Commendatore, Salvatore D’Acunto, Carlo Panico, and
there is a theory that may, for good reasons, be called ‘Classical’ economics Antonio Pinto have gone to great lengths to produce a comprehensive
as distinct from other kinds of economics, in particular ‘Neoclassical’ analysis of all the literature on the issue. Chapter 7 on Say’s law, by Fabio
economics and ‘Keynesian’ economics. This view could immediately be Petri, complements this analysis. As is argued in the first chapter of this
challenged with the indisputable heterogeneity and multi-layeredness of the book, the fact that the endowments of all resources, including capital and
writings of authors in these groups. Moreover, whilst regarding some aspects labour, are among the data of neoclassical theory imposes that this theory can
an author might be classified in one group, regarding some other aspects he consider growth only as exogenously directed. However, a sort of alternative
or she might be classified in another group. Therefore, I wish to make it clear exists; it consists in complementing neoclassical theory with a theory
from the outset that we are not so much concerned with elaborating a modelling the evolution of some endowments. Chapters 8 and 9 perform this
classification of authors, which in some cases would be an extremely task. Piero Manfredi and Luciano Fanti provide an analysis of the dynamics
difficult, if not impossible task. We are rather concerned with classifying of the working population within the Solovian model. Maria Rosaria Carillo
various analytical approaches to dealing with certain economic problems. studies the changes in the efficiency of work connected with social factors,
Our interest in these approaches is not dominantly historical; we rather as opposed to economic factors.
consider them as containing the key to a better explanation of important Thus Chapters 3–10 are mainly devoted to a ‘vertical’ or in-depth
economic phenomena. Our concern with classical economics is therefore analysis of four schools of thought, the ‘Classical’, the ‘Keynesian’, the
primarily a concern with its analytical potential which in our view has not yet ‘Neoclassical’ and the ‘Evolutionary’ School. By contrast, the remaining
been fully explored. chapters of the book are devoted to a ‘horizontal’ analysis of a number of
The book opens with a chapter by Kurz and Salvadori that summarises items connected with growth. Chapter 11, by Antonio D’Agata, explores the
their previous contributions and clarifies what we mean by ‘Classical’ and problem of legal barriers to entry and rent-seeking in Smith and in the
‘Neoclassical’ economics. Chapters 2 and 3 complete this methodological modern theory of growth. Chapters 12 and 13 investigate the problem of
analysis. Antonio D’Agata and Giuseppe Freni insert also ‘Keynesian’ technical change: Mauro Caminati proposes an ingenious method to classify
economics into the picture and find some other connections among these the modern literature whereas Maria Daniela Giammanco compares recent
schools of thought. Mario Pomini studies the emergence of endogenous results with some features that characterise the analysis of technical change
growth theory (as opposed to Neoclassical growth theory) from the point of proposed by Marx. Chapter 14, by Andrea Mario Lavezzi, compares the
view of Lakatosian categories. These chapters isolate and compare the modern contributions on the division of labour with the old literature, mainly
logical structures and the methodological underpinnings of old and new Adam Smith and Allyn Young. Chapters 15 and 16 analyse the connection
growth theories. They provide some well-defined guidelines that address the between growth and cycles: Serena Sordi surveys the macrodynamic models
analysis developed in the following chapters. whereas Davide Fiaschi and Serena Sordi survey the more recent literature
Chapters 4–9 analyse in greater detail the above-mentioned schools of on this topic. Tommaso Luzzati, in Chapter 17, is concerned with the
thought: Classical, Keynesian, Neoclassical. Chapter 10, by Santangelo, questions that the environment poses for growth theorists. Finally, Chapter
surveys the evolutionary point of view on growth and thus complements 18, by Salvatore Capasso, investigates the problems connected with the
Chapters 4–9. Chapter 4, by Giuseppe Freni, Fausto Gozzi, and Neri existence of financial intermediation.
Salvadori, can be read as an analysis of the problems that the extension to a
multi-sector economy poses for endogenous growth theorists, but it can also
be read both as a restatement of some solutions proposed by the theory of
production of ‘Classical’ orientation (see Kurz and Salvadori, 1995) and as a
complement to this theory when the growth rate is negative and depreciation
is by evaporation. Chapter 5, by Davide Fiaschi and Rodolfo Signorino,
investigates a problem concerning the ‘Classical’ growth model that has
rarely been on the agenda of scholars interested in modern developments of
the ‘Classical’ school (but see Pasinetti, 1981, pp. 69–70; 1993): the problem
of consumption patterns. Chapter 6 is a broad survey on ‘Keynesian’ theories
2 The Theory of Economic Growth: a ‘Classical’ Perspective
accumulation. We argue that in Ricardo the growth rate is endogenous and

may fall to zero when, during capital accumulation and population growth,
the rate of profit tends to fall due to diminishing returns in agriculture and
the exhaustion of some natural resources. Section 1.4 deals with linear
models of economic growth: the authors discussed include Robert Torrens,
1. Theories of economic growth: Karl Marx, Georg von Charasoff and John von Neumann. Section 1.5
old and new* provides a taxonomy of ‘classical’ cases in which the rate of profit, and thus
the rate of growth, need not fall to zero. Section 1.6 discusses ‘neoclassical’
ideas or models of exogenous growth. Section 1.7 classifies the recent
Heinz D. Kurz and Neri Salvadori literature on the so-called ‘new’ growth models (NGMs) into three groups
according to the route by which they try to avoid diminishing returns to
capital. Section 1.8 draws some conclusions and argues that the ‘new’
1.1. INTRODUCTION growth theory (NGT) shares some crucial elements of the classical approach
to the problem of growth and distribution.
Ever since the inception of systematic economic analysis at the time of the From the point of view of method adopted in this paper it should be
classical economists from William Petty to David Ricardo the problem of pointed out right at the beginning that the classical economists analysed the
economic growth – its sources, forms and effects – was high on the agenda economic system in motion essentially in terms of a sequence of long-period
of economists. In the real world the problem and the fact of economic growth positions of the economy reflecting discrete changes in the independent
is, of course, of much longer standing. Even in the more or less stationary variables, or ‘data’, determining the rate of profits, rents and normal
economies of antiquity the possibility, if not the fact, of economic expansion (relative) prices. These data concerned (a) the technical alternatives from
lingers at the back of certain considerations. Clay tablets from Mesopotamia which cost-minimizing producers can choose; (b) the overall level and the
provide information about social productivity by means of a simple input– composition of output; and (c) the real wage rate of common labour (taking
output calculation in terms of barley. The main question concerned the the available quantities of different qualities of land as given and non-
surplus product of barley the ancient society was able to generate, that is, the depletable). There was no presumption in the classical economists that the
excess of total output in a year with a normal harvest over the amount of economic system could be expected to converge to a dynamic steady state.
input of barley as seed or as a means of subsistence for labourers plus any While they illustrated certain concepts in terms of arguments that could be
other inputs needed in the society measured in terms of barley. From the reinterpreted as steady-state models, they did not think that the steady state
Surplus Rate, that is, the ratio of Surplus Product to Necessary Input, it is was of interest to describe and analyse real economic systems evolving in
obviously only a small step intellectually, but a huge step historically, to the historical time. We may therefore distinguish between two concepts of
concept of the rate of growth. This step was taken, at the latest, by ‘endogenous’ growth. On the one hand, we have the classical economists’
economists in the seventeenth century, most notably William Petty. view with their radical concept which sees the rate of growth at any moment
This chapter is devoted to a brief discussion of the characteristic features and also in the long run shaped by the interaction of people between one
of a selection of contributions to the problem under consideration. It another and with their environment. On the other hand, we have the much
summarizes previous contributions by the same authors. The interested more narrow view entertained in the majority of contributions to new growth
reader can see more detailed analyses in Kurz and Salvadori (1998b, 1999). theory, which argue, fully in line with the Solow model, that the long run is a
Section 1.2 summarizes some crucial features of Adam Smith’s views on steady state and add, contrary to the Solow model, that the steady-state
capital accumulation and economic growth. The emphasis is on two growth rate is determined from within the economic system and is not given
contradictory effects of capital accumulation contemplated by Smith: a from outside. In the following this crucial difference between the two
tendency of the rate of profit to fall due to the intensification of competition concepts of endogeneity has to be kept in mind. In the classical economists
among capital owners; and a tendency of the rate of profit to rise due to the we simply do not encounter the concept of an asymptotic growth rate of a
increase in productivity associated with the division of labour. Section 1.3 dynamic model except as the hypothetical case of a stationary state.
turns to David Ricardo’s approach to the theory of distribution and capital However, in order to have a sufficient basis for comparison between the
1
Theories of economic growth: old and new 3 4 The Theory of Economic Growth: a ‘Classical’ Perspective
different approaches to growth theory, old and new, we shall focus attention the market and thus upon capital accumulation. ‘The greatest improvement in
on models that are in fact steady-state models or ideas that are given a the productive powers of labour’, we are told, ‘seem to have been the effects
steady-state form. of the division of labour’ (WN I.i.1), both within given firms and industries
and, even more significantly, between them. In his analysis in the first three
chapters of book I of The Wealth of Nations Smith established the idea that
1.2. ADAM SMITH ON GROWTH there are increasing returns which are largely external to firms, that is,
broadly compatible with the classical hypothesis of a uniform rate of profit.
A characteristic feature of the classical approach is the view that production In the first chapter he made clear how powerful a device the division of
involves labour, produced means of production and natural resources. In labour is in increasing labour productivity, and analysed in some detail its
contrast to some contributions to modern growth theory none of these factors major features: (i) the improvement of the dexterity of workers; (ii) the
– labour, capital and land – were considered negligible other than in saving of time which is otherwise lost in passing from one sort of work to
conceptual experiments designed ‘to illustrate a principle’ (Ricardo). To another; and, most importantly, (iii) the invention of specific machinery (see
understand real growth processes one had to come to grips with the WN I.i.6–8). In the second chapter he argued that there is a certain
interrelated laws governing the growth of population, the pace of propensity in human nature ‘to truck, barter and exchange one thing for
accumulation and the rate and bias of technical innovation in an environment another’, which appears to be rooted in ‘the faculties of reason and speech’,
characterized by the scarcity of natural resources. At stake was an that gives occasion to the division of labour (WN I.ii.1–2). In the third
understanding of the working of a highly complex system. chapter the argument is completed by stressing that the division of labour is
limited by the extent of the market (see WN I.iii.1): a larger market generates
1.2.1. Capital Accumulation and the Division of Labour a larger division of labour among people and, therefore, among firms, and a
larger division of labour generates a larger productivity of labour for all
Adam Smith viewed the growth process as strictly endogenous (see also firms.
Lowe, [1954] 1987, p. 108, and Eltis, 1984, p. 69), placing special emphasis Despite the presence of increasing returns, Smith retained the concept of a
on the impact of capital accumulation on labour productivity. He began his general rate of profit. His argument appears to be implicitly based on the
inquiry into the Wealth of Nations by stating that income per capita hypothesis that each single firm operates at constant returns, while total
production is subject to increasing returns. Even though some examples
must in every nation be regulated by two different circumstances; first, by the provided by Smith relate more to the division of labour within firms than to
skill, dexterity, and judgment with which its labour is generally applied; and, the division of labour among firms, Smith appears to be correct in sustaining
secondly, by the proportion between the number of those who are employed in that some of the activities which were originally part of the division of labour
useful labour, and that of those who are not so employed (WN I.3). within the firm may eventually become a different ‘trade’ or ‘business’, so
that the division of labour within the firm is but a step towards the division of
According to Smith there is no upper limit to labour productivity. This is labour amongst firms. In the example of pin making at the beginning of
why Smith maintained that an investigation of the growth of income per chapter I, Smith pointed out that ‘in the way in which this business is now
capita is first and foremost an inquiry into ‘The causes of this improvement, carried on, not only the whole work is a peculiar trade, but it is divided into a
in the productive powers of labour, and the order, according to which its number of branches, of which the greater part are likewise peculiar trades’
produce is naturally distributed among the different ranks and conditions of (WN I.i.3).
men in the society’ (WN I.5). Smith’s analysis foreshadows the concepts of induced and embodied
Smith’s attention focused accordingly on the factors determining the technical progress, learning by doing, and learning by using. The invention
growth of labour productivity, that is, the factors affecting ‘the state of the of new machines and the improvement of known ones is said to be originally
skill, dexterity, and judgment with which labour is applied in any nation’ due to the workers in the production process and ‘those who had occasion to
(WN I.6). At this point the accumulation of capital enters into the picture, use the machines’ (WN I.i.9). At a more advanced stage of society making
because of Smith’s conviction that the key to the growth of labour machines ‘became the business of a peculiar trade’, engaging ‘philosophers
productivity is the division of labour which in turn depends on the extent of or men of speculation, whose trade it is, not to do any thing, but to observe
every thing; and who, upon that account, are often capable of combining cannot be traced back to an intensification of competition. Second, Smith
together the powers of the most distant and dissimilar objects’. Research and erroneously tried to carry an argument that is valid in a partial framework
development of new industrial designs becomes ‘the principal or sole trade over to a general framework. This problem was tackled by David Ricardo.
and occupation of a particular class of citizens’ (Ibidem). New technical Adam Smith explained economic growth thoroughly as an endogenous
knowledge is systematically created and economically used, with the phenomenon. The growth rate depends on the decisions and actions of
sciences becoming more and more involved in that process. The agents, especially their savings and investment behaviour, and the creativity
accumulation of capital propels this process forward, opens up new markets and innovativeness they come up with in given social and historical
and enlarges existing ones, increases effectual demand and is thus the main conditions and institutional settings. Special emphasis is placed on the
force behind economic and social development (WN V.i.e.26). Here we have endogenous creation of new knowledge that can be used economically. New
a dynamic notion of competition, conceived of as rivalry, which anticipates technical knowledge is treated as a good, which is or in the long run tends to
in important respects the views on competition of authors such as Karl Marx become a public good. There are no clear and obvious limits to growth. The
and Joseph Alois Schumpeter. Smith also anticipates the following two ideas additional work force required in the process of accumulation is generated by
that are prominent within the ‘new’ growth theory literature: (1) ‘new that process itself: labour power is a commodity, the quantity of which is
improvements of art’ are generated within the economic system by regulated by the effectual demand for it. Diminishing returns due to scarce
specialized activities; (2) new technical knowledge is or eventually will natural resources are set aside or taken to be compensated by the increase in
become a public good, that is, non-rival and non-excludable. productivity due to the division of labour.
Did Smith expect the endogenous growth factors to lose momentum as In the following much of the analysis will focus on whether savings (and
capital accumulates? He considered three potential limits to growth: an investment) behaviour has an impact on the long-run rate of economic
insufficient supply of workers, the scantiness of nature, and an erosion of the growth, or, in the framework of steady-state models, on the steady-state rate
motives of accumulation. Smith saw that the scarcity and potential depletion of growth. Accordingly, growth will be dubbed ‘exogenous’ if the long-run
of renewable and the depletion of exhaustible resources may constrain (or steady-state) rate of growth is independent of such behaviour. This
human productive activity and the growth of the economy (WN I.xi.i.3; see narrows somewhat the scope of our investigation, because the set of
also I.xi.d). At the time when he wrote, the limits to growth deriving from purposeful decisions and actions of agents includes, but is not co-extensive
nature were apparently still considered negligible. with, savings decisions (see, for example, Jones, 1995; and Eicher and
Smith also saw no danger that the process of accumulation might come to Turnovsky, 1999a).
an end because of an insufficient supply of labour and the ensuing
diminishing returns to capital. He rather advocated a view which was to
become prominent amongst the classical economists: the supply of labour is 1.3. DAVID RICARDO ON DIMINISHING RETURNS
generated within the socio-economic system, that is, endogenously. He drew
an analogy between the multiplication of animals and that of the inferior Ricardo set aside what may be called statically and dynamically increasing
ranks of people (see WN I.viii.39, 40). Smith envisaged the growth of the returns. The beneficial effects of capital accumulation on productivity
labour force as endogenous, the determinant being the rate of capital mediated through the extension of the division of labour play hardly any role
accumulation. Real wages are higher, the more rapidly capital accumulates. in his analysis. In modern parlance, the problems of externalities which
As to the impact of high and rising real wages on the rate of profit, it appears figured prominently in Smith’s analysis are given only scant attention. This
that we cannot say anything definite, given Smith’s opinion that ‘the same does not mean that Ricardo was of the opinion that they are of negligible
cause ... which raises the wages of labour, the increase of stock, tends to interest. One has to recall that Ricardo explicitly subscribed to much of
increase its productive powers, and to make a smaller quantity of labour Smith’s analysis and set himself the moderate task of correcting views of the
produce a greater quantity of work’ (WN I.viii.57). Scotsman that he deemed wrong. These concerned especially Smith’s view
Surprisingly, Smith came up with a definitive answer in chapter IX of of the long-term trend of profitability as capital accumulates. Ricardo was
book I. His explanation of a falling tendency of the rate of profit in terms of keen to show that, given the real wage rate, the rate of profits cannot fall as a
‘competition’ (WN I.ix.2) does not stand up to close examination.1 First, consequence of the ‘competition of capital’, as Smith had argued, but only
since Smith commonly presupposed free competition, a fall in profitability because of diminishing returns due the scarcity of land(s). Much of Ricardo’s
argument therefore was developed in terms of the implicit assumption that where rmin _ 0 is the minimum level of profitability, which, if reached, will
the set of (constant returns to scale) methods of production from which cost- arrest accumulation (see Ricardo, Works I, p. 120).
minimizing producers can choose, is given and constant. In such a Ricardo saw the rate of accumulation as endogenous. The demand for
framework the question then is how scarce natural resources affect labour is governed by the pace at which capital accumulates, the long-term
profitability as capital accumulates. The resulting vision is reflected in what supply of labour by the ‘Malthusian Law of Population’. Real wages may
Ricardo called the ‘natural course’ of events. rise, that is, the ‘market price of labour’ may rise above the ‘natural’ wage
As capital accumulates and population grows, and assuming the real wage rate. This is the case when capital accumulates rapidly, leading to an excess
rate of workers is given and constant, the rate of profit is bound to fall; due to demand for labour. As Ricardo put it, ‘notwithstanding the tendency of
extensive and intensive diminishing returns on land, ‘with every increased wages to conform to their natural rate, their market rate may, in an improving
portion of capital employed on it, there will be a decreased rate of society, for an indefinite period, be constantly above it’ (Ibidem, pp. 94–5).
production’ (Ricardo, Works I, p. 98). Since profits are a residual income If such a constellation prevails for some time a ratchet effect may make itself
based on the surplus product left after the used up means of production and felt: it is possible, Ricardo observed, that ‘custom renders absolute
the wage goods in the support of workers have been deducted from the social necessaries’ what in the past had been comforts or luxuries. Hence, the
product (net of rents), the ‘decreased rate of production’ involves a decrease natural wage is driven upward by persistently high levels of the actual wage
in profitability. On the assumption that there are only negligible savings out rate. Accordingly, the concept of ‘natural wage’ in Ricardo is a flexible one
of wages and rents, a falling rate of profit involves a falling rate of capital and must not be mistaken for a physiological minimum of subsistence.
accumulation. Hence, Ricardo’s ‘natural course’ of events will necessarily Setting aside the complex wage dynamics in Ricardo’s theory, that is,
end up in a stationary state. This path should not be identified with the actual assuming boldly a given and constant real wage rate and setting boldly the
path the economy is taking because technical progress will repeatedly offset minimum rate of profit equal to zero, we may illustrate Ricardo’s view of the
the impact of the ‘niggardliness of nature’ on the rate of profit (see Ricardo, long-run relationship between profitability and accumulation and thus
Works I, p. 120). growth in a schematic way. Figure 1.1, originally used by Kaldor (1955–56),
The assumption of a given real wage rate represents a first logical step in shows the marginal productivity of labour-cum-capital curve CEGH. It is
an approach to the problem of capital accumulation and income distribution decreasing since land is scarce: when labour-cum-capital increases, either
which proceeds in terms of distinct analytical stages (see Garegnani, 1990). less fertile qualities of land must be cultivated or the same qualities of land
The attention focuses first on abstract and general principles which are then must be cultivated with processes which require less land per unit of product,
gradually attuned to the concrete case or specific historical circumstances but are more costly in terms of labour-cum-capital. Let the real wage rate
under consideration. Economic theory is combined with historical analysis. equal OW. Then, if the amount of labour-cum-capital applied is L1, the area
Here we focus only on the first stage and set aside its historical part. The OCEL1 gives the product, OWDL1 gives total capital employed, and BCE
reader therefore will not be misled into thinking that in our view classical total rent.
political economy is co-extensive with or can be reduced to this first stage. It Profit is determined as a residual and corresponds to the rectangle WBED.
reaches far beyond it. As a consequence, the rate of profit can be determined as the ratio of the
Like Smith, Ricardo thought that saving and investment, that is, areas of two rectangles which have the same base and, therefore, it equals the
accumulation, would largely come from profits, whereas wages and rents ratio WB/OW. Let us now consider the case in which the amount of labour-
played a negligible role. Hence, as regards the dynamism of the economy cum-capital is larger, that is, L2. Then OCGL2 gives the product, OWFL2 the
attention should focus on profitability. Assuming that the marginal capital, ACG the rent, and WAGF the profits. The rate of profit has fallen to
propensity to accumulate out of profits, s, is given and constant, a ‘classical’ WA/OW. Obviously, if a positive profit rate implies a positive growth rate,
accumulation function can be formulated the economy will expand until labour-cum-capital has reached the level L .
At that point the profit rate is equal to zero and so is the growth rate. The
⎧s(r − rmin ) if r ≥ rmin system has arrived at the so-called stationary state: growth has come to an
g=⎨ end because profitability has.
⎩0 if r ≤ rmin
Marginal productivity of labour-cum-capital extensive and intensive diminishing returns make themselves felt and are not
counteracted by sufficient technical progress. This also shows that it is not a
necessary condition, for a theory to be considered a theory of endogenous
C growth, that it assumes some kind of increasing returns. This becomes clear
in Section 1.4 below. Interestingly, its main message was anticipated by
Ricardo.
E Ricardo contemplated the implications for income distribution and the rate
B
G of expansion of the economic system in the hypothetical case in which land
A
of the best quality is available in abundance. In one place he wrote:
H Profits do not necessarily fall with the increase of the quantity of capital because
W
D F the demand for capital is infinite and is governed by the same law as population
itself. They are both checked by the rise in the price of food, and the consequent
increase in the price of labour. If there were no such rise, what could prevent
population and capital from increasing without limit? (Ricardo, Works VI, p. 301)
_
0 L L L Labour-cum-capital If land of the best quality were available in abundance it would be a free
1 2
good and no rent would be paid for its use. In this case the curve of the graph
showing the marginal productivity of labour-cum-capital would be a
Figure 1.1: The one-commodity Ricardian model with land as an
horizontal line and the rate of profit would be constant whatever the amount
indispensable resource
of labour-cum-capital employed. As a consequence, other things being equal,
the growth rate would also be constant: the system could grow for ever at a
For both Smith and Ricardo the required size of the work force is essentially
rate that equals the given rate of profit times the propensity to accumulate.
generated by the accumulation process itself. In other words, labour power is
As the passage from Ricardo’s Works just quoted shows, Ricardo was
treated as a kind of producible commodity. It differs from other commodities
perfectly aware of this implication.
in that it is not produced in a capitalistic way in a special industry on a par
with other industries, but is the result of the interplay between the growth of
the working population and socioeconomic conditions. In the most simple
and abstract conceptualization possible, labour power is seen to be in elastic 1.4. LINEAR CLASSICAL MODELS OF PRODUCTION
supply at a given real wage basket. Increasing the number of baskets
available in the support of workers involves a proportional increase of the Central elements of classical analysis are the concept of production as a
work force. In this view the rate of growth of labour supply adjusts to any circular flow and the related concept of surplus product left after the wage
given rate of growth of labour demand without necessitating a variation in goods and what is necessary for the replacement of the used up means of
the real wage rate. production have been deducted from the annual output. This surplus can be
In a more sophisticated conceptualization, higher rates of growth of consumed or accumulated. With constant returns to scale and setting aside
labour supply presuppose higher levels of the real wage rate. But the basic the problem of scarce natural resources, the notion of an economy expanding
logic remains the same: in normal conditions the pace at which capital at a constant rate of growth was close at hand. In this section we shall
accumulates regulates the pace at which labour, a non-accumulable factor of mention some contributions to what may be called linear growth theory with
production, grows. Thus labour cannot put a limit to growth because it is a classical flavour.
generated within the growth process. The only limit to growth can come Robert Torrens in his Essay on the External Corn Trade clarified that the
from other non-accumulable factors of production. In other words, there is concept of surplus provides the key to an explanation of the rate of profit.
only endogenous growth in Ricardo. This growth is bound to lose Growth in the model by Torrens is both linear and endogenous; the rate of
momentum as the system hits its natural barriers, especially as soon as growth depends on the general rate of profit and the propensity to
accumulate. The same can be said of Marx’s theory of expanded therefore the rate of growth, does not fall to zero. There is perpetual growth
reproduction in chapter 21 of volume II of Capital (Marx, [1885] 1956).2 provided that the premises underlying the different cases hold infinitely. It
There Marx studied the conditions under which the system is capable of will be seen that while the cases discussed are all derived from a classical
reproducing itself on an upward spiralling level. The expansion of the framework of the analysis as it was developed by Adam Smith and David
economy at an endogenously determined rate of growth is possible. This rate Ricardo, the cases exhibit some striking similarities to the types of NGMs
depends on the proportion of the surplus value ploughed back into the discussed in Section 1.7.
productive system to increase the scale of operation. Marx stressed that the
accumulation of capital is ‘an element immanent in the capitalist process of 1.5.1. Constant Returns to Capital
production’ (Ibidem, p. 497; emphasis added). For, ‘the aim and compelling
motive of capitalist production’ is ‘the snatching of surplus-value and its As we have seen, the main ingredient to obtain a stationary state in the
capitalisation, i.e., accumulation’ (Ibidem, p. 507). Ricardian model is the existence of land available in limited supply. If land
The Russian mathematician Georg von Charasoff elaborated on Marx’s were not needed as an input or if land of the best quality were available in
analysis and was possibly the first to provide a clear statement of the abundance, then the graph giving the marginal productivity of labour-cum-
fundamental duality relationship between the system of prices and the rate of capital would be a horizontal line and therefore the rate of profit would be
profit on the one hand, and the system of quantities and the rate of growth on constant whatever the amount of labour-cum-capital. As a consequence, the
the other (see Charasoff, 1910). He developed his main argument within the growth rate would also be constant.
framework of an interdependent model of (single) production exhibiting all
the properties of the later input–output model, and which is fully specified in 1.5.2. Backstop Technology
terms of use values (rather than labour values as in the case of Marx) and
labour needed per unit of output. To assume that land is not useful in production or that it is available in given
The most sophisticated linear model of endogenous growth was quality and unlimited quantity is unnecessarily restrictive. It is enough to
elaborated by John von Neumann (1945) in a paper first published in assume that ‘land’, although useful in production, is not indispensable. In
German in 1937 and then translated into English in 1945. In it von Neumann other words, there is a technology that allows the production of the
assumed there are n goods produced by m constant returns–to–scale commodity without any ‘land’ input. With continuous substitution between
production processes. There is a problem of the choice of technique which labour-cum-capital and land, the marginal productivity of labour-cum-capital
consists in establishing which processes will actually be used and which not, would be continuously decreasing, but it would be bounded from below.
being ‘unprofitable’. Von Neumann (1945, pp. 1–2) took the real wage rate, With respect to the case depicted in Figure 1.1, the CEGH curve is
consisting of the ‘necessities of life’, to be given and paid at the beginning of decreasing, as in Figure 1.1, but the concavity is reversed and there is a
the uniform period of production, that is, he considered wages as part of the horizontal asymptote. In this case the profit rate and thus the growth rate are
capital advanced and thus as part of the physical real costs of production. In falling, but they could never fall below certain positive levels. The system
addition, he assumed ‘that all income in excess of necessities of life will be would grow indefinitely at a rate of growth that asymptotically approaches
reinvested’. In von Neumann’s model the rate of growth is determined the product of the given saving rate times the value of the (lower) boundary
endogenously.3 He set aside the problem of scarcity of all non-accumulable of the profit rate.
factors of production: while all primary factors other than labour (that is, all
natural resources) were taken to be available at whichever amount was 1.5.3. Increasing Returns to Capital
needed at zero price, labour was assumed to be available at the required
amount at a given real wage rate. The final case is that of increasing returns to labour-cum-capital as was
discussed, following Adam Smith, by Allyn Young (1928) and Nicholas
Kaldor (1957 and 1966). Taking the wage rate as given and constant, the rate
1.5. A CLASSIFICATION OF CASES of profit and the rate of growth will rise as more labour-cum-capital is
employed. To preserve the notion of a uniform rate of profit, it is necessary
We can now classify some broad cases in which the rate of profit, and to assume that the increasing returns are external to the firm and exclusively
connected with the expansion of the market as a whole and the social will remain of the same quality, and in the same general relations to one another,
division of labour. This implies that while in the case of decreasing returns though they are all increasing in volume (Ibidem, p. 306).
due to the scarcity of land the product was given by the area under the
marginal productivity curve, now the product associated with any given The resulting economic system grows at a constant rate which equals the
amount of labour-cum-capital is larger than or equal to that amount exogenous rate of growth of population.5 Income distribution and relative
multiplied by the corresponding level of output per unit of labour-cum- prices are the same as in the stationary economy. In modern parlance: the
capital. In any case, the sum of profits and wages equals the product of the system expands along a steady-state growth path.
given amount of labour-cum-capital multiplied by the corresponding level of We encounter essentially the same idea in Gustav Cassel’s ([1918] 1932)
output per unit of labour-cum-capital.4 As a consequence, the product is Theory of Social Economy. The model of exogenous growth delineated by
larger than the area under the marginal productivity curve. The cases of Cassel can be considered the proximate starting point of the development of
decreasing and increasing returns are therefore not symmetrical: with neoclassical growth theory. In chapter IV of book I of the treatise Cassel
increasing returns a rising real wage rate need not involve a falling general presented two models, one of a stationary economy, the other of an economy
rate of profit. growing along a steady-state path.
In his first model Cassel assumed that there are z (primary) factors of
production. The quantities of these resources and thus the amounts of
1.6. MODELS OF EXOGENOUS GROWTH services provided by them are taken to be in given supply. General
equilibrium is characterized by equality of supply and demand for each
The marginalist or ‘neoclassical’ school of economic thought seeks to factor service and for each good produced, and equality of the price of a
explain income distribution in a symmetrical way via the relative scarcities of good and its cost of production. The resulting sets of equations constitute
the factors of production, labour, ‘capital,’ and land. Interestingly, the idea of what is known as the ‘Walras–Cassel model’ (Dorfman, Samuelson and
exogenous growth which classical theory did not entertain is the starting Solow, 1958, p. 346). It satisfies the then applied criterion of completeness:
point of important early works in the marginalist tradition. there are as many equations as there are unknowns to be ascertained.
Cassel (1932, pp. 152–3) then turned to the model of a uniformly
1.6.1. Alfred Marshall and Gustav Cassel progressing economy. Although described only verbally, he introduced the
model in the following way:
The idea of an economic system growing exclusively because some
We must now take into consideration the society which is progressing at a uniform
exogenous factors make it grow has variously been put forward in the history
rate. In it, the quantities of the factors of production which are available in each
of economic thought as a standard of comparison. For example, in chapter V period ... are subject to a uniform increase. We shall represent by [g] the fixed rate
of book V of his Principles, first published in 1890, Alfred Marshall ([1890] of this increase, and of the uniform progress of the society generally.
1977, p. 305) introduced the ‘famous fiction of the “Stationary state” ... to
contrast the results which would be found there with those in the modern In Cassel’s view this generalization to the case of an economy growing at an
world’. By relaxing one after another of the rigid assumptions defining the exogenously given and constant rate does not cause substantial problems.
stationary state, Marshall sought to get gradually closer to the ‘actual The previously developed set of equations can easily be adapted
conditions of life’. The first relaxation concerned the premise of a constant appropriately, ‘so that the whole pricing problem is solved’. Cassel thus
(working) population: arrived at basically the same result as Marshall.
The Stationary state has just been taken to be one in which population is
1.6.2. Robert Solow, Trevor Swan and James Meade
stationary. But nearly all its distinctive features may be exhibited in a place where
population and wealth are both growing, provided they are growing at about the
same rate, and there is no scarcity of land: and provided also the methods of The neoclassical growth models of the 1950s and early 1960s differ from the
production and the conditions of trade change but little; and above all, where the growth version of the Walras–Cassel model in five important respects:
character of man himself is a constant quantity. For in such a state by far the most
important conditions of production and consumption, of exchange and distribution
1. they are macro-models with only one produced good which could be 1.7. THE ‘NEW’ MODELS OF ENDOGENOUS GROWTH
used both as a consumption good and as a capital good;
2. the number of primary factors of production is reduced to one, One of the key properties of the NGMs emphasized by their advocates is the
homogeneous labour (as in Solow, 1956 and 1963; Swan, 1956), or two, limitation of diminishing returns to capital. The first generation of NGMs
homogeneous labour and homogeneous land (as in Swan, 1956; Meade, defined the confines within which subsequent contributions to NGT were
1961); carried out. The attention focuses on the mechanism that prevents the returns
3. the all-purpose good is produced by means of labour, capital, that is, the to capital from falling (below a certain level).6
good itself, and possibly land;
4. there is a choice of technique, where technical alternatives are given by a 1.7.1. Constant Returns to Capital
macroeconomic production function, which is homogeneous of degree
one with positive and decreasing marginal productivities with respect to The first class of models set aside all non-accumulable factors of production
each factor of production; and such as labour and land and assume that all inputs in production are
5. planned saving, which is taken to be equal to planned investment at all accumulable, that is, ‘capital’ of some kind. The simplest version of this
times, is proportional to net income, that is, a ‘Keynesian’ saving class is the so-called ‘AK model’, which assumes that there is a linear
function is assumed. relationship between total output, Y, and a single factor capital, K, both
consisting of the same commodity:
Focusing attention on the models with a single primary factor (labour), in
steady-state equilibrium Y = AK, (1)
sf(k) = gk, where 1/A is the amount of that commodity required to produce one unit of
itself. Because of the linear form of the aggregate production function, these
where s is the (marginal and average) propensity to save, f(k) is the
models are also known as ‘linear models’. This model is immediately
production function per unit of labour or per capita, k is the capital–labour
recognized as the model dealt with in Subsection 1.5.1. The rate of return on
ratio (where labour is measured in terms of efficiency units), and g is the
capital r is given by
steady-state growth rate of capital (and labour, and income etc.). In steady-
state equilibrium output expands exactly as the exogenous factors make it Y
grow. Note that assuming s > 0 presupposes that the exogenous factors are r+δ= = A, (2)
K
growing at some positive rate. In these models the steady-state rate of growth
is exogenous. Outside steady-state equilibrium the rate of growth can be where δ is the exogenously given rate of depreciation. There is a large
shown to depend also on the behavioural parameter of the system, that is, the variety of models of this type in the literature. In the two-sector version in
propensity to save (and invest), but that parameter plays no role in Rebelo (1991) it is assumed that the capital good sector produces the capital
determining the long-term rate of growth. good by its own means and nothing else. It is also assumed that there is only
While these models are aptly described as models of exogenous growth, one method of production to produce the capital good. Therefore, the rate of
they can also be described as models of endogenous profitability. Since in profit is determined by technology alone. Then the saving-investment
the one-good framework adopted by the authors under consideration the rate mechanism jointly with the assumption of a uniform rate of growth, that is, a
of profit r equals the marginal productivity of capital, steady-state equilibrium, determines a relationship between the growth rate,
g, and the rate of profit, r. Rebelo (1991, pp. 504 and 506) obtains either
r = f’(k),
the two equations are able to determine a relationship between the rate of A−δ − ρ r − ρ
g= = , (3)
profit and the steady-state rate of growth. The following section shows that σ σ
the NGMs essentially reverse what is endogenous and what is exogenous. In or
other words, without over-exaggeration they can be called models of
endogenous growth and exogenous profitability. g = (A – δ)s = sr. (4)
Equation (3) is obtained when savings are determined on the assumption where h(k) is the per capita production function and b is a positive constant.
that there is an immortal representative agent maximizing the following The special case contemplated by them is
inter-temporal utility function
h(k) = f(k) + bk, (5)
∞
1
∫e
− ρt
⎡ c(t )1−σ − 1⎤⎦ dt , where f(k) is the conventional per capita production function. As capital
0
1−σ ⎣ accumulates and the capital–labour ratio rises, the marginal product of
capital will fall, approaching asymptotically b, its lower boundary. With a
subject to constraint (1), where ρ is the discount rate, or rate of time given propensity to save, s, and assuming capital never wears out, the steady-
preference, and 1/σ is the elasticity of substitution between present and state growth rate g is endogenously determined: g = sb. Assuming, on the
future consumption (1 σ > 0), and where Y = c(t) + K . Equation (4) is contrary, intertemporal utility maximization, the rate of growth is positive
obtained when the average propensity to save s is given. Hence, in this model
provided the technical parameter b is larger than the rate of time preference
the rate of profit is determined by technology alone and the saving- ρ. In the case in which it is larger, the steady-state rate of growth is given by
investment mechanism determines the growth rate. equation (3) with r = b.
King and Rebelo (1990) essentially followed the same avenue. Instead of It may be easily recognized that the difference between the model of
one kind of ‘capital’ they assumed that there are two kinds, real capital and Jones and Manuelli (1990) and that of Rebelo (1991) is the same as that
human capital, both of which are accumulable. There are two lines of existing between the case dealt with in Subsection 1.5.1 and that dealt with in
production, one for the social product and the real capital, which consist of Subsection 1.5.2.
quantities of the same commodity, and one for human capital. The
production functions relating to the two kinds of capital are assumed to be 1.7.3. Factors Counteracting Diminishing Returns to Capital
homogeneous of degree one and strictly concave. There are no diminishing
returns to (composite) capital for the reason that there is no non-accumulable Finally, there is a large class of models contemplating various factors
factor such as simple or unskilled labour that enters into the production of the counteracting any diminishing tendency of returns to capital. Here we shall
accumulable factors, investment goods and human capital. As in Rebelo’s be concerned only with the following two sub-classes: human capital
model the rate of profit is uniquely determined by the technology (and the formation and knowledge accumulation. In both kinds of models positive
maximization of profits which, because of the Non-substitution Theorem,7 external effects play an important part; they offset any fall in the marginal
implies that only one technique can be used in the long run); the growth rate
product of capital.
of the system is then endogenously determined by the saving–investment
equation. The greater the propensities to accumulate human and physical A. Human capital formation
capital, the higher is the growth rate. Models of the first sub-class attempt to formalize the role of human capital
formation in the process of growth. Elaborating on some ideas of Uzawa
1.7.2. Returns to Capital Bounded from Below (1965), Lucas (1988) assumed that agents have a choice between two ways
of spending their (non-leisure) time: to contribute to current production or to
The second class of models preserve the dualism of accumulable and non- accumulate human capital. With the accumulation of human capital there is
accumulable factors but restrict the impact of an accumulation of the former said to be associated an externality: the more human capital society as a
on their returns by modification of the aggregate production function. Jones whole has accumulated, the more productive each single member will be.
and Manuelli (1990), for example, allow for both labour and capital and even This is reflected in the following macroeconomic production function
assume convex technology, as the Solow model does. However, convex
technology requires only that the marginal product of capital is a decreasing Y = AKβ(uhN)1 – βh*γ, (6)
function of its stock, not that it vanishes as the amount of capital per worker
where the labour input consists of the number of workers, N, times the
tends towards infinity. Jones and Manuelli assume that
fraction of time spent working, u, times h which gives the labour input in
h(k) bk, each k 0, efficiency units. Finally, there is the term h*. This is designed to represent
the externality. The single agent takes h* as given in his or her optimizing.
However, for society as a whole the accumulation of human capital increases be withheld from producing current output. The basic idea of Romer’s (1986,
output both directly and indirectly, that is, through the externality. p. 1015) model is ‘that there is a trade-off between consumption today and
Lucas’s conceptualization of the process by means of which human knowledge that can be used to produce more consumption tomorrow’. He
capital is built up is the following: formalizes this idea in terms of a ‘research technology’ that produces
‘knowledge’ from forgone consumption. Knowledge is assumed to be
h = υh(1 – u), (7) cardinally measurable and not to depreciate: it is like perennial capital.
where υ is a positive constant. (Note that equation (7) can be interpreted as a Romer stipulates a research technology that is concave and homogeneous
‘production function’ of human capital.) of degree one,
Interestingly, it can be shown that if the above-mentioned externality is ki = G(Ii, ki), (8)
not present, that is, if γ in equation (6) equals zero, and therefore returns to
scale are constant and, as a consequence, the Non-substitution Theorem where Ii is an amount of forgone consumption in research by firm i and ki is
holds, endogenous growth in Lucas’s model is obtained in essentially the the firm’s current stock of knowledge. (Note that the forgone consumption
same way as in the models by Rebelo (1991) and King and Rebelo (1990): good is a capital good utilized in the production of ‘knowledge’.) The
the rate of profit is determined by technology and profit maximization alone; production function of the consumption good relative to firm i is
and for the predetermined level of the rate of profit the saving–investment
mechanism determines the rate of growth. Yet, as Lucas himself pointed out, Yi = F(ki, K, xi), (9)
endogenous growth is positive independent of the fact that there is the above- where K is the accumulated stock of knowledge in the economy as a whole
mentioned externality, that is, independent of the fact that γ is positive.8 and xi are all inputs different from knowledge. The function is taken to be
Therefore, while complicating the picture, increasing returns do not add homogeneous of degree one in ki and xi and homogeneous of a degree
substantially to it: growth is endogenous even if returns to scale are greater than one in ki and K. Romer (1986, p. 1019) assumes that ‘factors
constant. If returns to scale are not constant then the Non-substitution other than knowledge are in fixed supply’. This implies that ‘knowledge’ is
Theorem does not apply, implying that neither the competitive technique nor the only capital good utilized in the production of the consumption good.
the associated rate of profit are determined by technical alternatives and Spillovers from private research and development activities increase the
profit maximization alone. Nevertheless, these two factors still determine, in public stock of knowledge K.
steady states, a relationship between the rate of profit and the rate of growth. Assuming, contrary to Romer, that the above production function (9) is
This relationship together with the relationship between the same rates homogeneous of degree one in ki and K involves a constant marginal product
obtained from the saving–investment mechanism determines both variables. of capital: the diminishing returns to ki are exactly offset by the external
Although the analysis is more complex, essentially the same mechanism improvements in technology associated with capital accumulation. In this
applies as in the models dealt with in Subsection 1.7.1. case it can be shown that, similar to the NGMs previously dealt with, the rate
of profit is determined by technology and profit maximization alone,
B. Technical change provided, as is assumed by Romer, that the ratio K/ki equals the (given)
Models of the second sub-class attempt to portray technological change as number of firms. The saving–investment relation then determines
generated endogenously. The proximate starting point for this kind of model endogenously the growth rate. Once again, endogenous growth does not
was Arrow’s (1962) paper on ‘learning by doing’. Romer (1986) focuses on depend on an assumption about increasing returns with regard to
the role of a single state variable called ‘knowledge’ or ‘information’ and accumulable factors. Growth would be no more endogenous if increasing
assumes that the information contained in inventions and discoveries has the returns were to be assumed: such an assumption would only render the
property of being available to anybody to make use of it at the same time. In analysis a good deal more complicated. In particular, a steady-state
other words, information is considered essentially a non-rival good. Yet, it equilibrium does not exist, and in order for an equilibrium to exist the
need not be totally non-excludable, that is, it can be monopolized at least for marginal product of capital must be bounded from above. This is effected by
some time. It is around the two different aspects of publicness – non-rivalry Romer in terms of an ad hoc assumption regarding equation (8) (Ibidem,
and non-excludability – that the argument revolves. Discoveries are made in p. 1019). This assumption is not different from the one used in drawing
research and development departments of firms. This requires that resources
Figure 4, where the marginal product of corn is shown to be increasing with f(L, L*) as a function of L alone is either decreasing (if land is scarce) or constant (if land is
not scarce). The product at L* equals ∫0 f ( L, L*)dL , i.e., the area under the curve f(L, L*)
L*
the scale of production, but is bounded from above.
in the range [0, L*]. If (∂f ∂L *) > −( ∂f ∂L ) for L* = L, then the curve x = f(L, L*) is
increasing, but the product is, as stated in the text, larger than or equal to the sum of profits
and wages, which equals the product of the given amount of labour-cum-capital multiplied
1.8. CONCLUSION by the corresponding level of output per unit of labour-cum-capital.
5. It should be noted that Marshall (1977, book IV, ch. IV) saw reason to suppose that the
The NGMs revolve around a few simple and rather obvious ideas which have growth of population depended, among other things, on socioeconomic factors and thus
could not sensibly be treated, other than in an initial step of the analysis, as exogenous.
been anticipated by earlier economists, most notably Adam Smith and David
6. For a more detailed treatment of these models, see Kurz and Salvadori (1998b).
Ricardo. Many of the interesting aspects of the NGMs are related to the 7. We need a special case of the Non-substitution Theorem, because no primary factor (or a
classical perspective that their authors (often unwittingly) take on the primary factor with a zero remuneration) is assumed; see Kurz and Salvadori (1994).
problem of growth, whereas some of their shortcomings derive from the lack 8. For a demonstration of this, see Kurz and Salvadori (1998b).
of solutions to the problems of the neoclassical theory of growth which were
put into sharp relief during the 1960s and 1970s. It has also been hinted that
in some non-neoclassical approaches to the theory of accumulation and
growth, the endogeneity of the growth rate has always been taken for
granted. A brief look into the history of economic thought shows that from
Adam Smith via David Ricardo, Robert Torrens, Thomas Robert Malthus,
Karl Marx up to John von Neumann both the equilibrium and the actual rate
of capital accumulation and thus both the equilibrium and the actual rate of
growth of output as a whole were seen to depend on agents’ behaviour, that
is, endogenously determined. In this regard there is indeed nothing new
under the sun.
NOTES
* We should like to thank Mauro Caminati for valuable comments and suggestions. All
remaining errors and omissions are entirely our responsibility.
1. For an interesting different view placing special emphasis on Malthus’s interpretation of
Smith according to which Smith had ruled out constant and diminishing returns, see Negishi
(1993).
2. In Marx’s analysis, this theory is only a logical step toward a proper theory of accumulation.
Here we cannot deal with the latter and Marx’s ‘law’ of a falling tendency of the rate of
profits in Volume III of Capital. Here Marx argues that a tendency of the real wage rate
toward a socially and historically defined subsistence level is not due to a population
mechanism, but due to the presence of an ‘industrial reserve army of the unemployed’,
which is continually filled and re-filled by labour–saving technical progress.
3. This is one of the reasons why the conventional interpretation of that model as belonging to
the tradition established by the so-called ‘Walras–Cassel model’ cannot be sustained (see
Kurz and Salvadori, 1993). Cassel (1932) took as exogenously given the rates of growth of
all primary factors and assumed their continuous full employment (see Section 1.6 below).
Von Neumann never made this assumption.
4. Let x = f(L, L*) be the product of the last unit of labour-cum-capital, when L represents the
amount of labour-cum-capital employed and the division of labour is artificially kept fixed
at the level appropriate when the amount of labour-cum-capital employed is L*. Obviously,
the economic process. In fact, while the neoclassical theory of factor–income

distribution is hardly consistent with endogenous growth (Frankel, 1962),
classical economists, basing their distribution theory on non-economic
elements and assuming full reproducibility of means of production and
constant returns, constructed a theory of growth which was able to ensure the
2. The structure of growth models: possibility of persistent (endogenous) growth.
a comparative survey* While the work of Kurz and Salvadori is helpful in highlighting the
continuity of the endogenous growth theory with the classical tradition, its
scope is too restricted to allow full appreciation of the features of each theory
Antonio D’Agata and Giuseppe Freni in terms of common and idiosyncratic elements. On the other hand, a
complete idea of the basic structure of each theory is important for possible
cross-fertilisation between different streams of analysis. The aim of this
2.1. INTRODUCTION paper is to provide a first attempt in this direction by examining in a very
simple way the structure of the most important growth models. While we
The basic problem of growth theory is to describe the behaviour of an accept Kurz and Salvadori’s idea that the endogenous growth theory can be
expanding economy over time. The more traditional way to conceive growth considered as retrieving the classical view as far as distribution and the
is to consider it as due to the accumulation of capital. In its attempt to sources of growth are concerned, we shall try to point out the existence of
construct a theory of growth (see e.g. Solow, 1956), neoclassical economics common aspects and specific features of the models considered. To be more
sought to extend its static theory of distribution to a dynamic context, and in specific we shall emphasise that there is a continuity between classical
order to succeed in this attempt it had to assume decreasing returns with growth theory, neoclassical growth theory and endogenous growth theory as
respect to the accumulated factor (see e.g. Bertola, 1993, 1994). This far as the relationship between savings and investment is concerned, in that
assumption results in the accumulation process having only transitory effects all these models conceive the investment decision as always co-ordinated
on the rate of growth, whose long-run behaviour therefore remains with the saving decision while, as is well known, the Keynesian models
unexplained within the model and is characterised by the constancy of the propose a radically different view in which investment determines savings
capital/labour and product/labour ratios. As a consequence, empirically and thus the theory of growth proves intrinsically connected with the theory
relevant examples of permanent growth, like sustained increase in the per of the business cycle. However, as we will point out, there are elements of
capita stock of capital, are attributed to the ‘compensating influence of continuity between the classical tradition, the Keynesian one and the
residual factors that have been assumed away in the model’ (Kaldor, 1961, endogenous growth theory with regard to the view concerning the adaptation
p. 177). of the rate of growth of population and that of investment.
The properties of the neoclassical growth theory have always been The paper is organised as follows. In the next section the early
questioned not only on empirical but also on theoretical grounds. One of the contributions to growth theory by classical economists and von Neumann are
main criticisms has been that the rate of growth of economies should depend illustrated. In Section 2.3 the Keynesian tradition will be considered. Section
upon the thriftiness of the economy and that technical change should be the 2.4 deals with various streams of neoclassical growth theory, from the Solow
outcome of intentional decisions of economic agents. The recent literature on model to the discounted Ramsey models of Cass (1965) and Koopmans
the endogenous growth theory has been successful in dealing with such (1965). In Section 2.5 we shall review the Lucas and the 1990 Romer models
criticisms and has been able to construct a variety of models in which the rate of endogenous growth. In Section 2.6 we shall summarize the main aspects
of growth depends upon the saving decision of households and/or of each model. We seek to emphasize that our analysis is not exhaustive, and
technological change is the intentional or unintentional outcome of the several models are not considered. Among others, all overlapping generation
maximising behaviour of agents. models and models with altruism are excluded from analysis.
Kurz and Salvadori (1998b; 1999) and other authors have pointed out that
the endogenous growth theory represents a substantial break with respect to
the neoclassical theory by recovering various aspects of the classical view of
23
The structure of growth models: a comparative survey 25 26 The Theory of Economic Growth: a ‘Classical’ Perspective
2.2. ACCUMULATION-ORIENTED MODELS VI, p. 301). Moreover, the rate of profit is determined by the wage rate,
which in turn depends upon the conditions of the labour market. Classical
2.2.1. Classical Models of Economic Growth economists held that the rate of growth of population depends upon the wage
rate: a wage rate higher than the natural one, i.e. that which maintains
Classical economists centred their attention on the economic growth of constant population (Works, I, p. 91), yields a population increase
nations and explained this phenomenon through a theory based upon the (WN I.viii.34–42, Works, Ibidem). More specifically the higher the wage
class structure of the capitalist economy. They identify three classes, rate, the higher the rate of growth of population (WN I.viii.22).3
workers, capitalists and landowners, which have their own specific role in Hence Smith, anticipating Malthus, conceives the demographic law as a
the economic process. Workers own labour and sell it on the labour market technological rule for producing labour as any good: ‘… the demand for
for a subsistence wage. Landowners rent their land to capitalists in order to men, like that for any other commodity, necessarily regulates the production
obtain rent. Capitalists own the produced means of production and organise of men; quickens it when it goes on too slowly, and stops it when it advances
production by employing labour and renting land. Profits are their income. too fast’ (WN I.viii.40).
The behaviour of all these agents is governed by their attempt to get as much Within the classical framework it is possible to consider two kinds of
as possible from the resources they own. models. The first is the Ricardian model, which emphasises the tendency
As far as the use of the income is concerned, the usual interpretation of towards the stationary state due to the existence of scarce natural resources,
classical economists is that they conceive each class as being characterised the second model is rooted in the Smithian and Marxian tradition which
by specific behaviour (Kaldor, 1961, p. 180):1 while workers and landowners emphasises the progressive nature of economic growth. Both models are
substantially consume all the income they get, capitalists save and invest commonly interpreted as being characterised by the assumptions that capital
essentially the entire amount of profit. However, while workers buy mainly and labour are employed in fixed proportions, capitalists anticipate wages
subsistence goods, due to the low level of wages, landowners buy mainly that are entirely consumed by workers, landowners consume all rents and
luxury goods. An important aspect of the classical view on the consumption capitalists/entrepreneurs invest all profits. While the existing amount(s) of
decision is that allocation of income is not determined by preferences or by land(s) is (are) constant, labour supply is constant only in the short run,
the type of income earned but mainly by the social group to which those who whereas in the long run it is infinitely elastic at the natural wage rate w*.
receive the income belong. The sociological analysis of the rise of capitalism This means that the supply of labour can be eventually increased (or
carried out by Smith in Book III of The Wealth of Nations, makes this decreased) indefinitely at the wage rate w*, although the Malthusian law
position very clear by pointing out that capitalism arises from the emergence mentioned above regulates the rate of increase of population during any
of a class, the ‘merchants and artificers’, which ‘acted merely from a view to transitional phase.
their own interest’, by contrast ‘[T]o gratify the most childish vanity was the To illustrate the structure of the basic classical model, we consider a
sole motive of the great proprietors’ (WN III.iv.17).2 Capitalists are the class growing economy at generic time t. At the beginning of this period, the
which use their riches in order to improve their condition by accumulating economy is endowed with a set of available methods of production, Mt, with
them. The mechanism by which savings are transformed into investment can a given amount of physical capital, Xt, with a given amount of labour, Nt, and
be direct, if the savers are also entrepreneurs, or indirect, through the capital with a given amount of natural resources (land), T (T is a vector if land is of
market, if savers are not entrepreneurs. In any case classical economists different quality). The amount of land is assumed to be constant, while the
accept the view that all savings are transformed into investment (see e.g. other magnitudes can change over time. Set Mt must be interpreted as the set
Hagemann, 1998). However, while classical economists accept the idea that of methods describing the output per worker that can be obtained on the
savings can be equalised to investments through the capital market, it can be different pieces of land or on the same piece of land and that are available in
hardly said that they share the neoclassical view according to which this the economy. Every agent wants to employ all the resources he/she owns; in
equalisation is due to the adjustment of the interest rate (Hahn and Matthews, particular, capitalists end up employing the whole capital Xt. Competition
1964, pp. 12–15). In fact, for example, for Smith and Ricardo the interest among capitalists (workers) equalizes the rates of profits (the wage rate) on
rate is determined by the rate of profits (WN I.vi.18, Works, I, pp. 363–4), different industries, while competition among landowners drives the rent on
while the equalisation between savings and investment is ensured by the fact idle lands to zero. For the sake of simplicity, following Smith,we assume that
that ‘the demand for capital is infinite’ at the current rate of interest (Works,
the economy is growing and that the wage rate is at such a level as to ensure the circulating capital and the wage goods anticipated to workers,
a growth of population equal to the growth of demand for labour.4 respectively. The first function is illustrated in Figure 2.1, where it is
For a given amount of capital Xt, therefore, the wage rate wt determines indicated by Y(wt).
the set of cost-minimising methods of production, Mt*, and the rate of
profits. By activating the cost-minimising methods of production, capitalists
determine the allocation of total capital between physical capital (Kt) and
wage goods anticipated to workers (Wt), and the decisions concerning the
demand for labour (Lt). The net production Yt – Xt is determined as well and
is devoted to pay rents and profits. Rents are entirely spent in consumption
while profits are entirely saved and reinvested. At the beginning of time t+1,
the investment It determines the stock of capital Xt+1, the available
technology, Mt+1, may be different from the technology available the
preceding period,5 the supply of labour is changed according to an
exogenously given Malthusian rule, while the amount of natural resources
remains unchanged, by assumption.
The Ricardian model can be considered a particular version of the basic
classical model, and it is characterised by the further assumptions that the set
of available methods is constant over time and that land scarcity that goes
with capital accumulation pushes output per worker down below what it is
needed for reproduction. By contrast, the Smithian–Marxian model is
characterised by non-decreasing productivity of labour and capital.6 This is
justified by the fact that Smith and Marx do not lay any emphasis on the role
of natural resources as a factor limiting growth. The Ricardian and the
Smithian-Marxian model have been analysed by Kurz and Salvadori (this
volume Ch.1) to which we refer.
Here, we focus upon the short run adjustment towards the steady state Figure 2.1 – The Smithian–Marxian model
path with full employment in the Smithian-Marxian theory. The logic of this
process can be intuitively grasped by means of Figure 2.1.7 Let us assume Note that Y(wt) rotates clockwise as the wage rate increases. Suppose that the
that there is only one good produced by itself as circulating capital and by population rate of growth is Gn(wt). Since the capital/labour ratio is a
labour under constant returns, and that a single method of production is constant, if the stock of capital at time t is Xt, then in order to maintain full
available. Without loss of generality we can now suppose that the given employment at the given wage rate in the next period, it is necessary to
method uses v units of capital and 1 unit of labour per unit of output. Let Yt supply (1+ Gn(wt))Xt at the end of period t as capital. This requirement is
be the production level of period t. Then Yt = min (Kt/v, Lt), where Kt is the illustrated in Figure 2.1 by the curve X'(wt). This curve rotates anticlockwise
amount of circulating capital employed at time t and Lt is labour employed at as wt increases. The vertical distance between the curve Y(wt) and the 45°
time t (Wt = wtLt, where wt is the wage rate at time t, is therefore the wages curve indicates the surplus for each level of initial total capital. Since there is
capital anticipated to workers at time t). It is assumed that the fraction s of no rent, this surplus makes up net profits. Hence, curve s(Y(wt) – Xt) + Xt,
profits is saved. Time is discrete. With Xt representing the total capital giving the amount of savings out of profits plus the total capital, indicates the
employed at time t, if the wage rate is positive, then from the equilibrium capital stock that capitalists will employ at time t + 1. In the case illustrated
conditions Xt = Kt + wtLt, Yt = Kt/v = Lt one obtains Yt(wt) = Lt(wt) = Xt/(v+wt), in Figure 2.1, for example, for the level of total capital OX, the segment AB
Kt(wt)= vXt/(v+wt), Wt(wt) = wtXt/(v+wt). The first relation associates to each indicates the amount of net saving (i.e. net of total capital) necessary to
stock of total capital the demand of labour (or, equivalently, the level of ensure continuous full employment over time at the given wage rate (the
production), the second and the third associate to each level of total capital segment XB indicates the amount of total resources which must be saved in
order to ensure an increase in demand of labour equal to the increase in production techniques Mt which is stationary and is represented by a finite set
supply). This amount can be interpreted as the net demand for capital for of linear processes. Finally, Xt and Yt are vectors.
steady-state full-employment growth. The segment AC indicates the amount The equilibrium concept considered by von Neumann is the balanced
of net savings (i.e. net of total capital) available for employment in growth equilibrium in which the economy is growing at the maximum
productive activities. This magnitude can be interpreted as the net supply of technical rate. He provides sufficient conditions ensuring the existence of a
capital. For the sake of simplicity we assume that at time t there is full semi-positive equilibrium vector of activity levels, a semi-positive
employment of labour, i.e. Lt = Nt. In Figure 2.1, if total capital at time t is equilibrium price vector and a non-negative rate of interest which is equal to
OX, then at time t + 1 the demand of labour for full employment equilibrium the (maximum) rate of growth.
is given by the segment X"F while the supply of labour is X'E. Hence the The multisector von Neumann model can be considered the first complete
supply of capital is greater than the demand of capital for steady-state full- economic model in which the rate of growth is endogenously determined.
employment, wages will increase and also the rate of population growth will Some of the elements that characterize this model, as the assumption that all
increase. The former change yields a clockwise rotation of curve Y(wt) and, factors of production are producible, have been recently used in the modern
therefore, of curve s(Y(wt) – Xt) + Xt, and an anticlockwise rotation of curve endogenous growth literature to obtain persistent growth in absence of
X'(wt). The equilibrium between net demand and supply of capital for steady- technological change (e.g. Lucas, 1988; Jones and Manuelli, 1990; Rebelo,
state equilibrium will be reached when wages reach a value we such that 1991). This approach to generating endogenous growth will be discussed in
demand and supply of capital for steady state equalises: Section 2.5.
(1+ Gn(we)) = s/(v + we).
From this intuitive description, it can be seen that for classical economists 2.3. THE KEYNESIAN TRADITION
adjustment on the capital market between savings and investments is not the
same as in neoclassical economics, since it occurs mainly through 2.3.1. The Harrod–Domar Model
adjustments in the labour market. Moreover, unlike neoclassical growth
theory as we shall see, the rate of growth of the economy is determined by Harrod (1939, 1948) and Domar (1946) developed the first macroeconomic
the interplay between savings and population growth rate, the former being model to formally analyse the problem of growth. In so doing, particular
completely employed in investment and the latter being endogenously given attention is paid to make explicit the relationship between the consumption-
as an increasing function of the real wage rate. saving by households and the investment decision by entrepreneurs, although
these behaviours are not theoretically developed. In fact, the consumption-
2.2.2. The von Neumann Model saving decision is defined, following the Keynesian approach, by an
exogenously given propensity to consume, while the investment decision is
Von Neumann studied a multisector version of the classical Smithian– defined by the accelerator principle. In their model, production is obtained
Marxian model. From the formal point of view, this model is a multisector only by means of physical capital and labour.8 Given the usual Keynesian
linear model with only labour as the non-produced means of production and assumption of fixed prices, firms choose the best technique at the given
possibly with joint production. Von Neumann looks for an activity level prices. Thus generically there is only one cost-minimising technique, which
vector yielding the maximum rate of growth of the system and the associate implies that the capital/labour ratio and the capital/production ratio are
competitive price system. He deals with labour in exactly the same way as uniquely determined (By using both the normalisation and the notation of the
classical economists did, that is, labour does not appear explicitly in his previous section, we have Kt/Lt = Kt/Yt = v). Since Harrod and Domar,
model because it is ‘produced’ by linear technology by means of wage goods following Keynes, believe that the market mechanism is not able to attain
(Champernowne, 1945–46; see also Kurz and Salvadori, 1993). Hence full employment of labour, they focus only on the equilibrium of the goods
production is carried out by means of physical capital and wage good capital, market – which holds when savings are equal to the desired investment –
and the aim of the productive activity is exclusively the accumulation of rather than the general equilibrium on the goods and labour markets.
capital. The model is characterised by a zero value of rent and a set of An economy growing along a path with equilibrium on the goods market
is said to be on its warranted growth path. Along this path one obtains
GW = s/v where GW is called the warranted rate of growth of income, s is the labour at time t + 1, DE, greater than the supply of labour available at that
rate of saving and v is the capital/production ratio. The behavioural time, FG. Harrod and Domar, being concerned mainly with steady-state
hypothesis on producers and the Keynesian multiplier yield that the conditions, do not consider in detail what happens in this case, although it is
warranted growth path is unstable. If the warranted growth path ensures also suggested that an increase in wages and a subsequent inflationary process
the full employment of labour – a possibility which is just accidental in this could be generated.
model – the economy is said to be on the golden age growth path.
The decision structure of the model is the following. At time t the
economy is endowed with a technology Mt, which is assumed to be subject to
exogenous technical progress, a stock of capital Xt, and a given amount of
labour Nt. Entrepreneurs decide the level of investment at time t on the basis
of the acceleration principle, according to which the optimal level of
investment depends upon the expected desired level of capital employed at
time t + 1, Xte+1 , and the current level of capital stock. In equilibrium the
former must be equal to the actual level. By the multiplier principle, the
investment at time t determines the equilibrium level of production at this
time. Given the production function and the hypothesis on prices, the cost
minimising technique, Mt* , is chosen and the desired amounts of capital and
labour at time t are obtained, Kt /v = Lt = Yt. Assumed to be on a warranted
path, the current amount of capital Xt must be equal to the desired capital
stock Kt. In this model there is nothing ensuring the equilibrium on the labour
market, i.e. it is not necessary that Nt = Lt.
Aggregate income Yt determines the aggregate amount of consumption Ct
and hence the aggregate amount of saving St through the Keynesian
assumption of constant propensity to consume; i.e. Ct = (1 – s)Yt and St = sYt. Figure 2.2 – The accumulation process in the Harrod-Domar and Kaldor
Through the equilibrium condition on the goods market one obtains that models
sYt = It , i.e. the saving decision of households is coherent with the investment
decision of producers.
2.3.2. The Kaldor and Pasinetti Models
Figure 2.2 describes the process of accumulation in the Harrod–Domar
model. The level of production that ensures full employment of the capital
Kaldor (1954a, 1954b, 1961) holds that it is not saving, investment, technical
stock is described by curve Yt, while the saving function is function sYt. By
progress and population growth that are the causes of growth – these being
assuming that the decay rate of capital is δ, 0 ^ δ ^ 1, the curve δKt indicates
just features of growth – but the attitude of investing by society and in
the amount of resources which must be devoted to replacing the decayed
particular of entrepreneurs. In this he follows the Keynesian approach in
capital, while curve (n+ δ)Kt indicates the amount of resources necessary to
conceiving the expansion of the economy as driven by psychological and
ensure the reintegration of capital and an increase in the demand for labour
social factors like ‘human attitude to risk-taking and money-making’
equal to the rate of growth of population. According to Harrod and Domar’s
(Kaldor, 1954a, p. 228).
view, investment determines the level of income, Yt, which determines in
In seeking a growth theory which explains the real dynamics of
turn the net savings, i.e. the net supply of capital for full employment of
economies, Kaldor criticises Harrod–Domar’s model on the grounds that it
capital steady state, which is indicated by the segment AC. The current stock
explains only the growth of an acyclical economy with full employment of
of capital, which is optimal by the accelerator principle, OK, and the rate of
savings rather than the actual rate of growth of a system that does not
growth of population determine the net demand of capital for full
maintain a shifting equilibrium. Indeed, he held that in a system in which
employment of capital and labour steady state, indicated by segment AB. In
growth results from successive booms and slumps, the actual trend is
the case illustrated in Figure 2.2, the supply of capital yields a demand for
determined by the ‘natural rate’ of growth (Kaldor, 1954a). Because of the
sociological factors underlying the phenomenon of growth, he maintains, 2.4. NEOCLASSICAL GROWTH MODELS
following here the fundamental Schumpeterian intuition, that a satisfactory
growth theory cannot be constructed without a business cycle theory (Kaldor, 2.4.1. The Solow Model
1954a, 1954b, 1957). However, he never formally develops his position on
economic growth, and his major contribution consists in solving in an A different attempt to solve the stability problem of full-employment steady
original way the stability problem of the Harrod–Domar model. This is state is that of Solow (1956). He accomplishes this task by assuming a
accomplished by allowing the possibility that the economy can grow along a neoclassical production function that allows for flexible coefficients of
natural growth path through adjustments in the rate of saving due to changes production. By adopting a neoclassical framework, Solow changes the object
in the distributive shares between wages and profits and assuming that the of analysis with respect to the Keynesian growth theory; in his view the
population rate of growth is constant. The latter assumption is adopted only major problem is to construct a theory of general full-employment growth,
in his formal works, while in several non-technical articles (Kaldor, 1960) he and the most important concern is to ensure the convergence of the economy
and other post-Keynesian economists like Joan Robinson (Robinson, 1963) towards the natural growth path. Hence, growth theory has to explain the
share the classical view that the rate of growth of population depends upon potential growth of economies (Solow, 1999, 2000), without paying
the wage rate and, therefore, upon the rate of growth of the economy. attention, therefore, to cyclical trends of the economy and their possible
Kaldor’s contribution can be seen in Figure 2.2, where the usual effects on the long-run trend of the economy.
production function (Yt), supply of capital (sYt), demand of capital for steady Solow assumes that there is only capital and labour as factors of
state with full employment ((n + δ ) Kt) and demand of capital for production. The technology is represented by means of a neoclassical
reintegration (δ K t) and demand for labour curve (Lt) are depicted (we still production function with constant returns to scale, decreasing productivity
assume that the there is only one method of production). Note that now the with respect to physical capital and possibly labour-augmenting technical
average saving rate depends inversely upon the wage rate due to the relation progress. In order to construct a model that conciliates full employment of
s = s(wt) = swYt + (sp – sw)Πt (wt), resources with growth, Solow assumes that prices are flexible and therefore
all markets are cleared. In particular, the equilibrium on the capital market
where sp is the saving rate out of profits, sw the saving rate out of wages and yields that investments are equal to savings while the equilibrium on the
Πt (wt ) the profits (Kaldor, 1955–56), which are inversely related to the labour market yields that there is always full employment of labour.
wages. If the stock of capital is OK, then the supply of capital, KC, is higher Production is distributed between savings and consumption on the basis of
than the demand of capital ensuring full employment and constant a Keynesian saving rule. If savings are equal to the level of investment which
capital/labour ratio, KB. At time t + 1 the demand of labour, DE, is higher ensures the constancy of per capita capital with full employment, then the
than the supply of labour, FG, therefore wages increase. This yields an economy is in the steady state. Otherwise price adjustment on the capital
increase in s and a clockwise rotation of the curve s(wt )Yt, and Lt tends to Nt. market yields equalisation between savings and investments and appropriate
The equilibrium is attained when wt is such that s(wt )Yt = ( n + δ ) K t ; in this changes of the per capita capital until the steady state is attained. The
case, the supply and the demand of capital are equalised at the level which convergence process towards the steady state is ensured by the assumption of
ensures full employment. Although the equilibrating mechanism is described decreasing productivity of capital.
in terms of forces operating in the labour market, Kaldor (1961, pp. 196–7) At time t the economy is endowed with a technology set, a given amount
suggested that the process could equally take place in the goods market of capital and labour. Producers choose the best technique according to the
through changes in prices. ruling prices, and full employment of resources is ensured by their complete
Kaldor’s approach has been developed and pursued further by Pasinetti flexibility. The resulting income is either consumed or saved according to the
(1962) and by a fairly vast literature which, however, in emphasising the usual, exogenously given, propensity to save, s. Savings are completely
study of equilibrium paths have shifted attention from the original attempt to transformed into investments by the flexibility of the interest rate on capital
construct a growth theory out of a business cycle theory to the more market. Investment and the initial stock of capital determine the amount of
traditional view of constructing a theory of economic growth at the natural capital available at time t + 1. The technology set and the amount of labour
rate. available at this time are determined exogenously by the values of these
variables at time t, the former by assuming a positive rate of technical saving of households endogenous. The early models following this approach
progress, the latter by a positive rate of population growth. are characterised by a normative interpretation of the accumulation process
Figure 2.3 illustrates the accumulation process in the Solow model. In this in that the economic decisions concerning production and saving are taken
figure, all variables are expressed in per capita terms. Curve ( n + δ ) k by a planner choosing over an infinite horizon. An important peculiarity of
indicates the (per capita) demand of capital for a full-employment steady these early contributions is that the production side of the economy and the
state with constant per capita capital, while curve s f( k ) indicates the (per income employment side are not separated since a single agent, the planner,
capita) supply of capital. If the capital per capita is Ok', the supply of per has control both over production and saving decisions. The literature later
capita capital, k'A, exceeds the per capita demand of capital for a full- emphasised that the optimal path chosen by the planner coincides with the
employment steady state with constant per capita capital, k'B. Hence, in order path chosen by a perfectly competitive economy with many agents. This step
to maintain full employment on the labour market the rate of interest will is important in order to ensure that Ramsey’s approach is able to provide a
decrease and entrepreneurs find it profitable to increase the demand for positive theory of growth like the traditional models, such as Harrod–
capital. The price adjustment and the consequent demand increase is such as Domar’s and Solow’s. The decentralisation problem, moreover, opened the
to equalise demand with supply. At the new, higher level of capital per capita way to the construction of growth models in which the equilibrium path is
production will generate a new supply of capital and a new demand of capital analysed independently of the optimal, centralised path. This frees the theory
for a full-employment steady state with constant per capita capital will arise. from the hypothesis of perfectly competitive markets or from the assumption
If the former is equal to the latter, a steady state is attained, if the former is of the absence of external effects, and it is one of the most important
still greater than the latter a further decrease in the interest rate and of the per contributions of the recent endogenous growth theory.
capita capital occurs. And so on. In the traditional growth models à la Ramsey (Cass, 1965; Koopmans,
1965), the assumptions concerning the production function are the usual
neoclassical ones and the planner is endowed with a separable and stationary
utility function u(⋅) and a constant discount rate ρ. The optimal path, which
dynamically has the structure of a saddle point, is unique and converges
towards the steady-state path.
The basic structure of the normative version of Ramsey’s model can be
illustrated as follows. At time t the economy is endowed with a technology
set Mt and given amounts of capital Xt and labour Nt. The planner chooses the
best method of production Mt* and the entire amount of labour and capital to
produce the output, Yt. The planner then decides how to allocate the amount
Yt between consumption and savings which are immediately transformed into
investment, It. The stream of consumption levels (Ct )t∞=0 is chosen in such a
∞
way as to maximise the sum ∑ 0 ρ t u(Ct ) subject to the constraint
Xt+1 = Yt – (1 – δ ) Xt – Ct, where δ is the capital decay rate.9 Investment at
time t and the initial stock of capital Xt determine the stock of capital
available at time t + 1, Xt+1. The amount of labour available at time t + 1, Nt+1,
is determined by a demographic rule that is given exogenously.
Figure 2.3 – The Solow model in discrete time and without technical The accumulation process in Ramsey’s model is similar to that of Solow’s
progress described by Figure 2.3. Now, however, along the optimal path the rate of
saving changes over time and converges towards the long-run level
2.4.2. The Growth Model à la Ramsey associated with the steady state.
Inspired by the article by Ramsey (1928), several growth models have

been constructed in order to improve Solow’s model by making the rate of
2.5. ENDOGENOUS GROWTH MODELS reproducible factor by interpreting it as human capital. However, now an
additional decision has been taken by households concerning the amount of
The aim of the endogenous growth theory is twofold: first, to overcome the resources to employ in the accumulation of physical and human capital.
shortcomings of the Solow and Ramsey models which are unable to explain The following is a summary of the structure of the Lucas model. At time t
sustained growth, and second, to provide a rigorous model in which all the economy is endowed with a technical set for producing output Mt, a
variables which are crucial for growth, in particular savings, investment, and technical set for the production of human capital MH, given stocks of
technical knowledge, are the outcome of rational decisions. Hence the physical capital Xt and human capital Ht. Set MH is made up by only one
endogenous growth theory has adopted Ramsey’s model as a reference linear method of production and does not change over time. On the basis of
theoretical structure, in which saving is the outcome of a maximising agent the ruling prices, households choose the amount of human capital to employ
and the equilibrium growth path is seen as the consumption/saving trajectory in productive activities, uHt, and in the further accumulation of human capital
chosen by rational agents by solving an intertemporal optimisation problem. (1 – u)Ht. Firms, on the other hand, choose the best technique Mt* , the
As for the former, the endogenous growth literature points out that a amount of capital Kt and the amount of human capital to employ in
necessary condition for perpetual growth is that from the household’s point production. Flexibility of prices ensures full employment of factors.
of view the rate of interest should never be driven too low, and this is Households allocate the production Yt between consumption and savings
ensured if the productivity of accumulated factors does not decrease to zero according to their utility function, and savings are, as usual, transformed into
as accumulation proceeds (see for example, Jones and Manuelli, 1997). On investment through the capital market. At time t + 1 the economy will be
the contrary, if this case occurs, savings will be driven to a level that is not endowed with a technical set Mt+1, which is different from that at time t
enough for fuelling sustained growth. In this perspective, the main object of because of exogenous technical change and externalities from the
the endogenous growth theory has been to develop economically meaningful accumulation of human capital, a technical set MH, capital stock Xt+1
ways of ensuring non-decreasing returns to scale with respect to the determined by initial capital stock Xt and investment It, and, finally, a stock
accumulated factors. This has been accomplished either by removing the of human capital Ht+1 determined by investment (1 – u)Ht and the production
scarcity of natural resources or by introducing technical progress. As far as process in MH.
the former is concerned, for example, labour has been straightforwardly The accumulation process in the endogenous models considered here is
transformed into a fully reproducible resource, human capital. As for similar to that illustrated in Figure 2.2 for the Harrod–Domar model. Similar
technical progress, one of the main features of the endogenous growth theory to this model, along the optimal path the rate of saving is constant; however,
is the capacity of endogenise the investment decision yielding technological unlike the model, it is determined endogenously by the maximizing
progress which consists mainly in the introduction of new intermediate behaviour of the planner or of households. Moreover, the AK model, unlike
and/or final goods (Romer, 1990; Grossman and Helpman, 1991; Aghion and the Keynesian models, follows the classical and neoclassical tradition in
Howitt, 1992, 1998). conceiving investment as induced by savings, rather than the other way
round.
2.5.1. Accumulation of Physical and Human Capital
2.5.2. Technical Progress
The simplest endogenous growth model is the so-called AK model (see e.g.
Rebelo, 1991). It can be considered as a Ramsey model with the assumption In the Lucas model externalities are present but they are not essential to
that the production function is linear with respect to physical capital and that ensure sustained growth. An alternative approach to the accumulation of
scarce resources are not considered explicitly. Within the AK model it is factors to ensure sustained growth is the introduction of technical progress.
possible to obtain a constant positive rate of growth of the per capita Technical progress can be considered an improvement in technological
consumption along the optimal path. This path, moreover, can be rationalised knowledge incorporated in the new production function – usually due to
as the outcome of a decentralised perfectly competitive economy. unintentional effects like externalities.10 This class of models is important in
A second approach to obtain sustained growth is to introduce human the development of a growth theory since, because of externalities or explicit
capital (see e.g. Lucas, 1988). By following this approach, persistent growth introduction of imperfect competitive markets, it makes a substantial break
is obtained by transforming labour from a scarce resource into a fully with respect to the growth models à la Ramsey given that on the equilibrium
path there is no longer an optimal allocation of resources. This implies that producing ideas, obtains new ideas affecting the technology set for producing
analysis of the optimal path can no longer be obtained by means of the intermediate goods available in the next period. Intermediate goods are
normative approach but has to be carried out through a decentralised produced by employing the whole physical capital and the relevant
mechanism which requires a detailed description of the behaviour of agents technology. Intermediate goods, in turn, together with labour for the
and of the working of market mechanisms. production of final goods and the relevant technology, produce the final
The structure of the model with externalities is very similar to that of good. This is allocated by households into savings and consumption. The
Ramsey’s model, except that the existence is assumed of a finite number of former, through the capital market, are transformed completely into
firms whose production and accumulation decision affects positively and investment, which determines the stock of physical capital available in the
unintentionally the technology of all firms. This effect ensures the constancy next period, Xt+1. The production set for producing the final goods at time t is
of the productivity of the accumulated factors, thereby removing the assumed to be invariant over time, while the set for producing ideas at time
tendency towards a growth path in which the rate of growth is determined t + 1, MIt+1, is different from that at time t because of externalities from the
only by exogenous factors like population growth. existing amount of ideas. The latter is interpreted as learning by doing or as
Models with new products are usually three-sector models: a final good the public nature of researchers’ skills. Unlike externalities in Lucas’ model,
sector, an intermediate sector and an invention sector which produces here the existence of externalities is important in guaranteeing sustained
‘designs’ of new intermediate products. Competition is perfect in the final growth since it ensures the linearity of the production function in research for
and invention sectors, while that of intermediate goods is imperfectly new ideas.
competitive because there is a fixed cost associated to the purchase of new
designs. Imperfect competition in the intermediate sector is necessary to
ensure the existence of profit in this sector which, through fixed costs, will 2.6. CONCLUDING REMARKS
be transferred to the invention sector. These profits in turn are the incentive
for the invention sector to continue its activity. Such models endogenise the In this paper we surveyed in a simple and intuitive way different models for
production of new ideas and introduce the Schumpeterian idea that technical analysing the accumulation process conceived in various traditions from
progress is linked with imperfect competition. However, it is not possible to classical economics to the endogenous growth theory. Our main aim was to
consider this class of model a truly Schumpeterian one since the production highlight similarities and differences among these alternative theories. We
of new ideas is here conceived as a smooth process, while in Schumpeter pointed out that there is continuity from classical to endogenous growth
invention is strictly linked, as in Kaldor’s view, to business cycle. theory, partly through Keynesian theory (see also Hahn and Matthews, 1965,
In the basic model taken from Romer (1990), there exists only one final pp. 8–9), concerning the fact that the steady state is conceived as
good, which can be employed also as physical capital, infinite intermediate endogenously determined by the model. By contrast, neoclassical economists
goods and potentially producible ‘ideas’. Each intermediate good is produced see it as exogenously determined by factors considered outside the realm of
by means of physical capital and only one idea. The technology set for economic explanation. We also emphasised continuity between classical,
producing the consumption good is assumed to be stationary, while it is neoclassical and endogenous growth theory, as opposed to Keynesian theory,
assumed that the production sets for producing ideas and intermediate goods in terms of the saving/investment relationship. While the former theories
are made up by only one method of production. The method of production to conceive saving as wholly transformed into investment, and therefore,
produce ideas is linear and requires only labour, while the method of growth being determined by saving itself, Keynesian theory conceived
production of intermediate goods requires physical capital and ideas. It is investment as the source of growth and no relationship between the former
linear with respect to physical capital, although it has fixed costs due to the and the latter variable necessarily exists.
use of the corresponding idea. At time t the economy is endowed with
technology sets to produce new ‘ideas’, intermediate goods and the
consumption good indicated respectively by MIt, MINt and MC, a given stock NOTES
of physical capital Xt and a given amount of labour Nt. Households choose
* We would like to thank Heinz Kurz for valuable comments.
how much labour to allocate to produce consumption goods and to produce
1. Smith, for example, while allowing that wages and rents could be employed in savings,
ideas. The amount of research labour, together with the technology for maintains that capitalists are ‘naturally the most disposed to accumulate’ (WN IV.viii.61).
The structure of growth models: a comparative survey 41
2. For a more complete elaboration on this point see Rosenberg (1975).

3. See also Hollander (1973, p. 158) and Eltis (1984, p. 88).
4. This assumption does not imply that there is full employment of labour. An ‘industrial
reserve army’ is compatible with the case considered.
5. It seems reasonable to say that Marx considered explicitly investment to improve
technology. In the basic classical model here considered we shall neglect this fact and we
consider technical improvement due to non-deliberate economic decisions, like in Smith’s
3. Endogenous growth theory as a
view (see McNulty, 1968). Investment in R&D will be taken up in detail later, in dealing
with the endogenous growth theory.
Lakatosian case study
6. Smith holds that labour and capital have an increasing productivity (WN I.iii, I.viii.57, II)
and Marx seems to hold that labour and capital have constant productivity (Marx, 1956). Mario Pomini
7. Figure 2.2 does not appear to reflect some important features of the Marxian view of the
accumulation process; in particular, the full employment condition. However, this figure is
able to make the classical approach comparable to the more recent models as far as the
adjustments on the capital market is concerned. 3.1. INTRODUCTION: THE ISSUE
8. Harrod’s model is quite different from Domar’s. In what follow we shall refer to the
Harrod–Domar model because we consider only a simplified version of these models.
9. In the original version of the Ramsey (1928) model it was assumed that ρ = 1. The fact that
In the mid 1980s, growth theory experienced a remarkable revival and
future utilities were not discounted implied that many feasible paths had an infinite value. became once more a very active area of macroeconomic research. Starting
To compare these paths, Ramsey recast the maximisation problem as a minimisation from the seminal articles by Romer (1986) and Lucas (1988), the research
problem in which the objective was the sum of losses from a reference path (the Bliss point). took a precise direction in the sense that, in contrast to the earlier neo-
10. See e.g. Romer (1986), or the introduction of new products that can be either intermediate classical view, it called for an endogenous determination of technological
goods (Romer, 1990; Aghion and Howitt, 1992) or final goods (Grossman and Helpman,
1991).
change, which means an endogenous determination of the sources of growth.
This article will concentrate on the treatment of endogenous growth by
neo-classical growth theorists. It uses the Methodology of Scientific
Research Programmes (MSRP) proposed by Lakatos (1970) in order to
explain why the endogenous growth approach was not incorporated into the
neo-classical growth programme until the late 1980s, although the essential
features were well known during the 1960s. The resulting thesis is that the
new growth theory may be seen in terms of an extension of the neo-classical
research programme to incorporate theoretical elements which previously fell
beyond its scope.
MSRP was introduced into economics methodology by Latsis (1976) and
in the following decades was used intensively as a way to study the nature of
progress in many branches of economics.1 Even if the MSRP is not without
its critics among economic methodologists (for example Hands, 1990), it
remains a useful framework within which to analyse the evolution of
economic ideas.
3.2. THE DEVELOPMENT OF THE NEO-CLASSICAL

GROWTH RESEARCH PROGRAM
In this section and in the following, we will try to characterize neo-classical

growth theory in Lakatosian terms. The important organizing category in
42
Endogenous growth theory as a Lakatosian case study 43 44 The Theory of Economic Growth: a ‘Classical’ Perspective
Lakatos’ methodology is the scientific research programme, which is a more HC2) The dynamics of an economic system is determined by the
detailed version of Kuhn’s concept of paradigm. Specifically, it should be accumulation of the factors of production.
thought of a series of theories that make up a continuous whole because they HC3) The supply side of the economy is described by an aggregate
share some common elements. A research programme is made up essentially production function which allows complete substitutability of factors
of two elements: a hard core with its protective belt and its positive of production, which typically are labour and capital.
heuristics. The core hypotheses are viewed as unchangeable by those who HC4) The aggregate production function is characterised by constant returns
take part in the research programme and anomalies are incorporated into the to scale on the factors employed (the production function is
programme not by changing the indispensable hypotheses but rather by homogeneous of degree one).
modification of the auxiliary hypotheses, the initial conditions and the
observation set. These propositions are not the same in nature. The first two are wholly
Further, each research programme is characterised by a set of general conditions which can be found also in research programmes other
methodological rules: some of these (the positive heuristics) indicate what than neo-classical ones. HC1) was introduced by Harrod (1939) who was the
the researcher should do, whereas others (the negative heuristics) are first to pose the problem of formalising the way in which an economic
injunctions concerning what not to do. The negative heuristic has a purely system can reach a position of long run steady growth. For Harrod, the idea
protective role and aims to prevent the hard core’s propositions being of steady-state growth is nothing other than the adaptation to the dynamic
changed and tested. The positive heuristic outlines development directions of case of the notion of equilibrium which we find in statics, even if this move
the research programme and consists of a set of suggestions concerning how represented ‘a real revolution in mentality’ (Harrod, 1939, p. 15). Proposition
to change or modify some aspects of the research programme which may be HC2) goes even further back in the history of economic thought and
refuted (Lakatos, 1978, p. 110). translates the idea of the classical authors according to which economic
A research programme is not a static entity but evolves over time: new growth appears essentially as a circular process: in order for the economic
facts are discovered and new problems arise which require changes in the system to expand, part of the output must be saved and used to increase the
protective belt and the positive heuristic. It may be progressive or stock of factors of production, primarily, capital. Economic growth and
degenerating. If a research programme leads to the discovery of new facts factor accumulation are two processes which, from Adam Smith on, largely
which are successfully explained on the basis of the programme elements, coincide.
then there is a progressive problem shift. If, on the other hand, the Propositions HC3) and HC4) are typical aspects of the neo-classical
programme is running out of steam and the new hypotheses added are ad hoc approach. HC3) makes it possible to endogenise the capital–labour ratio,
hypotheses, and thus partial adjustments which do not increase the empirical which in Harrod was constant. Solow’s criticism of Harrod is that the latter
content of the programme, then the programme undergoes a degenerating chose to study long-run phenomena using short-run technical tools, such as
problem shift (Lakatos, 1978, p. 118). that which requires technical coefficients of production to be constant. HC4)
To illustrate the essential features of the neo-classical research programme comes from the theory of distribution and requires that output be shared
it will be easier to divide the exposition into two parts. In this section, we among the factors according to their marginal productivity. These two
will consider the short-run dynamic model. In the next section, we shall see propositions, on the basis of which the capital–labour ratio may vary and its
under which hypotheses the model can be extended to encompass long run variation is regulated by the quantity of available capital for the economy,
phenomena and obtain a theory of economic growth. enabled the young Solow to provide a brilliant solution to the problem left
The best illustration of the guidelines and the most important results of the unsolved by Harrod of instability on the economic growth path.
neoclassical growth programme remains Solow’s essay, Growth Theory: An The positive heuristic of the neo-classical research programme can be
Exposition (1970). From this it can be gleaned how the neo-classical hard- illustrated by the following propositions:
core2 organised around the following propositions:
PH1) Build models where the agents optimise.
HC1) Growth theory concerns itself with the conditions under which an PH2) Build models which allow predictions on the equilibrium states.
economy grows in steady-state conditions. PH3) Show up the logical supremacy of resource allocation over
distribution.
PH4) Do models where the long run dynamics is determined by the HC5) Technological progress tends to be a temporal exogenous trend which
accumulation of physical capital. increases output at a constant rate.
The first two propositions are typical of the neo-classical view, in the According to HC5), in the long run, the growth rate of the economy
broadest sense, and do not require any particular attention. EP3) expresses coincides with the growth rate of labour efficiency, to which must be added,
the fact that dynamic allocation of resources is determined purely by if necessary, the growth rate of the working population. The continuing
technology and by the initial stock of resources. Solow here shifts radically increase in labour efficiency creates new opportunities for profit which, in
from previous tradition which considered growth and supply as two equilibrium, exactly compensate its downward tendency following the
fundamentally inter-dependent processes and profit was the driving force in process of accumulation.
economic growth. EP4) is representative of the deeply-held conviction, From a methodological point of view it is not difficult to see how HC5) is
shared by those studying economic growth in the ‘50s and ‘60s, that a rather typical situation in which a theoretical model is adjusted to empirical
economic development was driven by the accumulation of capital and so by reality simply by the introduction of an ad hoc hypothesis. In the
industrialisation. This does not mean that the importance of other factors was methodological literature, the notion of ad hoc hypothesis refers to the
not recognised, but simply that they did not have any great weight in various stratagems used by researchers to introduce new assumptions solely
theoretical elaboration (Arndt, 1984). to save the model in the face of contrary empirical evidence. An hypothesis
is ad hoc when it cannot independently produce new predictions and hence
does not improve the empirical content of a theory. Following Popper, we
3.3. FROM THE DYNAMIC MODEL TO THE THEORY can say that the adjustment of a theory for the sole purpose of protecting it in
OF GROWTH the light of contrary evidence is not good scientific practice. The
modification introduced into the dynamic model with the addition of
It is Solow himself (1970, p. 34) who observed that the solution to the exogenous technological progress was ad hoc because it did not lead to any
problem of Harrodian instability, albeit an undoubted step ahead on the further predictions, backed up in turn by further checks and observations.
theoretical plane, does not provide a solution to Smith’s problem: to identify The empirical content of the theory was reduced rather than increased.
which factors determine growth in the long run. But there is also a further aspect beyond the typically Popperian, which
To move from the dynamic model to growth theory, as required by characterises an hypothesis as ad hoc and this aspect has been highlighted by
empirical evidence, we must add a further element to the model, so far economists. Lakatos claims that an hypothesis may be ad hoc not just
missing, which can support the dynamics over the long term. With his usual because it prevents the genuine falsification of a theory, but because it
clarity, Solow states that ‘there are two obvious candidates: technological conflicts with the programme’s heuristic and so weakens its internal
progress and increasing returns’ (1970, p. 34). Both exogenous technological coherence and unity. This is the case for theorists of rational expectations
progress and increasing returns to scale are tools which can offset the (Hands, 1988) who hold that other types of expectations have been damaged
consequences resulting from diminishing returns in the accumulated factor. by ad hoc assumptions, not so much because they are contrary to available
Exogenous technological progress obtains this effect via continuous evidence but rather that they are inconsistent with the principle of
innovation which postpones the production function and returns to scale via maximisation which represents the key assumption of neo-classical
those processes of cumulative causation which are linked to the increase in economics. The same criticism can be levelled at the hypothesis of
the size of the economy. In both cases, the result is an increase in per capita exogenous technological progress: it is an ad hoc hypothesis not because it
output and capital, with a single restriction that, to be compatible with the lacks the capacity to produce new empirical evidence but because it is not
steady-state condition, both returns to scale and technological progress must derived from the optimising behaviour of the individual agent and therefore
be introduced into the model in a very particular way. it is inconsistent in microeconomic terms.
As is well-known, Solow chose exogenous technological progress. So the It is a remarkable instance in the history of economic thought that a
list of hypotheses included in the hard-core must be supplemented by: research programme with such a flimsy empirical basis and supported by an
ad hoc assumption should have gained such an important position as to hold
the stage for two decades before being challenged by recent models. Once
again it is Solow who offers us the key to understanding the success of his The second argument advanced by Solow was the necessary conclusion of
approach to economic growth: the preceding one. The fact that technological progress is exogenous and in
particular takes the form of increased labour efficiency, makes it possible to
I shall concentrate on technological progress without taking returns to scale into
maintain all the model’s implications, from monotonous convergence to
account, for two reasons. In the first place, I reckon that technological progress
stability. As in the case of static equilibrium, also in dynamics the desired
must be the more important of the two [elements under consideration] in real
economy. It is difficult to believe that the US is enabled to increase output per results in terms of existence, uniqueness and stability of the equilibrium
man at something over 2 per cent a year by virtue of unexploited economies of solution are obtained.
scale. ... Second, it is possible to give theoretical reasons why technological
progress might be forced to assume a particular form required for the existence of
a steady state. They are excessively fancy reasons, not altogether believable. But 3.4. DEVELOPMENT OF THE NEO-CLASSICAL
that is more of a lead than we have on the side of increasing returns (Solow, 1970,
RESEARCH PROGRAMME
p. 43).
Here Solow clearly puts forward two different arguments, equally decisive For Lakatos a research programme is not a static entity. New facts are
for our ends. The first is a meta-analytical argument which can be defined as discovered, new problems emerge and consequently the protective belt
rhetorical. The second is the formal argument that shows how a mathematical undergoes some adjustment. Lakatos holds that a research programme should
structure may be found which is coherent with the basic viewpoint. be evaluated on the basis of its ability to evolve with time.
The argument whereby the technological progress hypothesis seems to In the 1960s the neo-classical research programme was enjoying most
Solow to have greater persuasive force must be qualified in that Solow is favour; new hypotheses were added to the base model which seemed to be
here referring not so much to growth theory as to the whole structure of neo- able to increase its ability to interpret and its analytical richness via
classical economics. What appeared convincing was the fact that the Solow subsequent progressive shifts. In economics, a useful criterion among the
parabola may be considered the extension to the dynamic case of the theory many available to establish whether a theory or a research programme is
of general economic equilibrium with all the supports of associated progressive consists in evaluating the relationship which exists at any
hypotheses, including the essential one regarding constant returns to scale. If moment between exogenous variables, that is, the inevitably arbitrary
the essential aim of growth theory was to offer a dynamic vision of the starting point of any research, and the endogenous variables, that is, the
Arrow–Debreu model, it follows naturally that the principle of increasing elements which the model sets out to explain. Generally speaking, we have a
returns was almost completely abandoned as it was incompatible with the moment of progress when the programme increases its scope and variables
theory of competitive equilibrium.3 which were originally considered exogenous are in a subsequent phase
So it was not empirical evidence or analysis of facts which indicated the endogenised.
research path to be followed but rather a theoretical vision in which the Seen from this standpoint, most of the theoretical research in the golden
economy tends naturally towards co-ordination of economic decisions age of the neo-classical programme concentrated on the attempt to make
through the price mechanism and is highly suspicious of intervention in the savings and technological progress endogenous. Endogenisation of savings
economy. All that could challenge such a perspective was set aside by was the work of a group of young economists of an analytical orientation,
methodological decision. This rhetorical vision which aimed to justify a among whom are Cass (1965), Uzawa (1965), Shell (1969), Arrow and Kurz
particular vision of the economic system hinging on free-play in the market (1970), Ryder (1967) who, in a relatively short time, successfully managed
was not without consequence for the development of the research to apply the techniques of dynamic optimisation. Progress was made at the
programme. Neo-classical theory was gradually consigned to irrelevance on level of mathematical formalisation but there was no equivalent advance in
the empirical plane and its position became even weaker when it was the analysis of the factors which determine economic growth, hence at an
strongly criticised and its theoretical foundation and internal coherence interpretative level. Retrospectively, Shell noted that ‘the success of the
challenged (Harris, 1980). On the other hand, the theory of economic Hamiltonian view in the analysis of economic growth has been rather
development as a separate discipline from growth theory was given a boost limited’ (Shell, 1987) since, even in the refined maximising version,
and great importance was held in it by all those elements which were ignored economic growth was completely independent of savings, or rather, of
by the rival model, starting with the active role of the state (Arndt, 1984). typical consumer preferences.
Let us now examine technological progress. As in the neo-classical view, research and innovation becomes an endogenous process. Within this
but not only, technological progress is the fundamental factor that determines approach we can distinguish three separate research paths, each of which has
economic growth, it comes as no surprise that the analysis of technological focused in different ways on knowledge as a factor which can be produced,
progress was the main area of research in growth modelling. In the literature owned and accumulated.4
of the 1960s, there is a clear awareness that technological progress was to a The first direction of research started with Arrow’s famous article in 1962
great extent an endogenous element but the question set was how to translate which was the fundamental contribution of the neo-classical approach to
this common conviction onto the analytical level. endogenous growth up to present-day models. The innovation Arrow
The endogenisation of technological progress followed basically two introduced was to hypothesise that labour efficiency was an increasing
paths. The first came directly from Kaldor’s observation (Kaldor, 1957) that function of work experience. As each work activity implies learning, workers
the means via which technological progress is realised is capital over time become more productive. In Arrow’s model, work experience is
accumulation. Kaldor’s idea that technological progress is incorporated in represented by the cumulative flow of new investment as it is in the
new capital goods underpinned the vintage approach in which capital was production of new capital goods that learning is shown. Arrow’s original
disaggregated by year of output. Within the neo-classical system this attempt idea was that learning-by-doing was a non-linear function of total
to link investment and technological progress took hold with an essay by investment.
Solow in which he developed the assumption that ‘technical innovation Arrow’s model represents the most important attempt to propose an
influences technological progress only if incorporated into new capital goods endogenous mechanism for growth in the neo-classical school. Although it
or via substitution of outdated equipment with the latest models’ (Solow, undoubtedly marked a step forward by introducing greater realism, it has two
1960, p. 200). Models with capital divided by year sparked off a lively body basic limitations which have reduced its theoretical use. In the first place,
of research (Wan, 1970) and constituted the dominant approach to the long-run growth is only partially endogenous due to the learning effect but
analysis of the relationship between technological progress and economic depends solely on working population dynamics as in Solow’s version. In the
growth. second place, its interpretative capacity is rather weak. The effect of learning
However, notwithstanding the various and fanciful ways of treating is not determined by an economic choice but is the involuntary by-product of
capital goods by years and the efforts made in this direction, in the long run the process of accumulation. With these two elements in mind, it is hardly
the growth rate ended up being constant over time and equal to the rate of surprising that Arrow’s approach has been considered an ingenious
productivity increase of the new investment. As Phelps observed, even sophistication of the neo-classical model but has not brought any substantial
according to the new view the ‘rate of development over the long-term changes to the prevailing paradigm.
depends on the rate of technological progress, not on the type of progress’ A second approach to endogenous growth was formulated by Uzawa
(Phelps, 1962, p. 256). In the models with capital incorporated, technological (1965). The novelty of Uzawa’s model lies in the fact that he explicitly
progress still ends up exogenous and in a steady-state. While the year-by- inserts, alongside the sector which produces the final good, a second sector
year approach had the merit of making the theory more realistic, it did not which is to produce new knowledge. Uzawa hypothesises that the efficiency
change the underlying view of economic growth. of this sector which produces new ideas is a concave increasing function of
the quota of labour employed in this sector. Thus the parameter which
measures work efficiency in the final goods sector, the usual parameter A,
3.5. THE OLD THEORY OF ENDOGENOUS GROWTH does not develop exogenously. The research sector employs labour to
produce new ideas which shift the production function of the final sector
The second road followed to endogenise technological progress focused on a upwards. How strong this form of technological progress will be depends on
closer consideration of knowledge as a factor of production. This was a the proportion of the labour force employed in the research and education
minor research direction but which fed an original research tradition. If sector. As the research sector requires only employment of labour, an
acquired knowledge becomes an essential factor of production, then its economy will grow at a higher rate if a higher proportion of the labour force
growth, and so technological progress, becomes dependent on the amount of is employed in the sector. Uzawa’s is the first model we have of endogenous
economic resources assigned to it and also on the way in which innovation growth in the modern sense, in which the driving sector is research and
spreads within the economic system. Economic growth backed up by accumulation of knowledge plays an essential role in long-run dynamics.
The third major contribution to the first wave of studies on endogenous 3.6. ELEMENTS OF THE NEW THEORY OF
growth was that of Shell. His approach is particularly relevant and is ENDOGENOUS GROWTH
expounded in a series of works which indicate a precise line of research
(Shell, 1966, 1967, 1976). The analytical viewpoint Shell adopts to The 1970s was a very difficult period for the neo-classical research
endogenise economic growth is very close to Uzawa’s in the sense that programme. The attempt to make the theoretical apparatus more realistic led
accumulation of knowledge, hence technological progress, is made to depend to an increase in its technical complexity, such as in the multi-sectoral
on the resources allocated to the research sector. The greater the resources models (Burmeister, 1980), which produced the opposite result. The core of
employed in the search for new ideas, the greater economic growth will be. the programme was no longer able to respond satisfactorily to theoretical and
However, what distinguishes the two models, and this is where Shell’s logical criticisms, starting with the adequacy of the notion of production
originality lies, is the assumption that the new ideas may be considered a function and capital as a factor of production (Harris, 1980). In addition,
factor of production in their own right. Shell develops an original form of the there was an external element: in the ‘70s it is macroeconomic theory which
function of production in which stock of knowledge is considered as a third abandons the study of long-run phenomena to concentrate on the short run
factor of production alongside labour and capital, Y = (F, K, A). Further, such as the theory of the economic cycle, stagflation, the impact of rational
Shell assumes that technology can be considered a public good, and this expectations and, more generally, the crisis of Keynesian orthodoxy.
assumption will play a central role also in the endogenous growth theory of The downturn in the neo-classical programme’s fortunes came with the
the 1980s. Shell argues: ‘Technical knowledge can be employed by any International Economic Association Congress in Jerusalem in 1973 which
economic agent without altering either its quantity or its quality. Thus, we was entirely given over to growth theory. As documented in Mirrlees’
must think of technical knowledge as a public good – primarily a public good introduction to a selection of the papers published the following year, this
in production’ (1974, p. 79). was a momentous occasion. Everyone who was actively involved in growth
If new ideas and patents are a non rival input, this requires that the theory took part, with Solow in the chair, and the range of discussion can be
hypothesis of returns to scale be abandoned and the aggregate production understood from the numerous sections into which the work was divided.
function show increasing returns to scale. Because of the presence of One section was dedicated to Growth and Technology and hinged on two
increasing returns it can no longer be hypothesised that firms operate in papers, one by Shell and another by Weizsacker, both dealing with the
perfect competition, for in this case the firms would take a loss. To obviate endogenous aspects of technological progress. In the following years, the
this problem, Shell adopts, like Arrow, the Marshallian hypothesis that interest in the neo-classical theory waned, at least in neoclassical circles and
returns to scale are external to the firm and under this condition there exists a this led Fisher to observe in a long and detailed review on developments in
unique and stable state equilibrium for capital (K) and technology (A). macroeconomics in the 1970s and ‘80s for the Economic Journal that ‘after a
Interestingly, also in Shell, as already in Arrow, increasing returns are not rapid development in the ‘50s and ‘60s, the theories of economic growth and
enough to guarantee a positive growth rate in the long term without capital have received little attention for almost two decades’ (Fisher, 1988,
exogenous technological progress. This happens because in both models p. 37).
returns to scale are increasing on the production function but decreasing on There has been a soaring revival of interest only recently. Endogenous
the real factor which is accumulated capital. This was a well-known result in growth theory is spreading and many mechanisms have been adopted by
the literature of the 1960s. With increasing returns to scale, the economy neo-classical authors to make growth an endogenous process. As it is not our
tends to experience explosive growth and thus it is impossible to find a stable aim to supply a review nor a classification but rather to analyse the evolution
equilibrium, but if, at the same time, marginal productivity of the of the idea of endogenous growth within the neo-classical research
accumulable factor is decreasing, then the system reaches equilibrium. But it programme, we shall limit ourselves to a consideration of the essential
is precisely this condition, which guarantees the steady state properties of the elements in Lucas’ and Romer’s approach, as their contributions remain
model, that prevents the system from showing endogenous growth. theoretical landmarks (Jones and Manuelli, 1997).
Stiglitz (1988) observed that all progress in growth theory must overcome
two types of obstacle. The first has to do with the mathematical constraints
which characterise every dynamic model, constraints which generally are not
to be found in static analysis. The second is of a different kind and calls into
question the view of economic processes underlying the analytical structure.

Each advance in economic theory requires that new intuitions be elaborated With economic growth now depending on the accumulation of ideas and no
which give an economic significance to advances on the analytical level. We longer on the accumulation of physical capital, the points in common
shall apply this interpretative key to the new growth theory as elaborated by between the old and new theories of endogenous growth are plain. The two
Romer and Lucas. approaches share the same plan even though the end results were to be very
As regards the analytical aspect, it is Lucas (1997) who clearly defines the different. What is clear in particular is that, albeit in a different form, both
essential ingredient in the new approach to economic growth models. To Romer and Lucas pick up again the concept of increasing returns to scale, or
obtain an endogenous growth model, in the present sense of the expression, rather, they give a different reply from Solow’s to the problem of inserting
the marginal product of the accumulated factor must be higher than the increasing returns into a general equilibrium model.
interest rate. If this is not the case, the sector which produces the Without going into the analytical details of the models, the basic idea
accumulable factor is not sufficiently productive to guarantee long-run behind Romer’s approach is outlined clearly and concisely in a short essay
product growth. As Lucas observes: ‘What lesson can we draw from the with the pertinent title: Are Non-Convexities Important to Understanding
failure of the neo-classical model? I think there are two. First, the villain is Growth? (1990). His thesis can be summarised in the following terms: if
the Law of Diminishing Returns. It is this feature that makes it hard to get technological progress consists in the accumulation of new ideas and
sustained growth in a model of a single economy … We have to find a way immaterial goods, then it is inevitable that we should turn to the principle of
to repeal this law, theoretically’ (Lucas, 1997, p. 68). increasing returns to scale. Knowledge in fact is a non-rival factor which has
In other words, returns on investment should not decline with a high production cost but it can be replicated without excessive cost. Here
accumulation but be independent of it and also sufficiently remunerative. we find Shell’s, and before him, Kaldor’s, idea that technology is a public
The crux of the problem, in an intertemporal context in which the usual good and that consequently the function of production is characterised by
representative agent maximises his utility, is that the conditions on increasing returns to scale. In Romer’s words:
technology, and thus on the supply side, must be defined appropriately so The oldest question in Economics is what causes growth. One of the oldest
that the return of the accumulated factor is not zero with the growth process. conjectures, built into Adam Smith’s story of the pin factory, is that non-
What Lucas calls for, and what the new models will supply in many ways, is convexities are important for growth […]. We now know how to fit this kind of
some modification of the traditional aggregate function of production which effect into an aggregate growth model, and we can already see that these models
generate many theoretical possibilities (Romer, 1990, p. 98).
moves in the direction required, which is that of contrasting the effect of
decreasing returns. So we must turn to the hypothesis of increasing, or at
But as we have already seen in Shell, technology as a public good cannot
least, non-decreasing, returns to scale, re-elaborated as necessary to satisfy
grant endogenous growth. Romer’s step forward consists in assuming that
the criterion of balanced growth.
increasing returns are due to an externality effect linked to the accumulation
If this is the common analytical skeleton of the new class of growth
of capital on the part of the single firm, that is, that it has the characteristics
models, there are a number of ways of reaching the required result according
of a public good but is produced privately. As new discoveries are non-
to the specifications of the way in which knowledge is used, produced and
excludable, the single firm produces knowledge from which all the other
distributed in the economy. Romer and Lucas, and later many others, could
firms can also benefit thanks to the circulation of information. This
not but return to reflect on the issue, but barely-sketched during the 1960s,
externality effect, practically irrelevant at a micro-economic level, becomes
concerning the concept of knowledge as an autonomous factor of production
the decisive factor on the macro-economic level and thus also for growth
which has particular characteristics that are totally different from other
theory. To ensure that the long-run growth rate is constant it is necessary to
traditional factors of production. To quote Lucas again and his criticism of
introduce an extremely particular condition. The aggregate effect due to
the Solow model:
capital accumulation must exactly balance the tendency to decreasing returns
The neo-classical model focuses on the capital accumulation decision, but it is which are found at micro level. Otherwise, it is impossible to achieve a
growth in ideas – not merely capital – that drives the system. This observation balanced growth situation and the economic system tends to explode. Here
suggests a shift of focus from decisions on capital accumulation to decisions that we find once again, albeit in a different form, the problem of instability
determine rate of production of ideas (Lucas, 1997, p. 68). which seems to characterise dynamic linear models from Harrod on.
Lucas’ model is less innovative than Romer’s on the interpretative level HC1) Growth theory concerns itself with the conditions under which an
but nonetheless has its roots in earlier endogenous growth literature. Lucas economy grows in steady-state conditions.
recognises his debt to Uzawa and presents his 1988 model as a variant of HC2) The dynamics of an economic system is determined by the
Uzawa’s (1965). In his words: ‘In 1965, Uzawa showed that a growth model accumulation of the factors of production.
based on human capital accumulation, without diminishing returns, can HC3) The supply side of the economy is described by an aggregate
produce sustained growth without the deus ex machina of exogenous production function which allows complete and immediate
technical change’ (Lucas, 1997, p. 68), Lucas gives Uzawa’s model a substitutability of the factors of production which are typically labour
microeconomic syntax which was totally missing from the original model. and capital.
The research sector is replaced by the notion of human capital, but, above all, HC4) The aggregate production function is characterised by constant returns
the accumulation of human capital depends linearly on the time which each to scale on the accumulated factor.
worker dedicates to study and training. With these modifications of HC5) Growth is determined by the accumulation of immaterial capital.
interpretation but with the same analytical structure, long-run growth rate per
man becomes endogenous in the sense that it depends on the fundamental That there should have been changes to the hard core of the research
parameters of the economic system, such as preferences and production programme begs the question whether there has been a progressive change in
function parameters. the research programme and therefore an internal adjustment, or whether we
To sum up, both Romer and Lucas completely re-orientate the neo- are faced with a new programme which relegates the Solovian model to the
classical growth theory, shifting the focus of the analysis from material attic.
resources to immaterial, picking up again the thread of the endogenous When Lakatos describes the creative shifts of a research programme it is
growth theory of the 1960s. Once on this road, the new growth theory also clear he is referring to changes in the positive heuristic while the hard core
found itself with the problem of how to incorporate increasing returns into remains intact. This could suggest that with the new endogenous growth
the theory of general economic equilibrium. Their reply was to completely theory we do not have a shift forward of the programme as in the case, for
set aside the problem of distribution, that is, the basic idea of the neo- example, of the endogenisation of savings in the 1960s or of the vintage
classical school that there is a correspondence between the quantity of a models, but a whole new growth research programme. It could be argued that
factor and its price, seeing as this was the obstacle which had to be the new growth theories have in fact created an alternative programme to the
overcome. dominant Solovian program.
But this rigid application of Lakatos’ approach would, however, lead us in
the wrong direction. We could hardly say that economists involved in the
endogenous growth research project are outside the neo-classical research
3.7. CHANGES TO THE RESEARCH PROGRAMME
programme, in the usual meaning of the term. The endogenous growth
approach shows a large degree of continuity in the neo-classical research
The new growth theory has had remarkable success in giving new energy to
programme and the greatest effort on the part of Lucas and his school is to
the neo-classical research programme, going back to the fundamental
prove the superiority of the neo-classical research programme over rival
question of the factors which determine economic growth and abandoning
programmes (Lucas, 1988).
the static vision of competitive economic equilibrium. The blossoming of the
This seeming paradox springs from the difficulty of precisely identifying
new models was made possible by a change which concerned the research
the elements of a research programme and can easily be overcome if we use
programme’s hard core. The principle of decreasing returns proved to be a
Remenyi’s suggestion, when discussing the application of Lakatos’
barrier to the understanding of growth. It was substituted by a new
methodology to economics (Remenyi, 1979), that the categories of a research
proposition which made it possible to view long-run growth as an
programme be made more flexible by introducing the idea of demi-cores. For
endogenous fact, that is, tied to the behaviour of economic agents. The core
Remenyi, each research programme generates in its development a series of
of the neo-classical research programme on growth now included the
specialties and sub-disciplines that have common features, each of which is
following propositions:
characterised by its own core, named demi-core. The protective belts of sub-
programs can overlap and, although their demi-cores may be distinct, they
share common elements mediated through the hard core. For Remenyi ‘the assuming that each producer benefits from the level of human capital.
demi-core is to the sub-discipline what the hard core is to the MSRP’ Besides the methodological implications that we analysed in the preceding
(Ibidem, p. 33). The important point is that the dynamics of a research section, the linearity in the production function produces some crucial
programme is determined by the evolution of the sub-programmes it can consequences that are worth considering.
generate which map out the heuristic path of the programme. He states: ‘It is The first is the troublesome result that a non-accumulable factor like
a fundamental result of the theory of core demi-core interaction that the labour does not have any role in production, for otherwise the growth rate
number of demi-core is not constant over time’ (Ibidem, p. 34), but the would depend upon the population level (scale effect), as with Arrow’s
heuristic of the core continually generates specialties and demi-core which model. This is quite evident in the simple version of this kind of model, the
testify to the vitality of a research programme. so called AK model (Rebelo, 1991), which starts from the premiss that there
This elaboration of Remenyi’s allows us to get out of the impasse into is a single factor of production, capital, and all the other factors are simply
which a rigid application of Lakatos’ methodology led us. The new growth eliminated by hypothesis. In turn, Romer and Lucas face this crucial problem
theory may be considered not so much a new research programme but rather only for very specific assumptions, arguing that the technological parameter
a new articulation of the neo-classical programme on economic growth depends on the economy’s average capital (human or physical) per worker,
which has led to the formation of a new demi-core capable of filling the gaps rather than aggregate capital stock, H or K. The upshot of this solution is the
in the previous one though belonging to the same research programme. totally unrealistic assumption that the production function no longer depends
Which element distinguishes the new demi-core from the previous one? on L, the labour force employed.
For the new growth theorists the answer is plain: the fundamental limit of the If it is hard to think of an economic process without non-accumulable
Solovian approach is to be found in the fact that it was lacking a micro- factors such as labour or natural resources, the difficulties for growth
economic theory of technological progress. The new theory of endogenous theorists are even tougher. One striking implication of linearity assumptions
growth may therefore be considered part of a more complex and ambitious is that labour’s share of income becomes asymptotically zero, which reopens
project carried on by the new neo-classical macroeconomics to rethink the issue of what factors govern income distribution (Bertola, 2000). It seems
macroeconomic analysis on the basis of the fundamental assumption that that the price to be paid in order to make the rate of growth endogenous is to
individual agents make optimal choices in markets which are linked to each abdicate to the traditional theory of income distribution in which relative
other and that these markets reach some sort of equilibrium. Equipped with factor prices reflect relative scarcity and the amount which each factor
tools from general equilibrium economics theory, growth economists tried to obtains from the national product is determined by technology and relative
solve the old problem of giving a serious microeconomic foundation to factor endowments.
macroeconomics. In the 1970s, the economists of the new classical The second interesting feature of NGMs is tied to the previous one in the
macroeconomics school developed a theory of the economic cycle as optimal sense that, if income distribution no longer depends upon the endowment
deviation of output around a trend; in the following decade they attempted to factor, then it is possible to consider profit as the relevant exogenous variable
explain the trend itself. As growth theory economists have often pointed out, that makes the micro foundations of the growth process possible. If this is
their contribution was that they successfully inserted the old idea of true, recent development can be characterized by a partial return to ideas that
increasing returns into a general economic equilibrium context. were prominent in growth theory prior to Solow’s model (Kurz and
Salvadori, 1998b, 1999). These models adopt a simplified version of the
classical notion of production as a circular flow and of profit as a surplus
3.8. FURTHER ASPECTS OF LINEAR GROWTH product. As in the classical tradition there is no limit to growth because it is
MODELS generated within the growth process and the growth rate is endogenously
determined assuming a relationship between the rate of growth and the rate
In the previous section we saw that in some of the recent NGMs the of profit. In the simple AK model, since the consumption good is produced
dominant idea is to drop the non-produced factors in the production function only by means of capital and the saving rate is constant, the growth rate is
so as to avoid any sources of decreasing returns assuming linearity in the simply the profit rate multiplied by the saving rate (Kurz and Salvadori,
production function. Romer obtains this result assuming that the 1998b, p. 76).
determinants of technological progress are non-rival factors and Lucas by
The genuinely novel element consists in the fact that profit is determined 3. The idea that the economic system is driven by exogenous factors precedes Solow’s
by technology alone and the growth rate of the system is then determined by contribution and it appears in several authors, see for instance Marshall and Cassel (Kurz
the saving–investment equation: the larger the propensity to accumulate, the and Salvadori, 1999, pp. 246–47)
4. For a more in-depth analysis of this point see Pomini (2000).
larger the growth rate. In the former approach, the endogenous aspect of
economic growth referred to various institutional, social and economic
mechanisms that were able to induce economic change. In NGMs these
mechanisms are generally based on some special technological relationship
placed in the equation that describes the accumulation of the relevant factor
in the long run.
3.9. CONCLUDING REMARKS
In this paper I have viewed the development of neoclassical growth theory

from a Lakatosian perspective in order to evaluate the relevance of
Lakatosian ideas for economics. The main conclusion that emerges is that the
rational reconstruction of neo-classical growth theory fits the facts.
In the first place, it has been possible to characterize neo-classical growth
theory as a genuine research programme based on capital accumulation and
exogenous progress. The evolution of this programme could be described in
terms of Lakatosian problem shifts that have concerned essentially the
question of making saving and technological progress endogenous variables.
Based on the key hypothesis of exogenous technical progress, the neo-
classical programme first flourished and later, in the 1970s, stagnated. The
current rebirth was determined by the return to earlier frameworks that were
primarily based on increasing returns to scale bound up with the process of
knowledge accumulation developed in the sixties. In Lakatosian terms, the
new models could be interpreted as a change that went beyond the impact on
the positive heuristic of the Solovian approach in that they effected a change
in the core propositions of the dominant programme. Thus, using Remenyi’s
suggestion, they resulted in the development of a new demi-core in the
neoclassical tradition that in certain ways mark a return to Harrod and to
traditions that go further back in the history of economic thought.
NOTES
1. See Backhouse (1998) for a discussion of the Lakatosian methodology applied to

economics.
2. Following Backhouse we consider only the propositions which are relevant to growth
theory. For a complete presentation see Remenyi (1979) and Weintraub (1985)
sector unbounded growth models have been provided by Dasgupta and Mitra
(1988), Dolmas (1996), Kaganovich (1998), Ossella (1999), and Freni,
Gozzi and Salvadori (2001, hereafter FGS). In FGS we studied a multi-sector
‘AK model’ in continuous time. We provided an existence result for optimal
strategies, a set of duality results, and a complete classification of the price-
4. Endogenous growth in a multi-sector supported steady states of the model. In doing so we used a number of
economy assumptions (for the full list of assumptions used in FGS see Section 4.3
below). A consequence of combining these assumptions with the lack of
primary resources that characterizes the ‘AK’ model was that for positive
Giuseppe Freni, Fausto Gozzi and Neri rates of growth (low discount rates) the structure of prices in the unique
Salvadori steady state of the model turned out to be invariant with respect to
preferences. Hence the independence of the interest rate from preferences
that holds in the one-sector ‘AK’ model carries over to the multi-sector
4.1. INTRODUCTION model we studied in FGS. But this result is not confirmed when negative
growth rates (i.e. high discount rates) are considered. On the contrary, we
There are at least three different approaches to endogenous growth (see Jones envisaged two other regimes, in one of which the prices in the steady state
and Manuelli, 1997). Two include non-convexities or externalities or both. depend dramatically on the discount rate. This proves that the introduction of
The third relies on convex models of growth in which, properly interpreted, a multiplicity of sectors entails problems which are not visible in a one-sector
the two welfare theorems hold (e.g., Jones and Manuelli, 1990 and Rebelo, model even if stability is not taken into account. However, the multi-sector
1991). The models in this last strand of literature are characterized by the fact linear growth model in FGS still has a very simple static structure and
that production is not limited by primary resources and hence the equilibrium predicts that the long-run rate of growth of an economy does not depend on
paths can show endogenous growth. For the sake of simplicity we call the initial conditions. Thus one could wonder if path dependence can be
‘convex models’ the models whose equilibria satisfy the conditions of the obtained in a linear endogenous growth framework or whether increasing
two Fundamental Theorems of Welfare Economics, and subdivide them into returns and/or imperfectly competitive markets are required to obtain it. In
‘bounded models’ (those whose feasible paths are limited by the availability the present paper we will discuss removing some of the assumptions made in
of natural resources) and ‘unbounded models’. In the last fifteen years, FGS in order to show that the structure of the steady-state set of the ‘AK’
models with explicit consumption and a production side in which ‘goods are model can be considerably enriched. Moreover, we provide several
made out of goods alone’1 have been widely used in the new growth theory, interpretative elements, relevant in dealing with convex unbounded models.
especially in that approach to endogenous growth based on the assumption The various building blocks of the model presented in FGS have different
that all production factors are reproducible (Lucas, 1988; Rebelo, 1991). It origins. In particular, the production side of the model has a clear classical
could be argued that any mechanism found in the literature to make sustained flavour (see Kurz and Salvadori, 1995) and is very close to the production
growth possible has essentially involved the assumption that there is a ‘core’ side of the von Neumann model, in which commodities are produced out of
of capital goods whose production does not require (either directly or each other because ‘[w]age costs are not considered as such, for laborers are
indirectly) non-producible factors. In fact the reduced form of most not separately considered any more than are farm animals’ (Champernowne,
endogenous growth models is linear or asymptotically linear in the 1945, p. 12). This structure hints at a ‘technological’ theory of the long-run
reproducible factors (see for example, Frankel, 1962; Romer, 1986, 1987, rate of interest, that ‘appears as the natural and optimum rate of organic
1990).2 Consequently, in the endogenous growth literature, static analysis is expansion of the system, and depends on the technical processes of
mainly centered around the concept of the ‘balanced growth path’, which, in production which are available’ (Ibidem). Nevertheless, the model has also a
this context, performs the role played by the stationary state in convex Ramsey-like preference side in which the optimal behaviour of the
bounded models. representative agent determines the system’s saving rate. Since we maintain
It is common in the literature on growth to study one- or two-sector the hypothesis that the behaviour of the representative agent does not affect
models. The exceptions are some convex models. Recent contributions to n- the technical conditions of production, our rational agent cannot be a worker,
61
Endogenous growth in a multi-sector economy 63 64 The Theory of Economic Growth: a ‘Classical’ Perspective
because the real wage is still ‘whatever is needed to persuade people to 4.2. A CLASS OF CONVEX MULTI-SECTOR
work’ (Ibidem, p. 16). Therefore, it is ‘[t]he question of consumption by the ENDOGENOUS GROWTH MODELS IN
propertied class’ (Ibidem) that properly arises in the model with an
CONTINUOUS TIME
intertemporal utility functional. However, even under this ‘representative
capitalist’ interpretation, the introduction of explicit consumption creates a
The commodity space is finite and the technology is stationary. Production
tension in the model as regards the forces determining the long-run rate of
consists in combining the productive services from the stocks to generate
interest.3 Both considerations of technology-and-cost arbitrage and
flows that add to the existing stocks. Decay and consumption, on the other
preferences concur in determining the growth rate, the profit rate and the
hand, drain away the stocks. The production set is generated by a finite
relative prices that prevail in the long-run equilibrium. Nevertheless, in the
number m, m _ 1, of independent activities (or processes), each of which can
model there is some space for the ‘classical’ opinion that ‘even if part of the
be run at any scale of operation. Hence, process j (j = 1, 2, ..., m) can be
income from property were spent on consumption, and not saved, the rate of
represented by a pair of n- dimensional input–output vectors:
interest would not necessarily be much affected: it might still be
approximately equal to the greatest expansion rate that would have been aTj → bTj ,
possible if all income from property had been saved’ (Ibidem).
Although the production framework used in FGS is the ‘simple linear where aTj ei ≥ 0 is the amount of productive services from stock i that
production model’ (Gale, 1960, p. 294) which excludes both joint production process j uses at the unitary level of activation and bTj ei ≥ 0 is the flow of
and choice of technique, the complete set of results we summarize in the commodity i produced by the same process at the same level of activation.
Classification Theorem provided in that paper is novel. The reason is that the Thus, we can summarize the production processes with a pair of mxn non-
study of multi-sector closed linear models of production with explicit negative matrices:4
consumption has been confined almost exclusively to the discrete-time A→ B.
framework (see McFadden, 1967; Atsumi, 1969 and the more recent
literature mentioned above), while we study a continuous-time model. As a For the sake of simplicity, we assume that the rate of decay of the stocks
consequence, some kinds of complications which in discrete time can be in production is given by a single constant δ x , δ x ≥ 0 . Stocks not used in
avoided at first (see however Atsumi, 1969, p. 270), arise from the production are ‘stored’. Stored commodities decay at the rate δ z , δ z ≥ 0 . To
beginning. These complications are generally connected with joint avoid jumps in the stocks we conceptualize ‘disposal’ as a storage process
production and decomposability, but in a continuous-time model they can with δ z ≥ δ x . Each stock can be produced and no productive service from
also be connected with the way fixed capital is formalized in order to avoid primary factors is relevant in the system. The way we handle storage (or
an infinite number of commodities (depreciation by evaporation). disposal) processes implies that we assume that, in addition to the m
The paper is organized as follows. In Section 4.2 we present the model in processes formally included in the technology, free disposal activities for the
a general format whereas in Section 4.3 we present the results obtained by productive services of the stocks are available.
FGS in a restrictive setting. Section 4.4 is devoted to a comparison with the We give a basic characterization of the technology by means of the
von Neumann–Sraffa–Morishima models developed mainly in the 1960s and following two classical von Neumann-like assumptions (Gale, 1960, p. 311):
1970s. Section 4.5 clarifies the extensions of the results by FGS which, on
the basis of the comparison, are expected to be easily obtained and what [HP1] Each column of matrix B is semipositive
should require more effort. It is also devoted to clarifying the differences in This assumption means that all commodities are reproducible and therefore
the analysis connected with relaxing some assumptions, allowing there is no primary factor.
decomposability of the input matrix, joint production, and multiple
consumption goods. Section 4.6 provides some conclusions. [HP2] Each row of matrix A is semipositive
This assumption means that no process can be activated without using (the
service of) some commodity as an input. It therefore implies that for each
t _0 the intensity levels of the production processes are bounded from above
by the existing stocks. It is convenient to assume that the system can grow at Note that if cm < n , then there are n − cm pure capital goods. Since the
a positive rate since this case is the most interesting one from an economic production side of the model is linear, the whole model is homogeneous if
point of view: the preference side is so. We therefore assume that the utility function is the
usual iso-elastic one:
[HP3] ∃ x ≥ 0, g > 0 : xT ⎡⎣ B − (δ x + g ) A⎤⎦ >= dT , where d is any non-negative
vector proportional to the vector of consumed commodities. 1
[HC1] u(⋅) = [v(⋅)]1−σ for σ > 0, σ ≠ 1 , u() = log[v(⋅)] for σ = 1 ;
FGS and Gozzi and Freni (2001) concentrate on the case in which there is no 1−σ
joint production in flow outputs. That is:
[HC2] v(⋅) : ℜc+m →ℜ+ is increasing, concave and homogeneous of degree
[FGS1] Each row of matrix B has one and only one positive element one. If cm > 1 , then v(⋅) is strictly concave.5
Note that in this case each positive element of B can be normalized to 1 Preferences are fully described by ρ, σ, and function v(⋅) . In the following
without loss of generality. Under Assumption [FGS1] processes can be we will say that v(⋅) describes the preferences concerning consumption at a
unambiguously linked to industries. If, moreover, each industry has one given moment in time, whereas ρ and σ describe preferences concerning
production process, then the input/output matrices are square and we will distribution of consumption (and saving) over time. Note that if cm = 1 , the
have: utility function v(c) := c is equivalent to any other. Hence, if 6
[FGS2] m = n (i. e. A is square and B = I)
[FGS5] cm = 1 ,
Another assumption about production we sometime use is:
then parameters ρ and σ completely describe the preference side of the
model.
[FGS3] xT ⎡⎣ B − (δ x + g ) A ⎤⎦ >
= 0T , g > −δ x , x ≥ 0, x ≠ 0 ⇒ xT B > 0T
Let s be the nx1 vector of stocks, x be the mx1 intensity vector and ĉ
It means that the services from each stock enter directly or indirectly into the be the n x 1 vector obtained from the consumption vector c by adding a zero
production of each commodity (see Kurz and Salvadori, 1998c, pp. 95–7). In component for each pure capital good. The evolution of the stocks is given
particular, if Assumptions [FGS1] and [FGS2] hold, [FGS3] is equivalent to by the following differential equation:
the assumption:
s T = xT B − δ x xT A − δ z (sT − xT A) − cˆ T
[FGS4] The square matrix A is indecomposable
with the constraints
As mentioned above, the way disposal is conceptualized implies the
assumption xT A ≤ sT , x ≥ 0, c ≥ 0.
[HP4] δz ≥ δx The common approach to the analysis of competitive equilibria in the
The discount rate, ρ ∈ℜ , and the instantaneous utility function, u, describe above setting is through the extension of the first and second welfare
the preference side of the economy because, as is quite common in the new theorems for finite dimensional economies. This strategy leads to investigate
growth theory, we are dealing with a single-consumer economy whose the link between the competitive equilibria of the system and the solutions, if
preferences can be represented by a utility functional U (c(t )) with the form there are any, to the problem:
∞ ∞
U (c(t )) = ∫ e− ρ t u(c(t ))dt , V (s* ) = sup ∫ e− ρ t u(c(t ))dt
0 0
where s T = xT B − δ x xT A − δ z (sT − xT A) − cˆ T (P)
u(⋅) : ℜc+m →ℜ ∪ {−∞}, 1 ≤ cm ≤ n . x A ≤ s , x ≥ 0, c ≥ 0, s(0) = s ≥ 0 given.

T T *
We will therefore be interested in the existence and characterization of the (see e.g. Bose, 1968; Weitzman, 1971; Jones and Manuelli, 1990; Rebelo,
paths for which the problem (P) has a solution. Moreover, since under the 1991). Second, models with adjustment costs (see e.g. Dolmas, 1996;
above assumptions (P) is a homogeneous program, our interest will lie also Ladron-de-Guevara, Ortigueira and Santos, 1999). Third, models in which
in the existence and characterization of the special paths that solve (P) and the technology is not polyhedral (e.g. Kaganovich, 1998, Jensen, 2000).
enjoy a steady state structure. These paths provide the simplest reference Finally, models with an infinite dimensional commodity space (Boldrin and
point for the analysis of the asymptotic behaviour of non-stationary paths. Levine, 2002).
In studying the optimal control problem (P), Hamiltonian formalism is
often used to introduce the ‘price’ variables. From an economic point of
view, this is a particularly significant procedure because it leads to the 4.3. STEADY STATES IN A ‘SIMPLE’ MULTI-SECTOR
introduction of competitive prices. Indeed, what are defined as competitive AK MODEL
paths are simply stock-price paths supporting the maximized Hamiltonian
(see Cass and Shell, 1976). In our context, since the optimal control problem In FGS Assumptions [HP1]–[HP3], [FGS1]–[FGS5], [HC1]–[HC2] hold.
(P) is autonomous, the discounted Hamiltonian is used. It is given by: Moreover, it is also assumed that all commodities are available at time 0 and
that the (unique) consumption good enters directly in its own production.
H D (s,v) = max ⎡⎣u(c) − cˆ T v ⎤⎦ − δ z sT v + max xT [B − (δ x − δ z )A]v . That is, if the consumption good is commodity 1,
c≥0 T T
x A≤ s
x≥ 0
[FGS6] s* > 0 ,
D
The linear programming problem involved in the definition of H (s,v) a11 > 0 .
[FGS7]
requires the existence of a vector q *(s,v) ∈ ℜn+ which is a solution to the
dual problem and can be interpreted as the vector of the equilibrium ‘rental’ The first of these two assumptions implies that there is an admissible
rates for the use of the stocks. Hence, if c * (v), x * (s,v) indicate a set of solution to problem (P) with a positive s for each t > 0. The second
controls solving the max problems involved in the definition of H D (s,v) , assumption guarantees that s > 0 for each t > 0 in each optimal solution
then the set of paths [s(t ),v(t ), c * (v(t )), x * (s(t ),v(t )), q * ( s(t ),v(t ))] ì, which starting at any s* > 0. This result is relevant since if s > 0, then q is bounded
satisfy appropriate continuity properties and solve and, therefore, prices v cannot jump. FGS first prove that an optimal path
exists if and only if
s ∈ ∂ HvC (⋅, ⋅) (1)
v − ρ v ∈ −∂ HsC (⋅, ⋅) (2)
[HE] σΓ > Γ − ρ
where
is called a competitive program, i.e. a critical point for the problem (P). A
transversality condition is then involved in the extension of the first and {
Γ = sup g : ∃ x ≥ 0 : xT ⎡⎣ B − (δ x + g ) A ⎤⎦ ≥ dT }
second welfare theorems to infinite horizon economies. The first welfare
theorem, in particular, will state that absolutely continuous competitive paths and e1 is the first unit-vector: a vector proportional to the vector of
with nonnegative prices that satisfy a suitable transversality condition are consumed commodities.7 Under the assumptions maintained by FGS it turns
optimal. It is well known, however, that a full converse of this result does not out that Γ = λPF −1
− δ x , where λPF is the Perron–Frobenius eigenvalue of
hold due to the possibility that absolutely continuity of prices cannot be matrix A.
granted for stock paths hitting the nonnegativity boundary. Moreover, prices The above existence theorem is completed with two theorems concerning
supporting non-interior stock paths can fail to exist altogether (examples of the optimality conditions for the problem at hand, that are the extensions to
this phenomenon are provided by FGS, Appendix D). the present framework of standard results holding in smooth-bounded
In closing this section, we should point out that there are some convex models. In particular it is proved (i) that a competitive program is optimal if
endogenous growth models that cannot be reduced at once to the present the usual condition that the value of the stocks converges to zero holds and
framework, a few of which are mentioned here. First, models with pure (ii) that for each optimal solution to problem (P) it is possible to find prices
consumption goods and/or non-reproducible resources that are not essential and rentals paths such that the optimal solution is also a competitive program
whose value of the stocks converges to zero. For formal statements and possible to deal with the ‘own rate of return of commodity i’ as the rate of
corresponding proofs see the original paper by FGS. profit which can be obtained by an investment measured in commodity i
As a step towards the study of the dynamics of the system, FGS analyzed getting a revenue measured in commodity i (see, for instance, Malinvaud,
the steady-state optimal solutions to problem (P), which are defined as 1953). FGS show that in any optimal solution9
optimal solutions [s(t ), cˆ (t )), xˆ (t )] to problem (P) for which there is a real
number g, a real number c0 , and a non-negative vector x 0 such that eTj v (t )
cˆ (t ) = c0 egt e1 , xˆ (t ) = x 0 egt , where e1 is the first unit vector. The definition of ri (t ) = ρ −
eTj v(t )
steady-state optimal solutions does not state that the supporting relative
prices are constant over time.8 However, FGS have proved that for each
where ri (t ) is the own rate of return of commodity i at time t. FGS also
steady-state solution there are a price path and a rental path such that
proved that in any optimal solution the growth rate of consumption equals
the ratio of the difference between the own rate of return of commodity 1 and
v (t ) = v(0)e− gσ t , q (t ) = ( ρ + δ z + gσ )v(0)e− gσ t .
ρ over σ :
c(t ) r1 (t ) − ρ
We will refer to these prices and rentals as steady-state price-rental paths. = . (4)
On the basis of these and other results, FGS reduce the problem of the c(t ) σ
optimal steady states to the analysis of finding scalars
−1 −1 In a steady-state solution supported by steady-state prices all the own rates
g ∈ [ σ1 (λPF − δ x − ρ ), λPF − δ x ) and c0 and vectors s0 , x 0 , and v 0 such that:
of return are equal to each other so we can call this common rate the ‘real
e1T v 0 = c0−σ (3a) rate of profit r'. Similarly, in a steady-state solution all the intensities of
operation of processes as well as consumption grow at the same rate so we
[I − ( ρ + δ x + gσ )A]v 0 ≤ 0 (3b) can call this common rate the ‘growth rate g'. Obviously, in a steady state
xT0 [I − ( ρ + δ x + gσ )A]v 0 = 0 (3c) r−ρ

g= . (5)
σ
x A−s ≤ 0
T
0
T
0 (3d)
From equations and inequalities (3) FGS obtained a Classification
xT0 [I − (g + δ x )A]v 0 = c0 e1T v 0 (3e)
Theorem, in which three different regimes are envisaged, depending on the
xT0 [I − (g + δ x )A] ≥ c0 e1T for g +δ z ≥ 0 (3f) values of the parameters involved. The Classification Theorem states a
particular relationship between the growth rate g and the rate of profit r. This
1 r – g relationship is drawn in Figure 4.1. In the first regime r is constant and
sT0 − {xT0 [I − (δ x −δ z )A]− c0 e1T }≥ 0 for g + δ z < 0 (3g) −1 −1
g +δ z equals λPF − δ x , whereas g varies in the range (−δ x , λPF − δ x ) . In the second
regime g is constant and equals −δ x , whereas r varies in the range
c0 >0 , v 0 ≥ 0 , x 0 ≥ 0 . (3h) [λPF−1
− δ x , a11−1 − δ x ] . In the third regime r is constant again and equals
−1
a11 − δ x , whereas g varies in the range (−∞, −δ x ) . This relationship is not to
Optimal steady state solutions with steady state support are relevant also be confused with another r–g relationship, that is, equation (5). The former
since in these states some relevant concepts, such as that of ‘real rate of depends on technology and on preferences concerning consumption at a
profit’ or ‘growth rate’, can be defined. The literature on growth often refers given moment in time and does not depend on preferences concerning
to steady states in order to convey some macroeconomic insights. Lucas distribution of consumption over time.10 The latter, by contrast, depends
(1988. p. 11), for instance, refers freely to steady state concepts as ‘the rate only on preferences concerning distribution of consumption over time. In a
of growth’ or the ‘real rate of profit’ under the explicit assumption of a fast steady state solution r and g are determined by the intersection between these
convergence to the steady state. two r – g relationships.
As a matter of fact there are profitability concepts which can be used with
reference to optimal solutions even if they are not steady states. It is always
g
4.4. A COMPARISON WITH THE VON NEUMANN–
SRAFFA–MORISHIMA MODELS
λ11−1 − δx
Let us consider the case of g + δ z ≥ 0 , taking account of the fact that in a
steady state equation (4) holds. Let us substitute r for ρ + gσ in inequality
(3b) and in equation (3c). Let us drop Assumptions [FGS1]–[FGS4] in order
to allow for choice of techniques and joint production and, therefore,
a11−1 − δ x substitute matrix B for I in inequalities (3b), (3g), (3h) and in equations (3c)
r
and (3f). Then we obtain from inequalities (3b), (3g), (3i) and equations (3c)
and (3f) exactly the model analyzed in the von Neumann–Sraffa–Morishima
− δx
literature developed in the 1960s and 1970s, with contributions up to now
(Kurz and Salvadori, 1995, summarize the whole approach):
Figure 4.1 [B − (r + δ x )A]v 0 ≤ 0 , xT0 [B − (r + δ x )A]v 0 = 0
The interpretation of the Classification Theorem is simple. If g > −δ x , then xT0 [B − (g + δ x )A] ≥ c0 e1T , xT0 [B − (g + δ x )A]v 0 = c0 e1T v 0
all commodities need to be produced and therefore n processes are to be
operated. Thus, the n equations relating prices and rate of profit relative to c0 > 0 , v 0 ≥ 0 , x 0 ≥ 0 .
the operated processes (the no arbitrage conditions) determine both the n – 1
relative prices and the real rate of profit. In this regime prices are The main difference consists in the fact that in the literature in question
proportional to v PF > 0 , that is the right eigenvector of matrix A there was almost always at least one primary factor called ‘labour’. In the
corresponding to λPF . If g < −δ x , the only process which is relevant is the cases in which no primary factor was taken into consideration, or the wage of
process producing commodity 1. Since the inputs used by this process are labourers was taken to be zero, an extra inequality was mentioned, which in
produced jointly by the process itself at a rate larger than the growth rate, all our case it is certainly satisfied since equation (3a) and inequality (3h) hold:
commodities used in the production of commodity 1 except commodity 1
xT0 Bv 0 > 0 .
itself have a zero price. In other words, production is reduced to the
production of commodity 1 by means of commodity 1 and free goods. The first problem to be analyzed is the following. The first regime
Hence, similar to the previous case the equation relating prices and the rate mentioned in the Classification Theorem is mentioned in the von Neumann–
of profit relative to the operated process producing commodity 1 can Sraffa–Morishima literature. Actually it is ‘the’ result that would have been
determine the rate of profit (apart from commodity 1, all commodities which predicted for an economy with stationary relative prices and a wage rate
are either produced or stored have a zero price). If g < −δ x , then once again equal to zero. But what about the other regimes? The literature in question
the only relevant process is that producing commodity 1, and the inputs used has rarely considered negative growth rates. Could one expect such a
by this process are produced jointly by the process itself. Yet this is realized difference?
at a rate equal to the growth rate and therefore these commodities (except The problem arises since for J < −δ [ the distinction between single
commodity 1) may have either a positive or a zero price. Those with a production and joint production is not relevant. This section is devoted to
positive price cannot be separately produced or stored; their existing stocks clarifying this point as a contribution to understanding the problems involved
can be regarded as stocks of ‘renewable’ resources for which a growth rate in multi-sector models. A graphic exposition in terms of two goods will be
of −δ x can be granted in the production of commodity 1. sufficient here. Consider a joint production process
D→E,
and let us represent it in a commodity space in which vectors a and b and combination of the operated processes meets the components of the
vector E − J + δ [ D appear. Figure 4.2 depicts four alternative vectors consumption vector with a positive price and overproduction of commodities
E − J + δ [ D , depending on the size of g: J
< −δ [ < J
< J
. In Figure 4.3, with a zero price.

instead, there are two single production processes, one producing commodity In a two-commodity economy the fact that the consumption vector is on
1 and one producing commodity 2. Also in this case in the figure we have the boundary of the cone means that there is a convex combination of
drawn alternative vectors E − J + δ D . It is immediately recognized that
L [ L
processes which can supply an amount of the required consumption without
when J + δ is positive, the cone consisting of convex combinations of
[
the condition that a commodity is overproduced, but this convex combination
vectors E − J + δ D (i = 1, 2) includes the positive orthant, whereas for
L [ L
is actually made up of only one process: no commodity needs to be
negative values of J + δ the same cone is included in the positive orthant.
[
overproduced, but the no arbitrage conditions determine a number of
Both processes are needed to produce the consumption vector at the required constraints lower than the number of the prices to be ascertained. Hence we
growth rate only if the consumption vector is internal to that cone. Let us have to study whether the fact that the consumption vector is on the cone
boundary is just by chance or whether there are economic forces at work to
2 2 impose this condition. In the former case prices are undetermined and vary in
a range. In the latter case the forces at work have an impact on prices which
b − (g'+ δx )a can be determined. The same argument is immediately applicable to an n-
commodity economy: once again the no-arbitrage conditions determine a
b 2 − (g'+ δx )a 2 number of relations among prices lower than that which would be necessary
a b2 to determine prices. Thus the economic forces which drive the consumption
vector to the boundary may determine the further constraints which are able
b 2 − (g''+ δx )a 2
b to complete the determination of prices. Otherwise prices are not fully
−1
b 2 − λ PFa 2
b1 − (g' + δ x)a1
determined and may vary in a range.
With these arguments in mind we can move on to analyze the last two
b − (g''+ δx )a regimes mentioned in the Classification Theorem. The third is clearly a case
b1
1 1 in which the consumption vector, which in our case is proportional to the
b − (g'''+ δx )a
first unit vector, is outside the cone: all commodities needed for the
b1 − (g'' + δx )a1
production of commodity 1 except commodity 1 itself are overproduced. The
−1
b1 − λPF a1 second regime mentioned in the Classification Theorem is clearly a case in
which the consumption vector is on the boundary of the cone. In this range
Figure 4.2 Figure 4.3 the no-arbitrage conditions are not able to fully determine prices; then in the
focus on the other cases: that in which the consumption vector is external to long run an increment in ρ pushes down the prices of the inputs of
the cone and that in which the consumption vector is on the boundary of the commodity 1 pushing the rate of profit r up in such a way that the increment
cone. in ρ is exactly compensated and the growth rate is unchanged. Note that if
In a two-commodity economy the fact that the consumption vector is the growth rate were pushed up, then the no-arbitrage conditions would
external to the above-mentioned cone means that no convex combination of impose prices proportional to Y ! , whereas if it were pushed down, then the
processes can supply an amount of the required consumption without no-arbitrage conditions would impose zero prices for all inputs of
overproduction of a commodity which, as a consequence, has a zero price. commodity 1 apart from commodity 1 itself.
Moreover, there is a problem of choice of technique even if the original
model contemplated a number of processes equal to the number of
commodities: as a matter of fact the model is equivalent to one in which the 4.5. GENERALIZATIONS
commodity with a price equal to zero does not exist, whereas the number of
processes is unchanged. The same argument is immediately applicable to an The arguments developed in the previous section suggest that
n-commodity economy: once again some prices need to be zero, a convex
• if there is joint production the second and third regime of the three regimes mentioned in the Classification Theorem, there is a fourth
Classification Theorem do not need to be connected with negative growth regime, in which commodity 2 behaves like a renewable resource which, if it
rates, is left to itself, grows at a rate higher than that of the consumption good. In
• if more than one commodity is consumed (in a single production setting), order to produce commodity 1, producers pick it just as fruits were collected
then in the Garden of Eden: at no cost. This example can also be used to show
- the first regime of the Classification Theorem may also hold for values what may happen when assumption [HP4] is dropped and, therefore, when
of the growth rate lower than −δ [ , storage is not part of the process of disposal but is effected in order to
- if the proportion in which commodities are consumed depends on preserve the commodities. This point is clarified at the end of Appendix C.
prices, then the second regime may determine a relationship between
the rate of profit and the growth rate which may be different from a
horizontal segment, 4.6. CONCLUDING REMARKS
- if there is continuous substitution in consumption, the third regime of
the Classification Theorem may not exist. This chapter has investigated a number of problems which are absent in any
single sector economy, but can be present in a multiple sector economy. This
Simple examples illustrating the above properties are easily constructed. In
has been done with the help of a generalization of the multi-sector ‘AK
the appendixes to this chapter we provide such examples. Appendix A
model’ in continuous time which we analyzed in a previous paper. This
presents two examples involving joint production. The first example shows
analysis has also shown how this model is connected to the von Neumann–
that the second regime of the Classification Theorem and part of the third can
Sraffa–Morishima linear models investigated in the sixties and seventies.
actually occur for positive growth rates. The second example shows that the
first regime does not need to exist when joint production is involved.
Appendix B presents a number of examples involving two consumption
commodities. These examples are related to the value assumed by a APPENDIX A
parameter. Three possibilities are envisaged. In all of them for high growth
rates (i.e. low discount rates) the rate of profit and the prices are determined Let (P) be the problem
as in the Classification Theorem studied in FGS. However for low growth ∞ c1−σ
rates (i.e. high discount rates), the relationship between prices and quantities V (s* ) = sup ∫ e− ρ t dt
0 1−σ
and growth and profit rates are very different from what was predicted by the
theorem. s T = xT B − δ x xT A − δ z (sT − xT A) − cT
In general, the arguments developed in the previous section suggest that xT A ≤ sT , x ≥ 0, c ≥ 0, s(0) = s* > 0 given.
many of the results obtained in long-period models of Classical inspiration where
like those of von Neumann and Sraffa can at least partly be imported in the
⎡1 1 ⎤ ⎡2 1⎤ ⎡1⎤
framework here presented. In particular the problem of choice of technique 11 A=⎢ ⎥ , B=⎢ ⎥ and c = c ⎢ ⎥ .
and that of joint production appear to be easily handled. Similarly, some ⎣1 0 ⎦ ⎣1 2⎦ ⎣0 ⎦
difficulties recognized in those models should have corresponding
It is easily checked that the steady-state optimal solutions can be
difficulties here. In particular we know that dropping Assumption [FGS4], or
represented in the same three regimes referred to in the Classification
its general form [FGS3], may lead to difficulties. The analog of these
Theorem. If δ [ < , the second regime of the Classification Theorem and
difficulties in the present framework can be illustrated with an example. In
part of the third occur for positive values of the growth rate.
Appendix C we present an example with two commodities: commodity 2
Let (P) be the same problem above, except that the input–output–matrices
enters directly into the production of both commodities, whereas commodity
are now given by
1 (which is the only commodity to be consumed) enters directly only into its
own production. This simple model is analyzed to illustrate the difficulties
that a decomposable matrix A can generate.12 In this example, besides the
⎡1 1
⎤ ⎡ 2 2⎤ 8(4 − 3α + α g) 1 + 6α + g − 2α g
A = ⎢4 2
⎥ and B = ⎢ x1 = c1 , x2 = c1 .
1 ⎥. (1 − α )(27 − 6g − g )
2
(1 − α )(27 − 6g − g2 )
⎣0 2 ⎦ ⎣ 0 1⎦
It is easily verified that the steady-state optimal solutions can be There are two critical values for the parameter α : 1/3, and 1/2. For
represented by two of the three regimes referred to in the Classification 0 < α < 1/3 vector x is positive if and only if g < [(3α − 4) α ] or
Theorem. The second process is inefficient with respect to the first process [(1 + 6α ) (2α − 1)] < g < 3 . For α = 1/3 vector x is positive for g < 3. For
and therefore it is never operated. Hence the first regime of the Classification 1/3 < α < 1/2 vector x is positive if and only if g < [(1 + 6α ) (2α − 1)] or
Theorem cannot exist. [(3α − 4) α ] < g < 3 . For 1/2 ^ α < 1 vector x is positive if and only if
[(3α − 4) α ] < g < 3 . If only the first process is operated, then
APPENDIX B v1 2(1 + r ) c1 g−7
= , = , 3 < r < 7 and g < –1.
v2 7−r c2 2(1 + g)
Let (P) be the problem
∞
V (s* ) = sup ∫ e− ρ t log(c11−α cα2 )dt Hence g = [7 + (8α − 1)r (8α − 7) + r ] and α < 1/2. If only the second
0 process is operated, then
s = x − s − c
v1 8 c (1 + g)
x A ≤ s , x ≥ 0, c ≥ 0, s(0) = s* ≥ 0 given
T T
= , 1 =− , r > 3 and g ^ –1.
v2 1 + r c2 8
⎡1 1
⎤ ⎡ c1 ⎤
where A = ⎢ 81 ⎥ and c = ⎢ ⎥ . Hence g = − [1 + (1 − α )r α ] . Figures 4.4 provide the relationship between
4
⎣8 0⎦ ⎣c2 ⎦ g and r for α ‘s in one of the three relevant ranges: 0 < α < 1/3,
1/3 < α < 1/2, 1/2 ^ α < 1, respectively.
If α = 0 or α = 1, there is a single consumption commodity. Hence, let 0 <
α < 1. The instantaneous Cobb–Douglas utility function determines g g
consumption share, in value, as constant and depending only on α : along

any optimal path 3 3
c1v1 1 − α
= r r
c2 v2 α
1 + 6α
2α − 1 3α − 4
and, as a consequence, both prices and both consumption levels are positive. α
Hence in a steady-state optimal solution
c1 = x1 − (1 + g)s1 > 0, c2 = x2 − (1 + g)s2 > 0 1 + 6α

3α − 4 2 α− 1
α
Let us partition all possible cases on the basis of operated processes. If both
processes are operated, then
v1 c1 1 − α
r = 3, =2 , = ,
v2 c2 2α
Figure 4.4a Figure 4.4b

g g
−1
3 a22 − δx
3α − 4 −1
a11 − δx
α r
− δx
Figures 4.4c
Figure 4.5
APPENDIX C
Let (P) be the problem

NOTES
∞ c1−σ
V (s ) = sup ∫ e
* − ρt
dt
0 1−σ 1. Champernowne (1945, p. 12) used this expression for the von Neumann (1945) model. In
s T = xT I − δ x xT A − δ z (sT − xT A) − cT the following we will borrow other expressions from this paper by Champernowne to
emphasize some similarities with the von Neumann model and the literature devoted to it.
xT A ≤ sT , x ≥ 0, c ≥ 0, s(0) = s* > 0 given 2. Some theorists even came to the conclusion that unbounded growth is more an assumption
about the linearity of the technology than a result of the models; see for instance Romer
where (1990, p. S84).
⎡a a12 ⎤ ⎡1⎤ 3. This is not the usual interpretation found in the recent literature. On the contrary it has been
A = ⎢ 11 , a11 < a22 and c = c ⎢ ⎥ .
a22 ⎥⎦
suggested that the AK model ‘becomes more plausible if we think of K in a broad sense to
⎣0 ⎣0 ⎦ include human capital’ (Barro and Sala-i-Martin, 1995, p. 39). Such two interpretations are
not so different as they seem at first sight. We will come back on this in footnote 4. It also
It is easily verified that the steady state optimal solutions can be possible to assume that the necessary subsistence of workers (or the inputs to produce
represented in four regimes which are depicted in Figure 4.5. Three regimes human capital) are included in the A of the AK model, whereas another part of wages, which
are the same mentioned in the Classification Theorem. The fourth regime is exceeds the necessaries, are subject to the choice of the ‘representative agent’, who, in this
characterized by r = a11−1 − δ x and −δ x < g < a22 −1
− δ x . In this regime case, does not need to be anymore a ‘representative capitalist’.
4. In this description of technology labour was not explicitly considered. On the contrary a
commodity 2 is (over)produced and its price equals 0. The growth rate process can be represented as
−1
cannot be equal or larger than a22 − δ x because otherwise the intensity of
operation of the process producing commodity 1 would be nought or D 7M ⊕ O M → E7M
,
negative (and therefore consumption would be nought or negative). where O refers to labour input. Then there are two ways to obtain the simbolism used in the
Process 2 in the example can be interpreted as a storage process for M
text, in the assumption that the real wage rate per unit of labour is defined by the vector w.
commodity 2 with a negative rate of decay. The existence of the fourth Let A^ be the usual material input matrix used in input–output analysis and let l be the input
regime, however, does not depend on the sign of the rate of decay, but on the vector of (simple) labour. Then the matrix A in the text can be seen either as A = A ˆ + lwT
−1
fact that a22 − δ x > −δ x . It is clear therefore that, even if matrix A is or as
indecomposable, something similar to the fourth regime in this example ⎡ $× O ⎤
$=
comes into existence whenever assumption [HP4] does not hold. ⎢⎣ Z ⎥⎦ .
In the latter alternative the last process referes to production’ of ‘labour’ or ‘human capital’.
Endogenous growth in a multi-sector economy 81
5. The use of iso-elastic utility functions goes back to Ramsey (1928) who studied this
‘interesting special case’ in section II of his 1928 paper.
6. Cases with multiple consumption goods are considered by Gozzi and Freni (2001).
7. We note that the utility function u is unbounded above or below or, in the log case, both
above and below. Moreover, the boundedness of feasible paths is not assumed. This
generates a non-trivial existence problem for (P) which is usually solved by the introduction
of an existence condition linking the technology with the preferences (see for example
5. Income distribution and consumption
8.
McFadden, 1967).
An example can clarify the issue. Let
patterns in a ‘classical’ growth model
⎡

⎤ ⎡ ⎤ Davide Fiaschi and Rodolfo Signorino
$ = ⎢ ⎥, ⎢ ⎥
V =
ρ = , X (F (W )) = F , % = , , δ [ = δ ] = ,

cˆ = c e1 and
⎢

⎥ ⎢ ⎥
⎣ ⎦ ⎢⎣ ⎥⎦
5.1. INTRODUCTION
in the (P) problem. It is easily checked that
Historians of the Industrial Revolution have not failed to study the role
⎡⎤ ⎡ ⎤ ⎡ ⎤ ⎡⎤ ⎡ ⎤ played by demand factors in the process of industrialization of an agricultural
Y = ⎢⎥ H
⎢ ⎥H
+ K ⎢− ⎥ H T = ⎢⎥ H + K ⎢− ⎥ H
W − W − W W W − W − W
V = V H ,
, F = H , [ =
,
economy. Landes (1969, ch. II) emphasizes the relation between income
⎢⎥ ⎢ ⎥ ⎢ ⎥ ⎢⎥ ⎢ ⎥
⎣⎦ ⎣ ⎦ ⎢⎣ ⎥⎦ ⎣⎦ ⎣ ⎦ distribution, consumption patterns and the growth of manufactures in the
eighteenth century England. In his view, the middle classes flourished thanks
for –1/2 < h < 1/2 is a steady-state optimal solution to problem (P) supported by a to favourable income and wealth distribution. The typical consumption
competitive rental-price path, so that despite the fact that the quantity side grows at rate 1/3, pattern of these classes consisted of commodities manufactured using mass
the relative prices and relative rentals do not need to be constant (they are so if and only if
h = 0).
production techniques with a high capital/labour ratio. Moreover, English
9. This formula is clearly a reminiscence of Fisher formula, when the own rate of return is farmers were used to eating a superior kind of food, such as white bread, and
interpreted as a ‘real’ rate of profit, the discount rate as a ‘nominal’ rate, and the fraction in to spending a smaller share of their income on food than their Continental
the RHS as an inflation rate counterparts. Thus English farmers had more money to spend on non-
10. Evans et al. (1998) refer to this relationship as a ‘technological’ relationship whereas they agricultural commodities. Such a consumption pattern is, for Landes, one of
refer to equation (5) as a ‘preferences’ relationship. However, as shown in the example of
Appendix C, the former relationship depends not only on technology, but also on
the key elements which favoured the industrialization of the English
preferences concerning consumption at a given moment in time. economy.
11. It is easily checked that the Classification Theorem can be generalized to allow the existence Landes’ point of view is all the more intriguing when confronted with that
of several processes for the production of each commodity (even a continuous number). If endorsed by some leading British classical economists who were direct
Assumptions [FGS1], [FGS3], [FGS5], [FGS6] hold and if it is further assumed that each witnesses of the historical facts that he studied many decades later. We refer
process producing commodity 1 uses commmodity 1 as an input, the Classification Theorem
holds with the following differences. In the first regime the rate of profit is λ − δ ,
in particular to Adam Smith and Thomas Robert Malthus (see Rosenberg,
1968; Brewer, 1998; and Marshall, 2000). In Book III of The Wealth of
[
where λ is the minimum of all the Perron–Frobenius eigenvalues of possible input square
matrices which can be obtained by peaking up for each commodity a process producing it; in Nations (hereafter WN), Smith reconstructs the progress of wealth in Europe
−
the third regime the rate of profit is D − δ , where a is the minimum of all coefficients
[ from the fall of the Roman Empire and highlights the role played by
relative to inputs of commodity 1 used in the production of commodity 1. landowners’ consumption patterns. In a stationary agricultural economy an
12. The case of decomposable matrices is not uncommon in the new growth literature: the
model by Lucas (1988), for instance, is of this type (for a similar remark, see also
increasing taste for ‘luxuries’, usually imported from abroad, provides
McKenzie, 1998, p. 11). landowners with a powerful stimulus to modify their routine economic
behaviour. Landowners are willing to change the lease conditions to their
tenants in order to allow the latter to implement more efficient agricultural
techniques and, consequently, to pay higher rents. Thus productivity in the
key sector of the economy, agriculture, increases. Landowners’ increasing
82
Income distribution and consumption patterns in a ‘classical’ growth model 83 84 The Theory of Economic Growth: a ‘Classical’ Perspective
expenditure on luxuries makes domestic production of these commodities The basic aim of our paper is to present a model in which the take-off of
profitable. According to Smith, in fact, ‘finer manufactures’ were introduced an agricultural economy as well as the long-run growth of an industrialized
into agricultural economies either through the gradual refinement of economy depends on income distribution and consumption patterns. Our
domestic primitive manufactures or through the imitation of foreign model is an extension of Murphy et al. (1989), up till now unduly neglected
manufactures (see infra Section 5.2). in contemporary literature. The main differences (crucial to our findings) are
Malthus’ analysis of the process of growth in an industrialized economy is the following: (i) agricultural productivity is a function of land distribution
found in Book II of his Principles of Political Economy (hereafter PPE). and of the availability of industrial goods; (ii) workers’ population is
According to Malthus, the basic obstacle which may slow down growth in an endogenous; and (iii) wage-earners may consume both agricultural goods
industrialized economy is the lack of ‘an adequate stimulus to the continued and industrial goods.
increase of wealth’ (Malthus, 1986, p. 288). This stimulus consists in an Besides economic historians and classical economists, another important
adequate level of ‘effectual demand’ mainly determined by income source of inspiration for our paper has been the work of Luigi Pasinetti
distribution and the structure of property rights. For Malthus a wide class of (1981, 1993) on structural economic dynamics. To put it in a nutshell,
relatively well-off farmers is able to generate a level of expenditure much according to Pasinetti, economic growth implies structural dynamics:
higher than that generated by few large landowners (when land ownership is demand composition greatly changes as GDP increases (the so-called
too highly concentrated) or by a multitude of poor peasants (when land Engel’s Law) and technical progress seldom displays the same rate in all
ownership is too fragmented). By the same token, the extent of internal and productive sectors. The economic structure of growing economies is thus
external trade and the level of consumption from Smithian ‘unproductive bound to change. Hence, to be empirically relevant, multisectoral models of
labourers’ need not be too low for effectual demand and industrial economic growth should dispose of the assumption of proportional growth.
production to grow pari passu (see infra Section 5.3). Moreover, as real Moreover, Pasinetti depicts technical progress and the evolution of the
wages increase, workers may develop a taste for manufactured commodities, patterns of demand, the two great dynamic forces driving growth in real
usually referred to as ‘conveniences’ or ‘comforts’, which may induce them world economies, as two intertwined phenomena:
to control their fertility and resist the temptation of ‘indolence or love of
ease’. As Gilbert points out: Since increases in per capita income necessarily imply non-proportional expansion
of demand, and since technical progress leads to increases in per capita incomes,
When men are seen exercising a free choice not to marry at the first opportunity, it the introduction of technical progress in any dynamic economic investigation
becomes more difficult to view them crudely as mere food-consumers and necessarily implies a non-proportional expansion of demand (Pasinetti, 1981,
children-producers (Gilbert, 1980, p. 90). p. 70).
Conversely,
In this scenario, workers’ rational decisions concerning their fertility and
consumption basket may become a crucial variable affecting the long-run
No commodity, whatever ingenious technique it may require, can be successfully
growth performance of an industrialized economy. By contrast, analysis of produced if its (real or imagined) utility for the consumers is not sufficient to
demand factors and, in particular, analysis of the relationship between justify its cost: it would remain unsold. The relevance itself of technical progress
income distribution, consumption patterns and growth is not high on the depends on potential demand: an increase of productivity, however large it may
contemporary research agenda. Among the few exceptions it is possible to be, loses much or even all of its meaning, if it takes place in the productive
mention Laitner (2000), Zweimuller (2000) and Kongsamut et al. (2001). process of a commodity for which demand can only be small or negligible. This
Laitner analyzes the structural change of an economy with two goods and means that any investigation into technical progress must necessarily imply some
non-homothetic preferences, focusing on the wealth effects involved by such hypotheses … on the evolution of consumers’ preferences as income increases
a transition. Zweimuller is interested in the relationship between demand (Ibidem, pp. 68–9).
composition and innovation and investigates the properties of the balanced
growth path. Finally, Kongsamut et al. study a multisectoral economy which Pasinetti focuses his analysis on the process of growth of industrial
shows structural change and whose income grows at a constant rate. economies. We make use of his intuitions on the importance of Engel’s Law
and increases of productivity also to study the development process of an
agricultural economy.
The paper has seven sections. Sections 5.2 and 5.3 briefly recall Smith’s provides the incentives which are the key to economic development (Brewer,
and Malthus’ points of view on the relation between property rights 1998, p. 81).
distribution, consumption expenditure, development and growth. Section 5.4
A widespread taste for finer commodities, in fact, creates incentives for
outlines the formal model whose equilibrium conditions are investigated in
landowners to modify dramatically their overall economic behaviour. On the
Section 5.5. In Section 5.6 we use our model to formalize three interesting
one hand, landowners are led to cut their expenditure on personal services
stages in the development process of an economy. Finally, Section 5.7 draws
and hospitality and to increase their expenditure on commodities; on the
some concluding remarks.
other hand they are willing to change the lease conditions to their tenants in
order to allow them to improve agricultural technique and thus to pay higher
rents:
5.2. LANDOWNERS’ CONSUMPTION AND THE
TAKE-OFF OF A STATIONARY AGRICULTURAL Farms were enlarged, and the occupiers of land, notwithstanding the complaints of
depopulation, reduced to the number necessary for cultivating it, according to the
ECONOMY imperfect state of cultivation and improvement in those times. By the removal of
the unnecessary mouths, and by exacting from the farmer the full value of the
According to Smith, ‘no large country ... ever did or could subsist without farm, a greater surplus ... was obtained for the proprietor, which the merchants and
some sort of manufactures being carried on in it. [...] This is even more manufacturers soon furnished him with a method of spending upon his own person
universally the case in those poor countries which are commonly said to have in the same manner as he had done the rest. The same cause continuing to operate,
no manufactures’ (WN, III.iii.17). Smith’s examples of primitive he was desirous to raise his rents above what his lands, in the actual state of their
manufactures carried on in all countries are clothing and housing. Of course, improvement, could afford. His tenants could agree to this upon one condition
the presence of such manufactures does not imply that the economy is an only, that they should be secured in their possession, for such a term of years as
industrial one. Thus a stationary agricultural economy may be defined as an might give them time to recover with profit whatever they should lay out in the
economy where there are no manufactures of the industrial type and where further improvement of the land. The expensive vanity of the landlord made him
willing to accept this condition; and hence the origin of long leases (WN,
rent, earned by landowners (the sovereign, the landlords and the clergy), is
III.iv.13).
spent on the consumption of personal services or of commodities produced
by foreign ‘finer manufactures’ and imported from abroad. Thus, according to Smith, the implementation and development of
Rent may be also partially saved and hoarded.1 It goes without saying that manufactures deriving from a widespread taste for finer commodities involve
in a stationary agricultural economy rent is the only kind of income which is a series of interesting economic phenomena: the widening of the commodity
not tied to the requirement of (re)production or of subsistence. Rent absorbs space leads to a significant improvement of the efficiency of agricultural
the whole surplus produced in the economy. production and to an increase of both national income and consumption.
It is to be stressed that landowners’ consumption behaviour is not Rosenberg summarizes Smith’s point of view in this regard:
explained by Smith in subjectivistic terms, e.g. as the preference for a certain
kind of commodities. Smith’s explanation runs in objectivistic terms since it The expansion in the range of alternatives for the disposition of the economic
involves the absence in the economy under scrutiny of alternative surplus had the immediate effects of 1) shifting the composition of consumer
expenditure flows away from services and towards goods; 2) shifting upward the
commodities to consume. Thus landowners’ consumption behaviour is
consumption functions of large property owners, who previously lived within their
caused by the narrowness of the commodity space at their disposal, which is incomes because of the limited scope afforded for the exercise of personal vanity;
a very important point in relation to the possibility of take-off of the and 3) the strength of the desire for these new goods provided a motive for
economy. Moreover, the absence of finer commodities to consume implies efficient cultivation which was previously lacking. The increased incentive
the absence of incentives to improve the organization of agricultural provided by the availability of new goods led to the elimination of known
production. As noted by Brewer, given known technology, land resources are inefficiencies which had previously been tolerated and to legal and institutional
under-employed and technological innovation is neglected: changes which, by strengthening economic incentives, Smith regarded as
indispensable to sustained economic growth (Rosenberg, 1968, p. 368).
agriculture had been ... under-performing because of indolence, caused by a lack
of attractive manufactures. A taste for ‘luxury’, and an opportunity to gratify it,
In a stationary agricultural economy the possibility of take-off is strictly tied Also Ricardo focuses mainly on supply factors as the main determinants
to the creation of conditions favourable to the implementation and of growth. For Ricardo, limits to growth are basically to be located in the
development of manufactures. In the economy under scrutiny, the traditional supply conditions of a crucial factor of production, land, whose quantity and
sectors, such as agriculture and personal services, have plenty of physical quality are assumed to be given and invariant. The logical chain underlying
resources which are idle or under-employed and which may be diverted to Ricardo’s argument may be briefly reconstructed as follows. In the long run
the new, growing sector, manufactures. Yet, the availability of idle physical the rate of real wages is at its historically determined level of subsistence.
resources is not sufficient for take-off. The presence is required of a class of Workers consume almost exclusively agricultural products (usually referred
agents within the economy who gain profits from the refinement of domestic to as ‘corn’). Thus in the long run the money price of ‘corn’ regulates the
primitive manufactures. This class gains benefits from the fact that the taste rate of money wages. Since the ‘natural’ or long-run normal price of each
for finer commodities has become so general as to occasion a considerable commodity is determined by its productive conditions, the normal price of
demand for them. In the case of an economy characterized by a wide external ‘corn’ is regulated by the state of cultivation. Thus diminishing returns in
trade, profits may also derive from the domestic production of those finer agriculture provoke an increasing ‘price of labour’ for industrial
commodities previously imported from abroad. entrepreneurs (the rate of money wages increases to compensate the rising
price of ‘corn’). Provided that 1) the rate of wages and the rate of profits
move in opposite directions and that 2) the rate of profits and the rate of
5.3. CONSUMPTION PATTERNS AND INCOME capital accumulation move in the same direction, an increasing population
DISTRIBUTION IN AN INDUSTRIALIZED provokes an increasing demand for ‘corn’, a rising price of ‘corn’ (as soon as
agriculture enters its diminishing returns stage), a rising rate of wages, a
ECONOMY
falling rate of profits and a falling rate of capital accumulation. Population
growth and capital accumulation cease as soon as the rate of wages and the
Once industrialization starts, the economic problem is constituted by the
rate of profits reach their ‘natural’ levels. If technical progress is unable to
persistence of growth, that is, by the cumulative processes which may sustain
counteract the action of diminishing returns in agriculture, the process of
or choke the expansion of manufactures. Smith apparently privileges supply
economic growth inevitably comes to a halt.
side factors in the capital and labour markets. Right at the beginning of WN
It is not difficult to find quotations within Ricardo’s texts which point to a
Smith declares that in each country per capita income is not regulated by
supremacy of supply factors in the analysis of growth. Perhaps the most
natural factors such as ‘soil, climate, or extent of territory’ but by the
explicit is the following:
productivity of its labourers, namely ‘first, by the skill, dexterity, and
judgment with which its labour is generally applied; and, secondly, by the
Profits do not necessarily fall with the increase of the quantity of capital because
proportion between the number of those who are employed in useful labour, the demand for capital is infinite and is governed by the same law as population
and that of those who are not so employed’. Since only part of total labour is itself. They are both checked by the rise in the price of food, and the consequent
applied to productive activities, the rate of growth of a country is greatly increase in the price of labour. If there were no such rise, what could prevent
affected by the factors which determine the share of ‘productive labourers’ population and capital from increasing without limit? (Ricardo, 1951–73, vol. VI,
and its dynamics. These factors are basically saving decisions: it is p. 301).
‘parsimony’, in fact, which determines the ‘funds destined for the
maintenance of productive labour’ (WN, II.iii.14). Yet, things are not that easy. One of the crucial assumptions in the chain of
Yet, Smith did not neglect the demand side in his analysis of economic reasonings sketched above is that workers consume (almost) only ‘corn’, that
growth. In a growing economy the leading sector is the manufacturing is, a basket of commodities produced in the agricultural sector and thus
sector: manufactures are characterized by considerable scale economies and subject to the law of diminishing return. What would happen if workers
a high rate of technical progress because they offer broader scope than reacted to an increasing real income with a reduction of their consumption of
agriculture for the process of division and specialization of labour. As is well ‘corn’, with a control of their fertility and with an increase in their
known, for Smith, the pace of this process is basically determined by demand consumption of ‘conveniences’, that is, manufactured commodities produced
factors (WN, I.iii). As Young remarked 152 years after Smith: ‘it would be under a regime of increasing returns? In short, what would happen if
wasteful to make a hammer to drive a single nail’ (Young, 1928, p. 530). workers’ consumption baskets abided by Engel’s Law, one of the most
certain empirical regularities in economics? In his paper on the mathematical falls to the share of the labourers employed’ (Malthus, 1986, p. 219).2
formulation of the Ricardian system, Pasinetti (1960) overtly denounces ‘the Malthus’ argument is that a low level of ‘effectual demand’ for manufactured
crudeness of Ricardo’s assumptions’: commodities may depress the rate of profits and the rate of capital
accumulation long before the exhaustion of fertile lands:
The economic theory of demand had not yet been developed, at [Ricardo’s] time,
and there is no question of substitution among wage goods in the Ricardian model. But it appears to me perfectly clear in theory, and universally confirmed by
The natural wage-rate is represented by a fixed basket of goods, to be accepted as experience, that the employment of capital may, and in fact often does, find a
given by factors lying outside economic investigation (Pasinetti, 1960, p. 90, limit, long before there is any real difficulty in procuring the means of subsistence;
Pasinetti’s emphasis). and that both capital and population may be at the same time, and for a period of
considerable length, redundant, compared with the effectual demand for produce
Ricardo himself was aware that to assume the normal consumption basket of (Malthus, 1986, p. 321).
workers as consisting only of ‘corn’ was too crude an assumption even for For Malthus, a low level of ‘effectual demand’ may derive from three
his times. He plainly acknowledges that workers’ standards of living are sources: (i) a ‘most unequal and vicious’ distribution of land resources, (ii)
constantly rising: barriers to internal and external trade and (iii) an insufficient amount of
consumption from Smithian unproductive labourers. Malthus explicitly
Many of the conveniences now enjoyed in an English cottage, would have been considers the distribution of land property as one of the main determinants of
thought luxuries at an earlier period of our history (Ricardo, 1951–73, vol. I,
the level of ‘effectual demand’. In his view
p. 97).
a very large proprietor, surrounded by very poor peasants, presents a distribution
He also acknowledges that in a growing economy the movement of relative of property most unfavourable to effectual demand. [...] Thirty or forty
prices favours an increasing consumption of manufactured commodities: proprietors, with incomes answering to between one thousand and five thousand a
year, would create a much more effectual demand for the necessaries,
From manufactured commodities always falling, and raw produce always rising, conveniences, and luxuries of life, than a single proprietor possessing a hundred
with the progress of society, such a disproportion in their relative value is at length thousand a year (Malthus, 1986, pp. 298–9).
created, that in rich countries a labourer, by the sacrifice of a very small quantity
only of his food, is able to provide liberally for all his other wants (Ibidem). Yet Malthus was not unaware that consumption expenditure may be
depressed by going too far in the redistribution of land property. In the final
Finally, Ricardo is perfectly aware that a widespread taste for conveniences part of Chapter VII, in fact, he discusses the economic consequences of the
among workers is the surest remedy against the evils deriving from the abolition of the right of primogeniture in England and France: a possible
action of the so-called Malthusian Law of population: long-run outcome of this abolition may be an excessive fragmentation of
The friends of humanity cannot but wish that in all countries the labouring classes land property which may have a negative effect on the development of a
should have a taste for comforts and enjoyments, and that they should be country.
stimulated by all legal means in their exertions to procure them (Ricardo, 1951– The following Chapter VIII is devoted to scrutinizing the relationship
73, vol. I, p. 100). between the extent of internal and external trade and domestic prosperity. 3
According to Malthus, ‘no country with a very confined market, internal as
Unfortunately, Ricardo did not develop these interesting insights any further. well as external, has ever been able to accumulate a large capital, because
By contrast, in Malthusian economics the rate of profits and the rate of such a market prevents the formation of those wants and tastes, and that
capital accumulation may decline even if top quality land is still available. desire to consume, which are absolutely necessary to keep up the market
Limits to growth for Malthus are basically to be located on the demand side. prices of commodities, and prevent the fall of profits’ (Malthus, 1986, p.
Malthus concedes to Ricardo that the scarcity of land of first quality is the 309).
limiting principle of profits; but he claims that the principle which actually Consumption expenditure out of wages may play for Malthus an
regulates the rate of profits is ‘the varying value of the produce of the same important role to sustain long-run growth. In Chapter IV of PPE Malthus
quantity of labour occasioned by the accidental or ordinary state of the admits that growing real wages may induce workers to modify their concept
demand and supply, by which a greater or smaller proportion of that produce
Income distribution and consumption patterns in a ‘classical’ growth model 91
of subsistence not only from a quantitative but also from a qualitative point
of view. In rapidly growing economies, real wages and, consequently,
workers’ power of purchasing manufactured commodities or ‘conveniences’
usually increase. Since the consumption of subsistence goods, usually
referred to by classical authors as ‘necessaries’ or more simply ‘corn’, is
proportional to the number of children to rear, a family of growing size
obliges workers to reduce the share of their income devoted to the
consumption of conveniences. Thus if workers develop a taste for
conveniences, then they would probably control their fertility and keep their
labour supply at least constant in the face of increasing real wages:
From high real wages, or the power of commanding a large portion of the
necessaries of life, two very different results may follow: one, that of a rapid
increase of population, in which case the high wages are chiefly spent in the
maintenance of large and frequent families; and the other, that of a decided
improvement in the modes of subsistence, and the conveniences and comforts
enjoyed, without a proportionate acceleration in the rate of increase (Malthus,
1986, p. 183). Figure 5.1 – Engel’s curves for food and non-agricultural goods
In this second scenario rising real wages provide an important source of Thus the consumption pattern of each agent depends on his/her personal
effectual demand for commodities produced by manufactures. income; as long as income is lower than a certain threshold, z, total income is
devoted to food consumption. Rich agents can consume both food and non-
agricultural goods; richer agents consume a wider range of goods than poorer
agents. Agents care about current consumption; saving and investment in our
5.4. THE BASIC MODEL model are carried on in the same period and there is no proper accumulation
of capital stock as in standard growth models. As shown below, we assume
In this section we present an extension of Murphy et al. (1989) whose basic that investment takes the form of a payment of a fixed cost in terms of wages
features are the following. The economy has an agricultural sector, a and has a depreciation rate equal to 1.
manufacturing sector and a personal services sector. The former produces a Food is produced by land and labour. We assume a fixed coefficient
homogeneous good, ‘food’, by means of a decreasing returns technology production function and decreasing returns to scale. Moreover, we assume
using land and labour. The manufacturing sector produces a continuum of that the supply of land is overabundant in relation to labour supply.
goods by means of an increasing returns technology using labour: we call Employment in agriculture is thus the only factor which determines the
these goods ‘industrial goods’. We assume that the very same goods may be aggregate production of food. We consider the price of food as numeraire, so
produced by a constant returns technology using labour in the personal that the following equality holds:
services sector: we call these goods ‘personal services’. The economy is
populated by a continuum of agents of measure L. We assume that each F ( LF , aF ) = R + wF LF (1)
agent devotes his/her first z units of income to food consumption; all the
remaining income (if any) is devoted to non-agricultural goods x. The latter where R is the rent, F the production function in the food sector, with
are ranked according to an index q and the marginal utility of good q is ∂F ∂LF > 0 and ∂ 2 F ∂LF < 0 , aF is a productivity parameter
2
decreasing in q. Every non-agricultural good is such that a consumption (∂F ∂aF > 0) and wF is the rate of wages in the food sector. We assume that
lower or greater than one unit does not produce any utility. More formally, wF negatively depends on LF , while R positively depends on LF :
let X = {x (q ) ∈ {0,1} with q ∈ [0, +∞ )} be the consumption set of non-
agricultural goods. The assumptions made on preferences imply that every
wF = w ( LF , aF ) (2) 5.5. EQUILIBRIUM ANALYSIS
RF = R ( LF , aF ) (3) In this section we analyze the equilibrium characteristics of our model, given
technology. Murphy et al. (1989b) consider labour in agriculture and labour
in manufactures as imperfect substitutes. According to historical observation,
where ∂w ∂LF < 0 , ∂w ∂aF > 0 , ∂R ∂LF > 0 and ∂R ∂aF > 0 .5 We stress they suppose that wM > wF because of an indirect cost borne by labourers to
that the assumption ∂w ∂LF < 0 alone does not imply by itself that work in a factory. Hence, if wM > z > wF then a flow of labour from
∂R ∂LF > 0 . agriculture to manufactures implies an increase of demand for industrial
Every good can be produced by two technologies. If a good is produced goods. Nonetheless, for the sake of simplicity, we assume that labourers can
by the increasing returns technology we consider it an industrial good or a costlessly move across sectors and that every agent supplies one unit of
convenience; if the good is produced by the constant returns technology we labour inelastically. This implies that wages in all sectors are equal:
consider it a personal service or a luxury good.6 In particular, aL units of
labour are necessary to produce one unit of good when the constant returns wF = wM = wL = w
technology is adopted, while aM < aL units of labour are necessary when the (5)
increasing returns technology is adopted. Yet, in order to use the latter
technology firms have to pay a fixed investment equal to C units of labour. The equilibrium condition in agriculture determines the level of
This cost can be considered an R&D activity which allows firms to discover employment in agriculture, LF , together with conditions (2) and (5). In
a new method of production and provides them with a monopoly power.7 By equilibrium the following relationship must hold:
contrast, luxury goods are sold in competitive markets.
min {w, z}( L − N ) + Nz = F ( LF , aF ) (6)
Though agents’ endowments of land and capital are different, for
simplicity we assume that the share of land owned by each agent is the same
From the above equality it is possible to derive the equilibrium
as his/her share of firms. This implies that there is no difference between
employment in agriculture (according to equation (2) w is a function of LF ).
capitalists and landowners and that rents and profits distributions are the
Finally since the following relationship must hold:
same. Once agents are ranked according to their shares in increasing order,
si is the share of agent i and G ( si ) the cumulative distribution of s. We wLF + R = min {w, z}( L − N ) + Nz
assume that a large part of population owns no property rights on land and/or
firms. Let L be the total population and N = L (1 − G ( s )) be the number of the level of rents in equilibrium is:
shareholders, where s = sL − N is the minimum positive share: L – N is the
number of agents having nothing but their labour. The generic agent’s R = min {w, z}( L − N ) + Nz − wLF (7)
income is given by y = w + s ( R + Π ) , where s ∈ {0, [s, ∞ )} and Π is the
aggregate profit. We assume that every shareholder can buy at least z units of In equilibrium the extent of industrialization is determined by the demand
food: for conveniences, given technology and prices. Murphy et al. (1989) show
that the equilibrium prices of all non-agricultural goods are the same and
y = w + s (R + Π ) > z (4) equal to the price which obtains when the good is produced by the constant
The value of N and the shareholders’ distribution are crucial in order to returns to scale technology. In particular, since the markets for luxuries are
determine the extent of demand of conveniences if w < z, that is, if wage- competitive then the price of a luxury good is set at aL w and the profit for a
earners can buy only food; otherwise the demand for conveniences depends monopolist is given by
on the size of population L. π = ( aL w − aM w ) O − Cw
where O is the market output. A good is produced if (expected) profits are

positive, that is, if O ≥ C ( aL − aM ) . Therefore, the minimum quantity which
makes it profitable to produce a good is O* , given by
To complete the description of equilibrium it is necessary to calculate total

profits. The latter are equal to
a w − am w ⎪⎧ ⎫⎪
*
s
Π = L ⎨( L − N ) max {w − z, 0} + ( R + Π ) ∫ sdG ( s )⎬ +
aL w ⎩⎪ s ⎭⎪
aL w − am w
+
aL w
{( N − N )(w − z ) + N s ( R + Π ) + N
* * * *
( w − z )} − CwQ*
where the first member in brackets is the demand (in nominal terms) from
labourers (residual income from food); the sum of the second and third
members is the demand from middle class shareholders and the sum of the
fourth and fifth members is the demand from the richest agents; ( aL w − aM w )
is the difference between the price and the average (and marginal) cost of
production and finally CwQ* represents all the fixed costs paid in the
economy.
Substitution from (10) yields:
{ (
Π = η ( L − N ) max {w − z,0} + ( R + Π ) S M + N − N * ( w − z ) ) }
where
aL w − aM w
η=
aL w
is the mark-up in the monopolist markets of industrial goods and

s*
S M = ∫ sdG ( s )
s
is the share of profits and rents owned by the middle class. Note that, given
N * ≤ N , if w > z then profits are always positive. This also holds if w < z
( )
since S ( R + Π ) > N − N ( z − w ) by assumption (4).8 Therefore N * ≤ N
M *
is a necessary and sufficient condition for the existence of a manufacturing

sector. Rearranging, we obtain:
Π =
{
η ( L − N ) max {w − z, 0} + RS M + ( N − N * ) ( w − z ) } (11)
1 − ηS M
Aggregate profits are a positive function of aggregate rents R, of the level

of wages w, of the aggregate share owned by the middle class S M and of
Figure 5.2 – Income distribution and consumption pattern
mark-up η. Finally, it is interesting to calculate employment in the subsistence (w > z in our model) or below subsistence (w < z) leads to an
manufacturing sector: increase (a decrease) in total population and agricultural employment.
Formally:
( L − N ) max {w − z, 0} + ( S M + N * s* )( R + Π ) + N ( w − z ) L = Δ ( w − z )
LM = (14)
aL w
where ( d Δ dw − z ) > 0 . Thus any w – z gap is supposed to be a short-run
which positively depends on S M + N * s* (the total share of profits plus rent phenomenon. Rebus sic stantibus, the economy described in our model
spent in conveniences) and negatively on aL . A manufacturing sector cannot would not be able to achieve long-run growth. Yet, as we have argued in
arise if (i) w > z and O* > L or if (ii) w < z and O* > N . If a manufacturing Section 5.3, Malthus, whose name is usually strictly associated to the Iron
sector cannot arise, then only the first three equations and the last are the Law, was perfectly aware that the availability of conveniences and luxuries
relevant ones, with LM = 0 . may affect wage-earners’ behaviour: given the size of their families, wage-
earners may devote (at least partially) the excess of their income over
subsistence to the consumption of conveniences. Thus, workers’ population
5.6. EXTENSIONS OF THE BASIC MODEL and workers’ consumption of food may stay almost stationary or may grow
not too fast in the face of increases in agricultural productivity and real
In this section we extend our basic model in two directions according to the wages. In terms of our model we must modify equation (14) and consider the
suggestions of Smith and Malthus: we make agricultural productivity and the change in labour force also as a function of the range of available
stock of the labour force L endogenous. In WN a stationary agricultural conveniences:
economy is characterized by a class of landowners that consume only luxury
∂L ∂L
goods and a class of wage-earners that consume only food. Smith argues that L = L ( w − z, Q ) , > 0, <0 (15)
a more egalitarian land distribution and different types of agricultural ∂w − z ∂Q
contracts may lead not only to a more egalitarian rent distribution among In Sections 5.2 and 5.3 we briefly recalled Smith’s and Malthus’ analyses
landowners but also to an increase of aggregate rent thanks to an increase of concerning the relations among income distribution, demand conditions and
agricultural productivity (WN, III.ii). Both factors provide potential growth. In what follows we show that classical analysis may be easily
entrepreneurs with the incentives to invest in the manufacturing sector and formalized in terms of our model. In particular we discuss three cases which
may lead the economy out of long-run stagnation (see also Baldwin, 1956 best fit our previous results. The first concerns the factors which prevent
and Strassman, 1956). Moreover, as we emphasized in section 5.2, Smith development in an agricultural economy. The second regards the transition
claims that the availability of a wider range of goods provides landowners from a stationary agricultural economy to an industrialized one and, finally,
with further incentives to increase the productivity of their lands.9 To model the third case concerns the conditions to be satisfied for an economy to
the relationship between productivity in agriculture on the one hand and land achieve positive long-run growth.
distribution and availability of industrial goods on the other we suppose that
aF positively depends on equality in land distribution and on the range of Case I: a stationary agricultural economy
industrial goods: Consider an agricultural economy (i.e. an economy with no manufactures of
the industrial type) characterized by a low level of agricultural productivity,
aF = aF (λ , Q ) ,
∂aF ∂a
<0, F >0 (13) that is, w < F ( LF , aF ) LF < z . In such an economy there is no demand for
∂λ ∂Q industrial goods if condition 4) does not hold, that is, if s ( R + Π ) + w < z . In
this case income is entirely devoted to food consumption and the conditions
where λ is an index of inequality (e.g. Gini index) of land distribution. for take-off do not occur. This kind of poverty trap may happen if land
Our basic model neglects a crucial element of classical economics. We property is too fragmented. A certain degree of inequality in land distribution
refer to the so-called Iron Law of wages and population (also called may create the conditions for take-off provided that it generates sufficient
Malthusian Law of population) according to which a rate of wages above expenditure out of rents for industrial goods (see equation (10)). An increase
in agricultural productivity may lead to w < z < F ( LF , aF ) LF . In this case, economy, that is, a shift from agriculture and personal services to
the increase in agricultural productivity may lead to an increase in aggregate manufactures. We have argued that manufactures start developing as soon as
rent and create the physical resources to be potentially employed in a decrease in λ causes an increase in aF and, consequently, an increase in R
manufactures. If inequality (4) holds, then manufactures may develop and w. Given total population and the related total demand for food, an
provided that land distribution is such that the economy generates a level of increase in aF implies a labour force surplus in agriculture. Labour resources
demand able to cover the fixed and variable costs of production of industrial are thus set free to move to new developing manufactures.
goods. Thus if N < N * , that is, if land property is too concentrated then
demand for industrial goods is insufficient to make investment in Case III: long-run growth
manufactures profitable. In such a situation the increase in aggregate rent, In the previous subsection we showed how a stationary agricultural economy
generated by the increase in agricultural productivity, is spent on personal characterized by a low level of agricultural productivity may start its process
services and/or consumption of ‘finer’ foreign commodities. Domestic of industrialization. Land redistribution and/or an increase in agricultural
manufactures do not arise. productivity may encourage shareholders to boost their demand for industrial
In this regard it may be objected that an increase in agricultural goods. Provided that such a demand reaches a critical level, it becomes
productivity may lead to an increase in wages and rents and not just rents. If profitable for some agents within the economy to become entrepreneurs, that
an increase of F ( LF , aF ) LF leads to w > z, then wage-earners may become is, to invest resources in the refinement of domestic manufactures or in the
industrial goods consumers provided that the so-called Iron Law of wages domestic production of those ‘finer’ commodities previously imported from
and population does not hold. As is well known, according to this law any abroad. Shareholders’ demand thus plays a fundamental role in the take-off
increase in the rate of wages above subsistence (historically determined by of an agricultural economy. Yet the explanation of long-run growth must
‘habits and customs’) eventually leads to an increase in workers’ population. include other factors. As highlighted by the modern literature on growth,
In our model this means that w would exceed z only temporarily (see Case III technological progress is undoubtedly one of the crucial elements affecting
below). long-run growth. While we leave the analysis of technological progress to
future research, in what follows we focus on the level of wages and the
Case II: from a stationary agricultural economy to an industrial one related consumption choices of wage-earners. In Section 5.4 we assumed that
A more egalitarian land distribution leads to an increase in N and, therefore, shareholders constitute only a small fraction of total population. Thus, in the
a decrease in λ. An increase in N may lead to N > N * thus making long-run, profits and employment in the manufacturing sector of the
investment in manufactures profitable. A decrease in λ positively affects aF economy may be supposed to depend heavily on wage-earners’ expenditure
and thus leads to an increase in rents and wages (see equations (2) and (3) on industrial goods. In terms of our model this means that in the long-run
and the sign of the relevant derivatives). The above may be considered as the (L – N)(w – z) becomes the crucial addendum in equations (11) and (12).
first phase of development. The second phase begins when manufactures In our model an increase in aF implies an increase in w (see equation (2)
start producing and thus industrial goods start being widely consumed in the and the relevant derivative) and, possibly, an increase in w over z. Wage-
economy. In this second phase the relation between aF and Q becomes the earners start consuming industrial goods. But according to the Iron Law, a
crucial factor. The widening of the consumption set, that is, the increase in Q positive w – z gap fosters a rise of total population, L. An increase in L
makes landowners willing to raise their disposable income. Landowners are implies an increase in food consumption. Accordingly, agricultural
thus ready to accept the introduction of more efficient agricultural techniques production and LF rise. But an increase in LF implies a decrease in w (see
(even if this involves a partial loss of power over their tenants). Agricultural equation (2) and the relevant derivative) and thus a decrease in the positive
productivity increases and this leads to an increase in rents and wages. w – z gap. This fact would have negative effects on the growth performance
Finally, the increase in labour and property incomes leads to an increase in of the economy since it is precisely (L – N)(w – z) which mainly supports
the number of industrial goods produced in the economy (see equation (10)). long-run growth if N is small in relation to L. According to equation (15) an
The growth of agricultural production and the growth of industrial increase in the variety of conveniences and luxuries, Q, may at least partially
production sustain each other: a virtuous circle of growth starts. balance the growth of population, L , driven by the excess of wage over
It is worth remarking that the development of a new productive sector subsistence, w – z. Total population may grow at a lower rate than that
requires a shift of the labour force from the traditional sectors of the
predicted by the Iron Law. Thus a positive w – z gap may become persistent size of their family and/or to reduce their labour supply. We have argued that
and display its positive effects on long-run growth. if shareholders constitute only a small portion of total population, then the
long-run growth performance of an economy depends on wage-earners’
consumption patterns.
5.7. FINAL REMARKS To conclude, we are well aware of several serious shortcomings in our
work. To mention a few: the model presented here is static and its dynamic
It would not be an overstatement to say that contemporary growth literature properties are still to be carefully studied. Moreover, we have not analyzed a
lacks a systematic analysis of the relation between income distribution, most important factor in the long-run, technological progress. Finally, we
consumption patterns, development and growth. This neglect of demand have not considered capital accumulation. Much work is thus left over to
factors may prove a serious lacuna: the long-run stagnation of an agricultural future research.
economy as well as the long-run growth of an industrial economy may
escape full comprehension. Our paper has tried to make good this lacuna. We
have elaborated a simple model, basically an extension of Murphy et al. NOTES
(1989), to examine: (i) some of the causes which force an agricultural
economy into a situation of long-run stagnation; (ii) some of the basic forces 1. As remarked by Rosenberg: ‘In a society where the finer manufactures are not available,
driving an economy during its transition from the agricultural stage to the opportunities for cultivating one’s vanity are necessarily limited. In the absence of such
industrial one and, finally; (iii) some of the conditions to be fulfilled for an commodities, large rental incomes are employed in hospitality, in the maintenance of a large
group of retainers, and in acts of bounty to one’s tenants. In spite of these acts of generosity,
industrial economy to perform a long-run growth. however, the typical behaviour of large landowners as late as the time of European
An agricultural economy may not escape from a poverty trap if feudalism was reasonably frugal. Large landowners were not extravagant, and it was even
agricultural productivity is too low. One of the possible causes of low common for them to save’ (Rosenberg, 1968, p. 367).
agricultural productivity is excessive fragmentation of land property. Since 2. The distinction between the ‘limiting principle of profits’ and the ‘regulating principle of
in our model agents spend the whole of their income on food if income is profits’ is drawn by Malthus in Book I, ch. V of PPE. See also Costabile and Rowthorn
(1985).
lower than a certain threshold, neither wage-earners nor shareholders buy 3. As is well-known, Malthus’ overall position on foreign trade is many-sided and not easy to
industrial goods. If agricultural productivity rises, rents and/or wages raise. grasp. We shall not deal with it in greater details since in our paper we consider a closed
The economy may thus be able to generate a level of expenditure for economy. Needless to say, the analysis of foreign trade could be an interesting extension of
industrial goods which makes their production profitable. Yet an increase in our model.
agricultural productivity does not necessarily lead the economy out of its 4. An example of agent’s preferences compatible with the Engel’s curves used in the text is the
following:
poverty trap. If only rents rise and land property is too concentrated, then
extra-income may be spent on the consumption of personal services and/or ⎧c for c ≤ z
foreign luxury goods. If only wages rise and the Iron Law of wages and ⎪
U= ⎨ ∞
⎪z + ∫ x (q ) q
−γ
population holds, then extra income is spent on extra consumption of food in dq for c > z
order to rear a growing family. In both cases domestic manufactures do not ⎩ 0
develop.
where x ( q ) ∈ {0,1}, q ∈ [0, ∞ ) , c is the food consumption, z > 0 is the threshold or the
However, if landowners acquire a strong taste for industrial goods they minimum amount of food consumption required before the consumption of non-agricultural
would strive to increase their disposable income. Thus they may be willing to goods begins; q is the index of the non-agricultural goods produced by the economy and
implement a change in agricultural techniques which raise agricultural x(q) assumes value 1 if good q is consumed or 0 otherwise. Since marginal utility of good q
productivity. In such a scenario a virtuous circle of development may start: is equal to 1/q, agents prefer to consume goods with a lower index.
5. Notice that we do not assume that factors are paid their marginal productivity.
the growth of the agricultural sector, led by the growing availability of (and
6. To clarify this point consider the following example: someone willing to hear a Mozart‘s
desire for) industrial goods, releases new resources for the growth of the symphony either may buy a CD or may buy a ticket to hear an orchestra playing live music.
manufacturing sector. Similarly, if wage-earners acquire a strong taste for 7. The interpretation given in the text may be questionable, since generally the cost of
industrial goods they would not strive to increase the share of their income innovation is payed once, while the return to investment is given by a flow of future profits.
devoted to food consumption. Thus they resist the temptation to increase the In our framework it is particularly difficult to model this aspect because future profits
Income distribution and consumption patterns in a ‘classical’ growth model 103
depend on the extent of demand, while the latter depends on future innovations, changes in
income distribution, etc. Zweimuller (2000) provides an analysis of this case, though limited
to balanced growth equilibria.
To prove it consider that assumption (4) implies that s ( R + Π ) S > ( z − w ) S , from
M M
8.
which
⎡
6. Keynesian theories of growth
( N − N ) + ∫ (s − s ) dG (s )⎤⎥
s*
s ( R + Π ) S > ( z − w ) ⎢s
M *
⎣ s ⎦
Pasquale Commendatore, Salvatore
and finally:
D’Acunto, Carlo Panico and Antonio Pinto
*
s
( z − w ) ∫ ( s − s ) dG ( s )
(R + Π )S M
(
− N−N
*
)(w − z ) > s
>0 6.1. INTRODUCTION
s
9. Malthus closely follows Smith’s argument: ‘Adam Smith has well described the slack kind This paper outlines the content of a Keynesian approach to the theory of
of cultivation which was likely to take place, and did in fact take place, among the great growth. While for other established traditions it is possible to talk of a theory
proprietors of the middle ages. But not only were they bad cultivators/ and improvers; and
of growth described by some specified models and contributions,1 for the
for a time perhaps deficient in a proper taste for manufactured products; yet, even if they had
possessed these tastes in the degree found to prevail at present, their inconsiderable numbers Keynesian tradition it is only possible to identify several lines of
would have prevented their demand from producing any important mass of such wealth. We development, which share the view that the economic system does not tend
hear of great splendour among princes and nobles in every period of history. The difficulty necessarily to full employment and that the different components of demand
was not so much to inspire the rich with a love of finery, as to break down their immense may affect the rate of growth of the economy.
properties, and to create a greater number of demanders in the middle ranks of life who were
As far as we know, there is no essay in the recent literature which seeks to
able and willing to purchase the results of productive labour. This, it is obvious, could only
be effected very gradually. That the increasing love of finery assisted considerably in reconstruct the content of a Keynesian approach to growth by describing the
accomplishing this object is highly probable; but these tastes alone, unaccompanied by a lines of research, which have historically emerged. In what follows an
better distribution of property in other respects, would have been quite inefficient’ (Malthus, attempt will be made to do so. This attempt outlines a unified framework that
1986, pp. 298–99). See also Maccabelli (1997). can deal with the influence of the different components of aggregate demand
on the rate of growth of an economic system that does not tend necessarily to
full employment. The specification of this unified framework makes it
possible to preserve the diversity of the ideas proposed by Keynesian authors
on what can be considered the most relevant factors at work.2 Moreover, it
shows Keynesian growth theorists as a homogeneous crew, sharing a positive
theoretical standpoint on the role of aggregate demand, rather than a group of
authors united by a critical attitude towards orthodoxy, but unable to present
a systematic challenge to the dominant theories.3
The paper is so organised. Section 6.2 aims to derive a unifying
framework for Keynesian theories of growth from the analyses proposed by
Harrod, the founder of modern growth theory. Sections 6.3, 6.4 and 6.5 deal
with the analyses underlining the influence on growth of three components of
effective demand, coming from the Government sector, the private sector, in
the form of autonomous investment (i.e. investments not directly generated
by savings), and the foreign sector. Section 6.6 draws some conclusions.
104
Keynesian theories of growth 105 106 The Theory of Economic Growth: a ‘Classical’ Perspective
6.2. HARROD AND THE FORMATION OF A activity on this subject and after the Great Depression he actively supported
KEYNESIAN FRAMEWORK FOR GROWTH Keynes’s proposals.6 By that time, Harrod had come to recognise the need
for deep political and theoretical changes. As Young (1989, pp. 30–8) points
THEORY
out in an unpublished paper written in 1933, Harrod stated that the Great
Depression had posed a new problem to economists and politicians. The
According to Varri (1990, p. 9), Harrod‘s contributions to growth have
previous recessions had not led the economy too far from full employment,
received less attention than they deserve. Recently, however, Young (1989)
nor had they cast doubts on the belief that the economy is able to return to it.
and Besomi (1999) have reconsidered his writings, taking advantage of the
The severity of the Great Depression had changed this situation. It had
availability of his papers at the Chiba University of Commerce in Ichikawa
jeopardised political stability and raised the problem both of a new political
(Japan) and clarifying the extent to which some of his writings have been
approach and of a new economic theory able to clarify whether market forces
misrepresented. They have refuted, in particular, the view that Harrod‘s
can lead the economy towards full employment or Government intervention
efforts to develop a theory of growth and dynamics were stimulated by his
is required to restore it.
work on imperfect competition and his dissatisfaction with the Austrian trade
As an initial contribution to these problems in 1933 Harrod published
cycle theory put forward by Hayek (see Kregel, 1980, p. 98; 1985, pp. 66–7).
International Economics. This book, as Young (1989, pp. 38–9) points out,
Moreover, they have confirmed the limits of the widespread belief that
sets the lines of analysis that Harrod developed in the following years. In
Harrod developed his analysis of growth by assuming absence of monetary
International Economics and in his 1936 The Trade Cycle, he moved from
influences and fixed technical coefficients and saving propensity, in order to
Keynes‘s Treatise (Young, 1989, pp. 48–50), to focus on the cyclical
establish the famous ‘knife-edge problem’ (Solow, 1956, 1970; for the
fluctuations of the economy around a line of steady growth. His aim was to
opposite interpretation, see Eisner, 1958, Asimakopulos and Weldon, 1965,
point out that competitive market forces may widen the gap between actual
Kregel, 1980, Asimakopulos,1985).
and equilibrium growth, independently of the destabilising influences of
In opposition to the first view, Young (1989, pp. 15–50) clarified that
monetary and credit factors, which had been underlined by the literature of
Harrod‘s efforts to develop a theory of growth and dynamics were mainly
the time. His 1939 essay on dynamics, again stimulated by the discussions
stimulated by his contacts with Keynes. These began in 1922, when Keynes
with Keynes (CW XIV, pp. 150–79), focused instead on the equilibrium
invited Harrod to study economics in Cambridge under his supervision
paths of the economy and on the factors determining the ‘warranted’ and the
(Phelps Brown, 1980, pp. 7–8). One year later, having read A Tract on
‘natural’ rates of growth. This study represented ‘a preliminary attempt to
Monetary Reform,
give an outline of a “dynamic” theory’ (Harrod, 1939, p. 254) and ‘a
necessary propaedeutic to trade–cycle study’ (p. 263).
Harrod took up Keynes‘s call for deeper research into the problems of the ‘credit
cycle’, and over the next few years produced a number of essays on the subject. In It moved from the condition of equilibrium in the commodities’ market.
these Harrod focused on the theoretical basis for – and policy options related to – In the most simplified case, that of an economic system without Government
issues raised by Keynes in the Tract (Young, 1989, p. 16). intervention and closed to non-residents, this condition is represented by the
equality between saving and investment decisions. In the formal presentation
According to Young, in these essays, some of which were never published, of his analysis, the saving propensity was taken as given. Yet Harrod (1939,
Harrod dealt with a problem that was central to Keynes‘s and other works of p. 276) made some reference to the influence of the interest rate on the
the time. Moving on from the idea that the economic system is stable and propensity to save and, in his following writings, he recalled the possibility
that negative influences on fluctuations only come from monetary and credit of using Ramsey‘s intertemporal approach on which to base this part of his
factors, attempts were made to identify a ‘neutral’ policy, i.e. a policy that analysis.7 The equation relative to investment, which introduces, according
can prevent monetary and credit disturbances from amplifying the to Sen (1970, pp. 11 and 23) and Asimakopulos and Weldon (1965, p. 67),
fluctuations of the economy. the major difference with other traditions, assumes that investment decisions
In those years Harrod also focused on Keynes‘s proposals for are taken independently of saving decisions and are not generated by them.
Government interventions.4 According to Phelps Brown (1980, pp. 13 and They depend on the ‘acceleration principle’ and on the degree of utilisation
18), Harrod first heard Keynes‘s proposals at the Liberal Summer School of of capital equipment, along the following lines:
August 1924.5 From then onwards, he closely followed Keynes‘s intellectual
i = k g* + f ( g − g−*1 ) (1) normal level, inducing entrepreneurs to reduce investment decisions. In both
situations, the rate of growth will be pushed further away from the warranted
where f (0) = 0 and df /dg > 0, i is the ratio between investment and the net level. This description was considered by Harrod (1939, pp. 263–4)
*
output of the economy, g is the current period expected rate of growth of equivalent to that developed by static theory when it is assumed that the
*
output, g−1 is the previous period expected rate of growth, g is the current market price exceeds (is lower than) the equilibrium price and the
period rate of growth, k is the equilibrium capital/output ratio. appearance in that market of an excess supply (an excess demand) tends to
Harrod used his analysis to study the ‘warranted’ rate of growth (gw ), restore equilibrium. These descriptions, unlike the ‘cobweb’ analysis in the
defined as that equilibrium rate which allows the normal utilisation of capital traditional supply and demand theory, do not represent a dynamic analysis of
equipment.8 He assumed that, along the warranted equilibrium path, disequilibrium. They just point out in an informal way that some centrifugal
expectations are realised (g−*1 = g) and the expected rates of growth are equal or centripetal forces come into operation as soon as disequilibrium occurs.
to the warranted rate (g* = g−*1 = gw ) . The following equations were thus used Most literature has interpreted this part of Harrod‘s work as the outcome
for the analysis of the warranted rate: of a dynamic analysis of stability. Sen (1979, p. 14), for instance, after
s = k gw (2) pointing out that Harrod‘s analysis only deals with the initial elements of this
problem and can be compatible with different analytical developments,
k = k (r), (k'(r) ≤ 0) (3) criticised his conclusions.
r = r0 (4)
There are many other ways in which Harrod‘s somewhat incomplete model can be
where s is the average propensity to save and r is the rate of interest. completed. Some confirm instability, while others either eliminate it or make it
The introduction of equation (3) and (4) points out, in opposition to a conditional on certain actual circumstances. In general, it will be fair to say that
widespread view, that Harrod did not develop his analysis of growth by Harrod‘s instability analysis over-stresses a local problem near the equilibrium
assuming absence of monetary influences and fixed technical coefficients. without carrying the story far enough, and extensions of his model with realistic
Equation (4) assumes that the rate of interest depends on the conduct of assumptions about the other factors involved tend to soften the blow (Sen, 1970,
monetary policy, which, according to Harrod, operates by stabilising this rate p. 14).
at some specified level.9 Equation (3) recognises the possibility of
substitution between factors of production. Harrod admitted the existence of Already in 1939, however, Harrod had stated that his analysis did not give a
decreasing marginal returns,10 but considered that this kind of substitution complete account of the problem, suggesting some lines along which a
was low, following the results reached by the Oxford Research Group, in dynamic analysis of the behaviour of the system can be developed.
which he actively participated.
Space forbids an application of this method of analysis to the successive phases of
From equation (2) one can derive
the trade cycle. In the course of it the values expressed by the symbols on the
s right-hand side of the equation undergo considerable change. As the actual growth
gw = (5) departs upwards or downwards from the warranted level, the warranted rate itself
k
moves and may chase the actual rate in either direction. The maximum rates of
The study of the ‘warranted’ rate was for Harrod a preliminary part of the advance or recession may be expected to occur at the moment when the chase is
analysis of the dynamic behaviour of the economy, which in 1939 was successful (Harrod, 1939, pp. 271–2).
presented through the following steps.
The first step dealt with the forces that start to operate as soon as the Moreover, in the subsequent years, Harrod (1948, p. 99) first claimed that he
economy gets out of equilibrium and expectations are not realised. was reluctant to enter the field of the dynamic analysis of disequilibrium
According to Harrod (1939, pp. 263–7), when the rate of growth differs from without developing the analysis of the equilibrium warranted path which,
the equilibrium warranted rate, some centrifugal forces operate. If the former according to him, had a higher degree of generality.11 He then rejected the
exceeds the latter, capital equipment is utilised above its normal level, view that his aim had been to raise a ‘knife-edge problem’12 confirming that
inducing entrepreneurs to increase their investment decisions, as pointed out he had only tried to underline the existence of some centrifugal forces
by equation (1). In the opposite case, capital equipment is utilised below its coming into play as soon as the economy gets out of equilibrium. The
reference to these forces did not exclude the existence of other forces, approach to this problem: it outlines a framework that much literature within
producing stabilising effects, which have to be analysed by considering, this tradition has subsequently adopted.
according to Harrod, that the ‘natural’ rate of growth represents the ‘ceiling’
limiting the expansion of the economy.
The second step of the analysis proposed by Harrod (1939) to study the 6.3. THE INFLUENCE OF THE GOVERNMENT
dynamic behaviour of the economy considered the existence of forces COMPONENT OF AGGREGATE DEMAND
pushing the ‘warranted’ rate of growth towards the ‘natural’ rate. This part of
Harrod‘s work was based on his assumptions on substitution between factors The need to take into account the influence of Government activity on
of production and on the determination of the interest rate. As stated above, growth was pointed out by Harrod (1939, pp. 269–70 and 275), who also
Harrod did not deny the existence of substitution between factors of gave some initial formal account of how this source of demand can affect the
production, but considered that it occurred to a small extent. After 1939, this equilibrium growth path of the economy. For him, Government policies have
idea was often restated: he claimed, with increasing emphasis, that he was to be used both to stabilise the economy and to achieve higher growth.
skeptical on the possibility of reaching full employment through reduction of
Policy in this field is usually appraised by reference to its power to combat
the interest rate.13 Moreover, he confirmed that the rate of interest tends to tendencies to oscillations. Our demonstration of the inherent instability of the
show some rigidity, since it depends on the conduct of monetary policy, dynamic equilibrium confirms the importance of this. But ... in addition to dealing
which, according to Harrod (1948, pp. 99–100; 1973, p. 67), operates by with the tendency to oscillation when it occurs, it may be desirable to have a long-
stabilising this rate at some specified level. This view of the interest rate, range policy designed to influence the relation between the proper warranted rate
which also took into account the attempts of the monetary authorities to of growth and the natural rate (Harrod, 1939, p. 275).
maintain the equilibrium of the balance of payments (Harrod, 1969, pp. 178
and 191; 1973, p. 75), raises the problem of the links between the theory of In 1939 Harrod claimed that both fiscal policy and variations in the long-
growth and that of distribution, since it was associated in Harrod‘s writings term interest rate have to be used to pursue this long-range objective, adding
with the idea that a persistent change in this rate leads to a similar variation that the latter are more appropriate than the former to this aim. The bank rate
in the rate of profit.14 The analysis of this problem, however, was little policy can be used instead to combat the runaway forces of the economy.
developed by the Oxford economist, who focused instead on the conclusion If permanent public works activity and a low long-term rate availed to bring the
that one cannot rely on the belief that the spontaneous operation of market proper warranted rate into line with the natural rate, variations in the short-term
forces always leads the economic system towards full employment. rate of interest might come into their own again as an ancillary method of dealing
This conclusion led to the third step of analysis relative to the role of with oscillations (Harrod, 1939, p. 276).15
effective demand and Government policy on growth. Harrod (1939) pointed
out that the warranted rate could be influenced by three different components This position was maintained in Harrod (1948, pp. 74–5 and 117–22), where
of effective demand coming from the Government sector, the private sector, he again identified fiscal policy with ‘public works’. In the subsequent
in the form of autonomous investment, and the foreign sector. Harrod (1939, writings these ideas were revised, claiming that it was advisable to rely on
pp. 269–74) gave some initial formal account of how these three sources of fiscal, rather than on monetary policy, to affect the equilibrium warranted
demand can affect the equilibrium path of the economy. Then, he focused on path, so as to bring it close to the natural path, and to conduct fiscal policy by
the Government sector and considered how policy can be used to stabilise changing the tax rates while keeping Government expenditure constant.
the economy and to achieve higher growth and employment. This new position was presented in Harrod (1964 and 1973), where he
To sum up, the recent studies on Harrod‘s papers clarify that his seminal also recalled that the conduct of policy is difficult owing to the complexity of
work on growth theory and dynamics was conceived as an extension of the objectives to be achieved (Harrod, 1964, pp. 913–15) and to the fact that
Keynes‘ analysis to a long-period context. It developed the view that the
even if the authorities had succeeded in maintaining a steady growth rate ... for a
economic system does not tend necessarily to full employment and that the
substantial period of time – a state of affairs not yet realised – and there was
different components of aggregate demand may affect the rate of growth of general confidence that their success would continue, this would not relieve the
the economy. His theory can be considered a prototype of a Keynesian entrepreneur of his major uncertainties ... Entrepreneurs usually have to cast their
bread upon the water (Harrod, 1964, p. 907).
He proposed to use the equilibrium condition of the commodity market to economy, h is the amount of Government‘s expenditure on goods and
study how Government policy has to be applied and suggested dealing with services, measured in terms of the net output of the economy (h ≥ 0).
this equation by taking the natural rate of growth as given, i.e. as the As Harrod suggests, this equation can be used either to study the factors
objective that the long-term policy has to pursue. Harrod (1973, p. 45) affecting the warranted rate of growth (in this case, g is taken as unknown,
considered Government intervention necessary, arguing that this view was while r and the policy parameters t and h are taken as given) or to analyse
becoming increasingly popular. how fiscal policy has to be applied to maintain reasonable full employment
or growth in accordance with the potential of the economy (in this case, g is
In the spectrum of countries ranging from individualism to socialism, the U.S.A.
taken as given at its natural level, while one policy parameter, say t, is
may be regarded as being at or near the individualist end. But even in that country
considered unknown).
‘monetary’ and ‘fiscal’ policies are regarded as legitimate weapons of
government, including the central bank. These policies serve to doctor the saving From equation (6) one can derive
ratio and to provide enough, neither more nor less, to maintain reasonably full
employment and growth in accordance with the growth potential of the economy s(1 − t + rb b) + t − h − rb b
g= . (7)
(Harrod, 1973, pp. 28–9; see also 1964, p. 906). k
He also underlined that the traditional position, which confines the use of It can be noticed that variations in the tax rate keep affecting growth even
these policies only ‘to ironing out the business cycle’, ‘implies too narrow a in the simplified case of a balanced Government budget and absence of
view of the duties of the authorities’ (Harrod, 1973, p. 29). Government bonds (t = h > 0 and b = 0), when equation (7) becomes
Finally, Harrod (1964, p. 906; 1973, pp. 102–3, 173 and 177) claimed that
fiscal policy was appropriate to achieve this long-term objective. It should be s(1 − t )
g= . (8)
used by varying the tax rates while keeping government expenditure constant k
(Harrod, 1973, p. 107). Monetary policy was appropriate instead to deal with
what he defined the short-term policy objective of correcting the divergence The influence of t on g does not depend on that of t on the propensity to
of the actual rate from the warranted rate and stabilising the fluctuations of save and on the capital–output ratio 16
the economy. Temporary variations in the short-term rate of interest operate The presence of Government debt and the interest rate in equation (7)
through their effects on the availability of credit in the markets (i.e. credit raises the problem of the relationships between growth and distribution and
rationing) (Harrod, 1964, pp. 912–3; 1973, pp. 178–9). On the other hand, between monetary and fiscal policy. Only the former problem is known to
permanent variations in the interest rate tend to be more effective in causing occupy a central place in the original development of the post Keynesian
similar variations in the rate of profit than in changing the capital–output theory of growth and distribution.17 Kaldor‘s 1958 Memorandum to the
ratio (Harrod, 1973, pp. 44, 78 and 111). Radcliffe Committee, however, considers both problems simultaneously.
The formal analysis used by Harrod to deal with these views was limited. The Memorandum describes how Government policy can affect stability
It can be developed as done in equation (6) below, which follows his and growth. It argues that monetary policy has to stabilise the short-term
proposal to study how to apply Government policy by using the equilibrium interest rates in order to avoid some ‘undesirable consequences’. The
condition of the commodities’ market, which in this case takes the form instability of the interest rates enhances financial speculation and reduces the
‘saving plus taxation is equal to investment plus Government expenditures’. ability of the markets to convey financial resources towards productive
enterprises. Moreover, it raises the risk premium to be paid on loans of
s (1 – t + rbb) + t = kg + h + rbb (6) longer maturity and leads to higher long-term interest rates. Higher long-term
interest rates, in turn, make the management of Government debt difficult.
where s is the private sector’s propensity to save (0 < s < 1), t is the average Moreover, they increase the probability that firms may not be able to pay
tax rate, defined in terms of the net output of the economy (0 < t < 1), rb is back their loans, making lending institutions and financial markets more
the interest rate on Government bonds, b is the amount of Government fragile. Finally, they tend to cause economic stagnation.
bonds in circulation, measured in terms of the net output of the economy To justify the tendency to stagnation Kaldor made reference to his theory
(b ≥ 0), k is the capital–output ratio (k > 0), g is the rate of growth of the of growth and distribution and to the ‘Cambridge equation’.
In a steadily growing economy the average rate of profit on investment can, in the Kaldor did not present his positions on the role of Government policy in a
first approximation, be taken as being equal to the rate of growth in the money formalised way. Nor can such a treatment be found in other literature of that
value of the gross national product divided by the proportion of profit saved … To time. His reference to the Cambridge equation must then be considered, as he
keep the process of investment going, the rate of profit must exceed the (long- himself stated, a first approximation rather than the result of a thorough
term) interest rates by some considerable margin (Kaldor, 1958, pp. 137–8)
treatment of this problem. The first formal presentation of the post
A monetary policy causing unstable interest rates raises the long-term rates Keynesian theory of growth and distribution, which explicitly introduced the
to a level considered by investors too high to keep accumulation going. Government sector, was provided by Steedman (1972). This article proved
Under these circumstances, stagnation prevails, unless the rate of profit is that in an analysis that assumes a balanced Government budget and no
raised too. According to Kaldor, this can be done through fiscal policy. outstanding bonds, the Cambridge equation holds in a larger number of cases
than the ‘dual theorem’ of Modigliani and Samuelson. Some years later,
If the rate of interest were higher than [the level that keeps investment going], the Fleck and Domenghino (1987), who challenged the validity of the
process of accumulation would be interrupted, and the economy would relapse Cambridge equation when the Government budget is not balanced,
into a slump. To get it out of the slump it would be necessary to stimulate the stimulated an intense debate on this subject. The debate has examined a large
propensity to consume – by tax cuts, for example – which would raise the rate of
number of cases, showing when the Cambridge equation holds and
profit and thus restore the incentive to invest (Kaldor, 1958, p. 138).18
confirming the conclusion that Steedman had previously reached.19
The post Keynesian theory of growth and distribution, to which Kaldor The results of the debate show how the views on the role of Government
greatly contributed, differs from Harrod‘s growth theory for the introduction policy that Kaldor presented in the Memorandum to the Radcliffe
of the saving propensities of different income groups and for the role Commission can be formally developed and clarify some features of his
attributed to distributive shares in restoring equilibrium conditions. proposals. Let us consider the case examined by Denicolò and Matteuzzi
According to some literature, this part of Kaldor‘s work departs from the (1990), in which the Cambridge equation holds. It refers to a closed economy
Keynesian tradition, since it does not reject the idea that market economies with two classes (workers and capitalists),20 where the Government sector
tend to full employment. finances its budget through the issue of bonds and the private sector finances
Kaldor‘s Memorandum to the Radcliffe Commission does not confirm its productive activity through the sale of shares to other components of the
this allegation (Kaldor, 1958, pp. 135–7 and pp. 141–2). It shows many private sector. Capitalists do not work: they earn their income through the
similarities with the views proposed by Harrod and the rest of Keynesian returns of their wealth. Moreover, the two classes have different saving
tradition on the role of Government policy. First of all, Kaldor considered propensities, can invest their wealth in shares representing real capital and in
Government policies necessary to pursue stability and growth. Secondly, Government bonds, and have the same portfolio structure (for the case of
thought that Government policies have to deal with a complex set of different portfolio structures, see Panico, 1993). To study what are the
objectives, which are interrelated – and often incompatible – among them. conditions allowing steady growth, we must specify the equilibrium
Thirdly, for Kaldor, monetary policy is the appropriate tool against the condition in the commodities’ market, the dynamic equilibrium conditions
fluctuations of the economy, while it is advisable to use fiscal policy to between the savings of the two classes and the growth of their wealth, and
pursue the long-range objective of sustained growth. Fourthly, when he the dynamic equilibrium condition between the Government budget and its
advocated fiscal policy, Kaldor referred to variations in the tax rate, rather debt. These conditions can be written as follows:
than to variations in the level of Government expenditure. Finally, like
sc(1 – t) α (rbb + rkk)+ s (1 – t) [1 + rbb – α (rbb + rkk)] + t = gk + h + rbb (9)
Harrod, Kaldor proposed to use the equilibrium condition of the
commodities’ market to deal with these problems and referred to it either to sc (1 – t) α (rbb + rkk) = g α (b + k) (10)
determine the growth path of the economy (considering the rate of growth as
unknown and the interest rate, the tax rate and Government expenditure as g b = h + rbb – t (11)
given) or to determine the intensity of fiscal policy appropriate to the
where sc is the propensity to save of the capitalist class (0 < sc < 1), t is the
achievement of a specific rate of growth (considering one policy parameter –
tax rate (0 < t < 1), which is assumed to be the same on all forms of income,
the tax rate – as unknown and the rate of growth as given).
α is the quota of wealth owned by the capitalist class (0 ≤ α ≤ 1), sw is the
propensity to save of the working class (0 < sw < sc), rb is the rate of interest Keynesian investment function and several investment-led growth theories
on bonds, b is the stock of Government bonds measured in terms of the net have been proposed. The first type of theory (labelled neo-Keynesian) was
output of the economy (b ≥ 0), g is the rate of growth, k is the capital/output proposed by Joan Robinson (1956, 1962) and Kaldor (1957 and 1961). They
ratio (k > 0), h is the Government expenditure on goods and services, are characterised by full capacity utilisation of plants, flexible income shares
measured in terms of net output (h ≥ 0), rk is the rate of return on real capital. and a functional relationship between the rate of capital accumulation and the
If we assume rb = rk = r, equation (10) becomes: rate of profits.21 A second group of theories (labelled Kaleckian) was
inspired by the works of Kalecki (1971) and Steindl (1952). They assume
sc (1 – t)r = g. (12) that firms under-utilise their productive capacity and apply mark-up
This confirms the validity of the Cambridge equation, taking into account procedures in determining prices. Moreover, capital accumulation is driven
the role of t, and allows one to calculate the value of t compatible with steady by profitability (through the rate of profits) and by effective demand
growth at the rate of interest fixed by the monetary authorities. (through the degree of capital utilisation). These investment-led growth
Equations (9)–(12) thus show how to develop in a formal way the views theories have been further elaborated in the literature. In what follows, an
proposed by Kaldor in his Memorandum to the Radcliffe Commission, where attempt is made to compare the alternative lines of development of
the lack of a formal analysis of how Government intervention can affect investment-led growth within the Keynesian tradition by introducing a
growth and distribution led the author to refer to a version of the Cambridge homogeneous set of equations which can be modified to take account of the
equation which, unlike equation (12), does not include the tax rate. As a assumptions relating to capital utilisation, income distribution and
consequence, Kaldor conceived the influence of tax variations on growth in investment determinants.
terms of their effect on the propensities to save. The analysis presented Let’s assume (i) a closed economy with no government intervention; (ii)
above, instead, clarifies how Government intervention can affect demand and two factors of production, labour and capital, with a fixed coefficient
growth independently of changes in the propensities to save and in the technology; (iii) flexible labour supply; (iv) absence of technological
capital–output ratio. It thus further elaborates Kaldor‘s attempt to describe progress and capital depreciation; (v) identical physical composition of
how fiscal policy can be used to maintain steady growth conditions. capital and product; (vi) homogeneous firms. The following equations can
Finally, the results of the recent debate on the role of the Government then be written
sector in the post Keynesian theory of growth and distribution clarify some 1 = wal + rkk (13)
other common elements of the classical and the Keynesian traditions (see
Panico, 1997, 1999). They allow reconciliation of two approaches to 1 ⎛ l 1⎞
distribution, which have been considered alternative (see Moss, 1978, p. 306; = min ⎜ , ⎟ (14)
k ⎝ al ak ⎠
Vianello, 1986, p. 86; Nell, 1988; Pasinetti, 1988; Pivetti, 1988; Wray, 1988;
Abraham–Frois, 1991, pp. 197 and 202). These are the approach proposed by ak
Kaldor and Pasinetti in their theory of growth and distribution and that u= (15)
k
implied by Sraffa‘s suggestion in Production of Commodities to take the rate
of profit, rather than the wage rate, as the independent variable in the min(wπ , wω) ≤ w ≤ max(wπ , wω) (16)
classical theory of prices and distribution.
s = s c rk k (17)
i
6.4. THE INFLUENCE OF AUTONOMOUS = γ (rk , u, g) (18)
k
INVESTMENT
s=i (19)
The introduction of an autonomous investment function is often considered
what differentiates a Keynesian theory of growth from other approaches. where k is the capital/output ratio, w is the real wage rate, rk is the rate of
There is, however, no agreement in the literature on what characterises a profits, l is the labour/capital ratio, al is the labour coefficient of production,
ak is the capital coefficient of production, u is the degree of capacity
utilisation, wπ is the wage firms are prepared to pay, wω is the wage workers i
are prepared to accept, s is the ratio between saving and output, i is the ratio = γ 0 + γ 1rk (22)
k
between investment and output, g is the rate of growth of income, sc is the
capitalists’ propensity to save, with 0 < sc ≤ 1. s=i (23)
According to equation (13) output (normalised to one) is distributed
between wage and profit recipients. Following expression (14), which By rearranging (20), one obtains the following expression
describes a fixed-coefficient (Leontief) type technology, the elastic labour 1 a
supply guarantees that the labour/output ratio always coincides with the rk = −w l (24)
corresponding technical coefficient, al = lk. Conversely, capital is not ak ak
necessarily fully utilised. It follows that output is not necessarily the
which describes the traditional long-term negative relationship between r and
maximum technologically possible, 1/k ≤ 1/ak. Expression (14) leaves open
w. Following Joan Robinson (1962), investors’ ‘animal spirits’ (encapsulated
the determination of the degree of capacity utilisation, defined in expression
in the constant coefficients γ 0 and γ 1) are prompted by expected profitability
(15) as the ratio between current demand and full capacity output. It is
and favoured by the availability of internal finance. This explains the
possible to envisage two cases. In the first, capacity is fully utilised, that is,
relationship (22) between desired investment and the rate of profits.
the equality u = 1 (1/k = 1/ak ) holds. In the second, some capacity is left idle
The model (20)–(23) is similar to that proposed by Marglin (1984a,
with the degree of capacity utilisation settling in any period at some level
1985b) to describe the contributions of Joan Robinson and Kaldor to growth
which does not necessarily equal one, that is, u ≤ 1 (1/k ≤ 1/ak ). Expression
theory. By imposing the equilibrium growth condition according to which all
(16) also leaves the wage rate open to two possible determinations. In the
the variables have to grow at the same rate, i/k = g, the solutions are
first case, workers’ and firms’ claims over the shares of income (in real
univocally determined:22
terms) are not inconsistent, wω ≤ w ≤ wπ. If follows that distribution and
growth are simultaneously determined. In the second case, workers and firms γ0
lay conflicting claims over income shares, wπ ≤ w ≤ wω (and wω ≠ wπ ). The rk = (25)
sc − γ 1
distribution between profits and wages depends on the relative power of
workers and firms. The way in which distribution is in fact determined γ0
depends on the institutional setting. Equation (17) clarifies that saving g = sc (26)
propensities differ between classes. According to expression (18), investment sc − γ 1
demand depends on profitability (through rk ), on the demand level (through
There are three major features of the neo-Keynesian analysis. The first is
u) and on demand growth (through g). Keynesian approaches to investment-
that distribution and growth are simultaneously determined. The second is
led growth differ inasmuch as they do not assign to each of the determinants
the transposition to the long run of the so-called ‘paradox of thrift’,
of investment the same prominence. Finally, equation (19) represents the
according to which an increase in the propensity to save induces a reduction
equilibrium condition saving equal to investment. The model (13)–(19) has
in the rate of growth and in the equilibrium rate of profits. Indeed, by
three degrees of freedom. The way in which it is closed differentiates the
differentiating expressions (25) and (26) with respect to sc one obtains
Keynesian approaches to investment-led growth.
The neo-Keynesian position is represented by the following equations drk γ0
derived from expressions (13)–(19) by assuming full capacity utilisation, =− <0 (27)
dsc (sc − γ 1 )2
u = 1(k = ak); endogenous income distribution, wω ≤ w ≤ wπ; and disregarding
the role of the rate of growth of demand in the investment function: dg γ 0γ 1
=− <0 (28)
1 = wal + rkak (20) dsc (sc − γ 1 )2
s = scrak (21) The third is the negative relationship between g and w. From (21), (23)
and (24), taking into account the equilibrium condition i/k = g, it follows that
dg sa i
=− c l <0 (29) = γ 0 + γ 1 rk + γ 2 u (34)
dw ak k
Lower levels of the wage rate correspond to higher accumulation. Profit s=i (35)
leads growth. According to expression (32), income distribution is determined outside
If the equilibrium solution w lies outside the interval wω ≤ w ≤ wπ, the the model according to the Kaleckian theory of distribution. It is assumed
neo-Keynesian analysis becomes overdetermined. When the left constraint is that firms, independently of workers’ wage resistance, fix prices through a
binding, w = wω > w and rk < rk , the economy suffers inflationary pressures, mark-up procedure securing profit margin π, wage rate wπ = (1/al ) – π /al and
because investment demand permanently exceeds saving, gak > i > s . Joan profit share rkk = 1 – wπ al = π.24 Moreover, using (31), a relationship may be
Robinson (1962) acknowledged this possibility by referring to an expressed between the rate of profits and the degree of capacity utilisation,
‘inflationary barrier’ (also named ‘real wage resistance‘), which represents
πu
the minimum level of the real wage rate organised labour is prepared to rk = (36)
accept without opposing rises in monetary wages.23 Conversely, when the ak
right constraint is binding, w = wπ < w and rk > rk , the economy is
stagnating since investment is too low (or saving is too high) for full capacity according to which rk is not univocally determined by income distribution as
growth, s > i > gak . This constraint may become operational when, it was, according to expression (24), in the neo-Keynesian model. Equation
following Kaldor (1957a), firms are – regardless of demand – not prepared to (34), a linear form of (18), postulates a relationship between capital
lower prices below that level which guarantees a minimum profit margin π, accumulation, the rate of profits and the degree of capital utilisation,
which determines wπ = (1/al ) − π /al and depends on the Kaleckian ‘degree of specified by the constant coefficients γ 0, γ 1 and γ 2.25 In Kaleckian writings
monopoly’. Note that the discrepancy between s and i can be reduced by the current rate of profits is relevant for investment decisions for two main
varying sc or (in the opposite direction) γ 0 and γ 1. reasons. It represents a proxy for expected profitability and also a source of
Unlike the neo-Keynesian approach, some economists (e.g. Rowthorn, internal financing.26 The level of capacity utilisation affects investment
1981; Dutt, 1984, 1987, 1990; Nell, 1985; Amadeo, 1986a, 1986b, 1987 and decisions both indirectly (acting through the rate of profits) and directly by
Lavoie, 1992, 1995), inspired by the works of Kalecki and Steindl, reflecting the state of demand.27
developed analyses in which firms are allowed to operate under long-run By imposing the equilibrium growth condition i/k = g, the solutions of
under-utilisation of production plants. In Kaleckian analyses demand affects equations (30)–(35) are univocally determined:
capital accumulation through changes in the degree of capacity utilisation. γ 0 ak
They assume, moreover, oligopolistic markets and conflicting claims over u= (37)
(1 − wπ al )(sc − γ 1 ) − γ 2 ak
income distribution, wω > wπ. This position can be represented by the
following equations derived from expressions (13)–(19) by assuming an
endogenous degree of capacity utilisation, u ≤ 1; exogenous income γ 0 (1 − wπ al )
rk = (38)
distribution, w = wπ; and disregarding the role of the rate of growth of (1 − wπ al )( sc − γ 1 ) − γ 2 ak
demand in the investment function:
scγ 0 (1 − wπ al )
1 = wal + rkk (30) g= (39)
(1 − wπ al )(sc − γ 1 ) − γ 2 ak
ak
u= (31)
k Note that the paradox of thrift is preserved, as shown by differentiating
expressions (38) and (39) with respect to sc,
w = wπ (32)
s = sc rk k (33) drk γ 0 (1 − wπ )2
=− <0 (40)
dsc [(1 − wπ al )(sc − γ 1 ) − γ 2 ak ]2
dg γ (1 − wπ al )[γ 1 (1 − wπ al ) + γ 2 ak ] Moving on from the relationship between the rate of profits and the
=− 0 <0 (41) degree of capacity utilisation (36), rk = πu/ak, Bhaduri and Marglin (1990)
dsc [(1 − wπ al )(sc − γ 1 ) − γ 2 ak ]2
amended the Kaleckian theory taking into account that investment reacts
The negative relationship between growth and the real wage rate, instead, differently to similar changes in profitability. In particular, at the same rate of
disappears. Equations (30)–(35) generate the so-called ‘paradox of costs’, profit investment decisions differ when profit margins are low and capacity
according to which an increase in costs, in the form of a higher wage rate, utilisation high and profit margins are high and capacity utilisation low.
implies higher profits and growth rates (see Rowthorn, 1981, p. 18 and Firms may not be willing to expand further productive capacity when excess
Lavoie, 1992, p. 307). By differentiating expressions (38) and (39) with capacity is already extensive. Consequently, equation (34) has to be replaced
respect to wπ, one obtains by the following
drk γ 0γ 2 ak al i
= >0 (42) = γ 0 + γ 1π + γ 2u (44)
dwπ [(1 − wπ al )(sc − γ 1 ) − γ 2 ak ]2 k
The solutions of the model (30)–(33), (35) and (44), considering that
dg scγ 0γ 2 ak al
= >0 (43) π = 1 – wal, are
dwπ [(1 − wπ al )(sc − γ 1 ) − γ 2 ak ]2
[γ 0 + γ 1 (1 − wπ al )]ak
u= (45)
The paradox of costs is caused by the fact that investment expenditures sc (1 − wπ al ) − γ 2 ak
are more sensitive to changes in effective demand (reflected by the degree of
capacity utilisation) induced by changes in distribution (reflected by the (1 − wπ al )[γ 0 + γ 1 (1 − wπ al )]
rk = (46)
wage share) than to changes in costs induced by changes in the wage rate sc (1 − wπ al ) − γ 2 ak
(and in the profit margin).
The analytical condition indicating when the paradox of costs occurs is sc (1 − wπ al )[γ 0 + γ 1 (1 − wπ al )]
given by the value of the elasticity ξ (u, π) < – 1. This elasticity measures the g= (47)
sc (1 − wπ al ) − γ 2 ak
sensitivity of effective demand to changes in distribution. From (36), in fact,
the inequalities drk /dπ < 0 and dg/dπ < 0 (and, therefore, drk /dwπ > 0 and By differentiating expressions (46) and (47) with respect to wπ, one
dg/dwπ > 0) imply ξ (u, π) < – 1. For the model (30)–(35), this condition obtains
always holds since, from (37),
drk [γ 2 uak − γ 1 (1 − wπ al )al ]al
π (sc − γ 1 ) = (48)
ξ (u , π ) = − < −1 dwπ sc (1 − wπ al ) − γ 2 ak
π (sc − γ 1 ) − γ 2 ak
Note finally that, when the wage rate exceeds the value dg s [γ ua − γ (1 − wπ al )al ]al
= c 2 k 1 (49)
dwπ sc (1 − wπ al ) − γ 2 ak
γ a
wπ > 1 − 2 k ,
sc − γ 1
The sign of the derivatives (48) and (49) depends on the parameters of the
the equilibrium solution u does not satisfy the condition u ≤ 1 and the model. It follows that the model modified with the investment function (44)
Kaleckian analysis becomes overdetermined. When the constraint u = 1 is is able to generate two alternative growth regimes. A wage-led growth
binding, firms cannot expand production to accommodate further rises in regime, characterised by drk dwπ > 0 and dg dwπ > 0 ( drk dπ < 0 and
demand. The disequilibrium between demand and supply, i > s > gak u , dg dπ < 0 ), prevails when u > γ 1π al γ 2 ak . The wage-led regime is
persists unless prices and profit margins rise and the wage share falls (see characterised by great responsiveness of effective demand to changes in
Rowthorn, 1981, p. 10). The neo-Keynesian adjustment mechanism is thus distribution, ξ (u , π ) < −1 . The overall effect of an increase in the wage rate
restored. on growth is positive because the positive effect of demand (induced by the
distribution in favour of workers) is greater than the negative effect of higher i

costs (generated by the increased wage rate or decreased profit margin). The = u – 1 + gu (55)
k
paradox of costs holds. Conversely, a profit-led growth regime, characterised
by drk dπ > 0 and dg dπ > 0 ( drk dwπ < 0 and dg dwπ < 0 ), prevails s=i (56)
when u < γ 1π al γ 2 ak . The profit-led regime is characterised by little
Equation (53) assigns a conventional nature to the wage rate. Unlike the
responsiveness of effective demand to changes in distribution ξ (u , π ) > −1 .
neo-Keynesian analysis, exemplified by equation (24), normal distribution,
Growth is enhanced by increases in the profit margin because the negative
and in particular the normal rate of profits, is independent of accumulation.
effect of changes in the wage share on demand is more than compensated by
rn = 1/ak – wω (al /ak) represents the normal rate of profits.33 According to
the inducement to invest caused by lower costs (lower wage rates). The
equation (55), investment expenditure is driven by an accelerator
negative relationship between w and rk and g holds as in the neo-Keynesian
mechanism. The latter involves the entrepreneur’s attempt to adjust
model.
productive capacity towards the planned degree (here corresponding to full
A recent attempt has been made to develop an approach (labelled neo
capacity) and to install capacity to adjust to (expected) demand growth.
Ricardian) to investment-led growth in line with the Classical theory of
From (50)–(56), by imposing the equilibrium growth condition u = 1, one
prices and distribution (see Vianello, 1985, 1989, 1996; Ciccone, 1986,
obtains the solutions:
1987; Committeri, 1986, 1987; Kurz, 1986, 1992; Garegnani, 1992; Serrano,
1995; Trezzini 1995, 1998; Garegnani and Palumbo, 1998; Ciampalini and rk = rn (57)
Vianello, 2000; Park, 2000; and Barbosa–Filho, 2000). In this approach the
‘normal’ income distribution, that is, the distribution corresponding to the g = sc rn (58)
degree of capacity utilisation desired by entrepreneurs (which is also labelled
‘normal‘),28 is determined by conventional or institutional factors.29 According to expressions (57) and (58), in equilibrium, the rate of profits
Moreover, the rate of growth of demand may affect investment decisions, as coincides with its normal value and the rate of growth is governed by that
a result of firms’ constant attempts to match productive capacity to expected level of saving, scrn, which corresponds to normal capacity utilisation or
demand. This feature is not explicitly taken into account in neo-Keynesian ‘capacity saving’. From this analysis it follows that, along the equilibrium
and Kaleckian analyses. Neo Ricardians also object that the Kaleckian path, effective demand does not affect growth.
approach has no adjustment mechanism between the current and normal To re-assign a role to demand the neo Ricardian literature has taken two
degree of capacity utilisation.30 However, they allow that these two routes. The first introduces in the equilibrium condition of the commodity
magnitudes may differ for long periods of time.31 market a component of demand that is independent of the level of income
An attempt to clarify the neo Ricardian position is made by introducing and its rate of change (Serrano, 1995; Park, 2000; and Barbosa–Filho,
the following equations derived from expressions (13)–(19) by assuming an 2000).34 The second abandons the use of equilibrium growth analysis and
endogenous degree of capacity utilisation, u ≤ 1; an exogenous income suggests the adoption of empirical and historical analyses, which are case-
distribution, w = wω; and disregarding the role of expected profitability in the specific, in order to identify the influence of the various components of
investment function:32 demand in different historical phases (see Garegnani, 1992; Ciampalini and
Vianello, 2000; and, for an example of historical analyses, Garegnani and
1 = wal + rkk (50)
Palumbo, 1992).
1 ⎛ l 1⎞
= min ⎜ , ⎟ (51)
k ⎝ al ak ⎠ 6.5. THE INFLUENCE OF THE EXTERNAL
a COMPONENT OF AGGREGATE DEMAND
u= k (52)
k
The analysis of the influence of the external components of demand is
w = wω (53) mainly based on the contributions of Harrod, Kaldor and Thirlwall, which
s = sc rk k (54) point out that the rate of growth of an open economy may be constrained by
its trade performance. Some insights into the role of external demand can the gold outflows caused by a trade deficit cannot affect relative prices, so
already be found however in Keynes’s writings on the British return to gold. that the ‘classical’ adjustment process does not work. In this case, the gold
In The Economic Consequences of Mr. Churchill (1925), Keynes claimed outflows would cause ‘real’ effects, and a poor trade performance may
that the return to the pre-war parity would have had a negative influence on therefore become a constraint to domestic activity and employment (Harrod,
the British trade, making a sharp reduction of money wages necessary to 1933, pp. 118 and 125). This view is formally depicted through the so-called
restore the competitiveness of the national industry on overseas markets. The ‘foreign trade multiplier’ (pp. 119–23), that is a causal relationship going
wage adjustment, however, would not have been painless: in the absence of a from exports to domestic output. Consider an economy with no Government
fall in the cost of living, workers’ resistance to wage reductions had to be sector and no saving and investment. In this case, income,Y, is spent either
overcome ‘by intensifying unemployment without limits’ (Keynes, 1925, pp. on home-made consumption goods, C, and imports, M:
211 and 218).
At the time, the theory of international trade was dominated by ‘classical’ Y=C+M (59)
thinking, according to which the balance of payments automatically adjusts National income is equal to the sale of domestic goods at home, C, and
through gold flows and consequent relative price movements: countries exports, X,:
experiencing a trade deficit would lose gold, causing an internal price
deflation which would induce a rise in exports and a fall in imports such as to Y=C+X (60)
restore equilibrium. According to Keynes, however, gold flows may fail to
If the country spends on imported commodities a stable fraction μ of its
restore the balance of payments equilibrium if wages and prices react slowly
income,
to changes in the quantity of money: in these cases, the ‘classical’
mechanism would not work, and interest rate adjustments have to come into M=μY (61)
play to ensure capital inflows sufficient to compensate for the trade deficit,
discouraging capital accumulation and slackening economic activity. substituting (61) in (60) and equating (59) and (60), we get:
In the following years, Keynes restated this view on various occasions. In
1
the evidence addressed to the Macmillan Committee, he went so far as to Y= X (62)
advocate protectionism as a remedy against recession, a provocative μ
suggestion in a laissez-faire oriented environment (Keynes, 1929, pp. 113–
7). The proposal testifies to the relevance Keynes attributed to the constraint The link with Keynes’ insights into the influence of international trade on
that the balance of payments can set to domestic prosperity. In his view, as domestic prosperity is straightforward: when deterioration of the trade
long as monetary policy was sacrificed to the achievement of external performance of a country, whether a reduction of exports or an increase in
equilibrium, Britain was inevitably condemned to stagnation (Keynes, 1929, the import propensity, occurs, the commodity market equilibrium is restored
pp. 56–7). To ‘release’ monetary policy from this task the British through a reduction of output. Thus, the country’s trade performance may
competitive performance in overseas markets had to be improved. This view constrain economic activity and employment.
also emerges in the General Theory. Harrod’s analysis of the dynamic adjustment of output following an
external shock also reflects Keynes’ line of reasoning: in the case of a current
In an economy subject to money contracts and customs more or less fixed over an account disequilibrium, the gold outflows would cause pressures on interest
appreciable period of time, where the quantity of domestic circulation and the rates, thus affecting investment in fixed and working capital and giving rise
domestic rate of interest are primarily determined by the balance of payments, …, to changes in domestic output (Harrod, 1933, pp. 135–7).
there is no orthodox means open to the authorities for countering unemployment at Harrod noted that, under the simplified assumptions of the model, the
home except by struggling for an export surplus and an import of the monetary commodity market equilibrium automatically implies X = M (Harrod, 1933,
metal at the expense of their neighbours (Keynes, 1936, p. 348). p. 120). He also clarified that the relationship between foreign trade
performance and domestic ouput still holds in a more general model taking
The idea that the trade performance of a country may affect its level of into account saving and investment, even if in this case the output
activity was restated by Harrod in his 1933 International Economics. Like adjustments may no longer be sufficient to assure balanced trade.
Keynes, Harrod analyzed the case of an economy with sticky wages, where
Other contributions to the study of the role of the external component of In 1975 Dixon and Thirlwall tried to embody in a formal model the view
aggregate demand in growth theories can be found in the 1960s with presented by Kaldor in his 1970 article. According to them, the working of
Kaldor’s work on growth rate differentials, where this analysis was the growth process in an open economy may be so depicted:
intertwined with that of cumulative causation.35 In these works, which had a
great impact on development studies and on the subsequent birth of the g = γ xˆ (63)
‘evolutionary literature’,36 Kaldor claimed that orthodox theory fails to
explain the divergence in growth rates among economies, which ‘are largely xˆ = ηx ( pˆ − pˆ f − eˆ ) + ε x g f (64)
accounted for by differences in the rates of growth of productivity’ (Kaldor,
1966, p. 104). The latter, in turn, are mainly due to the economies of scale pˆ = wˆ − aˆl + πˆ (65)
occurring within the industrial sector, whose rate of growth shows an
‘extraordinarily close correlation’ (Kaldor, 1978a, p. XVIII) with the rate of
aˆl = a0 + λ g (66)
growth of GDP and productivity.
In order to describe the actual performance of the economies, Kaldor
(1966; 1967; 1970; 1972) used the notion of ‘circular and cumulative where g is the rate of growth of the economy, x̂ the rate of growth of
causation’, introduced by Myrdal (1957), considering the dynamics of the exports, p̂ , pˆ f and ê are rates of change of domestic prices, foreign prices
industrial sector as the ‘engine of growth’. Following Young, Kaldor (1966 and exchange rates respectively, g f is the rate of growth of world income,
and 1967) described growth as a process generated by the interaction ŵ , aˆl and πˆ are rates of change of wages, labour productivity and mark-up
between demand and supply: the rate of growth is positively related to the factor respectively.
ability of supply to accommodate variations in demand and to the reaction of Equation (63) specifies Kaldor’s idea that the rate of growth of the
demand to changes in supply. Moreover, he clarified that economies move economy is directly related to the growth of exports.38 Equation (64) is the
through different stages of economic development. In an early stage, the dynamic formulation of a conventional multiplicative export function
demand for consumption goods plays the leading role in the growth process. relating the rate of growth of exports to the rates of change of relative prices
In the later stages, the leading forces are, respectively, the export of and world income, with ηx and εx being constant price and income
consumption goods, the demand for capital goods, and, finally, the export of elasticities. Equation (65) describes the rate of change of domestic prices as
capital goods (Kaldor, 1966, pp. 112–4). depending on changes in the unit labour costs and on changes in the mark-up
In his subsequent essays, Kaldor underlined other aspects of the growth factor. Finally, equation (66) describes the relation between the rate of
process. In 1970 he examined how growth depends on the rate of change of change of productivity and the rate of growth of output known in the
exports, by applying Hicks’ (1950) ‘super-multiplier’ to an open economy literature as the Verdoorn’s Law.39
and considering exports as the leading force, and consumption and The equilibrium solution of equations (63)–(66) is
investment as induced components. The rate of growth of exports, in turn,
γ [ηx (wˆ − a0 + πˆ − pˆ f − eˆ) + ε x g f ]
was assumed to depend on an external cause, the world rate of growth of g= (67)
demand, and on a domestic cause, the rate of change of production costs. An 1 + γηx λ
increase in world demand raises exports and domestic production through the
super-multiplier. Increasing returns in the export sector reduces costs, unless Dixon and Thirlwall (1975) also presented the model in terms of finite
a proportional rise in wages occurs. The reduction in costs further increases difference equations, deriving equation (67) as the steady growth solution.40
exports, setting up a cumulative process, which tends to broaden the gaps This equation can be used to describe the evolution of the rates of growth of
with other regions.37 different countries or of different regions within the same country. If one
For Kaldor, therefore, the demand coming from the foreign sector plays a assumes a given mark-up in each region and given and equal values of pˆ f ,
primary role in setting in motion the growth process, while the domestic gf, and ŵ in all regions,41 the differences in the rates of growth depend on
sources of demand mainly influence the competitiveness of the economy and the regional values of λ, γ, ηx, εx, and a0.
the intensity with which the external stimulus is transmitted to the rate of Owing to its ‘aggregate’ structure, the model (63)–(66) neglects the role
growth. of the sectoral composition of the economy and, therefore, it does not
adequately depict the richness of Kaldor’s views on growth, based on the
idea that the productive structure affects the overall rate of growth of coincidence between the composition of demand and the productive structure
productivity. Yet, the relevance of these ‘composition effects’ may be easily of the economy. In 1966 Kaldor related the degree of coincidence of the
taken into account by analysing how the sectoral composition of the productive structure to demand to the stage of development reached by a
economy affects the parameters of the model. country. The more a country can rely on a large capital goods sector, the
As to λ, Kaldor (1971) argued that it mainly depends on the composition lower will be the elasticity of imports, the higher the value of γ and the more
of demand and on the weight of the capital goods sector in the productive stimulating the effect of a given rate of change of exports. A country that has
structure. High investments and a large capital goods sector enhance reached a stage of development which allows it to be a net exporter of capital
productivity and the competitive performance of the economy in the world good can enjoy ‘explosive growth’, since ‘a fast rate of growth of external
markets.42 According to Kaldor (1966; 1967; 1971), the influence of the demand for the products of the ‘heavy industries’ is combined with the self–
composition of demand on productivity is due to the presence of variable generated growth of demand caused by their own expansion’ (Kaldor, 1966,
returns in the different sectors of the economy. The intensity of the effect on p. 114).
productivity thus crucially depends on the sectors towards which the demand An important and controversial issue concerns the factors affecting ηx and
for consumption and investment is directed, since increasing returns mainly εx. Kaldor (1971) considered price competitiveness the most important factor
occur in the capital goods sector. Moreover, the extent to which this sector is at work. In Kaldor (1978c) this position was abandoned, on account of the
able to accommodate demand is also important. High quotas of investment to fact that the worst performing countries in terms of relative prices after the
output and of the capital goods sector in the productive structure enhance 2nd World War proved to be the best performing in terms of exports
productivity changes, which, in turn, improve the international performance (McCombie and Thirlwall, 1994, pp. 262–300). Kaldor (1981) then
of the economy setting up and intensifying cumulative processes. concluded that the rate of growth of exports mainly depends on income
Kaldor (1971) referred to the role of composition of demand on long-term elasticity, which in turn depends on the innovative capacity of a country, that
growth in his policy analyses too. He distinguished between the concepts of is, the capacity of a country to differentiate its products. This innovative
‘consumption-led’ and ‘export-led’ growth, arguing that the latter is more capacity gives the economy a privileged position in foreign markets.
desirable than the former: consumption-led growth tends to have negative In their 1975 paper, Dixon and Thirlwall also tested their model on United
long-run effects on productivity, since it tends to raise the weight of non- Kingdom data, but the model gave rise to unsatisfactory approximation
increasing return sectors in the productive structure of the economy. This between fitted and actual values over the period 1951–66, since higher than
tends to worsen the international performance of the economy. Hence, as actual growth rates were systematically predicted. According to Thirlwall
stated in section 6.3 above, Kaldor claimed that Government intervention (1998, p. 194) this discrepancy could be explained by the neglect of the
should avoid the use of fiscal policy to increase the rate of growth and reduce balance-of-payments constraint, in that period a severe hurdle to Britain’s
unemployment. By making growth more dependent on the demand for growth performance. To make up for this failure, in 1979 Thirlwall worked
consumption, this policy generates the undesired consequences previously out an analytical model incorporating the external equilibrium condition,
recalled. In this case, he said, the authorities should intervene on the described by the following equation:
exchange rate, rather than through fiscal measures.43
Kaldor’s writings also hint at the factors affecting γ, which depends on the p X + F = pf M e (68)
quotas and elasticities of the various components of domestic demand to the where p is the export price index, pf the import price index, e the exchange
net output of the economy.44 The elasticity of the demand for consumption is rate and F the value of net capital flows measured in domestic currency.
influenced by productivity growth through the introduction of new products Expressing (68) in terms of rates of change, we get:
of large consumption (Kaldor, 1966, p. 113; 1981, p. 603; and Rowthorn,
1975, p. 899). When this occurs a higher value of γ and a more intense effect
θ ( pˆ + xˆ ) + (1 − θ ) fˆ = pˆ f + mˆ + eˆ (69)
of a given rate of growth of exports come about. For Kaldor (1971) tax
reduction too has a positive influence on γ, through its effect on
consumption.45 Yet, any stimulus to the latter variable has long-run negative where m̂ and f̂ denote respectively the rate of growth of imports and the
consequences, as stated above, since it makes the growth process rate of change of net capital flows, while θ and (1 – θ) are respectively the
consumption-led. Finally, the elasticity of imports depends on the degree of value of exports and capital inflows as a percentage of imports. If we specify
the demand for imports and exports through the conventional multiplicative capacity growth rate up to the ceiling represented by equation (73)
functions with constant elasticities, we may express the rate of change of According to this approach, capital and labour availability does not constrain
exports through equation (64) and the rate of change of imports by: growth, being to a large extent ‘endogenous’ to the economic system. 47
The relevance of equation (73) lies in the fact that it supplies a simple and
mˆ = ηm ( pˆ − pˆ f − eˆ) + ε m g (70) attractive explanation of why growth rates differ among countries. An
increase in world income generates a rate of growth that depends on the
where ηm and ηm are price and income elasticities respectively. value of each country’s εx/εm ratio. Since there are significant international
Substituting (64) and (70) in (69) and rearranging, we get: differences in this ratio (Houthakker and Magee, 1969), the same increase in
the world income gives rise to different growth rates among countries.
A relevant question, to which this strand of literature has not yet given a
θε x g f + (1 − θ )( fˆ − pˆ ) + (1 + θηx + ηm )( pˆ − eˆ − pˆ f )
gB = (71) conclusive answer, is what determines the εx/εm ratio. In some contributions,
εm Thirlwall (1979, p. 286 and 1991, p. 26) claims that the differences in this
ratio mainly reflect those in the patterns of productive specialization. This
where gB is the rate of growth consistent with equilibrium in the balance of way of interpreting the dynamic foreign trade multiplier has striking
payments. Basing his work on the extensive empirical evidence showing implications for the theory of uneven development. For example, assume a
long-run stability in the terms of trade,46 Thirlwall assumed that the simplified world where some countries only produce manufactured goods
contribution to growth of the price term in (71) is likely to be small. If for and others only produce primary goods. As the income elasticity of the
simplicity’s sake it is assumed to be zero, equation (71) reduces to: demand for manufactured goods, due to Engels’ Law, is higher than income
elasticity of the demand for primary goods, it would be εx/εm > 1 for countries
θε x g f + (1 − θ )( fˆ − pˆ )
gB = (72) producing manufactured goods and εx/εm < 1 for those producing primary
εm goods. According to this view, therefore, the pattern of specialisation is the
source of a process of cumulative divergence in GDP levels: countries
If we also assume that a country cannot finance its trade deficit through producing primary goods would be unable to grow at the same rate as those
capital inflows for a considerable length of time, the long-run equilibrium producing manufactured goods, owing to their tighter balance-of-payments
requires that θ = 1 (McCombie, 1998, pp. 229–32). Equation (72) changes constraint.
into Although attractive, this way of interpreting the foreign trade multipliers
has been poorly supported on empirical grounds,48 inducing Thirlwall to
εx
gB = gf (73) return to the topic and clarify that, for industrial countries, income elasticities
εm must also be made to depend on the supply characteristics of the goods
produced, such as their technical sophistication and quality (see Thirlwall,
which represents the dynamic version of Harrod’s foreign trade multiplier. 1991, p. 28 and 1998, p. 187). With this revision, the ‘cumulative
The economic meaning of equation (73) is that a poor trade performance divergence’ view rooted in the post-Keynesian tradition may be extended
constrains a country to grow at a slower pace than that allowed by the growth even to growth differentials among industrial countries: in Thirlwall’s view,
of internal demand and by resource availability. If g > gf, imports would indeed, an initial discrepancy in growth rates sets in motion the negative
grow quicker than exports, worsening the country’s trade account and feedback mechanisms associated with Verdoorn’s Law, which ‘will tend to
forcing policy-makers to intervene. When for various reasons (real wage- perpetuate initial differences in income elasticities associated with “inferior”
resistance and subsequent transmission of exchange rate variations on productive structures on the one hand and “superior” industrial structures on
domestic prices, product differentiation leading to small price elasticity of the other’ (Thirlwall, 1991, p. 27).49
demand for tradable goods, etc.) exchange rate devaluations prove Thirlwall’s 1979 analysis has been subsequently extended to take into
ineffective, the balance of payments adjustment takes place through internal account the role of international capital flows. Thirlwall and Hussain (1982)
demand deflation, which slackens the pace of growth (Thirlwall, 1979, used equation (72), instead of (73), to capture the experience of some
pp. 279–80). Analogously, if g < gf and the country is able to expand internal developing countries running persistent current account deficits, financed by
demand, the pressure of demand upon productive capacity may raise the
foreign investment. In some more recent contributions (Moreno Brid, 1998– reconstructing the content of a Keynesian approach to growth and describing
99, McCombie and Thirlwall, 1999), however, the use of equation (72) has the lines of development that have historically emerged, this paper has tried
been considered inappropriate for a steady-state analysis without imposing to underline the wealth of this tradition. At the same time, it has sought to
any restriction on the evolution path of foreign capital inflows, as the lack of outline the existence of some unifying elements which, while preserving the
this restriction may generate a path of foreign debt unsustainable in the long diversity of ideas and analyses, reduces the risk of interpreting the Keynesian
run. According to Moreno Brid (1998–99), international credit institutions literature as a disorganised set.
impose on developing countries borrowing restrictions based on some index
of their expected ability to repay the foreign loans. He therefore proposes a
different specification for the balance-of-payments constraint based on the NOTES
requirement of a constant ratio between the current account deficit and the
GDP, interpreted as a measure of a country’s creditworthiness. When this 1. The model proposed by Solow (1956) describes the neoclassical theory of growth. For the
restriction is added to the model, the dynamic foreign trade multiplier may classical tradition one can refer to the analyses proposed by Pasinetti (1960) and by
assume a value higher or lower than the standard one, depending on the Samuelson (1978). The analyses presented by Barro and Sala-i-Martin (1995) give the main
initial current account position of the country concerned. This revision has elements of the New Growth Theories.
2. This multiplicity of ideas and analyses is, according to some authors (e.g. Dow, 1985;
considerable implications for empirical analysis, clarifying that estimates of Hamouda and Harcourt, 1989; and Chick, 1995), a great merit of the Keynesian literature,
the εx/εm ratio may be significantly biased if they do not take into account the since it adds to the richness of this line of thought.
countries’ initial export/import ratio. 3. See Rochon (1999, pp. 64–9) for a collection of these criticisms against Keynesian
To sum up, the balance-of-payments constraint approach provides some economics, raised by authors like Solow, Backhouse, Dornbusch, Fisher, Felderer and
important insights into the analysis of the relationship between external Homburg.
4. ‘During the twenties many of us were deeply interested in Keynes’s advocacy of measures
demand and growth. While on theoretical grounds the relevance of the to promote fuller employment’ (Harrod, 1967, p. 316).
cumulative causation mechanism embodied in the model (63)–(66) cannot be 5. Harrod (1951, ch. IX, par. 3) recalls however Keynes’s article in the Nation, May 24, where
denied, the empirical evidence seems to show that the simpler formula the Cambridge economist presented for the first time his proposals for public works.
described by equation (73) suffices to capture the main ‘stylised facts’ 6. Phelps Brown (1980, p. 19) points out that since 1932, Harrod wrote several letters to The
relating to growth.50 As the analysis of the factors affecting the εx/εm ratio Times, in favour of Keynes’s proposals.
7. See Harrod (1948, p. 40; 1964, pp. 903 and 905–6). The similarity between Harrod’s and
seems to suggest, however, the balance-of-payments constraint approach Ramsey’s analysis of saving is underlined by Asimakopulos and Weldon (1965, pp. 66).
does not obscure the peculiar role played by the interaction between Harrod (1973, p. 20) also clarifies that ‘what each person chooses in regard to saving is
‘external’ and ‘internal’ factors underlined by Kaldor in his writings. governed by various institutional arrangements, which differ from country to country and
from time to time. There is the question of what the State will provide for future
contingencies – old age, ill health, unemployment, etc. – by current transfer payments as and
when they arise. The more ground that the State covers, the less will the individual feel it
6.6. CONCLUSIONS incumbent to provide for himself by saving. Personal saving will also be affected by the
degree the education of one’s children is subvented by the public authorities’.
Harrod’s seminal work on growth theory was conceived as an attempt to 8. According to Harrod (1939, p. 264), the warranted rate is the rate that, if it occurs, leaves
extend Keynes’s analysis. It moved from the Keynesian ideas that the producers satisfied, in the sense that for them ‘stock in hand and equipment available will be
exactly at the level they would wish to have them’.
economic system does not tend necessarily to full employment and that
9. Harrod (1948, p. 83) points out that his analysis of the warranted rate assumes the rate of
aggregate demand may affect the rate of growth of the economy. In interest constant. He referred to the realism of Keynes’s view on the behaviour of the
subsequent years, Keynesian economists developed this approach along interest rate (pp. 64–5), agreeing that this rate may be rigid (pp. 56–7) and unable to
several lines, focussing on the different components of aggregate demand decrease in such a way as to lead to full employment (pp. 70–1; 83–4; 97; 99).
and on their role in the growth process, by using several descriptive and 10. See Harrod (1939, pp. 258, 259 and 276). On page 276, in particular, Harrod explicitly
referred to an inverse relationship between k and r. In the Thirties the neoclassical
analytical methods. As stated above, this multiplicity of ideas and analyses
assumption of decreasing marginal returns was generally accepted. Sraffa’s critique of the
shows, according to some authors, the fertility of this line of thought. neoclassical theory of capital had not yet been elaborated. (See Panico, 1998, p. 177, fn. 55;
Conversely, an external observer may judge the lack of a unified framework and Panico, 2001, pp. 300 and 308–9 fn. 59, 60 and 61). As is well known, it was published
a weakness, considering the Keynesian literature a disorderly set. By in 1960 and discussed at length in the following decade.
11. Dealing with his analysis of the equilibrium warranted path, Harrod claimed: ‘I know of no 21. Kaldor (1955–56) and Pasinetti (1962), instead, assume that investment is exogenous. Their
alternative formulation, in the world of modern economic theory, of any dynamic principle models are characterised by full employment. According to some authors this assumption
of comparable generality. We must start with some generality however imperfect. We shall cannot be considered Keynesian (see Marglin, 1984a, p. 533–4 and Kurz, 1991, p. 422). In
never go ahead if we remain in a world of trivialities or fine points. It is useless to refine and section 3 above, however, we have pointed out that for Kaldor, full employment growth can
refine when there are no basic ideas present at all’ (Harrod, 1948, pp. 80–1). be achieved through suitable policy interventions. In the absence of government
12. As to the ‘knife-edge problem’ Harrod stated: ‘Nothing that I have ever written (or said) interventions, the economy does not necessarily grow at the full employment rate. Pasinetti,
justifies this description of my view’ (Harrod, 1973, p. 31; but see also pp. 31–45). on the other hand, explicitly investigates the conditions of steady growth at full
13. See Harrod (1948, pp. 132–3, 137–8 and 144; 1960, pp. 278–9, 283 and 285; 1964, pp. 910– employment. For a survey of the subsequent developments of the neo-Keynesian theory, see
13; 1973, pp. 68, 78, 80, 102). It should be noted too that, after 1960, Harrod thought that Baranzini (1991) and Panico and Salvadori (1993).
the major influence of the interest rate on investment is through the availability of finance, 22. The introduction of a non-linear form for expression (22) could generate multiple solutions,
owing to the fact that the credit markets are imperfect (information are asymmetrically some of them unstable. This is the case of Joan Robinson’s (1962) well-known ‘banana
distributed) and tend to react to the shortage or availability of credit (see Harrod, 1960, diagram’ which gives rise to two equilibria, one stable and one unstable.
pp. 278–9 and 292; 1964, pp. 912–13; 1973, pp. 44, 61, 179). 23. Marglin (1984a, 1984b) solved this type of overdetermination by introducing in the analysis
14. ‘Sustained low interest will presumably in the long run reduce the normal profit rate’ a new variable, the rate of inflation, depending on the discrepancy between s and i.
(Harrod, 1973, p. 111). And again: ‘If the market rate of interest rises considerably and stays According to this author, ‘equilibrium can be characterised in terms of investment, saving,
up for a substantial period, ... that may cause firms to increase the mark-up’ (p. 44; see also, and conventional wages, but to do so we must abandon the static characterisation of
p. 78). equilibrium in favour of a dynamic one. Using the disequilibrium dynamics of the two
15. Harrod (1964, p. 908) gave a somewhat different account of this point: ‘In the concluding systems, we can synthesise Marxian and Keynesian insights into a just-determined model in
pages of my first “Essay” I did recognise that there were two distinct problems of policy, which investment, saving, and the conventional wage jointly determine equilibrium’
namely: (i) the short-term one of preventing deviations from a steady growth rate, and (ii) (Marglin, 1984b, pp. 129–30).
the long-term one of bringing the warranted rate into line with the natural growth rate. I 24. Dutt (1987; 1990) presented a more refined resolution mechanism of conflicting claims
recognised that, if the warranted rate was not equal to the natural rate – and there is no between firms and workers which could generate a value of the wage rate between wπ and
reason why it should be – difficulties would inevitably arise. Thus, policy was required to w ω.
bring them together. My remarks on this subject were admittedly very sketchy. I suggested 25. In the Kaleckian literature these coefficients are not univocally interpreted. According to
that the long-term interest rate might be used to make the warranted rate adhere more closely Dutt (1984, p. 28), γ 0 and γ 1 accounts for the (constant) entrepreneurs’ desired degree of
to the natural rate, while “public works” (nowadays “fiscal policy”) and the short-term rate capacity utilisation. Lavoie (1992, 1995), instead, interpreted γ 0 as firms’ expected rate of
of interest should be used to deal with short-term deviations. All this was very loose. The growth of sales, which is not necessarily constant.
existence of the double problem was, however, recognised’. 26. These are the same reasons invoked by Joan Robinson (1962). See above.
16. Some recent contributions to the New Growth Theories consider, instead, the influence of 27. According to Steindl (1952), firms plan a reserve of excess capacity facing uncertainty. This
Government intervention on growth, be it a change in taxation or in expenditure, through its is to avoid the permanent loss of market share owing to the temporary inability to fulfil
effect on the propensity to save and on the capital–output ratio (see Barro, 1990). unexpected demand. Other reasons, invoked by the literature to justify firms’ planned excess
17. In his seminal contribution Kaldor (1955–56, p. 98) explicitly recognised the need to deal capacity, are: (i) seasonal fluctuations of demand; (ii) expected growth in demand; (iii)
with the State in the analysis of steady growth conditions. Yet, like other authors, he failed costly use of overtime work and night shifts or shifts involving unordinary hours or days;
to do so in most of his later work. (vi) indivisibility of plants and equipment. For a short review on this argument, see Lavoie
18. According to Kaldor (1958, pp. 136–7), the drawback of this solution is that in times of (1992, pp. 124–6).
inadequate demand the Government gradually transforms the economy into one of high 28. The normal degree of capacity utilisation, un, is ‘the degree of utilisation of capacity desired
consumption and low investment, with the undesirable consequences on long-run growth, by entrepreneurs, and on which, therefore, they base their investment decisions about the
which will be described in Section 6 below. size of a new plant relative to the output they expect to produce’ (Garegnani, 1992, p. 55).
19. In this debate, Pasinetti (1989a; 1989b) and Dalziel (1989; 1991a,b; 1991–92) examine the 29. In particular, income distribution can be determined either by referring to some
validity of the Cambridge equation by introducing into the analysis the Ricardian ‘conventional standard of life’, which affects the wage rate, or, alternatively, by the level of
debt/taxation equivalence. Denicolò and Matteuzzi (1990) and Panico (1993, 1997, 1999) the money interest rates, which affects the rate of profits, as suggested by Sraffa (1960,
consider the same topic by introducing into the analysis the existence of financial assets p. 33) and envisaged by Vianello (1996).
issued by the Government. Commendatore (1994, 1999a), instead, compares the limits of 30. On the absence of an adjusting mechanism between u and un, Committeri warned that if ‘the
validity of the dual and the Pasinetti theorem. “equilibrium” utilisation degree does not coincide with its normal level, and hence
20. Denicolò and Matteuzzi (1990) deal with the so-called ‘personal’ version of the post producers’ expectations are not being confirmed by experience … as the economy moves
Keynesian theory of growth and distribution. It may be noted, however, that the debate has away from the steady path, the model has nothing to say about the long-run tendencies of
considered different versions of the post Keynesian theory of growth and distribution: the capital accumulation’ (Committeri, 1986, p. 175). See also Ciampalini and Vianello (2000).
personal version, in terms of classes, the functional version, in terms of income groups, and 31. According to Garegnani (1992, p. 59), ‘the entrepreneurs will certainly attempt to bring
the institutional version, in terms of sectors of the economy (see Panico, 1997 and about, through investment, a capacity which can be used at the desired level. And the degree
Commendatore, 1999a, 1999b). of their success will depend on how well they will be able to forecast the outputs which it
will be convenient for them to produce. But given the initial arbitrary level of capacity that 41. See Dixon and Thirlwall (1975, p. 209). Notice that, on the contrary, Kaldor (1966, p. 147)
success will show only in shifting, so to speak, backward in time the deviation of the assumes that the differences in the rate of change of money wages of different regions do not
utilization of capacity from the desired level. Even correct foresight of future output will not counter-balance the reduction in costs due to the different rate of change of productivity.
eliminate average utilization of capacity at levels other than the desired one’. 42. To empirically estimate the influence of the composition of demand on productivity, Kaldor
32. Neo-Ricardians consider the normal rate of profits, rn, a more suitable variable than the (1966) also used an expression, which differs from our equation (66) only in introducing, as
current rate of profits, r, to capture the role of expected profitability in investment decisions. an additional variable, the ratio of investment to output. His analysis showed that this
See on this point Vianello (1996, p. 114). variable explained the divergence of the rate of change of productivity from the trend
33. We assume, for simplicity, that normal and full capacity utilisation coincide, un = 1. determined by the original equation (66). It explains the residual change in productivity, not
34. The independent component of aggregate demand can come from any sector of the explained by increasing returns.
economy. Notice that this analysis only shows that effective demand can affect the 43. In the subsequent years, Kaldor changed this position too: ‘In this respect I now feel I was
adjustment path towards equilibrium even if along this path u = 1 (see Park, 2000, pp. 11–16 mistaken. Events since 1971 have shown that the exchange rate is neither as easy to
and Barbosa-Filho, 2000, p. 31). As to the conclusion that equilibrium growth is governed manipulate nor as rewarding in its effect on the rate of growth of net exports as I have
by capacity saving, Park (2000, p. 8) and Barbosa-Filho (2000, p. 31) showed the existence thought’ (Kaldor, 1978a, p. XXV).
of two solutions of this analysis. The first, which is locally stable, confirms that growth is 44. Let Y = D + X, where Y is income, D is the demand for domestic products and X is exports.
governed by capacity saving. The second, which is unstable, implies that income grows at By definition γ = ω x ( d D + d X ) / d X , where ωx is the ratio of exports to income. Since
the same rate as the independent component of demand, if the latter has certain properties. ( d D / d X ) ω x = ( d D / d Y ) ( d Y / d X ) ω x = ( d D / d Y ) γ and ωx = 1 – ωD, we can write
35. For an analysis of Kaldor’s views on growth and cumulative causation, see Thirlwall (1987) γ = ( 1 – ω D ) / ( 1 – d D / d Y ) . Finally, from the definition of the income elasticity of
and Ricoy (1987; 1998). They describe several aspects of Kaldor’s position, including the demand for domestic products ε D , we get γ = ( 1 – ω D ) / ( 1 – ω D ε D ) .
role of technical progress and structural change, and his idea of growth as a path-dependent 45. This view was already presented in Kaldor (1958), as stated in Section 3 above.
process. In what follows, we mainly focus on the role of demand in the growth process, 46. See Wilson (1976), Ball, Burns and Laury (1977). Long-run stability in the terms of trade
paying less attention to other equally relevant aspects of his vision of the topic. may alternatively rely either on arbitrage or on wage-resistance forcing domestic prices to
36. This is the literature that moves from the contributions of Nelson and Winter (1974, 1977, move equiproportionately to exchange rate depreciations so that pˆ − pˆ f − eˆ = 0 (Thirlwall,
1982), examined by Santangelo’s essay in this volume. 1979, p. 283).
37. In 1972 Kaldor further integrated Young’s analysis with the Keynesian principle of effective 47. According to McCombie and Thirlwall (1994, 233), there are a number of possible
demand, examining the role played by the demand for investment and focusing on the mechanisms through which capacity growth may adjust to demand growth: ‘the
conditions allowing self-sustained growth. In this contribution, he argued that growth is a encouragement to invest which would augment the capital stock and bring with it
fragile process. In order to work it requires that several things simultaneously occur: technological progress; the supply of labour may increase by the entry of the workforce of
investors must have confidence in the expansion of the markets; the credit and financial people previously outside or from abroad; the movement of factors of production from low
sectors have to accommodate the needs of trade; the distributive sector has to bring about productivity to high productivity sectors, and the ability to import more may increase
price stability. According to Kaldor, after the 1930s, Government intervention secured the capacity by making domestic resources more productive’. On this point, see also Thirlwall
smooth working of the process by demand-management policies (Kaldor, 1972, p. 1252). (1986, pp. 48–9) and McCombie (1998, pp. 238–9).
38. As stated above, Kaldor borrowed this relationship from Hicks’ super-multiplier. Following 48. See McCombie (1993, p. 481), who quotes extensive empirical evidence showing that
standard notation, I + X = S + M is the commodity market equilibrium condition for an open income elasticities are not related to the differing product mixes of the exports of the various
economy without public sector. If we assume that S = sY, I = κY and M = μY, the countries.
equilibrium level of income is given by Y = αX, where α = 1/(s – κ + μ) is Hicks’ super- 49. It is worth noting that alternative ways of interpreting the foreign trade multipliers may lead
multiplier. In terms of rates of change, we get g = α ( X / Y ) xˆ . Since α = dY/dX and, by to less pessimistic conclusions. Bairam (1993), for example, shows the existence of a
definition, γ = (dY/dX) (X/Y), the rate of change of income simply reduces to (63). statistically significant inverse relationship between the εx/εm ratio and the stage of economic
39. Dixon and Thirlwall (1975, pp. 208–10) point out that a0 is determined by the autonomous development of the country, proxied by per-capita output. Such a relationship implies that
rate of disembodied technical progress, by the autonomous rate of capital accumulation per developing countries are less balance-of-payments constrained than developed countries,
worker and the extent to which technical progress is embodied in capital accumulation. λ is and therefore provides some support for the ‘catching-up’ hypothesis: if developing
instead determined by the induced rate of disembodied technical progress, by the degree to countries are able to grow quicker than developed ones, GDP levels will inevitably converge
which capital accumulation is induced by growth and the extent to which technical progress in the long-run.
is embodied in capital accumulation. 50. See McCombie and Thirlwall (1994, 434). Kaldor himself (1981, p. 602) admitted the utility
40. The stability condition of the model is⏐γηxλ⏐ < 1, which, in their opinion (1975, p. 208), of the simplified model. In the same essay, Kaldor assumed that the sum of the marginal
may be plausibly assumed to hold. As a consequence, since ηx < 0, in equation (5.9) g is propensities to consume and invest is equal to unity. This assumption transforms Hicks’
related positively to γ, a0, pˆ f , ê , εx, gf and λ, and negatively to ŵ and πˆ . The effects of supermultiplier into Harrod’s multiplier. If we also assume ηx = 0, equation (67) collapses to
variations of ηx are not determined. Notice too that recently Setterfield (1997) has presented the dynamic foreign trade multiplier. Note that the assumption c + κ = 1 has also been used
an analysis, similar to that of Dixon and Thirlwall (1975), in order to study the movements in the Cambridge Economic Policy Group model. On this point see also Targetti (1991).
of the economy out of equilibrium.
important influence on the growth rate of per capita output. Then another
possible cause appears of differences among nations in the growth rate of per
capita output: the different growth rate of aggregate demand. Sections 7.4
and 7.5 survey the main reasons to question the belief that the full
employment of resources, or the theory that the economy gravitates around a
7. Should the theory of endogenous NAIRU, are good starting points for the theory of growth. Section 7.4
growth be based on Say’s law and the discusses the relevance of the speed with which the economy converges, if at
all, toward the full employment of resources, and surveys some criticisms of
full employment of resources? the supposed tendency of the economy toward a NAIRU. Section 7.5
questions the ability of the rate of interest to bring investment into line with
Fabio Petri savings.2
7.1. INTRODUCTION* 7.2. THE POSSIBLE ROLE OF DEMAND IN OUTPUT

GROWTH
In the writings of the modern Endogenous (or New) Growth theory the term
‘growth’ is used to mean growth of per capita output, with a striking 7.2.1. Textbooks on growth theory such as those by Barro and Sala-i-Martin
definitional change relative to a couple of decades ago. This change reflects (1995) or Charles Jones (1998) present only one theory of long-run output
the belief that the theory of long-run aggregate output growth is a solved growth: investment is determined by the savings corresponding to the full
problem, there being no doubt that long-run output growth is, with sufficient employment of resources. This is also the growth theory in many
approximation, determined by the reinvestment of full-employment savings intermediate macroeconomics textbooks. Probably this is the sole view on
(or of the savings associated with a natural rate of unemployment or output growth to which many of to-day’s young economists have ever been
NAIRU).1 introduced.
I will argue that, on the contrary, the problem of the determinants of the I therefore start by indicating why and how a different approach,
growth of aggregate output is far from solved, because there are good recognizing an independent role of aggregate demand, is possible. A good
reasons to assign aggregate demand an extremely important autonomous way to do this is by commenting upon an interesting passage in a recent
role; and that Endogenous Growth theorists ought seriously to reconsider this paper by Robert Solow (1997) on the state of macroeconomics. Solow, after
question, because it may have relevant consequences also for the growth of reaffirming his faith in full-employment models (e.g. his own model) for the
per capita output, or more precisely, of labour productivity. study of long-run growth, admits that in the short run there may be
The structure of the paper is as follows. Section 7.2 delineates the deviations from full employment, and then continues:
premises for an alternative theory which makes output growth depend on the
evolution of aggregate demand. This is done in the form of a criticism of an One major weakness in the core of macroeconomics as I have represented it is the
lack of real coupling between the short-run picture and the long-run picture. ... A
opinion to the contrary recently expressed by Robert Solow. A numerical
more interesting question is whether a major episode in the growth of potential
example is presented to make the point that the flexibility of capacity
output can be driven from the demand side. Can demand create its own supply?
utilization and the adaptability of labour supply make the growth both of The magnitudes suggest that it would be awfully difficult for a surge of aggregate
output and of productive capacity depend on the growth of [the autonomous demand to generate enough investment to provide the capacity necessary to
components of] aggregate demand. Section 7.3 then argues that the faster is accommodate it. In special circumstances it might be done, say, in an economy
the growth of output, the greater is the effect of all the forces which influence that has a pool of labour (rural, foreign) that it can mobilize. It might also work if
the growth of labour productivity in Endogenous Growth models; therefore strong aggregate demand can induce a rise in total factor productivity... The
the moment the full-employment, normal-capacity-utilization assumption is demand-driven growth story sounds quite implausible to me under current
dropped and a possible autonomy of the evolution of aggregate demand is conditions: but it is an example of the kind of questions that needs to be asked
admitted, the growth rate of aggregate demand must be admitted to be an (Solow, 1997, pp. 231–2).
139
Endogenous growth, Say’s Law and the full employment of resources 141 142 The Theory of Economic Growth: a ‘Classical’ Perspective
Note that the possibility of an autonomous influence of aggregate demand on single industries, but also for entire sectors, in particular, for the capital-
long-run growth is admitted; the scepticism as to its plausibility appears goods-producing sector, and − the moment one admits (see below) that the
empirically motivated: but the reasons remain vague; nor, I will argue, can supply of labour is not usually fully utilised and can usually be increased (in
persuasive reasons be found. the short period perhaps by overtime work) − also for the entire economy.
First of all, Solow’s restriction of the problem only to ‘a surge of (Obviously the possibility of accelerating the rate of production is due to the
aggregate demand’ is unwarranted. Growth theory must also explain existence of inventories of intermediate goods, inventories which will be
episodes of lasting slowdowns of aggregate demand, and for these it seems initially run down but will be then rapidly reconstituted by the increase itself
clear that there is no impediment to a lengthy economic crisis discouraging in production.)
investment and thus decreasing capacity to the lower level appropriate to the But if the level of production is so variable, then − concentrating now on
lower level of demand (and this by itself suggests a potentially very the capital-goods sector, i.e. the sector whose production creates productive
important role of aggregate demand). But Solow’s thesis that ‘it would be capacity − the production of productive capacity, and therefore the rate of
awfully difficult for a surge of aggregate demand to generate enough growth of productive capacity, must be considered determined by demand,
investment to provide the capacity necessary to accommodate it’ appears to so the evolution over time of the overall productive capacity of an economy
be very often wrong also for accelerations of aggregate demand. I show this must be considered, in an analysis of growth, the result of demand, rather
with a numerical example, crude but I think sufficient to make the point. than a determinant of production.
7.2.2. As a preliminary to the example, let us note that an aspect of the 7.2.3. To make such statements more concrete and also stress the cumulative
functioning of industrial economies, which appears to be forgotten or effects of a higher or lower aggregate demand, let us use a numerical
underplayed in much current growth theory, is the adaptability of productive example. This is based on a multiplier–accelerator model of a closed
capacity to demand, due to the variability of the degree of utilization of economy with public expenditure and a balanced budget (G = T), and with
productive capacity (Garegnani and Palumbo, 1998; Trezzini, 1995). depreciation and a distinction between gross and net investment.4 The actual
Firms want to maintain spare capacity because that allows them to meet capital stock is K, the desired capital stock is K*. Net investment IN in each
expected, or possible, demand fluctuations, or expected or possible growths period t is performed to bring the capital stock at the beginning of the period,
of demand; and because very high rates of utilization cause higher average Kt, to its desired level by the end of the period, i.e. at the beginning of the
costs since overtime-labour and night-labour are more expensive.3 The first next period, Kt*+1 . The desired capital–output ratio Kt* Yt is 1, and firms are
group of reasons for spare capacity means that firms will produce more at no assumed to be myopic or very prudent and, when deciding investment for
extra average cost if only demand is, on average, higher than expected. Even period t, to expect for period t + 1 the level of gross output just observed, i.e.
the last reason is usually not sufficient to prevent an increase in production if Yt–1. Thus Kt*+1 = Yt −1 and net investment is governed by I N t = Y t – 1 – K t , while
the demand for the product increases, even if the price of the product remains gross investment It is equal to net investment plus depreciation: the latter is
constant, because in imperfectly competitive markets (the most frequent assumed to be 10% of Kt. Investment plans are realized and therefore Kt = Yt–2.
market form) the fear of losing market shares will make a higher degree of Consumption is equal, with a one-period lag, to 8/9 of after-tax income, and
capacity utilization than the long-period optimal one become convenient in the state is assumed to be able every period to tax for an amount exactly
the short run – and this without any redistribution of income away from equal to public expenditure, so Ct = (8/9)(Yt–1 – Gt–1). Output is assumed to
labour: the opposite is on the contrary the more likely case because of the adjust very rapidly to demand so that in every period Y = C + I + G. The
increased share of overtime wages. autonomous role in aggregate demand is taken by public expenditure G. The
Such variability in capacity utilization is of course what makes the economy is assumed to be initially stationary, with Y = 1000, K = K* = 1000,
production of goods capable of that rapid adaptation to changes in the level I = 100, IN = 0, G = 100, C = 800. Initially a zero growth rate is Harrod’s
or composition of demand, which is usually observed in reality. This warranted growth rate. Then from period 0 onwards public expenditure starts
confirms that the variability of capacity utilization is considerable not only increasing at a rate of 2% per period. The following table describes the initial
th th
downwards – what nobody would deny – but also, up to limits rarely evolution of the economy and the 15 and 16 periods:
reached, upwards. The implication of this variability is that the level of
production is quite variable in response to variations in demand, not only for
t Gt K t + 1 = Yt − 1
*
Kt = Kt
* deprec.= IN I (gross) Ct Yt Yt – Gt because of the likelihood of greater fluctuations of demand in the capital
0,1⋅Kt goods industry (the accelerator!), this industry can be presumed to be
particularly well prepared to adapt production to wide fluctuations of
–2 100 1000 1000 100 0 100 800 1000 900
demand.
–1 100 1000 1000 100 0 100 800 1000 900
The reasoning is just as applicable, or even more applicable, to the case
0 102 1000 1000 100 0 100 800 1002 900
where aggregate demand and production were not initially stationary, but
1 104.04 1002 1000 100 2 102 800 1006.04 902
were growing at a positive rate g and this rate then becomes g+2%. Even
2 106.12 1006.04 1002 100.2 4.04 104.06 801.78 1011.96 905.84
more applicable, because the positive growth rate g justifies the existence of
3 108.24 1011.96 1006.04 100.60 5.92 106.52 805.19 1019.95 911.71
planned underutilised capacity which would not hold in a stationary
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
economy: the construction of bigger plants than necessary at the moment of
15 137.28 1231.15 1203.97 120.40 27.18 147.58 974.72 1259.58 1122.30
completion, in the expectation of future increases of demand.
16 140.02 1259.58 1231.15 123.12 28.43 151.55 997.60 1289.17 1149.15
Thus the assumption that, starting from a situation of normal capacity
utilization, a faster growth of the autonomous components of aggregate
The growth of public expenditure (with a balanced budget) stimulates a demand (analogous examples might be construed for increases of exports, or
growth of Y which induces a growth of I at a rate which grows up to about of state-controlled investment in nationalized firms) will bring about a 2%-
2.8% (in periods 11 to 14) and then tends slowly (with damped oscillations) faster growth rate of output and of average productive capacity appears to
toward 2%, with an associated growth of Y and of K which also becomes meet no obstacle in the existing productive capacity.7 And an increase of two
initially higher than 2% and then tends toward 2% (it is about 2.3% in period points in the growth rate is quite a considerable increase: already a one-point
16); thus after 15 periods the capital stock has grown by over 20%.5 increase would be considered a great success by most nations.
Very importantly, the example shows that in order to achieve this result A simple reversal of the reasoning (assuming that the growth rate of the
there is no need for decreases either of consumption, or of the average autonomous components of aggregate demand becomes 2% lower) shows
propensity to consume: the increase of the rate of growth of capital without that an insufficient growth rate of demand may be responsible, after a few
decrease of the average propensity to consume is made possible by the years, for a very considerable loss of potential productive capacity, a loss
increase of Y/K, which makes it possible to increase I /K while also easily resulting in structural unemployment, but otherwise not easily
increasing C. perceptible, as it is not visible.
The example shows that there is little reason why this growth process The historical observation that productive capacity is not greatly
should run against bottlenecks. The average Y/K ratio remains all the time underutilized for very long periods can then be explained as due to the fact
close to 1, arriving at most at about 1.05, which means on average a capacity that, if productive capacity is excessive relative to demand, then net
utilization only 5% greater, something easily obtainable in most situations. 6 investment decreases and may become negative, the older plants are closed
The utilization rate in the capital-goods industry becomes initially greater down, and productive capacity shrinks or increases at a slower rate than
than that, of course (because when gross investment increases, it increases demand, thus tending to adapt to demand.
initially percentwise more than Y); but a more detailed example, separating Except in the event of an increase of the growth rate of aggregate demand
sectors and including hypotheses on the forces affecting the allocation of of more than two percentage points, or of an initial situation where capacity
investment among sectors, would have been necessary in order to estimate is already utilized more than normally, it seems therefore impossible to agree
by how much. What one can say on the basis of this example is that, before with Solow on the presumed difficulty demand would have in creating its
the growth process begins, the capital goods industry (assuming there too the own supply; the very rapid growth of countries like Korea becomes more
capital–output ratio to be 1) employs 1/10 of the capital stock, and that if this easily understandable. For the opposite cases of a prolonged slowdown of the
ratio remained the same afterwards, the utilization ratio in the capital goods growth of aggregate demand, there can be no doubt as to the (cumulative!)
industry would rise by 20%; but one can be certain that in fact it would rise effect on the loss of potential capital accumulation.
much less, because the higher capacity utilization will induce net investment
to go in greater proportion toward the capital goods industries, whose 7.2.4. It might be objected that my example discusses only the adaptation of
productive capacity will therefore increase faster. And anyway, precisely capital, while − strangely enough − Solow appears to have in mind above all
a difficulty with finding additional labour. I say strangely enough, because now to show that this different approach to output growth is also relevant for
for an economist like Solow who accepts the neoclassical theory of the question of the growth rate of per capita output or labour productivity.
distribution with the associated conception of capital–labour substitutability,
a given level of employment ought not to be an obstacle to an acceleration of
capital accumulation. The latter ought only to entail a gradual increase of the 7.3. ENDOGENOUS GROWTH WITHOUT THE FULL
ratio K/L .8 EMPLOYMENT OF RESOURCES
But even in a non-neoclassical approach, where real wages were
considered given and thus also relative prices and technology were 7.3.1. If we look at the reasons which in the recent models of endogenous
essentially given (as implicit in the numerical example presented above), and growth make indefinite growth of labour productivity possible even in the
where therefore a necessity would arise to increase employment if output absence of exogenous technical progress, we can see that the force of their
growth is faster than the growth of labour productivity, Solow’s scepticism action is influenced by the rate of growth of total output; therefore, whether
appears unwarranted in most cases. In the short period, there is nearly or not this rate is determined by the investment of full-employment savings
universal agreement on the fact that official unemployment rates, besides can make a significant difference to the rate of growth of per capita output.
being always positive, hide the presence of hidden unemployment, and that One of these reasons can be left aside as unacceptable, namely that
the rate of participation increases if labour demand increases; this is indeed proposed by Jones and Manuelli (1990). This is the hypothesis of a CES
part of the accepted explanation of Okun’s Law; furthermore, employed aggregate production function, which has the property that the marginal
workers generally do not object to temporary periods of overtime work and product of a factor is bounded above a strictly positive value. It had already
wages. If one turns to the ample time intervals relevant for the theory of been noticed as a possibility by Solow (1956) but then ignored in subsequent
long-run economic growth, historically one observes clear signs of a literature, evidently because it was judged implausible. Indeed, even
(spontaneous or engineered) tendency of the supply of labour to adapt to the neglecting for the sake of argument the illegitimacy of the treatment of
demand for labour, so that capitalist economies seem to have always been capital as a single factor, the hypothesis that the marginal product of capital
able to avoid a labour supply constraint. Agricultural underemployment, never goes below a strictly positive value even when labour employment
other pre-capitalist sectors, domestic labour have historically supplied the goes to zero appears to have no economically acceptable interpretation.
labour reserves necessary for the industrial revolution and for subsequent To summarise, the other reasons used to endogenise the growth of per
growth when population growth was not enough. When that was insufficient capita ouput are:
(as in post-war Germany, or in the USA), there were huge and carefully
regulated immigration flows, or sometimes policies promoting fertility (or, in • investment in the production of new knowledge
the opposite case of excessive population, combating fertility, as now in • investment in the production of human capital
China). It would seem therefore that historically for labour supply too, it is • externalities connected with the expansion of production, or with
true that in the long run demand has created − spontaneously or through production itself (e.g. learning by doing)
policy interventions − its own supply. The present pressure for immigration • division of labour (increase of specialisation) producing scale economies.
from poorer countries toward the industrialized ones suggests that there is
little obstacle to the same being true now and in the foreseeable future. Let us then examine the influence of the output growth rate on such causes of
Furthermore the policy question behind all this, i.e. whether the state per-capita output growth.
should or not be assigned a relevant role of regulation of aggregate demand,
becomes relevant above all in the periods of recession or crisis, when there is 7.3.2. Let us start by considering learning-by-doing à la Arrow (1962). If
a greater than usual excess of labour supply. one is to leave a role to aggregate demand, then as stated above the full
This section was intended to illustrate why long-run output growth may employment of labour should not be assumed (variability in labour
be viewed as determined by the evolution of aggregate demand, with labour employment in the short period is a prerequisite for variability in capacity
supply and capital accumulation adapting to aggregate demand, rather than utilization in response to variations in aggregate demand). Should one then
determining it via the investment of full-employment savings.9 I proceed assume a relationship between employment and the real wage? Keynes’s
approach was to assume a decreasing labour demand curve even in the short
period, thus maintaining a univocal decreasing relationship between real same approach may also be viewed as one possible simple way to formalize
wage and Y. But it has been repeatedly noted (e.g. Zenezini, 1990; a positive influence on labour productivity of the scale of production, and
Brandolini, 1995)10 that empirical evidence does not accord with this therefore of externalities or increasing returns connected with ‘the size of the
construction for the short period. The Cambridge critique of neoclassical market’. One sees at a glance − what was intuitive − that a faster growth of
capital theory shows that the labour demand curve is an unacceptable notion output means a faster growth of Y/L.
even for the long period.11 On the other hand, the increasingly widely Romer (1987) tries to capture the increasing-returns effects of an
accepted NAIRU approach shows that economists are prepared to admit the increasing ‘division of labour’ due to expansion of the size of the market, via
role of bargaining strength and other socio-political factors in the a more micro-founded approach. The production function for final output is
determination of the real wage. On this basis, I propose to assume a given the same as in the more often quoted Romer (1990): output is produced by
real wage (assumed now to increase through time in proportion with labour labour and by a number of capital goods of different design, whose
productivity), which determines technical choices, so that it is as if there productivity increases with the increase in the number of designs: the same
were in each period fixed technical coefficients. This has the advantage of stock K of capital, if subdivided among a greater number of different types of
allowing us to dispense with a discussion of what should replace that capital goods, counts as more capital. But rather than the resource–
indefensible theoretical construct, the aggregate production function. But in consuming production of new designs as in Romer (1990), it is the scale of
order not to lose all contact with New Growth models, let us still assume a aggregate output that determines the number of types of capital goods
single output (corn produced by corn and labour), as an approximation to a utilised, owing to U-shaped average cost curves in the production of capital
multi-sector economy where aggregation in value terms is possible as long as goods, which set a limit to the convenience of specialisation. The functioning
relative prices are given. Let us then suppose of the model is such that in the end it is as if final output were produced by a
production function of the usual type Y = Kα(AL) α, where A is the number
1–
Y = AK = BL. (1) of different types of capital goods, so one is formally back to labour-
augmenting technical progress.15 So one should find here too that an
Let us assume, as is often done, that technical progress is of the purely autonomous role of aggregate demand influences the growth rate of Y/L.
labour-augmenting type,12 i.e. that A is not affected by technical progress; Indeed Romer (1987) concludes the analysis of the model as follows: ‘any
Arrow’s learning-by-doing (depending, as is well known, on cumulated change that leads to an increase in savings − for example a tax subsidy, a
experience measured by cumulated investment) can then be formalized in decrease in the rate of impatience ρ, or a decrease in the intertemporal
this approach by assuming that, at the aggregate level, B is an increasing substitution parameter σ − will cause growth to speed up; the rate of
function of the level reached by the accumulation of capital: exogenous technological change will appear to increase’ (p. 62). But the
B = Kβ with β a positive constant. (2) increase in savings brings about this consequence only because it increases
the growth rate of output. Anything else that increases the growth rate of
Hence:13 output will have the same effect.
Y/ L = Kβ. (3) 7.3.4. Let us turn to human capital. In Lucas’s formalization of the role of
So per capita output increases with the growth of the stock of capital.14
It human capital the production of human capital requires no input apart from
is then clear that whatever accelerates the growth of K also accelerates the human capital, and there is no depreciation of human capital, what is not
growth of Y/L. If the growth of capital depends on the growth of aggregate plausible. But these limits of Lucas’s formalization are overcome by Rebelo
demand which at least partially depends on autonomous elements (e.g. (1991):
exports, or state expenditure), then the growth rate of aggregate demand will Y = ψK α(uHL)α
1–
influence the growth rate of Y/L. .

H = (1 – ψ)K1–β [(1 – u)HL]β – dH
7.3.3. Let us now consider the division of labour. Since in the above model it .
was assumed that Y/K was constant, it would have been equivalent to assume K = sKY – dK
that B was an increasing function of Y rather than of K. This means that the
where H is aggregate ‘human capital’.16 Rebelo (1991, pp. 508–10) like very little meaning. Thus even more than for the other arguments I am
Lucas assumes that L is constant, and shows that one obtains again Lucas’s interested here in the basic insight, rather than in the precise formalizations
result, that an increase of the fraction of labour power assigned to the which have appeared in the Endogenous Growth literature. The basic insight
production of human capital increases the growth rate of output and thus, is that advances of knowledge increase labour productivity but require
since labour employment is constant, also the growth rate of labour resources to be produced. Here, again, the role of the full employment
productivity. assumption is to establish a trade-off between producing more output or
Here the role of the full employment assumption is that of making it producing more research; a trade-off which no longer can be presumed to
necessary to choose between more production of current output, and more exist if one admits the generalized presence of unused resources whose level
production of human capital. The admission of the generalized presence of of utilization depends on the evolution of aggregate demand. Then more
unused resources would mean that it is possible to have more of both, or less resources dedicated to research can go together with more output: indeed, an
of both, depending on the evolution of aggregate demand. The increase in increase in output will be the likely effect of more investment in R&D, and
public expenditure considered in the example of Section 7.2.4 might for conversely a greater output, if the share of output dedicated to research is
example be an increase in education expenditure, thus bringing about both a constant, will mean more research.17
higher growth rate of Y and of H. In conclusion, on the basis of their own hypotheses as to the determinants
But even admitting that the allocation of resources to education is of the growth of labour productivity, Endogenous Growth theorists ought to
primarily the fruit of private choices, a positive influence of the evolution of admit that the question, whether one should accept the full (or NAIRU-level)
aggregate demand on educational attainments in all likelihood exists. The employment of resources, is a central question also for the explanation of the
'production' of skills owing to private investment is responsive both to growth rate of labour productivity.
employment opportunities, and to income, and both are influenced by
aggregate demand. The incentive to acquire specializations is higher when
there are good job prospects for them; and a higher level of employment is in 7.4. FULL EMPLOYMENT OR NAIRU? SOME FIRST
all likelihood conducive to more schooling, because more families can afford DOUBTS.
to send their children to school for more years.Also, some training is
performed inside firms, and as employment increases, firms will train a 7.4.1. What was argued in Sections 7.2 and 7.3 only indicates the relevance
greater number of workers. of the question of the determinants of output growth; one might agree, but
Therefore a higher level of employment and of Y, and a higher growth then add that there are good reasons to accept the full or NAIRU
rate of aggregate demand, can be trusted to go together with higher average employment of resources in long-run output growth theory.
levels of education, and hence, if these are believed to increase labour But is that really so? That this is what a majority of macroeconomists
productivity, with a higher labour productivity. appear to think is obviously no guarantee. Science does not proceed on the
basis of majority voting. Let us therefore ask whether the exclusion of a
7.3.5. I turn to investment in the production of knowledge. That the evolution relevant autonomous role of aggregate demand in long-run output growth
of knowledge may be measured by a scalar is of course nonsense; but even theory is really solidly based.
interpreting ‘knowledge’ as simply an index of labour productivity, and even Some doubt is immediately raised by the empirical observation of the
leaving aside the index problems in measuring the evolution of labour numerous historical episodes of output growing, for many years in a row, at
productivity in the face of changes in the quality and kinds of goods a rate in all likelihood far lower than that potentially attainable. Besides the
produced, still the assumption that increases of labour productivity may be Great Depression of the 1930s, one can mention European unemployment
connected by a production function to the resources dedicated to research, since 1974: it is difficult to believe that unemployment rates ten points higher
and furthermore by a production function which remains unchanged for very than a few years before entail no potential for greater output expansion.
long periods, appears at least for the moment to be purely an act of faith Another obvious case is Japan after 1990.18
which goes against the intrinsic unpredictability of discoveries. Therefore
steady growths (or, in more recent literature, constant growths, see e.g. 7.4.2. By themselves, these episodes might be judged insufficient to
Jones, 2002) derived on the basis of this assumption appear to me to have undermine the theory that market economies spontaneously tend to the full
employment of resources: it might be argued that this tendency is not very will be argued in Section 7.4, this assumption founders on both empirical and
fast, or sometimes meets obstacles which slow it down. theoretical grounds. But before getting to this issue, it is opportune to discuss
But once this is admitted, then the exclusion of a role of aggregate the increasingly adopted NAIRU approach.
demand in long-run growth becomes hard to defend: one would have to
argue that the periods of underemployment growth are compensated by 7.4.3. For theories of growth which deny the relevance of autonomous
periods of overemployment growth, so that the trend is sufficiently correctly elements of aggregate demand, the NAIRU approach only implies replacing
approximated by an assumption of normal frictional unemployment only (the the full employment of labour with the NAIRU; and it has two advantages.
practical meaning of full employment); and convincing arguments of this The first is that it does not need a decreasing labour demand curve in order to
type are not easily found. argue that the economy will gravitate around a definite rate of
Let us indeed consider the IS-LM model which, as shown by current unemployment; the second is that it is better able to admit a downward
textbooks, remains at the core of mainstream short-run macroeconomics. The stickiness of wages.
acceptance of this model makes it possible to argue − if the money supply But, to start with, the adaptability of labour supply to labour demand (see
can be considered sufficiently exogenous − that persistent unemployment is Section 7.2.4 above) weakens any attempt to determine a NAIRU growth
ultimately caused by the downward rigidity of money wages, which impedes path independent of autonomous aggregate demand influences. The economy
the working of the so-called ‘Keynes effect’ i.e. the effect of variations of the of the example in Section 7.2.3 would only need – if there were no labour
price level on aggregate demand through their effect on the demand for reserve in the nation – to open up to immigration, and then even accepting
money, hence on the interest rate, hence on investment. The well-known the NAIRU theory there would be no inflation, and on the contrary there
argument is that if money wages decreased in the presence of involuntary would be all the advantages for the growth of labour productivity discussed
unemployment, this would induce a decrease of the price level and then, in Section 7.3.
through the ‘Keynes effect’, an increase of investment and thus of Furthermore, there are empirical and theoretical reasons to doubt the
employment. (The Keynes effect is used against Keynes.) But in order to NAIRU approach, which indeed is encountering considerable opposition
graduate from this conclusion to the idea that Keynes’s analysis is only the even among otherwise mainstream economists. Empirically, a growing
explanation of the fluctuations around a trend which is sufficiently close to number of studies concludes that there is no stable relationship between
that of Solow’s growth model, it is necessary to argue that money wages are inflation and unemployment, and that the notion of a rate of unemployment,
not very rigid, and that the deviations from full employment are not very beyond which inflation would continually accelerate, is contradicted by the
great, and compensated by deviations of opposite sign; and there is nothing econometric evidence (see e.g. Setterfield et al., 1992; Rowley, 1995;
in the model which authorizes such an argument. First, there is nothing in the Galbraith, 1997; Lindbeck and Snower, 1999; Ball, 1999; Solow, 2000). On
IS–LM model which authorizes the thesis that, once a period of the more theoretical side, not only doubts have been advanced on the internal
unemployment caused by a recessionary shock has ended, there will be a consistency and on the stability of the models used to determine the NAIRU
period of overfull employment to compensate for it. Second, the observed (Sawyer, 1997); but also, quite radical general criticisms appear possible.
persistence of levels of unemployment not easily explainable as frictional or The increasing adoption of the NAIRU approach in place of the monetarist
voluntary would rather suggest – within this model – that the downward theory of the tendency to a natural rate of unemployment indicates that
rigidity of money wages is so strong as to maintain the economy far from full economists find it increasingly difficult to explain unemployment (in excess
employment most of the time. Now, if the level of money wages is given, of frictional unemployment) as essentially voluntary (in the sense that the
according to IS–LM theory (as traditionally conceived i.e. coupled, like in unemployed workers are supposedly unwilling to work at the ruling wage),
Keynes, with a decreasing labour demand curve) the real wage (and hence and increasingly admit the need to interpret inflation as mostly cost, rather
employment) is determined by the price level which depends on the level of than demand, inflation and therefore reflecting a distributive conflict. There
aggregate demand; therefore it is the evolution of the latter which is in the is thus an interesting rapprochement between the positions of mainstream
end responsible for the level and rate of growth of output. Furthermore, the and of conflict economists: unemployment is increasingly viewed as
working of the ‘Keynes effect’, which should bring the economy back to full influencing the bargaining strength of wage labour vis-à-vis firms.
employment if money wages were flexible, relies crucially on the assumption But then the undefinability or impermanence of the NAIRU should not
of a significant influence of the interest rate on aggregate investment, and, as come as a surprise, for at least two reasons. First, once it is admitted that
firms are compelled to fix cost-covering prices and that therefore, given the Therefore, if the 'Keynes effect' is judged implausible, the mainstream
other costs, an increase in monetary wages greater than the increase of labour approach to long-run output growth loses all residual credibility, because it
productivity obliges firms to raise prices, the ‘given-other-costs’ clause is loses the mechanism which should ensure Say's Law i.e. the adaptation of
clearly crucial, and most of the times illegitimate. Among the other costs investment to the level of savings deriving from the income associated with
there are: interest rates; the salaries of white collar workers, of managers, of the full or NAIRU level of resource utilization.19
external consultants (e.g. lawyers); taxes; public utilities; the prices of Let us then turn to a discussion of the plausibility of the ‘Keynes effect’.
imported inputs which depend inter alia on the exchange rate. Therefore
there are several degrees of freedom which make it possible that there may
be no need for firms to increase prices when money wages increase, or 7.5. AGAINST SAY’S LAW
contrariwise that firms may have to raise prices in spite of no money wage
increases. (That prices increase when employment increases may then 7.5.1. The ‘Keynes effect’ requires: (a) that changes in money wages
sometimes be due to the fact that, fearing inflation, the central bank raises the influence the price level in the same direction; (b) that changes in the volume
interest rate!) of monetary transactions change the demand for money; (c) that changes in
The second strictly connected reason is that in the relative bargaining the demand for money alter the rate of interest; (d) that changes in the rate of
strength of labour and firms many institutional and political elements play a interest alter aggregate investment in the opposite direction.
part, such as the preferences of government and of the monetary authorities, I have argued that (a) is highly debatable (§3.3). Doubts have been
the degree of unity of labour and of firms in the bargaining process, the advanced on (c) on the basis of 'endogenous money' arguments of various
expectations of trade unions on the effects of their action on employment. types.20 Here I discuss (d).
Thus it is conceivable that trade unions may agree to restrain their demands
for wage increases in exchange for policies promoting employment (this is 7.5.2. Why investment should be a negatively elastic function of the interest
generally agreed to have happened in the so-called ‘neo-corporatist’ rate is something on which there is at present considerable disagreement;
economies). But it is also conceivable that, without considerable variations in increasing numbers of economists are sceptical about the whole idea.
unemployment, a change in the political climate may cause sharp increases It is well known that the empirical evidence is unable convincingly to
of wage demands (as in May 1968 in France, or the so-called Hot Autumn of support the thesis that the rate of interest exerts a significant influence on
1969 in Italy) − or decreases, if e.g. there is a change of government in a investment. This conclusion of older empirical inquiries (see e.g. Junankar,
direction hostile to the labour movement (Pinochet's Chile). So many are the 1972) has not been disproved by later econometric research. The recent
historically-specific elements entering the picture, that it would be indeed survey of investment theory in the Journal of Economic Literature
surprising if great regularity were to be observed in the connection between concludes:
unemployment and inflation.
While there is clearly no uniformity in the results and the role of shocks remains
7.4.5. The theory of a spontaneous tendency towards the NAIRU again to be assessed, it appears to this author that, on balance, the response of
requires the so-called ‘Keynes effect’: a level of output higher than the investment to price variables tends to be small and unimportant relative to quantity
NAIRU level causes – it is argued – an acceleration of inflation (due to variables (Chirinko, 1993, p. 1906; also see Ibidem, pp. 1881, 1883, 1897, 1899).
money wage increases) which increases the demand for money and the rate
Thus the empirical evidence would appear to suggest that the negative
of interest, so investment decreases, and unemployment returns to the
influence of the rate of interest on aggregate investment, if it exists at all, is
NAIRU level. Without the spontaneous tendency towards the NAIRU which
too weak to justify the belief that investment adjusts to savings faster than
the Keynes effect should ensure, the thesis that the long-run growth trend is
does savings to investment via the Keynesian mechanism of variations of
determined by the NAIRU becomes even less credible, because it would
aggregate income. Edmond Malinvaud (1995, pp. 125–7) comes to the same
require government intervention to ensure the tendency to reach the NAIRU,
conclusion.21
and this is hardly credible, in view e.g. of the long historical periods when
It is only natural, then, to suspect that there might be something
concern with unemployment as a regulator of inflation was not found among
unconvincing in the theoretical arguments which support the expectation that
policy-makers.
investment ought to exhibit a negative elasticity vis-à-vis the interest rate.
The suspicion proves to be confirmed by theoretical analysis. For reasons of with reverse capital deepening and with the absence of the right to assume
space I can barely mention the problems; for detailed analyses the interested the full employment of labour.
reader is referred to Ackley (1978, chs 18 and 19), and to Petri (1997). Another derivation of a decreasing investment function, making no
reference to the ‘capital’ intensity of production, relies on Kalecki’s
7.5.3. The traditional support for the theory that investment is a decreasing ‘principle of increasing risk’ (and more recently has come to be called the
function of the interest rate was the belief in capital–labour substitution, with liquidity approach). Kalecki (1937) takes product prices as given and argues
capital treated as a homogeneous factor, an amount of value. As explained by that the level of investment is limited by the rising ‘cost of borrowing’
Garegnani (1990), the traditional demand-for-capital function was in fact the inclusive of risk, because more investment means more borrowing, hence a
demand for a capital/labour ratio in new plants, and it therefore implied a higher leverage, hence a higher risk of default and bankruptcy. This
demand for the (gross) investment necessary to achieve the desired approach is often used to argue a dependence of investment on sales or cash
capital/labour ratio in the employment of the flow of labour ‘freed’ by the flow i.e. on demand, but here we are interested in the implication (not much
closure of the plants reaching the end of their economic life. stressed − but admitted − by Kalecki) that a decrease in the basic rate of
This derivation of the investment function was undermined by the interest will increase retained profits, and thus induce firms to adopt a higher
discovery in the 1960s of the possibility of reverse capital deepening debt/asset ratio, i.e. to borrow more and invest more. This approach, again,
(Garegnani, 1990), which showed that there is no guarantee that the demand forgets that product prices decrease relative to money wages if the rate of
for capital is a regularly decreasing function of the rate of interest. But even interest decreases, and that therefore a decrease of the rate of interest also
before this discovery, the theory of investment was in disarray, because the decreases the rate of return on investment; retained profits do not increase.
traditional derivation needed a well-defined flow of ‘freed’ labour, i.e. Other approaches which have enjoyed or enjoy considerable popularity
needed the full employment of labour, which on the contrary could no longer are Jorgenson’s approach (Jorgenson, 1967) and the adjustment costs
be assumed after the acceptance of the IS–LM foundation of approach (see Söderstrom (1976), Abel (1990) for surveys of the vast
macroeconomics. Hence a number of attempts to derive the decreasing literature), both of which assume a given number of firms – an incredible
investment function without assuming full employment, all of which suffer assumption in aggregate investment theory! – and also make again the
from grave deficiencies. mistake of treating prices as independent of the interest rate, and are
One approach, named ‘array-of-opportunities’ by Gardner Ackley (1978), therefore clearly unacceptable.
argues that entrepreneurs have at each moment in front of them a series of Nowadays Tobin’s q enjoys greater popularity; but the derivation, from
investment projects, which they rank in order of decreasing internal rate of this approach, of a negative dependence of aggregate investment on the rate
return; they then adopt the projects with a rate of return not lower than the of interest rests either on adjustment costs, what has been argued to be
rate of interest; so a lower rate of interest means the adoption of more unacceptable, or on the increasing-supply-price-of-capital-goods approach of
projects, hence a greater aggregate investment. The basic weakness of this Lerner (1944), which is empirically more than doubtful, and which
approach is that it treats the returns from the investment projects as given furthermore, in order to explain why a lower interest rate makes the
independently of the rate of interest, as if prices could be treated as given. On aggregate of firms desire an increase of the capital stock, needs the
the contrary, competitive prices tend to equal costs of production; so if the traditional notion of capital–labour substitution undermined by reverse
rate of interest decreases, prices will tend to decrease too (relative to money capital deepening.
wages) because interest is one of the costs, so the rates of return will tend to In conclusion, the justifications of the view of aggregate investment as a
decrease as well. Furthermore, competition and the tendency of investment to decreasing function of the interest rate either ultimately rely on the
be allocated in greater proportion to the sectors where the rate of return is indefensible traditional marginalist conception of a decreasing demand
greater will tend to annul the differences in rates of return, making all of function for capital the value factor, or are theoretically indefensible even
them equal to the rate of interest (plus a risk allowance). So Ackley harshly aside from the criticisms of marginalist/neoclassical capital theory. In
criticises the ‘array-of-opportunities’ approach and concludes that the reason particular, again and again the mistake recurs, of treating the yields from
why a decrease in the interest rate increases investment can only lie in the investment projects as independent of the level of the interest rate, a mistake
induced increase in the desired capital-labour ratio (1978, p. 625, footnote pointed out also by the more attentive mainstream theorists, e.g. by Ackley
15). However one then runs into the problems mentioned above, connected (1978) and even by Jorgenson (1967, p. 152, quoting Alchian, 1955).
7.5.4. There is therefore no convincing theoretical basis for viewing the 6. U.S. Census Bureau data on capacity utilization for the fourth quarters of 1997 to 1999
interest rate as the price which brings investment into equality with savings. (http://www.census.gov/prod/2001pubs/mqc1-99.pdf) show an average utilization rate for
Say’s Law − the thesis that investment adapts to savings − loses its all manufacturing of respectively 76%, 73% and 74%, which can be compared with an
average utilization rate over the period 1967–99 of about 82% (and ‘popular wisdom’ that
foundation. The faith in the spontaneous tendency of market economies
inflation starts when utilization rates go above 85%). Presumably little difficulty would have
towards the full (or NAIRU) employment of resources must therefore be been found in those quarters with producing even 10% more; the constraint was clearly on
abandoned. the demand side. These are measures of the utilization rate relative to ‘normal’ production,
It is then natural to turn again to the principle of effective demand, i.e. to but firms also indicate a ‘national emergency’ potential production from 40% to 80% higher
the idea that it is variations in income which bring about equality between than this normal production.
7. Other obstacles might be a labour supply constraint (discussed below) or an inflation
savings and investment or more generally between aggregate income and
constraint (see Section 3 below); a balance-of-payments constraint would on the contrary be
aggregate demand. Output and its growth must then be explained through the proof of the influence of aggregate demand on growth.
evolution of the autonomous components of aggregate demand. This will in 8. And for the short run, it is Solow himself who has repeatedly argued that the observed
all likelihood make it easier to reconcile economic theory with historical unemployment is largely involuntary, implying little difficulty with increases of
episodes such as the Great Crisis of the 1930s, or persisting European employment.
9. See Garegnani and Palumbo (1998) for further implications of such an approach and for
unemployment, or the East Asian initial successes and present difficulties.
some remarks on historical evidence supporting it.
The theory of per capita growth will have to recast its insights accordingly. 10. Among older studies, Dunlop’s and Tarshis’ objections to Keynes are well known.
11. The Marshallian long period is that situation in which, owing to the greater speed with
which the relative quantities of different capital goods can change relative to the speed of
accumulation, one can treat the composition of capital as endogenously determined, while
NOTES neglecting the effects of accumulation. But then, how should one specify the given
employment of capital? The given capital stock is indeterminable, because since its
* Funding from the Ministero dell’Università is gratefully acknowledged. I thank Mauro composition must be treated as variable, it can only be measured as an amount of value; and
Caminati and Carlo Panico for their comments. a given value of capital K is an indefensible notion because values change with the real
1. This explains why little attention is given to distinguishing per capita output from labour wage. So it is impossible to determine the labour demand curve.
productivity i.e. output per unit of employed labour. (This is much less acceptable when one 12. This obviously unrealistic hypothesis, like others in this section, is only made for the
admits the possibility of ample variations of unemployment rates, and the possibility of an purpose of departing as little as possible from Endogenous Growth models.
influence of demand on participation rates, but these issues will not be discussed here, and 13. Romer (1986) assumes a knowledge-producing activity of firms, with knowledge spillovers
for brevity’s sake the term ‘per capita output’ will be used below to indicate average labour to other firms, but the final result is, as is well known, extremely similar to Arrow’s and, if
productivity). one assumes fixed coefficients as here in equation (1), then equations (2) and (3) appear also
2. Say’s Law states that investment adapts to savings rather than vice versa; as made clear by to represent Romer’s ideas as to the final result of what in that article he sees as the causes of
e.g. Ricardo, it does not imply the full employment of labour; it does so only in the technical progress (with the difference that Arrow would assume β < 1 while Romer would
neoclassical approach to distribution, which postulates a decreasing demand curve for labour assume β > 1). This may be the place for a comment on Romer’s article. Romer assumes that
and where Say’s Law guarantees that there will not be aggregate demand deficiencies to the consumption good is produced (by a given number of firms) by firm-specific knowledge
create obstacles to increases in employment when, owing to unemployment, real wages k, general knowledge K, and a vector x of other inputs given in amounts; and that each firm
decrease and the demand for labour therefore increases. On the other hand, the Pigou or real- produces private knowledge via the use of its own private knowledge and of forgone
balance effect might bring about the full employment of labour without the validity of Say’s consumption. The details of these assumptions are difficult to justify, and yet no justification
Law. This is why it was found necessary in the title to cite both Say’s Law and the full is offered: for example, why doesn’t general knowledge K also influence the production of
employment of resources. private knowledge? Why aren’t the rentals of the given factors included in the costs? Why is
3. Jacob Steindl has particularly insisted on the first group of reasons, Robin Marris on the there no mention of other factors in the production of knowledge? Why (in a model of long-
second. Another possible reason for excess capacity in oligopolistic industries is entry run growth!) is the number of firms given? Can we measure knowledge quantitatively so as
deterrence. to be able to say that knowledge has e.g. doubled? Should an article be accepted for
4. On the contrary the usual simplification is maintained of giving no role to inventories or to publication when assumptions as dubious as these receive no justification?
their fluctuations. 14. If the growth rate of K is constant, then the growth rate of Y/L is constant, decreasing, or
5. What happens is that Y and K grow at a higher rate than G until the G/Y ratio has sufficiently increasing depending on whether β is equal to, less, or more than 1, but this is of limited
decreased so as to make room for the higher I / Y ratio necessary for a 2% rate of growth interest here; even when β < 1 there is indefinite growth of Y/L, and the reason for wanting
with normal capacity utilization. In this example the multiplier–accelerator interaction does to obtain a constant growth of Y/L is only the determination of steady states, whose
not cause instability. empirical relevance is far from established and a priori highly doubtful on theoretical
grounds, given the difficulties of aggregation and of intertemporal comparisons of a Y 19. There is general consensus (see e.g. Patinkin, 1987) that the real-balance effect is too weak,
including all the time new products. and too fraught with difficulties (e.g. bankruptcies) caused by the required price decreases,
15. ‘... the economy will behave as if there is a form of exogenous, labour-augmenting to be relied upon as a plausible mechanism ensuring a sufficiently fast tendency of aggregate
technological change’ (Romer, 1987, p. 61–2). demand to increase when prices decrease. A recent calculation by Sawyer (1997, section 6)
16. It is striking how little space is devoted in the New Growth literature to justifying crucial based on the NAIRU model of Layard, Jackman and Nickell concludes that a decrease of the
assumptions. How legitimate is the representation of human capital as influencing the fiscal stance causing a 1% decrease of aggregate demand would require a 67% decrease of
efficiency units of labour through a multiplicative effect HL? Neither Lucas nor Rebelo pose money prices in order for the real balance effect to counterbalance it − assuming no perverse
the question. Treating human capital as a factor of production in the usual sense would effects of the price decrease on investment.
imply that human capital should run against decreasing returns when added to a constant 20. Besides the well-known modern defenders (e.g. Kaldor, Basil Moore) of the endogeneity of
quantity of labour. The multiplicative approach has been probably aided by the inherent the money supply owing to overdraft facilities or to the creation of money substitutes, it is
vagueness of the notion of ‘human capital’, which has to do with acquiring know-how, interesting to remember that recently David Romer (2000) proposed to give up the LM curve
something different from increasing the amount of an input. The correct analogy seems to be of the IS–LM model owing to the observation that central banks increasingly target the
with software: adding human capital is similar to adding more or better software to a interest rate rather than the money supply. Pivetti (1991, ch. 2) summarises evidence
computer. But then different ‘quantities’ of human capital mean that one is dealing with suggesting that this is not only a recent tendency as Romer suggests, but a nearly universal
different kinds of labour, and aggregability and measurement of increases of the stock of aspect of capitalism in industrialised countries.
human capital are highly dubious. Several authors write that this multiplicative approach has 21. See also Hall (1993, pp. 278–9): ’established models are unhelpful in understanding this
been suggested by studies (e.g. Mincer) showing that for each extra year of schooling wages [1990–91] recession, and probably most of its predecessors. ... In spite of low interest rates,
increase by approximately the same percentage, 10% for some authors, 7% for others. A firms cut all forms of investment. ... Little of this falls into the type of behaviour predicted
non-neoclassical economist would argue that wage differentials are not due to differences in by neoclassical models’.
marginal products, and the capital-theoretic criticisms support such a stance. But even
within a neoclassical framework that justification appears questionable. Education changes
one’s skills, i.e. changes the kind of labour one offers, so in a supply-and-demand approach
relative wages depend on the relative scarcities of the different kinds of skills, and it will be
the choices of individuals which will alter relative scarcities until an equal convenience of
investment in education is reached; in other words, the ‘marginal product’ of an extra year of
education might well be decreasing in the sense that an increase of K per person and of the
‘amount of education’ (assuming it were measurable) per person in the same proportion
might increase output less than proportionally, but in an economy where different degrees of
education coexist, since different skills are not perfectly substitutable, if the supply of the
more highly skilled labour is sufficiently decreased, its marginal product and wage will
increase, and this will indeed happen until the rate of return on one more year of education
becomes the one desired by consumers: so within a neoclassical growth framework the
observed ‘marginal product’ of education reflects consumer choices and not a technological
property of that ‘marginal product’.
17. The ‘non-scale’ recent literature (see Jones, 1999, for a simple characterization) argues that,
given the share of output going to research, a bigger output does not mean a higher rate of
growth of labour productivity if the increase in output goes together with a proportional
increase in the variety of goods produced, and if research is sector-specific and with no
spillover to other sectors. No doubt the reasoning behind this argument has some relevance,
but it can hardly be used to deny that a larger economy dedicating more resources to
research will be advantaged by its greater size. The different size of the economy is what
made it easy for the USA to get ahead of the USSR in military technology, by allocating to
military research a much larger budget. Also, a bigger economy does not generally have a
proportionally greater number of no-spillover sectors; e.g. in the USA many industries are
simply bigger than in European nations.
18. I am not claiming that mainstream macroeconomics is unable to find explanations for these
episodes; I only claim that resources were underutilized and therefore policies, maintaining
aggregate demand higher than it was, would have ensured a higher utilization of resources
and higher growth rate.
and Weil, 1999). A fairly accepted view is that Solow’s (1956) model is
hardly able to provide any explanation of the transition. This is expressed,
for instance, by Galor and Weil (1999), and Chu (1998, p. 133). The latter,
emphasising the gain permitted by ‘new growth’ models, such as the
Becker–Murphy–Tamura model (Becker et al., 1990), states:
8. The demographic transition and
... the explanation of the Demographic Transition has not been successful under
neo-classical models of balanced growth the neo-classical model of Solow, for it typically predicts a converging steady-
state growth rate of per capita income, which is incompatible with the diverging
Piero Manfredi and Luciano Fanti development paths among countries observed over the past 50 years. Moreover,
Solow’s model is also weak in predicting the relationship between income growth
rate and population growth rate. It is well known that in Solow’s model the steady
state rate of per-capita income is a decreasing function of the population growth
8.1. INTRODUCTION rate.
An aspect lying at the very heart of classical economics, which is lost in neo- Though generally accepted, the above reasoning is, to our mind, exaggerated.
classical theory, is, as Samuelson (1985) sharply claimed, the endogenous What chiefly matters is not whether Solow’s original model with its
interaction between population and the economy, exogenous supply of labour can predict the demographic transition, or
explain its timing and forms, obviously doesn’t. Rather the point is whether,
Once upon a time, throughout the heyday of classical economics, demography adequately equipped, i.e. by postulating sound ‘transitional’ hypotheses,
belonged to political economy. The supply of labour was one of the important Solow's model can offer insights into fundamental aspects of modern growth.
endogenous variables in the systems of Smith, Malthus, Mill and Marx. One The present paper is divided into three main parts. In the first part we
feature of neoclassical economics that distinguishes it from the classical version, is review some main aspects of the DT aiming to combine the distinct
the removal of population as a variable subject to economists’ analysis (p. 166). perspectives of historical demography and growth theorists. In the second
part we critically discuss the main strategies used to incorporate the
This ‘removal’ is evident in the ‘standard’ neoclassical model of economic Transition in economic models. Here the DT is essentially regarded as a
growth (Solow, 1956), in which the production factors, labour and capital, major change in mortality and fertility patterns, leading to a peculiar time
are fully exogenous. Such a switch in the focus of analysis, the roots of profile in the relationship between the growth rate of the population and
which date back to the birth of neoclassical theory, is only partly a output per-capita.1 We distinguish two main approaches: 1) the transition is
consequence of the ‘logical structure’ of neoclassical theory (Kurz and microfounded, as the outcome of the maximising behaviour of individuals
Salvadori, this volume, ch. 1). In fact already in his 1956 paper Solow stated (for instance Jones, 1999); 2) the transition is postulated on empirical bases,
that he was dissatisfied with the gross treatment given to labour, and emphasizing the asynchronous patterns of decay of mortality and fertility as
suggested the need for fully endogenising the dynamics of population and per-capita income increases (as in Strulik, 1997, 1999a, 1999b). We also use
the supply of labour. An important, though obvious, consequence of a ‘diffusionist’ argument to model the asynchronous decay of mortality and
population endogenisation is that Solow’s model (and in general all ‘proper’ fertility. The ‘diffusionist’ approach (Rosero-Bixby and Casterline, 1993)
neo-classical models) becomes, according to the taxonomy in Kurz and has often proved to be more effective than demand-supply mechanisms in
Salvadori (this volume, ch. 1), an endogenous growth model. explaining the transition.
In this paper we discuss some of the effects of endogenous population Finally, in the third part we investigate the effects of transitional
dynamics and labour supply within Solow’s model. We will do this by using assumptions on simple neo-classical models in the presence of both constant
the ‘case study’ of a topic which is at the core of endogenous growth theories (CRS), and decreasing (DRS) returns to scale. First we consider a traditional
with population, namely the economic explanation of the Demographic CRS Solow model. In the absence of technical progress, mere population
Transition (DT). The Demographic Transition currently represents a major mechanisms lead to the existence of multiple equilibria, with a stable ‘poor
challenge for scholars in demo-economics aiming to find endogenous equilibrium’, i.e. a poverty trap, coexisting with a stable rich (‘modern’)
explanations of the transitions between great historical regimes (see Galor
161
The demographic transition and neo-classical models of balanced growth 163 164 The Theory of Economic Growth: a ‘Classical’ Perspective
equilibrium. Moreover, the well-known ‘absolute convergence’ statement of reduction in mortality; b) an increasing trend in per-capita income and
Solow’s model is replaced by the more general result by which countries that wellbeing, c) a fall in the demand for labour in agriculture which was the
escaped the poverty trap with different endowments of capital would condition for a transformation of social organisation, from rural to urban.
possibly experience an initial divergent phase, which should convert to a The reduction in mortality favours the subsequent reduction in fertility, 2 via
convergent regime only in the long (perhaps very long) run. This result the great shift from quantity to quality of children (Galor and Weil, 2000).
provides a powerful explanatory tool for observed dynamics, and an answer The diffusion of birth control was the main tool through which fertility
to the critical difficulty of Solow’s model previously quoted from Chu. We decline became possible. The outcome of this process was a transient period
subsequently consider a basic neoclassical model with DRS. DRS lead, of fast population growth, due to the asynchronous decay of mortality and
compared to CRS, to important, and more empirically plausible, additional fertility, and per-capita income. Population growth eventually came to an
results, which have not, with the exception of Strulik (1999a), been stressed end, due to the continued decrease in fertility, whereas growth in per-capita
in the literature on growth: a) coexistence of stable poor ‘Malthusian income has been uninterrupted.
stagnation’ equilibria with stable states of ‘modern’ long term balanced Demographic approaches are mainly descriptive, focusing on
growth, as historically observed, suggesting that DRS models are better than heterogeneities in timing at onset, pace, and local features of the transition,
CRS as candidates for unified modelling of ‘very-long’ term economic and no attempt has been made to develop simple formal demoeconomic
growth; b) appearance of persistent endogenous oscillations around the explanations of the transition. Major questions concerning, for example, the
‘Malthusian stagnation’ equilibrium when the population transition is inevitability of the transition, have not been posed at all. Very recently the
considered jointly with the, quite well documented, saving transition. The transition has become the object of renewed interest by (demo-) economic
latter result, obtained via the most standard growth model, offers a pleasing growth theorists (Galor and Weil, 2000 and refs therein; Jones, 1999 and refs
alternative perspective, i.e. a purely endogenous one, to the traditional therein).
exogenous explanations of oscillations around the stagnant Malthusian In order to manage what has become a jungle of papers with a huge
equilibrium. It is also interesting for the debate on Malthusian oscillations, an number of different definitions, we introduce some fundamental definitions
important topic in the recent demoeconomic debate (Lee, 1997). into the endogenous growth theory with population, e.g. with respect to the
This paper is organised as follows. The DT is reviewed in Section 8.2. terms of ‘poverty trap’ and ‘demoeconomic regime’.
Section 8.3 discusses some different modelling frameworks of the transition. ‘Poverty trap’ traditionally (Barro and Sala-i-Martin, 1995) refers to a
Sections 8.4 and 8.5 study some CRS and DRS Solow-type models with DT, stable steady state whose ‘name’ can be motivated thus: 1) ‘poverty’ because
and are followed by concluding remarks. it has low levels of per capita output and capital stock, 2) ‘trap’ because if
agents attempt to break out of it then the economy always tends to return to
it. But the meaning of ‘trap’ is clearer if it co-exists with (at least) another
8.2. THE DEMOGRAPHIC TRANSITION AND ITS equilibrium with better properties in terms of ‘welfare’ which is either i)
ECONOMIC DETERMINANTS: A REVIEW repelling (so that when the economy could – due to some shock – approach
it, the trap would inexorably re-swallow it), or ii) attracting but with a basin
By the phrase Demographic Transition demographers mean the set of of attraction so far from the ‘poverty trap’ region as to be unattainable for the
dramatic changes by which, starting in Western Europe, particularly England economic variables. The search for models with endogenous mechanisms of
and France, during the 18th century, demographic systems moved from their escape from the poverty trap has been a major ‘target’ for growth theorists.3
Ancien Régime, in which stationarity was maintained via a highly It is easy to see that, by only focusing on the dynamics of per capita physical
‘expensive’ balance between high levels of fertility and mortality, to their capital as in the Solowian model, in order to attain the ‘target’ (for instance
modern regime, characterised by low fertility and mortality (Chesnais, 1992; in order to have the interval of rising average product, necessary to have
Livi-Bacci, 1997, and references therein). Following the standpoint of multiple equilibria) there are few possibilities: i) of assuming increasing
historical demography (Livi-Bacci, 1997, 1998) the major engine of the returns to scale for all the factors through learning by doing or other
whole process appears to be the well-documented achievement of a threshold externalities, 2) of assuming a non-linear saving function, 3) of suitably
in the level of technology (or an increase in its pace of growth) which, endogenising population growth. While the first two candidates have been
allowing an increase in agricultural surpluses, led to: a) a sharp ‘endogenous’ widely used, and in particular the first perhaps is, directly or indirectly, the
‘growth engine’ of all the endogenous growth theory, the third, however Tsiddon, 1998; Galor and Weil, 1996, 1998, 1999; Jones, 1998, 1999;
widely explored, has not been successful, to our knowledge, within the Hansen and Prescott, 1999; Kögel and Prskawetz, 2001).
simple framework of Solow’s model (1956). In the present work we attempt
to fill this gap.
According to Galor and Weil (1999) three distinct regimes have 8.3. REPRESENTATIONS OF THE DT IN GROWTH
characterised the process of economic development: the ‘Malthusian’ MODELS: OPTIMISING VERSUS DIFFUSIONIST
Regime, the ‘Post-Malthusian’ Regime, and the ‘Modern Growth’ Regime.
PERSPECTIVES
To fully understand these definitions it is useful to distinguish the
macroeconomic point of view from the demo-economic one: the first focuses
The present section discusses some main avenues through which the DT may
on the behaviour of income per capita and technological progress, while the
be modelled: 1) via full microfoundation; 2) via an empirically-based
second on the relationship between the level of income per capita and the
formulation.
growth rate of the population.
Moreover, the empirically-based avenue can emphasize either only the
• Malthusian Regime: asynchronous patterns of decay of mortality and fertility as per-capita
1) low technological progress and population growth, at least relative to income increases or the more general ‘diffusionist’ mechanism.
modern standards, and roughly constant income per capita; As we will see, in the final analysis, all these approaches lead to
2) positive relationship between income per capita and population endogenising the supply of labour via a humped function of per-capita
growth. income, that will be used for subsequent analyses in descriptive growth
• Post-Malthusian Regime: models à la Solow.
1) growing income per capita during this period, although not as rapidly
as during the Modern Growth regime; 8.3.1. The Demographic Transition as the Outcome of an Optimal
2) positive relationship between income per capita and population Choice
growth (still as in the Malthusian Regime).
• Modern Growth Regime: Here we enrich Jones’ (1999) formulation by adding an exogenous non-zero
1) steady growth in both income per capita and the level of technology; saving rate, in order to make Jones’ formulation compatible with the
2) negative relationship between the level of output and the growth rate accumulation side in Solow’s (1956) model. Each individual has, in each
of population. period of time, an endowment of one unit of labour which can be used to
obtain consumption or children. Let h denote the fraction of time spent
From the demoeconomic point of view the DT is represented by the working (1 – h = time spent producing children), w the earnings per unit of
transition from the Malthusian to the Modern Growth Regime. As we are time worked, s the (constant) saving rate, c° the (constant) subsistence
mainly interested in the relation between economic growth and DT, in this consumption, b° the constant number of children independent of the time
section we essentially consider the two latter regimes. The Post Malthusian spent on childrearing (i.e. a ‘natural’ fertility). According to Jones (1999) the
Regime would be interesting if the focus were on the Industrial Revolution, utility of individuals depends on consumption (c) and number of children (b)
technological progress and other macroeconomic aspects.4 as follows
Obviously research on the interaction between income growth and fertility
(1 − m)(c − c°)1−γ m(b − b°)1−η
is not new in economics (e.g. Razin and BenZion, 1975; Cigno, 1981, 1984a, u(c, b) = + 0 < m < 1 ; 0 < γ ,η ≤ 1 (1)
1984b; Becker, 1988) but only more recently the literature has focused on the 1− γ 1 −η
existence of different long-run regimes corresponding to major epochs of
mankind’s history5 (amongst others Becker et al., 1990; Azariadis and The individuals take w as given and solve the problem:
Drazen, 1991; Ehrlich-Lui, 1991; Kremer, 1993; Palivos, 1995; Tamura,
max u(c − c°, b − b°) (2)
1996; Yip and Zhang, 1997; Lucas, 1998; Strulik, 1997, 1999a; Dahan and c ,b, h
subject to the restraints:

i) c = (1 − s)wh; and ‘natural’ fertility and saving behaviour can explain different time-paths as
regards the DT in different countries (and ultimately, as we will see in the
ii) b = f (1 − h) ; next sections, also the escape of poverty and self-sustained growth).
The restraint ii) states that each unit of time spent producing children The previous development can be used to endogenise population growth
produces f births, with f > b°. in Solow’s (1956) model, in which the supply of labour L is assumed to be
A simple reformulation of constraints i) and ii) gives the usual budget exogenously growing. In general terms the quantity of labour supplied for
constraint production is related to the total population in the work age span (N) by
L = hN where h is the participation rate. By taking h as constant over time, it
b(1 − s)w simply holds that Lˆ = Nˆ . Let us assume (under some simplifying
c+ = (1 − s)w (3) assumptions) that Nˆ = b(w) − μ ( w) = n(w) , where b and μ respectively denote
f
the birth and death rates in the population. Obviously, even in the simplest
where (1 – s)w is disposable income and 1/f is the (per child) cost of case, i.e. μ taken as constant, the one-hump shape of b implies that n is one-
childrearing. Jones has shown that the relation between the optimal fertility hump as well. If n is also non-negative, which can be obtained by simply
rate and the wage, which in general is only defined in implicit terms, actually taking μ = b0, then n(w) faithfully mirrors the demographic transition.
defines b as a humped function of w, therefore mirroring a major stylised fact The present approach offers a route to endogenous modelling of fertility
of the Demographic Transition (see Jones, 1999 for details). In order to transition within descriptive growth models which, being in full closed form,
obtain a greater analytical insight, we develop Jones’ formulation for the offers potentially interesting analytical results.
special case6 γ = η, corresponding to the well-known utility function of the
‘Constant Elasticity of Substitution’-type, which allows the explicit solution: 8.3.2 Empirical Approaches to the DT: Asynchronous Decay of
Fertility (and Mortality) with Per-capita Income, and the
b=
(
b° f + Hf (1 − s)wη −1 η − c°w −1 η ) (4)
‘Diffusionist’ Perspective
Ω The simplest ‘model’ of the DT used by demographers represents the

transition as a historical phase of population growth separating the ancient
where H is a suitable positive constants and Ω is a positive function of the
and modern demographic eras, which are mainly characterised by
wage.
stationarity. This phase of growth is the outcome of the asynchronous
A differentiation of (4) shows that
decline, usually s-shaped, of mortality (falling first) and fertility over time
2
(Chesnais, 1992). This empirical fact, together with another empirical fact –
−
∂b Hf [c° − w(1 −η )(1 − s)] Hf [c° − w(1 − s)](1 − s)(1 −η )w the monotonically increasing time trend of per-capita income during the
η
= + + same historical period – implies that the corresponding rate of change of the
∂w η wΩ ηΩ 2 population, i.e. the difference n ( y ) = b ( y ) – μ ( y ) , turns out to be a humped
1
− function of per-capita income (as in the case of the microfounded behaviour
η
Hfb°(1 − s)(1 −η )w
+ shown in Section 8.3.1). This humped function proves to fit classical
ηΩ 2 ‘transitional’ data sets very well (Strulik, 1997; 1999a).7
The dynamic consequences of a humped n ( y ) function are far from
The investigation of the latter expression suggests that, ceteris paribus, a trivial. As Strulik (1997) correctly noticed the traditional demographic
high subsistence consumption, a high propensity to saving, and a high approaches to the DT are basically a descriptive comment of the peculiar
‘natural’ fertility mean that the hump in fertility occurs for larger wage time patterns of the components of population growth observed during the
values. This implies that these factors could have played a role in temporally transitional regime. This may lead to the wrong impression that the transition
delaying the onset of the population bulge (i.e. the attainment of the moment has simply been ‘inevitable’. As shown in the next section, even the simplest
of maximal population growth). In other words differences in ‘cultural’ (or neo-classical growth model suggests that this does not need to be the case.
‘preference’) factors such as the level of perceived minimal well-being,
Though the choice of s-shaped fertility and mortality functions over time as new variables the total population N = X + Y, and the controlled fraction
appears mainly motivated by empirical considerations, there is an important ϕ = Y/N (0 ^ ϕ ^ 1). We obtain the decoupled equations:
theoretical argument in favour of the ‘logistic curve’ approach of the
previous section, namely the diffusionist argument. There is a strong body of (
ϕ = (1 − ϕ ) α + ( β − (bX − bY ))ϕ ) (6a)
evidence (Livi-Bacci, 1997; 1998) suggesting that the main tool by which
fertility decline was made possible during the DT was the diffusion of the N
practice of birth control. The diffusionist paradigm is often taken as the basis = rY + (bX − bY )(1 − ϕ ) (6b)
N
for alternative explanations to those based on economic factors, which often
provide better fits of observed patterns and pace of the transition (Rosero- with the initial condition ϕ ( 0 ) = 0. There are two qualitatively interesting
Bixby and Casterline, 1993, and references therein).8 Here we consider a cases.10 First, if β > bX – bY , then (6a) only has the non-trivial equilibrium
model for the diffusion of birth control which justifies under very wide ϕ1=1 which is globally asymptotically stable (GAS). In other words, when
conditions the s-shaped time pattern for the birth rate during the transition. the effects of inter-human communication are sufficiently strong, then in the
Consider a population in which the transitional decrease in mortality has long term all the population choose, in relative terms, to move to birth
already occurred, and which is composed by two groups, those who are still control. This means that the higher fertility of the natural group is made
practising natural (uncontrolled) fertility, and those who are practising birth ineffective by the ‘migration’ toward the lower fertility group. If rY = 0 the
control. We assume that the ‘natural’ group has high fertility and would, if total population becomes stationary in the long term, whereas it continues to
left alone, grow exponentially, whereas the ‘controlled’ group has a lower grow if rY > 0. Second, if β < bX – bY , (6a) also has the equilibrium
fertility. Let X = X(t) and Y = Y(t) respectively denote the size of the natural ϕ2 = α / [ ( b X – b Y ) – β ] which is meaningful for β + α < bX – bY , i.e. when the
and controlled fertility subgroups. Individuals are assumed to move from the overall rate of information diffusion is below a prescribed threshold. In this
natural to the controlled group due to the diffusion of information on birth case ϕ2 is GAS, implying long-term coexistence between the two groups.
control, which occurs via both external (the action of the media) and internal Consider now the overall birth rate in our population, which by definition
(inter-human contacts) diffusion.9 Moreover, children are assumed to inherit is a weighted average of the birth rates of the natural and controlled fertility
the fertility attitudes of their parents at birth. The model is the following: subgroups: bN (t ) = bX (1 − ϕ ) + bY ϕ . Note that in both cases considered
above, the time behaviour of ϕ ( t) will be that of an s-shaped function very
⎧ ⎛ XY ⎞
⎪⎪ X = bX X − μ X − ⎜ α X + β N ⎟
much like the logistic curve. In the first case for instance, postulating
⎝ ⎠ ϕ ( 0 ) = 0, the overall fertility rate bN will decline logistically over time from
⎨ (5)
⎪Y = b Y − μY + α X + β XY its natural fertility level bX to its controlled level bY, as largely documented by
⎪⎩ Y
N the empirical evidence on the DT.
This result moreover suggests that the s-shaped relations between fertility
where μ is the mortality rate; bX, bY are, respectively, the fertility rates of the rate and per-capita income (rather than time) used for instance by Strulik
natural and controlled groups, with bX > bY ≥ μ; α is the rate of transition from (1997, 1999a, 2000) necessarily appears empirically due to the
the natural to the controlled group due to external information, β is the monotonically increasing historical time trend of per-capita income during
contact rate in internal diffusion. Model (5) is more general than other the transitional era.
models used for the diffusion of birth control in that it is not limited to the
situation of a stationary population (as in Rosero-Bixby and Casterline,
1993). 8.4. THE DEMOGRAPHIC TRANSITION AND SOLOW’S
What does model (5) predict? Let rX = bX − μ − α ; rY = bY − μ MODEL: THE CASE OF CONSTANT RETURNS TO
respectively denote the rates of growth of the two subpopulations in the SCALE
absence of diffusion. We assume rX ≥ rY because the opposite case would
imply that external diffusion alone is capable of quickly driving the natural In this section we investigate the dynamic consequences of the DT, within
group to a slower growth compared to the controlled group. Let us consider the Solow (1956) model without technical progress. According to the
discussion of Section 8.3, the DT may be, as a first approximation, of a constant saving rate. The function m1 (k) = s f ( k ) /k has the traditional
represented by replacing the exogenous rate of change of the supply of decreasing form whereas m 2 ( k ) = δ + n ( k ) mirrors the humped form of
labour with a positive humped function n(y) of per-capita income y. The n(k). The vertical distance m 1 ( k ) – m 2 ( k ) is the rate of growth of k over
ensuing ‘transitional’ Solow-type model is time, denoted as γk . Figure 8.1 shows that multiple equilibria may exist. The
behaviour of (8) is summarised by the following:
k = s ( y ) f ( k ) − (δ + n ( y )) k (7)
Proposition 1. Apart from the zero equilibrium, system (8) admits one or
where k denotes the capital labour ratio, f(k) is a CRS production function (in three non-trivial equilibria. In the former case let k1 be the unique positive
per capita terms) satisfying Inada’s condition, s is the saving rate (here equilibrium. Then k1 is always globally stable. In the latter case (see Figure
postulated, in contrast with the original Solow’s model, to be income- 8.1) let the three equilibria be klow<kmid<khigh. It is easy to show that kmid is
dependent, but will be simply assumed constant below), δ is the rate of unstable, whereas klow and khigh are locally stable, with respective basins of
capital depreciation, n(y) is the rate of change in the supply of labour. Since attraction: Bas(klow) = (0,kmed), Bas(klarge) = (kmid ,+∞).
y = f( k) under CRS, we obtain (by writing s ( k ) , n ( k ) as shortcuts for
s ( f( k ) ) , n ( f( k ) ) :
k = s ( k ) f ( k ) − (δ + n ( k )) k (8)
Clearly n has a humped relation with per-capita capital as well. We

assume that n starts from zero (or a slight positive value) at the beginning of
the transition, increases up to a maximum, and finally decreases up to a small
positive (or zero) asymptotic value at the end of the DT.11
Remark 1. Formulation (8) is well acknowledged in the literature on

economic growth with endogenous populations. Already Solow (1956)
considered it (see also Nelson, 1956, Niehans, 1963), though he concluded
that population endogenizaton, regardless of the complexity of n(.), had no
relevant effects on his main results. Nerlove and Raut (1998, p. 1127)
emphasize this point, thus raising strong skepticism about the approach:
it is clear that merely endogeneizing population growth at the macro-level does
not shed light on the shape of n and thus on the nature of the dynamics; a utility-
maximising model should be used to elucidate the nature of the function.
Figure 8.1 – The three equilibria in Solow’s model with transitional
Their remark is in principle correct and indeed we have fully elucidated the dynamics
microfounded nature of the shape of n in the previous Section 8.3.1 (see also
for other different attempts to give microfoundations to the shape of n, The proof easily follows from examination of the sign of the rate of change γk
Momota and Futagami, 2000 and Strulik, 1999b). (the arrows on the horizontal axis in Figure 8.1 denote the direction of
Nevetheless, we believe that an indisputable exception to the claim of motion of k). The following substantive results emerge:
Nerlove and Raut is represented by the humped ‘demographic transition’
hypothesis which is supported by extensive empirical evidence and thus may 1) Existence of a poverty trap
also justify the macro-level approach to the DT. In the case of multiple equilibria in (8), there is a ranking of equilibria (see
The analysis of (8) is straightforwardly carried out by the usual graphical Figure 8.1) in terms of all per-capita variables, so that the locally stable ‘low’
tools (Barro and Sala-i-Martin, 1995). Let us consider (Figure 8.1) the case equilibrium klow is a poverty trap (or a Malthusian equilibrium), whereas the
rich equilibrium khigh can be interpreted as a ‘modern’, post-transitional,

regime. Barro and Sala-i-Martin (1995) argue that poverty traps typically
arise as the result of the coexistence of regions of decreasing returns with
regions of increasing returns. In (8) the poverty trap arises from the shape of
the rate of population growth during the transition.
There is a second interesting case, namely when the m1 curve is so low
that the existence of kmid , khigh is prevented. In this case the unique equilibrium
k1 is a ‘very poor’ equilibrium as, ceteris paribus, k1<klow, and is stable.
Therefore k1 is still a poverty trap. This case can be considered historically
prior to the three equilibria case, representing that time-window of the
‘ancien régime’ stagnation during which accumulation was so low that the
existence itself of a richer regime was impossible. Finally, there is a third
case (not showed in Figure 8.1) when m1 is so high that the existence of the
poor equilibria klow , kmid is prevented, k1 becomes a ‘virtuous’ modern
equilibrium. This leads to the following remark concerning the non-
ineluctability of modernisation:
Remark 2: The existence of a poverty trap suggests that the traditional view,
which views modernisation (the industrial revolution and the DT) as an Figure 8.2 – Realistic (absolute) divergent/convergent patterns in the CRS
ineluctable process, is incorrect. Solow model with demographic transition
Once a poverty trap exists, a major problem is of course how to break out
of it. A common feature of models such as (8) is that the ‘escape’ may only This has led to the controversial concept of absolute convergence (Barro and
occur as a consequence of policies and/or external shocks, such as i) Sala-i-Martin, 1995, ch. 1): economies with lower per-capita capital are
domestic policies/shocks leading to an increase in the saving rate (as predicted to grow faster in per-capita terms, a fact which has often been
documented in Strulik, 1997, 1999b), ii) increases in the technological denied on empirical grounds (quotation from Chu in Section 8.1). In model
parameters tuning both the height and/or the shape of the m1 curve. Though (8) economies which in the end escaped the Malthusian trap do not exhibit
such external events are not the only explanations of the ‘escape’, they convergent paths. Consider two economies A, B which escaped the trap, i.e.
certainly took place in history, and played a role in giving the ‘big push’ to that after some external shock entered the (kmid , khigh) region with respective
investment in capital that allowed the escape from the Malthusian stagnation endowments kpoor<krich. (not shown on the figure). There is a whole region in
towards the Modern Regime. This is clearly expressed by Becker et al. the set (kmid , khigh) in which the two economies initially diverge, i.e. the richer
(1990, S33): ‘We believe that the West's primacy, which began in the XVII economy grows faster than the poorer one (the amplitude of such a region
century was partly due to a “lucky” timing of technological and political depends on the actual position of kpoor<krich). In other words richer countries
changes in West’. become, in an initial phase (the temporal length of which may be quite large)
even richer. Only at a later stage will the two economies enter a phase of
2) Realistic convergence patterns convergent dynamics as in Solow’s model. We argue that the present
Let us consider the rate of change of k in the region (kmid , klarge ) where escape mechanism offers the simplest explanation for currently observed paths of
from the trap has occurred. As Figure 8.2 shows, the rate of change of k, γk, rich versus developing countries. Such an explanation has the merit of being
increases from zero (at kmid) up to a maximum at k*, and then monotonically based on a huge body of demographic evidence, and therefore appears more
declines to zero again, as the system is attracted in the khigh equilibrium. The convincing of alternative explanations, such as that offered in Barro and
implications are noteworthy. In Solow’s (1956) model γk monotonically Sala-i-Martin (1995, ch. 1), which is based on the rather controversial
declines to zero as k increases toward its equilibrium value. coexistence of windows of increasing and decreasing returns. Therefore the
following remark holds:
Remark 3: the diverging growth paths among countries observed over the evidence (Livi-Bacci, 1997, 1998; see also the discussion in Section 8.2.1)
past 50 years can be wholly compatible with Solow’s model with DT, in that and the recent history of developing countries, a non-trivial investigation of
the length of the first phase after the escape from poverty can be quite large. the implications of the DT would imply the need to simultaneously
endogenise all the parameters of Solow’s model.12
8.4.1. Adding Technological Progress Consideration of further realism in descriptive neoclassical models of the
DT has led to several research efforts in very recent times. Strulik (1999a)
Let us consider exogenous technical progress at the constant rate α, considered a DRS Solow-type model for the demographic transition, saving
combined with CRS Cobb–Douglas technology, i.e. Y = Qeα t L (t ) K (t ) .
β 1− β
transition (as suggested by both theoretical and empirical evidence) and fixed
This leads to the following model in the per-capita variables k = K/L, y = Y/L wage regulation (typical of developing countries). Unfortunately, the fixed
wage regulation assumption, in which the workers are exposed to a forever
⎧ k y
⎪ = s −δ − n (y)
constant wage, requires further assumptions to explain why they should
⎪k k experience a transition in their demographic patterns as per-capita income
⎨
(9)

⎪ = α + (1 − β ) k
y increases without increasing wages, and moreover the possibility of the
⎪⎩ y k Modern regime equilibrium is vitiated by a violation of the supply of labour
constraint. Prskawetz et al. (2000) investigated, within a similar framework,
The previous system does not have non-trivial equilibria in the per-capita the role of human capital as an additional factor in the production function in
variables (if k / k = 0 , then y / y = α > 0 ). It has only a stable state of balanced the presence of a ‘transitional’ dynamics of the rate of technological
growth. This fact would lead to unrealistic predictions in terms of very long progress. However, their descriptive introduction of human capital in the
term growth. In particular, because of the well documented long-term growth production function is rather ad hoc. Strulik (1999b) considered the
in technology, it would imply a sustained economic growth much before than transition in a broader framework, with a microfounded approach to human
when it has been observed. This result suggests that steady technological capital, and a more general production function exhibiting both increasing
growth plus CRS cannot have been the rule in historical epochs. Hence, the and decreasing returns, while Strulik (2000) investigated the transition in a
hypothesis of CRS as the rule during very long-term economic development two-sector Solow-type economy.
appears as very implausible. Moreover, the long-term rate of growth of per- The model that is presented here may be considered the neo-classical
capita variables in the balanced growth regime is the rate of labour backbone of all the aforementioned contributions in that it is simpler while at
augmenting technical progress q = α / β, i.e., it is independent of the (long- the same time preserving the main dynamic features of richer formulations,
term) rate of change of the population. In other words, the CRS model would such as Strulik (2000) and Praskawetz et al. (2000). Moreover it does not
lead to the conclusion that the demographic transition played no role in long- suffer the contradiction of Strulik’s (1999a).
term growth. This suggests the following: Our model assumes the following DRS production function:
Y = Qeα t L (t ) K (t )
β γ
Remark 4: The CRS model is genetically unable to study the features of the Q > 0 ; β + γ − 1< 0; α > 0 (10)
DT and to explain the historical pattern of the economic ‘take-off’.
We moreover assume ‘transitional’ dynamics (over the same time
window) of: a) the rate of change of the population according to a positive
8.5. DEMOGRAPHIC TRANSITION AND THE SOLOW humped function of per-capita income, b) the saving rate, which is assumed
to be an increasing logistic function of y (as in Strulik, 1999a, 1999b, 2000;
MODEL UNDER DECREASING RETURNS TO
Praskawetz et al., 2000); c) the rate of change of technical progress α, taken
SCALE as an increasing logistic function of y, as in Prskawetz et al. (2000). Standard
Solow-type assumptions and straightforward manipulations lead to the
Motivated by the last remark of the previous section we consider now a following formulation in the per-capita variables x = K/P, y = Y/P where P is
model with Decreasing Returns to Scale (DRS). We acknowledge that the the total population size (or the total supply of labour)13
complexity of the DT would require much greater detail to take into account
the changing economic environment. As suggested by both the historical
x y The curves h1(y), h2(y) inherit the humped shape of n(y). Let us consider
= s (y) − δ − n (y) (11a) the standard case of stationary population at both the beginning and end of
x x
the transition: n(0) = n(∞) = 0.14 In this case h1(0) > h2(0), and h1(∞) > h2(∞).
y x There are two possibilities: a) the curve h1(y) always lies above h2(y) (no
= α ( y ) + γ − (1 − β − γ ) n ( y ) (11b)
y x equilibria), or b) the curve h2(y) intersects twice h1(y).
Case a) occurs for α > (1 – β – γ ) n Max, where nMax is the maximal growth
It is convenient for analysis to consider the variables U = y/x, and y, rate attained by the population during the DT, whereas b) occurs in the
obtaining: opposite case, α > (1 – β – γ ) n Max. Therefore no equilibria exist in case a),
whereas exactly two equilibria exist in case b). In this latter case let us
U
= α ( y ) + (1 − γ )δ + β n ( y ) − (1 − γ ) s ( y )U (12a) denote the equilibria by E1,E2, where E1 is the ‘poor’ equilibrium with smaller
U per-capita income. It is easy to show (see Appendix) that E1 is always locally
asymptotically stable (LAS), whereas E2 is always unstable. The nullclines
y
= α ( y ) − γδ + γ s ( y )U − (1 − β ) n ( y ) (12b) and the directions of motion of system (12) are drawn in Figure 8.3 for case
y b). To fully understand the dynamics of the system it is also useful to look
for states of balanced growth, i.e. asymptotic states characterised by
In order to gain a broad understanding of the results of the present section,
exponential growth of per-capita variables at the same constant rate q, and
it is worth making the following remark concerning the basic neo-classical
therefore by an asymptotic constant ratio U* = H. This leads to the following
DRS model (i.e. model (12) for exogenously determined s, n, α):
system in the quantities (H,q)
Remark 5 (neo-classical model with DRS). Under constant s,n,α model (12) α + (1 − γ )δ + β n (∞ ) − (1 − γ ) sH = 0 (14a)
has a unique globally asymptotically stable state of balanced evolution in
per-capita variables. The long term rate of change of per-capita variables is: q = α − γδ + γ sH − (1 − β ) n (∞ ) (14b)
α − (1 − β − γ ) n
q= giving:
(1 − γ )
α − (1 − β − γ ) n (∞ ) α + (1 − γ )δ + β n (∞ )
q= ; H= (15)
In particular balanced growth occurs for α − (1 − β − γ ) n > 0 , whereas 1− γ (1 − γ ) s
balanced decay occurs in the opposite case. In the absence of technical
progress (α = 0) the long term outcome is always balanced decay at the rate Examination of Figure 8.3 shows that the state of balanced growth (15) is
q = − ((1 − β − γ ) n ) / (1 − γ ) . LAS. Note that the asymptotically constant portion of the h1 curve represents
Let us now investigate (12) in a hierarchical manner, by starting from the the radius of the balanced growth state.15
basic situation in which only population is endogenous, whereas s, α are Let us summarise the main results by the following:
taken as constant. Subsequently, in addition to the endogenous population,
also the cases of, respectively, endogenous saving (Section 8.5.2) and Proposition 2. In the neo-classical model (12) with endogenous population
endogenous technical progress changes will be investigated (Section 8.5.3). dynamics, given the technological ratio γ / β , two main outcomes are
possible: a) the rate of change of technology (α) is so large16 as to always
8.5.1. Only Population Growth is Endogenous absorb the population bulge observed during the transition. In this case the
system always attains a long-term state of balanced growth, at the rate
Equilibrium analysis gives the non-trivial nullclines q = (α − (1 − β − γ )n(∞) ) (1 − γ ) ; b) the growth of technology is not
sufficiently fast to absorb the population bulge. In this case two equilibria (E1
α + (1 − γ )δ + β n ( y ) γδ − α + (1 − β ) n ( y )
U= = h1 ( y ) ; U = = h2 ( y ) (13) and E2) are possible and, depending on initial conditions, the economy will
(1 − γ ) s γs
be attracted in the Malthusian poverty trap E1, or will attain a long-term state Let us summarise the main results by the following:
of balanced growth.
Proposition 2. In the neo-classical model (12) with endogenous population
U dynamics, given the technological ratio γ / β , two main outcomes are
h 2 (y) possible: a) the rate of change of technology (α) is so large18 as to always
absorb the population bulge observed during the transition. In this case the
system always attains a long-term state of balanced growth, at the rate
q = (α − (1 − β − γ )n(∞) ) (1 − γ ) ; b) the growth of technology is not
sufficiently fast to absorb the population bulge. In this case two equilibria (E1
h 1 (y)
and E2) are possible and, depending on initial conditions, the economy will
E1 E2
be attracted in the Malthusian poverty trap E1, or will attain a long-term state
of balanced growth.
which are not necessarily humped, as in the previous case. Under the
standard assumption n(0) = n(∞) = 0, it holds that h3(0) > h4(0), and
h3(∞) > h4(∞), with h3(0) > h3(∞), and h4(0) > h4(∞).There are two interesting
cases depending on whether the initial population growth has been faster, or
slower, compared tothe saving rate growth. In the first case the curves
h3(y), h4(y) are, in most situations, one-humped, as in Section 8.5.1, whereas
in the second case both curves will be, in most situations, monotonically
decreasing in y. In terms of equilibria this implies that again, as in the
previous section, two main cases are possible: a) no equilibria, when the rate
Figure 8.3 – Equilibria and direction of motion for the neo-classical DRS of exogenous technical progress is much larger compared to the maximal rate
model with DT in case b) (two equilibria) of increase of the population during the transition, and b) two equilibria in
the opposite case. The dynamic analysis shows results quite similar to those
Remark 6. The previous result suggests that, contrary to the CRS case, the of Section 8.5.1 (the poor Malthusian state is usually LAS, whereas the
DRS framework allows: a) the simultaneous coexistence of stable ‘historical’ second equilibrium E2 is unstable) with a major novelty, e.g. the appearance
steady states with ‘modern’ long-term stable states of balanced growth, of stable oscillations around the Malthusian equilibrium. Such stable
characterised by different basins of attraction. That is: the DRS model offers oscillations appear through a Hopf bifurcation of the Malthusian equilibrium
a unified explanation of both Malthusian stagnation and modern growth. b) (see Appendix). The major substantive results are summarised in the
that only in the DRS framework the population dynamics affects per-capita following:
income growth. This shows to what extent the DT may have been crucial in
allowing modern regimes with fast economic growth. Proposition 3. In the neo-classical DRS model with endogenous population
and ‘saving transition’, in addition to the outcome formulated in Proposition
8.5.2. Endogenous Population and Saving 2, also the following outcome is possible: the stagnant Malthusian dynamics
around the Malthusian poverty trap E1 may occur through stationarity, or
We now move on to study the joint dynamic effects of population transition through stable oscillations. The appearance of stable oscillations is the
and ‘saving transition’.17 consequence of a quick change in the patterns of saving at the beginning of
The non-trivial nullclines are: the transition, which locally (but only locally) destabilises the Malthusian
equilibrium.
The previous proposition suggests that those persistent oscillations that
α + (1 − γ )δ + β n ( y ) γδ − α + (1 − β ) n ( y )
U= = h3 ( y ) ; U = = h4 ( y ) (16) scholars in economic history have usually explained through purely
(1 − γ ) s ( y ) γ s (y) exogenous arguments, plague or famine crises for instance, could have been
just part of the story. In fact the economic system was potentially able to lead 8.6. CONCLUDING REMARKS
to purely endogenous oscillations around the Malthusian stagnation
equilibrium, for instance via the mechanism embedded in the model of this The present paper, by re-emphasizing the endogeneity of the supply of
section. labour as in the classical economics, shows that also the standard
In sum we have shown that the DRS framework is capable of capturing neoclassical Solowian model, can offer insights into fundamental aspects of
two major features of long-term demo-economic evolution: the existence of a modern growth. We critically discuss several aspects related to the modelling
Malthusian, possibly oscillatory, regime and the crucial role played by the of the Demographic Transition, offering a microfounded explanation of the
DT in allowing modern regimes with fast economic growth. latter. In particular the basic Solow 1956 Constant Returns to Scale (CRS)
model, having been enriched with an assumption on the dynamics of the
8.5.3. The Joint Role of Endogenous Productivity Changes and supply of labour mimicking the DT, offers interesting insights into the
Population interrelationships between population dynamics and poverty traps, and a
simple, but interesting, generalisation of the widely debated notion of
Major studies in economic history (Bairoch, 1973) have amply documented convergence. Moreover when also (exogenous) technical progress is
that technical progress has not been constant in the long term and its time considered we argue that 1) the CRS model is genetically unable to study the
profile can be represented by an s-shaped curve, mirroring the initial phase of features of the DT and to explain the historical pattern of the economic ‘take-
very slow increase followed by substantial acceleration from the beginning off’; 2) by contrast, our neo-classical Decreasing Returns to Scale framework
of the industrial revolution, and by a decrease in the pace of growth in the is capable of showing the coexistence of a stable poor ‘Malthusian
more recent phase.19 The analysis may lead to two distinct cases depending stagnation’ (possibly oscillatory) equilibrium with stable states of ‘modern’
on whether the population increase temporally followed, rather than long-term balanced growth, as historically observed, and especially of
preceded, the increase in technology. For western countries the pace of showing to what extent the DT may have been crucial in allowing modern
technological growth is known to have started to increase long before the regimes with fast economic growth. This feature suggests that DRS models,
onset of the DT. In this case we again have the coexistence of a locally stable rather than CRS, are optimal candidates as unified models of ‘very-long’
Malthusian equilibrium, with an unstable intermediate equilibrium, with a term economic growth, and that the enrichment of the neoclassical growth
locally stable regime of balanced growth. Compared to Section 8.5.2 no models with themes belonging to classical economics, namely the supply of
persistent historical oscillations are possible (see Appendix). In the opposite labour dynamics and decreasing returns, can provide an interpretative tool of
case, in which the DT anticipated the technological take-off, and which is the DT and economic ‘take-off’.
relevant for many developing economies, a third equilibrium may appear
(oscillations are again ruled out) making the mechanisms of escape from the
poverty trap richer than the previous cases. This result was also obtained by APPENDIX: LOCAL STABILITY ANALYSIS OF
Prskawetz et al. (2000) but in a different model.
EQUILIBRIA IN THE NEO-CLASSICAL MODEL WITH
To sum up, our models presented above have ‘benchmarks’ with which to
compare all the models generating multiple equilibria (e.g. the models of DEMOGRAPHIC TRANSITION AND DECREASING
Becker et al., 1990 and Lucas, 1998). Although the two latter models are RETURNS TO SCALE
completely different from ours in that they are i) microfounded according to
a different history with respect to that described in Section 8.3.1, and ii) are The Jacobian of (12) evaluated at equilibrium
centred on the role of human capital accumulation, they share with our model
the focus on the population dynamics in a Malthusian and a Modern era, and (
⎛ − (1 − γ ) s ( y )U U α ' ( y ) + β n' ( y ) − (1 − γ ) s' ( y )U)⎞⎟
in particular the same interpretation as follows: the Malthusian and Modern J =⎜
eras are different steady states of the same model.
⎜
⎝ γ ys ( y ) ( ) ⎟⎠
y α ' ( y ) − (1 − β ) n' ( y ) + γ Us' ( y )
A) In the model with only endogenous population (Section 8.5.1) we obtain

Tr ( J ) = − (1 − γ ) sU1 − (1 − β ) n' ( y ) y , Det ( J ) = sUn' ( y ) y (1 − β − γ ) > 0 . NOTES
Since n’(y) > 0 in the Malthusian equilibrium E1, then Tr ( J1 ) < 0 , 1. In this section we do not consider other features not directly related to fertility and mortality,
whereas Det ( J1 ) = sUn' ( y1 ) y1 (1 − β − γ ) > 0 , showing that E1 is always LAS. such as technological progress dynamics and the saving ‘transition’, that the recent literature
– reviewed in the next section – embodies in the more general historical process connecting
Similarly, since n'(y) < 0 at E2, E2 is always unstable. demographic transition, industrial revolution and the modern growth era. Some possible
roles played by such phenomena are taken into account in the third part of this work.
B) In the model with endogenous population and saving (Section 8.5.2) we 2. In some cases however there is also evidence (Dyson and Murphy, 1985) that fertility
obtain: temporarily rose, thanks to increased well-being, before starting to decline.
3. The same terminology may express a different concept in other authors: e,g Kögel and
(
Tr ( J ) = − (1 − γ ) s ( y )U + y − (1 − β ) n' ( y ) + γ Us' ( y ) .) Prskawetz (2001, p. 2) claim that: ‘... economy will be trapped in a situation where no
sustained growth of per capita income can occur. This trap is commonly labeled Malthusian
trap’ and ‘we label the simultaneous take-off in economic growth and population growth as
Moreover: Det ( J ) = ys ( y )U (1 − β − γ ) n' ( y ) > 0 . The last expression escape from the Malthusian trap’. In our opinion these definitions are restrictive: they reflect
shows that a Hopf bifurcation occurs for some parameter constellations at the the emphasis on the ‘forever sustained’ exponential growth which pervades all the
endogenous growth theory.
Malthusian equilibrium. In fact the Trace may become negative for some 4. Finally, note that from the demoeconomic perspective, the Modern Regime is sometimes
parameter constellations, especially when the saving rate increases rapidly at defined also as Post-Classical, Anti-Malthusian, Beckerian.
the beginning of the transition era, whereas Det ( J1 ) > 0 . 5. Some issues associated with this recent literature are reviewed in Fanti and Manfredi (2001).
6. Notice that this utility function, by assuming γ = η < 1, shows a high elasticity of
C) In the presence of endogenous population and technological change substitution between consumption and children (e): indeed e = 1/η > 1 and the elasticity is
decreasing with η (at the limit when η tends to one, the elasticity tends to become unitary as
(Section 8.5.3) at the Malthusian equilibrium E1 the following holds: in the Cobb–Douglas case).
Let us summarise the main results by the following: 7. This argument is developed by Strulik (1997, 1999a) who models the population birth (b)
and death (μ) rates during the transitional regime by means of two asynchronously
Proposition 2. In the neo-classical model (12) with endogenous population decreasing logistic-alike functions of per-capita income y. Moreover, an attempt to provide a
dynamics, given the technological ratio γ / β , two main outcomes are theoretical justification based on an overlapping generation argument is also given in Strulik
(1999a).
possible: a) the rate of change of technology (α) is so large20 as to always 8. More general explanations seek to embody both types of effects (Retherford and Palmore,
absorb the population bulge observed during the transition. In this case the 1983).
system always attains a long-term state of balanced growth, at the rate 9. Rosero-Bixby and Casterline (1993) suggests three main types of effects which may lead
q = (α − (1 − β − γ )n(∞) ) (1 − γ ) ; b) the growth of technology is not people to switch to a different group: not only information flows, but also demonstration
sufficiently fast to absorb the population bulge. In this case two equilibria (E1 effects, based on the experience of other people which evidence the benefit of the transition,
and changes in normative contexts.
and E2) are possible and, depending on initial conditions, the economy will 10. Equations (6a)–(6b) can also be solved analytically.
be attracted in the Malthusian poverty trap E1, or will attain a long-term state 11. We will not be concerned with post-transitional (or ‘second’ demographic transition)
of balanced growth. phenomena (of which Italy and Spain in the past 15 years are major examples), by which the
rate of population change could even become sharply negative.
( )
Tr ( J ) = − (1 − γ ) sU1 − y (1 − β ) n' ( y1 ) − α ' ( y1 ) , 12. An example is the rate of growth of technology, which has not been constant since historical
times, but has rather experienced a historical evolution from an initial value close to zero,
and followed by a very slow increase, before the take-off of the modern era, which is complexly
correlated with growing per-capita income and knowledge. Similarly, the constancy of the
( )
Det ( J ) = sU1 y1 (1 − β − γ ) n' ( y1 ) − α ' ( y1 ) . parameters tuning the relative role of capital and labour in the production function is, at best,
a simplistic approximation, and so on.
13. The distinction between variables y = Y/P and the traditional Y/L is formally unnecessary
This implies that no Hopf bifurcation is possible at the Malthusian here but it is useful in more general contexts.
equilibrium. 14. Some differences would arise if n(0) > 0 and/or n(∞) > 0 (or even n(∞) < 0, according to the
second demographic transition). As the purpose of this paper is not taxonomical we do not
consider them here.
The demographic transition and neo-classical models of balanced growth 185
15. As regards the basins of attraction of the balanced growth state, note that the region to the
right of the unstable equilibrium E2 and above the curve h1 is positively invariant, as is the
region to the right of E2 between the two curves.
16. Alternatively one could reason in terms of the composite technological parameter
α / ( 1 – β –γ) versus the rate of population growth.
17. The latter has been documented empirically for some Western countries by Strulik (1997),
who has also given theoretical support to the fact that the mortality decline observed during
9. Human capital formation in the new
the DT could lead to an increase in the saving rate as well.
growth theory: the role of ‘social
α / ( 1 – β –γ) versus the rate of population growth.
19. For instance, as stated above, Prskawetz et al. (2000) used a logistic function.
factors’
α / ( 1 – β –γ) versus the rate of population growth. Maria Rosaria Carillo
9.1. INTRODUCTION
First-generation endogenous growth models, assuming human capital

accumulation as a major engine of growth, have grounded their analysis on
the Beckerian model of human capital, where homogeneous agents in the
presence of perfectly competitive markets forgo leisure and current income
in order to increase their knowledge and obtain a higher future income. This
approach envisages no role in the creation of human capital for any of the
phenomena tied to an individual’s social behaviour such as ‘peer effect’,
‘direct knowledge transmission’, ‘status-ranking’ of occupations, ‘network
relationship’, and so on, although the importance of such social phenomena
for individual formation has been widely recognized by the literature on
human capital.1 Probably, behind the recognition that human capital has the
distinctive feature of producing a large amount of externalities (Lucas, 1988)
lies the idea that direct social relations among agents themselves create
knowledge. Nevertheless, this phenomenon has not been explicitly
investigated, and the mechanisms by which the externalities are generated
remain entirely unexplained.
In recent years, a class of endogenous growth models have analysed in
depth how knowledge is formed and transmitted among individuals to give
rise to externalities. In doing so, they have highlighted the role played by
social relations in the creation of human capital, by assuming that the latter is
formed not only through an educational activity, but also through the
relations that arise among individuals. More specifically, they show that
social factors are further channels for the transmission of knowledge which
also modify its use and desirability.
The growth role of social factors has been largely analysed by the New
Growth Theory, without limiting the analysis to the effects on human capital
formation. Cole et all (1992) and Corneo and Jeanne (1997), for example,
186
Human capital formation in the new growth theory 187 188 The Theory of Economic Growth: a ‘Classical’ Perspective
have analysed the effects of status-seeking behaviour on wealth also reflects the reputation associated to each. This happens since individuals
accumulation and on the saving rate. Temple and Johnson (1998), using the take care of their social relations, and the total reward of an occupation is
Adelman and Morris data base, carry out an empirical analysis to test affected also by the relative position in the social ranking obtainable with
whether ‘social capital’ matters in determining economic performance, that occupation. ‘First the wages of labour vary with…the honourableness or
thereby confirming this hypothesis. Also Knack and Keefer (1997) found dishonourableness of the employment. … Honour makes a great part of the
evidence that trust and cooperation are associated with a higher economic reward of all honourable professions’ (Smith 1776, p. 202). Again he wrote
growth rate while Zak and Knack (2001), assuming that trust reduces ‘The public admiration … makes a considerable part of total reward in the
transaction costs, show that high trust societies have a higher investment rate professions of physic, a still greater perhaps in that of law; in poetry and
and produce more output than low trust societies. philosophy it makes almost the whole’ (Ibidem, p. 209).
A large number of papers follow this strand of literature and there is ever This hypothesis about individual behaviour has been adopted by several
growing attention towards this field of research. In this paper I shall consider strands of literature, yet the inclusion in an endogenous growth framework
only a particular aspect of this wide theme: the growth effects of social opens up further lines of research, since it makes it possible to analyse how
factors via their influence on the accumulation of human capital. Before social interdependence can interact with the growth process: according to
analysing this theme, I shall try to define what are social factors and the which behaviour will prevail and which relations between agents will
nature of the relation existing between them and human capital. An initial become stronger or weaker, it is possible to predict the evolution of ‘types’
problem is the lack of clarity over the definition of social factors. Moreover, of individuals within the population. Therefore this approach can lead
it will be apparent that also the nature of the relation between human capital naturally to an evolutionary analysis of economies and of their economic
and the latter factors is not clear. Several authors hypothesise that they performance.3 From these considerations, it is apparent that the possible
directly affect the human capital accumulation process, which occurs with developments of this approach are considerable.
important feedback effects (Coleman, 1988). However, others assume that This chapter, which will review this strand of literature, is organized as
social factors influence human capital accumulation only indirectly, since follows. The second section discusses the concepts of social interactions,
they are productive factors in the aggregate production function which are social capital, culture and ideology, and social status concern, all of which
complementary to, or substitutes of human capital (Glaeser et al., 2000; indicate the effects of social factors on human capital formation and its
Iyigun et al., 2001). diffusion among individuals. The third section surveys the analytical
Analysis of the relation between human capital and social factors is an methods proposed to include in the economic analysis the effects of social
important theoretical aspect, since it leads to very interesting results such as factors on economic growth via their influence on the labour factor. Here it
the persistence of heterogeneity of individuals and the possibility that social will be argued that a general framework, which can encompass in the
classes2 and a wage structure reflecting not only differences in productivity, economic analysis the effects of social factors, is still lacking. The fourth
but also the social organisation, may emerge. Both of them are themes section contains some proposals for a solution to this problem. The chapter
largely analysed by Classical Theory. concludes with some brief remarks.
Another important feature of this class of models which is a resurgence of
a classical theme, is the assumption that the behaviour of a rational agent also
depends on some extra economic factors almost always related to social 9.2. SOCIAL FACTORS AND HUMAN CAPITAL IN
relations with other agents. For example, in Fershtman et al. (1996) agents NEW GROWTH THEORY
take care of their social position, and the level of human capital is chosen by
considering also the effect on their social reputation, in Gradstein and Although there is no general agreement on the nature of the relation between
Justman (2000) agents undertake an education activity for conformism, while human capital and social factors, the most generally accepted idea is that
in Galor and Tsiddon (1994) parents transfer their human capital to children social factors affect human capital accumulation through different channels.
for altruism. The most important is direct social relations among agents, since individual
The social aspects underlying the agents’ economic behaviour are well human capital can be acquired not only through an educational activity
known by classical authors. Smith (1776) in his analysis of the nature of undertaken in school, but also from other individuals with whom they have
wage highlighted the fact that the reward structure of different occupations social contacts. Normally, individuals from whom knowledge can be
acquired are agents whose services can be bought in the market, but this is the utility or pay-off an individual receives from a given action depends
not always true. They can be relatives and parents, for example, who transmit directly on the choices of others in that individual’s reference group…’
their knowledge without receiving recompense for it. In such cases, (Ibidem, p. 235). These are therefore relations among individuals of
knowledge is transmitted because of the relations among individuals in the economic importance, because the action of one agent influences the chosen
absence of a market and without a price being formed for it. action of another agent with whom s/he is directly or indirectly linked. The
Another channel is the culture, the norms and beliefs that characterise a main hypothesis is that agents influence each other through their actions and
community. The latter constitute a considerable part of the human capital that not through other media like, for example, directly exchanged information.5
individuals possess, and are transmitted4 to all the members of the Another crucial hypothesis is that this influence comes about directly, in the
community only because they belong to it. sense that it does not operate through the market: the individual modifies
Finally, social factors may influence the level of education since they her/his rational choice simply by observing the actions of other agents.
modify the incentives to acquire new knowledge. An example is the case Generally the literature distinguishes between local and global
when individuals desire more human capital not to earn a higher income, but interactions. Global interactions arise when an agent is able to interact with
to acquire a higher social status or conform to their group. The incentives in any other agent in the economy. Local interactions are cases in which agents
this case lie in the social relations among individuals who attend to their interact only with some specific group of agents. In the latter case, a set of
relative social position. neighbours must be defined, and the notion of social distance between agents
Second-generation growth models have included these factors in their is also required.
analysis of human capital accumulation. However, the set of analytical One major difficulty concerns the way in which the effects of such
categories used until now, denoted here as social factors, is rather interactions can be included in the choice problem. Three different
heterogeneous. There are in fact models which have focused mainly on the approaches exist in the literature: (i) it is assumed that social interactions
effects of social interactions (Benabou, 1996 and Durlauf, 1996), where the modify the constraints under which the rational choice is made; (ii) they
term refers principally to interactions among single individuals or between influence the formation of expectations; (iii) or modify the preference
these and reference groups. Others have instead emphasised the role of social structure.
capital, by which is meant a broad and heterogeneous set of phenomena Constraint interactions occur when an agent’s action modifies the choice
including the social norms and institutions that characterize a society set of other agents. An example is provided by the congestion or spillover
(Coleman, 1988). Finally, some have included culture and ideology (Cozzi, effects due to knowledge diffusion. Interactions through expectation
1998; North, 1981; Iyigun et al., 2001) among society’s ‘social assets’ which formation are well known, and their analysis pervades information theory.
influence the formation of human capital. There is, therefore, a plethora of Economic analysis assumes that an agent forming expectations may seek to
concepts and analytical categories which seem difficult to sum up in a single draw lessons from observation of the actions chosen by others. Preference
term. However, all of them relate to phenomena that spring from direct interactions occur when an agent’s preference over the alternative in a choice
interdependence among individuals, ‘direct’ in the sense that it is not set depends on the actions chosen by other agents, examples being provided
mediated by market mechanisms. In what follows, I shall analyse the various by conformism, jealousy and envy. This way of modelling the interaction
analytical categories used to grasp the effects exerted by social factors on among agents can be traced back to Smith (1759) who recognised that agents
agent formation and on economic performance. interact through preferences6 and that the intensity of interactions varies
according to the strength of the relationship.7 Also this last idea is embedded
9.2.1. Social Interactions in social interaction models, since another key assumption is that the
intensity of the effects of social interactions depends on the strength of the
Social interactions constitute one of the most widely used analytical relationship among agents, usually captured by the social distance function,
categories to describe the effects of social factors on the labour supply in used to gain a measure of the social ‘nearness’ between two or more agents
terms of efficiency units. Their effects on the growth process have been and the degree of interaction between them.8
analysed mainly by Benabou (1996) and Durlauf (1996).
A clear definition of social interactions has been provided by Brock and
Durlauf (2001), as follows: ‘By social interactions we refer to the idea that
9.2.2. Social Capital suffer from the shortcoming of confusing social capital with its possible
positive effects. Moreover, many of the concepts outlined above (social
Social capital is another of the concepts used to specify the influence exerted norms, for instance, or the quality of the governing institutions) have already
by social factors on the behaviour of economic agents. Providing a definition been analysed in the literature without it being found necessary to introduce
of the term is difficult, since in this case there is no general agreement on its a new analytical category.
meaning in the literature. Moreover, the concept often overlaps with that of Although social capital may be a useful concept insofar as it underlines
social interactions. One of the best known definitions is provided by Putnam the sociality of individual action, and the effect of this on economic choices,
(1993, p. 167): ‘social capital ... refers to features of social organization, such it is either ill-defined or redundant. The definition proposed by Durlauf
as trust, norms, and networks, that can improve the efficiency of society by (1999a) deals with the former problem but makes the concept of social
facilitating coordinated actions’. capital very similar to that of social interactions. Involved once again are
A similar definition is provided by Coleman, to the effect that: ‘Social relations among agents not mediated by the market; only that in the case of
capital is defined by its functions. … (it) consists of some aspects of social social capital the role of the reference group is more stressed.
structures, that facilitates certain actions of actors. Like other forms of
capital, social capital is productive making possible the achievement of 9.2.3. Culture and Ideology
certain ends that in its absence would not be possible’ (1988, p. S98).
According to this second definition, social capital consists of the The concepts of culture and ideology (or religion) are two further analytical
mechanisms that facilitate the coordination of individual actions so that a categories introduced by the new growth literature to analyse the effects of
superior outcome is achieved. As Durlauf (1999a) and Woolcock (2000) social factors on the creation of new knowledge and on economic
have noted, the problem with this type of definition is that it concentrates on performance (Casson, 1993; Cozzi, 1998; Gray, 1996; Iyigun et al., 2001;
the possible effects while ignoring the mechanisms that create social capital. Lazear, 1999).
It thus confuses a positive outcome with what has made that outcome Culture is defined by anthropologists in a variety of ways, but whatever
possible. the definition, it usually includes some notion of shared values, beliefs,
A definition which overcomes this problem has been suggested by customs, rituals, language, and so on. Some authors (Gray, 1996)
Durlauf (1999a), for whom social capital is ‘the influence which the hypothesise that culture is a public good that affects the propensity of agents
characteristics and behaviours of one’s reference group have on one’s for various economic activities. Cozzi (1998) suggests that culture is a ‘social
assessments of alternative courses of behaviour’ (Ibidem, p. 2). Rather than asset’ that increases the productivity of labour measured in terms of
emphasising the ‘productivity’ of social capital, this definition stresses the efficiency units, and which also accelerates the pace of technological
‘sociality’ of individual behaviour. It highlights the important role of non- innovation. The reason is that, although culture does not have an immediate
market relationships in determining individual and collective behaviour, productive use, it shapes individuals’ behaviour and thereby their productive
allowing the sources of social capital to be separated analytically from its capacity.
consequences. Another analytical category often used is ideology (or religion) (North,
Besides the difficulty of coming up with a general definition of social 1981, Iyigun et al., 2001). According to some authors (Sacerdote and
capital, there is also the problem of defining the forms that it assumes. Glaeser, 2001, North, 1981), ideology is a particular form of knowledge
Woolcock (2000) proposes a scheme in which social capital may assume which enables generalizations to be made about the environment within
four dimensions: (i) the extent of horizontal associations; (ii) the nature of which agents operate and completes the factual information that they
social ties within communities (the degree of trust, peer effect, etc.); (iii) the possess. More specifically, by embracing an ideology, individuals increase
nature of the relation between civil society and the state and (iv) the quality their ability to acquire knowledge, and this affects positively the productivity
of the governing institutions. of labour factor.
Coleman (1988) proposes a similar scheme, although he places greater The positive relation between ideology and labour productivity has been
stress on the role of collective norms, concluding that social capital assumes hypothesised by several authors. Rosenberg and Birdzel (1986) point out that
three forms: (i) obligations and trustworthiness of structures; (ii) information the development of a moral system commensurate with wealth and capital
channels; (iii) norms and effective sanctions. Yet these classifications, too, accumulation can be traced to the Calvinist Reformation of Protestantism.9
Franke et al. (1991) and Gray (1996) find evidence for a positive correlation social status allows a high level of consumption of such goods to be
between certain Confucian values and economic growth in samples achieved. This implies that the demand for social status derives from the
comprising both Western and Asian countries. Iyigun et al. (2001) embed presence of social interactions by which exchanges of non-marketable goods
this idea in an endogenous growth model where ideology and education are and services occur. Hence it is only another aspect of the more general
substitutes and interact to influence technological progress. phenomenon of non-market interactions.
Although the concept of ideology is useful since it unifies in a single term However, the existence of decisions that affect consumption but are not
phenomena such as beliefs, moral and political attitudes that influence the mediated by price mechanisms can also lead to another type of behaviour, in
behaviour of an individual, no significant difference between this and the contrast to the concern for status: the desire to conform. A conformist
concept of culture seems to emerge. Both refer to a particular type of behaviour may emerge especially when there are activities that are
knowledge which is shared by a multitude of persons, and have a undertaken in groups. It refers to an inclination of an individual to behave
pronounced normative content. Furthermore, they do not seem to differ like the other agents belonging to his/her reference group. Also in relation to
greatly from the concept of social capital. this different mode of social interaction the effects on human capital
formation and on economic performance have been analysed. Gradstein and
9.2.4. Social Status and Conformism Justman (2000) propose a model where individuals gain utility from
conformist behaviour by reducing the social distance between themselves
Social status is a ranking of individuals (or group of individuals) in a given and their reference group. They show that such behaviour may have perverse
society, based on their traits, occupation, consumption, assets and actions effects on growth because it may reduce the returns on investment in
(Weiss and Fershtman, 1998). The sociological literature (Davis and Moore, education.
1945; Treiman, 1977) has shown that high social status is usually awarded to At the end of this discussion, it seems clear that most of the concepts
wealthier individuals and to those who have an occupation requiring a high presented here suffer from a lack of definitional clarity. Moreover, the
level of human capital. differences between them seem to be quite marginal. Social capital, for
The economic implication of this social phenomenon has long been example, is a result of interactions among agents: the trust-degree, which is a
recognised by economists,10 who have largely analysed also the implications dimension of it, is only a form assumed by social interactions. Also culture
for growth.11 The influence of social status on growth has been assigned and ideology or rank concern are only particular results of the latter. All the
principally to its effects on the saving rate (Cole et al., 1992; Corneo and concepts presented here are only different aspects of the same phenomenon:
Jeanne, 1997) and on the demand for positional goods (Funk, 1996; Hirsh, the social exchange of knowledge, information, etc. which occurs among
1976), while only few have recognised its growth effect via the influence on agents, social because it is not mediated by the market and which for this
human capital (Fershtman et al., 1996). In the model of Fershtman at al. reason the standard economic model, if not appropriately amended, is unable
(1996) agents attend to their social status, obtainable by undertaking an to capture.
occupation that gives high social prestige. Following Smith (1776), they
assume that the latter is an attribute of those occupations which require high
human capital. Therefore, the demand for social status may constitute a 9.3. ANALYTICAL METHODS TO ENCOMPASS
strong social incentive for the accumulation of human capital, which is added SOCIAL FACTORS IN ENDOGENOUS GROWTH
to the monetary incentive.
MODELS
On the existence of this positive relation between human capital
accumulation and concern for rank there is general agreement, but it is not so
Besides of the above-cited problems, analysis of the effects of social factors
clear why individuals are concerned about their relative social position.
also suffers from a lack of a generally accepted analytical framework. It is
According to Postlewaite (1998), the rank concern arises instrumentally
not an easy task to encompass social factors in economic analysis given the
because relative standing influences the consumption level. As a matter of
fact that some of them refer to aggregate concepts. Indeed literature has
fact, because of market imperfections not all goods or services can be
almost universally viewed social capital, culture, norms and beliefs as
acquired through the market. When the allocation of some goods or services
community level attributes, but since economic models are based on decision
such as information, education or other does not occur via the market, high
maker agents, aggregate definitions may impede the inclusion of social
factors in the economic framework. In spite of such difficulties, there have categories. A second problem, strictly linked to the first, is that it is not clear
been several attempts that can be summarised in two analytical strategies: in what way ideology or culture is formed and how their formation process
one where social factors modify the constraints under which the optimal can be embedded in an individual choice-based model.
choice is made, and another, more general strategy, where they modify the Another strand of literature which uses this analytical method involves
objective function or the preference structure. attention to social interactions. In these models one typically postulates that
human capital is formed by way of social relationships because these favour
9.3.1. Models where Social Factors Modify Constraints knowledge transmission among individuals (Benabou, 1996; Durlauf, 1996;
Galor and Tsiddon, 1997; Hassler and Rodriquez Mora, 2000). Different
An initial example of the adoption of this kind of strategy can be found in cases of social knowledge transmission are identified: one of great
models where social factors modify the rewards structure by favouring or importance is that occurring within the family because of the close and
reducing human capital accumulation rate. Papers by Acemoglu (1995), enduring relationships among its members. Yet also the relationships arising
Baumol (1990) and Murphy et al. (1991), for example, have analysed the between members of the same group or community are important channels
effects of social factors on the allocation of talents, showing that if these through which knowledge is transmitted. The models developed by Benabou
make rent-seeking activities more remunerative than productive ones, the (1996) and Durlauf (1996) attach greater importance to the latter channel of
economy grows at a lower rate.12 In all these models social factors influence knowledge transmission.
the individual human capital indirectly, since they affect the relative In particular, Benabou’s model highlights how articulate social relations
convenience of its allocation between different sectors. can be and how their possible results in terms of human capital level and
Another way of modelling the influence of social factors on human capital growth rate may depend on the forms which they take. He postulates that
is the assumption that social factors directly affect the ‘production social interactions at community level may give rise to very different results
technology’ of human capital. In this case the constraint which is modified is in terms of human capital according to which form they take. If individuals
not the rewards structure, but the human capital production function. with greater human capital exert considerable influence in the group (that is,
Several authors follow this approach. Cozzi (1998) for example, assumes the more educated members of the group are emulated), a community made
that culture directly affects the efficiency of labour factor, and Iyigun et al. up of heterogeneous individuals will be more efficient in terms of the human
(2001) follow a similar argument. According to these authors, education capital produced, because the ‘high tails’ of the distribution will prevail.
acquired through formal schooling and ideology are two productive factors Vice-versa, if the influence of those with a low human capital predominates,
that are substitutes in the human capital accumulation function. The idea an increase in the proportion of high-quality individuals will have a
behind this assumption is that a more sophisticated ideology allows us to negligible effect on human capital since the ‘low tails’ of the distribution will
obtain more accurate knowledge about the facts, such that individuals with predominate. These different types of social interactions can be formalised
different ideologies but the same level of education make different inferences by two production technologies of hi,t +1 . The first case is,
about the world.
In both models the production function of human capital can be hi,t +1 = di (i, l)ε (σ )hl ,t (2a)
represented as,
while the second case,
hi,t +1 = f ( Et ; I t ) (1)
di (i, l )hl ,t
hi ,t +1 = , (2b)
where Et is the education acquired in formal schooling and I t is the ε (σ )
ideology (or the culture) which can be ranked according to its degree of
sophistication.13 where di (i, l) is the social distance function between individual i and l
Although these models contain a clear improvement with respect to the reference group, here considered as an exogenous parameter, 0 ≤ σ < +∞ is
simple Beckerian model, given by the recognition of the role played by an index of the variability of the distribution of h over the group, ε ( σ) is an
social factors in the individual formation process, they still present several increasing function of this variability ( ε (0) = 1 ), and hl is the average level
shortcomings. First of all, they are based on poorly defined analytical of human capital in l group.
Equation (2a) states that when the influence of more able individuals hi,t +1 = h j ,t di (i, j )α ( I , h j ,t ) xt ( h j ,t ) (3)
prevails, the transmitted human capital is above the average of the group, and
the intensity of the transmission increases with an increase in variability of where xt is the amount of resources invested in education, d (i; j ) is the
the distribution of h over the group. Vice-versa, equation (2b) states that if social distance between parents and children, which is weighted by a factor
the influence of less able individuals prevails, the transmitted human capital α ( I , h j ,t ) . This factor is increasing in parents’ human capital because
is below the average, and an increase in variability further reduces this individuals with parents who possess high h learn more quickly and are able
transmission. to use their knowledge more efficiently, and depends on some family
Although neither this model analyses the formation of social interactions, characteristics (I ) , such as the possibility for the parents to spend many
it shows that their introduction helps to describe the formation of individual hours with children, or whether the family structure is a ‘nuclear’ or
human capital in a clearer and more detailed manner by allowing us to ‘extended’ one. These institutional characteristics, as Coleman (1988) has
distinguish cases where social relations have positive effects, from cases emphasised, influence the quality of family relationships and hence the
where their effects are negative. Moreover, through the social distance effectiveness of knowledge transmission.
function it is possible to make social interactions endogenous. In fact, In a more recent work, Galor and Tsiddon (1997) considered another
although in the Benabou model the social distance function is an exogenous feature affecting the transmission between parents and children. In this
parameter, the analysis could be extended by considering it as a choice paper, they assume that knowledge transmission depends also on the
variable which comes from a decisional process where an individual chooses occupations chosen by children. If they choose the same occupations as their
the optimal social distance. parents, then human capital transmission is complete.17 This happens
Another aspect greatly emphasised by this literature is that the effects of because in performing a job, individuals develop a set of interpersonal
social interactions on human capital accumulation may explain the dynamic relations that constitute a sort of ‘social capital’ to be used in their work,
of the relation between inequality and growth.14 In fact in the models of which can be transmitted to the children only if they remain in the same
Benabou (1996) and Durlauf (1996), an equilibrium may emerge, where sector.
different groups of agents are formed. In this case individuals differ in their This case is particularly interesting since it shows that individuals are not
level and possibility of further accumulating human capital. members of a single group, but they are involved in different types of social
This result is particularly interesting since these groups can be considered relations, and it may be relevant to analyse how the relations developed in
as different social classes, each being characterised by a different level of different contexts interact with each other. This idea has been well defined
human capital and a different possibility of further accumulating it. Such a by Gellner (1996), who introduced the concept of ‘modularity’ of
class structure, however, is radically different from that of the capitalists– individuals, which means that individuals define themselves by multiple
th
workers dichotomy that was prevalent in Europe in the 19 century and attributes associated with distinct spheres of social life. This implies that an
th
beginning of the 20 century, since in this case there would be no single class individual may belong to different groups. In some cases there is no relation
of workers, but a whole range of worker classes, each characterised by a between the attributes an individual has in a group with those that s/he has in
different level of human capital and a different access to the sources of another, but in other cases this relation exists, and it may influence inter-
human capital production.15 group interactions, as in the case analysed by Galor and Tsiddon (1997).
The models of Galor and Tsiddon (1994 and 1997) and of Hassler and
Rodriquez Mora (2000) mainly focused their attention on knowledge 9.3.2. Models with Social Factors which Modify Preferences
transmission occurring within a family. In Galor and Tsiddon’s (1994)
model, parental relations affect the level of human capital through two The analytical strategy followed by the models surveyed above, even if it is a
variables: the amount of parental knowledge and the resources invested by straightforward way to include social factors in economic analysis, is
the individual in education, which depends on the parents’ human capital. 16 actually rather limiting, since the decisional process which leads to the
Moreover, the authors assume that the parent/child relationship affects choice of a particular behaviour is not explicitly analysed.
human capital formation not only by transmitting directly human capital to A different way of modelling the growth effects of social factors is to
their offspring, but also by modifying their cognitive capacity. Formally, assume that they modify the preferences structure or, more generally, the
knowledge transmission at the family level can be written as: objective function. In this way the formation process of the social factor
emerges as a solution of an individual decisional problem. However, even if proposed by Glaeser et al. (2000); another can be found in recent papers
this method may be more suitable to overcome the above cited problems, belonging to the literature on non-market interactions. In what follows we
within the new growth theory it is not so widely used: the few papers which will discuss the relevant features of these two approaches that, if embedded
use it largely coincide with those that have focused on the demand for social in a growth model, may lead to important developments for the analysis of
status. the effects of social structure on economic performance.
Generally it is assumed that social status depends on some individual
traits, usually one’s wealth or the level of education, and that an agent 9.4.1. The Individual Social Capital Investment Approach
chooses the level of wealth or of education, in order to maximise his/her own
utility function, where the latter is defined not only over a set of market In this model agents can accumulate individual social capital in the same way
goods, but also over the social prestige s/he obtains. However, social prestige they do with human and physical capital. Individual social capital is defined
can be defined only in relative terms. Consequently, to establish the relative as ‘a person’s social characteristics … which enable him to reap market and
social position of an individual, it is necessary to identify the reference group non market returns form interactions with others. As such, individual social
with respect to which the individual’s social status is defined. Therefore, capital might be seen as the social component of human capital’ (Glaeser et
most of these models have to define ‘a priori’ the neighbour structure to al., 2000, p. 4). Moreover, this particular kind of investment is assumed to be
which the individual refers. time-consuming, and as such has an opportunity cost given by the hourly
Formally the individual choice problem is, wage.
This definition enables the authors to use the standard model of optimal
hi∗ ∈ arg max Ui ( p, s( hi , hl ) ) (4) individual investment decisions to analyse the formation of individual social
capital. By applying this very standard model, they are able to obtain
where p is a vector of personal characteristics, including income, hl is the interesting results: a relation between the lifecycle of an agent and the
average level of human capital in the reference group, s ( shi > 0 and shl < 0 ) individual investment in social capital, which is positive in the early stages of
is the social status function, implying that an individual obtains greater status life and negative in the latter stages; a negative relation between mobility and
if s/he differentiates her/himself from own reference group. social capital investment, which implies that what reduces mobility, such as
Although in this case the social factor (i.e. individual social status) is a homeownership, also increases social capital investment; a positive relation
choice variable whose level emerges as a solution of a maximisation between the patience rate across individuals and social capital investment,
problem, this framework is still not sufficiently general, since the formation which generates a reduced form correlation between the latter and human
of the reference group and, more generally, the structure of the social capital accumulation.
interaction environment do not emerge as an equilibrium solution. An important feature of individual social capital, characterising also
Once again, there is a clear need to elaborate a more general framework human and physical capital albeit to a lesser extent, is the presence of major
that is able to explain how economic variables interact with the social externality and positive complementarity effects. Complementarities raise the
environment, making the formation of the latter, at least partially, possibility that there exist multiple equilibria in the levels of social capital
endogenous to the model. Moreover, it would be worth constructing a investment, and explain how small differences in initial conditions may
general framework which can unify the various models and the different generate large divergences in long-run levels of social capital. Moreover, the
analytical categories used to analyse the effects of social factors on economic presence of externalities and complementarities implies that the transition to
performance. the aggregate level is not immediate. Indeed aggregate social capital is
defined as the average of individuals’ social capital, adjusted for all the
externalities, which are of a considerable amount and can be positive or
9.4. FURTHER DEVELOPMENTS negative. This latter feature depends on the type of individual social capital
accumulated. For example, joining a network is a form of individual social
Recently, there have been some interesting developments in two different capital that creates positive externalities, while status-seeking behaviour,
strands of literature that may constitute some possible solutions to the above which is another form, causes negative externalities to other agents in the
discussed problems. One is the model of individual social capital investment, reference group.
This approach has the great advantage of being able to give a unified An interesting contribution is provided by Glaeser and Scheinkman
interpretation for the different concepts used to analyse the effects of social (2000) who present a general model that is able to treat as special cases
factors on economic variables, using an analytical format very familiar to several of the better known models in this area. Hence, to clarify the main
economists. In this framework, social interactions are only different forms of features of this approach, I will refer to the analytical format presented in this
individual social capital: joining a group and status-seeking behaviour are model.
only two of the many forms of individual social capital. Moreover, this In Glaeser and Scheinkman’s model utility function includes the
model provides a coherent interpretation for the positive relationship individual’s actions, the actions of agents within the reference groups,
between human capital and social capital: these are two distinct accumulable individual’s personal characteristics and common prices. The reference
factors that show a positive relationship between them, because of an equal groups may include only individual’s closest neighbour or the entire
response to changes in the intertemporal preferences rate of individuals. economy. Therefore this framework can examine both local and global
Nevertheless, by adopting this approach one may lose a major aspect interactions. Formally the individual problem is
highlighted by most of the papers analysed here: the fact that individual
decisions are influenced by actions of other agents in a way that is not max Ui = Ui ( ai , Ai1 ,....., AiK ,θ i , p) (5)
regulated (because it is not possible) by the market. By assuming that one
where
can define a market price for the investment in social capital, it is difficult to
n
maintain this specific nature of personal relations, which cannot be regulated
AiK = ∑ γ iK, j a j
by the market and, consequently, cannot be regulated by a price mechanism. j =1
Even if this aspect is treated through externalities, which express the direct
effects of agents’ actions on others, nevertheless, a direct analysis of these ( γ iK, j ≥ 0 and ∑ γ iK, j = 1 ), ai ∈ I 18 is the agent i action, p is a vector of
phenomena could be more correct, because the incompleteness of markets parameters and θ i is a ‘taste shock’ of each agent. In other words the utility
does not mean that an individual is unaware of the effects of such personal of an agent i depends on his own chosen action, on a weighted average of the
interactions. By confining them within the narrow space of generic actions chosen by agents in his/her reference groups ( AiK ), on his/her taste
externality effects, one makes this intuition unable to clearly emerge. shock, and on a set of parameters.
These considerations suggest that the most suitable framework for An interesting result of this model is that multiple equilibria may arise
analysing the economic effects of social factors, that least distorts their real even with very little heterogeneity. In this case two populations with slightly
nature, may be the non-market interaction framework, where each person’s distinct realisations of the θ i ’s could exhibit very different average values of
actions change not only because of the change in fundamentals, but also the actions. This happens if the marginal utility of an agent’s own action is
because of the change in the behaviour of his/her own neighbours. more influenced by change in the average action of his/her peers than by a
change on his/her own action. In other words multiple equilibria occur when
9.4.2. Non-market Interactions Approach the group effect is strong enough. An implication of this result is that it is
possible to analyse how different groups of agents emerge as an equilibrium
Another possible solution may be supplied by the literature on social solution.
interactions, which uses the random field approach, imported from statistical This model also has the potential to facilitate a more rigorous analysis of
physics, where one typically postulates individual’s interdependence and social capital. In fact, if social capital is interpreted according to the
analyses the macrobehaviour that emerges. Follmer (1974) was the first definition proposed by Durlauf (1999a), by which social capital is ‘the
paper to use this framework. Other models inspired by statistical physics are influence which the behaviors of one’s reference groups have on one’s
Scheinkman and Woodford (1994), that studies the impact of independent assessment of alternative courses of behavior’ (Ibidem, p. 2), this can be
sectoral shocks on aggregate fluctuations, Glaeser et al. (1996) who use the identified, at least as a first approximation, as the weights in the AiK terms,
voter model to analyse the distribution of crime across American cities, and because the latter terms capture the influence that the average behaviour has
Brock and Durlauf (2001), who develop a model which extends the random on the optimal choice of an agent. Moreover, as Brock and Durlauf (2001)
field approach to global interactions, to the case of discrete choices. have shown, by choosing the weights parameters appropriately, it is also
possible to consider cases where agents wish to differentiate themselves from 5. This is a rather restrictive hypothesis, in fact, since humans do communicate about all
their own reference groups (status-seeking behaviour). manner of things. Nevertheless, the assumption proves very useful because it considerably
This approach could be very useful to highlight the sociality of an simplifies the analysis (Manski, 2000).
6. Smith (1759, p. 3) wrote: ‘How selfish soever man may be supposed, there are evidently
individual behaviour, providing, at the same time, a very general framework
some principles in his nature, which interest him in the fortune of others, and render their
that includes as special cases all the possible ways in which social factors happiness necessary to him, though he derives nothing from it, except the pleasure of seeing
influence individual economic decisions. Moreover, if it is applied to it’.
continuous actions, it can explain how human capital formation depends 7. ‘Every man feels his own pleasures and his own pains more sensibly than those of other
strictly on social relations, emphasising that individuals are aware of such people … After himself, the members of his own family, those who usually live in the same
house with him, his parents, his brothers and sisters, are naturally the objects of his warmest
influences and they can act to modify their social interaction environment.
affection’ (Smith, 1759, p. 321).
8. More specifically, given a social space which comprises the structure of the agents’
neighbourhoods, the social distance between them is defined as ‘the number of links in the
9.5. CONCLUSIONS shortest path between the agent’ (Kirman, 1999, p. 24). According to this definition, those
agents that are directly connected have a social distance which reaches the maximum value.
9. Max Weber (1930) developed the same idea, attributing the rise of capitalism and the
The literature analysed in the previous sections constitutes a new and Industrial Revolution to the Calvinist Reformation.
interesting strand of research which seeks to integrate non-market mediated 10. As we have already noted, Smith (1776) has recognised that individuals chose an occupation
social interactions into the analysis of economic growth. There has clearly also for the reputation that they can acquire. Marshall (1890) also have noted that: ‘The
been an excessive proliferation of concepts and analytical categories, none of desire to earn approval, or to avoid the contempt, of those around us is a stimulus to action
… in any class of persons … A professional man ... will be very sensitive to the approval or
which are well defined, with all referring to the same phenomenon: the
disapproval of those in the same occupation’ (1890, reprinted 1962, p. 19).
interdependence of agents not regulated by market mechanisms. In this paper 11. See for a comprehensive survey Weiss and Fersthman (1998).
I have set out to argue that by basing the analysis on the better defined 12. Carillo and Zazzaro (2001) have developed a neo-Schumpeterian growth model where social
concept of social interactions, it is possible to obtain a coherent and well- factors, such as the ‘professionalization process’ and the status seeking behaviour of
grounded analysis of the effects of social factors on economic growth. professionals, modify the reward structure reducing the convenience to devote human
capital to R&D sector and by slowing down the pace of technological innovation.
Consequently, an important stage will be to encompass in the endogenous
13. Iyigun et al. (2001) assume that ideology is in its turn affected by the level of education.
growth framework models designed to analyse non-market interactions. This assumption combined with the possibility that ideological beliefs affect the human
Finally, an interesting finding from this literature is that social interactions capital accumulation process generates a feedback loop between ideology and human
accentuate the heterogeneity of agents and may create distinct groups of capital.
agents. Each of these groups has its own specific rules of behaviour and 14. In fact, this approach gives rise to what has been denominated as the membership theory of
inequality (Durlauf, 1999b) according to which income distribution depends not only on
different levels of transmittable human capital. This entails the emergence of
individual characteristics, but also on characteristics of those groups to which an individual
a class structure based on the level and/or type of human capital possessed by belongs.
individuals and on the possibility of further accumulation of such capital. 15. It is interesting to note that this class structure seems to describe well what is observed in
th
most OECD countries in the last part of 20 century since the end of seventies. In fact in
almost all OECD countries it has been observed a strong increase in educational wage
differential, with a spectacular increase in the return to education (Acemoglu, 1999; Aghion,
NOTES Caroli and Garcia Pegnalosa, 1999; Goldin and Katz, 1999).
16. This happens because parents with high h are assumed to be willing to invest more in their
1. See for example the F-connection theory of Ben-Porath (1980). children’s education since they have altruistic preferences and the welfare of their children is
2. Where social classes are defined according to their function in the production process and a normal good.
the posses of an accumulating productive factor. 17. The same line of research has been pursued by Hassler and Rodriquez Mora (2000), who
3. In this line of research are Galor and Moav (2000) who propose an evolutionary framework also emphasise that parental transmission is complete if there is no social mobility. The
where some types of behaviours, such as preference for ‘quality’ children rather than for transmission of knowledge takes place between parents and children only within the
‘quantity’ children, emerge as a result of an interplay between the technological progress entrepreneurial class and it is maximum if there is no innovation.
and the fertility rate. 18. I can be a discrete set or an interval of a real line, therefore this framework can analyse
4. For example by the communication media, the school, etc. discrete as well as continuous choices.
evolutionary model can account for all aggregate regularities at the same
time, although a certain degree of consistency emerges between the different
models (Dosi and Nelson, 1994). This is due to the fact that economic
growth is a historical process, which interacts with, and is influenced by,
many factors outside the economic domain (e.g. culture, institutions, etc.)
10. The evolutionary perspective on (Verspagen, 2001). Moreover, due to the microeconomic perspective of
growth economic change adopted by evolutionary theory as well as the concern with
ex post explanation rather than ex ante prediction, all these models are rather
different in terms both of analytical structure and issues analysed. Therefore,
Grazia D. Santangelo this does not intend to be an exhaustive review of evolutionary growth
modelling. Nonetheless, we believe that the models herein discussed,
selected on the basis of classification made by Dosi and Nelson (1994),
10.1. INTRODUCTION provide a broad idea of the main features of the evolutionary perspective on
growth. The common denominator of all the models reviewed here can be
The interest in economic growth, central to the classical writers, re-emerged identified in the explicit firm-level microfoundation.1
during the post-World War II period. The revised version of Solow’s model The paper is organised in 6 main sections. The next section provides an
(1957) has long been the most successful attempt in explaining total output overview of evolutionary theory as an alternative to orthodoxy in economics
growth. However, the widely recognised weakness of the model was the as presented in Nelson and Winter’s seminal book (1982). The aim is to
identification of technological change with the ‘residual’, which did not provide a background for the following sections as well as an understanding
allow the importance of technological change to be specified. The main of the historical emergence of the evolutionary perspective. In this sense, it
problem appeared to be the impossibility of distinguishing an increase in per should be underlined that, after the 1982 watermark, evolutionary theory has
capita capital due to a move along the production function from an increase developed its own research agenda without benchmarking against orthodox
due to a shift in this function. results. Section 10.3 reviews the basic evolutionary growth model. Further
Endogenisation of technological change has been the main aim of new developments of evolutionary growth theory are covered in Section 10.4.
growth theory (NGT) which, according to the positivist tradition of Section 10.5 discusses the contributions of the evolutionary perspective to
neoclassical economics, has proceeded by testing particular theories against growth theory. A few conclusions are drawn in Section 10.6.
data rather than describing and explaining phenomena as they are observed
(Nelson, 1996). Therefore, although representing an advance by comparison
to ‘old’ neoclassical growth models, NGT still proposes an understanding of 10.2. AN OVERVIEW OF EVOLUTIONARY (VERSUS
economic growth as a smooth process based upon the concept of a ORTHODOX) THEORY
continuing equilibrium, failing to encompass the historical account of
technical change and related activities and institutions (Ibidem). By contrast, The emergence of evolutionary theory in the economic realm may be framed
evolutionary theory adopts a behavioural approach to individual firms which, within the widespread dissatisfaction with the state of economic theory. This
based on Simon’s (1978) bounded rationality, perceives individuals as sense of malaise was mainly directed towards the inability of orthodox
rational and maximising according to their limited mental processing theory2 to come to grips with empirical reality. In this context, the
abilities, information, and the complex environment in which they operate. evolutionary critique of the mainstream economic approach has addressed
This logical structure allows evolutionary theory to account for the three main fundamentals of the orthodox structural logic (Nelson and Winter,
macroeconomic phenomena explained by neoclassical theory by making 1982). First, the continued reliance on equilibrium analysis (even in its more
more realistic assumptions, as well as investigate new research questions. flexible forms) neglects phenomena associated with historical change, as
The aim of this paper is to provide a critical review of some selected disequilibrium situations are outside the focus of investigation. Second, the
evolutionary models which propose an alternative perspective to mainstream perfect-information assumption (even in its more relaxed forms) limits the
economic growth. It should be, however, underlined that no single explanatory power of the theory when facing complex realities by implying
205
The evolutionary perspective on growth 207 208 The Theory of Economic Growth: a ‘Classical’ Perspective
that economic agents can foresee all possible contingencies and weigh their evolutionary biological theory. Routines, defined as ‘a repetitive pattern of
consequences. Third, the economic rationality of the agent – in the sense that activity in an entire organisation’ (Nelson and Winter, 1982, 97), are
he/she optimises – encompasses clear anticipation, calculation and clarity of persistent features of the firm as well as determinants (together with the
risks taken by this actor in dealing with realistic complexity without external environment) of the firm’s behaviour. This implies that choice-
accounting for confusion about the situation, distraction and mistakes on the making is a firm-specific process. Like genes, routines are heritable as
part of the actor. tomorrow’s plants generated by today’s firms have the same characteristics
Evolutionary theory criticises neoclassical theory, as the latter does not (path-dependency). They are selectable as shown by the fact that firms
offer an explanation of why decisional rules are what they are due to its two owning certain routines perform better and increase their importance over
structural pillars: maximising behaviour and the concept of equilibrium. The time through a continuous selection. This selection mechanism is well
rejection of the notion of maximising behaviour can be traced back to the described by the biological concept of natural selection and evolution, hence
consequent rejection of the distinction made by orthodox theory between the the label ‘evolutionary’ applied to this new economic approach.3 In this
choice set and choosing. In fact, according to the evolutionary viewpoint, the sense, evolutionary theory is concerned with a dynamic rather than static
economic agent engaging in the decision-making process is not aware of all analysis. In this process, firms with different routines compete in a rival
possible alternatives he/she can have, but (some of) those alternatives will be selective process by attempting to differentiate one from another (variety).
discovered during this process by trial and error. In this sense, the Firms’ operations are framed within an uncertain environment by which they
determination of decisions and the decisional outcome involve a stochastic are affected (mutation) and to which they adapt by trial and error
element and may be non-optimal. Therefore, agents’ behaviour cannot be (adaptation). In this sense, the environment pushes firms into continuous
assumed to be prompt and rational. As far as the concept of equilibrium is learning and adaptation (both of which are firm-specific processes).
concerned, although the analysis of a stable equilibrium configuration in Competition is then understood in terms of dynamic selection over time
which particular forces no longer produce change is recognised as important, promoting economic change because new techniques progressively displace
the drastic narrowing of orthodox investigation to the equilibrium situation is outmoded ones. This change proceeds gradually and in a localised manner as
firmly rejected as it allows explanation of the relations among the efficient exploration and development of new techniques is likely to occur in the
survivors, but not the competitive struggle consumed in disequilibria. neighbourhood of the techniques already in use, as already highlighted by
Conversely, the latter aspect is captured in the evolutionary perspective by Atkinson and Stiglitz (1969) in their critique of the neoclassical concept of
the concept of selection. technological progress.
Going into the details of evolutionary economics, this is built upon Drawing on the Schumpeterian tradition, in evolutionary theory change
population thinking rather than typological thinking as in the orthodox arises partly from within the firm and part as the selective effect of the
approach. Indeed, the concern is with frequencies of behaviours that differ, environment. Due to the struggle for survival engaged by firms, technology
not with uniform behaviour. While in the population perspective variety is a is a crucial factor in the selection process. Successful firms are major
natural state, variations around the ideal type are regarded as accidental. technological innovators and/or imitators4 as in the capitalist reality
Each firm has its interpretation of economic opportunities and constraints technology-based competitiveness is a source of major advantages
and, consequently, different firms perform different actions in different (Schumpeter, 1942). This may be better understood when recalling the
ways. This is due to the fact that the firm’s capabilities are embedded in its definition of technology provided by evolutionary theory. Unlike the
organisational structure, which allows it to pursue certain strategies more orthodox tradition, evolutionists distinguish between a public and a tacit
easily than others. Unlike the orthodox theory, in a highly uncertain element of knowledge (Nelson and Winter, 1977, 1982). The former, better
environment there is no global optimisation over a given set of choices defined as potentially public knowledge, can be codified in patents,
comprising all objectively available alternatives. Drawing on the managerial blueprint, textbooks, etc., implying that ‘public’ technology can be easily
school, evolutionary theory goes on to claim that firms as such may not have transferred and, therefore, traded between firms – this corresponds to the
any objective in the sense that what is required to operate in the business neoclassical understanding of technology. The latter is a private element
world is just a procedure for determining the action to be taken, rather than a embodied in the firm’s organisational routines, expertise and skills acquired
defined object function. In evolutionary thinking, this procedure depends through a process of learning, and takes the shape of a firm-specific set of
upon the concept of ‘routines’ which can be understood as genes in practices. If the public aspect of technology can be easily transferred, the
tacit element is non-tradable and difficult to transfer as the result of a process of labour (capital) needed to produce one unit of output. Thus, at any point in
of learning-by-using (Rosenberg, 1982). Therefore, since tacit knowledge is time, a state of a firm is defined as a pair of inputs aLt, aKt and its capital stock
developed through an internal learning process, technology is also partially Kj. An industry state is simply a list of firms’ states [(r1.,K1.), …, (rP., KP.)]
context-specific in the sense that it develops linkages with the local (where P is the finite number of firms potentially in existence). An industry
environment.5 This aspect is strictly linked to the concept of the state in a given time period implies a certain aggregate capital stock ∑ . , a M
technological paradigm6 as well as the concept of institutions (understood in certain aggregate output in use ∑ . D and a certain aggregate labour
M .M
both a narrow – e.g. governments, regulatory bodies, etc. – and broad sense – demand ∑ D D ./M .M
e.g. universities, business culture, tacit patterns of behaviour, etc.). The prices of the output and inputs are determined by the industry state
according to a continuous demand price function, which maps the aggregate
capital, labour and output quantities into a list of non-negative prices of
10.3. EVOLUTIONARY GROWTH THEORY output, capital and labour which ensure market equilibrium. It is also
assumed that this function ensures positive profits for all firms for some
Drawing on the features of evolutionary theory outlined in Section 10.2, an industry states. It is assumed that the industry wage rate is determined
evolutionary model is characterised by dynamic analysis accounting for according to the following rule:
random elements that generate some variation in the variable in question c
(e.g. the firm or the technology) through a mechanism selecting on extant ⎛ Lt ⎞
w = a + b⎜ ⎟ (1)
variation (Dosi and Nelson, 1994). In this process, learning and discovery by ⎜ (1 + g )t ⎟
trial and error play a great role in defining adaptation and mutation in the ⎝ ⎠
variable in question. Based on Schumpeter’s (1942) idea of the centrality of where t is the time period, Lt is the aggregate demand of labour in the period
the firm in capitalist economic development, evolutionary growth models are and a, b, c and g are constants. If g = 0, labour supply conditions are
microfounded in the sense that firms are the key actors in making constants over time and the model as a whole is a Markov process with
investments to develop and adopt new technologies, and in the use of these constant transition probabilities. If g ≠ 0 , then labour supply conditions
technologies to produce goods and services. undergo changes and the model is still a Markov process but with time-
dependent transition probabilities. In turn, the wage rate allows the firm’s
10.3.1. The Basic Model gross profitability to be determined. Gross investment is determined by gross
profit.
The basic evolutionary growth model of Nelson and Winter (1982), In this context, natural selection is formalised by a finite Markov chain.
developed in different stages (Nelson and Winter, 1974, Nelson, et al., From one period to the next, the state of an individual firm changes
1976), is based upon the logical apparatus set forth in Winter (1971) where according to probabilistic rules that depend on the initial state of the firms
decisions, rules, search and selection mechanisms are provided analytically. and its profitability (determined by the initial state, the wage rate and
It is assumed that each firm has the same convex constant returns to scale constant parameters). Since the industry state is the list of firms’ states, the
technology represented by a production set Y in 3-dimensional Euclidean transition probability rules for individual firms define the transition
space where it is assumed that there are two inputs and one homogeneous probability in the set of industry states. It is assumed that there are extant
product. The generic element of Y is indicated by y. That is, each firm has the firms and potential entrants. As far as existing firms are concerned, transition
same technology defined by a 3×M matrix A (where M is the number of probability rules for productive techniques are specified on the grounds of
techniques). Each column is a list of inputs and an output flow that is feasible firms’ gross return on capital according to whether the firms belong to group
in a single plant in a single period, and two rows are strictly negative. 1 or group 2, defined as follows:
Moreover, in each period, a firm is characterised by a stock of capital Kj,
which is assumed to be entirely employed. This structure provides the basis GROUP 1: firms with positive capital in the current state satisfice with
for the mechanism of the decision rule (for production decisions). This rule is respect to their decision rules if they are sufficiently profitable –
represented simply by one column of A and can be formalised for each firm that is if they make a gross return on capital exceeding a target
T
(j) in time t by the vector rjt = (1, aLt, aKt) , where aLt (aKt) indicates the amount
level. Therefore, they retain the production technique of the state inputs and outputs, time path of the industry wage rate and firms’ and
with probability 1 (satisficing); industry rate of return on capital). Setting the original conditions of the
GROUP 2: firms that make a gross return on their capital less than the target model in order to make them correspond to the conditions revealed in
level undergo a probabilistic technique-change process. This Solow’s data, the two authors obtain a smoother aggregated ‘technical
process may entail either seeking incremental improvements in progress’ than that found by Solow in the real data for the US, thereby
the firm’s present methods (local search) or looking at what confirming the gradual and incremental character of technical change
other firms are doing (imitation), but not both at the same time.7 hypothesised by evolutionary theory. Moreover, the simulation model does
generate ‘technical progress’ with rising output per worker, a rising wage
(a) In the local search case, the search is local in the sense that the rate and a rising capital–labour ratio, and a roughly constant rate of return on
probability distribution of what is found is concentrated on techniques capital.
close to the current one. Distance between technique h and h’ is given However, it is worth emphasising that, although in the simulation model
by firms respond to profitability signals in making technical changes and
' '
D (h, h’) = WTL | log a Lh − log a Lh | + WTK | log a Kh − log a Kh | (2) investment decisions, they are not maximising profit. Conversely, emphasis
is placed on corporate behaviour: firms, that are doing well, relax by making
where WTL+WTK = 1. That is, the distance between the two only small changes when they do change their decision rules; firms subject to
techniques is a weighted average of the absolute differences in the logs payout constraints grow by maximising investments. Similarly, the model
of input coefficients. The closer the techniques to each other, the higher does not rely on the concept of equilibrium. At any point in time, there are
the probabilities for transition from one to another. different techniques used and rates of return obtained as a result of the fact
that there are always better techniques not being used. It assumes the
(b) In the case of imitation, the probability that firms looking at what other
existence of a set of physically possible techniques, a subset of which is not
firms are doing will find a particular technique is proportional to the
assumed to be known at each particular time. Rather, this subset is explored
fraction of total industry output produced by that technique in the
historically through an incremental process where non-market information
period in question. Alternative techniques turned up in the search
flows among firms play a major role and firms know only one technique at
process are adopted by the firm only if they yield a higher return per
the time.8
unit of capital than the firm’s current rule.
As far as entrance is concerned, potential entrants are classified into two

groups: 10.4. FURTHER DEVELOPMENTS: THE ROLE OF
TECHNOLOGY DIFFUSION
GROUP 1: firms with zero capital in the current state contemplating the use
of a production decision rule that implies a return to capital The Nelson and Winter (1982) model has been further developed, giving rise
exceeding the target level. These firms become actual entrants to two streams of evolutionary modelling attempting to tie together analysis
with a given size probability less than one; of diffusion patterns and productivity change explaining technological
GROUP 2: firms contemplating rules that yield capital return less than the asymmetries (i.e. gaps between firms in terms of costs of production and
target level and do not enter with probability 1. product characteristics), technological variety (i.e. diversity related to
By means of computer simulation, the authors used this analytical framework differences between firms in their search procedures, input combinations and
to show that an evolutionary model of the sort described above is able to products, despite similarity in their production costs) and behavioural
account for the macroeconomic patterns explained by neoclassical theory, as diversity (i.e. within the same industry firms show different strategies). The
represented by Solow’s 1957 article, where data on gross national product, issues addressed in these models concern the processes involved in economic
capital input, labour input and factor prices over forty years are considered. development. Although development issues lie outside the focus of this
Nelson and Winter (1982) generates the same macro aggregates by building survey, there is scope for including these models as their results (e.g. the
them up from microeconomic data (e.g. time paths of firms’ and industry significance of the learning process internal and external to the firm, of the
proprietary aspect of technology, and of embodied technological change) can determined as a dynamic compromise between the desired mark-up on unit
provide significant insights in explaining growth. costs and relative competitiveness.
In discussing these models, a distinction is made between models The core of the model concerns the comparison of two technological
allowing two technologies (an old and a new one) and models allowing a trajectories representing at any time the maximum productivities achievable
multitude of technologies. Common to both models is the elimination of the in best practice vintages of respective technologies. It is assumed that these
stochastic element of the new-technique generation focusing on a selection technologies change at some rate and that the second is always superior in
of techniques that are initially in use. In the discussion, emphasis is placed on productivity. Initially, all firms use technology 1, which is already mature in
the main results obtained, while providing a brief intuitive overview of the the sense that skills levels are saturated at 100 per cent. As far as technology
model. 2 is concerned, firms possess lower efficiency and are unaware of the margin
for further development. Therefore, they can only guess at the rate of further
10.4.1. Diffusion in a Two-technology Model improvements in efficiency and technological progress. This reflects the fact
that the productivity of a technology concerns specific expertise and
In the model proposed by Silverberg, Dosi and Orsenigo (1988), industry experience internal and external to the firm rather than the machines used.
level demand is taken as given and growing at some exponential rate. Firms Therefore, investment decisions rely on the ability of evaluating the
own some market shares of this demand at any given time, although market prospects for further development either by acquiring experience with the
shares may change over time with a characteristic time constant as a dynamic best practice technology or by waiting for the right moment in order to avoid
response to disparities in relative competitiveness. For each firm, possible development costs. The rate of change of the internal skills level of
competitiveness is defined as a linear combination of market price and firm i using the new technology, si, follows the rule
current delivery delay. Firms’ market share changes according to the
deviation of the firms’ competitiveness from the industrial average APi
si = si (1 − si ) if si > sp (3)
competitiveness. It is assumed that entrepreneurs are aware of the process of CPi + C
economic growth and technological change, whose developments are taken
into account when deciding on their fixed investments. Thus, the decision- where A is a parameter, Pi is the firm’s current production, CPi is its
making process is based on certain rules of thumb and animal spirits in the cumulated production with the new technology, C is a constant proportional
form of decisional rules governing replacement policy and expansion to corporate capital stock and sp is the level of skill generally available in the
capacity. Technical change is embodied in vintages. The capital stock of industry also to those firms which are not yet producing on the new
each firm is represented as an aggregate of gross investment made between technological trajectory. This ‘public skill’ can be thought of as skilled
the current period and the scrapping date. This capital stock may consist of labour and management moving between firms, operating instructions of
different technologies as well as different vintages of a single technology some industrial equipment diffused by manufacturers to users, industrial and
trajectory. The change of capital stock over time is defined as a linear trade publications, etc. Thus, the rate at which the level of generally available
combination of net expansion, gross investments and removals due to skill changes can be formalised as
scrapping. The desired level of capacity expansion may be set initially at any
( )
.
level, but revised over time in order to account for the deviation of the rate of s p = A s − sp (4)
capacity utilisation from its desired level. Labour is the only current
where s = ∑i fi si with constant parameter fi and N is the number of firms
production cost firms face. The total quantity of labour per unit of capital is N
given by the average of the technical labour-output coefficient at different in the industry. Firms enjoy this learning
. . externality even if they are not yet
times comprised between the current period and the scrapping date weighted employing the new technology: s i = s p if si = s p . In deciding whether to
by the vintage. Changes in the total quantity of labour over time are due to switch to the new technology firms may abandon their investment criteria
more productive equipment through investments and removal of equipment indicated above in order to take into consideration the gains in productivity
due to scrapping. Changes in the levels of production reflect deviation of the due to the installation of the new equipment and to the early proficiency in
current delivery date from an industry-wide standard level. Prices are use. Such considerations depend on how optimistic firms are about the future
development potential of the new technological trajectory. Thus, firms select
an ‘anticipation bonus’ (Xi) they award to the new technology in making their by each technique. No depreciation or scrapping is assumed. Assuming a
choice of technique. The new technology will be preferred to the old common wage rate, the rate of return on capital obtained by all firms using
according to the following rule: the α technique is denoted by rα and the return on capital for the whole
economy is denoted by r. A particular technique is the best practise
(P2 − P1 )si Xi technique if it yields the highest rate of return given the common wage rate.
≤ bi (5)
c1 − c2 The entrepreneurs will search for the most profitable technique to invest in,
but not all will be successful in that search. The underlying behavioural
where P1 and P2 are the prices of the old and new technique, respectively; c1 assumption is that entrepreneurs make different decisions as a result of
and c2 are the respective unit costs of the old and new technique and bi is the ‘retardation’ factors such as uncertainty about the profitability of the new
target payback period of firm i. Therefore, the new technique will be technique, costs and time involved in learning and using it, possible
preferred to the old if its adjusted productivity is higher and if it is cheaper protection, etc. Therefore, the adoption of the new technique may be delayed
(P2 < P1). If it is more expensive (P2 > P1) it will still be preferred provided until information about other firms’ experience is available. This behavioural
the difference in price is counterbalanced by lower unit costs (c2 < c1) within pattern, which is reflected in the entrepreneurs looking for greater
the desired payback period. profitability, who will not adopt the most profitable technique immediately
Due to the complex mathematical properties of the model, the authors run because of some retardation factors, can be formalised in the following
computer simulations in order to uncover some of its economic properties. In investment function for technique α:
doing so, they focus on the anticipation bonus reflecting expectations about
the future course of the new trajectory in order to explore the strategic Iα = σrα Kα + η (rα – r) Kα (6)
aspects of the diffusion process and the problem of the interdependence of
where σ < 1 is the saving ratio out of profits assumed to be the same for all
behaviour. If all firms anticipate incurring the development costs associated
techniques, Kα is the total capital stock embodied in technique type α and η
with bringing the new technology to commercial maturity, none of them will
is a constant. If η = 0, investments in technique α will come only from the
adopt the new technology. Conversely, if the rate of internal learning is
surplus generated by the α technique (σrα Kα ); if η ≠ 0, then investment in
accelerated, this will raise the dynamic appropriability of the innovation of
technique α will be greater or less than the surplus generated by the α
the early adopters, who show high efficiency levels. Equally the evolution of
technique depending on whether rα > r. That is, a more profitable technique
firm-specific and external skill levels allows middle and later adopters to
will attract investment which, conversely, will decline if the technique is not
start from higher initial levels of efficiency due to external learning, and to
very profitable. Given g = σr and gα = (σ + η) rα – ηr, the differential
overtake the early adopters. Therefore, technological innovation and
growth rate of each technique (gα – g) is given by
diffusion are characterised by collective effects and social tensions between
private and social gain, which impact on the aggregate economic growth. gα – g = (σ + η) (rα – r) (7)
The rate of internal and external learning affects firms’ investment decisions
concerning production techniques. In turn, this affects corporate productivity This implies that the difference in the growth rate of a particular technique
and growth, which impact on aggregate economic growth. α is proportional to the difference between the profitability of this technique
and the profitability of the whole economy.
10.4.2. Diffusion in a Multiple-technology Model The authors show that the rate of technological change, v, is the following
∂rα Ka 2 K
In Soete and Turner (1984), it is assumed that there is a large number of v=∑ + (σ + η ) ∑ (rα − r ) α (8)
firms producing a homogeneous good by using different N techniques. The a ∂t w K α K
number of different techniques may change over time as a result of new
technological innovations. Each technique uses two factors of production Two components can be distinguished: the first provides the weighted
(capital and labour). For each technique, the capital–output ratio and labour– average of the rate of technical progress for each technique α and can be
capital ratio may change over time due to an increase in efficiency. The referred to as disembodied technical change (i.e. accounting for the mere
capital stock of the economy is given by the sum of the capital stocks used occurrence of innovation); the second, which represents the ‘diffusion’ term
or technological change embodied in new investments, gives us the
covariance of the deviation of the rate of return with respect to the On the basis of the methodological distinction between formal and
distribution Kα/K. If there is a wide distribution of the rates of return of appreciative theory and under more realistic assumptions, the evolutionary
different techniques, the contribution of the diffusion terms to the overall rate perspective has been able to account for the facts explained by neoclassical
of technical change will be high. Indeed, in this case there will be a large theory as well as compile its own research agenda, thus providing some
number of entrepreneurs spread across the spectrum of rates of return. Thus, major contributions to growth theory. First of all, due to its microfounded
the number of entrepreneurs moving to new techniques will be high. This nature evolutionary theory treats the micro processes as fundamental and the
implies a rapid diffusion and a high rate of embodied technological progress. macro processes as aggregates. This implies that macro phenomena are the
Conversely, a sharp distribution of the rates of return of different techniques results of underlying micro phenomena, which play a major role in
will reduce the contribution of the diffusion term to the overall rate of explaining aggregated processes. The great emphasis placed on these micro
technical change, as the number of entrepreneurs moving to the new phenomena has allowed evolutionary theory to deal with structural change
technique will be small. Therefore, aggregate economic growth is related to issues more realistically. The adoption of an aggregated (Cobb-Douglas)
variation across technologies and their levels of diffusion. In turn, this production function – whose theoretical use is widely known to be logically
implies that technological progress embodied in new investments increases faulty (e.g. see Harcourt, 1972) – prevents neoclassical as well as new
the diffusion process of new techniques and accelerates aggregate growth. growth theory from understanding the effects of important distributive
changes that affect overall growth rates. The key difficulty lies in the fact
that the overall growth rate of aggregate output and labour productivity
10.5. EVOLUTIONARY CONTRIBUTIONS TO GROWTH depends on the cross-sectoral variations in the rates of output growth and
THEORY level of labour productivity (Cornwall and Cornwall, 1994). By treating the
economy as a homogeneous whole, new growth theory assumes that each
In what follows, evaluation of the overall contribution of evolutionary sector experiences the same changes in productivity growth. This fails to
growth theory is attempted beyond the models discussed in the previous capture the fact that rising productivity growth in only one sector impacts on
pages. average productivity growth according to the output share of the affected
According to Nelson and Winter (1982), the development of traditional sector and any resulting inter-sectoral labour reallocation. Evolutionary
growth theory along unsatisfactory lines may be attributed to the detachment analysis overcomes the limits of the orthodox theory, conducting a
between formal and appreciative theory (Ibidem). The latter tends to be close microeconomic analysis through the use of an evolving Leontief technology
to empirical work and concerns what the analyst thinks is going on. (as in Nelson and Winter’s model discussed in Section 10.3).10
Therefore, it can be usually expressed in terms of storytelling. The former Along these lines, the evolutionary perspective has contributed to more
appeals to data for support, as its main purpose is to set up an abstract realistic theorising on growth by adopting a behavioural approach in
structure enabling proposed logical connections to be explored, found and explaining the corporate decision-making process. In this perspective, firms
checked (Nelson and Winter, 1982; Nelson, 1998). If economic theorising is take their decisions on the grounds of some observable rules of thumb,
going well, then empirical facts influence appreciative theory, and which, taken as given, link environmental stimuli with corporate responses.
appreciative theory provides the basis for formal theory. However, according This allows the construction of a more realistic growth theory than the
to the above authors, in the history of economic thought what seems to have neoclassical and new growth theory, where the rules concerning the firm’s
happened is that formal theory and appreciative theory developed separately, decision-making process are deduced from maximisation. Adoption of a
with formal theory sourcing aspects of appreciative theory which could have behavioural approach also enhances the explanatory power of evolutionary
been straightforwardly formalised according to the available mathematical growth theory when considering the highly uncertain environment in which
tools and leaving the rest aside due to the lack of analytical tools to model the firm operates. When taking their decisions, firms know nothing about the
non-competitive general equilibria.9 This detachment marked a twofold possible alternatives and potential outcomes of success. Therefore, it is
development of the analysis of economic growth: an analysis of economic impossible for them to face ‘large, well defined production sets that extend
growth at macro level concerned with formalisation and an analysis of beyond the experiences range of operation’ (Nelson and Winter, 1982).
economic growth at micro level which is more concerned with empirical Conversely, firms proceed by trials and errors through search processes.
evidence.
A further contribution of evolutionary theory lies in the introduction of a attempts made to formalise the links between the micro and macro aspects of
dynamic concept of competition understood as evolving competitive growth according to an evolutionary logic are rather recent. Nonetheless,
advantage, which does not rely on equilibrium analysis as in the orthodox there have been some significant attempts to link explicitly macro and micro
theory. The latter understands competition as circumstances where no aspects (for a review see e.g. Silverberg and Verspagen, 1998). For instance,
relative competitive shifts or profits can be realised and assumes that the along these lines a more recent contribution has been provided by Iwai
system must be near or in this state. Instead, in the evolutionary perspective (2000), who has shown that as long as the state of technology retains a
competitions is defined in terms of conditions which continually change in feature of disequilibrium, the economy will keep generating positive profits
response to the strategies pursued by firms and feedback from the rest of the and, in turn, economic disequilibrium due to continuous innovation.
system. Moreover, some macro results achieved by NGT have been accounted for by
In this context, recognition of the proprietary aspects of technological evolutionary growth theory. However, viewing evolutionary growth theory
change (against the traditional conception of knowledge as a public good) as just able to explain phenomena explained by mainstream theory but with a
has been identified has a major element in understanding macroeconomic more realistic microfoundation is a ‘minimalist’ position: evolutionary
growth. The possibility of appropriating technological innovation generates growth theory should be seen as an approach extending along its own lines
differences in growth rates between firms and at aggregate level. The and focusing on its specific methodological features (Silverberg and
recognition by NGT of market imperfections, externalities from R&D Verspagen, 1998). Indeed, after Nelson and Winter’s 1982 formalisation,
investments or from education (Lucas, 1988), spillovers generating evolutionary (growth) models have developed along such lines (e.g. by
economies of scale (Romer, 1986) and of the proprietary aspect of attempting to endogenise the mutation and imitation process), rather than
technology – due to the profitability of R&D investments enabling the firm providing a benchmark against neoclassical results.
to appropriate a portion of the increase in productivity – can be seen as a
consequence of heterodox developments. Similarly, the rise of an
evolutionary alternative explanatory scheme has also promoted the 10.6. CONCLUSIONS
flourishing of neo-Schumpeterian models within the NGT realm as in the
case of Aghion and Howitt’s 1992 model, which treats technological change This paper reviews selected evolutionary models which propose an
in terms of a process of ‘creative destruction’. Therefore, the main alternative view to mainstream economic growth theory. The models
differences between these and the neoclassical models lie in more realistic discussed can be framed in a wider attempt to build up an economic theory
assumptions of the former over the latter. Nonetheless, recognition of these capable of explaining macroeconomic patterns under more realistic
phenomena as important determinants of growth cannot be regarded as a microeconomic assumptions than those made by orthodox theory. A major
novelty given that applied theory has long emphasised such aspects. element of these models is the explicit assumption of bounded rationality, as
Moreover, as noted by Aghion and Howitt (1994), new growth theory actors behave according to their routine and know nothing about what is
models still have some limitations owing to their reliance upon rational optimal. Similarly, the evolutionary character of such models lies in the fact
expectations – the assumption of perfect information is relaxed only by that firms select technologies by deciding which to introduce as well as by
treating uncertainty about the future in terms of a correctly specified deciding which one takes on board. In turn, on the basis of their fitness firms
probability distribution of possible future events – and the lack of attention to are selected in the market. Based on a theory of firms’ technological change,
institutions and transaction costs. NGT models have taken into account the evolutionary growth models seem to unfold the immediate source of growth
uncertainty that surrounds technological change, but without recognising that by providing some insights into the nature of technology, the processes
continuing technological advance implies a continuing state of driving technological change and the factors influencing corporate behaviour
disequilibrium. This is mainly due to the optimisation nature of the economic and effectiveness.
agents in the models. Therefore, although technical advance is regarded as The microfounded character of such models has allowed evolutionary
the main source of economic growth, its concept is understood as involving theory to avoid the criticism levelled at neoclassical and NGT analysis based
shifting equilibria, whose paths are foreseen by the actors involved. on macro aggregates. Moreover, the adoption of an approach based on
However, due to the evolutionary research agenda, which encompasses observable patterns of behaviour rather than a priori theoretical assumptions
formal theory as a further subsequent abstraction of appreciative theory, as well as the proprietary aspect of technology makes the evolutionary
The evolutionary perspective on growth 221
setting more realistic, thus enhancing its explanatory power. Even if these
features are specific to an evolutionary approach, some have been
incorporated into NGT given their relevance to the understanding of
economic growth. Similarly, realisation of technological change as a
‘creative destruction’ process has given rise to a neo-Schumpeterian stream
of models developed within NGT. Nonetheless, although the neo-
11. Competition, rent seeking and
Schumpeterian endogenous models have partially attempted to account for growth: Smith versus the endogenous
the limits of orthodox growth theory, they still rely heavily on traditional
neoclassical assumptions. growth theory
Antonio D’Agata*
NOTES
1. It should be, however, pointed out that other models adopting a higher level of aggregation
do exist (for a review see e.g. Silverberg and Verspagen, 1998).
11.1. INTRODUCTION
2. Following Nelson and Winter (1982) ‘the orthodoxy … represents a modern formalisation
and interpretation of the broader tradition of Western economic thought whose lines can be Following Solow’s interpretation, growth theory is devoted to explaining the
traced from Smith to Ricardo through Mill, Marshall, and Walras. Further, it is a theoretical growth of potential capacity of economies, not their effective growth (see
orthodoxy, concerned directly with the methods of economic analysis and only indirectly Solow, 1999, p. 639). The endogenous growth theory (EGT), however, has
with any specific questions of substance’.
been able to explain not only sustained growth of potential capacities of
3. It should be emphasised that in the literature (Foster, 1996) it has been remarked that
biological theory provides a general model that explains a process of endogenous change by economies, but, by making it clear that economic policies can affect the rate
the interaction of several mechanisms (e.g. endogenous change, selection retention).. Thus of growth, it has also been able to provide insights into the causes of
spurious analogies between biological and economic phenomena are rejected. different actual rates of growth among countries (see Rebelo, 1991).
4. As broadly discussed by Cantwell (1999), imitation also presumes firm-specific capabilities. As regards the explanation of the growth of potential capacity, the EGT
5. Technological advantage proceeds autonomously in terms of knowledge and inventions, but
has obtained persistent economic growth by removing the assumption of
‘innovation – i.e. application and diffusion of specific techniques in the productive sphere –
is very much determined by social conditions and economic profit decisions’ (Perez, 1985). decreasing returns on accumulated factors and by assuming that savings are
6. In a pioneering article, Dosi (1982) defines a technological paradigm as ‘a model of solution completely transformed into investment. These two features allowed Kurz
of selected technological problems based on selected principles derived from the natural and Salvadori (1998b) and, among others, D’Agata and Freni (this volume,
science and on selected material technologies’ (Ibidem, p. 152). Each paradigm shapes and ch. 2) to claim that the EGT view of the mechanics of growth can be traced
constrains the rates and direction of technological change regardless of market inducements.
back to the classical economists. As for the explanation of actual growth
7. These means through which firms could generate novelty have been further expanded in
other evolutionary ‘micro-models’ (for a review see e.g. Silverberg and Verspagen, 1998) capacity, one of the main conclusions is that the actual growth of economies
8. This leads to the evolutionary critique of growth accounting attempting to single out the depends crucially upon the rewards earned by entrepreneurial resources
relative contribution of different factors on growth given the impossibility of exactly through the free market. In fact, as entrepreneurs are not interested in profits
specifying the production set (Nelson, 1973, 1996). in themselves, rather in any kind of income (Baumol, 1990), they try to
9. In fact, while in the 1950s and 1960s neoclassical theory was concerned with bringing
escape the market mechanism – where profits are determined by competition
together bodies of formal theory into a specific growth context and with formalising
appreciative ideas in a naïf form, appreciative theory embarked on an analysis of micro- –, if by doing so they can earn higher returns.
aspects of technology whose results started questioning some of the fundamentals of The attempt by entrepreneurs to escape the competitive pressure of the
neoclassical economics (Nelson and Winter, 1982, Nelson, 1994). free market has been widely studied by the literature on public choice
10. The knife-edge problems of the Leontief production function affecting the Harrod–Domar (Tullock, 1967), which has introduced the distinction between rent-seeking
model are overcome by the Leontief technology by introducing flexibility through
and profit-seeking behaviour (Buchanan, 1980). The rent-seeking activity,
innovation (Andersen, 2001). In fact, within this context disequilibrium on the labour
market yields an active search by firm towards new methods of production which employ
which manifests itself through corruption, lobbying, etc., chiefly aims to
the most abundant factors of production more intensively. redirect policy proposal for the entrepreneur’s own advantage. In this way it
is possible to earn permanent rents by altering the workings of the market
222
Competition, rent seeking and growth 223 224 The Theory of Economic Growth: a ‘Classical’ Perspective
process towards a non-competitive structure (Colander, 1984, p. 3), usually 11.2. COMPETITION AND RENT-SEEKING
through the erection of legal barriers to entry (licences, quotas, tariffs, etc.).
A feature peculiar to rent-seeking is that it has purely redistributive effects The literature on rent-seeking starts with a paper by Gordon Tullock (1967),
and the relevant literature has shown that, within a static context, this activity who questioned the calculation of the welfare losses of the static monopoly.
usually yields inefficiencies (Rowley, Tollison and Tullock, 1988). He argued that the welfare losses are greater than the Harberger triangle
The EGT has extended the study of rent-seeking to a dynamic context and since they should also include the monopolist’s expenditure for gaining and
has developed quite a few models which study the effects of rent-seeking on maintaining the monopolistic position. Since then, the term rent-seeking has
the growth process (Murphy, Shleifer and Vishny, 1991, 1993; Pecorino, been variously defined (Brooks and Heijdra, 1989) with one of the most
1992; Rama, 1993). This literature has reached the general but not commonly used meanings being the following: rent seeking is ‘the pursuit of
unanimous, conclusion that, whatever the way in which rent-seeking is profits via the use of government coercion’ (Anderson, Rowly and Tollison,
carried out, rent-seeking behaviour tends to depress the rate of growth of the 1988, p. 100). By ‘government coercion’ is meant any activity by public
economy; by contrast, profit seeking usually boosts growth. authorities which alters the income distribution determined by the
It is our opinion that the idea that growth depends negatively on the competitive market mechanism through redistributive actions (for example,
existence of rent seeking can be traced back to classical economists, and in taxing and government subsidies; Sturzenegger and Tommasi, 1994) or by
particular to Smith, although the latter maintains a more specific view creating legal barriers to entry like import licences or tariffs (Krueger, 1974;
concerning the nature of rent seeking compared with that of the EGT. We Bhagwati and Srinivasan, 1980; McCormick, Shughart and Tollison, 1984).
shall argue that Smith considers rent seeking as oriented only towards the In any case government intervention upon economic activity gives rise to
creation of legal barriers to entry and that, according to this author, this rents benefiting particular groups. Therefore, if rent seeking yields higher
activity, in this specific form, although not wasteful in se, negatively affects returns than the usual productive activities, people compete through it by
economic growth. Smith and the proponents of the EGT also hold different diverting resources from the usual productive activities. Thus, rent seeking is
views as regards the mechanism through which rent-seeking affects growth, wasteful and creates inefficiencies.
Smith’s being much more elaborate that that of the EGT. Hence we can Rent seeking may therefore be seen as just an alternative form of profit
conclude that, from the explanatory point of view, the classical approach to seeking: while profit seeking is carried out by agents seeking to maximise
growth – both in its original formulation due to Smith, Ricardo and Marx and returns on their resources within the market system, rent seeking is carried
in its modern rehabilitation due to the EGT – is a more satisfactory theory out by agents who want to maximise the returns on their resources by
than the neoclassical one as it is able not only to explain the (sustained) exploiting the political power of coercion. The common opinion (Buchanan,
growth of potential capacity but also to provide insights into the growth of 1980) is that profit seeking operates through innovation and the shift of
actual capacity of economies. resources towards employment with the highest returns. This yields, albeit
The paper is organised as follows: the following section introduces the unintentionally, positive effects on the economy because it ensures an
concept of rent-seeking behaviour. Section 11.3 provides a description of a efficient allocation of resources and generation of technical progress. Since
simple model of endogenous growth in which entrepreneurs can pursue profit-seeking activities are oriented to generating competitive advantage,
profit seeking or rent-seeking. Section 11.4 is devoted to surveying Smith’s they may create barriers to entry as well; however, since the market process
view of monopolistic markets and we shall point out that legal barriers to does not ensure any protection, any barrier to entry usually turns out to be
entry in Smith’s analysis arise from ‘rent-seeking’ activity, and that Smith’s temporary and fades away through the competitive process of imitation. The
view of the role of this activity on growth is very similar to the EGT view. political system, by contrast, is able to alter the distribution of the market
Final conclusions are contained in Section 11.5. mechanism by redistributive action or by creation of permanent legal barriers
to entry that favour certain agents by allowing them to obtain, even in the
long run, higher returns for their resources than their market levels. Rent
seeking does not necessarily have beneficial effects on the economy, since it
ensures neither technical progress nor efficient allocation of resources.
An alternative definition of rent seeking requires the employment of
resources: rent seeking is ‘any redistributive activity that takes up resources’
(Murphy, Shleifer and Vishny, 1993, p. 409). Private and public rent seeking example, licenses or tariff barriers (Pecorino, 1992; Rama, 1993) in favour of
is distinguished: ‘Private rent-seeking takes the form of theft, piracy, sector 2. Therefore, rent seeking in sector 2 allows agent a to earn a rent
litigation, and other forms of transfer between private parties. Public rent equal to R(a) = r(a)K, where
seeking is either the redistribution from the private sector to the state, such as
¤a
taxation, or alternatively from the private sector to the government r (a) = R ( ) ,
bureaucrats who affect the fortunes of the private sector’ (Murphy, Shleifer ∫c ¤ ad ¬
and Vishny, 1993, p. 412). Private rent seeking has been widely used in
recent years to explain the existence of the state, private property and and where μ is the Lebesgue measure on I, Ω is the set of agents acting as
markets (Grossman and Kim, 1995). We do not consider private rent seeking rent-seekers, δ is a positive number and R(Ω ) indicates the total rent
here since it is not helpful in explaining the role of rent seeking, as an extracted in the economy. Rent R(Ω ) is assumed to be fraction t of the high
alternative activity to market-based competition, in the process of growth. value-added production of good 1, 0 ≤ t ≤ 1. Thus, by employing capital K in
Hence, below we shall focus only upon public rent seeking. sector 2, agent a yields a total income equal to
⎡ ¤ ⎤
11.3. RENT SEEKING AND GROWTH Y2 (a) + R(a) = ⎢ ρ + R ( )⎥ a ( K − L ) .
⎢
⎣ ∫c ¤ ad ¬ ⎥
⎦
Rent seeking, as analysed in the static framework and briefly surveyed in the
preceding section, has been extended to a dynamic framework by the EGT. Finally, we assume that each agent saves the fraction s (0 ≤ s ≤ 1) of his/her
Particular attention has been focused on the effects on growth of the different income and that capital does not depreciate.
ways in which rent seeking is carried out, for example by means of Using the assumptions adopted, agents with a higher index are also the
corruption (as in Barreto, 2000), or by means of lobbying (as in Mohtadi and most efficient in both production of goods 1 and 2. Hence set Ω is always a
Roe, 1998). In this section we shall provide a simple growth model with rent left interval of point 2. Suppose, therefore, that agents in interval [A, 2]
seeking. The model is deliberately as simple as possible in order to highlight produce good 2, hence total production in sector 1 is
the features of rent seeking and its effects on growth. Our analysis is an
extension of the two-sector model of Murphy, Shleifer and Vishny (1991, A A ( A2 -1) β
1993) to an explicitly growth context à la Harrod–Domar, as their model is
YT 1 ( A) = ∫
1
Y1 (a)da = ∫ β a ⋅ Kda =
1 2
K,
essentially static.
Let us assume a two-sector economy with a continuum of agents indicated total production in sector 2 is:
by the interval I = [1, 2].1 The two goods are indicated by 1 and 2 and either
2 2 (4- A2 ) β
good can be used as consumption and as a capital good. Good 1 is the YT 2 ( A) = ∫ Y2 (a)da = ∫ β a(K - L )da = (K - L ) ,
numeraire and, for the sake of simplicity, we assume that prices are fixed and
A A 2
that the price of good 2 is also equal to 1. Agent a ∈ I is endowed with a while total rent is
positive amount of capital K(a) = K , with the technology Y1(a) = βaK to
produce good 1 and with the technology Y2(a) = ρaK to produce good 2, ( A2 -1) β
where β > 0, ρ ≥ 0. We also assume that β > ρ, which means that production RT ([ A, 2]) = tYT 1 ( A) = t K.
2
of good 2 has a value-added lower than production of good 1. We also
assume that production of good 2 is carried out together with a rent seeking By using a* to indicate the agent for whom the production of good 1
activity which requires a fixed amount L (0 < L K ) of capital. The way in yields the same returns as the production of good 2, then a* must satisfy the
which rent seeking affects the distribution of income is usually described in following equality:
reduced form and is interpreted as due to a direct redistributive policy
through, for example, taxing or subsiding (Murphy, Shleifer and Vishny,
1991,1993) or due to the creation of legal barriers to entry through, for
⎡ ⎤ 11.4. SMITH’S VIEW ON RENT SEEKING,

¤
(1 − t ) β a * K = ⎢ ρ + R ([a*,2])⎥ a * ( K − L ) ; COMPETITION AND GROWTH
⎢ ⎥
⎣⎢ ∫[a*,2] ¤ ad ¬ ⎦⎥
In this section we show that according to Smith, rent seeking activities,
competition and growth are closely related, as free competition maximizes
⎡ (
t β K a *2 −1 ) ⎤⎥ a * K − L
i.e. ⎢
(1 − t ) β a * K = ρ +
⎢ 4 − a *2 ⎥
( ). the amount of resources devoted to the growth of wealth, while imperfect
competition negatively affects the growth of wealth by reducing the amount
⎣ ⎦
of resources employed in productive investment, and the average rate of
saving. While Smith has a narrower conception of rent seeking than the
If (1 − t ) β K > ρ (K − L ) the above equation has only one solution in the
EGT, as he deals only with rent seeking oriented towards the creation of
interval [1,2]; therefore, every agent with an index higher than a* will
legal barriers to entry, he has a more complex view of the way in which rent
produce good 2, while every agent with an index lower than a* will produce
seeking affects growth. However, Smith and the EGT share the view that
good 1.
rent seeking negatively affects growth.
An easy exercise of comparative statics shows that a* decreases as t or ρ
According to Smith, the growth of a nation depends upon technology
increases and L or β decreases. Therefore, the more effective the barriers to
(‘division of labour’, WN I.i) and upon the stock of capital available which,
entry (measured by t and by L), the higher is the number of agents who turn
in turn, depends upon the way in which labour is allocated between
to the production of good 2. In any case, the more talented people are those
‘productive’ and ‘unproductive’ employments (WN II.iii). Here we need just
who turn first to rent-seeking whenever it yields higher returns than
to recall that productive labour is any kind of labour which is employed to
producing good 1. Therefore, rent seeking diverts entrepreneurial resources
produce material goods, while unproductive labour is employed to produce
from the production of good 1 to the production of good 2. This has several
services which, although useful, are not considered to contribute to the
implications at the macroeconomic level. In fact, the total production is:
wealth of the economy (for details see Bowley, 1975, pp. 369–72, or Eltis,
1984, p. 78) and are substantially consumption activities (WN II.iii.6–7). In
( β − ρ )a *2 −( β − 4 ρ ) ⎡ 4 − a *2 ⎤
YT (a*) = YT 1 (a*) + YT 2 (a*) = K − ρL ⎢ ⎥. fact, while economic activities in which productive labour is employed yield
2 ⎣ 2 ⎦ profits (WN II.iii.6), those which employ unproductive labour do not yield
profits and are supported just by income that has been produced elsewhere
Hence total production and savings, sYT (a*), decrease as t increases and L (WN II.iii.7). Unsatisfactory as it may be, the distinction between productive
decreases as we assume that β > ρ. Moreover, the total stock of capital and unproductive labour is useful in that it highlights the Smithian idea that
employed in the production activity, (a*–1) K + (2 –a*)( K – L), decreases growth is determined by the way in which resources are employed between
as a* decreases. It follows, that the rate of capital growth, activities which increase income and activities which merely transfer it.
In chapter III of book II of The Wealth of Nations, after presenting the
( β − ρ )a *2 −( β − 4 ρ ) L ⎡ 4- a *2 ⎤
sYT (a*) K = s −ρ ⎢ ⎥ distinction between productive and unproductive labour as factors
2 K⎣ 2 ⎦ determining the accumulation of capital, Smith points out that the allocation
of labour in productive and unproductive employments depends upon
decreases as a* decreases. ‘parsimony’ and ‘frugality’ of individuals (WN II.iii.14); in fact:
This model identifies a major general effect of rent-seeking on growth: it
can make the high value-added activity unattractive by increasing the Parsimony, by increasing the fund which is destined to the maintenance of
opportunity cost in carrying out that activity. Thus, it diverts capital from productive hands, tends to increase the number of those hands whose labour adds
highly productive activities (production of good 1) towards lower productive to the value of the subject upon which it is bestowed. It tends therefore to increase
activities (production of good 2) or totally unproductive activities the exchangeable value of the annual produce of the land and labour of the country
(production of good 2 if ρ = 0). Moreover, the way in which the economy is (WN II.iii.17).
modelled indicates that the human resources which turn to rent-seeking are
usually the best resources available in the economy. By contrast,
The prodigal… [by] diminishing the funds destined for the employment of entry will be called ‘informational’ and is due to the imperfect information of
productive labour, …, necessarily diminishes, …, the quantity of that labour which the condition of demand, which allows producers to sell their product at a
adds a value to the subject upon which it is bestowed, and, consequently, the value price higher than its natural level. The third kind of barrier to entry will be
of the annual produce of the land and labour of the whole country (WN II.iii.20). called ‘technological’ and is due to the existence of technical improvements
which are kept secret and which allow the inventor to sell the goods at a
Prodigality in turn can be private and public. Actually, according to higher market price, calculated with respect to the old technology, than the
Smith, all public revenue is often employed in maintaining unproductive new natural price. The fourth kind of barrier to entry, which will be called
labourers (WN II.iii), although Smith recognises that this kind of ‘legal’, is due to the intervention of political authorities that, by law, are able
employment may be useful (WN V.i). Moreover, mistakes in business can be to impede free movement of resources (for example, by means of monopoly,
assimilated to prodigality as both destroy resources which can be employed statute of apprenticeship, etc.).
in productive uses (WN II.iii.26), although, in general, successful business Intersectoral competition has a crucial role in the classical theory of value
compensates its unsuccessful counterpart (WN II.iii.27). and distribution as it ensures that, in competitive sectors, i.e. in those sectors
It is our opinion that, according to Smith, the competitiveness of markets, in which there are no barriers to entry, market prices gravitate around their
in particular the existence of legal barriers to entry, has a crucial role in natural values (WN I.vii.15). Particularly important for our concern,
determining the proportion between productive and unproductive labour, and however, is the role that free competition has on the allocation of resources:
therefore the growth of the economy. In order to support this view we now in fact, a main point of Smith’s view of the capitalistic process is that
turn to the concept of competition developed in The Wealth of Nations. whenever the economic activity is left to the interest of individuals, the
The society that Smith has in mind in constructing his theory of value and allocation of free mobile resources among different activity, governed by the
distribution is a society divided into classes of resource owners: workers, ‘invisible hand’, turns out to be in accord with the ‘public interest’ (WN
capitalists and landowners, and the self-interest of each individual belonging IV.ii.9). Smith never defines explicitly what the ‘public interest’ is. However
to any of these classes is the origin of the competitive process. He conceives it is reasonable to interpret it as the maximization of the wealth of nations. In
competition as a process carried on within the market by agents who desire to fact, in the same chapter that contains the preceding statement Smith makes it
attain a particular goal. Smith uses this term mainly with two meanings: first, clear that people, in seeking their own advantage and in competitive
to describe the race amongst producers or consumers to sell or buy, conditions, employ their resources ‘as near home as possible’ and in such a
respectively, the quantity of goods they have produced or they desire to buy; way that ‘its produce may be of the greatest possible value’ (Ibidem). By
second, to describe the process of entry of resources into those markets in doing so, ‘every individual necessarily labours to render the annual revenue
which they can secure higher remuneration. The former kind of competition, of the society as great as he can’ (Ibidem).
called price competition, ensures the attainment of the market price (WN However, according to Smith not all barriers to entry are against the
I.vii.7–10). The latter kind of process, called intersectoral competition, is ‘public interest’. Indeed, while Smith deeply criticises legal barriers to entry,
based upon the idea that individuals are greedy and seek to earn the highest he seems to maintain a positive view of technological barriers to entry.2 As
remuneration for the resources they own through the market process. for the latter, Smith gives an assessment of technological barriers to entry
The working of intersectoral competition can be hampered by possible incidentally in Book V in discussing the only case in which he agrees to give
impediments to the free movement of resources, i.e. by ‘barriers to entry’. In a temporary legal monopoly to companies of merchants to establish new
the presence of barriers to entry, market prices no longer gravitate around trade with high risk countries. In this case, Smith justifies the granting of a
their natural values and the income of the resource owners who employ their monopoly ‘upon the same principles upon which a like monopoly of a new
resources in these sectors can be higher than the income of the same resource machine is granted to its inventor, and that of a new book to its author’ (WN
employed in competitive sectors even for long periods. Smith conceives of V.i.e.30). The justification for granting a temporary monopoly is that in all
four causes of barriers to entry: (i) the scarcity of a particular resource, (ii) these cases the monopoly recompenses traders and researchers for the risks
the secrets in trade (iii) the secrets in manufacturing and (iv) the privileges undertaken. However, Smith is very careful to emphasise that the monopoly
granted by law to an individual or to a trading company (WN I.vii.21–26). must be temporary in order to ensure such beneficial effects on the public
The first kind of barrier to entry will be called ‘natural’ and is due to the non- (Ibidem).
reproducibility of scarce natural resources. The second kind of barrier to
By contrast, as anticipated above, Smith expresses a radically different (Murphy, Shleifer and Vishny’s), it seems that Smith never explicitly
opinion with respect to permanent legal barriers to entry. In analysing trade considers the creation of legal barriers to entry as the outcome of a specific
with the colonies, Smith maintains that permanent monopolies affect economic activity which employs resources. Rather, this activity is always
negatively the growth of the economy as they reduce the amount of capital conceived within a productive activity; hence, any activity aimed at the
invested in productive labour and employ it in unproductive labour (WN creation of legal barriers to entry is not wasteful in se. However, like the
IV.vii.c.49). Smith holds that legal barriers to entry, like technological EGT, Smith believes that legal barriers to entry do have negative effects on
barriers to entry, are created under the pressure of private interest. This is income and growth, although, as we shall see immediately, he has a more
made clear in chapter X, where he points out that the resulting monopolies complex view than the EGT as far as this relationship is concerned.
and corporations are established not only because of the sovereign’s interest In chapter VII of book IV of The Wealth of Nations Smith provides a
in raising money (WN I.x.c.17), but also under the pressure of mercantile detailed account of the effects of legal barriers to entry – in particular the
interests: monopoly granted on colonial trade – on the growth of an economy. Apart
from specific negative effects that monopoly on trade produces due to the
The government of towns corporate was altogether in the hands of traders and peculiar nature of the barrier to entry, he pinpoints two general negative
artificers, and it was the manifest interest of every particular class of them, to effects, one direct and the other indirect, that monopolies have on the stock
prevent the market from being over-stocked (WN I.x.c.18; see also IV.ii.21). of capital employing productive labour. The direct negative effect is due to
the fact that monopolies, by impeding entry of new capital in markets,
Legal barriers to entry, however, are not required only for protecting the contract the stock of capital employing productive labour; the indirect
incumbent firms from domestic potential competition: negative effect is due to the fact that monopolies reduce prodigality. The first
effect supports Smith’s general view that barriers to entry divert resources
A great empire has been established for the sole purpose of raising up a nation of from their ‘natural’ allocation, which is the competitive one and which
customers who should be obliged to buy from the shops of our different producers,
ensures a higher level of productive labour with respect to the monopolistic
… for the sake of that little enhancement of price which this monopoly might
afford our producers, the home-consumers have been burdened with the whole
case (WN IV.ii). Thus, the presence of a monopoly reduces the amount of
expence of maintaining and defending that empire (WN IV.viii.53–4). capital in the country that employs productive labour. This implies that the
income generated in the country decreases, with a consequent decrease in the
Again, this is the outcome of the lobbying of the capitalists on political savings and in the rate of capital accumulation (WN IV.vii.c.57). Moreover,
institutions: Smith is very careful to stress that this contraction affects equally all kinds of
income, with the exception of the profits in monopolistic sectors. Indeed, a
It cannot be very difficult to determine who has been the contrivers of this whole reduction in the productive labour employed implies a reduction in wages
merchantile system, not the consumers,…, but the producers, whose interest has (Ibidem). Furthermore, the increase in the rate of profits in monopolistic
been so carefully attended to, and among this latter class have been by far the sectors, by depressing the remuneration of capital in other sectors relative to
principal architects (Ibidem. For other examples, see also WN IV.i.10, IV.iv.1 and the monopolised sectors, discourages improvement in land and this, in turn,
IV.v.1). ‘necessarily retards the natural increase of another great original source of
revenue, the rent of land’ (WN IV.vii.c.58. Finally, although monopoly
From the above textual evidence it seems reasonable to conclude that yields an increase of profits in the monopolised sector, Smith believes that
Smith conceives the creation of legal barriers to entry as the outcome of a total profits are lower than in the competitive situation because of the entry
rent seeking activity deliberately carried out by landlords and entrepreneurs constraints of capital in monopolised markets (WN IV.vii.c.59).
in their attempt to maximize rents or profits.3, 4 However, some important As already stated, the indirect effect of monopolies on growth is due to
differences must be emphasised between Smith’s view and the modern view the fact that they negatively affect parsimony (WN IV.vii.c.61). However,
of rent-seeking. First, Smith conceives rent seeking as only directed towards the reduction of savings is not caused only by the misbehaviour of capitalists,
the creation of barriers to entry. Moreover, consistent with part of the current but by a generalised reduction in the saving ratio as the monopolist’s
literature on rent-seeking (see Anderson, Rowly and Tollison’s definition of behaviour will be imitated by all the other classes (Ibidem). Finally, Smith
rent-seeking in Section 11.2), and in contrast with the alternative definition explicitly recognises rent-seeking activities directed towards the creation of
legal barriers to entry. Moreover, he also holds the idea that rent-seeking concerning the prodigality of rent seekers. It can be noted that in Smith’s
activity is detrimental to the growth of an economy, directly through the model the rate of growth of capital is:
reduction in the stock of capital in the economy and the consequent reduction
of productive labour employed and of the saving rate, and indirectly through sYT (a*) (s1 − s2 )a *2 −(s1 − 4s2 )
= β,
destruction of parsimony. Moreover, he explicitly recognises that the K 2
increase in profits in monopolistic sectors may reduce the incentive of
investing in activities that improve technology. thus, because of assumption (iii), it decreases as the production of good 2
The direct effect of legal barriers to entry on the stock of capital can be yields higher returns (a* decreases).8
illustrated graphically as follows.5 Consider an economy with two sectors,
say agriculture and industry. In Figure 11.1 the segment O–O’ measures the
total stock of capital available in the economy. Curve Πi indicates the rate of Πa
Πi
profit in industry as a function of the stock of capital employed in that sector,
while curve Πa, which should be read from origin O’, indicates the rate of
profit in agriculture as a function of the stock of capital employed in that C'
MW:OK C
sector. These curves are decreasing because Smith believes that the market ri'
price, and therefore, the rate of profit decrease as the production level
ra'
increases. Intersectoral competition ensures that fraction OA of capital
should be employed in industry and that fraction O ' A should be employed
in agriculture so that the common rate of profit r* = ra = ri is generated.
Following Smith, r* is also the rate of interest (WN I.ix.4) so no production
can yield in the long run a rate of profit lower than r*. If a legal barrier to ri ra
entry is introduced in industry which ensures that the producers earn ri’, then
a contraction of amount AB of stock in industry must occur. Eventually, the
total stock of capital in the economy will be equal to OB + O ' A , where OB 0 B A B' 0'
is employed in industry and O ' A in agriculture (the competitive sector).
As seen above, Smith also suggests that if agriculture yields lower returns Figure 11.1
than those earned in the monopolistic sector, then capital in that sector may
be withdrawn, thus causing reduction of rents. If this occurs, capital will also
be reduced in agriculture, and possibly to such a level that ensures a new,
higher, common rate of profit r’ = ri’ = ra’. If we interpret curve Πa as 11.5. CONCLUSIONS
illustrating the rate of profit on the stock of capital employed also for
improvements in agriculture,6 the amount of capital that must be withdrawn We have shown that the EGT and Smith share the view that legal barriers to
in agriculture in order to ensure the common rate of profit r’ is equal to AB ' . entry created through rent seeking negatively affect growth because they
Now the total amount of capital employed, therefore, is OB + O ' B ' .7 tend to increase the opportunity cost of pursuing productive activities, and
Smith’s view of the effects of rent seeking on growth via prodigality can therefore tend to employ resources in ‘unproductive’ or ‘less’ productive
be easily formalised by means of the model introduced in the preceding activities. We have also pointed out that there are differences between the
section. The Smithian growth model with legal barriers to entry due to rent EGT and Smith’s views concerning the mechanism through which rent
seeking makes the following assumptions: (i) L = 0; (ii) β s ρ ; and (iii) the seeking affects growth: unlike the EGT, according to Smith rent-seeking is
rate of saving of producers in sector 1, s1, is greater than the rate of saving of directed towards the creation of legal monopolies, and moreover it is not in
producers in sector 2, s2. Assumptions (i) and (ii) should capture the idea that se a wasteful activity as it is usually carried out within a productive activity.
for Smith rent seeking is not necessarily wasteful and it can be associated Finally, it is through the creation of such barriers to entry that, according to
with productive activities. Assumption (iii) should reflect Smith’s idea Smith, rent seeking has negative effects on growth. He identifies two
Competition, rent seeking and growth 235
negative effects on growth; the first, direct, reduces the amount of the overall
(productive) capital employed in the economy because barriers block entry
and because, as stated above, it increases the opportunity cost in pursuing
productive activities; moreover, it reduces the aggregate magnitude of each
kind of income. The second, indirect effect entails the reduction in
parsimony.
12. R&D models of economic growth
and the long-term evolution of
NOTES productivity and innovation
* I would like to thank the participants at the workshop organised by the MURST group Mauro Caminati
‘Classical Themes in Modern Growth Theory’ held in Rome 1–4 March 2001 for their
useful comments. I am indebted to Mario Lavezzi and Neri Salvadori for insightful
suggestions on a draft of this paper. The financial support of the MURST (Cofin 1999) and
the University of Catania (Fondi di Ateneo) is also gratefully acknowledged. 12.1. INTRODUCTION
1. Murphy, Shleifer and Vishny’s 1991 model is a special case of this model by setting ρ = 0
and L = K , i.e. they conceive rent-seeking as a purely redistributive activity.
The ratio between the number of scientists and engineers engaged in research
2. Smith seems to have expressed no specific opinion as far as natural and informational
barriers to entry are concerned. and development (R&D) and the level of total employment increased
3. For an interpretation along this line and for a more detailed analysis of the relationship dramatically in the U.S.A. in the second half of the twentieth century. Let us
between private interest and public intervention see Stigler (1975). call hL the ratio between employment outside of R&D and total employment.
4. Interestingly, Smith also considers that political authorities are petitioned not only to create In the U.S.A. (1 − hL) was nearly three times as large in 1993 as in 1950, with
legal barriers to entry, but also to increase rivals’ costs (WN I.ix.5).
a pronounced upward fluctuation in the period 1960–70 due to government-
5. The approach we adopt is essentially a partial equilibrium one. Although it is not consistent
with Smith’s overall analysis, which is of a general equilibrium nature, it seems to be what funded R&D. Jones (2002) estimates that from 1950 to 1993 there was an
he has in mind when he concludes that ‘[t]he increase of stock, …, tends to lower profits’ even larger rise in the researchers/employment ratio in the set of G-5
(WN I.ix.2: see also II.iv.8). countries (France, West Germany, Japan, the United Kingdom and the
6. In this case the reduction in the rate of profit as the capital stock increases is due not only to United States).
the consequent reduction in the price of produce due to increased production, but also to the
It is quite striking that the observed dramatic rise of R&D employment
reduction of the returns of improvements.
7. Smith also holds that the total amount of profits decreases, even if the rate of profit did not show up in the productivity figures. As is well known, the growth
increases. This means that according to Smith in Figure 11.1 area OriraO' is larger than area rate of GDP per hour tended to decline in the advanced countries after the
Ori'CB + B'C''r'aO'. 1950–70 ‘golden age’. The decline was less pronounced in the U.S.A.
8. A formal analysis of the effects of rent seeking on growth through reduction of the total because this country did not enjoy the boom in productivity from
stock of capital in productive activities seems to be a difficult task since it should be carried
technological catching up after the Second World War. Hence the U.S.
out within a general equilibrium analysis rather than a partial equilibrium one like that
employed by Smith (see footnote 5). experience provides a more telling indication of the relation between R&D
effort and productivity growth for a country located on the frontier of
technological knowledge.1
I will refer the mentioned rise in the researchers/employment ratio as
stylised fact (a) and the relatively constant growth rate of GDP per hour in
the second half of the twentieth century as stylised fact (b).2
The question discussed in this paper is how the R&D models developed
within the recent revival of general-equilibrium-growth theory cope with the
facts (a) and (b).3 A similar question was addressed in an influential paper
written by C.I. Jones and published in 1995. Jones observed how the R&D
growth models developed to that date displayed a ‘scale effect’ of the
236
R&D models of economic growth 237 238 The Theory of Economic Growth: a ‘Classical’ Perspective
number of researchers on the growth rate of GDP per capita. These models can be traced back to Karl Marx, I stress the relation between the arrival of
are criticised by Jones because the ‘scale effect’ is in striking contrast with new technologies and the growth of variety and show how this may help to
the evidence. In the same paper he builds a model, which he defines semi- explain the stylised facts (a) and (b).
endogenous, where innovations are still the outcome of purposeful and costly The focus of this paper is on steady-state results. When transitions
R&D effort, but the steady-state growth rate of output per capita is between different steady states are involved, e.g. in Section 12.5.1, the
completely determined by the technological parameters and the rate of implicit assumption is that transitional dynamics are monotonic. Eicher and
population growth. It is therefore independent of the level of population, of Turnovsky (1999b and 2001) show that non-monotonic transitional paths
preferences, and of policy variables that do not affect technology. The family may exist in the non-endogenous growth models with two endogenously
of R&D growth models with these properties is called here non-endogenous. accumulating factors, knowledge A and capital K. In what follows the
By contrast, the endogenous R&D models of general-equilibrium growth are endogenously accumulating factors are capital K, intensive technical
those where per-capita GDP growth depends upon preferences and/or policy knowledge A and extensive technical knowledge N. A general analysis of the
variables generally. The basic structure of the endogenous and non- transition dynamics for the R&D growth models of this type is still lacking.5
endogenous general-equilibrium models of economic growth is discussed in The discussion of how it may be relevant to the theme of this paper is left to
Sections 12.2, 12.3 and 12.4. future work.
Partly as a reaction to Jones’ critique, a second generation of endogenous An important caveat must be added. In what follows, the rigid supply
R&D growth models appeared in the late 1990s. In this second generation, orientation of the general-equilibrium models of economic growth is taken
besides ‘intensive’ innovations that increase the productivity of the for granted and is not questioned. This is not because the author is not aware
intermediate good produced in their sector of application, there are of the biases that are introduced when co-ordination problems or stability in
‘extensive’ innovations, that increase the number of intermediate goods. In the disequilibrium dynamics are disregarded. These issues are simply outside
steady-state equilibrium, the number of intermediate goods (hence of sectors) the scope of this paper.6 In a similar vein, the paper is unconcerned with the
grows at the population growth rate n, so that, in steady state, the number of criticism that may be levelled at the use of capital aggregates in theoretical
intensive-researchers per sector is constant. This implies that the ‘scale models. The attitude is simply to take the model predictions for what they are
effect’ on the rate of growth disappears. In other words, there is a dilution of and discuss their consistency with broadly defined stylised facts.
the ‘scale effect’ across the growing number of intermediate-good sectors.
The steady-state predictions of the second-generation endogenous and
also of the non-endogenous R&D growth models are still at odds with the 12.2. A UNIFYING REPRESENTATION OF
evidence presented at the beginning of this introduction. The dramatic long- TECHNOLOGY
term rise of the R&D employment share (1 − hL) reveals that the long-term
growth path of the U.S. economy cannot find a theoretical approximation In what follows I build a framework which embeds different views of the
through the hypothesis that the economy has been growing in the relation between output growth and the generation of new inputs, as may be
neighbourhood of a single steady-state path.4 The observed long-term rise of encountered in R&D growth models. This is done under a number of
(1 − hL) and the approximately constant rate of productivity growth may be simplifying assumptions about technology that still enable us to discuss
more consistent with the hypothesis of a transitions path induced by usually neglected issues, such as the role of complementarities and the
exogenous parameter changes. This issue is addressed in Section 12.5.1. My relation between technological compatibility and knowledge spillovers. The
conclusion here is that the changes in the technological parameters required main simplifying assumption is that the service characteristics of final output
to reconcile the stylised facts (a) and (b) above may be implausibly large. Y are unchanged throughout, that Y can be either consumed or accumulated
In Section 12.5.2 I suggest that the failure of the R&D growth models to in the form of capital and that it is produced by means of intermediate goods
reconcile the constant productivity growth with the long-term rise of the and labour. The number of available intermediate goods Nt changes through
R&D employment share can be interpreted as the result of technological time as a result of innovation activities.
assumptions that make abstraction from complementarity in production. Assume that the number of service-characteristic types that exist in nature
Following a system-like view of technology which owes much to the is finite. An intermediate good is a pair (v, Av) ∈ R+ , where v is the
2
contributions of Nathan Rosenberg, Joel Mokyr and many others and which intermediate–good variety (which identifies a class of functions performed
by the good, that is, a composition of the associated flow of service unit of a new good is acquired, it can be applied to the production of an
characteristics) and Av is the technological level, or generation, to which (v, indefinite number of units. If intermediate-goods production is otherwise
Av) belongs. In principle, we should expect Av to have only an ordinal subject to constant variable costs, we are faced with a clear case of increasing
meaning, possibly with the further ordinal implication that later generations returns.
of a variety are also more productive. This is not, however, the interpretation The input of the activity for producing one unit of (v, Av) is a quantity of
we find in the new-growth literature, where Av is an index leading to a capital K which depends positively on the technology level Av. K units of
cardinal productivity measure. The marginal product of (v, Av) is a known capital invested in the production of good (v, Av ) give rise to K Avω units of
time-invariant function of Av (and possibly other variables). This leads to a this good, where ω > 0, thus implying that more capital intensive methods
time-invariant production possibility frontier, describing the productive are required to produce intermediate goods of a later generation. Hence
potential of every possible present and future combination of intermediate K v = xv Avω . Howitt (1999) adopts a similar increasing-capital-intensity
goods. assumption and claims that capital used in intermediate-goods production
can be interpreted as human capital. The above specification implies that the
12.2.1. Production of Material Goods average and marginal cost, in terms of final output, of producing (v, Av) is
rAvω , where r is the rental price of capital. Since we abstract from
Final output Y is produced by means of intermediate goods and labour by depreciation, r is also the rate of interest.
perfectly competitive firms which, individually, face constant returns to The monopoly profit from producing xv, t is:
scale. Following the R&D growth literature, we introduce a set of
simplifying assumptions implying that at every date t only the highest (and π v,t = α (1 − α ) N tγ L(1Y ,−t α ) Av,t xαv,t
latest) available technology level Av t of each variety v is used. This will be the
case to the extent that the value of the productivity gain from using the latest Aghion and Howitt (1998, ch. 12) and Howitt (1999) obtain a monopoly
generation of a given variety invariably dominates the cost differential output which is uniform across varieties and independent of A,8 by setting
associated with the same choice. 7 In fact, these models assume a particular ω = 1. We hold to the latter simplifying assumption to obtain:
substitutability relation between intermediate goods, to the effect that they 2 γ 1
enter the production function in an additively separable form. Recalling our xv,t = α 1−α N t1−α LY ,t rtα −1 = xt (2)
simplifying assumptions, the individual production function is:
In equilibrium, final output Y is then:
⎡ Nt ⎤
⎢ ∫ Av,t x v ,t ∂v ⎥
γ 1-α α
Yt = N L
t Y ,t (1) 2α 1−α + γ 1
⎣⎢ v=0 ⎦⎥ Yt = N tγ L1Y−,αt Nt At xαt = α 1−α Nt 1−α LY ,t rtα −1 At (3)
where xv is a quantity of the intermediate-good variety v and LY is labour
employment in the production of final output. It follows from (1) that the where A t is the average technology level across intermediate goods:
marginal product at t of the intermediate good (v, Av, t) is α N tγ LY1-,αt Av,t xαv ,t−1 .
1 ⎡ Nt ⎤
It is independent of the inputs of the other intermediate goods, although it At = ⎢ ∫ Av,t ∂v ⎥
may depend, if γ ≠ 0, on the total number of intermediate goods cooperating N t ⎣⎢ v = 0 ⎦⎥
with it. The above form of independence is interpreted here as resulting from
the lack of technological complementarity between any two intermediate An equivalent equilibrium expression of Yt is obtained by observing that,
goods. if hK is the capital share employed in material, as opposed to knowledge,
Intermediate goods are produced by local monopolists through a different production, then in equilibrium we must have (hK, t Kt) / At = Nt xt. Hence:
set of activities. The reason why firms in the intermediate-good sector cannot
Yt = Ntγ ( hL ,t Lt )
1−α
be perfectly competitive is quite robust (Arrow, 1987 and 1998; Romer, N 1t −α A1t −α (hK ,t Kt )α (4)
1990). The right to produce a new intermediate good involves an innovation
cost that represents a fixed cost, because once the knowledge to produce a
It is then clear how the assumption γ = α − 1 (see, for instance, Aghion The returns offered by the investment of rival-resources in intensive R&D
and Howitt, 1998, ch. 12) sterilises the effects of the growing number of are constant or decreasing, depending on θ + ξ = 1 (Barro and Sala-i-Martin,
varieties on final output, which result from the additively separable way in 1995, ch. 7), or θ + ξ < 1. The second case arises if there is a congestion
which the single varieties enter the production function. Where these effects effect on the returns to R&D investment (Stokey, 1995; Howitt, 1999), with
are not sterilised, because (1 − α + γ) > 0, the production function the result that the larger the rival resources invested in research, the higher
corresponding to a constant technology level contains a form of increasing the probability that independent innovation efforts produce the same
returns due to specialisation, as measured by N. The best known example outcome.
along these lines is probably Romer (1990), which assumes γ = 0. The parameter χ is meant to capture how the arrival rate is affected by the
Recalling that in steady state the rate of interest is constant, and the labour frontier knowledge stock At Max. There are two main forces at work here,
and capital shares employed in the (final and intermediate) output sector are which act in opposite directions. Thus, we may split the parameter χ into two
also constant, equation (2) yields the steady-growth equation: components, χ = χ1 + χ2. χ1 is the so called ‘complexity effect’: more
advanced technology levels are progressively more difficult to discover as a
1−α + γ result of the increasing complexity of the search activity. Thus, we have
gY = gL + gN + gA
1−α χ1 < 0. This is the assumption we find in a number of search models of R&D-
based economic growth (Jovanovic and Rob, 1990; Stokey, 1995; Kortum,
where gi is the proportional instant rate of change of variable i. In particular, 1997).9 The parameter χ2 > 0 captures the ‘standing on giants’ shoulders’
if following Romer (1990) we set the restrictions γ = 0 and gA = 0, the above effect (see Merton, 1965; see also Caballero and Jaffe, 1993), which
relation boils down to gy = gL + gN , where it is apparent that the growth rate postulates that a higher frontier knowledge increases the probability of
of per-capita output is simply the growth rate in the number of specialised invention because an investment in intensive R&D creates the opportunity to
varieties. exploit a knowledge spillover from the technology frontier to the innovators.
This positive influence of knowledge on the innovation-success probability is
12.2.2. Intensive Innovations distinct from and indeed adds to the influence of the stock of ideas on the
size of the knowledge shift, which takes place if the innovation arrives (see
An intensive innovation in sector v arriving in the interval t + dt is the (5) above). To this extent, it is unclear what are the grounds for assuming
stochastic outcome of the innovation effort performed at t in this sector. The that the giants’ shoulders effect is positive and is close in absolute magnitude
innovation contributes to shifting the technology frontier according to to the complexity effect. We shall see nevertheless that the restriction
δ χ = χ1 + χ2 = 0 (or other equivalent condition) is characteristic of the R&D
A t Max = At Max (5) endogenous-growth models.
Nt
The main simplifying hypothesis introduced with (6) is that the success
and brings Av, t to the shifted frontier. Thus, access to the frontier technology probability of intensive R&D on variety v is independent of the distribution
level is available, but not costless, to every successful intensive innovator of the local stocks Av, t. Together with (2) this implies that the intensive
operating in sector v. The knowledge increment has elasticity +1 with respect research effort and the arrival rate are uniform across sectors. Other
to At Max and elasticity − 1 with respect to the number of sectors in the formulations (see, for instance, Barro and Sala-i-Martin, 1995, ch. 7) relate
economy (Aghion and Howitt, 1998, ch. 12). The idea is here that the higher the complexity effect and the giant’s shoulders effect for sector v to the local
the number of sectors, the lower the impact of an innovation in sector v on stock Av, t. The same property of a uniform equilibrium arrival rate is however
the technology frontier. imposed also in this case, by means of ad hoc restrictions introduced to this
The Poisson arrival rate of an intensive innovation in sector v at t is: end.
Since each agent engages in R&D independently of the agents in the same
φv,t = λ (uL ,v,t Lt )θ (uK ,v,t Kt )ξ AtχMax (6) or in other sectors and the equilibrium research effort is uniform across
sectors, the aggregate rate of intensive innovations is deterministic and
where ξ ≥ 0, θ > 0, λ is a constant, uL,v and uK,v are the fractions of total labour equals
and capital invested in intensive R&D on variety v.
θ ξ
θ ξ χ ⎛ L ⎞ ⎛ K ⎞
N tφv,t = N t λ (uL , v,t Lt ) (uK ,v,t K t ) At Max = N t λ ⎜ uL t ⎟ ⎜ uK , t ⎟ AtχMax (7)
⎝ Nt ⎠ ⎝ Nt ⎠
12.3. STEADY-GROWTH EQUATIONS
where uL and uK are the aggregate labour and capital shares invested in
A steady state, or balanced-growth path, is a particular constant-growth path
intensive R&D.
such that the growth rate of every variable is constant for ever. Since the
employment shares of the factors cannot exit the interval [0, 1], the definition
12.2.3. Extensive Innovations
immediately implies that the growth rate of such variables is zero on a
balanced path.
An ‘extensive’ innovation is the introduction of a new variety v. On the
The assumptions of Section 12.2.2 imply that the ratio (At Max / At)
assumption that there is an external effect such that the technical knowledge
converges to (1 + δ) (see Aghion and Howitt, 1998, p. 412). Assuming that
in the economy affects the technology level of a new variety, a-not-too-
convergence has already taken place, from (5) and (7) we obtain the
implausible restriction is that the technology level distribution of a new
following shift in the average technology level at time t, resulting from the
variety corresponds to the technology level distribution across the existing
intensive R&D in the N sectors:
varieties (Howitt, 1999). This implies that extensive innovations at t do not
affect the average technology level in the economy At. An assumption to the θ ξ
⎛ L ⎞ ⎛ K ⎞
same effect is that new varieties arriving at t have a deterministic technology A t ,Max = δλ ⎜ uL , t t ⎟ ⎜ uK , t t ⎟ Atχ +1 (9)
level At (Peretto, 1998). ⎝ Nt ⎠ ⎝ Nt ⎠
We assume that the extensive innovation effort is related to the creation of
new varieties by the deterministic law: Recalling that on a constant-growth path A t and At grow at the same rate,
using (4), (8) and (9) we write the steady-state growth equations:
N t = β ( zL ,t Lt ) Nt ( zK ,t Kt ) At ≡ φN ,t
ε τ ψ ν
(8)
gA[− χ] + (ξ + θ) gN − ξ gK = θ n (10)
where β is a constant and zL is the fraction of total labour employed in −ν gA + (1 − τ) gN − ψ gK = ε n (11)
extensive R&D. We impose the restriction ε > 0, ψ ≥ 0, τ ≥ 0, ν ≥ 0. The
case ε + ψ < 1 indicates that there are decreasing returns with respect to the −(1 − α) gA − (γ + 1 − α) gN + gK (1 − α) = (1 − α) n (12)
scale of the rival resources invested in extensive search. The restriction is
referred to as the ‘congestion hypothesis’. A positive τ bears the If we define the variables k ≡ K / N, l ≡ L / N, so that
interpretation that a higher number of varieties amounts to a wider gK = gk + gN, n = gl + gN,
knowledge base in the economy as a whole and therefore facilitates the
discovery of yet new varieties. If this is in itself quite plausible, far more then (10), (11) and (12) yield the following system:
questionable appear to be ‘point restrictions’ such as τ = 1, or τ = 0, as may
be found, for instance, in the pure variety-extension model of Romer (1990) ⎡ −χ 0 −ξ ⎤ ⎡ gA ⎤ ⎡ φ g1 ⎤
and in Peretto (1998), respectively. ⎢ −ν 1 − τ − ε − ψ −ψ ⎥⎥ ⎢⎢ gN ⎥⎥ = ⎢⎢ ε g1 ⎥⎥ (13)
⎢
The parameter ν indicates how the production of an extensive innovation ⎢⎣ −(1 − α ) −(γ + 1 − a) 1 − a ⎥⎦ ⎢⎣ gK ⎥⎦ ⎢⎣(1 − a)g1 ⎥⎦
flow N of technology level A is related to the size of the average technology
index A. If ν = 0, then the cost (in terms of rival resources invested in
extensive R&D) of producing a given innovation flow N with average 12.3.1. Endogenous R&D Growth
technology level A is independent of A (Peretto, 1998). If ν > 0 (< 0), this
cost is decreasing (increasing) in A. The restriction ν > 0 fits with the idea Let [I − Γ ] be the square matrix in the left-hand-side of (13). If [I − Γ ] has a
that the growth of technical knowledge along the quality dimension goes non-zero determinant, the steady-state growth rates of A, N and K are fully
hand in hand with a growing ‘complexity’ of technology, which has a determined by equations (13), hence by technology, given the exogenous
positive effect on the ease with which new varieties are discovered.10
growth rate of population. Thus Det [I − Γ ] ≠ 0 states that preferences do not population, equation (14) is reconciled with the lack of any scale effect on
have any bearing on the speed of steady-state growth and policy measures by the steady-state rate of growth, by introducing special assumptions which
a government are equally ineffective, unless they are able to affect the make sure that L / N is constant (Howitt, 1999), or at least converges to a
technological parameters. It is then apparent that the crucial assumption of fixed steady-state value (Peretto, 1998; Young, 1998). With the simplified
the endogenous R&D growth models is Det [I − Γ ] = 0. In this case the specification of equation (8) considered below (ν = 0, ψ = 0), the required
coefficients in (13) are linearly dependent and additional equations are restriction is τ + ε = 1. This implies:
necessary to determine the steady-state growth rates of the variables. One ε
missing equation is derived from the first-order conditions associated to the N t ⎛L ⎞
= β zLε ,t ⎜ t ⎟
utility-maximisation problem: Nt ⎝ Nt ⎠
∞
c1t −σ − 1 −( ρ − n ) t
Max : ∫ −σ e
t =0 1
∂t and using the steady-state condition gN = n, this yields
1
⎛ n ⎞ε
subject to the flow budget constraint that per-capita consumption at t c t is not mzL = ⎜ ⎟ (15)
negative and is constrained by wage and interest income minus the ⎝β⎠
accumulation of stocks at t (see Barro and Sala-i-Martin, 1995, ch. 2). ρ is
the rate of time preference and (1/σ) is the constant inter-temporal elasticity where m is the steady-state value of L/N. There are two different sets of
of substitution. In particular, the proportional growth rate of ct must satisfy: steady-state solutions of the endogenous model, as specified above, which
correspond to the possibility that: (i) the costs of one additional unit of labour
rt − ρ effort invested in extensive or intensive R&D are identical; (ii) these costs
gc (t ) =
σ are not identical. Case (i) is considered in the next section, case (ii) in
appendix A.
where c is per capita consumption. In steady state n + gc = gY = gK. We shall proceed under the further simplifying assumption γ = α − 1 (see
The restriction Det [I − Γ ] = 0 may be of course introduced in a number equation 12), so that gK = gA + n. Thus gc = gA = (r − ρ ) / σ .
of ways. The standard practice of endogenous growth models with intensive
R&D is to postulate the special case: χ = 0 and ξ = 0 (see, for instance,
Grossman and Helpman, 1991; Aghion and Howitt, 1992; Howitt, 1999; 12.3.1.1. Identical opportunity cost of effort in extensive and intensive
Peretto, 1998; Young, 1998; Barro and Sala-i-Martin, 1995; ch. 7). This is R&D
the case considered in the sequel of Section 12.3.1, yielding: Suppose the only cost of one additional unit of labour effort in extensive or
intensive research is the forgone opportunity of obtaining the wage rate w by
θ
A t ⎛ Lt ⎞ selling that unit in the labour market. Free entry in research implies that, if
= δλ ⎜ uL ,t ⎟ (14) the equilibrium levels of intensive and extensive R&D are positive, then the
At ⎝ Nt ⎠
private instantaneous marginal returns from innovation effort must be
identical between the two activities and must be equal to the wage rate w.
As is also revealed by the first equation of system (13), with χ = ξ = 0
consistency with steady state requires gN = n, that is, gl = 0. In particular, in θ −1 ε −1
the models where extensive innovations are not contemplated, so that N is ⎡ φ v .t ⎤ ⎛ Lt ⎞ ⎡ φN ,t ⎤ ⎛ Lt ⎞
⎢ ⎥ vv,t = λ ⎜ uL ,t ⎟ Vt = wt = ⎢ ⎥ VN ,t = β ⎜ zL ,t ⎟ VN ,t (16)
constant, it is assumed that L is also constant and there is a scale effect of the u L
⎣ L , v ,t t ⎦ ⎝ N t ⎠ z L
⎣ L ,t t ⎦ ⎝ N t ⎠
intensive-research employment level on the growth rates of A and Y. This where Vv, t = Vt is the expected value of a quality innovation in any sector v at
occurs in the pure quality expansion model of Grossman and Helpman time t, and VN, t is the expected value of an extensive innovation at time t.
(1991), Aghion and Howitt (1992) and Barro and Sala-i-Martin (1995, ch. 7). Moreover, with our production function (4) we have:
Jones (1995) draws attention to the lack of empirical corroboration for the
hypothesis of a scale effect on the growth rate. In models with a growing w = (1 − α)hL−α qαA (17)
where q ≡ K / AL. 1
Let vt ≡ Vt /At,Max and vN, t ≡ VN, t /At ; in other words, vt and vN,t are the gA = δ n ⎡⎣λ (1 + δ )θ β −1 ⎤⎦ 1−θ (24′)
productivity adjusted values at time t of an intensive and extensive
Thus we reach the striking conclusion that in the endogenous model as
innovation, respectively. From (16) and (17):
specified above, an identical marginal innovation cost for intensive and
1−θ extensive R&D makes (uL / zL) and gA depend only on technological
1 − α −α α ⎛ L ⎞
vt = hL q ⎜ uL ,t t ⎟ (18) parameters. Instead, the steady-state shares uL,, zL, and hL depend also on the
λ (1 + δ ) ⎝ N t ⎠ preference parameters ρ and σ. (See footnote 14, which refers to the special
case θ = ε.)
1−ε
1 − α −α α ⎛ L ⎞ The reason why the model still qualifies as endogenous is that a policy
vN , t = hL q ⎜ zL ,t t ⎟ (19) variable such as an innovation subsidy (see Aghion and Howitt, 1998,
β ⎝ N t ⎠
p. 419) would affect the rate of growth if it exerted an asymmetric influence
Moreover, one obtains the asset equations (see Aghion and Howitt, 1998, pp. on the cost from one additional unit of labour effort in extensive and
109–10): intensive R&D. For a discussion of this point, the reader is referred to the
case considered in appendix A, where the cost asymmetry does not arise
vt = [rt + φt ] vt − π t (20) from a policy variable, but from a slight generalisation of the innovation
technology considered above.
vN ,t = [rt + φt ] vN ,t − π t (21)
12.3.2. Non-endogenous R&D Growth
where πt is the productivity adjusted profit of a local monopolist. It is worth
recalling that, since an extensive innovation will be displaced by an intensive Referring back again to system (13), the crucial assumption of the non-
innovation in the same sector, the expected obsolescence rate takes the same endogenous R&D growth models is Det[I − Γ ] ≠ 0. In particular, referring to
value φ t for extensive and intensive innovations. the case [I − Γ ]− > 0, standard theorems of linear algebra lead to the
1
Since the productivity adjusted value of extensive and intensive following proposition, which shows that the result similar in spirit to be
innovations are identical in equilibrium, vt = vN,t , which in steady state can be found in Eicher and Turnovsky (1999a) extends to our economy with
written: expanding varieties and technology levels.
θ −1 θ ε −1 ε
(1 + δ )λ uL m = β zL m (22) Proposition 12.3.2.1: Assume Γ ≥ 0. Assume also that, for each row, the row
Using (14) and (15) we obtain: sum of the elements of Γ is positive and lower than 1. Then, for every n > 0,
1
there exist positive values gA, gN, gK that are solutions to (10), (11) and (12)
uL ⎡ θ −ε
⎤
−θ 1−θ and such that gl = n − gN > 0.
= ⎢(1 + δ )λ n ε β ε ⎥ (23) Recalling that 0 < α < 1, a quick look at equation (12) reveals:
zL ⎣ ⎦
1
Proposition 12.3.2.2: If, in addition to the assumptions of proposition
⎡ ⎤
−θ 1−θ
12.3.2.1, we have (γ + 1 − α) ≥ 0, then gK > n (positive per-capita-output
gA = δ ⎢λ (1 + δ )θ n −ε β ε ⎥
1
(24) growth).
⎣ ⎦
Remark 12.3.2.1: The if condition of Proposition 12.3.2.2 amounts to the
In the special, but convenient case θ = ε (23) and (24) simplify to:
existence of increasing returns to scale in the output sector. The assumption
uL 1 of Proposition 12.3.2.1 implies, but is not equivalent to, aggregate decreasing
= ⎡⎣(1 + δ )λβ −1 ⎤⎦ 1−θ (23′) returns to scale in extensive and intensive search.
zL
Thus, where the equations of system (13) are not linearly dependent (notably,
a condition of full measure in the relevant parameter space) the steady-state
growth rates of output, technology levels and varieties are completely 12.5. RESEARCH EMPLOYMENT AND PRODUCTIVITY
determined by population growth and the technological parameters. These
rates are therefore independent of preferences and of savings rates in A second and deeper problem is posed to the R&D growth models by the
particular. stylised facts (a) and (b) mentioned in the introduction. These stylised facts
The above propositions extend to a three-sector environment the formal are at variance with the possibility of approximating (if at a very aggregate
characterisation of the class of two-sector non-endogenous growth models level) the long term evolution of innovation activity and productivity growth
first laid down by Eicher and Turnovsky (1999a). From a formal viewpoint in the U.S. through the hypothesis that this economy has been growing in the
the seminal paper of Arrow (1962), where technology accumulation is driven neighbourhood of a single steady-state path. More specifically, endogenous
by learning rather than deliberate R&D investment, belongs to the same and non-endogenous models alike are faced with the problem of
class. Within the family of R&D growth models, the best-known non-
endogenous example is probably that provided by Jones (1995; see also (i) explaining how the rising researchers/employment ratio (1 − hL) can be
Jones, 1998 and 2002), where the author abstracts from the expansion of reconciled with the behaviour of productivity growth;
varieties, so that gN = 0 and gl = n > 0. In particular, Jones (1995) assumes ξ (ii) identifying the causes of the rising researchers/employment ratio.
= 0 (no physical capital input in R&D) and 0 < −χ < 1, so that his two-sector
A first way of answering these questions is to suppose that the rise in (1 − hL)
version of system (13) boils down to
corresponds to a transitions path with constant growth rate gA induced by
⎡ −χ 0 ⎤ ⎡gA⎤ ⎡ θ n ⎤ exogenous changes in one or more technological parameters.
⎢ ⎥⎢ ⎥ = ⎢ ⎥ A second and more ambitious way is much in the spirit of Pasinetti (1981)
⎢⎣ −(1 − α ) (1 − α )⎥⎦ ⎢⎣ g K ⎥⎦ ⎢⎣(1 − α )n ⎥⎦ and searches for rules of structural change that may get closer to explaining
the observed phenomena without resorting to exogenous parameter changes.
and the conditions of propositions 12.3.2.1 and 12.3.2.2 are trivially In the remainder of this paper we shall expand on these two lines of
satisfied. investigation.
Interestingly, the steady-state relation gc = gA = (r − ρ) / σ continues to To this end, I shall refer to the simplified versions of system (13) that
hold, but the direction of causality at work here is such that, given n, feature in ‘standard examples’ of endogenous and non-endogenous R&D
technology determines gA and r is then determined by gA and preferences. As growth models. In particular, physical capital is not an input to innovation
is discussed in appendix A, in the endogenous model with asymmetric cost activity, intensive and extensive, hence ξ = 0, ψ = 0; the productivity of the
of innovation effort between extensive and intensive R&D, technology and extensive innovation effort does not depend on the technology level A, that
preferences simultaneously determine gA and r. is, ν = 0; the aggregate production function does not depend on the number
of varieties N, thus γ = α − 1. In addition, I introduce the simplifying
restriction ε = θ, that is, the elasticity of innovation output with respect to
12.4. IS n AN UPPER BOUND FOR gN ? R&D labour effort is uniform across extensive and intensive innovations.
As it turns out, the available examples of endogenous and non-endogenous 12.5.1. Looking for Appropriate Parameter Changes
R&D growth models share the prediction that, in steady state, the expansion
of varieties proceeds at a pace which is not faster than the pace of population Referring to the U. S. experience in the second half of the twentieth century,
growth. In particular, gN = n in the endogenous and gN < n in the non- we may observe how the rate of interest and the capital output ratio have
endogenous models considered above. On closer examination, however, been ‘relatively constant’11 over the period. Since the model structure
implies σ gA + ρ = r = α K / Y, using stylised fact (b) we derive the restriction
2
these predictions are the by-product of quite special assumptions. Both the
endogenous and the non-endogenous model admit extensions such that gN that α has been constant; in this Section I am also led to formulate the
may be greater than n. The point is considered in appendix B. ‘working hypothesis’ that the preference parameters σ and ρ were unchanged
throughout. With this situation in mind I consider what, if any, changes of
the technological parameters of the non endogenous and endogenous models
can answer the issues posed under (i) and (ii) above.
12.5.1.1. Non-endogenous model Using (28), (27) and (25), a transition path with
With the assumptions of proposition 12.3.2.1 in place, in particular
δt β
0 < −χ < 1, ε + τ < 1, the non-endogenous model yields the steady-state ≠ 0, λ t ≠ 0, t ≠ 0
predictions: δt λt βt
gY = gA + n
and constant growth rates gA,t, gN,t satisfies:
εn
gN = δt δt λ t β t δt δt λ t β t
(1 − τ ) if < 0 , then + > ; if > 0 , then + < .
δt δt λ t β t δt δt λ t β t
θ (1 − τ − ε )n
gA = Moreover, the steady state share (uL + zL) is independent of λ and β and
− χ (1 − τ )
satisfies12 ∂(uL + zL) / ∂δ < 0, if σ is not too lower than 1. The above
The growth rate of per capita output is independent of δ, the proportional considerations suggest the conjecture that a transition path with rising share
productivity effect of quality innovations; it is also independent of λ and β, (uL + zL) and constant growth rate of productivity is explained by
the parameters that, for any given innovation effort, regulate the arrival rates δt δt λ t β t
of intensive and extensive innovations, respectively. < 0 and + > .
Since the (expected) productivity-adjusted values v t, v N, t of intensive and δt δt λ t β t
extensive innovations are identical, free entry in R&D implies the following To gain some understanding of the problems posed by this lone of
equilibrium condition at every date t: reasoning, it is worth observing that the dramatic rise of (uL + zL) would be
1 obtained through partly offsetting changes of uL and zL. Recalling (28), our
= ⎡ Atχ, Max(1+ δ ) N 1t −τ −ε ⎤
u L ,t λ 1− ε
conclusion here is that the rates of change of the technological parameters δ
(25)
z L ,t ⎣ β ⎦ and λ which are required to explain the stylised facts (a) and (b) may be
implausibly high.
On every equilibrium path sustained by smooth changes of λ, β and δ:
12.5.1.2. Endogenous model
⎡ δ λ β ⎛ u L ,t z L ,t ⎞
g A,t = ⎢ t + t − t + (θ − 1) ⎜
⎤ 1 In addition to the simplifying assumptions stated at the outset of section 12.5,
− ⎟ + (1 − τ − ε )g N ,t ⎥ (26) the endogenous model we are considering assumes χ = 0, ε + τ = 1. The
⎣⎢ 1 + δ t λ t β t ⎝ u L , t z L ,t ⎠ ⎦⎥ − χ innovation technology is that considered in section 12.3.1.1 generating a
symmetric cost from one additional unit of labour effort across extensive and
On a growth path with constant growth rates of A and N:
intensive innovations.13 Following the same line of reasoning explained
above, we obtain that a transition path with smooth changes of λ, β and δ and
⎡ β ⎛ z L ,t ⎞⎤
g N ,t ⎢ t + θ ⎜ + n ⎟ ⎥ (1 − τ )−1 (27) constant growth rates gA,t, gN,t satisfies (28) and
⎣⎢ β t ⎝ z L ,t ⎠ ⎦⎥
β
δt ⎛ ε ⎞
⎡ δ λ ⎛ u L ,t θ z L ,t 1 β t 1 − τ − ε ⎞ ⎤ 1 + λt = t + δt ⎜ ⎟.
g A, t = ⎢ t + t + θ ⎜ − − + n ⎟⎥ δ t λt β t ⎝ δ t(1 + δ t ) ⎠
⎣⎢ δ t λ t ⎝ u L ,t 1 − τ z L , t 1 − τ β t 1−τ ⎠ ⎥⎦ − χ The difficulties encountered by the line of reasoning under investigation
are therefore similar to those discussed for the non-endogenous model.
Substituting from (27) into (26) we obtain:
12.5.2. Growth and Structural Change
δ t uL ,t zL ,t
− = − (28) In a recent paper, Kongsamut, Rebelo and Xie (2001) suggest that the long
δ (1+ δ t ) uL ,t zL , t
term rise in the service-employment share has to do with changes in the
composition of consumers’ expenditure associated with the long term rise of We say that technology level A has been implemented if Av, t = Aj, t = A for all
per-capita income. A tradition in economic theory, from Kuznets (1957) to v, j ∈ [0, ΛA] .Variety v is necessary to the implementation of A if and only if
Pasinetti (1981) had already emphasized this order of phenomena. In a ∈ [0, ΛA].
similar vein, I introduce in this Section the hypothesis that the long-term rise If technology level A(t) is implemented at time t, there is an instantaneous
of the research employment share may be explained by a slow, almost knowledge spillover such that Av,t = A(t) for every v ∈ [0, ∞]. The
negligible secular rise of the intertemporal elasticity of substitution implementation of a higher technology level is instead costly, because it
associated to the long-term rise of GDP per capita14. In what follows, the requires the higher level is independently developed for every necessary
focus of my analysis is not that of giving a detailed specification of the variety as the result of a deliberate R&D effort. The number φv,t of intensive
hypothesis, but is that of suggesting a line of argument explaining how innovations in sector v at t evolves according to the deterministic process:
stylised facts (a) and (b) may be reconciled.
The explanation rests upon the complementarity between the goods and φv, t = λ (uL, v, t Lt)θ Av, t χ
methods used in production. However often neglected, the idea is far from If every innovation has a proportional effect δ on the technology level Av,t,
new. Perceptive remarks on the relevance of this notion can be found in we obtain:
Marx’s volume I of Capital. In chapter XV it is emphasised that the
θ
successful exploitation of new engineering and scientific principles in ⎛ L ⎞ χ +1
production required the emancipation of technology from the pre-existing set A v ,t = δλ ⎜ uL ,t t ⎟ Av ,t (29)
⎝ N t ⎠
of tools15 and that ‘a radical change in the mode of production in one sphere
of industry involves a similar change in other spheres’ (Marx, 1887, p. 362). Higher technology levels are of higher complexity and their
In the new-growth literature, the problem of complementarity between implementation requires a larger number of necessary intermediate inputs.
intermediate goods has been introduced in relation to the idea of a sequence Assume that the number of necessary varieties evolves according to:
of general-purpose technologies (GPTs). The adoption of a GPT requires the
previous creation of a set of intermediate goods that are specific to it.16 ΛA( t ) = Atη η>0
I suggest that a similar set of ideas can be conducive to phenomena of
structural change within a framework which is borrowed, with some This implies that, if gΛ(t) is the proportional growth rate of ΛA(t), then:
important variations or qualifications, from the R&D growth models
considered in this paper. gΛ(t) = η gA(t) (30)
For the sake of simplicity, let us assume away the problem of extensive
The strong complementarities of the form described above imply that the
R&D by assuming that at every date there is an unchanging continuum of
market implementation of a higher technology level will face a host of co-
intermediate-good varieties ordered on ℜ+. To employ these varieties in
ordination problems. Here we are not concerned with this feature, however
production, their appropriate technology level must be developed. [0, ΛA] is
important it may be. Our aim is simply to show that equilibrium paths on
the set of complementary intermediate-good inputs necessary to implement
which the productivity index At grows at a positive constant rate gA > 0 are
the technology level A in the production of final output. Nt is the number of
not steady states and have a rising share uL of R&D employment.
intermediate goods used at t. There is only one final good Y. Its production
In equilibrium, Nt = ΛA(t) . With gA constant, from (29) and (30) we obtain:
function is:
⎛ u ⎞
⎡∞ ⎤ − χ gA = θ ⎜⎜ n + L ,t − gΛ (t ) ⎟⎟
⎢ ∫ P( Av,t )x v ,t ∂v ⎥
α −1 1−α α
Yt = N L uL ,t
t Y ,t
⎣ v=0 ⎦ ⎝ ⎠
where P(Av, t) is the productivity index associated to the technology level Av,t ⎡ χ⎤ ⎛ uL ,t ⎞
of variety v. with P(Av, t) = A, if 0 ≤ v ≤ ΛA and Av, t = Aj, t = A for all v, j ∈ ⎢η − θ ⎥ gA = ⎜⎜ n + u ⎟⎟
⎣ ⎦ ⎝ L ,t ⎠
[0, ΛA]; P(Av, t) = 0 otherwise. This assumption formalises a strong form of
incompatibility between intermediate goods of a different technology level.
Recalling that the ‘congestion effect’ in R&D implies θ < 1, and that our intermediate goods of a different technology level yields the result that
considerations suggest χ < 0, it is easy to see that, given n, the higher the structural change in the form of a rising R&D employment share is a
value of η, the higher the growth rate uL ,t uL ,t required to elicit a given necessary condition for the sustained growth rate of productivity experienced
productivity growth gA . Thus, with η sufficiently large, the value gA ≈ 0.02 in the U.S. during the second half of the twentieth century.
prevailing in the period 1950–93 would not have been possible in the
presence of a constant labour share in R&D. Indeed, a growth rate gA of the
observed dimension cannot be a steady-state growth rate and cannot be APPENDIX A: ASYMMETRIC INNOVATION COST
sustained ‘for ever’.
If the argument above offers a tentative explanation of how the long-term Suppose that every unit of labour invested in R&D at time t is combined with
rise of the researchers/employment ratio can be reconciled with a constant a quantity of capital At,Max TA, in the case of intensive R&D and At TN in the
growth rate of productivity, what is yet to be explained is the source of the case of extensive R&D. In this section I assume TN ≠ TA. In other words,
rising researchers/employment ratio. labour and capital are perfectly complementary inputs to intensive and
Here I suggest, as a working hypothesis to be explored by future work, extensive innovation activities, but the ratio between the two inputs is
that the preference structure with constant inter-temporal elasticity of different in the two sets of activities, even after adjustment is made for the
substitution is replaced by a preference structure such that the rising per- productivity levels At,Max and At. The case TN = TA yields conditions identical
capita consumption causes a slowly rising inter-temporal elasticity of to those obtained in Section 12.3.1.1, with the understanding that terms K
substitution. Since hL is close to 1 and uL is close to zero,17 the required and q must be replaced everywhere with h K K and h K q , where hK is the
change in σ does not have to be large, since a very small, seemingly fraction of total capital employed in the output sector (to produce
negligible, shift away from employment in manufacturing in favour of intermediate goods). uK and zK are the fractions of total capital employed in
research is sufficient to explain that: intensive and extensive R&D, respectively. With this notation, and assuming
for simplicity θ = ε, the procedure followed in Section 12.3.1.1 yields:
(i) hL ,t hL ,t is negative but very close to zero, as in the data; 1
(ii) uL ,t uL ,t is positive and significantly large, as in the data. ⎧ ⎤⎫
1−θ
⎪ λ (1+ δ ) ⎢⎣(1−α ) +α ( α )
⎡
1
r 1−α
⎥⎦ ⎪
2
2 TN
uL
=⎨ ⎬
⎪ β ⎡(1−α ) +α 2⎛⎜ r ⎞⎟ ⎤ ⎪
1
12.6. CONCLUSIONS zL 1−α
TA
⎩ ⎢⎣ ⎝α ⎠ ⎥⎦ ⎭
2
In spite of the statements to the contrary (Jones, 1995; Aghion and Howitt,
1998, ch. 12), growth models that avoid the scale effect of R&D employment
Hence uL /zL is related to the steady-state rate of interest, which depends on
on productivity growth do not explain the evidence on R&D employment and
the preference parameters ρ and σ. In particular, it can be easily checked that
productivity growth in the U.S. Indeed, the stylised facts (a) and (b)
the sign of ∂( uL/zL) /∂r is positive if TN −TA > 0 and is negative if TN − TA < 0.
mentioned in the introduction are not easily reconciled within the standard
Moreover, similar considerations apply to the relation between gA and the
steady-state hypothesis.
rate of interest. We can write:
The first reason offered in this paper is that cross-sector research
r−ρ
spillovers are less extensive than is normally assumed in R&D models: After gA = = f (r , λ , δ , β ,α ,θ , TN , TA )
a new basic idea is first discovered, the development and profitable σ
implementation of the same idea in the production of final utput is a costly
If TN − TA ≠ 0, then r is a non-redundant argument of the function f( ) and,
process. A second reason is that the number of complementary inputs
given n, gA and r are simultaneously determined by technology and
necessary to implementa technology level A in the production of final output
preferences. If TN = TA the simultaneity collapses and gA is determined by
is likely to be an increasing function of A. The further assumption of
(24′).
complementarities in the form of strong incompatibilities between
APPENDIX B: EXISTENCE OF SOLUTIONS WITH gN > n 4. By definition, on a steady-state path the growth rate of every variable is constant for ever.
Since the employment shares are bounded between zero and one, their unique admissible
steady-state growth rate is zero.
It is enough to provide two examples: one for the endogenous model and one 5. Peretto (1998) reports on the transition dynamics of an R&D growth model where the
for the non-endogenous model. In both examples the simplifying restriction endogenously accumulating factors are only A and N. In the transition dynamics results of
γ = α − 1 holds so that Det[I − Γ ] = −(ξ + χ) (1 − τ − ε − ψ)(1 − α). Aghion and Howitt (1998, pp. 109–15), the endogenous factors are A and K.
For the endogenous model with χ = 0, ξ > 0, υ > 0, the crucial restriction 6. Still, in reading it, it is best to bear in mind what is implied by the seminal work by Jacob
Det[I − Γ ] = 0 is fulfilled by τ + ε + ψ = 1. In this case Schmookler on innovation and growth: the interest in the causes of the long-term growth of
GDP per capita, as distinguished from the GDP level, is at best only a partial justification for
N = N β zL ( L / N ) A
ε ε υ the rigid supply orientation of general-equilibrium growth models.
7. The assumption is not fully realistic. Even granting that Av amounts to a productivity index,
we should in general expect the flow of service characteristics associated with (v, Av) to
which in steady state requires ε (n − gN) + υgA = 0. If 0 < υ < ε, this yields depend upon the type and quantity of other intermediate goods with which (v, Av) co-
gKì = gA + n > gN. Since from (10) gK = gN − (θ / ξ) (n − gN) we have that gN > n operates within a production activity. If there are strong complementarities between different
and gA > 0 are consistent with a steady state path. intermediate goods, the best-practice technology level of variety v at t may not be the
highest available. Compatibility constraints may in fact imply that it is inefficient to use very
For the non-endogenous model it is sufficient to assume τ < 1,
different technology levels of complementary varieties in the same activity.
τ + ε + ψ > 1; ξ + χ > 0, ν and ψ sufficiently close to zero. Complementarities of this sort are simply ruled out in most (an exception is Helpman and
Trajtenberg (1994); see section 12.5.2 below) R&D growth models.
8. If 1 > ω, then the monopoly output is positively related to the technological advance Av,t.
9. Realistic as it may be, the positive correlation between the technology-frontier index and the
NOTES difficulty of search must be simply assumed and cannot find a micro foundation within a
formal framework which does not lend itself to consider the feedback of innovations on the
1. There are instances of R&D activities performed in a given country which exert their complexity of the search space.
productivity effects mainly outside the country: think of a new treatment for curing a 10. As before, since the present framework cancels from view the rising complexity of the
tropical disease discovered in the U.S.A. or in Germany. The view taken in this paper is that technology space, the treatment of this feature can be at best evocative.
this type of phenomenon is far from explaining the qualitative evidence presented in the text. 11. At least in the sense specified in the introduction to this paper.
I thank Francesco Pigliaru for drawing my attention to this point. θ n(1 − τ − ε )(1 + σδ ) − χδ [( ρ − n )(1 − τ − ε ) + ρε ]
−1
2. To reconcile facts (a) and (b), two candidates come to mind. (i) There has been a fall in the 12. (uL + z L ) = 1+
average effect of innovations on measured productivity. This may be at least partly due to (1 + δ )(1 − τ − ε )αθ n − χδεα n
the fact that official statistics underrate the qualitative changes in goods and the 13. The research employment shares are:
improvement in their service characteristics (Nordhaus, 1997). Alternatively, or in addition
1
to the previous cause, it may be increasingly difficult to produce the same proportional
⎡ λ⎤ 1− ε
improvement in the service characteristics of goods. Hence, the productivity gain tends to
⎢⎣(1 + δ ) β ⎥⎦
fall in the more recent innovations. Robert Gordon (2000) compares the effects on well- u = L 1
being of the ‘new economy’ to those produced by the great innovations during the second
industrial revolution. He concludes that the effects of the former do not bear comparison 1+
ρ ⎡1 + σδ + 1 ⎤ ⎡(1 + δ ) λ ⎤ 1−ε
with those of the latter. (ii) A different, but compatible, line of explanation is a fall in the α n ⎢⎣ (1 + δ )α ⎥⎦ ⎢⎣ β ⎥⎦
average productivity of R&D labour, as measured by the number of innovations per unit of
−1
research effort. A fall of this kind has certainly taken place, if the number of innovations is
⎡ ρ ⎡ σδ + 1 ⎤ ⎡ ⎤
1
λ⎤ 1− ε
z = ⎢1 + ⎥
measured through the number of patents, granted or applied for (Griliches, 1988, 1990).
1+ (1 + δ )
Measures of this type are strongly biased not only by changes in the ‘productive capacity’ of L
⎢ ⎥ ⎢
⎢⎣ α n ⎣ (1 + δ )α ⎦ ⎣ β ⎥⎦ ⎥⎦
institutional patent agencies (e.g. the U.S. Patent Office), but also by changes in the
propensity to apply for a patent. Microeconomic studies (Lanjow and Schankermann, 1999)
indicate that a lower fall of the productivity of R&D labour is obtained if the aggregate It can be easily verified that : ∂uL / ∂(λ/β) > 0; ∂zL / ∂(λ/β) < 0; ∂(uL + zL) / ∂(λ/β) < 0 and
innovation output is obtained by weighting patents by means of indicators of their ∂(uL + zL) / ∂δ < 0 if σ > [(1 + δ)ρ − n] / δn; ∂(uL + zL) / ∂(λ/β) > 0 and ∂(uL + zL) / ∂δ > 0 if
technological and economic importance. This is related to point (i) above. σ ≤ 1.
3. We shall not consider other families of models where growth is likewise driven by 14. The hypothesis implicitly assumes some measurement error leading to a (very) mild under-
innovations, let alone the huge microeconomic literature on R&D. evaluation of productivity growth. See above, note 2.
R&D models of economic growth 259
15. ‘It is only after considerable development of the science of mechanics, and accumulated
practical experience, that the form of a machine becomes settled entirely according to
mechanical principles, and emancipated from the traditional form of the tool that gave rise to
it’ (Marx, 1887, p. 362, n. 1).
16. When the GPT s first appears a labour share is shifted from manufacturing to R&D
(phase 1); next, after the intermediate goods required by s have been invented all
employment is shifted to manufacturing until the GPT (s + 1) arrives (phase 2). The idea is
13. Competition and technical change in
exploited by Helpman and Trajtenberg (1994) and Aghion and Howitt (1998) to study the
relation between growth and cycles. The notion of a steady state is correspondingly
Aghion & Howitt: a formalisation of
extended by these authors to the effect that in an economy with a constant population ‘a
steady-state equilibrium is one in which people choose to do the same amount of research
Marx’s ideas?
each time the economy is in phase 1 …’ (Aghion and Howitt, 1998, p. 248).
17. The U.S. researchers/employment ratio was 0.008 in 1993 (see Jones, 2000, p. 16). Maria Daniela Giammanco
13.1. INTRODUCTION
The aim of this paper is to identify some similarities between Aghion and
Howitt’s analysis and Marx’s work with respect to competition and technical
progress as presented in Aghion and Howitt’s 1992 model, and some of the
extensions proposed by the two authors in 1998.
The comparison is not based on analogies between the methods of
analysis, which reflect two different visions of the world. Marx investigates
the distribution of the surplus produced in the economy among conflicting
social classes and conflicting capitals. Aghion and Howitt propose a
neoclassical general equilibrium model whose parameters are individual
initial endowments, individual preferences, technology and the probabilistic
production function of new technology. The proposed comparison stems
rather from the identification of some fundamental assumptions
characterising Aghion and Howitt’s model, which are also to be found at the
basis of Marx’s idea of competition and technical progress.
Aghion and Howitt’s work belongs to the New Growth Theory (NGT)
tradition. Romer (1994) affirms that the origins of the NGT must be sought
not only in the effort to solve the ‘convergence problem’ but also in the
effort to build a valid alternative to perfect competition at a theoretic level.
The need to abandon perfect competition stems from the necessity to explain
one of the determinants of economic growth ignored in the neoclassical
model: endogenous technical progress. Aghion and Howitt (1992)
acknowledge the pioneering contributions to the endogenous growth theory
of Romer (1986c) and Lucas (1988) and propose a model in which economic
growth is driven solely by innovation resulting from the firm’s research
activity. Such an activity is performed because, if it meets with success,
monopoly rents are earned, though momentarily; these rents are possible
because knowledge, as in Romer (1990b), not only has a non-rival
260
Competition and technical change in Aghion and Howitt 261 262 The Theory of Economic Growth: a ‘Classical’ Perspective
component but is also a partially excludable good. Non-rivalry, together with creative destruction and to the revolutionising of the economic structure by
partial excludability, introduces increasing returns internal to the firm and new production methods: hence they call their model neo-Schumpeterian.1
therefore imperfect competition. Aghion and Howitt present a general equilibrium model, with perfectly
An investigation of the characterising assumptions made by Aghion and competitive final good and research sectors and a monopolised intermediate
Howitt will show that these assumptions are really close to the elements one. The final good is produced by the intermediate input together with the
which characterise Marx’s view of competition and technical change. A total endowment of unskilled labour of the economy whose amount is fixed.
companion paper (Giammanco, 2002) focuses on the two conceptions of As a consequence the final good production function is represented by a
competition proposed by Marx. In Marx’s analysis competition is a negative constant returns function, the only argument of which is the intermediate
force and also a process. From the idea of competition as a negative force, good multiplied by a parameter representing the intermediate good’s
destroying all barriers to the free development of capital follows a vision of productivity and increasing by the same amount with every successful
technical progress as a powerful weapon of capital in its struggle for innovation. For each firm the invention of a new intermediate input is a
complete command over labour power, reduced to simple labour. Marx’s random event governed by a Poisson distribution, with an arrival rate which
idea of competition as a process, inseparable from the capitalistic mode of is a constant returns to scale function, the arguments of which are researcher-
production where a complete command of capital over labour has already specific skilled labour and specialised labour, each invention being an
been established, leads to the idea of technical progress as a powerful tool in independent event at the aggregate level. The monopolistic intermediate
the struggle among capitals in the fully developed capitalistic society. sector produces the intermediate good using skilled labour together with a
Technical progress as an instrument in the struggle among capitals focuses design: to find the equilibrium means to determine, in every period, the
on a major aspect of Marx’s thought – the deliberate action of the allocation of the given amount of skilled labour between intermediate and
capitalist/innovator moved by the desire/need to innovate, which engenders research activities.
relentless technological change. The strongest analogy between Aghion and Howitt’s work and Marx’s
Aghion and Howitt do not refer to Marx as their inspiring Muse. analysis is that the mechanism generating growth in Aghion and Howitt’s
However, their work, like many within the NGT tradition, is stimulated by model is the mechanism which propels the accumulation of capital in Marx’s
the many empirical studies on technological change as a source of growth, analysis. In Aghion and Howitt’s work monopolistic competition in the
engendered by profit-rent seeking firms. This is a cornerstone of Marx’s intermediate sector drives growth. In this sector, perfect competition is
vision; those who know Marx might find it stimulating to encounter this and abandoned thanks to the introduction of the Schumpeterian idea of creative
some of his other crucial arguments treated formally within an elegant destruction. This idea is based on the assumption that knowledge is partially
general equilibrium construction. excludable: the firm that innovates obtains a patent that lasts forever, but will
The first section of the work presents Aghion and Howitt’s model of be used only till a better intermediate input is invented. Perfect competition
1992; it also outlines the fundamental issues underlying their idea of cannot prevail in the intermediate sector, because the cost of the patent is a
competition and technical progress, which play a crucial role in Marx’s fixed cost which the innovating firm will pay only if it can sell its output at a
analysis. The second section enlarges the analysis to some of extensions of price higher than the marginal cost. Technical progress by means of creative
the 1992 model proposed by Aghion and Howitt in 1998. The third section destruction makes the monopoly power of the successful innovating firm last
offers some conclusions. only for a period, and creates new rent opportunities for the next innovator.
Growth in Aghion and Howitt’s model is therefore engendered by the
prospect of monopoly rent which makes firms carry out research; it stems
13.2. AGHION AND HOWITT’S ANALYSIS from rivalry among innovating firms. In Aghion and Howitt’s analysis, the
innovating firm plays the role of the capitalist/innovator whose action
Aghion and Howitt (1992) focus their attention on technical progress continuously revolutionises the economic structure in Marx’s world. In both
improving the quality of products; by doing so, they consider the frameworks the existence of a patent system is crucial (on the patent system
obsolescence created by technical progress which consequently engenders which serves to meet capitalists’ demand for differentiation in Marx, see
losses as well as gains. They refer specifically to Schumpeter’s idea of Giammanco, 2002). A major aspect of Marx’s thought is in fact the
deliberate action of the capitalist/innovator moved by the desire/need to
innovate, which engenders relentless technological change; the introduction should create, reducing labour to simple labour. By rendering the worker
of a new technology within a productive sector raises labour productivity and indifferent to the precise nature of the work, this facilitates inter-sectoral
allows for transient extra-profits, until it becomes widespread, igniting an labour mobility – the crucial premise for capital mobility and the hallmark of
endless struggle in the production arena. As profit rates tend to equalise, this a fully developed capitalistic society (see Giammanco, 2002). It must be
attracts capitals towards the innovating sector and engenders ever-increasing said, however, that skilled labour in Aghion and Howitt’s analysis cannot be
demand for new and more efficient technologies (see Giammanco, 2002). considered as labour according to Marx, but as an asset which seeks the most
The strong analogy is therefore evident between Marx’s analysis, in remunerative use.
which continuous mutation in the economic structure is achieved by the On the implication of the skilled labour market equation which reflects the
incessant differentiation of each firm, and Aghion and Howitt’s process of frictionless nature of this market and assumes market clearing, it is worth
creative destruction, which engenders obsolescence and therefore winners referring to what Duménil and Lévi (1987) argue: the common characteristic
and losers. However, while according to Marx firms show differences in the of all dynamic equilibrium models consists in equality between supply and
mode of exploiting common knowledge (see Giammanco, 2002), Aghion demand; it is possible, however, to conceive of different market mechanisms
and Howitt, because of the general equilibrium framework chosen, do not which respectively lead to: i) an ex-ante equilibrium in which demand
formally tackle the problem of firm differentiation. In every period, the immediately matches supply, as in the Walrasian model with production; ii)
structure of the economy is characterised by perfect competition in all sectors an ex-post equilibrium which is only reached asymptotically. The supply is
but the intermediate one. Splitting their model into a perfectly competitive determined together with the prices of the goods before the demand is
section and a monopolistic one, Aghion and Howitt try to confine the known, and the information available is that of the previous period. The
analysis of firm differentiation to the study of the monopolistic sector. In this NGT uses the ex-ante equilibrium concept because it adopts an intertemporal
sector, however, what changes from period to period is the firm which plays equilibrium model, in which the Walrasian auctioneer determines the
the role of the incumbent: while the present incumbent does not try to complete sequence of events in advance. Marx uses the ex-post equilibrium
differentiate, outside firms strive to innovate and gain the monopoly position. concept: the competitive process is characterised by adaptation to
This, however, does not engender differences in the research function disequilibrium: the system gravitates around the equilibrium but does not
available. The economic change results in changes in the productivity reach it within the period considered.2
frontier of the final sector, which causes growth, and in a change of firm The interpretation of competitive equilibrium as an ex-ante or as an ex-
which assumes the role of incumbent in the intermediate sector. post equilibrium modifies the rationale behind the mechanism generating
Innovation as the mechanism generating growth may be further examined growth. In Aghion and Howitt’s model it is the amount of skilled labour
by looking at the backbone of Aghion and Howitt’s model. The problem of devoted to research which determines growth. The average growth rate is
finding the equilibrium is solved by satisfying an arbitrage condition in the determined through a forward-looking difference equation, which sets the
labour market, together with a labour market clearing equation. The arbitrage amount of skilled labour employed today in terms of the amount of skilled
condition establishes that for a positive amount of skilled labour used in labour employed tomorrow. What will be done tomorrow is known because
research, the hourly wage paid in manufacturing must equal the expected of the perfect foresight assumption: perfect foresight is ruled out in Marx’s
value of an hour’s research; the clearing equation states that the amount of analysis, in which technological progress is a path-dependent problem-
skilled labour used in research and the amount used in manufacturing must solving activity (see Giammanco, 2002).
add up to the given endowment of skilled labour. These two conditions However, Grossman and Helpman (1994) argue that the NGT is well
embody the idea that at the basis of the functioning of the model is perfect aware that firms struggle continuously at a micro level, and that growth is an
competition, characterised by the free movement of assets towards the most irregular and stochastic process; but they argue that ‘aggregation masks this
remunerative use and by the frictionless nature of markets. micro-level turbulence’ (p. 34). Ignoring the continuous change occurring in
The arbitrage condition reflects the fact that labour can be freely the competitive sector allows Aghion and Howitt to set the stage and give all
employed either in the research sector or in the manufacturing sector, the information necessary for the solution of the monopolistic maximization
imposing that skilled labour be equally remunerated in both sectors. This problem. This can be compared to the adoption of a long-run approach of
condition stylises Marx’s idea of the need for free mobility of capital towards classical inspiration,3 to concentrate on the general laws governing the
the most remunerative uses: it states what competition as a negative force
gravitational centres without investigating the analysis of disequilibrium knowledge increases as the economy develops. As firms expand, the
which still remains at the heart of the competitive process. accumulated experience enlarges, with more possibilities for exploitation and
Aghion and Howitt’s choice to introduce a Poisson distribution regulating for additions. There is also a strong link between collective labour, the
the outcome of research activity is a compromise between awareness of the possibility of exploiting ideas and scale economies: large plant-size allows
uncertain outcome of research and the need to model through a well-defined not only constant capital saving, due to increasing returns to scale, but also
probability function. Aghion and Howitt’s concern shows close links with the accumulation of practical experience, through the work of the combined
Marx’s notion of the uncertainty of the innovative action. Marx speaks of the collective worker, often necessary in order to exploit the common heritage of
daring capitalist and is aware that trail-blazers often go bankrupt while ideas (see Giammanco, 2002).
money-capitalists, who merely acquire machinery and buildings at a low
price from them, win the competitive struggle (see Giammanco, 2002). Per
se the uncertainty characterising the outcome of innovative activity in Marx’s 13.3. AGHION AND HOWITT (1998): SOME
analysis is not a statistically computable risk. However, a well-defined EXTENSIONS TO THE BASIC MODEL
probabilistic distribution governing the outcome of research is an analytical
necessity to model uncertainty within the limit imposed by a general Aghion and Howitt (1998, ch. 6) propose a further development to their 1992
equilibrium framework. model in which they try to represent the firm and its research effort. They
The aggregate production function of the final sector with its intermediate acknowledge that in order to exploit research-generated knowledge, a firm
good productivity parameter introduces in Aghion and Howitt’s analysis a must apply its own theoretical knowledge in practice, solve unexpected
scale effect that can be found, in slightly modified versions, in many New problems and grasp new opportunities. They therefore deal with
Growth Models (NGMs). Aghion and Howitt assume that the final output heterogeneity in the innovative structure by distinguishing between
grows with the growth in the level of aggregate knowledge. Every successful fundamental and secondary innovations, which are complementary to each
innovation allows for an increase in the final sector productivity by a other. They consider two extreme and co-existing kinds of research: R&D
constant proportional factor, i.e., it allows for the use of more efficient and learning by doing.4 R&D results in potential new products, whereas
methods in the production of the final good. The productivity rise due to learning by doing results in improving existing ones. With learning by doing,
successful innovation lasts forever, which introduces an important Aghion and Howitt model an element which is at the basis of Marx’s
intertemporal spillover effect in the model. This scale effect in Aghion and conception of technical change. Ideas can be produced not only by an R&D
Howitt’s work implies that each new innovation allows other researchers to activity, but also as the outcome of a learning activity within production.
begin operating on the next one. At the macro level, size is therefore Marx distinguishes between radical and incremental invention: a heritage of
important. It is worth stressing that, though the intertemporal spillover effect pre-existing ideas, universal labour, is a prerequisite for the birth of a new
is an aggregate phenomenon, it is limited to the final output sector, as the idea, thanks to which radical inventions may be produced. Each of these
systemic accumulation of knowledge engenders aggregate increasing returns engenders a sequence of incremental inventions, thanks to the learning
solely in the final output sector, while research activity is characterised by process of collective labour, and increases systemic knowledge (see
lack of memory (Aghion and Howitt, 1992, p. 327). This arises from the fact Giammanco, 2002). Aghion and Howitt introduce the idea of general
that the innovation arrival rate depends solely on present research and not on knowledge, ‘the common scientific, technological, and cultural heritage
the stock of past research. potentially available to everyone’ (p. 174), which is engendered by both
The scale effect in Aghion and Howitt’s model recalls Marx’s idea of the R&D and learning by doing and which is really close to Marx’s idea of
accumulation of experiences at a macro level, which become part of the universal labour.
general knowledge potentially available to everyone. In Marx the Aghion and Howitt propose a first variant of the basic model, in which
accumulation of general knowledge is one of the levers of the production of learning by doing is only accumulated at the macro level. They assume the
new ideas at the basis of technical progress. However in Marx’s analysis the existence of a constant mass of infinitely lived skilled labourers, each of
scale effect is both a micro and a macro phenomenon; magnitude is whom can choose to work either in research or in productive sectors. There
intimately connected with the capital accumulation process, being important is an R&D sector which invents intermediate goods by means of research
both at the systemic and at the firm level. The systemic accumulation of labour and general knowledge. Its outcome is ruled by a Poisson distribution.
The arguments of the aggregate production function of the final good sector progress in one industry’s mode of production revolutionises that of other
are the quantity of labour used in the production of each intermediate good of industries (see Giammanco, 2002).
different vintages, the quality of such intermediate goods, and the state of The interpretation of the arbitrage conditionis similar to that of the basic
general knowledge. In this sector secondary innovations are jointly produced model. In this model, however, a fixed Poisson upgrading rate has been
and are not internalised by each single firm: they improve the quality of introduced, which is the rate at which a worker can switch from producing an
already existing intermediate goods at the systemic level. The quality of the old to producing a new vintage intermediate good, or to producing research.5
last vintage intermediate good is zero, continuously increased by the From this it follows that only when skilled labour is free to move between
accumulation of systemic learning by doing. More recent vintage goods are research and the new vintage good sector, if it is exogenously upgraded,
potentially better, as they incorporate more general knowledge. Each firm must its expected remuneration be equal in both sectors. This makes more
producing an intermediate good of a specific vintage is a monopolist in that explicit the importance of freedom of circulation of skilled labour, without
sector: it must compete with other firms in order to hire production labour. which equality in remuneration ceases to exist, and creates a stronger link
As in the basic model, the steady-state growth rate is found by with Marx’s analysis, according to which mobility of labour is a necessary
determining the steady-state proportions of workers engaged in research and condition for capital accumulation, hence growth.
in production. This is obtained by means of an arbitrage condition in the Upgrading implies that the amount of production labour devoted to each
skilled labour market, determining the allocation of labour between research intermediate good will fall exponentially with an increase in the age of the
and production. intermediate good: obsolescence is represented by a crowding-out process
The importance of this variant of the basic model stems from the explicit rather than a creative destruction process. Thus, Aghion and Howitt propose
introduction of learning by doing, which endorses the following ideas: alternative dynamics to the process of creative destruction, interpreting
innovations are not necessarily radical but can also be incremental; capital struggle as a more gradual process, and take into account another
incremental technical progress can be a by-product of the production process aspect of technical progress present in Marx’s analysis: Marx believes that
which originates and develops new ideas aiming to solve practical problems; technological trajectories have unknown potentialities which are discovered
and the experience engendered by the production process accumulates at a only with the passing of time (see Giammanco, 2002).
systemic level. Aghion and Howitt make learning by doing appear indirectly Aghion and Howitt introduce the idea of product life-cycle: as old
in the aggregate production in two ways: by determining the quality level of intermediate goods incorporate less aggregate knowledge than new ones,
the intermediate products of each vintage, and by influencing the level of with the introduction of new intermediate goods the contribution to
general knowledge; the growth equation of general knowledge is in fact a production of old vintage goods decreases continuously as they become
function of the current flow of fundamental and secondary innovations and obsolete. This creates another analogy with Marx, who envisages a sort of
of the stock of previous general knowledge. The definition of general product life-cycle for each new machine, a prototype that can be improved in
knowledge and the equation which governs its growth formalise Marx’s the production phase (see Giammanco, 2002).
conception of the systemic accumulation of knowledge: ideas and production Aghion and Howitt also propose a further variant of the basic 1992 model,
experience become part of social knowledge accessible to everyone. in which learning by doing is completely internalised by the firm.6 Only the
The production function of learning by doing, regulating the quality firm that solves problems within the productive process is able to improve its
growth of intermediate goods of each vintage, not only models the product quality. This means that as production workers can appropriate some
accumulation of production experience at a macro level, which is strongly of the fruits of their learning-by-doing activity, they must be compensated
connected to Marx’s learning process of combined collective labour which accordingly in a competitive equilibrium. The introduction of internalised
increases systemic knowledge; it also sketches the idea of the learning by doing formalises another important element of Marx’s analysis:
interdependence among productive sectors. This is so because the the idea of learning through the accumulation of practical experience by the
simultaneous amelioration in quality of all products can be explained by the combined worker, which increases labour productivity within each firm.
close interdependence among sectors which forces technical progress to With internalised learning by doing, Aghion and Howitt find that the size
spread immediately. According to Marx, many of the productive sectors can of the adaptability parameter is positively related to the growth rate, as long
be seen as gears of the same complicated mechanism. As a consequence, as the growth of fundamental knowledge does not depend too much on
learning by doing. They explain this as a consequence of the increase in the
share of skilled labour devoted to a new vintage good, which increases production itself and is characterised by the struggle among capitals. In this
research productivity and reduces the cost of implementing an innovation.7 struggle, technical progress is a very powerful weapon: it is the outcome of
This result further stresses the importance of worker adaptability in Aghion deliberate research activity and allows the innovator/capitalist to increase his
and Howitt’s model8 and its analogy with the importance of labour profit. These ideas are developed by Aghion and Howitt’s 1992 analysis,
flexibility, stemming from the action of competition as a negative force, in which deals with creative destruction. Marx’s systemic accumulation of
Marx’s work. knowledge can be considered modelled, in Aghion and Howitt (1992), by the
Aghion and Howitt also study another fundamental/secondary dichotomy intertemporal spillover effect which increases productivity at the systemic
by focusing on research and development. Research produces fundamental level.
innovations, each consisting of a potential line of new products; development Aghion and Howitt’s 1998 extension, in which they introduce two forms
produces a workable plan for actually producing one such potential product. of learning by doing, matches an important notion of Marx’s analysis: the
They explicitly model Marx’s ideas on the difference between radical and difference between radical and incremental innovation. Learning by doing
incremental inventions. The method of analysis is analogous to that applied accumulated at a macro level also models the idea of Marx’s systemic
to learning by doing: also in this case there is a positive relation between accumulation of knowledge development, by means of accumulation of
growth and the workers’ adaptability, provided that the growth of experiences at a macro level. Learning by doing, increasing the productivity
fundamental knowledge does not depend too much on development. of already existing intermediate goods, also shapes the interdependence in
Aghion and Howitt (1998, ch. 7) consider not only product market technological advancement of different sectors of production present in
competition but also competition as increased freedom of entry in the Marx’s analysis. The learning by doing internal to each firm recalls the
research sector. They identify the competitiveness level with the entry cost, importance of the learning process by means of Marx’s combined worker.
and demonstrate that a higher degree of competitiveness in the research The relevance of the worker’s adaptability parameter models Marx’s idea of
sector stimulates growth because it stimulates innovation. They refer to their the importance of the extent of capital/labour mobility, which lies at the basis
basic model of 1992 and relax the assumption of constant returns to scale in of the action of competition as a ‘negative force’. The product life-cycle is
the research sector; this is still governed by a Poisson arrival rate, defined as also a problem found in Marx. The 1998 extension, in which Aghion and
an increasing and concave function of the total labour employed in the Howitt introduce barriers to entrance, models Marx’s idea of a minimum
industry minus the entry cost, multiplied research productivity. This allows plant-size required to implement inventions.
them to find, as in the basic model, the steady-state growth rate which While there are ideas characterising Marx’s treatment of competition and
responds positively to an increase in competitiveness, i.e. to a reduction in technical change that underlie Aghion and Howitt’s analysis, attention has
the entry cost. According to Aghion and Howitt this result vindicates also focused on the parts of Marx’s analysis that have not been developed by
Schumpeter’s claim that the higher the competition in research, the higher these authors. Although in their models competition drives firms to innovate
the growth. Also in Marx’s analysis the presence of barriers to entry in the in order to gain monopoly power, their analysis greatly simplifies the
more technologically advanced sectors is an element restricting the representation of the struggle for diversity characterising Marx’s work.
exploitation of technological advances which are potentially available to Moreover, Aghion and Howitt (1992) are aware of the uncertainty of the
everyone (see Giammanco, 2002). Accordingly, Aghion and Howitt, dealing innovation activity considered by Marx and they introduce its proxy, which
with barriers to entry in the research sector, model another crucial idea which is risk: the invention of a new intermediate input is a random event governed
characterises Marx’s analysis of capital accumulation. by a single Poisson distribution. This models the firm’s effort to innovate,
hence to differentiate, but does not take into account that each firm, through
research, will differ from other firms and have a distinct research production
13.4. CONCLUSION function. The same can be said of the 1998 extension which, with the entry
barriers hypothesis, envisages identical firms facing the same entry cost and
In the previous pages I have suggested that the ideas underlying Aghion and the same functional form of the arrival rate. These hypotheses are related to
Howitt’s 1992 model, and part of its extension of 1998, can be found in Aghion and Howitt’s reliance on a general economic equilibrium
Marx’s analysis of competition and technical change. The analysis shows construction, typical of the NGMs based on agents’ maximization, which as
that in Marx, competition as a process is identified as the capitalistic mode of argued by Nelson and Winter (1974) does not lend itself to deal with firms’
Competition and technical change in Aghion and Howitt 271
differences in technology, profitability, knowledge and luck. In equilibrium

no firm modifies its strategy.
In conclusion, it may be of interest to know that some of the best
contemporary literature on growth develops in a rigorous and elegant way
many ideas that loom large in Marx. The presence of many of Marx’s ideas
in Aghion and Howitt’s model proves the sharpness and modernity of
14. Division of labour and economic
Marx’s vision of the capitalistic society. Although Aghion and Howitt’s growth: Paul Romer’s contribution
formalisation can offer a clear view of some of Marx’s arguments, the two
authors do not tackle other crucial features present in his work. Marx in an historical perspective
observes history and proposes his explanation of the development of the
capitalist society, the multiform aspects of which cannot be bridled by the Andrea Lavezzi
general equilibrium approach. His analysis can still be a great source of
stimuli for the growth theorist, who should not abandon formalisation but
look for a more suitable analytical apparatus, which as suggested by Nelson
(1998), should be more attentive to Knightian uncertainty and firm
14.1. INTRODUCTION
organisation. The issue of economic growth regained importance in the economic
profession with the development of the endogenous growth theory (EGT).
Among the plethora of models that followed the seminal papers of Romer
NOTES (1986b) and Lucas (1988), an old hypothesis re-emerged: that economic
growth depends on increasing returns generated by the division of labour. As
1. Romer (1994) defines as neo-Schumpeterian all the New Growth Models that abandon the is well known, this was one of the main contributions of Adam Smith’s
hypothesis of perfect competition and whose discoveries are the outcome of a monopoly
Wealth of Nations, published in 1776. The resurgence of interest in economic
profit-seeking activity.
2. On competition as a tendency toward a predicted result, already present in Smith, see growth motivates this article, which puts in an historical perspective the
McNulty (1967 and 1968). modern approach to the division of labour and growth, represented by Romer
3. Duménil and Lévi (1985) demonstrate that, in the intertemporal equilibrium model with (1987). In particular we evaluate the modern approach from the perspective
infinite-time horizon, Walrasian prices are the same as production prices only in an of Allyn Young’s contribution, who developed the classical theory of Smith.
asymptotic position. Production prices are the prices depending on technology and
We are aware that, from a strictly methodological point of view, it may
distribution and independent of initial endowments and utility functions, i.e. of supply and
demand. This is in line with what Kurz and Salvadori ((1998b and 1999) argue on the not be completely legitimate to compare a ‘classical’ theory to a recent
equivalence between NGT and the classical authors in the long run. formal model. Indeed, the former was expounded only verbally, in particular
4. The implication of this model is that there exists a value of the growth rate beyond which an without the availability of modern mathematical tools. However, we attempt
increase in research, at the expense of secondary innovations, jeopardises growth. to identify the main characteristics of both approaches, which we argue
5. To shift from the production of an old vintage good to a new one is upgrading because the
correspond to quite different views of growth dynamics. In this respect, we
present value of the wage received by a worker who produces a good of vintage τ is a
constant value multiplied by the general state of knowledge. assert that the modern theory does not completely capture the insights found
6. The implication of this extension is that when firms do not internalise learning by doing it is in the classical theory, and that therefore the latter may still contribute to the
impossible for the level of research to be too high. development of a modern theory of economic growth.
7. This effect, by means of time discounting, dominates the reduction in the profitability of A companion paper (Lavezzi, 2002) reconstructs the Smith-Young (and
research due to the consequent increase in the rate at which a worker will quit the newly
Marshall) theory showing that its main features are:
discovered line of production.
8. Aghion and Howitt consider the positive relation between growth and adaptability the result
of an increase in the amount of research labour, and not, as in Lucas (1993), the result of 1. economic growth is endogenous: competition among profit-seeking
enhancing learning by doing. entrepreneurs pushes them to continually reorganize their productive
activities. This happens through two forms of division of labour: within
and among firms. The former is characterized by the introduction of
272
Division of labour and economic growth 273 274 The Theory of Economic Growth: a ‘Classical’ Perspective
organizational and technological changes, the latter by a change in the allocation of given resources; the second is that growth depends on an
structure of the economy. These processes generate an endogenous endogenous process of accumulation.
increase in the network of interdependent economic units, allowing for In particular, competition conflicts with endogenous growth in the
the exploitation of productivity gains due to specialization. Economic framework of the neoclassical model of Solow (1956). In the Solow model,
growth is the aggregate result. Capital accumulation fuels this process output is produced by means of capital and labour, under constant returns to
and is inextricable from technological progress and population growth. scale. In competitive markets factors receive in equilibrium the value of their
2. Growth has the nature of a cumulative, path-dependent process. Current marginal product and, by Euler’s theorem, the product is exhausted by the
opportunities for increasing productivity in the economy are determined remuneration of the factors of production. Hence, nothing can compensate
by the degree of division of labour attained in the past. for technological progress, the engine of long-run growth, which must
3. Economic growth has the characteristics of a disequilibrium process, in necessarily be exogenous. Following Romer, since economists first
which the data of the system, e.g. the number of commodities and the developed the economic models of perfect competition for their
technologies available, are continuously changing under the pressure of mathematical simplicity, they were precluded from studying growth as an
competition. The economy therefore appears as an inherently unstable endogenous process.
system, since qualitative changes constantly take place, activating The next step in Romer’s reconstruction is Marshall’s definition of
productive and technological feedbacks. In this context, increasing external economies: they derive from the aggregate production of a
returns are to be understood as ‘generalized’ or ‘macroeconomic’, being commodity, and accrue to individual firms which remain price takers. Romer
related to the size of the network of specialized, interdependent units. (1991, p. 87) remarks that these economies do not affect much of the
The growth of this network characterizes economic growth. neoclassical apparatus, but contribute to a better understanding of growth.
Then Romer (1991, p. 88) introduces Young’s contribution, whose roots
The latter two aspects make the classical approach essentially different from ‘go back to Marshall and even Smith’. He argues (ibidem) that Young: ‘[l]ike
the modern one. Also, we maintain that the latter differs from the former in Marshall, ..., called the beneficial effects arising from the introduction of a
the following aspects: (i) in the bias towards the supply side and the partial new good a positive external effect. Consequently, he tried to describe a
neglect of demand; (ii) in an excessively important role attached to fixed model of growth driven by aggregate increasing returns that were external to
costs; (iii) in the vague use of the term ‘external economies’, which may be individual firms’. Also, according to Romer (1989, p. 108) ‘Marshall and
instead replaced by ‘network externalities’. Young choose to describe specialization in terms of competitive equilibrium
The rest of the paper is organized as follows: Section 14.2 introduces the with externalities’. Romer specifies that the introduction of new goods is not
Romer (1987) model; Section 14.3 critically compares the old and new strictly equivalent to a Marshallian external economy (like ‘trade
approaches; Section 14.4 contains some concluding remarks. knowledge’), but its consideration can lead to models which behave exactly
like models with true externalities. Therefore, when the focus is on the
introduction of new goods, fixed costs enter the picture because it is
14.2. GROWTH AND SPECIALIZATION IN THE EGT: reasonable to assume the presence of a fixed cost when a new production is
THE ROMER MODEL started. For Romer (1989, p. 108), Marshall and Young’s story would be told
in a ‘more rigorous way in a model with fixed costs’. Such fixed costs may
In this section we analyse the model by Romer (1987), which represents the be interpreted as the costs necessary for the introduction of the commodity
formalization of growth and division of labour in the EGT. This model (e.g. research costs).1 Once incurred, they do not contribute to the marginal
explicitly refers to Allyn Young (see Romer 1989, p. 108), but a reference to cost of producing the good. Since in a competitive equilibrium the firm
Young cannot exclude Smith. For this reason we consider the Romer model equates marginal cost to price, and therefore would be unable to recover the
especially in the light of Young’s contribution, but Smith is not overlooked. fixed cost, what is needed is a departure from the competitive framework.
Let us consider first the historical roots of the model, as presented by Romer resorts to monopolistic competition, where (new) goods are
Romer (1989 and 1991). In particular, Romer (1991) outlines a history of differentiated but competition exists from potential producers, firms have
growth theory, from Adam Smith to the EGT. He claims that in Smith two market power but earn zero profits in equilibrium. Once this is admitted, the
conflicting ideas coexist: the first is that competition ensures an efficient delay in the exploitation of Smith and Young’s ideas is explained by the
technical difficulties involved in building dynamic, general equilibrium maximizes a discounted flow of utility.6 That is, consumers’ savings are
macro-models with non-competitive sectors. invested in Z and rented to the intermediate goods sector. Consumers also
Therefore, it seems that the main obstacles for the exploitation of certain inelastically supply a fixed amount of labour to the final good sector.
‘old’ insights relate to mathematical difficulties. In particular, they depend on Romer specifies a particular form of the function g(.) and of the cost
the requirement to represent growth dynamics as a general equilibrium function for the intermediate goods producers.7 Then he shows that: (i) the
process.2 However, we argue that the interpretation of Smith and, especially, quantity of the intermediate goods produced in equilibrium is x = 1 ; (ii) the
Young as advocates of the equilibrium approach may be challenged. equilibrium condition for the monopolistically competitive sector is:
Let us briefly present the building blocks of the Romer (1987) model. The M ( t) = Z ( t) ; (iii) the following equation:
economy has two sectors: a final good sector and an intermediate goods
. . .
sector. Only the final good is consumed. Intermediate goods are produced Y c Z 1− ρ
with the same technology using a capital good Z, owned by consumers in a = = = (3)
Y c Z σ
given quantity. Production of intermediate goods entails a quasi-fixed cost,
that is, no production at zero costs is feasible. The final good is produced represents the solution for the consumer problem. Here σ is the reciprocal of
under constant returns to scale, using intermediate goods and labour. In the the elasticity of intertemporal substitution and ρ is the intertemporal discount
intermediate sector a regime of monopolistic competition prevails: firms rate.
have market power but earn zero profits in equilibrium. When σ = 1, Romer obtains that the consumption level in equilibrium is
What is relevant is the functional form describing final good production, c ( t) = ( G + ρ ) Z ( t) so that an increase in impatience leads to an increase in
which must be such that ‘having more available [intermediate] goods is the level of consumption, a decrease in the level of savings and a reduction
useful’ (Romer, 1989, p. 108). This can be obtained with the following of the long-run growth rate.8 Finally, Romer (1987, pp. 61–2) points out that
production function:3 ‘this model is not one with a true positive externality, but it nonetheless
behaves exactly as if one were present ... the economy will behave as if there
Y = L1−α ∫ x(i)α di (1) is a form of exogenous, labour augmenting technological change’. He shows
ℜ+
that equation (2) can be rewritten as:
Here Y is the final good, L is labour, x(i) is the quantity of the good i, and 1–α 1–α α α
0<α<1. Thus the marginal product of each intermediate good is decreasing. 4 Y ( t) = M ( t) ( L ( t) N( t) ) = A Z ( t) L ( t) 1 – (4)
If all goods are produced in the same quantity x = N / M (which is the case where the constant A collects all the other constants.
here because of the symmetry of the model), where N is the total amount of In equation (4) the production function for aggregate output, although
intermediate goods and M is the range of goods actually produced, the postulated as a constant returns to scale function, actually appears as if an
production function becomes: external effect were present. Normalizing L to 1 returns the form of the
α α α familiar AK function which can be considered as the base for: ‘the simplest
Y=L1– N M1– (2)
endogenous growth model’ (Barro and Sala-i-Martin, 1995, p. 38).
In equation (2) output can increase without bound with M, given N and L. The next section analyses the elements which inspired Romer’s formal
Aggregate production appears as if increasing returns were present, which model, along with a critical assessment of its relation with the Smith–Young
is not the case if one considers equation (1). The range of intermediate inputs theory of the division of labour and growth.
is theoretically infinite but the assumption of a fixed cost in terms of Z,
whose quantity is given, guarantees that it is finite. The integration of a
power function in equation (1) is the specific form in which intermediate 14.3. EVALUATION OF THE ROMER MODEL
goods are assembled in this model for production of the final good.5
To generate dynamics in this model, it is necessary to establish a In this section we discuss the Romer model in the perspective of the Smith–
mechanism which supports a growing M. This is obtained by assuming that Z Young theory. We argue that the Romer model is a successful attempt to
is accumulated from forgone consumption by a representative individual who bring some relevant insights of such theory in the framework of EGT
models, but at the same time it differs from the ‘old’ perspective in important economies: ‘which manifest themselves in increasing returns are the
respects. economies of capitalistic or roundabout methods of production’. Young
It is fair to say that Romer himself is often cautious as to his simplifying focuses on one aspect of the general process of the division of labour, that is,
hypotheses, but it seems that some of his claims cannot be safely taken for the introduction of specialized machinery when labour has reached a certain
granted, in particular when he refers to Allyn Young. Romer follows Young degree of simplification, and the economies which stem from this process.10
with respect to a possible formalization of the way in which an increase in In particular, such economies depend on ‘large production’ and not on
the ‘roundaboutness’ in production can increase the growth rate. This ‘large-scale production’. Attention should therefore be placed neither on
produces a sort of ‘macroeconomic’ increasing returns, according to the individual firms, and on their negatively sloped cost curves, nor on
hypothesis on the imperfect substitutability of intermediate goods. However, individual industries. Young introduces what may be termed
it appears that this is done in a different perspective from Young’s. ‘macroeconomic increasing returns’,11 which can be understood from an
In particular we mostly concentrate on a comparison between Romer and analysis of the entire economy, considered as a large interactive system. The
Young, it being understood that the latter’s contribution is related to Smith’s. market is in fact defined by Young (Ibidem, p. 533) as: ‘an aggregate of
We organize the discussion around four points: productive activities, tied together by trade’.
Young (Ibidem, p. 531 and 533) takes an ‘inclusive view [of the market,
1) Romer chooses an equilibrium approach against the disequilibrium which is not] an outlet for the products of a particular industry, and therefore
approach of Young (and Smith). This is connected to the view of growth external to [an] industry, but [i]s the outlet for goods in general. [Therefore:]
as a path-dependent process. the size of the market is determined and defined by the volume of
2) The Romer model is essentially supply-oriented and demand does not production’. This immediately leads him to this reformulation of Smith’s
play an essential role as in Young (and Smith). theory: the division of labour is limited by the division of labour. Although
3) The emphasis on fixed costs is different. reminiscent of Say’s law, this argument is more far-reaching: 12 it asserts that
4) Young was more cautious than Romer on the use of the concept of both demand and supply are endogenously determined according to the
Marshallian external economies. degree of division of labour prevailing.13 This amounts to recognizing that
the extent of the market is at least partially endogenous.
14.3.1. Equilibrium or Disequilibrium? So, we arrive at the important implication that
First of all, the Romer model is cast into an intertemporal equilibrium setting, the counter forces which are continually defeating the forces which make for
while a reading of Young (1928) highlights that he strongly rejected the economic equilibrium are more pervasive and more deeply rooted in the
constitution of the modern economic system than we commonly realise. Not only
equilibrium approach to study endogenous economic growth. Young seemed
new or adventitious elements, coming in from the outside, but elements which are
on the contrary to advocate a disequilibrium theory of endogenous growth.9
permanent characteristics of the ways in which goods are produced make
Young (Ibidem, p. 528), discusses the Marshallian dichotomy on internal continuously for change. Every important advance in the organisation of
and external economies, arguing that the economies of a firm depend on production ... alters the conditions of industrial activity and initiates responses
what happens in an 'obscurer field' where: 'new products are appearing, firms elsewhere in the industrial structure which in turn have a further unsettling effect.
are assuming new tasks, and new industries are coming into being'. Hence This change becomes progressive and propagates itself in a cumulative way
(Ibidem, p. 531. Italics added).
No analysis of the forces making for economic equilibrium, forces which we might
say are tangential at any moment of time, will serve to illuminate this field, for Thus for Young, not only is economic growth endogenous, but also
movements away from equilibrium, departures from previous trends are
endogenous forces generate disequilibrium, in the sense that, in the growth
characteristics of it. Not much is to be gained by probing into it to see how
process, the structure of the economy and the technological opportunities
increasing returns show themselves in the costs of individual firms and in the
prices at which they offer their products (italics added). cannot a priori be considered fixed. Young (Ibidem, p. 533) the apparatus of
supply and demand is incapable of exploring this sort of dynamics, since it
This view on disequilibrium is related to the very specific idea of Young on may ‘divert attention to incidental or partial aspects of a process which ought
the functioning of a market economy. First, for Young (Ibidem, p. 531) the to be seen as a whole’.14 The use of an equilibrium approach to study growth
seems even to imply the impossibility of defining growth as an endogenous accumulation of capital, and cannot proceed unless some previous stage of
process (Ibidem, p. 535). division of labour has been reached. Therefore, growth is characterized by
Young (Ibidem, p. 534) then introduces the concept of reciprocal demand. path dependency: current opportunities to increase productivity by further
Reciprocal demands among firms are characterized by a certain level of subdivisions of labour, depend on the past stages of the division of labour.16
elasticity, to be interpreted as the capacity for the increased production of a Is there path dependency in the Romer model? The answer is in the
good to generate demand for other goods. affirmative in the sense that the model predicts lack of convergence, since
The elasticities are different for different products, so growth in the economies starting poor because they do not specialize stay poorer than
economy will be different among sectors. In any case ‘[e]ven with a economies that start richer, because they produce a larger range of
stationary population and in the absence of new discoveries in pure or intermediate goods. However, this type of model does not feature path
applied science there are no limits to the process of expansion except the dependency in the strict sense, since cumulative effects are absent: if the
limits beyond which demand is not elastic and returns do not increase’ saving rate, e.g., increases through a decrease in an exogenous factor like the
(Ibidem, p. 534). intertemporal discount rate, the range of intermediate goods increases as well
Differently, in the representation of the productive process, Romer adopts as the growth rate, but this per se does not preclude the saving rate, and the
an equilibrium approach, preserving the ‘one-way avenue’ from given range of intermediate goods, from decreasing in a subsequent period due to a
resources to final output, although by means of an intermediate sector. It is change in the same parameter of the opposite sign.
not clear that this can be taken as a faithful representation of the economy
which Young had in mind. He considered the economy as an ‘interrelated 14.3.2. Supply-side or Demand-side?
whole’, where the extent of the market is endogenous and feedbacks, for
instance in the form of ‘reciprocal demands’, among productive units are In the Romer model consumers save and invest in Z; this increases the range
continuously displacing the tendency towards equilibrium, when this is of intermediate goods which in turn increases production and income. In the
interpreted in an allocative sense.15 intermediate sector there are firms potentially active, but the decision of such
In the Romer model, there are two allocation problems: the first regards firms to produce is not due to a sudden increase in demand for their good. A
the allocation of the given resource Z among the intermediate goods potential demand is always existing for an infinite number of intermediate
producers; the second is the allocation problem of consumers between goods, because of the form of the production function for final goods.
consumption and saving. The first problem is solved imposing the zero profit Potentially active firms can become operative once the available quantity of
condition in the intermediate sector: in this case the results is an equilibrium Z makes it possible; thus it is savings that foster growth.
range of intermediate goods. The second is solved by utility maximization of The increase in specialization, i.e. in the number of intermediate goods,
consumers, given the paths of the rental price for Z and the price for the which is permitted by savings, increases Y, which is income earned by
consumption good. consumers and subsequently consumed or invested. Thus, the growth of Y is
In any case, production is never assumed to take place under increasing constrained by the supply of intermediate goods, in turn constrained by the
returns, due for instance to the continuous re-organization of the production availability of the primary resource Z.
process, to learning by doing, to improvements in the technology, as Let us briefly recall the role of demand in the process of growth based on
emphasized by Smith and Young (see Lavezzi, 2002). In the Romer model the division of labour. It is well known that in Smith and Young the main
increasing returns appear in the aggregate, as for Young, but they are limit to the division of labour, and thus to growth, is the extent of the market.
generated by a series of equilibrium conditions and depend on a particular According to Smith (WN I.i.1), a ‘natural’ predisposition for socio-
hypothesis on the way intermediate inputs are assembled. economic interactions allows individuals to specialize and obtain a gain from
In the Introduction we also argued that in the Smith–Young framework trading their surplus products, i.e. the production in excess of their own
growth based on the division of labour is a path-dependent, cumulative consumption, deriving from higher productivity due to the specialization.
process. This nature of growth emerges from the various quotations from The extent of the market comes into play here: an individual has the
Young, and can be inferred even from Adam Smith. In Smith it is capital incentive to specialize if he possesses ‘power of exchanging’ that surplus, i.e.
accumulation that transforms economic growth based on the division of if sufficient demand exists, allowing the agent to purchase other goods with
labour into a cumulative process: the division of labour is allowed by the the revenue from the disposal of his surplus product.17 Hence, economic
growth may be spurred by the creation of a network even of similar (i.e. not 14.3.4. External Economies or Network Externalities?
specialized) individuals. Once connected, they will sort themselves out in
different occupations, and increase their aggregate production. Thus, in the Romer claims that, as reported in Section 14.2, Young (and Marshall)
Smithian framework, demand and supply are at least on the same level.18 discussed growth and specialization in terms of competitive equilibrium with
Can we find evidence in the Romer model for the principle that ‘the positive external effects. However, this interpretation does not seem to be
division of labour is limited by the extent of the market’? What is certainly correct. From the previous discussion, we observe first that Young firmly
true is that the division of labour, that is the introduction of new intermediate rejected the approach to the study of economic growth based on equilibrium
goods in the production of the final good, is limited by fixed costs. However, of supply and demand.
when discussing the introduction of new, specialized machines (which At the same time he (1928, p. 528) considered the Marshallian distinction
corresponds to the intermediate goods in the Romer framework) as an aspect between internal and external economies, as implying just ‘a partial view’ on
of the process of the division of labour, Young mainly placed emphasis on the growth process. For Young (1928, pp. 527–8) the Marshallian distinction
another question: that is on the possibility of adopting more capital-intensive, is ‘fruitful’, but partially misleading.
highly productive methods, conditional on the possibility of selling a large It may well be that, when considering the positive effect generated outside
output In this case the absence of demand limits the division of labour; a productive unit, in that ‘obscurer field’ as Young called it, a more
Young seemed to be less concerned with resource-constraints faced by firms. appropriate concept than external economy for Young’s theory is that of
network externality. Consider the following definitions from Economides'
14.3.3. What Role Do Fixed Costs Have? Dictionary of Terms in Network Economics:19 ‘Networks: networks are
composed of complementary nodes and links ... Network externality: a
The latter point is also related to the question of the role of fixed costs in network exhibits network externalities when the value of a subscription to
Young’s theory. Young seemed to be aware of them in the discussion of the the network is higher when the network has more subscribers’.
introduction of machines, but did not emphasize their role. Sandilands (2000, In the Smith–Young framework the creation of a network of specialized
p. 315) comments: ‘Young did not say that specialization is limited by the production units (i.e. ‘complementary links’), or the increase in the number
presence of fixed costs, though he did say that specialization increasingly of ‘subscribers’ to an existing one, increases the ‘value of the subscription’
took the form of greater roundaboutness in the economy as whole. In his of the individual participant, in the sense that it may represent an increase of
theory, fixed costs and increased roundaboutness are not so much a its market. The discussion by Young on reciprocal demands implies that the
constraint on growth as its consequence’. network itself is subject to endogenous change, as growth of production may
The issue for Young (and Smith) was the creation of new markets or the lead to creation of new specialized sectors, activating feedbacks elsewhere,
extension of existing ones. Young (ibidem, p. 530) writes in a famous etc. The aggregate results should be endogenous economic growth.20
passage that 'it would be wasteful to make a hammer to drive a single nail' or
'to furnish a factory with an elaborate equipment of specially constructed
[machines]' if the market is not sufficiently large. According to this idea, 14.4. CONCLUDING REMARKS
noted also by Kaldor (1972, p. 1242), the capital–labour ratio chosen by
firms depends on the extent of the market and not only on relative factor Summing up the arguments put forward in the previous section, we believe
prices. Hence, the importance of the extent of the market for the division of that the Romer model did successfully capture some features of the growth
labour implies a role for the extent of the market in the choice of the optimal dynamics generated by the division of labour. However, our re-evaluation of
capital–labour ratio. the contributions of Allyn Young and Adam Smith indicates that some
At any rate, in the context of the whole of Young’s theory, his discussion aspects are in need of different formalizations.21
on the introduction of new specialized machines seems more relevant for its Both the old and new approach recognize the endogenous nature of
‘dynamic’ implications. Namely, for the choice of the firm’s internal economic growth. Both relate growth to an increasing complexity of the
organization and technology, and for the feedbacks on the environment, for economy, and introduce a sort of ‘macroeconomic’ increasing returns. The
instance for capital–goods industries. main differences relate to the view of growth based on division of labour as
an equilibrium or path-dependent, disequilibrium process. Other differences
regard the role of demand, of fixed costs and the suitable notion of external 8. This equilibrium growth rate is suboptimal: a policy which raises savings would positively
effect. We therefore conclude that the classical theory of division of labour affect long-run growth.
and growth may still provide insights for the development of a modern 9. In Lavezzi (2002) we argue that also the original theory of Adam Smith can be interpreted in
this way. See also Richardson (1975).
theory of economic growth.
10. Young also treats the division of labour among industries.
11. This definition appears in Currie (1997) and is compatible with the one recently proposed by
Buchanan and Yoon (2000, p. 45) with respect to Smith: ‘the Smithean proposition that
NOTES relates the division or specialization of labour to the extent of the market is best captured by
the notion of generalized increasing returns, which implies only that the degree of
specialization utilized increases with the size of the whole nexus of economic interaction
1. See also Romer (1990). thereby increasing the ratio of positively valued outputs to inputs’ (Italics in the original
2. As Romer (1989, p. 70) writes: ‘[g]rowth is a general equilibrium process’. text).
3. Dixit and Stiglitz (1977) introduced this form in a utility function to express preference for 12. Young (1999b, p. 145) in fact criticises Say’s law as such.
variety in consumption. Ethier (1982) proposed to utilize it for a production function. 13. This aspect is assumed in recent models such as Yang (1999), where the agents are
4. The more general form of equation (1) is: producer–consumers, and the structure of demand and supply is simultaneously determined
with the degree of division of labour.
⎛ x (i ) ⎞ di
Y=L ∫ℜ+
g⎜
⎝ L ⎠
⎟ 14. See also (Young, 1999a, p. 45): ‘Seeking for equilibrium conditions under increasing returns
is as good as looking for a mare’s nest. Certainly the matter cannot be explained by this
where it is required that g(.) is increasing and strictly concave, with g(0) = 0. curve apparatus, which does not see things “in their togetherness”’
5. Another functional form to aggregate intermediate goods appears in the literature, the CES 15. Moreover, interdependence among sectors in the Romer model appears in the sense that the
specification. In the CES specification, the production function can be expressed by: final good is produced by means of intermediate goods in one period, and becomes a factor
of production for them in the following period if not consumed. Again, this does not seem to
α be the story told by Young on the reciprocal effects triggered by increases in supply, which
⎧ 1
⎫
⎨⎡⎣ ∫ x (i) di ⎤⎦ θ ⎬
1−α θ
Y =L stimulate increases in demand, which in turn become increases in supply by other firms, etc.
⎩ ⎭ 16. Kaldor (1972, p. 1245, and 1975, p. 355) draws strong implications from the recognition of
this aspect, denying the possibility of defining growth as an equilibrium process.
where θ is a parameter reflecting the elasticity of substitution among different intermediate 17. Clearly, the same logic applies to a firm: specialization of operations is profitable if there is
inputs, given by ε =1/(1 – θ). When 0 < θ < 1, goods are imperfect substitutes (i.e. demand for the higher quantity of goods that the firm can produce by specializing.
1 < ε < ∞); when θ = 1 goods are instead perfect substitutes. Note that the formulation in 18. For further discussion of demand in the process of the division of labour see Lavezzi (2002).
equation (1) is simply obtained by putting θ = α. In the CES case, when all intermediate 19. Available at: http://www.stern.nyu.edu/networks/dictionary.html.
goods are produced in the same quantity x = N / M , we obtain: 20. In the present case, the network effects should be understood in a particular sense, with
1–α respect to the existing literature. In fact, we know that the choice to specialize in the
NαMα
.(1–θ)/θ
Y=L .
production of a particular good depends on others' choice to specialize in different goods.
Thus output can increase without bounds in M as long as 0 < θ < 1. A general form for this Thus we can call the benefit from specialization conformity effects (in the sense that
type of production function is the following: different agents make the choice of specialization), keeping in mind that the specialization is
in different activities. Typically, conformity effects are such that the benefits of an action,
γ τ for example the adoption of a technological standard (see, e.g., Arthur, 1987), increase in the
Y =L
1−α
⎡⎣ ∫ x (i ) di ⎤⎦ ⎡⎣ ∫ x(i) di ⎤⎦
θ δ
number of those making the same choice. In this sense the positive feedback effect (see
Agliardi, 1998) reinforcing the outcome of agents’ choices should be understood
This formulation preserves the homogeneity of degree α in x(i) and the positive relation correspondingly: the choice of agents to specialize is mutually reinforcing, but they are
between M and Y, if the following conditions are satisfied: involved in different activities, and therefore, e.g., adopt different technologies.
i) γ θ + τ δ = α , 0 < α , θ < 1 ; 21. For instance, the comment provided by Heal (1999, p. xxiii) supports this claim. Heal, after
ii) γ + τ > α . presenting the main features of Young‘s growth theory, writes: ‘[t]his seems an interesting
intuition, broadly consistent with casual empiricism, and not captured by any formal growth
The form chosen by Romer corresponds to the case in which: θ =α , γ = 1, τ = 0. The case of
models. It has some resemblance to evolutionary models in biology, where evolution leads
a CES specification is obtained when: γ = α /θ, τ = 0.
to increasing complexity and longer food chains’ (Italics added). Moreover, from an
6. A conventional isoelastic utility function U(c) is considered.
historical point of view, the Romer model of growth and specialization has been recently
7. In particular, g(.) is strictly concave on the interval [0,x0], and has a constant slope equal to 1
criticized by Sandilands (2000, p. 315), for not being able to: ‘fully capture Young’s view of
on the interval [x0, ∞]. In addition: g(0) = 0 and g'(x0) = 1. The intercept on the vertical axis
the links between fixed costs, specialization, external economies, and the economy-wide
obtained by prolonging the slope equal to 1, is indicated by G. The cost function is
2 external returns that make growth a semi-automatic, self-perpetuating process’.
h(x)=(1+x )/2.
profitable avenues such that more entrepreneurs are induced to innovate,

giving rise to an investment boom and driving growth for the economy as a
whole. Once the innovation is fully exploited, however, the economy
relapses into a depression which lasts until the accumulation of new ideas
creates a favourable climate for a new burst of innovating investment and so
15. The interaction between growth on.
and cycle in macrodynamic models Both Marx’s and Schumpeter’s theories, however, are descriptive rather
than analytical. Thus, the question of how to incorporate the Marxian
of the economy ‘distributional’ mechanism, or the Schumpeterian ‘innovational’ mechanism
or a combination of the two in a formalized dynamic model of the economy
Serena Sordi does not have a single, straightforward answer.1
Keeping this in mind, in the present paper we intend to review, with
regard to their capacity to represent the cyclical growth of the economy,
some early contributions to non-market clearing macrodynamics.2 This will
15.1. INTRODUCTION be done by analysing two prototype models. In Section 15.2, the simple
linear Keynesian (‘multiplier–accelerator’) model is presented and analysed
Since the beginning of the study of the dynamics of industrial capitalism,
with respect to its cyclical growth properties. Two results easily follow from
economic growth and business cycles have been seen as indissolubly linked.
a qualitative analysis of this model: first, there cannot be any interaction
In this regard, it is important to remember that Marx (1954, ch. 25) – ‘the
between the cyclical and the growth component of the solution; second, even
first economist seriously to study the cyclical aspect of capitalism’
as a ‘pure’ cyclical model, it is not at all satisfactory given that the
(Goodwin, 1986, p. 15) – considered the business cycle as nothing other than
representation of persistent fluctuations can only be obtained for a very
the basic way in which capitalist economies develop, due to the interaction
special combination of parameter values. Our purpose is to show that a non-
between the accumulation process (and the resulting growth of productive
linear version of the model, in addition to solving the second of the two
capacity) and the conflict over income distribution between capitalists and
problems, provides some interesting hints also with regard to the first, once
workers. Schumpeter (e.g., 1939), in turn, produced an integrated theory of
we assign some role to ‘innovational’ investment in the Schumpeterian
growth and business cycles in which economic fluctuations are nothing other
sense.
than the ‘form which progress takes in capitalist society’ (Schumpeter, 1927,
In Section 15.3 we outline Goodwin’s growth cycle model which provides
p. 295).
a formalisation of the Marxian reserve army mechanism described above.
As is well known, a crucial role in Marx’s (endogenous) explanation of
The original version of this model is known to suffer from two weaknesses.
the mutual conditioning of growth and cycle is played by the size of the
First, it is structurally unstable, in other words, even the smallest perturbation
reserve army of labour. In short, the story runs as follows (see Marx, 1954,
of its structure can destroy the closed orbit (cyclical) character of its solution;
p. 597). In periods of high rates of accumulation, the reserve army of labour
second, it neglects altogether any disequilibrium in the product market. To
shrinks, thus leading to an increase in labour’s bargaining power. This, in
conclude this section, we briefly discuss a modified version of the model
turn, causes a change in income distribution in favour of workers, entailing a
which, in relaxing Goodwin’s extreme (‘classical’) assumption about
decline of profits and a consequent decline of accumulation. But this leads to
savings, is shown to produce limit cycle solutions, thereby overcoming the
an increase in unemployment and pushes the wage share down. As a
first of the two weaknesses. In the attempt to overcome also the second, an
consequence, the profitability of real investment is restored, the rate of
integration of the two prototype models is then sketched in Section 15.4.
capital accumulation goes up and the sequential mechanism just described
Finally, Section 15.5 contains some conclusions and suggestions for further
can start again. The crucial role in Schumpeter’s explanation, on the other
research.
hand, is played by the concept of the ‘pioneering entrepreneur’ who carries
out an innovation, whether this is a new method of production or
transportation, a change in industrial organisation, a new product, the
opening up of a new market or a new source of materials. This opens up new
285
The interaction between growth and cycle in macrodynamic models 287 288 The Theory of Economic Growth: a ‘Classical’ Perspective
15.2. THE KEYNESIAN ‘MULTIPLIER–ACCELERATOR’ and the accelerator principle ( vY ) respectively, where v is the (constant)
MODEL desired capital–output ratio. In writing (3) and (4), we have assumed that
both error-adjustment mechanisms involve a simple exponential lag, of
15.2.1. A Linear Formulation of the Model lengths equal to ε and θ respectively (see Allen, 1967, pp. 76–9 and 94–5).
From (3) and (4), given (1) and (2), we obtain the following second-order
In most of the earlier macrodynamic models which followed the publication differential equation, the solution of which describes the dynamics of the
of Keynes’ General Theory (1936), the prevalent attitude was that of a national income
separate handling of business cycles and growth. These were typically linear
1 s 1
models based on purely endogenous relationships, aimed at explaining the Y + (ε + s − v )Y + Y = ⎡⎣ Ia (t ) + I a (t )⎤⎦ (5)
aggregate behaviour of consumers (through the multiplier) on the one hand ε ε ε
and that of entrepreneurs (through some version of the principle of the where s = 1 − c and where we have chosen the time unit so as to have θ = 1.
accelerator or some other theory of aggregate investment) on the other. The In the case in which Ia(t) = 0 for all t, (5) is a homogeneous differential
dynamics resulting from such an interaction between the multiplier and the equation. By a simple analysis of the roots of its characteristic equation, we
accelerator is either cyclical or monotonic and does not succeed therefore in can then conclude that the relation between the values of the parameters of
replicating the observed persistent cyclical growth of real economies. The equation (5) and the dynamics of the model is the one summarized in Table
only way out of this puzzle is to assume that the parameters of the model are 15.1 and shown graphically, for two different cases, in Figure 15.1.5 As
such that the solution is cyclical (with fluctuations of constant amplitude) anticipated, the resulting dynamics is either monotonic (for combinations of
and then to add to the model an autonomous component of aggregate the parameter values in regions A or E) or oscillating (for combinations of
expenditure, which grows exogenously through time. In this case, the the parameter values in regions B, C, or D).
‘multiplier–accelerator’ interaction implies constant amplitude fluctuations
of output around a growing trend, but, by construction, there cannot be any Table 15.1 – Intervals of parameter values and type of solution
interaction between the growth and the cyclical components of the
dynamics.3 Intervals of values of v Type of solution
All this can be easily illustrated by using the following (prototype) linear v ≤ ( ε − s )2 monotonic, convergent (A)
model of the ‘multiplier–accelerator’ interaction:4
( ε − s) < v < ε + s
2
oscillating, damped (B)
C = cY , 0 < c < 1 (1) v =ε +s oscillating, constant amplitude (C)
I = I i + I a (t ) ε + s < v < ( ε + s) 2
(2) oscillating, divergent (D)
v ≥ ( ε + s) 2
monotonic, divergent (E)
1
Y = ⎡⎣(C + I ) − Y ⎤⎦ , ε > 0 (3)
ε
θ
1
θ
{ }
Ii = ( vY − Ii ) = vY − ⎡⎣ I − I a (t )⎤⎦ , v ,θ > 0
1
(4)
Equation (1) is a standard Keynesian consumption function, where C

denotes aggregate consumption and Y the national income. Equation (2), on
the other hand, simply states that total investment (I) is the sum of an
induced component (Ii) and an autonomous component (Ia (t)). The basic
dynamic ingredients of the model are the two error-adjustment mechanisms
(3) and (4), according to which the aggregate supply and the induced
investment adjust to their desired levels, determined by total demand (C + I) Figure 15.1: Regions of parameter values and type of solution
On the other hand, if we assume that there is an autonomous component 15.2.2. A Non-linear Formulation of the Model with ‘Innovational’
of investment which increases over time, for example, such that Investment
I a (t ) = a0 + a1t, a0 , a1 > 0 The previous analysis demonstrates that the simple ‘multiplier–accelerator’
model cannot produce both growth and cycle: in the case considered, for
from (5) we obtain the following non-homogeneous differential equation example, it proves to be a pure cyclical model, such that the representation of
cyclical growth can only be achieved by superimposing a trend on it.8 On the
1 s 1 basis of what was stressed in Section 15.1, however, this is hardly surprising
Y + (ε + s − v )Y + Y = ( a2 + a1t ) (6) given that in such a model (i) only the product market is formalised, whereas
ε ε ε the labour market is neglected altogether and (ii) any kind of ‘innovational’
investment in the Schumpeterian sense is ignored. Given (i), it appears that
where a2 = a0 + a1. In this case, choosing a combination of values for the the consideration of the conflict-over-distribution mechanism would
parameters of equation (6) along the straight line C of Figure 15.1, we obtain certainly require a complete reformulation of the model and we will not
the representation of (constant amplitude) fluctuations around a growing attempt such a task here. Thus, it remains to be seen whether it is possible to
trend.6 incorporate into the model the other − ‘Schumpeterian’ − mechanism.
By construction, however, there cannot be any interaction between the In order to do that, let us consider a non-linear version of the model (see
cycle component (Figure 15.2(i)) and the growth component (Figure 15.2(ii)) Goodwin, 1951), which at least enables us to represent persistent fluctuations
of the dynamics in that they are simply superimposed (Figure 15.2(iii)).7 (limit cycles) for a larger range of parameter values. This is easily done by
replacing the desired level of investment vY in the error-adjustment
mechanism (4) with the non-linear function φ (Y ) , where φ is as shown in
Figure 15.3(i). The rationale for the ‘sigmoid’ shape is the fact that
investment in fixed capital has an upper limit ( Iimax ), given by the maximum
rate of investment allowed by existing productive capacity, and a lower limit
( − Iimin ), corresponding to zero gross investment. Thus, it is plausible to
assume that the acceleration principle determines the desired level of the
capital stock only over some middle range, but passes to complete
inflexibility at either extreme.9 Equation (5) becomes
s
Y + ⎡⎣(ε + s )Y − φ (Y )⎤⎦ + Y = ⎡⎣ Ia (t ) + I a (t )⎤⎦
1 1
(7)
ε ε ε
From (7), assuming (as in Goodwin, 1951, p. 12) that autonomous

investment is constant and equal to I a* for all t, we obtain
s
Yc + ⎡⎣( ε + s )Y c − φ (Y c )⎤⎦ + Y c = 0
1
(8)
ε ε
where Y c = Y − I a* / s . As is well known (see, for example, Gandolfo, 1997,

Figure 15.2: Dynamics of the linear multiplier–accelerator model with a pp. 440–1), in the case in which v > ε + s (i.e., the equilibrium is locally
linear trend and oscillations of constant amplitude (for ε = 2, s = 0.25, unstable) the solution of equation (8) is a limit cycle, describing persistent
v = 2.25 , a0 = 90, a1 = 10) fluctuations of national income around the (constant) equilibrium I a* / s (see
Figure 15.4).
To assess the importance of this result, however, it is useful to go back to the

meaning of autonomous – as opposite to induced – investment. In the model
we are considering, such a component includes all that investment in fixed
capital that is not explained by the acceleration principle, mainly, therefore,
innovational investment in the Schumpeterian sense. Thus, it is not at all
satisfactory to assume that it is constant over time. Rather, a better, although
rough, way of formalizing Schumpeter’s idea of clustering of innovations is
to assume, as suggested by Goodwin (1946, p. 97), that such a component of
investment is a periodic function of time of the kind
I a (t ) = b ⎡⎣1 + sin (ct )⎤⎦ (9)
where the two parameters b and c determine the amplitude and the frequency
of innovational investment respectively (see Figure 15.3(ii)).
An example of the simulation of the model with the ‘forcing’ effect of
innovational investment (9) is given in Figure 15.5. As we see, the result is
now that the autonomous (‘innovational’) investment interacts with the
cyclical dynamics of the model in such a way as to generate cyclical growth,
Figure 15.3: (i) The induced and (ii) the autonomous component of although, for the parameter values chosen, the latter is only transient.10
investment
Figure 15.5: Dynamics of the non-linear accelerator model with

Figure 15.4: The limit cycle of the non-linear accelerator model (for ε = 2, s innovational (‘forcing‘) investment (for ε = 2, s = 0.25, v = 3 , b = 20,
= 0.25, v = 3, I a* = 100) c = 3 .5 )
15.3. THE ‘MARXIAN’ GROWTH CYCLE MODEL
Since the late Sixties, a different approach to growth cycles has been
developed, based on Goodwin’s (1967) growth cycle model.11 We call this
approach ‘Marxian’, in order to stress that in it the crucial role is played by
the Marxian reserve-army-of-labour mechanism we described in the
introduction. Taking account of the vast literature that has appeared since the
publication of the original contribution (hereafter, OVM = Original Version
of the Model), in what follows we propose and analyse two generalisations
of the OVM.
Before doing that, let us briefly recall that the OVM gives rise to the
following dynamical system of the Lotka–Volterra type
⎛1 1 ⎞
E = ( g − gn ) E = ⎜ − gn − U ⎟ E (10)
⎝v v ⎠
U = ⎡⎣ f ( E ) − α ⎤⎦ U ≈ ⎡⎣ − (γ + α ) + ρ E ⎤⎦ U (11)
Figure 15.6: Growth cycles of the OVM
where, apart from the notation already introduced in the previous section, L
stands for employment, A = Y/L, labour productivity, N, the labour force, gn = It is obvious that, in this case, the capacity to generate growth cycles is
α + β, the natural rate of growth, W, the real wage, U = WL/Y = W/A, the intrinsic to the model and is not due to the choice of any particular form for
share of wages, E = L/N, the employment rate, S, total savings, and g, the rate the functions of the model. We believe, however, that two aspects reduce the
of growth of output, and where it is assumed that importance of this result. First, the fact that the positive equilibrium point of
the model is a centre implies that the fluctuations of E and U around their
A = A0 exp (α t ) equilibrium values are of an amplitude which fully depends on initial
conditions. In the existing literature on the topic, however, one can find a
number of contributions in which the OVM is modified in such a way as to
N = N 0 exp ( β t )
generate limit cycle dynamics of the relevant variables. We present and
discuss one of these extensions of the OVM in Section 15.3.1. Second, and
Equations (10) and (11) easily follow from these basic assumptions more importantly, the OVM neglects altogether any effective-demand
together with the assumption of a Phillips curve for the real wage dynamics considerations. There cannot be, therefore, any role, in the generation of the
growth cycles, for adjustments to product market disequilibrium. This is in
Wˆ = f ( E ) ≈ −γ + ρ E, γ , ρ > 0 sharp contrast with the Keynesian model presented in Section 15.2. A
‘hybrid’ version of the OVM, which attempts to introduce a Keynesian
and a classical assumption about savings behaviour, according to which all flavour while preserving the capacity of the model to generate growth cycles,
profits are saved and invested and all wages consumed. is outlined in Section 15.4.
As is well known, the solutions in E and U of equations (10)–(11) are
cyclical (the positive equilibrium point being a centre) (see Figures 15.6(i) 15.3.1. A Modified Version of the Model with Differential Savings
and 15.6(ii)) and, given that g = (1 − U ) / v , this implies that the rate of
growth of the economy is also cyclical (Figure 15.6(ii)). Thus, output is In Sordi (2001), the author proposed a modified version of Goodwin’s model
subject to growth cycles as shown in Figure 15.6(iii). in which the ‘distributional’ mechanism plays a role in the dynamics also via
savings behaviour. She studied the case in which both capitalists and workers
save a fraction of their incomes and showed that the model can produce ⎡ Δs s Δs ⎤
persistent oscillations (limit cycles) and even chaotic dynamics if, in X = ⎢ − c U − X (1 − U )⎥ X (15)
⎣ v v v ⎦
introducing such a modification of Goodwin’s classical assumption, (i) we
take account of Pasinetti’s criticism (e.g. 1962) of Kaldor’s (1956) approach with positive singular point (E , U , X ) = (E , 1 − v gn /sc, sc(gn − sw / v )/Δsgn)
* * * *
to differential savings; and (ii) we consider a more general version of the where E is the value of the employment rate for which h(E ) = α.
* *
Phillips curve, according to which the rate of growth of real wages depends It is worth noting that the positive equilibrium guarantees steady-state
not only on the level of the rate of employment, but also on its rate of results that have a Pasinettian–Kaldorian ‘flavour’:
change.
To introduce this first ‘Modified Version of the Model’ (MVM1), with the • it guarantees a steady-state growth of the system at a warranted rate equal
to the natural rate and is such that the Cambridge equation r = (1 −
*
help of some additional notation, we write
U )/ v = gn/sc is satisfied;
*
Sc = sc Pc = sc rKc • in order to be economically meaningful, it requires that the Pasinettian

case holds
Sw = sw (WL + Pw ) = sw (WL + rK w )
0 ≤ sw < vgn < sc ≤ 1 (16)
Y = WL + Pw + Pc
• it is such that the steady-state growth path is characterized by a positive
P Y − WL 1 − U (constant) rate of unemployment equal to (1 − E ) rather than by full
*
r= = =
K K v employment.
sw + Δ s (1 − U ) X However, and more importantly given our purposes, when condition (16) is
g= * * *
v satisfied, the system may not converge to (E , U , X ), but rather persistently
fluctuate around it (see Figure 15.7).13
and12
( )
Wˆ = f1 E, Eˆ = h ( E ) + δ Eˆ , h′ ( E ) > 0, h′′ ( E ) > 0, δ > 0 (12) 15.4. A HYBRID VERSION OF THE GROWTH CYCLE
MODEL
where X = Kc/K is the proportion of capital held by capitalists, Pw, workers’
profits, Pc, capitalists’ profits, P = Pc + Pw, total profits, r = P/K, the rate of The MVM1, apart from the generalized Phillips curve and the assumption
profit, sw and Sw, workers’ propensity to save and savings, sc and Sc, concerning differential savings, maintains all the other simplifying
capitalists’ propensity to save and savings, S = Sw + Sc, total savings, Δs = sc − assumptions of the OVM. In particular, it assumes a permanent product
sw > 0 and where, to simplify, we have assumed that the function f1 in (12) is market equilibrium and does not have an independent investment function. In
additive. the attempt to integrate the two different types of dynamic model considered
It is then not too difficult to show (see Sordi, 2001, p. 101) that the in this chapter (‘Keynesian’ and ‘Marxian’),14 we next propose a second
MVM1 reduces to the following 3D-dynamical system in E, U, and X modified version of the model (MVM2), in which (i) investment in fixed
capital is explained by an ‘accelerator-type’ mechanism;15 and (ii) the
⎡s Δs ⎤
E = ⎢ w − gn + X (1 − U )⎥ E (13) product market does not clear at all times; rather it is governed by a
⎣ v v ⎦ simplified version of the error-adjustment mechanism (3) of the ‘multiplier–
accelerator’ model studied in Section 15.2.
U = ⎡⎣ h ( E ) + δ Eˆ − α ⎤⎦ U (14)
Kw = 0 ↔ X = 1
S = sc (1 − U ) Y
Thus, expressing (19) in terms of rates of growth
gn η
gY = + ⎡ v − v − sc (1 − U )⎤⎦ (20)
1 − η v τ 1 − η vτ ⎣
Taking account of (18) and (20), it is then not too difficult to show that
the MVM2 reduces to the following complete 3D-dynamical system in the
three endogenous variables E, U and v19
⎧ η vτ gn η ⎫
E = ⎨ + ⎡⎣ v − v − sc (1 − U )⎤⎦ ⎬ E (21)
⎩ 1 − η vτ 1 − η vτ ⎭
U = ⎡⎣ h ( E ) + δ Eˆ − α ⎤⎦ U = ⎡⎣ F ( E, U, v ) − α ⎤⎦ U (22)
Figure 15.7: Growth cycles of the MVM1 (for sc = 0.9, sw = 0.0134, v = 2.57 , ⎧ gn η ⎫
α = 0.0221, β = 0.0037, γ = 0.9, ρ = 1, δ = 0.02 and X(0) = 1) v = v − v + (τ v − v ) ⎨ + ⎡⎣ v − v − sc (1 − U )⎤⎦ ⎬ (23)
⎩ 1 − η vτ 1 − η vτ ⎭
First of all, with regard to induced investment,16 we slightly modify (4) by where the function F(E, U, v) in equation (22) is such that FE(E, U, v) = h′(E)
assuming that the desired level of the capital stock Kd is determined by the > 0, FU(E, U, v) = δηsc/(1 − ητ v ), Fv(E, U, v) = −δη/(1 − ητ v ).
flexible accelerator (see Goodwin, 1948), expressed in terms of expected The positive equilibrium point for this version of the model proves to be
output Ye:17 (E , U , v ), where, as in the MVM1, E = E , whereas we now have
** ** ** ** *
I = K = K d − K = vY e − K (17)
gn v** ** (1 + τ gn ) v
U ** = 1 − ,v =
Assuming that Ye is related to realized output according to an sc (1 + gn )
extrapolative mechanism (see Gandolfo, 1997, p. 210), we can write
It is worth noting that, as in the MVM1, the equilibrium values just
Y e = Y + τ Y , τ > 0 obtained guarantee interesting steady-state results. In particular
• they guarantee a steady-state growth of output and capital stock at a rate
from which, inserting into (17), we obtain
equal to the natural rate
I = ( v + vτ gY − v )Y (18)
gn η ⎡
where gY = Yˆ and v = K/Y.
gY** = +
1 − η vτ 1 − η vτ ⎣
( )
v − v** − sc 1 − U ** ⎤⎦ = gn
Second, we consider the following error-adjustment mechanism for
v + vτ gY** − v**
disequilibrium in the product market18 gK** = = gn
v**
Y = gnY + η ( I − S ) , η > 0 (19)
• in order to be economically meaningful, they require that the Pasinettian
where, as in the OVM, we assume that sw = 1, which implies: case 0 < gn v** < sc ≤ 1 holds;
• they imply the Cambridge equation:

C=−
( )
h′ E * U **η sc E * (1 + gn )
(27)
1 − η vτ
1 − U ** gn v** 1 g
r ** = **
= **
= n
v sc v sc Under the following two assumptions:
However, as in the MVM1, the system may not converge to the positive 1
Assumption 1: 1 − ητ v < 0 ↔ τ > >0
equilibrium point, but rather persistently fluctuate around it. The next two ηv
sections are devoted to the analysis of the dynamics of the MVM2, from both
a qualitative and numerical point of view. Assumption 2: 1 − η v** − δη scU ** + (1 − ητ v ) gn < 0
15.4.1. Qualitative Analysis of the Dynamics from (25), (26) and (27) it follows that all coefficients of (24) are positive so
* ** **
that the sign of the expression AB − C is undetermined. In particular, in terms
Linearising the dynamical system (21)–(23) at (E , U , v ), we obtain of the parameters of the model, it is easy to check that we have AB − C < 0
according as to whether
⎡ E ⎤ ⎡ E − E* ⎤ ⎡ 0 a12 a13 ⎤ ⎡ E − E* ⎤
⎢ ⎥ ** ⎢ ** ⎥ ⎢ ⎢ ** ⎥
⎢U ⎥ = J ⎢U − U ⎥ = ⎢ a21 a22 a23 ⎥⎥ ⎢U − U ⎥ F (τ ) = − ⎡⎣(τ v − v** ) − δ scU ** ⎤⎦ ⎡⎣δη (1 + gn ) + η E * h′( E * ) ⎤⎦
⎢ v ⎥ ⎢ v − v** ⎥ ⎢⎣ 0 a32 a33 ⎥⎦ ⎢ v − v** ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦
−δ (1 −η vτ )(1 + gn ) < 0
2
(28)
where
We are then in a position to prove the following proposition:
∂E η s E* ∂E η E*
a12 = = c , a13 = =−
∂U 1 − η vτ ∂v 1 − η vτ Proposition 1: Under Assumptions 1 and 2, if expectations are extrapolative
and such that
∂U ∂U δη scU ** ∂U δηU **
a21 =
∂E
( )
= h′ E * U ** > 0, a22 = =
∂U 1 − η vτ
, a23 =
∂v
=−
1 − η vτ
1
<τ ≤
v** (1 + τ gn )
= ↔
1
<τ ≤1 (29)
ηv v (1 + gn ) ηv
a32 =
∂v
=
( )
τ v − v** η sc
, a33 =
∂v
= − (1 + gn ) −
(
τ v − v** η ) the positive equilibrium (E*, U**, v**) of the dynamical system of the MVM2 is
∂U 1 − η vτ ∂v 1 − η vτ locally stable.
Thus, the characteristic equation of the linearised system is Proof: We already noted that, under Assumptions 1 and 2, all coefficients of
the characteristic equation (24) are positive. Moreover, when (29) holds,
λ + Aλ + Bλ + C = 0
3 2
(24)
from (28) it follows that we also have
where
F(τ) > 0 ↔ AB − C > 0
−η v − δη scU + 1 + gn − ητ vgn
** **
A= (25) All Routh–Hurwitz conditions for (local) stability of the positive
1 − η vτ equilibrium point are therefore satisfied. L/
η scU ** ⎡ However, when condition (29) does not hold and the only restrictions on the
B=− δ (1 + gn ) + E * h′ ( E * )⎤⎦ (26) value of τ are those implied by Assumptions 1 and 2, we can establish the
1 − η vτ ⎣
possibility of persistent cyclical paths of the variables of the model. More
precisely, we can establish the following result:
Proposition 2: Under Assumptions 1 and 2, there exists a value of the v**) is a stable focus. This implies that, after a certain period of time, when
parameter τ = τH > 1 at which the dynamical system (21)–(23) of the MVM2 the growth cycle has completely dampened out, output starts to grow
undergoes a Hopf bifurcation. exponentially at a rate equal to the natural rate gn = 0.0258 (see Figures
15.8(iii) and 15.8(iv)).
Proof: It is easy to adjust to the present case the method of proof of the Hopf
bifurcation adopted in previous contributions, for example in Asada (1995).
To this end, let us note that, under Assumptions 1 and 2, all coefficients of
the characteristic equation (24) are positive. Moreover, from (28) we know
that AB − C < 0 according to whether F(τ) < 0. Proposition 1 implies that
AB − C > 0 for all values of τ ≤ 1. Let us instead indicate with τ > 1 a value
of the parameter for which F (τ ) < 0 so that A(τ ) B(τ ) − C(τ ) < 0 and the
equilibrium is locally unstable. By continuity, this means that there exists at
least one value of the parameter τ = τ H ∈(1,τ ) at which
A(τ H ) B(τ H ) − C(τ H ) = 0 (30)
∂ [ A(τ ) B(τ ) − C(τ )]

≠0 (31)
∂τ τ =τ H
Now, it is possible to prove (see Asada, 1995, p. 248) that, under

Assumption 1, (30) is a necessary and sufficient condition for (24) to have a
pair of purely imaginary roots λ1,2 = ±iω (ω ≠ 0). Then it is also
straightforward to prove (see Asada, 1995, pp. 267–8) that condition (31) Figure 15.8: Locally stable positive equilibrium of the MVM2
implies that the real parts of the complex roots for τ = τH cross the real axis at
non-zero speed. We can therefore conclude that the dynamical system Convergence to a stable limit cycle. Proposition 2 proves only the existence
(21)–(23) undergoes a Hopf bifurcation at τ = τH. L part of the Hopf bifurcation theorem and says nothing about the uniqueness
and stability of the closed orbits. The numerical simulations we have
15.4.2. Numerical Simulations performed for a large number of initial conditions, however, seem to suggest
that the emerging closed orbit is stable (i.e., the Hopf bifurcation super-
We finally turn to the illustration of the analytical findings of Section 15.4.1 critical) and unique. For example, with initial conditions E(0) = 0.91,
with the help of numerical simulations. U(0) = 0.91 and v(0) = 2.4 and the same set of parameter values as before
(except that τ = 1.0211, such that Assumptions 1 and 2 but not condition (29)
Cyclical convergence to the positive equilibrium. As a starting point of our are satisfied), the trajectory converges to a limit cycle (see Figures 15.9(i)
numerical analysis, we choose η = 1, sc = 0.9, v = 2.57, τ = 0.75, α = 0.0221, and 15.9(ii)). Thus, in this case, output is subject to persistent cyclical
β = 0.0037, γ = 0.9, ρ = 1 and δ = 0.02 which imply (E*, U**, v**) = (0.9221, growth (see Figures 15.9(iii) and 15.9(iv)).
0.9268, 2.5538) and are such that Assumptions 1 and 2 and condition (29) of
Proposition 1 are satisfied.
As was to be expected, the trajectory (starting, for example, from initial 15.5. CONCLUSIONS
conditions equal to E(0) = 0.81, U(0) = 0.81 and v(0) = 2.4) converges to the
positive equilibrium (see Figures 15.8(i) and 15.8(ii)). The convergence In the wake of the recent spurt of interest in the study of the interaction
* **
appears to be cyclical, so that we can conclude that the equilibrium (E , U , between economic growth and economic fluctuations, in this chapter we
limit cycle solutions. Thus, both overcome the first of the two drawbacks of
the OVM mentioned above.
It is worth recalling that the increase in the dimensionality of the
dynamical system in the MVM1 is simply due to the introduction into the
OVM of the hypothesis of differential savings along Kaldorian–Pasinettian
lines, together with a more general specification of the Phillips curve: the rest
of the OVM is left unchanged. In the MVM2, on the other hand, such an
increase is due to the introduction into the OVM of an independent
investment function (which corresponds to the flexible accelerator), together
with the assumption of disequilibrium also in the product market. In this
sense, the MVM2 represents an attempt at integrating the two prototype
models we considered in the paper in such a way that also the second of the
two shortcomings of the OVM is overcome. In performing this integration,
however, we considered a simplified version of the Keynesian prototype
model. It remains to be ascertained whether the integrated model can
generate cyclical growth paths in the case in which a more general
specification of investment behaviour is used. By this we mean a
specification which, besides an error-adjustment mechanism for induced
investment, takes account of the interpretation of the autonomous component
Figure 15.9: Stable limit cycle of the MVM2 of investment outlined in Section 15.2.2, and tries to develop it. This issue,
given the recent revival of Schumpeterian ideas within the endogenous
presented two prototype models belonging to the non-market clearing growth approach, appears to be worth investigating further.
approach to macrodynamics, and analysed them with regard to their capacity
to generate cyclical growth paths of the relevant variables.
First of all, we considered the simple Keynesian ‘multiplier–accelerator’
NOTES
model which, while providing basic insights into the role played in the
dynamics of the economy by effective demand problems, can generate
1. This was stressed, in many occasions, by Goodwin. See, for example, Goodwin (1986). See
cyclical growth paths only when an exogenous trend is superimposed on it. also Kaldor (1954, pp. 53–6).
We showed that the version of the model with a non-linear formulation of the 2. One should be aware that, more recently, in the last two decades or so, there has been a
accelerator can generate cyclical growth paths also in the case in which we renewed interest in the problem of the interrelation of economic growth and business cycles,
assume that the autonomous (‘innovational’) component of investment is a within both the literature on real business cycles (e.g., Kydland and Prescott, 1982) and the
periodic function of time. This, as a first approximation, can be interpreted as literature on endogenous growth (see Aghion and Howitt, 1998, ch. 8). This more recent
(market clearing) literature is critically surveyed by Fiaschi and Sordi in their contribution
a way of introducing into the model Schumpeter’s idea of clustering of to this volume.
innovations. As a second prototype model, we then presented Goodwin’s 3. This feature of linear models is one of the most critically debated by the ‘pioneers’ of
model which, by contrast, provides a neat explanation of growth cycles based economic dynamic modelling such as Goodwin (e.g., 1953, 1955), Harrod (e.g., 1939,
on the Marxian reserve army mechanism. The OVM is known to be 1951), Kaldor (1954) and Kalecki (e.g., 1968). See also Pasinetti (1960) and Allen (1967,
structurally unstable and it focuses only on the labour market, in sharp ch. 18 and 19).
4. In order to guarantee continuity with what follows, we consider a continuous-time
contrast with the other prototype model. In the attempt to overcome these formulation of the model. That is, rather than Samuelson’s (1939) original formulation or its
two drawbacks of the OVM, we proposed two extensions of it and showed extension by Hicks (1950), we consider the version of the model studied by Phillips (1954).
that both modified versions of the model (MVM1 and MVM2) entail a three- As usual, in what follows a dot over a variable (e.g., z ) indicates its derivative with respect
dimensional dynamical system. In both cases, then, by application of the to time (dz/dt), whereas a hat (e.g., ẑ ) indicates its rate of growth ( z / z ).
Hopf bifurcation theorem, we were able to prove that the system can admit 5. In presenting these results, we are simply adjusting to our case and notation the presentation
of Phillips’s model made by Allen in his celebrated book on Macroeconomic Theory (1967,
The interaction between growth and cycle in macrodynamic models 305
pp. 328–33). In short, given the characteristic equation of (5), λ + (1/ε)(ε + s − v )λ + (s/ε)
2
1/2
= 0, and the expression for its roots, λ1,2 = (1/2ε)[( v − ε − s) ± Δ ], where Δ = ( v − ε − s) −
2
4sε, the intervals of parameter values listed in Table 15.1 are obtained by jointly considering
the conditions under which these roots are real or complex (Δ ≥ 0 or Δ < 0) and the stability
condition ( v < ε + s). On the application of this kind of qualitative analysis to Phillips’s
6.
model, see also Flaschel (1993, pp. 100–01).
In passing, it should be noted that this (with v exactly equal to ε + s) is the only case in
16. Real business cycle models,
which the simple linear multiplier–accelerator model we are considering is able to represent
oscillations which last in time, neither exploding, nor dying away.
endogenous growth models and
7.
c
Mathematically, the cyclical component Y (t) is given by the general solution of the
p
homogenous equation, whereas the growth or trend component Y (t) is given by the
cyclical growth: a critical survey
particular solution of the non-homogenous equation.
8. On this type of criticism, see Kaldor (1954). See also Aghion and Howitt (1998, p. 234). Davide Fiaschi and Serena Sordi
9. See Goodwin (1951, p. 9). In the numerical simulations which follow, we will use the
following functional form for φ:
⎧ I
max
+I
min
⎫ 16.1. INTRODUCTION
φ (Y ) = ⎨ − 1⎬ I
min
⎩I exp[ −( I +I ) vY / I I ] + I ⎭
max max min max min min
Since the early contributions to the topic, business cycles have been
which, as required by the nonlinear accelerator, is such that φ (0) = 0, φ′ (0) = v , considered as essentially connected with the development of capitalist
φ (−∞) = −Imin < 0 and φ (+∞) = Imax > 0. For a similar specification, see Allen (1967, pp. 378– economies. In the early 1980s the issue of the relationship between growth
80). and cycles was addressed within a market clearing environment by the real
10. As stressed by Gandolfo (1997, p. 522), forced oscillators of this kind have recently
attracted the attention of students of economic dynamics for their capacity to generate
business cycle (RBC) literature, and, more recently, in endogenous growth
transient chaotic dynamics. Of course, given an appropriate choice of parameter values, this (EG) models.1
is also the case for equation (7) with (9). In the articles belonging to the RBC literature, which are based on the
11. In Aghion’s and Howitt’s opinion (1998, p. 234), this is perhaps the first model in which the neoclassical model of (optimal) capital accumulation augmented by
occurrence of economic fluctuations was modelled as a deterministic consequence of the technology shocks, it is common to find assertions like ‘our approach
accumulation (i.e., growth) process; more specifically, of the variations in income
distribution this process induces over time.
integrates growth and business cycle theory’ (Kydland and Prescott, 1982,
12. To the best of my knowledge, this more general version of the Phillips curve was first p. 1345) or ‘real business cycles theory (...) holds considerable promise for
introduced in the OVM by Cugno and Montrucchio (1982, p. 97). enhancing our understanding of economic fluctuations and growth as well as
13. See Sordi (2001), where this is proved by applying to the dynamical system (13)–(15) the their interaction’ (King, Plosser and Rebelo, 1988, p. 196). Thus, it appears
Hopf bifurcation theorem. For the numerical simulation, we have used a linear that such an integration of growth and business cycle theory is understood as
approximation of the generalized Phillips Curve (12), of the type f1 ( E, Eˆ ) ≈ −γ + ρ E + δ Eˆ .
14. In doing this, we are following a suggestion made by Flaschel, Franke and Semmler (1997,
one of the most, if not the most, important achievement of the analysis.
p. 101), who maintain that ‘a proper integration of a Goodwin growth cycle with the The relationship between growth and cycles has also been tackled from a
multiplier–accelerator and the study of the dynamics arising from it is still an open problem different point of view; some contributions in the EG literature focus on the
and should be addressed by future research’. possibility of generating non-linear (periodic) dynamics as the effect of
15. For other contributions in which an independent investment function has been introduced introducing endogenous growth into an otherwise neoclassical growth model.
into the ‘growth–cycle’ framework see, for example, Glombowski and Krüger (1988),
Rampa and Rampa (1988) and Wolfstetter (1982). See also Flaschel (1988, 1993).
In the whole of this variegated literature on RBC and EG, one can
16. In order to keep things simple, we assume that Ia(t) = 0 for all t. distinguish at least three main different approaches to the study of the
As is easy to verify, (17) and (4) are equivalent when Y = Y and θ = 1. interaction between growth and cycles:
e
17.
18. For a similar assumption made in a different extension of Goodwin’s model, see
Glombowski and Krüger (1988, p. 427). 1. Starting with the contributions by Kydland and Prescott (1982) and Long
19. The derivation of this and other calculations in the paper are available from the author upon and Plosser (1983), RBC theorists have studied the interaction between
request. growth and cycles within a stochastic business cycles framework, where
cycles are generated by continuous exogenous shocks to technology;
306
Real business cycle models, endogenous growth models and cyclical growth 307 308 The Theory of Economic Growth: a ‘Classical’ Perspective
2. Some EG theorists analyse the implications of a sounder microfoundation can account for most of the cyclical component. This conclusion, however,
of technological progress on the relationship between growth and cycles. can be criticised from many points of view. We focus on the assumption of
Examples are Aghion and Saint-Paul (1998a), where cycles are generated exogenous technological change. Stadler (1990) and Aghion and Saint-Paul
in a model with a Schumpeterian flavour, and Stadler (1990), where (1998a) provide two interesting contributions, whose findings substantially
growth is generated by a learning-by-doing process; differ from RBC results. The next section is devoted to the exposition of a
3. Finally, the possibility of multiple steady states in EG models has given standard RBC model, which then will be used as a term of comparison for
importance to the analysis of out-of-steady-state dynamics (cycles) in a the two models of endogenous technological progress presented in Section
deterministic framework (see Greiner and Semmler, 1996a, 1996b; and 16.2.2.
Benhabib and Perli, 1994).
16.2.1. A Basic Model of RBC
The aim of this chapter is to survey and compare these contributions, starting
from the stochastic approaches 1 and 2 (Sections 16.2) and continuing with In this section we present a basic RBC model, referring, in particular, to the
the deterministic approach 3 (Section 16.3). Section 16.4 concludes and classical contribution by Christiano and Eichenbaum (1992), in which the
gives some suggestions for further research. authors analyse an RBC model with government expenditure and
endogenous labour supply. For the sake of simplicity, in our presentation, we
ignore the government expenditure.
16.2. GROWTH MODELS WITH STOCHASTIC BUSINESS The model economy considered is composed by an infinitely living
CYCLES representative agent, which maximises the discount sum of instantaneous
utility. To keep things simple we consider a log instantaneous utility
The first attempts to explain business cycles on the basis of stochastic shocks function, U = ln ct + φ V (lt ) , where ct is the per capita consumption and lt is
are due to Frisch (1933) and Slutsky (1937). While the latter showed how the leisure at period t.
sum of random components generates cycles similar to empirical The production side of the economy is characterised by a competitive
fluctuations, the former presented technical innovations as exogenous market where each firm produces homogeneous output according to a Cobb–
perturbations to the available level of technological progress. In the same Douglas constant returns to scale technology:
period Schumpeter (1939) identified in the continuous introduction of new
yt = ( At zt ht )
1−θ θ
kt (1)
innovations, and the resulting shocks to productivity, the source of growth of
a country. In this view growth and business cycle are generated by the same where yt is per capita output, At, an index of long-run deterministic
source and therefore they must be jointly analysed. Unlike Frisch, however, technological progress, zt, an index of short-run cyclical productivity,
Schumpeter considered innovations as driven by economic factors, i.e., as ht = L − lt are the worked hours, and kt is per capita capital.
endogenous.2 These contributions are the main inspiration of the modern Then it is assumed that productivity evolves according to the following
theory of RBC, where the business cycle is seen as a phenomenon essentially stochastic process3
due to shocks to the real part of the economy and long-run growth as the
cumulative sum of such shocks. zt = zt −1 exp (γ + ε t ) (2)
The seminal contribution by Kydland and Prescott (1982) starts from the
idea of analysing – within the neoclassical framework of optimising agents – (
where γ > 0 is a constant drift and ε t
N 0, σ ε2 . )
the behaviour of an economy which is converging toward its long-run From the assumption of competitive markets it follows that factors are
equilibrium but which is continually shocked by random disturbances. The paid according to their marginal productivity, so that
standard RBC model is essentially a neoclassical growth model in which
rt = θ ⎡( At zt ht ) kt ⎤ − δ
1−θ θ −1
exogenous technological progress is modelled as a stochastic process (see ⎣ ⎦
Cooley and Prescott, 1995). The aim is to simulate series whose properties
wt = (1 − θ ) ⎡⎣ At zt ht kt ⎤⎦
1−θ 1−θ −θ θ
are similar to those of observed series. Their conclusion is that the business
cycle is essentially a real phenomenon and that the neoclassical framework
where δ is the depreciation rate of capital.
In this economy the first welfare theorem holds, so that it is convenient to Table 16.1 we report the standard deviations of our simulated series after
solve the competitive allocation as a social planner problem, that is4 they are logged and detrended by the Hodridk–Prescott filter to extract the
transitory component at frequency 4–6 years and, for comparison, we also
max W = E0 ∑ t = 0 β t {ln ct + φ V (lt )}
∞
(3) report the corresponding statistics relative to US data (see Canova, 1998).6
{ct , lt }t∞=0
As we see, the simulated series show standard deviations normalised with
respect to the standard deviation of output that in some cases overestimate
kt +1 = ( Azt ht ) ktθ − ct + (1 − δ ) kt
1−θ
subject to the real values (consumption and productivity) while in others underestimate
k0 = k (0) them (investment, hours worked and real wages).
Since zt follows a stochastic path, a closed form solution to problem (3) Table 16.1 – Standard deviations, sources US estimates: Canova (1998)
does not exist in general, but only for a particular configuration of the
parameters (θ = δ = 1 and V(lt) = ln lt). The procedure generally used consists Statistic of σ Simulated data US data 1955:3–1986:3
in calculating the steady-state equilibrium (which corresponds to the locus σc/σy 0.6174 0.49
which the economy converges to if it were not subject to productivity σi/σy 2.2055 2.82
shocks) and in approximating problem (3) around this steady state.5 In order σh/σy 0.3710 1.06
to calculate the steady-state equilibrium, it is useful to normalise each σw/σy 0.6785 0.70
growing variable with respect to its long-run growth rate. By so doing, we σy/n/σy 0.6949 0.49
obtain the following normalised variables:
Another point is the correct sign and magnitude in the cross-correlation
kt c y z
kˆt = , cˆt = t , yˆt = t , zˆt = t among variables. In Table 16.2 we report such cross-correlation for
At −1 zt −1 At z t At zt zt −1 simulated series (firstly detrended and then filtered by the Hodrick–Prescott
The solution to problem (3) is given by the following difference equations procedure) and for US data. Table 16.2 shows the main drawback of the
RBC model, namely, the discordance between simulated and real series with
ht +1 = qh kˆtnh zˆt h ; kˆt +1 = qk kˆtnk zˆtmk
m
(4) respect to the cross-correlation of hours worked and productivity with
output. The negative empirical correlation between hours worked and
where the coefficients qi, ni and mi for i = h, m are non-linear functions of the productivity suggests that the labour market does not work as a perfectly
original parameters of the model. competitive market. Moreover, simulations predict a strong correlation
The system of difference equations (3) – together with the stochastic between productivity and output whereas the empirical evidence shows only
process (2) – fully describes the dynamics of model. As already stressed, a a slightly positive correlation. According to Canova (1998, p. 503) this
closed form solution does not exist and qi, ni and mi must be determined by correlation changes over time so that there is ‘the need for theoretical work
numerical simulations. Christiano and Eichenbaum (1992, p. 441) provides a to provide reasons for why this phenomenon occurs’.
numerical simulation for which per capita capital shows a stochastic trend
(the average growth rate of which equals to the value of the drift of Table 16.2 – Correlations, sources US estimates: Canova (1998)
stochastic process). On the contrary, hours worked do not show any trend.
Further the system proves to be stable. Statistic of σ Simulated data US data 1955:3–1986:3
The properties of the artificial model economy are analysed by means of corr(c, y) 0.9481 0.75
numerical simulations. In so doing, the purpose is to match some empirical corr(I, y) 0.9677 0.91
regularities or stylised facts (see Canova, 1998). These regularities generally corr(h, y) 0.8890 0.88
refer to differences in the variance of some relevant variables and/or cross- corr(w, y) 0.9705 0.81
correlations among the latter. corr(y/h, y) 0.9684 0.10
Let us turn our attention to the problem of the choice of proper indicators corr(h, w) 0.7691 0.67
for business cycles. We start from the standard deviations of the variables. In corr(h, y/h) 0.7887 –0.24
The latter point suggests that the current RBC theory needs a sounder yt = z χ ht1−θ aηt , with 0 < θ < 1, χ > 0,η > 0 (6)
microfoundation of technological progress. This will be the argument of the
next section. where zt is the productivity shock and at the per capita accumulated
knowledge available to the firm. The latter is accumulated both by an
16.2.2. Microfoundation of Technological Progress exogenous component and as a by-product of production:9
at = at1−−1δ a ytλ−1 (7)

The exogeneity of technological progress was the most serious criticism
faced by exogenous growth models à la Solow. In the same way some where λ, δa > 0.
authors argue that a business cycle theory needs a microfoundation of The productivity shock follows a stochastic process similar to (2) but
innovation. Moreover, as we showed in the previous section, the RBC model without drift:
predicts a strong positive correlation between productivity and output, which
contrasts with the empirical evidence.7 Thus, endogenising technological zt = zt −1 exp(ε t ) ε t
N 0,σ ε2 ( ) (8)
progress may help to understand the causes of this phenomenon.
Firstly, we analyse the model proposed by Stadler (1990), in which Given the assumption of exogenous knowledge accumulation, the
technological progress is viewed as a learning-by-doing process, i.e., as a by- intertemporal maximization problem (5) of the representative firm becomes a
product of production. In so doing, our purpose is to stress that, even if static optimisation problem. Therefore the demand curve of the firm is given
productivity is always strongly procyclical, this has crucial implications for by the first order condition (FOC) to the problem of the maximisation of
business cycles analysis, in particular for the effect of monetary policy on profits at time t, that is:
long-run productivity.
1/ θ
Secondly, we consider the contribution by Aghion and Saint-Paul ⎡ (1 − θ ) ztχ aηt ⎤
(1998b), in which technological progress is endogenised within a h =⎢
t
d
⎥ (9)
Schumpeterian framework. The focus in this case is on the relationship ⎢⎣ wt / pt ⎥⎦
between productivity and the growth rate of output; it will be shown that
productivity can be both procyclical and countercyclical. Labour supply is assumed to be10
Stadler (1990) proposes a model in which growth is generated only φ2
⎛w ⎞
through learning by doing. Business cycles are generated by shocks to hts = exp (φ1 ) ⎜ t ⎟ with φ1 ,φ2 > 0 (10)
productivity similar to those in the model presented above but without a ⎝ pt ⎠
positive drift. In the model economy there is no fixed capital and the only
accumulated factor is knowledge, which is accumulated through learning by Stadler (1990) then assumes that neither consumer nor firm knows the
doing. The learning-by-doing process is external to the firm and is a by- level of price at period t when they bargain in the labour market to set the
product of production. In such a framework, the consumer plays no role in wage at the beginning of the period. Thus, they formulate their choice
the allocation of resources. As a consequence, the condition for according to their expectations of the level of prices at time t, given their
maximisation of profits of the representative firm determines the competitive information set (which included the level of productivity zt–1).
allocation. In our case, we obtain the following expression for the log of aggregate
The representative firm solves the following problem: supply yts
⎧⎪ ∞ ⎡ ⎛ wt ⎞ ⎤ ⎫⎪ (
yts = f0 + f1 zt + f2 zt −1 + f3 at + f4 p t − p te ) (11)
V = max ∞
E0 ⎨∑ β t ⎢ yt − ⎜ ⎟ ht ⎥⎬ (5)
{ht }t =0
⎩⎪ t = 0 ⎣ ⎝ pt ⎠ ⎦ ⎭⎪ where ‘~’ over a variable denotes the log of the variable,
f0 = ( 1 − θ ) [ φ 2 lo g ( 1 − θ ) + φ 1 ] /( 1 + φ 2 θ ) , f1 = χ / θ , f2 = −(1−θ)χ/θ(1+φ2θ),
where 0 < β < 1 is a discount factor,8
wt, the monetary wage, pt, the price
f3 = ( 1 − θ ) φ 2 η / ( 1 + φ 2 θ ) and f4 = ( 1 − θ ) / θ . Hence aggregate supply
level and ht, the labour employed by the firm.
depends on productivity shocks, zt , accumulated knowledge, at , and on the
It is assumed that the production function is:
difference between actual price and expected price p t − p te ; it is to be noted
that an expected increase in demand can lead to an increase in supply of ⎛ δa ⎞ ⎛ λ q0 ⎞

E0 [ yt ] = q0 ⎜ ⎟ (1 − δ a + λ q2 )
t
output. We next deal with this point. ⎟ + q2 ⎜ a0 +
On the basis of the quantitative theory, following Stadler (1990) we ⎝ δ a − λ q2 ⎠ ⎝ λ q2 − δ a ⎠
assume that aggregate demand is given by

The existence of a positive long-run growth of output depends on the
m value of 1 − δa + λq2. If 1 − δa + λq2 > 1, then the growth rate of output is
ytd = t (12) ever increasing; otherwise, the long-run growth rate is zero and the level of
pt
output (in log) converges to q0δa/(δa − λq2).
where mt is per capita money stock. mt is assumed not to be observable at the From equation (17) it follows that if the economy achieves positive long-
beginning of period t but follows the stochastic process run growth, i.e. 1 − δa + λq2 > 1, then both monetary and real shocks have
long-run effects. However they work through two different channels.
mt = mt −1 exp (γ m + ζ t ) ζ t
N 0, σ ζ2 ( ) (13) Monetary shocks affect the long-run output through the stock of accumulated
knowledge, while real shocks work both directly on output and indirectly on
where γm > 0 is the drift in money supply. the stock of accumulated knowledge. In this regard assume η = 0, i.e. the
In equilibrium we have: stock of accumulated knowledge is not relevant in production; this implies
that q2 = 0 (the long-run growth rate is equal to zero). From (16) we have:
ζ t − f1ε t
p t − p te = (14) t −1
1 + f4 yt = q0 + q1 ∑ ε i + q3ζ t + q4ε t
i =1
Substituting (14) in (11) yields equilibrium (log) output:

Therefore real shocks ε have long-run effects, while monetary shocks ζ
yt = q0 + q1 zt −1 + q2 at + q3ζ t + q4ε t (15) have only one-period effects. This case corresponds to the standard RBC
model of Section 16.2.1, provided that the stochastic process of productivity
where q 0 = [ φ 2 lo g ( 1 − θ ) + φ 1 ] /( 1 + θφ 2 } ) , q 1 = χ ( 1 + φ 2 ) /( 1 + φ 2 θ ) , has a positive drift.
q2=η(1+φ2)/(1+φ2θ), q3 = 1−θ and q4=χ. Finally, it is to be noted that monetary shocks alone can account for the
The dynamics of output is driven by monetary and real shocks (ζ and ε, unit root detected in empirical data. Consider the case where real shocks
respectively), by productivity shocks, z , and by the stock of accumulated have no effect on output, that is χ = 0, from which q1 = q4 = 0. From (16) we
knowledge, a . Equation (15), provided that at is time-constant, makes it have
clear why monetary shocks cannot have a long-run effect in an RBC model:
in fact, only the level of productivity z is relevant in determining the long- ⎡ t
t− j ⎤
yt = q0 ⎢1 + λ ∑ (1 − δ a + λ q2 ) ⎥ + q2 (1 − δ a + λ q2 ) a 0
t
run behaviour of y , while the monetary shock ζ has only a short-run effect. ⎣ j=0 ⎦
However, if a incorporates also the short-run dynamics of y , as in our t
+ λ q3 ∑ (1 − δ a + λ q2 )
t− j
model, then monetary shocks can have a long-run effect. ζ j + q3ζ t
j=0
Substituting (7) and (8) in (15) leads to:
and, since 1 − δa + λq2 > 1, monetary shocks ζ have long-run effects and yt
t −1
⎧ t
yt =q0 + q1 ∑ ε i + q2 ⎨(1 − δ a + λ q2 ) a0 + λ ∑ ( 1 − δ a + λ q2 ) ⋅
t t− j
presents a unit root (or more than a unit root in our case where growth rate is
i =1 ⎩ j =0 ever increasing).
(16)
⎡ j −1
⎤ ⎫⎪
⎢q0 + q1 ∑ ε i + q3ζ j + q4ε j ⎥ ⎬ + q3ζ t + q4ε t 16.2.2.1. Endogenous growth theory and the Schumpeterian approach
⎣ i =1 ⎦ ⎭⎪ to innovation
The relationship between productivity and growth is hotly debated in the
From equation (16) we can calculate the expected level of long-run (log) literature. According to Schumpeter, recessions are required to eliminate
output: inefficient firms from the market, so that the final effect of a decrease in
economic activity is an increase in overall productivity (see Caballero and Given that the equilibrium in the monopolist market i is given by yi,t = di,t,
Hammour, 1991). There are however factors suggesting that productivity is we then have:
procyclical, like learning by doing, demands spillovers and capital market
−1/ ψ ⎛ z ⎞ 1/ψ (ψ −1) /ψ
imperfections that constrain investment in the R&D sector (see Stiglitz, pi,t = ⎡⎣1 − μ ( vi,t )⎤⎦ exp ⎜ − i,t ⎟ Yt Pt (20)
1993). Moreover, Bean (1990) stresses that, if the reorganization of the firm ⎝ ψ ⎠
is costly in terms of output, during a recession the opportunity cost of such
an activity is lower and therefore the procyclical pattern of productivity is where pi,t is the price of good i produced by firm i.
magnified. The empirical evidence for the relationship between productivity Given the actual level of productivity, firm i must choose the optimal
and output is mixed. Canova (1998), for example, shows that productivity increase in productivity; the latter is the result of the following problem:12
was countercyclical up to the mid 1960s and procyclical afterwards.
In what follows we discuss a simplified version of the Aghion and Saint- vi
{
V ( zi ,t ) = max π i ,t dt + (1 − rdt ) Et ⎡⎣V ( zi ,t + vi dt )⎤⎦ } (21)
Paul (1998b) model, which provides an example of how productivity can be
countercyclical if innovations are endogenised in a Schumpeterian fashion. where V is the value of the firm, r is the constant interest rate at which agents
The basic idea is that the cost of innovation is lower in recession than in a can lend or rent their resources. The future value of the firm depends on the
boom if the act of innovating drains resources from production. This level of prices, which, in turn, depend on the aggregate demand index; the
framework corresponds to the case analysed by Hall (1999), who assumes expectation operator reflects the possible uncertainty of the latter variable.
that the increases in productivity are the result of internal reorganization of a To determine the equilibrium, market entry and exit conditions must be
firm, which negatively affects the current level of output.11 specified. Following Aghion and Saint-Paul (1998b) we assume that the firm
In the economy at period t there are Nt firms and Nt different goods; each must bear a fixed cost equal to C to enter the market, while the firm has a
firm is a monopolist in its own market. Each is characterised by a level of liquidation value equal to τC, where τ < 1.13 Finally, it is assumed that new
productivity zi,t and it is assumed that firms cannot change their employed firms show the same level of productivity. In the equilibrium the number of
workforce and production capacity. Thus, the gross product of each firm is firms will be constant if the expected value of a firm V is greater than τC, but
determined only by its productivity, which is given by exp(zi,t). lower than C, that is
Let vi,t = dzi,t/dt be the change in productivity. Each firm can modify its
productivity by sacrificing part of its production; a change in productivity N = 0 ↔ V ∈ [τ C, C ] (22)
equal to vi,t involves a proportional drop of output equal to μ = μ ( vi,t), where
the authors assume that μ ( 0) = 0, μ′ > 0, μ″ > 0. In equilibrium we expect that V ∈[τC, C] since V greater than C drives
For firm i, demand at period t is given by: new firms to enter the market, causing V to decrease up to C, while a firm’s
value lower than τC leads firms to exit, causing V to increase up to τC.
−ψ
⎛ D ⎞⎛ p ⎞ Consider the symmetric equilibrium where pi,t = pt and zi,t = zt and
di,t = ⎜ t ⎟ ⎜ i,t ⎟ (17) therefore vi,t = vt for all i.14 From (22) we obtain the level of profits for each
⎝ Pt ⎠ ⎝ P t ⎠ firm:
where ψ > 1, Dt is an aggregate demand index and Pt, an aggregate price Dt
index, defined by πt = = dt (23)
Nt
1/ (1−ψ )
⎛ Nt ⎞ where dt = Dt/Nt is the demand for the single firm.
Pt = ⎜ ∫ pi1,−t ψ di ⎟ (18)
⎜ ⎟ Aghion and Saint-Paul (1998b) assume that cycles are generated by
⎝0 ⎠
fluctuations in the level of aggregate demand Dt. In particular, it is assumed
that there exist only two states, E, expansion and R, recession, and that the
The net output of firm i, yi,t, is given by:
probability of jumping from state R to E follows a Poisson process and is
yi,t = ⎡⎣1 − μ ( vi,t )⎤⎦ exp ( zi,t ) (19) given by ε, while from state E to R it is given by ζ.
In the stochastic steady state for each possible state E and R, all variables, cost of innovating, measured in terms of lower output, is higher in expansion
except for z and p, are constant and the economy stays for a fraction of time because the level of demand (and therefore of profits) is greater than in
equal to ε/(ζ + ε) in expansion and for a fraction equal to ζ/(ζ + ε) in recession. This is the main finding of the model: recession can have a
recession. In such a framework the problem of the firm (21) becomes:15 positive impact on the growth rate of productivity. Finally, from (27) it
follows that ∂vj/∂r < 0, j = E, R, ∂vR/∂ε > 0, ∂vR/∂ζ < 0, ∂vR/∂τ < 0 and
⎪ { ⎣ }
⎧ Vt R = max v π tR dt + (1 − rdt ) ⎡(1 − ε dt )Vt R+ dt + ε dtVt E+ dt ⎤
⎦ ∂vE/∂τ > 0.
⎨ E (24) The business cycle affects the growth rate of the economy because of the
{
⎪⎩Vt = max v π t dt + (1 − rdt ) ⎡⎣(1 − ζ dt )Vt + dt + ζ dtVt + dt ⎤⎦
E E R
} different increases in productivity which characterise the two states. The
average growth rate of the economy is given by:
where π tj = Dtj / Nt = dtj for j = E, R.
To close the model we must calculate the level of per capita demand in ⎛ ζ ⎞ R ⎛ ε ⎞ E
g=⎜ ⎟v +⎜ ⎟v (28)
the two states; in fact, since the number of firms is endogenous, a higher ⎝ζ +ε ⎠ ⎝ζ +ε ⎠
level of aggregate demand may not be matched by a higher level of per
capita demand. With regard to the average growth rate of the economy, Aghion and Saint-
We assume that recession is deep enough to lead some firms to exit; in Paul (1998b, p. 333) underline the importance of the following three effects:
this case the free entry condition implies that
1. Composition effect: the time that the economy spends in expansion, ε/(ζ +
Vt = τ C; V = C
R E
(25) ε), and in recession, ζ/(ζ + ε), given vR > vE, crucially affects the average
growth rate;
By inserting (25) into problem (24), we obtain the levels of profits for 2. Return effect: in expansion per capita demand is higher than in recession
each of the two states: and therefore the longer the economy spends in expansion, the higher are
the increases in productivity;
π R = C ⎡⎣rτ − ε (1 − τ )⎤⎦ ; π E = C ⎡⎣r + ζ (1 − τ )⎤⎦ 3. Cost of capital effect: since the firm does not recoup all costs of entry in
the case of exit, both the time it spends in expansion and in recession and
from which the levels of per capita demand also easily follow (see equation the liquidation value affect the incentive to increase productivity.
(23)):
Given all this, we can conclude that the three main parameters (ε, ζ, τ) of
d = C ⎡⎣rτ − ε (1 − τ )⎤⎦ ; d = C ⎡⎣r + ζ (1 − τ )⎤⎦
R E
(26) the model affect the average growth rate in the following way:
• ∂g/∂ε > 0 if dvR/dε > (vR − vE)/(ζ + ε); in the case of an increase in ε, the
From these expressions, it follows that dE > dR, i.e. per capita demand is
composition effect is negative (ζ(vE − vR)/(ζ + ε) < 0), whereas the sum of
2
higher in expansion than in recession.

the return and of the cost of capital effects is positive
From equations (26) and the FOCs of problem (24), we obtain:
([ζ /( ζ + ε ) ] /( d v R /d ε ) > 0 ), so that ∂g /∂ ε > 0 if the latter is greater than
μ ′ (vR ) τ
the former.
= • ∂g /∂ ζ < 0 if − d v E /d ζ > ( v R − v E ) /( ζ + ε ) ; in the case of an increase in ζ
1− μ v ( ) R
rτ − ε (1 − τ ) the composition effect is positive ε ( v − v R ) /( ζ + ε ) > 0 ), whereas the
E 2
(27)
μ ′ (v )E sum of the return and of the cost of capital effects are negative
1 ( [ ε /( ζ + ε ) ] ( d v E /d ζ ) < 0 ), so that ∂g /∂ ζ < 0 if the former is greater than
=
1 − μ ( v ) r + ζ (1 − τ )
E
the latter.
• ∂g /∂ τ > 0 if d v E /d τ > − ( ζ / ε ) ( d v R /d τ ) ; the cost of capital effect is
which provide implicit solutions for vR and vE. From (27), given the positive for vE ( [ ε /( ζ + ε ) ] ( d v E /d τ ) > 0 ) and negative for
assumptions on μ, it follows that vR > vE: this means that productivity in this v R ( [ ζ /( ζ ε ) ] d v R /d τ ) < 0 ), so that ∂g /∂τ > 0 if the former is greater than
model is countercyclical. The intuition is straightforward: the opportunity the latter.
To sum up, the idea that expansion is better that recession for growth is present the results of a recent contribution by Mattana and Venturi (1999), in
challenged in this model; our intuition is based on the fact that during a which they show that periodic solutions may emerge in the Lucas model.
recession the reorganization of a firm in order to increase its productivity is
less costly; this has implications for both the cross-correlation between 16.3.1. Persistent Cycles in One-Sector EG Models
output and productivity (which is negative rather than positive) and the effect
of recession on average growth rate (which is positive rather than negative). The model considered by Greiner and Semmler (1996a, 1996b) is a one-
sector EG model of the Romer type (with learning by doing) in which,
however, it is assumed that one unit of investment has different effects
16.3. ENDOGENOUS GROWTH MODELS WITH concerning the building up of physical capital and knowledge. This implies
DETERMINISTIC CYCLES that the two variables cannot be merged into a single variable.
The production possibilities of the model economy (in per capita terms)
Since the late 1980s, following seminal contributions by Lucas (1988) and are given by
Romer (1986a, 1990), a large number of articles have focused on the Y
dynamics of EG models, in a deterministic context. This line of research was = bAα k1−α , b > 0, α ∈ ( 0,1) (29)
L
strongly motivated by the fact that the original contributions by Romer and
Lucas focused on steady state only and, in addition, neglected the stability where A stands for the stock of knowledge, K, the stock of physical capital,
properties of the steady state. L, labour force, k = K/L. In what follows, to simplify, we choose b = 1.
We now turn to the analysis of this literature. Our main purpose is to Assuming that L grows exponentially in time at a constant rate equal to
check whether there exist results concerning the emergence of persistent n > 0, the equation for the evolution of k is the following:
cycles similar to those obtained by Benhabib and Nishimura (1979) for the
conventional exogenous growth models. A positive answer to this question k = i − (δ + n ) k
would imply that EG models offer an additional (deterministic) approach to
the study of cyclical growth in a market-clearing context, alternative to the where δ > 0 is the rate of depreciation, i = I/L and I, gross investment.
stochastic approach presented in Section 16.2. With regard to the stock of knowledge, it is assumed that it accumulates
In reviewing this literature on EG, it is useful to distinguish between one- according to a learning-by-doing process à la Arrow, in the formulation
sector models and two-sector models. In models belonging to the first class given by Levhari (1966). In addition, it is assumed that the contribution – to
(e.g., Romer, 1986a), the accumulation of knowledge is only a by-product of the formation of knowledge – of gross investment further back in time is
production activities and EG is generated by mechanisms of learning by smaller than that of recent gross investment. Hence (see Greiner and
doing, by externalities or by increasing returns. The models belonging to the Semmler, 1996a, p. 82):
second class, on the other hand, starting either from Lucas (1988) or from t
Romer (1990), generate EG by assuming an intentional allocation of A (t ) = ρ ∫ exp ⎡⎣ ρ ( s − t )⎤⎦ i ( s ) ds, ρ > 0 (30)
resources for the accumulation of human capital or an intentional R&D effort −∞
for increasing the level of technological progress. or

In what follows, we first review the existing literature on this topic with
regard to one-sector EG models (Section 16.3.1). We will focus in particular (
A = ρ (i − A ) = ρ Aα k 1−α − c − A ) (31)
on recent contributions by Greiner and Semmler (e.g., 1996a, 1996b), which
aim to show that a basic model of EG with learning by doing (which is a where ρ > 0 represents the weight given to more recent levels of gross
modified version of the Romer 1986a model) may produce a rich array of investment and c stands for per capita consumption.16
outcomes, such as multiple steady states, indeterminacy of equilibria or even Following Greiner and Semmler (1996a, p.82), we limit ourselves to
persistent cycles of the state variables. Then, in Section 16.3.2 we tackle the analysing the competitive situation, in which the evolution of knowledge is
same problem with regard to two-sector EG models. In this case, we discuss not explicitly taken into account by the representative agent when solving the
the classical contribution by Benhabib and Perli (1994) and we briefly
optimisation problem. Normalising so as to have L(0) = 1, the latter is the A balanced growth path for the original system is obtained as a rest point
following:17 of the reduced system at which kA / k A = cA / cA = 0 so that A / A = k / k = c / c .
∞
As shown by the authors (see Proposition 2 in Greiner and Semmler
Max
c ∫ exp ⎡⎣− ( β − n ) t ⎤⎦u (c(t ) ) dt, β > 0
0
(32) 1996a, p. 85 and Theorem 1 and Theorem 2 in Greiner and Semmler 1996b,
p. 110), in the case of positive per capita growth,
subject to: k = Aα k 1−α − c − (δ + n ) k . (i) if δ + n ≥ (δ + β ) / σ , there exists a unique steady state (c*A , k A* ) which
proves to be saddle stable and such that:
From the current-value Hamiltonian for problem (32), we then obtain the
following set of necessary FOCs for an optimum:18 k A*1−α − ρ k A*2−α − (δ + n ) k A* + ρ k A*
c =
*
A
1 − ρ k A*
u′ ( c ) − λ = 0 ↔ u′ ( c ) = λ (33)
(1 − α ) ⎡ (δ + β ) ⎤ ⎡ (δ + β ) ⎤
⎛ A⎞
α −
σ
(ρk *
A )
− 1 + k A* ⎢ ρ −
σ
*1+α
⎥ + ρ kA ⎢ − (δ + n ) +
σ ⎦
⎥=0
λ = λ (δ + β ) − λ (1 − α ) ⎜ ⎟ (34) ⎣ ⎦ ⎣
⎝k⎠
We are now in a position to derive the differential equations system which (ii) if δ + n < (δ + β ) / σ , there exist two steady states, (c*A1 , k A* 1 ) and
describes the dynamics of the model economy. From (33)–(34), one obtains (c*A2 , k A* 2 ) with k A* 1 < k A* 2 , such that the path associated with the second
can be anything except a saddle path.
⎡ u′′ (c ) c ⎤ c c ⎛ A⎞
α
⎢ ⎥ = −σ = (δ + β ) − (1 − α ) ⎜ ⎟
⎣⎢ u′ (c ) ⎦⎥ c
Given our aim, we are mainly interested in case (ii) in which the second of
c ⎝k⎠
the two steady states is either completely stable or unstable. To understand
why it is so, let us note that complete stability requires that the Jacobian of
which, together with (31) and (33), gives
the linearised system at (c*A2 , k A* 2 )
c ( 1 − α ) ⎛ A ⎞ δ + β
α
= − (35)
c σ ⎜⎝ k ⎟⎠ σ ⎡ −α k A*−2α − (1 − α ) ρ k A*12−α + c*A2 / k A* 2 ρ k A* 2 − 1⎤
J =
*
⎢ ⎥
⎢ − ⎡α (1 − α ) / σ ⎤ c*A2 k A*−2α −1 − (1 − α ) ρ c*A2 k A*−2α
2
⎣ ⎣ ⎦ ρ c*A2 ⎥⎦
k ⎛ A ⎞ ⎛ c ⎞
α
= ⎜ ⎟ − ⎜ ⎟ − (δ + n ) (36)
k ⎝ k ⎠ ⎝k⎠ is such that
c*A2
A
α −1
⎛ A⎞ ⎛c⎞ trJ*2 = −α k A*−2α − (1 − α ) ρ k A*12−α + + ρ c*A2 < 0
= ρ⎜ ⎟ − ρ⎜ ⎟− ρ (37) k A* 2
A ⎝k⎠ ⎝ A⎠
Following a standard practice, the order of system (35)–(37) can be ⎡ c* ⎤
det J*2 = ⎢ −α k A*−2α − (1 − α ) ρ k A*12−α + A* 2 ⎥ ρ c*A2
reduced by performing a change of variables with k A = k / A and cA = c / A . ⎣ k A2 ⎦
Hence:
⎡α (1 − α ) * *−α −1 ⎤
kA
(
+ ρ k A* 2 − 1 ⎢ ) cA2 k A2 + (1 − α ) ρ c*A2 k A*−2α ⎥ > 0
c
= k A−α − A − (δ + n ) − ρ k 1A−α + ρ (1 + cA ) (38) ⎣ σ ⎦
kA kA
The basic fact is that, as one of the parameters, e.g. ρ, varies, there may
cA (1 − α ) −α (δ + β ) exist a value ρH for which trJ*2 ( ρ H ) = 0 and det J*2 ( ρ H ) > 0 . When this
= kA − − ρ k 1A−α + ρ (1 + cA ) (39) happens, the dynamics of the system may undergo a qualitative change,
cA σ σ
known in the literature as Hopf bifurcation:19 the model economy does not also Barro and Sala-i-Martin, 1995 and Arnold, 1997), whereas Asada,
reach the steady-state growth rate, but rather persistently fluctuates around Semmler and Novak (1998) obtain the same result for the social planner
it. problem of the Romer model. Second, starting with the important
In terms of the parameters of the model, we have:20 contributions by Benhabib and Perli (1994) and Xie (1994), interesting
dynamics (included multiple equilibria, indeterminacy of equilibria and even
α k A*−2α + (1 − α ) ρ H k A*12−α the emergence of periodic solutions) has been shown for the market solution
trJ*2 = 0 ↔ c*A2 = (40)
(ρ H + k A*−21 ) of the Lucas model when externalities from human capital are considered.
Given the aim of the present chapter, we now turn to a brief description and
α k A*−2α + (1 − α ) ρ H k A*12−α analysis of this second set of results.
det J*2 > 0 ↔ c*A2 = In the model by Lucas (1988) the optimisation problem that has to be
(ρ H + k A*−21 ) solved by the representative agent is the following
α (1 − α )(1 − ρ H k ) + ρ Hσ k
* *
∞
⎛ c1−σ − 1 ⎞
> A2 A2
⎟ exp ( − β t ) dt
(41) max ∫ ⎜ (43)
ρ Hσ k *
1−σ ⎠
0⎝
A2 c ( t ),u ( t )
↔ − (1 − α − σ )αρ H2 k A*22 + ρ H σ k A* 2 (1 − α ) + α (1 − α ) < 0

k = k1−α (uh ) haγ − c h = δ h (1 − u )
α
subject to:
The existence of a Hopf bifurcation also requires that
k ( 0 ) = k0 > 0, h ( 0 ) = h0 > 0
d
trJ*2 ( ρ ) ≠0 (42)
dρ where c is per capita consumption, β, a positive discount factor, σ, the
ρ =ρH
inverse of the intertemporal elasticity of substitution, k, physical capital, h,
Using parameter values that satisfy conditions (41) and (42), numerical human capital, u, the fraction of labour allocated to the production of
simulations show that indeed the model may generate limit cycles which, in physical capital (so that uh is the fraction of effective labour), δ, a positive
addition, prove to be stable (see Greiner and Semmler, 1996a, pp. 91–96 and technology parameter, 1 − α, the share of capital, and γ, a positive externality
1996, pp. 111–16). The conclusion is that the model may generate persistent parameter in the production of human capital.
cycles in the growth rate of per capita variables, e.g. per capita output, For problem (43) we obtain the following set of necessary FOCs for an
namely, persistent growth cycles. interior solution:21
c −σ = λ1 (44)
16.3.2. Persistent Cycles in Two-Sector EG Models
α c −σ k 1−α hα +γ −1u −α
It is a fact that, with few exceptions of the kind considered in the previous = λ2 (45)
δ
subsection, the contributions on out-of-steady-state dynamics in EG models
have focussed on two-sector models, for both the case with human capital λ1 = βλ1 − λ1 (1 − α ) k −α hα +γ uα (46)
(see, for example, Caballé and Santos, 1993, Mulligan and Sala-i-Martin,
1993, Benhabib and Perli, 1994, Boldrin and Rustichini, 1994, Xie, 1994, λ2 = βλ2 − λ1α k1−α hα +γ −1uα − λ2δ (1 − u ) (47)
Barro and Sala-i-Martin, 1995, ch. 5, and Arnold, 1997) and with R&D (see,
for example, Benhabib, Perli and Xie, 1994, Asada, Semmler and Novak, whereas the transversality condition is22
1998, and Arnold, 2000a, 2000b). The result is a large body of literature in
lim exp ( − β t )[λ1k + λ2 h ] = 0
which it is possible to find some clear-cut results. First, no interesting out-of- t →∞
steady-state dynamics is usually found for the social planner solution of the
From (44)–(47), with simple algebraic manipulation, we derive the
models. This is shown, for example, by Caballé and Santos (1993) who find
dynamical system of the model. In order to do so, note that from (44) and
saddle-path stability for the Lucas model in the absence of externalities (see
(46) we obtain:
⎛ 1 − α ⎞ −α α +γ α ⎛β ⎞ α (β − δ ) δ (α + γ )
c = ⎜ ⎟ ck h u −⎜ ⎟c (48) x2* = 1 − , x3* = η x2* +
⎝ σ ⎠ ⎝σ ⎠ δ ⎡⎣γ − σ (α + γ )⎤⎦ 1−α
whereas from (45) and (47)
The study of the local stability of the linearised system leads to some
−σ c c + (1 − α ) k k + (α + γ − 1) h −1h − (1 − α ) u −1u = β − δ
−1 −1 major results that can be summarised as follows (see Propositions 1 and 2 in
Benhabib and Perli, 1994, p. 123–24)
Inserting in the latter expression (48) and the constraints of (43), we 1. For values of the parameters such that 0 < β < δ and σ > (1 + β / ψ ), the
finally obtain Jacobian of the system has one eigenvalue with a negative real part and
αψ two eigenvalues with a positive real part. In this case the competitive
x2 =η x22 + x2 − x2 x3 equilibrium path is locally unique: given the initial conditions of k and h,
1−α there exists only one value of u and c that drives the economy towards its
where x2 = u, x3 = c/k , η = (1 − α − γ)δ / ( 1 − α), ψ = −(α + γ)δ / α < 0. (determinate) steady-state path;
We then define 2. For some other values, namely for δ < β < −ψ, σ < (1 + β / ψ ) and
γ > 1 − α, one eigenvalue has a positive real part and two eigenvalues
x1 = h − (α +γ ) / α k have a negative real part. In this case there is always a continuum of
equilibria in the neighbourhood of the indeterminate steady-state path:
from which, taking the derivative with respect to time of both sides and given the initial conditions of k and h, there exist infinitely many values
inserting the two constraints of (43) of u and c that drive the economy towards the BGP;
3. When β and σ are as in the previous case, but 0 < γ ≤ 1 − α, there is
x1 =ψ (1 − x2 ) x1 − x1 x3 + x1 x2α
either (i) one positive eigenvalue and two eigenvalues with negative real
Finally parts or (ii) three eigenvalues with positive real parts. In the latter case
there is complete instability and no equilibrium path leads to the BGP.
x3 = x32 + φ x3 x1−α x2α − ξ x3
As stressed by Benhabib and Perli (1994, p. 124), of particularly interest is
where φ = ⎡⎣(1 − α ) / σ ⎤⎦ − 1 and ξ = β / σ . Case 3, in which there is a basic change in the roots structure of the Jacobian.
This happens when, as one of the parameters (for example, the externality
Thus the results of Benhabib and Perli (1994) can be discussed by parameter γ) changes, the real parts of the two complex eigenvalues change
analysing the following reduced dynamical system sign. At the value γ = γH for which the real parts of the two complex
eigenvalues are zero the system may undergo a Hopf bifurcation such that it
x1 = x11−α x2α +ψ (1 − x2 ) x1 − x1 x3 (49) persistently fluctuates around the steady-state path. However, although this
possibility is mentioned by Benhabib and Perli (1994, p. 124), they do not
⎛ ψα ⎞ investigate the matter any further but rather concentrate on the indeterminacy
x2 = η x22 − ⎜ ⎟ x2 − x2 x3 (50)
⎝ φ ⎠ results.
An interesting analysis of the case is instead contained in a recent
x3 = x32 + φ x1−α x2α x3 − ξ x3 (51) contribution by Mattana and Venturi (1999), where the authors, after a
general presentation of Benhabib and Perli’s results, concentrate on the
the steady-state values of which are23 possibility of emergence of periodic solutions and achieve results worth
−1/ α
mentioning. They establish analytically the existence of closed orbits (see
⎡ (1 − α )ξ − δ (α + γ ) − δ (1 − α − γ ) x2* ⎤ Theorem 1 in Mattana and Venturi, 1999, p. 270) for the reduced system
x1* = ⎢ ⎥ x2*
⎣⎢ (1 − α )φ ⎦⎥
(46)–(48). Moreover (see Theorem 2), they establish, with the help of
numerical simulation, that the closed orbits emerging from the steady state
can be either sub-critical (i.e., repelling) or super-critical (i.e., attracting). In

∑ (1 + n ) {ln c + φ V ( lt )} dt
t
∞ t
max W = β
particular, when σ is ‘small’, the numerical simulations show that the {c t
, lt }
∞
t=0
t =0 t
supercritical seems to prevail.

subject to the intertemporal budget constraint
kt +1 = ( At zt ht ) k t − ct + (1 − δ − n ) k t
1−θ θ
16.4. CONCLUSIONS
In this case, writing βˆ = β (1 + n ) and δˆ = δ + n , the problem becomes equivalent to (3).
In this paper we explored the possibility of generating persistent growth 5. The details of such a procedure are in an appendix available upon request.
cycles in a market-clearing framework, distinguishing between stochastic 6. The purpose in filtering simulated series is to make them comparable with the actual time
and deterministic models. As regards the first class of models, we focussed series, that is to consider cycles of the same frequencies for both.
7. As shown by Christiano and Eichenbaum (1992), this result holds for a wide range of
on the novelty originating from the new growth theory. In particular, we
parameters.
stressed both the role played by monetary factors and the consequences of 8. β depends on the interest rate and, in a model with capital accumulation, would be an
modelling R&D activity. With regard to the second class, we analysed the endogenous variable; however, in the framework considered by Stadler, without fixed
conditions for the emergence of periodic orbits in EG models which, in such capital, the constancy of β does not appear a restrictive assumption.
a framework, represents growth cycles. 9. To avoid scale effects, which are typical in endogenous growth models, we suppose that
every firm has a positive externality proportional to per capita capital knowledge stock
The inverse empirical relationship between productivity and output is a
available in the economy.
major challenge for business-cycle researchers. The representative agent 10. A possible microfoundation of such labour supply can be performed by considering a
framework, which is common to all contributions considered in this chapter, representative agent that maximises an instantaneous utility function whose arguments are
does not appear well-suited to this goal. As suggested by Lippi (1993), this consumption and leisure; moreover labour is assumed to be paid to the real wage rate w/p,
may require the consideration of the dynamics of the productivity of firms which represents the only source of income for the agent. The choice of the labour supply
(10) is due to its analytical tractability. We also notice that the total amount of labour which
and of their interaction.
is supplied has no upper bound; a more plausible formulation would entail total available
Finally, analysis of out-of-steady-state dynamics in EG models has shown labour being fixed and finite. For example a possible formulation for the labour supply
would be ht = [1 + ( wt / pt ) ] L , which is the solution to the following consumer
−ν −1
interesting results, but the latter appear to be applied to a very narrow set of s
problem: U = ct ( L - ht ) , s. to ct = ( wt / pt ) ht .
ν ν
EG models. A direction for future research is to explore the possibility of
periodic orbits in other types of deterministic EG models, such as R&D 11. The increase in productivity, however, could also be the result of employing in the
production process new resources brought onto the market (see Stiglitz, 1993). We analyse
Schumpeterian models, in an attempt to bridge the gap between stochastic
this more general case in an Appendix available on the authors’ websites, in which the firm
and deterministic approaches. can choose between these two methods of increasing its productivity.
12. Problem (21) is derived from the standard Bellman equation for a firm whose goal is to
maximizes its value, given by the discounted sum of future expected profits at rate r.
13. Aghion and Saint-Paul (1998b) highlight how the presence of a liquidation value
NOTES independent of productivity can generate an exit effect, which leads firms to underinvesting;
for the sake of simplicity we ignore this aspect.
1. In another contribution to this volume, Sordi analyses the problem of the interrelation 14. Notice that by restricting our attention to symmetric equilibrium we are excluding analysis
between growth and cycle in a non-market clearing framework. of the dynamics of cross-section productivity along business cycles. A number of other
2. Frisch (1933) argued that his theory of the business cycle was supported by the empirical contributions focus on this point. See, for example, Caballero and Hammour
Schumpeterian idea of innovations as a cause of economic fluctuations. However, he (1994) and Davis and Haltiwanger (1992).
thought that Schumpeter was considering as exogenous the process of innovation, while 15. In the formualtion of the firm problem we have implicitly assumed that there is no ‘exit
Schumpeter was actually distinguishing between scientific discovery, not driven by effect’, in other words, that the value of the firm in recession is never below τ C . See Aghion
economic forces, and innovation, i.e. the economic implementation of a scientific discovery, and Saint-Paul (1998a, p. 326).
which depends on economic factors. 16. The relation between this formulation and the original approach by Romer (1986a) is
3. The form of this stochastic process is suggested by the empirical evidence about the discussed in Greiner and Semmler (1996a, p. 82).
presence of a unit root in time series. 17. As usual, it is assumed that the per capita utility function is such that u′(·) > 0, u″(·) < 0,
4. We assume that population is constant. In the case in which the population increases at a l i m c → 0 u ′ ( · ) = 0 , u ″ ( c ) c / u ′ ( c ) ≡ − σ , constant.
constant rate equal to n, the representative agent’s problem becomes: 18. Clearly, conditions (33)–(34) are also sufficient for an optimum if the following
transversality condition is satisfied
Real business cycle models, endogenous growth models and cyclical growth 329
( )
lim t →∞ exp [− ( β − n ) t ] λ ( t ) k (t ) − k (t ) ≥ 0
*
*
where k is the optimal value of per capita capital stock.
19. For an introduction to the concept of Hopf bifurcation, see Gandolfo (1997, ch. 25).
20. From condition (41) it follows that α + σ < 1 is a necessary condition for a Hopf bifurcation
to occur. 17. Growth theory and the environment:
21. To obtain these conditions we have taken account of the fact that, in equilibrium, we must
have h = ha. how to include matter without
22. Benhabib and Perli (1984, p. 117–18) have shown that the maximised Hamiltonian is jointly
concave in (k, h). This implies that the above conditions are also sufficient for problem (43). making it really matter
23. It is possible to show that these values satisfy the transversality condition. See Benhabib and
Perli (1994, p. 122).
Tommaso Luzzati
17.1. INTRODUCTION
It is sometimes useful to start by asking why questions are asked. In this
respect, the question of the limits to growth is a very ‘natural’ one. Our
everyday experience is that growth eventually ceases in any process, at least
as concerns natural processes. At a more abstract level, this has a
correspondence in sciences, where growth processes are modelled or
conceived as exhibiting logistic (or similar) trends. Examples ranges from
biology, e.g. cell cultures, to business analyses, e.g. the life-curve of products.
Two mechanisms that end growth in natural processes are easily
identified. One is due to the lack of nutrients/inputs or other external factors
feeding the process. The other one is internal, the inhibition that occurs due
to (by) products of the process, a ‘poisoning’.
As regards environmental limits to economic growth we can find both
types of limits. Scarcity of resources might leave production without its
material basis, while waste and pollution might ‘poison’ the environment
within which the economic process takes place and make further economic
growth undesirable.
The classical economists perceived the limits from natural inputs – the
times were not ripe enough to think of a ‘self-poisoning economy’ due to
environmental degradation. What is relevant here, however, is that the
question of ‘limits to growth’ was central to their research agenda. Their
analyses of the increasing size (and structural changes) of their economies
also included the causes that would have stopped it. That growth would have
stopped was not in dispute. Ricardo, for example,
disliked the idea of the stationary state […] and saw two factors that might, at least
temporarily in his view, delay the stationary state. The first was international trade
[…]. The other […] was technical change […]. However, Ricardo, like
330
Growth theory and the environment 331 332 The Theory of Economic Growth: a ‘Classical’ Perspective
contemporary theorists of limited economic growth, viewed both free trade and pendence between economic processes and ecosystems failed to receive
technical change as only temporary stopgaps delaying, but not preventing, the much attention (Common, 1997).
arrival of the stationary state (Foley and Michl, 1999, p. 161) Consequently, the problem was initially conceived and modelled by
economists as being mainly one of economic growth with exhaustible
Industrial revolution – its deep specialisation, its strong urbanisation and its resources. Such a problem fitted well within the mainstream approach, of
opulence – dissociated ‘Western man’ progressively from his environment which a major issue is the allocation of scarce resources. The answer within
and from nature in general. This affected many economists too as, from the that framework was substitution of scarce inputs of production via changes in
neoclassical revolution onwards, most economic theory (see Martinez Alier, relative prices. Limits to growth would not be binding if, in the long run,
1987 for the exceptions) disregarded the question of the material basis of exhaustible inputs were (markedly) substituted. In a market economy, the
economic process. After a century man looked so powerful that economists occurrence of such a process was seen as highly probable. Any increasing
had an almost unanimous reaction when ‘The Limits to Growth’ (Meadows scarcity of particular materials would lead to a progressive increase in
et al., 1972) appeared. They regarded such a question as not to be put on the relative prices. Forced by this pressure, technology and science would find
research agenda since, consistent with their experience, they perceived better extraction techniques, new inputs, and new productive processes. At
unlimited economic growth as obvious. This was a U-turn with respect to the same time, progress in abatement techniques would reduce pollution.
classical economists. Technological progress actually occurred, materials reservoirs were
Forced to tackle the question, mainstream economics gave its answer shown to be greater than expected, new materials and processes are being
shortly after the publication of the report in question. The analytical continually introduced, production is often cleaner. However, the argument
framework was the well-established neoclassical growth theory while the still looks weak as nothing guarantees that progress will be strong and quick
focus, consistent with the ongoing debate, was on exhaustible resources. The enough to prevent resource scarcity from becoming binding. To believe that
new growth theory has tackled the question again since the 1990s. However, ‘salvation’ will come from advancement in technology is an extrapolation of
in a world where waste and pollution are leading actors, economists what occurred in the recent past based on static expectations, resembling
understood that lack of inputs can be a minor problem as compared to more an act of faith in human power than a scientific argument.
environmental deterioration. As a consequence, the environment is modelled Doubts in that faith also emerge when looking at the formal models built.
mainly as a renewable resource and beliefs are more cautious, that is, the One reason for being suspicious is the simplistic view of the environment as
existence of environmental limits is considered at least as possible. a mere resource input to production. On the contrary, the environment and
Does the mainstream growth theory hit its target, to show that ‘there ecological systems provide many functions for humans and life in general, in
might actually be no environmental limits to economic growth’? The present a setting where there are sizeable risks of irreversibility, thresholds and
paper aims to answer this question. catastrophes. Thus, technological progress needs not merely to find
substitutes for resources, but also to offset the general damage generated by a
highly disturbed environment on human society. The analysis should then
17.2. THE FIRST REACTION TO ‘LIMITS TO GROWTH’ also consider the possibility of a halt due to ‘poisoning’. Another reason
concerns the realism of the conditions that are found to be necessary for
It is well known that the report ‘Limits to Growth’ provoked strong criticism unlimited growth. As we will see, this applies also to the analyses developed
adverse reactions from many economists who considered it nonsense (see, in the 1990s within the New Growth Theory.
e.g., Beckerman, 1972). If they did not show methodological consistency,1 it The first strand of literature is well represented by three seminal papers,
is also true that much criticism was forcefully argued. At the same time it Dasgupta and Heal (1974), Solow (1974), and Stiglitz (1974), published in a
was mainly based on the narrow interpretation of the report that prevailed in special issue of the Review of Economic Studies. They develop one-sector
the general debate. In particular the focus remained on its quantitative models of neoclassical growth with exogenous technical progress where,
predictions (interpreted as actual rather than ceteris paribus predictions) and along with capital and labour, also a depletable resource is included in the
on the problem of resource scarcity (it was about the time of the first oil production function. Production is specialised with CES functions. Utility
crisis). Both the qualitative mechanisms outlined in the analysis and its depends only on consumption. A central planner maximises the present value
attention towards the issue of pollution as well as the general interde- of the utility function of the representative agent.
The results depend crucially on the elasticity of substitution between development suggested the possibility of making market growth and the
capital services and the resource. For elasticity greater than one, production health of ecosystems compatible. ‘No environmental limits to growth’ was
is possible even without the resource. The depletable resource does not not warranted anymore and became rather a desired goal. In this new
constitute a limit since it can be progressively substituted by capital. For atmosphere it was possible for a joint group of economists and ecologists
elasticity less than or equal to one these models allow sustained utility or (Arrow et al., 1995) to work together and agree that there are ‘limits to the
even optimal growth only in the presence of a strong enough and increasing carrying capacity of the planet’ (Ibidem, p. 520).
resource-augmenting technical progress (see Stiglitz, 1974).
Neither way to circumvent the problem seems very convincing. It is hard
to think that the whole economic process could run without matter/energy. 17.4. NGT AND THE ENVIRONMENT
On the other hand, with Toman et al. (1996, p. 146), one can doubt
When introducing the chapter on growth and the environment in their
whether it is realistic to make such a conception of technical progress that
squeezes a constant flow of [...] services out of an evershrinking flow of resource textbook Aghion and Howitt claim that endogenous growth theory
service inputs. is also inherently more suitable for addressing the problems of sustainable
development than is the neoclassical theory, because whether or not growth can be
In both cases the issue is whether it is realistic for the amount of used
sustained is the central question to which endogenous growth theory is addressed
resources to tend asymptotically to zero, which also entails a ratio of
(Aghion and Howitt, 1998, p. 151).
‘exhaustible resource to Income’ that tends to zero. The basic requirement
for unlimited growth in this class of model is therefore a very simple one, a Nonetheless a quick look both at reference databases and at textbooks reveals
progressive decoupling of income from its material basis. that the environmental question is not so central in the research agenda of
new growth theory. Most recent surveys include the above-mentioned
chapter by Aghion and Howitt (1998), and Smulders (1999). These surveys
17.3. A BETTER REPRESENTATION OF THE contain two archetypal models used by new growth theory to take the
ENVIRONMENT environment into account. This section will first describe and compare them,
and then attempt to provide a critical assessment.
From the 1970s, economics gradually broadened its view about the relation The first framework was set forth by Stokey (1998), in a paper aimed at
between the environment and the economy. First, pollution started to be providing analytical foundations for the Environmental Kuznets Curve.2
included in the models (e.g. Forster, 1980). More importantly, the general Aghion and Howitt further elaborated on it. In this framework the critical
interdependence between ecosystems and economies started to become environmental factor is pollution. Pollution (treated as a flow or as a stock)
accepted. A new discipline was born towards the end of the 80s, Ecological negatively affects utility while being a joint product of production.3 A
Economics, in order to take explicitly into account this interdependence. In growing income will then increase both consumption and pollution.
general terms, it was acknowledged that the environment is not merely a Whenever, as reasonable, the welfare increase due to an increase in
source of resources. The environment provides services of waste absorption consumption is more than offset by the welfare loss due to an increase in
and general ecosystem maintenance; it directly enters the utility function pollution, then economic growth will become undesirable. The way out that
both due to its amenity value and its effects on health, and it affects Stokey suggests is a progressive shift towards cleaner technologies, whose
production. Moreover, pollution abatement is a major economic activity. adoption, however, is costly as it entails reductions of output.
At the same time, confidence in the non-existence of environmental limits The technology side is as follows:4
to growth started to decline. The progressive deterioration of the Y=f(.)z (1)
environment, the appearance of the idea of sustainable development (WCED,
1987) and, perhaps, the collapse of the Soviet Union, raised the issue of the Where Y is income z, z∈[0,1], is a (direct) index of pollution intensity of
possible environmental bankruptcy of the market economy. Environmental production, f(.) is potential income. This equation reflects the assumption
degradation was becoming increasingly evident, global, and damaging, the that cleaner technologies entail costs.
market economy was no longer under discussion, the notion of sustainable
Pollution,5 X, is assumed to be an increasing and convex function of N = g(N)–R with g ( 0 ) ≤ 0 , g N N < 0 (4)
actual output, given potential output f(.). Stokey uses the following
Despite appearances, this is not very different from the Stokey/Aghion–
X= f(.)zβ with β >1 (2) Howitt framework. When modelling pollution as a stock, P, its dynamic will
be ruled by the absorption capacity of the environment, h(P), which again
By combining the two equations, actual output can be obtained as can be assumed to show a logistic trend, and by the rate of pollution (i.e.
pollution added by new production). Thus, one can write a differential
Y=f(.)1–1/β X1/β (3) equation9 analogous to (4)
which highlights the role of pollution as a production factor in this model.
P = h ( P ) – X ( Y, z) with h ( 0 ) ≤ 0 , h N N < 0 (5)
A standard utility function, whose arguments are consumption and
pollution,6 and specific functional forms for potential output, f(.), complete
The analogy goes beyond mere formality as pollution stock is in fact
the model. Stokey analyses two different functions for f(.), an AK production
inversely related to environmental quality. Moreover the rate of pollution, X,
function, f(.) = AK, and a Cobb–Douglas with exogenous technical progress,
is positively related to the extraction rate, R, as, due to the law of
f( . ) = A K α e g t with 0 < α < 1 .
conservation of mass and the degradation of matter caused by the economic
To get an intuition of the outcome of the model it must be noted that the
process, pollution is by the amount of material inflows (extraction).
marginal product of capital is MPK = B Y/K, where B = ( 1 – 1 / β ) in the AK
With regard to preferences, Smulders assumes, as usual, that the
case, and B = α ( 1 – 1 / β ) in the Cobb–Douglas one. If f( . ) = A K then the
representative agent has a utility function that depends positively both on
average product of capital is Az, implying MP K = ( 1 – 1 / β ) A z . As optimal z
consumption, and on environmental quality.10
exhibits a decreasing path,7 the MPK will fall below the rate of time
Production, Y, is a positive function of extracted resources, state of the
preference and make investment not attractive anymore. Growth, while being
environment, and capital. All inputs are essential. The long-run marginal
technologically feasible (it is possible to choose a rate of change of z such
productivity of capital does not go to zero. As usual, production is partly
that output grows and pollution declines), is not optimal. On the contrary, in
consumed and partly accumulated within new capital. Capital, H, includes
the presence of exogenous technical progress the optimal outcome can be
both physical and human capital
unbounded growth. This is because the average product of capital,
M P K = α ( 1 – 1 / β ) Y / K , is constant along the balanced growth path. Similar Y (N, R, H) (6)
conclusions are obtained within the endogenous growth Schumpeterian
framework developed by Aghion and Howitt (1998, p. 151–71). with Y ( 0 , R , H) = Y ( N, 0 , H ) = Y ( N , R , 0 ) = 0 and YN ≥ 0, YR > 0, YH > 0
A distinctive feature of Stokey’s paper is that pollution can increase Given this setting, unlimited growth with a non-deteriorating environment
without bounds. This is unrealistic due to the existence of critical ecological can be optimal (in the standard representative agent framework and
thresholds ‘below which environmental quality cannot fall without starting in depending on the utility function) if the economy lives off a constant level of
motion an irreversible and cumulative deterioration entailing a prohibitive ‘extraction’ which is consistent with ecological stability, R=E(N) (see eq.
cost’ (Ibidem, p. 157). In the presence of such thresholds, income can grow (4)). Constant levels of both environmental quality and resources enter
without bounds only if optimal pollution intensity, z*, tends to zero at an production while the economy can be optimally fuelled by levels of man-
appropriate speed (see eqs. (1) and (2)). With a CES utility function made capital, which are optimally increasing in virtue of the absence of
additively separable in consumption and pollution, this is optimal only if the decreasing returns. This is illustrated by a simple example Smulders (Ibidem,
elasticity of the marginal utility of consumption is greater than one. The p. 613–14).11 illustrates this through a simple example where constant
same holds for the Schumpeterian model (see Ibidem, p. 161). returns to human capital are assumed. In this case the production function
Smulders’ (1999) archetypal model starts by modelling the environment becomes Y ( N , R , H ) = y ( N , R ) H so that the whole model ends up as an AK
as a renewable resource. It assumes that it is possible to define a variable, model. If, for simplicity’s sake, a constant saving propensity is assumed, s,
environmental quality, N, that, as usual, follows a spontaneous logistic the accumulation equation becomes + = s y( N , R ) H . Dividing by H, the
growth trend8 that is altered by ‘extraction’ of resource, R, for production long-run growth rate of a balanced growth path is obtained.
purposes.
This brief analysis has shown that the two frameworks described in this must be bounded from above so that growing income will make the ratio
section differ only with respect to the ‘extended’ production function, i.e., ‘Material Inflows’/’Income’ fall asymptotically to zero. As we will see in the
the production function that also includes the ‘use’ of the environment. Such next section, this requirement (which is similar to that found in the
differences explain the different theoretical predictions. neoclassical growth theory with depletable resources, see Section 17.2.) can
It may be interesting to look at the mechanisms behind the two production be obtained in a simpler way.
functions. With reference to Stokey/Aghion–Howitt one can raise the doubt,
as an external critique, that technology at the level of the whole economy can
be considered as a choice variable. ‘Production’ of new technology needs 17.5. BOUNDED MATERIAL INFLOWS
time and is the outcome of an evolutionary process. When looking
specifically at the model, equation (1) does not seem to have robust As Georgescu-Roegen emphasised, the economic process consists in
justifications since a cleaner technology does not necessarily entail a cost, at dissipating matter and energy, that is, in producing waste. Thus, material
least in static terms. This is true both at the individual level (e.g. a single inputs13 matter not so much as input for production, but as they start
production process that is less polluting due to a reduced use of materials is becoming waste from the very beginning of their use.
actually cheaper) and in terms of aggregate output (e.g. abatement costs enter Bearing this in mind, it is easy to understand how economic growth can
GDP, contributing positively to the relationship between cleaner technology occur. Let I be the impact, the harmfulness, of the human system. It is
and higher income). Secondly, as emphasised by Aghion and Howitt reasonable to impose, I ≤ I as beyond I the impact on human welfare is
M M
themselves (Ibidem, p. 162), the above-mentioned condition on the elasticity undesirable and/or the ecosystems enter a catastrophic involution. z is the
of marginal utility of consumption (>1) is problematic in many macro- ‘dirtiness’ of the technology, bounded to be strictly positive for the reasons
economic models. Finally, as for neoclassical growth theory (see Section seen before, z ≥ z >0. M is total material use. As in the debate during the
m
17.2) with exhaustible resources, doubts can be raised about the consistency 1970s, let the impact be given by the product of some measures of the
of the model with physical laws. Actually it is particularly striking that the material scale, M, and the state of the technology. Then the following
pollution intensity for the whole economic process is assumed not to be inequalities hold:
bounded from below.
Smulders’ archetypal model is neater as it allows us to go straight to the IM ≥ I = Mz ≥ Mzm (7)
central issue for unbounded growth, the need for the economy to be run by
implying an upper bound to M, M ≤ I M /z m . M is unbounded only if one
processes whose interference with the environment becomes small to the
assumes there is no upper bound in I, or if zm tend to zero, exactly those
point of disappearing. Smulders directly assumes that production does not
hypotheses used in the Stokey and Aghion–Howitt approach.
necessarily entail pollution as a joint product. Given a minimum requirement
In a more descriptive way the need to decouple matter and income can be
of natural inflows, R, production can be increased indefinitely by increases in
summarised as follows. Material inputs, after remaining for a while within
(human) capital, H, is assumed not to affect the environment.
the economy, go back to nature as waste. This affects the natural
The following conclusion can be drawn. Although endogenous growth
environment in ways that can be harmful to human welfare both directly and
models avoid simplistic representations of the links between the economy
indirectly via reduced productive efficiency:
and the environment, the conditions for unlimited growth are built on
attempts to break exactly those links, that is, on attempts to decouple matter a b c
from the economy. This is obtained in one case, Stokey and Aghion–Howitt, INFLOWS OUTFLOWS(waste) HARMFULNESS
by assuming that technology can make production non-polluting,12 and in
There are three ways to reduce the negative consequences on welfare. First,
the other, Smulders, by directly assuming that production need not affect the
one can break the link ‘c’, between waste and its harmfulness by improving
environment.
the ‘quality’ of waste, thanks to ‘end-of-pipe’ tools, such as treatment of
As it is undesirable to have ever-increasing pollution, and as there are
toxic waste, emission abatement, and so on. Second, there is some space for
limits to the cleanness of average technology, the conditions for continual
increasing the persistence of matter within the economic system (arrow b),
growth identified by the new growth theory can be summarised by the idea
by promoting, for example, product durability, repairing, and recycling.
that the economy must become decoupled from matter. Material inflows
Third, less material can be used (arrow a) thanks, for example, to increases to decouple. This feeling underlies, for example, relevant empirical research
in material efficiency or in durability (b). that is ‘desperately seeking (environmental) Kuznets’.16 The idea we are
If material inflows/outflows are left to follow the growth of income, given entering a ‘weightless economy’17 is due both to progress in pollution
the limits to the possibility of preventing goods deteriorating into waste, and abatement techniques and the appearance of many new products of the
the limits to the ‘regenerative’ capacity of nature, the improvement in ‘end- ‘knowledge economy’, where high value is embedded in a few bits or in
of-pipe’ technologies must make the average impact per unit of waste tend to some lines of code.
zero. The lack of plausibility of an economy running on ‘non-polluting’ The fact that this is probably an illusion may be easily understood by
technology has already been discussed in the previous section. Consequently, examining the literature on the EKC (see references mentioned in note 2) or
if income has to grow indefinitely (and as long as our environment remains by looking at the increasing trends in material requirements of the developed
our planet), material (and energy) throughput of the economy must be economies (see, e.g., Adriansee et al., 1997). The reasons for this illusion are
bounded from above. In other words, income has to become decoupled from manifold. For example, although some new products are almost immaterial,
its material basis and the ratio material input / income has to tend to zero. it cannot be forgotten that their consumption needs complements which are
The answer to the question whether it is possible for income to grow highly material. Software alone is useless, we need hardware, whose
indefinitely given a constant amount of material inflows ‘lies in what we production requires a large quantity of material inflows, which soon becomes
mean by output’ (Petith, 2001, p. 15). Somewhat surprisingly, this question obsolete due to progress in software itself. Another reason is the small size of
has not received much attention in the debate, including the recent forum many end-products, which is often misleading as small size does not
‘Georgescu Roegen vs Solow/Stiglitz’ organized by Daly on Ecological necessarily entail low material inflows. Generally speaking, the wrong
Economics in 1997 (issue 22). Actually physical output, income and welfare perception seems to be arising from a mistake in shifting from one
are almost used as synonyms while, on the contrary the real issue is how our hierarchical level to another, as small improvements visible at the individual
society maps physical output both onto income and onto value/welfare. Prof. level are believed to hold also at the level of the whole economy (see, for
G. Fuà, in his last masterpiece, taught us how problematic it is to look for a example, Giampietro and Mayumi, 2000).
mapping between income and welfare in rich countries (Fuà, 1993). To add Future trends of total throughput are uncertain. There could be a reduction
‘matter’ to the picture looks even more problematic. in the throughput of developed countries, although such a reduction would
However some considerations can be made on what is occurring and its easily be more than offset by the increase in the throughput of the rest of the
perception. As regards perception it is illustrative to look again at Smulders world as and when it starts growing.
(1999, p. 610) and read in the introduction that
environmental and natural resource constraints did not turn the historical growth 17.6. CONCLUSION
process into stagnation. Instead, accumulation of human knowledge […] allowed
the economy to expand within the fixed physical system of the earth. […] (man)
Aghion and Howitt claim that
continually creates new knowledge to derive more value from a given amount of
physical resources (emphasis added). … endogenous growth theory […] does imply that with enough innovations, and
the right direction of innovations, such an outcome (sustainable development) is at
Such a position reveals firstly a sort of ‘linear thinking’ as it seems to least within the realm of possibility (Aghion and Howitt, 1998, p.151, emphasis
suggest the prediction of perpetual growth merely based on the (relatively added),
recent) past. More importantly, the available amount of physical resource is
not distinguished by the amount of resources actually used. This is not to while Stokey (1998) is even more optimistic about environmental limits to
deny that knowledge can be considered the engine of growth. However, the growth. However, for the reasons seen above, the conditions that the growth
fuel has been the dramatic growing amount of materials (particularly fossil theory sees as necessary for unbounded growth do not look very much
energy materials14) occurring in the past two centuries. More value has been ‘within the realm of possibility’. This is not a novelty in economics, as both
created from a growing, rather than a given, amount of physical resources. rationalisation rather than explanation is often the theoretical purpose and the
However, Smulders cannot be blamed for the above statements as they realism of hypotheses is seldom an issue.
constitute a shared perception,15 the perception that the economy has started
Growth theory and the environment 341
Moreover, there is a problem of methodological consistency. The

principle of Occam’s razor, often invoked by economists, constitutes a
problem for most of the literature surveyed here. A major outcome of its
elegant formal models and optimisation techniques, unbounded growth, can
be obtained in a much simpler way. The obvious necessary condition for
unbounded growth is to make the economy ultimately grow out of a constant
18. Modelling growth and financial
material throughput. Whether or not this can occur is an open question. intermediation through information
However, we need to be conscious that it is not true that economic growth
was supported by a finite amount of physical resources. Material throughput frictions: a critical survey
(especially fossil fuels) has dramatically increased and is still on the increase.
Salvatore Capasso
NOTES
18.1. INTRODUCTION
1. As was emphasised by Georgescu-Roegen (1976, p. 21–2) much of what neoclassical
economists said could also have been used against their analysis.
2. The EKC is a supposed hump-shaped empirical relationship between income and pollution, Considerable empirical evidence has shown strong linkages between real and
a relationship that, despite claims by the author, is far from being proved (see, for example, financial development. As economies grow, the relative size and complexity
Stern, 1997; de Bruyn and Heintz, 1999; de Bruyn, 2000). of financial systems tend to increase. New markets and financial instruments
3. Wastes could affect production so greatly as to make growth not even feasible. However, as develop, while the role of financial intermediaries tends to change. Financial
the most immediate effects of pollution are on welfare, the focus is here (as in the literature)
intermediation, very limited in the early stages of economic development,
on the optimality of growth.
4. Notation in Stokey is slightly different. becomes increasingly important with economic growth. However, as
5. Alternatively, ’increase in pollution‘ when pollution is modelled as a stock. economies continue to grow, better organised financial markets facilitate the
6. The utility is assumed additively separable in consumption and pollution. direct transfer of resource between lenders and borrowers: stock markets
7. Stokey shows that optimal z is decreasing in potential outcome, f(.). develop and financial intermediaries play a decreasing role, in relative terms,
8. As E is continuous and E(0) ≤ 0, the existence is admitted of a threshold below which the
in the credit market.
renewable resource enters a process of progressive deterioration.
9. Actually, Aghion and Howitt use an equation similar to (4) except for using ‘pollution’ The idea that financial markets affect the real allocation of resources and
instead of ‘extraction rate’. influence capital accumulation and growth is a very old one in economics.
10. U(C, N) with U(C,0) = U(0, N)= -∞ and Uc > 0 , UN ≥ 0 Bagehot (1862) firmly believed that capital in England was more productive
11 If constant returns to human capital are assumed the production function becomes than in other countries because, in England, larger and better organised
Y ( N , R , H ) = y ( N , R ) H (the whole model ends up as an AK model). With a constant saving
capital markets were channelling resources towards more productive
propensity, s, the accumulation equation becomes H = s y ( N , R ) H . Dividing by H, the
long-run growth rate of a balanced growth path is obtained. investments. Schumpeter (1934), on the other hand, stressed the role of
12. More precisely, in Stokey matter does not matter also in another way, as, depending on the financial intermediation, and in particular of banks, in improving resource
utility, pollution can grow without bounds without consequences for ecological systems. allocation and enhancing the aggregate productivity of capital. More
13 Material inputs are to be considered key indicators of the interference provoked by man on recently, Hicks (1969), in highlighting the importance of financial markets,
natural processes, as is particularly emphasised by research on dematerialization (see, e.g.,
suggested that the industrial revolution was not the result of innovations and
Schmidt-Bleck, 1994) which started in the 1990s at the Wuppertal Institute for Climate and
the Environment. the development of new technologies, but rather the result of the expansion
14. See, e.g., Cohen (1995). of financial systems that allowed the applications of these technologies.
15. The same confusion between available and actually used resources is also in Aghion Howitt, In recent years, new research has attempted to provide a rigorous
1998, p. 151: ‘If it had not been for resource-saving innovations it is unlikely that our finite theoretical interpretation of the linkages between the real and financial side
planet could have supported the expansion in material welfare’ (emphasis added).
of the economy. In the wake of the works by Gurley and Shaw (1955, 1960,
16. This is the title of a recent paper by Galeotti and Lanza (1999).
17. The term ‘weightless economy’ is used, e.g., by Quah (1999).
1967), Goldsmith (1969) and McKinnon (1973), a great number of studies
342
Modelling growth, financial intermediation and information frictions 343 344 The Theory of Economic Growth: a ‘Classical’ Perspective
have attempted, in the last two decades, to give a theoretical explanation for quite different analytical structures. Moreover, such differences can be
the positive empirical correlation between financial development and growth. magnified by the fact that the structure of the financial market becomes
In a standard Arrow–Debreu framework, in which markets are perfect, significantly complex under the assumption of information asymmetries,
agents are fully informed and there are no transaction costs, financial markets and, therefore, specific simplifying assumptions are required.
play no role in the allocation of real resources. In order to investigate the This chapter will review the literature on finance and growth, which is
channels of interaction between the financial and the real sector, economists fundamentally based on the assumption of asymmetric information, by
have modified this framework by introducing some kinds of market frictions, critically assessing the nature and consistency of the other most common
such as liquidity costs, transaction costs or imperfect information. Indeed, in assumptions. The objective is to clarify issues related to modelling
the presence of market frictions financial markets can affect the allocation of procedures, which might appear peculiar to these models, and to identify the
resources and the process of economic development. Thus, for example, similarities that lie behind apparently very different analytical set ups. Filling
financial markets can affect growth by reducing liquidity risks, or by an existing gap in the literature, this chapter provides a general representation
increasing the flow of savings and by channelling such resources towards of the latest results of the research on finance and growth. Attention is
more productive alternatives (Greenwood and Jovanovich, 1990; Levine, focused on the elements that characterise the optimal financial contract, the
1991; Bencivenga and Smith, 1991; Saint-Paul, 1992; Blackburn and Hung, structure of the real side of the economy and, most importantly, on the
1998 among others). Levine (1997) and Becsi and Wang (1997) provide an channels of interactions between financial and real sectors. In doing so, the
extended review of this literature. chapter highlights the analytical and conceptual mechanisms of interplay
Following the developments of the research in financial economics on the between financial development and growth and the possible avenues for
theory of the optimal financial contract, the most recent literature has focused further research in the area. It also provides a general illustrative model
on the assumption of information asymmetries between agents in modelling which attempts to capture the basic working of these models in a single
finance and growth. These studies integrate microeconomic models of general framework.
optimal financial contract under information asymmetry into dynamic The chapter is organised as follows. In Section 18.2 we analyse the credit
general equilibrium models, with very interesting results. market structure under information asymmetries. After classifying
In the presence of information asymmetries between lenders (typically information asymmetries into two broad categories (ex ante and ex post), we
households) and borrowers (typically firms) different informational problems examine, in each case in turn, the informational problems that might arise,
might arise, and the exchange of resources can become costly, sometimes to and the possible solutions in terms of the optimal financial contract. In
such an extent as to prevent capital markets from functioning. For example, Section 18.3 we describe the basic analytical structure and discuss the main
problems of adverse selection might arise when firms have the possibility of assumptions in these models: agents’ endowments, technologies and
hiding their expected profits or their level of efficiency. Problems of moral investment opportunities. Section 18.4 studies the channels of interactions
hazard, instead, might arise because of the incentive of firms to misreport the between capital accumulation and financial development and the possible
actual return on their investments. These informational problems generate dynamics of the models. In Section 18.5 we present a simple general model
agency costs, and the financial contract is the result of agents’ attempts to which attempts to compact the different features of these models in a unique
reduce these costs. The contract, as well as the credit market structure, is, analytical framework. Some concluding remarks are included in section 18.6.
therefore, endogenously determined. The link between growth and credit
market structure arises because growth can affect the level of agency costs
and hence the financial arrangements, while the structure of the credit market 18.2. THE CREDIT MARKET UNDER INFORMATION
affects growth because it determines the amount of resources invested and FRICTIONS
the allocation of capital. Thus, for instance, capital accumulation can reduce
the level of credit rationing because it increases the cost of rationing. Credit The assumption that credit markets are characterised by strong informational
rationing, on the other hand, can affect growth because it reduces the amount asymmetries between borrowers and lenders finds large support in everyday
of savings channelled to investments. experience and has a stronghold in basic intuition. It is natural to think, for
Since these models are built on specific assumptions regarding agents’ example, that an entrepreneur seeking external funding knows much more
endowments and the nature of information distribution, they can display about his own activities and prospective profits than the bank or the investor
that is willing to supply the funding. In addition, this assumption is not a arise. Under information asymmetry, a single unique contract for all
secondary one and can have significant consequences for the agent’s borrowers (pooling equilibrium) might be unfeasible or, certainly, very
financial arrangements. In recent years, it has been widely shown that the costly since it risks attracting only bad or inefficient borrowers, driving the
optimal financial contract depends strongly on the nature of informational good or efficient ones out of the market (a typical lemons’ problem as
problems in the economy. described by Akerlof, 1970). It has been argued, for example, in an attempt
On logical grounds information asymmetries can be classified into two to explain credit rationing, that even in the presence of an excess demand for
categories. In the first, lenders cannot observe the borrower's type – i.e. the credit, banks might not find it optimal to increase the interest rate on their
borrower’s characteristics – (ex ante information asymmetry); in the second, loans, but prefer to credit ration, since higher interest rates might attract a
lenders cannot observe the outcome of the borrower's activity or action (ex greater number of riskier agents, with the result of obtaining a lower
post information asymmetry). The distinction is important given that the set expected return on the loans (Keeton, 1979; Stiglitz and Weiss, 1981).
of problems and the possible solutions arising in each case can be very Certainly, it is true that under the assumption of information asymmetry
different. Typically, ex ante information asymmetries generate problems of borrowers might have the incentive to hide their type and pretend to be some
adverse selection. Ex post information asymmetries, instead, generate other in order to obtain contracts with better conditions.
problems of moral hazard. As a consequence, the financial contracts that One possibility lenders have of solving ex ante informational problems is
originate in each of these cases can substantially differ, with important to design different contracts for each type and to make sure that each
implications for the interpretation of the co-evolution of financial and borrower will optimally prefer the contract designed for his/her own type
economic development, given that the relationship between the financial over all the others. A logical consequence of this one-to-one matching is that
sector and the real sector depends on the nature of the financial contract. the choice of the contract will be self-revealing of the borrower’s type. The
Admittedly, even though such informational problems are completely practical result is that borrowers are ‘separated’, i.e. grouped, according to
different and easily identifiable on theoretical grounds, in practical terms their type. Of course, for separation to occur borrowers must be
they often coexist and may be very difficult to disentangle. It is only for heterogeneous and possess specific features beyond the given differences in
reasons of clarity and simplification that much of the literature on finance their types. By exploiting those differences, contracts can be made more or
and growth has focused on one kind of information friction at a time. We less attractive for one type or another. Thus, if borrowers have access to
will now proceed to analyse each of these cases in turn. alternative investment opportunities with different rates of return and
therefore have different costs when denied credit, separation can be achieved
18.2.1. Models that Are Based on ex ante Information Asymmetries by means of credit rationing. In Bencivenga and Smith (1993), for example,
and Bose and Cohtren (1996, 1997), ‘bad’ borrowers (with lower expected
The typical framework with ex ante information asymmetries (Bencivenga returns on their project) also have a higher alternative cost when credit
and Smith, 1993, Bose and Cohtren 1996, 1997) involves the presence of rationed. As a consequence, the contract designed for ‘good’ borrowers can
different types of borrowers. The borrower’s type is the borrower’s private be made unattractive to bad borrowers by including a high enough
information and is identified with the ability of the borrower to repay the probability of credit rationing.
lender. Usually, this corresponds to the expected return on a project for The level of credit rationing will depend on the degree of information
which the borrower needs external funding. Even though the borrower’s type asymmetry, as well as on other factors, such as loan size and alternative
is unobservable, a solution to the informational problem can eventually be costs. It follows that if one or more of these factors change with capital
found given that lenders know the distribution of the different types in the accumulation, the optimal level of credit rationing will change as well. So,
economy and therefore have a knowledge of the expected returns on the loan for example, if the net return on the borrower’s project is increasing with the
which they will eventually issue. The possible solutions, however, will also loan size, the larger the loan size, the more costly it will be not to finance the
depend on other specific assumptions, such as the agent’s endowments, project – the cost of credit rationing. Yet the loan size, the amount of
opportunity costs, returns on technologies: assumptions that describe the resources available to be transferred, is in itself an increasing function of
general structure of the economy. capital accumulation.
The informational gap between lenders and borrowers generates agency Separation via credit rationing is only one of the possible channels
costs. Depending on the nature of these agency costs different contracts can through which ex ante informational problems can be solved. It is realistic to
think that very often lenders have the possibility to directly verify borrower’s Interestingly, it has been demonstrated (Townsend, 1979; Gale and
type through a costly process. Banks, for example, have all the tools to study Hellwig, 1985) that in a standard costly state verification framework (CSV),
the financial position of their clients prior to the issue of a loan. Based on this where agents are risk-neutral and monitoring costs do not depend on the
idea, Bose and Cohtren (1996, 1997) assume that lenders are endowed with a project’s expected return, the optimal repayment takes only the form of debt
screening technology which allows the borrower’s type to be determined – a fixed repayment independent of the project’s actual return. Thus in the
without uncertainty and to solve, at outset and radically, the informational standard CSV framework, apart from debt, there is no space for any other
problem. However, screening is costly and, consequently, it is not optimal financial instruments. It is clear that, in order to explain the evolution of
for lenders to screen in each and every contingency. The contract, therefore, other forms of financial markets and instruments, this framework needs to be
will involve a probability of screening that is just sufficient to deter bad or modified.
inefficient borrowers masquerading as good or efficient ones. Crucially, in Following extensive empirical evidence, the most recent research has
Bose and Cohtren the cost of screening depends on the loan size. This makes been oriented towards the description of more specific features of the
it clear that as the loan size increases because of capital accumulation, the linkages between financial development and economic growth. These recent
return on the screening contract will change. studies attempt to provide an account of the development of new financial
instruments that seems to be the result of capital accumulation. Economists
18.2.2. Models that Are Based on ex post Information Asymmetries have tried to examine how and why equity markets appear to develop
relatively late in the process of capital accumulation, what happens to the
The situation is completely different when borrowers are homogeneous – i.e. debt/equity ratio when growth occurs, and the new role of financial
they have the same type – but the return on their project is stochastic and intermediation following the development of more direct forms of lending,
unobservable; or, alternatively, when borrowers are homogeneous but it is such as stock markets or bond markets.
not possible to observe their actions. In these circumstances the lender will Along this line of research, Boyd and Smith (1996, 1998) show that by
face ex post informational problems. modifying the standard CSV framework with the assumption that borrowers
Under ex post informational asymmetries, the informational gap can be have access to lower return observable technology, as well as to higher return
filled if lenders have access to costly monitoring technology that allows them unobservable technology, the optimal financial contract will involve not only
to observe the actual return on the borrowers’ projects, or borrower’s action. debt but also equity, a form of repayment that is a function of the project’s
As with screening, given that monitoring is costly, it can be shown that it is actual return.1 Most importantly, they show that while the issue of equity is
not optimal for the lender to monitor in each and every state. The financial associated to the use of observable technology, the issue of debt is associated
contract will therefore determine, together with all other contractual elements to the use of unobservable technology. As a result, an increase in monitoring
such as loan size and repayment, the contingency states in which monitoring costs, which makes unobservable technology more costly, will result in a
occurs. Monitoring, as well as screening, is technically the means by which more intense use of observable technology and, ultimately, in a
the contracts create the incentives for borrowers to truthfully reveal the proportionately higher issue of equity over debt. The positive correlation
hidden action or the actual return. between the equity/debt ratio and economic growth is explained (Boyd and
In Bernanke and Gertler (1989) the ex post agency cost can be lowered Smith, 1996, 1998) under the assumption that monitoring costs depend
when borrowers can provide collateral. As in a standard ex post information positively on capital accumulation.
asymmetries framework, lenders cannot observe the borrower’s actual As in the case when the return on the project is unobservable, it is
production, and hence the contract involves a probability of monitoring. The possible to show that debt still remains the dominant form of repayment
expected return on the borrower’s project, the cost of monitoring and the when the borrower’s action is unobservable, in which case a problem of
collateral determine the set of states in which monitoring occurs. Bernanke moral hazard might arise. Very interestingly, debt might lose its dominant
and Gertler show that if the value of the collateral changes, as happens position if there is more than one action that cannot be observed, and the
during business fluctuations, the probability of monitoring will change as lender faces multiple moral hazard problems. Blackburn, Bose and Capasso
well. Once again, this very intuitive result stresses the strict links between the (2001) show that under the assumption of two unobservable actions, the
real side of the economy and the financial structure. optimal financial contract is a combination of debt and equity. These authors
also assume that one of these actions can alternatively be observed via a
costly process. Therefore, lenders can either observe one action by sustaining Under a different perspective, the assumption of a double production
this cost, in which case the optimal contract is a debt contract; or they can process can be justified if one thinks of the economy as being vertically
leave both actions unobservable, in which case the optimal repayment will be integrated, with information asymmetries concerning only the first stage of
a combination of debt and equity. One can now understand that if this cost of production. Two goods, capital and consumption, are produced with two
observation is an increasing function of capital accumulation, the optimal different technologies, and the former is the intermediate good used in the
contract might be only debt for low levels of capital accumulation, and a production of the latter.
combination of equity and debt for higher levels of capital accumulation. A As outlined, since physical capital is used in output production, and since
different explanation of the co-evolution of stock markets and economic production of capital requires external funding, the optimal financial contract
growth thus emerges. and the structure of the credit market will affect output production by
affecting the flow of capital produced. At the same time, the technology for
output production will affect the shape of the optimal contract, and hence the
18.3. BASIC FRAMEWORK financial market structure, since it determines the prices of capital and labour
and, ultimately, the equilibrium choice of the contract.
Given the complex structure that credit markets can assume under The standard working of these models is the following. Agents can be
information asymmetries, modelling the interrelationship between finance borrowers or lenders. Young borrowers have access to projects for capital
and economic growth requires the use of many simplifying assumptions. production and require external funding.2 Suppliers of such funding are
These assumptions are adjusted to specific modelling needs and can make young lenders who, instead, are endowed with labour. Lenders supply this
the financial markets configuration widely differ from one model to another. labour in the sector for output production and obtain a wage income which
Notwithstanding such differences, the majority of these models share can be lent out to borrowers. Once borrowers obtain the loan, they produce
common features and exhibit many similarities in their basic structure. capital which is sold to output producers. It is important to mention that,
At the outset, one cannot but notice that in the literature the prevalent while output is produced instantly (at the same time that production starts),
approach in modelling growth and financial development is that of capital takes one period to be produced and can be sold only in the period
‘overlapping generations’. It is clear that when interrelationships between following the beginning of production. This has very important implications
agents become very complex due to the presence of informational and may create problems of inconsistency in the results. In fact, while the
asymmetries, the overlapping generations framework greatly simplifies the price of labour (the size of the loan) is well known at the time the contract is
understanding of such contractual arrangements and their dynamics. The designed, the price of capital, considered in the contract, is based on
results, however, do not hinge on this particular approach and could expectations.3 It is clear that such expectations must be consistent with the
presumably be reiterated with representative agents. actual realised return on capital. This problem can find different solutions.
Simplification is, again, the main rationale behind the common The most simple is to assume that the output production function displays
assumption of a production activity divided into two distinct processes: constant returns to the aggregate capital stock, as in many endogenous
production of output (i.e. consumption good) and production of accumulable growth models (Romer, 1986). In fact, if this is the case, the marginal
factor (typically physical capital). The technology describing the former product of capital (and the price of capital) in equilibrium will be given and
process employs as inputs, among others, the physical capital produced in the constant (Bose and Cohtren, 1996; Blackburn, Bose and Capasso, 2001).
latter process. This establishes the main connections between the two Alternatively, one has to make sure that the realised price of capital is always
technologies. Most importantly, the imperfections in the credit market consistent with agents’ expectations (Bose and Cohtren, 1997). As in a Nash
concern only the financing of the projects producing physical capital and not subgame perfect equilibrium, the equilibrium optimal contract must be
the technology producing output. As a consequence, informational problems consistent with agents’ beliefs.
do not concern the prices of the factors of production, such as capital and Borrowers will repay the loan to lenders once the capital has been
labour, as determined by the output production function. This will greatly produced. On logical grounds one can assume, indifferently, that loans are
simplify the design of the financial contract since agents take such prices as repaid in terms of capital or output. That is, the contract might bind
given. The rate of growth of the economy will ultimately be determined by borrowers to repay lenders directly with the capital produced, or it might
capital accumulation as determined by the technology to produce capital. require that borrowers repay lenders in units of output, after having sold the
capital to output producers. The two alternatives, equivalent on logical corresponding credit market structure, will change when those market
grounds, can produce different results in practice. For instance, under the conditions change because of economic growth. In turn, the credit market
assumption that borrowers have all the bargaining power, the repayment is can influence economic growth since it can affect the flow of resources
determined according to the lenders’ reservation utility which is very often devoted to investment. This explains the co-evolution of real and financial
represented by an alternative investment technology. If the reservation utility development.
is in units of output, then the price of capital will enter the repayment In every period, given the set of available information and the new market
function only when the contract requires borrowers to repay in units of conditions, financial contracts are redesigned. Each generation of young
capital (and vice versa). In this case, if the price of capital is not constant in agents (borrowers and lenders) meets in the credit market where funds are
equilibrium many analytical problems might arise. transferred from lenders to borrowers under a pre-specified contract.
Agents are usually assumed to derive positive utility only from old age The contract, as seen, depends on variables such as the expected price of
consumption. The consequence is that all first period income is invested and capital, the amount of available savings – which, in turn, are functions of the
there is no consumption in young age. This seemingly very unrealistic wage rate –, and the verification costs. Accordingly, agents will design the
assumption meets the need to eliminate problems of consumption-saving financial contract in order to achieve the highest possible payoff and hence
choice. Under this assumption all income is saved and, as a consequence, if the highest expected utility. Of course, optimisation implies that if different
there is any effect on growth from financial markets, it does not stem from forms of contract are available, the equilibrium contract will be that
savings mobilisation. This allows one to focus only on the effects of associated with the highest payoff. Thus, if there were two forms of available
information asymmetry and credit market imperfections on capital contracts to choose from, credit rationing and screening for instance (Bose
accumulation and growth. Introducing the consumption-saving problem does and Cohtren, 1996, 1997), agents will abide, in equilibrium, by the one that
not influence the marginal choice on the optimal contract, apart from the delivers the highest payoff: either rationing or screening. Obviously, the
natural reduction in the availability of aggregate savings. The same rationale equilibrium contract can also be a combination of different contracts when
applies to the assumption of risk neutrality. Assuming risk-neutral agents this combination delivers an expected payoff higher than that achievable
eliminates the problem of risk sharing and allows concentration on specific through a single contract. If we think of equity and debt, for example, as two
informational problems. different forms of repayment each relating to a specific contract, a
Finally, it is worth mentioning the issue regarding bargaining power. In combination of the two – a repayment which consists partly of equity and
all these models, the contract is designed either under the assumption that partly of debt – might be strictly preferred to a repayment only in the form of
borrowers have all the bargaining power (Bose and Cohtren, 1996, 1997; debt or equity (Boyd and Smith, 1996, 1998 ; Blackburn, Bose and Capasso,
Bencivenga and Smith, 1993) or that lenders have all the bargaining power 2001).
(Blackburn, Bose and Capasso, 2001; Boyd and Smith, 1996, 1998). Apart It is intuitive that the optimal contract will not necessarily be the same as
from empirical considerations regarding which of the two parties really has capital accumulates and market conditions vary. With growth occurring, the
the power in credit markets, both assumptions deliver a Pareto efficient wage rate, the price of capital and the verification costs might change, with
situation and are analytically equivalent. The substantial difference is that, in the result that the optimal structure of the financial contracts and the
one case, lenders are driven to their reservation utility, in the other, associated payoffs will change as well.
borrowers. The level of economic activity can be measured in different ways. As
capital accumulates, given a fixed labour supply, the wage rate increases.4
Hence the amount of resources available to lenders increases, as do the loan
18.4. CAPITAL ACCUMULATION AND FINANCIAL size and the project size.5 The price of capital, on the other hand, decreases
DEVELOPMENT: THE NEXUS with capital accumulation or remains constant (in equilibrium), depending on
the form of the output production function. Thus, for example, if the cost of
As outlined, under the assumption of asymmetric information in the credit rationing is a decreasing function of the level of economic activity
market, the optimal financial arrangement is the result of agents’ efforts to (Bencivenga and Smith, 1993; Bernanke and Gertler, 1989), with growth
reduce agency costs. Given that such financial arrangements depend on occurring, the optimal contract might involve a lower level of credit
specific market conditions, the optimal financial contract, and the rationing. Then again, if screening involves a cost which is a decreasing
function of the loan size, agents might prefer a rationing contract to a of growth. Importantly, the effect of financial markets on growth does not
screening one at a low level of capital accumulation, and screening to necessarily feed through to a permanent increase in the growth rate
rationing at a higher level (Bose and Cohtren, 1996, 1997). This could (Bencivenga and Smith, 1993; Bose and Cohtren, 1996). Indeed, such effects
explain why in richer economies the level of credit rationing is clearly lower could not be analysed in a model that displays exogenous growth. Very
than in poorer ones. often, as reported in the literature, the effects of financial market
Even if monitoring costs are assumed to be fixed in units of output, they development on growth prove to be effects on the capital accumulation path.
might appear to be increasing if expressed in units of capital when the price Thus a change in the financial contract from rationing to screening (Bose and
of capital is decreasing because of capital accumulation. As a consequence, Cohtren 1997), or from debt to debt and equity (Blackburn, Bose and
with economic growth, agents prefer to reduce monitoring costs by Capasso, 2001) will free resources and will push the capital accumulation
reallocating investment from costly unobservable technologies towards path upwards, temporarily boosting the rate of growth.
observable ones (Boyd and Smith, 1996, 1998). The financial implication, as
shown by Boyd and Smith, is a contract with a higher proportion of
repayment in the form of equity rather than debt. Similarly, if monitoring 18.5. A SIMPLE GENERAL MODEL
costs are an increasing function of the wage rate, then with growth occurring,
lenders might find it optimal not to sustain these costs and directly observe We will now illustrate through a simple framework the central structure of
the borrower’s actions. Hence lenders will face more complex forms of the models on finance and growth which are based on the assumption of
moral hazard and the financial outcome will be a contract that involves forms information asymmetries. Given the complexity of the financial market in
of financial instruments, such as equity, other than debt (Blackburn, Bose itself and the many different assumptions made in these models, the attempt
and Capasso, 2001). This mechanism could explain the emergence of stock requires a large dose of generalisation. The main objective is to clarify the
markets at later stages of economic development. general working of these models and to highlight their dynamics.
Recalling that, usually, the output production function is separated from The model is characterised by the following equations:
the capital production function, the rate of economic growth is a function of
the rate of capital accumulation as determined by the capital production C = C( Rt ; lt ;π t ;φt ; γ t ;ζ t ) (1)
function. Indeed, the amount of resources invested in the latter (the amount
of available savings) and, therefore, the amount of future capital, depends on Rt = Et + Dt (2)
the actual amount of capital available in the output production technology.6 lt = l(wt) (3)
It is worth stressing, at this stage, that constant returns to aggregate capital in
the output production function, as in standard endogenous growth models, do π t = (1 − α ) z + α z( ρ t +1 ; wt ) (4)
not necessarily imply an endogenous rate of growth. If the capital production
function displays decreasing returns to capital, the rate of growth will be φt = (1 − β )g + β g( ρ t +1 ; wt ) (5)
exogenous as in the Solow–Swan model.
As already outlined, the relationship between the level of economic γt = γ(ρt+1, wt) (6)
activity and financial development is bi-directional. Financial markets can
affect the level of economic activity by determining the amount of funds E(Kt +1 ) = G ⎡⎣C(⋅); T ⎤⎦ (7)
devoted to investment and capital accumulation. It is immediately obvious
that credit rationing, by reducing the amount of projects undertaken or the Yt = F (Kt , Lt ) (8)
scale of the projects, can limit capital accumulation. However, also screening
and monitoring costs can lower the amount of resources accumulated when wt = FL (9)
these costs consist of a net loss for the economic system and do not result in Equation (1) captures the form and elements of the contract, C(⋅). As
a simple transfer of resources. The general idea is that, under information already argued, the elements that a contract eventually determines include:
asymmetries, agency costs influence investment allocation by deviating the the repayment, Rt, the size of the loan, lt, the probability of rationing, πt, the
flow of investment towards less productive activity, thereby affecting the rate probability of screening, φt, and the required amount of collateral, γt. It is
clear that while the repayment and loan size are essential elements of a credit produced according to (8). The output is then used to pay labour and capital
contract (i.e. each and every contract needs to determine repayment and loan their price, (9). Young lenders and borrowers meet in the credit market
size), the presence of other contractual arrangements depends on the where, for a given wage rate and the expected price of capital, the optimal
assumptions regarding the credit market. As seen above, under the contract is determined according to (1). Borrowers who are granted credit,
assumption of information asymmetries, the contractual arrangement is the l(wt), produce capital E(Kt+1) which will be available in the next period. Once
result of the level of agency costs, endowments and technologies, which can capital has been produced, borrowers will repay the loan. The loan will be
take very different forms. Some of these contractual elements have been repaid either in terms of capital or output. In the first case after producing
explicitly considered (probability of rationing, screening etc.). However, capital, borrowers immediately repay the lenders and sell the rest of capital
many others could be included depending on the assumptions. The variable ζt to new output producers. In the second case, borrowers first sell the capital to
captures any other element that might enter the contract. output producers and then repay the lenders. With the new stock of capital
The form of repayment is specified by equation (2). The repayment can be and with the labour supplied by the new young generation of lenders output
in the form of debt, Dt, equity, Et, or a combination of the two.7 Since the is produced and the cycle restarts.
resources in the credit market are supplied by lenders who sell their labour to
output producers, the loan size depends on the wage rate, wt, (eq. (3)).
The probability of credit rationing, eq.(4), and the probability of 18.6. CONCLUSIONS
screening, eq.(5), can be either constant (α = 0; β = 0) or they can depend on
the price of capital, ρt+1, and/or on the price of labour (α = 1; β = 1). In Bose Empirical evidence strongly supports the view that financial markets play a
and Cohtren (1996, 1997) both the probability of screening and the key role in the process of capital accumulation. The development of financial
probability of credit rationing are constant. In this case, the effect of growth markets seems to be positively correlated with the process of economic
on the credit market feeds through to a switch in the equilibrium contract growth. This view is not a new one in economics. Classical economists were
(from rationing to screening), rather than the level of credit rationing (as, convinced of the importance of the financial structure for its effects on the
instead, occurs in Bencivenga and Smith, 1993). The collateral, γt, is a allocation of real resources and, therefore, on the average return on
function of the price of capital and/or the wage rate. A change in either of investments. However, only recently have economists attempted to provide a
these two variables might affect the value of the collateral and, consequently, theoretical explanation of the linkages between financial and real sectors of
the borrower’s creditworthiness and the level of agency costs. an economy within fully articulated general equilibrium models of growth.
The expected level of capital is determined by equation (7). This depends The most recent research has focused attention on the assumption of
not only on the technology to produce capital, T , but also on the contract, asymmetric information in the credit market. Indeed, the agency costs that
C(⋅) . Clearly, the form of the financial contract can affect the amount of stem from such informational problems can determine financial
capital produced by simply determining the flow of resources devoted to arrangements that prove very different from those based on the assumption
investment. The technology to produce capital, on the other hand, delivers a of fully or symmetrically informed agents. Most importantly, under the
stochastic return, which can be either observable or unobservable, and it assumption of asymmetric information, the structure of the financial system
directly affects the amount of capital produced. is the result of agents’ attempts to reduce these agency costs, and can be
The last three equations describe the output production function and the viewed as endogenously determined. As a result, this framework highlights
price of the factors of production, labour and capital. The output production issues that cannot be investigated and analysed in a framework where the
function, as outlined, may display decreasing returns in each factor (as in a structure of the financial market is exogenously imposed.
standard neoclassical growth model) or constant returns to aggregate capital This chapter has provided a critical analysis of the most important
(production function à la Romer). Therefore, the one in (8) is a very general assumptions in this developing literature, in an attempt to reveal the
production function. Equations (9) can be interpreted in the following way: if underlying common analytical framework of seemingly very different
markets for capital and labour are competitive, then factors will be paid their models. The inductive process which lies behind this attempt, illustrated the
marginal productivity. general theoretical mechanism that explains the interactions between
The working of the model is the following. Given an initial level of financial development and economic growth.
capital stock, K0, and the initial labour supply from young lenders, output is
Modelling growth, financial intermediation and information frictions 357
It is important to note that the role played by financial markets, as it

appears in the literature, is essentially not very far from the role that classical
economists thought financial markets were playing in the economic system.
Financial markets can affect the process of capital accumulation through
specialisation and by channelling resources towards more productive
investments. However, it is also clear that these models, so very rich in their
References
analytical structure, have the power to highlight specific features of financial
systems and, consequently, provide the possibility to understand specific Abel, A.B. (1990), ‘‘Consumption and investment’, in B.M. Friedman and F.H. Hahn
aspects of the co-evolution of financial development and economic growth. (eds), Handbook of Monetary Economics, vol. II, New York: Elsevier Science, pp.
Indeed, on the basis of the potentially high explanatory power of this 725–78.
analytical framework, the focus of research has gradually shifted from very
Abraham-Frois, G. (1991), ‘Corporate behaviour, valuation ratio and macroeconomic
general questions, such as the role of financial intermediation in the process
analysis’, in Nell and Semmler (1991), pp. 190–204.
of economic development, to more specific issues, such as the endogenous
development of other forms of financial instruments and markets, and their Acemoglu, D. (1995), ‘Reward structures and the allocation of talents’, European
impact on growth. Economic Review, 39, 17–33.
Acemoglu, D. (1999), ‘Changes in unemployment and wage inequality: an alternative
theory and some evidence’, American Economic Review, 89, 1259–78.
NOTES Ackley, G. (1978), Macroeconomics: Theory and Policy, London: Macmillan.
1. We recall that in the standard CSV framework the borrower has access to a ‘single’ Adriansee, A., S. Bringezu, A. Hammond, Y. Moriguchi, E. Rodenberg, D, Rogich
unobservable technology whose return can be observed by the lender only by sustaining a and H. Schützt (1997), Resource Flows: The Material Basis of Industrial
monitoring, or verification, cost (for this reason ‘costly state verification’). Economies, Washington, DC: World Resource Institute.
2. From now on, for simplicity, we will base the discussion on the assumption of two-period-
Aghion, P., E. Caroli and C. Garcia Penalosa (1999), ‘Inequality and economic
lived overlapping generations.
3. Recall that funds to young borrowers are supplied by young lenders out of their wage growth: the perspective of the new growth theories’, Journal of Economic
income, which is well known at the time of the contract. On the other hand, the return on the Literature, 37, 1615–60.
project – the amount of capital produced – will only be available one period after the
Aghion, P. and P. Howitt (1992), ‘A model of growth through creative destruction’,
contract is designed.
Econometrica, 60, 323–51.
4. The assumption of a fixed labour supply is common. The results would be the same if labour
supply was growing at a lower rate than capital. Aghion, P. and P. Howitt (1994), ‘Endogenous technical change: the Schumpeterian
5. If there is no consumption-saving problem and all income is saved, the increase in wage perspective’, in L.L. Pasinetti and R.M. Solow (eds) Economic Growth and the
results in an equivalent increase in loan size. This, of course, requires that lenders have no
Structure of Long-Term Development, London: Macmillan, pp. 118–31.
alternative investment opportunities with higher return and that the return on the project is a
positive function of the resources invested. Aghion, P. and P. Howitt (1998), Endogenous Growth Theory, Cambridge,
6. Recall that capital takes more than one period to be produced. The amount of today’s capital Massachusetts: MIT Press.
used in output production together with labour, determines the amount of today’s savings
available for investment. These resources employed in the capital production function will Aghion, P. and G. Saint-Paul (1998a), ‘Uncovering some casual relationships
determine tomorrow’s capital. between productivity growth and the structure of economic fluctuations: a
7. The form of repayment is assumed to be exogenously given. However, since our objective tentative survey’, Labour, 12(2), 279–303.
goes beyond the study of the financial arrangement in itself, and it consists in showing the
relationship between real and financial development in a general equilibrium framework, Aghion, P. and G. Saint-Paul (1998b), ‘Virtues of bad times: interaction between
these forms of repayment can be justified by the simple observation that in almost every productivity growth and economic fluctuations’, Macroeconomic Dynamics, 2(3),
financial system equity and debt are the most common forms of financial instruments. 322–44.
Moreover, there is a vast micro literature in finance that has rigorously proved that the
optimal repayment takes generally either the form of debt or equity. Agliardi, E. (1998), Positive Feedbacks Economies, London: Macmillan.
358
References 359 360 The Theory of Economic Growth: a ‘Classical’ Perspective
Alchian, A. (1955), ‘The rate of interest, Fisher’s rate of return over costs and Asada, T., W. Semmler and A.J. Novak (1998), ‘Endogenous growth and the
Keynes’s internal rate of return’, American Economic Review, 45(5), 938–45. balanced growth equilibrium’, Research in Economics, 52(2), 189–212.
Allen, R.G.D. (1967), Macro-Economic Theory. A Mathematical Treatment, London: Asimakopulos, A. (1985), ‘Harrod on Harrod: the evolution of “A Line of Steady
Macmillan. Growth”‘, History of Political Economy, 17(4), 619–35.
Amadeo, E.J. (1986a) ‘The role of capacity utilization in long-period analysis’, Asimakopulos, A. and J.C. Weldon (1965), ‘A synoptic view of some simple models
Political Economy, 2(2), 147–85. of growth’, Canadian Journal of Economics and Political Science, 31(1), 52–79.
Amadeo, E.J. (1986b), ‘Notes on capacity utilization, distribution and accumulation’, Atkinson, A.B. and J.E. Stiglitz (1969), ‘A new view of technological change’,
Contributions to Political Economy, 5, 83–94. Economic Journal, 69, 573–78.
Amadeo, E.J. (1987), ‘Expectations in a steady-state model of capacity utilization’, Atsumi, H. (1969), ‘The efficient capital programme for a maintainable utility level’,
Political Economy, 3(1), 75–89. Review of Economic Studies, 36, 263–87.
Andersen, E.S. (2001), ‘Toward a Multiactivity Generalisation of the Nelson–Winter Azariadis, C. and A. Drazen (1991), ‘Threshold externalities and economic
model’, Paper presented at the 2001 Nelson and Winter Conference, Aalborg. development’, Quarterly Journal of Economics, 105, 501–26.
Anderson, G.M., C.K. Rowly and R.D. Tollison (1988), ‘Rent seeking and the Backhouse, R. (1998), Explorations in Economic Methodology, London: Routledge.
restriction of human exchange’, Journal of Legal Studies, 27, 83–100.
Bagehot, W. (1962), Lombard Street, 1st edn 1873, Homewood IL: Richard D. Irwin.
Arndt, H.Z. (1984), Economic Development, Chicago: The University of Chicago
Press. Bairam, E.I. (1993), ‘Income elasticities of exports and imports: a re-examination of
the empirical evidence’, Applied Economics, 25(1), 71–4.
Arnold, L.G. (2000a), ‘Stability of the market equilibrium in Romer’s model of
endogenous technical change: a complete characterization’, Journal of Bairoch, P. (1973), ‘Agriculture and the industrial revolution’, in C.M. Cipolla (ed.),
Macroeconomics, 22(1), 69–84. The Fontana Economic History of Europe, vol. 3, The Industrial Revolution,
Glasgow: Collins.
Arnold, L.G. (2000b), ‘Endogenous technological change: a note on stability’,
Economic Theory, 16, 219–26. Baldwin, R.E. (1956), ‘Patterns of development in newly settled regions’, The
Manchester School, 24, 161–79.
Arrow, K.J. (1962), ‘The economic implications of learning by doing’, Review of
Ball, L. (1999), ‘Aggregate demand and long-term unemployment’, Brookings Papers
Economic Studies, 29, 155–73.
on Economic Activity, n. 2, 189–251.
Arrow, K.J. (1998), ‘Innovation and increasing returns to scale’, in K.J. Arrow, Y-K.
Ball, R.J., T. Burns and J.S.E. Laury (1977), ‘The role of exchange rate changes in
Ng, and X. Yang (eds), Increasing Returns and Economic Analysis, London:
balance of payments adjustment: the U.K. Case’, Economic Journal, 87, 1–29.
Macmillan, pp. 403–8.
Baranzini, M. (1991), A Theory of Wealth Distribution and Accumulation, Oxford:
Arrow, K.J., B. Bolin, R. Costanza, P. Dasgupta, C. Folke, C.S. Holling, B-O.
Clarendon Press.
Jansson, S. Levin, K-G. Maler, C. Perrings and D. Pimentel (1995), ‘Economic
growth, carrying capacity and the environment’, Science, 268, 520–21. Barbosa-Filho, N.H. (2000), ‘A note on the theory of demand-led growth’,
Contributions to Political Economy, 19, 19–32.
Arrow, K.J. and Kurz, M. (1970), Public Investment, the Rate of Return, and Fiscal
Policy, Baltimore: Resources for the Future. Barreto, R. (2000), ‘Endogenous corruption in a neoclassical growth model’,
European Economic Review, 44, 35–60.
Arthur, W.B. (1987), ‘Self-reinforcing mechanisms in economics’, in P.W. Anderson,
K.J. Arrow and D. Pines (eds), The Economy as an Evolving Complex System, Barro, R.J. (1990), ‘Government Spending in a Simple Model of Endogenous
Addison Wesley Publishing Company, 9–31. Growth’, Journal of Political Economy, 98(5), 103–25.
Asada, T. (1995), ‘Kaldorian dynamics in an open economy’, Journal of Economics, Barro R.J. and X. Sala-i-Martin (1995), Economic Growth, New York: McGraw-Hill.
62(3), 239–69.
Baumol, W.J. (1990), ‘Entrepreneurship: productive, unproductive and destructive’,
Journal of Political Economy, 101, 274–98.
Baumol, W.J. (1990), ‘Entrepreneurship: productive, unproductive and destructive’, Bertola, G. (1994), ‘Theories of saving and economic growth’, Ricerche Economiche,
Journal of Political Economy, 98, 893–921. 48, 257–77.
Bean, C. (1990), ‘Endogenous growth and the procyclical behaviour of productivity’, Bertola, G. (2000), ‘Macroeconomics of distribution and growth’, in B.A. Atkinson
European Economic Review, 34(2–3), 355–63. and F. Bourguison (eds), Handbook of Income Distribution, Amsterdam: Elsevier
Science, pp. 477–540.
Becker, G. (1988), ‘Family economics and macro behaviour’, American Economic
Review, 7, 1–78. Besomi, D. (1999), The Making of Harrod’s Dynamics, London: Macmillan.
Becker, G.S. and R.J. Barro (1988), ‘A Reformulation of the economic theory of Bhaduri, A. and S. Marglin (1990), ‘Unemployment and the real wage: the economic
fertility’, Quarterly Journal of Economics, 103(1), 1–25. basis for contesting political ideologies’, Cambridge Journal of Economics, 14(4),
375–93.
Becker, G.S., K.M. Murphy and R. Tamura (1990), ‘Human capital, fertility, and
economic growth, Journal of Political Economy, 98(5), S12–S37. Bhagwati, J.N. and T.N. Srinivasan (1980), ‘Revenue seeking: a generalization of the
theory of tariffs’, Journal of Political Economy, 88, 1069–87.
Beckerman, W.B. (1972), ‘Economic development and the environment: a false
dilemma’, International Conciliation, 586, 57–71, reprinted in W.B. Beckerman Blackburn, K. and V. Hung (1998), ‘A theory of growth, financial development and
(1992), Growth, the Environment and the Distribution of Incomes: Essays by a trade’, Economica, 65, 107–24.
Sceptical Optimist, Aldershot: Edward Elgar.
Blackburn, K., N. Bose, and S. Capasso (2001), ‘Financial development, financial
Becsi, Z. and P. Wang (1997), ‘Financial development and growth’, Economic choice and economic growth’, CGBC Discussion Paper, University of
Review, Federal Reserve Bank of Atlanta, 46–62. Manchester.
Benabou, R. (1996), ‘Heterogeneity, stratification, and growth: macroeconomic Blaug, M. (1975), ‘Kuhn versus Lakatos, or paradigms versus research programmes
implications of community structure and education finance’, American Economic in the history of economics’, History of Political Economy, 7, 399–433.
Review, 86, 584–609.
Blaug, M. (1980), The Methodology of Economics, Cambridge: Cambridge University
Bencivenga, V.R. and B. Smith (1991), ‘Financial intermediation and endogenous Press.
growth’, Review of Economic Studies, 58, 195–209.
Boldrin, M. and A. Rustichini (1994), ‘Growth and indeterminacy in dynamic models
Bencivenga, V.R. and B. Smith (1993), ‘Some consequences of credit rationing in an with externalities’, Econometrica, 62(2), 323–42.
endogenous growth model’, Journal of Economic Dynamics and Control, 17, 97–
Boldrin, M. and D.K. Levine (2002), ‘Factor saving innovation’, Journal of Economic
122.
Theory, forthcoming.
Benhabib, J. and K. Nishimura (1979), ‘The Hopf bifurcation and stability of closed
Bose, N. and R. Cothren (1996), ‘Equilibrium loan contracts and endogenous growth
orbits in multisector models of optimal economic growth’, Journal of Economic
in the presence of asymmetric information’, Journal of Monetary Economics, 38,
Theory, 21(3), 421–44.
363–76.
Benhabib J. and R. Perli (1994), ‘Uniqueness and indeterminacy: on the dynamics of
Bose, N. and R. Cothren (1997), ‘Asymmetric information and loan contracts in a
endogenous growth’, Journal of Economic Theory, 63(1), 113–42.
neoclassical growth model’, Journal of Money, Credit, and Banking, 29, 423–39.
Benhabib J., R. Perli and D. Xie (1994), ‘Monopolistic competition, indeterminacy,
Bose, S. (1968), ‘Optimal growth and investment allocation’, Review of Economic
and growth’, Ricerche Economiche, 48(4), 279–98.
Studies, 35, 465–80.
Ben-Porath, Y. (1980), ‘The F-connection: families, friends, and firms and the
Bowley, M.(1975), ‘Some aspects of the treatment of capital in the wealth of nations’,
organization of exchange’, Population Development Review, 6, 1–30.
in Skinner and Wilson (1975), pp. 361–76.
Bernanke, B. and M. Gertler (1989), ‘Agency costs, net worth, and business
Boyd, J. and B. Smith (1996), ‘The coevolution of real and financial sectors in the
fluctuation’, American Economic Review, 79, 14–31.
growth process’, The World Bank Economic Review, 10(2), 371–96.
Bertola, G. (1993), ‘Factor shares and savings in endogenous growth’, American
Boyd, J. and B. Smith (1998), ‘The evolution of debt and equity markets in economic
Economic Review, 83, 1184–98.
development’, Economic Theory, 12, 519–60.
Brandolini, A. (1995) ‘In search of a stylised fact: do real wages exhibit a consistent Cass, D. and K. Shell (1976), ‘The structure and stability of competitive dynamical
pattern of cyclical variability?’, Journal of Economic Surveys, 9(2), 103–63. systems’, Journal of Economic Theory, 12, 10–18.
Brewer, A. (1998), ‘Luxury and economic development: David Hume and Adam Cassel, G. (1903), The Nature and Necessity of Interest, London: Macmillan.
Smith’, Scottish Journal of Political Economy, 45, 78–98.
Cassel, G. (1932), The Theory of Social Economy (first German edn 1918), New
Britto, R. (1974), ‘Some recent developments in the theory of economic growth: an York: Harcourt Brace.
interpretation’, Journal of Economic Literature, 12, 1343–67.
Casson, M. (1993), ‘Cultural determinants of economic performance’, Journal of
Brock, W.A. and S. Durlauf (2001), ‘Discrete choice with social interactions’, Review Comparative Economics, 17, 418–42.
of Economic Studies, 68, 235–60.
Champernowne, D.G. (1945–56), ‘A note on J. v. Neumann’s Article “A model of
Brooks, M.A. and B.J. Heijdra (1989), ‘An exploration on rent seeking’, The economic equilibrium”‘, Review of Economic Studies, 13, 1–9.
Economic Record, 65, 32–50.
Charasoff, G. (1910), Das System des Marxismus: Darstellung und Kritik, Berlin:
Buchanan, J.M.. (1980), ‘Rent-seeking and profit-seeking’, in J.M. Buchanan, R.D. H. Bondy.
Tollison and G. Tullock (eds), Towards a Theory of the Rent-Seeking Society,
College Station, Texas A&M University Press, pp. 3–15. Chesnais, J. (1992), The Demographic Transition, Oxford: Oxford University Press.
Buchanan, J.M. and Y.J. Yoon (2000), ‘A Smithian perspective on increasing Chick, V. (1995), ‘Is there a case for post Keynesian economics?’, Scottish Journal of
returns’, Journal of the History of Economic Thought, 22, 43–8. Political Economy, 42(1), 20–36.
Chirinko, R.S. (1993), ‘Business fixed investment spending: modeling strategies,
Burgstaller, A. (1994), Property and Prices: Towards a Unified Theory of Value,
empirical results, and policy implications’, Journal of Economic Literature, 31(4),
Cambridge: Cambridge University Press.
1875–911.
Burmaister, E. and A.R. Dobell (1970), Mathematical Theories of Economic Growth,
Christiano, L. and M. Eichenbaum (1992), ‘Current real-business-cycle theories and
New York: Macmillan.
aggregate labour-market fluctuations’, American Economic Review, 82(3), 430–
Caballé, J. and M.S. Santos (1993), ‘On endogenous growth with physical and human 50.
capital’, Journal of Political Economy, 101(6), 1042–67.
Chu, C.C.Y. (1998), Population Dynamics. A New Economic Approach, Oxford:
Caballero, R. and M. Hammour (1994), ‘The cleansing effect of recessions’, Oxford University Press.
American Economic Review, 84(5), 1350–68.
Ciampalini, A. and F. Vianello, (2000), ‘Concorrenza, accumulazione del capitale e
Caballero, R.J. and A.B. Jaffe (1993), ‘How high are the giant’s shoulders: an saggio del profitto’, in M. Pivetti (ed.), Piero Sraffa, Roma: Carocci, pp. 363–98.
assessment of knowledge spillovers and creative destruction in a model of
Ciccone, R. (1986), ‘Accumulation and capital utilization: some critical
economic growth’, NBER. Macroeconomics Annual, 15–74.
considerations on Joan Robinson’s theory of distribution’, Political Economy,
Canova F. (1998), ‘Detrending and business cycles facts’, Journal of Monetary 2(1), 17–36.
Economics, 41(3), 475–512.
Ciccone, R. (1987), ‘Accumulation, capacity utilization and distribution: a reply’,
Cantwell, J.A. (1998), ‘Innovation as the principal source of growth in the global Political Economy, 3(1), 97–111.
economy’, in D. Archibugi, J. Howells and J. Michie (eds) Innovation Policy in a
Cigno, A. (1981), ‘Growth with exhaustible resources and endogenous population’,
Global Economy, Cambridge: Cambridge University Press, pp. 225–41.
Review of Economic Studies, 48, 281–87.
Carillo, M.R. and A. Zazzaro (2001), ‘Innovation, rent-seeking and social prestige:
Cigno, A. (1984a), ‘Consumption vs. procreation in economic growth’, in
towards a dynamic theory of professions’, paper presented at the Conference Old
G. Steinmann (ed.), Economic Consequences Change of Population in
and New Growth Theories: an Assessment, held in Pisa, October, 2001.
Industrialised Countries, Berlin, Heidelberg, New York: Springer-Verlag, pp. 29–
Cass, D. (1965), ‘Optimum growth in an aggregative model of capital accumulation’, 49.
Review of Economic Studies, 32, 233–40.
Cigno, A. (1984b), ‘Further implication of learning-by-doing: the effect of population
on per-capita income’, Bulletin of Economic Research, 36(2), 97–108.
Cohen, J.E. (1995), How Many People Can the Earth Support?, New York and Currie, L. (1997), ‘Implications of an endogenous theory of growth in allyn young’s
London: Norton. macroeconomic concept of increasing returns’, History of Political Economy, 29,
413–43.
Colander, D.C. (ed.)(1984), Neoclassical Political Economy, Cambridge: Ballinger
Publishing Company. Dahan, M. and D. Tsiddon (1998), ‘Demographic transition, income distribution, and
economic growth’, Journal of Economic Growth, 3, 29–52.
Cole, L., G. Mailath and A. Postlewaite (1992), ‘Social norms, saving behaviour and
growth’, Journal Political Economy, 104, 946–59. Daly, H.E. (ed.) (1973), Economics, Ecology, Ethics: Toward a Steady State
Economy, San Francisco: Freeman.
Coleman, J. (1988), ‘Social capital in the creation of human capital’, American
Journal of Sociology, 94, S95–S121. Dalziel, P.C. (1989), ‘Cambridge (U.K.) versus Cambridge (Mass.): a Keynesian
solution of “Pasinetti’s paradox”: comment’. Journal of Post Keynesian
Commendatore, P. (1994), ‘Sulla esistenza di un’economia a due classi in un modello
Economics, 11(4), 648–53.
Post Keynesiano di crescita e distribuzione con settore pubblico e attività
finanziarie’, Studi Economici, 51(3), 5–38. Dalziel, P.C. (1991–92), ‘Does government activity invalidate the Cambridge
theorem of the rate of profit? A reconciliation’, Journal of Post Keynesian
Commendatore, P. (1999a), Government Deficits (and Financial Activities) in Post
Economics, 14(2), 225–31.
Keynesian Theories of Growth and Distribution, PhD dissertation, University of
Naples ‘Federico II’. Dalziel, P.C. (1991a), ‘A generalisation and simplification of the Cambridge theorem
with budget deficits’, Cambridge Journal of Economics, 15(3), 287–300.
Commendatore, P. (1999b), ‘Pasinetti and dual equilibria in a post Keynesian model
of growth and institutional distribution’, Oxford Economic Papers, 51(1), 223–36. Dalziel, P.C. (1991b), ‘A note on Dalziel’s model of long run distributive
equilibrium: reply’, Journal of Post Keynesian Economics, 14(1), 121–4.
Committeri, M. (1986), ‘Some comments on recent contributions on capital
accumulation, income distribution and capacity utilization’, Political Economy, Dasgupta, P. and G.M. Heal (1974), ‘The optimal depletion of exhaustible resources,
2(2), 161–86. Review of Economic Studies, 41, 3–28.
Committeri, M. (1987), ‘Capital utilization, distribution and accumulation: a rejoinder Dasgupta, S. and T. Mitra (1988), ‘Intertemporal optimality in a closed linear model
to Amadeo’, Political Economy, 3(1), 91–5. of production’, Journal of Economic Theory, 45, 288–315.
Common, M.S. (1997), ‘Roles for ecology in ecological economics and sustainable Davis, K. and G.J. Moore (1945), ‘Some principles of stratification’, American
development’, in N. Klomp and I. Lunt (eds) Frontiers in Ecology: Building the Sociological Review, 10, 242–9.
Links, Amsterdam: Elsevier Science.
Davis, S. and J.C. Haltiwanger (1992), ‘Gross job creation, gross job destruction, and
Cooley, T.F. and E.C. Prescott (1995), ‘Economic growth and business cycles, in employment reallocation’, Quarterly Journal of Economics, 107(3), 819–64.
T.F. Cooley and E.C. Prescott (eds), Frontiers of Business Cycles, New Jersey:
de Bruyn, S.M. (2000), Economic Growth and the Environment: An Empirical
Princeton University Press, pp. 1–38.
Analysis, Dordrecht: Kluwer.
Corneo, G. and O. Jeanne (1997), ‘On the relative wealth effects and the optimality of
de Bruyn, S.M. and R.J. Heintz (1999), ‘The environmental Kuznets Curve’, in Van
growth’, Economic Letters, 70, 37–51.
den Bergh J. (ed.), Handbook of Environmental and Resource Economics,
Cornwall, J. and W. Cornwall (1994), ‘Growth theory and economic structure’, Aldershot: Edward Elgar, pp. 656–77.
Economica, 61, 237–51.
Denicolò, V. and M. Matteuzzi (1990), ‘Public debt and the Pasinetti paradox’,
Costabile, L. and B. Rowthorn (1985), ‘Malthus’s theory of wages and growth’, Cambridge Journal of Economics, 14(3), 339–44.
Economic Journal, 95, 418–37.
Diamond, D. (1984), ‘Financial intermediation and delegated monitoring’, Review of
Cozzi, G. (1998), ‘Culture as a bubble’, Journal of Political Economy, 106, 376–94. Economic Studies, 51, 393–414.
Cugno, F. and L. Montrucchio (1982), ‘Cyclical growth and inflation: a qualitative Diamond, P. (ed.), (1990), Growth, Productivity, Unemployment, Cambridge,
approach to Goodwin’s model with money prices’, Economic Notes, n. 3, 93–107. Massachusetts: MIT Press.
Dinopoulos E. and P. Thompson (1998), ‘Schumpeterian growth without scale Eicher, T.S. and S.J, Turnovsky (1999b), ‘Convergence in a two-sector non-scale
effects’, Journal of Economic Growth, 3, 313–37. growth model’, Journal of Economic Growth, 4, 413–28.
Dixit, A. and J. Stiglitz (1977), ‘Monopolistic competition and optimum product Eicher, T.S. and S.J. Turnovsky (2001), ‘Transition dynamics in a two-sector
diversity’, American Economic Review, 67, 297–308. nonscale rowth model’, Journal of Economic Dynamics and Control, 25, 85–113.
Dixon, R. and A.P. Thirlwall (1975), ‘A model of regional growth-rate differences on Eisner, R. (1958), ‘On growth models and the neoclassical resurgence’, Economic
Kaldorian lines’, Oxford Economic Papers, 27(2), 201–14. Journal, 68, 707–21.
Dolmas, J. (1996), ‘Endogenous growth in multisector Ramsey models’, International Eltis, W. (1984), The Classical Theory of Economic Growth, London: Macmillan.
Erlich, I. and F.T. Lui (1991), Intergenerational trade, longevity and economic
Domar, E. (1946), ‘Capital expansion, rate of growth, and employment’, growth, Journal of Political Economy, 99, 1029–59.
Ethier, W. (1982), ‘National and international returns to scale in the modern theory of
Dosi, G. (1982), ‘Technological paradigms and technological trajectories’, Research international trade’, American Economic Review, 72, 389–405.
Policy, 11, 147–62.
Evans, G.W., S. Honkapohja and P.M. Romer (1998), ‘Growth cycles’, American
Dosi, G. and R.R. Nelson (1994), ‘An introduction to evolutionary theory in Economic Review, 88, 495–515.
economics’, Journal of Evolutionary Economics, 4(3), 153–72.
Fair, R.C. (1999), ‘Estimated inflation costs had European unemployment been
Dow, S. (1985), Macroeconomic Thought: A Methodological Approach, Oxford: reduced in the 1980s by macro policies’, Journal of Macroeconomics, 21(1),
Basil Blackwell. Winter.
Duménil, G. and D. Lévi (1985),‘The classicals and the neoclassicals: a rejoinder to Fanti, L. and P. Manfredi (2001), ‘Population dynamics, political economy and “old”
Frank Hahn’, Cambridge Journal of Economics, 9, 327–45. and “new” growth theory’, Working Paper, Dipartimento di Statistica e
Matematica Applicata all’Economia, University of Pisa
Duménil, G. and D. Lévi (1987), ‘The dynamics of competition: a restoration of the
classical analysis’, Cambridge Journal of Economics, 11, 133–64. Fershtman, C., K. Murphy and Y. Weiss (1996), ‘Social status, education and
growth’, Journal of Political Economy, 104, 108–32.
Durlauf, S.N. (1996), ‘A theory of persistent income inequality’, Journal of Economic
Growth, 1, 75–93. Fisher, S. (1988), ‘Recent development in macroeconomics’, Economic Journal, vol.
98, pp. 294–339.
Durlauf, S.N. (1999a), ‘The case against social capital’, Focus, 20, 1–5.
Flaschel, P. (1988), ‘Fiscal policy in an accelerator-augmented classical growth
Durlauf, S.N. (1999b), ‘Membership theory of inequality’, in E. Brezis and P. Temin cycle’, in P. Flaschel and M. Krüger (eds), Recent Approaches to Economic
(eds), Elites, Minorities and Economic Growth, Amsterdam: Elsevier Science. Dynamics, Frankfurt am Main: Peter Lang, pp. 80–99.
Dutt, A.K. (1984) ‘Stagnation, income distribution and monopoly power’, Cambridge Flaschel, P. (1993), Macrodynamics. Income Distribution, Effective Demand and
Journal of Economics, 8(1), 25–40. Cyclical Growth, Frankfurt am Main: Peter Lang.
Dutt, A.K. (1987), ‘Alternative closures again: a comment on ‘growth, distribution Flaschel, P., R. Franke and W. Semmler (1997), Dynamic Macroeconomics.
and inflation’, Cambridge Journal of Economics, 11(1), 75–82. Instability, Fluctuations, and Growth in Monetary Economies, Cambridge,
Dutt, A.K. (1990), Growth, Distribution and Uneven Development, Cambridge: Massachusetts: MIT Press.
Cambridge University Press. Fleck, F.H. and C.M. Domenghino (1987), ‘Cambridge (UK) versus Cambridge
Dyson, T. and M. Murphy (1985), ‘The onset of fertility transition’, Population and (Mass.): a Keynesian solution to the “Pasinetti paradox”‘, Journal of Post
Development Review, 11, 399–440. Keynesian Economics, 10(1), 22–36.
Eicher, T.S. and S.J. Turnovsky (1999a), ‘Non-scale models of economic growth’, Foley, D.K. and T.R. Michl (1999), Growth and Distribution, Cambridge,
Economic Journal, 109, 394–415. Massachusetts: Harvard University Press.
Follmer, H. (1974), ‘Random economies with many interacting agents’, Journal of Galor, O. and D.N. Weil (1996), ‘The gender gap, fertility, and growth’, American
Mathematical Economics, 1, 51–62. Economic Review, 86(3), 374–87.
Forster, B.A. (1980), ‘Optimal energy use in a polluted environment’, Journal of Galor, O. and D.N. Weil (1998), ‘Population, technology and growth: from the
Environmental Economics and Management, 7 (4), 321–33. Malthusian regime to the demographic transition’, NBER Working Paper n. 6811,
November.
Foster, J. (1996), ‘Analytical foundation of evolutionary economics’. University of
Queensland Department of Economics Discussion Papers, n. 198. Galor, O. and D.N. Weil (1999), ‘From Malthusian stagnation to modern growth’,
American Economic Review, 89, 150–4.
Franke, R.H, G.H. Hofstede and M.H. Bond (1991), ‘Cultural roots of economic
performance: a research note’, Strategic Management Journal, 12, 165–73. Galor, O. and D.N. Weil (2000), ‘Population, technology and growth: from the
Malthusian regime to the demographic transition and beyond’, American
Frankel, M. (1962), ‘The production function in allocation and growth: a synthesis’,
Economic Review, forthcoming.
American Economic Review, 52, 995–1022
Gandolfo, G. (1997), Economic Dynamics. Study Edition, Berlin: Springer-Verlag.
Freni, G., F. Gozzi and N. Salvadori (2001), ‘A multisector ‘AK model’ with
endogenous growth: existence and characterization of optimal paths and steady Garegnani, P. (1990a), ‘Quantity of capital’, in J. Eatwell, M. Milgate and
state analysis’, Studi e Ricerche del Dipartimento di Scienze Economiche P. Newman (eds), Capital Theory, London: Macmillan.
dell’Università di Pisa, n. 75, http://www-dse.ec.unipi.it/salvadori/pdf/FGS-
Garegnani, P. (1990b), ‘Sraffa: classical versus marginalist analysis’, in K. Bharadwaj
web.pdf.
and B. Schefold (eds), Essays on Piero Sraffa: Critical Perspectives on the
Frisch, R. (1993), ‘Propagation problems and impulse problems in dynamic Revival of Classical Theory, London: Unwin Hyman, pp. 112–41
economics, in Economic Essays in Honour of Gustav Cassel, London: Allen and
Garegnani, P. (1992), ‘Some notes for an analysis of accumulation’, in J. Halevi, D.
Unwin, pp. 171–205.
Laibman and E. Nell (eds), Beyond the Steady State: A Revival of Growth Theory,
Fuà, G. (1993), Crescita economica: le insidie delle cifre, Bologna: Il Mulino. London: Macmillan, pp. 47–72.
Funk, P. (1996), ‘Relative needs, and the long run economic problem’, unpublished Garegnani, P. and A. Palumbo (1998), ‘Accumulation of capital’, in Kurz and
manuscript, Paris: Delta. Salvadori (1998e), pp. 10–18.
Futagami, K. and K. Mino (1993), ‘Threshold externalities and cyclical growth in a Gellner, E. (1996), Condition for Liberty, New York: Penguin Books.
stylized model of capital accumulation’, Economic Letters, 41(1), 99–105. Georgescu-Roegen, N. (1976), Energy and Economic Myths, New York and Oxford:
Galbraith, J.K. (1997), ‘Time to ditch the NAIRU’, Journal of Economic Pergamon Press.
Perspectives, 11(1), 93–108.
Giammanco, M.D. (2002), ‘Competition and technical change in Marx: two
Gale, D. (1960), The Theory of Linear Economic Models, New York: McGraw-Hill. perspectives’, History of Economic Ideas, forthcoming.
Gale, D. and M. Hellwig (1985), ‘Incentive-compatible debt contracts: the one-period Giampietro, M. and K. Mayumi (2000), ‘Jevons’ paradox, scaling in societal
problem’, Review of Economic Studie, 52, 647–63. metabolism and the fairy tale of Kuznet curves’, paper presented at Transitions
Towards a Sustainable Europe, 3rd Biennial Conference of the European Society
Galeotti, M. and A. Lanza (1999), ‘Desperately seeking (environmental) Kuznets’, for Ecological Economics, Vienna, 3–6 May 2000.
Crenos Working Papers 99/1, available at http://www.crenos.unica.it/working/
99_1.html Gilbert, G. (1980), ‘Economic growth and the poor in Malthus’ Essay on Population’,
History of Political Economy, 12, 83–96.
Galor, O. and O. Moav (2000), ‘Natural selection and the origin of economic growth’,
Brown University Working Paper, 2000–18. Glaeser, E. and J. Scheinkman (2000), ‘Non-market interactions’, NBER Working
Paper n. 8053.
Galor, O. and D. Tsiddon (1994), ‘Human capital distribution, technological progress
and economic growth’, CEPR Discussion Paper, 971. Glaeser, E., B. Sacerdote and J. Scheinkman (1996), ‘Crime and social interactions’,
Quarterly Journal of Economics, 61, 507–48.
Galor, O. and D. Tsiddon (1997), ‘Technological progress, mobility and economic
growth’, American Economic Review, 87, 363–82. Glaeser, E., D. Laibson and B Sacerdote (2000), ‘The economic approach to social
capital’, NBER Working Paper n. 7728.
Glombowski, J. and M. Krüger (1988), ‘A Short-period growth cycle model’, Greiner, A. and W. Semmler (1996a), ‘Multiple steady states, Indeterminacy, and
Recherches Economiques de Louvain, 54(4), 423–38. cycles in a basic model of endogenous growth’, Journal of Economics, 63(1), 79–
99.
Goldin, C. and L. F. Katz (1999), ‘The returns to skill across the twentieth century
United States’, NBER Working Paper n. 7126. Greiner, A. and W. Semmler (1996b), ‘Saddle path stability, fluctuations, and
indeterminacy in economic growth’, Studies in Nonlinear Dynamics and
Goldsmith, R.Y. (1969), Financial Structure and Development, New Haven and
Econometrics, 1(2), 105–18.
London: Yale University Press.
Griliches Z. (1989), ‘Patents: recent trends and puzzles’, Brooking Papers on
Goodwin, R.M. (1946), ‘Innovations and the irregularity of economic cycles’, Review
Economic Activity, Microeconomics, 291–330.
of Economic Statistics, 28(2), 95–104.
Griliches Z. (1990), ‘Patent statistics as economic indicators: a survey’, Journal of
Goodwin, R.M. (1948), ‘Secular and cyclical aspects of the multiplier and the
Economic Literature, 28, 1612–707.
accelerator’, in Income, Employment and Public Policy. Essays in Honor of Alvin
H. Hansen, New York: W.W. Norton & C., pp. 108–32. Grossman, G.M. and E. Helpman (1991), Innovation and Growth in the Global
Economy, Cambridge, Massachusetts: MIT Press.
Goodwin, R.M. (1951), ‘The nonlinear accelerator and the persistence of business
cycles’, Econometrica, 19(1), 1–17. Grossman, G.M. and E. Helpman (1994), ‘Endogenous innovation in the theory of
growth’, Journal of Economic Perspectives, 8(1), 23–44.
Goodwin, R.M. (1953), ‘The problem of trend and cycle’, Yorkshire Bulletin of
Economic and Social Research, 5(2), 89–97. Grossman, H.I. and M. Kim (1995), ‘Swords or plowshares? a theory of the security
of claims to property’, Journal of Political Economy, 103, 1275–88.
Goodwin, R.M. (1955), ‘A model of cyclical growth’, in E. Lundberg (ed.), The
Business Cycle in the Post-War World. Proceedings of a Conference held in Gurley, J.G. and E.S. Shaw (1955), ‘Financial aspects of economic development’,
Oxford by the International Economic Association on ‘Business Cycle Problems’, American Economic Review, 45, 515–38.
September 1952, London: Macmillan, pp. 203–21.
Gurley, J.G. and E.S. Shaw (1960), Money in a Theory of Finance, Washington, DC:
Goodwin, R.M. (1967), ‘A growth cycle’, in C.H. Feinstein (ed.), Socialism, Brookings Institute.
Capitalism and Economic Growth. Essays Presented to Maurice Dobb,
Gurley, J.G. and E.S. Shaw (1967), ‘Financial structure and economic development’
Cambridge: Cambridge University Press, pp. 54–8.
Economic Development and Cultural Change, 15, 257–68.
Goodwin, R.M. (1986), ‘The M-K-S-system: the functioning and evolution of
Hagemann, H. (1998), ‘General glut controversy?, in Kurz and Salvadori (1998e), pp.
capitalism’, in H.-J. Wagener and J.W. Drucker (eds), The Economic Law of
320–4.
Motion of Modern Society. A Marx-Keynes-Schumpeter Centennial, Cambridge:
Cambridge University Press, pp. 14–21. Hahn, F.H. and R. Matthews (1964), ‘The theory of economic growth: a survey’,
Economic Journal, 74, 779–901, reprinted in Surveys of Economic Theory,
Gordon R. (2000), ‘Does the ‘new economy’ measure up to the great inventions of
Growth and Development, London: Macmillan, 1965.
the past?’, NBER Working Paper n. W7833.
Hall, R.E. (1993), ‘Macro theory and the recession of 1990–91’, American Economic
Gozzi, F. and G. Freni (2001), ‘On a dynamic non-substitution theorem and other
Review, 83(2), 275–9.
issues in Burgstaller’s “Property and Prices”‘, Metroeconomica, 52(2), 181–96.
Hall, R.E. (1999), ‘Reorganizations’, NBER Working Paper n. 7181.
Gradstein, M. and M. Justman (2000), ‘Human capital, social capital and public
schooling’, European Economic Review, 44, 879–90. Hamberg, D. (1971), Models of Economic Growth, New York: Harper and Row.
Gray, H.P. (1996), ‘Culture and economic performance: policy as an interventing Hamouda, O. and G. Harcourt (1989), ‘Post-Keynesianism: from criticism to
variable’, Journal of Comparative Economics, 23(3), 278–91. coherence?’, in J. Pheby (ed.), New Directions in Post-Keynesian Economics,
Aldershot: Edward Elgar, pp. 1–34.
Greenwood, J. and B. Jovanovic (1990), ‘Financial development, growth, and the
distribution of income’, Journal of Political Economy, 98, 1076–107. Hands, D.W. (1988), ‘Ad hocness in economics and the Popperian tradition’, in N. De
Marchi (ed.), The Popperian Legacy in Economics, Cambridge: Cambridge
University Press, pp. 121–37.
Hands, D.W. (1990), ‘Second thoughts on “second thoughts”: reconsidering the Hicks, J.R. (1950), A Contribution to the Theory of the Trade Cycle, Oxford:
Lakatosian progress of The General Theory’, Review of Political Economy, 2, 69– Clarendon Press.
81.
Hicks, J. (1969), A Theory of Economic History, Oxford: Clarendon Press.
Hansen, G.D. and E.C. Prescott (1999), From Malthus to Solow, Federal Reserve
Hirsh, F. (1976), Social Limits to Growth, Cambridge, Massachusetts: Harvard
Bank of Minneapolis, Research Department Staff Report 257.
University Press.
Harcourt, G.C. (1972), Some Capital Controversies in the Theory of Capital.
Hollander, S. (1973), The Economics of Adam Smith, London: Heinemann
Cambridge: Cambridge University Press.
Educational Books.
Harris, D.J. (1980), ‘A postmortem on the neo-classical parable’, in E.J. Nell (ed.)
Houthakker, H. and S. Magee (1969), ‘Income and price elasticities in world trade’,
Growth, Profits, and Property, Cambridge: Cambridge University Press, pp. 43–
Review of Economics and Statistics, 51(2), 111–25.
63.
Howitt P. (1999), ‘Steady endogenous growth with population and R&D growing’,
Harrod, R.F. (1933), International Economics, Cambridge: Cambridge University
Press.
Iwai, K. (2000), ‘A contribution to the evolutionary theory of innovation, imitation
Harrod, R.F. (1936), The Trade Cycle: An Essay, Oxford: Clarendon Press.
and growth’, Journal of Economic Behavior and Organization, 43, 167–98.
Harrod, R.F. (1939), ‘An essay in dynamic theory’, Economic Journal, 49(193), 14–
Iyigun, M.F., H.N. Mocan and A.L. Owen (2001), ‘Ideology, human capital and
33. Reprinted in R.F. Harrod (1972) Economic Essays, 2nd edition, London:
growth’, unpublished manuscript, University of Colorado.
Macmillan.
Jones, C.I. (1995), ‘R&D based models of economic growth’, Journal of Political
Harrod, R.F. (1948), Towards a Dynamic Economics, London: Macmillan.
Economy, 103, 759–84.
Harrod, R.F. (1951), ‘Notes on trade cycle theory’, Economic Journal, 61(242), 261–
Jones, C.I. (1998a), Introduction to Economic Growth, New York and London;
75.
W.W. Norton.
Harrod, R.F. (1951), The Life of John Maynard Keynes, London: Macmillan.
Jones, C.I. (1998b), Population and Ideas: A Theory of Endogenous Growth,
Harrod, R.F. (1960), ‘A second essay in dynamic theory’, Economic Journal, 70, Department of Economics, Stanford University, mimeo.
277–93. Jones, C.I. (1999a), ‘Growth: with or without scale effects?’, American Economic
Harrod, R.F. (1964), ‘Are monetary and fiscal policies enough?’, Economic Journal, Review, 89(2), 139–44.
74, 903–15. Jones, C.I. (1999b), ‘Was an Industrial Revolution Inevitable? Economic Growth
Harrod, R.F. (1967), ‘Increasing returns’, in R. Kuenne (ed.), Monopolistic over the Very Long Run’, NBER Working Paper n. 7375.
Competition: Essays in Honour of Edward Chamberlain, London: John Wiley and Jones, C.I. (2002), “Sources of economic growth in a world of ideas”, American
Sons,. Reprinted in R.F. Harrod, Economic Essays, London: Macmillan, 1972, Economic Review, 92, 220–239.
pp. 302–17.
Jones, L. and R. Manuelli (1990), ‘A convex model of equilibrium growth: theory
Harrod, R.F. (1969), Money, London: Macmillan. and policy implications’, Journal of Political Economy, 98, 1008–38.
Harrod, R.F. (1973), Economic Dynamics, London: Macmillan. Jones, L.E. and R. Manuelli (1997), ‘The sources of growth’, Journal of Economic
Dynamics and Control, 21, 75–114.
Hassler, J. and J. Rodriquez Mora (2000), ‘Intelligence, social mobility, and growth’,
American Economic Review, 90(4), 888–908. Jorgenson, D.W. (1967), ‘The theory of investment behavior’, in R. Ferber (ed.),
Determinants of investment behaviour, New York: NBER, Columbia University
Heal, G. (1999), ‘Introduction’, in G. Heal (ed.), The Economics of Increasing Press.
Returns, Cheltenham: Edward Elgar, pp. xiii–xxvii.
Jovanovich B. and R. Rob (1990), ‘Long waves and short waves: growth through
Helpman E. and M. Trajtenberg (1994), ‘A time to sow and a time to reap: growth intensive and extensive search’, Econometrica, 58, 1391–409.
based on general purpose technologies’, CEPR Working Paper, n. 1080.
Junankar, P.N. (1972), Investment: Theories and evidence, London: Macmillan.
Kaganovich, M. (1998), ‘Sustained endogenous growth with decreasing returns and Kaldor, N. (1981), ‘The role of increasing returns, technical progress and cumulative
heterogeneous capital’, Journal of Economic Dynamics and Control, 22, 1575– causation in the theory of international trade and economic growth’, Economie
603. Appliquée, 34, 593–617.
Kaldor, N. (1954a), ‘The relation of economic growth and cyclical fluctuation’, Kalecki, M. (1937), ‘The principle of increasing risk’, Economica,4(16), 440–447.
Economic Journal, 64(253), 53–71, reprinted in Kaldor (1980), pp..
Kalecki, M. (1968), ‘Trend and business cycles reconsidered’, Economic Journal,
Kaldor, N. (1954b), ‘Characteristics of economic development’, reprinted in Kaldor 78(310), 263–76.
(1980),pp. 233–42.
Kalecki, M. (1971), Selected Essays in the Dynamics of the Capitalist Economy,
Kaldor, N. (1955–56), ‘Alternative theories of distribution’, Review of Economic Cambridge: Cambridge University Press.
Studies, 23(2), 83–100.
Keeton, W. (1979), Equilibrium Credit Rationing, New York: Garland Publishing
Kaldor, N. (1957a), ‘A model of economic growth’, Economic Journal, 67, 591–624. Company.
Kaldor, N. (1957b), ‘Capitalist evolution in the light of Keynesian economics’, Keynes, J.M. (1973–89), The Collected Writings of J.M. Keynes, 30 vols, edited by D.
Sankhya (May 1957)???, reprinted in Kaldor (1980), pp. 243–58. Moggridge, London: Macmillan. Referred to in the text as CW, volume number,
page number.
Kaldor, N. (1958), ‘Monetary policy, economic stability and growth: a memorandum
submitted to the Radcliffe Committee on the working of the monetary system’, 23, King, R.G. and R. Levine (1993a), ‘Finance and growth: Schumpeter might be right’,
Principal Memoranda of Evidence, Cmnd 827, H.M.S.O, London, 1960, 146–53, Quarterly Journal of Economics, 108, 717–37.
reprinted in N. Kaldor, Essays on the Economic Policy I, London: Duckworth.
King, R.G. and R. Levine (1993b), ‘Finance entrepreneurship, and growth: theory and
Kaldor, N. (1961), ‘Capital accumulation and economic growth’, in F.A. Lutz and evidence’, Journal of Monetary Economics, 32, 513–42.
D.C. Hague (eds), The Theory of Capital, London: Macmillan, pp. 177–222.
King, R.G. and S. Rebelo (1990), ‘Public policy and economic growth: developing
Kaldor, N. (1966), Causes of the Slow Rate of Growth of the United Kingdom: An neoclassical implications’, Journal of Political Economy, 98, 126–50.
Inaugural Lecture, Cambridge: Cambridge University Press, reprinted in Kaldor
King, R.G., C.I. Plosser and S.T. Rebelo (1988), ‘Production, growth and business
(1978a).
cycles. I. The basic neoclassical model’, Journal of Monetary Economics, 21(2–3),
Kaldor, N. (1967), Strategic Factors in Economic Development, New York: Ithaca, 195–232.
Cornell University.
Kirman, A. (1999), ‘Interaction and markets’, in M. Gallegati and A. Kirman (eds)
Kaldor, N. (1970), ‘The case for regional policies’, Scottish Journal of Political Beyond the Representative Agent, Chelthenam: Edward Elgar Publishing, pp. 1–
Economy, 17, 337–48, reprinted in Kaldor (1978a). 44.
Kaldor, N. (1971), ‘Conflicts in national economic objectives’, Economic Journal, Knack, S. and P. Keefer (1997), ‘Does social capital have an economic payoff? A
81, 1–16, reprinted in Kaldor (1978a). cross-country investigation’, Quarterly Journal of Economics, 112, 1252–88.
Kaldor, N. (1972), ‘The irrelevance of equilibrium economics, Economic Journal, 82, Knodel, J.A. (1977), ‘Age patterns of fertility and the fertility transition: evidence
1237–55. from Europe and Asia’, Population Studies, 31(2), 219–50.
Kaldor, N. (1975), ‘What is wrong with economic theory’, Quarterly Journal of Kögel, T. and A. Prskawetz (2001), ‘Agricultural Productivity Growth and Escape
Economics, 89, 347–57. from the Malthusian Trap’, Journal of Economic Growth, 6(4), 337–57.
Kaldor, N. (1978a), Further Essays on Economic Theory, London: Duckworth. Kongsamut P., S. Rebelo, and D. Xie (2001), ‘Beyond balanced growth’, Review of
Economic Studies, 68(4), 869–82.
Kaldor, N. (1978b), Further Essays on Applied Economics, London: Duckworth.
Koopmans, T.C. (1965), ‘On the concept of optimal economic growth’, in The
Kaldor, N. (1978c), ‘The effects of devaluation on trade in manufactures’, in Kaldor Econometric Approach to Development Planning, Amsterdam: North-Holland,
(1978b). pp. 225–87.
Kaldor, N. (1980), Essays on Economic Stability and Growth, London: Duckworth.
Kortum S.S. (1997), ‘Research, patenting and technological change’, Econometrica, the Eleventh World Congress of the International Economic Association, Tunis,
65, 1389–419. vol. 4: ‘Economic Behaviour and Design’, London: Macmillan, pp. 225–61.
Kregel, J.A. (1980), ‘Economic dynamics and the theory of steady growth: an Kydland, F.E. and E.C. Prescott (1982), ‘Time to build and aggregate fluctuations’,
historical essay on Harrod’s “knife-edge”‘, History of Political Economy, 12(1), Econometrica, 50(6), 1345–70.
97–123.
Ladron-de-Guevara, A., S. Ortigueira and M.S. Santos (1999), ‘A two-sector model
Kregel, J.A. (1985), ‘Harrod and Keynes: increasing returns, the theory of of endogenous growth with leisure’, Review of Economic Studies, 66, 609–31.
employment, and dynamic economics’, in G.C. Harcourt (ed.), Keynes and his
Laitner, J. (2000), ‘Structural change and economic growth’, Review of Economic
Contemporaries, London: Macmillan, pp. 66–88.
Studies, 67, 545–61.
Kremer, M. (1993), ‘Population growth and technological change: one million B.C. to
Lakatos, I. (1978), The Methodology of Scientific Research Programmes, Cambridge:
1990’, Quarterly Journal of Economics, 108, 681–716.
Cambridge University Press.
Krueger, A.O. (1974), ‘The political economy of the rent-seeking society’, American
Landes, D.S. (1969). The Unbound Prometheus. Cambridge: Cambridge University
Press.
Kurz, H.D. (1986), ‘Normal positions and capital utilization’, Political Economy,
Lanjow J.O. and M. Schankerman (1999), ‘The quality of ideas: measuring
2(1), 37–54.
innovation with multiple indicators’, NBER Working Paper n. 7345.
Kurz, H.D. (1991), ‘Technical change, growth and distribution: a steady state
Latsis, S.J. (1976), Method and Appraisal in Economics, Cambridge: Cambridge
approach to unsteady growth on Kaldorian lines’, in Nell and Semmler (1991),
University Press.
pp. 421–48.
Lavezzi, A. (2002), ‘ Smith, Marshall and Young on Division of Labor and Economic
Kurz, H.D. (1992), ‘Accumulation, effective demand and income distribution’, in
Growth’, European Journal of the History of Economic Thought, forthcoming.
J. Halevi, D. Laibman and E. Nell (eds), Beyond the Steady State: A Revival of
Growth Theory, London: Macmillan, pp. 73–95. Lavoie, M. (1992), Foundations of Post-Keynesian Economic Analysis, Aldershot:
Edward Elgar.
Kurz, H.D. and N. Salvadori (1993), ‘Von Neumann’s growth model and the
“classical” tradition’, The European Journal of the History of Economic Thought, Lavoie, M. (1995), ‘The Kaleckian model of growth and distribution and its neo-
1, 129–60. Ricardian and Neo-Marxian critiques’, Cambridge Journal of Economics, 19(6),
789–818.
Kurz, H.D. and N. Salvadori (1994), ‘The non-substitution theorem: making good a
lacuna’, Journal of Economics, 59, 97–103. Lazear, E.P. (1999), ‘Culture and language’, Journal of Political Economy, 107, S95–
1126.
Kurz, H.D. and N. Salvadori (1995), Theory of Production: A Long-Period Analysis,
Cambridge: Cambridge University Press. Lee, R.D. (1986), ‘Malthus and Boserup: a dynamic synthesisp’, in D. Coleman and
R. Schofield (eds), The State of Population Theory: Forward from Malthus,
Kurz, H.D. and N. Salvadori (1998a), ‘“Endogenous” growth models and the
Oxford: Basil Blackwell, pp. 96–130.
“classical” tradition’, in Kurz and Salvadori (1998d), pp. 66–89.
Lee, R.D. (1997), ‘Population dynamics: equilibrium, disequilibrium, and
Kurz, H.D. and N. Salvadori (1998b), ‘Postscript (to ‘The non-substitution theorem:
consequences of fluctuations’, in M.R. Rosenzweig and O. Stark (eds), Handbook
making good a lacuna’) in Kurz and Salvadori (1998d), pp. 95–9.
of Population and Family Economics, Amsterdam: North-Holland, pp. 1063–116.
Kurz, H.D. and N. Salvadori (1998c), Understanding “Classical” Economics. Studies
Lerner, A.P. (1944), The Economics of Control, London: Macmillan.
in the Long-Period Theory, London and New York: Routledge.
Levhari, D. (1966), ‘Extensions of Arrow’s learning by doing’, Review of Economic
Kurz, H.D. and N. Salvadori (eds), (1998d), The Elgar Companion to Classical
Studies, 33(1), 117–31.
Economics, vol. I, Chelthenam: Edward Elgar.
Levine R. (1991), ‘Stock markets, growth and tax policy’, The Journal of Finance,
Kurz, H.D. and N. Salvadori (1999), ‘Theories of “endogenous” growth in historical
46, 1445–65.
perspective’, in M.R. Sertel (ed.) Contemporary Economic Issues. Proceedings of
Levine, R. (1997), ‘Financial development and economic growth: views and agenda’, Manski, C.F. (2000), ‘Economic analysis of social interactions’’, Journal of
Journal of Economic Literature, 35, 688–726. Economic Perspectives, 14, 115–36.
Levine, R. and S. Zervos (1996), ‘Stock market development and long-run growth’, Marglin, S.A. (1984a), Growth, Distribution and Prices, Cambridge, Massachusetts:
The World Bank Economic Review, 10(2), 223–39. Harvard University Press.
Levine, R. and S. Zervos (1998), ‘Stock Markets, Banks and Economic Growth’, Marglin, S.A. (1984b), ‘Growth, distribution, and inflation: a centennial synthesis’,
American Economic Review, 88, 537–57. Cambridge Journal of Economics, 8(2), 115–44.
Lindbeck, A. and D.J. Snower (1999), ‘Price dynamics and production lags’, Marshall, A. (1890), Principles of Economics, 8th edn 1920. Reprint, reset (1977),
American Economic Review, 89(2), May. London: Macmillan.
Lippi, M. (1993), ‘Sviluppi recenti del modello lineare-stocastico in macroeconomia, Marshall, M.G. (2000), ‘Luxury, economic development, and work motivation: David
con particolare attenzione alle teorie del ciclo reale’, in A. Medio (ed.), Analisi Hume, Adam Smith, and J.R. McCulloch’, History of Political Economy, 32, 631–
dinamica in economia: modelli non-lineari e lineari stocastici, Bologna: Il 48.
Mulino, pp. 35–52.
Martinez, A. (1987), Ecological Economics: Energy, Environment and Society,
Livi-Bacci, M. (1997). A Concise History of World Population, second edition, Oxford: Blackwell.
Oxford: Blackwell.
Marx, K. (1954), Capital, vol. I, Moscow: Progress Publishers. English translation of
Livi-Bacci, M. (1998), La popolazione nella storia d’Europa, Bari: Laterza. Das Kapital, vol. I (1867), Hamburg: Meissner.
Long, J. and C. Plosser (1983), ‘Real business cycles’, Journal of Political Economy, Marx, K. (1956), Capital, vol. II, Moscow: Progress Publishers. English translation of
91(1), 39–69. Das Kapital, vol. II, edited by F. Engels (1885), Hamburg: Meissner.
Lowe, A. (1954), ‘The classical theory of economic growth’, Social Research, 21, Mattana, P. and B. Venturi (1999), ‘Existence and stability of periodic solutions in
127–58, reprinted in A. Lowe, Essays in Political Economics: Public Control in a the dynamics of endogenous growth’, Rivista internazionale di Scienze
Democratic Society, edited by A. Oakley, Brighton: Wheatsheaf Books, 1987. Economiche e Sociali, 46(2), 259–84.
Lowe, A. (1975), ‘Adam Smith’s Theory of Equilibrium Growth’, in Skinner and McCombie, J.S.L. (1998), ‘Harrod, economic growth and international trade’, in G.
Wilson (1975), pp. 415–25. Rampa, L. Stella, A.P. Thirlwall (eds), Economic Dynamics, Trade and Growth.
Essays on Harrodian Themes, London: Macmillan Press, pp. 212–44.
Lucas, R.E. (1988), ‘On the mechanics of economic development’, Journal of
Monetary Economics, 22(1), 3–42. McCombie, J.S.L. and A.P. Thirlwall (1994), Economic Growth and the Balance-of-
Payments Constraint, New York: St.Martin’s Press.
Lucas, R.E. (1993), ‘Making a miracle’, Econometrica, 61, 251–72.
McCombie, J.S.L. and A.P. Thirlwall (1999), ‘Growth in an international context: a
Lucas, R.E. (1997), ‘On the mechanics of economic development’, in D.P. Jacobs,
post Keynesian view’, in J. Deprez and J.T. Harvey (eds), Foundations of
E. Kalai, M.I. Kamien (eds), Frontiers on Research in Economic Theory,
International Economics. Post Keynesian Perspectives, London: Routledge,
Cambridge: Cambridge University Press, pp. 61–71.
pp. 35–90.
Lucas, R.E. (1998), The Industrial Revolution: Past and Future, Unpublished
McFadden, D. (1967), ‘The evaluation of development programmes’, Review of
Manuscript, University of Chicago.
Maccabelli, T. (1997). Il progresso della ricchezza. Economia, politica e religione in
McGrattan, E.R. (1990), ‘Solving the stochastic growth model by linear-quadratic
T.R. Malthus, Milano: Giuffrè Editore.
approximation’, Journal of Business & Economic Statistics, 8(1), 41–4.
Malinvaud, E. (1995), ‘About investment in macroeconomics’, in J.-P. Fitoussi (ed.),
Economics in a Changing World, vol. 5: Economic Growth and Capital and McKenzie, L.W. (1998), ‘Turnpikes’, American Economic Review, 88, 1–14.
Labour Markets, New York: St. Martin’s Press, pp. 123–41. McKinnon, R.I. (1973), Money and Capital in Economic Development, Washington,
Malthus, T.R. (1986). The Works of Thomas Robert Malthus, edited by E.A. Wrigley D.C.: Brookings Institution.
and D. Souden, London: William Pickering.
McNulty, P.J. (1967), ‘A note on the history of competition’, Journal of Political Nell, E.J. and W. Semmler (1991) (eds), Nicholas Kaldor and the Mainstream
Economy, 75, 396–9. Economics, London: Macmillan.
McNulty, P.J. (1968), ‘Economic theory and the meaning of competition’, Quarterly Nelson, C. and C. Plosser (1982), ‘Trends and random walks in macroeconomic time
Journal of Economics, 82, 639–56. series: some evidence and implications’, Journal of Monetary Economics, 10(2),
139–62.
Meade, J.E. (1961), A Neoclassical Theory of Economic Growth, London: Allen and
Unwin. Nelson, R.R. (1956), ‘A theory of the low level equilibrium trap in under-developed
economies’, American Economic Review, 46, 894–908.
Meadows, D.H., D.L. Meadows, J. Randers and W. Behrens III (1972), The Limits to
Growth, New York: Universe Book. Nelson, R.R. (1973), ‘Recent exercises in growth accounting: new understanding or
dead end?’, American Economic Review, 63(3), 462–8.
Merton R. (1965), On the Shoulders of Giants: A Shandean Postscript, New York:
Free Press. Nelson, R.R. (1994), ‘What has been the matter with neoclassical growth theory’, in
G. Silverberg and L. Soete (eds), The Economics of Growth and Technical
Mirrlees, J. and Stern, N. (1974) (eds), Models of economic growth, London:
Change, Aldershot: Edward Elgar, pp. 290–334.
Macmillan.
Nelson, R.R. (1996), The Source of Economic Growth, Cambridge, Massachusetts:
Mohtadi, H. and T. Roe (1998), ‘Growth, lobbying and public goods’, European
Harvard University Press.
Nelson, R.R. (1998), ‘The agenda for growth theory: a different point of view’,
Momota, A. and K. Futagami (2000), ‘Demographic transition pattern in a small
Cambridge Journal of Economics, 22, 497–520.
country’, Economic Letters, 67, 231–7.
Nelson, R.R. and S.G. Winter (1974), ‘Neoclassical vs. evolutionary theories of
Moreno Brid, J.C. (1998–99), ‘On capital flows and the balance-of-payments
economic growth: critiques and prospectus’, Economic Journal, 84, 886–905.
constrained growth model’, Journal of Post Keynesian Economics, 21(2), 283–98.
Nelson, R.R. and S.G. Winter (1977), ‘In search of a useful theory of innovation’,
Moss, S.J. (1978), ‘The post Keynesian theory of income distribution in the corporate
Research Policy, 6, 37–76.
economy’, Australian Economic Papers, 38(4), 303–22.
Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change,
Mulligan, C.B. and X. Sala-i-Martin (1993), ‘Transitional dynamics in two-sector
Cambridge, Massachusetts: Harvard University Press.
models of endogenous growth’, Quarterly Journal of Economics, 108(3), 739–73.
Nelson, R.R., S.G. Winter and L.H. Schuette (1976), ‘Technological change in an
Murphy K., A. Shleifer, and R. Vishny (1989a), ‘Income distribution, market size,
evolutionary model’, Quarterly Journal of Economics, 90, 90–118.
and industrialization’, Quartely Journal of Economics, 104, 537–64.
Nerlove, M. and L.K. Raut (1997), ‘Growth models with endogenous population: a
Murphy K., A. Shleifer and R. Vishny (1989b), ‘Industrialization and the big push’,
general framework’, in M.R. Rosenzweig and O. Stark (eds), Handbook of
Population and Family Fconomics, Amsterdam: North-Holland, pp. 1117–74.
Murphy, K.D., A. Shleifer and R. Vishny (1991), ‘The allocation of talent:
Niehans, J. (1963), ‘Economic growth with two endogenous factors’, Quarterly
implication for growth’, Quarterly Journal of Economics, 106, 503–30.
Journal of Economics, 77, 349–71.
Murphy, K.D., A. Shleifer and R. Vishny (1993), ‘Why is rent-seeking so costly to
Nordhaus W. (1997), ‘Do real-output and real-wage measures capture reality? the
growth’, American Economic Review, Papers and Proceedings, 83, 409–14
history of light suggests not’, in T. Bresnahan and R.J. Gordon (eds), The
Negishi, T. (1993), ‘A Smithian growth model and Malthus’s optimal propensity to Economics of New Goods, Chicago: University of Chicago Press, pp. 29–66.
save’, The European Journal of the History of Economic Thought, 1, pp. 115–27.
North, D. (1981), Structure and Change in Economic History, New York:
Nell, E.J. (1985), ‘Jean Baptiste Marglin: a comment on “Growth, distribution and W.W. Norton and Company, Inc.
inflation”‘, Cambridge Journal of Economics, 9(2), 173–8.
Ossella, I. (1999), ‘A turnpike property of optimal programs for a class of simple
Nell, E.J. (1988), ‘Does the rate of interest determine the rate of profit?’ Political linear models of production’, Economic Theory, 14, 597–607.
Economy, 4(2), 263–7.
Palivos, T. (1995), ‘Endogenous fertility, multiple growth paths, and economic Pecorino, P. (1992), ‘Rent seeking and growth: the case of growth through human
convergence’, Journal of Economic Dynamics and Control, 19, 1489–1510. capital’, Canadian Journal of Economics, 25, 944–56.
Panico, C. (1993), ‘Two alternative approaches to financial model building’, Peretto P.F. (1998), ‘Technological change and population growth’, Journal of
Metroeconomica, 44(2), 93–133. Economic Growth, 3, 283–311.
Panico, C. (1997), ‘Government deficits in the post keynesian theories of growth and Perez, C. (1985), ‘Microelectronics, long waves and world structural change: new
distribution’, Contributions to Political Economy, 16, 61–86. perspective for developing countries’, World Development, 13, 441–63.
Panico, C. (1999), ‘The government sector in the post keynesian theory of growth and Petith, H. (2001), ‘Georgescu-Roegen versus Solow/Stiglitz and the convergence to
personal distribution’, in G. Mongiovi and F. Petri (eds), Value, Distribution and the Cobb-Douglas’, Working Paper 481.01 of the Departament d’Economia i
Capital, Essays in Honour of P. Garegnani, London: Routledge, pp. 339–53. d’Història Economíca, Universitat Autonoma de Barcellona.
Panico, C. (2001), ‘Monetary analysis in Sraffa’s writings’, in T. Cozzi and R. Petri, F. (1997), ‘On the theory of aggregate investment as a function of the rate of
Marchionatti (eds), Piero Sraffa’s Political Economy: A Centenary Estimate, interest’, Quaderni del Dipartimento di Economia Politica, Università di Siena,
London: Routledge, pp. 285–310. no. 215 (http://www.econ-pol.unisi.it/quaderni/215.pdf).
Panico, C. and N. Salvadori (eds), (1993), Post Keynesian Theory of Growth and Phelps Brown, H. (1980), ‘Sir Roy Harrod: a biographical memoir’, Economic
Distribution, Aldershot: Edward Elgar. Journal, 90, 1–33.
Park, M.-S. (2000), ‘Autonomous demand and the warranted rate of growth’, Phelps, E. (1962), ‘The new view of investment: a neo-classical analysis’, Quarterly
Contributions to Political Economy, 19, 1–18. Journal of Economics, 76, 548–67.
Pasinetti, L.L. (1960a), ‘A mathematical formulation of the Ricardian system’, Phillips, A.W. (1954), ‘Stabilisation policy in a closed economy’, Economic Journal,
Review of Economic Studies, 27(2), 78–98. 64(254), 290–323.
Pasinetti, L.L. (1960b), ‘Cyclical fluctuations and economic growth’, Oxford Pivetti, M. (1988), ‘On the monetary explanation of distribution: a rejoinder to Nell
Economic Papers, 12(2) 18–50. and Wray’, Political Economy, 4(2), 275–83.
Pasinetti, L.L. (1962), ‘Rate of profit and income distribution in relation to the rate of Pivetti, M. (1991), An Essay on Money and Distribution, London: Macmillan.
economic growth’, Review of Economic Studies, 29(4), 103–20. Pomini, Mario (2000), ‘Some notes on the early neo-classical growth models’,
Pasinetti, L.L. (1981), Structural Change and Economic Growth: A Theoretical Essay History of Economics Ideas, 8, 123–41.
on the Dynamics of the Wealth of Nations, Cambridge: Cambridge University Porter M.E. and S. Stern (2000), ‘Measuring the ‘ideas’ production function: evidence
Press. from international patent output’, NBER Working Paper n. 7891.
Pasinetti, L.L. (1988), ‘Sraffa on income distribution’, Cambridge Journal of Postlewaite, A. (1998), ‘Social status, norms and economic performance’, European
Economics, 12(1), 135–8. Economic Review, 42, 779–800.
Pasinetti, L.L. (1989a), ‘Government deficit spending is not incompatible with the Prskawetz, A., G. Steinmann G. and G. Feichtinger (2000), ‘Human capital,
Cambridge theorem of the rate of profit: a reply to Fleck and Domenghino’, technological progress and the demographic transition’, Mathematical Population
Journal of Post Keynesian Economics, 11(4), 640–7. Studies, 7(4), 343–63.
Pasinetti, L.L. (1989b), ‘Ricardian debt-taxation equivalence in the Kaldor theory of Putnam, R. (1993), ‘The prosperous community: social capital and public life’, The
profits and income distribution’, Cambridge Journal of Economics, 13(1), 25–36. American Prospect, 13, 35–42.
Pasinetti, L.L. (1993). Structural Economic Dynamics. A Theory of Economic Quah, D. (1999), ‘The weightless economy in economic development’, Centre for
Consequences of Human Learning. Cambridge: Cambridge University Press. Economic Performance Discussion Paper n. 417, London School of Economics
Patinkin, D. (1987), entry ‘Real Balances’, in J. Eatwell, M. Milgate and P. Newman Rama, M. (1993), ‘Rent seeking and economic growth. a theoretical model and some
(eds), The New Palgrave Dictionary of Economics, London: Macmillan.
empirical evidence’, Journal of Development Economics, 42, 35–50.
Ramanathan, R. (1982), Introduction to a Theory of Economic Growth, Berlin: Romer, P.M. (1989), ‘Capital accumulation in the theory of long-run growth’, in R.
Springer-Verlag. Barro (ed.), Modern Business Cycle Theory, Cambridge, Massachusetts: Harvard
University Press, pp. 51–127.
Rampa, G. and L. Rampa (1988), ‘Fluctuating growth with induced innovations’,
Metroeconomica, 39(1), 31–42. Romer, P.M. (1990a), ‘Are non convexities important to understanding growth?’,
NBER Working Paper n. 3089.
Ramsey, F.P. (1928), ‘A mathematical theory of saving’, Economic Journal, 38,
543–59. Romer, P.M. (1990b), ‘Endogenous technological change’, Journal of Political
Economy, 98, 71–102.
Razin, A. and U. Ben-Zion (1975), ‘An intergenerational model of population
growth’, American Economic Review, 65(5), 923–33. Romer, P.M. (1991), ‘Increasing returns and new developments in the theory of
growth’, in W.A. Barnett, B. Cornet and C. d’Aspremont (eds), Equilibrium
Rebelo, S. (1991), ‘Long run policy analysis and long run growth’, Journal of
Theory and Applications, Cambridge: Cambridge University Press, pp. 83–110.
Political Economy, 99, 500–21.
Romer, P.M. (1994), ‘The origins of endogenous growth’, Journal of Economic
Remenyi J., (1979), ‘Core demi-core interaction: toward a general theory of
Perspectives, 8(1), 3–22.
disciplinary and subdisciplinary growth’, History of Political Economy, 11(1), 31–
61. Rosenberg, N. (1968), ‘Adam Smith, consumer tastes, and economic growth’,
Retherford, R.D. and J.A. Palmore (1983), ‘Diffusion processes affecting fertility
regulation’, in R.A. Bulatao and R.D. Lee (eds), Determinants of Fertility in Rosenberg, N. (1975), ‘Adam Smith on profits – paradox lost and regained’, in
Developing Countries, vol. 2, New York: Academic Press, pp. 295–339. Skinner and Wilson (1975), pp. 377–89.
Ricardo, D. (1951–73). The Works and Correspondence of David Ricardo, edited by Rosenberg, N. (1982), Inside the Black Box: Technology and Economics, Cambridge:
Piero Sraffa with the collaboration of Maurice H. Dobb, 11 Vols, Cambridge: Cambridge University Press.
Cambridge University Press. Referred to in the text as Works, volume number,
Rosenberg, N. and L.E. Birdzell Jr. (1986), How the West Grew Rich, New York:
page number.
Basic Books.
Richardson, G.B. (1975), ‘Adam Smith on competition and increasing returns’, in
Rosero-Bixby, L. and J.H. Casterline (1993), ‘Modelling diffusion effects in fertility
Skinner and Wilson (1975), pp. 350–360.
transition’, Population Studies, 47, 147–67.
Ricoy, C. (1987), ‘Cumulative causation’, in J.L. Eatwell, M. Milgate and P.
Rowley, C.K., R.D. Tollison and G. Tullock (1988), The Political Economy of Rent-
Newman (eds), The New Palgrave. A Dictionary of Economics, vol. I, London:
Seeking, Boston: Kluwer.
Macmillan, pp. 730–36.
Rowley, R. (1995), ‘History, structure and the wandering natural rate of
Ricoy, C. (1998), ‘Cumulative causation’, in Kurz and Salvadori (1998e), pp. 202–8.
unemployment’, Economie Appliquée, 48(1), 133–55.
Robinson, J. (1956), The Accumulation of Capital, London: Macmillan.
Rowthorn, R.E. (1975), ‘A reply to Lord Kaldor’s comment’, Economic Journal, 85,
Robinson, J. (1962), Essays in the Theory of Economic Growth, London: Macmillan. 897–901.
Rochon, L.P. (1999), Credit, Money and Production, Chelthenam: Edward Elgar. Rowthorn, R.E. (1981), ‘Demand, real wages and economic growth’, Thames Papers
in Political Economy, Autumn, 1–39.
Romer, D. (2000), ‘Keynesian macroeconomics without the LM Curve’, Journal of
Economic Perspectives, 14(2), 149–69. Ryder, H. (1967), ‘Optimal accumulation and trade in a open economy of moderate
size’, in K. Shell (ed.), Essays on the Theory of Optimal Economic Growth,
Romer, P.M. (1986b), ‘Cake eating, chattering and jumps: existence results for
Cambridge, Massachusetts: MIT Press, pp. 87–117.
variational problems’, Econometrica, 54(4), 897–908.
Sacerdote, B. and E.L. Glaeser (2001), ‘Education and religion’, NBER Working
Romer, P.M. (1986c), ‘Increasing returns and long-run growth’, Journal of Political
Paper n. 8080.
Economy, 94(5), 1002–37.
Saint-Paul, G., (1992), ‘Technological choice, financial markets and economic
Romer, P.M. (1987), ‘Growth based on increasing returns due to specialization’,
development’, European Economic Review, 36, 763–81.
Sala-i-Martin, X. (1990), ‘Lecture notes on economic growth (II): Five prototype Setterfield, M., D.V. Gordon and L. Osberg (1992), ‘Searching for a will o’ the wisp:
models of endogenous growth’, NBER Working Paper n. 3564. an empirical study of the NAIRU in Canada’, European Economic Review, 36(1),
119–36.
Salvadori, N. (1998), ‘A linear multisector model of “endogenous” growth and the
problem of capital’, Metroeconomica, 49, 319–35. Shell, K. (1966), ‘Towards a theory of inventive activity and capital accumulation’,
Samuelson, P.A. (1939), ‘A synthesis of the principle of acceleration and the
multiplier’, Journal of Political Economy, 47, 786–97. Shell, K. (1967), ‘A model of inventive activity and capital accumulation’, in Essays
on the Theory of Optimal Economic Growth, Cambridge, Massachusetts: MIT
Samuelson, P.A. (1978), ‘The canonical classical model of political economy’, Press, pp. 67–85.
Journal of Economic Literature, 16(4), 1415–34.
Shell, K. (1974), ‘Inventive activity, industrial organisation and economic growth’, in
Samuelson, P.A. (1985), ‘Modes of thought in economics and biology’, AEA J. Mirrlees and N. Stern (eds.), Models of Economic Growth, London: Macmillan,
proceedings, American Economic Review, 75(2), 166–72. pp. 77–97.
Sandilands, R.J. (2000), ‘Perspectives on Allyn Young in theories of endogenous Shell, K. (1987), ‘Hamiltonians’, in J. Eatwell, M. Milgate and P. Newman (eds),
growth’, Journal of the History of Economic Thought, 22, 309–28. New Palgrave, Vol. 2, New York: W.W. Norton, pp. 588–90.
Sawyer, M. (1997), The NAIRU: A Critical Appraisal, The Jerome Levy Economics Sheshinski, E. (1967), ‘Optimal accumulation with learning by doing’, in K. Shell
Institute Working Paper n. 201.
(ed.), Essays on the Theory of Optimal Economic Growth, Cambridge,
Scheinkman, J. and M. Woodford (1994), ‘Self-organised critically and economic Massachusetts: MIT Press, pp. 35–53.
fluctuations’, American Economic Review, 84, 417–21.
Silverberg, G. and B. Verspagen (1998), ‘Economic growth and economic evolution:
Schmidt-Bleck, F. (1994), Wieviel Umwelt Braucht der Mensch? MIPS – Das Mass a modelling perspective’, in F. Schweitzer and G. Silverberg (eds),
fuer Oekologisches Wirtschaften, Basel/Berlin: Birkhaeuser. Selbstorganisation. Jahrbuch für Komplexität in den Natur-, Sozial- und
Geisteswissenschaften, Berlin: Duncker & Humblot, pp. 239–64.
Schmookler J. (1966), Invention and Economic Growth, Cambridge, Massachusetts:
Harvard University Press. Silverberg, G., G. Dosi and L. Orsenigo (1988), ‘Innovation, diversity, and diffusion:
a self-organising model’, Economic Journal, 98, 1032–54.
Schumpeter, J.A. (1927), ‘The explanation of the business cycle’, Economica, 7(21),
286–311. Simon, H.A. (1978), ‘Rationality as process and as product of thought’, American
Economic Review, 68, p. 1–16.
Schumpeter, J.A. (1934), The Theory of Economic Development, Leipzig, Cambridge,
Massachusetts: Harvard University Press. Skinner, A.S. and T. Wilson (eds), (1975), Essays on Adam Smith, Oxford: Clarendon
Press.
Schumpeter, J.A. (1939), Business Cycles. A Theoretical, Historical and Statistical
Analysis of Capitalist Process, New York: McGraw-Hill. Slutsky, E. (1937), ‘The summation of random causes as the source of cyclic
process’, Econometrica, 5(2), 105–46.
Schumpeter, J.A. (1942), Capitalism, Socialism and Democracy, London: Allen and
Unwin. Smith, A. (1976), The Theory of Moral Sentiments, 1st edn 1759, Vol. I of The
Glasgow Edition of the Works and Correspondence of Adam Smith, edited by A.L.
Segerstrom P. S. (1998), ‘Endogenous growth without scale effects’, American Macfie and D.D. Raphael, In the text quoted as TMS.
Smith, A. (1976), An Inquiry into the Nature and Causes of the Wealth of Nations, 1st
Sen, A.K. (1970), Growth Economics, Harmondsworth: Penguin. edn 1776, Vol. II of The Glasgow Edition of the Works and Correspondence of
Serrano, F. (1995), ‘Long period effective demand and the Sraffian multiplier’, Adam Smith, edited by R.H.Campbell, A.S. Skinner and W.B. Todd, Oxford:
Contributions to Political Economy, 14, 67–90. Oxford University Press. Referred to in the text as WN, book number, chapter
number, section number, paragraph number.
Setterfield, M. (1997), ‘“History versus equilibrium” and the theory of economic
growth’, Cambridge Journal of Economics, 21(3), 365–78. Smulders, S. (1999), ‘Endogenous growth theory and the environment’, in Van den
Bergh J. (ed.), Handbook of Environmental and resource economics, Chelthenam:
Edward Elgar, pp. 656–77.
Snowdon, B. and H.R. Vane (1999), Conversation with Leading Economists, Steindl, J. (1952), Maturity and Stagnation in American Capitalism, Oxford: Basil
Cheltenham: Edward Elgar. Blackwell.
Söderstrom, H. (1976), ‘Production and investment under costs of adjustment: a Stern, D.I. (1997), ‘Progress on the environmental Kuznets Curve?’, Environment and
survey’, Zeitschrift für Nationalökonomie, 36(4), 369–88. Development Economics, 3, 175–98.
Soete, L and R. Turner (1984), ‘Technological diffusion and the rate of technological Stigler, G. (1975), ‘Smith’s travels on the ship of state’ in Skinner and Wilson (1975),
change’, Economic Journal, 94, 612–23. pp. 237–46.
Solow, R. (1956), ‘A contribution to the theory of economic growth’, Quarterly Stiglitz, J.E. (1974), ‘Growth with exhaustible natural resources: efficient and optimal
Journal of Economics, 70(1), 65–94. growth paths’, Review of Economic Studies, 41, 123–37.
Solow, R.M. (1957), ‘Technical change and the aggregate production function’, Stiglitz, J.E. (1986), ‘Comments: some retrospective views on growth theory’, in
Review of Economics and Statistics, 39, 312–20. D. Diamond (ed.), Growth, Productivity, Unemployment, Cambridge,
Massachusetts: MIT Press, pp. 50–67.
Solow, R.M. (1960), ‘Investment and technical progress’, in K. Arrow et al. (eds),
Mathematical Methods in the social sciences, Stanford: Stanford University Press, Stiglitz, J.E. (1993), ‘Endogenous growth and cycles’, NBER Working Paper n. 4286.
pp. 89–104.
Stiglitz, J.E. and A. Weiss (1981), ‘Credit rationing in markets with imperfect
Solow, R.M. (1970), Growth Theory. An Exposition, Oxford: Oxford University information’, American Economic Review, 71, 393–410.
Press.
Stokey N.I. (1995), ‘R&D and economic growth’, Review of Economic Studies, 62,
Solow, R.M. (1974), ‘Intergenerational equity and exhaustible resources’, Review of 469–89.
Stokey, N.L. (1998), ‘Are there limits to growth?’, International Economic Review,
Solow, R.M. (1979), ‘Alternative approaches to macroeconomic theory: a partial 39, 1–31
view’, Canadian Journal of Economics, 12(3), 339–54.
Strassman, P. W. (1956), ‘Economic growth and income distribution’, Quarterly
Solow, R.M. (1997), ‘Is there a core of usable macroeconomics we should all believe Journal of Economics, 70, 425–40.
in?’, American Economic Review, 87(2), 230–2.
Strulik, H. (1997), ‘Learning by doing, population pressure, and the theory of
Solow, R.M. (1999), ‘Neoclassical growth theory’, in J.B. Taylor and M. Woodford demographic transition’, Journal of Population Economics, 10, 285–98.
(eds), Handbook of Macroeconomics, Vol. 1A, Amsterdam: Elsevier, pp. 637–67.
Strulik, H. (1999a), ‘Demographic transition, stagnation, and demoeconomic cycles
Solow, R.M. (2000a), ‘The neoclassical theory of growth and distribution’, Banca in a model for the less developed economy’, Journal of Macroeconomics, 21(2),
Nazionale del Lavoro Quarterly Review, 53, 349–81. 397–413.
Solow, R.M. (2000b), ‘Toward a macroeconomics of the medium run’, Journal of Strulik, H. (1999b), ‘On the mechanics of economic development and non
Economic Perspectives, 14(1),pages. development’, Cambridge University DAE-EO-Paper 9904.
Sordi, S. (2001), ‘Growth cycles when workers save: a reformulation of goodwin’s Strulik, H. (2000), ‘On demographic transition, structural change, and economic
model along Kaldorian–Pasinettian lines’, Central European Journal of growth and stagnation’, Mathematical Population Studies, 8, 333–56.
Operations Research, 9(1–2), 97–117.
Sturzenegger, F. and M. Tommasi (1994), ‘The distribution of political power, the
Sraffa, P. (1960), Production of Commodities by Means of Commodities: Prelude to a costs of rent-seeking, and economic growth’, Economic Inquiry, 23, 236–48.
Critique of Economic Theory, Cambridge: Cambridge University Press.
Swan, T.W. (1956), ‘Economic growth and capital accumulation’, Economic Record,
Stadler, G. (1990), ‘Business cycle models with endogenous technology’, American 32, 334–61.
Economic Review, 80(4), 763–78.
Tamura, R. (1996), Human Capital, Specialization, and Industrial Revolution,
Steedman, I. (1972), ‘The state and the outcome of the Pasinetti process’, Economic Clemson University, mimeo.
Journal, 82, 1387–95.
Targetti, F. (1991), ‘Change and continuity in Kaldor’s thought on growth and
distribution’, in Nell and Semmler (1991), 411–20.
Temple, J. and P. Johnson (1998), ‘Social capability and economic growth’, Varri, P. (1990), ‘Introduzione’, in Harrod, R.F. Dinamica Economica, Bologna: Il
Quarterly Journal of Economics, 113(3), 965–90. Mulino, pp. 10–37
Thirlwall, A.P. (1979), ‘The balance of payments constraint as an explanation of Verspagen, B. (2001), ‘Economic growth and technological change: an evolutionary
international growth rate differences’, BNL Quarterly Review, n. 128. interpretation’, STI Working Papers 2001/1.
Thirlwall, A.P. (1983), ‘Foreign trade elasticities in centre-periphery models of Vianello, F. (1985), ‘The pace of accumulation’, Political Economy, 1(1), 69–87.
growth and development’, BNL Quarterly Review, n. 146, 249–61.
Vianello, F. (1989), ‘Effective demand and the rate of profit: some thoughts on Marx,
Thirlwall, A.P. (1986), ‘A general model of growth and development on Kaldorian Kalecki and Sraffa’, in M. Sebastiani (ed.), Kalecki’s Relevance Today, New
lines’, Oxford Economic Papers, 38(2), 199–219. York: St. Martin’s Press, pp. 164–90.
Thirlwall, A.P. (1987), Nicholas Kaldor, Brighton: Wheatsheaf Press. Vianello, F. (1996), ‘Joan Robinson on normal prices (and the normal rate of
profits)’, in M.C. Marcuzzo, L.L. Pasinetti and A. Roncaglia (eds), The Economics
Thirlwall, A.P. (1991), ‘Professor Krugman’s 45 degree rule’, Journal of Post
of Joan Robinson, London: Routledge, pp. 112–20.
Keynesian Economics, 13(1), 23–8.
Neumann, J. von (1945), ‘A model of general economic equilibrium’, Review of
Thirlwall, A.P. (1998), ‘The balance of payments and growth: from mercantilism, to
Economic Studies, 13, 1–9. English translation of ‘Über ein ökonomisches
Keynes to Harrod and beyond’, in G. Rampa, L. Stella and A.P. Thirlwall (eds),
Gleichungssystem und eine Verallgemeinerung des Brouwerschen
Economic Dynamics, Trade and Growth. Essays on Harrodian Themes, London:
Fixpunktsatzes’, in Ergebnisse eines mathematischen Kolloquiums, 8 (1937), 73–
Macmillan Press, pp. 179–204
83.
Thirlwall, A.P. and M.N. Hussain (1982), ‘The balance of payments constraint,
Von Waizsaker, C.C. (1974), ‘Notes on endogenous growth of productivity’, in
capital flows and growth rate differences between developing countries’, Oxford
J. Mirrlees and N. Stern (eds.), Models of Economic Growth, London: Macmillan,
Economic Papers, 34(3), 498–510.
pp. 101–11.
Toman, M.A., J. Pezzey and J. Krautkraemer (1995), ‘Neoclassical economic growth
Wan, H. (1970), Economic Growth, New York: Arcourt-Bruce.
theory and sustainability, in D.W. Bromley (ed.)., Handbook of Environmental
Economics, Oxford: Blackwell, pp. 139–65. Weber, M. (1930), Protestant Ethic and the Spirit of Capitalism, trans. T. Parson,
New York: Scribner & Sons.
Townsend, R. (1979) ‘Optimal contracts and competitive markets with costly state
verification’, Journal of Economic Theory, 21(2), 265–93. Weiss, Y. and C. Fershtman (1998), ‘Social status and economic performance: a
survey’, European Economic Review, 42, 801–20.
Treiman, D. (1977), Occupational Prestige in Comparative Perspective, New York:
Academic Press. Weitzman, M.L. (1971), ‘Shiftable versus non-shiftable capital: a synthesis’,
Trezzini, A. (1995), ‘Capacity utilisation in the long run and the autonomous
components of aggregate demand’, Contributions to Political Economy, 14, 33– Wilson, T. (1976), ‘Effective devaluation and inflation’, Oxford Economic Papers,
66. 28(1), 1–24.
Trezzini A. (1998), ‘Capacity utilisation in the long run: some further considerations’, Winter, S.G. (1971), ‘Satisficing, selection and the innovating remnant’. Quarterly,
Contributions to Political Economy, 19, 19–32. Journal of Economics, 85, 237–61.
Tullock, G. (1967), ‘Welfare costs of tariffs, monopolies and theft’, Western Wolfstetter, E. (1982), ‘Fiscal Policy and the classical growth cycle’, Zeitschrift für
Economic Journal, 5, 224–32. Nationalökonomie, 42(4), 375–93.
U.S. Census Bureau (2001), Survey of Plant Capacity 1999, U.S. Department of Woolcock, K. (2000), ‘The place of social capital in understanding social and
Commerce, March. economic outcome’, mimeo, The World Bank.
Uzawa, H. (1965), ‘Optimum Technical change in an aggregate model of economic World Commission on Environment and Development (Brundtland Commission)
growth’, International Economic Review, 6, 18–31. (1987), Our Common Future, Oxford: Oxford University Press.
References 393
Wray, L.R. (1988), ‘The monetary explanation of distribution: a critique of Pivetti’,

political economy, 4(2), 269–73.
Xie, D. (1994), ‘Divergence in economic performance: transitional dynamics with
multiple equilibria’, Journal of Economic Theory, 63(1), 97–112.
Yang, X. (1999), ‘The division of labor, investment and capital’, Metroeconomica,
50, 301–24.
Yip, C.K. and J. Zhang (1997), ‘A simple endogenous growth model with
endogenous fertility: indeterminacy and uniqueness’, Journal of Population
Economics, 10, 97–110.
Young, A.A. (1928), ‘Increasing returns and economic progress’, Economic Journal,
38, 527–42.
Young, A.A. (1998), ‘Growth without scale effects’, Journal of Political Economy,
106, 41–63.
Young, A.A. (1999a), ‘Particular expenses and supply curves, (Nicholas Kaldor’s
notes on Allyn Young’s LSE Lectures, 1927–29)’, in P.G. Mehrling and R.J.
Sandilands (eds), Money and Growth. Selected Papers of Allyn Abbot Young,
London: Routledge, pp. 391–8.
Young, A.A. (1999b), ‘Supply and demand’ (from the Encyclopedia Britannica,
1928), in P.G. Mehrling and R.J. Sandilands (eds), Money and Growth. Selected
Papers of Allyn Abbot Young, London: Routledge, pp. 143–6.
Young, W. (1989), Harrod and his Trade Cycle Group, London: Macmillan.
Zak, P.L. and S. Knack (2001), ‘Trust and growth’, Economic Journal, 111, 295–321.
Zenezini, M. (1990), ‘Esistono davvero le funzioni neoclassiche di domanda di
lavoro?’, Politica Economica, 8(April), 19–59.
Zweimuller, J. (2000), ‘Schumpterian entrepreneurs meet Engel’s law: the impact of
inequality on innovation-driven growth’, Journal of Economic Growth, 5, 185–
206.

The Theory of Economic Growth - A 'Classical' Perspective PDF

Uploaded by

Copyright:

Available Formats

The Theory of Economic Growth - A 'Classical' Perspective PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Theory of Economic Growth - A 'Classical' Perspective PDF

Uploaded by

Copyright:

Available Formats

The Theory of

10. The evolutionary perspective on growth 205

11. Competition, rent seeking and growth: Smith versus the

12. R&D models of economic growth and the long-term

9. Human capital formation in the new growth theory:

accumulation. We argue that in Ricardo the growth rate is endogenous and

the economic process. In fact, while the neoclassical theory of factor–income

Inspired by the article by Ramsey (1928), several growth models have

2. For a more complete elaboration on this point see Rosenberg (1975).

3.2. THE DEVELOPMENT OF THE NEO-CLASSICAL

In this section and in the following, we will try to characterize neo-classical

question the view of economic processes underlying the analytical structure.

3.9. CONCLUDING REMARKS

In this paper I have viewed the development of neoclassical growth theory

1. See Backhouse (1998) for a discussion of the Lakatosian methodology applied to

where s T = xT B − δ x xT A − δ z (sT − xT A) − cˆ T (P)

u(⋅) : ℜc+m →ℜ ∪ {−∞}, 1 ≤ cm ≤ n . x A ≤ s , x ≥ 0, c ≥ 0, s(0) = s ≥ 0 given.

xT0 [I − ( ρ + δ x + gσ )A]v 0 = 0 (3c) r−ρ

Figure 4.1 [B − (r + δ x )A]v 0 ≤ 0 , xT0 [B − (r + δ x )A]v 0 = 0

. In Figure 4.3, with a zero price.

consumption share, in value, as constant and depending only on α : along

c1 = x1 − (1 + g)s1 > 0, c2 = x2 − (1 + g)s2 > 0 1 + 6α

Figure 4.4a Figure 4.4b

Let (P) be the problem

wF = w ( LF , aF ) (2) 5.5. EQUILIBRIUM ANALYSIS

where O is the market output. A good is produced if (expected) profits are

To complete the description of equilibrium it is necessary to calculate total

is the mark-up in the monopolist markets of industrial goods and

is a necessary and sufficient condition for the existence of a manufacturing

Aggregate profits are a positive function of aggregate rents R, of the level

distribution in favour of workers) is greater than the negative effect of higher i

7.1. INTRODUCTION* 7.2. THE POSSIBLE ROLE OF DEMAND IN OUTPUT

influence the growth rate of Y/L. .

subject to the restraints:

Ω The simplest ‘model’ of the DT used by demographers represents the

Clearly n has a humped relation with per-capita capital as well. We

Remark 1. Formulation (8) is well acknowledged in the literature on

rich equilibrium khigh can be interpreted as a ‘modern’, post-transitional,

A) In the model with only endogenous population (Section 8.5.1) we obtain

Tr ( J ) = − (1 − γ ) sU1 − (1 − β ) n' ( y ) y , Det ( J ) = sUn' ( y ) y (1 − β − γ ) > 0 . NOTES

First-generation endogenous growth models, assuming human capital

As far as entrance is concerned, potential entrants are classified into two

⎡ ⎤ 11.4. SMITH’S VIEW ON RENT SEEKING,

differences in technology, profitability, knowledge and luck. In equilibrium

profitable avenues such that more entrepreneurs are induced to innovate,

Equation (1) is a standard Keynesian consumption function, where C

From (7), assuming (as in Goodwin, 1951, p. 12) that autonomous

where Y c = Y − I a* / s . As is well known (see, for example, Gandolfo, 1997,

To assess the importance of this result, however, it is useful to go back to the

I a (t ) = b ⎡⎣1 + sin (ct )⎤⎦ (9)

Figure 15.5: Dynamics of the non-linear accelerator model with

15.3. THE ‘MARXIAN’ GROWTH CYCLE MODEL

Sc = sc Pc = sc rKc • in order to be economically meaningful, it requires that the Pasinettian

Thus, expressing (19) in terms of rates of growth

• they imply the Cambridge equation:

A(τ H ) B(τ H ) − C(τ H ) = 0 (30)

∂ [ A(τ ) B(τ ) − C(τ )]

Now, it is possible to prove (see Asada, 1995, p. 248) that, under

at = at1−−1δ a ytλ−1 (7)

where s T = xT B − δ x xT A − δ z (sT − xT A) − cˆ T (P)

↔ − (1 − α − σ )αρ H2 k A22 + ρ H σ k A 2 (1 − α ) + α (1 − α ) < 0