74247 HD82 .C472 1986 c.l2 Chenery, Hollis Burnley. Industrialization and growth a comparative study I Industrialization and Growth A WORLD BANK RESEARCH PUBLICATION with contributions by Gershon Feder Yuji Kubo Jeffrey Lewis Jaime de Melo Mieko Nishimizu INDUSTRIALIZATION AND GROWTH A Comparative Study Hollis Chenery Sherman Robinson Moshe Syrquin PUBLISHED FOR THE WORLD BANK Oxford University Press Oxford University Press NEW YORK OXFORD LONDON GLASGOW TORONTO MELBOURNE WELLINGTON HONG KONG TOKYO KUALA LUMPUR SINGAPORE JAKARTA DELHI BOMBAY CALCUTTA MADRAS KARACHI NAIROBI DAR ES SALAAM CAPE TOWN © 1986 The International Bank for Reconstruction and Development I THE WORLD BANK 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Manufactured in the United States of America First printing October 1986 The findings, interpretations, and conclusions expressed in this study are the results of country economic analysis or research done by the World Bank, but they are entirely those of the authors and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Library of Congress Cataloging-in-Publication Data Chenery, Hollis Burnley. Industrialization and growth. Bibliography: p. Includes index. 1. Economic development. 2. Industrialization. I. Robinson, Sherman. II. Syrquin, Moshe. III. International Bank for Reconstruction and Development. IV. Title. HD82.C472 1986 338.9 86-21837 ISBN 0-19-520547-2 Contents Preface tx 1 Introduction 1 The Issues 1 The Research Design 5 A Reader's Guide 7 Part I Structural Transformation 11 2 Growth and Transformation 13 HOLLIS CHENERY The Sources of Growth 14 Equilibrium Growth 16 Disequilibrium Growth 27 Structural Transformation 31 3 Typical Patterns of Transformation 37 HOLLIS CHENERY MOSHE SYRQUIN Modeling the Structural Transformation 38 Common Features of Industrialization 54 Dynamics of the Transformation 68 Appendix. Comparisons of Real Income across Countries and over Time 78 4 The Semi-Industrial Countries 84 HOLLIS CHENERY MOSHE SYRQUIN Types of Industrialization 85 Strategy and Performance 94 The Timing of Structural Change 99 Appendix. The Index of Trade Orientation 114 Part II The Experience of Industrialization 119 5 The Methodology of Multisector Comparative Analysis 121 YUJI KUBO SHERMAN ROBINSON MOSHE SYRQUIN The Static Input-Output Model 122 Sources of Growth and Structural Change 128 A Dynamic Computable General Equilibrium Model 143 The Methodology of Model-Based Comparisons 145 v vi CONTENTS 6 Trade Strategies and Growth Episodes 148 YUJI KUBO JAIME DEMELO SHERMAN ROBINSON Initial Conditions and Aggregate Performance 149 Sources of Change in Demand and Output 153 Trade Policy Regimes 165 Trade and Industry 172 • Conclusions 187 7 Interdependence and Industrial Structure 188 YUJI KUBO JAIME DE MELO SHERMAN ROBINSON MOSHE SYRQUIN Aggregate Measures of Structural Change 189 Interindustry Linkages and the Complexity of Production 202 Openness and Comparative Advantage in Manufacturing 209 ' Trade and Input Use 213 Conclusions 223 Part III Productivity and Structural Change 227 8 Productivity Growth and Factor Reallocation 229 MOSHE SYRQUIN The Growth of Output and Factor Use 230 Factor Productivity Growth 242 Resource Reallocation as a Source of Growth 252 Productivity and Growth: Demand-Supply Interactions 259 9 Growth in Semi-Industrial Countries: A Statistical Analysis 263 GERSHON FEDER The Disequilibrium Model 262 A Synthesis: Disequilibrium among Industry, Nonindustry, Export, and Nonexport Sources 277 Alternative Interpretations of the Results 280 10 Productivity Growth in Manufacturing 283 MIEKO NISHIMIZU SHERMAN ROBINSON The Analytical Framework for Measuring Total Factor Productivity 286 Growth and Productivity Change in the Manufacturing Industries 288 Trade Strategies and TFP Growth 298 CONTENTS vii Part IV Development Strategy 309 11 Alternative Routes to Development 311 HOLLIS CHENERY JEFFREY LEWIS JAIME DE MELO SHERMAN ROBINSON Three Development Strategies 312 A Dynamic Computable General Equilibrium Model 313 Macroeconomics of Alternative Strategies 320 Prices, Incentives, and Structural Change 331 Conclusions 337 Appendix A. Equations of the CGE Model 340 Appendix B. The Critical Parameters of the CGE Model 344 12 Growth and Structure: A Synthesis 348 The Structural Transformation 348 The Role of Industrialization 350 Productivity 354 Policy 355 References 3 61 Index 379 Preface DEvELoPMENT is now conceived as the successful transformation of the structure of an economy. In his historical studies of modern economic growth, Kuznets (1966) identified the shift of resources from agriculture to industry as the central feature of this transformation. While the postwar experience of developing countries shows industrialization to be highly correlated with rising income, it also reveals substantial differences due to resource endowments and government policies. The causal factors behind these relations are a matter of dispute. The sources of industrialization range from the need to adapt the composition of supply to shifts in domestic demand to the exploitation of comparative advantage in labor-intensive activities. In the past decade such historical trends have been modified as some countries have accelerated their indus- trialization to offset worsening terms of trade, while favored primary producers have suffered from "Dutch disease" and a tendency to deindus- trialize. Both the long-term tendency of middle-income countries to indus- trialize and variations from it need to be evaluated in designing develop- ment policy. The present volume is one of a series of studies of the structural trans- formation that have been supported by the World Bank. Earlier work by Chenery and Syrquin (1975, 1980) extended the Kuznets research pro- gram to cover the postwar development patterns of lower-income coun- tries and also developed a methodology for comparing the sources of industrialization in different economies. A broader approach to the analy- sis of structural change, using a general equilibrium framework, was developed by Dervis, de Melo, and Robinson (1982); this enabled them to compare the effects of different policies. These two lines of analysis have been consolidated in the present volume, which tries to explain the post- war experience of semi-industrial countries. Research for this study was undertaken in two stages, which were designed to take advantage of the expanding data on the productive structure of developing countries. 1 The centerpiece of the first stage was the development of a comparable set of input-output accounts for nine semi-industrial economies in a framework that facilitates economy com- 1. The first stage corresponds to World Bank research project 671-32, "A Comparative Study of Sources of Industrial Growth and Structural Change"; the second to research project 671-79, "The Sources of Growth and Productivity Change." Selected findings of these studies are listed in chapter 1 of this volume and in the World Bank's annual Abstracts of Current Studies for 1978-83. ix X PREFACE parisons over time. This stage led to a series of publications on the methodology and findings for individual economies. The second stage of our research focused on the growth of total factor productivity: the apparent differences between developed and less de- veloped countries, the effects of alternative development strategies, and the variation among sectors. In this context, more detailed studies were carried out for four economies-Japan, the Republic of Korea, Turkey, and Yugoslavia. Computable general equilibrium models, in addition to our standard input-output model, were used to test the relative importance of different factors. The design of both phases of research was carried out in collaboration with the authors of the principal country studies in order to reconcile the theoretically desirable and the empirically feasible approaches. Our main collaborators in the first stage were Bela Balassa, Merih Celasun, Kwang Suk Kim, Shirley W. Y. Kuo, Yuji Kubo, Jeffrey Lewis, Jaime de Melo, Tsunehiko Watanabe, and Larry Westphal. Gershon Feder, Mieko Nishi- mizu, and Shujiro Urata joined us in the second stage. All contributed to the specialized studies that provide the background for this volume. We are also indebted to a number of other colleagues and friends for their comments on successive drafts of the manuscript. We should espe- cially like to thank Jeffrey Williamson, whose many comments were invaluable in producing a more coherent whole. Others include Irma Adelman, Alain de Janvry, Kemal Dervis, Shantayana Devarajan, Frank Lysy, Dwight Perkins, Betty Sadoulet, Lance Taylor, Simon Teitel, and Adrian Wood. We have also benefited greatly from a series of able research assistants: Tim Condon, Hazel Elkington, Kathy Jordan, Murat Koprulu, Maria Kutcher, and Narayana Poduval. For their help in typing many drafts, we thank Trinidad Angeles, Nenita Bencio, Cheryl Cvetic, Teresita Kaman- tigue, Isabelle Kim, and Kim Tran. Our editor, Jeanne Rosen, patiently and expertly took us in hand and greatly improved the final product. Like most comparative studies, this final product is both more and less than might have been hoped for. Initial hypotheses tend to explain less than was expected, yet the first results have led to new and more interest- ing questions. Although our views have converged in the process, we have found it useful to identify the authors of each chapter rather than seek a common interpretation. 1 Introduction THE RELATION between industrialization and economic growth is a subject of continuing controversy. Historically, the rise in the share of manufacturing in output and employment as per capita income increases, and the corresponding decline of agriculture, are among the best documented generalizations about development. But how does this trans- formation of the structure of production affect the rate of growth and the distribution of its benefits? And what has been the effect of policies designed to accelerate this shift or to alter its composition? These and related questions are still in dispute. This book attempts to clarify the role of industrialization in develop- ment by conducting a series of comparative studies of semi-industrial economies. The studies address three main topics: industrialization as a stage in the overall transformation that constitutes modern economic growth; the similarities and differences in the experiences of nine indus- trializing economies; and the relation between rising productivity and structural change. This chapter traces the background of the principal issues to be considered; summarizes the studies' main findings; and gives the reader a guide to the countries chosen, the topics discussed, and the analytical techniques employed. The Issues Many of the issues considered in this volume were debated intensively but inconclusively in the 1950s. Although these debates defined the logic of different positions, they failed to resolve important questions about the role of industrialization or the choice of a development strategy. Only after the experience of the 1960s and the early 1970s was assessed did the empirical questions become more focused and the relevant policy issues emerge. A decade later, we can distinguish more clearly among the prin- cipal approaches to development that have been tried and analyze the role played by industrialization in each. Advantages of Industrialization Early arguments for accelerated industrialization were based largely on the assumed properties of technology in manufacturing and related sec- tors. Authors such as Rosenstein-Rodan (1943, 1961) and Mandelbaum (1945) stressed the importance of economies of scale and growth of productivity in manufacturing and the cumulative benefits that these bring about in the form of external economies. With some additional assump- tions, these factors argue for balanced investment in heavy as well as light 1 2 INTRODUCTION industry and a reduced share of manufactured imports in the gross na- tional product (Mahalanobis 1955). Industrialization was also advocated by Prebisch (1950) and Singer (1950) to offset the supposed disadvantages of specialization in primary production and the associated secular deterioration in the terms of trade. Nurkse (1961), noting the limited world demand for exports of primary products and the rising domestic demand for manufactured goods, pro- posed a policy of balanced growth of the industrial and primary sectors. In modified form, this approach has been a fruitful starting point for subse- quent empirical work on the ways in which the conditions of demand limit the patterns of development. These arguments for early industrialization were criticized by Viner (1952) and Fleming (1955) as conflicting with the neoclassical analysis of comparative advantage and the benefits of specialization. More recently, the longer-term comparative advantage of many developing countries has been seen to lie in those branches of manufacturing that use the growing skills of the labor force. The neoclassical position has thus come to be identified with the export of labor-intensive manufactured goods induced by outward-oriented trade policies. Nurkse's arguments for balanced growth were clarified by Scitovsky (1954, 1959) and Streeten (1959). This discussion weakened the case for limiting trade and stressed the differences in country conditions of demand and supply. Perhaps more important, it showed the limitations of partial equilibrium analysis as a means of resolving such economywide questions. Changing Perceptions Kuznets (1957, 1966) and his followers have put the issues of industrial growth into a broader perspective. Instead of focusing narrowly on the allocation of resources, Kuznets describes the increase in industrial output as part of the general transformation that he identifies as "modern eco- nomic growth." In this context, industrialization is not only a response to changing demand and supply conditions but also a principal means of acquiring modern technology. The style of comparative analysis initiated by Clark (1940) and Kuznets led to a number of empirical generalizations-"stylized facts"-that underlie much subsequent work. Although the earlier attempts to identify theoretical grounds for advocating a particular pattern of resource alloca- tion were inconclusive, Kuznets's empirical findings have stimulated a search for the causes and implications of these stylized facts. Much of this book is an attempt to interpret the interaction of the principal factors that cause the structural changes reflected in Kuznets's cross-country and time-series patterns. The sustained growth of the world economy from the mid-1950s to the mid-1970s led to a more optimistic view of the benefits of trade for INTRODUC'{ION 3 developing countries than had prevailed. As manufactured exports from developing countries grew at more than 10 percent a year, the assumption that export markets were limited became much less tenable. At the same time, comparative studies of the effects of import substitution by Macario (1964), Bruton (1970), and Little, Scitovsky, and Scott (1970) demon- strated that these policies become less and less efficient if they are main- tained for long periods. The argument for shifting from an inward-oriented to an outward- oriented strategy was greatly strengthened by the success of a small group of "newly industrializing economies" and particularly of the four East Asian economies in this group: Hong Kong, the Republic of Korea, Singapore, and Taiwan. 1 These four "superexporters" have followed a new pattern of industrialization characterized by very rapid growth of manufacturing based on increasing participation in the international econ- omy. Their experience has raised questions about the extent to which this pattern is suitable for larger economies, whether it depends on the particu- lar social and political features of East Asian societies, and so forth. In our search for answers to such questions, we shall try to identify the common sources of growth and structural change in these and other industrializing countries and to pinpoint the main differences. 2 Although outward-oriented policies have received great attention in recent years, they have been only one of several ingredients in successful development strategies. Japan, the original model of export-led growth, has been more notable in the postwar3 period for attaining a great increase in productivity than for having a particularly open economy. Among countries with favorable natural resource endowments, Malaysia and Thailand have pursued strategies of delayed industrialization; after a late start, manufacturing has increased rapidly in response to the growth of domestic demand, without the distortions introduced by premature attempts to accelerate industrialization. Even among countries with very autarkic policies in the early 1960s, one finds examples (Brazil, Spain, and Turkey) of relatively successful development based on a shift to more neutral trade policies and the growth of manufactured exports. These relatively large economies resemble the early Nurkse model of balanced growth, in which industrialization arises primarily from the growth of domestic demand; at the same time, export growth must be fast enough to 1. An OECD (Organisation for Economic Co-operation and Development) study, The Impact of the Newly Industrializing Countries on Production and Trade in Manufactures (1979), is the most comprehensive analysis of this group as a whole. The ten economies that it covers are Hong Kong, Korea, Singapore, and Taiwan (East Asia); Greece, Portugal, Spain, and Yugoslavia (Mediterranean); and Brazil and Mexico (Latin America). 2. Ranis (1981) compares Latin American and East Asian trade policies and their con- sequences. 3. Throughout the book, "postwar" and "prewar" refer in general to the times after and before World War II; the exact year varies depending on the economy under discussion. 4 INTRODUCTION avoid serious trade bottlenecks. Although all these cases incorporate shifts to more outward-oriented policies, they do not exhibit the extreme spe- cialization characteristic of the four East Asian superexporters. The adaptability of countries pursuing different strategies has been tested in the past decade by the periodic disruption and slower growth of the world economy. Those economic structures distorted by past efforts to reduce imports have proven vulnerable because of the limited margin for additional import substitution. Furthermore, it has been more difficult and disruptive of growth for inward-oriented countries to shift to export expansion than for more outward-oriented countries to make the smaller structural adjustments needed. The prospect of slower growth of world trade is therefore causing strategies of both export orientation and import substitution to be reassessed. The experience of the past twenty years has been subject to varying interpretations. Balassa (1981), Ranis (1981 ), and Little (1982) see the growth of the East Asian superexporters as the predictable result of freeing trade and removing discrimination against exports; their success points the way for others to follow. But skeptics such as Diaz-Alejandro (1975), Datta-Choudhuri (1981), and Streeten (1982) stress that the strong role of government, favorable human resource endowments, and the expansive international environment of the 1960s are special features that limit the transferability of the superexporters' experience. Despite these differing views, there is considerable common ground among economists debating development strategies. Krueger (1984) points out that the main issue between the proponents of neoclassical trade policy and its critics is not whether to industrialize but what form the industrialization should take. It is also generally agreed that the static gains achieved in shifting to more efficient patterns of trade are not the most important reason for the high growth rates of successful countries. Increased productivity is even more important. Identifying the virtues of outward-oriented trade policies is thus only a starting point for unraveling the causal links between industrialization and economic growth. Questions Addressed The first of the three sets of questions that we shall take up in this book concerns the uniform features of development patterns, especially the tendency to shift from primary production to manufacturing. How essen- tial is industrialization for development? What is the importance of changes in demand in comparison with changes in such supply-side factors as capital accumulation and comparative advantage? Before the answers to these questions can be applied to policy analysis, the causal links among them need to be better understood. The second set of questions concerns the relation between growth and structural change. Although neoclassical theory points to the significance of changes in factor supplies and productivity, studies of developing INTRODUCTION 5 countries show that changes in demand and trade are equally important to continued growth. Industrialization thus can be viewed as a way to satisfy similar patterns of demand growth with varying combinations of factor supplies. In analyzing these issues, we must recognize the importance of initial conditions and of the differences in resource allocation associated with such structural features as the size of an economy and its natural resource endowment. The third set of questions concerns the means by which policies are carried out. Is excessive import substitution inefficient mainly because it raises the cost of traded goods or because it is associated with other policies that lead to balance of payments bottlenecks or reduce productiv- ity growth? This type of question is perhaps the most difficult, but it must be addressed if we are to use the results of comparative studies to improve policy choices. The Research Design Two main approaches have been used to study the issues outlined above. One stems from trade theory and stresses external policies. The other stems from planning models and stresses the internal aspects of resource allocation. In designing the research for the present volume, a central problem was to combine elements of both techniques in a method- ology that also takes into account the data available for the industrializing economies. One technique for comparative country analysis has evolved over the past fifteen years in a series of studies on trade and development policy, beginning with the pioneering work of Little, Scitovsky, and Scott (1970). Succeeding studies have taken up specific topics that relate trade to de- velopment: the levels of effective protection (Balassa 1971), the impact of different trade regimes (Donges 1976; Bhagwati 1978; Krueger 1978; Balassa 1981), and the indirect effects of these regimes on employment (Krueger 1983 ). These authors have held their studies to manageable proportions by focusing on the implications of trade theory and policy without examining other circumstances affecting growth and resource allocation in any detail. A second approach is based on multisectoral models designed to study changing patterns of resource allocation. Originally developed to isolate the implications of alternative policies in individual countries, such models have rarely been applied to comparative studies. The early versions were based on open input-output systems, in which exports were usually treated as exogenous. Subsequent studies have used linear programming or com- putable general equilibrium models, which integrate the internal and external policies affecting resource allocation but are more demanding of data. Although multisectoral models are typically used to analyze the economic structure and policies of a given country, they have occasionally been extended to examine more general questions. Representative exam- 6 INTRODUCTION pies are the analysis of comparative advantage in a general equilibrium context for Israel (Bruno 1966) and for India (Weisskopf 1971), and the analysis of policies affecting income distribution for Korea (Adelman and Robinson 1978). The two techniques are essentially complementary. The various policy comparisons (Little, Scitovsky, and Scott 1970; Krueger 1978, 1983; Bhagwati 1978) rely on classifications of trade policy regimes in selected countries and on statistical descriptions of their effects. Although empiri- cal models are not used explicitly, trade theory provides a unifying analyt- ical framework. Multisectoral analysis, in contrast, is designed to study phenomena revealed only by models that are explicitly based on a social accounting framework. It is particularly suited to the study of industrialization because the indirect effects of structural change often outweigh the direct effects that are visible in the national accounts aggre- gates. This book began as an attempt to generalize from the growing number of multisectoral analyses of developing countries. Such studies now exist for twenty countries and cover periods of up to twenty years. From this group, we have selected eight semi-industrial economies that have had at least average growth rates throughout the postwar period; they form a spectrum of trade and development policies from inward-oriented (Mex- ico and Turkey) to outward-oriented (Korea, prewar Japan, and Taiwan). (The sample comprises a quarter of the semi-industrial group identified in chapter 4. Postwar Japan and Norway were added to illustrate the shift to a mature economy.) To prepare the book, studies of these selected economies were under- taken to analyze growth and structural change in the postwar period with the common accounting framework and methodology described in chap- ters 3 and 5. The authors were encouraged to extend the analyses to problems of particular interest for each economy. The main publications on individual economies that have resulted from these efforts range from full-scale monographs on Korea, Taiwan, and Turkey to articles on Israel, Japan, Norway, and Yugoslavia. (Only the standard accounts used in country comparisons were prepared for Colombia and Mexico.) These publications are: Colombia "Sources of Growth Data for Colombia" (de Melo 1983) Israel "Sources of Growth Data for Israel" (Frankel 1983) "Economic Growth and Structural Change in Israel: An International Perspective" (Syrquin 1984) Japan "Role of Industrialization in Japanese Development" (Chenery and Watanabe 1976) INTRODUCTION 7 Korea "Industrialization and Structural Change in Korea" (Kim 1978) Growth and Structural Transformation (Kim and Roemer 1979) Mexico "Sources of Growth Data for Mexico" (Syrquin 1983) Norway "Accounting for Economic Growth: The Case of Norway" (Balassa 1979a) Taiwan "Economic Growth and Structural Change in the Republic of China" (Kuo 1979) The Taiwan Economy in Transition (Kuo 1983) Turkey Sources of Industrial Growth and Structural Change: The Case of Turkey (Celasun 1983) The Foreign Exchange Gap, Growth and Industrial Strategy in Turkey 1973-1983 (Dervis and Robinson 1978) Yugoslavia Yugoslavia: Self-Management Socialism and the Challenges of De- velopment (Schrenk, Ardalan, and El Tatawy 1979) In addition, uniform data on the sources of growth in the form described in chapter 5 were compiled for all countries by Yuji Kubo and Jeffrey Lewis in collaboration with the authors listed above (see Kubo 1983). A Reader's Guide This book consists of several types of comparative study that are unified by common themes. Nearly half of it is devoted to what can be learned from the nine economies that were studied in some detail. These results are presented in part II and chapter 10. The rest of the book comprises four comparative studies in which we develop the principal hypotheses and test their applicability to a broad spectrum of semi-industrial countries. These cross-country studies are: • A comparison of the results of So low-type estimates of the sources of aggregate growth in thirty-nine economies (chapter 2). • A multisector simulation of the structural transformation of a repre- sentative semi-industrial economy (chapter 3). A dynamic version of this model is developed in chapter 8 to study the effects of resource reallocation on productivity growth. • An econometric model of the sources of growth under disequilibrium conditions, estimated from data for thirty-four semi-industrial econo- mies (chapter 9). 8 INTRODUCTION • A computable general equilibrium (CGE) model of the effects of alternative external policies on transformation and growth, based on data for Korea and other economies in our sample (chapter 11). Table 1-1 shows how these studies are fitted together in the four parts of the book. The first two are combined in part I to provide an overview of the structural transformation and its relation to aggregate growth. Chap- ter 2 justifies our focus on semi-industrial countries by showing the characteristic differences between their growth processes and those of more mature economies. The cross-country model is then used to analyze the common features of the transformation in chapter 3 and to establish a basis for a typology of all semi-industrial countries in chapter 4. The three chapters in part I together provide a background for characterizing the nine economies in our sample. Table 1-1. The Organization of the Book Part of book Topic Analysis' Results' I. Structural Sources of growth: Mature versus Varying role of transformation demand versus supply semi-industrial: productivity 39 economies (2) growth (2) Uniformity of Cross-country Components of transformation model (3) transformation (3) Semi-industrial Country economies (4) typology (4) II. Experience of Causes Methodology of Ourward- versus industrialization comparisons (5) inward-oriented strategies (6) Similarities Multisectoral Industrial models: structure (7) 9 economies (6, 7) III. Productivity Decomposition Dynamic cross- Measurement of and structural of productivity section model (8) reallocation change growth effects (8, 9) Importance of Econometric Country versus reallocation analysis: 34 sector effects (8, effects semi-industrial 10) economies (9) Productivity by sector: 4 economies (10) IV. Development Policy CGE model of Effects of policy strategy instruments representative changes (11) strategies (11) Synthesis Synthesis of supply Role of price and demand changes (11) factors (12) a. Numbers in parentheses refer to chapters. INTRODUCTION 9 Part II analyzes the postwar experience of industrialization as it is reflected in comparable multisectoral models of the nine economies. This approach has the advantage of relating changes in the internal structure of demand and production to changes in the external structure of trade and resource inflows. This, in turn, facilitates detailed comparisons between inward- and outward-oriented strategies in which the relative importance of the several sources of growth and structural change can be measured. Part II also reveals several aspects of industrialization that have not received much attention. Particularly notable is the growing importance of intermediate demand for industrial inputs in all the economies studied. This type of structural change is incorporated into the cross-country model and helps to explain the rapid growth of heavy industry. Part III is concerned with the relation between industrialization and productivity growth. For this purpose, the multisectoral model is extended in chapter 8 to include the use of labor and capital by sector. The principal issues discussed are the differences in productivity growth among coun- tries and sectors and the effects of reallocating resources. Industrializing economies differ from mature economies in that the reallocation of resources from sectors of lower productivity to sectors of higher productivity can make an important contribution to overall growth. Chapters 8 and 9 examine the quantitative significance of several such shifts: from agriculture to industry, from light to heavy manufactur- ing, from domestic markets to foreign ones. In addition to the detailed comparison of four countries Oapan, Korea, Turkey, and Yugoslavia), part III contains an econometric model in which the effects of such shifts are estimated for the semi-industrial economies as a group (chapter 9). The studies of factor use and productivity have led to some revisions in our original hypotheses. Although sectoral differences in productivity growth are significant, differences among countries are equally so. Total factor productivity growth tends to be higher in all sectors in countries of high growth. In the econometric model, the overall effect of these country differences can be measured as country residuals from the predicted rate of growth and can be associated with general economic policies. Part IV develops a general equilibrium framework for the analysis of individual policy instruments. This allows the outward-oriented package to be broken down into several components, including the effects of capital inflows as well as increased exports. Efficient and inefficient methods of achieving import substitution are also simulated. In this way, we can compare a spectrum of policy combinations ranging from a styl- ized version of the superexporters based on the experience of Korea and Taiwan to representative combinations of protection and lower capital inflows based on the experience of Mexico and Turkey (chapter 11). These results, in turn, make possible a synthesis of the demand and supply sides of growth accounting. PART I Structural Transfortnation THE POTENTIAL of the concept of structural transformation to unify the study of various aspects of development is demonstrated by the use of such terms as agricultural transformation, industrialization, demographic tran- sition, and urbanization, each of which describes one or more dimensions of the overall transformation process. Part I incorporates the transforma- tion of demand, trade, production, and employment into a single framework for analyzing long-term growth phenomena. Our analysis focuses on the interaction between growth and structural change. What are the main features of the transformation that affect the ways in which economies grow and that distinguish developed from developing countries? Two of the best known of these structural relations are: • Engel's law of the declining share of food in consumption • Lewis's hypothesis of the elastic supply of labor in most developing countries. Our findings suggest several other relations of comparable importance: • Balassa's "stages of comparative advantage," derived from the Heck- scher-Ohlin model • Kuznets's observation of systematic differences in the level and growth of labor productivity by sector • The demographic transition, a set of factors that produces first a rise and then a decline in population growth as per capita income rises. Taken together, these income-related structural changes imply that the growth processes of developing economies may differ substantially from those of advanced economies. Chapter 2 explores this hypothesis first by extending the techniques of growth accounting. The Solow methodology (Solow 1957) is applied to a 11 12 STRUCTURAL TRANSFORMATION large sample of developing and developed countries; this reveals some characteristic differences between the two groups in their sources of growth. These sources are then disaggregated to investigate the interaction between changes in the composition of demand and trade on the one hand and factor supplies and productivity growth on the other. Chapter 3 attempts to model the factors that produce the typical pat- terns of the structural transformation. The model incorporates stylized facts of the kind indicated above and is estimated from cross-country data. It allows for differences in factors (such as size and resource base) that affect the structural transformation, as well as for differences between inward- and outward-oriented development policies. Several representa- tive patterns of resource allocation are simulated by increasing the level of income over the range of the transformation. The main purpose of these cross-country simulations is to provide a standard of comparison for the studies of individual economies in part II. They also suggest several generalizations about the uniformity of the transformation itself. For example, industrialization (measured by a rise in the share of manufacturing in gross national product [GNP]) must occur unless there is a sufficient rise in the value of exports of primary products or services to outweigh both the Engel effects of income growth on demand and the intensified use of industrial inputs. Although the phe- nomenon of deindustrialization (or the Dutch disease) has been observed in the short run, there are few examples of it persisting in developing countries for more than a decade. The longer-term analysis of chapter 3 indicates that the main issue of transformation is not whether countries need to industrialize but when and in what manner. Chapter 4 addresses the problem of using these results for policy analy- sis. It proposes a general typology of industrialization based on the main distinctions that have emerged in simulating patterns of resource alloca- tion. A category of industrializing or semi-industrial economies, in- termediate between the less developed and the developed economies, emerges from these simulations. Its characteristics are then used to identify all the semi-industrial economies in the period 1960-80. Finally, distinc- tions in size, resource endowment, and trade policies are shown to be important in comparing the effects of development policies. 2 Growth and Transformation HOLLIS CHENERY THERE ARE TWO contrasting views of the way economic growth occurs. In the neoclassical tradition, GNP rises as the result of the long-term effects of capital formation, labor force expansion, and technological change, which are assumed to take place under conditions of competitive equilib- rium. Shifts in demand and the movement of resources from one sector to another are considered relatively unimportant because labor and capital produce equal marginal returns in all uses. In the second, broader view, economic growth is regarded as one aspect of the transformation of the structure of production that is required to meet changing demands and to make more productive use of technology. Given imperfect foresight and limits to factor mobility, structural changes are most likely to occur under conditions of disequilibrium; this is particu- larly true in factor markets. Thus a shift of labor and capital from less productive to more productive sectors can accelerate growth. Although this type of structural analysis has not received the same rigorous formula- tion as general equilibrium theory, it can provide a basis for empirical analysis. When general equilibrium is not treated as axiomatic, the question of how much the reallocation of resources to sectors of higher productivity contributes to growth becomes an empirical one. It is likely to be more .,; important for developing countries than for developed ones to recognize the potential of reallocation, for developing countries show more pro- nounced symptoms of disequilibrium in factor markets as well as more rapid change in the structure of production. This chapter sets forth the background for analyzing the relation be- tween industrialization-or, more broadly, the structural transformation of an economy-and the growth of per capita income. The results of neoclassical studies of equilibrium growth in both developed and develop- ing countries are summarized, after which the effects of changing patterns of demand and trade in promoting or limiting the shift of resources to more productive uses are considered. As the chapter will show, determin- ing the sources of growth calls for a synthesis of demand and supply factors and the use of multisectoral models. Some fundamental differences between the growth processes of de- veloping or transitional economies and those of mature, industrial econo- mies emerge from this survey. In particular, disequilibrium phenomena are shown to be more significant for the former than for the latter. Thus, 13 14 STRUCTURAL TRANSFORMATION although neoclassical theory is a useful starting point for the study of growth, it must be modified substantially if it is to explain the essential features of economies in the process of transformation. The Sources of Growth Measurement of the sources of economic growth has progressed greatly since the pioneering work of Abramovitz (1956), Solow (1957), and Denison (1962). The main objective has been to estimate the relative contributions of the growth of capital and labor inputs (corrected for quality changes) on the one hand and of total factor productivity on the other. There are now many studies of the industrial countries, covering much of the postwar period, that use variants of neoclassical theory. This methodology has also been applied to a growing number of semi- industrial countries, and therefore some of the differences in the growth processes of the two groups can be identified. Because the study of the disequilibrium aspects of growth requires a more detailed model than that of equilibrium growth, the principal econ- ometric efforts have tested the significance of these aspects in explaining differences in growth among countries. This work, which is reviewed below, has established the importance of moving resources from lower- productivity to higher-productivity uses-for example, by expanding ex- ports or by turning from agriculture to industry. These shifts are more important sources of growth in developing than in developed countries. This empirical work suggests some answers to several questions of concern in this book: • How useful is the neoclassical methodology as applied to developing countries? Are there significant differences among groups of countries that should be taken into account? • Which departures from the general equilibrium framework appear to be most significant? To what extent is more explicit analysis of the changing composition of demand and trade needed? • Are there systematic variations with per capita income in the factors affecting growth that should be allowed for? In considering these questions, it may be useful to contrast the assump- tions underlying neoclassical and structural views of the sources of growth. Since the basic assumptions of neoclassical theory are well known, they can serve as a point of departure in explaining the hypotheses of the structural approach. The most important distinction between the two views is between their systemic assumptions rather than between any one of their elements. Neoclassical theory assumes the efficient allocation of resources (Pareto optimality) over time from the point of view of both producers and consumers. At any given moment it is impossible to increase aggregate output by shifting labor and capital from one sector to another: realloca- GROWTH AND TRANSFORMATION 15 tion takes place only as the economy expands. In contrast, the structural approach does not assume fully optimal resource allocation; conse- quently, there may be systematic variations in the returns to labor and capital in different uses. Some of the assumptions that contribute to this basic distinction are outlined in table 2-1. Maintaining equilibrium in the face of shifts in internal demands and in external trade is helped by high elasticities of substitution among both commodities and factors and by rapid responses to market signals. Neoclassical theory assumes that the economic system has sufficient flexibility to maintain equilibrium prices, whereas the structural approach identifies some conditions that make complete adjust- ment unlikely. One of the best documented sources of disequilibrium is the duality of the labor market-a duality which has been accentuated in many developing countries by a population growing too rapidly to be absorbed in the high-productivity sectors of the economy. The result is an elastic supply of unskilled labor concentrated in the agricultural and service sectors. A second widely studied source of disequilibrium is the failure to reallocate resources efficiently to increase exports or replace imports. The factors contributing to a chronic balance of payments deficit include the tendency for import demands to expand more rapidly than total GNP, the lack of incentives for producers to enter new markets, and shortsighted Table 2-1. Alternative Views of Growth Neoclassical approach Structural approach Assumptions Factor returns equal marginal productivity Income-related changes in internal demand in all uses No economies of scale Constrained external markets and lags in adjustment Perfect foresight and continuous equilib- Transformation of productive structure rium in all markets producing disequilibria in factor markets Empirical implications Relatively high elasticities of substitution Low price elasticities and lags in adjust- in demand and trade ment Limited need for sector disaggregation Segmented factor markets Lags in adopting new technology Sources of growth Capital accumulation Neoclassical sources plus: Increase in labor quantity and quality Reallocation of resources to higher- productivity sectors Increase in intermediate inputs Total factor productivity growth within Economies of scale and learning by sectors doing Reduction of internal and external bot- tlenecks 16 STRUCTURAL TRANSFORMATION policies that favor import substitution over export expansion. Whatever the factors limiting balance of payments adjustment in the past, there is little doubt that these factors have been a source of disequilibrium in many developing countries and have impeded growth. Although the level of income in a hypothetical neoclassical economy is by definition higher than it would be under any set of disequilibrium assumptions, the growth potential of this economy may be less over time. Disequilibrium phenomena such as segmented factor markets and lags in adjustment imply a potential for accelerating growth by reducing bot- tlenecks and reallocating resources to sectors of higher productivity. This potential is likely to be greater in developing countries-which are subject to greater disequilibrating shocks and have greater market disequilib- rium-than in developed countries. In addition, developing countries can take advantage of the more productive technology available from ad- vanced countries. These two factors offer a plausible explanation for the acceleration of growth that has been noted in many industrializing coun- tries. In summary, the structural approach focuses on differences among sectors of the economy that may inhibit the equilibrating adjustments in resource allocation implied by neoclassical theory. Disequilibrium is more often manifested by the differences in returns to labor and capital in different uses than by the shortages and surpluses that indicate the com- plete failure of markets to clear. In contrast, neoclassical theory assumes that equilibrium is maintained over time, which limits the sources of growth to factors on the supply side. Equilibrium Growth The assumptions of competitive equilibrium that underlie neoclassical theory are a convenient starting point for growth analysis because they permit any group of inputs to be aggregated on the basis of their marginal productivities. 1 For economywide studies, all primary inputs can be categorized as either capital or labor. Each of these can then be consoli- dated on the basis of its share in the total product. The difference between the growth of total output and the weighted average growth of capital and labor serves as a measure of the increase in total factor productivity for the economy as a whole. This procedure is sufficiently general to permit comparisons among studies using different methodologies, so long as they maintain the assumptions of competitive equilibrium. Analyses designed to measure the importance of these three sources of growth have now been carried out for thirty-nine economies for several periods. They indicate that the growth of capital, labor, and productivity are of comparable importance for the sample as a whole but vary signifi- 1. This section also provides a background for the country studies in chapter 10 and was written in consultation with Mieko Nishimizu. GROWTH AND TRANSFORMATION 17 candy with the structure of an economy and the effectiveness of its policies. Methodology The methodology commonly used to estimate the sources of growth in a neoclassical framework has evolved from Solow's basic formulation (1957). An aggregate production function of the following general form is assumed: 2 (2-1) Q = F (K, L, t) where Q is the aggregate output of the economy, K and L are aggregate capital and labor inputs, and tis time. The simplest assumption about the effects of time (and the one made in most of the studies reported in table 2-2 below) is that technical progress is neutral in the (Hicksian) sense that it raises the output achievable from a given combination of capital and labor without affecting their relative marginal products. On this assump- tion, the production function can be written as (2-1a) The three sources of output growth can then be derived by differentiat- ing this equation with respect to time and dividing by Q: Q = A +A aF k. +A aF L Q A aK Q aL Q' where dots indicate time derivatives. Substituting 13K = (a Q/ aK)(KI Q) and 13L = (aQ!aL)(L/Q) gives the basic neoclassical growth equation:2 (2-2) where Gv, GK, and GL are the growth rates of aggregate output (value added), capital, and labor. The growth of total factor productivity (TFP), G A' is defined as the difference between Gv and the weighted sum of input growth, 13KGK + 13 L G v Each input coefficient, 13;, is defined as the elasticity of output with respect to input i, thus indicating the effect on output growth of a 1 percent increase in the growth of that input. This formulation can be extended to any number of inputs. Under conditions of competitive equilibrium, each factor receives its marginal product, so that the real wage, wlp, equals aQ I aL. This assump- tion leads to the important result that in equilibrium the coefficient 13L is also equal to the share of labor in the total product, wLipQ. Similarly, 13K = rK/pQ. In the absence of economies of scale, the sum of all the share 2. This formulation follows Solow (1957). The conditions underlying it and alternative assumptions are discussed in Branson (1979) and Nadiri (1970). A more general approach using a flexible form of the production function is discussed in chapter 10. 18 STRUCTURAL TRANSFORMATION coefficients is equal to unity. Because output elasticities can rarely be estimated directly, these product shares based on equilibrium assumptions are normally used to estimate the growth equation 2-2. Starting with the work of Denison (1962) and Griliches and Jorgenson (1966), major improvements in empirical analysis of the sources of growth have come from subdividing capital and labor by type and weighting them by their imputed returns. This can be done either for different types of capital assets or for the labor characteristics that affect productivity (education, age, sex, and so on). The formulation adopted by Elias (1978) in studies of many of the developing economies in our sample is repre- sentative: (2-3) Gv = 13LGL + 13L~ w· ~ { (f) + 13KGK + 13k~~(~) + GA where the effects of varying returns on different kinds of capital assets and on different qualities of labor are reflected in the terms w;l w and r/ r. Although this method of treating labor quality differences is generally accepted, the extent to which it should be followed in measuring the growth of capital is a matter of dispute. 3 For the group of developed countries listed in table 2-2 below, the effect of the quality terms in equation 2-3 is to add 25 percent to the growth of the capital stock and to offset the decline in the average hours worked. Compared to studies that do not include quality changes, this method increases the contribution of factor inputs and reduces measured factor productivity growth by a corresponding amount. For the developing countries, Elias (1978) attributes a smaller proportion of input growth to quality improvements; in some studies, this element is omitted entirely. Despite these conceptual differences, it will be shown that the Solow- Denison framework is a valuable basis for comparing the growth pro- cesses of developing and developed countries. Developing versus Developed Countries Similarities in the growth processes of developed countries have been explored in several recent studies. Earlier results tended to stress the relatively small proportion of growth accounted for by the increase in capital and labor, which leaves a large unexplained residual. Fuller allow- ance for quality improvements in the estimates of Christensen, Cummings, 3. Equation 2-3 is a simplified version of the methodology used by Christensen, Cum- mings, and Jorgenson (1980), which stems from earlier studies by Griliches and Jorgenson (1966, 1967). GROWTH AND TRANSFORMATION 19 and Jorgenson (1980), which are used here, has reduced the residual for the developed countries to an average of about half of total growth. 4 Few efforts have been made to compare the performance of any sub- stantial group of developing countries with that of industrial countries. In a pioneer attempt, Bruton (1967) analyzed data for five Latin American countries and concluded that total factor productivity growth (the re- sidual) was much lower in this group than in developed countries. Nadiri (1972) reached a similar conclusion for the period 1950-62. This test will be repeated for the larger sample used here. My analysis tries to identify the main differences in the sources of growth between developed and developing countries by focusing on the role of factor inputs compared with productivity growth. Since many studies provide only rough estimates of the shares of capital and labor in factor inputs, the analysis will consolidate the two into a single factor input, F. The growth equation 2-2 is then simplified to (2-2a) where G v = growth of value added, G F = (3G K + (1 - !3) G L equals the combined contribution of factor inputs to growth, and G A = growth of TFP. The average values of these variables for a sample of thirty-nine economies are reported in table 2-2 along with the underlying estimates for capital and labor shares. These thirty-nine economies include virtually all the cases for which roughly comparable studies have been published. The sample contains one or two observations for each economy (depending on the availability of studies for different periods) and is divided into three groups based on conventional distinctions between developed, developing, and centrally planned economies. All twenty developing economies fall into the semi- industrial category (defined in chapter 4). 5 Some preliminary distinctions among the three groups can be drawn from the average values of the variables given in table 2-2. The developed economies are characterized by little growth of labor inputs (1.1 percent), moderate growth of capital (5.2 percent) and output (5.4 percent), and a relatively large contribution of TFP to aggregate growth (50 percent). The developing economies, in contrast, have high growth of labor inputs (3.3 percent), a higher total factor growth (4.3 percent), and a relatively small contribution of TFP to aggregate growth (30 percent). The centrally planned economies are in most respects closer to the 4. Using the Denison methodology, Kendrick (1982) estimates that increased total factor productivity accounted for about two-thirds of the growth in GNP of nine industrial countries for the period 1960-73, compared with the 50 percent estimated by Christensen, Cummings, and Jorgenson for this period. 5. With respect to income level, Ireland, Israel, and Spain are at the upper end of the semi-industrial category; Ecuador, Honduras, and India are at the lower end. Table 2-2. The Growth of Output, Inputs, and Total Factor Productivity (percent) Total Growth TFP factor input Growth Growth of of of Capital Labor value Growth Growth capital labor income income added rate rate input input share share Economy Years (Gv) (GA) Share (GF) Share (GK) (Gd (~K) (~d Source Developed Belgium 1949-59 2.95 2.05 69.5 0.90 30.5 2.55 0.25 30.0 70.0 ECEb Canada 1947-60 5.20 3.50 32.5 1.70 67.6 6.80 1.10 42.0 58.0 CCJ' 1960-73 5.10 1.80 35.3 3.30 64.7 4.90 2.00 44.9 55.1 CCJ Denmark 1950-62 3.51 1.64 46.7 1.87 53.3 3.84 1.21 25.0 75.0 Dd N 1950-60 4.90 2.90 2.00 40.4 4.70 a France 59.5 0.30 38.2 61.8 CCJ 1960-73 5.90 3.00 50.8 2.90 49.2 6.30 0.40 41.7 58.3 CCJ Germany, Fed. Rep. 1950-60 8.20 3.60 56.8 4.70 43.0 6.90 1.60 36.7 63.3 CCJ 1960-73 5.40 3.00 55.6 2.40 44.4 7.00 -0.70 40.1 59.9 CCJ Italy 1952-60 6.00 3.80 62.7 2.30 37.5 3.30 1.60 40.5 59.5 CCJ 1960-73 4.80 3.10 64.6 1.60 35.4 5.40 -0.70 38.3 61.7 CCJ Japan 1960-73 10.90 4.50 41.3 6.40 58.7 11.50 2.70 41.5 58.5 CCJ Netherlands 1951-60 5.00 2.30 46.5 2.70 53.6 4.00 1.40 47.0 53.0 CCJ 1960-73 5.60 2.60 46.4 3.00 53.6 6.60 0.30 42.9 57.1 Norway 1953-65 5.40 2.88 53.3 2.52 46.7 5.10 0.80 40.0 60.0 BB' Sweden 1949-59 3.40 2.50 73.5 0.90 26.5 2.00 0.50 30.0 70.0 ECE United Kingdom 1949-59 2.50 1.20 48.0 1.30 52.0 3.10 0.60 30.0 70.0 ECE 1960-73 3.80 2.10 55.3 1.70 44.7 4.60 0.00 38.7 61.3 CCJ United States 1947-60 3.70 1.40 37.5 2.30 62.9 4.00 1.40 39.3 60.7 CCJ 1960-73 4.30 1.30 30.2 3.00 69.8 4.00 2.20 41.4 58.6 CCJ Average 5.40 2.70 49.0 2.70 51.0 5.20 1.10 38.5 61.5 Developing Argentina 1950-60 3.30 1.05 31.8 2.25 68.2 2.65 1.10 - - E' 1960-74 4.10 0.70 17.1 3.30 82.9 3.80 2.20 - - E Brazil 1950-60 6.80 3.65 53.7 3.15 46.3 3.10 2.80 - - E 1960-74 7.30 1.60 21.9 5.70 78.1 7.50 3.30 - E Chile 1950-60 3.50 0.85 24.3 2.65 75.7 2.60 2.50 - - E 1960-74 4.40 1.20 27.3 3.20 72.7 4.20 1.90 - E Colombia 1950-60 4.60 0.95 20.7 3.65 79.3 4.25 2.75 - E 1960-74 5.60 2.10 37.5 3.50 62.5 3.90 2.80 - - E Ecuador 1950-62 4.72 2.18 46.2 2.54 53.8 2.82 3.41 38.0 62.0 Co• Greece 1951-65 6.90 2.39 34.5 4.52 65.5 7.10 2.80 40.0 60.0 BB Honduras 1930-62 4.52 1.40 31.0 3.12 69.0 3.65 2.93 26.0 74.0 Co Hong Kong 1955-60 8.25 2.40 29.1 5.85 70.9 4.68 6.63 40.0 60.0 Chh 1960-70 9.10 4.28 47.0 4.82 53.0 7.60 2.97 40.0 60.0 Ch India' 1959/60 N -78/79 6.24 -0.18 -2.9 6.42 102.9 4.77 1.65 52.5 47.5 Ah' Ireland 1953-65 4.70 2.00 42.6 2.70 57.4 4.20 1.70 40.0 60.0 BB Israel 1952-58 9.80 3.90 39.8 5.90 60.2 11.80 3.20 30.0 70.0 Au' 1960-65 11.00 3.40 30.9 7.60 69.1 13.10 5.00 30.0 70.0 Gk Korea, Rep. 1955-60 4.22 2.00 47.4 2.22 52.6 2.18 2.25 40.0 60.0 Ch 1960-73 9.70 4.10 42.3 5.50 57.7 6.60 5.00 36.7 63.3 CCJ Mexico 1950-60 5.65 1.60 28.3 4.05 71.7 5.20 2.65 - - E 1960-74 5.60 2.10 37.5 3.50 62.5 3.90 2.80 - - E Peru 1950-60 4.50 -0.70 -15.6 5.20 115.6 7.65 2.70 - - E 1960-70 5.30 1.50 28.3 3.90 71.7 4.40 2.70 - - E Philippines 1947-65 5.75 2.50 43.5 3.25 56.5 - - - - L' Singapore 1972-80 8.00 -0.009 -0.1 8.01 100.1 9.48 5.52 61.1 38.9 Tm Spain 1959-65 11.20 5.02 44.8 6.18 55.2 8.70 4.50 40.0 60.0 BB Taiwan 1955-60 5.24 3.12 59.5 2.12 40.5 2.68 1.75 40.0 60.0 Ch (Table continues on the following page.) 24 STRUCTURAL TRANSFORMATION Figure 2-1. Relationships between Value Added Growth and Factor Input Growth 12 11 -.::;- u c:: Q) 10 ..." .... Q) All developing E:- 9 countries ...c:: .... ~ 8 0 .... Oil 7 -o Q) -o 6 -o 4 3 2 1 0 1 2 3 4 5 6 7 8 Factor input growth (percent) Key: • Developed economies o Centrally planned economies .._ Developing economies o Hong Kong, Israel, Korea, Spain, and Taiwan regression lines of equation 2-5. Figure 2-2, which shows the relation between factor inputs and productivity growth in each economy in a form corresponding to equation 2-2a, plots lines of constant aggregate growth as the sum of GA and GF. The second graph compares factor inputs and TFP as sources of growth in the three country groups, whereas figure 2-1 focuses on the relation between input and output/ Figure 2-2 shows that most of the developed countries fit within a small cluster, A, defined by relatively low factor growth, with TFP accounting for between 50 and 70 percent of overall growth. Japan is the chief exception; it has not only double the average growth rate for a developed country but also a higher proportion resulting from factor inputs. 8 (The United States 7. Figure 2-2 omits the centrally planned economies and India because their data are limited to manufacturing. All the observations from table 2-2 appear in figure 2-1, but only the most recent observation for each country appears in figure 2-2. In all but a few cases (Brazil and Korea), the most recent observation gives a good indication of the country's sources of growth for the whole period. 8. Japan fits the production function estimated for this group, but it is so far out of the range of the other observations that this has little significance. GROWTH AND TRANSFORMATION 25 Figure 2-2. Relationship between Total Factor Productivity Growth and Total Factor Input Growth 8 ··." "~v=8 Gv = 10 Gv = 12 7 ··." ''"·"·' "'· ~. ·~. ' 6 Gv = 6 '"·"·., ~ c ··~ , Spain "· '~ong~ong~~ •)a~ (1) 5 0 u .... (1) .E: ...c .... " :an~ 0 , ~ 4 " Gv = 4 . . ·.. Germany " "'·· Ko or ~@ Israel 0 Jf~rway ~ .... ell p.. 0 Italy ... ~ 3 " • •. \ ranee , /Nether! an ·. '" Sweden Ot:tited @ .x /Philippines , " Belgiu~ Kin~omEcuado~ • • - - Gfes,ce T k 2 • • •a Colombia "" ur ey Denmark• reland •Cana Mexico. · , Hd~ 4 on ura . , ·~ • Peru ® B Brazil '"'"' 1 United Stat~hile · Argentina • .. Venezuela 2 3 4 5 6 7 8 Total factor input growth (percent) and Canada have a somewhat higher share of labor and a somewhat lower share of TFP than the European countries.) The developing countries in figure 2-2 divide into two clusters. The larger one, B, is characterized by TFP growth between 0.5 and 2.0 percent. The smaller one, C, is composed of five developing economies plus Japan, with aggregate growth, Gv, averaging more than 10 percent. This per- formance was achieved by both higher factor inputs and higher factor productivity than the typical developing countries. The origins of the differences between clusters Band Care explored in the next section of this chapter, which tests the effects of resource reallocation as a source of growth. To pursue this analysis a step further, the nine observations for the five developing economies in cluster C-Hong Kong, Israel, Korea, Spain, and Taiwan-were treated as a separate group in the regression analysis (indicated by a dummy variable for each economy); the results are given in 26 STRUCTURAL TRANSFORMATION Table 2-3. Cross-Country Estimates of Growth and Productivity (percent) Production Productivity function and growth (Gv = et 0 + etpGp; (GA = f3o + f3vGv; equation 2-5) equation 2-6) Economy group eta etp R2 N f3o f3v Developed (cluster A) 1.71 1.35 0.83 19 0.63 0.39 0.66 (0.41)b (0.15) (0.36) (0.07) All developing 1.76 1.07 0.59 31 -0.78 0.45 0.48 (0.76) (0.17) (0.57) (0.09) Developing (cluster B) 2.74 0.68 22 0.28 0.21 0.61 (0.61) (0.14) (0.87) (0.15) Developing (cluster C)' 2.49 1.23 9 1.07 0.30 (1.13) (0.23) (1.48) (0.20) Centrally planned 0.98 1.38 0.77 7 -0.53 0.42 0.68 (manufacturing only) (1.79) (0.30) (1.18) (0.13) All economies 1.89 1.12 0.73 57 0.16 0.36 0.45 (0.40) (0.09) (0.36) (0.05) a. Values computed by using a dummy variable for cluster C economies. b. Standard errors in parentheses. Source: Table 2-2. table 2-3. The estimates for C resemble those for the developed economies, as shown in figure 2-1, whereas the estimates forB imply that an increase in factor inputs has a less than proportional effect on growth. These estimates also give some support to the Kaldor (1967)-Verdoorn (1949) hypothesis that TFP is a function of overall growth, however achieved. 9 The corresponding regression equation is (2-6) The results of estimating this equation for the present sample are also given in table 2-3. Although they show a significant association between productivity increase and overall growth in each group, this does not tell much about the causal factors involved. (The validity of this hypothesis is examined on a sectoral basis in chapter 10.) Both sets of regressions point to the relative inefficiency of the growth processes of the typical developing economies in cluster B. In terms of the production functions of figure 2-1, a representative developing economy with relatively high factor input growth of 5 percent can expect aggregate growth of about 6 percent, whereas the regressions for either developed economies (cluster A) or efficient developing economies (cluster C) predict 9. The Kaldor-Verdoorn hypothesis was actually formulated in terms of labor productiv- ity growth. GROWTH AND TRANSFORMATION 27 growth of about 9 percent. Later chapters will compare in detail four members of cluster C-Israel, Japan, Korea, and Taiwan-and three members of B-Colombia, Mexico, and Turkey-to identify some of the sources of this difference. Disequilibrium Growth Disequilibrium growth is characteristic of an economic system that exhibits significant departures from the neoclassical assumptions de- scribed above. It falls into the category of the "theory of second best," where for various reasons the optimal (equilibrium) solution is unattain- able. In empirical terms, the main questions are whether one can identify the effects of disequilibrium in factor or product markets and incorporate them into an analysis of growth and development. The growth characteristics of an inflexible economy are in general the opposite of those of neoclassical theory, which implicitly assumes a high degree of substitution among both commodities and factors of produc- tion. For example, a fixed coefficient model will almost automatically generate capital shortages and labor surpluses in a developing economy that has a relatively high growth of the labor force; yet this problem is virtually ruled out by neoclassical assumptions. Similarly, the growth of income produces a more than proportional rise in the demand for manu- factured goods and a resulting tendency for manufactured imports to outrun exports. The structural adjustments through export expansion, import substitution, and capital inflows needed to maintain a balance of payments equilibrium will reduce growth unless they are carried out efficiently. (This problem, which has been studied in the literature on two-gap models, is taken up in detail in part II.) In each area of potential disequilibrium, the actual performance of developing countries lies somewhere between the extremes of flexibility and inflexibility assumed by the neoclassical and the input-output systems respectively. The next section tests the importance of disequilibrium fac- tors by comparing the results of statistical studies that incorporate them with those that do not. The Effects of Disequilibrium To incorporate the effects of disequilibrium into the study of growth, I shall attempt to establish which of the several factors suggested are of general significance for developing countries. Those shown to be impor- tant can then be studied in more detail for individual countries. Several economists have done regression analyses of large samples of countries with the common objective of testing the significance of structur- al variables in explaining growth rates. (This approach was first applied in Hagen and Havrylyshyn 1969; Robinson 1969, 1971; and Chenery, Elkington, and Sims 1970. Data were for different periods between 1950 and 1965.) The main factors tested were: 28 STRUCTURAL TRANSFORMATION Neoclassical variables Growth of capital stock Growth of the labor force (or population) Improvements in the quality of labor (or a rise in the level of education) Structural variables Reallocation of labor and capital Growth of exports Capital inflow (two-gap hypothesis) Level of development Each study started from a version of the neoclassical growth formula given in equation (2-2) and added other explanatory variables. Because analysts were forced to use proxies for the underlying factors, the regres- sion coefficients in these equations cannot necessarily be identified with parameters in a specific model. The cross-country regression equations are of the general form (2-7) Gy= a0 + a 1 (~) + a2 GL + a 3 X 3 + a4 XA where II¥= the ratio of investment to GNP (a proxy for the growth of capital stock) GL = growth of the labor force X 3 =a measure of increase in labor quality (or education) 10 XA = a measure of the shift of labor or capital out of agriculture XE =a measure of the growth of exports XF = a measure of the balance of payments deficit X 0 =a measure of the level of development. The use of only the first two explanatory variables yields results that can be compared with the time-series estimates of neoclassical growth for indi- vidual countries. In making this comparison, it should be noted that the coefficient for the investment term, a 1 , can be identified with the marginal productivity of capital only to the extent that capital-output ratios are the 11 ,J same in all countries. Because this form of estimation does not rely on the assumption that factors are paid their marginal products, it gives some indication of the effects of this assumption. The results of four studies covering periods before 1965 are given in table 2-4. Each study illustrates the effects of adding one or more struc- tural variables to a model that includes only the growth of capital and 10. This variable was omitted from the summary of results in table 2-3 because it was not found to be significant in the studies surveyed. 11. To estimate the aggregate production function, the variable should be 1/K-which is not observable-instead of 1/Y, as discussed by Hagen and Havrylyshyn (1969). GROWTH AND TRANSFORMATION 29 Table 2-4. Growth Regressions, 1950-73 Regres- Sam- Explanatory variables sian pie equation size IIY L A E F In y DEVELOPING ECONOMIES; EARLY PERIODS 1958-66 (Robinson 1971) 1 39 0.17* * 0.55 0.22 2 39 0.15** 0.37 1.70*. 0.35 3 39 0.09 0.35 2.01. * 0.17** 0.52 1955-60 (Hagen and Havrylyshyn 1969) 4 33 0.21 *. 0.40 0.07 0.29 5 33 0.16 0.34 0.25* 0.03 0.11 0.54 1950-59 (Chenery, Elkington, and Sims 1970) 6 31 0.15* 0.16* 0.22 7 31 0.18** - 0.02** 0.36 Sa 31 0.14* * 0.59*. 0.75* * 0.21 * * 0.09* -0.01** 0.77 8b 19' 0.08 1.09** 0.37 0.84 1955-63 (Feder; chapter 9) 9 31b 0.14** 0.74** 0.26 10 27b 0.05 0.88** 0.82 ** 0.19 11 31b 0.09 0.57* 0.37 0.30 SEMI-INDUSTRIAL ECONOMIES, 1964-73 (Feder; chapter 9) 12 30 0.25* * 0.78* 0.46 13 30 0.14** 0.43* 0.80** 0.67 14 34 0.10** 0.59** 0.30** 0.79 15 32 0.11* 0.74** 0.90* * 0.23* * 0.81 Empty cells: Not applicable. *Significant at 10 percent with two-tailed test. **Significant at 5 percent with two-tailed test. Note: Dependent variable: annual growth of GNP. a. Developed economies. b. Semi-industrial economies. Sources: Equation 2-7; tables 9-1, 9-4, and 9-6. labor. The studies cover samples of thirty to forty developing countries for periods of five to ten years. 12 In almost all cases the addition of structural variables (as in regression equations 3, 5, and 8a in table 2-4) substantially improved the explanation of differences in growth rates among developing countries. The propor- tion of the variance explained by each equation typically increases from about 0.25 to more than 0.50. Feder's results for the period 1964-73 (table 2-4; also see chapter 9) show even greater improvement when the sample is limited to the semi-industrial economies. Because each explana- tory variable is correlated to some extent with per capita income, the 12. Hagen and Havrylyshyn (1969) and Chenery, Elkington, and Sims (1970) made separate estimates for the 1950s and for 1960-65. In almost all cases the fits for 1960-65 were worse, so I rely mainly on the studies for longer periods. Chenery, Elkington, and Sims also present comparable results for developed countries that will be discussed below. 30 STRUCTURAL TRANSFORMATION Table 2-5. Sources of Growth in Developing Economies, Two Studies, 1958-73 1958-66 (Robinson 1971) 1964-73 (Feder; chapter 9) Sample Sample Source mean 2 3 mean 12 13 14 15 Investment (1/Y) 0.168 2.90 2.56 1.56 0.201 2.80 2.14 2.20 4.97 (59)' (51) (31) (44) (33) (34) (78) Labor (L!L) 2.74 1.49 1.00 0.95 2.07 0.89 1.39 1.72 1.62 (30) (20) (19) (14) (22) (27) (25) Reallocation 0.77 0.90 2.00 0.50 (16) (18) (31) (8) Exports 0.70 1.85 1.96 (14) (29) (31) Residual 0.56 0.62 0.84 -0.18 0.72 1.04 0,01 (11) (13) (17) (-3) (11) (16) (0) Total growth 4.95 4.95 4.95 6.41 6.41 6.43 6.39 Empty cells: Not applicable. a. Figures in parentheses are percentages of total growth. Sources: Table 2-4; Robinson (1971, table 6); and tables 9-2 and 9-7 . .; values of most regression coefficients tend to decline as additional vari- ables are included. This is illustrated by the coefficients for labor and capital in each of the studies cited. Although almost all the structural variables that were tested make a statistically significant contribution to this explanation, little can be said about the underlying causal relations. When the mean values of each explanatory variable are inserted into the regression equations, the equa- tions give a breakdown of the implied sources of growth for the given period. To show the relative importance of each factor, the main findings of Robinson and Feder are compared in this way in table 2-5. 13 Five conclusions emerge: • The growth of capital is still the most important single factor, but its relative contribution is reduced from well over 50 percent of average growth in the neoclassical model to 30-40 percent in the structural formulations. • The growth of the labor force is similarly reduced in importance; in some developing-country samples, it is no longer statistically signifi- cant. These findings are consistent with the evidence that many de- veloping countries are characterized by surplus labor. • The reallocation of capital and labor from agriculture to more pro- ductive sectors accounts for about 20 percent of average growth. • The growth of exports makes a significant contribution to growth for 13. The sample means are shown for labor and capital, the only variables specified in the same form in both studies. GROWTH AND TRANSFORMATION 31 all developing countries in the period 1964-73; however, it does not appear to have been significant before 1960. If both factor realloca- v tion and export expansion are included in the same regression, the ' latter appears to be more important. (This interaction is discussed in chapter 9.) • The capital inflow (excess of imports over exports) shows a significant effect on growth in two of the studies cited in table 2-4, in addition to its effects on investment and exports. This finding gives some support to the two-gap hypothesis that imports may constitute a limit to growth. (Chapter 9 gives an alternative interpretation of the results for 1964-73 according to this hypothesis.) Developing versus Developed Countries Virtually all the structural factors in table 2-4 that affect the rate of growth are correlated with the level of development. That other factors are also significant is shown by adding per capita income as an explanatory variable in table 2-4 (equations 7 and 8a), which improves the regression results for developing countries although it has little effect on the de- veloped group. The extent of multicolinearity makes it necessary to esti- mate specific models-as is done in chapter 9-before attributing causal significance to any particular statistical finding. Despite these qualifications, the cross-country studies of growth yield two general results. First, they identify several aspects of structural change that affect the rate of growth and that have a varying importance at different levels of development. Taken together, these lead to a pattern of an accelerating and then a declining rate of growth as per capita income rises. Second, the structural factors in the cross-country regressions are all more significant for the developing countries than for the developed ones, whereas the growth of the labor force has more effect in developed countries. Investment is the only source of growth shown to be important for both groups, although one may speculate about the relative impor- tance of technological improvements. These comparisons suggest not only that there are somewhat dissimilar sets of growth factors for developing and developed countries but also that there is a need for different research strategies to more clearly identify each set. Although cross-country regressions and time-series analyses yield fairly consistent results for the developed countries, they give two quite different pictures for developing countries. To reconcile the two, it is necessary to take explicit account of the changing structure of demand and production, something omitted from the conventional description of the sources of growth. Structural Transformation The structural transformation of a developing economy may be defined as the set of changes in the composition of demand, trade, production, and 32 STRUCTURAL TRANSFORMATION factor use that takes place as per capita income increases. A main thesis of this book is that to understand country differences in sources and rates of growth, the transformation as a whole must be analyzed. More specifi- cally, changes in demand and trade may affect the sources of growth as much as the changes in factor supply that have been stressed so far. The central role of international trade in the structural transformation can be revealed only if the sectors that produce tradable commodities are isolated so that the relations between demand, trade, and productivity growth can be examined. Differences in resource endowments among countries are also manifested in variations in the patterns of trade over time. This section outlines the relations between changes in demand and trade on the one hand and the sources of growth on the other and suggests ways of integrating this demand-side analysis with the supply-side approach. Accounting for Sectoral Differences The sources-of-growth methodology referred to in the first section of this chapter relies on a combination of accounting identities and a few economic assumptions in order to compare the growth processes of differ- ent countries. I shall first disaggregate this supply-side analysis by sector and then combine it with a corresponding breakdown of demand and trade. The result is a demand-side view of the factors leading to structural change and growth that is consistent with supply-side analysis. The con- struction of comparable statistical accounts for the economies in the sample will in turn make possible comparative analysis and the develop- ment of more complete models. The disaggregation of the sources of growth on the supply side is quite straightforward. In the equilibrium version, a growth equation of the form of equation (2-2) can be specified for each sector of the economy: (2-8) G; = ~KpK; + ~LpL; +X.; where each term has a meaning similar to the aggregate model, equation 2-3. The growth of the economy is given by a weighted average of the sectoral growth rates: (2-9) where the weights are the average shares of each sector, p;, derived from the analysis of demand given below. When the analysis of supply is separated from the analysis of demand, these sectoral weights must be exogenously given. This disaggregated analysis can be used with different forms of production functions and can include sector-specific inputs (such as natural resources) that may set a limit to the growth of particular sectors. The corresponding system of growth accounting from the demand side GROWTH AND TRANSFORMATION 33 is developed in chapters 3 and 5. It is based on the following accounting identity for each sector of production (see equation 3-3): X;= D;+ (E;- M;) + IX; 1 I where X; = gross output of sector i D; =domestic final demand (consumption plus investment) (E;- M;) = net trade (exports minus imports) X; 1 = a; 1X 1 = intermediate use of commodity i by sector j (a; 1 is assumed to vary with the level of per capita income). The properties of the input-output system make it possible to eliminate intermediate demand as a separate source of growth by attributing it to the elements of final demand (see equation 5-11). In this way, the increase in production of sector i is equated to the sum of four factors: • The expansion of domestic demand (DD ), which includes the direct demand for commodity i plus the indirect effects on sector i of the expansion of domestic demand in other sectors • Export expansion (EE), or the total effect on output from sector i of increasing exports • Import substitution (Is), or the total effect on output from sector i of increasing the proportion of demand in each sector that is supplied from domestic production • Technological change (10), or the total effect on sector i of changing input-output coefficients throughout the economy as wages and in- come levels rise. Of these four factors, the only one with a strong basis in theory is domestic demand, for which generalized systems of Engel functions have been "#' estimated in many countries. 14 There are three formal similarities between the two approaches to accounting for growth. First, the supply-side analysis states the require- ment for a combination of inputs to produce a given commodity; the demand-side approach specifies the need for that commodity throughout the economy to "produce" a given level of GNP. Second, the supply decomposition allocates productivity growth by sector as a residual; the demand decomposition measures a similar factor-technological change-by changes in input-output coefficients throughout the economy. Third, each approach can be applied as a first approximation by ignoring price effects (or assuming constant prices). Once prices are explicitly introduced, the separation between supply-side and demand-side analysis breaks down because both are necessary to determine relative prices. 14. See Lluch, Powell, and Williams (1977). 34 STRUCTURAL TRANSFORMATION The importance of disaggregation depends on the sectoral differences in either production functions or demand conditions. On the supply side, large differences in production functions exist among agriculture, mining, manufacturing, utilities, and services. On the demand side, large differ- ences stem from income elasticities, tradability, and the extent of in- termediate use. These are illustrated in the prototype model of the trans- formation developed in the next chapter. Interactions between Supply and Demand The two accounting systems presented above describe the results of the interaction of the several factors affecting economic growth and structural change. Although an accounting system cannot disentangle causal rela- tions, it can indicate the relative importance of various factors. This chapter therefore concludes by juxtaposing the results of applying demand and supply decompositions to a single set of data taken from the prototype model (set forth in chapter 3). Table 2-6 shows a supply-side decomposition of the sources of growth Table 2-6. Sectoral Sources of Growth, Supply-Side Decomposition, Income Range $560-$1,120 Aver- Contri- age bution Capital Labor TFP Sector sector to growth growth share growth Sector i3K; i3K:GK; j3 I"i i3L:GL; (A;) (G,) (p,) (p,G,) Primary 2.69 0.41 1.00 4.09 17.4 0.71 Agriculture 0.46 2.41 0.54 0.26 0.86 3.53 14.8 0.52 Mining 0.50 4.26 0.50 1.19 1.81 7.26 2.6 0.19 Manufacturing 3.62 1.86 2.14 7.57 27.4 2.07 Light industry 0.42 2.87 0.58 1.94 2.11 6.91 16.7 1.15 Heavy industry 0.54 4.79 0.46 1.74 2.19 8.60 10.7 0.92 Nontradables 2.46 2.14 1.77 6.37 55.2 3.51 Social overhead 0.55 3.29 0.45 1.34 1.96 6.59 15.4 1.01 Services 0.35 2.16 0.65 2.46 1.66 6.28 39.8 2.50 Total economy Aggregate 0.43 2.71 0.57 1.31 2.28 6.30 100 6.30 (43)' (21) (36) Average of 2.82 1.77 1.72 sectors (45) (28) (27) Reallocation effect 0.56 (9) Note: Empty cells mean not applicable. Each row gives the breakdown of sector growth using equation 2-8. The last column gives the breakdown of total growth (6.30 percent) using equation 2-9. a. Figures in parentheses are shares of total growth. Source: Chapter 8. GROWTH AND TRANSFORMATION 35 for six sectors of a representative economy. For each sector, the rate of growth is determined (as in equation 2-8) as the weighted sum of the growth of factor inputs plus total factor productivity growth. These elements are aggregated (as in equation 2-9) to give a total growth of 6.30 percent over the specified income range. Table 2-7 gives a corresponding breakdown of the sources of growth from the demand side (using equation 5-16). In this case, the growth of each sector is expressed as a share of the total increment in GNP so that it can be compared with the initial composition of output. If a given rate of aggregate growth is assumed, the table shows the typical breakdown of the sources of demand growth starting from a middle-income level of $5 60. The most notable differences in the sectoral results are between primary production and manufacturing and between tradables and nontradables (social overhead and services). Exports and import substitution together account for 30 percent of the growth of tradables but have a relatively small effect on the growth of nontradables. Both exports and changes in -1 input coefficients reinforce the growth of demand for manufactures. T abies 2-6 and 2-7 also reconcile the two ways of accounting for the aggregate growth rate of 6.30 percent. On the supply side, the column totals in table 2-6 attribute 45 percent to the growth of capital, 28 percent to the growth of labor, and 27 percent to the growth of total factor / productivity. The row totals in table 2-7 (incremental share of value added) attribute 33 percent to the growth of manufacturing, 11 percent to\ the growth of primary production, 40 percent to the growth of services, 1 and 16 percent to the growth of social overhead capital. / These statements cannot be taken as separate explanations except under/ an extreme assumption either that demand adjusts fully to the increase in Table 2-7. Sectoral Sources of Growth, Demand-Side Decomposition, Income Range $560-$1,120 Share of Contribution to Growth (annual percent) value added (percent) Sources Total, Sector Initial Incremental Demand Trade 10 p;G; Primary 20.6 10.9 0.51 0.31 -0.11 0.71 Manufacturing 24.4 33.2 1.50 0.53 0.04 2.07 Light industry 15.8 18.4 0.88 0.25 0.02 1.15 Heavy industry 8.6 14.8 0.62 0.28 0.02 0.92 Nontradables 54.9 55.9 3.14 0.39 -0.01 3.52 Social overhead 15.0 16.2 0.91 0.11 0.00 1.02 Services 39.9 39.7 2.23 0.28 -0.01 2.50 Total 100.0 100.0 5.15 1.23 -0.08 6.30 Source: Cross-country model in chapters 3 and 6. 36 STRUCTURAL TRANSFORMATION supply by sector or, conversely, that supply adjusts to the pattern of the increase in demand. In all other cases, together they describe the outcome of equilibrating adjustments. Chapter 3 extends the analysis of the sectoral sources of growth to the whole structural transformation (see figure 3-7 below) to show the interac- tion between the level of income and the sources of growth by sector. The rise of industry-defined below as manufacturing plus social overhead facilities-becomes the dominant source of growth over the later stages of the transformation. An explanation of the factors determining these changes, and of the ways in which they are influenced by both the initial structure and government policy, constitutes the principal agenda for this book. 3 Typical Patterns of Transformation HOLLIS CHENERY MOSHE SYRQUIN CHAPTER 2 ARGUED that the growth processes of a developing country can best be understood as part of the overall transformation of its eco- nomic structure. This interdependence works in both directions: income growth causes changes in the composition of domestic demand and pro- duction, and, conversely, rising investment rates and the reallocation of labor tend to increase aggregate growth. The transformation is by no means uniform across countries, however, for it is affected by resource endowments and the initial structure of the economy as well as by the choice of development policies. In extreme cases, large structural changes may be associated with little or no growth. To pursue this argument further, it is necessary to move beyond the traditional growth accounting approaches of chapter 2 toward a more explicit model of the structural transformation that incorporates some of the most basic underlying relations. In doing so, our main objective is to design a form of analysis that takes advantage of both the aggregate data available for large numbers of countries and the detailed time series available for individual countries. This chapter constitutes a first step in this direction. It presents a simple multisectoral model designed to simulate the effects of changing demand and trade on the structure of production. Since these are aspects of the transformation for which data are relatively plentiful, the underlying structural relations can be estimated from cross-country data as well as from data for individual countries over time. This enables us to identify some common features of the transformation to be used as benchmarks in comparing the experience of various countries. Having allowed for these common elements, we can better analyze the sources of variations in the transformation. We shall concentrate on the interplay between two sets of factors: the pattern of specialization and the level of income already achieved. Differences in specialization are largely the result of a country's size, resource endowments, and trade policies. Different trade patterns, in turn, are associated with variants of the production pattern in which industrialization is either accelerated or retarded. We shall simulate three representative patterns that embody the main differences in country experience and that set the stage for the comparative studies of later chapters. 37 38 STRUCTURAL TRANSFORMATION Modeling the Structural Transformation The structure of an economy can be defined by its supplies of productive factors-labor, capital, and natural resources-and their employment in different uses or sectors. The term structural transformation encompasses the changes in the economic structure that lead to, and are caused by, a rise in the national product, together with the proximate causes of these changes. A narrow definition of proximate causes would include the accumulation of capital and skills, the effects of rising income on the composition of demand, and changes in comparative advantage. A broader definition would incorporate some aspects of productivity growth and the effects of government policies on resource allocation. The Conceptual Framework To simplify the problem of modeling the structural transformation, we divide the task into two parts: an explanation of the rate of growth, and an explanation of the changes in economic structure. Although the two parts are interrelated, they can be separated initially by treating one or the other as exogenously given. This procedure was followed in chapter 2, which analyzed the effects of specified structural changes on growth. Here the procedure is reversed; we shall begin by trying to explain changes in the economic structure with GNP growing at a given rate. Our basic measure of the economic structure is the share of GNP origi- nating in each sector of the economy (p; = V/V). The main purpose of a model of structural transformation is to explain the variations in Pi and V; as per capita GNP rises. By estimating the model from cross-country data, we attempt to simulate the full range of the transformation, which typi- cally involves at least a tenfold increase in per capita GNP. In this way, cross-country analysis supplements studies of individual developing coun- tries, which typically cover only 20 to 30 percent of this range. The study of patterns of economic development, initiated by Clark and Kuznets, has led to the identification and measurement of a number of structural changes associated with rising income. In an earlier study (Chenery and Syrquin 1975), we estimated many of these relations for the postwar period in a uniform econometric framework that takes into account a country's size and resources as well as its per capita income. The degree of similarity found between time-series studies and cross-country estimates of the principal relations provides support for the use of cross- country estimates in modeling these features, although care must be taken in interpreting the results. The most notable feature of the structural transformation, confirmed in both cross-country and time-series studies, is the rise in the share of manufacturing, Pm, in GNP and the corresponding decline in the share of agriculture, Pa· The reallocation of capital and labor from rural to urban TYPICAL PATTERNS OF TRANSFORMATION 39 areas, along with many related aspects of industrialization, stems from this basic change in the productive structure. The various hypotheses advanced to explain this process can be grouped as demand explanations, based on generalizations of Engel's law; trade explanations, based on shifts in comparative advantage as capital and skills are accumulated; and techno- logical explanations, which include the substitution of processed for natu- ral materials and the effects of differential rates of productivity growth. Our first objective is to capture the interactions among these three sets of factors as they affect the structure of production. This results in what can be described as a model of industrialization. We shall take the rates of growth of GNP and population as given, thus deferring consideration of the effects of structural change on growth to the final section of this chapter. In simplified terms, the static model of industrialization explains structural change for a given period, whereas the more complete dynamic model incorporates the interaction between demand and supply in explaining changes of growth rates for several periods. The static model of industrialization is intended for use in the historical analysis of individual countries as well as for cross-country simulations. This dual purpose requires that definitions of economic sectors and of the structural relations among variables be compatible across countries. Each approach has its strengths and weaknesses. On the one hand, because individual country models incorporate some behavioral relations omitted from the cross-country model, they can be used to analyze the effects of government policy. On the other hand, since the cross-country model covers a much wider range of income levels, it can be used to analyze the structural transformation as a whole. In this sense the two types of application are complementary; each helps to generalize the results of the other. Elements of the Model Two traditions of multisectoral analysis can be drawn on in modeling structural transformation: the input-output approach of Leontief (1951) and the applied general equilibrium approach pioneered by Johansen (1960). The main difference between them is the explicit incorporation of price effects by Johansen, which makes his model much more demanding of data but potentially more useful for policy analysis. This book draws on extensive experimentation with both types of analysis. The input-output system was first applied to the analysis of structural change in the American economy by Leontief and others (19 53). The main problem studied was the effect of changes in input coefficients between 1919 and 1939 on the structure of production and labor use, with exthnal trade and domestic demands held constant. A similar procedure was followed by Chenery, Shishido, and Watanabe (1962) in tracing the transformation of the structll'fe of production in Japan between 1914 and 1954 to changes in demand, trade, and technology. Attempts to generalize 40 STRUCTURAL TRANSFORMATION the results of this study led to the development of the cross-country model of transformation presented here, to which the experience of Japan and other countries will be compared. Johansen's 1960 study of the Norwegian economy addressed the causes and effects of nonproportional growth among sectors in more general terms. While retaining Leontief's input-output system to describe interin- dustry relations, he included demand and production functions that de- pend on relative prices. This general equilibrium approach has been developed further in a number of recent studies of developing countries designed to simulate the effects of alternative policies (see chapters 5 and 11). Although a general equilibrium approach is clearly preferable on theoretical grounds, the choice must also be tailored to both the analytical objectives and the available data. Significant differences between the two methods arise only in cases where relative prices change substantially. In the long term, the most important of these is the rising cost of labor, which leads to the substitution of capital for labor and to shifts in comparative advantage. Whereas applied general equilibrium models can, in principle, distinguish between capital-labor substitution and technological change, input-output models lump them together. For historical analysis, the potential advantages of a general equilibrium approach are likely to be offset by the limited data on prices and capital stocks. In analyzing alternative policies, however, general equilibrium models yield important insights, even when the relevant functions must be approximated from scattered observations. In the present volume, the input-output approach is used in this chapter and in the country studies of part II, whereas a general equilibrium model is used for the policy com- parisons of chapter 11. THE ACCOUNTING FRAMEWORK. The accounting framework under- lying the model must specify a breakdown of the national income and product accounts and their relation to international trade and factor use. A simple input-output system that satisfies these requirements is shown in table 3-1. The accounting unit is the productive sector, which can be defined either in aggregate terms (primary, industry, services) or with sufficient disaggregation to distinguish industries and products at the two-digit or even three-digit level of the standard international classifica- tions of production and trade. Table 3-1 represents a consolidation of elements of the national and international accounts in a standard input-output format. The breakdown of the gross national product by use is shown under (I), interindustry transactions under (II), and the sectoral origins of the GNP under (III). (The sectoral value added is broken down into returns to capital and to labor when data are available.) TYPICAL PATTERNS OF TRANSFORMATION 41 Table 3-1. The Accounting Framework Use Intermediate Source use Final use Total (producing sector) •.• j ..• n c I G E M X (II) (I) 1 X11 XI; X!n cl I! G1 E1 M1 XI x,l X;; Xin c, I, G, E, M, X, n Xn! Xn; Xnn en In Gn En Mn Xn (III) Value added VI V; vn v Capital stock K! K; Kn K Labor L! L; Ln L Total XI X; Xn c I G E M Symbols c, Private consumption w, ~,X,; I, Investment X; Output G, Government consumption V; Value added E, Exports K; Capital stock M, Imports L; Labor X,; Commodity i used by sector j Producing sector U; ~.x,; Using sector Five identities can be derived from the elements in table 3-1: First, for gross domestic product by use: (3-1) Y = (C +I+ G)+ (E-M)= D- F where Y is gross domestic product, D is domestic final demand, and F is capital inflow. Second, for gross domestic product by source: (3-2) Third, for commodity supply and use: (3-3) X;= 'i;X;; + D; + E;- M; = W; + D; + T; where T; is net trade by sector. Fourth, for the use of intermediate and primary inputs by sector: (3-4) Fifth, for aggregate sources and uses: (3-5) 'i;X; = 'i;U; + 'i1V; = 'i;W; + 'iP; + 'i;T; where - 'i;T; = F, the net capital inflow. 42 STRUCTURAL TRANSFORMATION These identities are illustrated in tables 3-4,3-5, and 3-6 below, which give estimates for three benchmark levels of income ($140, $560, and $2,100 in 1970 dollars) corresponding to the beginning, middle, and end of the transformation process. In these and other tables, the accounts are shown in a condensed eight-sector format. Two principles have been followed in designing these sectoral disaggregations: to focus on those sectors that are important to the analysis of industrialization and structural change; and to allow for the use of data from the maximum number of countries, including those for which only sectoral totals are available. These criteria lead to a substantial disaggregation of the manufacturing sectors, which are the main focus of the analysis, but much more aggregate treatment of the rest of the econ- omy. In the basic sectoral classification adopted as our standard form (see table 3-2), manufacturing constitutes fourteen out of the twenty-three sectors. The aggregation of the results into eight sectors for presentational purposes maintains the distinction between heavy and light industry used in the U.N. international standard industrial classification (ISIC). 1 STRUCTURAL RELATIONS. The model of industrialization is designed to show how factors directly related to the level of income generate changes in the structure of production and factor use. Starting from the accounting framework of table 3-1, we do this in two steps. First, the five aggregate components of the gross domestic product in equation 3-1 are expressed as functions of per capita income and other exogenous variables. Second, each aggregate is broken down by sector, so that each component of demand and trade is expressed as a function of the corresponding aggre- gate. This two-stage formulation takes advantage of the much larger body of data available for cross-country estimates of the aggregate relations. Since the application of the model to individual countries is described in chapter 5, the present discussion is limited to the cross-country version, whose purpose is to explain the stylized facts of industrialization. The exogenous variables in the cross-country model are of two kinds: universal factors, which vary fairly uniformly with the level of income, and particular factors, which produce the chief differences in development patterns among countries. The universal factors include: YIN, gross domestic product (GDP) per capita; KIN, capital stock per capita; and SIN, skills per capita. The particular factors, which serve to distinguish differ- ent allocation patterns, include: N, population size; RIN, natural re- sources per capita; and <1>, allocation policies. 1. The characteristics of industrial sectors are examined further in chapter 7. In principle, a corresponding disaggregation of agriculture would be desirable to allow for differences in demand, production functions, and trade. Although this can be done for individual countries, it has not proved feasible for cross-country comparisons. TYPICAL PATTERNS OF TRANSFORMATION 43 Table 3-2. Sector Classification and Aggregation Four sectors Eight sectors Twenty-three sectors !SIC' Tradables I. Primary 1. Agriculture 1. Agriculture 01-04 2. Mining 2. Coal and oil 12,13 3. Other mining 14, 19 II. Manufacturing Light industry 3. Food processing 4. Food, drinks, and and tobacco tobacco 20-22 4. Consumer goods 5. Textiles 23 6. Clothing 24 7. Lumber and wood products 25, 26 8. Paper and printing 27,28 9. Leather products 29 10. Miscellaneous manufacturing 39 Heavy industry 5. Producer goods 11. Rubber products 30 12. Chemicals 31 13. Coal and petro- leum products 32 14. Nonmetallic minerals 33 15. Metal products 34,35 6. Machinery 16. Machinery 36,37 17. Transport equipment 38 Nontradables III. Social overhead 7. Social overhead 18. Construction 40 19. Electricity, gas, and water 51, 52 20. Transport and communication 71-73 IV. Services 8. Services 21. Trade 61 22. Real estate 64 23. Other services 62, 63 81-84, 90 a. From United Nations (1958). DOMESTIC DEMAND. To analyze the variation in domestic demand with level of income, we rewrite equation 3-1 as (3-6) Y + F = C(y,) + l(y,) + G(y,) where F is the trade balance (M - E). It is assumed that each component of demand is a function of the level of income, y, and of a set of allocation policies, <1>. The theoretical basis for this relation is strongest for the aggregate consumption function, C(y), but the accounting identity im- 44 STRUCTURAL TRANSFORMATION plies that any variable that affects one component of demand must affect the others as well. Similarly, although raising the level of investment is a typical purpose of , the total effect is to shift resources from public and private consumption to investment at a given level of income. Next, it is assumed that each sectoral component of domestic demand is a function of the corresponding total: (3-7) C;=C;(C) I;= I;(I) G;=G;(G). The form of these equations is illustrated by the several variants of the linear expenditure system, which has been estimated from both cross- country and time-series data. 2 These and other equations in the cross- country model are expressed on a per capita basis. Two phenomena dominate the shift in the overall pattern of demand as income rises. The most important is the decline of the share of food in private consumption, which permits all other components to rise. The second is the increase in the share of investment in GDP. Overall, the income-related changes in domestic demand summarized in equations 3-6 and 3-7 will be shown to account for, at most, half the observed rise in the share of manufacturing in GDP, which is the principal phenomenon to be explained. EXTERNAL TRADE. The analysis of shifts in external trade parallels that of domestic demand. It starts from the identity E + F = M. Each of these aggregates is then expressed as a function of variables that have been shown to affect the level of trade. For intercountry analysis, these include natural resource endowments, R, and population size, N, each of which has a more important effect on the share of exports in GDP than does per capita income. The third component, F, is largely determined by national and international policies. These relations can be stated symbolically as (3-8) E(R, y, N, ) + F() = M(y, N, ). Rich natural resources tend to raise the level of exports and imports, whereas a large domestic market tends to decrease the share of trade. The Heckscher-Ohlin factor-proportions theory of trade provides a starting point for breaking down E and M by sector. The assumption that the composition of exports depends on the relative availability of natural resources, R, physical capital, K, and human capital, S, yields the follow- ing expression for exports of commodity i: (3-9) E; = E;(E, R, K, S, ). 2. The extended linear expenditure system of Lluch, Powell, and Williams (1977) demon- strates this approach and provides a basis for evaluating the results. TYPICAL PATTERNS OF TRANSFORMATION 45 A similar set of factors affects the composition of imports: (3-10) M; = M;(M, K, S, <1>). Although statistical studies by Balassa (1979b) and others provide empirical support for this type of explanation of trade patterns, the lack of acceptable measures of factor intensities-particularly of resource availa- bility-makes it difficult to estimate these equations directly. Instead, as explained below, we have made separate estimates for each of three groups of countries, which differ significantly in size and factor endow- ments. On balance, the change in trading patterns with rising income is as important to the explanation of industrialization as is the shift in domestic demand. Differences in trading patterns associated with factor endow- ments and policy approaches also constitute the largest source of variation in the patterns of structural change, which are analyzed in the last section of this chapter. TECHNOLOGY AND SUBSTITUTION. The rising cost of labor and the availability of new technology affect industrialization in two ways: through direct substitution of capital for labor in each sector and through changes in input-output coefficients. Technological change in sectors such as agriculture and services leads to an increase in the intermediate use of commodities because of the mechanization of hand operations. In addi- tion, technological change tends to increase the use of processed inputs, such as electric power or steel, in all sectors. We have incorporated the common tendencies observed in many countries into the estimates of the cross-country model. 3 The net effect is to make the input coefficients, and hence intermediate demand, a function of per capita income: X,1 = a;;(y)X;. COMPOSITION OF OUTPUT. The three sets of relations for sectoral de- mand, trade, and technological change can be substituted in the account- ing identity 3-3 to give the following equation for the output of each sector: (3-11) X;= C;(y,) + G;(y,) + I;(y,) + E;(y,N,) - M;(y,N,) + !.;a;;(y)X;. The level of output in each sector is then determined by solving the input-output system of equations: (3-12) where D; = C; + G; + I;; T; = E; - M;, and r;;(y) is an element of the Leontief inverse matrix for income level y (see chapter 5). 3. The rationale for this procedure is given in Syrquin (1976) and Chenery and Syrquin (1980). 46 STRUCTURAL TRANSFORMATION Finally, the value added m each sector is proportional to sectoral output: (3-13) V; = V;(y)X; where the value added ratio, v;, may vary with per capita income. The specification of the production function underlying this input-output rela- tionship is discussed below. Cross-Country Simulation The purpose of simulating the effects of rising income levels in the cross-country model is to establish an average or standard pattern with which individual country experience can be compared. In this way, com- parative analysis can be extended beyond individual structural features to the processes by which structural change comes about. For this purpose, it is more important to have a consistent set of estimates of the economic structure at successive points in the transformation than to have more accurate estimates of isolated features. The set of general relations outlined above indicates the underlying logic of the model that we have estimated from cross-country data. It traces changes in the economic structure to the evolution of two main factors: the level of total demand (or per capita income) and the composition of factor supply (capital, skills, and natural resources per person). Rises in income lead to fairly uniform changes in the composition of demand, whereas changes in factor supply lead to shifts in trade patterns and technology. Since changes in supply conditions are less uniform across countries, they are the major sources of differences in development patterns as well as the main focus of development policies. The variation in factor endow- ments has provided the basis for several typologies of development strategies, an approach we shall also adopt in an attempt to identify more homogeneous subgroups of developing countries. ESTIMATION. The procedure for estimating the model from cross- country data has been described in Syrquin and Elkington (1978) and Chenery and Syrquin (1980). 4 It is designed to establish a standard of comparison for country studies; to decompose the standard pattern into components that can be derived from the equations of the general model; and to use data from several country samples-which has led to the specification of the model at two levels of aggregation. Consistency among estimates of the aggregate variables in equations 3-6 and 3-8 is achieved by specifying the same regression equation for each component of GDP, denoted by x, and by using the same sample of countries. The equation is of the following form: (3-14) x =a+ ~ 1 In y + ~ 2 (ln y) 2 + 'Yt InN+ "{ 2 (in N) 2 + EF. 4. A summary of the more important parameters in the model is given in Chenery and Syrquin (1980). TYPICAL PATTERNS OF TRANSFORMATION 47 Estimates of this equation for the period 1950-70 are given in Chenery and Syrquin (1975) and give rise to the standard pattern derived for that period. To identify the main sources of deviation from the standard pattern, we have tested two alternatives. The first is to introduce additional exogenous variables into the specification of the structural relations, as illustrated by population size, N, and capital inflow, F, in equation 3-14. The other alternative is to use the variables representing differences in factor propor- tions and trade policies to identify different types of economies. After experimenting with both approaches/ we have chosen the more general typological method in specifying representative patterns for the present study. The breakdown of each aggregate by sector as specified in equations 3-7 and 3-9 is derived primarily from a sample of fifteen countries having input-output accounts. For the major components-food and nonfood in consumption, and primary products and manufactures in exports and imports-these have been supplemented by data from other countries, so that greater attention is given to the chief sources of variation. For our purposes, the main test of this set of estimates lies in the overall performance of the model in simulating long-term structural change. Its ability to replicate cross-country patterns of production has been shown in Chenery (1979, pp. 85-90). At the four-sector level, the agreement be- tween observed and simulated structural changes appears to be quite satisfactory. The discrepancies are larger with less aggregation because the country samples are smaller and less representative of low-income econo- mies. The overall performance of the model is examined in the next section. INDUSTRIALIZATION. The cross-country model will be used for two purposes: to give an overview of the interaction among the factors leading to industrialization and to provide a basis for comparing the nine econo- mies that will be studied in detail in subsequent chapters. We shall first simulate the standard pattern of industrialization, which is defined by a set of solutions to equation 3-12 that correspond to average demand and trade patterns at each level of income, with population size and capital inflow held constant. Country comparisons will then be made for each of the common features. The income levels used in these simulations covered the complete transi- tion from less developed to mature industrial economies. This range is broken down into six periods, each of which is defined by the interval between successive benchmark levels of per capita income (see table 3-3). Most of the simulations cover only the first four periods, which encompass the transition from a less developed to a mature economy. All the semi- industrial countries identified in chapter 4 now fall into the income range 5. See Chenery and Syrquin (1975, chap. 4). 48 STRUCTURAL TRANSFORMATION Table 3-3. Breakdown of Transition for Simulations Income range (dollars per capita)b Period' 1964 dollars 1970 dollars 100 140 200 280 2 400 560 3 800 1,120 4 1,500 2,100 5 2,400 3,360 6 3,600 5,040 Note: Another period, 0, is introduced in chapter 8; see footnote 3 in that chapter for explana- tion. a. Semi-industrial countries typically fall into periods 2, 3, and 4. b. The benchmark income ranges were originally defined in 1964 dollars in Chenery and Syrquin (1975). They are expressed here in the 1970 dollars used throughout this book, with a conversion factor of 1.4. A comparison factor of about 2.6 can be used to convert these 1970 dollars to 1982 prices. Also see the appendix to this chapter. of periods 2 to 4. Periods 5 and 6 have been added to analyze the changes related to the cessation of growth in the share of manufacturing in output and employment, which now characterizes virtually all mature industrial economies. Tables 3-4,3-5, and 3-6 summarize three solutions to the cross-country model, corresponding to the beginning, middle, and end of the structural transformation. The results are presented in the form of equation 3-11, which includes three components for each sector: domestic final demand, exports and imports, and intermediate demand. Changes in the composi- tion of output and value added can be traced to changes in the composition of each of these elements. Before proceeding to a more detailed analysis, we shall give an overview of transformation as a whole. Table 3-7 compares the economic structure at level 5 (table 3-6) with that at level 1 (table 3-4) using a four-sector aggregation. To focus on structural change, each element is expressed as a percentage of GDP. In this form, the increment in total output of each sector from level 1 to level 5 remains equal to the sum of the three components in equation 3-3: (3_ 15 ) (X/_ X/)= (D/ _ D/) + (T/ _ T/) + (W/ _ W/) ys yt ys y1 ys y1 ys y1 Although aggregate value added remains constant at 100 percent, total Table 3-4. Standard Solution to the Cross-Country Model, Income Level 1 ($140 per capita) (dollars per capita) Domestic final demand Per- Trade Inter- centage mediate Gross Value Per- Consump- Invest- Govern- Total of total Ex- Im- Net demand output added centage Sector tion ment ment demand demand ports ports trade (W) (X) (V) ofV Primary 1. Agriculture 25.5 0 0.5 26 18 21 3 18 17 61 51 37 2. Mining 0 0 0.5 1 0 1 1 0 2 3 2 1 Subtotal' 26 0 1.0 270 18 22 4 18 19 64 53 38 -1:>. Manufacturing \0 3. Food 15 0 1 16 11 0 3 -3 10 23 7 5 4. Consumer goods 11 0 0 11 8 1 4 -3 9 17 8 6 5. Producer goods 3 0 0 3 2 1 7 -6 11 8 4 3 6. Machinery 0 7 1 8 6 0 8 -8 1 1 1 Subtotal' 29 7 2 38 27 2 22 -20 31 49 20 15 Nontradables 7. Social overhead 7 12 1 20 14 1 1 0 7 27 15 11 8. Services 41 1 16 58 40 3 4 -1 13 70 51 36 Total' 102 21 20 143 100 28 31 -3 70 210 140 100 Percentage of total final demand 71 15 14 100 a. Totals may not add because of rounding. Source: World Bank data. Table 3-5. Standard Solution to the Cross-Country Model, Income Level 3 ($5 60 per capita) (dollars per capita) Domestic final demand Per- Trade Inter- centage mediate Gross Value Per- Consump- Invest- Govern- Total of total Net demand output added centage Sector tion ment ment demand demand Exports Imports trade (W) (X) (V) ofV Primary 1. Agriculture 45 0 1 46 8 62 21 41 62 149 101 18 2. Mining 1 0 1 2 0 3 6 -3 19 18 15 3 Subtotal' 46 0 2 48 8 65 27 38 81 167 116 21 Manufacturing '"" 0 3. Food 61 0 4 65 11 6 11 -5 48 108 33 5 4. Consumer goods 51 0 3 54 10 14 13 1 65 120 56 10 5. Producer goods 18 0 1 19 3 15 31 -16 78 81 38 7 6. Machinery 2 39 1 42 8 1 36 -35 13 20 10 2 Subtotal' 132 39 9 180 32 36 91 -55 204 329 137 24 Nontradables 7. Social overhead 32 65 7 104 18 10 7 3 42 149 84 15 8. Services 161 6 68 235 42 24 17 7 65 307 223 40 Total' 371 110 86 567 100 135 142 -7 392 952 560 100 Percentage of total final demand 65 20 15 a. Totals may not add because of rounding. Source: World Bank data. Table 3-6. Standard Solution to the Cross-Country Model, Income Level 5 ($2,1 00 per capita) (dollars per capita) Domestic final demand Per- Trade Inter- centage mediate Gross Value Per- Consump- Invest- Govern- Total of total Ex- Im- Net demand output added centage Sector tion ment ment demand demand ports ports trade (W) (X) (V) ofV Primary 1. Agriculture 64 0 4 68 3 117 101 16 89 253 137 6 2. Mining 19 0 4 23 1 12 78 -66 125 82 62 3 Subtotal' 83 0 8 91 4 129 179 -50 294 335 199 9 v, Manufacturing ..... 3. Food 186 0 19 205 10 59 32 27 180 412 126 6 4. Consumer goods 202 0 15 217 10 91 45 46 332 595 270 13 5. Producer goods 104 0 6 110 5 95 104 -9 480 581 263 12 6. Machinery 15 172 4 191 9 60 132 -72 73 192 99 5 Subtotal' 507 172 44 723 34 305 313 -8 1,065 1,780 758 36 Nontradables 7. Social overhead 115 283 26 424 20 41 28 13 150 587 330 16 8. Services 549 25 300 874 42 105 72 33 219 1,126 813 39 Total·' 1,254 480 378 2,112 100 580 592 -12 1,728 3,828 2,100 100 Percentage of total final demand 59 23 18 100 a. Totals may not add because of rounding. Source: World Bank data. Table 3-7. Structural Change during the Transformation, Income Level 5 Compared with Levell (percentage of GOP) Domestic Intermediate demand (D) Net trade (T) demand (W) Gross output (X)' Value added (V) --- Ini- Incre- Ini- Incre- Ini- Incre- Ini- Incre- Ini- Incre- Sector tial Final ment tial Final ment tial Final ment tial Final ment tial Final ment -- Tradables v, -14 -2 -15 -30 -29 N Primary 18 4 13 14 14 0 46 16 38 9 Manufacturing 28 34 +6 -14 0 +14 22 51 +29 36 85 +49 15 36 +21 Nontradables Social overhead 14 20 +6 0 1 +1 5 7 +2 20 28 +8 11 16 +5 Services 42 42 0 -1 2 +1 9 10 +1 50 53 +4 36 39 +3 Total' 102 100 -2 -2 -1 +1 50 82 +32 151 182 +31 100 100 0 Note: Period 1 is defined by a GOP level of $140 per capita (table 3-4); period 5 is defined by a level of $2,100 (table 3-6). a. X= D + T + W. b. Totals may not add because of rounding. Source: Tables 3-4, 3-6. TYPICAL PATTERNS OF TRANSFORMATION 53 gross output increases from 151 to 182 percent of GDP, an increase reflecting the growth in intermediate use of industrial products. Industrialization is commonly measured by the rise in the share of manufacturing in GDP, shown in table 3-7 as an increase from 15 to 36 percent. In a general-equilibrium context, however, industrialization is a property of the system as a whole, in which the fall in the share of primary production from 3 8 to 9 percent is offset by a rise in social overhead as well as in manufacturing. The share of services, in constant prices, changes little. 6 The causes of the rise in manufacturing differ considerably from those of the decline in primary output. An initial account of these differences is given by equation 3-15, in which intermediate demand is treated as a separate factor (which is later decomposed into separate sources). Table 3-7 shows that the decline of the primary share can be attributed equally to the fall in domestic demand and to the shift in net trade/ Intermediate demand for primary products remains constant in relation to GDP. The proximate causes of the rise in manufacturing are quite different. The increase in intermediate demand accounts for more than half the total, whereas the rise in the share of domestic demand is only 12 percent of the rise in gross output. Even after the growth of intermediate demand is distributed between domestic demand and trade (see chapter 5), the former will be shown to account for less than half the rise of the industrial share. This finding requires a substantial revision in the common view that industrialization is largely explained by Engel effects. couNTRY COMPARISONS. The changing economic structure that is simulated by the cross-country model can be thought of as the path that a hypothetical country, whose behavior is based on postwar relations, would follow as its income rises. Because the earlier transformations of the now-developed countries have taken place over a century or more, we could not hope to validate these relations from historical experience even if the relevant data were available. Instead, the validity of this approach can best be judged by the extent to which it captures important aspects of the experience of the countries in the postwar era for which it has been estimated. The cross-country model is used in this book as one of the ways of generalizing about the experience of a large sample of developing coun- 6. At this level of aggregation, the changes in value added simulated by the model are quite close to those derived from cross-country regressions (Chenery 1979, pp. 85-90). Because the regressions are based on current prices in each country, they reflect the rise in the relative price of services. Kravis (1984, p. 30) shows that the share of services in domestic final demand remains quite constant across rising income levels when valued in constant interna- tional prices. 7. In this aggregation, food processing is treated as part of manufacturing. Since tables 3-4, 3-5, and 3-6 show that the share of processed food remains almost constant, shifting it to agriculture would have little effect on this result. 54 STRUCTURAL TRANSFORMATION tries. Its results are considered in conjunction with detailed country studies in which we can come closer to inferring causal relationships between policies and performance. In this context, we are as much interested in the differences between individual country experience and the standard pat- tern as we are in the similarities. The nine economies in our sample were chosen to cover the full range of transformation that has been simulated above. On the basis of their per capita incomes, Colombia, Korea, Taiwan, and Turkey fell in period 1 in the 1950s and Mexico and Yugoslavia in period 2. At the upper end of the scale, Israel, Japan, and Norway reached income levels exceeding $2,000 by the early 1970s. 8 Because identical accounts have been constructed (in constant 1970 prices) for benchmark years in each of these economies, a comparison can be made with the cross-country simulations for each element of the economic structure. The common features of industrializa- tion that can be identified in this way are discussed in the following section. Although the cross-country model is generally consistent with the styl- ized facts of postwar industrialization, the limitations of this approach should be kept in mind when comparing country experience with the model simulations. Cross-country estimates of value added by sector have been compared with historical changes in developed countries by Temin (1967) and Kuznets (1971) and with postwar time series for developing countries by Chenery and Syrquin (1975). Both types of comparison point to a significant time-related shift from primary production to services that is separate from the effects of rising income. This may be largely the result of rising relative prices of services. For industry, however, the cross-section and time-series estimates are quite similar and the time trends are rela- tively small. 9 Common Features of Industrialization The model of industrialization traces the rise of industry to shifts in domestic demand, the growing intermediate use of industrial products, and the transformation of comparative advantage as factor proportions change. Although these phenomena can be observed in virtually all de- veloping countries, their relative importance varies according to each 8. As discussed in the appendix to this chapter, the methodology of comparative analysis can be improved by the use of international conversion factors for national income based on purchasing power rather than on exchange rates when the former become more widely available. Within our sample, the purchasing power of Israel is somewhat overstated and that of Japan understated by the use of 1970 exchange rates. However, the comparisons in this chapter are limited to changes over time, which are not significantly affected by differences in conversion factors of the magnitude that have been observed. 9. See Chenery and Syrquin (1975, p. 124) for a comparison of predictions with and without time trends. Extension of our previous work (Syrquin 1985c) up to 1982 indicates that the recent, more turbulent period has not significantly affected the nature of the transformation. A negative time shift is still in evidence for agricultural production, but it is now offset by a time-related increase in mining rather than in services. TYPICAL PATTERNS OF TRANSFORMATION 55 country's initial structure, natural resource endowment, and development policies. This section discusses these three phenomena as they are simulated by our model under typical assumptions and as they are reflected in the experience of the nine economies in our sample. In addition to explaining the model results, this procedure brings out the differences among these economies. The causes of variation in patterns of industrialization and sources of growth are explored further in the final section. Shifts in Domestic Demand Figures 3-1 and 3-2 show the estimated changes in the principal compo- nents of domestic demand-food and nonfood consumption, investment, Figure 3-1. Changes in Main Components of Demand in Sample Economies 90 80 70 60 "" 8 50 40 30 20 10 0 100 150 200 300 400 600 800 1,000 1,500 2,000 3,000 Per capita GNP (dollars) 56 STRUCTURAL TRANSFORMATION Figure 3-2. Changes in Shares of Food and Nonfood Consumption in Final Demand 0 ~---L---L~--L-~~~-L-L-L~----~---L~--~ 100 150 200 300 400 600 800 1,000 1,500 2,000 3,000 Per capita GNP (dollars) and government consumption-for the standard pattern and for each economy. In almost all cases, the largest single change in demand is the fall in the share of food consumption, which is therefore treated as a separate component in figure 3-2. In the average patterns shown in figure 3-2 and table 3-8, this share falls from 29 percent of national income at level 1 ($140) to 19 percent at level 3 ($560) and to 13 percent at level 5 ($2,100). (Note: Unless otherwise indicated, all dollars are 1970 dollars.) The reduction is even steeper for Israel, Japan, Korea, and Turkey in relation to their levels of income. The total fall in the share of food in domestic demand (by 16 percentage points) is offset in the standard pattern by a rise in the share of investment TYPICAL PATTERNS OF TRANSFORMATION 57 (by 8 points) plus a rise in nonfood and government consumption (by 9 points). However, differences in initial conditions and development poli- cies produce substantial variation in this pattern. In Japan, Korea, and Taiwan, which had the highest rates of GNP growth, virtually the entire fall in food consumption was offset by a rise in investment of more than 10 percentage points. At the other extreme, Israel, Norway, and Yugoslavia, which had already reached investment levels of over 30 percent in the mid-1950s, chose to offset the fall in food consumption primarily by an increase in public consumption. The remaining economies-Colombia, Mexico, and Turkey-followed the standard pattern. Not only does the proportion of nonfood demand change substantially with rising income, so does its composition. In contrast, there is relatively little variation in the commodity breakdown of either investment or government expenditure. 10 The combined effects of rising income on the composition of total domestic demand are shown in table 3-8, which compares the cross- country estimates for the six main sectors to the changes in each economy. Two major shifts occur in virtually all cases: first, a substantial fall in the share of food demand with the rise in per capita income; and second, increases in producer goods, machinery, and social overhead, produced by rises in both investment and consumer demand. The same trends are shown in the cross-country model, although there is somewhat less departure from proportional growth. This is due in part to the fact that this sample of semi-industrial economies is growing faster than the average; consequently, the share of investment-related sectors has increased more rapidly. Rise in Intermediate Use As countries industrialize, their productive structures become more "roundabout" in the sense that a higher proportion of output is sold to other producers rather than to final users. As with final demand, this phenomenon can be broken down into two parts: first, a shift in output mix toward manufacturing and other sectors that use more intermediate inputs; and second, technological changes within a sector that lead to a greater use of intermediate inputs. The second aspect is illustrated by the increased use of manufactured inputs in agriculture and transportation that accompanies increasing mechanization. The tendency for the share of intermediate use of commodities to increase with rising income is found throughout our sample and has also been observed in cross-country studies (Chenery 1963). As seen in figure 3-3 and table 3-7, the industrialization model simulates a rise in the 10. The principal inputs into investment are construction (59 percent) and machinery (36 percent), which are assumed in the model to be fixed. Within the sample, the share of machinery varies between 35 and 40 percent of investment costs. Table 3-8. Shares of Final Domestic Demand in Sample Economies Shares of demand (percent) Benchmark Per years capita Agri· Con- Pro- Social and income Change culture sumer ducer Ma- over- v, Economy' change (dollars) (percent) and food goods goods chinery head Services Oo Colombia 1953 274 34.9 8.1 3.5 6.4 12.7 32.0 1970 369 30.2 9.5 6.2 6.0 14.8 33.3 35 -4.7 1.4 2.7 -0.4 2.1 1.3 Mexico 1950 380 25.1 10.8 4.0 5.6 10.7 42.0 1975 736 19.5 8.4 6.1 12.2 14.8 37.0 94 -5.6 -2.4 2.1 6.6 4.1 -5.0 Turkey 1953 239 40.1 6.2 4.5 6.0 18.3 24.7 1973 461 26.1 7.6 9.3 8.8 20.3 27.6 93 -14.0 1.4 4.8 2.8 2.0 2.9 Yugoslavia 1962 469 33.9 8.3 2.0 10.2 28.6 16.3 1972 781 22.8 11.0 4.2 16.0 25.2 19.6 67 -11.1 2.7 2.2 5.8 -3.4 3.3 Japan 1955 500 27.0 7.9 0.9 5.8 15.3 43.5 1970 1,897 11.6 7.4 3.9 18.2 23.5 35.4 279 -15.4 -0.5 3.0 12.4 8.2 -8.1 Korea 1955 131 46.1 9.4 1.9 3.7 10.6 28.0 1973 323 33.1 9.9 4.7 10.9 18.2 23.1 147 -13.0 0.5 2.8 7.2 7.6 -4.9 Taiwan 1956 203 36.1 2.8 5.2 4.1 9.6 41.6 1971 426 24.5 8.5 5.4 14.7 14.1 32.6 110 -11.6 5.7 0.2 10.6 4.5 -9.0 Israel 1958 1,067 23.0 10.0 4.6 10.6 24.6 27.2 1972 2,372 12.6 8.7 16.1 18.5 26.1 18.1 122 -10.4 -1.3 11.5 7.9 1.5 -9.1 Norway 1953 1,171 20.9 12.3 16.7 1.9 20.0 28.0 1969 2,769 14.9 12.4 20.0 2.5 28.5 31.6 136 -6.0 0.1 3.3 0.6 -1.5 3.6 Cross- Levell 140 29 8 2 6 14 40 economy Level 3 560 19 10 3 8 18 42 model Level 1 to v, level 3 300 -10 +2 +1 +2 +4 +2 \0 Level 5 2,100 13 10 5 9 20 42 Level 1 to level 5 1,400 -16 +2 +3 +3 +6 +2 Note: Mining is omitted because of negligible final use. a. Listed according to broad development strategy, in descending order from inward- to outward-oriented, as discussed in chapter 4 (see especially table 4-3). This order of listing for the semi-industrial economies will be used throughout the tables in the book. Source: World Bank data. 60 STRUCTURAL TRANSFORMATION Figure 3-3. Share of Intermediate Use in Gross Output, Sample Economies 60r---,--------r--------r-------~------~--~ 50 40 .... ::I 0.. .... ::I 0 "' "' 30 ... 0 Oil .:: .., ... - --- __ ..... ----- <'l ..c Vl -- Standard manufacturing 20 -- Japan (psewar) 10 Colombia / Israel 140 280 560 1,120 2,100 Per capita GNP (dollars) average share of intermediate use in total domestic demand from 33 to 45 percent over the period of transformation. The four higher-income econo- mies follow this pattern fairly closely, whereas the five poorer ones show more rapid increases. II In addition to this overall rise in intermediate use, the substitution of manufactured goods for primary inputs is uniformly observed in our sample. This phenomenon is also captured in the general model, as shown in figure 3-3. The model produces a gradual decline in the intermediate use 11. Although there are not enough observations to analyze this aspect of technological change in greater detail, this comparison suggests that we may have underestimated its effects in the lower income levels. This phenomenon is discussed further in chapter 7. TYPICAL PATTERNS OF TRANSFORMATION 61 of primary products. A more rapid decline has occurred in economies that started from relatively high levels of primary inputs: Japan, Taiwan, Turkey, and Yugoslavia. The increase in the intermediate use of manufac- tures tends to be more rapid than is simulated in the standard pattern. The largest deviations are associated with the very rapid transformation that has taken place in Korea and Taiwan. To separate the effects of changing input coefficients from the effects of changing output composition, table 3-9 gives two calculations of in- termediate demand for each economy. The first shows the actual change in Table 3-9. Actual versus Predicted Intermediate Use in Sample Economies (constant output shares) Initial Primary Manufacturing Total and terminal Pre- Pre- Pre· Economy year Actual dieted Actual dieted Actual dieted Colombia 1953 5.9 6.7 12.3 13.3 28.7 30.9 1970 6.3 6.3 18.5 18.7 36.3 36.2 Change 0.4 -0.4 6.2 5.4 7.6 5.3 Mexico 1950 10.1 10.0 12.7 13.4 33.2 33.5 1975 8.5 9.0 20.9 18.9 40.3 38.4 Change -1.6 -1.0 8.2 5.5 7.1 4.9 Turkey 1963 16.5 14.8 12.6 12.3 35.9 34.5 1973 10.3 11.1 18.5 15.7 38.8 36.4 Change -6.2 -3.7 5.9 3.4 2.9 1.9 Yugoslavia 1962 17.7 17.5 17.8 14.9 46.2 41.4 1972 11.3 14.8 27.3 19.2 49.3 43.3 Change -6.4 -2.7 9.5 4.3 3.1 1.9 Japan 1955 14.6 10.0 24.4 20.5 48.0 43.2 1970 6.0 9.0 36.6 27.6 54.3 47.8 Change -8.6 -1.0 12.2 7.1 6.3 4.6 Korea 1963 11.8 11.7 24.5 29.6 43.6 49.9 1973 7.0 7.0 35.0 33.5 50.4 48.7 Change -4.8 -4.7 10.5 3.9 6.8 -1.2 Taiwan 1956 17.0 16.3 10.2 22.5 35.9 49.4 1971 10.4 10.6 32.1 30.9 50.8 50.0 Change -6.6 -5.7 21.9 8.4 14.9 0.6 Israel 1958 5.3 8.2 18.0 25.2 39.5 48.0 1972 4.3 6.0 27.9 32.2 45.4 49.5 Change -1.0 -2.2 9.9 7.0 5.9 1.5 Norway 1953 8.7 9.4 20.3 22.0 44.7 47.0 1969 7.3 9.7 23.7 26.6 46.9 52.4 Change -1.4 0.3 3.4 4.6 2.2 5.4 Note: Intermediate use is defined as ~ a;i (Xi/X) (k), where k indicates the actual or constant shares of the using sectors. For Colombia, Japan, Mexico, Turkey, and Yugoslavia, the standard of comparison was the average output mix of Mexico (1960) and Turkey (1963). For Korea, Israel, Norway, and Taiwan, the standard was the average of Korea (1970) and Taiwan (1971). Source: World Bank data. 62 STRUCTURAL TRANSFORMATION intermediate use between the extreme benchmark years, and the second shows the predicted effect of holding the composition of output constant. Thus the second column isolates the effects of changing coefficients, which account for well over half the total change in most cases. In agriculture, evidence is available for a large number of countries on the rising share of intermediate inputs (or decline in value added) in the total value of output. Figure 3-4 gives data on the value added ratio in Figure 3-4. Value Added Ratio in Agriculture 80 .s:: ~ "" "'0 60 140 280 560 1,120 2,100 Per capita GNP (dollars) TYPICAL PATTERNS OF TRANSFORMATION 63 agriculture, as well as the estimated relation to the level of per capita income. 12 There is a decline in the value added ratio--and a corresponding rise in intermediate use in agriculture-in the entire sample except for Turkey. The regression equation implies that during the course of the transformation intermediate inputs into agriculture typically increase from less than 20 percent to more than 45 percent of the value of output. This relation is incorporated into the general model. In summary, the combination of rising purchases by other sectors, together with substitution of manufactured for primary commodities, produces rapid growth in the intermediate demand for manufactured goods. This aspect of industrialization can only be analyzed by an interin- dustry model, which makes it possible to trace growing intermediate use to the changes in final demand and trade that are responsible for it. Although our estimates of changes in input coefficients are fairly arbitrary, they provide a much better representation of the stylized facts than do models that ignore this aspect of structural change. Changes in Comparative Advantage The third major source of industrialization is the transformation of international trade. Through import substitution and the expansion of manufactured exports, developing countries shift away from the spe- cialization in primary products that is characteristic of early stages of development. Underlying this shift are changes in supply conditions- accumulation of skills and physical capital plus the greater availability of intermediate inputs-as well as economies of scale based on a growing domestic market for manufactured goods. As table 3-7 shows, these changes typically contribute more to increasing the share of manufactur- ing in total output than do the changes in domestic final demand. Although shifts in comparative advantage ultimately affect all develop- ing countries, their magnitude and timing vary greatly. Countries with small populations have relatively specialized economies and a high share of trade in GNP, but the trade share declines markedly with increasing population size. A country's natural resources, and how they are ex- ploited, have a substantial impact on its comparative advantage; this is most pronounced in small countries and at low income levels. The effects of historical and geographical factors tend to be accentuated by differences in national policies: large countries have been prone to adopt inward- oriented policies, which appear more feasible to them than to small countries. The effects of size and resource endowments on trading patterns, being complex, can be brought out more readily by subdividing the sample and estimating separate patterns for each group. In an earlier study (Chenery 12. See Syrquin and Elkington (1978). The regression equation used in the cross-country model is Vag= 1.388- .111 In y, where Vag is the value added ratio in agriculture. 64 STRUCTURAL TRANSFORMATION and Syrquin 1975), we tested a two-way classification of countries based on their population size and their relative specialization in primary or manufactured exports. Because large countries (those with populations of more than 20 million in 1970) are-with few exceptions-considerably less specialized than small ones, they were treated as a single group in estimating typical trade patterns. Figure 3-5 shows the average export patterns of the three main country groups: large (L); small and oriented toward manufactured exports (SM); small and oriented toward primary exports (sP). The figure compares the estimated export pattern in each group with the standard pattern derived from the pooled regression for all countries; representative population sizes are used in each case. 13 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. This group of large developing countries is more homogeneous than the full sample, as shown by the smaller standard deviations from the average trade and production pat- terns (see chapter 4). Only in a few cases (such as Korea) do differences in resources and trade policies lead to substantial departures. The small countries are subdivided according to the natural resource content of their exports into primary-oriented and manufacturing- oriented trade patterns. 14 Although the indexes used to make this division are rather arbitrary, the typical patterns that result are not much affected by the choice of criteria. Figure 3-5 shows that the differences between the trade patterns for the two groups are very substantial. In the SM group, manufactured exports overtake primary exports at an income level of $250, which parallels the transformation of domestic demand. The typical SP economy, on the other hand, maintains a strong comparative advantage in primary exports throughout the transformation, while manufactured exports make little or no contribution to the rise of industry. In this case industrialization is largely the result of the demand effects of rising in- come. The identification of three typical trade patterns not only improves the explanatory power of the regressions but also brings out important varia- tions in the role of trade in industrialization. These are explored further in the final section of this chapter, in which the effects of policy differences are also considered. Reallocation of Capital and Labor The shift in the composition of output with rising income is reflected in varying degree in the reallocation of labor and capital from primary 13. The population sizes are specified as 5 million for small countries, 40 million for large, and 10 million for the pooled sample, which are close to the median values for each. 14. The basis for this division is described in Chenery and Syrquin (1975, pp. 68-69) and in chapter 4. Figure 3-5. Trade Patterns for Three Groups of Countries Primary exports 20 ------ ----- , ,, , ,, .................... ..., SP 15 1- ~ ~ ~ p.. 0 " ...... 0 ~ Standard --- --·- 10 I- "' <1J .... ? c:: <)) __ 40 r- ...... _____ . ... -- -- --- -- ---- u ..- <)) ~ 30 r- --::.-=-.::: ...-- Manufacturing ----- ""0 <)) ""0 ""0 20 r- - -- 10 ~·--·-·--· Social overhead ~:i~:;;,---- 0 140 280 560 1,120 2,100 Per capita GNP (dollars) 70 60 '>? c:: <)) 50 ... u <)) ~ ... c:: 40 <)) ........ 6 ....... 30 ....... >-. 0 ...... .... - ---- 0.. 6 Manufacturing ' ~ 20 10 Social overhead 0 140 280 560 1,120 2,100 Per capita GNP (dollars) 40 . ·-·-:-1-·-·-·-·- '>? c:: <)) ~--·-·-· Social overhead -·-·-·- - ... --- u <)) 30 Services ~ ~ Manufacturing u ... 0 20 ~---------- -----:..:=---- ---~- -c-:: ______ ... Vl -;; ... 10 ~------- ~----- -------- Primary ·o.. . 2 ~· u 1 s:: <1) ·o $<1) o Israel <1) > 0 ·a ell Q) o]apan ~ D oM~ -1 -2 '"""', '0 ugoslavia •Sout frica -3 Gv- 2 Gv = 4 -4 4.0 5.0 6.0 7.0 8.0 9.0 Predicted growth rate (Gv) Key: • Primary-oriented • Inward-oriented D Neutral o Industry-oriented Gv Growth rate in percent M, 1 Medians Source: chapter 9. 97 98 STRUCTURAL TRANSFORMATION try-oriented economies with their high efficiency, whereas Mexico and Turkey are representative of large, inward-oriented economies. In addi- tion to Yugoslavia, Israel is an outlier in achieving high growth through high factor inputs rather than through high efficiency. In general, the eight economies that we shall study in considerable detail appear as somewhat above average performers of the three strategies they represent. 18 Sectoral Sources of Growth Although industrialization plays an important part in the development of all semi-industrial countries, its timing and contribution to growth vary greatly with the country's initial structure and development strategy. Having illustrated the typical effects of specialization in the previous chapter, we now examine some aspects of intercountry variation in the context of the semi-industrial typology of table 4-3. In chapter 3, the causes of industrialization were traced to three factors: shifts in final demand, increased intermediate use of industrial products, and shifts in comparative advantage away from primary production. In the early stage of development, the need to industrialize can be offset for long periods by increasing primary exports; in the late stage, the rise of manufactures ceases because final demand for manufacturing no longer grows more rapidly than GNP. In between is a period of industrialization during which a substantial shift from primary production to manufactur- ing takes place. How uniform are these tendencies? Has industrialization always been the path followed in practice? We shall examine these questions first for the two decades ending in 1973, a period when the world economy was expanding rapidly and international policies were generally supportive of growth. As elsewhere in this chapter, our primary concern is to character- ize the group of all semi-industrial economies as a background for our subsequent study of the smaller sample. A complete set of data on sectoral sources of growth for the semi- industrial economies is given in table 4-4. It includes the average shares and growth rates for the three main sectors: primary, manufacturing, and services (including social overhead). The industry index in the last column shows the relative contribution of manufacturing to the production of tradable commodities. The contribution of manufacturing to total growth at various income levels is shown in figure 4-2 for the sample (including Norway) and for some developed countries. 19 Against this background, we can summarize some of the principal relations of these factors to the semi-industrial typology. 18. The exception is Yugoslavia, which is the only centrally planned economy included. 19. For the economies in our sample, the periods shown in figure 4-2 are those in table 3-11, except for the initial period in Japan, which is here 1914-55. The last period in Japan refers to 1970-79 (World Bank 1982). Forthe industrial countries, the information generally refers to 1950-60, 1960-70, and 1970-77 and is taken from World Bank (1980c). THE SEMI-INDUSTRIAL COUNTRIES 99 In the eight industry-oriented economies in table 4-4, manufacturing makes a larger contribution to growth than in other countries at the same income level. The effect of the rising share of manufacturing on aggregate growth is shown in figure 4-2 for Japan (over a longer period) as well as for other members of our sample. Korea and Taiwan stand out as examples of rapid industrialization. In the eight primary-oriented economies, manufacturing uniformly con- tributes less to growth than is typical of other countries at the same income level. But the inhibiting effects of the growth of primary exports on manufacturing (the Dutch disease) are outweighed by the accelerated growth of domestic income: there is no case among the semi-industrial economies in which the share of manufacturing declined from 1953 to 1973. 20 The relative contribution to growth of the sectors with limited trade, namely social overhead and services, is more or less constant across strategies because it is determined mainly by domestic demand. 21 The predicted effect of high capital inflows in raising the share of services is noticeable in some small countries, including Costa Rica, Greece, Israel, and Syria. The Timing of Structural Change Each development strategy implies a different sequence of changes in the composition of production, trade, employment, and other structural fea- tures. The policies adopted to carry out a given strategy are designed to stimulate certain aspects of this sequence-such as the shift from imports to production of manufactured goods-and inhibit others. But the success of a given set of policies also depends on its indirect effects on trade, employment, and income distribution throughout the economy-effects which are often more important than those for which the strategy was designed. Issues of timing affect virtually all aspects of development strategy. The exploitation of natural resources for export tends to slow the development of other tradable goods and hence the transformation of the labor force. Direct attempts to accelerate the development of manufacturing, in con- trast, tend to distort prices and inhibit the evolution of trade in accordance with comparative advantage. Before these phenomena are studied in de- tail, therefore, it is helpful to have an overview of the typical relations among different aspects of structural change in industrializing countries. How similar are the sequences of structural change among countries? This question will be addressed by looking first at the differences between large and small economies and then at the effects of different trade 20. Symptoms of the Dutch disease are evident for Ecuador and Iraq in table 4-6, which extends the analysis of structural change through 1980. 21. Service exports are substantial, however, in Hong Kong, Singapore, and Tunisia. Table 4-4. Sectoral Sources of Growth in Semi-Industrial Economies, 1953-73 Value added shares (percent of Annual growth rates GDP)b Contribution to Size, type, Per Pri· Manu· Ser· Pri· Manu- growth of GDP and name Popu· capita mary facturing vices mary facturing Industry of economy Years lation' GDP GDP (Gp) (Gm} (G,) (pp) (pm} pPGP PmGm p,G, index' Outward·, primary-oriented Large, LP "'Iran 1955-73 2.9 6.9 9.8 13.4 12.8 5.4 43 13 5.75 1.73 2.35 0.23 South Africa 1953-73 3.1 2.2 5.3 3.5 6.6 5.5 24 25 0.84 1.64 2.81 0.66 Small, SP Venezuela 1953-73 3.3 2.9 6.2 7.6 7.5 5.2 26 18 1.97 1.38 2.90 0.41 Malaysia 1955-73 2.8 2.9 5.7 5.0 8.4 5.6 42 15 2.09 1.27 2.41 0.38 ...... •Jraq 1953-73 3.3 5.4 8.7 7.0 10.1 11.1 53 11 3.74 1.11 3.95 0.23 0 • Algeria 1953-73 3.2 2.5 5.7 4.9 6.6 5.8 29 22 1.42 1.46 2.83 0.51 0 •Ecuador 1955-73 3.3 2.1 5.4 4.8 7.1 5.0 36 20 1.72 1.42 2.21 0.45 • Ivory Coast 1960-73 3.7 3.9 7.6 4.2 11.9 8.9 36 16 1.53 1.91 4.24 0.56 Inward-oriented Large, u Argentina 1953-73 1.4 2.2 3.6 2.6 4.4 3.4 18 34 0.48 1.49 1.62 0.76 Mexico 1953-73 3.5 2.9 6.4 3.6 7.2 6.9 17 29 0.60 2.08 3.76 0.78 Brazil 1953-73 2.9 3.9 6.8 4.2 8.4 7.1 21 27 0.89 2.22 3.72 0.71 Turkey 1953-73 2.6 3.3 5.9 3.4 9.6 6.9 41 18 1.39 1.76 2.81 0.56 •Jndia 1953-73 2.3 1.5 3.8 3.9 3.1 4.1 52 19 2.03 0.60 1.18 0.23 Small, SI Uruguay 1955-73 0.5 -0.2. 0.3 1.4 0.7 -0.3 17 26 0.24 0.18 -0.17 0.43 Chile 1955-73 2.0 2.0 4.0 4.2 6.8 2.4 18 29 0.76 1.98 1.27 0.72 "'Peru 1953-73 2.9 2.8 5.7 3.2 7.7 6.5 32 20 1.02 1.56 3.12 0.61 *Dominican Republic 1955-73 2.9 2.9 5.8 4.4 6.4 6.2 24 22 1.05 1.41 3.36 0.57 •syria 1956-73 3.4 3.3 6.7 2.3 8.0 8.2 25 21 0.57 1.67 4.46 0.75 Neutral Large, LN Spain 1954--73 1.1 5.4 6.5 3.1 7.0 7.4 18 35 0.54 2.43 3.53 0.82 Colombia 1953-73 2.8 2.1 4.9 3.5 6.6 5.2 36 21 1.25 1.41 2.23 0.53 *Philippines 1953-73 3.0 2.6 5.6 5.8 6.2 5.1 38 23 2.19 1.42 2.01 0.39 •Thailand 1953-73 3.1 3.7 6.8 5.5 7.0 8.5 45 18 2.47 1.29 3.12 0.34 *Egypt 1954--73 2.5 1.8 4.3 4.1 7.9 3.2 34 17 1.38 1.38 1.57 0.50 Small, SN Greece 1953-73 0.6 6.4 7.0 4.5 9.7 7.3 28 25 1.25 2.38 3.47 0.66 Ireland 1953-73 0.7 3.1 3.8 1.6 4.7 4.6 25 30 0.40 1.43 2.05 0.78 Costa Rica 1953-73 3.1 3.0 6.1 3.7 6.5 7.3 30 20 1.09 1.32 3.66 0.55 "'Tunisia 1961-73 2.2 3.6 5.8 6.8 7.0 5.0 27 16 1.86 1.12 2.83 0.38 *Guatemala 1953-73 3.2 1.9 5.1 4.4 6.1 5.2 30 16 1.34 0.98 2.78 0.42 "'Morocco 1953-73 2.4 0.3 2.7 2.0 2.8 3.2 33 18 0.65 0.51 1.58 0.44 Outward-, industry-oriented Large, LM "-" Yugoslavia 1953-73 1.0 4.9 5.9 3.5 5.8 7.7 24 42 0.83 2.43 2.66 0.75 0 "-" Korea 1953-73 2.3 5.2 7.5 4.5 13.4 7.7 40 20 1.78 2.64 3.13 0.60 Small, SM Israel 1953-73 3.3 5.5 8.8 5.7 9.5 8.9 9 34 0.49 3.27 5.07 0.87 Singapore 1960-73 2.0 7.3 9.3 5.4 15.4 7.6 3 23 0.17 3.53 5.63 0.95 Hong Kong 1953-73 2.2 5.8 8.0 5.1 9.2 7.6 3 31 0.14 2.89 5.00 0.95 Portugal 1953-73 0.2 7.4 7.6 4.0 9.8 8.0 25 35 1.00 3.43 3.19 0.77 Taiwan 1953-73 2.8 5.4 8.2 2.7 12.8 8.4 28 31 0.74 4.00 3.47 0.84 *Kenya 1954--73 3.2 2.1 5.3 3.3 6.3 6.7 39 18 1.28 1.16 2.88 0.47 • Marginal economies. See table 4·2 for definition. Note: Economies in each group are in order of highest to lowest per capita income. Economies included in sample in part II are in italics. a. 1960-75. b. Average of the initial and terminal years. c. PmGm/(ppGp + PmGml· Sources: Population, World Bank (1977); output, Poduval (1978). 102 STRUCTURAL TRANSFORMATION Figure 4-2. Contribution of Manufacturing to Growth in Selected Developed and Sample Developing Economies Taiwan c . .... I Ep/E = 0.67 I I ;::l I "'0 0 .... I 0.. >- I ;; 0.133 / TO = 2(0.18 - 0.67) = - 0.98 a ·;::: 0.. I D I I I, I I I I "-" II I ' ---- ---- I I I I L _____ _ I ------- 1 -- ------ 0.133 0.24 Exports of manufactured products as share of GDP Argentina "" 0 " ,..... 0 ... .... I TO = 2(0.76 - 0.52) = 0.48 t I ;::l "'0 0 .... 0.. t' .t.l ;; I / / ....... ---- M (high f) '/ Exports of manufactured products as share of GDP Apart from dynamic considerations, the critical values of the indicators for inclusion in the various groups are: • Outward-, primary-oriented: EL > 1.5 and TO > 0. Economies with EL between 1.0 and 1.5 and a large positive TO index were also put in this group. • Inward-oriented: EL < 0.5 and TO > 0. Economies with EL between 0.5 and 1.0 and a large positive TO index were also put in this group. • Outward-, industry-oriented, normal capital inflow: Relatively high EL and TO < 0, or normal EL with a large negative TO index. • Outward-, industry-oriented, high capital inflow: High levels of capi- tal inflow tend to be associated with low export ratios, especially for exports of primary products (Chenery and Syrquin 1975, p. 43). Therefore, a country with a high capital inflow (F) ratio but otherwise balanced will have a low EL value and a negative TO index. To accommodate these effects, we did not include in the industry- oriented group such countries as Greece with high F and a small negative TO index. Figure 4-8 sketches the boundaries of the classification. Point D repre- sents the predicted export composition and level for the country being classified. The changes in policies and in the indicators between 1965 and 1975 were significant for Algeria, Brazil, Ecuador, Egypt, Greece, and Spain. PART II The Experience of Industrialization IN THE NEXT three chapters, we turn our attention to the experience of a small sample of semi-industrial economies during the postwar period. In contrast with the long-run focus of the analysis in chapters 2 to 4, we shall make comparisons over time within a single economy, as well as compari- sons across the sample. By concentrating on shorter subperiods, we are able to examine in more detail the impacts of the various development strategies discussed in chapter 4. Our analysis will focus on the character- istics of industrialization and on the effects of different trade strategies on growth as well as on the interactions between aggregate growth and the sectoral structure of demand, production, and trade. In looking at a small sample and relatively short subperiods, it is always difficult to discern general patterns and to draw lessons for other coun- tries. Our goal is to examine the impact of the choice of a development strategy on growth and structural change. To help us discriminate between effects that reflect the choice of a development strategy and circumstances that are peculiar to individual economies, we use the cross-country model described in chapter 3 as a benchmark with which the experiences of the individual economies in our sample can be compared. Initial conditions such as country size, resource base, and income are controlled for. To study the characteristics of industrialization, one must consider the interlinked nature of the economy. Especially important during indus- trialization is the changing pattern of demand for goods as intermediate inputs. The greater use of intermediate goods in production reflects both increasing specialization and increasing complexity of interindustry rela- tions. These trends are part of the defining characteristics of industrializa- tion. The appropriate analytical framework for examining such trends is provided by the input-output accounts and models based on them. In chapter 5, we discuss a variety of models that start from the basic input-output accounts. In our analysis, such models serve two functions. 119 120 THE EXPERIENCE OF INDUSTRIALIZATION First, a model imposes a theoretical structure on historical data. It can be used as a tool for analyzing historical experience by providing a framework for sorting out the relative magnitudes of various important effects. Second, a model can serve as an empirical laboratory for doing controlled experiments. Such counterfactual analysis can provide impor- tant insights into the nature of the processes under study. In chapters 6 and 7, we use the approaches described in chapter 5 to analyze the experiences of our sample of economies: Colombia, Israel, Japan, Korea, Mexico, Norway, Taiwan, Turkey, and Yugoslavia. Inclu- sion in the sample was determined by availability of comparable input- output data in real terms for more than two benchmark years. As we have discussed in chapter 4, the sample is representative of the semi-industrial countries and spans a variety of development experiences in the postwar period. In chapter 6, we analyze the relation between different trade strategies and the aggregate structure of growth. The intent is to relate aggregate performance to policy choices and, more particularly, to explore the links between trade strategies, demand structure, and economywide growth. An input-output model enables us to identify which sectors and sources of demand serve to drive the economy forward as "engines of growth." In chapter 7, we turn to the sectoral structure of production and to the determinants of structural change. In examining changes in the composi- tion of output induced by changes in demand, trade, and production technology, and in searching for systematic relations between these changes and industrialization, we turn to such structural parameters as input-output coefficients, import coefficients, the pattern of trade, and expenditure shares for final demand. We also consider the relation be- tween the speed and character of structural change and an economy's trade strategy. 5 The Methodology of Multisector Comparative Analysis YUJI KUBO SHERMAN ROBINSON MOSHE SYRQUIN THE FIRST PART of this book presented a general analysis of growth and structural change in semi-industrial countries. The variety of development strategies and of levels of performance indicate that there is no one "best" or "necessary" way to achieve economic development. But patterns are discernible, and some strategies appear to be more successful than others. Comparative analysis reveals important lessons and uniformities that invite a more detailed analysis of the mechanisms at work. In this chapter, we systematically discuss the methods of multisector comparative analysis. Although this will take us back over some of the ground covered in part I, the discussion here will center on methodology. No actual data will be presented except for a simple numerical example. Comparisons that focus on the structure of production must capture the interlinked nature of the economy and, in particular, the flow of products for intermediate as well as final use. Greater use of intermediate goods in production reflects increasing specialization (that is, more division of labor) and increasing complexity, features that are among the defining characteristics of the process of industrialization. The national product accounts, which measure value added, purposely net out flows of in- termediate goods and so cannot provide a suitable framework for analyz- ing intersectoral linkages. We must turn, therefore, to the input-output accounts and to models based on them. The input-output accounts present, in tabular form, a picture of all the flows of products in the economy in a given year and thus provide a static view or "snapshot" of the economy. Although such a view is useful for making structural comparisons, it does not suffice for analyzing dynamic processes. In this chapter, we start from the static input-output model and go on to discuss two ways to make comparisons over time. The first is to compare changes between two points in time-the method of comparative statics. The second is to build models that explicitly capture dynamic processes and to use them to provide simulation laboratories for making dynamic comparisons. The numerical example used in the following sections consists of a 121 122 THE EXPERIENCE OF INDUSTRIALIZATION four-sector hypothetical economy that expands by 50 percent: per capita income goes up from $400 in the initial year, tb to $600 in the terminal year, t 2 • The four sectors are primary, light industry, heavy industry, and services. The numbers in the example are arbitrary, although the relative magnitudes roughly reflect the main features of typical semi-industrial countries as presented in chapters 3 and 4. (The data are presented in table 5-l below.) The Static Input-Output Model In this section, we first present the accounting elements for a compara- tive analysis of economic structure. This will be followed by a discussion of some of the problems that arise in such an analysis. Material Balance Equations The starting point for making structural comparisons is the material balance equation of the input-output accounts (which was given in slightly different form as equation 3-3): (5-l) where X; gross output of sector i W; intermediate demand for the output of sector i D; final demand for the output of sector i E; export demand for the output of sector i M; total imports of products classified in sector i Assuming that each sector produces only one output and that intermedi- ate inputs are required in a fixed proportion to output in each sector, we can write the demand for intermediate inputs by a sector as a function of its output: (5-2) where X;; is the intermediate use of commodity i by sector j and a;; is the corresponding input-output coefficient. Typically, input-output data are presented with imports classified as either competitive-that is, perfect substitutes-or as noncompetitive. If they are noncompetitive, then they are not grouped with domestic prod- ucts but are viewed as a nonproduced input into a sector, analogous to labor and capital. In the data in part II, we have consistently regarded all imports as competitive so that they share the same sector classification as domestic production.' 1. For some countries, such as Korea, this adjustment required the reclassification of data for noncompetitive sectors in various years. See Kim (1977) for the details of how this was done. THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 123 Imports of commodity i, M;, are demanded for intermediate use, Mw, and for final use, Mf. In equation 5-1 they appear in the total import supply and as part of both intermediate and final demand. Let u'f and u{ stand for the domestic supply ratios (the proportion of intermediate and of final demand produced domestically). Substituting these ratios and equation 5-2 in equation 5-1, we obtain the material balance equation for domestic production: (5-3) and similarly for imports: (5-4) M; = m'fW; + m{D; where the import coefficients are defined as m; = (1 - u;) for both intermediate and final goods. Three points should be noted about equations 5-3 and 5-4. First, exports are netted out of production in defining the domestic supply ratios. This is appropriate when there is no direct re-export of imports. Second, the formulation implicitly assumes that imports and domestic goods with the same sector classification are alternative sources of supply and are perfect substitutes in all uses. But for many intermediate and capital goods such an assumption might be incorrect. Another approach that allows for less than perfect substitution between imports and domes- tic goods is explored in chapter 11. Third, the domestic supply ratio for intermediate use, u'f, is assumed to be the same for all sectors using commodity i as an input but to be different from the domestic supply ratio for final use, u{. In the application in chapters 6 and 7, information about this distinction between intermediate and final imports was not available for the entire sample; therefore, a unique coefficient was defined for each sector irrespective of the source of demand. (For some economies, full import matrices detailing the sectoral use of intermediate imports of commodity i were available, and these were used in the calculations for table 7-7.) The presentation in this chapter will continue to distinguish between intermediate and final imports as in equations 5-3 and 5-4, with occasional references to the more complete formulation. 2 Equations 5-3 and 5-4 can be conveniently restated in matrix notation: (5-5) X= awAX + ufD + E and (5-6) where ' over a variable denotes a diagonal matrix and A is the matrix of input-output coefficients. 2. The general approach is developed in Syrquin (1976, 1985b); the extended equations are given in detail in Kubo (1980). 124 THE EXPERIENCE OF INDUSTRIALIZATION The A matrix represents the technology of interindustry relations. It has a domestic component and an imported one: (5-7) where Ad= irA= domestic input-output matrix Am= mw A= import matrix of intermediate use. 3 For the domestic material balances, Ad is the relevant matrix. The system in equation 5-5 can be solved to yield the domestic production needed to satisfy a specific level of domestic and export demand with a given technology, A, and import structure, aw and ar. The solution is: (5-8) where R is the inverse of the identity matrix minus the matrix of domestic coefficients. Table 5-1 presents the material balance accounts of our hypothetical economy at an initial year, tb and a terminal year, t 2. The structure of this hypothetical economy and its change over time are designed to reflect the main features of an average semi-industrial country. Table 5-2 summa- rizes the structural parameters, that is, the input-output matrices and the domestic supply shares. Also shown is the domestic inverse matrix of the terminal year, which is used in the computations below. In the initial period, value added in manufacturing is already larger than value added in primary production; together, these tradables account for more than half of total GDP. During the period, manufacturing grows at a faster than average rate and primary production at a slower than average rate. Services grow at the same rate as total GDP. Chapters 6 and 7 demonstrate the importance of further disaggregation within manufacturing. In this chapter, two branches are distinguished- light industry and heavy industry. This distinction serves primarily to illustrate the reduced scope for import substitution in light industry, which is reflected in the large share of domestic supply in that sector at tb when almost half the demands for heavy industry products are supplied by imports. These figures are representative of the initial conditions in the sample in chapters 6 and 7. The technology matrix, A, becomes denser over time. This shows up in the reduction in the value added coefficient for the economy and for all sectors except services, for which it remains constant. Other representative features of the hypothetical economy are mentioned below in the discus- sion of the results. Equation 5-8 and the associated import balance equation 5-6 can provide a framework for consistency planning. The simple input-output 3. These approximations are needed when a full import matrix is not available. THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 125 Table 5-l. Material Balances for a Hypothetical Economy (dollars per capita) Sector D+ E -Mf-Mw= Y+W= X v Initial year Primary 51 40 2 12 77 77 154 100 Light industry 113 14 8 11 108 92 200 80 Heavy industry 69 10 33 32 14 73 87 40 Services 181 20 0 0 201 64 265 180 Total 414 84 43 55 400 306 706 400 Terminal year Primary 85 51 4 23 109 115 224 123 Light industry 197 28 10 19 196 192 388 132 Heavy industry 110 22 44 62 26 162 188 75 Services 237 32 0 0 269 128 397 270 Total 629 133 58 104 600 597 1,197 600 Note: Variables referto equation 5-1. Yis final demand for products of a sector; Vis value added. model can be used to derive the domestic production and imports required to support any projected level of final demand and exports. The analysis, of course, is static, and projections of investment demand must be pro- vided exogenously along with projections of other components of final demand. This framework can then be extended to include dynamic pro- cesses either through comparative statics or through the design of explicit dynamic models. Before considering such extensions, however, we discuss some of the problems of using the basic framework for making intertem- poral or intercountry comparisons. Problems of Structural Comparisons In virtually all countries, input-output data are gathered and tabulated as nominal flows; that is, the entries in the table indicate nominal pay- ments by a column account to a row account. Corresponding to each payment is a flow of real goods and services from a row account to a column account. The problem is to start from the nominal flows and derive the corresponding real flows so that they are comparable over time and across countries. To see the nature of the problem, define the nominal balance equation explicitly, including prices (but ignoring exports and imports): (5-9) P;X; = P;X;i + P;D; where P; is the price of a good i and X;i is the flow of intermediate goods from sector i to sector j. The input-output coefficients are given by X;i a··=- '' XI Table 5-2. Structural Parameters for a Hypothetical Economy, Initial and Terminal Years (all entries multiplied by 100) Input-output matrixes for initial and terminal years Domestic supply shares Inverse matrix for terminal year Ab by sector Az, by sector Final Intermediate (R 2 =(I- u'f A 2 ) - 1), by sector ...... Sector 1 2 3 4 1 2 3 4 u~ u~ u'f u2 1 2 3 4 N u 0\ 1. Primary 13 10 11 8 96 95 85 80 lll9 17 13 2. Light industry 3. Heavy industry 4. Services I; a;; 35 22 16 9 60 16 20 8 54 ·ll 11 32 l'; 12 9 45 25 18 12 66 19 24 9 60 ·ll 11 32 93 52 100 95 60 100 88 56 100 90 62 100 14 13 15 138 20 22 31 123 18 'l 17 11~ Value added ratio, v; 65 40 46 68 55 34 40 68 THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 127 and the ratio of nominal intermediate flows to nominal output is given by P;a;; = P;X;; (5-10) P; P;X; For a set of nominal input-output accounts, it is convenient and traditional to define the units of the real flows so that all prices equal one. The coefficient a;; is defined as a dollar's worth of input from sector i required to produce a dollar's worth of output in sector j. When relative prices differ across countries or over time, the units of the various real magni- tudes are no longer the same and cannot be compared. Ideally, the data for all economies and periods should be deflated to a comparable set of base prices-presumably a set of world prices for some representative year. Unfortunately, such comprehensive data are not avail- able, and the comparative analysis relies instead on deflating the input- output tables over time within each economy so that the flows are ex- pressed in constant domestic prices. 4 Thus, although direct comparisons of economies' production levels are not possible, comparisons of production structures and of growth and structural change can be made. The most straightforward way to deflate to constant prices is to divide the material balance equation for each sector (equation 5-9) by an index of that sector's output prices. Intrasectoral disaggregation, both by source and by destination, has permitted the use of more refined procedures in several of the economies in our sample. 5 Nonetheless, the limitations of the available data have resulted in some estimates that are rather crude. All such comparative work entails a tradeoff between the quality of the data and the scope of the comparison. Although acknowledging the inclusion of data of inferior quality for some countries, we feel that the gains from wider country coverage outweigh the drawbacks. A problem of consistency arises from the deflation of the material flows. If value added at constant prices is taken to be the residual when the constant price value of intermediate inputs is subtracted from the constant price value of output, then it is possible for the real value added in some sectors to be negative. The problem is that the concept of real value added is theoretically ambiguous, with no obvious single way to measure it: For our comparative work, we define real sectoral value added by deflating current price value added by an index of the sector's own output prices. The resulting set of input-output flows in real terms does not satisfy the accounting identities: real GDP calculated as the sum of real sectoral value added will not, in general, equal real GDP calculated as the sum of sectoral 4. World price data, however, are available for Korea and Taiwan. See Kim (1977) and Kuo (1979). 5. Starting from equation 5-10, separate deflators can be defined for each entry in the input-output matrix. This approach has been used for Israel, Japan, Korea, Taiwan and-on an ad hoc basis-for Turkey. 6. See David (1962) and Bruno (1978). 128 THE EXPERIENCE OF INDUSTRIALIZATION net final demand. Such consistency is not required for any of our compara- tive analyses. With our focus on comparing production structures, it see.tp.s best to use a deflation procedure that is theoretically appropriate on the output side. Exports and imports raise additional problems. Ideally, exports and imports should be valued in domestic market prices. In producers' prices, exports are typically entered at their f.o.b. (free on board) value whereas imports are expressed at their c.i.f. (cost, insurance, and freight) value plus tariffs. For identical products, f.o.b. export prices do not equal the corre- sponding domestic producers' prices when, as often occurs, there is dif- ferential pricing. In turn, tariff rates do not equal the corresponding realized rates of nominal protection under various circumstances known to have prevailed for many products in all the economies in our sample. Corrections have been attempted for most of these economies, with varying degrees of sophistication/ The problems of deflation and data adjustment are sensitive to the level of aggregation at which the comparative analysis is pursued. In general, the goal has been to define an aggregation scheme that is comparable across countries and periods and has as many sectors as possible. For each economy, we started with the most disaggregated data available; the range is from 118 sectors in Korea to 20 sectors in Colombia. The finest dis- aggregation that is perfectly comparable across the entire sample is 14 sectors. We have also used 8-sector and 4-sector aggregations for compari- sons when finer detail is not needed. The various sectors and aggregations are defined in table 3-2 above. Table 5-3 indicates the nature of the underlying data. From tables 3-2 and 5-3, it can be seen that the compari- sons are roughly at the two-digit ISIC level, with some at the three-digit level. At these levels, the comparisons are probably relatively insensitive to random measurement errors owing to distortions, although systematic variations across economies or periods would still cause problems. In the following chapters, only distinct and robust results are emphasized. The data base is described in Kubo (1983). Sources of Growth and Structural Change The static input-output model described above has been used as an analytical framework for making structural comparisons. The extensive data set described above, however, will be used to go beyond static comparisons to the analysis of growth and structural change over time during the process of economic development. In this section, we first discuss some issues in the measurement of sources of growth and struc- tural change and then describe a comparative statics framework that facilitates intertemporal comparisons. 7. See the studies listed in chapter 1 for a detailed discussion. THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 129 Table 5-3. Data Characteristics of the Sample Economies Import World Number Price matrix prices of de- avail- avail- Economy sectors' f/atorb able? able? Benchmark years' Colombia 20 1958 No No 1953, 1966, 1970 Mexico 45 1960 No No 1950, 1960, 1970, 1975 Turkey 25 1968 No No 1953, 1958, 1963, 1968, 1973 Yugoslavia 29 1972 No No 1962, 1966, 1972 Japan 41 1965 Yes No 1914, 1935, 1955, 1960, 1965, 1970 Korea 118 1968 Yes Yes 1955, 1963, 1970, 1973 Taiwan 58 1971 Yes Yes 1956, 1961, 1966, 1971 Israel 84 1965 Yes No 1958, 1965, 1972 Norway 25 1955/1961 Yes No 1953, 1961, 1969 a. Refers to the lowest level of aggregation at which at least part of the data were available. For further details, see Kubo (1983). b. Base year for price deflator. For Norway the base year is 1955 for the 1953 and 1961 tables and 1961 for the 1961 and 1969 tables. c. Years for which data are available. Source: World Bank data; described in Kubo (1983). Growth Accounting The process of development entails both the growth of aggregate output and changes in the structure of the economy. Variations in the structures of production and factor use can only be analyzed in a multisectoral framework. Such a framework is also required for studying the interrela- tions between structural change and growth. That is, looking at growth only at the aggregate level or simply adding up the sectoral results leaves out an important dimension of the process. Studies of growth, whether at the sectoral or aggregate level, try to identify its determinants by one of two approaches. The first approach is to build a model (ideally a general equilibrium one) incorporating behav- ioral, technological, and institutional relations as well as assumptions about the function of markets (see chapter 11). The second approach, more limited but simpler to implement, derives the proximate sources of growth from identity-based decompositions and a few economic assump- tions. The following examples all refer to segments of the same accounting framework (see table 3-1). The first two were presented in chapter 2 as applications of the growth accounting approach. Basic equation or identity Growth accounting equation Y; = f(K;,L;,t) (2-10) G;= 13KiGK; + 13uGLi + ~; Y=iY; (2-11) Gv= ip;G; (5-1) X;= W; + D; + E;- M; (5-la) dX=dW+dD+aE-aM 130 THE EXPERIENCE OF INDUSTRIALIZATION The first growth accounting equation, 2-10, is the standard Abramo- vitz-Solow-Denison decomposition of the sources of growth. It starts from a production function, not from an accounting identity, and relies more on economic theory than the other equations do. 8 The problem with this approach is that a substantial source of growth (A. or TFP growth) has been obtained as a residual, which implies that important factors affecting growth were not incorporated into the original production function. The second equation 2-11, is illustrated graphically in figure 3-7. It allocates growth to its sectoral origins. In combination with equation 2-10, it gives the aggregate growth accounting from the supply side (see table 2-7 and chapter 8). The calculation of the sources of growth from the demand side starts from the material balance equation 5-1. Its corresponding growth accounting equation, shown in terms of increments, 9 matches the change in output with changes in its various uses and in imports. This very simple and only mildly illuminating decomposition is the starting point of the more elaborate approach in the following section. As presented below in equation 5-11, the equation is derived from an accounting identity with- out any further assumptions. Its relation to economic theory is limited to the economic basis underlying the selection of categories in national income accounting. Adding to this accounting identity the assumption of a linear technology for interindustry relations-represented by the matrix A-and an import structure that reflects the imperfect substitutability of similar commodities with different sources of supply (domestic or im- ported)-represented by the sets of coefficients t/ and aw -we obtain the more elaborate accounting identity in equation 5-8. The derivation of the growth accounting relation corresponding to 5-8 is the subject of the next section. Before turning to that analysis, we offer some observations on the growth accounting approach. • The identification of the "sources" of growth does not imply that causal relations have been established or that these proximate sources are exogenous; they may, in turn, be the result of more fundamental underlying causes. They are, however, first approximations and give orders of magnitude of the various effects. Often they provide essen- tial starting points for more structured models and a guide for further research. • When growth accounting starts from an identity, one must consider whether the categories are analytically relevant and whether the weights associated with them are economically meaningful. 8. See, however, Taylor (1979, chapter 6), which has a useful discussion of identity-based decompositions. 9. This is to facilitate the comparisons in the following sections. The expression in growth rates can be easily derived from the one in absolute changes. THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 131 Figure 5-1. Measures of Growth and Structural Change X} L--L-----L--------~----------------------x~ X\ X~ X~ • In interpreting the results, it is important to determine how indepen- dent the various sources-such as TFP growth and capital accumula- tion-are from each other and how responsive they are to policy instruments. • The supply-side and the demand-side decompositions are partial and complementary representations of the process of growth. Only in extreme cases can we rely on just one of them. Chapter 2 and part III include some attempts at integrating the two. Chapters 6 and 7 focus on the demand decomposition of the sources of growth and structural change. 10 To clarify the measures of growth and structural change used in these chapters, we give a graphic representation and then develop the algebra of a more extensive decomposition methodology. A Graphic Representation Figure 5-1 is a graph of growth and structural change in a two-sector model. Initially, the economy is at point I and is producing X~ and Xi (the subscript refers to the sector and the superscript to the period). Later, the economy is at point II and is producing Xi and X}:. Aggregate output has 10. For further discussion of these issues, see chapters 2 and 11 and the final section of this chapter. 132 THE EXPERIENCE OF INDUSTRIALIZATION grown from X 1 = X~ + X~ to X 2 = Xj + X~, where real output is defined simply as the sum of sectoral outputs in constant prices. The change in aggregate output is given by the sum of the changes in sectoral output: JlX = JlX 1 + JlX 2 (where JlX; = X7 - X}). The change in aggregate output, however, conceals significant changes in the structure of production. To measure this structural change, the movement from point I to point II is decomposed into two steps. First, the economy is assumed to grow so that all sectors expand proportionately. This balanced growth takes the economy from point I to point II', with outputs Xi and X~, and the same aggregate output as at point II. Then, holding aggregate output fixed, the structure of production is changed; the changes in sectoral production generated by moving from point II' to point II are given by 8X 1 and 8X 2 • Henceforth, we shall use the notational convention that small delta, 8, represents a change in structure and large delta, Jl, a change in output between the initial and terminal points. Note that since aggregate output is fixed along the line connecting II and II', the sum of changes in 8 over all sectors must equal zero. With only two sectors, 8X 1 must be equal in magnitude and opposite in sign to 8X 2 • It is possible to devise a number of summary measures of structural change based on the sectoral measures of 8. One summary measure from figure 5-l is the "distance" between points II and II' (denoted by II8XII). The measure II8XII is the Euclidean distance between points II and II' and is given by the square root of the sum of squares of the sectoral8 measures. 11 The measure of distance generalizes naturally to the multisector case and will be expressed as a ratio to a measure of aggregate output. In chapter 7, we use such summary measures of structural change not only for gross output but also for separate components of supply and demand. In all cases, the baseline against which structural change is defined is that of balanced growth, such that sector shares remain fixed. Thus "structural change" is defined as deviations from proportional growth while "growth" alone refers to the change in output. Decomposing Growth and Structural Change We next present the equations for separating the sources of growth and structural change. The calculations are illustrated with our numerical example. OUTPUT GROWTH. Equation 5-8 solves for domestic production given final demand-both domestic, D, and exports, E-domestic supply ratios, u~ and uj, and the input-output coefficients, a;;· Between two periods, a change in output, JlX, depends on changes in the sets of structural param- 11. Another measure, used by Chenery, Shishido, and Watanabe (1962), is the average of the absolute values of the sectoral measures. THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 133 eters, u~ aw, and A. After some algebraic manipulations, the change in sectoral outputs can be written as: 12 LlX = R2uf LiD+ R2LiE + R2LiufD1 + R2LiuwW1 + R2u'2 L1AX1 or, for sector i: (5-11) LlX; = 'i.r;; 2uf2LiD; =domestic demand expansion (oo) J + 'i.r;; 2 LlE; =export expansion (EE) J + 'i.r;; 2LiufD;; =import substitution of final goods (rsf), J + 'i.r;; 2 LlujW;; = import substitution of intermediate I goods (ISw) + 'i.r;; 2uj2fLia;kXk; = changes in input-output (10) coef- 1 ficients where R is the inverse of the identity matrix minus the matrix of domestic input-output coefficients [R = (1- Ad) - 1], r;; is an element of R, and the number subscripts refer to the periods. 13 The first two terms on the right-hand side of equation 5-11 are changes in the output of sector i induced by the expansion of domestic demand and exports in all sectors, given a constant import structure. The third and fourth terms measure the direct and indirect effects of changes in the import structure of final and of intermediate goods. The last term gives the direct and indirect effects of changes in the total (domestic and imported) matrix of input-output coefficients, which represent the widening and deepening of interindustry relations brought about by the changing mix of intermediate input requirements. The changes in input-output coefficients are caused, in turn, by changes in production technology as well as by 12. With continuous time and denoting for any variable z, dz/dt by D.z, the time derivative of equation (5-8) is D.X = R(!/D.D + M/D + D.E) + D.R(i/D +E). The derivative of an inverse matrix is given by Since AR (!/D + E) = AX = W, substituting in the first equation in this note gives the result in the text. In discrete fixed weights decompositions, interaction terms arise, but not when the weights of both periods are mixed as in equation 5-11. This index number problem is further discussed in the text. In this and later equations, we present one version of the weights. In all applications, an average of the Paasche and Laspeyres versions is used. 13. This derivation is taken from Syrquin (1976). It elaborates the direct type decomposi- tion first developed in Chenery (1960), which was incorporated into an explicit interindustry framework in Chenery, Shishido, and Watanabe (1962) and in Chenery (1969). 134 THE EXPERIENCE OF INDUSTRIALIZATION substitution among various inputs (perhaps in response to changes in relative prices), although we cannot separate these two effects without more information. All the terms in equation 5-11 give the total effects of changes, including the indirect linkages through intermediate input flows in the input-output matrix. It is also possible to define a direct decomposition in which the changes in the demand for intermediate inputs appear as an independent source of total sectoral demand, as in the growth accounting equation 5-1 above. Such an approach disregards the linked nature of the economy and is useful when concentrating on the behavior of particular sectors. 14 In this book, we generally use the total decomposition measures. In chapters 3 and 10, the direct measure is used for analysis that does not focus on linkages. Four points are worth noting about the decomposition equation 5-12. First, import substitution is defined as arising from changes in the ratio of imports to total demand for each sector. This treatment is natural given the view, discussed earlier, that imports are imperfect substitutes for domestic goods, so that the source of supply is an integral part of the economic structure. 15 With this definition, imports and domestically pro- duced goods in each sector can then be treated separately, while the aggregate contribution of import substitution to growth is sensitive to the level of sectoral disaggregation. For example, it is possible for import substitution to be positive in every sector but for the ratio of total imports to total demand still to increase because of changes in the sectoral com- position of demand. Second, since the domestic supply ratios for intermediate use are assumed to be the same for all users, the import substitution term for intermediates, ISw, is an average of the changes in the import matrix. Data permitting, it is possible to incorporate these separate changes into a more elaborate import substitution effect. Since such data are available for only a few economies (see table 5-3), we have not used this refinement in our comparative work. 16 Third, the effect of changes in input-output coefficients includes changes in the total coefficients without distinguishing between imported and 14. Frank, Kim, and Westphal (1975) argue that the direct measures are more relevant to assess how producers in specific sectors react to various incentive policies. Direct and total measures are compared in Syrquin (1976) and in Chenery (1979). 15. This treatment differs conceptually from the treatment in Chenery, Shishido, and Watanabe (1962), where imports are considered as perfect substitutes for domestic goods. There is an extensive literature on the appropriate definition of import substitution in multisector models; see Arrow (1954), Desai (1969), Morley and Smith (1970), Fane (1971, 1973), Syrquin (1976), and Balassa (1979a). 16. The algebra of the decomposition equation, including this refinement, has been developed by Syrquin (1976, 1985b). The available import matrices are used in chapter 7 (see tables 7-7 and 7-8). Kubo (1980) also summarizes the algebra of the measures used in chapters 6 and 7. THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 135 domestically produced goods. Thus, the input-output coefficients may remain constant (.:lA = 0 ), in which case the last term in the decomposi- tion will be zero, even though changes in domestic supply ratios result in changes in uwA (and hence in R ). Changes in input-output technology are defined as changes in the aggregate coefficients, including imports, while any changes in the intermediate domestic supply ratios are included in the import substitution term. For example, a change in the total amount of steel required to produce a car is classified as a change in the input-output coefficient, while a change in the mix of domestically produced and imported steel is classified as import substitution. Fourth, an index number problem is implicit in the decomposition equation because the decomposition can be defined either by the terminal- year structural coefficients and the initial-year volume weights (as in equation 5-11) or by the initial-year structural coefficients and the termi- nal-year volume weights. The two versions are analogous to Paasche and Laspeyres price indexes. In the analysis below, both indexes have been computed separately for the decomposition in each period, and averages of the two are presented. 17 The total change in sectoral output is decomposed into its sources by category of demand. The total change in output equals the sum of the changes in each sector and can also be decomposed either by sector or by category of demand. The relations (with the two IS terms combined) can be shown schematically: + + + + Reading down columns gives the sectoral composition of each demand category; reading across rows gives the decomposition of changes in sectoral demand by different demand categories. When making compari- sons across countries and time periods, it is convenient to divide the entire table by !-.:lX; so that all components across sectors and demand catego- ries sum to 100. Alternatively, it is sometimes convenient to divide the rows by .:lX; and then look at the percentage shares of the contribution of each demand category to the change in sectoral output. Both presentations are used in this book (see chapters 2, 6, and 7). 17. It is possible to define an appropriate Divisia index by specifying the time path of the different variables. See, for example, Fane (1971, 1973). Syrquin and Urata (1985), in a similar context, use mean weights to compute a discrete approximation to the Divisia index. 136 THE EXPERIENCE OF INDUSTRIALIZATION Computing equation 5-11 for the hypothetical economy yields three forms of the sources of growth (see table 5-4): absolute growth in dollars, the percentage of each sector's growth in output, and the percentage of the total change in output (as in the schematic above and in table 2-7). Tables 5-l and 5-2 provide all the data needed to perform the computa- tions in table 5-4. For example, in panel A of table 5-4, heavy industry output expanded by $101, of which $7 originated in increased demands from import substitution of final goods throughout the system: IS£(3) = 'llJ;LlufD; 1 = 0.13 ( -0.01 X 51) + 0.20 (0.02 X 113) + 1.23 (0.08 X 69) + 0.07 (0.00 X 181) = 7.2 In light industry, the domestic supply ratio increased by 0.02. Had final demand remained at $113, this would have called for an increased produc- tion of $2.26 ( = 0.02 X 113 ), which in turn would have placed indirect demands on heavy industry through the inverse matrix (r 3 ,2 = 0.20) of $0.45. Adding up all the direct and indirect effects, we obtain a total effect Table 5-4. The Sources of Growth Source of growth Sector DD EE IS f !Sw 10 A. As absolute growth in dollars ~X Primary 59 18 1 -4 -4 70 Light industry 131 27 5 3 22 188 Heavy industry 55 20 7 5 14 101 Services 92 21 1 1 17 132 Total 337 86 14 5 49 491 Rate' B. As percentage of sectoral output growth Total Primary 4.7 85 26 1 -6 -6 100 Light industry 8.3 70 14 3 12 100 Heavy industry 9.6 54 20 7 5 14 100 Services 5.1 70 16 1 12 100 Total 6.6 69 18 3 9 100 Incre- mental C. As percentage of total output growth shares Primary 12 4 0 -1 -1 14 Light industry 27 5 1 1 4 38 Heavy industry 11 4 2 1 3 21 Services 19 5 0 0 3 27 Total 69 18 3 1 9 100 a. Average annual growth rate for an eight-year period computed as 1/s(inX2 /X 1 ). Source: Data from tables 5-1 and 5-2. THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 137 of $7. This includes the decrease in demand from agriculture because of its negative import substitution (or import liberalization). The other numbers in table 5-4 can be similarly computed. Note that the last term in equation 5 -11-the effect of changes in input-output coefficients-involves at each step the change in all the coefficients in the matrix. Panel B of table 5-4 shows the relative contribution of the various sources of growth to sectoral output growth. In the numerical example, domestic demand expansion accounts for 85 percent of the growth of output in agriculture, trade effects for 21 percent, and changes in input- output coefficients for - 6 percent. This decline in the demand for agri- cultural output for intermediate use is due primarily to the uniform decline in the coefficients of the agricultural row in the table (compare A 1 and A 2 in table 5-2). A similar result is commonly found in country studies (see chapter 6), and it reflects the substitution of fabricated for natural mate- rials and the effect of changes in relative prices. Input coefficients into agriculture (the first column in the A matrix) generally rise as production becomes more roundabout. This is conveniently summarized by the change in the value added ratio shown in table 5-2 (see also figure 3-4 and chapter 6). Panel C of table 5-4 presents all entries as a proportion of the expansion of aggregate output. The last column shows the incremental sectoral shares. MEASURES OF TRADE EFFECTS. There is an important asymmetry in the way we measure the effects of exports and of import substitution, which we have to keep in mind when comparing the results in chapters 6, 7, and 10. Exports 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. The importance of this distinction varies by sector and period within a country. The values for domestic supply shares in table 5-2 are representative of the experience in the economies in our sample. Domestic shares for final and intermediate manufactured goods in the initial year are about 90 percent for light industry and 50 percent for heavy industry, or import shares of 10 and 50 percent. The scope for import substitution is ample in heavy industry but not in light industry. This can be compared with the import coefficients for several countries in our sample, which are shown in table 5-5. By the early 1950s, the potential for further import substitution in light industry was severely restricted except in Korea, where rapid import substitution came after 1955. Significant import substitution in heavy industry took place during the period covered by our study except in postwar Japan, where it occurred earlier. Another element of asymmetry in measuring exports and import sub- stitution is caused by aggregation. The flow of exports is independent of the level of aggregation, but sectoral import coefficients may hide signifi- 138 THE EXPERIENCE OF INDUSTRIALIZATION Table 5-5. Import Coefficients for the Sample Import coefficient (percent) Country Year Light industry Heavy industry Mexico 1929 22 75 1950 6 46 Turkey 1953 2 34 1963 2 22 Japan 1914 3 27 1955 2 4 Korea 1955 22 69 1963 7 42 Sources: For Mexico 1929, Villarreal (1976). All other figures are World Bank data; described in Kubo (1983). cant intrasectoral variation. Episodes of trade liberalization usually regis- ter a rapid increase in exports and in imported inputs. Sectoral import coefficients change little or even rise; this implies negative import substitu- tion. Any increase in intraindustry specialization and import substitution of specific products within sectors would fail to appear in the data, except at high levels of disaggregation. STRUCTURAL CHANGE. Equation 5-11 decomposes the change in sec- toral output. In the terminology used earlier, it represents a tool for analyzing the sources of growth from the demand side. The analysis of structural change calls for a different decomposition that measures the deviations from proportional growth. This sectoral () measure for a vari- able X is defined as ()X = X 2 - X.Xt. where X. is the proportional change in national income between two years. It is defined as the ratio of total national income in the second year to that in the first. In making compari- sons, data are available for periods of varying lengths. Thus, X. need not be an annual rate of growth but depends on the number of years in the period. Because of the linearity of the input-output system, if all the elements of domestic demand, exports, and imports were to expand at the same rate, X., then with a given input-output matrix, output would expand at the exact same rate in each sector, and the structure of production would be unchanged. Changes in this structure can therefore be traced back to deviations from proportional growth in domestic demand and exports and to changes in domestic supply ratios and in input-output coefficients. From equation 5-8, a() decomposition can be derived analogous to the .l decomposition in equation 5-11. After some algebraic manipulations, the decomposition (in matrix notation) is given by: (5-12) ()X= R 2 u{()D + R 2 ()E + R2 .luf()D 1 + R2 .luw()W1 + R 2 u2.lAX.X 1 • (a) (b) (d) THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 139 The deviation from proportional growth in output in sector i is seen to be the sum of five sets of structural changes: a. Effects of deviations in domestic demand in all sectors, with a constant import structure in all sectors b. Effects of deviations in exports in all sectors, with a constant import structure in all sectors c1 • Direct and indirect effects of changes in the import structure of final goods c2 • Direct and indirect effects of changes in the import structure of intermediate goods d. Effects of changes in the total (domestic and imported) matrix of input-output coefficients (dA ). In equation 5-12, the material balance equation in the terminal year is compared not with its counterpart in the initial year but with the hypothet- ical material balance equation in the terminal year under the assumption of balanced growth. The decomposition differs from the d version not only in measuring output growth as deviations from proportional growth but also in measuring growth of domestic final demand and exports as deviations. In the decomposition of the deviation measure, import substitution and technical change are still defined in terms of changes in domestic supply ratios and input-output coefficients, respectively. Thus, the last three terms in equation 5-12 are almost identical with the corresponding terms in the d decomposition. Given that the nonproportional components of domestic demand and export growth are generally smaller than their increments, the relative importance of import substitution and of changes in input-output coefficients will be greater in accounting for compositional changes in output than in accounting for total growth. In the diagrammatic presentation of the 8 measure in figure 5-1, it was noted that the sum of the deviations from proportional growth across all sectors must be zero. This result will also hold for the decomposition equation if A (the change in national income between two years) is defined in terms of total sectoral production. If the input-output coefficients do not change (dA = 0), then-under the assumption of balanced growth- the ratio of the aggregate value of intermediate goods to value added remains constant over time, and both value added and gross output will grow at the same rate. If the input-output coefficients do change, however, then even under the assumption of balanced growth, total value added and gross output will grow at different rates. In this case, since we define A as the change in total value added, the sum of the 8 measures across sectors will not equal zero. The difference from zero will reflect the change in the ratio of aggregate value added to total output. Table 5-6 presents the sources of deviations from balanced growth in the same format in which table 5-4 presents the sources of growth. To 140 THE EXPERIENCE OF INDUSTRIALIZATION Table 5-6. Deviation from Balanced Growth (A. = Yz l¥1 = 1.5) Source of deviation Sector DD EE rsf JSw 10 8X A. Absolute deviation in dollars Primary 12 -8 1 -6 -6 -7 Light industry 33 11 7 5 32 88 Heavy industry 8.5 9 11 8 21 57.5 Services' -33 4 2 1 25.5 -0.5 B. As percent of sectoral output deviation from balanced growth Total Primary 172 -114 14 -86 -86 -100 Light industry 37 13 8 6 36 100 Heavy industry 15 16 19 14 36 100 Services' + + + + -100 8x, C. As percent of total output growth ~X Primary 2.4 -1.6 0.2 -1.2 -1.2 -1.4 Light industry 6.7 2.2 1.4 1.0 6.5 17.8 Heavy industry 1.7 1.8 2.2 1.6 4.3 11.6 Services' -6.7 0.8 0.4 0.2 5.2 -0.1 a. Because of the small size of the sectoral deviation, only the signs of the sources of deviation from proportional growth are shown and not the relative magnitudes. Source: Data from tables 5-1 and 5-2. clarify the computations, we explicitly derive the effect of nonpropor- tional export expansion on light industry: EE(2) = !.;r2;'6E; = !.;r2;(E2;- X.Et;) = 0.14 [51 - (1.5 X 40)] + 1.38 [28- (1.5 X 14)] + 0.31 [22- (1.5 X 10)] + 0.17 [32- (1.5 X 20)] = 10.9 In explaining structural change, import substitution emerges as much more important than in the decompositions of total expansion. Domestic demand, as expected, is no longer the dominant effect. Trade and Input Use The various output decompositions start from the material balance equation for domestic production (5-3). Analogous decompositions for imports can be derived from the balance equation for imports (5-4). 18 This equation analyzes imports by sector of origin. Intermediate imports can also be studied by sector of use by being considered as inputs into the 18. The equations can be found in Syrquin (1976). THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 141 sector's production process. This approach recognizes explicitly that to sustain an increase in output, imports-which are imperfect substitutes for domestic products-are going to be needed as intermediate inputs. Ideally, an import matrix should be used in this analysis. When one is not avail- able, it can be approximated by rnA-as in some entries in table 7-7--or by ,nwA -as in the example in this chapter. Considering intermediate imports by sector of use will help us to examine the relation between export expansion and its import requirements. IMPORT CONTENT OF EXPORTS. The production needed for a given level of exports requires imports as intermediate inputs. We now develop a measure of the total import content of exports (ICE). Let Q = AmR, where Am is an import matrix. A typical element of Q-q;1-gives the total imports of commodity i needed to generate one unit of final product of sector j. The column sums of Q then give the total import content in one unit of final product in each sector. The Q matrix for the terminal year in our example is given in table 5-7. One dollar of final product from the light industry sector (sector 2) calls for 3.6 cents of imported primary goods, 5.2 cents of imported light industry products, and so on. Adding up all the requirements, we find a total import bill of 20 cents or an import content of 20 percent. The equation for the import content of exports (ICE) is ICE = e' Q se, where e' is the unit row vector (1, ... , 1) and se is the column vector of export shares. Since e' Q equals the row of column sums, we find in our case that ICE 2 = 0.161(0.38) + 0.20(0.21) + 0.238(0.17) + 0.072(0.24) = 0.16 where the figures in parentheses are the export shares from table 5-1. Each dollar of exports generates imports of 16 cents. The actual computations in our sample appear in table 7-7. Table 5-7. Trade and Input Use: Import Content of Final Demand (percent) Sector Sector 2 3 4 1. 2. 3. 4. Primary Light industry Heavy industry Services Total l44 3.4 6.7 1.6 3.6 3.9 5.2 5.8 2.1 9.2 11.7 2.8 2.0 2.4 0.7 16.1 20.0 23.8 7.2 '] Note: Q = AmR. Source: Data from tables 5-1 and 5-2. 142 THE EXPERIENCE OF INDUSTRIALIZATION MANUFACTURING AND THE BALANCE OF TRADE. Our final decomposi- tion computes the effect of manufacturing on the balance of trade. Each sector within manufacturing makes a positive contribution to the trade balance-that is, earns foreign exchange-by exporting, and it generates demands for foreign exchange for two uses: to purchase intermediate imports to sustain production whether for export or not, and to purchase final competitive imports of goods similar to those produced and exported by the sector (intraindustry trade). The various components can be summarized for any period as: BPC; = ll.E;- aM;w- ll.M( where BPC; balance of trade contribution by sector i ll.E; change in exports of sector i aM;w change in intermediate imports used by sector i ll.MfI change in final imports of commodities of sector i To allow comparisons of the various components across countries and over time, the figures are divided by the change in exports of total manu- facturing. The results of such calculations appear in table 5-8 (and in table 7-8). The first column of table 5-8 gives the change in exports for two sectors as a percentage of the change in total manufacturing exports. The second column gives the increase in imports of final goods from the same sector. The minus sign means that foreign exchange flows out. The third column gives the change in intermediate imports from all sectors generated by the expansion of total output (not just exports) in each sector. The net effect on the balance of trade is reported in the last column. In our example, the manufacturing sector in all its manifestations in the economy made a negative contribution to the trade balance. PRIMARY INPUTS. In the discussion so far, the emphasis has been on the material balance equations and, therefore, on the structure of demand for domestic production and imports. We have ignored the cost structure of production on the supply side, except for intermediate input requirements. Primary inputs-labor and capital-and payments for these services- value added-have not been considered at all. The ratios of value added, employment, and capital stock to gross output have often been considered as structural parameters reflecting income generation, labor productivity, and capital requirements associ- ated with production processes. Trends in these parameters mirror changes in the factor intensity of production as well as shifts in resource allocation that are the result of changes in production technology, relative prices, demand conditions, and so forth. Such trends crucially affect the sectoral pattern of factor use and value added. In turn, growth and THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 143 Table 5-8. Trade and Input Use: Manufacturing and the Balance of Trade (percentage of total manufacturing exports) Intermediate Contribution Final imports to balance Exports imports used of trade Sector (EJ (- M{) (- M;*w) (= BPC) Light industry 54 -8 -85 -39 Heavy industry 46 -42 -50 -46 Total manufacturing 100 -so -135 -85 Source: Data from tables 5-1 and 5-2. structural change generate increased demands for primary and intermedi- ate inputs and bring about changes in the volume and sectoral structure of value added. To sort out the various effects of output growth and changes in struc- tural parameters on factor use, the same decomposition framework used above can be applied. For example, the level of employment, L;, can be expressed by the product of the labor-output ratio (the inverse of labor productivity), l;, and output, X;: (5-13) The change in employment over time can be separated into two effects, one related to a change in the labor coefficient and the other to growth in production: (5-14) The change in output can be further decomposed, as in equation 5-11. The last term in equation 5-14 shows the effect on employment of a change in labor productivity. Labor coefficients usually decline during industrialization, so the last term will generally be negative. The change in labor coefficients reflects an aspect of technological change quite different from that captured by changes in input-output coefficients, although the effects may be related. The decomposition treats the changes in lover time as exogenous. No explanation of the source of such changes in labor productivity is offered. An adequate treatment of factor inputs and their contribution to growth requires an analytical framework different from that used so far. Such a framework will be sketched in part III. A Dynamic Computable General Equilibrium Model In the comparative statics analysis, there are no explicit links between the two points being compared. The input-output model for each year is essentially timeless and does not consider the dynamic forces that move 144 THE EXPERIENCE OF INDUSTRIALIZATION the economy from the earlier to the later state. To extend the framework requires an explicit model of the dynamic linkages. 19 In this section, we consider a dynamic computable general equilibrium model, which explic- itly includes market interactions that work through price mechanisms. 20 Such a model is used in chapter 11 to analyze the medium- to long-term effect on an economy's performance and structure of different choices of development strategy. A CGE model differs from the input-output model in two essential respects. First, many of the linear relationships in the input-output model are replaced by nonlinear functions that incorporate possibilities for sub- stitution in both production and demand. Second, and perhaps more important, the model simulates the workings of the markets for labor, goods, and foreign exchange and so embodies prices and market mecha- nisms as main elements of the economic system. Given the behavioral and technological assumptions, the model endogenously determines wages, profits, product prices, and the exchange rate; sectoral production, em- ployment, consumption, investment, exports, and imports; and the nominal flow of funds, including the accounts of the government, the private sector, and foreign trade. The price system in the CGE model is thus much more elaborate than in the input-output model and requires that the model be fully "closed" in the sense that all elements determining supply and demand are included. Conceptually, the dynamic CGE model used in chapter 11 consists of two distinct submodels: a static, within-period model and a between- period model that provides the needed intertemporallinks. The between- period model takes as exogenous all the variables solved in previous periods and generates all the variables that the within-period model takes as exogenous in the next period. The overall model is thus recursive in time; to solve for the current period requires only solutions from previous periods. The dynamic model uses two kinds of intertemporal equations: be- havioral equations, which depend on the history generated by the model, and time-trend equations, which impose exogenous trends on some vari- ables. The specification of the intertemporal equations depends largely on the focus of the analysis. A model with a five-year time horizon that studies year-to-year adjustments will be very different from a model that focuses 19. An extension of the static input-output model that explicitly incorporates capital accumulation but not incentives and markets is the dynamic input-output model. Chapter 2 of Dervis, de Melo, and Robinson (1982) has a useful presentation of the model. Kubo, Robinson, and Urata (1986) apply such a model to two of the countries in this study. There have also been attempts-often using linear programming formulations-to extend the basic input-output model to include more possibilities for choice and consideration of costs and prices. For a survey, see Blitzer, Clark, and Taylor (1975). 20. Early CGE models for planning started from the work ofJohansen (1960). For a survey of issues and models applied to developing countries, see Dervis, de Melo, and Robinson (1982) and Robinson (1986). THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 145 on cumulative processes affecting structural change in the long run, such as the model used in chapter 11. The recursive structure inherent in using separate within-period and between-period models implies that the dynamic model does not achieve full intertemporal equilibrium; that is, all expectations are not realized. Instead, the intertemporal model provides imperfect dynamic adjustments to intertemporal disequilibria that emerge from the within-period solu- tions of the static CGE model. During a long period, however, the dynamic model does converge to a growth path with equilibrium characteristics. The details of the dynamic specification are discussed in chapter 11. The Methodology of Model-Based Comparisons The models that have been presented in this chapter offer a variety of tools for conducting comparative analysis. Any useful comparative framework should allow one to ask "what if" questions that reveal interesting features of the economic structure being analyzed. For exam- ple, the simple input-output model makes it possible to capture the inter- dependence in an economy arising from the flow of intermediate inputs among sectors; one can describe the indirect as well as the direct demands for gross production arising from a given structure of final demand under the assumption of fixed input-output coefficients. The "what if" question in this case is simple: what are the total sectoral demands if the input- output coefficients are fixed? Nonetheless, even such a simple model is useful in comparing production across countries since it relates variations in production structures to differences in the structure of final demand and in the input-output coefficients. The comparative statics analysis, which decomposes changes in gross production over time and allocates them among various sources of de- mand, provides a somewhat more sophisticated comparative framework. Although apparently very simple, the underlying model embodies a host of behavioral as well as technological assumptions. Any comparative analy- sis based on this structure is contingent on these assumptions, which need to be clearly understood. As we shall see in chapters 6 and 7, the comparative statics framework permits a much richer analysis than is possible with the simple input- output model. The assumptions relating to technology and foreign trade, while strong, are theoretically justifiable, and they facilitate comparative analysis that focuses on the relative importance of internal and external influences on growth and structural change. Such an approach is crucial to understanding the process of development as it has occurred in the semi- industrial countries in the postwar period. CGE models differ from input-output models not only in degree but also in kind. The simpler models are valuable because they permit the separate analysis of particular effects, comparable in a sense to partial derivatives in mathematics: holding all other variables constant, what is the effect of 146 THE EXPERIENCE OF INDUSTRIALIZATION changes in one variable? But such a model cannot be used to ask a complex counterfactual question such as: what would have happened in a particu- lar country if it had followed a different development strategy from the one it actually pursued? Such a "what if" question is different in kind from the sort of decomposition analysis that can be pursued with simpler models. Counterfactual comparative analysis requires a model structure that can capture all the important effects that impinge on the question under consideration. The model serves as a simulation laboratory, providing a framework in which one can do controlled empirical experiments. It is used to make projections conditional on the specification of a set of exogenous variables and parameters that constitute an "interesting" ex- periment-for example, the specification of an alternative development strategy. The results are then compared with a base run or with actual history to see what would have happened differently. The realism of the comparison depends, first, on how well the model specification captures all the essential features of the economy and, second, on how accurately the set of variables and parameters that define the experiment is specified. A model suitable for counterfactual analysis can also be used for doing sensitivity analysis and simple decomposition analysis. Indeed, such ex- periments play an important part in validating a model, that is, in deter- mining whether a model's behavioral structure accurately reflects the important factors under consideration. Although a complex model can be used for simple analysis, one should always remember Occam's razor: when two competing hypotheses both explain the facts, pick the simpler. For comparative modeling, the principle might be restated as: use the simplest and clearest model sufficient to do the job. For example, if relative prices do not change, than a CGE model will not add much to the simpler input-output system. As social scientists, economists are generally hindered in their enquiries by being unable to do controlled experiments. They must make do with an analysis of the experiments that history has seen fit to provide. In this situation, an empirical model can serve two important functions. First, the model can be a tool for analyzing historical experience to sort out the relative magnitudes of various important effects. The model im- poses a theoretical structure on the historical data. It is a separate problem to test the validity of the theoretical structure-that is a task for statistical analysis. In the work described in the rest of part II, the opportunities for such statistical analysis are limited because of the relatively few observa- tions available over time. Second, the model can provide an empirical laboratory for doing con- trolled experiments. Since history's experiments are limited, the model can be designed to capture the important structural features of an economy and then be used for experimental work. Such counterfactual experiments can provide important insights into the nature of the processes under THE METHODOLOGY OF MULTISECTOR COMPARATIVE ANALYSIS 147 study-insights otherwise not obtainable. Although any structural model must of necessity simplify reality, the complexity that can be captured in such an empirical model often defies analysis by means of theoretical models. The potentially important interactions in even a simple mul- tisector, general equilibrium model usually cannot be sorted out with partial equilibrium tools, although such theoretical analysis is useful in pointing to the principal effects that need to be evaluated empirically. Inevitably, models that focus on one set of issues must neglect others; it is neither feasible nor desirable to try to capture all the complexity of a developing economy in one model. The models described in this chapter are not intended to provide an all-purpose framework suitable for analyz- ing any problem, but rather a framework suitable for analyzing the long-term features of industrialization, growth, and structural change. Quite different models would be required to analyze problems of short- term adjustment or income distribution. Given the need for simplification, a model must be constructed from basic building blocks that capture the important forces at work. A step-by- step approach, using a variety of models, seems best for analyzing com- plex, dynamic processes. Part I laid the groundwork for such an approach by reviewing the experience of a large number of countries in an effort to identify certain universal features or stylized facts that characterize the process of industrialization. The multisector models described in this chapter provide the framework for a more detailed analysis of the experi- ence of a few economies. These models focus on intersectorallinkages and can provide a more complete picture of industrialization and growth than is possible in a more aggregate analysis. 6 Trade Strategies and Growth Episodes YUJI KUBO JAIME DE MELO SHERMAN ROBINSON IN CHAPTER 4, systematic variations in the development patterns of semi-industrial countries were found to be associated with differences in income, country size, resource base, and trade orientation. Though sug- gestive, these findings do not enable us to identify the impact of particular policy choices on patterns of development. The purpose of this chapter is to link country policy choices with the structural changes that occur during industrialization while taking into account initial conditions that delineate the boundaries of what is feasible. The underlying analytical framework is the input-output methodology discussed in chapter 5. Focusing on the sources of growth from the demand side, we turn in this chapter to the connection between different development strategies-particularly trade strategies-and the aggregate structure of demand. Chapter 7 probes the relation between growth and various measures of intersectorallinkages and structural change at a more disaggregated level. Development strategies have been categorized in various ways. They can be described in terms of factor accumulation; for example, natural- resource-based intensive, physical capital intensive, or human capital intensive. Or they can be described in terms of income distribution; for example, "grow first, redistribute later" or "redistribute first, grow later." Strategies have also been defined by their sectoral emphasis; for example, agriculture-led or industry-led development. Finally, strategies have been defined in terms of trade policies, as, for example, export-led growth or import substituting industrialization. In this chapter, we concentrate on development strategies defined in terms of trade policies. When we speak of a change in development strategy, we shall be referring primarily to a change in foreign trade regime involving instruments such as the exchange rate and quantitative and price controls on imports and exports. We shall also be looking at policies other than trade policies and shall use the term policy regime to refer to the application of the panoply of government instruments to the economy. Recall that in chapter 3, "typical" countries were defined by trade orienta- 148 TRADE STRATEGIES AND GROWTH EPISODES 149 tion as well as size: primary versus manufacturing exporters and small versus large countries. In chapter 4 (table 4-3), this categorization was further refined for the semi-industrial countries by distinguishing four trade strategies: primary-oriented, inward-oriented, manufactur- ing-oriented, and neutral (the last being the residual group which included Colombia). In that discussion (with the exception of Colombia), the economies in our sample were classified as either inward-oriented or industry-oriented on the basis of long-term trends rather than specific policy choices. We examine the effect of changes in trade strategies with the input-output methodology introduced in chapter 5. Because input- output data are available for only a few benchmark years in each econ- omy, our analysis of the impact of different policy regimes has to take place for "episodes" defined by the years for which these input-output data are available. We begin with a brief discussion of the aggregate performance of our sample of nine economies during the postwar period. This is followed by a comparison of the performance of the sample with that of other semi- industrial countries. Next, we show that in most cases the data-determined benchmark years defining the episodes are characterized by coherent policy regimes and so can be seen as reflecting particular trade strategies. This allows us, later on in the chapter, to use an episode as the unit of analysis for identifying typologies of trade strategies and to examine the growth patterns of the nine economies under different trade strategies. Finally, we examine the sequence of trade strategies through time. Initial Conditions and Aggregate Performance Table 6-1 presents some basic indicators for the benchmark years for each economy in the sample. These were years of rapid expansion in world trade, especially in manufactures, and of generally good performance by developing countries. For example, from 1960 to 1970 the middle-income oil-importing countries achieved an average annual growth rate of GDP of 5.9 percent, compared with an average rate of 5.1 percent for the indus- trial market economies (World Bank 1981, table 2, pp. 136-37). For the same period, total merchandise exports for the middle-income oil- importing countries grew at 6.3 percent a year and imports by 7.7 percent a year (World Bank 1981, p. 148). The data in table 6-1 indicate that our sample performed well, even by the standards of middle-income countries. The growth rates of per capita GNP are generally respectable and rising during the period. A number of factors contributed to their growth performance. Structural change was important and will be a major theme of the analysis below. Also, average investment rates generally rose, sometimes dramatically. The two excep- tions, Norway and Yugoslavia, started with very high rates, which then fell to about 25 percent. Korea and Taiwan benefited from substantial foreign assistance in the early periods, which was later replaced by direct Table 6-1. Comparative Economic Indicators for Sample Economies Population Per capita GNP' Average Average Percentage of GDP Number annual annual of growth growth Primary Industry people rateb Level rateb Invest- value value ..... Economy Year' (million) (percent) (dollars) (percent) Imports Exports ment added added "' 0 Colombia 1953 12.5 274 - - 15.3 40.4 18.0 1966 18.4 3.0 330 1.4 15.1 12.1 20.4 32.4 22.4 1970 20.6 2.9 369 2.8 15.8 14.2 21.5 30.7 23.0 Mexico 1950 26.3 380 13.9 14.1 13.5 23.1 24.6 1960 36.0 3.2 479 2.9 12.8 11.3 20.1 17.5 26.7 1970 50.4 3.4 670 3.4 10.1 8.1 19.6 12.7 30.9 1975 60.0 3.5 751 2.3 10.9 7.7 24.4 1.2 32.6 Turkey 1953 22.8 239 - 12.4 51.1 12.1 1963 29.7 2.7 319 2.9 10.3 5.5 15.4 40.7 19.0 1968 33.5 2.4 377 3.3 7.5 5.3 18.0 32.3 24.4 1973 38.3 2.7 461 4.1 10.0 7.6 19.0 31.0 24.5 Yugoslavia 1962 18.8 469 17.1 16.0 30.9 24.4 40.1 1966 19.6 1.0 581 5.5 20.5 19.5 25.5 24.6 40.1 1972 20.8 1.0 781 5.1 24.1 22.0 26.3 16.8 42.2 Japan 1914 52.4 265 17.3 17.7 16.6 28.3 28.8 1935 69.2 1.3 416 2.2 21.5 21.5 18.1 16.6 34.4 1955 89.0 500 10.1 10.7 24.7 25.0 26.5 1960 94.1 1.1 753 8.5 10.6 11.1 33.7 15.1 38.2 1965 98.9 1.0 1,159 9.0 9.3 10.8 32.9 10.7 37.9 1970 104.3 1.1 1,897 10.4 9.8 11.2 39.4 7.2 43.2 Korea 1955 21.6 131 10.1 1.7 12.0 48.1 13.1 1963 27.0 2.8 149 1.6 16.4 4.8 18.6 46.8 16.9 1970 31.4 2.2 250 7.7 24.9 14.8 27.3 32.4 25.5 1973 32.9 1.6 323 8.9 35.0 31.7 26.0 28.8 29.7 Taiwan 1956 9.2 203 - - 15.9 29.8 24.3 1961 11.0 3.6 231 2.6 19.9 12.8 19.8 29.3 25.4 1966 12.8 3.1 305 5.7 21.5 20.6 23.1 24.2 29.8 1971 14.8 2.9 426 6.9 34.2 36.8 26.1 14.8 39.3 ...... Israel 1958 2.0 1,067 - - 26.8 13.0 31.7 v, 1965 2.6 1,587 31.9 18.9 27.1 8.2 36.3 ...... 3.6 5.8 1972 3.1 2.3 2,317 6.5 40.1 27.3 28.8 6.3 38.8 Norway 1953 3.4 1,171 - - 29.5 15.5 35.0 1961 3.6 0.9 2,028 7.1 42.6 39.7 31.1 9.9 32.6 1969 3.9 0.8 2,769 4.0 38.5 41.2 25.0 6.2 34.0 - Not available. a. Except for prewar Japan, the economy experiences fall into the period from the 1950--53 Korean war to the 1973 oil crisis. The special characteristics of this period have been discussed in earlier chapters, especially chapter 4. b. Growth rates refer to the period that starts from the previous benchmark year. c. In real 1970 dollars. Sources: World Bank (1976) and World Bank data. See Kubo and Robinson (1984, table 1). 152 THE EXPERIENCE OF INDUSTRIALIZATION foreign investment (in response to policy incentives and good perfor- mance). Except for Mexico, population growth rates were constant or declining. And the stable economic and political environment that pre- vailed between the 1950-53 Korean war and the oil crisis of 1973 pro- vided a favorable environment for international trade and economic growth.' In addition to the initial diversity of these economies in population, per capita income, and production structure, there was great diversity in the role of foreign trade. The neutral and inward-oriented countries (Co- lombia, Mexico, and Turkey) and the largest country (postwar Japan) started with relatively low shares of imports and exports in GDP and experienced little change during the period. 2 The outward-oriented econo- mies (Israel, Korea, and Taiwan) had widely differing shares of trade in GDP initially; all three dramatically increased these shares during the period. Pre-1945 Japan and Yugoslavia occupy an intermediate position. They started with similar shares, which then rose moderately but signifi- cantly. Table 6-2 presents the sectoral composition of imports and exports. The similarity of the economies in the structure of their imports is significant. Except for Japan, they depended greatly on imports of heavy industrial products (including intermediates and machinery) throughout the period. None was a large importer of consumer goods. (Only Israel and Korea had initial import shares of more than 20 percent in this category.) Thus, by the initial year, the economies in the sample had largely completed the "easy" or "primary" stage of import substitution, in which imported consumer goods are replaced by domestic substitutes. The diversity on the export side is greater. Three countries (Colombia, Mexico, and Turkey) are primary exporters; Turkey moved toward a significant share of manufacturing exports only at the end of the period. Korea and Taiwan are the now classic cases of rapid growth led by manufacturing exports. The other countries show a variety of intermedi- ate patterns. The long-run cross-country model in chapter 3 attributes the rising share of manufacturing in value added to three causes: shifts in the composition of demand, changes in input-output technology that lead to increased demand for intermediate inputs, and changes in comparative advantage. We shall explore the second and third cause in more detail in chapter 7. Here, we analyze the first cause by examining the role of shifts in the structure of demand, using the cross-country model as a benchmark with which to compare the economies in our sample. 1. W. A. Lewis (1980) notes the unusually high growth in trade between industrial and developing countries during the period and speculates on the implications of, and remedies for, a slowdown in what he calls "the engine of growth." 2. Norway is both the most developed and the most open of the sample economies and had little change in trade shares during the period. TRADE STRATEGIES AND GROWTH EPISODES 153 Sources of Change in Demand and Output The input-output model is used to trace the indirect linkages by which final demand for the output of one sector generates demand for the output of other sectors. Differentiating among categories of demand, we decom- pose the change in a sector's output during a period into changes arising from different sources of demand, taking into account indirect effects as well. The decomposition formula was discussed in detail in chapter 5. The decomposition used in this and the next chapter differs slightly from the presentation in chapter 5 in that the same average domestic supply ratios are assumed for all types of demand. The decomposition allocates the change in total output to four different factors: domestic demand expan- sion (DD), export expansion (EE), import substitution (Is), and changes in input-output coefficients (10). The first two terms (DD and EE) arise from direct changes in demand, whereas the third and fourth terms (IS and 10) arise from changes in coefficients. The treatment of exports and imports is not symmetrical. Import substitution is defined in terms of changes in import coefficients. These can be seen as technical coefficients, which seems especially appropriate for semi-industrial economies in which most imports are capital goods and intermediates that are difficult to replace with domestic products. Other treatments of exports and imports have been used in the literature and are discussed in chapter 5. The total change in aggregate output equals the sum of the changes in the individual sectors and can thus be decomposed either by sector or by category of demand. In chapter 5, we presented a schematic with one row for each sector and five columns giving the change in sectoral output and its decomposition into the four separate terms. Again, DDt + EEt + 1St +lOt = dXt DD2 + EE2 + IS2 + I02 = dX2 DDn + EEn + ISn + IOn = dXn IDD;+ IEE; + !.IS; + IIO; = IdX; Reading down the columns gives the sectoral composition of each demand category; reading across the rows gives the decomposition of changes in sectoral demand by demand category. To make comparisons across coun- tries and time periods, it is convenient to divide the entire table by the sum of sectoral output changes ( IdX;) so that all components across sectors and demand categories sum to 100. Alternatively, it is sometimes conve- nient to divide the rows by the change in each sectoral output (dX;) and then look at the shares of the contribution of each demand category to the change in sectoral output. Table 6-2. Composition of Imports and Exports of Sample Economies for Selected Years (percent) Import shares Export shares Pri- Light Heavy Ser- Pri- Light Heavy Ser- ..... mary industry industry vices mary industry industry v, Economy Year VICeS '""" Colombia 1953 0.2 17.0 77.6 5.2 86.0 0.8 1.2 12.0 1966 14.8 11.1 61.8 12.3 71.1 6.3 6.0 16.6 1970 4.6 13.1 69.1 13.2 78.3 6.7 1.9 13.1 Mexico' 1950 11.7 16.6 71.7 0.0 54.3 27.2 1.5 17.0 1960 10.2 12.5 77.4 0.0 52.7 31.1 4.8 11.5 1970 8.6 15.1 76.3 0.0 45.4 19.2 21.6 13.9 1975 11.9 10.4 77.6 0.0 34.4 18.0 30.9 16.7 Turkey 1953 6.6 6.3 85.3 1.8 78.1 3.0 4.5 14.4 1963 17.3 7.3 65.9 9.6 58.2 8.2 3.8 29.8 1968 7.9 5.1 75.3 11.8 45.7 19.6 3.1 31.6 1973 12.0 3.0 78.8 6.2 26.9 27.4 6.8 39.0 Yugoslavia 1962 17.1 10.4 68.3 4.1 11.9 21.9 30.2 35.9 1966 20.3 11.3 64.9 3.5 14.9 20.7 33.0 31.4 1972 15.8 14.4 66.8 3.0 11.2 25.3 40.2 23.2 Japan 1914 45.0 10.9 40.0 4.0 10.7 43.8 7.0 38.5 1935 53.7 10.9 28.0 7.3 2.8 42.2 12.2 42.8 1955 62.5 18.0 21.1 -1.7 3.0 45.6 28.5 22.8 1960 53.5 14.2 33.4 -1.1 3.7 44.3 40.0 12.0 1965 56.8 17.5 26.7 -1.0 2.2 26.7 58.4 12.7 1970 49.7 16.1 26.5 7.7 1.1 18.4 58.6 22.0 Korea 1955 1.7 38.5 59.6 0.1 23.8 17.2 2.4 56.6 1963 28.8 13.3 55.7 2.2 20.2 34.2 14.3 31.3 1970 24.1 19.6 55.7 0.7 8.7 55.0 14.5 21.8 1973 17.0 19.0 62.5 1.5 3.8 48.5 31.0 16.8 Taiwan 1956 23.4 11.4 59.8 5.4 6.6 78.8 9.4 5.2 1961 22.1 16.4 56.0 5.5 7.1 66.0 10.0 16.9 1966 18.6 16.0 60.5 4.9 9.1 49.4 20.0 21.6 1971 20.3 17.6 58.6 3.5 4.9 48.6 34.0 12.6 ..... Israel 1958 9.7 21.8 45.8 22.7 21.3 33.4 10.5 34.8 "' 1965 7.0 17.8 47.0 28.2 12.7 30.9 9.6 46.8 "' 1972 3.2 16.6 67.3 13.0 10.2 35.2 14.4 40.2 Norway 1953 13.7 15.0 46.4 25.0 6.9 20.3 20.4 52.4 1961 11.3 16.3 51.0 21.6 5.0 16.9 25.2 53.0 1969 10.1 20.2 50.7 19.0 3.2 15.9 34.3 46.6 a. Data for services in Mexico refers to net exports and are given under the export column. Services imports are defined as zero. Source: World Bank data; described in Kubo (1983). Table 6-3. Sources of Growth for Archetypal Economies (percent) Per capita Large (L) Small manufacturing (sM) Small primary (sP) income and sector DD EE IS 10 Total DD EE IS 10 Total DD EE IS IO Total $140-$280 Primary 11.0 2.4 .0 -.8 12.6 9.4 5.7 -1.9 -.5 12.7 10.0 12.9 .0 -.5 22.4 Light industry 19.8 2.7 1.1 1.7 25.3 18.2 8.5 1.1 1.5 29.3 18.4 1.8 1.8 1.7 23.7 Heavy industry 9.5 1.5 3.0 1.3 15.3 2.5 1.2 -.6 .2 3.3 4.0 .8 2.6 .8 8.2 Services 42.9 2.6 .5 .8 46.8 44.0 10.1 -.7 1.3 54.7 40.9 3.2 .5 1.1 45.7 Total 83.2 9.2 4.6 3.0 100.0 74.1 25.5 -2.1 2.5 100.0 73.3 18.7 4.9 3.1 100.0 $280-$560 Primary 8.8 1.5 -.6 -.7 9.0 6.3 4.5 -2.0 -.9 7.9 7.7 11.4 -.4 -.5 18.2 ..... Light industry 20.3 3.3 .6 1.4 25.6 18.2 9.8 1.2 1.1 30.3 18.9 2.5 1.3 1.4 24.1 Heavy industry 12.7 2.8 2.0 1.6 19.1 3.6 2.9 2.8 .5 9.8 6.9 1.3 3.1 1.1 12.4 "" 0\ Services 42.8 2.9 .2 .4 46.3 41.9 10.3 -.3 .1 52.0 40.7 3.3 .4 .9 45.3 Total 84.6 10.5 2.2 2.7 100.0 70.0 27.5 1.7 .8 100.0 74.2 18.5 4.4 2.9 100.0 $5 60-$1,120 Primary 7.3 1.0 -1.1 -.8 6.4 4.3 3.1 -2.1 -.7 4.6 6.3 9.0 -.7 -.7 13.9 Light industry 20.5 4.0 .2 .9 25.6 17.5 10.6 1.3 .9 30.3 18.8 3.4 .9 1.1 24.2 Heavy industry 15.3 4.3 1.4 1.8 22.8 6.5 5.5 4.3 .9 17.2 10.9 1.6 3.3 1.6 17.4 Services 42.0 3.3 -.1 .0 45.2 39.4 9.2 -.1 -.6 47.9 40.8 3.0 .3 .4 44.5 Total 85.1 12.6 .4 1.9 100.0 67.7 28.4 3.4 .5 100.0 76.8 17.0 3.8 2.4 100.0 $1,120-$2,100 Primary 6.2 .8 -1.2 -.6 5.2 3.2 1.8 -1.7 -.5 2.8 5.5 5.9 -.6 -.7 10.1 Light industry 20.2 5.1 -.1 .6 25.8 16.4 10.6 1.4 .9 29.3 18.1 4.5 .7 1.0 24.3 Heavy industry 17.4 5.6 1.0 2.1 26.1 10.2 8.7 5.4 1.4 25.7 15.7 1.5 3.3 2.4 22.9 Services 40.2 3.6 -.3 -.6 42.9 36.1 7.0 .2 -1.1 42.2 40.4 2.4 .2 -.3 42.7 Total 84.0 15.1 -.6 1.5 100.0 65.9 28.1 5.3 .7 100.0 79.7 14.3 3.6 2.4 100.0 Source: World Bank data; computations from cross-country model. See chapter 3 for description. TRADE STRATEGIES AND GROWTH EPISODES 157 Long-Run Comparisons Tables 6-3 and 6-4 present the decomposition results in the same format as the schematic at the four-sector level. Table 6-3 does this for the cross-country model as applied to three different country patterns or archetypes-large (L), small manufacturing exporter (sM), and small pri- mary exporter (sP)-and four per capita income intervals. The results are described in more detail in chapter 3; see especially table 3-10 and figure 3-9. Table 6-4 presents the results for the sample. Figures 6-1 and 6-2 summarize the sources of growth for the whole economy, as well as the changes in sectoral incremental output shares (LlX;II.:lX;) as income rises for the three archetypes. The variations in decomposition for each arche- type are very smooth, which reflects the cumulative effect of the indus- trialization processes described in chapter 3. The relative contribution to total growth of the primary sector declines in each archetype as income rises, while that of heavy industry (intermediates and machinery) rises as a result of the combined influence of the falling share of food in consump- tion, the increased share of investment in GDP, and the increased use of fabricated goods as intermediate inputs throughout the economy. The effects of these changes in the composition of final demand are sup- plemented on the supply side by the accumulation of physical capital and skills, which shifts comparative advantage from goods using primary resources to manufactured products. The three archetypes generated with the cross-country model are useful in delineating the role of initial conditions in determining the pattern and timing of industrialization. The breakdown between large and small coun- tries takes into account the influence of size on the openness of an econ- omy, while the breakdown between primary and manufacturing exporters takes into account the effect of natural resources on a country's compara- tive advantage. As figure 6-2 illustrates, all three converge toward a common industrial structure as they grow; the sectoral contributions to growth at the income range $1,120-$2,100 are very similar. 3 Figure 6-3 (derived from table 6-4) presents the sectoral contributions to total change in output for the sample economies. These results can be compared with those for the archetypes shown in figure 6-2. The econo- mies fall into distinct groups. Korea and Taiwan have strikingly similar patterns, distinguished by the large contribution of light industry. The other open economies, Israel and Norway, also have very similar patterns but are more characteristic of advanced countries in that they have a large contribution by services. 4 Postwar Japan and Yugoslavia show a large 3. See figure 6-2 and table 6-3. For a further discussion of this convergence, also see Chenery and Taylor (1968) and chapter 3. 4. The large contribution by services in the case of Israel is partly a reflection of large capital inflows; in the case of Norway, it reflects a long-standing comparative advantage in shipping. 158 THE EXPERIENCE OF INDUSTRIALIZATION Table 6-4. Sources of Growth for Sample Economies (percent) Economy and Growth sector rate DD EE IS 10 Total Colombia (1953-70) Primary 4.5 12.7 9.1 0.5 0.1 22.4 Light industry 6.8 15.3 1.5 1.5 3.0 21.3 Heavy industry 11.1 9.0 0.8 4.5 1.7 16.0 Services 5.5 36.7 2.6 0.3 0.7 40.3 Total 5.9 73.7 14.0 6.8 5.5 100.0 Mexico (1950-75) Primary 4.8 12.8 0.7 -0.3 -0.5 12.7 Light industry 6.0 17.7 0.4 0.5 0.7 19.3 Heavy industry 10.8 16.7 1.8 2.9 1.3 22.7 Services 6.4 43.7 0.7 0.4 0.5 45.3 Total 6.5 90.9 3.6 3.5 2.0 100.0 Turkey (1953-73) Primary 2.5 14.9 1.2 0.2 -4.5 11.8 Light industry 6.7 15.3 2.1 0.4 1.7 19.5 Heavy industry 9.6 18.8 0.9 1.6 3.4 24.7 Services 6.7 38.3 2.8 0.2 2.7 44.0 Total 5.9 87.3 7.0 2.4 3.3 100.0 Yugoslavia (1962-72) Primary 2.6 10.1 3.9 -3.2 -4.6 6.2 Light industry 11.0 17.7 6.2 -2.6 1.3 22.6 Heavy industry 13.6 23.6 12.2 -6.1 4.4 34.1 Services 8.8 33.0 5.2 -1.3 0.2 37.1 Total 8.7 84.4 27.5 -13.2 1.3 100.0 Japan (1914-35) Primary 1.9 7.6 2.8 -2.3 2.3 10.4 Light industry 4.6 15.8 10.5 0.2 -0.4 26.1 Heavy industry 8.1 15.2 4.4 1.9 -3.3 18.2 Services 4.2 35.2 9.0 -0.2 1.3 45.3 Total 4.1 73.8 26.7 -0.4 -0.1 100.0 contribution by heavy industry but probably for different reasons. As is typical for a socialist country, Yugoslavia pursued policies strongly favor- ing heavy industry. Japan is a large country which was completing its transformation into a mature industrial economy during this period. Prewar Japan, Mexico, and Turkey had more balanced sectoral contribu- tions to growth-a pattern one would expect of large economies. Co- lombia is the only country in the sample that followed a pattern character- istic of a primary-oriented economy. Role of Exports Figure 6-1 shows that one of the main differences in the sources of growth among the three archetypes is the relative importance of exports. As expected, small countries rely more heavily on exports to make up for TRADE STRATEGIES AND GROWTH EPISODES 159 Economy and Growth sector rate DD EE IS 10 Total japan (1955-72) Primary 2.2 4.4 0.5 -1.5 -1.9 1.5 Light industry 8.6 14.4 2.0 -0.7 1.1 16.8 Heavy industry 18.0 30.9 8.4 -0.1 3.3 42.5 Services 11.4 35.7 3.0 -0.8 1.3 39.2 Total 11.5 85.4 13.9 -3.1 3.8 100.0 Korea (1955-73) Primary 5.7 12.0 3.0 -1.7 -2.5 10.8 Light industry 13.6 19.7 15.1 0.0 -1.9 32.9 Heavy industry 22.1 11.1 10.7 1.4 1.9 25.1 Services 10.3 25.6 6.2 0.2 -0.8 31.2 Total 11.2 68.4 35.0 -0.1 -3.3 100.0 Taiwan (1956-71) Primary 7.1 8.8 5.3 -2.0 -1.8 10.3 Light industry 13.6 12.7 17.5 0.6 2.0 32.8 Heavy industry 22.5 10.2 13.5 2.4 1.0 27.1 Services 9.7 23.6 7.1 0.1 -1.0 29.8 Total 12.0 55.3 43.4 1.1 0.2 100.0 Israel (1958-72) Primary 6.4 2.6 3.6 -0.3 -0.4 5.5 Light industry 11.2 11.3 12.0 -2.0 1.2 22.5 Heavy industry 14.3 18.7 6.3 -6.6 2.6 21.0 Services 8.9 39.3 13.9 -1.6 -0.6 51.0 Total 9.9 71.9 35.8 -10.5 2.8 100.0 Norway (1953-69) Primary 2.5 3.8 2.4 -1.7 0.3 4.8 Light industry 3.7 14.0 6.2 -5.7 3.0 17.5 Heavy industry 7.2 10.7 15.6 -2.2 2.2 26.3 Services 4.8 31.9 21.5 -2.7 0.7 51.4 Total 4.7 60.4 45.7 -12.3 6.2 100.0 Source: World Bank data; described in Kubo (1983). the limited size of the domestic market, especially at an early stage of industrialization. Moreover, small manufacturing economies exhibit an increasing reliance on exports for growth throughout the transition. The experiences of our sample are compared with those of the three archetypes in figure 6-4. The length of the bars shows the contribution of export expansion to growth as a percentage of the total change in gross output. The bars also indicate the contributions of the four aggregate sectors: primary, light industry, heavy industry, and services. The archetypal econ- omies behave as expected: in the L pattern, contributions by all sectors are balanced; in the SP pattern, a heavy bias toward primary exports is evident; and in the SM pattern, manufactured exports-especially light manufactures-play a large role. Figure 6-1. Sources of Growth for Archetypal Economies Import substitution (Is) 6 ,• SM ~---~--::::=:-----=-_--· r . . . . . . . . . . . ............... SP ... ..................... ' .................................... ,,' .......... .............. _ / ,' -- ___ ..,.,~ ----------------:::~~------ -- ----· ,,,'' -2 -4L___________L __ _ _ _ _ _ _ _ _ _L __ _ _ _ _ _ _ _ ~ 140-280 280-560 560-1,120 1,120-2,100 Per capita income (dollars) Input-output change (10) 4 , ------- .......... -......_ ~------- SP '',,,'',,,',,, -......_..._ _ _ _ _ _ _ _ _ _ L ..... ____ _ ............................... SM 140-280 280-560 560-1,120 I, 120-2, I 00 Per capita income (dollars) Domestic demand (DD) 90 § g 85 ----- -------------- L :-£ 80 ~ 8 J::b.-_-_-_-__-_-_-_-_-_-_-_- ------------ ...... __ _ 65L___________L __ _ _ _ _ _ _ _ _ _L __ _ _ _ _ _ _ _ ----- ~ 140-280 280-560 560-1,120 1,120-2,100 Per capita income (dollars) Export expansion (EE) 29 -----------------------L § 24 ] ·~ 8 19F------------ ~ ~ ~ 14 --------------------------------------- 140-280 280-560 560-1,120 1,120-2,100 Per capita income (dollars) 160 Figure 6-2. Sectoral Contributions to Growth for Archetypal Economies Large (L) 140-280 280-560 560-1,120 1,120-2,100 Small manufacturing (sM) Vi' .... ..$ 140-280 0 ~ <1.l E 280-560 0 u .5 ... 0: ·c.. 560-1,120 0: u .... <1.l 1,120-2,100 ~ Primary (sP) 140-280 280-560 560-1,120 1,120-2,100 0 20 40 60 80 100 Percentage contribution Key: • Primary II Heavy industry • Light industry II Services 161 162 THE EXPERIENCE OF INDUSTRIALIZATION Figure 6-3. Sectoral Contributions to Growth for Sample Economies Colombia Mexico Turkey Yugoslavia Japan (prewar) Japan (postwar) Korea Taiwan Israel Norway 0 20 40 60 80 100 Percentage contribution Key: • Primary • Light industry ~ Heavy industry 0 Services As with the sectoral contributions to total output growth, Korea and Taiwan are again very similar and could be classified as extreme SM economies with relatively larger contributions from heavy industry than in the average SM pattern. Israel and Norway are comparable but show larger contributions by service exports. Mexico and Turkey are the most closed economies, followed by Colombia, which had significant primary exports and is the closest to the SP archetype. Yugoslavia and prewar Japan had a similar aggregate role for exports, but Yugoslavia had a stronger emphasis on heavy industry. Prewar Japan is the closest to the SM archetype. Postwar Japan again looks like a large industrial economy with a significant role for trade, especially in heavy industry. Israel, Korea, TRADE STRATEGIES AND GROWTH EPISODES 163 Figure 6-4. Contribution of Export Expansion to Total Output Change of Sample Economies Colombia Mexico Turkey Yugoslavia Japan (prewar) Japan (postwar) Korea Taiwan Israel Norway Archetypes ($280-$560)' L SM --·· _._... SP 0 10 •• 20 30 40 50 Percentage contribution Key: • Primary EiJ Heavy industry • Light industry 0 Services a. Only one income range is plotted for the archetypes. This is reasonably close to the income range for most of the sample economies (see table 6-1). 164 THE EXPERIENCE OF INDUSTRIALIZATION prewar Japan, and Taiwan show a dominant contribution by manufac- tured exports, led by light industrial exports which are largely labor- intensive. Postwar Japan, Norway, and Yugoslavia favored heavy indus- trial exports. Comparing these groups of economies with the cross-country model, we see that the model underestimates the contribution of heavy industry at the income range in question, particularly for Korea, Taiwan, Yugoslavia, and postwar Japan. For the latter two, this can be partly explained by their particular circumstances: Yugoslavia followed a socialist development strategy emphasizing the production of capital goods, while Japan was restoring its industrial base after the war. But for Korea and Taiwan, the contribution of both light and heavy industry far exceeded that predicted for the typical trade patterns of the model. 5 Policies and the choice of development strategy must have accounted for the early industrialization of these two economies. Finally, prewar Japan, Mexico, and Turkey displayed sectoral contributions close to those predicted from the L pattern in the cross-country model. Colombia is the only country that closely resembles the primary-oriented archetype. For both overall sectoral contributions to growth (figures 6-2 and 6-3) and the role of exports (figure 6-4), a comparison of the experience of the sample economies with the cross-country model confirms some of our earlier conclusions. Colombia follows the SP industrialization pattern, while postwar Japan, Mexico, and Turkey follow the L pattern. In Mexico and Turkey, however, the contribution of export expansion is only about half that predicted by the cross-country model. These countries were much more autarkic than typical countries of their size and per capita income. For Japan, the period 1955-70 was one of import liberalization rather than of import substitution as assumed by the cross-country model. In contrast, prewar Japan fits quite closely the SM pattern, which raises the question of whether this was also true for Mexico and Turkey during the prewar period. Although these countries, like Japan, pursued development strategies based on the substitution of light manufactures for imports, their development strategies were generally more autarkic. They did not have close colonial ties, as did Japan, that permitted an especially advan- tageous exchange of raw materials for manufactured products. The remaining economies in the group are outliers on the high side in comparison with the cross-country pattern. Although the choice of bench- mark years may be partly responsible, the discrepancy is caused both by the effects of the development strategy for these economies in relation to the "average" development strategy embodied in the cross-section regres- sions and by special exogenous circumstances, some of which were men- tioned earlier. Since the cross-country model was designed to explore the 5. Korea and Taiwan started the period with per capita incomes of $131 and $203, respectively (see table 6-1). TRADE STRATEGIES AND GROWTH EPISODES 165 process of industrialization that takes place mostly in the transition from a per capita income level of $280-$1,120, it is not surprising that Israel and Norway-with per capita incomes exceeding $1,000 in the initial year- deviate from the predicted pattern. But for Korea, Taiwan, and Yugosla- via, an explanation must lie in a further scrutiny of their choices of development strategy. Trade Policy Regimes The analysis so far has been concerned with long-run changes. To consider the effect of different development strategies on the nature and structure of growth, we need to look at episodes that can be characterized by a distinct development strategy. The choice of episodes in the sample economies was determined by the availability of input-output data for benchmark years. We are lucky in that, for our sample, most of the episodes do correspond to periods that reflect a specific development strategy. The benchmark years often fall at or close to a time when an economy underwent a significant shift in its policy regime. This makes it fruitful to use the episodes to explore the relation between different development strategies and the nature of growth in the economy. A country's development strategy is determined by its choice of policies and its institutional environment. It is also affected by exogenous events such as wars, droughts, and the state of the world economy. To charac- terize a period as reflecting a particular choice of development strategy implies that the policies chosen reflect a coherent policy regime. A policy may produce different results given different initial conditions, institu- tional settings, and conditions in the world economy. But within each economy at a given time, one can sensibly characterize policies as being consistent or inconsistent with a development strategy. In chapter 4, the semi-industrial countries were categorized by three long-term development strategies: inward-oriented, outward-oriented, and neutral. In examining episodes within our smaller sample, it is useful to refine these categories somewhat. We distinguish three trade strategies, one that is inward-oriented and two that are outward-oriented. The first is an import substitution strategy and is characterized by policies that bias production incentives against exports and toward the home market. One of the two outward-oriented strategies, export promotion, is characterized by policies that give roughly equal and positive incentives to production both for export and for the substitution of imports. In contrast with the import substitution strategy, export promotion entails no bias against sales to foreign markets compared with sales in the domestic market. 6 The second outward-oriented strategy, trade liberalization, is characterized by 6. In the Bhagwati-Krueger terminology, effective exchange rates for exporting and for import substituting activities are close to one another. See Krueger (1978) and Bhagwati (1978). 166 THE EXPERIENCE OF INDUSTRIALIZATION policies that give negligible incentives to both import substituting and exporting activities. This case corresponds to a relatively free trade regime, with few quantitative controls or price-related measures. This classification of development strategies omits important policy and institutional influences that are reflected in the contributions to growth discussed later in the chapter. Particularly important is the role of policies affecting the functioning of factor markets. For example, a change in development strategy, whereby incentives are shifted toward exporting, will not result in increased exports if labor mobility is low and the required investment in new activities is not forthcoming. Moreover, any one of the strategies defined above can be accomplished by a wide variety of policy regimes. Thus, an export-oriented strategy can be achieved by direct quantitative intervention in the market (for example, by applying export targets at the firm level) or by providing price incentives to export sales. Establishing the "equivalence," however defined, of different policy re- gimes for a development strategy is difficult, even for the simple cases constructed in the theoretical literature/ Episodes and Policy Phases For each economy in the sample, figure 6-5 relates several episodes framed by two benchmark years for the important policy phases. When no major shifts in policy occurred, no break is indicated. This summary figure seeks to establish whether the benchmark years imposed by the availability of input-output information are unusual in some important way; for example, did they fall in a time of war, acute political instability, poor harvest, or major shifts in policy? For Colombia, the two episodes (1953-66 and 1966-70) correspond to a period when great changes were taking place in the economy. First, the declining trend in coffee prices, which had been under way since the mid-1950s, was reversed about 1966. Second, that benchmark year was also unusual because it corresponded to a brief but intense liberalization effort, which was followed by a major reform in March 1967 that resulted in a less inward-oriented development strategy. The introduction of a crawling peg and of long-lasting incentives for nontraditional exports inaugurated a reorientation in Colombia's development strategy. 8 This mix of regimes led us in chapter 4 to classify Colombia as having pursued a neutral strategy. The three episodes in Mexico (1950-60, 1960-70, and 1970-75) con- stitute the longest time span covered in this study for any country. Throughout the period, Mexico continuously pursued an inward-oriented 7. For example, see Bhagwati (1978) for a good discussion of the equivalence of tariffs and quotas. 8. The incentives took the form of freely negotiable and tax exempt certificates equal to 15 percent of the export value. Figure 6-5. Benchmark Years, Major Events, Policy Regimes, and Trade Strategies of Sample Economies Economy 1950 1955 1960 1965 1970 1975 Colombia Benchmark year 1953 1966 Major event Po!Jcyreg~me Trade strategy Mextco Benchmark year 1950 1960 1970 1975 Pohcy regime Trade strategy Turkey Benchmark year 1953 1963 1968 1973 Major event Pohcy regime Tradesrrategy ER Yugoslavia Benchmark year 1962 1966 1972 Major event Polley regime Trade strategy Japan Benchmark year 1955 1960 1965 1970 Pohcy regtme Trade strategy Korea Benchmark year Major event Pohcyreg~me Trade strategy m• 1<1 I 5 ,, - 1963 1970 1973 Tmwan Benchmark year 1956 1961 1966 1971 Pol1cy reg1me - Trade strategy L*''l:i;:w;j;,'!h ,,,~,, Israel Benchmark year 1958 1965 1972 Major event Pohcy reg~me I& • Trade strategy Norway Benchmark year 1953 1961 1969 Policy reg~me Trade strategy 167 168 THE EXPERIENCE OF INDUSTRIALIZATION development strategy based on import substitution. Having benefited, as did Colombia, from natural protection during World War II, Mexico had already achieved considerable import substitution in consumer goods by 1950, the initial benchmark year. Significant breaks occurred in 1956 and 1970-the latter corresponding to the third benchmark year. The period in between was one of growth with stability, low inflation, and a fixed exchange rate that was maintained until 1975. Throughout, quantitative restrictions on imports were increasingly relied on to control the current account when foreign exchange imbalances arose. No real attempt at rationalizing or removing foreign exchange controls took place until1975 so that, from the point of view of policy, Mexico's strategy up until1970 can be characterized as continuously inward-oriented. During the 1970- 75 episode, Mexico experienced increasing inflation and foreign exchange pressures, which ended the previous period of growth with stability. Throughout, few incentives were provided to exporters. The three episodes in Turkey (1953-63, 1963-68, and 1968-73) cover two major cycles in Turkish postwar economic history, the first culminat- ing in the devaluation of 1958 and the second in the devaluation of 1970. The earlier crisis followed a period of very high inflation, while the later crisis followed a period of remarkable price stability. Like Colombia, Turkey experienced a rapid succession of stop-go policies during the 1953-63 period. During this period, multiple exchange rates, quantitative restrictions, and import surcharges were the main instruments used to control the demand for foreign exchange, though some export incentives were added. Occasionally, these controls were partially removed. On the whole, however, the trend was toward more and more control as the exchange rate became increasingly overvalued. During the 1963-68 period, the exchange rate was unified and a consistent import substitution strategy was pursued. An important shift occurred after the 1970 devalua- tion when, for the first time in its postwar history, Turkey enjoyed a relative abundance of foreign exchange because of a strong export re- sponse to the shift in incentives and unusually high levels of foreign remittances. Notwithstanding this brief liberalization, the basic strategy in the postwar period was to pursue successive phases of import substitution, first in consumer goods and later in capital goods. Export incentives were not sustained over any length of time because of a chronically overvalued exchange rate. Yugoslavia's two episodes (1962-66 and 1966-72) came well after the intensive period of import substitution during the 1950s, when basic industry was established. Until1965, Yugoslavia had a system of multiple exchange rates with export subsidies and quantitative restrictions. Rising inflation led to the devaluation of 1965; that year also saw a shift in policy as the foreign exchange regime was rationalized at the same time that a successful stabilization policy was pursued. The result was a unified exchange rate and a reduction in quantitative restrictions on imports. This TRADE STRATEGIES AND GROWTH EPISODES 169 liberalization of foreign trade was also accompanied by a liberalization of the internal domestic market as the dismantling of the government invest- ment fund led to less central control of the allocation of investment. Thus, Yugoslavia, which had already achieved a substantial substitution of imports by the initial benchmark year, progressively liberalized its foreign and domestic economic policies throughout the period (see Tyson 1980). After 1945, Japan pursued a stable and successful growth-oriented strategy that combined export incentives with substantial protection for import-competing industries that shifted increasingly away from tariffs and toward quantitative controls of imports. In common with Norway and in contrast with the other economies in the sample, Japan made no important change in its development strategy during the period, although one could argue that the emphasis on growth-oriented policies was height- ened around 1960 with the adoption of the so-called doubling national income plan. The three Korean episodes (1956-63, 1963-70, and 1970-73) straddle one important shift in policy regime in 1965-67 and an earlier, less important shift about 1958, the year that marked the end of post-Korean war reconstruction. The early 1960s were marked by political instability, successive bad harvests, and short-lived efforts to liberalize the foreign trade regime. About 1965, significant reforms in domestic economic poli- cies and foreign trade policies were introduced. On the domestic front, tax and interest rate reforms raised the savings rate, while strong incentives to exporters helped Korea achieve phenomenal rates of export growth up to the end of the period. 9 Korea's economic policy was not one of continuous liberalization throughout the period. The strategy included strong price and nonprice incentives for exporting accompanied by selective measures to promote import substitution. Taiwan's economic policies were among the most consistent in the group. Beginning with a period of hyperinflation in the early 1950s, Taiwan continuously and progressively liberalized and rationalized its domestic and foreign economic policies. Thus there was continuity throughout the three episodes (1956-62, 1962-66, and 1966-71) with one important change in policy regime following the "nineteen reforms" of 1960-61. These reforms marked a move toward an outward-oriented development strategy since effective export incentives were provided and sustained. By the end of the 1960s, as a result of rapid growth, Taiwan ended its period of labor surplus. Much like Taiwan, and in contrast with the other countries, Israel pursued sustained economic policies that guided the economy away from direct controls. Israel's two episodes (1958-65 and 1965-72) straddle the 9. On the relationship between financial reforms and savings, see Cole and Lyman (1971) and Brown (1973). For a revisionist interpretation of the interest rate reform, see Giovanni (1983). 170 THE EXPERIENCE OF INDUSTRIALIZATION reforms of 1962, which marked the replacement of quantitative restric- tions with price controls and represented a move toward trade liberaliza- tion. In addition to its small size, Israel is also distinctive in the group because of its wars and its political situation, which have necessitated heavy spending on defense. Thus, one can characterize Israel's economic policies as a progressive and sustained liberalization following an acute rationing of resources in the early years of its existence as a state. Norway consistently followed an outward-oriented strategy after the war. Moreover, Norway is the only one of the nine economies that achieved a truly liberalized trade regime. Indeed, as early as 1954, nominal and effective rates of protection on manufactured imports were less than 10 percent, while quantitative restrictions on imports were also being eliminated. Furthermore, realistic exchange rates were maintained throughout the period, which resulted in a continuously outward-oriented development strategy with a free-trade regime. The final line for each economy in figure 6-5 notes the trade strategies pursued during the period. 10 To support a trade strategy, a policy regime must be viewed by the economic actors in the system as persisting long enough to justify undertaking large investment projects and a significant reallocation of labor. Short-term stabilization policies whose goal is to control inflation and to establish short-run macroeconomic balance should be excluded from consideration. Unfortunately, adverse me- dium-term effects result from the stop-go policies often seen in countries suffering from foreign exchange shortages exacerbated by quantitative restrictions, insufficient exchange rate flexibility, and high inflation. In our sample, Colombia, Turkey, and Yugoslavia clearly suffered from such a regime at times, with adverse consequences for economic growth. Even in such cases, however, it is possible to discern long-term trends in policy by looking at trends in incentive measures such as effective rates of protection and effective exchange rates. Incentive Policies Most of the economies in our sample have been examined in detail in case studies. Five of them (Colombia, Israel, Korea, Taiwan, and Turkey) were included in the sample studied in the Bhagwati-Krueger project sponsored by the National Bureau of Economic Research (NBER).'' For these studies, comparable measures of effective exchange rates were calcu- lated for imports and exports. When properly measured, effective ex- change rates reflect the actual domestic currency costs for importers and 10. The classification is judgmental and is based on a number of comparative studies. See Little, Scitovsky, and Scott (1970), Bergsman (1979), World Bank (1981), Balassa and others (1982), and Balassa and associates (1971), as well as the NBER series edited by Bhagwati and Krueger. 11. For summaries of the project results, see Bhagwati (1978) and Krueger (1978). The data for Taiwan are drawn from a more recent study by Kuo (1983). TRADE STRATEGIES AND GROWTH EPISODES 171 Table 6-5. Ratio of Effective Exchange Rates for Imports to Exports of Selected Sample Economies Year Colombia' Turkey' Korea Taiwanb Israel 1954 1.48 1.04 1958 1.28 1.66 0.56 0.99 1960 1.02 0.68 0.95 1.00 1962 0.88 2.11 0.97 0.98 1.15 1965 1.02 2.39 0.96 0.98 1.15 1968 0.94 2.14 0.85 1.07 1.02 1971 1.00 0.82 1.14 1.01 - Not available. a. The rate for exports refers to "nontraditional" (manufacturing) exports. b. The ratio is for real effective exchange rates. See Kuo (1983, table 14-4). Sources: Colombia, Diaz-Alejandro (1976); Turkey, Krueger (1974); Korea, Frank, Kim, and Westphal (1975); Taiwan, Kuo (1983); Israel, Michaely (1975). the actual domestic currency receipts from export sales for producers in a sector. These rates should, therefore, take into account tariffs, import surcharges, the value of any import premiums, the tariff equivalent of quantitative restrictions, and all types of export subsidies including those of preferential access by exporters to loans and intermediate inputs. The ratio of the effective exchange rates for exports and for imports measures the bias in incentives resulting from the trade regime. 12 The data on effective exchange rates for the five sample economies in the 13 NBER study are summarized in table 6-5. From these data, certain pat- terns emerge. In Korea after 1964, the ratio of the effective exchange rate for imports to that for exports was relatively stable, with a slight down- ward trend indicating an increasing bias in favor of exports. In Taiwan, the ratio was near one (no bias), but with a slight upward trend. In Colombia, there was a distinct cycle, first falling to below one and then rising to about one. Israel also shows a cycle, but with less variation than Colombia. In both countries, there was a move in the latter part of the period to remove incentive biases against exports. In Turkey, the variance in effective exchange rates was quite high; there was a strong bias against exports throughout the period (but data are missing for 1970-73, when incentives evidently favored exports). An import substitution strategy reflects policies that give rise to incen- tives with a consistent and marked bias against exports. Export promotion implies either no bias or a bias in favor of exports. A liberalization strategy 12. See Bhagwati (1978) and McKinnon (1980) for discussions of trade bias. 13. Other incentive indicators have been calculated for some of the economies in the sample. Calculations of sectoral effective rates of protection (ERP) are available for some of the sample but only for scattered years. See Balassa and others (1982), and Balassa and Associates (1971). 172 THE EXPERIENCE OF INDUSTRIALIZATION implies-in addition to no bias toward either exporting or importing- low levels of protection and the maintenance of an equilibrium real exchange rate so that there is no bias toward producing tradables or nontradables. In figure 6-5, the points demarcating a shift in trade strategy are arbitrary in that such policy shifts usually unfold over a period of at least a few years. In some cases (Israel, Korea, and Yugoslavia, for exam- ple), the shift in strategy was accompanied by widespread economic changes and reforms that took several years to implement. If each de- marcation point is regarded as a zone, in almost every case a benchmark year lies in the zone. Thus, by good fortune, the episodes generally delin- eate periods characterized by consistent trade strategies. Economies that pursued gradual shifts in policy regimes include Israel, Japan, Korea, Taiwan, and Yugoslavia. In contrast, Colombia's shift in policy took place rapidly after the March 1967 reforms. Before that date, there had been a succession of stop-go policy cycles, reflecting brief episodes of stabilization and liberalization in the foreign trade regime. In Turkey, a major change in orientation took place about 1970, with a liberalization that was sustained for a few years (until the oil crisis) because of improved foreign exchange availability caused by rising work- ers' remittances. Exports were helped indirectly through the lower cost of imported intermediate inputs as import rationing became less severe and remittances permitted the exchange rate to be somewhat undervalued. In Mexico, one cannot discern any significant change in trade strategy. Inflation was very low in 1956-70, but without any shift in incentives. Throughout the period, domestic industries were protected, and the incen- tives to export were few. Norway maintained a liberal trade strategy during the entire period. Trade and Industry The preceding discussion presumed a causal link between incentives and the relative effect of import substitution and export expansion policies. But such a link is likely only for marginal activities, that is, for activities having either positive demand elasticities (in the case of imports) or positive supply elasticities (in the case of exports). This means excluding both what are referred to as noncompetitive imports and most exports derived from the exploitation of natural resources whose supply is fixed, such as mining and other extractive activities. Exports of primary com- modities such as coffee, which require special growing conditions and have no close substitutes, would also fall into the category of intramar- ginal export activities. Thus a considerable proportion of primary exports should by and large be fairly insensitive to changes in incentives of the kind discussed earlier since they would be exported under a variety of circum- stances and incentives. It is mostly in the manufacturing sector that we expect an association between incentives and contributions to growth. Figures 6-6 and 6-7 indicate the contributions of export expansion and Figure 6-6. Contribution of Export Expansion by Episodes, Selected Sample Economies 1 Colombia 2 1 Mexico 2 3 1 Turkey 2 3 1 Yugoslavia 2 0 5 10 15 20 25 Percent 1 2 Japan 3 4 1 Korea 2 3 1 Taiwan 2 3 1 Israel 2 1 Norway 2 0 10 20 30 40 50 Percent Key: • Heavy industry [J Light industry Note: Numbers beside economy names refer to episodes delineated by benchmark years (see table 6-6). 173 174 THE EXPERIENCE OF INDUSTRIALIZATION Figure 6-7. Contribution of Import Substitution by Episode, Selected Sample Economies 10 5 ... t:: ~ or-~~~--~~~~~~--~--~LL~------~ <1) ~ -5 -10 1 2 1 2 3 1 2 3 1 2 Colombia Mexico Turkey Yugoslavia 15 Key: B Heavy industry £ZI Light industry 10 5 ... t:: <1) ... u <1) 0 ~ -5 -10 -15 1 2 3 4 1 2 3 1 2 3 1 2 1 2 Japan Korea Taiwan Israel Norway Note: Numbers above economy names refer to episodes delineated by benchmark years (see table 6-6). TRADE STRATEGIES AND GROWTH EPISODES 175 import substitution for light and heavy industry (expressed as a percentage of the change in economywide total gross output). The data cover each episode for the sample economies and can be compared with figures 6-3 and 6-4, which cover the entire period. Table 6-6 gives the sources of change in total manufacturing output for each episode, expressed as a percentage of the change in manufacturing output (so that the various contributions add up to 100 for the sector). Typology of Trade Strategies Table 6-6 indicates that most of the sample economies went through an episode marked by a sizable effect of import substitution on manufactur- ing output growth. In Korea and Taiwan, this period corresponded to an Table 6-6. Sources of Growth in Manufacturing Output for Sample Economies Sourceb Years Growth Economy (episode) rate' DD EE IS 10 Colombia 1953-66 (1) 8.3 60.2 6.8 22.2 10.8 1966-70 (2) 7.4 75.7 4.7 4.2 15.3 Mexico 1950-60 (1) 7.0 71.6 3.1 10.9 14.5 1960-70 (2) 8.6 86.1 4.0 10.9 -0.9 1970-75 (3) 7.2 81.4 7.9 2.4 8.3 Turkey 1953-63 (1) 6.4 80.9 2.4 9.1 7.6 1963-68 (2) 9.9 75.1 4.5 10.5 9.9 1968-73 (3) 9.6 76.2 10.4 -1.6 15.0 Yugoslavia' 1962-66 (1) 16.6 73.7 24.8 -5.0 6.5 1966-72 (2) 9.1 72.1 37.6 -22.1 12.4 Japan 1914-35 (1) 5.5 70.0 33.6 4.7 -8.4 1955-60 (2) 12.6 76.2 11.9 -3.3 15.2 1960-65 (3) 10.8 82.4 21.8 -0.4 -3.8 1965-70 (4) 16.5 74.4 17.5 -1.5 9.6 Korea 1955-63 (1) 10.4 57.4 11.5 42.2 -11.2 1963-70 (2) 18.9 70.0 30.2 -0.6 0.4 1970-73 (3) 23.8 39.0 61.7 -2.6 1.9 Taiwan 1956-61 (1) 11.2 34.7 27.5 25.5 12.3 1961-66 (2) 16.6 49.1 44.6 1.6 4.7 1966-71 (3) 21.1 34.8 57.1 3.8 4.3 Israel 1958-65 (1) 13.6 57.0 26.5 11.7 4.8 1965-72 (2) 11.3 75.8 50.0 -36.6 10.8 Norway 1953-61 (1) 5.0 65.1 36.5 -16.1 14.4 1961-69 (2) 5.3 51.0 58.3 -19.4 10.0 a. Average annual growth rates of total manufacturing gross output. b. Expressed as percentages of change in total gross manufacturing output; add up to 100 percent. DD is domestic demand expansion, EE is export expansion, IS is import substitution, and 10 is change in input-output coefficients. Source: World Bank data, described in Kubo (1983). 176 THE EXPERIENCE OF INDUSTRIALIZATION early phase of industrialization from the mid-1950s to the early 1960s. In the relatively closed economies-Colombia, Mexico, and Turkey-the import substitution phase continued for many years. From figure 6-7, it appears that in the closed economies, and also in Israel, import substitu- tion was concentrated in heavy industry, whereas in Korea and Taiwan it was important in both heavy and light industry. In many other episodes, changes in input-output coefficients (ro) made a significant contribution to growth, with magnitudes comparable to those for import substitution. As discussed in chapter 5, such a positive con- tribution reflects an increase in input-output coefficients, that is, a deepen- ing in interindustry linkages. This phenomenon appears to be a universal characteristic of industrialization and will be discussed in more detail in the next chapter. In the closed economies, export expansion was not an important source of output change in any episode. Both Mexico and Turkey had episodes in which exports increased in importance, but the numbers are significant only in relation to their past values and not in comparison with the rest of the sample. In all the other economies, the role of exports was large in all episodes, but with significant variations across economies, episodes, and sectors. Korea and Taiwan stand out as the star performers. They started with dramatic increases in exports of light industrial products and then shifted toward exports of heavy industrial products. Although less drama- tic, this shift in exports from light to heavy manufacturing is also evident in Israel, Japan, and Norway. Only in Yoguslavia was the contribution of heavy industrial exports greater in the early period, which probably reflects a pattern typical of socialist economies with their emphasis on the development of heavy industry. The observed shifts in the contribution to growth from export expan- sion and import substitution across episodes parallel the shifts in develop- ment strategy discussed above (see figure 6-5). The episodes-twenty-four in all for the nine economies-can generally be categorized as belonging to one of three distinct trade strategies: import substitution, export promo- tion, or trade liberalization. In table 6-7, the episodes are arrayed accord- ing to the contribution of export expansion and of import substitution to the total change in manufacturing output. Within each group, the episodes are listed in ascending order of the contribution of export expansion. In the initial episodes for Colombia, Korea, and Taiwan, import sub- stitution made a large contribution, which is consistent with their policy regimes. In Mexico and Turkey, significant but smaller contributions of import substitution characterized the first two episodes. Both countries had also pursued an import substitution strategy in the preceding decades. In Turkey's 1968-73 episode, when exports were promoted, the export contribution rose to 10.4 percent. This episode, although short, did reflect a shift in incentives and perhaps provided an indication of what Turkey TRADE STRATEGIES AND GROWTH EPISODES 177 Table 6-7. Typology of Trade Strategies by Contributions to Growth (percent) Import substitution Export Row expansion -37 to -16 -5 to +5 9 to 12 22 to 43 total 2 to 9 Colombia (2) Turkey (1) Colombia (1) Mexico (3) Mexico (1) Mexico (2) Turkey (2) 7 10 to 28 Turkey (3) Israel (1) Korea (1) Japan (2) Taiwan (1) Japan (4) Japan (3) Yugoslavia (1) 8 30 to 62 Norway (1) Korea (2) Yugoslavia (2) Japan (1) Israel (2) Taiwan (2) Norway (2) Taiwan (3) Korea (3) 9 Column total 4 12 5 3 24 Note: Economies are arrayed according to contributions of export expansion and import substitution to total manufacturing growth. Within groups, episodes are listed in increasing order of the contribution of export expansion. Numbers in parentheses refer to episodes as defined in table 6-6. Source: Table 6-6. could accomplish, as confirmed by the effect of its shift toward an open development strategy in the 1980s. Colombia is an exception: its shift toward export promotion policies in the last episode is not reflected in the results because the effects had only begun to be felt and because of the importance of coffee. An additional explanation is provided by Morawetz (1981), who did a case study of the clothing sector. He argues that Colombian clothing exporters were not able to meet the quality and delivery schedule requirements of the U.S. market even though price incentives existed for part of the period. In seventeen out of twenty-four episodes, export expansion contributed more than 10 percent to total output growth. In the early episodes in Israel, Korea, and Taiwan, moderate export expansion was coupled with substantial import substitution. More commonly, however, export expan- sion contributed heavily to growth when incentives favoring import sub- stitution were removed, so that import substitution made little contri- bution. In four episodes for Israel, Norway, and Yugoslavia, export expansion was coupled with large negative import substitution. These episodes can 178 THE EXPERIENCE OF INDUSTRIALIZATION be characterized as reflecting trade liberalization policies and as resulting in increased exports and imports. 14 The sizable effect of changes in input- output coefficients for the liberalization episodes in both Israel and Nor- way may well reflect technological changes in response to international competition, a theme that will be explored in chapter 7. Sequencing of Trade Strategies In addition to a clustering of economies according to the contribution of import substitution and export expansion to growth, it also appears from table 6-7 that trade contributions in these economies follow a distinct sequence. Figure 6-8 plots the export expansion and import substitution contributions for all the episodes and shows clearly that periods of signifi- cant export expansion are almost always preceded by periods of strong import substitution. Japan is the only exception. It is by far the largest economy in the sample, so the small contribution of import substitution, either positive or 11egative, is not surprising. From figure 6-8, it is possible to distinguish three groups of economies. The first group comprises Israel, Korea, and Taiwan-economies in which a period of strong import substitution was followed by a period of export- led growth. The second group comprises Japan, Norway, and Yugosla- via-economies in which the growing role of export expansion in later periods combined with either continued import liberalization (Norway and Yugoslavia) or no significant import substitution (Japan). Colombia, Mexico, and Turkey constitute a third group-economies in which, although the role of import substitution fell over time, export expansion was never very important and growth rates were relatively low. Mexico and Turkey did have episodes with increased export expansion, and Turkey perhaps shares some of the features of the second group. For the first group, the pattern of observed sequencing of import sub- stitution and export expansion can be attributed to the continuation of a shift in policy toward an export-led development strategy and to the availability of foreign capital at the time when the change in policy was taking place. 15 Indeed, the timely availability of relatively large amounts of foreign exchange is a distinguishing characteristic of Israel, Korea, and Taiwan. For the second group, the situation was somewhat different since these countries were already semi-industrialized in the initial period and are therefore not strictly comparable to the other economies in the sample. It is not surprising to find that countries which are members of the General Agreement on Tariffs and Trade (GATT)-and which participated in the multilateral tariff reductions in the 1960s and 1970s-experienced import 14. Israel is discussed by Michaely (1975), Yugoslavia by Tyson (1980), and Norway by Balassa (1979a). 15. Some have argued that the threat of U.S. aid withdrawal was important in shaping the change in policies. See Krueger (1980b) on Korea. Figure 6-8. Trade Sequences in Manufacturing of Selected Sample Economies, by Episode c 0 15 ·.::::: ::l ..D ·;::: i5 I 0 u Turkey (3) I ~ 10 "' .... .:: < I I < 20 < '"'"' 0 -40 -30 -20 -10 0 10 20 30 40 50 Import substitution (percentage contribution) Note: Numbers in parentheses refer to episodes. 179 180 THE EXPERIENCE OF INDUSTRIALIZATION liberalization with increased export expansion. For the third group, the minor role of export expansion combined with decreasing import sub- stitution can be attributed to less favorable initial conditions and to an evident reluctance to shift toward a more open development strategy with a reduced bias against exports. In contrast with the first group, the supplies of both foreign exchange and human resources were less favorable in the initial periods and at the time when a shift in strategy seemed desirable. The observed sequencing raises an immediate question: is a period of significant import substitution, with strong protection of domestic manu- facturing, necessary to build the industrial base required for a later shift to an open development strategy led by export expansion? The question is related to a version of the infant industry argument for protection: is a period of protection for manufacturing required to allow firms to grow up before they must face the real world of international markets? It is difficult to generalize from so few cases, and a complete analysis requires consid- eration of productivity growth, which we discuss in part III. Nevertheless, the evidence presented so far is consistent with the infant industry argu- ment, and it is worthwhile looking at some cases in more detail. 16 From table 6-7, it appears that Korea and Taiwan are the two econo- mies that experienced the most dramatic switch from import substitution to export expansion. Israel and Turkey also followed such a pattern, but less dramatically. Israel is a very small country and something of a special case, so we do not consider it further. For Korea, Taiwan, and Turkey, we turn to a more detailed analysis of the role of export expansion and import substitution in the growth of the manufacturing sectors. Table 6-8 presents data on the relative contributions of import substi- tution and export expansion as percentages of total sectoral output growth during each episode for the three economies. The table also shows the total change in sectoral output as a percentage of the change in aggregate (economywide) output in each episode. The manufacturing sectors are listed in descending order of contribution to total output change, with the leading sectors at the top of each table. For Turkey, the table lists all the manufacturing sectors. For Korea and Taiwan, the miscellaneous and unallocated sectors are omitted. 17 Even though these economies underwent major shifts in development strategy, table 6-8 indicates that the leading sectors changed little. With a few exceptions, the first five or six sectors led in all episodes. In Korea, machinery moved from sixth to eighth place and then shot up to first place 16. For other analyses of the infant industry argument and sequencing from import substitution to export expansion, see Westphal (1982) and Balassa (1979b). 17. See chapter 5 for a detailed reconciliation of the sector definitions across all the economies. There are some differences in sector definitions. Paper and printing are aggre- gated in Taiwan and Turkey. Textiles and clothing are aggregated in Turkey, but leather products are a separate sector. Leather products are included with clothing in Korea and Taiwan. Kubo and Robinson (1984) also discuss the sources of growth in particular sectors. TRADE STRATEGIES AND GROWTH EPISODES 181 in its contribution to aggregate output growth. Taiwan shows a similar pattern, with machinery moving up as an increasingly important leading sector. In Turkey, machinery actually fell in rank from fifth to seventh place. In Taiwan, the role of food processing declined while that of clothing increased dramatically. In Turkey and Korea, there was less change in the relative importance of these "light" sectors. Although there was relatively little change in the ranks of the leading sectors, there were significant changes in all three economies in the relative roles of import substitution and export expansion. 18 Figures 6-9, 6-10, and 6-11 plot the import substitution and export expansion contribu- tions from table 6-8. In Korea, the sequencing is quite dramatic. In textiles, chemicals, machinery, petroleum products, paper products, and wood products, import substitution was the main source of growth in the first episode, while export expansion became the main source in the two later episodes. In Taiwan, import substitution in the first episode was very large for basic metals, chemicals, machinery, nonmetallic minerals, clothing, paper products, petroleum products, and transport equipment. In later episodes, the contribution of export expansion to all these sectors was large. The changeover in Taiwan was less dramatic than in Korea, how- ever. In many sectors, export expansion was significant in the first episode, so that import substitution and export expansion occurred simulta- neously. In Turkey, there was significant import substitution in one or both of the first two episodes in petroleum products, basic metals, machinery, nonme- tallic minerals, rubber products, and transport equipment. Of these six sectors, four had significant contributions of export expansion in the last episode (but not machinery and transport equipment). The numbers are much smaller for both contributions than in Korea and Taiwan, but the sequencing is evident. In addition, two sectors (food processing and leather products) had significant export expansion contributions during the earlier episodes. In contrast with what happened in Korea and Taiwan, the machinery sector in Turkey did not "mature" during this period-that is, the first episode of extensive import substitution was not followed by any significant export expansion. The export expansion episode in Turkey was not sustained. The country went through yet another cycle of increasing bias toward import substitu- tion and away from exporting, followed by a crisis in 1977-78. 19 Recently, however, Turkey has shifted toward a more open development strategy, and it now appears that exports, especially of manufactures, are respond- ing successfully. The period of inward-oriented development was quite prolonged in comparison with other semi-industrial countries, but the 18. See Chenery (1980), who uses the same data and discusses this issue in more detail. 19. See Dervis and Robinson (1982) and Celasun (1983) for an analysis of this period. Table 6-8. Leading Sectors and Trade Sequences for Selected Semi-Industrial Economies (percentage contribution) 1955-63 1963-70 1970--73 ..... Contribution Contribution Contribution 00 N Output Output Output Sector change Rank IS EE change Rank IS EE change Rank IS EE Korea Food processing 12.6 1 17.7 7.1 9.5 1 -2.9 9.7 11.3 2 1.9 17.5 Textiles 6.5 2 80.7 20.2 4.3 4 -14.3 72.0 6.9 6 -13.7 99.1 Clothing 6.1 3 5.7 6.1 6.0 2 -0.8 45.1 8.6 3 -5.2 70.7 Chemicals 5.4 4 72.8 2.7 4.8 3 8.4 21.2 8.0 4 5.9 44.8 Basic metals 4.5 5 4.1 28.0 3.3 5 11.3 20.4 7.3 5 -13.8 76.1 Machinery 4.3 6 52.7 4.3 2.5 8 -28.7 36.6 12.2 1 2.1 69.1 Petroleum products 3.0 7 97.5 3.1 3.0 6 30.4 17.3 2.1 7 -7.2 40.9 Paper and products 2.3 8 95.5 1.9 1.1 11 -24.2 19.3 1.6 9 -1.9 53.2 Nonmetallic minerals 2.3 9 14.0 2.0 1.9 9 7.2 11.2 1.7 8 -1.7 41.0 Printing 2.0 10 7.1 2.1 0.5 12 -7.5 18.3 0.9 11 4.2 42.8 Transport equipment 1.7 11 10.6 7.7 2.5 7 7.0 7.5 0.8 12 0.3 70.1 Wood and products 1.7 12 64.4 24.0 1.7 10 -0.3 62.0 1.5 10 -0.8 112.5 Taiwan Food processing 10.5 1 -0.7 35.8 12.7 1 -7.5 29.7 5.8 5 2.1 28.0 Textiles 7.5 2 8.2 41.1 6.6 2 -2.3 67.1 12.9 1 4.0 64.6 Basic metals 3.0 3 63.5 25.0 3.8 5 -9.1 63.2 4.5 6 13.2 59.5 Wood and products 2.9 4 6.9 33.1 1.7 11 0.8 108.8 3.6 7 0.4 66.0 Chemicals 2.5 5 98.0 13.3 5.9 4 10.5 31.2 7.3 4 3.8 52.5 Machinery 1.8 6 23.8 18.5 6.0 3 32.9 31.3 12.4 2 3.9 65.3 Nonmetallic minerals 1.7 7 21.0 32.1 2.6 7 -0.5 48.1 1.1 11 8.1 6.2 Clothing 1.6 8 30.1 23.7 2.2 9 0.0 43.1 7.7 3 -2.4 67.3 Paper and products 1.6 9 13.8 26.2 2.7 6 3.9 23.2 2.5 9 -0.6 30.6 Petroleum products 1.4 10 14.4 33.1 2.5 8 -7.9 24.3 2.3 10 3.5 46.6 Transport equipment 1.0 11 3.4 10.7 2.1 10 -10.0 22.6 2.7 8 18.5 28.5 Turkey Food processing 11.7 1 -3.0 2.8 6.9 3 6.6 24.2 9.6 1 -0.9 19.4 Petroleum products 5.5 2 21.1 3.5 4.5 4 1.7 1.2 6.7 2 1.7 10.6 ...... 7.2 2 4.5 2.5 4 -0.2 22.0 00 Textiles and clothing 4.8 3 8.3 1.6 6.0 w Basic metals 3.0 4 13.5 -2.5 8.7 1 17.7 1.8 6.3 3 -21.0 6.7 Machinery 2.1 5 67.6 0.0 3.3 6 18.9 0.4 4.3 7 8.8 1.9 Chemicals 1.4 6 -24.7 -0.2 3.5 5 -2.3 1.8 4.8 6 -5.7 3.8 Nonmetallic minerals 1.3 7 36.7 0.9 1.7 9 -0.3 1.1 2.7 8 4.2 7.5 Wood and products 1.2 8 2.7 0.8 1.5 11 7.3 0.9 1.4 10 -0.4 3.5 Rubber and products 1.0 9 8.7 1.5 1.3 12 31.3 1.1 1.2 11 -0.9 9.3 Leather and products 0.8 10 5.1 19.0 2.5 7 0.5 -6.8 0.5 12 -11.3 20.9 Paper and printing 0.8 11 -0.9 1.4 1.9 8 1.0 1.1 1.8 9 19.2 4.6 Transport equipment 0.5 12 -39.0 5.2 1.6 10 44.2 1.3 5.0 5 0.4 2.0 Source: World Bank data; described in Kubo (1983). Percentage contribution Percentage contribution ., ~ ...... ...... ...... i)Q' = ..., '0"" 0 '0"" m•• '0"" '0"" 0 0 0 0 0 (1) 0\ .................. I LOOO; I \0 \0 \0 Food processing tr1 Food processing ~ 00 w'"" 0\ 0\'"" >< i ~ I I I "1::l- "1::l- '-10\0\ w 00 w Textiles ... 0 .... Textiles 0 .... .... "1::1 "' >< "' ;: ... 0 ..... Clothing ., "1::l- Clothing <::!"' .... "' :::, ;::! ~· ;::: Chemicals "' a· ;: k ;::! Chemicals g. trl <::!"' ;::! ~ '< "tj- Basic metals Basic metals <::!"' ..... "1::l- ... '< ... 0 ..... 00 "' a· "1::l- ~ Machinery ~:>... Machinery "' .... a· Q .... :::, Petroleum products Petroleum products Jg "' "' ~· Paper and paper products Paper and paper products ~ Nonmetallic minerals Nonmetallic minerals ... 0 "' :::, Printing Printing Transport equipment Transport equipment Wood and wood products Wood and wood products Percentage contribution Percentage contribution "Tl ::>::: c;q· "' ~ ...... ...... I ...... = ..., :!1.:: VI 0 VI VI VI 0 (1) 1!11188 0 0 0 0 0 0 0 0 .................. I I 9" ...... \0 \0 \0 0\ 0\ VI Food pmre,.ing t"r1 Food processing ...... ;:;, 0\ ...... 0\ ~ ~ r I I I ~ ~ ....... '-.10\0\ ...... 0\ ...... 0 ... ... 0 ~ Textiles ..... Textiles ..... 'l;:t ~ "' "' ;:: ... 0 .... Basic metals ~ I:> ;::: "' a· Basic metals ab <::!- "' ..... ;:;.· ;:: I::> ;:::: l:l.. Wood and wood products J .!& . Uii ;::: Wood and wood products ~ g. t"rl ~ ,... ;::: p <::!- 'l;:t '< ...... Chemid•~ ~ Chemicals <::!- '< ... 0 .... 00 ... "' ~ '"" ~~ o· ... (J Machinery . s*'""'' $:)... Machinery "' a· ;:s-o I::> $:)... ;:::: ()q "' Nonmetallic minerals Clothing ££i. .m.d~I!AI Nonmetallic metals -l Clothing l =- r- "' ;:i" ~ Paper and paper products - - . . Paper and paper products Petroleum products l -i r ... ~- I::> ;:::: Transport equipment -R,. Transport equipment -i -t- ~ Percentage contribution Percentage contribution ~ ~ ~ ":'; I ....,. C ~~~ ~~~ x; ....; J• I ,_. IJ .. I N 7 7 ~ v.> v, 0 0 v, 0 o 0 [? (il ~ :-' r Food processing I Food processing I I I ~ ~ ~ ;:;:l ~ e:; Petroleum products ~..,,·:.I ~ Petroleum products ~ ~ Textiles and clothing -1 ~· ..1 . .:. . 1 ~ .z ~ Textiles and clothing r ~ ~ -. 0 ~ ~ Basic metals -l ~ g· Basic metals ~ ~- ~ ....,. ~ Machinery j b· ~ ~ ~ Machinery ~ ~ ~ ~ ~ Q ~ ~ i d ~- Q. 0\ Chemicals Nonmetallic minerals Wood and wood products -l j b Ja Chemicals Nonmetallic minerals Wood and wood products r ~ ~:).., i ~ ;:i· Rubber and rubber products !-,;,' , 1 Rubber and rubber products ,._ ~ Leather and leather products -1 V., :• ., ·.I Leather and leather products J '< j 6 6 r Paper and printing Paper and printing Transport equipment Transport equipment ~ TRADE STRATEGIES AND GROWTH EPISODES 187 results discussed here indicate that the economy might well have been able to shift strategies successfully in the early 1970s. The advent of the oil crises and the government's misguided response to them delayed the shift for a decade. Conclusions This chapter has expanded on the long-run comparisons in chapters 2, 3, and 4. With a smaller sample of economies but with more detailed data, we have analyzed the interrelations between the choice of a development strategy, its supporting policy regimes, and economic performance at both the aggregate and sectoral levels. The results for these economies have been compared with those for the archetypal economies modeled in chap- ter 3. The analysis has focused both on long-run trends and on shorter- term episodes that can be classified by their trade strategies. The analysis of the interrelations between the choice of a development strategy, its supporting policy regimes, and economic performance was of necessity heuristic. The input-output model that provides the framework for the analysis does not directly incorporate the links between policy choices, incentives, and performance. Although many of the links we discussed are suggestive, formal modeling of them is beyond the scope of this chapter. Chapter 11 analyzes some of these policy links with a more elaborate model of a single country. The multisectoral approach used in this chapter has shown how econo- mies pursuing different development strategies exhibit different kinds of structural change. In particular, the analysis has focused on leading sectors and on the roles of different sources of demand as engines of growth. In the next chapter, we focus more on the composition of output and on how the choice of a development strategy affects structural change. The analysis of trade sequences raises issues about technological change and productivity. In Korea, Taiwan, and Turkey, there were episodes when import substitution was an important source of growth for many manufacturing sectors; these were followed by episodes when export expansion took over as a significant driving force. The results in this chapter for our sample are consistent with the view that a period of import substitution and deepening of input-output relationships is required be- fore an economy can compete successfully in world markets and so shift to either an export-led or open development strategy. This is a theme that will be considered further in later chapters. 7 Interdependence and Industrial Structure YUJI KUBO JAIME DEMELO SHERMAN ROBINSON MOSHE SYRQUIN CHAPTER 6 FOCUSED on economywide growth and on the roles of different sectors and sources of demand as the engines of that growth. We identified several typologies of growth and related policy regimes-based largely on aggregate data-which could be used to characterize distinctive subperiods or episodes in the development of the nine economies in our sample. In this chapter, we examine the development experience of these economies from a different angle, stressing sectoral structure and the determinants of structural change. "Structural change" is a frequently used term in economics, but it is nevertheless hard to pin down. 1 In the absence of a formal model, it is convenient to define structural change as any shifts in the composition of various economic aggregates. In speaking of the causes or sources of structural change, however, one must have in mind some model of the economy that specifies underlying stable relationships-even if only in imprecise terms. Because we have an input-output model, we can be more precise in the measurement of structural change, although the scope of the model limits the aspects of structural change that we can cover. In the static model, the input-output coefficients are assumed to be given, as are the level and sectoral composition of the components of final demand-consumption, investment, imports, and exports. The model then determines the level and sectoral composition of total production. This treatment implies that changes in the structure of these exogenous elements are the sources or causes of structural change. Thus we shall study changes in the sectoral composition of aggregate output that are induced by changes in demand, trade, and production technology. The emphasis is on the role of such structural parameters as input-output 1. See Machlup (1963), and Chenery (1979, pp. 108-09) for a discussion that distin- guishes between the structure of an economy and the structure of a formal mathematical model of an economy. 188 INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 189 coefficients, import coefficients, and the composition of trade and final demand. 2 One of the main features of structural transformation is a rapid rise in the relative importance of manufacturing production accompanied by a relative decline in primary production. This chapter is concerned with the various aspects of this transformation. We begin the chapter by looking at some summary measures of structural change over roughly a twenty-year period for the economies in the sample and by setting out the long-term regularities that characterize their growth experience. We then trace these regularities back to changes in three different structural relations: com- position of demand, intermediate input technology, and size and composi- tion of trade. We shall also draw on data for the various benchmark episodes discussed in chapter 6 to examine the relation between these structural changes and the choice of a development strategy. Aggregate Measures of Structural Change Chapter 3 explored the nature of the long-run transformation of typical developing economies and divided the process into a series of stages or phases. This section examines the industrialization phase and several concomitant structural changes in the sample economies. Industrialization According to the model used in chapter 3, the industrialization phase of structural transformation typically entails a shift in the manufacturing share of GDP from about 19 percent to 36 percent, which is associated with an increase in per capita income from $280 to $2,100. In the model, the process was assumed to require about fifty years, with GNP growing at a rate of about 6.2 percent a year (or 3.9 percent per capita). These figures imply that the share of manufacturing increases by an average of 3.2 percentage points each decade throughout the period (which defines the "rate of industrialization"). This performance is about average compared with the record during the nineteenth century of those countries that are now industrialized; how- ever, the variance in historical performance is quite high. Kuznets (1966, table 3-1) estimates rates of industrialization for these countries that range from about 1 to 6 percentage points a decade, with Canada and Italy at the low end and Sweden and the United States at the high end. Figure 7-1 presents comparable data on the ten-year rates of indus- trialization (that is, decadal changes in the share of manufacturing in GDP) 2. In this chapter, given data limitations, we are not able to analyze separately the roles of investment and consumption demand. Kubo, Robinson, and Urata (1986) consider the role of investment in the framework of a dynamic input-output model applied to two countries in the sample, Korea and Turkey. 190 THE EXPERIENCE OF INDUSTRIALIZATION Figure 7-1. Change in Share of Manufacturing in GDP, Semi-Industrial Economies, 1953-73 40 I I I 0 I Yugoslavia : : Argentina 30 I 0 I 0 Norway I 0 Spain 0 Ireland -.::;- I M.ex1co t:: I!) 1 0 Portugal u .... 0 . Israe I I I!) Braz1 1 0 I 0 Japan Eo 0 South Africa 0 o- Uruguay Hong Kong I!) .... 20 ~ i 0Peru ....c:: India "' 0 Dom,inican 0 0 Philippines I 3 - E 4 0 0 u = 1 w Korea 2 3 1 Taiwan 2 ' ~,, ' 3 1 Israel 2 1 Norway 2 : 0 5 10 15 20 25 Percent creased with time. In their respective postwar recoveries, each had an episode during which the role of heavy industry was large followed by an episode during which its role declined dramatically. Then, in the final episode in each country, heavy industry again led nonproportional growth. In Colombia, the role of heavy industry declined dramatically during the two episodes for which there are data. In contrast, Israel and Norway, both more developed than the other economies in the sample, showed little change in the role of heavy industry. The role of light industry varied widely. In the large, more closed economies (Japan, Mexico, and Turkey) light industry was not a signifi- cant source of structural change. In the export-led economies (Korea and Taiwan), it was much more significant. In Taiwan, light industry (espe- cially food processing) grew very rapidly in the early episode. Next came a INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 199 Figure 7-4. Output Deviations from Balanced Growth in Light Industry, Sample Economies Colombia~ 1 Mexico 2 3 1 Turkey 2 3 - 4 a 0 c 1 0 u >-Ll Korea 2 3 1 Taiwan 2 3 1 Israel 2 1 Norway 2 -10 -5 0 5 10 15 20 Percent shift toward heavy industry (1961-66) followed by a new spurt (1966-71) in which light and heavy industry were both leading sources of structural change. In Korea, light and heavy industry moved together in their effects on nonproportional growth. Korea and Taiwan differ markedly from the rest of the sample. The magnitude of their output deviations is larger, and light industry plays a much more important role in the process (accounting for about half their nonproportional growth of manufacturing). The differences are related to the dramatic effects of the export-led development strategies pursued by these two economies, which represent extreme examples of the typical small manufacturing (sM) pattern discussed in chapter 4. Figure 7-5 presents the decompositions into sources, by category of demand, of the output deviations for heavy and light industry for most of Figure 7-5. Output Deviations of Heavy and Light Industry from Balanced Growth, Sample Economies Colombta Mexico 10 1950-60 1960-70 1970-75 ""' .:l 1953-66 Yugoslavia 1966--70 15 Turkey 10 c ~ -10 -15L--L----~----L_ ___ L_ _ _ _~_ _ -5 ~ ~ .; J 1953-63 ~ -< 1963-68 1968-73 1962-66 1966-72 Korea japan 20 15 1914-35 1955-60 1960-65 1955-63 1963-70 1970-73 1965-70 Israel Tarwan 15 15 10 - 10 ~ ~ 1956--61 ~ - 1961-66 1966--71 -15 - 20,L__j__ _...L.__ ___[_ _...L.__----,l_ ~ ~ ~ ~ Norway :1: - ~ - 15 1958-65 1965-72 10 Key: D Domestic demand 0 Export expansion 1m Import substitutiOn • Change in input-output coefficient 1958-61 1961-69 200 INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 201 the economies and episodes. The data are shown as percentages of the total deviation for the economy, so the sum of the contributions for each subsector across demand categories equals the contributions given in figures 7-3 and 7-4. Looking first at Colombia, Mexico, and Turkey, we note that export expansion, when significant, occurred in light industry, whereas import substitution was more important in heavy industry for most of the epi- sodes. In Turkey's brief phase of export-led growth (1968-73), the dra- matic increase in the contribution of exports to nonproportional growth of light industry was largely offset by a decrease in the contribution of expanding domestic demand. The export boom represented a diversion of goods from the domestic market with little effect on the overall share of light industry in output. 15 In contrast, the export expansion episodes in Korea and Taiwan were also led by light industry but with no offsetting decline in the expansion of domestic demand. The result was a significant increase in the share of light industry, as noted in figures 7-3 and 7-4-an observation that differenti- ates these two economies from the rest of the sample. Korea and Taiwan also differ from Colombia, Mexico, and Turkey in that export expansion made an increasingly significant contribution to structural change in heavy industry. In this respect, they resemble Yugoslavia and Japan. The role of import substitution varied in the sample. In the more advanced countries-Yugoslavia, Japan, Israel, and Norway-import substitution made a low or negative contribution to structural change in both manufacturing subsectors. In the rest of the sample, import substitu- tion was important in both subsectors in the early episodes but then declined in importance (with the exception of heavy industry in Co- lombia). By the end of the entire period, these countries appear to have exploited the available opportunities for "easy" import substitution. Less contribution to structural change from this source is likely in the future. The role of changes in input-output coefficients varied widely across economies and episodes. For nearly every economy and subsector, there were episodes during which changes in input-output coefficients had a significant effect on structural change. At this level of aggregation, how- ever, there does not appear to be any systematic relation between this source of deviation and other sources. As we discussed earlier (chapters 3 and 6), regularities in changes in the technology of intermediate inputs go along with development, but these changes are rooted in the increasing interdependence of the economy and are systemic. In the next section, we do a systemwide exploration of these changes in intermediate input tech- nology. 15. In this case, note the negative contribution of import substitution in heavy manufac- turing during Turkey's export expansion phase, when at least some of the foreign exchange was spent on imports of intermediate and capital goods. 202 THE EXPERIENCE OF INDUSTRIALIZATION Interindustry Linkages and the Complexity of Production This section explores the implications of the observation that the share of total output devoted to intermediate use increases markedly as part of the structural changes that accompany growth. First, we study the relation between output growth and the rate of change of the intermediate use ratio at the sectoral level and separate the influence of differences in output composition on the intermediate input use ratio. Second, we examine how the observed increases in the use of intermediate inputs are related to changes in input-output coefficients. Such increases in interindustry link- ages were first studied by Hirschman (19 58), who focused on forward and backward linkages to identify key sectors that, in his view of "unbalanced growth," would spread an initial growth impulse in specific sectors throughout the economy. We develop linkage measures that are used to compare interindustry linkages in the sample of nine economies. Intermediate Input Use, Growth, and Output Composition In the discussion of the cross-country simulation model in chapter 3, it was noted that the average share of intermediate use in total domestic demand increases from about 33 to 45 percent during the period of the transformation (some fifty years). The typical change in the intermediate use ratio is about 12 percentage points over a fifty-year period; this implies a change in the ratio of 2.4 percentage points per decade. From table 7-1 (third column), it can be seen that all of the sample except Norway and Turkey have achieved faster rates than this. In figure 3-3, the sample appears to divide into several groups according to two criteria: level of development and role of trade. Colombia, Mexico, and Turkey-large countries with relatively low trade shares-start out with lower initial intermediate use ratios and do not succeed in catching up with the other economies by the end of the period. Colombia is the most open of these three, and it exhibits the largest change in this ratio. Korea and Taiwan both experience dramatic changes in the intermediate use ratio and, by the end of the period, they are similar in structure to one another and to Japan. Israel and Norway-both small, open economies- are more developed initially and end up very close to each other. Norway, which started out with a higher ratio than Israel, changes the least of any country in the sample. There is also a positive relation at the sectoral level between output growth and the rate of change of the intermediate use ratio. Pooling data for three sectors (consumer goods, producer goods, and machinery) and ten economies (treating prewar Japan as a separate entry) for a total of thirty observations gives the following regression for the relation: Gw = -0.024 + 0.485 Gx, (- 3.5) (9.5) INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 203 where Gw is the annual growth rate of the sectoral intermediate use ratio, Gx is the growth rate of sectoral output, and the numbers in parentheses are t ratios. 16 In general, the growth rate of producer goods and machinery exceeded that of consumer goods, and their use as intermediate inputs also ex- panded more rapidly. Taiwan is an exception-in it, consumer goods grew faster-but the relation still holds. Moreover, intermediate demand for consumer goods grew faster in Taiwan than it did for producer goods-a result that arises from Taiwan's export structure and the inclusion of textiles (an intermediate good) with clothing in the consumer goods sector. These results reaffirm the importance of intermediate demand as a source of growth and structural change. The intermediate use ratio depends both on the density of the matrix of input-output coefficients and on the structure of production. To separate these effects, we constructed two standard output vectors designed to reflect the main differences in output composition by type of economy: an outward-oriented vector (an average of Taiwan in 1971 and Korea in 1970) and an inward-oriented vector (an average of Mexico in 1960 and Turkey in 1963). In terms of the typology in chapter 3, the first fits the small-industry-oriented (sM) pattern and the second the large (L) pattern. The principal difference in their composition of production is that the first has a larger share of manufacturing and the second a larger share of primary production and services. These two standard vectors were multi- plied by the input-output coefficient matrices of each economy for every subperiod to generate variations in intermediate demand that depend only on variations in intermediate input technology. A similar decomposition is discussed in chapter 3 for the cross-country model. The results of this exercise are revealing. First, in every case, the SM output structure generates greater demand for intermediate inputs than the L structure, since manufacturing is much more demanding of in- termediate inputs. Second, in every case, the actual intermediate use ratio in the initial year was closer to that generated by the output vector for the L pattern, and in the terminal year it was closer to that generated by the SM pattern. Over time, all the economies became industrialized, moving from primary production to industry. Third, in most cases, well over half the change in intermediate use ratios can be explained by changes in input- output coefficients; the rest resulted from changes in output composition. Fourth, there is significant variation in input-output relations across the sample. Application of the standard output vectors yields consistently higher intermediate use ratios for Korea, Taiwan, Japan, and Israel- ranging from 43 to 50 percent-than for Turkey, Colombia, and Mex- ico-ranging from 31 to 38 percent (shown in table 3-9). 16. Similar results have been found in other studies. See, for example, Chenery, Shishido, and Watanabe (1962) and Vaccara and Simon (1968). 204 THE EXPERIENCE OF INDUSTRIALIZATION Linkage Measures Various measures of linkages have been developed in the literature. 17 If principal sectors are focused on, as suggested by Hirschman (1958), attention is drawn away from the systemic phenomenon of increased interdependence or from the deepening of interindustry links commonly observed over time and across countries. 18 To focus on systemic properties and make comparisons among countries, we develop measures of interin- dustry linkages for the entire system. The approach we use is based on the Leontief inverse matrix. 19 The aggregate linkage measure, L, is defined as L = I'i.r;;f;- 1, where r;; are the elements of the Leontief inverse (I - A)- 1 and f; are the elements of a standardized final demand vector (consisting of shares that sum to one). Each column of the Leontief inverse matrix describes the amount of goods directly and indirectly required to deliver one unit of sectoral output to final demand. The first term of L is a weighted sum of these column sums, with the composition of final demand as weights. Hence it shows the total value of products directly and indirectly needed for the economy to deliver a unit of aggregate final demand. The L measure is defined as this term minus one, which is the value of intermediate goods required to produce an aggregate unit of final demand with given composition. In general, the denser the input-output matrix, the higher is L 20 This measure is sensitive to the composition of final demand. For purposes of comparison across the sample, we use a standard or average final demand vector for a semi-industrial country that is drawn from the cross-country model described in chapter 3. 21 In this way, the comparisons will focus on the input-output structures of the different economies and hold constant their final demand structures. There are two ways to calculate the aggregate linkage measure. One is to use the input-output matrix inclusive of imported intermediates so that all intermediate inputs are included in measurements of overall linkages. A second is to use the input-output matrix exclusive of imports-that is, the domestic matrix-so that only those interindustry linkages arising from domestic industries are included. The difference between the two measures reflects the role of imported intermediates in production. 17. See Rasmussen (1965), Chenery and Watanabe (1958), Yotopoulos and Nugent (1973), Shultz (1982), and Martin and Rodriguez (1979). 18. Chenery and Watanabe (1958); Deutsch, Syrquin, and Urata (1986); and Robinson and Markandya (1973) provide intercountry comparisons of systemwide linkage measures based on comparative input-output data. 19. It is common in this literature to start from the Leontief inverse. Our approach is similar to that of Rasmussen (1965), although there are significant differences. 20. Kubo (1982) discusses this and related measures in detail. He considers their relation to the Frobenius root of an input-output matrix. See also Kubo (1985). 21. It is the final demand vector for a typical country with per capita income of $560. See chapter 4 for the data. INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 205 Table 7-3. Interindustry Linkages of Sample Economies per Unit of Final Demand with Standard Composition Overall Domestic linkages linkages Economy Year (percent) (percent) Colombia 1953 50.0 37.2 1966 65.4 52.3 1970 69.0 53.9 Mexico 1950 54.3 40.5 1960 68.9 51.3 1970 63.9 52.0 1975 69.5 54.2 Turkey 1963 52.1 46.4 1968 56.7 51.5 1973 59.6 52.8 Yugoslavia 1962 82.2 67.9 1966 79.5 61.9 1972 87.3 59.4 Japan 1955 89.9 81.3 1960 94.5 82.7 1965 74.6 82.4 1970 106.3 88.7 Korea 1963 89.9 60.9 1970 89.8 58.7 1973 92.8 54.6 Taiwan 1956 76.5 42.6 1961 85.9 55.0 1966 92.9 55.7 1971 93.7 55.2 Israel 1958 83.7 53.8 1965 78.6 50.5 1972 101.5 48.1 Norway 1953 66.7 40.8 1961 77.9 47.8 1969 87.2 47.6 Note: See text for explanation of the linkage measure. Source: World Bank data; described in Kubo (1983). A Comparison of Interindustry Linkages across Countries The results of computing the two linkage measures-overall and domes- tic-for the nine economies are given in table 7-3. The sample forms two distinct groups. On the one hand, Yugoslavia, Japan, Korea, Taiwan, Israel, and Norway consistently have an overall linkage value of more than 75 (that is, $75 worth of intermediate inputs are needed to support the gross production necessary to deliver $100 worth of final demand of a standard composition). On the other hand, Colombia, Mexico, and Tur- 206 THE EXPERIENCE OF INDUSTRIALIZATION key have values less than 75 in all years. The values for Norway increase during the period and overlap the two groups. The average of the overall linkage measure is about 90 for the first group and 60 for the second. Thus for the same final demand, the first group requires about 50 percent more intermediate inputs than does the second group-a remarkable difference. For all the economies, overall linkages increase systematically over time-which is consistent with the results presented earlier and with conventional views about the process of development. A growing use of intermediate inputs is associated with an increasingly complex economic system, which is characterized by the prevalence of more roundabout means of production typical of developed economies. The first group shows rapid increases in overall linkages, and it approaches the level of Japan by the end of the period. 22 The second column of table 7-3 gives the domestic linkage measure and provides a clue to why Korea, Taiwan, and Israel differ from Colombia, Mexico, and Turkey. If we restrict our attention to domestic linkages, the two groups look similar. The indicator of domestic linkages is about 50 for both groups, with much less pronounced change over time than the overall linkages. Furthermore, the two groups differ much less in domestic link- ages than in overall linkages: the average is 53.5 for the first group and 49.2 for the second. We must look to differences in trade strategies for an explanation. Table 7-4 presents data on the volume and structure of trade. In Korea and Taiwan, the rapid increase in export earnings led to an equally rapid increase in imports, which consisted largely of intermediate and capital goods. 23 Rapid export expansion enabled these economies to introduce at an early stage-and to maintain-systemic linkage structures and, implic- itly, the underlying technologies typical of economies at a much higher level of development. Through imported inputs, they could achieve tech- nological structures that would have been difficult or impossible to achieve through domestic production. 24 Even in the early years before exports expanded, Korea and Taiwan achieved high levels of imported inputs, which they were able to finance through high levels of foreign capital infi.:. '." and foreign aid. Israel also benefited from substantial foreign capital inflow during this period, but with lower shares of intermediate and capital goods imports. 22. Robinson and Markandya (1973) rank Japan as being as "complex" an economy (according to their measures) as was the United States by about 1960. A direct comparison with the United States and with European countries would be helpful but is beyond the scope of this book. 23. Capital goods are defined as machinery and transport equipment. Intermediate goods include rubber and chemical products, coal and oil products, nonmetallic minerals, and basic metals. 24. In later chapters, we explore further the implications of trade strategies for output and productivity growth in an explicitly dynamic framework. INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 207 Table 7-4. The Volume and Structure of Trade in the Sample Economies Share in total Exports Imports imports (percent) b as per- as per- centage centage Capital Intermediate Economy Year of GDP' of GDP' goods goods Colombia 1955 12.4 14.3 (43.8) (33.9) 1966 12.1 15.1 28.9 32.8 1970 14.2 15.8 35.7 33.4 Mexico 1950 14.1 13.9 44.6 27.2 1960 11.3 12.8 50.9 26.5 1970 8.1 10.1 54.9 21.5 1975 7.7 10.9 55.7 21.9 Turkey 1963 5.5 10.3 38.8 27.2 1968 5.3 7.5 39.1 36.2 1973 7.6 10.0 35.6 43.2 Yugoslavia 1962 16.0 17.1 48.3 20.0 1966 19.5 20.5 39.4 25.5 1972 22.0 24.1 36.7 30.1 Japan 1955 10.7 10.1 8.6 12.7 1960 11.1 10.6 8.9 24.5 1965 10.8 9.3 8.5 18.2 1970 11.2 9.8 9.5 17.0 Korea 1963 4.8 16.4 21.4 34.3 1970 14.8 24.9 29.4 26.2 1973 31.7 35.0 32.6 29.9 Taiwan 1955 8.3 12.6 (21.6) (38.2) 1961 12.8 19.9 32.7 23.3 1966 20.6 21.5 31.4 29.1 1971 36.8 34.2 32.5 26.2 Israel 1955 11.5 32.8 (28.2) (17.6) 1965 18.9 31.9 26.6 20.5 1972 27.3 40.1 27.9 39.4 Norway 1955 40.7 43.6 (26.5) (19.9) 1961 39.7 42.6 29.5 20.1 1969 41.2 38.5 28.0 22.7 a. Taken from World Bank (1976). b. Based on World Bank data; described in Kubo (1983). Figures in parentheses correspond to input-output year nearest to 1955 (1956 for Taiwan, 1953 for Colombia and Norway, and 1958 for Israel). In Colombia, Mexico, and Turkey, the overall and domestic linkage measures are not very different. These countries emphasized import sub- stitution and through conscious policy choices discriminated against ex- ports; therefore, their export growth was not able to generate enough foreign exchange to meet their growing import needs. These countries have higher shares of intermediate and capital goods in total imports 208 THE EXPERIENCE OF INDUSTRIALIZATION because, through a variety of controls, they allowed only essential imports. As their levels of foreign capital inflow were also relatively low, quantity restrictions and other import control measures were commonly used to cope with foreign exchange imbalances. Their lack of imports forced these countries to rely on domestic production and domestic technology. Their reliance on domestic technology in turn is reflected in the structures of their interindustry linkages. Given the difficulties of import substitution in intermediate goods, the result was a much lower level of overall linkages. Japan stands out as having exceptionally strong domestic linkages. Considering its low dependence on imports for most manufactured prod- ucts, the small difference between its domestic and overall linkages is not surprising. This high level of domestic linkages was already in place by the mid-1950s. Japan's interindustry structure was by then established, and no significant increases occurred until the end of the period. Yugoslavia is similar to Japan in overall linkages and is less dependent on imported intermediates than are Korea, Taiwan, and Israel. As noted earlier, this probably reflects a development strategy that strongly emphasized indus- trialization, which is typical of many socialist countries that sought to establish a broad industrial base at an early stage. In the late 1960s and early 1970s, the domestic linkages declined somewhat in Yugoslavia while overall linkages increased; this reflects the effects of the import liberaliza- tion that took place during that period (see chapter 6). Finally, Norway exhibits one of the lowest levels of domestic linkages, although its overall linkages are comparable with those of other developed countries. A small country (with a population of 4 million in 1973), it did not try to develop a broadly based industrial structure; it specialized and relied heavily on imports. Also, in the postwar period, Norway adopted increasingly liberal trade policies and reduced its controls on imports. 25 Thus the large discrepancy between its domestic and overall linkage measures reflects both country size and policy choices. A similar explana- tion applies to Israel, where the discrepancy between domestic and overall linkages increased significantly in the 1965-72 episode. There appears to be a strong relation between a country's interindustry structure (and its evolution over time) and the choice of a development strategy. When an inward-oriented, import substitution strategy is adopted, domestic linkages increase but overall linkages remain relatively low. When an open development strategy with high levels of both exports and imports is chosen, overall linkages grow rapidly even though domestic linkages do not change much. High exports permit an economy to import the intermediate input structure of an advanced economy and perhaps facilitates rapid growth. 25. See Balassa (1979a) for further discussion of Norway. INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 209 Openness and Comparative Advantage in Manufacturing International trade plays a variety of roles in industrialization. On the output side, it permits a country to separate or "delink" the structure of production from the structure of demand. In trade theory, this separation is usually analyzed in terms of welfare gains to consumers. But trade also affects the input side. The analysis of the structure of intermediate inputs in the previous section indicates that an important gain from trade is that it allows a country to achieve through imports a more advanced intermedi- ate input technology than would be possible through domestic production alone. This gain arises from the import side only and is consistent with the two-gap view of foreign exchange constraints on growth. The imported intermediate inputs are difficult or impossible to replace with domestic production, especially in the early phases of industrialization. In this section we study further the pattern of trade by comparing the degree of openness across countries and the patterns of comparative advantage within manufacturing. Measuring Openness and Trade Structure To analyze the implications of delinking, we need a measure of the degree of openness of the economy. A natural choice is the ratio of the sum of exports and imports to total supply on the domestic market: T = ~; (E; + M;) I (X; + M;), where i is summed over the fourteen-sector aggregation. If all domestic production is traded, T approaches 100 percent; if it is all sold on the domestic market, T approaches zero. Thus, open economies have high values of T and closed economies have low values. To reap the gains from pursuing comparative advantage, a country must alter its structure of production in response to changes in trade opportuni- ties. Two different types of specialization can be observed. The first type is sectoral specialization with an increasing volume of interindustry trade, that is, high exports in some sectors and high imports in others. This is the pattern predicted by classical (and neoclassical) trade theory. The second type, which is consistent with "neotechnological" or "product cycle" views of trade, and especially with the new theory of international trade under imperfect competition, is specialization through diversification within sectors. 26 We would thus expect to see increasing intraindustry trade with both higher exports and higher imports in the manufacturing sectors. Even though our data base is not sufficiently disaggregated to support a 26. The theory of comparative advantage in imperfectly competitive markets is expound- ed in Helpman and Krugman (1985). 210 THE EXPERIENCE OF INDUSTRIALIZATION detailed analysis of intraindustry trade, it is possible to generate a measure of the extent of such trade. Within a sector, the extent of such trade is measured by the absolute difference between the value of exports and imports. The lower the difference, the more nearly balanced are sectoral exports and imports and the more important is intraindustry trade. Sum- ming over the manufacturing sectors, a summary measure is provided by: B = 1.0 - I; I E; - M; I I ( E; + M; ), where the index i is summed over the manufacturing sectors. In the absence of intraindustry trade, either ex- ports, E, or imports, M, is zero in each sector, the second term will equal one, and B will equal zero. If there is exactly balanced intraindustry trade, E will equal Min every sector and B will equal one (or 100 percent). 27 Table 7-5 gives the indicators of openness, T, and of the extent of intrasectoral trade, B, for the sample of economies and benchmark years. The extent of openness is, as might be expected, related in the initial year to domestic market size: a ranking by initial aggregate GDP is negatively correlated to a ranking by degree of openness (see table 6-1). During the period, however, the economies followed quite different patterns. As discussed in chapter 4, the importance of trade increased dramati- cally in the postwar period, with most semi-industrial economies becom- ing more open. In the sample, however, Colombia, Mexico, and Turkey had low and decreasing (or roughly constant) measures of openness during the period under study. Japan, with an initial value forT about equal to that of Turkey, also registered only a slight increase. All the other econo- mies showed significant increases in openness, with Israel, Korea, and Taiwan the leaders. With the exception of Turkey, every economy in the sample shows an increase, sometimes dramatic, in intraindustry trade. Thus within the manufacturing sector, increasing volumes of foreign trade were reflected for each sector in a rise in both imports and exports. And with the exception of Japan, Yugoslavia, and Norway, every economy started with a very low value for the index of intraindustry trade (in part because of a large trade imbalance within manufacturing). The more developed coun- tries in the group (Norway, Japan, and Yugoslavia) started with more intraindustry trade, which is typical of developed countries. Remarkably, Korea, Taiwan, and (to a lesser extent) Israel caught up with this pattern during the period. In contrast, Colombia, Mexico, and Turkey stand out as having maintained closed economies; they did not achieve much change in terms either of openness or of diversification in their patterns of manu- facturing trade. 28 27. The measure is sensitive to aggregation, so we compute it at the same level of aggregation for all economies. It is also sensitive to the overall balance of trade. See Grube! and Lloyd (1975) for a discussion of the measure and possible alternatives. 28. Part of the explanation for the small change in the degree of intraindustry trade for these countries is that they maintained negative trade balances in the manufacturing sector during the entire period. INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 211 Table 7-5. Openness and Intraindustry Trade of the Manufacturing Sector of the Sample Economies Measure (percent)' T (degree of B (summary openness of of intrasectora/ Economy Year economy) trade) Colombia 1953 17.0 2.4 1966 15.5 21.7 1970 16.9 15.5 Mexico 1950 14.8 18.3 1960 11.7 13.8 1970 10.1 26.7 1975 11.8 28.3 Turkey 1963 7.7 17.2 1968 7.0 8.3 1973 9.3 13.4 Yugoslavia 1962 12.7 64.2 1966 16.4 68.2 1972 22.3 67.9 Japan 1955 7.6 46.0 1960 7.7 52.3 1965 9.1 54.3 1970 10.9 60.1 Korea 1955 15.5 6.2 1963 15.2 31.2 1970 21.7 42.6 1973 30.2 59.3 Taiwan 1956 17.1 21.1 1961 18.4 33.3 1966 25.9 45.4 1971 33.1 63.9 Israel 1958 22.3 26.3 1965 26.9 38.2 1968 33.1 46.7 1972 39.4 39.2 Norway 1953 34.7 44.1 1961 35.6 44.4 1969 41.0 53.7 a. The T and B measures are defined in the text. Source: World Bank data; described in Kubo (1983). Comparative Advantage within Manufacturing In the discussion so far, changes in trade diversification have been considered by using aggregate measures to summarize variations calcu- lated at the fourteen-sector level. We now consider the impact of changes in subsectoral shares in total exports and total imports for the eight 212 THE EXPERIENCE OF INDUSTRIALIZATION manufacturing subsectors. The idea is that such changes reflect a country's "revealed comparative advantage." The cause of such changes may lie in differing factor proportions, as assumed by classical trade theory, or in other neotechnology or imperfect competition explanations. 29 Our interest is less in explaining the changes than in establishing the stylized facts. Table 7-6 presents the results for the eight manufacturing subsectors of each economy during the entire period. With imports and exports treated symmetrically, each entry in the table gives the rank within manufacturing for each subsector in the initial and terminal years, followed by the percentage change in share between the two benchmark years. For exam- ple, the entry for food exports for Israel (the first two columns) indicates that food ranked fourth for both benchmark years among the manufactur- ing subsectors in exports even though the share of food in total manufac- turing exports rose by 4.1 percentage points. The last two rows of the table summarize the change in comparative advantage for each sector for the entire sample by presenting the average of the figures in the corresponding column. Similarly, the last column provides a summary measure of com- positional changes in trade within the manufacturing sector for each economy by summing the absolute values of changes in shares across the sub sectors. An examination of table 7-6 reveals some striking similarities in the pattern of comparative advantage. On the one hand, subsectors that lose in revealed comparative advantage for exports in all nine economies also lose on the import side; that is, subsectors that supply a declining share of manufacturing exports also tend to receive a smaller share of manufactur- ing imports. These subsectors are food, textiles, paper, and other light manufacturing. Nonmetallic mineral products are an exception. On the other hand, subsectors that provide an increasing share of exports also tend to receive an increasing share of imports. These subsectors are rubber, metal products, and machinery. The tendency for diversification reflected by increased intrasectoral trade, which was discussed above, is clearly evident across the subsectors. As the bottom row indicates, processed food (the largest exporting subsector in the initial year for several economies) averages the greatest decline in share of manufacturing exports (11.1 percentage points). Sub- sectors that gain in export share include rubber and especially machinery (which rises an average of 13.9 percentage points). On the import side, machinery, rubber, and metal products maintain the highest import shares throughout the period. The last column in table 7-6 indicates that in general there was a much greater change in the structure of manufacturing exports than of imports. 29. The idea behind the notion of revealed comparative advantage is somewhat con- troversial. See Balassa (1965, 1977b). INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 213 These results are in accord with the economywide results discussed earlier (see table 7-1). Turkey and Yugoslavia are exceptions; the imports and exports of each had about the same magnitude of structural change. Trade and Input Use We finish our discussion of trade by analyzing the relation between trade and sectoral input use. First, we turn to the relation between export expansion and the requirements for imported intermediate inputs. Second, we explore the relative contribution to changes in the balance of payments at the sectoral level of changes in intrasectoral trade (sectoral exports and imports) and in sectoral demand for imported intermediate inputs. In this analysis, we focus on the manufacturing sector both as a source of foreign exchange through exports and as a demander through direct and induced imports. Import Content of Exports At the economywide level, we have discussed the importance of in- creased imports. We noted that increased intermediate imports permitted the deepening of input-output linkages, a form of technological change characteristic of industrialization. One of the advantages of pursuing an export-led development strategy is that foreign exchange earnings can be used to increase intermediate imports and so hasten the deepening process, which would otherwise require domestically produced intermediates. In- deed, it is often argued that the expansion of manufactured exports requires more "advanced" production technologies and high-quality in- termediate inputs, both of which must be imported since-at least in the initial stages of export expansion-domestic producers cannot manufac- ture them. In this view, intermediate imports embody advanced technol- ogy, and their use is part of the process of technology transfer. To analyze the relation between exports and imported intermediate inputs, we first develop a measure of the direct and indirect import content of exports, or ICE. The measure, discussed in chapter 5, is defined thus: ICE= eAmRdse = eQSe where e =unit row vector (1, ... , 1) Am =matrix of intermediate import coefficients Rd = (I - Ad)-\ the inverse of the matrix of domestic input-output coefficients se = column vector of export shares Q = AmRd. The ICE measure indicates the value of imported intermediates (in base year prices) needed to produce a unit value of exports of a given sectoral composition. In other words, it is the share of the cost of imported Table 7-6. Changing Pattern of Trade in Manufacturing Subsectors Food Textiles Paper - Exports (rank) Imports (rank) Exports (rank) Imports (rank) Exports (rank) Imports (rank) Economy I' Tb Change' I T Change I T Change I T Change I T Change I T Change N ..... -4 33.1 0.6 1 -5.9 Colombia 8 2 7 5 3 5 7 -3.4 1 7 5.5 6 4 4.0 Mexico 1 2 -47.6 7 6 -1.0 2 5 -10.2 6 7 -3.5 8 8 1.0 5 5 -2.2 Turkey 1 1 12.8 4 5 -3.4 2 2 -1.1 6 4 -1.2 7 5 0.1 5 7 -1.1 Yugoslavia 3 6 -5.1 5 5 1.5 4 2 -6.2 4 4 2.0 8 7 1.0 8 8 0.8 Japan 5 6 -4.4 1 5 -16.5 1 5 -27.5 6 6 3.9 8 8 0.1 7 7 0.7 Korea 3 6 -17.7 5 6 -5.4 2 2 11.0 3 4 -5.7 7 7 1.4 6 7 -1.6 Taiwan 1 4 -67.4 4 6 -2.6 2 1 22.3 5 4 7.4 7 3 0.6 7 7 -0.2 Israel 4 4 4.1 2 4 -11.3 3 2 13.5 5 5 0.7 7 7 -0.1 7 8 -1.6 Norway 4 5 -7.8 5 6 0.4 7 7 2.2 3 3 -2.8 1 4 -13.1 8 7 2.1 Average change, all economies -11.1 -4.2 1.2 -0.3 -0.4 0.1 Average rank, all economies 3.3 4.0 4.4 5.3 2.7 3.4 4.8 4.9 6.0 6.9 6.5 6.7 Light industry Rubber Nonmetallic minerals Exports (rank) Imports (rank) Exports (rank) Imports (rank) Exports (rank) Imports (rank) Economy I T Change I T Change I T Change I T Change I T Change I T Change Colombia 6 5 5.4 4 6 -3.3 3 1 26.3 3 2 7.8 4 8 -36.3 8 8 -0.2 Mexico 3 6 -1.3 4 4 -0.4 4 3 11.0 3 2 -1.8 6 1 2.1 8 8 -0.6 Turkey 6 7 0.2 8 8 -0.4 4 4 -6.4 2 2 12.5 5 3 2.6 7 6 -0.3 Yugoslavia 2 5 -5.6 6 7 0.1 6 4 4.6 2 2 3.1 7 8 -1.5 7 6 1.7 Japan 2 4 -5.7 5 4 4.2 6 3 5.3 3 1 5.1 2 7 -2.2 8 8 -0.1 Korea 1 3 -24.1 4 5 -3.8 8 5 5.7 1 2 -17.3 6 7 1.4 8 8 -0.4 Taiwan 5 3 11.4 6 5 2.5 4 5 5.7 1 3 -14.4 8 7 1.3 8 8 -1.6 Israel 1 1 -22.7 6 6 -0.3 2 3 0.7 4 3 3.1 6 6 -3.0 8 7 1.4 Norway 6 6 1.0 6 5 4.7 2 2 -4.3 2 2 2.1 8 8 0.9 7 8 -0.3 N ...... Average '-'> change, all economies -4.3 0.4 5.4 -0.1 -3.9 -0.4 Average rank, all economies 3.6 4.4 5.4 5.6 4.3 3.3 2.3 2.1 6.3 6.6 7.7 7.4 (Table continues on the following page.) Table 7-6 (continued) Metals Machinery Absolute change Exports (rank) Imports (rank) Exports (rank) Imports (rank) (percent) Economy I T Change I T Change I T Change I T Change Imports Exports - Colombia 2 4 -28.6 2 3 -2.9 5 6 0.5 1 1 -2.7 141.6 24.90 Mexico 7 4 0.7 2 3 -3.4 1 5 34.1 1 1 12.9 118.0 25.80 Turkey 3 3 -9.5 3 3 3.4 8 6 1.4 1 1 -9.5 34.1 31.80 Yugoslavia 5 3 3.4 3 3 7.0 1 1 -3.0 1 1 -16.2 30.4 32.40 Japan 3 2 3.1 4 3 2.7 4 1 31.3 2 2 0.4 79.6 34.00 Korea 4 4 1.8 7 3 12.1 5 2 20.5 2 1 21.5 83.6 67.54 N ..... Taiwan 3 6 -0.9 3 2 -3.3 3 2 24.4 2 1 12.3 136.7 44.30 0\ Israel 8 6 3.3 3 1 17.3 5 5 4.1 1 2 -8.5 51.5 43.10 Norway 3 1 10.1 4 4 -2.2 5 3 11.4 1 1 -3.7 50.8 18.00 Average change, all economies (percent) -0.7 3.4 13.9 0.7 Average rank, all economies 4.2 3.7 3.4 2.8 4.6 3.0 1.3 1.2 Note: All rankings are for manufacturing subsectors within each economy. See text for explanation. a. Initial year. b. Terminal year. c. Percentage point change in shares. Source: World Bank data; described in Kubo (1983). INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 217 intermediate inputs in the value of exports. The ICE shares are given in table 7-7 for various benchmark years for the economies in the sample. The table also gives the import content of domestic final demand, which is calculated by replacing se with a vector of final demand shares. For those economies for which we did not have data on imported intermediate input coefficients, we approximated the import matrix by assuming the same average import coefficient across each row. Comparing the approxima- Table 7-7. Import Content of Domestic Final Demand and Exports (percent) Import content Domestic Exports Economy Year final demand (ICE) Colombia 1953 7.0 4.1 1966 6.5 4.4 1970 7.3 3.7 Mexico 1950 6.6 5.2 1960 7.4 5.5 1970 6.1 6.5 1975 8.3 10.5 Turkey 1963 3.7 2.7 1968 3.2 2.3 1973 4.7 3.9 Yugoslavia 1962 6.0 9.6 1966 8.6 11.9 1972 14.1 18.7 Japan 1955 4.2 6.6 1960 5.8 9.3 1965 6.5 9.6 1970 8.5 10.0 Korea 1963 11.2 15.8 1970 14.8 18.7 1973 17.9 25.5 Taiwan 1956 9.7 13.6 1961 9.8 12.9 1966 14.3 19.7 1971 17.9 25.0 Israel 1958 12.8 12.1 1965 13.0 11.1 1972 27.0 21.2 Norway 1953 18.7 16.3 1961 22.5 18.8 1969 23.4 21.9 Source: World Bank data; described in Kubo (1983). 218 THE EXPERIENCE OF INDUSTRIALIZATION tion with the measure, given the full matrix, indicates that results differ somewhat with these two methods but that the basic trends stand. 30 An examination of the table reveals three patterns. Israel and Norway are small, open economies with high and stable ICE shares. Colombia, Mexico (up to 1970), Turkey, and Japan are large, relatively closed economies with low and stable ICE shares. Korea and Taiwan are charac- terized by rapidly growing manufactured exports, accompanied by a sharp increase in ICE shares. Yugoslavia is a special case; its ICE share starts low but rises significantly, a pattern between that of the large economies and that of Korea and Taiwan. In the last period in both Turkey and Mexico, when the share of manufactured exports in total exports rose significantly, the ICE share rose also (see table 7-4). These results are consistent with the view that a rapid expansion of manufactured exports requires increased intermediate imports. They are also consistent with the two-gap model, which treats foreign exchange availability as a separate constraint on growth. When import substitution is difficult, imported intermediates are needed to support manufacturing output, especially exports. It is also interesting to note that, in both Korea and Taiwan, the import content of exports is much greater than the import content of final de- mand. In the closed economies (Colombia, Turkey, and Mexico), the pattern is reversed although the differences are not as great. The structure of production needed for an export-led growth strategy uses more im- ports. Also, as discussed above, those economies that pursued an export- led strategy tended to increase intraindustry trade, with increased imports and exports falling into the same sector classification. Manufacturing and the Balance of Trade In table 7-8, we decompose the contribution of the manufacturing sector to the overall balance of trade. The decomposition measure, dis- cussed in chapter 5, is based on the following equation applied to the manufacturing sector: where change in the balance of payments change in manufacturing exports change in manufacturing imports for final demand change in intermediate imports needed by the manufacturing sector. The difference between the first two columns represents the change in net 30. See Kubo (1981) for a discussion of the measures and of the accuracy of the approx- imation. See also Kubo (1985). INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 219 Table 7-8. Contribution of the Manufacturing Sector to the Overall Balance of Trade, Selected Sample Economies in Comparison with Cross-Country Model (percentage of total manufacturing exports) Change In In In In inter- balance Economy and total final mediate of subsector exports' importsb imports' paymentsd Mexico 1950-60 Food processing 75 -3 -20 98 Consumer goods -14 3 78 -95 Producer goods 39 30 217 -208 Machinery 0 326 199 -525 Total manufacturing 100 356 474 -730 1960-70 Food processing -29 95 28 -152 Consumer goods 14 47 79 -112 Producer goods 43 -9 125 -73 Machinery 72 322 162 -412 Total manufacturing 100 455 394 -749 1970-75 Food processing 2 -7 44 -35 Consumer goods 19 16 16 -13 Producer goods 32 37 94 -99 Machinery 47 186 150 -289 Total manufacturing 100 232 304 -436 Japan 1955-60 Food processing -6 16 -9 Consumer goods 37 -19 25 31 Producer goods 21 -19 77 -37 Machinery 41 5 10 26 Total manufacturing 100 -39 128 11 1960-65 Food processing 2 -1 12 -9 Consumer goods 8 -2 13 -3 Producer goods 46 -31 42 35 Machinery 44 5 3 36 Total manufacturing 100 -29 70 59 1965-70 Food processing 1 -3 6 -2 Consumer goods 16 1 16 -1 Producer goods 29 -25 46 8 Machinery 54 6 8 40 Total manufacturing 100 -21 76 45 (Table continues on the following page.) 220 THE EXPERIENCE OF INDUSTRIALIZATION Table 7-8 (continued) Change In In In In inter- balance Economy and total final mediate of subsector exports' importsb imports' paymentsd Korea 1955-63 Food processing 16 4 71 -59 Consumer goods 51 -26 6 71 Producer goods 28 -18 61 -15 Machinery 5 83 22 -100 Total manufacturing 100 43 160 -103 1963-70 Food processing 5 2 15 -12 Consumer goods 76 4 58 14 Producer goods 10 2 32 -24 Machinery 9 43 16 -50 Total manufacturing 100 51 121 -72 1970-73 Food processing 2 -1 10 -7 Consumer goods 51 3 17 31 Producer goods 18 0 20 -2 Machinery 29 14 22 -7 Total manufacturing 100 16 69 15 Taiwan 1956-61 Food processing 38 -11 10 39 Consumer goods 45 10 40 -5 Producer goods 13 -53 27 39 Machinery 4 57 15 -68 Total manufacturing 100 3 95 5 1961-66 Food processing 46 6 10 30 Consumer goods 15 3 23 -11 Producer goods 24 4 33 -13 Machinery 15 35 15 -35 Total manufacturing 100 48 81 -29 1966-71 Food processing 5 1 8 -4 Consumer goods 49 4 20 25 Producer goods 15 3 14 -2 Machinery 31 14 20 -3 Total manufacturing 100 22 62 16 Israel 1958-65 Food processing 7 4 19 -16 Consumer goods 70 9 26 35 Producer goods 22 9 33 -20 Machinery 1 37 26 -62 Total manufacturing 100 59 104 -63 INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 221 Change In In In In inter- balance Economy and total final mediate of subsector exports' importsb imports' paymentsd 1965-72 Food processing 17 5 20 -8 Consumer goods 51 13 25 13 Producer goods 21 85 36 -100 Machinery 11 61 36 -86 Total manufacturing 100 164 117 -181 Cross-country model' Period 2 ($280-560) Food processing 17 12 20 -15 Consumer goods 37 9 20 8 Producer goods 42 19 24 -1 Machinery 4 54 7 -57 Total manufacturing 100 94 71 -65 Period 4 ($1,120-2,100) Food processing 20 4 11 5 Consumer goods 28 5 14 9 Producer goods 28 7 24 -3 Machinery 24 24 5 -5 Total manufacturing 100 40 54 6 Note: Economies included are those for which detailed import data are available. a. For four manufacturing subsectors. b. For final demand for four manufacruring subsectors. c. Imports from all subsectors including nonmanufacturing. This column measures change in imported intermediates needed to sustain total sectoral production-not just production for sectoral final demand. d. Difference between first column and sum of second and third columns. e. See chapter 3 for the model. Source: World Bank data; described in Kubo (1983). trade in the subsector excluding intermediate import demand from other subsectors. 31 Table 7-8 shows that, in general, even with growth in manufacturing exports, the manufacturing sector exerts a net drain on the balance of trade when both direct and indirect effects are taken into account. The pattern in Korea and Taiwan is revealing. At the time of rapid acceleration of manufacturing exports (the second episode in both cases), there was 31. Because it includes only the imports for final demand, the difference does not measure intrasectoral trade. 222 THE EXPERIENCE OF INDUSTRIALIZATION also a dramatic increase in induced intermediate imports to support manufacturing production, coupled with an increase in intrasectoral trade. As a result, in 1963-70 in Korea, a $100 increase in manufacturing exports is associated with a $72 worsening in the balance of trade. In 1961-66 in Taiwan, the effect is a $29 worsening in the balance of trade. Only in the final episode in both Korea and Taiwan does the manufactur- ing sector make a positive contribution to the change in the balance of trade because of a decline in the change in both intermediate and final imports. This is consistent with the view that these economies achieved successful import substitution in basic intermediates and in capital goods during the last phases of their export-led growth strategies. The causes of this net negative contribution of manufacturing to changes in the balance of payments under an export-led development strategy are difficult to disentangle. An economy pursuing an export-led strategy generally has a high foreign capital inflow, which permits in- creased imports. There is a correlation between large imports of intermedi- ate goods and rapid growth, which is consistent with the two-gap model of foreign exchange constraints. Whether the causal chain runs from higher exports to higher imports to higher growth, or vice versa, or both, is impossible to determine from this analysis. These links are a central concern of later chapters. Postwar Japan is a large country which decreased its manufacturing imports for final demand in absolute terms throughout the period. The result is that, even though induced changes in intermediate imports were large, the net contribution of changes in exports to the balance of trade was positive. Mexico, in contrast, is representative of countries that have pursued inward-oriented development strategies. Increases in manufacturing ex- ports were associated with large increases in intermediate imports to support the expansion of manufacturing. All three episodes also saw large increases in final demand imports of machinery, which were undoubtedly mostly for investment. Thus, unlike Korea and Taiwan, Mexico was unable to reduce its need to import intermediate and investment goods. This inward-oriented development strategy led to an industrial structure highly dependent on crucial imports. As a result, economies such as Mexico and Turkey are more susceptible to a shock emanating from the foreign sector than are more open economies because they have little room for adjustment. 32 This result is also consistent with the analysis of linkages above. Israel is an example of a small, open economy. With its expansion of manufacturing, including increases in exports, intermediate imports grew even faster than in Korea and Taiwan. Changes in final demand for 32. For an analysis of the reaction of Turkey and Yugoslavia to foreign shocks, see Dervis and Robinson (1982), Celasun (1983), and Robinson and Tyson (1984). INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 223 imports were also large and increased over time, so the net effect on the change in its balance of trade was very negative. In this case, which is not typical of semi-industrial countries, the main driving force was the very high level of foreign capital inflow, including aid. The cross-country model provides a convenient comparison. The two periods tabulated are those that begin and end the industrialization phase. In the earlier period, manufacturing growth and increases in manufac- tured exports are associated with large increases in machinery imports for investment, which leads to a negative effect on the change in the net trade balance. In the later period, increases in machinery imports for final demand are much less; and the effect of changes in manufactured exports on the trade balance are positive even though changes in intermediate imports from manufacturing growth are the same in both periods. The standard pattern of the cross-country model thus behaves more like a moderated version of Korea and Taiwan than like an inward-oriented economy such as Mexico or Turkey. Conclusions The postwar experience of semi-industrial countries with growth and structural change reveals a variety of development paths. All these coun- tries industrialized and in general did so at faster rates than would have been expected from historical comparisons with the developed countries. The nine sample economies exhibit a wide range of initial conditions and development experiences during the period. Their experiences are typical but span the range of semi-industrial countries. They thus provide enough diversity to permit some tentative conclusions to be drawn. The process of industrialization is associated with structural changes beyond a simple increase in the share of manufacturing in output. Changes in the structure of final demand, in international trade, and in the use of intermediate inputs all contribute to the evolution of an economy. Although some basic long-run forces are common to all countries, differ- ences in initial conditions and in the choice of a development strategy affect how these three components interact and the rate at which the process unfolds. The main common forces include changes in the structure of demand as per capita income increases and changes in technology affecting both the primary factors (land, labor, and capital) and intermediate inputs. Engel's law-that with a rise in income the structure of demand moves strongly in favor of manufactures-is clearly valid for these economies; such changes in demand are a strong force for industrialization. Equally important are the increases in demand for intermediate inputs that accompany growth. Such increases arise both from a shift in the structure of production toward manufacturing subsectors that are heavy users of intermediate inputs and from changes in technology that lead to a more specialized and complex economy. 224 THE EXPERIENCE OF INDUSTRIALIZATION International trade has been very important for the semi-industrial countries in the postwar period. The countries that chose and were able to pursue open development strategies based on manufactured exports grew faster and achieved more rapid rates of structural change than did those that followed inward-oriented strategies. 33 Their improved performances are related to two different effects of international trade on an economy. On the export side, trade permits a country to specialize and thereby expand production in specific sectors beyond the constraints imposed by limited domestic demand. Expansion of manufactured exports, in turn, may promote economies of scale and hence more rapid industrialization. Thus exports permit faster change in the structure of production in coun- tries that pursue open strategies than in those that pursue inward-oriented ones. On the import side, equally important forces are at work. An increased availability of foreign exchange-from whatever source-enables an economy to import intermediate and capital goods that are difficult if not impossible to produce domestically at a relatively early stage of develop- ment. Industrialization entails the increasing use of "modern" intermedi- ate inputs, but the needed technology can be acquired much faster by increasing imports than by expanding domestic production alone. Accord- ingly, the economies committed to open development strategies changed their input-output structures faster than did those that followed in- ward-oriented ones. To achieve an open development strategy with strong reliance on manu- factured exports calls for more than an appropriate choice of policies. Initial conditions are important, and more than just trade policy is in- volved. Economies that enjoyed high rates of growth of manufactured exports typically began with relatively high shares of manufacturing in total output (except for Korea, which started its spurt with a relatively low share). There is also some evidence of sequencing in that the economies successfully pursuing open development strategies did so after a period of significant import substitution. If it is to succeed, export expansion in manufactures evidently requires a substantial industrial base that includes increased intermediate linkages (which Korea achieved, even though the initial share was low). The development of such a base usually entails a period of significant import substitution, although it is possible to import some of the technology as the process unfolds. These conclusions are consistent with arguments for supporting infant industries, especially if they generate technology transfers and interindustry linkages. 34 33. We have not considered the primary exporter pattern since, with the possible excep- tion of Colombia, none of the sample fits this pattern. See chapter 4 for a broader discussion of alternative strategies. 34. Westphal (1982) provides a related argument for the protection of infant industries. See also the analysis of the relation between total factor productivity growth and choice of development strategy in chapter 11 as well as Teitel and Thoumi (1986). INTERDEPENDENCE AND INDUSTRIAL STRUCTURE 225 An open development strategy promises benefits by delinking the struc- ture of production from the structure of domestic demand. Countries can pursue two types of specialization: intersectoral, with increased exports in some sectors and increased imports in others; and intrasectoral, with increased imports and exports in the same sectors. Both trends are evident in the sample, but increases in intrasectoral trade are especially marked. Also evident are both specialization in export sectors and increased re- liance on imported intermediates and capital goods within the manufac- turing subsectors. The net effect of manufacturing growth on the balance of trade is mixed. Rapid industrialization tends to increase the demand for imported in- termediates, and manufactured exports are often concentrated in the sectors most dependent on imports. The net effect of manufacturing on the balance of payments is negative for much of the industrialization process. Only in the later stages, as an economy achieves import substitution in the "harder" heavy manufacturing sectors, does the net contribution to the balance of trade of changes in manufacturing turn positive. In an export-led or open development strategy, export performance, foreign capital inflows, imports, and aggregate growth interact. Although the nature of the causal links among these forces is difficult to disentangle, it is apparent that an open development strategy, if successful, demands a difficult balancing of forces and careful timing. The gains from success are great, but the difficulties of managing the required structural changes and concomitant pressures on the balance of trade are also great. PART III Productivity and Structural Change PART I I usED a common input-output framework to compare the experi- ence of industrialization in eight semi-industrial economies. In part III, the analysis is expanded to cover the growth of factor productivity and changes in the structure of production and factor use. As in other parts of the book, a variety of approaches is presented and then applied to diverse samples determined by the availability of comparable information. The growth of factor productivity for the whole economy often includes a structural component that arises when resources are reallocated from activities of lower productivity to activities of higher productivity. Two types of reallocation are explored: the shift of labor and other inputs from primary production to manufacturing (chapters 8 and 9) and the increase in the ratio of production for export to production for the domestic market (chapters 9 and 10). An increase in the weight of the export sector, especially of manufactured exports, can accelerate aggregate productivity if marginal factor products are higher in that sector than elsewhere in the economy. In chapters 9 and 10, it is argued that an outward orientation- which is associated with increasing the share of production for export- leads directly to higher rates of total factor productivity growth at the sectoral level in addition to contributing to higher aggregate productivity through the narrowing of productivity differentials. Of the sources of industrialization on the supply side, differential pro- ductivity growth is probably the best established. During industrializa- tion, productivity growth is expected to be higher in manufacturing than in other sectors; within manufacturing, a higher rate is commonly ex- pected in heavy industry than in light industry. Such expectations are largely confirmed in chapters 8 and 10. Surprisingly, however, both chap- ters identify an even stronger country and period effect. The 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 227 228 PRODUCTIVITY AND STRUCTURAL CHANGE countries in periods of rapid growth of aggregate productivity. This finding suggests that the overall economic environment, which includes general macroeconomic and trade policies, is an important factor in explaining differences in productivity growth. Chapter 8 extends the cross-country model that was described in chap- ter 3 and considers explicitly the long-term patterns of factor accumula- tion and productivity growth. Expanding the analysis to the supply side in a disaggregated framework, it offers an estimate of the contribution of resource reallocation and other effects of structural change to the growth of productivity and output. Chapter 9 presents a statistical analysis of the sources of growth. It focuses on a much shorter period (1964-73) but broadens the coverage to all the semi-industrial countries identified in chapter 4. The results show that during this period the semi-industrial countries differed significantly in performance and in the determinants of growth from both the lower- income countries and the industrial countries. Chapter 10 estimates the growth of total factor productivity within manufacturing at the sectoral level in four of the countries in our sample. The variation in productivity growth among sectors is shown to be related to output growth. This is not, however, the result of the simple Verdoorn effect, which implies that any expansion of the market, regardless of source, improves a sector's productivity performance. Rather, it appears to be a rise in exports that leads to higher rates of total factor productivity growth, while greater import substitution may even hamper an improve- ment in productivity. 8 Productivity Growth and Factor Reallocation MOSHE SYRQUIN CHAPTERS 2 AND 3 provided a framework for analyzing the relation between the structural transformation of an economy and the growth of its per capita income. In chapter 3, a simple multisectoral model designed to trace the effects of changes in demand and trade on the structure of production and factor use was presented. The estimation of the model from cross-country data revealed some typical patterns of transformation, which were then used as benchmarks in comparing the experiences of various economies. This chapter, building on that analysis, considers the typical patterns of productivity growth within sectors and the reallocation of resources among sectors during successive stages of transformation. This dynamic extension of the industrialization model, also based on cross-country data, becomes a frame of reference for evaluating country experience in this chapter. The relations on the supply side in the countries studied between sectoral accumulation and productivity growth on the one hand and the level of development on the other hand are more erratic and less well documented than are the relations on the demand side underlying the simulations in chapter 3. Also, the country information available for analyzing these relations is scarcer and less reliable. The results, therefore, are more speculative than those in chapter 3; only broad trends and orders of magnitude are emphasized. As in an earlier study (Chenery and Syrquin 1975), the stylized facts and the simulated links are regarded as reduced forms consistent with a variety of econometric structures. The range of possible structures is then nar- rowed by considering not only production but also demand and produc- tivity in a general equilibrium framework. The reduced forms also reveal the implications of partial assumptions about parameters and trends and provide useful starting points for structural modeling (as is done in chapter 11). The cross-country model of transformation is designed to cover the complete range of the transition. Although it is based on information referring mainly to the postwar period, it illustrates the interaction among the sets of processes that define economic development from the onset of self-sustained growth to the advanced stage characteristic of Western European countries in recent decades. It also helps to identify a transition 229 230 PRODUCTIVITY AND STRUCTURAL CHANGE stage of growth acceleration in output and productivity growth during which the structure of the economy is radically transformed. Such struc- tural transformation is not well captured in price endogenous models, which are usually calibrated to replicate observed change during a limited portion of the transition only. This chapter begins by deriving the rate of growth in the model and goes on to estimate factor productivity growth and the contribution of resource reallocation to aggregate growth. The final section returns to a considera- tion of the interactions between demand and supply that was begun in chapter 2. The Growth of Output and Factor Use The growth of aggregate output depends on the growth of inputs and on the efficiency of their allocation and use. A convenient summary of the outcome of the growth process is the Harrod-Domar accounting equation relating the accumulation of capital to the productivity of capital. Let IIV stand for the share of gross investment in GDP, 5 for the ratio of deprecia- tion to output, and h* for the incremental capital-output ratio (ICOR). At every point in time, aggregate value added growth, Gv, satisfies the equation G _ IIV- 5 (8-1) v- h* The two elements in equation 8-1, net investment and the capital coefficient, both vary systematically with per capita income. A rise in investment was shown in chapter 3 to be a main component of the shift in the structure of demand. The share of depreciation also appears to in- crease with per capita income (Kuznets 1966), but without significantly affecting the rise in the share of net investment in GDP. The evidence for the aggregate ICOR is derived indirectly from the variation in capital-output ratios at the sectoral level. The reduced-form specification of sectoral capital requirements in equation 3-16, which relates capital-output ratios to income, is the most natural one for the input-output system in chapter 3. The approach to factor requirements usually used by planners, it is easy to implement with commonly available data. There are virtually no guidelines from economic theory about what to expect for the secular variation in capital-output ratios. From the empiri- cal work of Kuznets (1961b) and others and from recent cross-country data, several systematic trends-or stylized facts-for broad sectors can be suggested. In agriculture, the capital-output ratio, excluding land, is rel- /atively low in the initial stages and increases with income. The opposite 1 pattern is typical for social overhead, the capital coefficient of which declines with rising income from a very high initial level dictated by indivisibilities in basic infrastructure. In manufacturing, although there PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 231 Table 8-1. Labor and Capital Requirements per Unit of Gross Output for the Cross-Country Model Per capita income (dollars) Sector 140 280 560 1,120 2,100 3,360 5,040 Capital coefficients (capital per unit of output) Agriculture 1.00 1.20 1.44 1.72 2.03 2.18 2.32 Mining 0.70 0.86 1.06 1.31 1.58 1.70 1.81 Manufacturing 0.63 0.66 0.69 0.71 0.74 0.75 0.76 Food processing 0.45 0.45 0.45 0.45 0.45 Consumer goods 0.70 0.70 0.70 0.70 0.70 Producer goods 1.00 1.00 1.00 1.00 1.00 Machinery 0.65 0.65 0.65 0.65 0.65 Social overhead 4.30 3.88 3.50 3.15 3.00 2.93 2.87 Services 1.30 1.27 1.24 1.21 1.19 1.19 1.19 Capital-output ratio (K/X) 1.44 1.45 1.43 1.37 1.32 1.31 1.32 Capital-value added ratio (KIV) 2.17 2.34 2.43 2.43 2.40 2.38 2.36 Labor coefficients (workers per million dollars of output) Agriculture 3,610 2,150 1,280 685 300 163 95 Mining 2,100 917 400 175 84 50 32 Manufacturing 678 343 167 85 45 29 20 Food processing 280 168 98 63 42 Consumer goods 1,155 540 252 126 70 Producer goods 770 315 133 57 28 Machinery 980 413 175 76 35 Social overhead 840 470 260 140 84 53 36 Services 890 590 385 252 160 108 76 Participation rate (LIN) 0.352 0.388 0.419 0.426 0.400 0.400 0.417 - Not available. Source: Syrquin (1986b). are few clear trends, the capital-output ratio tends to be higher for pro- ducer goods than for light industrial products or machinery. For the ~/ relation of sectoral capital-output ratios to income, see table 8-1. For the economy as a whole, the capital-output ratio, h, is a weighted average of the sectoral h;'s, with the output weights, p;, derived in chap- ter 3: (8-2) The marginal capital-output ratio, h'', can then be derived from the average h and substituted into equation 8-1 to determine the implied growth rate for any income level. 1 Use of the Harrod-Do mar growth equation 8-1 does not imply the adoption of a specific growth theory or the assumption of a technologi- 1. The growth rate for an income interval in table 8-2 below is set at the mean value of the instantaneous initial and terminal rates. Table 8-2. Growth Rates of Aggregate Output and Inputs of the Cross-Country Model Initial year's values Annual growth rate (percent) Per Net Length capita invest- Per Labor of time income ment Popu- capita Value produc- interval Period (dollars) share !COR' lation income added Capital Labor tivity (years) Stage 0 100-140 8 2.20 2.55 1.26 3.81 3.90 2.56 1.25 27 N 1 140-280 10 2.30 2.78 2.02 4.80 5.03 3.06 1.74 35 w 2 280-560 13 2.42 2.50 3.17 5.67 5.84 2.85 2.82 22 N 3 560-1,120 15 2.45 2.20 4.10 6.30 6.29 2.30 4.00 17 ~ II 4 1,120-2,100 16 2.40 2.00 4.58 6.58 6.52 1.81 4.77 14 5 2,100-3,360 16 2.39 1.50 4.71 6.21 6.11 1.40 4.81 10 } IIJ 6 3,360-5,040 14 2.39 1.00 4.60 5.60 5.50 1.47 4.13 9 Note: Chapter 3 describes the cross-country model; table 3-3 describes periods. Growth rates are continuously compounded. They are computed by lit In X y!X0 , where 0 and T are initial and terminal years for any variable X. The growth of per capita income for a period is set equal to the mean of the instantaneous growth rates for the initial and terminal benchmarks, calculated by the ratio of the net investment share to the !COR'. The corresponding figures for the last income benchmark ($5,040) are 13 percent for investment and 2.39 for the !COR. I, II, and III correspond to the three stages in figure 3-7. a. ICOR is the incremental capital-output ratio. Source: World Bank data. PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 233 cally fixed capital coefficient. The growth equation is an ex post identity implicit in the set of reduced-form relations presented above and in chapter 3. It is one of a set of relations consistent with the observed average patterns of structural change. Other patterns may relate differently to growth, but such differences are difficult to infer from the present model. A structural model making explicit the behavioral and technological rela- tions is needed for such extrapolations. The main predetermined variable in this model is per capita income. To derive its growth rate from the growth of total output, the pattern of population growth at various income levels must be estimated. Demographers and economists have described the fall in birth and death rates with an increasing level of development as the "demographic transi- tion." Historically and in cross sections, the fall in mortality rates has preceded the decline in fertility; the result is an initial acceleration of population growth followed by a slowdown. Here I assume a smooth transition incorporating these observations. 2 The specification of a growth rate of income for each interval converts the sectoral and total variations in output and inputs from chapter 3 into growth rates. The aggregate results appear in table 8-2. According to the illustrative calculations in the last column of this table, presented as rough orders of magnitude only, it takes about sixty years for a typical country to go through the first two periods and about fifty more for it to complete the industrializing stage represented by periods 2 through 4. 3 The use of purchasing power parities instead of exchange rates as the conversion factor reduces the income gap over the full range of the transition by about 30 percent. At the real growth rates in table 8-2, the time it takes for a typical country to go through the periods in the table is also reduced by about 30 percent (see appendix to chapter 3). Growth Acceleration The growth rates of per capita income and total value added in table 8-2 show an initial acceleration followed by a slowdown at higher income 2. Recent research has advanced our understanding of the determinants and timing of the demographic transition. The World Development Report 1984 (World Bank 1984) presents the state of the art. Although the population growth rates in table 8-2 seem to be representa- tive of the long-run demographic transition, the timing in specific cases has shown great variation around the trend. For specific examples, see World Bank (1984, figs. 4-1, 4-2). 3. The income range in the si-mulations of chapter 3 starts at $140 in 1970 dollars, equivalent to $100 in 1964 prices (see chapter 3 ). At this income level, the share of agriculture in value added has already come down substantially, and net investment accounts for about 10 percent of GOP. In this chapter, which deals with factor use and productivity growth, it proved convenient to illustrate an even earlier period where agriculture takes up about half total value added, net investment is below 8 percent, and growth is significantly lower. The basic information on investment and on the sectoral distribution of value added and employ- ment was taken from Chenery and Syrquin (1975, pp. 20-21). There, we presented average figures for countries with levels of income below $100 per capita. The mean income for this group was close to $70 in 1964 prices, which amounts to about $100 at 1970 prices. The period $100-$140 is labeled period 0. The other six periods are as in chapter 3. 234 PRODUCTIVITY AND STRUCTURAL CHANGE levels. The acceleration is the result of the increase in the rate of capital accumulation and of productivity growth (or technological change), which prevents the capital-output ratio from rising. In this long-term model the investment rate and the productivity of capital vary with the 1 level of income but not with its growth rate. Various authors have argued, however, that saving and investment rates increase with the growth rate, while capital-output ratios decrease with it. 4 Incorporating these virtuous feedbacks would result in faster acceleration than that shown in table 8-2. Growth acceleration is an important aspect of theories of economic development. Evidence of long-term acceleration of growth comes from the early experience of today's developed countries and from comparisons among countries grouped by income level. First, as Kuznets (1971) points out, when today's developed countries entered the stage known as modern economic growth, they must have experienced a significant acceleration of growth. Backward extrapolations of the growth rates during the epoch of modern economic growth suggest implausibly low levels of income in the not too distant past. In addition, for the more recent case of Japan, growth acceleration has been thoroughly documented and studied (Ohkawa and Rosovsky 1973, among others). The 1950-75 period, for which the model is calibrated, was one of very fast growth in almost all regions and groups of countries. This accelera- tion, however, was in part a response to the disruption of world trade and production in the preceding period and to the destruction of the war. It cannot be explained only by acceleration during the transition to indus- trialization, since it encompassed countries at all income levels. Indirect corroboration of growth acceleration at medium income levels comes from cross-country comparisons of growth over shorter periods of time. Table 8-3 presents the experience of groups of countries, ranked by income, since 1950. The middle-income group as defined in the World Development Report 1983 and the transitional and newly developed groups as defined by Chenery (1977) correspond roughly to the semi- industrial group described in chapter 4. (Italy and Japan are included in the newly developed group.) In Morawetz's classification, the third and fourth groups-developing countries with a per capita income of more than $520-correspond to the semi-industrial group. In every column in the table, growth is faster in the middle-income groups and slows down some at higher income levels. This pattern holds for both total and per capita income, and it holds for every subperiod. The pattern of growth in the model presented in table 8-2 agrees closely with this postwar configuration. 4. The effect on saving can be derived from the permanent income theory and the life cycle theory. Simple acceleration models imply a similar relation for investment. For the inverse relation between capital-output ratios and the growth rate, see Vanek and Studenmund (1968) and Chenery and Eckstein (1970). PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 235 Table 8-3. Growth of Total and Per Capita Income of All but Centrally Planned Countries by Income Group Per capita Annual growth Annual growth income of rate of per capita rate of GDP country group income (percent) (percent) Morawetz (1977) classification (197 5 dollars) 1950-60 1960-75 Less than $265 1.4 1.0 $266-$520 2.2 2.3 $521-$1,075 2.4 3.7 More than $1,075 (developing) 3.2 5.8 OECD 3.0 3.5 Chenery (1977) classification 1950-60 1960-73 1950-60 1960-73 Less developed 2.0 1.8 4.2 4.3 Transitional 2.7 3.9 5.6 6.8 Newly developed 5.7 7.0 6.9 8.1 Old developed 2.6 3.4 3.9 4.4 World Development Report 1983 classification 1960-81 1960-81 India 1.4 3.5 China 3.4' 5.4 Other low-income 0.8 4.1 Lower middle-income 3.4 5.3 Upper middle-income 4.2 6.0 Industrial market economies 3.4 4.0 a. Computed from the growth rates of GDP and population. Sectoral Growth Rates Aggregate growth is a summary of sectoral growth. The main purpos~ of this chapter is to show how variations in aggregate growth result from\ what happens at the sectoral level and from the shift in sectoral weights in the generation of output and in factor use. ,, The comparative static runs of the cross-country model in chapter 3 derived the sectoral structure of production for each income benchmark. Sectoral factor requirements were determined by income-related factor- output coefficients in equations 3-16 and 3-17. As with the capital-output ratios, the pattern of labor-output ratios is based on the observed associa- tion of these ratios with per capita income across countries and over time. Certain stylized facts about the changes in labor productivity (the inverse of the labor-output ratio) appear to be reasonably robust. First, labor\ productivity increases in virtually all productive sectors but not at a \ uniform rate. Second, the level of labor productivity in agriculture is uniformly lower than in manufacturing or services. The gap in productiv- 236 PRODUCTIVITY AND STRUCTURAL CHANGE ity commonly widens in the initial stage and then begins to narrow as the ( increase of labor productivity in agriculture accelerates, exceeding the I average growth. The increase in productivity at this stage is the result of faster output growth and an eventual absolute decline in labor in agricul- ture. Third, average labor productivity is higher, and increases at a faster rate, in heavy industry than in light industry. The relation of labor-output ratios and income, based on cross-country data, is given in table 8-1. Tables 8-4 and 8-5 present the growth rates of value added, labor productivity, and factor proportions for three sectors for the cross-country model and for the sample of economies used throughout this book. The uneven expansion of sectors translates into the changes in the structure of production analyzed in chapters 3, 6, and 7. The relation between aggregate and sectoral growth can be derived from the definitions of total output, V, and labor productivity, y: V= l;V; and where 'Y; is the employment share in sector i. Differentiating with respect to time gives the relations between aggregate and sectoral growth rates: Table 8-4. Growth Rates of Output and Inputs for the Cross-Country Model by Sector (average annual growth rates in percent) Period' Measure 0 2 3 4 5 6 Value added Primary 3.03 3.98 4.13 4.11 3.66 3.16 3.24 Manufacturing 5.36 5.65 6.79 7.57 7.84 6.18 4.95 Services 4.15 5.07 5.90 6.37 6.45 6.67 6.23 Total 3.81 4.80 5.67 6.30 6.58 6.21 5.60 Labor productivity Primary 0.74 1.32 2.03 3.54 5.71 6.22 5.83 Manufacturing 1.26 2.05 3.36 4.15 4.33 4.47 4.25 Services 1.31 1.33 2.13 2.77 3.20 4.12 3.85 Total 1.25 1.74 2.82 4.00 4.77 4.81 4.13 Capital-labor ratio Primary 0.73 2.11 3.32 5.03 7.18 7.26 7.02 Manufacturing 1.21 2.14 3.34 4.27 4.43 4.50 4.20 Services 1.18 1.24 1.94 2.46 3.04 3.84 3.52 Total 1.33 1.96 2.99 3.99 4.71 4.71 4.03 Note: Services includes social overhead. a. Periods are explained in table 3-3. PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 237 (8-3) and (8-4) Gy = l;p;Gy; + l;p;G"Y; where p; V;IV is the share of sector i in total output (GDP). The growth rate of total output, Gv, equals the sum of sectoral growth rates weighted by the sectoral output shares. The growth rate of aggregate labor productivity, GY, has two components. The first term averages the sectoral growth rates of output per worker. The second term measures the contribution to aggregate labor productivity growth of employment shiftS''\ among sectors with different labor productivities. I denote this second term by A(y) and label it the "gross allocation effect." ) When aggregate growth accelerates, manufacturing typically leads the way, growing faster than any other sector. Because of its initial low output share, however, its contribution to growth, PmGm, is at first modest and below that of primary production (see figure 3-7). At low income levels, agriculture is the dominant activity; aggregate growth is therefore closely related to performance in that sector. As manufacturing increases its output share, its faster growth rate pulls up the aggregate growth rate of output and labor productivity. If all sectors had the same production function and faced the same factor prices, and if resources were perfectly mobile, then labor productivity would follow the same pattern in all sectors. In actuality, both in the model simulations and in the observations in figure 8-1, labor productivity is significantly lower in agriculture than in the rest of the economy. In the early stages of development the growth of productivity in agriculture lags behind that of other sectors, which widens further the productivity gap. These sectoral differences in the average product of labor are partly a reflection of differences in the nature of the production function (which lead to different factor proportions) and in the rate of technological change. But they also stem 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, includ- ing handicrafts and services. When the industrial sector accelerates its growth in response to domes- tic demand and to changes in comparative advantage (usually with some help from commercial policies), the productivity gap tends to increase. 5 Labor starts to shift out of agriculture, at first in relative terms and eventually in absolute terms, but with a lag. Since productivity in agricul- ture rises even at this stage, a surplus of labor results. 5. In some cases where manufacturing was heavily protected, the policy-induced rise in the relative price of manufactured goods accounts for part of the acceleration of growth in that sector and for part of the increase in the productivity gap between agriculture and industry. Table 8-5. Growth Rates of Output and Inputs of Sample Economies by Sector (average annual growth rates in percent) Average for japan advanced Colombia Mexico Turkey Yugoslavia Korea Taiwan Israel countries' Measure (1953-70) (1950-75) (1963-73) (1962-72) 1915-35 1955-70 (1963-73) (1961-71) (1958-72) (1963-73) Value added Primary 4.18 4.16 1.77 2.33 1.69 2.03 6.39 3.23 6.23 1.9 Manufacturing 7.41 7.34 8.72 11.55 5.00 14.47 20.15 15.81 12.06 5.9 Services 5.11 6.56 7.67 9.51 4.18 10.83 15.72 9.62 9.00 4.6 Total 5.17 6.24 6.18 8.16 3.49 10.87 13.70 10.49 9.42 4.8 N Labor productivity <..., Oo Primary 3.62 3.43 2.30 3.55 2.04 5.47 6.44 2.69 7.73 6.1 Manufacturing 4.11 2.96 3.94 6.16 3.07 9.40 9.79 7.40 6.81 5.3 Services 1.28 2.34 0.97 10.67 1.58 6.27 6.47 4.59 4.43 2.2 Total 2.99 3.76 4.86 8.57 2.66 8.71 8.81 6.71 5.48 3.7 Capital-labor ratio Primary - 4.01 - - - 10.08 10.20 4.22 8.39 8.1 Manufacturing - 2.00 - - - 8.15 8.76 6.29 3.18 3.8 Services - 1.70 - - 4.83 -0.35 0.20 6.96 1.7 Total - 3.31 - - - 7.81 5.02 2.82 5.94 3.0 - Not available. Note: Services includes social overhead. Employment in Colombia is for 1951-70; labor productivity in Yugoslavia is for 1966-72. a. The advanced countries are Belgium, Canada, Denmark, France, Italy, Netherlands, United Kingdom, and United States. Sources: Employment in Colombia is from World Bank data. The average for advanced countries was computed from the data in Stein and Lee (1977). All other data are from World Bank and are described in Kubo (1983). PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 239 Figure 8-1. Relative Labor Productivity of Sample Economies and. Cross-Country Model ------ , .......... , .... .......... / / /----- ........................ ' ....... ..... , / "'""------,- ,...,..,.. -;><;......' .......... Manufacturing sector (model) -- 1.5 1- - ...... ......... ...... ......... / ' ......... Services sector (model) ......... ......... United Kingdom 1·or- ", ...... ......... ___ Belgium Kor,ea Colom~ Japan ?---~~-- Y"ugoslavia~srael ... T'---....__ .r ,,(prewar~o- ~~ ......... _l~~i.:a_11~--- J '\... ~ ~~~~~~ ,~.~ 0.5 1- ·---~--~---- ~' 4::::Jil::: , .,_ -- -------~"""~ Umted States T ur key · 'M ex1co Japan Primary sector 1 (postwar) (model) o~~·--~--~~U-~--~~~~~~~----~·~'--~~~~~·' 100 140 280 560 1,120 2,100 3,360 5,040 Per capita GNP (dollars) a. Index signifies labor productivity in a sector relative to labor productivity for the whole economy. The pattern of relative productivity in figure 8-1 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 oq the other is greatest in the middle-income range ($500-$1,000), which is\ typically the period of greatest inequality of income (Chenery and others·, 1974). 6 It is also the period when, because of the productivity gap itself,·. resource shifts can make their largest contribution to aggregate growth, as • discussed below. In a second phase, once migration and capital accumulation have signifi- cantly reduced the surplus of labor, relative wages increase and a catch-up process takes place in agriculture. 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 6. The widening gap in the case of Korea and Taiwan reflects their exceptionally fast growth in manufacturing. It does not necessarily imply a more unequal distribution of income. In Taiwan, for example, inequality during this period was reduced according to all measures except the productivity gap. The explanation lies in part in the rapid increase in the share of off-farm income in the total income of farm families. See Kuo (1984). 240 PRODUCTIVITY AND STRUCTURAL CHANGE productivity gap. Israel and the OECD countries represent this phase of transformation (see figure 8-1). The acceleration of aggregate growth is only partly caused by the shift of weights implicit in differential growth rates by sector; output growth accelerates in the model in all sectors other than agriculture. The accelera- tion of labor productivity encompasses agriculture as well, especially at medium income levels, at which the agricultural labor force begins to decline in absolute terms. This similarity in productivity growth across ,sectors suggests that there are underlying factors determining the overall process of transformation. This qualifies the notion of one sector leading the process without regard to conditions elsewhere in the economy/ Limits on Unbalanced Growth During the transformation, growth proceeds at an uneven rate from sector to sector. But sectoral interdependence imposes certain constraints which if violated may retard growth. A sector is said to be a leading sector when its rate of growth exceeds the average rate for a period long enough to raise overall growth toward its rate and when it spreads its dynamism through substantial links to other sectors. A large sector, such as agricul- ture at very low income levels, is almost by definition excluded from this role. By its very size, however, agriculture will initially have a larger weight in determining overall performance than any emerging but still small sector. 8 Historically, agriculture has often been instrumental in igniting growth, 9 but eventually the leading role has shifted to other sectors, predominantly manufacturing. The potential expansion of the primary sector has been limited on the demand side by low income and price elasticities for food and by the substitution of fabricated for natural raw materials and on the supply side by the availability of natural resources and technology and by low produc- tivity growth when underemployed labor is available. When these con- straints have been overcome, it has usually been through the discovery of natural resources (or price booms) and heavy reliance on the export of primary goods with little or no processing. The literature on the "staple" theory of growth gives various examples of episodes of fast growth based on the exploitation of natural resources and their export. Such episodes 7. A similar country effect or period effect is also observed in the following section for total factor productivity growth: differences across countries and over time for all sectors are more significant than differences across sectors for a given country and period. 8. Differences in the availability of arable land and natural resources, as well as in climate, labor supply, and the ownership structure, make it difficult to generalize about the agri- cultural sector, particularly at low income levels. Although the overall trend described is fairly representative, exceptions can probably be found for each specific component. 9. This has been true of newly settled areas specializing in primary exports. Argentina, Australia, Brazil, and Canada are examples of countries that went through episodes of this type of growth. PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 241 often did not lead to continuous development even when they extended over a long period of time. A measure of their success in producing self-sustained growth is the strength of the linkages to the manufacturing sector and the impetus given to the transformation of the economy. 10 If agriculture's capacity to sustain high rates of growth is limited, too slow an expansion of this sector can nevertheless retard the growth of the economy Y The interdependence of agriculture and manufacturing has received much attention in the development literature. 12 Agriculture has several contributions to make during transformation: it must provide food for the urban population, raw materials for industry, markets for some of the industrial products, revenue for the state, and foreign exchange to cover the import requirements of industrialization while the capacity to export manufactured goods is being developed. 13 Finally, as the economy industrializes there is a net flow of resources from agriculture to the expanding sectors. An elastic supply of labor facilitates the rapid expansion of manufacturing and of supporting infra- structure and services at both the early stages and higher income levels. 14 Agriculture can also finance the initial investment in manufacturing and infrastructure. The problem of maintaining the right intersectoral balance given the scarcity of capital has been aptly summarized by Kuznets: "One of the crucial problems of modern economic growth is how to extract from the product of agriculture a surplus for the financing of capital formation necessary for industrial growth without at the same time blighting the growth of agriculture, under conditions where no easy quid pro quo for surplus is available in the country." 15 The importance of maintaining a proper balance among sectors is clear. Economies that have fostered the development of agriculture through government investment, extension services, and nondiscriminatory price policies have better growth performances than those that have stimulated industry and neglected agriculture. Israel, Malaysia, and Taiwan are ex- amples of the first group and Argentina, Chile, and Uruguay of the second. In the more successful economies, sectoral transformation was marked by a rapid growth of agriculture, surpassed by an even faster growth of manufacturing. In the less successful cases, agriculture often stagnated. In 10. On the staple approach see, for example, Baldwin (1956), Watkins (1963), and Caves (1965). In the more recent literature, resource discoveries and price booms are discussed in relation to the Dutch disease (Corden and Neary 1982; Roemer 1985). 11. Non traded goods are similar to agriculture in that they are not best suited for leading growth but can retard growth if their expansion is inadequate. 12. See, for example, Johnston (1970), Johnston and Kilby (1975), and Meier (1984, pp. 427-31). 13. International trade makes these requirements less rigid, particularly if alternative sources of foreign exchange are available, such as mineral exports or foreign capital. 14. For the role of an elastic labor supply in the postwar growth performance of Western Europe, see Kindleberger (1967) and Cornwall (1977). 15. Kuznets (1961a, p. 115), quoted in Johnston (1970). 242 PRODUCTIVITY AND STRUCTURAL CHANGE /:{~ecent comparative study of a large number of countries for the period / 1960-79, Hwa (1983) shows that the growth rate of agriculture is signifi- 1 . candy associated with industrial growth and with the rate of increase in \ factor productivity. 16 The transformation of the production structure from primary activities to manufacturing has taken place in virtually all the developed countries and is very much in evidence now in the transitional economies. As indicated in chapter 4, the timing of this transformation varies greatly and depends chiefly on the degree of participation in international trade-the main source of increased specialization. Factor Productivity Growth In part II, the growth of total output and of sectoral outputs was analyzed in terms of changes in the level and structure of the expenditure (demand) components of GDP. The rate of growth of total output and of labor productivity can also be accounted for by changes in three sets of interacting supply-side factors: • Changes in primary inputs, or capital accumulation and the ex- pansion of employmenrt 7 • Increases in the efficiency of factor use within sectors, or total factor productivity growth • Resource reallocation across sectors. This section presents a detailed supply-side growth accounting that decomposes output growth into factor contributions and productivity growth, first by sector and then for the whole economy. The aggregate v/ growth of productivity will be seen to depend in part on the production structure of the economy, even in cases where the sectoral rates of produc- tivity growth are exogenously given. Aggregate Results The data on the growth of output and inputs by sector presented in tables 8-4 and 8-5 can be used to derive the implied rates of total factor productivity growth if they are supplemented by minimal assumptions about a neoclassical production function. The production function approach complements the fixed coefficients formulation of chapter 3. At the cost of additional assumptions, it yields additional results as to the pattern of TFP growth by sector and the contribution of intersectoral resource reallocation to aggregate productivity and output growth. In each sector, let output (value added) be produced by a differentiable 16. This resembles the country effect mentioned above. 17. In accounting for the growth of gross output, intermediate inputs also have to be considered. For the relation between growth accounting for gross output and for value added, see Syrquin (1985a). PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 243 production function, with constant returns to scale in capital and labor, of the form used in chapter 2: (8-5) or in intensive units (with k = Kl L): (8-6) The rate of TFP growth, or technical change, is assumed to be Hicks neutraP 8 and proceeds at a constant rate within periods. For each sector and period, the rates of change in output, inputs, and TFP growth are related by the growth accounting equation: (8-7) where 'A is the rate of TFP growth and a and (1 - a) are the elasticities of output with respect to capital and labor. To calculate TFP growth given the growth of output and inputs for a period, all that is needed is an estimate of the elasticity of capital, a. The exact functional form of the production function need not be specified. A point estimate for a is compatible with almost any constant-returns production function. The elasticities are not\ necessarily assumed to equal the factor shares in value added. Instead, I\ recognize the possibility of imperfect mobility, segmented labor markets, , surplus labor, and other sources of friction and disequilibrium preventing ! the equalization of factor returns across sectors or of factor shares and/ elasticities within sectors. The values assumed for the elasticities in the computations below were derived from cross-country information and in most sectors vary over time according to available information on sectoral elasticities of substitution. These values appear in table 8-6. 19 Substituting in equation 8-7 the value for the elasticity, a;, gives the rate of TFP growth, 'A;. 20 The usual procedure in short-run models is to regard 18. Technical change is Hicks neutral if it affects the level of output but not the marginal rates of substitution between pairs of inputs. 19. Since land expansion is an important factor in the growth of output in agriculture at low income levels, I assume that any residual productivity in period 1 was attributable to land. In period 2 three-quarters of the residual is credited to land; in period 3, half. Beyond period 3, land expansion is not considered, in accordance with the common practice in growth accounting studies. See, for example, Denison and Chung (1976). 20. This is the common procedure in growth accounting. The relation of this approach to the presentation of factor-output ratios above can be clarified by restating the growth decomposition in intensive units: GY; = ct;Gk; +A; where Gy. is the growth of labor productivity. Substituting Gk = Gh + Gy (h being the capital-ou'tput ratio) yields G .=~Gh +-A_,_ Y, 1- ct; ; 1- ct; Given the sectoral h;'s, this is an equation in three unknowns: y;, A;, and ct;. Assuming a value (Note continues on the following page.) 244 PRODUCTIVITY AND STRUCTURAL CHANGE TFP growth as exogenous and to derive output and labor productivity growth. Here the procedure is reversed; labor productivity growth is predetermined by estimating from cross-country data the variation in labor output ratios (I = 1/y ), and TFP growth is then determined residu- .;' ally. This procedure does not imply anything about the causality or the exogeneity of technical change. Like other growth accounting exercises, it is useful in singling out potentially important effects and in providing orders of magnitude. 21 The derived pattern of sectoral growth of TFP shown in table 8-6 is the one consistent with all the other blocks of the model. The complete model is best seen as a consistent simultaneous system that complements the more structural but data-intensive approach in chapter 11. Compared with the general equilibrium model in that chapter, this one calls for less information on specific parameters and fewer behavioral and institutional assumptions, but it stops at the reduced-form stage and is less useful for policy simulations. Its main virtue is that it incorporates into one system the various elements of growth and structure, from Engel effects to dif- ferential productivity growth. It illustrates in a simple and direct way the general equilibrium nature of the interrelations: changes in any one com- ponent imply variations throughout the system. To apply this approach, equation 8-7 was calculated at the eight-sector level for all periods. 22 The results are tabulated in table 8-6. For manufac- turing, only the total is shown since the results for its subsectors were not very different. The aggregate sources of growth are computed directly from the aggregate rates of growth of output and inputs, with the output- weighted average elasticities used as weights. When the sectoral elasticities equal the distributive shares, their output-weighted averages correspond to the factor shares in total output. To distinguish the two concepts, I denote the directly computed aggregate rate of TFP growth by l\ and the average of the sectoral rates by ~ p)\;. The results in table 8-6 illustrate the general pattern of change in the sectoral sources of growth at various income levels. Although any indi- vidual figure may have a large margin of error, the overall pattern appears for a; yields a relation between the growth of labor productivity, Gy, and total factor productivity, 'A.;. To an estimated or assumed value of one there corresponds a unique value of the other. 21. The important advances in the last twenty-five years in productivity research and in related concepts such as human capital, embodiment, and learning effect were in no small measure due to the large "residuals" reported by Abramovitz (1956) and Solow (1957). 22. The elasticities in equation 8-7 for any period are assumed to be mean weights of initial and terminal values; growth rates are computed by log differences. The index for productivity growth is thus a Tornqvist-Theil quantity index, which has been used as a discrete approximation of the Divisia index. Diewert (1976) has shown that it is exact for a homogeneous translog function, and since this can provide a second-order approximation to an arbitrary function, it is also superlative (see Diewert for definitions and references). PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 245 to be consistent with the experience of many countries. I turn now to the main results, some of which have previously been summarized in figure 3-8. In the early periods of the transformation (represented in table 8-6 by the income levels $100-140 and $140-$280), the rate of productivity growth is quite modest and accounts for no more than 10 to 15 percent of total output growth. The initial acceleration of growth comes largely as a result of faster input growth, primarily capital accumulation. It is at this stage that most of the increase in the net investment share takes place, but since the elasticity of capital normally declines, the contribution of capital accumulation to growth stabilizes or declines. Since output growth con- tinues to accelerate for a considerable period, the growth rate of factor productivity must increase and must also account for an increasing pro- portion of growth. This is basically the argument that leads Kuznets to expect the rate of factor productivity growth in most countries to be higher in the twentieth / century than it was in the nineteenth (Kuznets 1971, p. 75). In most of the countries for which long-term records are available, productivity growth did accelerate over time. A long-term acceleration has been documented for Canada and Norway (Kuznets 1971), the United Kingdom (Matthews, Feinstein, and Odling-Smee 1982), the United States (Abramovitz and David 1973), and Japan (Ohkawa and Rosovsky 1973). In the early periods in these studies, the growth of total factor productivity accounted for about 30 to 40 percent of the growth of output, and factor inputs, primarily capital, accounted for more than half. The postwar finding of . very large contributions of TFP growth and correspondingly low contribu- \ tions of capital and labor is therefore a recent phenomenon in the indus- · trial countries. It does not characterize the earlier experiences of these countries or the early periods of the transformation in developing coun- tries. At higher income levels, the contribution of capital declines and that of factor productivity increases (discussed in figure 3-8). In chapter 2, estimates of the sources of growth for a large number of economies were summarized in figure 2-2, which plots lines of constant output growth as the sum of the combined (weighted) growth of factor inputs, Gp, and of total factor productivity, X.. These estimates cover fairly short time intervals during the postwar period. The same kind of figure can be used for comparing the long-run relation of productivity and output growth for the cross-country model and for countries for which the required information is available. Figure 8-2 adapts figure 2-2, retaining its three country clusters. The very high growth rates of the cluster C economies in figure 8-2 were achieved both by fast input expansion and by high productivity growth. Such growth rates are certainly exceptional when compared with the experiences of the developed countries and are even so when compared with the universal fast growth in the postwar period. Table 8-6. Sources of Growth: Inputs and Productivity Annual growth Contributions to Per capita rate (percent) output growth (percent) Elasticity income of Period (dollars) Sector Output TFP TFP Capital Labor Land capital 0 100-140 Agriculture 2.98 0 0 39 47 14 0.40 Mining 4.31 0.75 17 54 29 - 0.50 N Manufacturing 5.36 0.53 10 59 31 - 0.60 4:. 0\ Social overhead Services } 4.15 0.72 17 48 35 - 0.50 Total 3.81 0.44 11 48 36 5 0.47 140-280 Agriculture 3.90 0.11 3 48 41 8 0.40 Mining 5.51 0.90 16 56 28 - 0.50 Manufacturing 5.65 0.91 16 54 29 - 0.54 Social overhead 5.36 0.88 16 57 27 - 0.60 Services 4.97 0.69 14 44 42 - 0.45 Total 4.80 0.72 15 49 34 2 0.47 2 280-560 Agriculture 3.92 0.23 6 58 30 6 0.43 Mining 6.50 1.43 22 57 21 - 0.50 Manufacturing 6.79 1.61 24 51 25 - 0.51 Social overhead 6.19 1.41 23 54 23 - 0.58 Services 5.80 1.20 21 39 40 - 0.40 Total 5.67 1.40 25 47 27 1 0.46 3 560-1,120 Agriculture 3.53 0.86 24 68 8 0.46 Mining 7.26 1.81 25 59 16 - 0.50 Manufacturing 7.57 2.11 28 47 25 - 0.46 Social overhead 6.59 1.96 30 50 20 - 0.55 Services 6.28 1.66 26 34 40 0.35 Total 6.30 2.28 36 43 21 - 0.43 4 1,120-2,100 Agriculture 2.68 1.46 54 87 -41 - 0.50 Mining 6.74 2.53 38 55 7 - 0.50 Manufacturing 7.84 2.52 32 43 25 - 0.42 Social overhead 6.57 1.89 29 48 23 - 0.50 Services 6.40 2.19 34 30 36 - 0.30 Total 6.58 2.92 44 39 15 - 0.39 5 2,100-3,360 Agriculture 1.57 1.55 99 109 -108 - 0.52 Mining 6.07 1.78 29 58 13 0.48 N Manufacturing 6.18 2.67 43 40 17 0.40 ~ '-I Social overhead 6.37 2.68 42 45 13 - 0.47 Services 6.78 2.99 44 25 31 0.25 Total 6.21 3.11 50 35 15 - 0.36 6 3,360-5,040 Agriculture 2.06 1.49 72 96 -70 0.52 Mining 4.79 1.92 40 59 1 - 0.48 Manufacturing 4.95 2.79 56 35 9 - 0.35 Social overhead 5.60 2.43 43 42 15 - 0.44 Services 6.48 2.91 45 25 30 0.25 Total 5.60 2.80 50 32 18 - 0.33 - Negligible or zero. Figure 8-2. Growth of Inputs and Total Factor Productivity for Cross-Country Model and Countries that Have Long-Term Records Gv = 12 A.!Gv = 0.40 United Kingdom 1 2 3 4 5 6 7 8 9 10 11 12 Total factor input growth (percent) (GF) Note: Gv is growth of value added in percent. The information for the seven periods of the model comes from table 8·6. The country data refer to the following periods: Country Period Source Canada 1891-1926 Kuznets (1971) 1926-56 France 1913--49 Carre, Dubois, and Malinvaud (1975) 1949-63 Israel 0 ewish Palestine) 1922--47 Syrquin (1986) 1950-72 Japan 1887-1922 Ohkawa and Takamatsu (1983) 1917-38 1953-71 Norway 1879-99 Kuznets (1971) 1899-1956 United Kingdom 1856-1913 Matthews, Feinstein, and Odling-Smee (1982) 1924-73, excluding 1937-51 United States 1800-1905 Abramovitz and David (1973) 1905-67 The rays from the origin show the proportion of total output growth caused by TFP growth. The three country clusters are reproduced from table 2-2 without the actual observations. Group A includes industrial countries. Group B is representative of most developing and centrally planned economies. Group C consists of semi-industrial economies with very high growth rates. 248 PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 249 The data for the seven countries in figure 8-2 indicate a long-run acceleration of productivity growth and, except for France, an increase in productivity's contribution to output growth. The pattern derived from the cross-country model is plotted for periods 0-6. The initial acceleration of aggregate growth is the result of faster input expansion. At medium income levels, productivity growth accelerates, pulling with it the growth of aggregate output. The typical country resembles the countries in cluster Bin periods 2 and 3 ($280-$1,120), which cover the bulk of the indus- trializing stage II. In the later periods-identified as the developed stage III-output growth diminishes but productivity continues to increase its share of growth, attaining levels similar to those in cluster A. (See figure 3-7 for explanation of the three stages.) Sectoral Results In agriculture, factor productivity growth is negligible at low income levels. Once the modern sectors are able to absorb a substantial portion of the excess labor in agriculture, however, capital intensity and productivity rise rapidly. The percentage of the labor force employed in agriculture shows a continuous decline, and from an income level of about $1,000 it falls in absolute terms as well. This decline, observed in virtually all countries, reflects not only the low elasticity of the demand for food but also significant increases in productivity and mechanization in the agri- cultural sector. This increase in the growth of factor productivity in agriculture was observed in the three countries for which long-term records are avail- able-the United States, the United Kingdom, and Japan (table 8-7). In Israel, the agricultural sector started in 1950 with relatively high produc- tivity and a high priority in the allocation of investment. It also had a very high rate of TFP growth, an important element in the successful perfor- mance of the Israeli economy during the period. Unbalanced productivity growth is one of the reasons on the supply side behind the shift in comparative advantage and the transformation of the structure of production. In the model, TFP growth in manufacturing exceeds TFP growth in agriculture by more than 1 percentage point for most periods (see table 8-6). 23 The last column of table 8-7 indicates the gap between productivity growth in manufacturing and in agriculture for some countries. The gap is 23. Total factor productivity growth at the value added level is higher than at the gross output level by a factor equal to the value added ratio; see Syrquin (1985a). Since this ratio is usually higher in agriculture than in manufacturing, the productivity differential between the two sectors is correspondingly lower for gross output. In the model, the differential is reduced at the gross output level to about one-third of a percentage point. If the value added ratio in agriculture declines over time (see chapter 3), the overestimate of the productivity growth gap for value added also declines. A constant or falling gap at the value added level could then hide an increase in the gap at the gross output level, contrary to the argument in Williamson and Linden (1980, p. 173). 250 PRODUCTIVITY AND STRUCTURAL CHANGE Table 8-7. Total Factor Productivity Growth in Agriculture and Manufacturing for Countries that Have Long-Term Records (percent) Annual rate of TFP growth Productivity Country Agriculture Manufacturing growth gap and years (a} (m) (X.m ~ X.a} United States 1839~59 1869~1929 } 0.3 1.9 1.6 1929~66 1.8 1.9 0.1 1966~73 1.0 2.3 1.3 United Kingdom 1856~1913 0.6 0.7 0.1 1924-73 2.6 1.8 ~0.8 japan 1887~1904 1.0 1.1 0.1 1904-19 1.4 2.1 0.7 1919~38 0.7 2.9 2.2 1954-76 2.4 6.1 3.7 Israel 1950-60 6.4 0.5 ~5.9 1960~72 5.4 5.9 0.5 Sources: United States, Williamson and Lindert (1980); United Kingdom, Matthews, Feinstein, and Odling-Smee (1982); Japan, Ohkawa and Takamatsu (1983); Israel, Gaathon (1971) for 1950~60, and Metzer (1986) for 1969~72. large in the case of the United States and Japan for most of the periods reported. In Japan, this gap rises sharply over time. In the United King- dom, the gap is negative for most of the twentieth century; a combination of substantial productivity growth in agriculture and unimpressive pro- ductivity growth in industry is responsible. The small or negative gap in Israel has more to do with the performance of agriculture than with a lag in industry, particularly after 1960. In the industrializing stage in the model, productivity growth in manu- facturing exceeds that in services; in the developed stage, however, the situation is reversed. Because of the well-known problems of measuring output in services, I do not elaborate on these results but only note that a positive gap, such as the one at low and medium income levels, is an important element in explaining systematic departures of exchange rates from purchasing power parities in terms of a "differential productivity model" (see appendix to chapter 3 ). The negative gap at high income levels has much to do with the rise in the relative price of services, which Kravis, Heston, and Summers (1983), have argued wholly accounts for the rise in the output share of services. Figure 8-3 presents the elements in the variation of productivity growth PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 251 Figure 8-3. Components of TFP Growth for Cross-Country Model 6.5 6.0 Gv (growth of value added) 5.0 Growth of factor input GF = aGK + (1 - Ci)GL 4.0 -.::;- ... !::: u .... ... _e,. ... ... 3.0 ~ (TFP growth-aggregate) o;l .... ..c: ~ 0 .... 0 2.0 PNTA.NT (contribution of nontradables sector 1.5 to TFP growth) 1.0 . - · - · PMAM (contribution of 0.8 ,...---·- manufacturing sector 0.6 ..,..... . ..,..... • to TFP growth) 0.4 PAAA (contribution of 0.2 ------------ agricultural sector ,360 5 ,040 to TFP growth) 140 280 560 1,120 2,100 3 Per capita GNP (dollars) during transformation. In all sectors in table 8-6, productivity growth accelerates throughout the transition, reaching a peak during period 4. The acceleration of the aggregate rate of productivity growth, X, discussed above, reflects this phenomenon, but it also involves two additional ele- ments. The first is the shift in weights from sectors with slower productiv- ity growth (such as agriculture) to sectors with faster productivity growth (such as manufacturing). The second appears when the aggregate rate, X, exceeds the weighted average of the sectoral rates, I p;h;. This difference measures the contribution to growth of resource reallocation among 252 PRODUCTIVITY AND STRUCTURAL CHANGE sectors with different marginal productivities, which is called total real- location effect (TRE; see figure). This element is discussed extensively in the following section, where it is shown to contribute significantly to the acceleration of output and productivity growth during the transforma- tion. Resource Reallocation as a Source of Growth 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 unbalanced productivity growth, and even by government policies. Various types of resource reallocation appear to have a significant effect on aggregate productivity growth. One of them, the shift from production for the domestic market to production for export, played a prominent role in the analysis in part II. The effect of this type of reallocation on growth in semi-industrial countries is estimated in chapter 9. A different type of reallocation, from market to nonmarket activities, or from the business to the government sector, appears in discussions of deindustrialization in industrial countries. In that case, the reallocation is seen as a source of the slowdown in productivity growth. In this section, I concentrate on in- tersectoral factor reallocation when factor returns differ across sectors. 24 The literature includes several formulations for measuring the effect of resource shifts on productivity growth. A simple approximation of this effect is the gross allocation effect, A ( y ), which appears in equation 8-4 as a component of the growth of aggregate labor productivity. 25 lt measures the growth in aggregate labor productivity that would have taken place with the observed labor shifts, G"Yi' had the relative labor productivities remained constant. The data required to compute this approximation are readily available, which is why it is often used. The gross allocation effect is a partial measure since it ignores factors other than labor and computes the gains and losses from employment shifts in terms of average and not marginal products. 26 A positive A ( y) can arise in a dynamic context even when resources are optimally allocated before and after the shift. This can be illustrated with a 24. This discussion borrows liberally from Syrquin (1984). 25. Equivalent formulations are used by Kuznets (1957), OECD (1965), and Ohkawa and Rosovsky (1963), among others. 26. Denison (1967) tries to correct for these two shortcomings with a differential weight- ing of the gains and losses from reallocation. See Syrquin (1984) for a description and evaluation of Denison's approach. PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 253 simple example of a small country producing only two goods (at fixed international prices). By Rybczynski's theorem, an increase in the aggre- gate capital input under equilibrium (equal marginal products across sectors) will lead to a reallocation of labor toward the more capital- intensive good, which is also the one with a larger average product (in value terms). In each sector, marginal and average labor products remain constant, while total labor productivity rises by an amount equal to the gross allocation effect, A ( y ). Since, by assumption, no resources are misallocated either before or after the increase in capital, the gains in labor productivity cannot be the result of the reallocation itself. A complete accounting, such as that presented below, would attribute the increase in labor productivity to the accumulation of capital. Tables 8-8 and 8-9 compare some estimates of A (y) with the more complete estimates of the reallocation effect. An intuitive way of deriving a comprehensive measure of the impact of resource shifts on growth is to start from an expression of the gross allocation effect equivalent (in the continuous case) to that in equation 8-4: 1 . (8-8) A(y) = ~p;G-y; =-V~L;(Y;- y) 1 I where a dot over a variable signifies a time derivative. This expression makes clear the dependence of A ( y) on intersectoral differences in average product per worker. Substituting marginal products, fL, for average prod- ucts, y, and adding an equivalent term for the input of capital, or for any other input in the production function in equation 8-5, give the measure sought-namely, the total reallocation effect or TRE: (8-9) where fL.I (f K.) I equals the marginal product of L ( K) in sector i and fL ( f K) is the economywide average of these figures. TRE is precisely the amount by which the aggregate rate of TFP growth, X, exceeds the weighted average of the sectoral rates, ~p;X.;, when the reallocation of resources leads to a reduction in the extent of disequilib- rium (see Syrquin 1984). 27 This will be the case if, on the average, sectors with above-average marginal products of labor or capital increase their share in total employment or capital. It was seen above that a positive A ( y) does not necessarily indicate the existence of disequilibria. But the differences in equation 8-9, and with them the total reallocation effect, 27. Massell (1961), who apparently was the first to publish a similar derivation, calls the weighted average of TFP growth (Zp,A.;) "intraindustry technical change" and the shift terms "interindustry technical change." Table 8-8. Contribution of Resource Allocation to Productivity Growth for the Cross-Country Model Reallocation measures Annual growth rate as percentage of (percent) Share of labor productivity TRE in relation TRE due Per Aggre- Average of to growth of to net Gross capita gate = sectoral allocation allocation Net + TRE income TFP rates Output TFP effect effect allocation N Period (dollars) (!\) (2 p,A.,) (TRE!Gv) (TREI!\) (percent) A(y) effect "" '-l::.. 0 100-140 0.44 0.40 0.04 1 9 100+ 15.0 5.7 1 140-280 0.72 0.57 0.15 3 20 69 17.0 5.5 2 280-560 1.40 1.11 0.29 5 21 85 16.5 9.0 3 560-1,120 2.28 1.72 0.56 9 25 89 20.0 12.5 4 1,120-2,100 2.92 2.17 0.75 11 26 90 20.0 14.0 5 2,100-3,360 3.11 2.71 0.40 6 13 83 9.0 7.5 6 3,360-5,040 2.80 2.72 0.08 2 3 75 0.0 1.5 PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 255 Table 8-9. Contribution of Reallocation to Labor Productivity Growth for Selected Sample and Other Economies Percentage of labor productivity growth Net allocation Aggre- Gross Net effect in gate allocation allocation relation to capital effect effect TFP growth etas- Economy Years (A[y]) (NA) (NAJ"i..) ticity (a) Sample Mexico 1950-75 30 13 22 0.48 Turkey 1963-73 61 Yugoslavia 1966-72 30 Japan (prewar) 1915-35 24 Japan (postwar) 1955-70 16 10 16 0.43 Korea 1963-73 16 5 6 0.40 Taiwan 1961-71 15 7 8 0.43 Israel 1958-72 5 6 11 0.40 Developed Italy 1963-73 14 9 14 0.40 France 1963-73 14 6 14 0.42 Belgium 1963-73 3 3 4 0.42 Canada 1963-73 8 7.5 7 0.45 Netherlands 1963-73 -4 -7 0.42 United Kingdom 1963-73 -1 1.5 0.40 - Not available. Note: For the sample economies, the level of aggregation is three sectors for prewar Japan, six sectors for Korea, and eight sectors for the others. Social overhead is omitted in Korea because of problems with the data on capital. Capital data were not available for Turkey, Yugoslavia, and prewar Japan. For the developed economies, the level of aggregation is four or five sectors. Sources: For sample economies: World Bank data; described in Kubo (1983). For developed economies, Stein and Lee (1977). The values for (i are based on Christensen, Cummings, and Jorgenson (1980). vanish when marginal productivities of labor and capital are equal throughout the economy. A more convenient formulation for estimation than equation 8-9 can be derived directly by focusing on the growth of labor productivity. If the sectoral growth accounting equation (the first equation in footnote 20 above) is substituted into the expression for the growth of aggregate labor productivity 8-4 and the results are compared with those of a growth accounting exercise performed directly at the aggregate level, another formula for TRE is given: (8-10) TRE = ~- ~p;A; = l_~k('/;(fK;- fK) + ~~i('XJ- ii~ki) t y t t y = A ( k) + net allocation. 256 PRODUCTIVITY AND STRUCTURAL CHANGE Under full equilibrium (equal marginal products across sectors) both terms equal zero and the difference vanishes. The first term on the right-hand side, A ( k ), measures the difference between the contribution of sectoral capital accumulation to labor pro- ductivity growth and the contribution that would have resulted had all sectoral marginal productivities of capital equaled fK· If k; (not just K;) increases faster in sectors with a higher than average marginal product, the term will be positive. This reflects the contribution of narrowing the existing disequilibrium through capital deepening in high-productivity sectors. The second term, labeled the net allocation effect, has two components. The first component is just the gross allocation effect, A ( y ). 28 The second component corrects for the deficiencies of the gross effect. It recognizes that for average products not to change, some net investment will usually be required. If after a reallocation of labor each worker is equipped with the same amount of capital as those who worked in the sector before, the aggregate capital-labor ratio will show a net change given by l ~;k;lk, even if the capital intensity in each sector remains unchanged ( k; = 0). This change in the aggregate capital-labor ratio, when multiplied by the average share of capital, a, gives the expected change in output per worker. In full equilibrium, this expected effect exactly cancels out the actual gross allocation effect, and the net allocation effect vanishes. 29 Out of equilib- rium, a shift improving the allocation of labor will contribute to aggregate labor productivity over and above the expected gain through the capital requirements it generates. The Gains from Reallocation: Empirical Results Estimates of the contribution of resource reallocation to growth are presented in table 8-8 for the dynamic model and in table 8-9 for the sample and for some advanced countries. The total reallocation effect is positive in every period. In its contribu- tion to growth, it reaches a maximum during period 4 (representing high-income semi-industrial economies), when it amounts to 11 percent of output growth and more than 15 percent of labor productivity growth. The results suggest that TRE is a significant component of aggregate TFP growth, particularly in the industrializing stage. 30 Its pattern across 28. 'i.;~;y/y = 'i.;~;/-y;p; =A (y). 29. The net allocation term is equal to l!y 'i.;~;(y;- fKk;). In equilibrium fK = fK, and therefore the term in parentheses reduces to fL. Since by assumption fL = fL an'd by definition 'i.;~, = 0, the result in the text follows. ' ' 30. The figures in the fifth column of table 8-8, which show the relation of TRE to ~, are probably too conservative. Syrquin (1984) and Kelley and Williamson (1984) review a variety of available estimates, many of which exceed the highest figure in that column. For their simulations, Kelley and Williamson chose a figure of 0.31 as representative of the experience of developing countries. PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 257 periods resembles the initial acceleration and subsequent slowdown in the growth of output and labor productivity. In the final period-which corresponds roughly to Western Europe since the late 1960s-the effect of resource reallocation almost disappears. Part of the accompanying pro- ductivity slowdown reflects the exhaustion of the shift out of agriculture as a potential source of growth. If productivity gains continue to decline at income 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 negative-allocation effect may become important as labor continues to shift into services, a low-productivity sector. The total reallocation effect can be decomposed into two terms, as in equation 8-10. The second term in the formula, the net allocation effect, accounts for 70 to 90 percent of TRE after the first period. This net effect can be computed without any information on sectoral production func- tions. Besides data on outputs and inputs, its computation requires only one parameter-an estimate of the aggregate elasticity of capital, &. It is therefore possible to calculate the net allocation effect for many econo- mies, as is done in table 8-9. When evaluating these results, however, it must be recalled that they refer to only a part of the total reallocation effect, though it may well be the predominant part, as the numbers in table 8-8 suggest. The last two columns of table 8-8 compare the gross and the net allocation effects in relation to the growth of output per worker. The gross effect generally exceeds the net effect, but in the last period the gross effect vanishes while the net effect remains positive. 31 The comparison of the two measures shows that the gross allocation effect is only a crude approxima- tion of the gains from reducing disequilibrium through resource shifts. I turn now to some country comparisons of resource reallocation and productivity growth. Table 8-9 presents the contribution to labor growth of intersectoral shifts in the simple form of the gross allocation effect, A ( y ), as well as of the net allocation effect of equation 8-10. The net effect is also shown as a share of the aggregate rate of growth of TFP, i.., implied by the aggregate figures for labor productivity and capital intensity together with the assumed value for &. The gross allocation effect accounts for a significant part of the growth in labor productivity. The relative contribution is much lower in the industrial countries, and it shows a sizable decline in Japan over time. When capital accumulation is taken into account in the net effect, the contribution to growth is reduced to about 10 percent in the sample 32 and 31. By the last period, the large shift from industry to services actually reduces capital requirements, thus allowing for a positive effect of labor reallocation by saving capital, even when average products of labor do not differ across sectors. 32. The figure for Korea may indicate that the initial disequilibrium was low. The estimate, however, is much affected by the data on capital in the service sector. When only the 258 PRODUCTIVITY AND STRUCTURAL CHANGE to less than 10 percent in the industrial countries. This latter group exhibits an interesting contrast between France and Italy on the one hand and the Netherlands and the United Kingdom on the other. At the begin- ning of the period, the first pair's employment in agriculture is still high and productivity of labor in agriculture low. So these countries can yet benefit from the shift out of agriculture. But the second pair's employment in agriculture is quite low and productivity high; the shift is into services, which yields no gain in growth and may even retard it. Compared with the aggregate rate of TFP growth, il., the contribution of resource reallocation is more significant. In the rough calculations for the sample, the figures are generally close to or above the highest figure in the cross-country model (table 8-8). In the group of industrial countries, resource reallocation accounts for a smaller fraction of il., reaching a high of 14 percent in France and Italy. Assessing the Results The effect of resource reallocation can be identified only at the aggregate level. When total output growth is aggregated from the sectoral results, there is no place for resource reallocation as an independent source of growth. Any such effect has already been accounted for in the contribution of inputs. It may nonetheless be of interest to present the effect separately, but in interpreting the results the resource shifts must be recognized as not being exogenous. Potential gains from reallocation may be present, as in the case of embodied technological change. But to realize the gains, some triggering mechanisms, with costs of their own-migration, investment, and so on-may be needed. The estimated contributions of structural change to growth probably understate the effect of resource shifts. The broad definition of sectors, even in fairly disaggregated studies, hides all factor reallocations within those sectors. This is important for industrial economies and for rapidly growing ones. About Taiwan, for example, Kuznets (1979, p. 73) argues that the high rate of growth of product per worker called for "a much greater rate of shift [than the] one now suggested in the three-sector classification and that the shifts from old to new subbranches within these sectors are particularly neglected." Another potential source of underestimation of the importance of re- source shifts lies in the static and partial nature of the measures. According to Cornwall (1977, p. 124), who regards the manufacturing sector as the engine of growth, "these estimates ... assume that the level and rate of primary and manufacturing sectors are considered, the net allocation effect is twice as high as the figure in table 8-9. In a recent application of Denison's method to Korea, Kim and Park (1984) give an estimate of the contribution of "contraction of agricultural inputs" equal to 12 percent of the growth of national income per person employed during 1963-72. PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 259 growth of productivity in the sector expanding inputs and output are independent of the expansion process itself. This rules out the possibility of various economies of scale in manufacturing." A more general and dynamic approach would explicitly recognize that resource shifts may facilitate or directly trigger higher productivity growth. A possible source of overestimation of the importance of resource reallocation is the assumption of input homogeneity. Differences in re- turns to labor and capital across sectors may reflect quality differences as well as disequilibrium. A reallocation of resources from sectors with low returns to those with high returns would then reflect a reduction in the misallocation of resources and an improvement in the average quality of inputs. The total reallocation effect includes both the reduction of dis- equilibrium and the upgrading of the quality of inputs. The contribution of resource reallocation to growth could also be overstated if the higher productivity observed in manufacturing when it is compared with agricul- ture reflects, in part, a distorted domestic price structure. A complete evaluation of the role of compositional changes in develop- ment would have to include a systematic treatment of the relevant mechanisms. Rapid shifts may help accelerate growth, but at the same time they may not be feasible without high rates of growth and investment. Similarly, just as the effect of resource shifts on growth may be underesti- mated, so the contribution of investment to growth may also be underesti- mated if the increase in flexibility caused by the higher investment rates is not considered. Productivity and Growth: Demand-Supply Interactions In the extreme supply approach, exogenous productivity growth drives the economy. Combined with the growth of inputs, this leads to output growth. It has already been shown that for two reasons the aggregate rate of TFP growth, ~' cannot be regarded as exogenous even when the sectoral rates, A;, are. First, the output weights needed to aggregate the sectoral rates are clearly not independent of demand. Second, the aggregate rate of productivity growth includes as a component the gains from resource shifts in a disequilibrium situation. In addition to these two reasons, if the sectoral rates of TFP growth are themselves endogenously determined, the extreme supply-determined approach to growth analysis becomes unten- able. Some additional links between the demand side and productivity growth at the aggregate as well as sectoral levels can be derived by a priori reasoning and well-established empirical regularities. In most cases, however, serious identification problems make it difficult to determine causality. The empirical associations are indicative only of demand-supply interactions, which probably reinforce each other regardless of the initiat- ing mechanism. 260 PRODUCTIVITY AND STRUCTURAL CHANGE Economic Environment In disaggregated cross-country comparisons or in a single country over time, one finds a high degree of uniformity of productivity change across sectors for a given country or period. Although the diversity among industries is far from negligible, it is difficult to point to particular indus- tries as being consistently the best or worst performers irrespective of time and place. In his study of long-term growth in the United Kingdom, Matthews (1974) found no support for the view that certain industries have persistently high rates of technical progress. Similarly, in his study of the United States, Kendrick (1961, p. 178) concluded that" ... there are certain forces that promote productive efficiency throughout the econ- omy." In the detailed comparison of four countries reported in chapter 10, productivity growth is found to be uniformly higher in some countries than others. This country effect emerges much more clearly than does any sector effect. This suggests that, in addition to any intrinsic productivity potential of an industry, the overall economic environment is important in explaining the general level of productivity growth. This country effect (or period effect within a country) is strongly associ- ated with general macroeconomic policies. It has been observed that productivity growth in developed countries has a large cyclical component related to capacity utilization. In developing countries, productivity growth is affected by stop-go episodes caused by balance of payments problems. Such cyclical phenomena will not only affect the measured rate of TFP growth, but will probably also influence the pace of advance of technological innovation (Nelson 1981). 33 A large part of the better per- formance of outward-oriented economies can probably be traced to their successfully having avoided stop-go cycles by preventing foreign exchange shortages. Denison (1967) and Maddison (1980) also stress the impor- tance of general macroeconomic policies for technological progress. Endogenous Technical Change The rate of technological progress in a sector or activity is in part a response to changes in economic variables within the sector. A high rate of investment is essential for realizing the potential gains from capital- embodied technological change. The rates of investment and output growth are key elements in theories of endogenous technical change such as Arrow's "learning by doing" (1962)-in which labor productivity increases with the level of cumulative gross investment or output-and Schmookler's argument (1966) that inventive activity gravitates toward industries experiencing rapid growth of demand. Learning effects play an important role in rationalizations of the high 33. Kornai (1980) has also emphasized the beneficial aspects of "taut" utilization of capacity. PRODUCTIVITY GROWTH AND FACTOR REALLOCATION 261 correlation commonly found between the rates of growth of productivity and output, either across industries (primarily manufacturing) or for a given industry over time. 34 The association is often presented as capturing the effect on measured productivity of endogenous technical change and economies of scale through the expansion of output. A faster rate of output growth facilitates the adoption of new technology, leads to a reduction in the average age of the capital stock, enhances efficiency by learning, and increases productivity through economies of scale, both static and dynamic. 35 In this view, the impact of the growth of output on the growth of productivity incorporates most of the demand-supply links discussed above. In numerous empirical studies, this approach has proved successful in accounting for interindustry and intercountry differences in the growth of labor productivity. The causal chain could run from faster productivity growth to faster output growth through changes in relative prices with elastic demands. 36 Given the complexity of the relation, how- ever, it is not likely that the direction of causality at the aggregate level- the level at which the relation is usually examined-can be determined. Some authors recognize the reciprocal nature of the relation but conclude after comparing the strength of various correlations (productivity and prices, prices and quantities, and so on) that the dominant influence goes from output to productivity (Kendrick 1961; Kennedy 1971). 37 A similar debate over causality is found in the literature on trade. Is a good export performance the result of productivity growth, or does the growth of exports contribute to a rise in productivity? Here again the relation is probably reciprocal. For the typical semi-industrial country that has gone through a period of inward-oriented industrialization, an inter- nal change that raises productivity might be needed before exports can expand significantly. Sustained growth of exports, on the other hand, may contribute to further productivity growth in a variety of ways: through economies of scale, through relaxation of the foreign exchange constraint, or through a host of positive side effects that are sometimes said to accompany an outward orientation (see, for example, chapter 9 and Krueger 1983). 38 34. In the literature, this relation often appears as Verdoorn's law (1949), which is examined in chapter 2. It is also one of Kaldor's "laws" (1967). See Kennedy (1971) for a critical evaluation, as well as the recent symposium in the Journal of Post-Keynesian Economics (1983). 35. Ohkawa and Rosovsky (1973, p. 77) discuss the organizational changes that must take place for "a faster expansion of the market to lead to more rapid productivity." They also make clear that it is increasing the scale relative to the size of the market that is relevant, rather than increasing returns to scale on the production side. 36. If technical progress is directed to the more price-elastic goods, then the link through prices from productivity to output growth is in part an endogenous response to demand characteristics (see Kennedy 1971, chap. 6). 37. See Caves (1968) for the opposite view for the United Kingdom. 38. Kelley and Williamson (1974) review the historical literature and the Japanese experi- ence. Also relevant for Japan is Kanamori's study (1968): by examining fifty-five manufactur- (Note continues on the following page.) 262 PRODUCTIVITY AND STRUCTURAL CHANGE One impqrtant factor usually ignored in the simple correlations between productivity growth and either output or export expansion is the type of output or trade involved. For example, a rise in domestic demand stimu- lated by a budget deficit would have a different effect than would a rise in world demand for machinery (Cornwall1977). Similarly, a greater expo- sure to trade might lead to faster productivity growth in the case of producer goods but not in the case of consumer goods or of manufacturing overall (Wragg and Robertson 1977; quoted in Caves 1980). This point is explicitly considered in chapter 10, which will distinguish between export expansion and import substitution as sources of output growth. ing industries for 1955-64, he shows that, at the sectoral level, higher rates of export increase were associated with high rates of investment growth and of domestic demand expansion. 9 Growth in Semi-Industrial Countries: A Statistical Analysis GERSHON FEDER THIS CHAPTER seeks to identify the factors that account for the growth experience of the group of semi-industrial countries described in chapter 4. The analysis is based on cross-section regression analysis. The simplest approach-the one presented in chapter 2-assumes an underlying aggre- gate production function, and it attributes changes in output to changes in the capital stock and in the labor force. But most researchers recognize that other elements also contribute to the variation in growth experience among countries. Thus Wheeler (1980) emphasizes human capital and the quality of labor; Hagen and Havrylyshyn (1969) introduce a set of vari- ables accounting for school enrollment, social mobility, communications, and so on; and Balassa (1978), Michalopoulos and Jay (1973), and Tyler (1981) stress the importance of exports. This chapter extends the analysis begun in chapter 8. It focuses on the role of disequilibrium-that is, of differences in productivity among sec- tors. Growth comes not only from increasing aggregate inputs but also from reallocating resources to more productive sectors. This is the central hypothesis of a study by Robinson (1971), who constructs a model that explicitly incorporates the shift of resources from the traditional (less efficient) to the modern (more efficient) sectors. The possibility of two types of disequilibrium is considered: that be- tween industrial and nonindustrial sectors and that between export and nonexport sectors. First, a general two-sector disequilibrium model is described, and the results based on the regression specifications suggested by the model are presented for each type of disequilibrium. Second, a four-sector framework is developed to allow an estimation that incorpo- rates the two types of disequilibrium simultaneously. Third, the possibility that the results reflect some other underlying explanation is considered. The Disequilibrium Model The first task is to construct an analytical framework that will both facilitate the design of the empirical work on sources of economic growth A partial version of this chapter appeared in the journal of Development Economics 12 (1983): 59-73. It is reprinted here with permission of the North-Holland Publishing Com- pany. 263 264 PRODUCTIVITY AND STRUCTURAL CHANGE and suggest possible interpretations of the estimated parameters. In de- veloping this framework, one has to bear in mind the likely limitations in the data available for a cross section of developing countries. Variables have to be specified at a high level of aggregation. Nevertheless, the formulation offered here should make possible an assessment of the effect of resource reallocation on growth so that the disequilibrium hypothesis can be tested. Suppose the economy consists of two sectors with outputs Y1 and Y2 • Each sector's output is determined by a production function that depends on sectoral inputs: (9-1) Y; = F;(K;, L;) where i = 1 or 2, K; and L; denote sectoral capital and labor inputs, respectively. The change in output over time is thus: (9-2) where Fk and Fi denote marginal factor productivities and I; denotes sectoral investment. Were the economy in equilibrium with optimal re- source allocation, marginal factor productivities would be equal across sectors. 1 But this formulation does not impose equilibrium. Rather, mar- ginal factor productivities are assumed to differ by a given proportion such that (9-3) and p2 (9-4) __b,=1+j.L Fl where 8 and 1-1 can take any sign. Let Y, which is the sum of Y1 and ¥ 21 denote GDP. Then equation 9-2 implies that 2 2 2 (9-5) L\Y= 2 .:lY;= 2 Fkl;+ 2 F{L\L;. i=l i=l i=l Substituting equations 9-3 and 9-4 in equation 9-5 gives (9-6) L\Y= Fklt + (1 + 8)Fki2 + FlaLt + (1 + ~-L)FlL\L2 = fk(l 1 + 12) + F2 (L\L 1 + L\L2) + 8Fkl2 + 1-1FlllL 2. Next, if!= 11 + 12 and L\L = L\L 1 + L\L 21 equation 9-6 can be rewritten: 1. Since constant prices are assumed, no distinction is made between quantities and values. GROWTH IN SEMI-INDUSTRIAL COUNTRIES 265 (9-7) b.Y = Fk I+ F{ b.L +-B- + Fk [z +-1J--Fl b.L2 1+B 1+1J- = Fk I+ F{ b.L + _B_ (Fk I2 + F[ b.L2) 1+B + (-IJ--- _B_)fi b.L2. 1+~J- 1+B Following arguments similar to those presented by Bruno (1968), I suppose that a linear relation exists between the marginal productivity of labor in a given sector and the average output per worker in the economy: . y (9-8) FL = 13;·-. L If equation 9-7 is divided by Y and the results in equations 9-2 and 9-8 are used (and Gz stands for the growth rate of a variable z), the result is: t I G B G Y2 fJ. B Gy=FK·-+13t' L+--·( y ·-)+(-----)·132 2 Y Y 1+B 1+~J- 1+B (9-9) L2 ·(GL2 · - ) . L Equation 9-9 reduces to the familiar neoclassical growth equation when B = fJ. = 0, that is, when marginal factor productivities are equal across sectors. In the more general case, however, it offers the possibility of estimating the contribution to growth of resource shifts from sectors of low productivity to sectors of high productivity. The data requirements are modest: the growth ratios of aggregate investment and labor are commonly used in the neoclassical formulation, and the growth rate of sectoral output is usually available. The sectoral growth rate of labor (the last term on the right-hand side of equation 9-9) may be available for at least some sectors for a large group of countries. Note that, if a case can be made for a particular sectoral decomposition of the economy, and if the ratios of marginal productivities are equal across sectors (that is, if B = !J-), then only sectoral growth rates are required to assess the effect of factor shifts. Even though equation 9-9 is fairly straightforward and is valid subject to the assumptions made earlier in this section, its estimation by regression analysis raises a host of econometric problems. Some of these problems are common in cross-country studies of sources of growth (see the detailed discussions in Hagen and Havrylyshyn 1969 and in Chenery, Elkington, and Sims 1970); they include errors in variables, simultaneity, variation of 266 PRODUCTIVITY AND STRUCTURAL CHANGE parameters across countries, and so on. 2 The use of data averaged over time ameliorates but does not eliminate some of these problems. More- over, results of regressions using the variables indicated by the present analysis could in fact reflect different underlying frameworks, with dif- ferent interpretations for the estimated parameters. Although the present formulation introduces disequilibrium, which in general provides a more realistic description of an economy than would otherwise be possible, the analysis assumes that all the economies studied are in the same type of disequilibrium. Yet more than one type of disequilibrium may be present, and some economies may be affected more by one type than another. These qualifications and limitations should be borne in mind when the results in the following sections are considered. Disequilibrium between Industrial and Nonindustrial Sectors The dual-economy model of growth suggests a relevant two-sector decomposition of the economy into traditional and modern sectors. The common argument (supported in several empirical studies) is that mar- ginal labor productivity is higher in the modern sector; therefore, expand- ing the labor force in the modern sector through shifts out of the tradi- tional sector makes a contribution to growth. In many dual-economy models, the traditional sector is assumed to produce with labor but without capital, so the issue of the differential marginal productivity of capital is not discussed. But in fact capital is used in all sectors of the economy, and persistent differences in the marginal productivity of capital across sectors are possible. Empirical evidence is not as abundant as in the case of labor. DeMelo (1977) provides figures for Colombia which suggest that its industrial sector enjoys rates of return on capital higher than those observed for agriculture and services. 3 Robin- son (1971) has constructed a model of the sources of growth that allows for higher marginal productivities of both capital and labor in the indus- trial sector than in the nonindustrial sector. His empirical results suggest substantial differences between these two sectors, although the statistical significance for capital parameters is on the borderline. (See also chap- ter 8.) This chapter uses two definitions of the industrial sector. The narrower definition includes the manufacturing sector only; the broader one com- bines manufacturing and construction. 4 The two sets of results obtained do not seem to differ substantially. 2. An application of equation 9-9 that uses a cross-country sample implictly assumes that the marginal productivities of capital are identical throughout the sample; thus they can be treated as parameters. In addition, the parameters f3 1 , f3 2 , o, and fLare assumed to be identical for all countries. 3. DeMelo (1977, p. 400) comments, however, that reliance on shaky estimates of capital stocks may seriously bias the calculated rates of return. 4. These two sectors comprise "industry" in the analysis of patterns of development in Chenery and Syrquin (1975). GROWTH IN SEMI-INDUSTRIAL COUNTRIES 267 The equation estimated in this section is therefore (9-10) Gy = ao + ar!. + azGL + a3 Y (eM M) Y + a GL 4 m Lm L where M denotes industrial output and Lm denotes industrial labor force. Interpretation of the parameters a; follows from equation 9-9. As is common in cross-section studies, averages over a long time are used for both GDP growth and explanatory variables. The reasoning is that nonsystematic changes are thereby averaged out and the problem of lags between investment and production is reduced. But the post-1973 era has been subjected to effects not properly covered by the simple disequilibrium model developed above. This, and data problems for the years before 1964, dictated that the analysis be confined largely to the period 1964-73, although some results for an earlier period will also be discussed. The analytical framework is designed for cross-country data. Yet it may be argued that grouping all developing countries in one sample would excessively stretch the assumption about the constancy of parameters across the sample. The group of semi-industrial economies identified in chapter 4 seems to offer a relatively homogeneous sample. Some results for another group of developing countries substantiate the decision to focus on the semi-industrial economies. Table 9-1 reports regression results comparing the common neoclassical formulation (which assumes equal marginal factor productivities across sectors) with the formulation of equation 9-10. Several observations fol- low from these results: • The two variants of the disequilibrium model-referred to in the table as models I and II-produce very similar parameter estimates. • The disequilibrium formulation explains the variability of average growth rates much better than the simple neoclassical model does. (The adjusted R 2 is about 50 percent higher.) This is consistent with the indication of a significant disparity between marginal factor pro- ductivities in favor of the industrial sector. (The coefficient of GM · [M/Y] is significantly larger than zero at the 99 percent confidence level.) • The lack of statistical significance for the parameter associated with industrial labor growth, GL m · (Lm/L ), seems to suggest that the ratio of marginal labor productivities in the two sectors is not much dif- ferent from the ratio of marginal capital productivities (that is, J.L = & in the notation of equation 9-9). 5 • The coefficient associated with the investment variable, 1/Y, is inter- preted in the simple neoclassical formulation as the average of the marginal productivities of capital for all sectors of the economy. In the 5. However, this indication does not amount to a rigorous test of the hypothesis fJ. = 5. 268 PRODUCTIVITY AND STRUCTURAL CHANGE Table 9-1. Regression Results for Semi-Industrial Countries, 1964-73; Neoclassical and Industry/Nonindustry Disequilibrium Models Disequilibrium Disequilibrium model I model II (equation 9-1 0; (equation 9-10; industry defined industry defined Variable or Neoclassical as manufacturing as manufacturing result model only) and construction) IIY 0.247 0.139 0.112 (4.059) (2.558) (2.125) GL 0.779 0.429 0.383 (3.457) (2.023) (1.922) GM. (M/Y) 0.801 0.871 (3.406) (4.105) GLm 0 (Lm/L) 0.791 0.729 (1.116) (1.109) Constant -0.002 0.007 0.010 (0.128) (0.652) (0.935) Adjusted R 2 0.460 0.665 0.707 Standard error of regression (percent) 1.616 1.272 1.190 Number of observations 30 30 30 Note: Numbers in parentheses are t values. Source: World Bank data. disequilibrium formulation, this coefficient is interpreted as the mar- ginal productivity of capital in the less productive of the two sectors of the economy. This apparently affords a partial explanation for the sizable decline in the magnitude of the investment coefficient when the disequilibrium specification is used. The decomposition of the sources of growth is reported in table 9-2. The contribution of industrial expansion is substantial-about 2 percentage points, or almost a third of total growth. But this needs to be interpreted carefully. The 2 percent contribution measures the gain from the greater productivity of factors of production in the industrial sector. It is equal to the difference between the actual growth rate and the hypothetical growth rate if marginal productivities in all sectors were the same as those esti- mated for the nonindustrial sector. A better assessment of the effect of the reallocation of resources can be derived from this hypothetical calculation: holding sectoral labor inputs and the capital stock constant, suppose that capital amounting to 1 percent of GDP is transferred from the nonindustrial sector to the indus- GROWTH IN SEMI-INDUSTRIAL COUNTRIES 269 Table 9-2. Sources of Growth in Semi-Industrial Countries, 1964-73; Neoclassical and Industry/Nonindustry Disequilibrium Models Contribution to growth Disequilibrium model (equation 9-10; Variable or Sample Neoclassical industry defined result mean model as manufacturing) II¥ 20.09 4.971 2.797 GL 2.07 1.616 0.888 GM. (MIY) 1.84 475 1. } 2 005 GL '(Lm/L) 0.67 0.530 . co;;.'stant -0.176 0.720 GOP growth' 6.410 6.410 Note: All numbers are multiplied by 100. a. Numbers may not add to totals because of rounding. Source: World Bank data. trial sector. This would yield an increase of about 0.5 percent in GDP, which is a substantial gain from reallocation alone. 6 In view of the magnitude of marginal factor differentials, the question arises why market forces have not brought about a much greater shift of resources, which would shrink the gaps. The answer may lie partly in constraints on factor mobility caused by segmented markets, government intervention in investment allocation, continuing technological change, and so on. But it is also conceivable that some part of the marginal productivity of factors in manufacturing is not taken into account by the individual economic agents because of intrasectoral externalities. Specifi- cally, the tendency of industrial activities to be concentrated in urban centers gives rise to positive agglomeration effects-both vertical and horizontal-that are not fully considered by individual firms. Nonmanu- facturing activities seem to enjoy many fewer benefits of this type. Other beneficial effects may be generated through training and on-the-job learn- ing. If skilled workers and managers are highly mobile within a sector, there will be additional externalities not fully reflected in a firm's calcula- tions. 6. The real extent of these gains, however, may be biased upward because of distorted prices. If the industrial sector is protected from foreign competition, then domestic prices in that sector are higher than world prices. The real difference between marginal factor productivities (if measured in world prices) would be lower than the difference implied by the present estimates (where GOP growth was calculated from a time series in constant domestic prices). 270 PRODUCTIVITY AND STRUCTURAL CHANGE Table 9-3. Regression Results for Less Developed Countries, 1964- 73; Neoclassical and Industry!Nonindustry Disequilibrium Models Disequilibrium Disequilibrium model I model II (equation 9-10; (equation 9-1 0; industry defined industry defined Variable or Neoclassical as manufacturing as manufacturing result model only) and construction) IIY 0.055 -0.016 -0.012 (0.841) (0.245) (0.186) GL 0.541 0.379 0.409 (1.354) (0.964) (1.065) GM. (M/Y) 1.603 1.382 (2.032) (2.384) GLm • (Lm/L) 2.384 1.742 (1.406) (1.023) Constant 0.024 0.019 0.018 (1.910) (1.580) (1.585) Adjusted R 2 0.032 0.166 0.205 Standard error of regression (percent) 1.835 1.703 1.663 Number of observations 33 33 33 Note: Numbers in parentheses are t values. Source: World Bank data. Experimentation with another group of less developed countries that are not semi-industrialized tended to support the hypothesis that the semi-industrial countries are more homogeneous and are therefore a more appropriate group for this type of analysis. As is apparent from table 9-3, the neoclassical formulation and the two disequilibrium models explain very little of the variability in GDP growth rates among these less developed countries. Although the coefficient associated with the growth of indus- trial output differs significantly from zero, the point estimate is clearly not compatible with the underlying model, for it cannot theoretically exceed the value of one/ The lack of statistical significance for any of the other explanatory variables and the overall weakness of the results may suggest the need for a different growth model that emphasizes other factors more 7. The hypothesis that this coefficient is in fact 0.8 or 0.9 (the values obtained for semi-industrial countries) cannot, however, be rejected. Furthermore, a Chow test implies that the hypothesis that the two groups of developing countries have the same parameters for equation 9-10 can also not be rejected. But this is not strong enough evidence to suggest that the two samples should be combined. The estimates for the less developed countries have such a high variance that many alternative hypotheses cannot be rejected. GROWTH IN SEMI-INDUSTRIAL COUNTRIES 271 Table 9-4. Regression Results for Semi-Industrial Countries, 1955-63; Neoclassical and Industry!Nonindustry Disequilibrium Models Disequilibrium Disequilibrium model I model II (equation 9-1 0; (equation 9-1 0; industry defined industry defined Variable or Neoclassical as manufacturing as manufacturing result model only) and construction) I/Y 0.135 0.067 0.053 (2.261) (0.816) (0.656) GL 0.740 0.848 0.878 (2.549) (2.264) (2.417) GM. (M/Y) 0.977 0.822 (1.603) (2.046) Constant 0.007 0.007 0.007 (0.538) (0.332) (0.366) Adjusted R 2 0.258 0.134 0.186 Standard error of regression (percent) 1.770 2.013 1.952 Number of observations 31 27 27 Note: Numbers in parentheses are t values. Source: World Bank data. relevant for such economies. A discussion of these issues, however, is beyond the scope of this chapter. Consideration of an earlier time period for the group of semi-industrial economies was complicated by data problems. Since labor force data (whether aggregate or by industrial sector) were not readily available, the growth of the total labor force was approximated by population growth (a practice quite common in cross-country studies of the sources of growth). Estimating sectoral labor growth demanded the assumption that 8 = J-l (see equation 9-9)-that is, that the ratios of respective marginal factor productivities are equal across sectors. As indicated earlier, for the period 1964-73 there was no strong indication of a substantial difference be- tween 8 and J-l· The results for the period 1955-63, presented in table 9-4, indicate that none of the models explains much of the variation in growth. The coefficient of investment (which according to the analytical framework measures marginal capital productivity in the nonindustrial sector) is small in absolute value, but it has a high standard deviation. In general, the hypothesis that parameters have remained constant over time cannot be rejected. But the low R 2 may imply that important variables were left out, which could bias the estimates. Also, differences in the 272 PRODUCTIVITY AND STRUCTURAL CHANGE parameters among countries may have been larger. No convincing argu- ment exists, therefore, for pooling observations over time. 8 In summary, the results reported in this section suggest that a substan- tial difference existed in the period 1964-73 between marginal factor productivities in the industrial and in the nonindustrial sectors of the sample group of semi-industrial countries. Consequently, those countries that pursued accelerated industrial growth tended-all else being equal- to grow faster than other countries in the group since resource allocation was closer to being optimal. Disequilibrium between Export and Nonexport Sectors The relation between export performance and economic growth has been a subject of considerable interest to development economists in recent years. (The presentation in this section draws on my earlier paper [Feder 1983].) A substantial body of literature suggests that distinguishing between outward-oriented and inward-oriented sectors might be useful in comparing countries' growth experiences. Empirical comparisons of countries tend to demonstrate that developing countries with favorable export growth records have generally enjoyed higher rates of growth of national income than other developing countries. Since exports are a component of aggregate output, a positive correlation coefficient is to be expected (Kravis 1970). But several empirical studies argue that rising exports contribute more to GDP growth than the change in the volume of exports alone would suggest (Balassa 1978; Heller and Porter 1978; Michaely 1977; Michalopoulos and Jay 1973). Explanations for these observations have been discussed by many econ- omists. They point to various benefits of export activity-such as greater capacity utilization, economies of scale, incentives for technological im- provements, and more efficient management-that arise from competitive pressures abroad (see Balassa 1978; Bhagwati and Srinivasan 1978; Kees- ing 1967, 1979; and Krueger 1980). These discussions imply substantial differences between marginal factor productivities in outward-oriented and inward-oriented industries, with the former having the higher factor productivity. It follows that countries which have adopted policies less biased against exports have benefited from resource allocation that is closer to being optimal-and from faster growth. In accordance with the disequilibrium framework developed earlier in this chapter, an economy is viewed as being composed of two distinct sectors, one producing for the domestic market and the other for the foreign market. 9 Two modifications to the model summarized by equation 8. Results for the less developed countries in the period 1955-63 were very poor. The adjusted R 2 was close to zero. There was no point in comparing these results with those for the later period or for the semi-industrial countries. 9. Clearly this is an abstraction, as many firms produce for both domestic and foreign markets. It may be argued that, even so, the domestically marketed output of such firms has GROWTH IN SEMI-INDUSTRIAL COUNTRIES 273 9-9 are needed. The first is dictated by the lack of information on the growth of labor in the export sector. This means that equation 9-9 can be used only under the assumption that o = f1, that is, that the ratio of marginal capital productivities in the two sectors is the same as the ratio of marginal labor productivities. If the main sources of disequilibrium are the distortions in product markets (rather than in factor markets), then the imposition of o = f1 is not a source of serious bias. The second modification to the original framework relates to the spec- ification of externalities. In the case of the disequilibrium between the industrial and nonindustrial sectors, an argument was made for the possi- bility of substantial intrasectoral externalities in manufacturing. In the case of the disequilibrium between the export and nonexpert sectors, an argument can be made for significant intersectoral externalities. These follow from the beneficial effects of export activities on other sectors in the economy through the development of efficient and internationally com- petitive management, the introduction of improved production tech- niques, the training of skilled workers, and the spillover consequences of scale expansion (Keesing 1967, p. 311; 1979, pp. 4, 5). For modeling purposes, such effects are best represented by introducing in equation 9-1 the volume of output of the export sector as a factor affecting output of the nonexport sector. Equation 9-2 can then be rewritten: (9-11) Y1 =FkAKt +F{ALt +NAY2 and (9-12) where 1 denotes the nonexport sector, 2 the export sector, and F:k the marginal effect of exports on the output of the nonexpert sector (that is, aY1/aY2). Combining equations 9-11 and 9-12 with equations 9-3, 9-4, and 9-8, and assuming o = f1, yields a slightly modified version of equation 9-9: (9-13) Gy= Fk.!. y + ~1·GL + ( 0 - -+ 1+0 F:k)(Gy2 y 2). y If F:k is treated as a constant parameter, equation 9-13 is not different from equation 9-9 for econometric purposes, except for the assumption o = f1· But if a specific form is adopted for F:k -one which hypothesizes that the extent of intersectoral externalities depends on the size of the export and nonexport sectors-then the specification of the econometric model is affected. Suppose that the relation between nonexports and exports is governed by a fixed elasticity, e, such that the same qualities (and factor productivities) as the exported output. To the extent that the growth of exports represents a good approximation of changes in the volume of production of such firms, the results are still valid. 274 PRODUCTIVITY AND STRUCTURAL CHANGE Table 9-5. Regression Results for Semi-Industrial Countries, 1964-73; Neoclassical and Export!Nonexport Disequilibrium Models Disequilibrium Disequilibrium Variable Neoclassical model III model IV or result model (equation 9-13) (equation 9-15) II¥ 0.242 0.148 0.104 (3.670) (3.008) (2.674) GL 0.612 0.606 0.593 (1.827) (2.563) (3.283) Gx • (X/Y) 0.446 0.302 (5.676) (4.496) Gx 0.145 (4.732) Constant 0.000 0.010 0.010 (0.005) (0.768) (1.092) Adjusted R 2 0.273 0.638 0.789 Standard error of regression (percent) 1.827 1.289 0.985 Number of observations 34 34 34 Note: Numbers in parentheses are t values. Source: World Bank data. (9-14) ¥1 = Y~ ·H(Kt. L1) where His some function of sectoral inputs. Then Fi = 9Y 1/Y2 , and since by definition Y1 = Y - ¥ 2 , equation 9-13 can be written: (9-15) where X = Yl> the output of the export sector. This formulation allows the externality effect to be separated from other effects that may cause deviation between marginal factor productivities, albeit at the cost of adopting a specific form for Fi . Results will be presented for both equations 9-13 and 9-15, which will be labeled dis- equilibrium models III and IV respectively. 10 The same considerations that led me in the preceding section to focus on semi-industrial economies in the period 1964-73 apply to this analysis, as will be shown. 11 The results in table 9-5 suggest that the export sector has 10. The work of Chenery, Elkington, and Sims (1970) includes some estimates that are identical to equation 9-13. The papers by Balassa (1978), Michalopoulos and Jay (1973), and Tyler (1981) include the rate of growth of exports and are thus equivalent to equation 9-15 under the assumption that 8/(1 + 8) = 0. 11. All results in this section use population growth to approximate labor force growth. Population data are available for the full period 1964-73, labor force data only for the period 1960-70. GROWTH IN SEMI-INDUSTRIAL COUNTRIES 275 higher factor productivity at the margin. Comparing models III and IV tends to support the formulation that allows variability among countries in the extent of intersectoral externalities. As in the preceding section, the coefficient associated with investment decreases when a disequilibrium formulation replaces the neoclassical formulation. The explanation is, as before, that in the neoclassical model this parameter represents some "average" marginal productivity of capi- tal, whereas in the disequilibrium model it represents the marginal produc- tivity in the less productive sector. The difference in productivity between the export and nonexport sec- tors is substantial, especially in economies where the export sector is small. It is true that the "average" productivity differential factor (as implied by the parameter of G x · [X /Y] in model Ill) is only about half the estimated coefficient associated with industrial expansion. But the differential, when measured in world prices, is understated in the export/nonexport model because of the protection granted to the nonexport sector in most coun- tries and the distorted prices that result. Table 9-6 presents the sources of growth for the semi-industrial econo- mies as implied by model IV. The higher productivity of resources in the export sector contributes 1.85 percentage points to GDP growth, or about 30 percent of overall growth. Put differently, average growth would be lower by 1. 85 percentage points if all sectors of the economy had the same marginal factor productivities as those observed in the nonexport sector. Thus, the success of export-led growth-as experienced by Korea and Taiwan-comes largely from the shift of resources into high-productivity sectors and from the establishment of new export-oriented and efficient industries. Table 9-6. Sources of Growth in Semi-Industrial Countries, 1964-73; Neoclassical and Export/Nonexport Disequilibrium Models Contribution to growth Neo- Disequilibrium Variable' Sample classical model IV or result mean model (equation 9-15) 1/Y 20.61 4.986 2.141 GL 2.35 1.439 1.394 Gx • (X/Y) 2.21 0.668 } 1 851 Gx 8.16 1.183 . Constant 0.008 1.044 GDP growth' 6.43 6.43 Note: All numbers are multiplied by 100. The small differences in the figures for sample means and for the neoclassical model from those in table 9-2 are due to differences in samples. a. Numbers may not sum to totals because of rounding. Source: World Bank data. 276 PRODUCTIVITY AND STRUCTURAL CHANGE Table 9-7. Regression Results for Less Developed Countries, 1964-73; Neoclassical and Export!Nonexport Disequilibrium Models Disequilibrium Disequilibrium Variable Neoclassical model III model IV or result model (equation 9-13) (equation 9-15) II¥ 0.106 0.059 0.053 (2.037) (1.163) (0.983) GL 0.518 0.727 0.710 (0.840) (1.269) (1.222) Gx • (XIY) 0.652 0.793 (2.811) (1.830) Gx -0.042 (0.388) Constant 0.013 0.007 0.009 (0.769) (0.430) (0.517) Adjusted R 2 0.074 0.214 0.196 Standard error of regression (percent) 1.879 1.732 1.752 Number of observations 42 42 42 Note: Numbers in parentbeses are t values. Source: World Bank data. The export/nonexport disequilibrium model seems to provide a reason- able explanation for much of the variation in growth among semi- industrial countries. Experiments with the sample of non-semi-industrial developing countries yielded results similar to those obtained for this group using the industry/nonindustry disequilibrium model: the under- lying production function framework does not explain much of the varia- tion in growth (see table 9-7). Also, although the coefficient of Gx · (X/Y) is significantly different from zero, the coefficients of investment and labor growth are not significant. Assessment of the superior marginal productiv- ity of factors employed in the export sector is therefore not meaningful. 12 The results for the period 1955-63 for semi-industrial countries are not very different from the results for 1964-73 with respect to parameter estimates, but the significance of the estimates is generally low and the unexplained variation is large (see table 9-8). Regressions for the same period for less developed countries provide a very poor explanation of the growth experience. Thus for both the export/nonexport and industry/non- industry models it seems that data problems as well as the influence of 12. The hypothesis that the parameters of the corresponding regressions for semi- industrial and less developed countries are identical is rejected by a Chow test for models III and IV. This substantiates the decision to treat these two groups separately. GROWTH IN SEMI-INDUSTRIAL COUNTRIES 277 Table 9-8. Regression Results for Semi-Industrial Countries, 1955-63; Neoclassical and Export!Nonexport Disequilibrium Models Disequilibrium Disequilibrium Variable Neoclassical model III model IV or result model (equation 9-13) (equation 9-15) IIY 0.135 0.108 0.092 (2.261) (1.789) (1.447) GL 0.740 0.536 0.569 (2.549) (1.756) (1.837) Gx • (X/Y) 0.429 0.370 (1.704) (1.408) Gx 0.056 (0.829) Constant 0.007 0.011 0.010 (0.538) (0.856) (0.777) Adjusted R 2 0.258 0.305 0.297 Standard error of regression (percent) 1.770 1.713 1.723 Number of observations 31 31 31 Note: Numbers in parentheses are t values. Source: World Bank data. other factors on growth make the early period less suitable for analysis. Regressions for the years 1974-77 for both groups of countries have a low R 2 , and no insight is gained since the parameters of both investment and labor growth have large standard deviations. In summary, the export/nonexport disequilibrium model seems to pro- vide a suitable framework for analyzing the growth of semi-industrial countries but not of other developing countries. The results suggest that, for the group of semi-industrial countries in the period 1964-73, substan- tial differences existed between marginal factor productivities in the ex- port and nonexport sectors. These differences were in part caused by externalities generated by the export sector and benefiting the nonexport sector. The shift of resources toward export industries is thus one of the important sources of growth in this group. A Synthesis: Disequilibrium among Industry, Nonindustry, Export, and Nonexport Sectors The existence of significant differences between marginal factor produc- tivities in industrial and nonindustrial sectors as well as in export and nonexport sectors has been demonstrated. It is natural to attempt a reformulation of the analytical framework to allow the coexistence of both sources of disequilibrium. In this section, we construct a four-sector framework. With some simplifying assumptions, this permits a reesti- 278 PRODUCTIVITY AND STRUCTURAL CHANGE mation that takes into account both types of disequilibrium simulta- neously. The economy is assumed to consist of four sectors: • Nonindustrial goods for the domestic market (denoted by superscript nm and subscript nx) • Nonindustrial goods for export (denoted by superscript nm and sub- script x) • Industrial goods for the domestic market (denoted by superscript m and subscript nx) • Industrial goods for export (denoted by a superscript m and subscript x). To facilitate the derivation of the model and allow for data limitations, two simplifications are introduced. First, the ratios of marginal capital productivities between any two sectors are assumed to equal the cor- responding ratios of marginal labor productivities. 13 Second, marginal intersectoral externality effects of each of the two export sectors on the other two sectors are fixed. With sectoral outputs denoted by Yi (where j = m and nm, and i = x and nx ), the externality effects are given by these six equalities: aY':,'; )\nm aY':,'; )\m X X aY;m aY'; aY::'x ._..;,., aY'::x 1-L'; ay;m aY'; aYr; aY:m = 'Y;m = "('; ay;m aY'; The assumed relations between marginal productivities are (9-16) MPK'::x = MPL';:x =1+8 MPKn,: MPLn,: (9-17) MPKnm MPLnm x =_ _x_ =1+6 MPK':,'; MPL':,'; and (9-18) MPK'; MPLm =--X- = 1 +'T]. MPK'::x MPL'::x The growth of GDP can now be expressed as 13. This assumption was already used in the formulation of the export/nonexport dis- equilibrium model. GROWTH IN SEMI-INDUSTRIAL COUNTRIES 279 -- + a1) Gx·~ + _&_GM· M 6 Gy= MPKnm nxy .!. + j3·GL + (1+6 Y 1+& Y (9-19) +[ TJ - _e_ + a 2 - a 1 ]· Gv';! · ¥:' (1 + TJ) (1 + &) 1+e x Y where a =A.nm+_1_,nm+ 1 "'lnm 1 X 1 + &r-x (1 + TJ) (1 + &) • IX a = A.m + _1_,.m + __ 1_. "'lm 2 X 1 + & r-x (1 +e) IX X= Y;' + y;m M= ¥;'+ Y'::x and j3 is a fixed ratio between marginal labor productivity in the nonin- dustrial domestic sector and average output per worker in the economy (analogous to the formulation in equation 9-8). Equation 9-19, which will be estimated in this section, includes one additional variable: the share- weighted growth of manufactured exports. Whereas the parameters of all other variables are expected to be greater than zero (if the marginal productivity of factors in industrial and export sectors is higher than it is in nonindustrial and nonexport sectors), the parameter associated with ex- ports of manufactured goods could be positive or negative, depending on the relative magnitudes of the other parameters in the equation. Table 9-9 presents the estimates of equation 9-19 for the group of semi-industrial countries using each of the definitions of the industrial sector discussed earlier: manufacturing only, which will be labeled dis- equilibrium model V, and manufacturing and construction, which will be labeled disequilibrium model VI. As in earlier regressions, the results are not much affected by how the industrial sector is defined. Furthermore, comparisons with tables 9-1 and 9-5 show that parameter estimates are not much different. As predicted, the parameters associated with export growth and industrial growth are significantly positive. The parameter of manufactured export growth is negative but not significantly different from zero. A direct estimate of 'T] (the parameter incorporating the hypothesized higher productivity of manufacturing industries that pro- duce exports over other manufacturing industries) cannot be obtained, but some indication that manufactured exports have higher marginal produc- tivity than other manufactured goods can be inferred. Adding 6/(1 + 6) + a 1 to the parameter of manufactured export growth yields an estimate of 'T]/(1 + TJ) • (1 + &) + a 2 (seeequation9-19),whichaccordingtotable9-9 is approximately 0.18. Since & is positive, the figure 0.18 is lower than 'T]/(1 + TJ) + a 2 • But since the parameter of manufactured export growth 280 PRODUCTIVITY AND STRUCTURAL CHANGE Table 9-9. Regression Results for Semi-Industrial Countries, 1964-73; Four-Sector Disequilibrium Models Disequilibrium model Variable or result (Equation 9-19) (Equation 9-19) V' VI' flY 0.105 0.082 0.135 0.109 (2.048) (1.833) (2.959) (2.656) GL 0.598 0.580 0.766 0.741 (2.442) (2.745) (3.730) (4.150) Gx • (X/Y) 0.338 0.312 0.246 0.228 (3.240) (3.437) (2.959) (3.231) GM. (M/Y) 0.832 0.898 0.809 0.898 (3.421) (4.804) (3.681) (5.072) cv;; • (Y;'IY) -0.153 -0.132 (1.227) (1.258) Constant 0.009 0.010 -0.002 0.000 (0.641) (0.813) (0.132) (0.006) Adjusted R 2 0.723 0.791 0.752 0.809 Standard error of regression (percent) 1.071 0.929 1.072 0.940 Number of observations 29 29 32 32 Note: Numbers in parentheses are t values. Source: World Bank data. has a large standard deviation, many numerical values can theoretically apply. Next, this parameter was assumed to be approximately zero, 14 thereby excluding manufactured export growth, and models V and VI were reesti- mated. The results, labeled models V' and VI' in table 9-9, indicate the same order of magnitude for estimated parameters. The conclusions drawn from these results, therefore, confirm what was said above about the sources of growth: the shift of resources into industrial and export sectors-that is, into sectors with higher productivity-characterizes countries that grow faster. The sources-of-growth accounting detailed in table 9-10 indicates that growth would have been lower by more than 2 percent if all sectors had had the same productivity as the domestic- oriented nonindustrial sector. Alternative Interpretations of the Results The interpretation given to the results so far is both plausible and consistent with the underlying framework. But it is not necessarily the only 14. This would still be consistent with positive values for TJ and the externality effects generated by manufacturing exports. GROWTH IN SEMI-INDUSTRIAL COUNTRIES 281 Table 9-10. Sources of Growth in Semi-Industrial Countries, 1964-73: Models V' and VI' Variable Sample Disequilibrium Disequilibrium or result mean model V' model VI' 1/Y 20.18 2.727 2.195 GL 2.32 1.778 1.720 Gx • (X/Y) 2.21 GM (M/Y) 0.542} 2 031 1.489 . OS03) Manufacturing 1.84 2.461 Manufacturing and construction 2.18 1.958 Constant -0.150 0.010 GOP growth 6.386 6.386 Note: All numbers are multiplied by 100. Source: World Bank data. possible interpretation. Alternative frameworks can be constructed that would yield specifications similar to the ones used. This is demonstrated below for the export/nonexport disequilibrium model, for which there is an alternative a priori reasoning that is plausible. Foreign exchange may be considered an important determinant of GDP growth. In fact, the simple trade-gap model implies that when the foreign exchange constraint is binding, growth depends only on foreign exchange inflows (from aid, borrowing, and exports). Even in a less rigid model, larger amounts of foreign exchange allow more flexibility and efficiency in production: bottlenecks are minimized, the need for using lower-quality domestic components is reduced, and the pressure for inefficient import substitution may be lower. It is, therefore, possible that the significant positive association between export growth and GDP growth indicated by the earlier analysis stems from the fact that exports are the main source of foreign exchange. Let us construct a model that considers the foreign exchange aspect of exports. Suppose that GDP is generated subject to the production function (9-20) Y= F[min (:d, ~m), Wm, L] where Kd and Km denote domestic and foreign-made capital, Wm denotes imports of intermediate goods, and L denotes labor. Assume that the supply of domestic investment is linearly related to GDP: (9-21) Kd = 'Yo + 'Y1 Y. Assume also that consumption imports, M 0 are a fixed proportion of GDP: (9-22) Efficiency considerations dictate that 282 PRODUCTIVITY AND STRUCTURAL CHANGE (9-23) where MK denotes imports of capital goods and the dots over the variables denote time derivatives: k = dKidt, and so forth. The change in GDP is (9-24) With total supply of foreign exchange denoted by S, it must hold that (9-25) w = s- M - MK = s- &y- o.Kd m c 13 = s- &y- ~ . 13 ' "'1 . Y. Given equations 9-21 and 9-25, and denoting (o. -y 1ll3) =-y, equation 9-24 can be written as . . . . . (9-26) Y = F 1 ·Kd + F2 • [S- (& + -y) Y] + F3 L. Note that I=Kd + Km = (1 + o./13) kd (by equation 9.22). Rearranging the terms of equation 9-26 yields (9-27) Manipulating equation 9-27 and adopting an assumption analogous to equation 9-8 regarding the relation between marginal labor productivity and average output per worker, one eventually obtains I S (9-28) Gy = a 1 · - + a2 · Gs·- + a3 · GL. y y Note that Gs · (SlY) = Gx · (X/Y) + Gp ·FlY, where F denotes foreign exchange inflows other than export revenues. If Gx · (XIY) is highly correlated with Gs · (SlY), then estimation of equation 9-28 will yield results not much different from equation 9-15. Indeed, an estimate of equation 9-28 for the semi-industrial countries in the period 1964--73 yields Gv = 0.009 + 0.139 ·!. + 0.404· Gs·~ + 0.704GL; (JP = 0.626). y y (0.682) (2.738) (5.50) (2.924) Comparison with table 9-5 verifies that no substantial difference exists between the two formulations. The conclusion is that both the earlier explanation based on marginal productivity differences and the trade-gap explanation may be valid simul- taneously. For some countries, one explanation may be more appropriate than the other, but this cannot be discerned from the results reported in this chapter. 10 Productivity Growth in Manufacturing MIEKO NISHIMIZU SHERMAN ROBINSON THE LAST two chapters explored a number of issues related to sectoral differences in factor productivity and the contribution of structural change to aggregate growth. In developing countries, there are many constraints on how fast employable resources can grow and on how easily they can be transferred across sectors. Aggregate growth depends not only on factor accumulation and its sectoral allocation, but also on total factor produc- tivity growth. In a constrained economy, achieving rapid rates of TFP growth is therefore a real issue in alleviating economic bottlenecks. Fur- thermore, as discussed in chapters 6 and 7, an important part of the "catching up" process involves exploiting changing comparative advan- tage, which provides a significant driving force for structural change. Differential sectoral rates of TFP growth are crucial determinants of evolv- ing comparative advantage and have a great effect on both growth and structural change in the medium to long run. Two issues concerning TFP growth are especially relevant for develop- ment policy. First, what range of TFP growth rates can one reasonably expect? Confidence intervals for TFP growth rates can in principle be obtained from historical records of firms, industries, or economies operat- ing under varying production environments. They provide information useful for answering various questions about development. For example: First, what is the appropriate duration of infant industry protection or promotion policies? Is five years too short? Is twenty years too long? Should the duration be uniform among industries, or should it differ from industry to industry? Second, what are the causes or sources of TFP growth? For example, does protection from competing imports destroy incentives to improve efficiency in production? Can some policies improve productivity-for example, subsidies tailored to specific factors such as fiscal incentives for accelerated depreciation or support for employee training? Over the years, the empirical literature on TFP change has accumulated a This chapter originally appeared in a slightly different form in the journal of Development Economics 16 (1984): 177-206. It is reprinted here with permission. 283 284 PRODUCTIVITY AND STRUCTURAL CHANGE substantial body of stylized facts about the contribution of productivity change and factor input growth to economic performance in various economies.' Perhaps the most significant stylized fact to emerge is the importance of TFP change in contributing to growth: as much as one-third to one-half of growth in output can be attributed to TFP change. Until quite recently, much of what we knew was in terms of macro aggregates. 2 There is now, however, a small but growing empirical literature on TFP change at a disaggregated level. 3 The first objective of this chapter is to add to this body of stylized facts by analyzing time-series data developed at the World Bank on TFP growth at the sectoral level within manufacturing for three countries: Korea, Turkey, and Yugoslavia. We include Japan in the sample as a comparator; data for it were developed by Jorgenson and Nishimizu (1981). In contrast to the growing stock of empirical estimates on TFP growth, sufficient evidence has not yet been accumulated to establish the causes of productivity change. As surveyed and discussed extensively by Nelson (1981), the literature on productivity change offers a wide variety of possible causes but no consensus as to which deserve most attention. In the development literature, the role of trade policy in increasing growth and efficiency has long been a main theme. Therefore, the second and more important objective of this chapter is to examine the effects of various development strategies, especially trade policies, on sectoral TFP growth. Our analysis is exploratory and considers several of the suggested hypoth- eses-of which there is certainly no shortage. Indeed, it is difficult to sort out the differences among them and to define the appropriate measures and tests required to assess each one. Our analysis does indicate important links between trade policies and productivity performance, and it raises some issues for further research. One hypothesis put forward in the literature is that a positive relation exists between productivity change and the rate of growth of output. Expressed in terms of labor productivity, this relation has been called 1. For a review of the literature, see Nadiri (1970, 1972). For an excellent critical survey of the productivity literature, see Nelson (1981). 2. See, for example, Christensen and Jorgenson (1973); Christensen, Cummings, and Jorgenson (1980); Denison (1967, 1974); Denison and Chung (1976); Ezaki and Jorgenson (1973 ); and Griliches and Jorgenson (1967) for studies on developed countries. For develop- ing countries, see Christensen, Cummings, and Jorgenson (1980), who included Korea in their international comparison, and Bruton (1967) and Elias (1978), who studied Latin American countries. See also Robinson (1971), Feder (chapter 9), and studies cited by them as well as by Nadiri (1972). 3. See, for example, Kendrick (1961, 1973) and Gollop and Jorgenson (1980) for the United States; Nishimizu and Hulten (1978) and Kuroda and Imamura (1981) for Japan; and a comparative study of the United States and Japan by Jorgenson and Nishimizu (1981). In addition, there are productivity studies of regulated industries or firms in the United States and Canada; see, for example, Cowing and Stevenson, eds. (1981). A comprehensive study of Indian manufacturing industries was made by Ahluwalia (1985). See also Ezaki (1975) on the Philippines, Kuo (1983) on Taiwan, and Kim and Son (1979) on Korea. PRODUCTIVITY GROWTH IN MANUFACTURING 285 Verdoorn's law after P. J. Verdoorn, who suggested it in 1949. Among those who have investigated this relation, Kaldor (1967) has argued that the fundamental explanation for it lies in economies of scale. 4 He has also noted that it is observed most prominently in manufacturing and other industrial activities. In developing countries, economies of scale and size of market have long been considered important in determining growth and structural change. 5 The existence of scale economies, or any other justifica- tion for Verdoorn's law, implies that widening the market through trade should lead to reductions in production costs. The argument is usually made in terms of the benefits of an expansion in demand through increased exports. Although the argument depends on the size of domestic markets, it should in principle apply to import substitution as well. A quite different trade policy hypothesis is that opening up to interna- tional competition will spur increases in domestic efficiency. There is an implicit challenge-response mechanism induced by competition; domes- tic industries are forced to adopt new technologies, to reduce "X- inefficiency," and generally to reduce costs wherever possible. According to this argument, export expansion is good and so is import liberalization. While a policy of increasing imports may restrict the market for domestic goods, it also increases competition and hence induces greater efficiency. The converse is also widely asserted: protectionist policies designed to promote import substitution reduce competitiveness and lead to ineffi- ciency in production. One must be careful not to overstate the argument. Infant industry protection, by definition, is afforded to high-cost industries that cannot compete with imports until they "grow up" and become internationally competitive. Yet export promotion policies such as exces- sive export subsidies may distort incentives and lead to increasing ineffi- ciency. It is important to focus on the causal mechanism assumed to be working: export expansion and import substitution policies may increase or decrease TFP (levels or growth rates) depending on their impact on competitive, cost-reducing incentives to producers in the medium to long run. The literature on foreign exchange constraints provides yet another hypothesis for a link between trade and productivity. A stylized fact characterizing developing countries is that intermediate and capital goods are not very substitutable with domestically produced goods. In a sense, these imported inputs embody technologies that are unavailable to domes- tic producers and can only be attained through imports. Policies that limit the availability of such imports, or make them more expensive, will lead to poor productivity performance. In contrast, policies that increase the availability of imported inputs or lower their costs-such as increased foreign aid or an export-led development strategy-will reduce costs for 4. See also Salter (1960) and Kaldor (1961). 5. See Balassa (1967) and Chenery and Westphal (1979). 286 PRODUCTIVITY AND STRUCTURAL CHANGE domestic industries and lead to better productivity performance. In this view, exports are important only as a source of foreign exchange; they permit industries to buy inputs that can be produced domestically only at a much greater cost, if at all. These hypotheses about possible links between alternative development strategies distinguished by trade policies and TFP growth are not mutually exclusive. They may all be true, and the postulated effects need not be independent of one another. Given the current state of knowledge, it is not possible to discriminate finely among these hypotheses. Indeed, it is not even possible to state with any real confidence what is the direction of causation. It is just as likely, for example, that exogenous TFP growth in a sector generates a shift in the supply curve and, if domestic demand is limited, provides a strong incentive to open up export markets. The possible relationships are myriad and probably have to be sorted out case by case. In this chapter, we begin by looking at the experiences of four countries and try to identify similarities and differences among them at the sectoral level. We then explore some of the hypotheses discussed above by analyz- ing additional data on the nature of the development process in these countries, which at different times have pursued a variety of development strategies and supporting trade policy regimes. This variety yields experi- ments in which different effects dominate the results and enables us to explore the relative importance of factors such as import substitution and export expansion. Before considering the empirical results, however, we first discuss the nature of the TFP measures we use. The measures embody some strong assumptions that affect how they should be interpreted and what they capture as productivity change. The Analytical Framework for Measuring Total Factor Productivity The analytical framework for TFP measurement is founded on the economic theory of cost and production. In recent decades, developments in the field of productivity research have been accompanied by advances in closely related areas of economic theory and measurement. They include duality theory, the theory of index numbers, and the development of flexible functional forms that are less restrictive in representing economic relationships such as production functions and cost functions. 6 Advances in these areas have strengthened the theoretical foundations of TFP measurement. 7 Indexes of TFP change are usually given in terms of output per unit of 6. Caves, Christensen, and Diewert (1981, 1982a,b) provide a good and concise summary of the literature and references. See also Gollop and Jorgenson (1980). 7. In this section, we shall provide a brief exposition of the analytical framework. For a more detailed and technical discussion, see the references cited above. PRODUCTIVITY GROWTH IN MANUFACTURING 287 total factor inputs and are functions of scale elasticities, output and input elasticities, and quantities (or prices) of outputs and inputs. It is usually assumed that output and input markets are competitive and that firms maximize profit subject to a constant-returns-to-scale production function and to market prices that are taken as parameters. Under these assump- tions, output and input elasticities are equivalent to the observed cost shares of factor inputs and revenue shares of each output produced. 8 The index of TFP change can then be computed using only the prices and quantities of outputs and inputs; it equals the difference between revenue- share-weighted output growth rates and cost-share-weighted input growth rates. There is an extensive literature on the choice of an appropri- ate index of TFP change. 9 Essentially, one must specify something about the form of the production function (or, alternatively, of the cost function) in order to justify a particular form of an index. We have chosen the translog production function and the resulting translog index number in our methodology. 10 This framework for TFP measurement has some shortcomings since the simple stylization of production and markets ignores a number of factors and constraints that may be important. In his review of the productivity literature, Nelson (1981) provides detailed criticism and evaluation of the approach. Several issues he raises are worth emphasizing since they affect the interpretation of our empirical results. A production process can be seen as the application of technology to the production of goods and services. Technology, however, is more than machines, tools, and equipment. It may be embodied in workers and managers, in the physical characteristics of material inputs, or in proce- dures and organizational principles that determine how various inputs are combined. It may also be embodied in produced outputs themselves. As Nelson discusses at length, TFP changes may therefore result from all sorts of changes in this broadly interpreted technology as applied to the produc- tion process. Nelson also points out that production takes place within "production environments" that are defined by the nature of the markets for inputs and 8. When these assumptions are not tenable-as in Yugoslavia-direct estimates of output and input elasticities and scale elasticities must be generated. See Nishimizu and Page (1982). 9. For a survey on the theory of index numbers, see Diewert (1979). 10. For a detailed exposition of this approach, see Diewert (1976) for a theoretical discussion and Gollop and Jorgenson (1980) for applications. There is an issue of whether one should work with a value added production function (excluding intermediate input) or with a gross production function (including intermediate input). We have chosen the gross production function approach because we believe that intermediate inputs matter in sectoral TFP change and that it is misleading to assume that intermediate inputs are separable from capital and labor. There is an extensive literature on this issue, but the most comprehensive treatment can be found in Gollop and Jorgenson (1979). Gollop and Jorgenson (1979) also provide a comprehensive treatment and survey of the literature on aggregation over sectoral TFP estimates and on the impact of intersectoral resource shifts on TFP change at the macro level. 288 PRODUCTIVITY AND STRUCTURAL CHANGE outputs and by a set of market and nonmarket constraints such as govern- ment policies. Changes in production environments ultimately affect pro- ductivity performance by altering production constraints through changes in prices, quantities, or qualities of inputs and outputs. They may also have an important shorter-run impact on TFP changes during the process of adjustment to new conditions in production environments. Our empirical results on TFP change thus should not be interpreted as measuring technical change only in the sense of a shift in the frontier of production possibilities because of the implementation of a new genera- tion of technical knowledge. Instead, the measures must be interpreted quite broadly to include such factors as industrial and plant organization, engineering know-how, or changes in response to disruptions in the production process that affect capacity utilization in the short run. The measures really treat production units as a black box. We measure the inputs and the outputs but make no real attempt to describe exactly what is going on inside the plant gate. Figuring out how the black box works is important, but it is beyond the scope of this book. 11 We seek to delineate the stylized facts at a fairly aggregate level and will necessarily be modest in our attempts to generalize and to discern causal links. Growth and Productivity Change in the Manufacturing Industries In this section, we attempt to distinguish systematic patterns of output, input, and TFP growth in the manufacturing industries of Japan, Korea, Turkey, and Yugoslavia. The period we consider is from the late 1950s to the late 1970s. One characteristic that unites the development experiences of these four countries during this period is that they are all semi- industrialized countries Gapan graduated to industrial status during the 1960s). Among the four countries, however, a variety of development strategies and supporting trade policy regimes has been followed. If factors related to stage of development or trade policy have a significant effect on productivity performance, we should be able to see systematic similarities and differences among the four countries caused by these factors. Our empirical analysis is based on data for Japan developed by Jorgen- son and Nishimizu (1981) and on data on TFP growth in manufacturing industries in Korea, Turkey, and Yugoslavia developed at the World Bank. 12 To summarize, data on gross output, labor, capital, and material 11. See Nelson (1981) for a survey of the relevant literature. Research toward this objective is under way at the World Bank in two research projects: "Acquisition of Techno- logical Capability" (RPO 672-48) by Carl Dahlman and Larry Westphal, and "Productivity Change in Infant Industries" (RPO 672-86) by Mieko Nishimizu and John M. Page, Jr. 12. The data come from two World Bank research projects: "Sources of Growth and Productivity Change: A Comparative Analysis" and "Productivity Change in Yugoslavia." More detailed discussions of the results are available in separate papers. See Krueger and Tuncer (1980b) on Turkey, Rhee (1980) on Korea, and Nishimizu and Page (1982) on Yugoslavia for a more comprehensive comparison of the countries. PRODUCTIVITY GROWTH IN MANUFACTURING 289 input in current and constant prices were assembled for the manufacturing industries in these four countries. 13 Conceptually similar methodologies were used in defining the variables and in aggregating to achieve comparability. 14 Gross output and material input by industry are in con- stant 1970 prices in each country. Capital is defined as net capital stock at replacement cost in 1970 prices and includes all nonresidential structures and producers' durables. Land and inventories unfortunately could not be included because of the unavailability of data for the three developing countries. Labor is defined as persons employed since, again, data on hours worked were not readily available for the developing countries. A summary of the industry estimates is presented in tables 10-1, 10-2, and 10-3. In an essay on economic growth, Kaldor (1967, p. 7) stated-with empirical support-that "fast rates of growth are almost invariably associ- ated with the fast rate of growth of the secondary sector, mainly manufac- turing, and ... this is an attribute of an intermediate stage of develop- ment." Table 10-4 presents the average annual growth rates of TFP and gross output and of capital, labor, and material input, as well as the standard sources-of-growth decomposition, for aggregate manufacturing in Japan, Korea, Turkey, and Yugoslavia. These countries all demonstrate rapid growth in manufacturing and seem to fit Kaldor's stylization of being at an intermediate stage of development. Figure 10-1 plots output growth rates of different industries for each country (from table 10-1), along with the sample mean and sample standard deviation. The figure shows that the rapid manufacturing growth in these countries is the result of many industries growing uniformly fast. More than two-thirds of all industries have growth rates that range from 9 to 15 percent in Japan, 17 to 27 percent in Korea, 9 to 18 percent in Turkey, and 6 to 13 percent in Yugoslavia. 15 In Japan, Korea, and Turkey, there are no industries char- acterized by slow growth; and in Yugoslavia, only two industries have growth rates of less than 6 percent a year. Earlier, we noted that size of market may be an important factor in determining growth and productivity change. One way in which this scale effect comes about is through interindustry linkages. Balassa (1967, p. 97) has argued that "cost reductions tend to have a cumulative effect: im- provements in particular industries are transmitted to other sectors through input-output relationships and through the effects of higher in- comes on the demand for consumer goods." These intersectoral links, 13. Data for Korea include 52 manufacturing industries; for Turkey, 33; for Yugoslavia, 19; and for Japan, 21. We have aggregated the data for each country to 16 comparable sectors (roughly the !SIC two-digit classification). 14. The methodology for Yugoslavia differs from that for the others because it could not be assumed that cost-share data reflected the workings of a competitive market (see Nishi- mizu and Page 1982). 15. Note that the sample means of figure 10-1 differ from aggregate manufacturing output growth given in table 10-1 (first row) since the latter is computed as a weighted average. 290 PRODUCTIVITY AND STRUCTURAL CHANGE Table 10-1. Output, Input, and Total Factor Productivity Growth in Four Countries by Industry (percent) Japan (1955-73) Korea (1960-77) Industry 2 3 4 5 2 3 4 5 Food processing 9.36 9.96 3.22 7.11 2.21 16.09 8.50 4.49 13.24 5.26 Textile 7.49 5.98 1.42 7.06 1.70 18.88 13.09 6.68 16.40 4.51 Apparel 12.52 16.23 6.48 11.28 1.94 23.34 22.11 12.75 22.98 1.62 Leather 11.15 8.45 5.09 11.63 0.95 25.20 14.78 18.91 25.46 2.80 Lumber and wood 7.94 7.45 1.98 7.88 1.12 16.32 5.56 4.89 13.00 5.62 Furniture 11.83 9.65 4.97 14.73 -0.09 13.49 4.93 3.74 11.90 4.88 Paper 11.25 10.75 4.96 10.38 1.62 19.41 6.73 7.61 19.37 4.52 Chemicals 12.23 10.86 2.38 10.73 2.50 21.33 14.42 5.93 19.46 4.49 Petroleum and coal 15.28 13.58 3.31 16.69 -0.43 22.81 20.40 2.24 24.06 0.68 Rubber 9.79 14.08 5.14 11.71 -1.22 20.90 16.80 11.02 15.44 5.88 Stone, clay, and glass 12.43 13.22 4.30 12.30 1.73 18.93 11.12 7.20 18.73 4.53 Basic metals 12.11 13.08 4.50 11.85 0.96 25.68 25.58 4.90 25.52 1.87 Fabricated metals 14.33 16.35 7.30 15.20 0.84 22.19 12.49 10.17 19.01 6.01 Machinery 15.90 13.87 6.12 14.56 3.14 23.01 13.31 7.88 21.91 5.73 Electrical machinery 18.26 12.20 7.68 15.72 4.42 36.00 25.87 17.48 31.88 7.25 Transportation equipment 16.69 13.27 6.25 15.89 2.53 28.68 13.64 8.66 30.76 5.10 Note: Column 1 is gross output growth; column 2 is capital input growth; column 3 is labor input growth; column 4 is material input growth; and column 5 is total factor productivity growth. while significant in developed countries, are especially important in de- veloping countries that are undergoing major changes in input-output structure and in the composition of final demand as part of the process of development. 16 It is characteristic of the intermediate stage of development that the share of intermediate demand in total gross production increases significantly over time. This trend would lead one to expect consistently higher output growth in the sectors turning out producer goods across all four countries. And, without prejudging causation, where one sees high output growth one also expects to see high TFP growth. Although it is difficult to map our industry classification strictly accord- ing to producer goods sectors, we can divide the industries into the following four groups: consumer goods, light intermediates, heavy in- termediates, and investment goods. For each country, industries are ranked with respect to both output and TFP growth. The use of such rankings allows us to separate the differences among countries in the 16. Rapid structural change, especially if it also leads to sustained disequilibrium in the factor markets (that is, different marginal productivities across sectors), also has a profound effect on aggregate growth. We shall not pursue this issue further; see chapters 8 and 9 and Robinson (1971). PRODUCTIVITY GROWTH IN MANUFACTURING 291 Turkey (1963-76) Yugoslavia (1965-78) 2 3 4 5 2 3 4 5 Yugoslav industry 8.47 8.30 3.39 6.40 1.91 7.20 7.28 4.55 8.24 -0.65 Food processing 5.74 7.47 -2.04 13.89 -1.71 Tobacco 9.47 10.88 3.35 8.09 1.44 9.77 7.78 3.50 12.87 -0.17 Textile and apparel 18.30 14.80 8.46 17.63 2.74 6.41 16.39 3.25 6.41 -0.98 11.69 8.21 5.29 15.45 -0.14 Leather 7.35 11.28 4.92 8.39 -1.20 10.85 7.89 1.94 15.45 -0.60 Lumber and wood, 12.37 19.13 4.34 9.28 3.23 furniture 13.53 12.34 4.24 13.93 1.41 10.77 7.18 3.64 13.01 0.07 Paper 15.23 12.13 7.65 15.55 1.62 12.14 8.19 4.15 14.06 0.10 Chemicals 16.60 17.68 -0.81 14.99 0.45 10.09 9.32 1.02 12.72 0.18 Petroleum 1.32 6.40 -2.91 5.05 1.10 Coal 19.19 13.29 3.59 15.85 5.80 13.19 10.74 5.36 17.55 2.35 Rubber 12.80 13.91 7.05 13.66 0.26 9.90 8.05 1.94 13.70 -0.05 Building materials 8.90 6.99 2.08 12.64 1.72 Nonmetallic minerals 14.98 14.52 11.41 14.62 0.87 6.08 6.85 0.37 7.84 -0.63 Ferrous metals 7.54 8.00 1.13 9.85 -0.65 Nonferrous metals 7.57 9.68 -0.88 6.55 1.51 12.58 7.35 4.18 16.31 0.60 Metal products 17.61 13.64 13.97 17.81 1.33 19.34 19.44 10.99 17.76 1.83 15.55 10.78 4.28 19.29 -0.25 Electrical machinery 19.48 16.05 7.51 19.65 3.33 3.09 6.52 1.35 5.21 -0.25 Shipbuilding Source: World Bank data. average growth rates. Table 10-5 presents these industry rankings. Table 10-6 further summarizes the results by giving a frequency count of indus- tries in each ranking across countries within the four industrial groups. The industry ranking of output and TFP growth arranged in this manner shows a strikingly similar pattern across the four countries of faster growth in heavy industries and slower growth in light industries. Invest- ment goods industries are the fastest growing, followed by heavy in- termediate goods industries and then the two light industry groups. Ken- dall's (multiple) rank correlation coefficient, which measures the similarity in rankings across all four countries together, is 0.75 for output growth excluding Yugoslavia (for which the industry classification differs some- what from the other three countries) and 0.52 including Yugoslavia. A similar but weaker correlation is observed in the industry ranking of TFP growth rates among the four countries. Similarly, a chi-square test treating table 10-6 as a contingency table yields values of 25.4 and 15.7 for the output growth and TFP growth respectively. These values indicate a signifi- cant association between the four industrial groups and their rank accord- ing to both output and TFP growth, although the latter is significant at only the 90-95 percent confidence level. Given these broad similarities, what are the main differences among the Table 10-2. Sources of Growth by Industry (percent) Japan (1955-73) Korea (1960-77) Turkey (1963-76) Yugoslavia (1965-78) Industry 2 3 4 2 3 4 2 3 4 2 3 4 Yugoslav industry Food processing 23.5 19.7 3.7 52.9 32.6 17.9 2.1 47.2 22.6 24.1 3.0 50.2 -9.0 1.7 4.5 102.8 Food processing -29.8 69.8 -3.8 63.8 Tobacco Textile 22.6 4.9 2.8 69.5 23.8 16.3 3.8 55.9 15.2 26.1 4.9 53.7 -1.7 20.9 6.6 74.1 Textile and apparel Apparel 15.5 9.2 10.2 65.0 6.9 21.7 17.1 64.1 14.9 13.8 8.2 62.9 Leather 8.5 4.7 8.0 78.6 11.1 11.8 8.0 68.9 -15.2 34.6 4.6 76.0 -1.1 3.3 14.0 83.7 Leather Lumber and wood 14.1 6.8 4.0 74.9 34.4 6.5 1.8 57.2 -16.2 33.7 9.5 72.9 -5.5 2.1 5.5 97.8 Lumber and wood, Furniture -0.7 4.9 11.4 84.3 36.1 8.3 5.1 50.3 26.1 35.2 -4.1 28.2 furniture Paper 14.4 9.2 6.7 69.5 23.3 8.8 3.4 64.3 10.4 28.2 5.3 55.9 0.6 3.4 7.2 88.7 Paper N Chemicals 20.4 19.6 2.4 57.3 21.0 21.1 2.5 55.3 10.6 22.8 6.1 60.4 0.8 0.0 6.9 92.2 Chemicals \() N Petroleum and coal -2.7 12.4 1.0 89.3 2.9 22.8 0.9 73.1 2.7 58.5 -0.1 38.9 1.8 11.5 0.1 86.5 Petroleum 82.8 63.6 -138.3 91.8 Coal Rubber -12.4 27.2 7.4 77.7 28.1 14.8 6.5 50.4 30.2 21.3 1.8 46.5 17.7 0.0 22.5 59.7 Rubber Stone, clay, and glass 13.9 18.1 8.1 59.7 23.9 21.1 5.0 49.9 2.0 38.8 10.4 48.6 -0.4 41.0 1.8 57.5 Building materials 19.3 17.4 9.4 53.6 Nonmetallic minerals Basic metals 7.9 12.8 4.3 74.9 7.2 18.1 0.9 73.6 5.8 30.1 8.9 55.1 -10.4 7.2 0.4 102.6 Ferrous metals -8.6 9.2 1.9 97.4 Nonferrous metals Fabricated metals 5.8 14.7 11.8 67.5 27.0 13.3 6.5 53.0 19.9 28.2 -2.2 54.1 4.8 3.8 10.7 80.6 Metal products Machinery 19.7 12.4 8.7 59.0 24.9 14.1 5.3 55.5 7.5 18.2 9.6 64.5 Electrical machinery 24.2 10.1 7.6 57.9 20.1 19.9 5.6 54.1 9.4 26.6 7.8 56.1 -1.5 2.0 5.0 94.4 Electrical machinery Transportation equipment 15.1 10.3 6.1 68.3 17.7 11.5 3.8 66.9 17.0 17.8 7.4 57.6 -7.9 47.3 24.6 35.9 Shipbuilding Note: Column 1 is total factor productiviry growth divided by gross output growth; column 2 is the contribution of capital input growth divided by gross output growth; column 3 is the contribution of labor input growth divided by gross output growth; and column 4 is the contribution of material input growth divided by gross output growth. Source: World Bank data. Table 10-3. Factor Input Shares by Industry (percent) Yugoslavia japan (1955-73) Korea (1960--77) Turkey (1963-76) (1965-78) Industry 1 2 3 1 2 3 1 2 3 1 2 3 Yugoslav industry Food processing 18.6 10.9 69.8 34.0 7.6 57.4 24.6 7.4 66.5 1.6 7.3 89.8 Food processing 53.7 10.8 26.3 Tobacco Textile 6.2 14.8 73.4 23.6 10.8 64.4 22.7 13.7 62.9 26.3 18.5 56.3 Textile and apparel Apparel 7.1 19.8 72.2 23.0 13.1 65.2 17.1 17.8 65.4 Leather 6.3 17.7 75.4 20.2 10.8 68.2 13.5 9.2 76.0 4.8 31.0 63.4 Leather Lumber and wood 7.4 16.2 75.5 19.1 6.1 71.8 22.0 14.2 63.9 3.0 30.9 67.0 Lumber and wood, Furniture 6.1 27.2 67.8 22.7 18.9 57.1 21.7 13.2 63.3 furniture N \0 Paper 9.9 15.3 75.4 25.6 8.9 64.5 31.0 17.0 54.3 5.2 21.4 73.5 Paper w Chemicals 22.2 12.6 65.3 31.3 9.1 60.6 28.7 12.3 59.2 0.0 20.2 79.7 Chemicals Petroleum and coal 13.9 4.5 81.8 25.5 9.8 64.5 55.0 3.7 43.1 13.5 17.5 68.9 Petroleum 13.1 62.9 24.2 Coal Rubber 19.0 14.2 65.0 18.5 12.3 68.3 30.9 10.0 56.3 0.0 55.4 44.9 Rubber Stone, clay, and glass 17.1 23.5 60.4 36.0 13.2 50.5 35.7 19.0 45.6 50.6 9.8 41.6 Building materials 22.3 40.4 37.8 Nonmetallic minerals Basic metals 11.9 11.6 76.6 18.2 5.3 74.1 31.1 11.7 56.5 8.7 13.3 78.1 Ferrous metals 11.9 15.4 74.2 Nonferrous metals Fabricated metals 12.9 23.3 63.7 23.7 14.4 61.9 21.3 14.9 64.3 6.5 32.3 62.2 Metal products Machinery 14.3 22.7 64.5 24.5 15.5 58.4 23.5 12.2 63.8 Electrical machinery 15.2 18.2 67.4 27.8 11.7 61.2 26.5 13.7 61.1 3.0 18.2 76.2 Electrical machinery Transportation equipment 13.0 16.3 71.8 24.2 12.6 62.4 21.7 19.3 57.2 22.4 56.3 21.3 Shipbuilding Note: Column 1 is capital input share; column 2 is labor input share; and column 3 is material input share. Source: World Bank data. 294 PRODUCTIVITY AND STRUCTURAL CHANGE Table 10-4. Sources of Growth for the Manufacturing Sector (percent per year) japan Korea Turkey Yugoslavia Measure (1955-73) (1960-77) (1963-76) (1965-78) Gross output 11.59 17.94 10.71 9.78 Capital input 10.84 12.98 11.24 7.72 Labor input 4.50 5.32 5.05 2.99 Material input 10.41 16.29 9.29 11.55 Weighted capital input' 1.51 (0.130) 3.50 (0.195) 3.23 (0.302) 0.78 (0.080) Weighted labor input' 0.70 (0.060) 0.46 (0.026) 0.55 (0.051) 0.67 (0.069) Weighted material input' 7.34 (0.633) 10.28 (0.573) 5.60 (0.523) 7.85 (0.802) Total factor productivity change' 2.04 (0.176) 3.71 (0.207) 1.33 (0.124) 0.48 (0.049) a. Ratios of weighted capital, labor, and material input growth as well as total factor productiv- ity change to gross output growth are given in parentheses. Source: Tables 10-1 and 10-2. four countries? In particular, are there systematic differences in productiv- ity performance by industries among the four countries? The country differences in TFP growth can be summarized statistically. For this pur- pose, we estimate a log-linear time-trend equation for TFP change over the individual industry's annual time series pooled across countries. The ordinary least squares regression (with standard errors given in paren- theses) is In TFP = 0.0085 + 0.0194t + 0.0177Kt- 0.0105Tt- 0.0195Yt (0.0054) (0.0014) (0.0020) (0.0023) (0.0020) where R 2 is 0.475; the sample size is 1,054; the variable tis time; and the variables K, T, and Yare country dummies set to one for Korea, Turkey, and Yugoslavia respectively and to zero otherwise. All estimated coef- ficients are significant at the 99 percent level, other than the intercept term (as should be expected since the level index of TFP is one in the base year). These results indicate that the sectoral TFP growth rates in Korea, Turkey, and Yugoslavia differ significantly from the rate in Japan. In Korea, TFP growth is 1. 77 percent above Japan, in Turkey it is 1.05 percent below Japan, and in Yugoslavia it is 1.95 percent below Japan. The difference in TFP growth rates is also statistically significant between Korea and Turkey, Korea and Yugoslavia, and Turkey and Yugoslavia. Furthermore, Yugo- slavia's TFP growth rate is the only one not significantly different from zero. Returning briefly to the aggregate manufacturing estimates in table 10-4, we note that differences in TFP growth reflect another marked difference in the manufacturing growth process of these four countries- that between the relative importance of TFP growth and of factor input growth in output growth (see the last four rows in table 10-4). Japan and Korea are similar in that TFP change is as important as capital and labor PRODUCTIVITY GROWTH IN MANUFACTURING 295 Figure 10-1. Distribution of Sectoral Growth Rates Japan (1955-73) ··- ·-·. ... ·- .. x = 12.41; s = 3.05 Korea (1960-77) • .... - ..... x = 22.01; s = 5.38 Turkey (1963-76) x = 13.66; s = 4.64 Yugoslavia (1965-78) x= 9.20; s = 3.68 .. ···- . I I I I I I I 0 5 10 15 20 25 30 35 40 Rate of output growth (percent) Note: Each dot represents a sector. Sector definitions are given in table 10-1. The variable x is a sample mean; s is standard deviation. input growth combined. 17 For Turkey, although the rate of TFP growth is respectable, its contribution to output growth is significantly less than that of capital and labor combined. In sharp contrast, Yugoslavia's manufac- turing growth involves very little TFP growth-virtually all growth is derived from increases in the quantity of inputs. The relative importance of TFP growth on the one hand and of capital and labor growth on the other at the disaggregated industry level can be examined in figure 10-2. From the data in table 10-2, the share of TFP growth in output growth is plotted against the share of capital plus labor growth in output growth in each industry by country, with the 45-degree line indicating equal contributions. These industry results also show that Japan and Korea differ from Turkey and Yugoslavia as noted above. In Turkey and Yugoslavia, the individual industry results mirror the aggre- gate pattern; with only one exception, the contribution of sectoral TFP growth is less important than that of capital and labor input combined. In Yugoslavia, almost all industries derive their growth in output from increases in factor inputs, with zero or negative contribution from TFP growth. 17. Since the contribution of material input growth to gross output growth is always the dominating factor in manufacturing, we shall focus on the relative importance of TFP, capital, and labor growth in our discussion. Table 10-5. Ranking of Gross Output Growth and Total Factor Productivity Growth Gross output growth Total factor productivity growth N \0 0\ Rank 1-4 Rank 5-8 Rank 9-12 Rank 13-16 Rank 1-4 Rank 5-8 Rank 9-12 Rank 13-16 Industry J K T Y J K T Y J K T Y J K T Y J K T Y J K T Y J K T Y J K T Y Consumer goods Food processing' @@@@ KD®0 16 Textileh @@ @@ 8 @®<£ Apparelh 4 ®0 11 4 6 9 15 Leather 4 6 12 16 8 12 @@ Light intermediates Lumber and wood' 7 @@@ 5 10 @@ Furniture' 7@@ 16 3 8 @ @ Paper @@®~ 7®@@ Heavy intermediates Chemicals ® 0s 9 4 06 12 Petroleum and coaJd 4 10 5 15@ 13 00 Rubber 10 13 00 16 Stone, clay, and glass' 7 @@ 13 7 9 14 Basic metals 1 3 8 9 15 @ @ @ 15 Investment goods Fabricated metals• 14 0 8 13 Machinery• 0 00 11 Electrical machinery 000 00 6 11 Transport equipmenth 000 16 0 0 7 11 N \0 Note: In columns, 1 stands for 1a pan, K for Korea, T forT urkey, andY for Yugoslavia. Ran kings are grouped in quartiles; circled rankings indicate that there is a similar 'I industry ranking in more than one country within each quartile. Kendall's rank correlation coefficient (excluding Yugoslavia) is 0.75 for output growth and 0.47 for TFP growth; significant at the 99.5 percent and 75 percent levels, respectively. Kendall's rank correlation coefficient (including Yugoslavia) is 0.52 for output growth and 0.30 for input growth; significant at the 99 percent and 75 percent levels, respectively. a. Excluding tobacco in Yugoslavia. e. Average of building materials and nonmetallic minerals for Yugoslavia. b. Textile and apparel are considered a tie ranking for Yugoslavia. f. Average of ferrous and nonferrous metals for Yugoslavia. c. Lumber and wood and furniture are considered a tie ranking for Yugoslavia. g. Fabricated metals and machinery are considered a tie ranking for Yugoslavia. d. Excluding coal for Yugoslavia. h. Shipbuilding only for Yugoslavia. Source: World Bank data. 298 PRODUCTIVITY AND STRUCTURAL CHANGE Table 10-6. Sectoral Frequencies across Countries of Ranks of Output and TFP Growth Rank Row Aggregate sector 1--4 5-8 9-12 13-16 sum Gross output growth Consumer goods 2 3 4 7 16 Light intermediates 0 2 6 4 12 Heavy intermediates 4 7 6 3 20 Investment goods 10 4 0 2 16 Column sum 16 16 16 16 64 TFP growth Consumer goods 6 5 4 16 Light intermediates 1 3 4 4 12 Heavy intermediates 5 4 4 7 20 Investment goods 9 3 3 1 16 Column sum 16 16 16 16 64 Note: For industries included in each aggregate sector, see table 10-5. Source: Table 10-5. The differences in productivity performance among countries might be at least partly caused by the nature of the economic policies pursued by each country. One important element distinguishing Japan, Korea, Tur- key, and Yugoslavia from each other is the choice of trade policies in their development strategies. Korea and Yugoslavia (in that order) have manu- facturing sectors that are relatively more open to trade, whereas Turkey and Japan are relatively more closed-the former by design and the latter mainly because of the size of the domestic market. As we shall discuss in the next section, Korea's development strategy has been distinguished by strong export promotion policies, often applied to selected industries. Turkey has long pursued import substitution policies for much of its manufacturing sector, many industries in which are dominated by state enterprises. Yugoslavia has pursued strong import liberalization accom- panied by export expansion. In addition, it has long sought regional and sectoral equalization of productive performance by designing wage, em- ployment, and investment policies to affect all industries similarly. Japan has made use of mixed export promotion and import substitution policies at different times. In the next section, we shall examine the relation between the growth and productivity performance of manufacturing in- dustries and the effect of choosing either an open, export-led development strategy or a closed, import substitution strategy. Trade Strategies and TFP Growth In the introduction to this chapter, we discussed three hypotheses linking TFP growth and trade policies. First, there is a positive link between PRODUCTIVITY GROWTH IN MANUFACTURING 299 Figure 10-2. Relative Contributions of TFP and Primary Inputs to Growth of Output for Disaggregated Industries ..0 Japan ..0 Korea ~ ~ 0 40 0 .... .... 40 Oil // Oil • / ~ ,.. 30 / ~ ,.. 30 f- , / / ."'-- • / "' / "' ...... • '/• / 0 t:: ... .:2 ;::l 20 - • 10 - // .,,• .. ~ 0 t:: .:2 10 f- 20 f- / / / /•• ~ • ..0 / ..0 / • ... ·;::: t:: 0 ... ·;::: t:: 0 0 -10- u 8-10r- (l) (l) ~ -20- ... ~- 20 r- ... t:: t:: (l) u -30 ::l- 30 I I I I I .... (l) i:l. 0 10 20 30 40 50 60 70 ~ 0 10 20 30 40 so 60 70 Percentage contribution of Percentage contribution of capital and labor growth capital and labor growth ... ..0 Turkey ..0 Yugoslavia ::: ~ 0 .... 40 0 .... 40r---------~--------, / / . Oil Oil / / ~ ,.. 30 r- •/ / G: 30- ,.. / "' . .. / / /. 0 20 f- / 0 20 / •• • / t:: ... .:2 ;::l ..0 10 1- / IL/ / • / .:2 ;::l .' t:: ... 10- / // , / . ... ·;::: 0 ~------------~----~ ~ ... 0 • t:: 0 u -10 1- .. t:: 8-10 (l) ... • ~-20 l l l l l l c 1:l - 30 l _l_ l l_ _l _l_ 10 20 30 40 50 60 70 ~ 0 10 20 30 40 so 60 70 Percentage contribution of i:l. Percentage contribution of capital and labor growth capital and labor growth higher exports or (depending on the size of the domestic market) increased import substitution and TFP growth, arising from Verdoorn's law and from the role of export expansion and import substitution policies in increasing the size of the market. Second, there is a positive link between higher exports and TFP growth and a negative link with import substitu- tion (or a positive link with import liberalization), arising from competi- tive cost-reducing incentives or lack thereof. Third, there is a positive link between export expansion, import liberalization, and TFP growth, arising from the importance of foreign exchange constraints and nonsubstitutable imports of intermediate inputs and capital goods. 300 PRODUCTIVITY AND STRUCTURAL CHANGE It is likely that what is observed is the net effect of all these hypothesized forces. As noted earlier, the hypotheses are not mutually exclusive, and distinguishing among them can be quite difficult. They can all be seen as involving a supply response in terms of TFP change to changes in two components of demand: export expansion and import substitution. Tak- ing these components as exogenous, or as determined by exogenous policy regimes, we can then relate TFP growth to changes in the sources of demand growth. One must be very cautious, however, in implying the direction of causality in the relation. For example, it may be that higher rates of exogenous TFP change lead to rapid growth in demand through lower costs and prices. Regardless of causality, however, the existence of any statistically significant relation will provide an interesting starting point for further investigation. The single equation model to be estimated is TFPG = [3 0 + f3EExEE + [3 15x 1s + E, where TFPG, X££, and x 15 are, respectively, annual rates of TFP growth, output growth allocated to export expansion, and output growth allocated to import substitution, and E is the random disturbance term. 18 For each industry in Japan, Korea, Turkey, and Yugoslavia, we use as the dependent variable our estimate of annual rate of TFP change. 19 For the explanatory variables, we combine our estimates of annual output growth rates with demand-side sources of growth decomposition measures. 2° For each industry, total demand can be decomposed into four terms: llX = iitllD + iitll W + llE + llii(Dt+ 1 + Wt+ 1 ) where uis the diagonal matrix of domestic demand ratios (that is, ratios of domestic demand to domestic plus import demand); D, W, and E are final demand, intermediate de- mand, and export demand respectively; and the subscript refers to the time period. 21 The third and fourth terms in the decomposition give the export expansion and import substitution components of demand changes. Di- viding each of these two terms by llX, we obtain share measures of export 18. Strictly speaking, the model should also include output growth allocated to domestic demand growth. We found, however, high co linearity between export and domestic demand growth (whereas no such co linearity problem arose between import substitution and domes- tic demand) for most industries in all countries except japan. It therefore becomes difficult to make a clear statistical distinction between the effect of export expansion and of domestic demand in these cases. Although we can sum the two growth rates, this imposes equality of coefficients between them. We choose instead to omit the domestic demand growth in the analysis below, although it may result in biased estimates particularly of export coefficients, and ask that care be taken in interpreting our results. 19. Note that in tables 10-8 and 10-9 we aggregate to thirteen industries (from those appearing in table 10-5) to achieve consistent data on demand components for all countries. 20. See chapter 5 for a description of the methodology and the data. 21. There is also a total decomposition equation which uses the input-output matrix and therefore incorporates indirect linkages into the decomposition. Since we are concerned with the supply response of individual sectors to changes in demand, the direct decomposition equation is more appropriate for our purpose than the total decomposition equation. There is also an index number problem arising from the choice of initial or terminal weights. We use an average of the analogous Paasche and Laspeyres indexes. See chapter 5 for further discussion of the measures. PRODUCTIVITY GROWTH IN MANUFACTURING 301 Table 10-7. Decomposition of Growth of Manufacturing Demand (percent) Manufacturing Growth decomposition' Domestic Export Import Economy Output Growth demand expan- substi- and years share' rateb expansion szon tution japan 1955-60 47.2 12.6 95.4 5.8 -1.2 1960-65 50.4 10.8 90.2 9.9 -0.1 1965-70 54.6 16.5 92.1 8.1 -0.2 Korea 1955-63 32.1 10.4 64.3 7.2 28.5 1963-70 41.9 18.9 81.8 18.0 0.2 1970-73 49.6 23.8 62.9 38.1 -1.0 Turkey 1953-63 27.9 6.4 90.6 1.3 8.1 1963-68 31.8 9.9 89.6 3.2 7.2 1968-73 36.5 9.4 94.2 6.7 -0.9 Yugoslavia 1962-66 39.0 16.6 90.0 12.7 -2.7 1966-72 45.0 9.1 91.5 21.2 -12.7 a. Average share of manufacturing in aggregate gross output during the period. b. Average annual rate of growth of gross output. c. The decomposition methodology is described in the text and in chapter 5. The three compo- nents sum to 100 percent. The first, domestic demand expansion, includes both intermediate and final demand. Source: World Bank data; described in Kubo (1983). expansion and import substitution in gross output changes for each indus- try in each country. We then multiply these share measures by the annual growth rates of gross output of each industry. 22 Table 10-7 provides a summary of the decomposition results at the aggregate level for the four countries. Note, first, the considerable varia- tion in the relative roles of domestic demand expansion, export expansion, and import substitution, both over time and across countries. In every country the role of export expansion increases over time-dramatically so in Korea and Yugoslavia-and in every country but Japan the role of import substitution decreases. Yugoslavia actually shows significant im- port liberalization (that is, negative import substitution). Korea and Tur- 22. Since the share measures are based on data for a few episodes, we apply the nearest benchmark shares to output growth rates for intervening years in the benchmark periods. The effect is to assume that the share measures reflect a regime that is uniform for each period. Although not ideal, this procedure does provide measures of the two explanatory variables. The episodes for each country are: Japan (1955-60, 1960-65, 1965-70), Korea (1963-70, 1970-73), Turkey (1963-68, 1968-73), and Yugoslavia (1966-72, 1972-78). See chapter 6 for further analysis of the relation between policy regimes and episodes. 302 PRODUCTIVITY AND STRUCTURAL CHANGE key appear to have distinct phases, with a period characterized by signifi- cant import substitution followed by a period of export expansion- although Turkey's short export expansion phase is hardly dramatic, espe- cially compared with Korea's. In Japan, although export expansion is significant in all three periods, the country is very large and the domestic market is the dominant component of demand. All in all, these four countries represent a variety of development experiences. With the excep- tion of Japan, each one underwent a significant shift in development strategy during the period under study, so they should constitute a good sample for statistical analysis. Whether or not trade policies are tailored to particular industries, there is no a priori reason to expect that the manner in which they affect productivity performance is similar for different manufacturing indus- tries. Applying covariance analyses to our panel data reveals significant differences in the estimated regressions among industries in each country and across countries in each industry. Therefore, we report a separate regression for each industry in each country in table 10-8. The regressions reported in table 10-8 indicate that, in general, substan- tial portions of the variation in TFP growth rates are explained by output growth allocated to export expansion and import substitution in Korea, Turkey, and Yugoslavia-but (interestingly) not in Japan. There are also significant differences among manufacturing industries. In Korea, 13 to 83 percent of the variance in TFP change is explained by export expansion and import substitution. Only three industries show less than 30 percent of the variance explained. In Turkey, the range is 13 to 95 percent, with only three industries less than 30 percent. In Yugoslavia, industries range from a low of 3 percent to a high of 93 percent, with four industries less than 30 percent. In Japan, in contrast, the range is from 2 to 41 percent, with all industries other than textiles and apparel showing less than 30 percent of variance in TFP change explained. Note also in table 10-8 that all the statistically significant constant terms ((3 0 ) for Korea, Turkey, and Yugo- slavia are negative, whereas they are all positive for Japan. Negative constant terms in the three developing countries imply reductions in TFP levels (that is, increases in the unit cost of production) unless they are offset by sufficiently large positive contributions from a growth in output through export expansion or import substitution. These striking contrasts between Korea, Turkey, and Yugoslavia on the one hand and Japan on the other point to the relative importance of trade and trade policies in the three developing countries. The results are also consistent with the view that domestic demand has been the prime source of growth in Japan. We also observe in table 10-8 that the estimated elasticities of TFP change with respect to growth through export expansion and import substitution are distinctly larger (in absolute values) in Turkey than in the other countries. Also in Turkey, in all industries except paper products, the elasticities with respect to export expansion are greater than those with PRODUCTIVITY GROWTH IN MANUFACTURING 303 respect to import substitution. Turkey is probably the most closed econ- omy of the four, and these results emphasize the importance of trade at the margin for such an economy. 23 Table 10-9 presents a summary of the regression results and gives only the signs of the statistically significant estimated coefficients. The table also provides an indication of export-oriented and import-competing industries in each country. Export-oriented industries are defined as those with exports greater than 10 percent of total production and import- competing industries as those with imports greater than 10 percent of total domestic supply. Aggregate export and import shares are also given; they indicate the relative openness of the manufacturing sector in these coun- tries. Table 10-9 reveals some interesting results that are hard to observe in table 10-8. Of the twenty-eight cases for which statistically significant elasticities with respect to export expansion are estimated, only two are negative. In contrast, thirteen out of twenty-one significant elasticities with respect to import substitution are negative. Import substitution re- gimes seem to be negatively correlated with TFP change, whereas export expansion regimes seem to be positively correlated with TFP change. In Korea, no industry suffers from export expansion, and those industries that benefit from it are concentrated in the light manufacturing and heavy intermediate categories. In Turkey, the concentration shifts down toward heavy industries. Yugoslavia shows no clear pattern of concentration. Paper products in Turkey and petroleum and coal products in Yugoslavia are the only two industries that show an adverse impact on productivity from export expansion. In Japan, only two industries benefit significantly from export expansion. Industries that experience a significant effect from import substitution are concentrated in Korea and Turkey. More of these are heavy industries in Turkey than in Korea. Only four industries in Yugoslavia and one in Japan show a significant effect from import sub- stitution. These results support some of the hypotheses outlined above and sug- gest four new hypotheses worth examining in future work. First, the results do not confirm the simple version ofVerdoorn's law, which implies that any expansion of the market, regardless of source, improves produc- tivity performance. There are significant and strong differences in the effect of export expansion versus import substitution. Second, the results are consistent with the hypothesis that export expansion leads to higher TFP growth through economies of scale or through competitive incentives. Third, the results are also consistent with the converse hypothesis that increased import substitution leads to lower TFP growth, perhaps by reducing competitive cost-reduction incentives; or that import liberaliza- 23. See Celasun (1983) for an analysis of the structure of Turkish growth during the period under study. Table 10-8. Effect of Export Expansion and Import Substitution on Total Factor Productivity Changes, Multiple Regression Results Japan (1955-73) Korea (1960-77) Durbin- Durbin- Industry ~0 13•• 13.. Rl Watson statistic ~0 ~ .. ~ .. Rl Watson statistic Food processing -0.013 13.740 -4.534 0.189 (1.441) -0.030 11.164. -1.212 0.352 2.042 (0.023) (12.796) (3.318) (0.030) (0.034) (4.088) (7.234) (0.064) Textile and apparel 0.036. 6.399· -3.291 0.408 (1.685) -0.005 0.224• -1.437* 0.236 2.020 (0.017) (2.491) (3.680) (0.061) (0.015) (0.129) (0.766) (0.054) Leather 0.000 .543 -1.252 0.023 1.631 -0.088 1.305• 9.233 0.340 1.567 (0.023) (3.349) (3.550) (0.046) (0.053) (0.541) (7.042) (0.132) w Lumber and wood, -0.008 7.026 -1.848 0.045 2.219 0.003 0.518 .. -0.729 0.402 2.579 a furniture (0.016) (5.649) (3.085) (0.046) (0.019) (0.130) (20.301) (0.090) ~ Paper -0.001 4.341 -8.659 0.034 1.982 -0.042· 3.051 •• -1.790 .. 0.835 1.997 (0.026) (6.262) (14.770) (0.040) (0.015) (0.876) (0.300) (0.047) Chemicals -0.ot8 4.127* -2.351 0.187 1.920 -0.068 .. 1.729 5.096 .. 0.822 2.200 (0.024) (2.245) (1.626) (0.036) (0.018) (2.795) (1.334) (0.044) Petroleum and coal -0.002 1.411 -2.039 0.094 1.643 -0.161• 99.614 .. -24.019 .. 0.568 1.419 (0.011) (2.173) (1.959) (0.029) (0.056) (27.424) (7.023) (0.142) Rubber -O.ot8 -0.070 -18.812 0.037 2.016 -0.017 0.682• 3.582 0.421 2.033 (0.017) (0.678) (24.831) (0.043) (0.034) (0.303) (18.995) (0.078) Stone, clay, 0.010 1.673 -5.265 0.106 1.779 -0.055 2.133* 6.651* 0.372 1.582 and glass (0.008) (1.492) (18.631) (0.027) (0.041) (1.232) (2.344) (0.077) Basic metals 0.021 -1.316 4.593 0.102 1.489 0.016 0.241 -0.298 0.129 2.484 (0.022) (1.970) (3.537) (0.061) (0.021) (0.333) (0.516) (0.049) Fabricated metals 0.003 1.352 -0.671 0.076 1.799 0.029 0.483. 0.019 0.136 1.999 and machinery (0.012) (0.875) (2.517) (0.046) (0.020) (0.265) (0.335) (0.088) Electrical machinery 0.048 .. 0.205 -7.185. 0.213 2.146 -0.024 0.375 -0.614* 0.320 1.870 (0.011) (0.554) (3.765) (0.021) (0.044) (0.253) (0.324) (0.089) Transportation 0.027* -0.214 3.501 0.158 1.974 -0.012 0.314 3.475* 0.439 1.705 equipment (0.010) (0.377) (2.145) (0.017) (0.025) (0.243) (1.175) (0.068) Turkey (1963-76) Yugoslavia (1965-78) Durbin- Durbin- Industry ~0 [3 •• [3,. Rz Watson statistic ~0 ~ .. ~" R> Watson statistic Food processing -0.064* 6.416* -9.917* 0.774 2.871 -0.024 .. 0.703 .. 0.056 0.605 2.201 (0.020) (1.100) (4.138) (0.048) (0.007) (0.233) (1.128) (0.036) Textile and apparel -0.008 0.665 -2.248 0.202 2.221 -0.011* 0.478* 0.170 0.387 1.738 (0.020) (0.430) (1.436) (0.082) (0.004) (0.190) (0.265) (0.008) Leather -0.059* 26.639* -40.567* 0.821 (1.734) -0.005 0.081 -0.625 0.449 2.004 (0.023) (10.643) (13.711) (0.066) (0.007) (0.161) (0.575) (0.016) Lumber and wood, 0.009 8.539 -4.503 0.125 2.007 -0.012 0.493 0.139 0.108 2.040 furniture (0.019) (6.011) (6.189) (0.082) (0.007) (0.475) (0.713) (0.016) Paper -0.070 -3800.580* 163.570* 0.496 1.757 -0.009 -1.186 -1.988 0.169 1.900 (0.043) (1291.160) (54.911) (0.110) (0.009) (1.271) (1.600) (0.021) Chemicals -0.069** 0.798 -10.198 0.741 2.929 -0.001 -0.124 -0.264 0.032 1.396 (0.018) (55.418) (9.707) (0.032) (0.013) (0.864) (0.664) (0.027) w Petroleum and coal -0.144 .. 373.362* - 852.478* 0.794 1.917 0.002 -0.748* -0.199* 0.226 1.777 0 v, (0.037) (60.403) (138.861) (0.102) (0.009) (0.288) (0.080) (0.044) Rubber -0.084 .. 134.499* 2.029* 0.952 2.018 -0.013 2.522 .. 0.138 0.772 2.775 (0.018) (21.511) (0.144) (0.040) (0.008) (0.465) (0.154) (0.017) Stone, clay, -0.067* 32.684 .. -38.329 .. 0.643 1.601 -0.005 2.005 .. -0.805* 0.576 1.806 and glass (0.021) (7.702) (9.463) (0.044) (0.004) (0.364) (0.289) (0.018) Basic metals -0.068 36.644 .. 4.715 .. 0.843 (1.724) -0.008 0.004 -0.100 0.222 (1.707) (0.038) (5.522) (0.868) (0.064) (0.006) (0.109) (0.131) (0.021) Fabricated metals -0.005 13.165* 0.109 0.224 2.308 -0.013* 1.119 .. 0.052 0.700 1.855 and machinery (0.018) (5.197) (0.674) (0.072) (0.004) (0.233) (0.079) (0.007) Electrical machinery -0.000 30.900 .. -2.002 0.621 2.952 -0.011 2.222 .. -9.051* 0.515 1.783 (0.027) (7.856) (1.366) (0.054) (0.010) (0.682) (3.106) (0.018) Transportation -0.017 221.755 .. 0.471* 0.560 1.629 -0.030* -0.048 14.654 .. 0.929 1.883 equipment (0.022) (67.220) (0.221) (0.060) (0.012) (0.044) (1.604) (0.029) • • and * imply coefficient significantly different from zero at 99 and 90 percent levels, respectively. Note: Regression equation: TFPG = J3 0 + J3,,x,, + J3.sx,, + E. Standard errors of coefficients are reported in parentheses below each coefficient. Standard errors of estimate are reported in parentheses below Durbin-Watson statistics. Durbin-Watson statistics in parentheses imply that the Cochrane-Orcutt correction was applied. See also notes a-h in table 10-5. Source: World Bank data. Table 10-9. Summary of Regression Results Export expansion Import substitution Industry Japan Korea Turkey Yugoslavia Japan Korea Turkey Yugoslavia Food processing + + + Me Textiles and apparel +Eo +EO +EO -Me Me Leather +EO + EO Lumber and wood, furniture +EO EO Paper + -Me + Me Chemicals + EO +Me Me Me Petroleum and coal + + -Me w Rubber EO +EO + + + Me a Stone, clay, and glass +EO + +EO + -Me 0'\ Basic metals EO + EO Me +Me Me Fabricated metals and machinery +EO + +Eo Me Me Me Electrical machinery + +EO - -Me Me -Me Transport equipment EO + EO +Me +Me +Me Total manufacturing share of export/production 0.081 0.254 0.037 0.164 Total manufacturing share of import/domestic supply 0.044 0.278 0.112 0.237 Note: EO, export-oriented industry (exports greater than 10 percent of total production); Me, import-competing industry (imports greater than 10 percent of total domestic supply, that is, imports plus total production less exports). These export and import shares were computed for 1973 in Korea and Turkey, for 1972 in Yugoslavia, and for 1970 in Japan. Source: World Bank data. PRODUCTIVITY GROWTH IN MANUFACTURING 307 tion leads to higher TFP growth, by increasing competitive cost-reduction incentives. Fourth, the results are also consistent with the hypothesis that export expansion and import liberalization increase TFP growth by relax- ing the foreign exchange constraint and facilitating imports of nonsubsti- tutable intermediate and capital goods. Such results provide some interesting material for the debate on infant industry protection policies. In every case but one in Korea and in Yugo- slavia, and in every case in Turkey, sectors with a statistically significant negative effect of import substitution on TFP growth are also sectors with a significant positive effect of export expansion. Westphal (1982) has re- cently revived the infant industry argument for selective protection by noting a strong link at the micro level between protection and export performance. He concludes (p. 274) that he has identified "one possible reason why the industrial sector in a country like Korea, following an outward-looking strategy, performs so well; namely, the possibility that its selectively promoted infant industries exhibit superior performance as a result of their export activity." Our results for Korea are consistent with this argument. 24 Krueger and Tuncer (1982) consider the standard infant industry argu- ment in Turkey and conclude ( p. 1149) that "input per unit of output must fall more rapidly in more protected industries if there is to be any rationale for infant industry protection. In the Turkish case, there was no such tendency over the period covered." In Turkey, in contrast to Korea, the export expansion phase was very short and not that strong. It can be stated, therefore, that the positive effect of export expansion on TFP growth that we found did not offset the negative effect of import substitu- tion. Our results are consistent with those of Krueger and Tuncer, but we would be more diffident in concluding that protection was not justified. The positive relation between TFP growth and export performance in Turkey indicates the possibility that it could have followed the Korean example of selective protection, with export performance providing a test of success. Indeed, it still might do so. One final point is worth noting. In both Korea and Turkey, an import substitution phase was followed by a phase with significant export con- tribution to growth. Although the Turkish export phase from 1970-73 turned out to be abortive, largely because the government allowed incen- tives to move against exports, the country is currently entering a new period of rapid and successful export promotion. The observed phasing suggests the hypothesis that a period of protected import substitution is 24. Our results may be somewhat distorted, however, by aggregation problems. Within any one of our "sectors," exports and import substitutes may be very different products. The colinearity issue discussed in footnote 18 needs to be recalled, and care should be taken in interpreting our comments on export expansion, since the estimates may be biased to reflect the domestic demand effect. See also Westphal and others (1985). 308 PRODUCTIVITY AND STRUCTURAL CHANGE useful-perhaps even necessary-to build a base from which a successful export drive, with associated positive TFP growth, can be launched. 25 Westphal's argument holds out the hope that the benefits of export expan- sion for TFP growth can be realized simultaneously with the protection phase, but only if the incentives are tied to export performance. Such was not the case in Turkey, nor was there such an intention on the part of the policymakers. But the question of whether a period of protected import substitution, with associated negative effect on TFP growth, is worth the costs is not so easily answered. The crucial policy issue is one of timing. How long must one wait for an infant to mature? And is it possible to devise a policy mix that hastens the maturation process by tying policy incentives to performance (especially when it comes to exports)? The results we have presented have raised as many questions as they have answered. At this stage in productivity research, such a state of affairs is probably desirable. There is a real need to coordinate research at the micro and aggregate levels. The sort of stylized facts we have been con- sidering must be tested against work at the micro level to see if they make sense. Similarly, the micro work must be tested against comparative data at more aggregate levels to see what kind of generalizations are reason- able. Also unresolved are questions about the interdependence of different policies. A development strategy implies a coordinated effort to devise a consistent set of policies covering many areas. By definition, such a strategy affects a large part of the total economic activity in a country. The existence of linkages and externalities implies that it will be difficult if not impossible to consider the effect of such strategies in a partial equilibrium framework. The work that has been done indicates that studying how different development strategies and TFP growth are related is important, if not crucial, to gaining an understanding of what constitutes a successful development strategy. 25. See also chapter 6 and Kubo and Robinson (1984 ), who present data on such phasing in other countries, and Balassa (1979b). PART IV Developn1ent Strategy THROUGHOUT THIS BOOK, we have used a variety of analytical tech- niques to investigate the forms that structural transformation has taken in different countries. In part I, we studied a large group of semi-industrial countries by applying the techniques of growth accounting, augmented by a simple model of stuctural change. In part II, we probed the experience of industrialization in nine economies in more detail, using an input-output framework to make comparisons among different policy episodes. These efforts have led to a better understanding of the causes of industrialization and the identification of characteristic sequences of structural change. Up to now, the causal relation between policy and performance has been inferred from each economy's trade regimes and capital inflows, which help to determine the exogenous elements in input-output models. While this procedure serves to establish quantitative links between various aspects of structural change, it leaves many questions about the effects of policy on performance unanswered. For this purpose, we need to know not only the historical results of a particular set of policies but also the probable consequences of varying these policies. In other words, we need to establish the essential links between specific policy instruments and the other variables in the model. Chapter 11 is an exploratory attempt to use general equilibrium model- ing for this purpose. We shall draw on experimental applications of computable general equilibrium models for two countries in our sample- Korea and Turkey-to develop some quantifiable relations between de- velopment policy and economic performance. Although the general equilibrium approach requires some specification of all commodity and factor markets, we shall concentrate on the external aspects that have received primary emphasis in earlier chapters, namely trade policy and capital inflow. General equilibrium modeling of long-run growth is still in an ex- perimental stage. Our results are best interpreted as quantitative illustra- tions of phenomena that can otherwise be described only in general terms. 309 310 DEVELOPMENT STRATEGY Chapter 11 thus serves two purposes: to determine how the conclusions from the fixed coefficient models of earlier chapters are modified by the introduction of price effects and to characterize development strategies in terms of changing relative prices as well as quantities. Chapter 12 explores the implications of our findings for the choice of a development strategy and highlights some unanswered questions for further research. 11 Alternative Routes to Development HOLLIS CHENERY JEFFREY LEWIS JAIME DE MELO SHERMAN ROBINSON UP To Now, our analysis has focused on real activity in the economy and its implications for aggregate growth, factor accumulation, resource allocation, productivity growth, and changes in the structure of produc- tion and of demand. The role of market mechanisms and of relative prices in determining resource allocation and structural change has been left in the background. Yet in mixed economies, the policy instruments that are designed to promote development work through markets and prices. In this chapter, therefore, we explicitly consider how market mechanisms and relative prices affect industrialization under different development strategies. Similarly, the models we have used up to now to provide the framework for the analysis have focused on real variables and neglected the role of prices. In this chapter, we turn to the Walrasian model, in which market- clearing prices achieve equilibrium in a set of interdependent commodity and factor markets. Specifically, we use a computable general equilibrium (CGE) model that simulates the operation of a market economy and into which price incentive policies such as taxes, subsidies, and tariffs are explicitly incorporated. With this framework, we can sort out some of the main causal mechanisms that operate through changes in relative prices and that determine the effects of different policy choices. Our approach is to use a CGE model of a single country as a simulation laboratory for doing controlled experiments designed to explore different development strategies. Chapter 3 used a model of several representative or archetypal economies that was based entirely on comparative data. In this chapter, we start instead with data for a particular country, Korea, in 1963. At that time, Korea had just completed a period of growth based primarily on import substitution and was poised for a major shift in development strategy. In many ways, its economic structure at that time was typical of those of other semi-industrial countries setting out on a path of rapid industrialization. Where Korea differs from the average, we have adjusted the data to obtain a more typical economy. We are interested in 311 312 DEVELOPMENT STRATEGY creating a stylized version of Korea for the purpose of comparative analy- sis, not in analyzing the strategic choices available to Korea in 1963. 1 The first section defines three development strategies that differ mainly in their trade policies and that span the range of strategies actually fol- lowed in semi-industrial countries. We also consider how differences in the external environment-such as the nature of export markets and access to foreign capital-can affect the success of a given strategy choice. The second section describes the theoretical structure of the dynamic CGE model, and the third section analyzes the macroeconomic features of some model experiments designed to isolate the chief mechanisms at work. In the fourth section, we provide a more detailed analysis of the experimental results at the sectoral level, focusing on the role of changes in commodity and factor prices and in the exchange rate. Three Development Strategies We have seen in earlier chapters that even in those economies in which manufactured exports burgeoned following the shift to an outward- oriented development strategy, rapid industrialization led to increasing demands for imports of intermediate and capital goods. Only late in the process did the industrial sector become a net contributor of foreign exchange. In an industrializing economy, the balance of payments pres- sure arising from increases in import demand can be met in three ways: by import substitution, by export expansion, or by increased foreign borrow- ing. Either the demand for imports must be limited or the supply of foreign exchange must be increased. As shown in chapter 6, different trade strategies can be classified according to the relative importance they give to each of these components. We define three options that cover a spectrum wide enough to encom- pass the experience of most semi-industrial countries. These three options provide the starting point for the experiments with the CGE model. In all cases, we assume an economy that has achieved a significant industrial base and that is not rich in natural resources-and so cannot generate ample supplies of foreign exchange through primary exports. The first option is the strategy of export expansion, which is illustrated in an extreme form by the experience of Korea after 1963. Starting from a situation in which production for the domestic market had been strongly favored, Korea implemented a strategy that called for, first, reduced protection to imports; second, real devaluation to provide incentives to shift resources toward the tradable sectors; and third, elimination of incentives with a bias against exporting. Among the economies that have pursued such a strategy for a shorter or longer time, the most success- ful-in addition to Korea-include Malaysia, Singapore, and Taiwan. 1. For an analysis of the Korean case, see Kim and Roemer (1979). Chapters 4, 6, and 7 provide comparative data that include Korea. ALTERNATIVE ROUTES TO DEVELOPMENT 313 Common features of their experience are export expansion well in excess of GNP growth, substantial reduction of the bias in incentives against exports, and sufficient foreign capital inflow to permit a sustained period of trade liberalization. The second option is the strategy of import substitution, the elements of which are also taken from the experience of countries that have followed it with some success-Mexico and Turkey in our sample. Its three essential characteristics are a limitation on imports through both tariff protection and foreign exchange rationing; maintenance of an overvalued real ex- change rate, which exacerbates the bias in incentives against exporting; and a relatively low foreign capital inflow, the result primarily of credit- worthiness constraints arising from the low level of exports. These two options represent extreme cases. In terms of policy choice, they are mutually exclusive. The protectionist policies supporting the import substitution strategy necessarily generate a bias against sales abroad and in favor of the domestic market; this applies to nontradables as well as tradables. Slower export growth also leads to less foreign borrowing, and this leads to still more limitations on imports. Between the two extremes of export expansion and import substitution, a third option, a balanced strategy, combines elements of both. This alternative calls for more equal adjustments in the three components and a phasing of capital inflows. Import demand is limited through exchange rate policy rather than through tariff protection and foreign exchange rationing. The net effect is less bias against exporting and hence more exports than in the import substitution strategy. Since removing the bias against exports generally takes time, the strategy leads to more foreign borrowing in the early periods to finance more imports and so requires less devaluation of the real exchange rate. The components of the strategy include a reduction in import growth as a result of devaluation, but less than in the second option; an increase in exports, but less than in the first option; and higher foreign capital inflows in the early period. For coun- tries with access to foreign capital, this strategy is less demanding than either the first option, which calls for a rapid shift of resources toward exports, or the second option, which calls for severe constraints on im- ports. In the sample, Israel is the best example of a country pursuing such a strategy. 2 A Dynamic Computable General Equilibrium Model The computable general equilibrium model presented in this section provides a framework for making systematic comparisons among the 2. Although the choice among these three options is primarily a domestic policy matter, external factors may impinge on a country's ability to pursue a given strategy successfully. For example, export performance is responsive to economic conditions and policies in the developed countries, and access to foreign borrowing is also affected by world economic conditions. 314 DEVELOPMENT STRATEGY three strategies. Given our emphasis on alternative trade strategies, the model focuses on the markets for tradable commodities and on the incen- tives facing domestic producers and demanders of imports. Moreover, since our concern is with the effect of alternative development strategies on growth and structural change in the long run, the model is simulated for a twenty-year period (in five four-year intervals). 3 The long-run CGE model incorporates the market mechanisms through which domestic policy choices affect incentives, and it endogenizes the supply and demand reac- tions of domestic economic actors to such policies. Because the model is for a single economy, external conditions are reflected in exogenous variables. In the development literature, CGE models trace their lineage back to the multisector input-output models widely applied to problems of planning in developing countries in the 1960s. 4 While firmly based on the founda- tion of Walrasian general equilibrium theory, CGE models can also be seen as a logical culmination of a trend in the literature on planning models to add more and more substitutability and nonlinearity to the basic input- output model. 5 The models tend to be highly nonlinear-to have neoclas- sical production and expenditure functions-and to incorporate a variety of substitution possibilities in production, demand, and trade. CGE models applied to developed countries have generally stayed rel- atively close to the Walrasian paradigm. 6 In applications to developing countries, however, most researchers have introduced certain structuralist features into CGE models to capture the stylized facts characterizing these countries. Our model is very much in this tradition; it starts from a family of models developed by Dervis, de Melo, and Robinson to explore ques- tions of foreign trade policy in semi-industrial countries characterized by many structural rigidities. 7 The model is presented in three stages. First, we outline the model structure, distinguishing between the static part, during which an equilibrium is achieved, and the dynamic part, which updates exogenous variables and parameters. A description of markets, agents in 3. The model is not designed to explore the short-run problems of making a transition to a new development strategy. Such issues of "structural adjustment"-in the terminology of the World Bank-are better addressed with an annual model designed to track the adjustment process in more detail. 4. The model of the Norwegian economy developed by Johansen (1960) was the first empirical implementation of a general equilibrium model in a developed country. 5. There are also theoretical similarities between CGE models that simulate a multisector market equilibrium and planning models, either linear or nonlinear, that specify an explicit objective function within the framework of a programming model. See Ginsburgh and Robinson (1984) for a discussion of the relationship. 6. For a survey of CGE models focusing on issues of tax policy and international trade in developed countries, see Shaven and Whalley (1984). 7. See Dervis, de Melo, and Robinson (1982) for a detailed discussion of the structure and theoretical properties of CGE models in general and of models applied to problems of foreign trade in particular. See also Robinson (1986). ALTERNATIVE ROUTES TO DEVELOPMENT 315 the markets, and functional forms governing agents' behavior follows. Second, we outline the adjustment mechanisms that operate under each of the three development strategies defined above. Third, we describe the dynamic processes that drive the model forward in time. Structure of the Model The dynamic model consists of two parts. First, there is a static CGE model which solves for a one-year equilibrium. In this model, a set of markets for factors, commodities, and foreign exchange is assumed to clear subject to a variety of structural rigidities and to choices of exoge- nous variables, including policy parameters. Given these constraints, the static equilibrium represents an optimum for producers and consumers. Second, intertemporallinkage equations update exogenous variables and parameters that are dependent on policy choices and specify cumulative dynamic processes such as factor accumulation and productivity growth. The intertemporal equations provide all exogenous variables needed for the next period (four years later) by the CGE model, which is then solved for a new equilibrium. The model is thus solved forward in a dynamically recursive fashion, with each static solution depending only on current and past variables. The model does not incorporate any behavioral role for future expectations, and in its present form it cannot be used to explore issues of dynamic optimality except through sensitivity analysis. Table 11-1 schematically organizes the main features of the model around blocks of equations and equilibrium conditions. The first two columns describe the overall structure of the within-period CGE model, while the third column summarizes the cumulative processes incorporated into the dynamic part. The equilibrium conditions in the CGE model include a supply-demand balance in three different types of market: labor, commodities, and foreign exchange. A fourth macroeconomic equilibrium condition is a balance between investment and savings-the macro "clo- sure" of the model. A detailed description of the mathematical equations of the CGE model, supplementing the briefer discussion here, will be found in appendix A to this chapter. The CGE model simulates the working of a market economy. In each period, it solves for wages, prices, and an exchange rate (or import premium rate) that clear the markets for labor, commodities, and foreign exchange. The model is Walrasian in that only relative prices matter. The numeraire against which all relative prices are measured is defined as an index of domestic prices. 8 The model also satisfies Walras's law so that, by construction, there cannot be a situation of aggregate excess demand or supply. Thus the model cannot address macro issues such as the role of 8. This choice is especially important in interpreting the role of the exchange rate, which will be discussed in more detail below. 316 DEVELOPMENT STRATEGY Table 11-1. Schematic Outline of the Static and Dynamic CGE Models Static model Dynamic model: Economic Principal Structural cumulative relations relations features processes Factor markets Labor Labor demand Segmented rural- Labor force equations urban labor growth markets Capital Marginal Fixed sectoral Capital stock product capital stocks growth equations Product markets Production Production Productivity functions growth Demand Expenditure Composition functions changes Foreign trade Exports Export Segmented World market supply domestic and trends functions export markets Imports Trade Imperfect Induced import aggregation substitutability substitution functions Trade balance Exchange rate Foreign exchange Sequence of or premium rationing, exogenous capital inflows rate inflow Macroeconomic balance Savings-investment Domestic Trends in savings rates savings rates External capital Endogenous Fixed exchange foreign rate capital inflows - Not applicable. inflation or Keynesian unemployment. Given its long-run focus, it is appropriate to ignore such cyclical effects. 9 Except for the structuralist features listed in table 11-1, the model is very neoclassical in spirit. Sectoral production is given by mixed two-level constant elasticity of substitution (CES) and linear functions. Intermediate inputs are required according to fixed input-output coefficients, aggregate labor and capital are combined to create value added according to a CES function, aggregate labor is a CES aggregation of labor of different types, 9. There are CGE models that have been designed to explore issues of inflation and unemployment, with some strain to the Walrasian paradigm. SeeJ. D. Lewis (1986) for such a model and for a discussion of the modeling issues raised. ALTERNATIVE ROUTES TO DEVELOPMENT 317 and the aggregate capital used in each sector is a linear aggregation of capital goods from different sectors. Sectors are assumed to maximize profits, and labor demand functions come from the first order conditions equating the wage with the marginal revenue product of labor of each category. The labor market is segmented, with four distinct categories of labor: agricultural, unskilled, skilled (in the industrial sector), and service- oriented. Migration from the agricultural sector is specified exogenously over time; there is no mobility within periods. There is full employment of aggregate labor in each period, with the sectoral allocation of labor by different categories determined endogenously. 10 Sectoral capital stocks are fixed within periods; they change over time given aggregate growth of the capital stock and the sectoral allocation of investment. Investment in the industrial sectors is allocated endogenously to make sectoral rental rates approximately equal by the terminal year (although the rates differ across sectors in the intervening years). Expenditure functions for demanders arise from Cobb-Douglas utility functions, which yield constant expenditure shares and unitary income elasticities of demand within periods. To capture the impact of Engel's law, exogenous trends are imposed on the expenditure shares dynami- cally. Thus, income elasticities all equal one within periods but differ from one over time. The model determines the flow of funds to all economic agents, including wage earners, recipients of capital income, and govern- ment (whose income consists of tax revenue). The supply of exports by sector is a function of the ratio of the price in domestic currency of exports (determined by the world price, the exchange rate, and any subsidies) to the price of output sold in the domestic market. This treatment partially segments the export and domestic markets. Prices in the two markets are linked but need not be identical. Imports and domestic products are assumed to be imperfect substitutes-an assump- tion widely used in CGE models of trade. Imports and domestic goods are combined according to a CES trade aggregation function, with consumers demanding the resulting composite good. 11 The trade substitution elastic- ity determines the extent to which import shares adjust in response to changes in relative prices. For both exports and imports, the world price in dollars is assumed to be constant-the small country assumption. Adjustment Mechanisms under Alternative Strategies There are three mechanisms by which the CGE model can achieve equilibrium in the balance of trade under different assumptions about the 10. The model thus has no surplus labor or underemployment. It would be feasible to specify a model with surplus labor (and a fixed real wage). See de Melo and Robinson (1982), who explore the effect of trade policy on income distribution in such a model. 11. See Dervis, de Melo, and Robinson (1982) and de Melo and Robinson (1985) for a discussion of the implications of this treatment. Armington (1969) used this specification in (Note continues on the following page.) 318 DEVELOPMENT STRATEGY availability of foreign borrowing. These alternative mechanisms are used in different experiments depending on the issues being addressed. Two of them assume a fixed foreign capital inflow, reflected in an exogenous value for the balance of trade, so that adjustment depends primarily on relative price effects. For the first mechanism, endogenous variation in the real exchange rate provides the equilibrating mechanism. The real exchange rate is defined as the relative price of tradables and nontradables. Since we use an index of domestic prices as numeraire, variations in the nominal exchange rate in the model directly affect the ratio of the price-in domestic currency-of imports and exports to the price of domestic sales and so represent a change in the real exchange rate. For example, a devaluation raises the domestic price of imports and exports relative to domestic sales and so encourages exports and import substitution. The model determines the equilibrium real exchange rate by manipulating the nominal rate relative to the fixed numeraire index of domestic prices. For the second mechanism, import rationing provides the equilibrating mechanism. We assume that the rationing is efficient in that the marginal value of a dollar is the same across sectors: this is equivalent to assuming that import licenses can be sold in an open market. 12 In effect, a uniform import premium is imposed on top of any official tariffs; the premium rate is determined endogenously to equate import demand with the available aggregate supply, given exports and the exogenous foreign capital inflow. The result is the same as a devaluation applied only to imports, and the scheme yields a bias in incentives against exporting since it operates like a tariff. For the third mechanism, we assume that the exchange rate is fixed and the balance of trade is endogenous, so that foreign capital inflow adjusts. Given the numeraire, this specification effectively fixes the real exchange rate. With a fixed relative price, the model achieves equilibrium through a quantity adjustment mechanism-in this case, through changes in foreign capital inflow. Because of the dual role of foreign capital inflow in permit- ting both increased investment and imports (as in two-gap models), exter- nal capital plays an important macroeconomic role as well. The balance between savings and investment in the model is achieved by setting total investment equal to the sum of domestic and foreign savings .13 estimating import demand functions: the trade aggregation function is sometimes called an Armington function. 12. This treatment is a proxy for the wide range of second-best quantitative and licensing restrictions that are often imposed in developing, as well as developed, economies. For a discussion of modeling alternative rationing schemes, see Dervis, de Melo, and Robinson (1982). 13. This specification is termed neoclassical closure. Much of the debate on the appropri- ate macro closure revolves around issues of adjustments in the short to medium run in models that allow unemployment by, for example, assuming a fixed wage. Given our long-run focus and full-employment assumption, the neoclassical closure is an obvious choice. See Rattso (1982), Lysy (1983), and Robinson (1986) for surveys of the issues involved. ALTERNATIVE ROUTES TO DEVELOPMENT 319 Domestic savings is modeled as a rising function of real GDP, so that a higher growth rate is associated with an increase in the domestic savings effort. The specification of foreign savings reflects empirical evidence that the propensity to save out of foreign capital inflows is less than one. In the CGE model, 40 percent of the inflow goes directly into savings. The remainder is funneled to capitalists, who in turn save a fraction; the remainder is consumed or taxed. The net effect is that the overall marginal savings rate from external resources is about 0.6, a value similar to that estimated by Chenery and Syrquin (1975). Cumulative Dynamic Processes The dynamic equations capture cumulative processes that drive the CGE model forward in time. These processes reflect three different types of forces: exogenous trends, policy choices, and past history incorporating solutions of the model for previous periods. The variables and parameters that are updated dynamically in the model can be classified into three categories: • Updated by exogenous trends Aggregate labor force growth Sectoral total factor productivity growth Input-output coefficients Government consumption shares Private consumption shares • Updated by policy choices Tariff rates Foreign capital inflows Exchange rate • Updated by economic behavior Sectoral investment allocation Sectoral capital stocks Labor force allocation by category Sectoral export shares Sectoral import ratios Variables updated by exogenous trends follow the same dynamic path in all experiments. The aggregate labor force is assumed to grow 3 percent annually. Total factor productivity growth rates are drawn from the results discussed in chapter 10. The other parameters-input-output coef- ficients and expenditure shares-are interpolated in each period between exogenously specified initial and terminal values. Variables updated by policy choices are the chief policy instruments of the three trade strategies. While the CGE model includes a large number of policy instruments, the variations in these three define the three trade strategies. We vary different combinations of these instruments and solve others endogenously to meet certain targets. 320 DEVELOPMENT STRATEGY Variables updated by economic behavior have values that are generated as part of the history of the model. The sectoral allocation of investment is assumed to adjust over time to equate rental rates in the industrial sectors by the terminal year. 14 Sectoral capital stocks in any year depend on investment allocation and the depreciation rate. Whereas aggregate growth of the labor force is exogenous, the composition by category is determined by a combination of two factors: exogenously specified migra- tion of agricultural labor to one of the three categories of urban labor and migration among categories in response to changing real wage differen- tials. Over time, real wage differentials will induce an offsetting migration response. Within a period, export supply shares and import demand shares re- spond to changes in relative prices. In some experiments, we also add a trend component to these shares. For exports, the assumption is that changes in world demand or increased market penetration lead to in- creased exports without any change in relative prices in the domestic market. With only supply behavior included in the model, such a specifica- tion is necessary to capture actual export behavior under the export expansion trade strategy .15 The dynamic specification of import demand also varies according to the trade strategy pursued. Under an import substitution strategy or a balanced strategy, it is assumed that the long-run trade substitution elas- ticity is higher than the short-run elasticity. 16 The effect is to allow for "successful" import substitution: the efficient ratio of imports to domestic supply for a given set of relative prices will gradually fall. Demanders can replace imports with domestic production more easily between periods than within periods and so avoid experiencing diminishing returns to further import substitution. Macroeconomics of Alternative Strategies In this section we describe the simulations of the three strategies in general terms and discuss their policy differences. All simulations begin from a common starting point in the base year, after which policy choices and the paths of exogenous variables differ. The purpose is to measure the effect of each strategy under different conditions, to illustrate the effects of various policy packages, and to examine the strains that emerge from the interactions among groups of policies. 14. In the base year, by construction, the economy is assumed to start from an initial solution with equal rental rates across the industrial sectors. 15. We chose not to include both supply and demand functions for exports because we did not wish to endogenize the effects of international terms of trade. Models with such behavior are described in Dervis, de Melo, and Robinson (1982). 16. There is a long-run envelope curve for each level of composite good (analogous to isoquants in production functions) with the short-run curves applicable to each period tangent to the long-run curve at a point. ALTERNATIVE ROUTES TO DEVELOPMENT 321 Table 11-2. Representative Trade Strategies Import Export substitution Balanced promotion Components (IS) (B) (EP) Trade policy Sectoral trends Import reduction Import reduction Export penetration/ import liberalization Main policy Tariffs/import Real exchange Real exchange instruments rationing rate rate Trade bias Inward Neutral Outward Trade shares Low/falling Constant Rising Borrowing policy Limited by Maintain Sustain trade low exports neutrality liberalization Indirect effects Productivity growth Low/intermediate Intermediate High Capital goods High cost World prices World prices Examples (1960s)' Turkey Israel Korea Mexico Thailand Taiwan Philippines Tunisia Singapore Colombia Greece Malaysia Argentina Brazil a. Examples are taken from chapter 4. Economies in italics are those analyzed in chapters 6 and 7. The main features of each strategy are compared in table 11-2, and the influence of these factors on growth is summarized in table 11-3. 17 Each strategy is characterized by typical exogenous elements, the policy rules for maintaining external balance, and other policy choices. A successful export promotion (EP) strategy is reflected in an average export growth of 14 percent annually. 18 Given the pattern and cumulative level of capital inflow, the real exchange rate varies to achieve external balance. Over a twenty-year period, this strategy results in substantial trade liberalization, shown by the rise in the share of imports in GOP from 14 to 24 percent. The capital inflow required to finance import require- ments declines steadily from an initial level of 8 percent of GOP to less than 1 percent in the terminal year. GDP growth averages 6.5 percent, of which 1. 7 percent is the result of assumed growth of total factor productivityY 17. The numbers cited in table 11-3 and in the description of strategies are for a moderate level of foreign capital inflow for each strategy. The implications of different cumulative inflows are discussed more fully below. 18. Of this 14 percent, about 2.5 percent is attributable to market penetration and other effects not reflected in relative prices. 19. While this GOP growth is representative of the outward-oriented economies identified in table 4-4, it is less than the growth rate of 8 to 9 percent observed in Korea and Taiwan during the 1960s. 322 DEVELOPMENT STRATEGY Table 11-3. Macroeconomic Indicators of Alternative Trade Strategies Import Export substitution Balanced promotion Indicator (IS-2) (B-2) (EP-2) Average ratio of capital inflow to GOP 4.5 4.4 4.1 Ratio of imports to GOP, terminal period 10.7 18.1 24.0 Incremental capital-output ratio, terminal period 3.26 3.02 2.95 Export growth rate 7.9 10.3 14.1 Import growth rate 4.5 6.4 9.3 GOP growth rate 5.7 6.2 6.5 Note: In the abbreviations IS-2, B-2, and EP-2, the number "2" refers to one of four levels of capital inflow, defined in table 11-4 as 1,900 million 1964 dollars. All figures are percentages except for the incremental capital-output ratio. The import substitution (Is) (inward-oriented) strategy maintains exter- nal balance by rationing foreign exchange through an import premium that provides incentives for import substitution. By increasing the cost of imported inputs, this strategy also makes exports less profitable. To maintain this policy over the twenty-year period, the import premium rises rapidly. This steadily increases the incentive bias in favor of import substitution over export promotion. The implications of this bias for relative prices are discussed more fully in the next section. The macroeconomic effect of the IS strategy is a considerable closing of the economy. Imports fall from 14 percent of GDP to under 11 percent. The incremental capital-output ratio is 10 percent higher in the terminal year than it is in the EP strategy, and average GDP growth declines by nearly a percentage point. The balanced (B) strategy contains elements of both the EP and IS strategies. The balanced strategy eliminates the inefficiency and price distortions associated with protection in the IS strategy without calling for the rapid growth of exports of the EP strategy. This is accomplished by combining successful import substitution with devaluation of the real exchange rate, which encourages exports and avoids the anti-export bias inherent in the IS strategy. Elimination of this bias increases export growth to 10 percent, which permits substantially more rapid growth in imports. 20 GDP growth over the period is 6.2 percent, only slightly less than growth with the EP strategy. In summary, the balanced strategy combines efficient import substitu- tion with moderate export expansion. In some of the countries pursuing 20. The balanced strategy simulations do not include the export growth through market penetration that is part of the export promotion simulations. ALTERNATIVE ROUTES TO DEVELOPMENT 323 this strategy, access to more external borrowing and the resulting increase in investment can offset the gains from further specialization that occur in the export-led strategy. The choice between the two strategies then de- pends on the preferences of a country and the feasibility of implementing the necessary trade and borrowing policies. In considering the three alternative strategies, it is important to empha- size that the sectoral rates of total factor productivity growth used in each strategy have been held constant. This assumption conflicts with the conclusions of chapters 2, 8, and 9, in which it was argued that the import restrictions, price distortions, and loss of specialization inherent in the aggressive pursuit of an import substitution strategy result in significantly lower rates of total factor productivity growth. We have assumed similar productivity growth to facilitate comparisons across strategies; however, we explore in other experiments how economic performance is affected by assuming lower factor productivity growth rates in the IS strategy. Elements of Policy To clarify the effects of the main elements of external policy-the trade bias, degree of openness, and capital inflow-it is useful to vary them separately while holding other aspects constant. This procedure is illus- trated in table 11-4, which gives solutions for each strategy with the cumulative capital inflow held constant at four different levels. This set of simulations is used to identify the effects of individual policy changes when all else remains unchanged. TRADE BIAS. The effect on growth of eliminating the trade bias (holding capital inflow constant) is shown in table 11-4 by comparing the balanced strategy (s) with the inward-oriented strategy (Is) in column 1. The import premium as an endogenous equilibrating variable is replaced by a flexible exchange rate, so that external balance is achieved both by increasing exports and by reducing imports in proportion to their respective elastici- ties. The exchange rate devaluation needed is only 35 percent instead of the 170 percent premium needed in IS. More than half the adjustment in the B strategy takes place through expanding exports. This shift to a neutral trade policy raises GDP growth from 5.2 to 5.9 percent. In addition to the small static allocative gains from more efficient trade, the decline in the relative cost of investment goods leads to a significant rise in the real rate of investment and contributes to GDP growth. 21 The marginal benefits of eliminating the trade bias decline as the capital inflow rises because the more plentiful supply of foreign exchange reduces the need for import substitution. Whereas the elimination of trade bias at 21. This phenomenon will be analyzed in more detail in the section below on prices, incentives, and structural change. 324 DEVELOPMENT STRATEGY Table 11-4. Trade Strategies, Capita/Inflows, and Growth Level of cumulative capital inflow (millions of 1964 dollars) 900 1,900 3,000 4,500 Strategy (1) (2) (3) (4) Import substitution (Is) GOP growth rate (percent) 5.2 5.7 6.2 Elasticity of imports to GOP 0.65 0.79 0.97 Ratio of imports to GOP, terminal period (percent) 9.4 10.7 13.4 Ratio of total debt to GOP, terminal period (percent) 0.31 0.54 0.78 Ratio of average capital inflow to average GOP (percent) 2.9 4.5 6.3 Balanced (B) GOP growth rate (percent) 5.9 6.2 6.6 Elasticity of imports to GOP 1.07 1.05 1.17 Ratio of imports to GOP, terminal period (percent) 17.6 18.1 17.7 Ratio of total debt to GOP, terminal period (percent) 0.27 0.50 1.10 Ratio of average capital inflow to average GOP (percent) 2.7 4.4 8.4 Export promotion (EP) GOP growth rate (percent) 6.3 6.5 6.9 Elasticity of imports to GOP 1.49 1.43 1.50 Ratio of imports to GOP, terminal period (percent) 23.9 24.0 24.0 Ratio of total debt to GOP, terminal period (percent) 0.26 0.47 1.02 Ratio of average capital inflow to average GOP (percent) 2.6 4.1 8.4 - Figures are not available because simulations with that cumulative capital inflow were not undertaken. Note: Levels of capital inflow are identified by column numbers. low levels of capital inflow (rs-1 versus B-1, where "1" refers to column 1 in table 11-4) increases GDP growth by 0. 7 percentage points, doing so at moderate levels of inflow (rs-2 versus B-2) increases GDP growth by only 0.5 percentage points. With capital inflows as high as in B-4, the premium falls to zero, eliminating the bias between the rs and B strategies. OPENNESS. An increase in exports and imports beyond the level spec- ified by the balanced strategy leads to greater specialization and lower ALTERNATIVE ROUTES TO DEVELOPMENT 325 resource costs for tradable goods as a whole. This result is the essence of trade liberalization (see chapter 6). The increase in export growth from 10 percent in B-2 to 14 percent in EP-2 leads to cumulative levels of both imports and exports that are 50 percent higher over twenty years. The aggregate effect of this greater specialization is to increase the growth rate by 0.4 percentage points, which is about half the effect of eliminating the trade bias at low levels of capital inflow. Although the increase in openness (as measured by the ratio of imports to GDP) from B-1 to EP-1 is similar to the increase from Is-1 to B-1, the reduction in the resource cost for tradable goods is considerably less. CAPITAL INFLOW. External resources perform two distinct functions in a development strategy: they add to the level of investment and they supply additional imports. In modeling the first function, we have assumed that 60 to 70 percent of the capital inflow represents a net addition to investment, in line with econometric estimates of this effect. 22 On the trade side, capital inflows are not a perfect substitute for exports since they do not have the cumulative benefits of market penetration and eventually need to be repaid. In addition, slower export growth limits the amount of external borrowing that can be sustained over the long run. 23 Borrowing does alleviate the need to undertake types of import substitu- tion or export subsidization that may prove inefficient once export per- formance improves. It also allows a lower real exchange rate in the early period of import substitution. The examples in table 11-4 suggest that an increase in capital inflow may add half a percentage point to the long-run growth rate within the feasible limit of the debt-export ratio. This increased growth from higher borrowing capability adds to the direct benefits of shifting from an in- ward-oriented to an outward-oriented strategy. Dynamic Growth Paths We now turn from a consideration of the cumulative effects of alterna- tive strategies to an examination of the dynamic aspects of growth. How do trade bias, openness, and capital inflows interact to determine the pattern of growth under each strategy? Table 11-5 and figures 11-1, 11-2, and 11-3 show the dynamic effect of changes in these policy components. The effects of trade policy on the openness of the economy (as measured by the ratio of imports to GDP) and on the growth of GDP are shown in 22. The proportion varies somewhat depending on the level of real GOP since the domestic savings effort is a rising function of national income. 23. In the model simulations, we approximate the borrowing constraint by restricting cumulative foreign borrowing to four times the level of exports. Among major borrowers, this was approximately the ratio of Argentina, Mexico, and the Phillipines in 1983. It was exceeded by Brazil (5.3) and Turkey (4.5). In contrast, the outward-oriented countries (for example, Korea, Malaysia, Thailand, and Yugoslavia) had ratios ranging from 0.9 to 2.0. 326 DEVELOPMENT STRATEGY Table 11-5. Dynamics of Alternative Strategies Import Export substitution Balanced promotion (Is-2) (B-2) (EP-2) Period or Base variable Time 0' Time2 Time 5 Time2 TimeS Time2 Time 5 Variable-internal aspects Percent of GDP Domestic savings 8.6 13.8 24.3 13.5 24.7 13.5 26.3 Foreign savings 5.1 3.5 1.9 4.2 2.3 4.4 0.6 Investment 13.7 17.3 26.2 17.7 26.9 17.9 26.9 Real investment 13.7 17.2 23.6 17.5 26.4 18.0 27.1 Growth rateb Primary output 3.5 2.9 3.3 2.9 3.5 3.2 Manufacturing output 7.4 9.1 8.9 11.6 9.1 11.5 GDP 5.6 6.4 5.7 7.3 6.0 7.9 Incremental capital- GDP ratio 2.83 3.26 2.78 3.02 2.65 2.95 Variable-external aspects Percent of GDP Import share 14.1 11.6 10.7 15.1 18.1 17.3 24.0 Export share 5.7 6.3 8.1 8.7 15.1 10.5 23.3 Trade balance 8.4 5.4 2.6 6.5 3.0 6.8 0.8 Growth rateb Imports 3.5 6.8 5.0 9.8 8.8 10.7 Exports 7.3 9.5 9.6 11.3 14.2 14.4 Exchange rate 1.00 1.00 1.00 1.19 1.22 1.00 1.00 Premium rate (percent) 0 35 118 0 0 0 0 Note: The cumulative trade balance in all runs is the same. a. Each twenty-year simulation is composed of five four-year "times." Time 0 is the base year; time 2 is eight years later; time 5 is the last year in the twenty-year simulation. b. Growth rates for time 2 are for the first eight years; growth rates for time 5 are for the last four years. figure 11-1. In the import substitution strategy (Is-2), the initial import ratio of 14 percent falls to 11 percent by the fourth four-year period. 24 This decline in openness is accompanied by lower GDP growth. The balanced strategy (B-2) is characterized by a steady rise in the import ratio, which reaches 18 percent by the end of the twenty-year span; in the export promotion strategy (EP-2), the import ratio reaches 24 percent. Figure 11-2 shows the changing import ratio, and the relative impor- tance of exports and of capital inflows in financing imports, for the three trade strategies. The moderate capital inflow of EP-2 is sufficient to main- tain a constant real exchange rate over the twenty years. 25 External capital 24. Recall that each twenty-year simulation with the model is composed of five four-year periods. Declines of this magnitude have been observed in Argentina, Brazil, Turkey, and some other countries following implementation of an inward-oriented policy. Such declines eventually end either because of the increasing cost of import substitution or because of a change in policy to offset the anti-export bias in selected sectors. 25. This result was achieved by construction to provide a base for comparisons. The model was run with a fixed exchange rate and endogenous capital inflow. Figure 11-1. Import Requirements under Alternative Strategies "" Ci 0.20 " ... 0 ... "' ... 0.15 0 0.. .§ ....... 0.10 0 .9 ... <1S ~ 0.05 0 500 1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,400 Per capita GDP (dollars) Figure 11-2. Financing of Imports- under Alternative Strategies 0.14 "" Ci " ... 0 0.12 ~ 0 <::;::: 0.10 .5 -;; ... ·o.. 0.08 "' u t:: 0.06 0.0 ... 'Qj 0 ....... 0.04 ....... 0 0 ·c 0.02 <1S ~ 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 Ratio of exports to GDP Note: See table 11-4 for key to simulations in figure. Dashed lines indicate ratios of investment to GDP. 327 328 DEVELOPMENT STRATEGY Figure 11-3. Financing of Investment under Alternative Strategies 0.22 p., 0.18 0 " ... 0 "' l:lO t:: 0.:14 ·:;: o:S "' t:: l:lO ·;:; 0.10 .... c8 ,_..., 0 0 0.06 ·c rn ~ 0.02 o~~-L~L-~-L~--~_L~--L-~~~L_-L~-- 0.02 0.06 0.10 0.14 0.18 0.22 0.26 0.30 Ratio of domestic savings to GDP Note: See table 11-4 for key to simulations in figure. Dashed lines indicate ratios of investment to GDP. is more concentrated in the earlier periods, during which it sustains higher imports and more rapid growth. 26 With the high capital inflow of EP-4, the importance of exports in GDP actually declines in the second period, as the large capital inflow results in a revaluation of the exchange rate and a corresponding decline in export growth. 27 The moderate inflow of B-2 is also distributed in a manner which maintains a fairly constant real exchange rate from the second period onward. 28 It is characteristic of this strategy to reduce dependence on external capital more slowly. Although the opening up of the economy takes place more slowly than in EP-2, the corresponding loss in aggregate growth is relatively small. With the initial increase in external dependence associated with the higher inflow of EP-4, dependence is actually lower in B-4 than in EP-4 in the second period. The export growth of EP-4 rapidly 26. Analysis of the optimal distribution of capital inflows (Chenery and MacEwan 1966) shows that this pattern persists under a wide variety of assumptions. 27. This initial decline in the contribution of exports under an export promotion strategy is in keeping with the empirical conclusions of chapter 7. 28. Although the CGE model does not determine optimal aid patterns, experiments with alternative paths for capital inflow showed very little improvement over the pattern in B-2. ALTERNATIVE ROUTES TO DEVELOPMENT 329 overtakes the balanced path, however, so that from the third period onward, the balanced path is more dependent on external inflows. After twenty years, dependence on external resources in B-4 remains unchanged. Table 11-5 compares the economy's performance during time 2, after the major dislocations associated with policy changes have occurred, to that during time 5, which reflects the cumulative effects of twenty years of growth and structural change. From this table, which supplements figures 11-2 and 11-3, we can distinguish the common features of all strategies as well as the main differences among them: • Acceleration of aggregate growth takes place in all strategies. GDP growth rates increase nearly 30 percent in the two neutral strategies (EP-2 and B-2). The increase is less with import substitution. • Accelerated growth has two main sources: a rise in the domestic savings rate for a given income level and increasing openness of the economy. Whereas the first effect is similar in all strategies, the second occurs only in the two neutral strategies. • Several factors act to offset accelerated growth. Higher domestic savings are partially offset by declining foreign savings (figure 11-3) and by rising incremental capital-output ratios. Greater openness is partly offset by rising import costs, particularly under the import substitution strategy. • The combined effects of these factors is to widen the range of growth rates, since the slower growth that characterizes the inward-oriented strategy is more affected by the adverse factors than the other strategies. For the runs reported in table 11-4, GDP growth rates toward the end of the period range from less than 6 percent to more than 8 percent, three times the range observed at the beginning. A summary of the effects on terminal GDP and terminal capital stocks of varying both the choice of strategy and the level of capital inflow is given in figure 11-4. The greater efficiency of the outward-oriented strategies is shown by the higher output-capital ratios of the EP and B curves compared with the IS curve. The relative slopes of these curves reveal the marginal productivity of external capital in each case. Since the inward-oriented strategy becomes increasingly inefficient at low import levels, the marginal productivity of external resources is correspondingly higher, as indicated by the steeper slope of the IS curve starting from the low inflow of IS-1. Productivity Growth and Trade Strategies Evidence from earlier chapters points to a strong positive association between the outward orientation of an economy and the growth rates of total factor productivity in the economy. In the characterizations of alternative strategies undertaken thus far, however, no distinction has been made in the sectoral rates of productivity growth according to the strategy pursued. If inward-oriented, import substitution strategies do 330 DEVELOPMENT STRATEGY Figure 11-4. Productivity of Capital under Alternative Strategies 4,400 .----,---------,-----,-----T"::::oo------, " EP-4 = 4,500 4,200 4,000 "' c ~ . 3,8oo 6 D 3,000 5,500 6,000 6,500 7,000 7,500 8,000 8,500 9,000 9,500 Capital stock, terminal time (millions of dollars) Note: See table 11-4 for key to simulations in figure. The points on the dashed lines indicate terminal-time GDP and capital stocks under different strategies for a given level of cumulative capital inflow (D). Lines further from the origin indicate higher capital inflow. The slopes of the solid lines indicate the marginal productiviry of capital under each strategy. imply lower rates of productivity growth, then the differences in growth performance by strategy will be larger than those described above. To provide some indication of how economic performance is affected by the lower productivity growth associated with inward-oriented strategies, table 11-6 compares IS-1 with a scenario (1s-1L) in which sectoral rates of productivity growth are reduced by half. All other policies are the same, so that differences in performance are attributable to the lower productivity growth. The reduction in TFP growth slows GDP growth by about 1 percent. 29 Nominal investment, which rises with GDP growth, remains lower as well. Incremental capital-output ratios are substantially higher in Is-lL, with even larger differentials observed in earlier periods. The slower growth does reduce import requirements, so that the premium rate required to achieve external balance is actually lower than in IS-1 ( 134 percent com- pared with 170 percent), although it remains quite high. The openness of the economy is largely unchanged, although export and import growth is slower since GDP growth has dropped. GDP growth accelerates less, in- 29. The TFP growth is assumed to apply only to value added. Changes in intermediate input coefficients are specified separately and are not altered in this experiment. ALTERNATIVE ROUTES TO DEVELOPMENT 331 Table 11-6. The Effect of Lower Productivity Growth Normal Low productivity productivity Variable (Is-1) (1s-1L) Percentage of GDP, terminal year Domestic savings 23.2 19.5 Foreign savings 1.0 1.3 Investment 24.2 20.8 Real investment 20.9 18.0 Imports 9.4 9.7 Exports 7.9 7.8 Growth rate Primary output 3.0 2.0 Manufacturing output 8.8 7.5 GDP 5.2 4.0 Imports 7.3 6.0 Exports 3.4 2.4 Other values Incremental capital- GDP ratio 3.14 3.38 Premium rate (percent) 170 134 Note: For a description of these experiments, see the text. Further discussion of these results can be found in appendix B. creasing from 3.5 percent in the first four-year period to 4.5 percent in the final four-year period. By incorporating the adverse impact of inward- oriented strategies on productivity growth, rs-1 Lis more representative of the experience of countries pursuing such strategies over a long time. Although not reported here, additional experiments were conducted to analyze the sensitivity of the results to such main elasticity assumptions as trade substitution elasticities and capital-labor substitution elasticities. 30 The principal finding is that the rs strategy is more adversely affected than the others by reductions in the substitution elasticities: when both trade substitution and capital-labor elasticities are reduced by 50 percent, the GDP growth acceleration disappears completely. In the two neutral strategies (EP and B), acceleration is reduced but not eliminated. As sub- stitution possibilities are reduced, the economy behaves more like the traditional two-gap model. Prices, Incentives, and Structural Change The macroeconomic policy choices discussed in the previous section work primarily through their influence on relative prices. Since sectoral 30. These sensitivity experiments are discussed in detail in appendix B to this chapter. 332 DEVELOPMENT STRATEGY factor allocation responds to price incentives, we can trace the mecha- nisms through which policy changes affect structural change and hence the growth of the economy. After a discussion of the structural features of the economy and the sectoral sources of growth, we focus on the relative price movements unds:rlying the alternative trade strategies. In particular, we examine the exchange rate regime and the bias it creates toward producing for export markets or for domestic import substitution, and the relative price of capital goods and its influence on real capital accumulation. Structural Change and Development Strategy Each strategy leads to a different economic structure in the terminal period, primarily because of the differing pattern of international trade. In particular, the degree of tradability of a sector-defined here in terms of the share of imports and exports in domestic supply-together with the responsiveness of output supply determine the distribution of adjustment across sectors. For example, more tradable sectors will benefit from a depreciation of the real exchange rate since both import substitution and export expansion will be stimulated. The columns of table 11-7 contain measures of sectoral trade depen- dence and supply responsiveness. Sectors with low value added ratios are more adversely affected than others by a rise in the price of intermediate goods. 31 The ratio of imported intermediates to total intermediates mea- sures the dependence of each producing sector on imported goods-the higher the ratio, the greater the impact of an increase in the prices of imported intermediate goods. Since capital stocks are fixed within a given period, the capital-labor ratio is a proxy for the supply elasticity of the sector. The trade substitution elasticity measures the substitutability of domestic for imported goods, which together with the composition of final demand will determine demand elasticities. 32 To summarize the pattern of structural change implicit in the three strategies, we decompose the growth of output using the sources-of- growth methodology developed in chapter 5. Sectoral output growth is decomposed into four components: domestic demand expansion, export expansion, import substitution, and changes in input-output coefficients. 33 Table 11-8 contains the sectoral growth decompositions for the full twenty-year period. (The three service sectors have been aggregated.) The 31. This is especially true in a model such as this one that assumes fixed input-output coefficients, since no price-responsive substitution between intermediates and value added is possible. Note, however, that there is substitution between imports and domestic goods within each cell of the input-output matrix. 32. The composition of final demand is important because different components of final demand have different elasticities of demand. Intermediate demand has a zero price elasticity, while consumption has an elasticity of one. 33. Since the changes in input-output coefficients are exogenous and identical in all strategies, the contribution of this component varies little, with aggregate differences ex- plained by variations in the composition of output. ALTERNATIVE ROUTES TO DEVELOPMENT 333 Table 11-7. Sectoral Trade Dependence and Elasticities Imported Gross Value inter- Export Import Capital- Trade output added mediate supply supply labor elastic- Sector structure ratio ratio ratio ratio ratio ity Primary 32.6 73.7 10.8 2.4 10.9 0.22 1.10 Food processing 9.4 10.8 8.2 3.7 6.4 0.21 1.30 Consumer goods 16.0 33.6 6.3 3.0 3.4 0.63 1.10 Intermediate goods 7.9 33.6 12.7 4.7 38.7 0.68 0.60 Machinery 3.0 42.4 20.3 2.6 64.5 0.49 0.50 Construction 5.6 33.3 16.2 2.0 0.0 0.12 0.33 Social overhead 4.7 51.3 15.4 7.5 1.1 2.19 0.33 Services 20.8 69.6 5.5 3.9 1.1 1.02 0.33 Total 100.0 53.2 10.0 3.3 9.9 0.51 - Not applicable. Note: Imported intermediate ratio = ratio of imported intermediates to total intermediates Export supply ratio = ratio of exports to domestic output Import supply ratio = ratio of imports to domestic supply (domestic supply = domestic output- exports) Trade elasticity = elasticity between imports and domestic goods Source: World Bank data. economywide results reveal the importance of trade strategy to the pattern of growth in the economy. With an export-led strategy, export expansion accounts for 27 percent of the overall expansion in output, while import substitution is negligible. In the more neutral balanced strategy, the con- tribution of export expansion falls to 13 percent, only half that of the export-led strategy but still substantially larger than the 6 percent con- tribution of import substitution. Finally, in the import substitution strategy, the import substitution contribution reaches 12 percent, sur- passing the export contribution of 8 percent. The sources of growth presented in table 11-8 are comparable with the empirical results presented in chapter 6 (see, for example, table 6-4). The typical strategies analyzed here avoid the extreme episodes observed for limited periods in certain economies; this reflects the focus on long-run structural change. The negligible contribution of import substitution to growth under the export promotion strategy differs from the experience of some outward-oriented economies and arises from the assumption of a degree of trade liberalization as part of the strategy. Effects of the Trade Incentive Bias One direct impact of alternative development policies is reflected in the incentives to domestic producers to sell products in domestic rather than foreign markets. The policies that influence these incentives include sec- toral policy instruments such as tariffs and subsidies as well as macroeco- nomic variables such as the real exchange rate. 334 DEVELOPMENT STRATEGY The upper graph of figure 11-5 shows the incentive to substitute for imports, represented by the ratio of import to domestic prices. 34 The lower graph portrays the bias of the trade regime, which corresponds to the ratio of import prices to export prices-in domestic currency-and summarizes the aggregate incentives for import substitution and for export. A higher bias value signals relatively greater incentives to produce import substi- tutes rather than exports. 35 The differences in strategy are apparent in the figure. With the EP strategy, the rapid growth in export earnings achieved through market penetration and increased market shares is sufficient to finance import needs without substantial import substitution. Price incentives further encourage increased exports as average tariffs are lowered from 17 to 11 percent by the end of the period. The successful penetration of export markets and the availability of adequate capital inflows obviates the need for increased export incentives that would require real devaluations of the exchange rate. 36 With the B strategy, an exclusive reliance on export expansion is re- placed by more neutral policies that increase exports as well as gradually reduce import dependence through successful import substitution. With lower export growth and the same inflow as with EP-2, external balance is achieved through devaluations of the real exchange rate. Since devaluation affects export and import prices uniformly, the bias of the trade regime remains unchanged. The real devaluation necessary is about 20-25 per- cent, and it occurs in the early portion of the twenty-year period. 37 The IS strategy assumes the same successful import substitution as in the B strategy. The real exchange rate is held constant, however, so that adjustment to the foreign exchange constraint occurs entirely through import reduction brought about by a rising import premium. Since this premium affects import prices but not export prices, the bias of the trade regime more than doubles as increasingly costly and inefficient import 34. Although the results shown are only for a moderate level of capital inflow, the pattern is largely unchanged for other levels of inflow as well. 35. A measure of the bias is given by B, defined as: M; PWM; (1 + TM; + PR)ER B=--- ---, E; PWE; ( 1 + TE,) ER where M and E are base-year weights, PWE and PWM are world prices, TE and TM are subsidies and tariffs, PR is the import premium, and ER is the exchange rate. In the current case, however, with no export subsidies and no change in world prices, the bias is determined as B = (1 + ATM + PR), where ATM is the appropriate average tariff rate. 36. This pattern is in fact characteristic of the East Asian superexporters. Westphal (1978) shows, for example, that the real exchange rate in Korea remained constant throughout a period of substantial penetration of foreign markets. 37. This policy is representative of the crawling peg policies pursued by some Latin American and other countries starting in the mid-1960s; the real exchange rate was held constant, with periodic discrete devaluations to increase export competitiveness and curtail imports. ALTERNATIVE ROUTES TO DEVELOPMENT 335 Table 11-8. Sources of Output Growth for Twenty- Year Simulation Period Percentage contribution Share Strategy of Import Change in and Output output Domestic substitu- input-output sector growth change demand Exports tion coefficients EP-2 Agriculture 3.2 9.0 103.4 31.2 -4.6 -30.2 Food processing 7.3 9.1 79.4 23.4 -0.7 -2.1 Consumer goods 9.4 24.9 64.4 38.1 0.2 -2.6 Intermediate goods 12.4 23.0 51.0 30.9 1.7 16.4 Machinery 12.6 8.9 70.1 16.4 -1.5 14.9 Services 6.6 25.1 89.5 14.9 0.1 -4.5 Total 7.5 100.0 73.0 26.7 -0.2 0.4 B-2 Agriculture 3.2 9.4 101.3 16.6 13.2 -31.1 Food processing 6.8 8.6 89.2 10.0 3.4 -2.6 Consumer goods 8.1 20.0 85.5 13.9 3.8 -3.2 Intermediate goods 12.6 25.5 56.2 17.0 9.4 17.5 Machinery 12.9 10.1 70.4 6.4 6.4 16.8 Services 6.5 26.4 90.4 11.6 2.3 -4.3 Total 7.2 100.0 79.6 13.2 5.9 1.2 IS-2 Agriculture 3.2 10.2 95.3 8.8 26.8 -30.8 Food processing 6.6 8.7 89.8 5.4 7.5 -2.7 Consumer goods 7.9 20.1 86.7 9.1 7.1 -3.0 Intermediate goods 12.3 25.9 52.6 11.4 17.0 19.1 Machinery 12.2 9.5 65.4 3.5 12.2 18.9 Services 6.2 25.6 92.4 7.6 4.3 -4.3 Total 6.9 100.0 78.4 8.4 11.5 1.7 Note: The first column gives sectoral output growth rates. The second column expresses sectoral growth as a share of total change in output. The remaining four columns show the percentage contribution of each demand component to sectoral output change and sum to 100.0 for each sector. substitution opportunities are pursued. Thus, although the large foreign borrowing in the base year (8 percent of GDP) has been substantially reduced, evidence of continued external imbalance is apparent in the import premium of nearly 200 percent in the terminal period. 38 The Cost of Capital and Growth One important implication of the choice of a development strategy-an implication stemming from sectoral interdependence-concerns the rela- 38. In practice, of course, such continued disequilibrium would almost inevitably result in major policy changes as the economy was increasingly crippled by a shortage of foreign exchange. The results described here illustrate the strains inherent in such circumstances but have nothing to say about how the situation would be resolved. 336 DEVELOPMENT STRATEGY Figure 11-5. The Level of Incentives <1J .::! .... 0.. 2.0 IS-2 u ·c Incentives for import substitution