Patterns of Structural Change

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/247317518
Patterns of Structural Change
Article in Handbook of Development Economics · December 1988

DOI: 10.1016/S1573-4471(88)01010-1
CITATIONS READS
241 7,596
1 author:
Moshe Syrquin
University of Miami
53 PUBLICATIONS 787 CITATIONS
SEE PROFILE
All content following this page was uploaded by Moshe Syrquin on 09 April 2018.
The user has requested enhancement of the downloaded file.

Chapter 7
PATTERNS OF STRUCTURAL CHANGE
MOSHE SYRQUIN*
Bar-llan University
Contents
0. Introduction 205
1. The study of structural change 206
1.1. Structural change in economic history 209
1.2. Structural change in development economics 211
1.3. The need for a typology 214
2. Empirical research on the structural transformation 216
2.1. Bases for comparative analysis 216
2.2. A unique path of development? 217
2.3. The methodology of comparative analysis 218
2.4. Time-series vs. cross-section studies 221
3. Patterns of growth and accumulation 223
3.1. Growth patterns 223
3.2. Accumulation 225
4. Changes in sector proportions 228
4.1. The accounting framework 230
4.2. Final demand 231
4.3. Intermediate demand 231
4.4. Trade 232
4.5. Structure of production: Broad sectors 235
4.6. Post-war patterns 239
4.7. Manufacturing: Disaggregated results 242
5. Structure and growth 243
5.1. Sectoral contributions to growth 244
5.2. Typology of development patterns 248
*I am grateful to Hollis Chenery, James Ito-Adler, Howard Pack, T.N. Sfinivasan, Lance Taylor,
and Adrian Wood for their helpful comments and suggestions, and to Yosi Deutsch for help and
advice on the empirical sections.
Handbook of Development Economics, Volume L Edited by H. Chenery and T.N. Srinivasan

© Elsevier Science Publishers B. K, 1988
204 Ch. 7: Patterns of Structural Change
Contents (continued)
6. A c c o u n t i n g for the transformation 250

6.1. Growth accounting: Demand side decomposition 250
6.2. Growth accounting: Supply side 254
6.3. Resource shifts and productivity growth 255
7. Relative prices and exchange rate conversions 258
7.1. Relativeprices 259
7.2. Exchange-rate conversions 262
8. A p p r o a c h e s to policy 265
References 268
Ch. 7: Patterns of Structural Change 205
O. Introduction
Development economics can be characterized as dealing with issues of structure

and growth in less developed countries. 1 Analysis of structure appears in two
variants. The first, and more recent, is concerned with the functioning of
economies, their markets, institutions, mechanisms for allocating resources, in-
come generation and its distribution, etc. This is primarily a micro approach,
solidly anchored in economic theory with little emphasis on economic history or
long-run processes of structural change. In the second variant, economic develop-
ment is seen as an interrelated set of long-run processes of structural trans-
formation that accompany growth. The central features of this approach are
economy-wide phenomena such as industrialization, urbanization, and agricultur-
al transformation, regarded as elements of what Kuznets identified as "modern
economic growth". This is essentially a comparative approach deriving its infor-
mation from the historical evolution of the advanced economies and from
intercountry associations of structural changes and growth.
Most of the early literature on development was "structural" in this sense.
More recently, the supply shocks of the 1970s and stabilization programs shifted
attention to shorter-run issues. However, even in this framework it is important
to regard the long-run tendencies of the economy. This and the following
chapters on structural transformation are mostly about such long-run processes.
This chapter deals with the patterns of structural transformation during the
transition from a low income, agrarian rural economy to an industrial urban
economy with substantially higher per capita income. 2 Section 1 reviews the basic
concepts of the empirical research program 3 into the economic structure of
developing countries during the transition process, which originated with the
monumental work of Simon Kuznets. Section 2 deals with methodological issues
of empirical research on structural transformation. A summary of the main
stylized facts of development, with emphasis on growth, accumulation and sector
proportions, is presented in Sections 3, 4 and 5. Attempts to model and explain
the transformation are presented in Section 6. In Section 7 relative prices are
considered, and Section 8 deals with the role of the state in facilitating, fostering,
or at times hampering, an efficient transformation.
1Lewis (1984) defines the scope of development economics as dealing "with the structure and
behavior of economieswhere output per head is less than 1980 US$ 2,000" (p. 1).
2Issues related to the adequacy of seeing the process of developmentas a transition (smooth or
otherwise) and to the uniqueness of the path are discussed below.
3This follows Chenery's (1986a) characterizationof the work on structural transformation as a
research program.
206 M. Syrquin
1. The study of structural change
There are many uses of the concepts of structure and structural change in
economics. Some of them have a clear meaning or are made clear by the context,
while others are vague or worse. 4 This chapter deals with structural change in
development economics and, therefore, employs the concepts and meanings
prevalent in the field which, as explained below, differ from the concept in
econometrics of the structure of a model. A brief discussion may be useful to
clarify the main uses of the terms in this chapter.
The most c o m m o n use of structure in development and in economic history
refers to the relative importance of sectors in the economy in terms of production
and factor use. Industrialization is then the central process of structural change.
In this sense - structure as the composition of an aggregate - the term is also
applied to other aggregates that have some bearing on the process of {ndustriali-
zation such as demand and trade)
Following common use, structure also refers to some ratios derived from
technological or behavioral relations. I n p u t - o u t p u t coefficients are an example of
the former and the aggregate saving ratio of the latter. The principal changes in
structure emphasized in the development literature are increases in the rates of
accumulation (Rostow, Lewis); shifts in the sectoral composition of economic
activity (industrialization) focusing initially on the allocation of employment
(Fisher, Clark) and later on production and factor use in general (Kuznets,
Chenery); and changes in the location of economic activity (urbanization) and
other concomitant aspects of industrialization (demographic transition, income
distribution).
The interrelated processes of structural change that accompany economic
development are jointly referred to as the structural transformation.
The accumulation of physical and human capital and shifts in the composition
of demand, trade, production, and employment are described [following Chenery
(1986a)] as the economic core of the transformation, while the related socio-eco-
nomic processes are identified as peripheral. It is in this sense that the term
structure is used in this chapter. Its scope is restricted to those economic aspects
deemed to be relevant for the analysis of growth although the process of modern
economic growth is clearly more encompassing. In addition to the elements of
transformation mentioned above, for example, it considers changes in institutions
by which structural change is achieved. This wider framework is often acknowl-
4Machlup (1963) is still the best source for fhe various ways in which the terms have been used and
abused in economics.
5Kuznets (1959, p. 162) defined structure as "... a relatively coherent framework of interrelated
parts, each with a distinctive role but harnessed to a set of common goals".
Ch. 7." Patterns of Structural Change 207
edged, though seldom represented in empirical work [for notable exceptions see
Kuznets (1971) and Adelman and Morris (1967)]. 6
The structure of a model in the econometric sense implies something else. A
model is an abstraction, a simplified representation of an economy or of a certain
aspect of it. When formally laid out, the postulated relations and the parameters
of the equations represent the structure of the model. Clearly, structure in this
sense is model-specific, and structural change reflects how general the model is. A
better specified, more comprehensive model, captures endogenously what in a
narrower model would appear as change in structure. At the risk of belaboring
this point, two examples are presented, referring to accumulation and to sector
proportions.
(1) Changes in structural parameters may reflect specification errors, or omitted
variables. A doubling of the saving ratio may appear as a structural change in a
model where the saving relation to income was assumed to be linear or in a
model that omitted growth effects on saving. In alternative models the same rise
in the saving ratio would be described as a movement along unchanged relations.
In this paper, a long-term increase in the saving ratio is always regarded as an
element of structural transformation.
(2) Changes in sector proportions are implied by a variety of models. In a
small open economy producing two tradable goods, capital accumulation leads to
a change in the relative weight of the two sectors in an unambiguous predictable
way. In development economics, and in this Chapter, such a change in the
relative importance of sectors is defined as a structural change.
A definition, when clear and used consistently, may or may not be useful but it
is pointless to argue about its correctness.
Sections 2 and 3 present empirical results on patterns of structural change. The
estimated relations can be interpreted as reduced forms from an underlying
structural model. Reduced forms are most useful for studying structural change
(as defined in this chapter) when the underlying model is stable, that is, when the
structure (in the econometric sense) does not change.
The preceding discussion is also pertinent to the concept of equilibrium. 7 As
with the structure of a model, equilibrium can only be defined in relation to a
specific model and to the variables included. A position can be both, one of
equilibrium for a certain model and one of disequilibrium for a different model.
6North (1981) gives an interpretation of structural change in economic history as institutional

change, but almost completely omits shifts in the structure of production and factor use. The
contribution of institutional change to economic growth is the subject of Matthews' presidential
address to the Royal Economic Society (1986).
71 adopt Machlup's definition of equilibrium as "as constellation of selected interrelated variables
so adjusted to one another that no inherent tendency to change prevails in the model which they
constitute" [Machlup (1963, p. 54)].
208 M. Svrquin
In the cobweb model, for example, a position "will be both an equilibrium or a

disequilibrium, depending on the length of time that is taken into account"
[Machlup (1963, p. 52)].
The main instance where the issue appears in this chapter is the case of
intersectoral differences in factor returns. They are here regarded as evidence of
disequilibrium which is to say, a temporary situation "that cannot endure if left
alone, that must give rise to further change" [Machlup (1963, p. 67)]. Change
takes place in the form of migration for example. Though temporary, the
disequilibrium could be of long duration if migration is slow, or if the initial
shock that created the disequilibrium (differential productivity growth say)
appeared as a recurrent phenomenon. It is always possible to build a wider
model; one that considers migration, barriers to mobility, costs of adjustment
and so on, for which the existence of gaps in marginal products among sectors
would not be incompatible with equilibrium. The solution of the model would
include an equilibrium rate of migration, etc.
Wage gaps among sectors could also be the result of an exogenously imposed
constraint that prevents their disappearance. It is doubtful whether we want to
refer to this situation as disequilibrium, unless we equate equilibrium with a
standard of performance.
An obvious reason for studying structural change is that it is at the center of
modern economic growth. It is, therefore, an essential ingredient for describing
the process and for the construction of any comprehensive theory of develop-
ment. More important is the hypothesis that growth and structural change are
strongly interrelated. Most writers recognize their interdependence, and some
emphasize the necessity of structural changes for growth. For Kuznets, "some
structural changes, not only in economic but also in social institutions and
beliefs, are required, without which modern economic growth would be impossi-
ble" [Kuznets (1971, p. 348, emphasis in the original)]. Chenery views economic
development "as a set of interrelated changes in the structure of an economy that
are required for its continued growth" [Chenery (1979, p. xvi)]. The interdepen-
dence also appears as a cumulative process: "Sectoral redistribution of output
and employment is both a necessary condition and a concomitant of productivity
growth" [Abramovitz (1983, p. 85)]; or more guardedly: "Neither structural
change nor growth in GDP is an exogenous variable; both result from a complex
of interacting causes on the supply side and the demand side" [Matthews,
Feinstein and Odling-Smee (1982, p. 250)].
In the absence of a continuous equalization of factor returns across sectors, the
reallocation of resources to sectors of higher productivity contributes to growth.
In such disequilibrium situations, structural change becomes a potential source of
growth if it leads to a fuller or better utilization of resources. The potential gains
are likely to be more important for developing countries than for developed ones
since the former exhibit more pronounced symptoms of disequilibrium and can
achieve faster rates of structural change. In a dynamic context the gains can be
far from negligible, accounting for as much as one-third of the measured growth
in total factor productivity (see Section 6.3.2).
Much of the interest in the structural transformation derives from its possible
implications for development policy. The range of policy analysis in the study of
growth and transformation is vast. At the analytic end there are attempts to
assess the effects of policies pursued on the observed, historical patterns of
structural change. At the prescriptive end the roles assigned to policy have
included that of facilitating change by, for example, removing barriers to mobil-
ity; coordinating the changes in demand, trade, production, and primary inputs
to prevent bottlenecks from arising; and directing and stimulating change which
otherwise might be defeated by inadequate or non-existent markets. All these
approaches are represented in the policy sections in Chapters 8-11. A general
discussion of economic policy and structural change appears in Section 8.
1.1. Structural change in economic history
The following sections present a selective review of the evolution of thought on

structural change in economic history and in theories of development. They
distinguish between structural change seen as an economy-wide phenomenon
interacting with aggregate growth, and partial changes which may well be
important elements in the overall framework but are analyzed as isolated
phenomena. The most famous instance of the latter is Engel's study of consump-
tion expenditures.
Economic historians have tended to concentrate on specific features, while
paying less attention to more general or to economy-wide aspects of transforma-
tion. This assertion is based on the authority of a noted economic historian:
The historians, however, have generally neglected structural Change as an
historical phenomenon, and, in particular, have usually ignored its role in
economic growth . . . . The main characteristic of the analyses of economic
growth by the historians has been their almost unqualified belief in the prime
importance of the industrial sector with its increasing physical capital forma-
tion embodying improved technology. Generally, there has been a comparative
neglect of other factors. . . . and almost complete neglect of structural change as
a source of increasing productivity [Hartwell (1973, pp. 389, 391)].
One historical approach that emphasizes structural transformation is the
"stages approach"8 represented in the modem theories of growth and develop-
ment by Rostow (1960). The central stage-the take-off-features two elements
8For a reviewof "stages"theoriesof growth see Hoselitz(1960).

210 M. Syrquin
of discontinuity that, in one way or another, appear in most discussions on

structural change. 9 One is the sharp increase (doubling) of the rate of capital
accumulation, the other is the concept of the leading sector, with its implied
transformation of the productive structure. Although it made a great impact on
contemporary theories of development, this approach fell into disrepute after it
was severely criticized. Even today writers using the terms "phases" or "stages"
invariably make sure todisassociate themselves from the original. The main
criticisms focused on the notion of a unique path of development, on the absence
of endogenous mechanisms of transition between stages, and on the concept of
necessary "prerequisites" for the take-off.
The last point was dealt with by Gerschenkron in the title essay of his 1962
collection and in the postscript. He pointed out that processes of rapid industri-
alization started, in several European countries, from different levels of economic
backwardness, and made the course and character of industrialization depend on
the initial degree of such backwardness. The focus is then shifted from the set of
prerequisites to a search for ways in which a missing prerequisite was substituted
for and to an attempt to link the pattern of substitution with the degree of
backwardness. This is illustrated, for example, in Gerschenkron's analysis of the
sources of capital supply for industrialization; summarized in his propostion that
"the more backward a country's economy [on the eve of its industrialization], the
greater was the part played by special institutional factors [banks, the State]
designed to increase supply of capital to the nascent industries" [Gerschenkron
(1962, p. 354)]. The possibility of a variety of routes to reach a certain target is
central to the notion of alternative paths of development. It can be formulated as
a testable proposition: the variance of a certain process among countries is
expected to be larger the more disaggregated the level of analysis.
A leading sector propagates growth through its "linkages" to other segments of
the economy. This term, coined by Hirschman (1958), is at the core of the staples
approach1° derived in the context of the economic history of Canada and other
regions of "recent settlement".11 In this approach, growth and structural change
are interpreted in terms of the production characteristics of the dominant natural
resource (the staple) being exploited for external markets. The literature on the
staple theory of growth contains examples of episodes of fast. growth based on
the exploitation of natural resources extending over long periods of time that,
nevertheless, did not lead to continuous development when structural transfor-
9The "two theories" of take-off follows the interpretation of Fishlow (1965).

10In a later study, Hirschman (1977), generalized the linkage approach to development with special
reference to the staples. Leading sectors and cumulative processes are also featured in Myrdal (1957)
and Perroux (1955).
11See Baldwin (1956), Watldns (1963), and Caves (1965). In the more recent literature, shorter run
impact of resource discoveries and price booms are discussed in relation to the Dutch disease [Corden
and Neary, (1982), Roemer (1985)]. Stephen Lewis discusses the approach in Chapter 30, Volume II,
of this Handbook.
mation (including the social and institutional structure) did not keep apace.
Episodes of "growth without development" followed by stagnation are now
common in dependency approaches that identify socio-economic structures as the
root cause of underdevelopment.
In the staple approach there is surplus to be vented in response to external
demand. This key role given to demand in igniting the process, is apparently the
common view in economic history. Landes (1969, p. 77) sums up his discussion of
demand and supply factors in the industrial revolution in Britain thus: "it was in
large measure the pressure of demand on the mode of production that called
forth the new techniques in Britain, and the abundant, responsive supply of the
factors that made possible their rapid exploitation and diffusion".
1.2. Structural change in development economics
The emphasis on demand, reinforced by Keynesian theory, greatly influenced the

early writings on development economics. The dynamic version of the Keynesian
model (Harrod-Domar); dual-economy models (Lewis); demand complementar-
ity, balanced growth, and "big-push"; these are among the central concepts of
the 1950s. They all feature demand as the centerpiece of analysis and relegate
trade to a minor position. In this vintage of development economics we recognize
the two components of the economic core of the transformation: accumulation
and sectoral composition. Both had policy implications, the former at the
aggregate level and the latter, by its nature, at a disaggregated level but in an
economy-wide framework. Accelerating and sustaining growth required increas-
ing the rates of accumulation and maintaining sectoral balance to prevent
disequilibrium in product markets, or to overcome disequilibrium prevailing in
factor markets.
At about the same time neoclassical theory responded to the structuralist
challenge posed by the Harrod-Domar model and came out with its own growth
theory from the supply side [Solow (1956)]. Its implications were starkly different.
In the aggregate version, there is no surplus labor and long-run growth is
independent of the saving rate. In multisectoral versions of the von-Neumann
type, growth, still independent of the saving rate, proceeds in a balanced fashion
and no disequilibrium is allowed. These features may account for the initial
limited impact of neoclassical growth theory on development economics. To-
gether with steady-state theory, overboard went prices, incentives and market
behavior, until the excesses of the policies pursued and the drastic change in the
prices of fuels and other primary commodities in the early 1970s led to the
neoclassical resurgence.
I return now to the two main components of the transformation in develop-
ment economics: accumulation and sectoral composition. In the 1950s accumula-
212 M. Syrquin
tion almost invariably referred to physical capital in commodity production and
infrastructure. Capital appeared as the critical factor in the H a r r o d - D o m a r
model. As mentioned above, Rostow (1960) emphasized the sharp increase in the
rate of investment during the take-off stage. A doubling in the investment rate
was also seen as indispensable by Arthur Lewis: " T h e central problem in the
theory of economic development is to understand the process by which a
community which was previously saving and investing 4 or 5 per cent of its
national income or less, converts itself into an economy where voluntary saving is
running at about 12 to 15 per cent of national income or more" [Lewis (1954, p.
155)1.
In Lewis's model the shift of resources to the m o d e m sector increased the
profit share in income and thus raised the saving rate. " T h e central fact of
economic development is that the distribution of incomes is altered in favour of
the saving class" (p. 156). Two important early developments can be seen as
attempts to specify the role of capital. For the closed economy, Mahalanobis
(1953) argued that with non-shiftable capital the key planning problem is the
allocation of investment between sectors producing consumption and production
goods, or how many machines to use in making machines} 2 For the open
economy Chenery and his collaborators introduced foreign-exchange require-
ments as an additional constraint on growth besides the limitation imposed by
savings [Chenery and Bruno (1962), Chenery and Strout (1966)]. These develop-
ments were made an integral part of disaggregated dynamic input-output models
and of optimizing multisector programming models} 3
The main message of these studies and of the later emphasis on human
resources is that a sustained increase in rates of accumulation, while not suffi-
cient, is a necessary requirement for long-run growth and transformation.
Modern analyses of sectoral transformation originated with Fisher (1935, 1939)
and Clark (1940), and dealt with sectoral shifts in the composition of the labor
force. As in most areas in economics one can find precursors of their ideas in
earlier writings} 4 However, they were probably the first to deal with the process
of reallocation during the epoch of modern economic growth, and to use the form
of sectoral division ( p r i m a l - s e c o n d a r y - t e r t i a r y ) which, in one way or another,
is still with us today. The specific division into sectors was influenced by the
conditions and debates in New Zealand. Its value was questioned in New
Zealand is in particular and for low income countries in general. The latter
primarily by Bauer and Yamey (1951), who argued that in the early stages of
12A similar model was developedin the mid-1920sby the Soviet economist Feldman. The modern
reformulation is due to Domar (1957). See also Taylor (1979).
~3These models and many more are su~'eyed in Taylor (1975).
14Sir William Petty and FfiedrichList among others. See Clark (1940) and Hoselitz (1960).
15Fisher's 1939 paper was largely devoted to justifying and clarifyinghis definition of "tertiary
production" first introduced in 1933 and popularizedby his 1935 book.
development individuals are engaged in a variety of trades, and it is therefore
difficult to assign them to any one occupation. With m o d e m economic growth the
size of the market increases and with it the degree of specialization and the
differentiation of occupations. The difficulty - an empirical one - diminishes in
importance as development proceeds.
Clark's approach was predominantly empirical, but he did relate the observed
shifts to differential productivity growth and Engel effects, the two principal
elements in subsequent attempts to account for the transformation in the
structure of production. Clark and Fisher studied economy-wide phenomena.
Hoffmann's (1958) attempt to derive a law of industrialization was similar in
approach but restricted to industry only. In the early phases of industrialization,
according to Hoffmann, the ratio of consumer goods to producer goods is as high
as 4 to 1. The ratio declines steadily during the process and reaches a value of
about 1 to 1 or less, in the more advanced stages. As with many such efforts, the
main value of Hoffmann's analysis was probably in the careful compilation of
historical data and in stimulating discussion. 16
In addition to the need to increase the rate of capital formation, theories of
development of the 1950s stressed sectoral differences. In Lewis's model sectoral
differences appear as traditional versus modern sectors, and in Nurkse (1953) and
Rosenstein-Rodan (1943, 1961) as a requirement for balanced growth. These
approaches shared some views of the functioning of less developed economies:
labor surplus in agriculture, low mobility of factors, price-inelastic demands,
export pessimism, and a general distrust of the market. These are the hallmarks
of what Little (1982) characterizes as the "structuralist view". They found a
policy echo in the advocacy of inward-oriented strategies and planning primarily
in Latin America and South Asia. Although this is a valid characterization of the
prevailing views at the time, it remains necessary to recognize the diversity and
evolution of ideas along a continuum.
On the empirical side, studies of long-run transformation are best represented
by Kuznets' synthesis of modern economic growth in a series of seminal papers. 17
Kuznets established the stylized facts of structural transformation, but was
reluctant to offer a theory of development. He saw his analysis as an essential
building block towards such a theory. His approach, however, is quite different
from the empiricism of earlier writers. Economic theory guided his choice of
concepts and the all-encompassing interpretations that accompanied every statis-
tical finding. His essays on modern economic growth are a compendium of ideas
on growth, transformation, distribution, ideology, institutions, and their interrela-
tions. General equilibrium modellers can find in these essays a rich source for
16For a critical evaluation see Kuznets (1966).

17"Quantitative Aspects of the Economic Growth of Nations" (1956-67). More compact state-
ments are the 1966 monograph and the Nobel lecture (1973).
214 M. Syrquin
ideas, a guide to specification, and to the long-run relations against which to
calibrate their models.
The amount of information assembled by Kuznets was enormous, but he did
not use formal statistical techniques in its analysis. This task was later taken up
by Chenery (1960). Before that, however, Chenery had contributed to the study
of investment criteria in a multisectoral framework in a situation of disequi-
librium, and to the development of input-output and programming models. For
Chenery, the possibility of disequilibrium and differences among sectors implied
the need for coordination to anticipate potential bottlenecks. TM His 1960 "Pat-
terns of Industrial Growth" fit well in this approach as an attempt to determine
the " n o r m a l " transformation in the structure of production as income grows. In
subsequent studies the "normal" pattern gives way to a standard or expected one
devoid of normative connotations: "Probably the most that should be claimed
for statistical comparisons of this sort is that they are helpful in diagnosing the
structural problems of a given country and in suggesting feasible growth patterns"
[Chenery and Syrquin (1975, p. 137)]. The patterns of production structure were
initially derived from a general equilibrium system as reduced forms and, in later
studies, were integrated with demand and trade in wider models. 19
In the earlier models of Chenery, prices (accounting, shadow or otherwise) are
explicitly considered, some (the exchange rate) playing an important allocative
role. In the patterns approach they are silent bystanders. The reasons for not
explicitly considering prices and the impact of this on the long-run patterns are
discussed below.
1.3. The need for a typology
An additional feature of Chenery's approach, important for the links between

growth and change, is the issue of typology. In the 1960 study only income and
size were considered as explanatory variables. Shortly thereafter, natural re-
sources were added to the analysis [Chenery (1964)], but they proved difficult to
handle within the framework of a unique cross-country regression. The main
problem was one of measuring economic resources. The various proxies of
available resources missed the fact that in practice, the decision to exploit them
and the extent to which the strategy of development was made dependent on
tSBecause of these contributions and his sojourn in 1957 at the Economic Commission for Latin
America Chenery appears as one of the originators of "structuralism'" in Little (1982) and Arndt
(1985). To this Chenery demurs, as he sees himself standing in the middle of the continuum rather
than at one end of it. See Chenery (1975, 1986a).
19The wider approach was first presented in an historical study of Japan [Chenery,Shishido, and
Watanabe (1962)] and later in cross-country models of industrialization [Chenery and Watanabe
(1965), Taylor (1969), Chenery and Syrquin (1980, 1986a)].
them were policy decisions. In addition, the impact of resources appeared not to
be additive to that of income growth, z° The proposed solution was to determine
alternative patterns for types of countries grouped according to characteristics
such as size and degree of reliance on natural resources. Chenery and Taylor
(1968) adopted a three-way classification: large, small-primary oriented, and
small-industry oriented. This scheme was then expanded into a typology of
development strategies in Chenery (1973) and Chenery and Syrquin (1975,
1986b).
The use of a typology in the analysis of long-run transformation was also
advocated by Ranis (1984). Unlike Chenery, who began with a large sample of
countries and then classified them, Ranis started with the historical experience of
a single country and then progressively built up the typology by adding ad-
ditional countries. The two approaches complement each other and some conver-
gence appears to have been achieved [see Ranis (1984)]. One last example of the
typology approach in long-run studies is the classification of agricultural paths of
transformation in Hayami and Ruttan (1985), discussed by Timmer in Chapter 8
of this Handbook. 21
Japan, as the one developed country not of European origin, has attracted a
great deal of attention. Ohkawa has been a central figure in various collaborative
enterprises to explain long-run growth and transformation in Japan, and the
lessons to be derived from this experience for other developing economies, zz
Other studies of Japan that influenced research on patterns of structural change
were Fei and Ranis (1964) on Japan as a surplus-labor economy and the general
equilibrium model of Kelley, Williamson and Cheetham (1972).
A perennial issue in development economics is the relevance of the past (of
today's developed economies) for the future (of LDCs). In research on long-run
transformation in the Kuznets tradition, this issue, while not absent, is not on
center stage. To assess the relevance of historical patterns, they have to be
studied first. Some very useful research on long-run structural change refers to
developed countries. Examples include Kuznets' work on modern economic
growth, the long-run country studies in the SSRC project, 23 the comparative
analyses of transformation in Europe by Svennilson (1954) and Kindleberger
(1967), and the recent work of Maddison (1982) on phases of capitalist develop-
ment.
2°Interrelations of income effects and resources could be accounted for by interaction terms, but
the simplicity of the original presentation would be lost.
2~Lance Taylor informs me that Lenin had a similar classification in the form of alternative
"roads" of agricultural development.
22Ohkawa and Rosovsky (1973); Ohkawa and Ranis (1985).
23The project was directed by Abramovitz and Kuznets. Publications have appeared on Japan by
Ohkawa and Rosovsky (1973), the United States by Abramovitz and David (1973), France by Carte,
Dubois and Malinvaud (1975) and, after a long gestation period, the United Kingdom by Matthews,
Feinstein and Odling-Smee (1982).
216 hi. Syrquin
2. Empirical research on the structural transformation
This section amplifies the discussion of structural change, focusing on the main
recurring debates on the concepts and measurement of the transformation.
2.1. Bases for comparative analysis
The empirical research program on growth and transformation originating with

Kuznets deals with long-run processes in a comparatioe framework. The long-run
aspect, which would appear to be self-evident in studying development, is much
de-emphasized in some recent research with a micro, neoclassical orientation.
Analysis of the comparative experience of nations varying in "size, location
and historical heritage" is essential for establishing "common features and
patterns" and for identifying "divergences from such patterns" [Kuznets (1959)].
The crucial question is why do we expect to find any uniform patterns? Kuznets
addressed this question in various publications. Probably his best and most
thorough answer appears in a little-known paper published in the proceedings of
a conference on "The Comparative Study of Economic Growth and Structure"
[Kuznets (1959)]. 24 The rationale for the comparative study of structure and
growth "is conditioned on the existence of common, transnational factors, and a
mechanism of interaction among nations that will produce some systematic order
in the way modern economic growth can be expected to spread around the
world" (p. 170).
The principal transnational factors, "those potentially common to the world,"
are three:
(1) The industrial system, that is, the system of production based on the
application of the technological potential afforded by modern science. Some of
the requirements of the system are some minimum level of literacy, a non-familial,
impersonal type of organization, and a high degree of urbanization.
(2) A community of human wants and aspirations. This is illustrated by the
relatively weak resistance to the spread of modern technology in reduction in
death rates, by the generality of Engel's law, and by the widespread desire for
higher standards of economic performance and levels of living.
(3) Organization of the world into nation-states.
The way the transnational factors affect the pattern of growth is conditioned
by national factors such as size, location, natural resources, and historical
heritage. "The consideration of the national elements thus leads directly to an
24In this section I rely heavily on this paper, including the phrases in quotation marks in the
previous paragraph. Chenery (1986a) derives from this study the core of the research program
originating with Kuznets.
emphasis on the distinctive structure and on the differences in growth patterns"
[Kuznets (1959, p. 166)]. Finally, there are the international factors relating to the
various channels of interdependence among the different nations. The "complex
of international relations can best be viewed, as a mechanism of transmission of
the unequal impact of economic growth" (p. 167).
The "crucial point" stressed by Kuznets is that "if there were no substantial
transnational factors, there would be no common features of significance in the
economic growth of nations and comparative study would be hardly warranted"
(p. 170).
The same idea appears in different form in Chenery (1960). Universal factors,
of which he lists five, 25 lead us to expect uniform patterns of development, while
particular factors and policy are behind divergences from a common path. In the
analysis of development the sources of diversity are no less important than those
leading to uniformity.
2.2. A unique path of development?
The presence of transnational (or universal) factors is the basis for expecting
uniformities in the growth process. But national (or particular) factors recognized
from the outset make clear the inevitability of differences at some level. The
comparative approach thus suggests uniformities at a broad (macro) level of
analysis or aggregation, but allows for variations at a lower (micro) level. This
notion resembles Gerschenkron's (1962) concept of substitutability (among "pre-
requisites") expanded to recognize that its nature depends not only on "relative
backwardness", but also on other national particular factors.
The possibility of lower-level substitutability can be illustrated by the concept
of a transition from an agricultural to an industrial economy. The transition may
be likened to "something like a historical likelihood or near-necessity", 26 but one
that allows for differences in many dimensions such as, industrial composition,
timing or sequencing of changes during the process, and sources of financing of
capital accumulation. For example, in resource-rich countries, the shift of re-
sources away from primary production lags behind that in other .countries at low
income levels. At higher income levels, however, the difference in the degree of
industrialization is much reduced. The availability of resources also affects the
composition of industrial production. Resource-rich countries tend to emphasize
more the capital-intensive branches of industry because of the relatively high
25"Among the universal factors are: (1) common technological knowledge; (2) similar human
wants; (3) access to the same markets for imports and exports; (4) the accumulationof capital as the
level of income increase [sic]; (5) the increase of skills, broadly defined, as income increases"
[Chenery (1960, p. 626)].
26Solow (1977, p. 493),
218 M. Syrquin
price of labor in comparison to their degree of industrialization. At the other
extreme, resource-poor countries have had to search for a substitute for primary
exports as a source of foreign exchange. Early development of manufactured
exports and a heavier reliance on foreign capital can thus be seen as largely a
response to the absence of natural resources.
A different illustration comes from the field of economic history. The English
model has sometimes been regarded as the one path for growth and, accordingly,
"countries have been described as backward because their coal and iron and
cotton production are relatively low". 27 O'Brien and Keyder (1978), however,
reject this view. In a study on comparative long-run economic growth in France
and Britain they argue that "[e]conomic theory lends no support to assumptions,
that there is one definable and optimal path to higher per capita incomes and still
less to the implicit notion that this path can be identified with British industrial-
ization as it proceeded from 1780 to 1914" (p. 18). Instead, they argue, "there is
more than one way of transition from an agricultural to an industrial economy
and from rural to urban society" (p. 196). The notion of a broad transition is not
questioned, but the view that the path of transition is unique is challenged.
A central thesis of the research program on transformation, therefore, goes
beyond the question of whether or not countries need to industrialize; and
focuses on the problem of when and in what manner it will take place. 28
The presence of the transnational factors that clearly identify the epoch of
modern economic growth, is the principal justification for expecting uniformities
across countries in long-run patterns of transformation. But nowhere in such
comparative analysis is it implied that there is a single unique path through which
all economies have to pass.
2.3. The methodology of comparative analysis
It is neither feasible nor desirable to try to capture all the complexity of the
transformation in one model. It is, therefore, pointless to search for one best
approach for comparative analysis. Often alternative approaches, which are
portrayed as competitive (if not mutually exclusive), are best seen as complemen-
tary. In this section two salient issues are elucidated: the level of analysis
(micro-macro), and the contrast between modeling and description. 29
270'Brien and Keyder (1978, p. 16).

28Some advocatesof an agriculture-basedstrategyof developmentsuch as Mellor (1986), aware of
the natural limits to that strategy, neverthelessrecognizethat "[e]conomicdevelopmentis a process
by which an economyis transformedfrom one that is dominantlyrural and agricultural to one that is
dominantly urban, industrial and servicein composition."(p. 67).
29Other issues related to the methodologyof multisector comparative analysis are examined in
Kubo, Robinson and Syrquin(1986).
2.3.1. The level of analysis
The appropriate level of analysis is closely related to the question of the unit of
analysis in the study of modern economic growth; is it the individual, an
industry, a region, or the nation? 3° In the process of development large numbers
of households, firms, and other agents interact within a framework determined
largely by the sovereign nation-state. Many long-term decisions that affect
economic growth are made for the nation by its agencies of the nation-state. Any
analysis dealing with policy aspects has to consider the decision-making units as
well as the universe affected by such decisions. 31
The nation-state can be distinguished from other regional aggregates by the
presence of various discontinuities relating to political authority, language, reli-
gion, and institutions. These discolatinuities are usually reinforced by trade
policies.
A further consideration is that modern economic growth is a manifestation of
the industrial system. Its inception requires significant social changes and is
bound to encounter opposition and resistance. In this context the state is the key
decision-maker.
Virtually all of the empirical work on structural change and much of that on
development in general, has been conducted at the macro level. Disaggregation,
essential for structural analysis, has usually proceeded to industries at a 2, 3 or
higher digit level, but not beyond that. As with macroeconomics in developed
countries, in the study of development, attempts have been made to provide a
firmer micro foundation for macro theory. This has taken two forms. In econ-
omy-wide analysis, microtheory has guided the choice of relations and variables
ranging from the centrality accorded to Engel effects to elaborate general
equilibrium models: in the neoclassical tradition 32 or emphasizing structuralist
features. 33
A second approach deals directly with the behavior of individuals. It is
represented in this Handbook by the chapters on human resource development
and labor markets in Part 3 and to a large extent by Gersovitz's chapter on
savings in this Part. These chapters attest to the progress made in understanding
important features of the functioning of economies in less developed countries.
However, for the analyses of economy-wide structural change, the micro ap-
proach faces a problem of aggregation. At a more aggregate level, results are
3°The issue of the nation as the unit of analysis is addressed in Kuznets (1951, 1966) and
Svennilson (1960). See also the discussion by Perkins and Syrqnin in Chapter 33, Volume II, of this
Handbook.
31A significant force at the national level nationalism has been interpreted as a public good
explaining the universal drive to industrialize [Johnson (1965)].
32Kelley and Williamson (1984), for example.
33Tayloret al. (1984), for example.
220 M. Syrquin
sensitive to the distribution of micro characteristics among individuals. Few
studies worry about the problem at all. When they do address the issue (as
Gersovitz does) they mostly reach inconclusive results and yet, the corresponding
macro relations are often quite robust. It appears that micro analysis has to be
supplemented by macro research. Aggregation is not necessarily bad; 34 some
effects can only be determined at a macro level. An example in the present
context is the effect of country size on the level and composition of foreign trade,
clearly an aggregate, cross-country effect.
2.3.2. Models vs. description
The distinction between a modelling approach and a purely descriptive one is

largely a bogus issue. Statistical-descriptive analyses of modern economic growth
are usually based on general models even if their structure is not always explicitly
presented. Thus, Kuznets (1959) emphasized the need for, at least, a skeleton
model that would give us some framework of the field; and Chenery (1960) began
with a general model before estimating reduced-form equations.
A variety of approaches is demanded by the complexity of the topic and the
paucity of information, in light of our inability to carry out controlled experi-
ments. The historical, statistical, and theoretical work on transformation can be
treated as a single research program, all involving some interaction between
generalizations from comparative studies and elaborations from theoretical anal-
ysis [Chenery (1986a)].
The statistical approach focuses on the search for uniform features of develop-
ment ("stylized facts") and on the main sources of growth and change. Its
sources of information are long historical series from developed countries, shorter
time series from developing countries, and cross-country comparisons. As is the
case with other partial approaches, this approach is subject to limitations.
Markets and prices are seldom studied directly; therefore, the links to policy are
indirect. Uniform patterns reveal associations, but cannot determine causality.
This is only partially remedied in the analysis of identity-based decompositions
of observed changes into their proximate sources, as illustrated in Section 6.
Price endogenous models address some of the deficiencies of statistical analy-
sis, usually focusing on a specific issue within a country. They rely on the
historical experience of an economy, but have to do without the variation
afforded by the comparative approach across countries. A computable general
equilibrium model is better suited to confront issues of causality and to probe
beyond "proximate" sources of growth by endogenizing effects regarded as
exogenous in the statistical approach. The extent to which additional relations
34Paraphrasing the title of Grunfeld and Griliches (1960), See also Chapter 8 by Timmer in this
Handbook.
can be incorporated into a model is basically limited by the available data. The
main contribution of applied general equilibrium models has been the ability to
go beyond partial effects to general equilibrium analysis. To date, however, such
models have contributed little to unravel the threads between structural change
and growth. Computable general equilibrium models are most useful for short- or
medium-run studies. For long-term analysis their strength- the ability to model
market behavior- becomes a weakness. Long-run transformation involves signifi-
cant changes in the nature and working of markets and institutions. In this type
of analysis a closer collaboration with the statistical approach has a potentially
high pay-off. General equilibrium models can borrow some stylized facts and in
return can act as a laboratory for assessing their robustness and their sensitivity
to variables downplayed in the statistical approach (prices for example). 35
2.4. Time-series vs. cross-section studies
Empirical research on the features of modern economic growth started with

comparisons of long-term series in developed countries, and then turned to
cross-country comparisons of less developed countries for a given year or for
short periods. The reason for limiting the time horizon was of course, the absence
of long-term records in LDCs. Cross-section relations, it was implied or hoped,
were good substitutes for relations within countries over time: "cross-section
analysis of national structures at a given point in time.., a d d . . , an insight into
current structural differences, viewed partly as points in the process of growth,
caught, as it were, at different stages and phases" [Kuznets (1959, p. 174)].
This view, held by many analysts at the time, now seems unrealistic in its
expectation of the congruence between cross-section and time-series patterns.
After all, in the same paper, Kuznets stressed the sources of diversity among
countries. Moreover, even if one huge dynamic model could perfectly capture the
experience of every single country, we would still not expect a perfect fit in a
simple relation between structure and level of per capita income. And indeed,
Kuznets grew increasingly skeptical of the ability to identify long-term changes
(past or future) in cross-section differentials. 36 His reasons were only in part
statistical (sample coverage, short-term disturbances, non-linearities). A more
fundamental reason was that cross-section analysis does not take into account
technological innovations and changes in consumer tastes. Additional reasons for
expecting differences between time-series and cross-section relations are the
35For a recent example of an eclectic approach that combines a variety of models to analyze
long-term features of industrialization with a common framework, see Chenery, Robinson and
Syrquin (1986).
36See, for example, Kuznets (1966, pp. 431-437) and (1971, pp. 182-198).
222 M. Syrquin
impact of varying policies, the omission of dynamic effects (besides those

mentioned by Kuznets), and changes in the international environment.
For similar reasons, it is also unrealistic to expect identical time-series relations
across countries.37At best we can expect a high degree of uniformity in the nature
of the relations (not in specific coefficients) reflecting the operation of common
universal factors; and discrepancies that can be interpreted. 38 Kuznets' skepti-
cism was a case of looking at the half-empty glass. Other sludies that compared
long-term series of production structure to cross-section results, concluded that
the degree of comparability between the two was high [Chenery and Taylor
(1968)], especially at the aggregate level. At a more disaggregated industrial
classification, results were mixed [Maizels (1963)].
As the time series of developing countries accumulated, the cross-country
framework was expanded to include a comparison of short time series for a large
number of countries. This allowed examination of the stability of cross-section
results, and the estimation of uniform time shifts.
Instead of searching for a unique pattern, attention was turned to the de-
termination of average patterns over time, and to an exploration of the relation
between time-series and cross-section patterns. 39 As was the case in studies of
saving and production, discrepancies between time-series and cross-section esti-
mates have proved to be fruitful sources of ideas.
An important argument for joint estimation of cross-section and time-series
estimates is that it may improve the efficiency of estimation, but this is not the
only consideration. Some effects are primarily, if not solely, cross-sectional, either
because some variables vary little over time, but differ widely among countries, or
because the effect in question depends, in part, on the relative magnitude of a
variable among countries. An example of the latter is the well-documented
inverse association of country size and trade shares, examined at length by
Perkins and Syrquin in Chapter 33 of this Handbook.
In contrasting cross-section and time-series results it is customary to interpret
the former as reflecting long-term adjustments, and the latter as short-term or
37This is precisely what Eckaus (1978) seems to require from such analysis. In his review of the
Chenery and Syrquin (1975) study on patterns of development he writes: "What has not been
demonstrated is that the hypothesis can be rejected that the coefficients estimated separately from
time series data for each country are different. Yet that is the crucial issue, and until it is settled it will
require some faith to believe that development patterns common to all developing countries have
been discovered" (p. 624). Now, it is true that "it cannot be concluded, with any conclusiveness that
there is an unique pattern of development for each of the dependent variables" (p. 624, emphasis
added) but, as emphasized repeatedly in the text, there is no reason to expect a unique pattern to
begin with.
3S,,The discrepancies are, therefore, to be studied- not removed. To put is paradoxically, the value
of the cross-section may lie not in its capacity to predict correctly the magnitude of changes over
time; but rather in its revelation of the discrepancy between its implication and the observed
historical change" [Kuznets (1971, p. 198)].
39See Chenery and Taylor (1968), Chenery and Syrquin (1975), and Syrquin and Chenery (1986).
partial adjustments to changes in exogenous variables. The cross-section ap-
proach was originally intended as a response to the limited data in developing
countries. Comparisons of economic structure across countries are now regarded
as useful in their own right. This is illustrated in recent studies of patterns of
development applied to nineteenth-century Europe [Adelman and Morris (1984),
Crafts (1984)].
3. Patterns of growth and accumulation
Empirical regularities accompanying development are commonly referred to as

"stylized facts". 4° The term can be applied to the list of features taken by
Kuznets to characterize modern economic growth. It is a more modest label than
the "historical laws" of the past (Engel, Hoffmann, Sombart, Wagner), but more
comprehensive in that they refer to system-wide phenomena. These "facts" are
empirical regularities observed "in a sufficient number of cases to call for an
explanation that would account for them.., independently of whether they fit
into the general framework of received theory or not" [Kaldor (1985, pp. 8-9)]
Various authors have published summary lists of the principal stylized facts of
growth and transformation. 41 Some characteristics appear in all lists (e.g. the
decline in the share of agriculture in output and employment), while others
appear to reflect the idiosyncratic tastes of the selector. Some of the uniform
features of development have held up quite well when confronted with new data
and more countries, and have therefore acquired the status of stylized fact. Other
apparent uniformities may not be so universally applicable and are still the
subject of debate in the literature. In the presentation below both types are
included, although I will try to point out those cases where the patterns are more
conjectural than established.
This and the next two sections present a selection of patterns of growth and
transformation derived from econometric studies, from multisectoral models
within countries, and from a combination of both in the form of an open
Leontief model of industrialization based on cross-country information, used to
simulate patterns of structural change over the transition.
3.1. Growth patterns
The most obvious characteristic of modern economic growth in developed

countries is the high rate of growth of total and per capita product; significantly
4°The term was apparently coined by Kaldor (1961) in his summary of observations about the
growth of industrial economies.
41Kuznets (1966, ch. 10; 1973), Perkins (1981), Taylor (1986).
224 M. Syrquin
higher than in previous periods. An acceleration of growth in middle-income

countries and in the industrial economies, was also observed during the post-
World War II period until the early 1970s. This acceleration was partly a
response to the disruption of world trade and production in the preceding period
and to the destruction of the war, but it also reflected the spread of modem
economic growth to transitional countries, better known as semi-industrial coun-
tries or newly industrialized countries.
Indirect corroboration of growth acceleration at medium income levels comes
from cross-country comparisons of growth over shorter periods of time. In every
decade since 1950 middle-income countries have grown faster than the groups of
countries with lower or higher income levels [Syrquin (1986a)]. The fast rate of
growth of total output in developing countries has followed the even faster rate
of manufacturing growth. One implication of growth acceleration in medium-
income countries has been a substantial geographical redistribution of world
manufacturing production. Between 1950 and 1973 developed countries saw their
combined share of manufacturing decline from 72 percent to 56 percent, while
the shares of transitional groups were rising by more than 50 percent [Chenery
(1977)].
A common view in the 1930s, not totally extinct at the present, regarded the
industrialization of primary-producing countries with some alarm. The argument
was that the spread of technological advance would narrow the differences in
relative costs and thereby eliminate the basis for international trade. Britain and
other advanced countries would lose their export markets. This view, which goes
back to Torrens, was represented in academic discussions by Keynes and
Robertson. 42 The facts were not kind to it. After World War II, world trade in
manufactures expanded enormously, largely in the form of intra-industry trade
among industrial countries, but also through the significant participation of many
semi-industrial economies. The proportion of foreign trade to output also rose
significantly during the century preceding World War I. This increase in the share
of foreign trade appears as one of the characteristics of modern economic growth
in Kuznets' summary [Kuznets (1966, p. 498)].
The process of economic growth can be formally described as the result of the
expansion in productive resources and the increase in the efficiency of their use.
During the transition (or in the epoch of modern economic growth) the growth of
inputs-labor and capital-also accelerates, 43 but the most important element
accounting for output growth in developed countries by far, has been the growth
of total factor productivity (TFP). However, it would be wrong to deduce from
42The same argument is sometimes known as Sombart's "law" of the declining importance of
international trade. For an evaluation and references see Syrquin (1978). The policy debates are
discussed in Maizels (1963).
43On the demographic transition, see World Bank (1984); on capital, see below the section on
accumulation.
this result that capital accumulation is not an important factor for development.
First, studies of productivity growth in developing countries, have shown that
factor inputs account for a much higher proportion of growth than in advanced
countries. This is due in part to the observation that the share of value added
imputed to labor is higher in rich countries than in poor [one 9f Taylor's (1986)
stylized facts]. Other reasons are the role of capital accumulation as a carrier of
technological change, and its status as a necessary factor for intersectoral
resource shifts. In addition to embodiment effects, a high rate of investment may
be required to sustain aggregate demand and prevent idle capacity from arising.
These observations point to a limitation of sources-of-growth analysis. The
sources considered are (usually) assumed to act independently from each other,
usually ignoring links and interactions among them. The missing link in this case,
is the relation between measured productivity growth and capital accumulation.
Evidence from micro studies (see Chapter 9 by Pack in this Handbook)
suggests another reason for the low measured growth of factor productivity in
various developing countries: the resources deployed are used inefficiently rela-
tive to both international best practice and the best domestic firms.
The very large contribution of productivity growth to output expansion in
developed countries is a relatively recent pheriomenon. In most of the countries
for which long-term records are available, factor productivity growth accelerated
over time to a larger extent than output growth, thereby raising its relative
contribution [see Syrquin (1986a) for references].
At the sectoral level most evidence indicates faster TFP growth in the in-
dustrial-modern sector than in agriculture. However, the high rate of productivity
growth has been pervasive, encompassing all major production sectors. As
Kuznets (1966, p. 491) pointed out in relation to the experience in developed
countries, even "[i]f the rise in output per unit of input in agriculture was lower
than that in industry, it was still so large compared with premodern levels that
one can speak of an agricultural as well as of an industrial revolution". Recent
studies have also identified strong "country" and "period" effects. Rates of labor
and total factor productivity growth tend to be uniformly higher across sectors in
countries with good average performance as well as within countries in periods of
rapid growth of aggregate productivity. This finding suggests' that the overall
economic environment, which includes general macroeconomic and trade poli-
cies, is an important factor in explaining differences in productiv.ity growth.
3.2. Accumulation
Accumulation refers to the use of resources to increase the productive capacity of

an economy. Indicators of accumulation include rates of saving; investment in
physical capital, in research and development, and in the development of human
226 M. Syrquin
resources (health, education); and investment in other public services which

augment productivity. This section focuses on aggregate saving and investment
patterns. Most of the long-run results reported below apply to both savings and
investment, even if their underlying determinants differ.
The importance accorded to rising rates of accumulation in development
economics was mentioned before. Even if "capital fundamentalism" no longer
enjoys wide currency, the critical role of saving and investment in income growth
is still recognized. 44 During the epoch of modern economic growth, and over the
transition range, there is a significant rise in the share of saving and investment in
GDP.
Kuznets (1961, 1966) analyzed long-term trends in capital formation propor-
tions 45 in ten countries. In most countries he found a significant secular rise in
capital formation proportions. The two notable exceptions were the United
Kingdom and the United States. In view of the subsequent impact of this finding
for the United States on the study of consumption and saving, its exceptional
nature is worth emphasizing. 46 Moreover, even in this case, according to Kuznets,
a rising proportion would have been found if the records could be extended
further back.
In a study on patterns of development in nineteenth century Europe, Crafts
(1984) pooled time-series and cross-section data for 17 countries and found a
significant income effect for the investment ratio. A dummy variable for Britain
improved the fit, indicating Britain's low investment share after accounting for
income.
Cross-country studies for the post-World War II period reveal significant
income effects for saving and investment, both among countries and over time
within countries. Syrquin and Chenery (1986) report estimates of development
patterns for samples of over 100 countries during 1950-83. 47 The expected total
44On "capital fundamentalism" see, for example, Gillis et al. (1983).

45Kuznets dealt with four measures of capital formationproportions incorporating the distinctions
between gross and net, and between domestic and national. The main findings apply to all four
definitions. Gersovitz discusses other issues and problems of measurement.
46Gersovitz (in Chapter 10 of this Handbook) reviews studies on the aggregate saving proportion
of national income, and finds support for the proposition that aggregate saving is affected by
demographic variables and the rate of growth of per capita income, but not by the level of per capita
income. This proposition is based on studies by Left (on the aggregate domestic saving ratio) and by
Modigliani (on the private saving ratio). In the present context, the former is more relevant. Left
(1969) argued that the effect of income in cross-section reflects demographicvariables correlated with
income. The regression results confirmed the importance of the demographicmeasures, although the
coefficient of per capita income remained significant even in their presence. But even if the net effect
of income had disappeared, the question about the existence of a long-run relation between the saving
ratio and income would still remain.
47The study excludes centrally planned economiesexcept Hungary and Yugoslavia. It updates and
extends the earlier work of Chenery and Syrquin (1975).
change over the transition as derived from the pooled regression, is about 8
percentage points of G D P for the investment share and about 11 percentage
points for the saving proportion. This difference between the changes in saving
and investment, reflects the tendency for the inflow of foreign capital (measured
by the current account deficit) to decline over the transition. A similar secular
decline took place in advanced countries [Kuznets (1961)].
The accumulation of data in developing countries now makes it feasible to
estimate time-series relations in a large number of countries. Average time-series
effects can be obtained in a covariance framework by adding a set of country
d u m m y variables. This assigns to each country its own intercept and yields slopes
which are weighted averages of the within-country slopes [see Chenery and
Syrquin (1975)]. The average income effects on investment and saving in the time
series are positive and significantly larger than the implied effects across coun-
tries. That is, whatever the initial shares were, they tended to go up since 1950.
The higher short-run propensities are in accord with predictions from theories of
saving behavior and are similar to Houthakker's (1965) results for personal
saving.
Are these average results representative of the experience in most countries?
Simple regressions of the investment share on the log of per capita income were
run for 106 countries. The distribution of the income coefficients (ignoring their
statistical significance) is as follows: negative slopes were obtained in 20 cases, 7
of them in industrial countries; in 33 countries the slope came out positive but
smaller than 0.10, and in the remainder 54 countries the coefficient exceeded
+ 0.10.49 A significant tendency was also found for the slope to decline with per
capita income. This suggests a logistic-type pattern with a constant ratio at high
income levels, but certainly not in the industrializing stage. Focusing on an
alleged constant saving share seems a misplaced goal for developing countries.
W h a t we need is better understanding of the underlying sources of an increasing
share for which per capita income may be only a proxy.
The previous discussion referred to both saving and investment. They were
treated together since they are highly correlated among countries and over time.
Feldstein and Horioka (1980) argue that such a correlation is critical in examin-
48In Chenery and Syrquin (1975) the transition was represented by the income interval $100 to
$1000 in 1964 US dollars, based on the observation that about 75 - 80 percent of the transformation
takes place within this range, In Syrquin and Chenery (1986) the interval in 1980 US dollars is $300
to $4000. The revised figures account for inflation since 1964, and reflect the observation that
exchange rates in developingcountries have tended to depreciate relative to the average for industrial
economies [see Syrquin (1985) and Wood (1986)]. The figures and tables below refer to a slightly
longer interval: $300 to $4500.
This generalized rise in saving and investment proportions refutes Left's contention (1969, p.
886) that "savings rates have generally not shown an upward trend" according to the time-seriesdata
in most countries.
228 M. Syrquin
ing the implications of the emergence of a world capital market in the last
decades. If national savings go to a common pool from which they then flow to
the various national centers according to differentials in return, as a perfect
market would require, we should find a low correlation between domestic
investment and saving. Even among industrial countries a high correlation was
found.
So far the evidence reported referred mostly to current price shares. Changes in
the relative price of investment goods are examined in Section 7.1.4 below. The
main conclusion there is that the secular rise in the investment share is a real
phenomenon and not due to a price effect.
The increase in overall accumulation rates 5° at a faster pace than population or
employment, results in changes in factor proportions and in comparative ad-
vantage with implications for the sectoral allocation of economic activity.
4. Changes in sector proportions
Changes in the sectoral composition of production are the most prominent

feature of structural transformation. Associated with income growth are shifts in
demand, trade, and factor use. These interact with the pattern of productivity
growth, the availability of natural resources, and government policies, to de-
termine the pace and nature of industrialization. The development of the
i n p u t - o u t p u t approach stimulated analysis of individual aspects of transforma-
tion in a multisectoral, economy-wide framework. At the same time as Kuznets'
work on modern economic growth, but using more formal statistical techniques, a
series of cross-country comparative studies were published, mostly by researchers
in or around Stanford University. Houthakker (1957) examined the universality
of Engel's law, Chenery (1960) established uniform patterns of trade and in-
dustrialization, and together with Watanabe [Chenery and Watanabe (1958)]
searched for similarities across countries in interindustry relations. These effects
were integrated in multisectoral models following the spread of input-output
analysis to developing countries. Multisectoral country models were used to
analyze the effects of government policy, but usually covered only a small part of
the transition range.
As a supplement to studies of individual countries, models were estimated
from cross-country data to simulate the full range of the transformation, which
typically involves at least a tenfold increase in per capita income. One such
cross-country prototype model initially applied to Japan [Chenery, Shishido, and
Watanabe (1962)], has proved useful for an integrated presentation of patterns of
5°On the rising investmentin human resourcessee the chapters in Part 3 of this Handbook. Studies
with a longer-run focus include World Bank (1980).
structural change [Chenery and Syrquin (1980, 1986a)]. The cross-country model
consists of two parts. The first is an open Leontief model in which growth rates
of output and population are taken as given. Its solutions are estimates of
structural change for a given income range. In the second part, growth of inputs,
outputs, and productivity are considered in a multisectoral framework. Although
weak on causality and market behavior, the model is a useful complement to
direct estimates of patterns of structural change, whether econometric or Kuznets'
type comparisons. If the latter are seen as reduced forms, the model presents a set
of relations that might underlie the reduced forms. Its main virtue is that it
incorporates into one system the various elements of growth and structure, from
Engel effects to differential productivity growth. It illustrates in a simple and
direct way the general equilibrium nature of the interrelations: changes in any
one component imply variations throughout the system.
The stylized facts on sector proportions presented below are ,based on the
cross-country model, on long-term comparisons for industrial countries of Kuznets
and others, and on econometric estimates for a large number of countries since
1950, all within a common accounting framework.
In the comparisons of long-term historical trends of production structure to
more recent information, there is an important data limitation (besides those in
Section 2.4) that has to be emphasized, namely, the treatment of quality changes
and of new products in particular.
The problem is not just one of aggregation. In addressing the issue, Kuznets
found it "frustrating that the available sectoral classifications fail to separate new
industries from old, and distinguish those affected by technological innovations"
[Kuznets (1971, p. 315)]. This limitation constrains our analysis particularly if,
following Kuznets, we identify science-based technological change as the prime
mover of modern economic growth. An implication is that "both the true rate of
shift in production structure and its connection with the high rate of aggregate
growth are grossly underestimated" [Kuznets (1971, p. 315)].51
New products do not just substitute for old ones, but they tend to increase the
variety of similar goods commonly grouped under the same classification. The
income elasticity for variety is larger than one [Jackson (1984)]. The increase in
variety of a generic good has various implications. To mention two: (1) The
increase in intra-industry trade in manufacturing among industrial countries can
be largely explained by product differentiation coupled with economies of scale
[Helpman and Krugman (1985)]. (2) Behrman and Deolalikar, in Chapter 14 of
this Handbook, point to the heterogeneity in nutrient quality among foods
defined as "flee". Since consumers' demand for food variety increases with
51New commodities are central to the Schumpeterian analysis. They were also emphasized by
Young (1928). On the input side, they often appear as embodied technical change.
230 M. Syrquin
income, estimated income elasticities derived from aggregate food demand equa-
tions overestimate the critical income elasticity for calories (nutrients).
4.1. The accounting framework
The accounting framework specifies a sectoral breakdown of the national income

and product accounts in a simple interindustry system. The accounting unit is the
productive sector defined in aggregate terms (primary, industry, services) or in
the finer distinctions of standard international classifications of production and
trade. 52 The elements of sectoral transformation are linked by the following
accounting identities. First, total gross domestic product by use:
Y = ( C + I + G) + ( E - M ) = D + T, (4.1)
where Y is gross domestic product, C is private consumption, G is government

consumption, I is gross investment, E is exports, M is imports, 53 D is domestic
final demand, and T is net trade.
At the sectoral level we start with the material balance equation of the
input-output accounts:
Xi = W~+ D, + T~, (4.2)
where Xi is gross output of sector i, and W~ is intermediate demand for the

output of sector i, (D and T are defined above).
Looking at a sector as a producing unit,
Xj = Uj + b' (4.3)
Vj = vj Xj, (4.4)
where Uj is intermediate purchases by sector j, Vj is value added in sector j, and

vj is the value-added ratio in sector j.
Adding up GDP by source,
V = Y'~ Vj= Y. (4.5)
Patterns of industrialization are chiefly concerned with changes in the distribu-

tion of the sectoral V/s. It is clear from eqs. (4.2) and (4.3) that industrialization
52Sectors may be distinguishedaccordingto variouscriteria. The issue has recentlybeen addressed

in Taylor (1986) and Syrquin(1986b).
S3All imports are treated as competitive. For alternative approaches see Kubo, Robinson and
Syrquin (1986).
Ch. 7: Patterns t~f Structural Change 231
has to be analyzed in conjunction with changes in the structures of demand (final
and intermediate) and trade. Since the stylized facts of the allocational aspect of
the transformation have appeared in various places in the literature, s4 in this
section only a brief account of the principal features will be presented.
4.2. Final demand
Among the most uniform changes in demand affecting industrialization, are the
decline in the share of food in consumption and the rise in the share of resources
allocated to investment. Thirty years ago Houthakker (1957) highlighted the
universality of Engers law in commemorating the centenary of the law, which
additional cross-section and time-series studies have continued to confirm. At low
income levels, food consumption accounts for as much as 40 percent of GDP and
total private consumption for about 75 percent. Over the whole transition both
shares decline; food consumption by more than 20 percentage points (of GDP)
and total consumption by somewhat less. The rise in the shares of non-food
consumption and investment imply a shift in demand away from agricultural
goods and to industrial commodities and nontradables.
4.3. Intermediate demand
The largest element in the material balance equations (4.2), is the use of
intermediate products, which in the aggregate accounts for over 40 percent of
total gross output in most countries. Analysis of the evolution of interindustry
relations has not received as much attention as other aspects of the transforma-
tion. This is surprising since, almost three decades ago, Chenery and Watanabe
(1958), demonstrated the feasibility and the value of a comparative analysis of
production structures. Chenery and Watanabe compared input-output tables of
Italy, Japan, Norway, and the United States for years around 1950 and searched
for similarities in patterns of interdependence among sectors. A recent study
[Deutsch and Syrquin (1986)] generalized the approach and applied it to a much
larger sample (83 tables from 30 countries). In addition to similarities in
interindustry relations, the study also found some systematic changes associated
with the level of development.
During the process of development, the total use of intermediates relative to
total gross output, tends to rise, while varying its composition. The relative use of
primary products as intermediates declines, while the uses of intermediates from
heavy industry and services go up. Most of the overall rise in intermediate use is
54Kuznets(I966), Cheneryand Syrquin(1975), Syrquinand Chenery(1986).

232 M. Syrquin
not due to changes in the composition of output but rather to increases in the
density of the i n p u t - o u t p u t matrices. These trends reflect the evolution to a more
complex system with a higher degree of fabrication, and the shift from handicrafts
to factory production. 55 The latter can also be observed in the change in the
distribution of firms by size. The increase in the use of intermediate services is
indicative of the dependence of industrial growth on a parallel expansion of
modern services. This relation provides an additional explanation to those based
on income elasticities, government expansion, and productivity growth, for the
rising shares of services in employment and output.
The preceding results referred to the use of a sector's output as an intermediate
input (a row measure). Looking at total intermediate purchases by a sector (a
column measure) a systematic trend has been observed in agriculture. The share
of intermediate inputs in the total value of output increases significantly with the
level of income. Technical change in the sector and a rising relative price of labor,
induce a more mechanized structure of production and a more intensive use of
inputs from outside the sector-fuels, fertilizers and capital goods. During the
course of the transformation the value-added ratio in agriculture (the counterpart
of the ratio of intermediates purchased to gross output) typically goes down from
close to 80 percent to less than 55 percent of the value of output [Chenery and
Syrquin (1986a). See also Chapter 8 by Timmer in this Handbook.].
4.4. Trade
In a relatively closed economy the structure of production has to conform closely

to the structure of demand, as stressed in the balanced-growth approach of the
1950s. In the open economy we also have to consider the level and composition
of international trade. The extent of a country's participation in the international
economy is only weakly related to the level of development across countries, in
spite of the fact that over time the association has clearly been positive except for
the period between the two wars. The principal determinant of the share of trade
in income across countries is the size of the economy. 56 This relation, among the
more robust of the empirical regularities in comparative analysis, cannot be
found in traditional trade models. Furthermore, it is probably inconsistent with
such models [Deardorff (1984)].
In small countries the shares of trade and capital inflows in G D P are relatively
high, domestic markets relatively small, and the production structure, therefore,
tends to be more specialized than in larger countries. The commodity composi-
55The trends also reflect substitution effectsdue to changes in relative prices. Unfortunately, there
is almost no information on price structures by sector among countries to be able to make some
general statements on substitution effects, let alone quantify them.
S6Size is usually represented by population. For other measures and a general discussion of the
effects of size, see Chapter 33, Volume II, by Perkins and Syrquin in this Handbook.
tion of trade and the type of specialization are largely determined by the
availability of natural resources, by traditional factor proportions, and by policy.
In practice, the evolution of comparative advantage and commercial policies have
combined to create an export pattern that reinforces the shift from primary goods
to industry, implicit in the pattern of domestic demand. The strength and timing
of the reorientation of exports have not been the same across countries; small
countries lacking a broad base of natural resources, had to develop manufactured
exports at an earlier stage than resource-rich countries, where specialization in
primary exports persists to a much later stage of development. Large countries
have shifted away from the specialization in primary products through import
substitution. These countries have been prone to adopt inward-oriented policies,
which appear more feasible for them than for small countries.
Because of the diversity in trade patterns, it has proven useful to subdivide
countries into more homogeneous categories, and to estimate separate patterns
for each group. A two-way classification (further discussed in Section 5.2 below)
based on population size and the relative specialization in primary or manufac-
tured exports, yields four types: large-primary oriented (LP), large-manufactured
oriented (LM), small-primary (SP), and small-manufacturing (SM). The diversity
of trade levels and specialization is illustrated in Figure 7.1, which shows the
average export patterns of the groups, including the predicted pattern for the
combined sample. Large countries (those with populations of more than 15
million in 1970) are shown as a single group since - with few exceptions - they are
considerably less specialized than small ones.
The large-country pattern has less than half the share of exports of the pooled
regression, and the shift from primary to manufactured exports takes place at a
lower income level. Among small countries, in the SM group manufactured
exports overtake primary exports quite early in the transition. The typical SP
economy, on the other hand, maintains a strong comparative advantage in
primary exports throughout the transformation.
Natural resources and size influence the timing of the shift from primary to
manufactured exports and the commodity composition of trade in manufactures.
The expectation of the shift itself is based on predictions from trade theories as
to the likely evolution of comparative advantage. The more rapid growth of all
types of capital, relative to natural resources and unskilled labor, facilitates the
development of manufactured exports and the replacement of manufactured
imports by domestic production. 5v The shift is also supported by Linder's (1961)
theory of representative demand, according to which comparative advantage is
acquired through production for domestic markets. An increase in the relative
importance of manufactures in total exports took place in the historical experi-
ence of the industrial countries [Maizels (1963)].
57Strictlyspeaking, the changein factorproportionshas to be comparedwith the one takingplace

in tradingpartners.
234 M. Syrquin
PRIMARY EXPORTS
20
123
LL
0
"1-
2~00 600 1200 2400 4500

PER CAPITAGNP (dollars)
MANUFA~D EXPORTS
20
c~
It.
0 10
Go
ILl
-1-
~o0 6oo 1200 24oo45oo

PER CAPITA GNP (dollars)
Figure 7.1. Trade patterns for three groups of countries. Key: L = large country pattern; SM = small,
manufacturing-oriented country pattern; SP = small, primary-oriented country pattern. Source:
Chenery and Syrquin (1986a).
Changes in the commodity composition of trade within manufacturing have

also resulted in some regularities over time and across countries although it is
i m p o r t a n t to stress the differences in the international environment. Today,
industrialization takes place in the presence of countries already industrialized.
T h e reliance on industrial imports in today's developing countries cannot help
b u t differ f r o m that in Britain and other industrial countries in the past. Still,
some b r o a d generalizations are possible.
L o w i n c o m e countries depend heavily on industrial imports. The c o m m o n
experience has been an early substitution of imports in light industry s8 in most
countries. Large countries have then proceeded to institute import substitution in
heavy industry and machinery to a greater extent than small countries.
S8The classification of industries into light and heavy is partly arbitrary and follows conventional
usage. The division in this Chapter is as in Chenery, Robinson and Syrquin (1986), and is based on
the 1958 ISIC classification. Light industry includes food (20-22) and consumer goods (23-29, 39).
Heavy industry includes producer goods (30-35) and machinery (36-38).
When a country begins to export manufactures these usually come from light
industry, except for simple processed products based on natural resources
(metals). At a later stage (often much later) exports of heavy industry become
feasible, and then tend to rapidly increase their share in industrial exports. Japan
is the best example of this pattern. In resource-poor countries (the SM pattern)
light industry exports become important at an early stage. In resource-rich
economies (the SP pattern) the need to develop manufactured exports is less
apparent. When the shift to manufactured exports takes place at higher income
levels, wages are relatively high and often preclude the fast growth of light
industry exports. The preceding discussion was based on the premise that light
and heavy industries differ in their factor intensities; there is some evidence that
this is the case. Heavy industries on the whole tend to be more capital and skill
intensive, enjoy faster productivity growth, and are more prone to exhibit
increasing returns to scale [ECE (1977), Balassa (1979)].
The changing pattern of comparative advantage in the process of economic
development was the subject of a study by Balassa (1979). Comparative ad-
vantage was defined in terms of relative export performance. By analyzing export
performance in 30 countries; the factor intensity of 184 product categories, and
such country characteristics as physical and human capital endowments, Balassa
(1979, p. 141) showed that
inter-country differences in the structure of exports are in large part explained

by differences in physical and human capital endowments. The results lend
support to the 'stages' approach to comparative advantage, according to which
the structure of exports changes with the accumulation of physical and human
capital. [See also Learner (1984).]
Among industrial countries, trade in manufactures increased very rapidly in
the postwar period. Simple factor-proportions theories cannot account for the
high and increasing share of intra-industry trade during the period. Newer
approaches emphasizing economies of scale and product differentiation have
been developed for this task [Helpman and Krugman (1985)]. These approaches
should prove useful for developing countries but, as shown by Balassa's study
mentioned above, conventional explanations still go a long way in accounting for
the evolution of comparative advantage in development. [Trade patterns in
developing countries during the 1970s are e:,.amined in McCarthy, Taylor and
Talati (1987).]
4.5. Structure of production: Broad sectors
The change in the commodity composition of trade reinforces the changes in final
and intermediate demands to produce a more pronounced shift in production
236 M. Syrquin
from primary activities to manufacturing and services. This shift is the centerpiece
of the transformation and has been validated in the long-term experience in the
industrial countries, and in virtually all countries in the postwar period.
Before proceeding to a more detailed analysis, I shall give an overview of the
transformation in the production of commodities and how the various elements
in eqs. (4.2) and (4.4) fit together. To focus on structural change, eqs. (4.2) and
(4.4) are combined and each dement is expressed as a share of G D P ( = V):
V J V = v,(W~/V + Di/V + TJV). (4.6)
Changes in sectoral shares in value added can be related to changes in the

composition of demand (intermediate and final), changes in the composition of
trade, and changes in the value-added coefficient:
A(V///V) = ~ / [ A ( W / / V ) +A(Di//V) + A(T///V)] + (V~)Ai)i//o i (4.7)
(a bar over a variable means that its value is set at the mean of the initial and
terminal levels).
Both sides of eq. (4.7) are independently calculated for the whole transition
range ($300 to $4000 in 1980 US$). The various components are derived from
econometric estimates of cross-country patterns for the period 1950-83 [W and v
from Deutsch and Syrquin (1986); all others from Syrquin and Chenery (1986)].
Table 7.1 shows predicted values at the end points, for the variables most
relevant for computing eq. (4.7). In the computations I assume, as an approxima-
tion, that food consumption generates demands from the primary sector only,
and that manufacturing supplies one-half of non-food consumption and invest-
ment (the other half represents construction and other non-tradables). With this
assumption I can now present a concise summary of the dimensions of the
transformation and a first approximation of its main determinants.
Over the course of the transition there is a significant shift in value-added from
primary production to manufacturing and nontradables. The average patterns in
Table 7.1 show a close correspondence between the directly estimated shift (the
last row) and the one calculated by the right-hand side of eq. (4.7). Changes in
domestic demand (Engel effects) directly account for less than one half of the
change in structure, and changes in net trade for about 10 percent on the
average. 59 The contribution of intermediates has two components. First, there is
a significant increase in the demand for manufacturing products to be used as
intermediates, and a decline in the relative use of intermediate inputs from the
primary sectors. The second component refers to variations in the ratio of
value-added to gross output in a sector. In agriculture this ratio tends to decline
59The decline in food consumptionas a share of GDP equals 0.20. The assumption in the text is
that the primary gross output falls by the same amount. Multiplying this value by the mean
value-added ratio (0.71), yields0.14 which is less than half of the total implied change in the primary
share in GDP (0.30). The other results in the text are calculated in a similar way.
Table 7.1
Accounting for the transformation: A first approximation
Predicted shares Contribution to

of GDP at change in share of:
$300 $4000 Change Primary Manufacturing
A. Final demand
Food consumption 39 19 - 20 - 20
Non-food consumption 34 42 8
1/2
Investment 18 26 8
B. Intermediate demand
Primary 23 17 - 6 -6
Manufacturing 25 43 18 18
C. Trade
Primary: exports 14 11 - 3
imports 6 8 2
net trade 8 -3 - -5
Manuf.: exports 1 9 8
imports 12 15 3
net trade - IT -6 5
Changes in gross
output A( X i / V ) -31 31
D. Value-added ratios av~ Mean v~ kvffvi

Primary 0.81 0.61 0.20 0.71 - 0.28
Manufacturing 0.33 0.36 0.03 0.35 0.09
Output Mean V/,/V

Primary 44 16 --28 0.30
Manufacturing 12 24 12 0.18
Implied changes in shares = ~iA(X~/V) -22 11

+ (vJ v) Av#o -8 2
-30 13
Directly estimated change -28 12
Source: Syrquin and Chenery (1986).
with the rise in income, or equivalently, the use of purchased intermediate inputs
per unit of output tends to increase. As shown in the table, this factor accounts
for about one-fourth of the decline in the share of primary production in total
GDP. In an input-output model, the variation in intermediate uses can be
further attributed to changes in final demand, trade and input-output coeffi-
cients. Such a decomposition is presented below in Section 6.1.
For a more complete picture of long-term changes in the structures of
production and factor use, I now draw on the results from the cross-country
model in Chenery and Syrquin (1986a). Changes in the composition of value
added, labor, and capital that are generated by the model are shown in Figure
7.2. Although the same basic pattern can be detected in each, it is exaggerated in
238 M. Syrquin
SOCIAL OVERHEAD - -~ ]
PRIMARY . . . .
MANUFACTURING . . . .
SERVICES -- -
A. VALUE ADDED
t ~ I I"--1 '"1 ''~ ] I I
40
"~.~.~ ~ .... I - ~ - - ~ - ~ "
~0 ~ ~ ~ ~ ~ " ~ , ~
I ] I .L L I I 1 I I t l [ I J
8. EMPLOYMENT
"--..... I I I r I 7 i I I 1 i 1 I
60- ~ . ~ /
/
/
/
40
/
/
20 1 f /
; I ~ I I, ; ] I I._1. 1..... ] I ].
C. CAPITAL STOCK
40 I I I I
~ J
20
0 ..... I I I _ 1 ,i
500 600 1200 2400 4500 9(300
PER CAPITA GNP
Figure 7.2. Simulation of value-added, employment, and capital (shares). Source: Chenery and
Syrquin (1986a).
the case of employment and minimized in the case of capital. These differences
are due to variations in rates of productivity growth and in factor proportions
among sectors. The typical employment pattern reflects the lag in the movement
of workers out of agriculture and the correspondingly lower growth in labor
productivity in this sector during most of the transformation. The rise of
employment in industry is much smaller than the decline in agriculture, and
consequently most of the shift is from agriculture to services.
The pattern of capital use shows a much higher proportion in social overhead,
which is larger than primary production and manufacturing combined. Because
this difference in capital intensity persists at all income levels, the shift from
primary production to manufacturing appears less pronounced. A more detailed
breakdown would show a corresponding shift from infrastructure supporting
primary production to infrastructure supporting industry.
Figure 7.2 illustrates the shift from primary activities to manufacturing during
the transition. The figure also portrays the decline in the share of manufacturing
in output and factor use at higher income levels. Such a decline has taken place
in the last 20 years in virtually all industrial countries, and has become known as
de-industrialization.
The cross-country relations on which the factor use patterns are based, are
more erratic and less well documented than those for demand and output. Also,
the country information available for analyzing those relations is scarcer and less
reliable. The results, therefore, are mostly illustrative, and only broad trends and
orders of magnitude are emphasized. Similar caveats appear in Kuznets (1966)
particularly for the data on capital (reproducible wealth). The patterns in Figure
7.2 present a picture similar to the long-term patterns in advanced countries as
described by Kuznets (1966, 1971), and to the estimated patterns for nineteenth-
century Europe of Adelman and Morris (1984) and Crafts (1984).
4. 6. Post-war patterns
With the accumulation of data in developing countries it has now become

feasible to analyze individual time series for a large number of countries. To what
extent have the post-war experiences been similar to the long-term patterns? To
address this question, simple regressions for the structures of output and employ-
ment were estimated in about 100 countries with data for parts or whole of the
period 1950-83. The equation estimated was:
x = a + b In y, (4.8)
where x stands for a share in GDP or employment and y for per capita income.
Output shares were defined in current and in constant prices. The estimates of b
are measures of structural change with respect to income per capita, but not
necessarily with respect to time, save for the case of a constant growth rate of per
240 ~. Syr~.~in
Table 7.2
Time series relations averaged by income groups
Lower Upper Industrial

Low middle middle market
Variable and sector income income income economies
Value-added
Current price shares
No. of countries 28 30 21 18
Agriculture - 0.24 - 0.17 - 0.12 - 0.09
No. with b > 0 3 4 0 0
Manufacturing 0.06 0.05 0.03 - 0.05
No. with b < 0 6 5 7 13
Constant price shares
Agriculture - 0.19 - 0.14 - 0.11 - 0.05
No. with b > 0 2 4 0 1
Manufacturing 0.07 0.04 0.06 - 0.03
No. with b < 0 3 4 1 6
Employment
Agriculture - 0.10 - 0.20 - 0.22 - 0.18
No. with b > 0 6 2 0 0
Industry 0.05 0.07 0.08 0.01
No. with b < 0 5 4 2 9
Source: Syrquin (1986b).
capita income. The results, summarized in Table 7.2, are presented as unweighted
averages of the individual income slopes for groups of countries ranked by level
of development.6°
The most striking result is the almost universal inverse association of income
and the share of agriculture in income and employment. Of the 97 countries for
which adequate time series were available, the income coefficients for the share in
value-added at current prices come out positive in only seven cases. In three of
them (Liberia, Nicaragua, and Zambia), the estimated coefficient did not differ
significantly from zero. In another three (Niger, Senegal, and Somalia), per capita
income fell during the period; hence the positive coefficient signifies that the
share of agriculture diminished in spite of the decline in income. The seventh,
Burma, is the only true exception to this general phenomenon.
The average income slopes of the share of manufacturing at current prices is
positive in developing countries, but diminishes with the level of income. There
are many more exceptions in this case than was true with agriculture. In almost
one-third of the cases recorded, the estimated slope is negative. It is instructive to
identify the main cases with negative income elasticities. Among the very low
6°The estimates by country are given in Syrquin (1986b). This section is based on that paper.
income countries we find some with negative growth (Niger, Somalia). In

oil-exporting countries (Algeria, Congo, Egypt, Iraq, Iran, Libya, and Saudi
Arabia), the decline in industry is the result of the oil boom - Dutch disease. In a
third group there was a fall in the manufacturing share, but from extremely high
initial values (Hungary, Israel, Yugoslavia). Finally, as mentioned above, in
virtually every industrial country there was evidence of de-industrialization at
some point during the period. For the period as a whole, negative slopes were
estimated in 13 of the 18 countries defined as industrial and with the required
data.
The decline in the share of employment in agriculture follows the decline in
value-added but with a lag. Since initially the share of employment exceeds the
share in output, labor productivity in agriculture declines. In the upper-middle
income group, relative labor productivity in agriculture often improves. It is
interesting to note the large size of the income slope of agriculture employment in
industrial countries.
Comparing the results for value-added in current and constant prices we find
that, in almost every case the prices of agriculture and manufacturing relative to
the price of total GDP, declined in the period. At constant prices, the decline in
agriculture's share was smaller and the increase in manufacturing larger than was
true of the current price shares. Offsetting these changes in relative prices were
the relative increases in the prices of mining and non-tradables. A significant part
of the shift in industrial countries, from tradables to services, was, therefore, a
price effect. The results in Table 7.2 suggest that the association of growth with a
reallocation of economic activity away from agriculture is among the most robust
of the stylized facts of development.
The post-war decline in the share of agriculture in the labor force has been as
pronounced as the similar decline that took place in the historical experience of
today's advanced countries. But, it is sometimes argued, today's migrating
workers mostly go into services and not into manufacturing. The figures in Table
7.2 confirm that only between one-third and one-half of the decline in agricul-
ture's share was taken up by industry (which includes mining, construction, and
public utilities). Has this pattern really been that different from the experience of
the industrial countries? Thanks to the efforts of Maddison (1980), we have
information on the sectoral composition of employment in sixteen advanced
countries at two distant dates: 1870 and 1950. The averages of the income slopes
between the two points are -0.27 for agriculture and 0.09 for industry. These
values indicate a faster rate of transformation with respect to income (not
necessarily per unit of time, since the growth rate of income per capita was much
lower). The income slope for industry is slightly higher than the ones in Table 7.2
for middle income countries. As a proportion of the corresponding slope for
agriculture it does not differ at all. Almost 30 years ago, Kuznets' analysis of
long-term trends in the industrial distribution of the labor force led him to a
242 M. Syrquin
similar conclusion: "In fact, in most countries the substantial decline in the share
of the A sector is compensated by a substantial rise in the share of the S
sector- not by a rise in the share of the M sector" (1957, p. 32).
Cross-country patterns are not always a reliable guide to variation over time.
In the case of employment structure, however, the correspondence between the
two has, on the average, been quite high.
4. 7. Manufacturing: Disaggregated results
During the process of industrialization the composition of the manufacturing

sector changes considerably. At a more disaggregated level, country-specific
features and policies become more prominent in determining the pattern of
specialization. Nevertheless, a high degree of uniformity still remains among
countries and over long periods of time. Most disaggregated analyses have been
done at the two-digit level of the international standard industrial classification
(ISIC), and have usually included an attempt to group sectors into homogeneous
categories, differing either in the demand for their products, their technology, or
in their dynamism. Examples are Hoffmann's (1958) division into consumer and
producer goods, and Chenery and Taylor (1968) grouping into early, middle, and
late branches. This latter is done according to the stage at which they make their
main contribution to the rise of industry.61
Table 7.3 shows the predicted change in the structure of manufacturing,
derived from regression analysis of the pooled time series since 1950.62As income
rises, the composition of manufacturing shifts from light to heavy industry. The
early increase in light industry is generally the result of the expansion of domestic
demand, and the opportunities for import substitution which are exhausted at an
early stage. Heavy industry is composed of goods purchased by other sectors as
intermediates or capital goods, and durable consumer goods with high income
elasticities of demand.
Some indication on the changes in intermediate demands for the products of
the two aggregate groups can be found in Deutsch and Syrquin (1986). In Table
7.1 it was reported that the increase in the ratio of manufacturing intermediates
to GDP over the transition equals 18 percentage points. Of these, only 3 points
originate in light industry, while the other 15 points come from heavy industry. A
more detailed analysis of the sources of change appears below in Section 6.1.
In advanced countries, the long-term change in the structure of the industrial
sector was generally in the same direction as the one found in cross-country
61Disaggregatedindustrialpatterns also appear in Chenery(1960), Kuznets(1966, 1971), Maizels

(1963), and UNIDO (1979, 1983).
62Based on Syrquin and Chenery (1986), where the early-middle-late classification is also
presented. In this studywe reliedin part on an unpublishedstudyby Prakash and Robinson(1979).
Table 7.3
Structure of manufacturing (percent)
ISIC Income per capita (1980US $)

Sector code 300 1000 4000
Food, beverages, tobacco 31 35 25 17

Textiles, clothing 32 22 18 13
Wood and products 33 3 3 4
Paper and printing 34 3 4 7
Other 1 1 2
Light industry 64 51 43
Chemicals and rubber 35 20 19 17
Non-metallic minerals 36 4 6 5
Basic metals 37 4 7 8
Metal products, machinery 38 8 17 27
Heavy industry 36 49 57
Share of manufacturing in GDP 12 19 24
Source: Syrquin and Chenery (1986).
comparisons in recent times. The most significant difference refers to chemicals,

whose share in total manufacturing shows little association with income across
countries, whereas in long time series it was the fastest growing component
[Maizels (1963)]. The difference is largely due to the very heterogeneous make up
of the sector. The relatively large share of chemicals observed in today's develop-
ing countries consists largely of "simply processed goods, such as soap, candles,
matches and paint" [Maizels (1963, p. 53)].
5. Structure and growth
To this point the analysis of the transformation has been in static terms. The
simulations of the cross-country model, on which the patterns in Figure 7.2 are
based, can be seen as elements of a comparative-statics analysis. The next section
is based on a dynamic extension of the cross-country model, designed to simulate
typical changes in productivity growth and the effects of alternative patterns of
specialization.
The static model is solved for various income benchmarks going from $300 to
$4500 (in 1980 US$). In the dynamic version for each income interval (between
adjacent benchmarks), the implied aggregate rate of growth is derived from the
income-related investment rate and the incremental capital-output ratio. The
latter is aggregated from the income-related sectoral capital-output ratios, with
output weights obtained from the static solutions. Sectoral growth rates are then
244 M. Syrquin
> 8 I I l L I l
STAGE Z~ STAGE 'It' INDUSTRIAUZATION ISTAGE
PRIMARY I Ig~
PRODUCTION DEVELOPED
"1" .....-
k- 6
/:
0¢r
tD
0 4 I
i- I
Z
o I ~.vmEs .--..---<~""-'F-
_ _ . ~ . ~ M/~TI.~r,.~ ..~--T
I-"
z o ~ L 2[ j 3. L 4 js-~--Z-f-rt
o ° ° ~ ~ o o
o o
PER CAPITA G N P (dollors) "
Figure 7.3. Sectoral sources of growth. Source: Chenery and Syrquin (1986a).
derived by combining outputs at various income levels with the aggregate growth
rate [see Chenery and Syrquin (1986a), and Syrquin (1986a) for further details].
5.1. Sectoral contributions to growth
During the transformation, growth proceeds at an uneven rate from sector to

sector. The relation between aggregate and sectoral growth can be derived by
differentiating with respect to time the definition of total output, V = ZV~, and
expressing the result in growth terms:
gv = Ep,gv,, (5.1)
where g vi and g v are the growth rates of V/and V, respectively, and the weights
are sectoral output shares, Pi = V~/V. The results of computing eq. (5.1) for the
whole range of the transformation, are summarized in Figure 7.3. It shows the
contribution of each of the major sectors to aggregate growth. 63
The acceleration of growth until late in the transition is related to the rise in
the rate of investment and to the acceleration of factor productivity growth. As
63 The figure plots data for each period at the income level of the terminal years. This and the
following section are based on Chenery and Syrquin (1986a).
shown below, the latter reflects the reallocation of resources to sectors with
higher productivity.
Figure 7.3 can be regarded as the dynamic version of Figure 7.2(A). The
contribution of each sector to growth is measured by its average share of total
GDP, #i, weighted by its growth rate, gw. If all growth rates were identical in a
given period, the relative contribution of each sector to aggregate growth would
merely equal its share of GDP. The contribution of a rapidly growing sector, such
as manufacturing in the early periods, is greater than its static share, whereas that
of primary production is less.
In using these results for country comparisons, a further distinction should be
made between outputs that are tradable, i.e. primary products and manufactures,
and those that are essentially non-tradables such as social overhead and most
services. Since imports and exports increase the range of choice of resource
allocation, differences among countries are concentrated in the traclable sectors
and in the non-tradables related to them.
Figure 7.3 distinguishes three stages of transformation: (1) primary produc-
tion; (2) industrialization; and (3) the developed economy. Because there are no
significant discontinuities in the processes that lead from one to the next, the
dividing lines between them are somewhat arbitrary. 64 The extent to which
different trade patterns influence this sequence is considered after summarizing
the principal features of each stage. In the discussion I also refer to typical shifts
in the sources of growth presented more extensively in Section 6.1.
Stage 1: Primary production. The first stage of the transformation is identified
by the predominance of primary activities- principally agriculture- as the main
source of the increasing output of tradable goods. Even though primary produc-
tion typically grows more slowly than manufacturing, this difference is more than
offset at low income levels by the limited demand for manufactured goods. The
large weight of agriculture in value-added is also one of the main reasons for
slower overall growth during this stage.
On the supply side, Stage 1 is characterized by low to moderate rates of capital
accumulation, accelerating growth of the labor force, and very low growth in
total factor productivity. As shown below, it is the absence of productivity
growth more than low investment rates that causes the lower aggregate rates of
growth in Stage 1. Although comprehensive studies of productivity growth in
poor countries are few, this conclusion is consistent with the evidence cited in
Chenery (1986b).
Stage 2: Industrialization. The second stage of the transformation is char-
acterized by the shift of the center of gravity of the economy away from primary
production and toward manufacturing. The main indicator of this shift is the
64The criteria used to distinguish the semi-industrial economies (Stage II) are discussed in Chenery
and Syrquin (1986b).
246 M. S y r q u i n
5.0
~M J
& 2.5
;'-" sPJ
0 2.0
r~
1.5 "" / 7
Z
o PRIMARY
I"- CJ~NTRIBUTION
,.° '~gp) -
rn
I---
Z
\SF
0 0.5
0
$M
500-600 600-1200 1200-2400 2400-4.500
PER CAPITA GNP (dollers)
Figure 7.4. Sectoral contributions to growth: alternative trade patterns. Source: Chenery and Syrquin
(1986a).
relative importance of the contribution of manufacturing to growth, as shown in

Figure 7.3. In the standard pattern generated by the dynamic model, manufactur-
ing makes a larger contribution to growth than primary production above an
income level of about $1200. This shift occurs at lower or higher levels of income
depending on the resource endowments and trade policies of different countries,
as shown in Figure 7.4.
On the supply side, the contribution of capital accumulation remains high for
most of Stage 2 because the rise in the rate of investment tends to offset the
decline in the weight of capital in the sectoral production functions.
Stage 3: The developed economy. The transition from Stage 2 to Stage 3 can be
identified in several ways. On the demand side, income elasticities for manufac-
tured goods decline, and at some point their share in domestic demand starts to
fall. Although this tendency is offset for a while by the continued growth of
exports, it is ultimately reflected in a decline in the share of manufacturing in
GDP and in the labor force. This turning point has occurred in virtually all the
industrial market economies within the past twenty years.
On the supply side, the main difference between Stage 2 and Stage 3 is the
decline in the combined contribution of factor inputs as conventionally mea-
sured. The contribution of capital falls because of both its slower growth and its
declining weight. In addition, because of a slowdown in population growth, only
a few developed countries still have significant increases in their labor force.
Table 7.4
Sectoral shares and contributions to growth: 1960-80 (percent)
Income No. of Social

group countries Primary Manufacturing overhead Services GDP
A. Shares in value-added."
A verage for 1960- 80
Low y 13 45 13 11 31 100
Lower mid-y 24 32 14 14 40 100
Upper mid-y 16 23 22 15 40 100
Industrial 15 8 27 17 48 100
B. Annual growth rates 1960-80
Low y 2.5 4.7 5.3 5.3 3.8
Lower mid-v 3.4 6.6 6.5 5.1 5.0
Upper mid-y 4.4 7.9 8.2 6.6 6.4
Industrial 1.6 4.8 4.2 4.2 4.2
C. Sectoral contributions to growth
Low y 28 15 16 41 100
Lower mid-y 24 19 16 41 100
Upper mid-y 14 26 19 41 100
Industrial 3 30 17 50 100
Source: World Bank data.
In more developed countries, total factor productivity growth is less associated

with industrialization and more widely diffused throughout the economy than in
Stage 2. The most notable change is in agriculture, which has shifted from being
a sector of low productivity growth to being the sector of highest growth of labor
productivity in most of the developed countries. The underlying cause is the
continued movement of labor out of agriculture and the closing of the wage gap
between agriculture and other sectors, which has stimulated the substitution of
capital for labor as well as technological improvements.
Table 7.4 summarizes the experience during 1960-80 of a large number of
countries grouped by income level. In all the groups, growth in the primary sector
was lower than aggregate growth, and the opposite was true of manufacturing.
Consequently, the dynamic sectoral share (or sectoral contribution to growth) of
primary production was lower than the mean static share over the period, while
in manufacturing the incremental share exceeded the average one. Both the static
and the incremental shares show a clear association with income across groups
for the primary and manufacturing sectors. In industrial countries, services
account for half of total GDP growth. The current price share of services in those
countries has risen even more because of the rise in the relative price of services.
By the middle of the period covered the manufacturing share started to decline in
virtually all advanced countries.
The figures in Table 7.4 illustrate the acceleration of growth with the level of
income among developing countries and the subsequent slowdown in the
248 M. Syrquin
industrialized stage. This pattern encompasses all of the broad sectors dis-
tinguished in the table.
5.2. Typology of development patterns
Average patterns of development are a useful starting point. They provide an

initial reference point stressing the uniformities of the transformation. Various
other factors have a systematic effect on resource allocation, affecting the timing
and sequence of structural change more than its overall nature.
The identification of systematic sources of diversity leads naturally to attempts
to divide the sample into more homogeneous groups. In the literature on long-run
comparative patterns of development, one finds two main approaches to con-
structing a typology. The first one lets a purely statistical method, such as
clustering analysis, come up with the various types. 65 The second approach reties
more on theoretical arguments and on a priori judgements. There are two
variants of this latter approach, differing on whether they start from individual
cases representative of ideal types, or from the whole sample and then proceed to
subdivide it. Ranis (1984) refers to the first variant as the "comparative historical
approach" and to the second as the "econometric comparative patterns ap-
proach". 66 The latter variant is now illustrated.
Differences in international specialization were identified above as the primary
source of deviation from the average growth patterns described in the previous
section. They include the effects of factor proportions on comparative advantage
as well as policy decisions about the levels of trade and external capital inflows.
The alternative trade patterns in Figure 7.1 lead to significant differences in the
sectoral contributions to growth. These will now be presented assuming the same
aggregate growth rates for all trade patterns, followed by some estimates of
comparative performance during the period 1950-83.
The different trade patterns have a large impact on the timing of industrializa-
tion. These effects are brought out in Figure 7.4, which shows the growth
contributions of primary production and manufacturing under each trade pat-
tern. (Non-tradables are omitted since they show relatively tittle variation from
the standard pattern of Figure 7.3.) The timing of industrialization is measured
by the beginning of Stage 2, the income level at which the growth contribution of
manufacturing begins to exceed that of primary production.
With this measure, large countries typically reach the semi-industrial stage at
an income level of about $550, the standard pattern at about $850, and small
65See, for example, UNIDO (1979).

66The first is illustrated by Lewis (1954), Fei and Ranis (1964), and Kelley, Williamson and
Cheetham (1972). The second by Chenery and Taylor (1968), Chenery and Syrquin (1975, 1986b) and
by Adelman and Morris (1984).
Table 7.5
A n n u a l growth rate of G D P 1950-83: Simple averages (percent)
Trade No. of
Size orientation Openness countries g,; S.D
Primary In 10 4.94 1.35

Out 5 5.12 2.13
LP 15 5.00
Large anufacturing In 6 4.73 1.2t
t~Primary
LM
Out
In
Out
8
14
29
27
23
5.26
5.04
5.02
3.58
5.01
1.51
1.48
1.94
SP 50 4.24
Small ~Manufacturing In 17 4.74 1.85
Out 1_00 5.7_~3 2.43
SM 27 5.11
77 4.54
All primary 65 4.42
All manufacturing 41 5.09
All inward 60 4.28
All outward 46 5.22
ALL 106 4.67 1.90
Note." Growth rates within countries are OLS estimates. The number of annual
observations varies from 14 to 34.
Source: World Bank data.
primary exporters at $1300. In the L pattern, early industrialization results from

the widespread policy of import substitution in manufacturing. In the SP pattern,
manufactured imports are replaced much more slowly. The SM pattern is more
complex because it typically starts with a relatively high inflow of external
capital, which is later replaced as a source of foreign exchange by the rapid
growth of exports of light manufactures.
Table 7.5 presents a rough picture of comparative performance since 1950,
based on unweighted averages of growth rates (estimated by OLS regressions).
Three dimensions are considered in the typology.
Size: Economies are separated into small and large on the basis of their
population size in 1965. The dividing line is 15 million.
Openness: Inward and outward orientations are distinguished according to the
level of merchandise exports relative to the value predicted by regressions.
Trade orientation: To capture the combined effect of the abundance of re-
sources and trade policy an index of trade orientation was defined. It compares
the actual commodity composition of exports to the one predicted for a country
250 M. Syrquin
of similar income and size. Countries are classified into primary or manufactur-
ing oriented [see Chenery and Syrquin (1986b)].
The averages in Table 7.5 mask the variation within groups. Still, the differ-
ences are of interest, and quite suggestive. During the period 1950-83, large
countries seemed to have performed better than small °ones [see Perkins and
Syrquin, Chapter 33¢ Volume II, in this Handbook); a manufacturing specializa-
tion outperformed the primary specialization, and an outward orientation ex-
hibited faster growth than an inward orientation. The superiority of the outward
orientation is in evidence for all four types in the table (LP, LM, SP, and SM).
6. Accounting for the transformation
The transformation in the structure of production is part of the process of

economic development. To determine its basic sources, therefore, a general model
incorporating the interdependence between growth and structural change is
required. Some attempts at such general modelling are reviewed below, after
presenting results from simpler attempts to decompose growth and change into
their more proximate sources.
6.1. Growth accounting: Demand-side decomposition
The computation of sources of growth from the demand side starts from the
material balance equations (4.2), reproduced below in vector notation, with trade
separated into exports and imports:
X= W+ D+ E-M. (6.1)
Assuming a linear input-output technology (A) for intermediate demands and

letting u i represent the share of total domestic demand of sector i supplied from
domestic sources, we can substitute into (6.1) and solve for X:
X = R ( aD + E), (6.2)
where R = ( I - A) -1 and the "hat" over u indicates a diagonal matrix.

From this accounting identity we derive a growth accounting system relating
output growth to its sources on the demand side, 67 by differentiating (6.2) with
67The "demand"labelhas becomecustomaryfor this approach.Trade and comparativeadvantage

are more supplythan demanddriven.
Ch. 7."Patternsof StructuralChange 251
respect to time (~ = dx/dt):
2 = R~D + R E + R~( D + W) + R~t,4X. (6.3)

(a) (b) (c) (d)
The four factors measure the total (direct and induced) effects of:
(a) domestic demand expansion (DD),
(b) export expansion (EE),
(c) import substitution (IS), and
(d) changes in input-output coefficients (IO).
This decomposition elaborates the formulation in eq. (4.7) by assigning the
change in intermediates to the effects of demand, trade, and technology
throughout the system. Eq. (4.7) computed changes in sector proportions, whereas
eq. (6.3) refers to absolute changes in output. The latter can easily be adapted to
consider changes in structure by comparing its results to a balanced growth case
where all the right-hand elements in the equation expand at the same rate with
unchanged import and input-output coefficients. This is an identity-based de-
composition that provides logically consistent definitions of concepts such as
import substitution and balanced growth.
I now present results based on eq. (6.3), for the transition range in the
cross-country model and for a group of semi-industrial countries in the post-war
period.
6.1.1. The shift from primary to manufacturing
Table 7.6 summarizes information on changes in shares in gross output, and the
sources of those changes. 68 The results from the cross-country model suggest that
the fall in the primary share is mostly due to demand (Engel effects) at low
income levels, and to trade effects afterwards. Trade effects combine a lower than
proportional expansion in primary exports, and, at higher income levels, import
liberalization (defined as an increase in the sectorat import coefficient). Changes
in input-output coefficients contribute to the decline in the primary share at all
income levels. This effect captures the substitution of fabricated materials for
natural resources induced by technology and relative prices.
The rise in manufacturing share, which best represents the process of industri-
alization, is due less to high income elasticities and more to trade and technology.
Import substitution is quite significant at all income levels. A more disaggregated
analysis would show early import substitution in consumer goods, shifting to
68The results in Table 7.6 refer to deviationsfroma balanced-growthpath focusing,therefore,on

changes in shares. In Table 7.7 the resultspertain to absolutechangesand referonly to manufactur-
ing.
252 M. Syrquin
Table 7.6
Deviations from balanced growth: Gross output (percent)
A. Decline in share of B. Increase in share

Primary of Manufacturing
Change Change
in Source in Source
share Demand Trade I-O share Demand Trade I-O
Simulated results
Income interval
(1980 US $)
I. $300 $600 - 7 68 27 5 5 23 50 27
II. $600 $1200 -6 49 43 8 6 15 55 30
III. $1200 $2400 - 5 30 59 11 6 10 62 28
IV. $2400 $4500 -4 13 76 11 6 7 63 30
Country results a
Korea 1955-73 - 22.3 68 4 28 27.6 17 82 1
Taiwan 1956-71 - 12.2 55 16 29 23.5 2 84 14
Japan 1914-35 - 10.4 92 19 - 11 9.3 58 55 - 13
Turkey 1953-73 - 18.5 64 10 26 12.5 27 40 33
Mexico 1950-75 - 7.6 40 48 12 9.4 35 40 25
Yugoslavia 1962-72 - 12.7 58 12 31 12.6 64 4 32
Japan 1955-70 - 11.7 42 27 31 12.2 40 13 47
Israel 1958-72 - 3.9 123 - 36 13 11.0 45 11 44
Norway 1953-69 - 3.4 66 43 - 9 1.3 - 21 44 77
aCountries ranked by per capita income in the initial year.

Sources: Simulated results: Chenery and Syrquin (1980). Country results: Kubo et al. (1986).
producer and capital goods at higher levels of development. The increase in the
overall density of the input-output matrix that accompanies development is
especially important in heavy industry [Deutsch and Syrquin (1986)]. It shows
here as large contributions of the IO term to the rise in the manufacturing share.
The country results show more variation than the smooth simulated patterns,
but in general they conform to the generalizations based on the latter. A
significant difference is the faster pace of industrialization in semi-industrial
countries than in the typical case simulated by the cross-country model. This
stands out particularly in the case of the East Asian super-exporters, Korea and
Taiwan.
6.1.2. Manufacturing: A sequence of phases?
So far the contributions of export expansion and import substitution have been
presented together as a general trade effect. They are now considered separately
to assess their relative importance to the growth of manufacturing. Before
discussing the results in Table 7.7, two conceptual points are worth noting. First,
Table 7.7
Trade sources of growth in manufacturing output (percent)
Contributions to output growth by type

Large Small-manufacturing Small-primary
Income (L) (SM) (SP)
interval a EE IS EE IS EE IS
I 10 10 30 2 8 14
II 14 6 32 10 10 12
III 17 3 34 12 12 10
IV 21 2 35 12 13 8
EE IS
Korea 1955-63 1~
1963-73 48 - 2
Taiwan 1956-61 27 26
1961-71 57 4
Japan 1914-35 34 5
1955-70 18 - 2
Turkey 1953-63 2 9
1963-73 8 3
Mexico 1950-60 3 11
1970-75 8 3
Israel 1958-65 27 12
1965±72 50 - 37
Norway 1953-69 50 - 18
aIntervals as in Table 7.6.

EE: contribution of export expansion.
IS: contribution of import substitution.
Sources: Background information for Chenery, Robinson and Syrquin (1986).
import substitution is defined as arising from changes in the ratio of imports to

total demand for each sector. It is an approach with a long tradition in
development, and it fits well with the view that imports are imperfect substitutes
for domestic goods, so that the source of supply is an integral part of the
economic structure. 69 In the present context the measure incorporates economy-
wide effects through the interindustry relations. Different domestic supply ratios
(u) are assumed for final and for intermediate demands, and for the latter, when
the data are available, the input-output matrix is separated into domestic and
imported components. In eq. (6.3) for simplicity, for each sector a unique u
coefficient is presented. [The complete formulation appears in Syrquin (1976),
where alternative approaches to the measurement of import substitution are
discussed.]
69See de Melo and Robinson (1985) for a discussion of the implications of this treatment.
254 M. Syrquin
Second, this section deals with the sources of output growth rather than
deviations from a balanced growth path as in Table 7.6. In this case there is an
important asymmetry in the way we measure the effects of exports and of import
substitution. Exports, in eq. (6.3), enter as a flow and their potential contribution
is, in principle, unbounded; this is not the case for imports, which appear as
ratios to final or intermediate demand. Typical u coefficients in semi-industrial
economies in the 1950s were around 90 percent for light industry and 50 percent
for heavy industry. The scope for further import substitution was ample in heavy
industry but not in light industry.
Table 7.7 shows the relative contributions of export expansion (EE) and import
substitution (IS) to the growth of manufacturing gross output. The long-term
simulated patterns are shown by type. As expected, in large countries exports are
less important especially at an early stage of industrialization. At such a stage
import substitution is quite significant in large and SP economies. Most of the
countries that vigorously pursued an import substitution strategy in the post-war
period, were relatively large (Brazil, India) a n d / o r primary oriented (Chile,
Uruguay).
Country results are shown for two periods (except for Norway). Trade contri-
butions appear to have followed a distinct sequence: periods of significant export
expansion preceded by periods of strong import substitution. This sequence
appears most clearly in Korea and Taiwan where it can be related to the changes
in trade strategy in the early 1960s. Similar results were also found at a more
disaggregated level [Kubo, de Melo and Robinson (1986)]. The results suggest
than an economy may have to develop an industrial base and acquire a certain
technological mastery before it can pursue manufactured exports on a significant
scale. 7° The crucial question is then not one of export promotion versus import
substitution, but rather one of designing the latter to avoid inefficient production
and delays in shifting out of it.
6.2. Growth accounting: Supply side
The previous section presented sources of growth from the demand side. As with
other growth-accounting exercises, it starts from an accounting identity and
decomposes growth or structural change into its proximate sources without
necessarily implying causality. An alternative growth-accounting approach, this
time from the supply side, is the Abramovitz-Solow-Denison decomposition of
70For other analysesof sequencingand its relation to infant industryarguments,see Balassa(1979)

and Westphal (1982). Along similar lines, it has been argued that in the large countries of Latin
America import substitutionprovided "a preamble to the export stage" [Teitel and Thoumi (1986)].
See also Chapter 31, VolumeII, by Bruton in this Handbook.
Table 7.8
Changes in sector proportions in the cross-country model:
Supply-side accounting (percent)
Differences between manufacturing and

agriculture in the growth of:
Income Value Factor Factor
interval a added input productivity
I 1.8 1.0 0.8

II 2.9 1.5 1.4
III 4.1 2.9 1.2
IV 5.1 4.1 1.0
alntervals as in Table 7.6.

Source: Syrquin (1986a).
the sources of growth into the effect of factor accumulation and productivity
growth.
Changes in sector proportions, which clearly imply differential rates of sectoral
growth, can be related in this approach to differential expansion of inputs and of
total factor productivity. Some orders of magnitude for agriculture and manufac-
turing, based on the long-run model of industrialization, are presented in Table
7.8. The results indicate the importance of differential productivity growth in
accounting for the shift in activity from agriculture to manufacturing. Only in the
higher income interval, when labor is declining absolutely in agriculture and
productivity has risen significantly, is the differential in factor input growth the
dominant source of change.
6. 3. Resource shifts and productivity growth
The supply-side analysis of sources of growth focuses on the growth of factor

productivity. At the aggregate level, productivity growth cannot be analyzed
independently of demand aspects. Formally, the measured rate of aggregate
productivity growth (~) equals a weighted average of the sectoral rates (?~i) with
output weights (Oi), plus a factor measuring the effect of intersectoral resource
shifts (RE = reallocation effect):
= E # i X i + RE. (6.4)
The reallocation effect (RE), when positive, shows the increase in efficiency
that results when resources (labor and capital) move from sectors with lower to
sectors with higher marginal productivity, reducing the extent of disequilibrium.
256 M. Syrquin
I I I f F I I
2-.0
"~'~" MANUFACTURING SECTOR

1.5
SERVICES SECTOR"O~
x
w
121
z 1.0
0.5
~MARY SECTOR
0 I [ I I I I I
300 600 1200 24430 450(3 72(30 10800
PER CAPITA GNP (dollars)
Figure 7.5. Relative labor productivity. Note: Index signifies labor productivity in a sector relative to
labor productivity for the whole economy. Source: Syrquin (1986a).
The effect vanishes when resources are optimally allocated before and after the
shift.
6. 3.1. Relative labor productivity
A partial indicator of differential returns across sectors is the pattern of relative

labor productivity, obtained by dividing a sector's share in value-added by its
share in employment. It is partial because it refers to average and not marginal
products and considers only one input (labor). vt The pattern of relative produc-
tivity in the model simulations is illustrated in Figure 7.5.
If all sectors had the same production function and faced the same factor
prices, and if resources were prefectly mobile, then labor productivity would
follow the same pattern in all sectors. In the model simulations in Figure 7.5,
labor productivity is significantly lower in agriculture than in the rest of the
economy. 72 In the early stages of development the growth of productivity in
agriculture lags behind that of other sectors, which further widens the productiv-
ity gap. These sectoral differences in the average product of labor reflect
differences in the nature of the procluction function (which lead to different
factor proportions) and in the rate of technological change. But they also stem
71See Syrquin (1984) for a comparison of approaches to measure the contribution of intersectoral
resource shifts.
72Syrquin (1986a) presents data for various countries and compares them to the model simulations.
In our earlier study [Chenery and Syrquin (1975)] we derived the patterns of relative productivity
from regressions. The results for the primary sector were very similar, for manufacturing less so.
from the low mobility of resources, a condition that lies behind the persistence of
disequilibrium phenomena such as surplus labor in agriculture and other low
productivity activities, including handicrafts and services.
When the industrial sector accelerates its growth in response to domestic
demand and to changes in comparative advantage (usually with some help from
commercial policies), the productivity gap tends to increase. Labor starts to shift
out of agriculture, at first in relative terms and eventually in absolute terms, but
with a lag. Since productivity in agriculture rises even at this stage, a surplus of
labor results.
The pattern of relative productivity in Figure 7.5 is related to and resembles
the Kuznets curve of income inequality. The productivity gap between primary
production on the one hand and industry and services on the other is greatest in
the middle income range, which is typically the period of greatest inequality of
income. It is also the period when, because of the productivity gap itself, resource
shifts can make their largest contribution to aggregate growth.
In a second phase, once migration and capital accumulation have significantly
reduced the surplus of labor, relative wages in agriculture increase and a catch-up
process takes place. Capital intensity in this sector then increases faster than in
other sectors. This is coupled with the continuing growth in factor productivity.
As a result, agriculture begins to reduce the productivity gap.
Crafts (1984) estimated output and employment patterns in agriculture from
data for nineteenth-century Europe, and calculated the implied productivity gap
between the primary sector and the rest of the economy. He found that the gap
narrowed substantially early in the transition, in contrast to "twentieth-century
countries" [from Chenery and Syrquin (1975)] where the gap grows throughout.
However, when comparing labor productivity in the primary sector to the
economy-wide figure the results are quite similar. In both cases, relative produc-
tivity tends to decline within the range studied by Crafts ($300 to $900 in 1970
US$).
6.3.2. The effect of reallocation of resources
Illustrative estimates of the contribution of resource reallocation to growth are

presented in Table 7.9. The results suggest that RE is a significant component of
aggregate growth of output and TFP particularly in the industrializing stage. At
its peak, it amounts to 11 percent of output growth and to almost 30 percent of
aggregate TFP. Its pattern across periods resembles the initial acceleration and
subsequent slowdown in the growth of output. In the last interval in the
table-which corresponds roughly to Western Europe since the late 1960s-the
effect of resource reallocation almost disappears. Part of the accompanying
productivity slowdown reflects the exhaustion of the shift out of agriculture as a
potential source of growth. If productivity gains continue to decline at income
258 M. Syrquin
Table 7.9
Contribution of resource shifts to productivity growth
for the cross-country model
Annual growth rate (percent)

Income Aggregate Reallocation
interval a GDP TFP (~) effect (RE)
I 4.8 0.7 0.15

II 5.7 1.4 0.29
III 6.3 2.3 0.56
IV 6.6 2.9 0.75
Higher income 5.6 2.8 0.08
alntervals as in Table 7.6. The higher income interval

corresponds roughly to $8000 to $12000.
Source: Syrquin (1986a).
levels beyond those in the simulation, however, it will no longer be attributable,

even partially, to the reduced shift out of agriculture. A different- and probably
negative - allocation effect may become important as labor continues to shift into
services.
In the presence of significant differences in factor returns across sectors,
structural change becomes an essential element in accounting for the rate and
pattern of growth. On the one hand, that change can retard growth if its pace is
too slow or its direction inefficient. On the other hand, it can contribute to
growth if it improves the allocation of resources. Market forces tend to move the
economic system toward equilibrium, but they are blunted by inflexibility in the
system and high adjustment costs, by shocks from external events and unbal-
anced productivity growth, and even by government policies (see Chapter 11 by
Williamson in this Handbook).
7. Relative prices and exchange rate conversions
Previous sections presented patterns of structural change, namely, systematic

associations of economic structure with the level of development. With few
exceptions, prices have not appeared directly in the analysis. For comparative
analysis of long-run transformation, the crucial question is not whether relative
prices do not change (they do), but whether they change in a systematic way
during the process of development. If such associations can be identified we may
want to add them to our list of stylized facts and to examine their impact on the
estimated patterns of change.
At a point in time, the internal structure of relative prices varies across
countries. The estimated income effects in cross-country patterns incorporate
price effects (and various other effects) correlated with per capita income. To the
extent that the association is expected to continue to hold, the combined total
income effects are of interest on their own, even if they no longer represent pure,
real income effects.
This section examines some key relative prices and the issues of exchange-rate
conversion of incomes.
7.1. Relative prices
There are various relative prices that can be expected to change in a regular
fashion (the wage-rental, for example). Of these, some are particularly relevant
for the study of structural change.
7.1.1. Consumer demand
The balanced-growth approach in the early literature on development

[Rosenstein-Rodan (1943), Nurkse (1953)] is commonly associated with price-in-
elastic demands. In this context, it is instructive to quote from Nurkse's posthu-
mous "Notes on 'Unbalanced Growth'" (1959, pp. 296-297):
Shortages in specific factor supplies, among other possible causes, can and will
naturally produce changes in relative prices, to which consumer demand will
tend to adjust itself. To this extent the output expansion will deviate from the
path of income elasticity, which assumes a constancy of relative prices. No one
has denied that price-elasticities, too, help to determine a community's pattern
of demand. But changes in relative prices have no close or determinate relation to
economic growth as such, whereas changes in income are a direct reflection and
measure of growth. Hence, the prominence given to income elasticities in a
macro-economic approach to the problem of international growth economics
(emphasis added).
7.1.2. Primary commodities
The internal and external terms of trade between primary commodities and
manufacturing have figured prominently in the development literature. Central to
the Prebisch-Singer advocacy of import-substitution industrialization was the
alleged secular tendency for the (external) terms of trade to move against primary
production. While a general tendency of deterioration in the terms of trade has
not been supported by the evidence, the debate is still far from over. The prices
of some primary products have fallen, but for other products sharp increases
were recorded, especially after 1970.
260 M. Syrquin
Regarding the internal terms of trade two approaches can be distinguished.
Work on the Soviet national income expected the prices of products based on
natural resources to rise relative to industrial products because of diminishing
returns in primary production and technological change in industry.73 A negative
association between relative prices of commodities and production levels implies
that the measurement of real growth over long periods of time will differ
according to whether we use initial or terminal year prices. This systematic
difference became known as the "Gerschenkron effect".
The second approach does not predict a tendency, rather it advocates turning
the internal terms of trade against agriculture to foster industrialization. It was
one of Preobrazhensky's policy proposals in the 1920s and was widely followed
by inward-oriented developing economies in the post-war period. TM
If the "Gerschenkron effect" predominates, that is, if over the transition the
relative price of primary commodities rises, then the extent of the real transfor-
mation in production, as estimated from current shares, is underestimated. The
real shift in economic activity is partly offset by an opposite price trend.
However, if the terms of trade are turned by policy against the primary sector,
the observed reallocation can only partially be regarded as real. Little, Scitovsky
and Scott (1970) analyzed the contributions of the primary and manufacturing
sectors to the growth of GDP, in six countries that pursued an import substitu-
tion strategy during the 1950s. They compared the contributions as convention-
ally measured and as measured after allowing for the effects of protection. In
every case the contribution of the primary sector was understated and that of
manufacturing overstated when ignoring protection. The results from a much
larger sample in Table 7.2, do not reveal any clear tendency in the terms of trade
for the longer period examined.
7.1.3. The price of food
Food is a primary-based commodity and therefore the previous discussion

applies to it, but since it is singled out by various analysts I shall add some
further observations.
In Lewis's (1954) model of development in the dual economy, a rising price of
food could choke off the process. Evidence on the importance of the relative price
of food appears in the comparative study of consumption patterns by Lluch,
Powell and Williams (1977). A decrease in the price of food was found to lead to
an increase in the demand for food, to an increase in the aggregate demand for
commodities other than food, and to a decrease in saving. The expected effects
73Perkins (1981).
74See Sah and Stiglitz(1984). In the USSR, however,the proposal was apparently not adopted
[EUman(1975)].
decline in importance at higher income levels. The various effects are quantita-
tively significant, but for them to affect the long-run patterns a systematic trend
in the price of food needs to be established. A general association of the price of
food with the level of income has neither been established over time nor across
countries, except for some partial samples. An interesting example of the latter is
Fishlow's (1973) comparison of American, British, and French expenditures of
urban households around 1900. Fishlow documented smaller American expendi-
tures on foodstuffs and related them to lower food prices. Since this took place at
a time when income was relatively low and the share of food still high, it gave the
low food prices more leverage, turning them into a significant factor in the rapid
rate of industrialization in the nineteenth century.
7.1.4. Price of investment goods
It is not always the case that constant-price shares are the relevant measures for
analysis. Saving behavior, for example, is more meaningful when expressed as
shares in current prices, while for comparisons of the productivity of investment
it is the constant share which is more relevant.
In the long-term records of developed countries analyzed by Kuznets (1961)
there was a tendency for the relative price of capital goods to rise, thus reducing
somewhat the secular increase in real capital formation proportions. 75 However,
" a n upward trend in the latter was still evident in most countries" (p. 54). The
increase in the relative price of investment was due to construction and not to
producers' durable equipment or inventories (p. 13). This implies a faster rate of
productivity growth in the tradable component of investment than in the non-
tradable one - a key stylized fact of the "differential productivity model" used to
account for the association of the national price level with the level of income
[Kravis and Lipsey (1983), and Section 7.2 below].
For a more recent period we have the price comparisons of the International
Comparisons Project (ICP) of Kravis and his associates. Summers and Heston
(1984) use ICP data to derive price relatives of broad expenditure categories for a
large number of countries during 1950-80. Regression results across countries
show a negative relation between the relative price of investment and real per
capita income, contrary to the long-term positive association found by Kuznets.
The results imply that the difference among countries in the real share of
investment is larger than the differential in nominal or current price shares.
Within countries, however, the experience since 1950 shows no decline in the
75The experience'in Japan was quite different. The impressiverise in the current share of capital
formation was accompaniedby a downward drift in the relativeprice of capital goods. The increase
in capital formation and saving shares was thereforeeven more impressivein constant prices. For an
endogenous explanation of accumulation and growth acceleration in Japan see Williamson and
de Bever (1978).
262 M. Syrquin
relative price of investment. The relative overvaluation of investment in low and

middle income countries has probably widened. Syrquin and Chenery (1986) also
report regressions of investment shares at constant 1970 prices. The estimated
income effects did not differ significantly from those at current prices.
7.1.5. Price o f services - non-tradables
The relative price of non-tradables increases significantly with the level of income
across countries and over time. Since this is the central element in the explana-
tion of the systematic departure of exchange rates from purchasing power parities
(PPPs) it is discussed in the next section on this topic.
7.2. Exchange-rate conversions
In the study of patterns of structural change the principal variable is the level of
development. In this chapter it has usually been measured by income per capita
in 1980 US dollars at official exchange rates. 76
It is well known that exchange rates are an imperfect measure of the purchas-
ing power of the various national currencies. Price structures vary across coun-
tries, and exchange rates at best reflect only the prices of internationally traded
goods in the absence of impediments to trade. The alternative to using exchange
rates as conversion factors for international comparisons is to reprice the local
components of income in every country at a uniform set of prices. This is
equivalent to the use of purchasing power parities (PPPs) as conversion factors,
where PPPs, following Kravis (1984), refer to the number of currency units
required to buy one US dollar's worth of output. PPPs are a weighted average of
the ratios of domestic price (Pi) to the price of the same good i in the country
chosen as numeraire (Pi*) or to its price in some international units.
There have been various attempts to estimate real incomes for small samples of
countries in which a high-income country was designated as numeraire. Until
recently, similar estimates for a large number of developing countries were not
available. This deficiency has begun to be remedied by the International Com-
parisons Project (ICP). In a series of studies, the ICP has estimated real incomes
for an ever increasing number of countries [see, for example, Kravis (1984)].
Phase III of the project, which includes 34 countries, became available in 1982
[Kravis, Heston and Summers (1982)]. Results of phase IV, covering more than
60 countries, were published recently [United Nations (1986)].
The principal finding of the ICP and similar studies is that differences across
countries in incomes converted at exchange rates tend to exaggerate the real
differences in income.
76This sectiondrawson the appendixto Cheneryand Syrquin(1986a).

Let YE stand for GDP converted at exchange rate and YR for real income. If
GDP is expressed in a country's own currency, the two income figures are
obtained as:
YE = G D P / e , YR = GDP/PPP,
where e is the exchange rate and PPP the purchasing power parity rate.
The systematic divergence between YE and YR per capita has been analyzed
in two equivalent ways:
(1) The ratio Y R / Y E , labeled by Kravis and associates the "exchange rate
deviation index" (ERD), declines steadily with the level of income per capita
from levels close to 3 for very low income countries to a value of I for the United
States.
(2) The reciprocal of the ERD is the national price level (PPP/e), which
increases systematically with per capita income.77
Several explanations have been proposed to account for the systematic pattern
of the ERD or of its inverse, the price level. The main one, which is most relevant
for our purposes, is the "differential productivity model", of which there are
various versions going back at least to Ricardo [see Kravis and Lipsey (1983)]. Its
modern version [see Balassa (1964) and Samuelson (1964)], accepts the law of one
price as approximately valid for traded goods but introduces a sector producing
non-traded goods. Assuming that productivity differences among countries are
larger for traded than for non-traded goods, that the production of the latter is
relatively more labor-intensive, and that labor is relatively more abundant
(relatively less expensive) in low income countries, the relative price of non-traded
goods can be expected to be positively associated with per capita income. Since
the prices of traded goods are similar across countries, the model also predicts
that the national price level will be higher in high income countries.
The differential productivity model relies on real factors related to the struc-
ture of the economy. In addition to the level of per capita income, other such
influences that have been suggested for examination include the industrial
structure of output and employment and the degree of openness of the economy.
These long-run factors determine the underlying price level, while short-run,
mainly monetary, variables are the cause of deviations from that basic level
[Kravis and Lipsey (1983, p. 10)].
The rise in the relative price of non-traded goods with the level of income
across countries can also be derived from a model where differences in total
factor productivity among sectors do not vary in a systematic way with income,
Bhagwati (1984) has recently shown that the price effect is compatible with the
traditional factor-proportion model in trade theory.
77The variation across countries in the national price level is analyzed in Kravis and Lipsey (1983)
and Clague (1985).
264 M. Syrquin
Effects of using e3cchange rates
There are two types of international comparisons that could be affected by using
exchange rates as conversion factors rather than purchasing power parities. The
first one relates measures of economic structure to per capita income across
countries and is essentially a static comparison. The second one refers to the
dynamic simulation of an economy over time.
Static comparisons include studies of the structure of demand and production
across countries as well as comparative static simulations of changes in structure
with incomes.
The patterns using exchange rates show the average transformation in relation
to income changes which incorporate a systematic variation in prices. If the price
variation is not correlated with per capita income among countries, then the use
of Y E instead of YR may affect the precision of the estimates but will not
necessarily bias the results. When the deviation of YE from YR is systematic, the
results will be biased. However, a close relation between YE and YR per capita
implies that there is a simple transformation between the results using YE and
the ones that would obtain if YR were to be used.
When the bias in using exchange rates affects all components of G D P equally,
the transformation in economic structure will take place over a narrower real
income range than the one shown here. The pace of the transformation is
therefore underestimated in our analysis.
A major component of the deviation between YE and YR is caused by
differences in price structures. To correct for these differences, separate correc-
tion factors are needed for different components of GDP. Although the ICP
examined expenditure categories, it does not provide information on the relative
prices of goods by sector of origin. 78 For this a production approach to real
income is necessary [see Maddison (1983) and Maddison and van Ark (1987)].
Once the time element comes in, the deviation of YE from YR becomes more
significant. Growth rate calculations and comparisons, such as those in Figures
7.3 and 7.4, are intended to be in real terms, whereas the changes across
benchmark levels of YE incorporate a price effect. The growth rates between
benchmark levels of YE in Figures 7.3 and 7.4 give the time required to traverse
the distance between the various levels of YE. The results are overestimates since
they do not take into account the variation in the price level during the same
interval. The quantitative results are sensitive to the use of exchange rates, but
the important question is the extent to which the qualitative results about the
transformation are also affected. As mentioned above, the use of exchange rates
as conversion factors stretches the income range studied. A uniform stretching
78See,nevertheless,the attempt by Kravis, Heston and Summers(1983) to use expendituredata to

examine the share of servicesin GDP. Ram (1985) compared estimates of production structure using
"conventional" and "real" GDP per capita. Sectoralshares were not correctedsince the required data
are not available.
would be the least damaging to the nature of the results and the easiest to
incorporate into the analysis. A uniform stretching implies a logarithmic relation
between Y R and YE, and most of the short-cut studies have postulated such a
logarithmic relation and estimated it with quite good results. Based on the results
of Phase III of the ICP, I estimated the following relation for 27 economies:
In YR = 2.81 + 0.68 In YE, ~2 = 0.97.

(16.7) (28.19)
The implication of the results is that real growth amounts to 68 percent of

growth in income converted at exchange rates at all income levels. The remaining
32 percent represents the systematic price effect.
The effect of switching to income converted at PPPs instead of exchange rates
on the income intervals in Tables 7.8 and 7.9 would be to reduce all of them
uniformly by 32 percent, but, most important, all the acceleration and decelera-
tion effects on growth would still remain.
8. Approaches to policy
Policy analysis is most useful when applied to concrete situations as illustrated,

for issues related to transformation, in the various chapters in Parts 2 and 6 of
this Handbook. This concluding section draws on these chapters and offers some
general observations on the role of the state in modern economic growth.
Structural change is a necessary corollary of economic growth but it is a
disruptive process. By definition it implies that various segments of the economy
grow at different rates. Groups of the population attached to slow growing
segments lose out relatively to those in faster growing sectors. In addition,
structural change requires the population to adapt itself to different working
conditions and styles of life (factory production, impersonal firms, urbanization).
As Kuznets noted, "policy action and institutional changes are required [to
minimize] the costs of, and resistance to, the structural shifts implicit in, and
required for, a high rate of growth" [Kuznets (1979, p. 130)]. For modern
economic growth the sovereign state has to act "as a clearing house for necessary
institutional innovations; as an agency for resolution of conflicts among group
interests; and as a major entrepreneur for the socially required infrastructure"
[Kuznets (1971, p. 346)].
The above describes what may be called the "minimal development state". 79At
the other extreme we find a state for which a certain pattern of structural change
791n the context of developmentit is a minimal state even if it goes well beyond the one in classical
liberal theory "limited to the functionsof protecting all its citizens against violence, theft, and fraud,
and to the enforcementof contracts, and so on" [Nozick (1974, p. 26)].
266 M. Syrquin
becomes itself an objective, to be actively pursued or even imposed. This may

take various forms: structural change may be fostered as the means to further
growth and other aims as in ECLA's advocacy of import substitution or in the
Stalinist approach to industrialization- or it could be dictated by ideology even
at the expense of growth, as in the extreme case of Cambodia. A less extreme,
and more positive case of determining structure exogenously is found in the early
years of the Israeli state. Ideology (return to the land) and security considera-
tions, called for an increase in agriculture's share in employment reversing the
transformation which would otherwise have taken place [Syrquin (1986c)]. Ad-
justment between demand and supply was left to foreign capital inflows in the
Israeli case and to coercion in Cambodia.
In between the two extremes there is a more pragmatic approach that tries to
anticipate structural change, facilitating it or accelerating it, by removing ob-
stacles and correcting for market failures. Following Pack and Westphal (1986),
its representatives can be labeled "industrial strategists". Strategists worry about
coordination of supply and demand and the design of adjustment mechanisms.
These are "policy measures that are adopted to avoid disequilibrium and thus
achieve more rapid development" [Chenery (1979, p. 62)]. 80
Issues tackled by strategists referred initially to factor accumulation (saving,
labor surplus), and sectoral balance (agriculture vs. industry) with trade a minor
consideration. At a later stage, trade strategy became the central issue.
Dissenters on the left regarded the approach as too narrow. Within mainstream
economics the main challenge came from the neoclassicals who typically grew up
in the tradition of international economics, in contrast to the strategists who
came from a planning tradition. 81
Comparative policy analyses relevant for the study of transformation have
been dominated by trade issues. The main examples are the studies of the import
substitution strategy of industrialization [Little, Scitovsky and Scott (1970),
Balassa and associates (1971)]; the analysis of trade regimes by Krueger (1978)
and Bhagwati (1978), and the studies of strategies in semi-industrial countries
[Balassa and associates (1982)].
The typology of specialization patterns, used in previous sections, also empha-
sized trade aspects. The main reason is that trade policy has become the main
instrument used by governments to influence resource allocation, and that
differences in trade policy have led to substantial variation in economic structures
and on performance [Chenery and Syrquin (1986b)]. The chapters in Part 6 of
this Handbook look at country experience. The grouping of countries or the
strategies analyzed are also trade determined. The other chapters in this part deal
8°Disequilibrium has also been regarded as beneficial because it ellicits creative responses
[Hirschman (1958)] and generates desirable forces such as innovation and competition [Kornai
(1971)].81 See also Hahn's (1973) review of Kornal"'s book.
The following paragraphs owe much to Pack and Westphal (1986).
with external policy, but also emphasize internal aspects chiefly related to
technological change.
The neoclassical approach focuses on the overall policy regime (it should be
neutral not discriminatory) and, in one way or another, recommends "getting the
prices right". This is important not just for the static gains of a more efficient
allocation of resources, but more so for fostering technical change and dynamic
efficiency.
Strategists have incorporated many of the neoclassical prescriptions, but
remain convinced of the indispensability of selective intervention. They see
industrialization as a process of technological change, with learning related to
experience as a major source of externality. The generation of local technological
capability becomes crucial and the consequence is to focus decisions about
strategy and intervention on areas where technological capability either does not
exist or exists in at least a rudimentary form but is not being deployed effectively.
The neoclassical emphasis on choices among policy instruments is therefore
regarded as incomplete. No less important are "the different ways of using the
same policy instruments.., whether they are used promotionally or restrictively"
[Pack and Westphal (1986, p. 103)]. This distinction is related to a missing link
stressed in various chapters, namely, the management factor. Selective interven-
tion "works" only if its primary objective is a dynamically efficient process of
industrialization. If it is not, then selective intervention may be counterproduc-
tive and the best advice is probably to adhere to the neoclassical prescription.
"However, it should be noted that the factors responsible for a government's
inability to intervene effectively may also preclude its following the neoclassical
prescription" [Pack and Westphal (1986, p. 104)].
Managerial factors are brought out clearly in Mason's (1984) comparison of
development policy and its implementation in Egypt and Korea. He considers
differences in various managerial factors such as government objectives, the share
of public enterprises in the public sector, managerial techniques in the public
sector, implementation of policy, etc. The first one he regards as the most
important factor: in Korea the primary purpose of government was to facilitate
growth, while in Egypt it was not. 82 To this he adds one non-economic influence
of "possibly overwhelming importance"-the cultural milieu. Cultural and social
factors assume a similar conditioning role in the staple theory of growth. Exports
of primary commodities (staples) can ignite a cumulative process of development
only if no inhibiting traditions are present and if society and its institutions are
favorably predisposed toward development [Roemer (1970)].
The importance of institutions and institutional change for modern economic
growth is increasingly being recognized. Morris and Adelman (1988) go as far as
82This iS similar to the argument in Adelman and Morris (1967), mentioned approvingly by
Mason, that the single most important factor has been the leadership commitment to economic
development.
268 M. Syrquin
identifying institutional change as the single most important differentiator of

development performance among groups of countries. 83 The time may be ripe for
trying to establish the stylized facts of institutional change and their role in the
structural transformation.
References
Abramovitz, M. (1983) 'Notes on international differences in productivity growth rates', in: D.C.
Mueller, ed., The political economy of growth. New Haven: Yale University Press.
Abramovitz, M. and David, P.A. (1973) 'Reinterpreting economic growth: Parables and realities',
American Economic Review, 63:428-439.
Adelman, I. and Morris, C.T. (1967) Economic growth and social equity in developing countries.
Stanford: Stanford University Press.
Adelman, I. and Morris, C,T. (1984) 'Patterns of economic growth, 1850-1914, or Chenery-Syrquin
in perspective', in: M. Syrquin, L. Taylor and L.E. Westphal, eds., Economic structure and
performance: Essays in honor of Hollis B. Chenery. New York: Academic Press.
Arndt, H.W. (1985) 'The origins of structuralism', Worm Development, 13:151-159.
Balassa, B. (1964) 'The purchasing power parity doctrine: A reappraisal', Journal of Political
Economy, 72:584-596.
Balassa, B. (1979) " ' A stages approach' to comparative advantage", in: I. Adelman, ed., Economic
growth and resources, Vol. 4: National and International Policies. London: Macmillan.
Balassa, B., and Associates (1971) The structure of protection in developing countries. Baltimore, MD:
Johns Hopkins University Press.
Balassa, B., and Associates (1982) Development strategies in semi-industrial countries. Baltimore, MD:
Johns Hopkins University Press.
Baldwin, R.E. (1956) 'Patterns of development in newly settled regions', Manchester School of
Economic and Social Studies, 24:161-179.
Bauer, P. and Yamey, B.S. (1951) 'Economic progress and occupational distribution', Economic
Journal, 61:741-755.
Bhagwati, J.N. (1978) Foreign trade regimes and economic development: Anatomy and consequences of
exchange control regimes. Cambridge, MA: Ballinger.
Bhagwati, J.N. (1984) 'Why are services cheaper in the poor countries?', Economic Journal,
94:279-286.
Carre, J.J., Dubois, P. and Malinvaud, E. (1975) French economic growth. Stanford, CA: Stanford
University Press.
Caves, R.E. (1965) "'Vent for surplus' models of trade and growth", in: R.E. Baldwin et al. eds.,
Trade, growth and the balance of payments. Chicago: Rand McNally.
Chenery, H.B. (1960) 'Patterns of industrial growth', American Economic Review, 50:624-654.
Chenery, H.B, (1964) 'Land: The effects of resources on economic growth', in: K. Berrill, ed.,
Economic development with special reference to East Asia. New York: St. Martin's.
Chenery, H.B. (1973) 'Alternative strategies for development', IBRD working paper no. 165, World
Bank, Washington, DC. Revised version published in: Y. Ramati, ed., Economic growth in
developing countries material and human resources, Proceedings of the Seventh Rehovot Conference.
New York: Praeger (1975).
Chenery, H.B. (1975) 'The strncturalist approach to development policy', American Economic
Association Papers and Proceedings, 65:310-316.
Chenery, H.B. (1977) 'Transitional growth and world industrialization', in: B. Ohlin, P.O. Hesselborn
and P.M. Wijkman, eds,, The international allocation of economic activity. London: Macmillan.
83See also Matthews (1986) and Olson (1982).

Chenery, H.B. (1979) Structural change and development policy. New York: Oxford University Press.
Chenery, H.B. (1986a) 'Structural transformation: A program of research', Harvard Institute for
International Development, Discussion Paper no. 232, processed.
Chenery, H.B. (1986b) 'Growth and transformation', in: H. Chenery, S. Robinson and M. Syrquin,
Industrialization and growth. New York: Oxford University Press.
Chenery, H.B. and Bruno, M. (1962) 'Development alternatives in an open economy: The case of
Israel', Economic Journal, 72:79-103.
Chenery, H.B. and Strout, A. (1966) 'Foreign assistance and economic development', American
Economic Review, 56:679-733.
Chenery, H.B. and Syrquin, M. (1975) Patterns of det;elopment, 1950-1970. London: Oxford
University Press.
Chenery, H.B. and Syrquin, M. (1980) 'A comparative analysis of industrial growth', in: R.C.O.
Matthews, ed., Economic growth and resources, Vol. 2: Trends and Factors. New York: Macmillan.
Chenery, H.B. and Syrquin, M. (1986a) 'Typical patterns of transformation', in: H. Chenery, S.
Robinson and M. Syrquin, Industrialization and growth. New York: Oxford University Press.
Chenery, H.B. and Syrquin, M. (1986b) 'The semi-industrial countries', in: H. Chenery, S. Robinson
and M. Syrquin, Industrialization and growth. New York: Oxford University Press.
Chenery, H.B. and Taylor, L. (1968) 'Development patterns among countries and over time', Review
of Economics and Statistics, 50:391-416.
Chenery, H.B. and Watanabe, T. (1958) 'International comparisons of the structure of production',
Econometrica, 26:487-521.
Chenery, H.B. and Watanabe, T. (1965) 'The process of industrialization', Paper read at the World
Congress of the Econometric Society, Rome, Italy, September 9-14.
Chenery, H.B., Robinson, S. and Syrquin, M. (1986) Industrialization and growth: A comparative"
study. New York: Oxford University Press.
Chenery, H.B., Shishido, S. and Watanabe, T. (1962) 'The pattern of Japanese growth, 1914-1954',
Econometrica, 30:98-139.
Clague, C.K. (1985) 'A model of real national price levels', Southern Economic Journal, 51:998-1017.
Clark, C. (1940) The conditions of economic progress. London: Macmillan.
Corden, W.M. and Neary, J.P. (1982) 'Booming sector and de-industrialization in a small open
economy', Economic Journal, 92:825-848.
Crafts, N.F.R. (1984) 'Patterns of development in nineteenth century Europe', Oxford Economic
Papers, 36:438-458.
Deardorff, A.V. (1984) 'Testing trade theories and predicting trade flows', in: R.W. Jones and P.B.
Kenen, eds., Handbook of international economics, Vol. 1. Amsterdam: North-Holland.
Deutsch, J. and Syrquin, M. (1986) 'Economic development and the structure of production',
processed.
Domar, E. (1957) 'A Soviet model of growth', in: Essays in the theory of economic growth. New York:
Oxford University Press.
Eckaus, R. (1978) 'Review of Chenery and Syrquin, patterns of development, 1950-1970', Economic
Development and Cultural Change, 26:621-625.
ECE (Economic Commission for Europe) (1977) Structure and change in European industry. New
York: United Nations.
Ellman, M.J. (1975) 'Agricultural surplus and increase in investment: USSR 1928-32', Economic
Journal, 85:844-863.
Fei, J.C.H. and Ranis, G. (1964) Development of the labor surplus economy. Homewood, IL: Irwin.
Feldstein, M. and Horioka, C. (1980) 'Domestic saving and international capital flows', Economic
Journal, 90:314-329.
Fisher, A.G.B. (1935) The clash of progress and security. London: Macmillan.
Fisher, A.G.B. (1939) 'Production, primary, secondary and tertiary', Economic Record, 15:24-38.
Fishlow, A. (1965) 'Empty economic stages?', Economic Journal, 75:112-125.
Fishlow, A. (1973) 'Comparative consumption patterns, the extent of the market, and alternative
development strategies', in: E.B. Ayal, ed., Micro aspects of development. New York: Praeger.
Gerschenkron, A. (1962) Economic backwardness in historical perspective. Cambridge, MA: Belknap.
Gillis, M., Perkins, D.H., Roemer, M.R. and Snodgrass, D.R. (1983) Economics of development. New
York and London: Norton.
270 M, Syrquin
Grunfeld, Y. and Griliches, Z. (1960) 'Is aggregation necessarily bad?', Review of Economics and
Statistics, 42:1-13.
Hahn, F.H. (1973) 'Review of Kornai, Anti-equilibrium', Economica, 40:323-330.
Hartwell, R.M. (1973) 'The service revolution: The growth of services in modern economy', in: C.M.
Cipolla, ed., The Fontana economic history of Europe: The industrial revolution. Glasgow: Collins.
Hayami, Y. and Ruttan, V. (1985) Agricultural development: An international perspective. Baltimore,
MD: Johns Hopkins University Press.
Helpman, E. and Krugman, P.R. (1985) Market structure and foreign trade. Cambridge, MA: MIT
Press.
Hirschman, A.O. (1958) The strategy of economic development. New Haven, CT: Yale University
Press.
Hirschman, A.O. (1977) 'A generalized linkage approach to development, with special reference to
staples', Economic Development and Cultural Change, 25(supplement):67-98.
Hoffmann, W.G. (1958) The growth of industrial economies. Manchester: Manchester University Press.
Hosefitz, B.F. (1960) 'Theories of stages of economic growth', in: B.F. Hoselitz, ed., Theories of
economic growth. New York: The Free Press.
Houthakker, H.S. (1957) 'An international comparison of household expenditure patterns: Commem-
orating the centenary of Engel's law', Econometrica, 25:532-551.
Houthakker, H.S. (1965) 'On some determinants of savings in the developed and underdeveloped
countries', in: E.A.G. Robinson, ed., Problems of economic development. London: Macmillan.
Jackson, L.F. (1984) 'Hierarchic demand and the Engel curve for variety', Review of Economics and
Statistics, 66:8-15.
Johnson, H.G. (1965) 'A theoretical model of economic nationalism in new and developing states',
Political Science Quarterly, 80:169-185.
Kaldor, N. (1961) 'Capital accumulation and economic growth', in: F.A. Lutz and D.C. Hague, eds.,
The theory of capital. London: Macmillan.
Kaldor, N. (1985) Economics without equilibrium. New York: M.E. Sharpe, Inc.
Kelley, A.C. and Williamson, J.G. (1984) What drives third world city growth? A dynamic general
equilibrium approach. Princeton, NJ: Princeton University Press.
Kelley, A.C., Williamson, J.G. and Cheetham, R. (1972) Dualistic economic development. Chicago:
University of Chicago Press.
Kindleberger, C.P. (1967) Europe's postwar growth: The role of labor supply. Cambridge, M.A:
Harvard University Press.
Komai, J. (1971) Anti-equilibrium. Amsterdam: North-Holland.
Kravis, I.B. (1984) 'Comparative studies of national incomes and prices', Journal of Economic
Literature, 22:1-39.
Kravis, I.B. and Lipsey, R.E. (1983) Towards an explanation of national price levels. Special Studies in
International Finance 52. Princeton, NJ: Princeton University Press.
Kravis, I.B., Heston, A.W. and Summers, R. (1982) World product and income: International compari-
sons of real gross product. Baltimore, MD: Johns Hopkins University Press.
Kravis, I.B, Heston, A.W. and Summers, R. (1983) 'The share of services in economic growth', in:
F.G. Adams and B. Hickman~ eds., Global econometrics. Cambridge, MA: MIT Press.
Krueger, A.O. (1978) Foreign trade regimes and economic development: Liberalization attempts and
consequences. New York: National Bureau of Economic Research.
Kubo, Y., de Melo, J. and Robinson, S. (1986) 'Trade strategies and growth episodes', in: H.B.
Chenery, S. Robinson and M. Syrqnin, Industrialization and growth. New York: Oxford University
Press.
Kubo, Y., Robinson, S. and Syrquin, M. (1986) 'The methodology of multisector comparative
analysis', in: H.B. Chenery, S. Robinson and M. Syrqnin, Industrialization and growth. New York:
Oxford University Press.
Kuznets, S. (1951) 'The state as a unit in the study of economic growth', Journal of Economic History,
11:25-41.
Kuznets, S. (1956-1967) 'Quantitative aspects of the economic growth of nations', Economic
development and cultural change (a series of 10 articles).
Kuznets, S. (1957) 'Quantitative aspects of the economic growth of nations: II. Industrial distribution
of national product and labor force', Economic Development and Cultural Change, 5:supplement.
Kuznets, S. (1959) 'On comparative study of economic structure and growth of nations', in: National
Bureau of Economic Research, The comparative study of economic growth and structure. New York:
NBER.
Kuznets, S. (1961) 'Quantitative aspects of the economic growth of nations: IV. Long term trends in
capital formation proportions', Economic Development and Cultural Change, 9:1-124.
Kuznets, S. (1966) Modern economic growth. New Haven, CT: Yale University Press.
Kuznets, S. (1971) Economic growth of nations: Total output and production structure. Cambridge, MA:
Kuznets, S. (1973) 'Modern economic growth: Findings and reflections', American Economic Review,
63:247-258.
Kuznets, S. (1979) 'Growth and structural shifts', in: W. Galenson, ed., Economic growth and
structural change in Taiwan: The postwar experience of the Republic of China. Ithaca, NY: Cornell
University Press.
Landes, D. (1969) The unbound Prometheus: Technological change and industrial development in
Western Europe from 1750 to the present. Cambridge: Cambridge University Press.
Learner, E.E. (1984) Sources of international comparative advantage. Cambridge, MA: MIT Press.
Left, N.H. (1969) 'Dependency rates and savings rates', American Economic Review, 59:886-896.
Lewis, W.A. (1954) 'Economic development with unlimited supplies of labor', Manchester School of
Economic and Social Studies, 22:139-191.
Lewis, W.A. (1984) 'The state of development theory', American Economic Review, 74:1-10.
Linder, S.B. (1961) An essay on trade and transformation. New York: Wiley.
Little, I.M.D. (1982) Economic development: Theory, policy and international relations. New York:
Basic Books.
Little, I.M.D., Scitovsky, T. and Scott, M. (1970) Industry and trade in some developing countries: A
comparative study. London: Oxford University Press.
Lluch, C.A., Powell, A. and Williams, R.A. (1977) Patterns in household demand and savings. New
York: Oxford University Press.
Machlup, F. (1963) Essays in economic semantics. Englewood Cliffs, N J: Prentice-Hall.
Maddison, A. (1980) 'Economic growth and structural change in the advanced countries', in: I.
Leveson and J.W. Wheeler, eds., Western economies in transition. Boulder, CO: Westview Press.
Maddison, A. (1982) Phases of capitalist development. New York: Oxford University Press.
Maddison, A. (1983) 'A comparison of levels of GDP per capita in developed and developing
countries, 1700-1980', Journal of Economic History, 43:27-41.
Maddison, A. and van Ark, B. (1987) 'International comparisons of purchasing power, real output,
and productivity in manufacturing: A pilot study for Brazil, Mexico, and the U.S.A., 1975',
Groningen, processed.
Mahalanobis, P.C. (1953) 'Some observations on the process of growth of national income', Sankhy~,
12:307-312.
Maizels, A. (1963) Industrial growth and world trade. Cambridge: Cambridge University Press.
Mason, E.S. (1984) 'The Chenery analysis and some other considerations', in: M. Syrquin, L. Taylor
and L.E. Westphal, eds., Economic structure and performance: Essays in honor of Hollis B. Chenery.
New York: Academic Press.
Matthews, R.C.O. (1986) 'The economics of institutions and the sources of growth,' Economic
Journal, 96:903-918.
Matthews, R.C.O., Feinstein, C. and Odling-Smee, C. (1982) British economic growth. Oxford: Oxford
University Press.
McCarthy, F.D., Taylor, L, and Talati, C. (1987) 'Trade patterns in developing countries, 1964-82',
Journal of Development Economics, to appear.
Mellor, J.W. (1986) 'Agriculture on the road to industrialization', in: J.P. Lewis and V. Kallab, eds.,
Development strategies reconsidered. New Brunswick, NJ: Transaction Books for the Overseas
Development Council.
Melo, J. de and Robinson, S. (1985) 'Product differentiation and trade dependence of the domestic
price system in computable general equilibrium trade models', in: T. Peeters, P. Praet and P.
Reding, eds., International trade and exchange rates in the late eighties. Amsterdam: North-Holland.
Morris, C.T. and Adelman, I. (1988) Comparative patterns of economic development, 1850-1914.
Baltimore, MD: Johns Hopkins University Press.
272 M. Syrquin
Myrdal, G. (1957) Economic theoo' and under-developed regions. London: Duckworth.

National Bureau of Economic Research (1959) The comparative stud), of economic growth and
structure. New York: NBER.
North, D.C. (1981) Structure and change in economic history. New York: Norton.
Nozick, R. (1974) Anarchy, state, and utopia, New York: Basic Books.
Nurkse, R. (1953) Problems of capital formation in underdeveloped countries. Oxford: Basil Blackwell.
Nurkse, R. (1959) "Notes on 'unbalanced growth'" Oxford Economic Papers, 7:295-297.
O'Brien, P.K. and Keyder, C. (1978) Economic growth in Britain and France, 1780-1914. London:
Allen and Unwin.
Ohkawa, K. and Ranis, G., eds., (1985) Japan and the developing countries: A comparative analysis.
Oxford: Basil Blackwell.
Ohkawa, K. and Rosovsky, H. (1973) Japanese economic growth: Trend acceleration in the twentieth
century. Stanford, CA: Stanford University Press.
Olson, M. (1982) The rise and decline of nations. New Haven, CT: Yale University Press.
Pack, H. and Westphal, L.E. (1986) 'Industrial strategy and technological change: Theory vs. reality',
Journal of Development Economics, 22:87-128.
Perkins, D.H. (1981) 'Three decades of international quantitative comparisons', Harvard Institute for
International Development.
Perroux, F. (1955) "Note sur la notion de 'PSle de croissance' ", MatOriaux pour une analyse de la
croissance economique. Cahiers de l'Institute de Science Economique Applique, Serie D, No. 8.
Prakash, V. and Robinson, S. (1979) 'A cross-country analysis of patterns of industrial growth',
Development Economies Department, The World Bank, Washington, DC.
Ram, R. (1985) "Conventional and 'real' GDP per capita in cross-country studies of production
structure", Journal of Development Economics, 18:463-477.
Ranis, G. (1984) 'Typology in development theory: Retrospective and prospects', in: M. Syrquin, L.
Taylor and L.E. Westphal, eds., Economic structure and performance: Essays in honor of Hollis B.
Chenery. New York: Academic Press.
Roemer, M. (1970) Fishing for growth: Export-led development in Peru, 1950-1967. Cambridge, MA:
Roemer, M. (1985) 'Dutch disease in developing countries: Taking bitter medicine', in: M. Lundahl,
ed., The primary sector in economic development. New York: St. Martin's.
Rosenstein-Rodan, P. (1943) 'Problems of industrialization in Eastern and South-Eastern Europe',
Economic Journal, 53:202-211.
Rosenstein-Rodan, P. (1961) "Notes on the theory of the 'big push' ", in: H.S. Ellis and H.C. Wallich,
eds., Economic development for Latin America. London: Macmillan.
Rostow, W.W. (1960) The stages of economic growth: A non-communist manifesto. Cambridge:
Cambridge University Press.
Sah, R.K. and Stiglitz, J.E. (1984) 'The economics of price scissors', American Economic Review,
74:124-138.
Samuelson, P.A. (1964) 'Theoretical notes on trade problems', Review of Economics and Statistics,
46:145-154.
Solow, R.M. (1956) 'A contribution to the theory of economic growth', Quarterly Journal of
Economics, 70:65-94.
Solow, R.M. (1977) 'Comment of Chenery (1977)', in: B. Ohlin, P.O. Hesselborn and P.M. Wijkman,
eds., The international allocation of economic activity. London: Macmillan.
Summers, R. and Heston, A.W. (1984) 'International comparisons of real product and its composi-
tion: 1950-80', Review of Income and Wealth, 30:207-262.
Svennilson, I. (1954) Growth and stagnation in the European economy. Geneva: United Nations
Economic Commission for Europe.
Svennilson, I. (1960) 'The concept of the nation and its relevance to economic analysis', in: E.A.G.
Robinson, ed., Economic consequences of the size of nations. London: Macmillan.
Syrquin, M. (1976) 'Sources of industrial growth and change: An alternative measure', paper
presented at the European Meeting of the Econometric Society, Helsinki, Finland, August.
Syrquin, M. (1978) 'The share of trade and economic development', processed. Hebrew version in: N.
Halevi and Y. Kop, eds., Issues in economics 1977. Jerusalem: Falk Institute.
Ch, 7: Patterns of Structural Change 273
Syrquin, M. (1984) 'Resource reallocation and productivity growth', in: M. Syrquin, L. Taylor and
L.E. Westphal, eds., Economic structure and performance: Essays in honor of Hollis B. Chene~,. New
York: Academic Press.
Syrquin, M. (1985) 'Patterns of development since 1960: A comparison for China', processed.
Syrquin, M. (1986a) 'Productivity growth and factor reallocation', in: H.B. Chenery, S. Robinson and
M. Syrquin, eds., Industrialization and growth, New York: Oxford.
Syrquin, M. (1986b) 'Sector proportions and economic development: The evidence since 1950', paper
presented at the 8th World Congress of the International Economic Association, New Delhi, India.
Syrquin, M. (1986c) 'Economic growth and structural change in Israel: An international perspective',
in: Y. Ben Porath, ed., The Israeli economy: Maturing through crises. Cambridge, MA: Harvard
University Press.
Syrquin, M. and Chenery, H.B. (1986) 'Patterns of development: 1950 to 1983', processed, World
Bank.
Taylor, L. (1969) 'Development patterns: A simulation study', Quarterly Journal of Economics,
83:220-241.
Taylor, L. (1975) 'Theoretical foundations and technical implications', in: C.R. Blitzer, P.B. Clark
and L. Taylor, eds., Economy-wide models and development planning. London:Oxford University
Press.
Taylor, L. (1979) Macro models for developing countries. New York: McGraw-Hill.
Taylor, L. (1986) 'Theories of sectoral balance', paper presented at the 8th World Congress of the
International Economic Association, New Delhi, India.
Taylor, L., Sarkar, H. and Rattso, J. (1984) 'Macroeconomic adjustment in a computable general
equilibrium model for India', in: M. Syrquin, L. Taylor and L.E. Westphal, eds., Economic structure
and performance: Essays in honor of Hollis B. Chenery. New York: Academic Press.
Teitel, S. and Thoumi, F.E. (1986) 'From import substitution to exports: The manufacturing exports
experience of Argentina and Brazil', Economic Development and Cultural Change, 34:455-490.
UNIDO (1979) World industry since 1960." Progress and prospects, New York: United Nations.
UNIDO (1983) Industry in a changing world. New York: United Nations.
United Nations (1986) Worm comparisons of purchasing power and real product for 1980, phase IV of
the international comparison project. New York: United Nations.
Watkins, M.H. (1963) 'A staple theory of economy growth', Canadian Journal of Economics and
Political Science, 29:141-158.
Westphal, L.E. (1982) 'Fostering technological mastery by means of selective industry promotion', in:
M. Syrquin and S. Teitel, eds., Trade, stability, technology and equi~ in Latin America. New York:
Academic Press.
Williamson, J.G. and de Bever, L.J. (1978) 'Saving, accumulation and modern economic growth: The
contemporary relevance of Japanese history', Journal of Japanese Studies, 4:125-167.
Wood, A. (1986) 'Puzzling trends in real exchange rates: A preliminary analysis', processed.
World Bank (1980) Worm development report 1980. New York: Oxford University Press.
World Bank (1984) World development report 1984. New York: Oxford University Press.
Young, A.A. (1928) 'Increasing returns and economic progress', Economic Journal, 38:117-132.
View publication stats

Patterns of Structural Change

Uploaded by

Copyright:

Available Formats

Patterns of Structural Change

Uploaded by

Document Information

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

Patterns of Structural Change

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Patterns of Structural Change

Article in Handbook of Development Economics · December 1988

The user has requested enhancement of the downloaded file.

PATTERNS OF STRUCTURAL CHANGE

Handbook of Development Economics, Volume L Edited by H. Chenery and T.N. Srinivasan

6. A c c o u n t i n g for the transformation 250

Development economics can be characterized as dealing with issues of structure

1. The study of structural change

6North (1981) gives an interpretation of structural change in economic history as institutional

In the cobweb model, for example, a position "will be both an equilibrium or a

1.1. Structural change in economic history

The following sections present a selective review of the evolution of thought on

8For a reviewof "stages"theoriesof growth see Hoselitz(1960).

of discontinuity that, in one way or another, appear in most discussions on

9The "two theories" of take-off follows the interpretation of Fishlow (1965).

1.2. Structural change in development economics

The emphasis on demand, reinforced by Keynesian theory, greatly influenced the

16For a critical evaluation see Kuznets (1966).

1.3. The need for a typology

An additional feature of Chenery's approach, important for the links between

2. Empirical research on the structural transformation

2.1. Bases for comparative analysis

The empirical research program on growth and transformation originating with

2.2. A unique path of development?

2.3. The methodology of comparative analysis

270'Brien and Keyder (1978, p. 16).

2.3.2. Models vs. description

The distinction between a modelling approach and a purely descriptive one is

2.4. Time-series vs. cross-section studies

Empirical research on the features of modern economic growth started with

impact of varying policies, the omission of dynamic effects (besides those

3. Patterns of growth and accumulation

Empirical regularities accompanying development are commonly referred to as

3.1. Growth patterns

The most obvious characteristic of modern economic growth in developed

higher than in previous periods. An acceleration of growth in middle-income

Accumulation refers to the use of resources to increase the productive capacity of

resources (health, education); and investment in other public services which

44On "capital fundamentalism" see, for example, Gillis et al. (1983).

4. Changes in sector proportions

Changes in the sectoral composition of production are the most prominent

4.1. The accounting framework

The accounting framework specifies a sectoral breakdown of the national income

where Y is gross domestic product, C is private consumption, G is government

Xi = W~+ D, + T~, (4.2)

where Xi is gross output of sector i, and W~ is intermediate demand for the

where Uj is intermediate purchases by sector j, Vj is value added in sector j, and

V = Y'~ Vj= Y. (4.5)

Patterns of industrialization are chiefly concerned with changes in the distribu-

52Sectors may be distinguishedaccordingto variouscriteria. The issue has recentlybeen addressed

4.2. Final demand

4.3. Intermediate demand

54Kuznets(I966), Cheneryand Syrquin(1975), Syrquinand Chenery(1986).

In a relatively closed economy the structure of production has to conform closely

57Strictlyspeaking, the changein factorproportionshas to be comparedwith the one takingplace

2~00 600 1200 2400 4500

~o0 6oo 1200 24oo45oo

Changes in the commodity composition of trade within manufacturing have

inter-country differences in the structure of exports are in large part explained