Structural Change and Development Policy A World Bank Research Publication with the assistance if Hazel Elkington including contributions 0/ Montek S. Ahluwalia Michael Bruno N icho las G. Carter Arthur MacEwan Alan M. Strout William J. Raduchel Larry E. Westphal 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 © 1979 by 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. The views and interpretations in this book are the author's and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf. Library of Congress Cataloging in Publication Data Chenery, Hollis Burnley. Structural change and development policy. Bibliography: p. 497 Includes index. 1. Economic development. 2. Underdeveloped areas. 3. Comparative economics. I. Elkington, Hazel, joint author. II. International Bank for Reconstruction and Development. III. Title. HD82.C474 1979 338.9 79-18026 ISBN 0-19-520094-2 ISBN 0-19-520095-0 pbk. Acknowledgments The publisher, the author, and the \Vorld Bank acknowledge with thanks permission to reprint or adapt the fonowing copyrighted material: "Substitution in Planning Models" by Hollis B. Chenery and William J. Raduchel. Reprinted by permission of the publishers from Studies in Development Planning, edited by Hollis B. Chenery. Cambridge, Mass.: Harvard University Press. Copyright © 1971 by the President and Fel- lows of Harvard College. "The Interdependence of Investment Decisions" by Hollis B. Chenery. Reprinted from The Allocation of Economic Resources, by Moses Abramovitz and others, with the permission of the publishers, Stanford University Press. © 1959 by the Board of Trustees of the Leland Stan- ford Junior University. "Economies of Scale and Investment over Time" by Honis B. Chenery and Larry E. "Westphal. Reprinted by permission from Public Econom- ics, edited by J. Margolis. © 1969 by the International Economic Asso- ciation, st. Martin's Press, Inc., and Macmillan & Co., London and Basingstoke. "Comparative Advantage and Development Policy" by Hollis B. Chenery. Reprinted by permission from American Economic Review, vol. 51, no. I (March 1961), © 1961 by the American Economic Association. "Development Alternatives in an Open Economy: The Case of Israel" by Hollis B. Chenery and Michael Bruno. Reprinted by permission from The Economic Journal, vol. 72, no. 285 (March 1962), © 1962 by Cambridge University Press. "Optimal Patterns of Growth and Aid: The Case of Pakistan" by Hollis B. Chenery and Arthur MacEwan. Reprinted by permission from The Theory and Design of Economic Development, edited by Irma Adelman and Erik Thorbecke. © 1966 by The Johns Hopkins University Press. "Foreign Assistance and Economic Development" by Hollis B. Chenery and Alan M. Strout. Reprinted by permission from American Economic Review, vol. 56, no. 4 (September 1966), © 1966 by the American Economic Association. "Growth and Poverty in Developing Countries" by Montek S. Ahluwalia, Nicholas G. Carter, and Hollis B. Chenery. Reprinted from The Jour- nal of Development Economics, vol. 6, no. 3 (September 1979) by arrangement with the publisher, North-Holland Publishing Company. © 1979 by the International Bank for Reconstruction and Development. v Contents List of Figures x List of Tables xiii Preface xvi Part One. A Framework for Policy Introduction 1 1. Economic Growth and Structural Change 5 Dimensions of the Transition 6 Alternative Patterns of Specialization 21 Transitional Countries 29 Implications 44 2. Models of the Transition 46 Alternative Modes of Analysis 47 Development Phenomena 55 An Approach to Policy 60 Part Two. Internal Structure Introduction 67 3. The Process of Industrialization 70 A Model of Industrialization 71 Patterns of Specialization 90 Sources of Industrialization 108 Overview 140 vii viii CONTENTS 4. Substitution and Structural Change 143 WITH WILLIAM J. RADUCHEL Substitution among Commodities and Factors 144 A Planning Model with Substitution 146 Surplus Labor and External Capital 165 Implications 172 5. The Interdependence of Investment Decisions 173 The Problem 174 The Model 177 Measurement of the Effects of Interdependence 188 Evaluation 204 Appendix. The Calculation of Prices and Total Costs 213 6. Economies of Scale and Investment over Time 217 WITH LARRY E. WESTPHAL Statement of the Problem 218 Formulating the Model 221 Investment Patterns 235 Conclusions 266 Part Three. External Structure Introduction 269 7. Comparative Advantage and Development Policy 272 Conflicts between Trade Theory and Growth Theory 273 The Measurement of Optimum Resource Al1ocation 281 Comparative Advantage and Balance in Development Programs 299 Conclusions 306 8. Development Alternatives in an Open Economy: The Case of Israel 309 WITH MICHAEL BRUNO Limits to Growth in Israel 310 A Model of Development Alternatives 314 Bottlenecks and Sector Limitations 334 The Choice of Policy 337 The Productivity of Foreign Assistance 339 9. Optimal Patterns of Growth and Aid: The Case of Pakistan 342 WITH ARTHUR MACEWAN The Analytical Framework 343 The Model 347 Growth Alternatives for Pakistan 355 Development and Assistance Strategy 376 CONTENTS ix Part Four. International Development Policy Introduction 379 10. Foreign Assistance and Economic Development 382 WITH ALAN M. STROUT Aid and the Transition to Sustained Growth 383 Prospects for the Transition 412 International Assistance Policies 437 Postscript 443 II. Growth and Poverty in Developing Countries 456 WITH MONTEK S. AHLUWALIA AND NICHOLAS G. CARTER The Dimensions of Global Poverty 457 Consequences of Existing Policies 469 The Scope for Improvement 484 References 497 Index 515 Figures 1-1. Accumulation Processes 11 1-2. Transformation of Demand 15 1-3. Transformation of Trade 16 lA. Transformation of Production 17 1-5. Transformation of Employment 19 1-6. Productivity of Labor by Sector 20 1-7. Alternative Patterns of Specialization 25 1-8. Examples of Primary Specialization, 1950-75 36 1-9. Examples of Import Substitution, 1950-75 39 1-10. Examples of Balanced Development, 1950-75 41 I-II. Examples of Industry Specialization, 1950-75 43 3-1. Structure of Trade 78 3-2. Structure of Production 87 3-3. Comparison of Observed Export Patterns 93 3-4. Simulated Patterns of Tradable Output 97 3-5 (part a). Sources of Output Growth by Sector: Average Pattern 116-17 3-5 (part b). Sources of Deviations from Proportional Growth: Average Pattern 118-19 3-6 (part a). Sources of Output Growth by Sector: Large Country Pattern 122-23 3-6 (part b). Sources of Deviations from Proportional Growth: Large Country Pattern 124-25 3-7 (part a) . Sources of Output Growth by Sector: Small Primary Pattern 126-27 3-7 (part b). Sources of Deviations from Proportional Growth: Small Primarv Pattern 128-29 3-8 (part a). Sources of O~tput Growth by Sector: Small Manufacturing Pattern 132-33 3-8 (part b) . Sources of Deviations from Proportional Growth: Small Manufacturing Pattern 134-35 3-9. Sources of Growth in Value Added: Average Pattern 136-37 x FIGURES xi 4-l. Long-run Production Functions (P 2b ) : Unit Value Added 156 4-2. Net Trade Functions (T 2 ) 157 4-3. Demand Functions (D 1 ): Unadjusted 158 4-4. Isoquants for GNP 1,000 159 4-5. Variation in Capital Intensity with Factor Prices 160 4-6. Output Effects of Capital Inflow 170 4-7. Employment Effects of Capital Inflow 171 5-1. Minimum Scale of Investment in Sector I: Metal Products 200 5-2. Minimum Scale of Investment in Sector 2: Steel Production 201 5-3. Effects of Demand on Optimal Investment 202 6-1. Best Solution for Model I [A. Levels of Capacity, Production, and Trade; B. Sources of Financing for Investment; C. Uses of Investment; D. Sources and Uses of Foreign Exchange] 240-44 6-2. Best Solution for Model II [Parts A-D as above] 246-50 6-3. Best Solution for Model III [Parts A-D as above] 252-56 6-4. Growth of Consumption over Preplan Value in the Best Solutions 257 6-5. Effect of Plant Size on Unit Cost of Investment 258 6-6. Model I, Pattern B3 (Solution Number 24): Levels of Capacity, Production, and Trade 262-63 8-1. Limits on S-V Axes 328 8-2. Limits on E-V Axes 329 8-3. Limits on F-V Axes 330 8-4. Welfare Analysis 338 9-1. Growth of GNP in Pakistan's Perspective Plans and in Two Solutions to the Model 361 9-2. Saving and Investment in the Basic Solution to the Model 362 9-3. Exports and Imports in the Basic Solution to the Model 363 9-4. Composition of Investment in the Basic Solution to the Model 364 9-5. Optimal Patterns of Aid: Varying Aid Conditions 365 9-6. Effects of Varying the Supply of External Capital .'370 9-7. Marginal Productivity of Aid 375 9-8. Optimal Pattern of Aid: Savings and Investment Growth 376 10-1. Illustrative Growth Paths: Pakistan, Past and Projected 403 10-2. Gap Equalization through Additional Import Substitution (Model 2) : Pakistan, 1962-75 408 10-3. Total Capital Inflow Required to Reach Self-sustaining Growth 411 10-4. Foreign Capital Requirements of Fifty Developing Countries: Alternative Development Patterns, 1962-75 427 11-1. The Kuznets Curve with Country Observations 467 11-2. Alternative Simulations, 1975-2000 475 xii FIGURES 11-3. Schematic Representation of Causal Relations Underlying Simulation Models 476 11-4. Effects on Poverty of Improving Distribution 493 11-5. Poverty Profile, 1960--2000 494 11-6. Tradeoff between Growth and Distribution 495 Tables I-I. Dimensions of the Transition 12-13 1-2. Alternative Patterns of Specialization 24 1-3. A Typology of Development Strategies, 1965 30-33 2-1. Examples of Development Phenomena 57 3-1. Summary of Simulation Procedure 76-77 3-2. Demand Response to Rising Income: Base Case 80 3-3. Trade Response to Rising Income: Base Case 81 3-4. Total Demand and Supply at Income Level of $400: Base Case 84 3-5. Comparison of Growth Elasticities from Model to Regression Results 88-89 3-6. Examples of Trade Patterns, 1965 94-95 3-7. Solutions for Large Countries: Income Level of $400 100-01 3-8. Solutions for Small Primary-oriented Countries: Income Level of $400 104-05 3-9. Solutions for Small Industry-oriented Countries: Ineome Level of $400 106-07 3-10. Decomposition of Sector Growth: Alternative Methods 112 3-l I. Incremental Decomposition of Value Added, $200-$400 138-39 4-1. Illustrative Estimates of Parameters 154-55 4-2. Effects of Direct Substitution in Production, Medium Demand and Trade: Model I 162-63 4-3. Effects of Indirect Substitution through Trade and Demand 164 4-4. Effects of Increasing External Capital 168-69 5-1. The Model of Production 180-81 5-2. Basic Solutions to the Model 183 5-3. Effects of Coordination on Profits and Prices: Case I 187 5-4. Economies of Scale in Steel Production 192 5-5. Input Functions for Labor and Capital: Case II 194 xiii xiv TABLES 5-6. Direct Factor Use by Sector and External Economies: Cases I and II 195 5-7. Cost of Supplying in Each Sector under Various Assumptions 199 5-8. Production and Imports of Crude Steel 203 5-9. Relative Costs of Production of Hypothetical Steel Plants 203 5-10. Calculation of Dual Variables in Cases I and II 214 5-11. Calculation of Dual Variables for Equation (5.17) 215 6-1. Tableau of the First Cycle of Activity 226-29 6-2. Classification of Investment Patterns 236 6-3. \Velfare Values for the Solutions Obtained 238-39 6-4. Effects of External Resources 259 7-1. Evaluation of Production and Import Activities by Accounting Prices 292-93 8-1. Israel's Economic Growth, 19;0-59 312-13 8-2. Selected Solutions to the Model 332 8-3. Comparison of Alternative Development Programs 336 9-1. Base Year Data 357 9-2. Value of Structural Parameters 358-59 9-3. Growth Rates and Significant Ratios for the Basic Solution of the Model and the Two Versions of the Pakistan Plan 360 9-4. Quantities in the Basic Solution 366-67 9-5. Shadow Prices in the Basic Solution 368 9-6. Effects of Varying Aid Supply 372-73 10-1. Distribution of Parameter Values, Thirty-one Country Sample 388 10-2. Summary of Phase Formulas for Foreign Capital Inflow (F t ) and Cumulative Foreign Capital Inflow (l,Ft) as Ratio to Initial GNP (Vo) 398-99 10-3. Structural Parameters for Pakistan Projections 400 10-4. An Example of Phase Developments for Pakistan, 1962-75 401 10-5. Effects of Import Substitution Policy on Capital Inflow, Pakistan, 1962-75 406 10-6. Indicators of Progress in Attaining Self-sustaining Growth, 1957-62 416-17 10-7. Value of Parameters Used in Projections 420-25 10-8. Proportion of Countries with Foreign Capital Requirements Determined by Investment-Saving Gaps 426 10-9. Aggregate Projections for 1962-75 428-29 10-10. Regional Projections, 1970 and 1975 430-33 10-11. Comparison of Modell and Model 2 Projections of 1975 Capital Inflow 436 TABLES XV 10-12. Summary of Structural Parameters and Growth Rates 447 10-13. Aggregate Growth, 1960-70 448 10-14. Comparison of External Aspects of Growth Projections 450-53 11-1. Sample Panel: Per Capita GNP, Population, and Poverty 460-61 11-2. Sample Panel: Indexes of Growth and Distribution (Base Case) 472-73 11-3. Growth and Poverty in Developing Countries 478-79 11-4. Growth Lags within and among Country Groups 480 11-5. Changes in Income Distribution in Selected Countries 482 11-6. Alternative Seenarios for 2000 490-91 Preface ECONOMIC DEVELOPMENT can be viewed as a set of interrelated changes in the structure of an economy that are required for its continued growth. They involve the composition of demand, production, and employment as well as the external structure of trade and capital flows. Taken together, these structural changes define the transforma- tion of a traditional to a modem economic system. Although development economics initially was concerned with the conditions for starting growth in a primitive economy, this problem now has been resolved in one way or another by most countries. The experience of growth at 5 or 6 percent a year, which has been typical of the past two decades, has lcd to a shift in emphasis to the avoid- ance of various forms of structural disequilibrium and to securing a wider distribution of the benefits of growth. The need to "restruc- ture" economic relations is now seen to be at the heart of interna- tional as well as national problems of development. The study of structural problems requires a revision of traditional forms of economic analysis. These problems imply additional con- straints on the functioning of the economic system that need to be specified empirically for individual countries or groups of countries. Policy conclusions are correspondingly more specific and less subject to generalization than are the criteria that emerge from purely neo- classical formulations. The approach to development policy that is illustrated in this vo]- ume has emerged from two parallel lines of research. The first con- sists in developing models that incorporate basic features of the resource endowments, productive structure, and policy constraints of a particular country. Studies based on structural characteristics of Japan, Chile, Israel, and Pakistan illustrate this procedure. These models are used either to analyze the sources of growth and structural change in the past or to simulate the effects of alternative policies in the future. xvi PREFACE xvii The second line of research consists in comparative econometric studies that seek to identify uniform patterns of change in the struc- ture of demand, production, and trade, as well as the effects on these patterns of size and other country characteristics. 1 The "stylized facts" of development that are identified in this way provide a basis for generalizing from country-based models. This procedure has been fol- lowed in studies of industrialization, the use of external capital, and the relations between growth and poverty. Identification of the effects of structural differences among develop- ing countries helps in formulating national and international policy. For example, the importance of international specialization, of capital inflows, and of economies of scale varies greatly between small and large countries. Similarly, the presence or absence of mineral and agri- cultural resources substantially affects the logical sequence of struc- tural changes and the policies needed to promote them. A typology based on such structural characteristics makes it possible to proceed from the analysis of indiVidual countries to more general results. Several types of conclusion emerge from these studies. The first are generalizations about individual phenomena, such as the sources of industrialization, the effects of economies of scale on resource alloca- tion, and the productivity of external resources. The second are gen- eralizations about development strategy in different types of countries, such as small primary exporters or large diversified economies. 'Dlese results can be incorporated in studies of individual countries or used in the design of international policies. The policy studies in this volume are grouped in three sections that focus in turn on the internal structure of demand and production, the external structure of trade and capital flows, and the international aspects of development policy. Structural phenomena are analyzed first in the context of one or more countries2 -then in a more general form. The topics treated are summarized in the introductions to each section.s 1. The analysis of development patterns has been published in Chenery and Syrquin (1975), which provides a background for the present volume. 2. These studies constitute chapters 5, 8, 9, and 10, which are reproduced with only editorial changes from the original versions. 3. Chapters 1, 2, 3, and 11 have not been previously published; chapters 4 and 6 have been extensively revised to incorporate results of more recent work. A post· xviii PREFACE Much of this work has been done in a collaborative form with suc- cessive research groups at Stanford and Harvard universities and the \Vor1d Bank. I am heavily indebted to my collaborators for the op- portunity to explore a wider range of country experience and of ana- lytical technique than would otherwise have been possible. The earlier stages of this work owe much to the suggestions of Irma Adelman, Kenneth Arrow, Hendrik Houthakker, Edward Mason, Tibor Scitov- sky, and Jan Tinbergen. I am particularly indebted to Tsunehiko \Vatanabe, my long-time collaborator in the study of industrialization in Japan and elsewhere. My attempts to produce a degree of synthesis from the studies assembled here have benefited from comments by Montek Ahluwalia, Bela Balassa, Michael Bruno, \V. Arthur Lewis, Ian Little, Sherman Robinson, Moises Syrquin, Lance Taylor, Lynda Thoman, Jeffrey Williamson, and Larry Westphal. Hazel Elkington is responsible for most of the statistical work and Nancy Jewett for the preparation of the manuscript. The final text was edited for publication by Brian J. Svikhart, who also supervised design and production of the book; Harry Einhorn read and corrected proof, Raphael Blow and Pensri Kimpitak prepared the charts, and Florence Robinson indexed the text. Alternation between academia and government over the past twenty years has had both costs and benefits. I hope there are benefits in relevance to policy that may be set against the reduction in theoretical rigor that this sequence has undoubtedly produced. Finally, it should be made clear that the views on policy expressed here are personal and do not implicate the World Bank. HOLLIS CHENERY !'v1enemsha, Massachusetts September 1979 script to chapter 10 compares the 1964 projections of trade and aid requirements for 1975 with what actually happened. In the interests of brevity, technical ap- pendixes to chapters 3, 4, 6, and 11 are not included here, but are available from the author on request. Part One A Framework for Policy THE STUDY OF DEVELOPMENT POLICY has arisen from the decline of the colonial system and its replacement by independent governments. Early attempts in this field were direct applications of neoclassical and Keynesian methods of analysis that had been developed for mature economies. These methods are gradually being replaced by more eclectic formulations rooted in the experience of the developing countries themselves. This reorientation of policy thinking has also brought a substantial change in the perception of the basic issues involved. Although early discussion about development had focused on the problems of ini- tiating growth and increasing investment, subsequent experience has shown that the developing countries as a group have been able to expand their economies at unprecedented rates in the period follow- ing \Vorld \Var II. Although starting to grow is still a problem for the least developed, the central questions for most developing coun- tries concern the management of the rapid structural changes that are required both to sustain growth and to improve its distribution. In dealing with these questions, it is useful to conceive of develop- ment as a transition from traditional to modern forms of economic organization, rather than as a process of uniform expansion. This transition is taking place more rapidly in many developing countries than it did in the industrializing economies of the nineteenth century because of the availability of more efficient technOlogy, imported industrial goods, and other influences from the advanced countries. To make effective llse of these opportunities requires coordinated changes in demand, production, trade, and the allocation of capital and labor. Many development problems can be traced to the emer- 2 INTRODUCTION TO PART ONE gence of disequilibria in either commodity or factor markets that reflect a lack of consistency in allocation decisions. Development policy is thus increasingly concerned with the management of struc- tural change and with the development of feasible combinations of market forces and government intervention that make this task possible. Part one of this volume constitutes a general introduction to the analysis of structural change. An earlier study of development pat- terns analyzed the postwar experience of a number of developing countries. 1 Portions of that work which are relevant for development policy are summarized in chapter 1. Such comparative studies lead to the identification of characteristic features of the development process, which are termed here development phenomena. Some of the principal development phenomena-such as the persistent imbal- ance between supplies of capital and labor, the consequences of scale economies and limited market size, and the dual role of external resources-are described in chapter 2. These phenomena provide the focus for the analysis of structural change and development policy that appears in subsequent chapters. The studies in this volume identify allocational problems that are characteristic of industrializing countries and propose analytical tech- niques for dealing with them. Like other applied fields, development economics is more concerned with the observable characteristics of developing countries than with the properties of highly abstract mod- els. Although attempts to apply the standard neoclassical theories of trade and growth to poor countries have provided some general in- sights into their problems, this deductive approach is not well suited to the study of structural change and the possibilities of disequilibrium. Case studies of transitional countries provide a necessary starting point for empirical analysis. Such studies must be supplemented by some form of comparative analysis to reach conclusions of more gen- eral validity. The chapters in this volume incorporate several ap- proaches to this problem. In the earlier studies simple models were developed to analyze a particular set of relations in a given country. Examples include an analysis of the sources of industrialization in Japan (chapter 3), the effects of economies of scale in Chile (chapter 5), and the role of external capital in Israel and Pakistan (chapters 1. Chenery and Syrquin (1975). INTRODUCTION TO PART ONE 3 8 and 9). Such country studies take advantage of a specific data base for estimating structural parameters and for a realistic specification of policy options; they also develop analytical methods that can be ap- plied elsewhere. Several methods of generalization are illustrated here. One ap- proach is to replace the statistical estimates in a country-based model with more representative parameters derived from cross-country data. This procedure is used in chapter 3 to analyze the process of indus- trialization under alternative assumptions about resource endow- ments and trading possibilities. Instead of explaining the historical changes observed in a particular country, this approach leads to explanations of general phenomena that are common to a number of countries. A second form of generalization is to estimate similar models for several countries. ll1is approach is particularly useful in studying the effects of external factors-such as export growth and the allocation of external capital-that influence a number of countries. This method forms the basis for the analysis of international development policy in part four of this book. The systematic analysis of development phenomena in different country settings leads naturally to a reexamination of the subject matter of development economics. This volume starts with an assess- ment of the "stylized facts" of development as they have emerged from statistical comparisons and country studies. In the period fol- lowing World War II there has been an increasing differentiation among developing countries as a result of their varying policies and degrees of success. The result has been to blur the distinction between the developed and the more advanced of the developing countries and to stimulate a search for other distinctions. The present volume makes use of a typology, based on resource endowments and interna- tional policies, which has proved helpful in analyzing different pat- terns of change in production and trade and the policies related to them. The accumulation of empirical generalizations about developing countries has also led to more realistic specifications of development models. Although these were initially described as "planning models," they have gradually become the principal mode of theorizing about development. 'Theoretical conclusions derived from such models are typically arrived at by numerical analysis to determine the implica- tions of the alternative values of structural parameters that they 4 INTRODUCTION TO PART ONE incorporate. Several examples of this form of theorizing are given here in studies of the effeets of substitution, economies of scale, and capital inflows. Although the focus of this volume ranges from individual sectors to international policies, it has a common concern with the inter- dependence of various aspects of the development process. The anal- ysis is designed to bring out the nature of interdependence and the relative importance of policies designed to influence given objectives. Given realistic limits to actions by governments, development strate- gies based on a combination of national and international policies are usually more effective than those relying too heavily on any single form of intervention. In this sense there are few development prob- lems that can be treated in isolation. Chapter I Economic Growth and Structural Change THE FIELD OF DEVELOPMENT ECONOMICS has been transformed over the past ten or fifteen years by a rapid increase in empirical knowledge. Theories advanced in the 1950s have been tested, reformulated, or discarded in the light of statistical analysis of their basic hypotheses. Out of this process has come a more comprehensive view of develop- ment as a set of interrelated changes in the structure of an economy. The starting point for this reformulation can be found in Arthur Lewis's (1954) concept of development as a transition from tradi- tional to modern forms of production and economic behavior. Over the period of the transition, the supply of unskilled labor is elastic; profits, savings, and investment are rising; industry grows more rapidly than agriculture; and the pattem of international trade is gradually transformed as the comparative advantage of a country changes. As Lewis points out, analysis of the transition requires some modification in the hypotheses that are relevant to advanced economies and a form of theory that is neither neoclassical nor Keynesian. In various forms the concept of a transition from a traditional to a developed economy has provided the basic organizing principle for both empirical and theoretical analysis. Starting with the work of Kuznets and Lewis, the transition has been measured by the accumu- lation of physical and human capital and by the transformation of the structure of demand, production, trade, and employment as the level of income rises. 1bese phenomena were studied first in the historical experience of the advanced countries and from intercountry compari- sons. More recently it has become possible to extend these results by analyzing the experience of many developing countries over the past quarter century. This introductory chapter presents the main conclusions of a series 5 6 ECONOMIC GROWTH AND STRUCTURAL CHANGE of studies of development patterns that were designed to provide an empirical basis for models of development.! The general dimensions of the postwar transition are described in the first section of thc chapter (pages 6-21). Since differences in resource endowments and international specialization lead to different development policies, they form the basis for the typology of development strategies devel- oped in the second section (pages 21-29). Countries representing each type of development strategy are discussed in the third section (pages 29-44), which makes it possible to generalize more readily from the studies of individual countries that are given in later chapters. Dimensions of the Transition The transition from a traditional to a developed economy can be defined in general terms as the set of changes in the economic struc- ture required to sustain a continued increase in income and social welfare. 2 Since these requirements depend both on social objectives and on possibilities for production and trade, they vary somewhat from country to country. Nevertheless, a number of factors lead to considerable uniformity in the transition in all countries during a given historical period: (a) similar changes in consumer demand with rising income (Engel functions); (b) the necessity to accumulate physical and human capital to increase per capita output; ( c) access of all countries to similar technology; and ( d) access to international trade. There is also a significant list of reasons to expect somewhat differ- ent patterns of structural change: (a) variation in social objectives and in the choice of policies; (b) variations in natural resource endowments; 1. Chenery and Watanabe (1958), Chenery (1960, 1964), Chenery and Taylor (1968), Chenery and Syrquin (l975), and Chenery (1977). 2. This section is based largely on Chenery and Taylor (1968) and Chenery and Syrquin (1975). A full discussion of concepts, measurement, and results ap- pears in the latter. Useful critiques of these attempts to measure development pat- terns are included in reviews of the Chenery and Syrquin work by Diaz Alejandro (1976) and Eckaus (1978). DIMENSIONS OF THE TRANSITION 7 (c) variation in country size; ( d) disparity in access to external capital; and ( e) changcs in the uniform factors ovcr time. Interest in the transition is enhanced by the empirical observation that in a given historical period the factors leading to uniformity seem to predominate. The sources of diversity are of equal concern for policy, however, since they lead to variations in the optimal strategy for a given country. Some economists have interpreted the statistical evidence on the uniformity of the transformation of production to imply that all countries must industrialize at a certain rate to de- velop. A more perceptive analysis, however, shows that there is con- siderable choice of when to industrialize and in which sectors. In establishing the dimensions of the transition, it is therefore important to identify the extent and sources of diversity as well as the central tendencies. To facilitate empirical analysis, the transition can be thought of as comprising a number of interrelated processes. In many cases these processes can be defined by giving particular forms to the structural relations in a general equilibrium system. The most straightforward example is provided by Engel's law, which can be derived by estimat- ing the demand for food as a function of income and price. If similar estimates can be made for all commodity groups, the results describe a general process by which the composition of demand changes as income rises. From an econometric point of view, these relations should be estimated together from a uniform specification because their components must add up to total expenditure. If variables other than income (such as relative prices or family size) affect con- sumption patterns, they must also be specified uniformly for all components.3 The processes that make up the transition include changes in vir- tually all economic functions: the increase in productive capacity (accumulation of capital and skills); transformation of resource use (demand, production, trade, and factor use); and such socioeconomic processes as urbanization, the distribution of income, and the demo- graphic transition. The present overview is limited to the more nar- rowly economic processes of accumulation and transformation that 3. A comprehensive analysis of the structure of demand based on intercountry data is given by LIuch, Powell, and Williams (1977); following this logic, they also include saving to achieve a consistent analysis of total income use. 8 ECONOMIC GROWTH AND STRUCTURAL CHANGE have been reasonably well measured and form the core of most models of development.4 Estimation Each development process can be described econometrically by a set of one or more equations in which the dependent variable is a dimension of the economic structure and the independent variables include the level of per capita income and indexes of some of the sources of diversity indicated above. For the latter, satisfactory meas- ures are available only for size of population and the capital inflow. rThere are no simple measures for the other sources of variation that have been identified, such as natural resources and country policies. Instead of including them in the regression equation, therefore, these factors are used as a basis for classifying countries into different pat- terns of specialization. The basic idea of a transition from an underdeveloped to a devel- oped economic structure can be represented by a logistic curve. In intercountry regressions this has been approximated by the following quadratic forma: (1.1) X=a ,BllnY+,B2{lny)2 y1lnN yz(lnN)2 +}; OiTi eF, where x = dependent variable, y = per capita gross national product (GNP) in 1964 dollars, N = population in millions, F = net resource inflow as a share of gross domestic product (GDP),and Ti = time period. This equation has been fitted for the 1950-70 period to data for 101 countries each having a population of over 1 million in 1960. \Vh.en applied to processes that are measured by shares of production, con- sumption, exports, or other aggregates, this formulation has the useful 4. Chenery and Syrquin (1975) also give summary measures for some of the principal socioeconomic processes. 5. This equation allows for only one of the two asymptotes of the logistic curve, but it adequately describes the middle and upper income range. A similar equation is used for time series analysis with population omitted. DIMENSIONS OF THE TRANSITION 9 property that the regression coefficients for each component add up to unity. (This results in functions of the form illustrated in figure 1-2 for each component.) The results summarized here are designed to depict the transition as a whole rather than to analyze its separate features. Many of the causal relations are explored in chapter 3 by simulating alternative patterns of industrialization. Processes of accumulation Accumulation may be broadly defined as the use of resources to increase the productive capacity of the econorny. Such uses include investing in physical capital, improving the quality of human capital (for example, education, health, and nutrition), and accumulating knowledge. 6 Although the accumulation of capital is an essential fea- ture of all theories of development, the extent to which the several types of capital can be substituted for each other is only now being explored empirically. To get an indication of total resources used for accumulation, it is necessary to add to the conventional measure of investment in physi- cal capital some portion of the public and private expenditures that contribute to increasing human capital.7 A rough indication of the total investment in human capital can be derived by assuming it to be a given percentage of total government expenditure. There are two types of explanation for the observed increase in the shares of savings and of government revenue with rising income. Classical theories of saving, as well as much of the literature on taxa- tion, trace both phenomena to a change in the composition of in- come recipients that accompanies rising levels of income. Saving out of profits is higher than that from wages, and aggregate saving rises as the share of profits increases. Similarly, sectors such as mining, industry, and foreign investment are more readily taxed than agricul- ture, and the ability of the government to raise taxes increases with the share of the readily taxable sectors. This type of explanation does not, however, exclude the direct effect of rising personal income, as 6. Some of the characteristics of a theory of growth based on this broad concept of capital are discussed by Johnson (1968). 7. Empirical work has concentrated on educational expenditure; the productive effects of health and other elements have not been widely studied. 10 ECONOMIC GROWTH AND STRUCTURAL CHANGE embodied in Keynesian savings functions and progressive tax struc- tures. Empirical evidence supports both approaches. 8 The best available measures of accumulation for intercountry anal- ysis are the shares of saving, investment, and government revenue in GDP and indexes of school enrollment. 9 Some of the cross-country estimates of these measures are given in table 1-1 and are displayed graphically in figure 1-1. The latter shows that a substantial increase in investment and in school enrollment takes place in the early phases of the transition (below $300 in 1964 prices)1° and a smaller increase thereafter. The uniformity of these processes among countries is shown by the standard errors of estimates (SEE) in table 1-1, which are approximately 25 percent of the value of each index at the mid- point of the transition. To test the validity of interpreting cross-country patterns as indi- cations of long-term trends, table 1-1 compares the increases in accu- mulation predicted from both time-series and cross-country regres- sions. Since it is not possible to separate income effects and time trends with any accuracy, the comparison is made by predicting the change in each variable, starting from its value at an income of $300 per capita and assuming a growth in per capita income of 2.5 percent for twenty years.n For all three measures of accumulation, the time-series regressions give increases that are 25-30 percent higher than the pure cross-section estimates from the same sample. 12 Both time-series and cross-section 8. The literature on saving in developing countries is reviewed by Mikesell and Zinser (1973) and Saito (1977). Evidence on taxation is evaluated by Lotz and Morss (1967) and Musgrave (1969). 9. The latter are more useful than educational expenditure (which is also ana- lyzed by Chenery and Syrquin) since they relate more directly to the investment process. 10. All the graphs show income levels in both 1964 U.S. dollars (as originally measured) and in 1976 prices, which are approximately donble the 1964 figures. The 1964 data are used in the discussion unless otherwise noted. 11. This procedure combines time trends and income effects into a composite "growth" effect. The comparison is taken from Chenery and Syrquin (1975, table 23), which shows the two elements separately. 12. The time-series regression equation assumes a common function for all countries but a separate intercept for each. TIle equation therefore measures the average effects of increasing income with capital inflow held constant. A positive effect of rising income on saving is also found in 75 percent of the separate regres- sions for individual countries. DIMENSIONS OF THE TRANSITION 11 Figure 1-1. Accumulation Processes 60 ,---------------~------------------~ Population :::: 10 million' persons · Y at midpoint 90 50 SO c: 0 '.0 70 '" '3 0. l'. 40 60 0 0. A C) ..... 0 ..... 50 0 ~ OJ , ,, 30 ~ ... <.> OJ , ,, Investment Po. 20 20 .,._ ... "".,.",. .-.--- .. .. - 10 , Human capital 70 100 200 300 500 SOO 1,500 (140) (200) (400) (600) (l,000) (1,600) (3,000) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Source: Chenery and Syrquin (1975, figures 1, 2, and 3). results therefore contradict the hypothesis of a constant aggregate propensity to save in transitional countries although it may be quite plausible for rich countries. Transformation of commodity demand and supply Transformation of the supply and use of resources is the dominant feature of the transition to which virtually all other aspects can be related. This transformation can be analyzed at either the commodity level or the factor level. In growth theory, primary attention is given to the allocation of labor and capital; commodities are significant only to the extent that they have different production or demand functions. In empirical analysis the emphasis is reversed. It is possible to give a fairly complete picture of the change in composition of commodity demand, production, and trade, but only indirect and partial measures are available for the use of capital, natural resources, and labor. In 12 ECONOMIC GROWTH AND STRUCTURAL CHAKGE Table 1-1. Dimensions of the Transition Share of CDP estimated value at Total change Component of the economic $70 $300 $1,500 (3 1) structure (1) (3 ) (4) Accumulation Investment (1 b) 0.136 0.203 0.234 1.72 Saving (la) 0.103 0.190 0.233 2.26 Government revenue (2a) 0.125 0.202 0.307 2.46 Index of school enrollment (3b) 0.244 0.637 0.863 3.54 Transformation of demand Food consumption (4c) 0.444 0.275 0.167 0.38 Other private consumption (4aAc) 0.335 0.392 0.457 1.36 Government consumption (4b) 0.119 0.135 0.141 1.18 Transformation of trade Imports (6e) 0.205 0.243 0.250 1.22 Primary exports (6b) 0.130 0.131 0.058 0.45 Manufactured exports (6c) 0.011 0.046 0.131 11.91 Services exports (6d) 0.028 0.048 0.059 2.11 Resource inflow (Ie) 0.033 0.012 0.001 0.03 Transformation of production Primary (Sa) 0.522 0.266 0.127 0.24 Industry (5b ) 0.125 0.251 0.379 3.03 Utilities (5c) 0.053 0.079 0.109 2.06 Services (5d) 0.300 0.403 0.386 1.29 Transformation of employment Shares by sector Primary (7a) 0.712 0.489 0.159 0.22 Industry (7b) 0.078 0.206 0.368 4.72 Services (7c) 0.210 0.304 0.473 2.25 Relative productivity Primarv sectors 0.733 0.548 0.799 n.a. Indust;y sectors 1.603 1.200 1.059 n.a. Urbanization (8) 0.128 0.439 0.658 5.14 -- .. -~.- Source: Chenery and Syrquin (1975, tables 3,4,5,7, and 23, and figure 9). n.a. Not available. Note: Proeesses are numbered as in the original. DIMENSIONS OF THE TRANSITION 13 Uniformity Increase (percentage) SEE ($300-$500) at $300 (per- Cross- Time- centage) R2 section series (5) (6) (7) (8) Accumulation 25 0.40 14 19 Investment (1 b) 26 0.71 15 19 Saving (la) 25 0.64 23 32 Government revenue (2a) 21 0.72 n.a. n.a. Index of school enrollment (3b ) Transformation of demand 15 0.82 n.a. n.a. Food consumption (4c) 17 0.54 n.a. n.a. Other private consumption (4a-4c) 31 0.15 n.a. n.a. Government consumption (4b) Transformation of trade 47 0.34 10 5 Imports (6e) 46 0.67 n.a. n.a. Primary exports (6b) 139 0.31 n.a. n.a. Manufactured exports (6c) 79 0.21 n.a. n.a. Services exports (6d) 0.08 n.a. n.a. Resource inflow (1 c ) Transformation of production 30 0.75 -27 -29 Primary (5a) 23 0.71 13 13 Ind ustry (5b) 33 0.32 n.a. n.a. Utilities (5c) 20 0.30 8 6 Services (5d) Transformation of employment Shares by sector 24 0.75 n.a. n.a. Primary (7a) 31 0.74 n.a. n.a. Industry (7b) 29 0.54 n.a. n.a. Services (7c) Relative productivity Primary sectors Industry sectors 29 0.67 n.a. n.a. Urbanization (8) 14 ECONOMIC GROWTH AND STRUCTURAL CHANGE describing the dimensions of the transition, therefore, it is more en- lightening to start from the commodity level and then proceed to the transformation of factor use. Systematic analysis of demand, production, and trade requires re- conciling the three different accounting systems in international use and also retaining a maximum number of countries in each sample. The following classification of commodities is therefore used for re- gression analysis: International Standard Valtie added in Industrial Con- production Classification Exports sumption Yp Agriculture and mining (0 + 1) E" Primary C, Food Y", Manufacturing and construction (2 - 4) Em Industry } Y" Utilities (5 + 7) (non traded) en, Nonfood Y. Services (6 + 8 81) E. Services Primary exports include the initial processing of minerals and agricul- tural products (as defined in the UNCTAD "A" concept) to combine all resource-determined commodity exports. 1bese measures are linked by the following accounting identities: (1.2) Y=V p + Vm V u+ V s, (1.3) Y = (C f + C n,) + G 1+ (E M), and ( 1.4) where Y is gross national product and E is total exports. To analyze the relations between the sources and uses of GNP, it is necessary to include production and trade in intermediate goods in a system of interindustry accounts. This system forms the basis for the more detailed analysis of the transformation in chapter 3.I3 The results of applying the standard regression equation (1.1) to the major components of demand, trade, and production are sum- marized in table 1-1 and figures 1-2, 1-3, and 1-4. Although each re- 13. Since the sample sizes are greatly reduced by disaggregation, the aggregate relations presented here serve as control totals for the subsequent disaggregation of production, demand, and trade in chapter 3. DIMENSIONS OF THE TRANSITION 15 Figure 1-2. Transformation of Demand 90 .--------------------,---------------------, Population 10 million perso'ns e= Y at midpoint 80 ,, ,, 70 , ,, -- --- --- -- Total private ..... _ _ consumption 0.. Q -.:: 60 --_ ... --_ ... o v OJ) E @50 ~ -_ ........ -- "CI Food conrumption ,, ~ 40 SV '"0 u ·z Nonfood [1 30 S conrumption 0 Q 20 .................. :::::: -- ..... _- ....... 10 Government conrumption 0 70 100 200 300 SOO 800 UOO (140) (200) (400) (600) (1,000) (1,600) (3.000) Per capita GNP in 1964 U.S, dollars (1976 dollars in parentheses) Source: Chenery and Syrquin (1975, figure 4) , fleets a different aspect of resource allocation, they show a general similarity that can be traced to two underlying factors: (a) changes in the composition of domestic demand and (b) changes in comparative advantage resulting from the accumulation processes described above. lne rise in the share of industry in total output, which can be taken as the central feature of the transformation, results from the inter- action of both sets of factors. lne largest and most uniform of these changes is the decline in 16 ECONOMIC GROWTH AND STRUCTURAL CHANCE Figure ]-3. Transformation of Trade .= Population 10 million persol;s Y at midpoint 0.. 30 § ..... ~ 25 OJ) ~ '" 20 u & 15 ~~ Total exports (E) OJ v ---- ------4--_ £ 10 Primary exports (Ep) C u Services exports (E,) E u 5 £2 ill 70 100 200 300 500 800 1,500 ( 140) (200) (400) (600) (1,000) (1,600) (3,000) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Source: Chenery and Syrquin (197:;, figure 6). food consumption from over 40 percent to only 17 percent of total domestic demand in the course of the transition. Figure 1-2 shows that the low income elasticity of demand for food allows all the other major components of demand-nonfood consumption, government consumption, and investment-to increase their shares of total demand. Imports and exports are the main sources of variation in the de- mand and supply of commodities. Even when allowance is made for country size, table 1-1 shows much less uniformity in all of the trade measures, particularly the share of manufactured exports. This is attributable to variation in both resource endowments and govern- ment policies. Since the whole course of the transition is affected by differences in patterns of specialization, they are discussed in more detail in the second section of this chapter (pages 21-29). Tbe average transformation of production shown in figure 1-4 is the net result of the transformation of both demand and trade. In DIMENSIONS OF THE TRANSITION 17 aggregate terms, table 1-1 shows that the rise of industry is remark- ably uniform despite the variability of trade patterns. This is beeause of the greater weight of domestic demand as compared with trade. Furthermore, time-series and cross-section regressions give essentially the same results in the postwar period. However, subsequent analysis will show that substantial variation in the composition and timing of industrialization is associated with different patterns of specialization. Transformation of factor use The transformation of factor use may be broken down into three components: (a) a change in overall factor proportions through the accu- mulation of physical capital and skills; (b) reallocation of these factors among productive sectors in varying proportions; and ( c) increases in total factor productivity by sector. Figure 1-4. Transformation of Production A. <:: '-' 60 '0 50 Primary .= Population = 10 million persons Y at midpoint ;;:j 40 ~~~~~~ Q,) :g , ······· '" ~- .- '" 30 ,',. ..::. Services 20 .. .' ' ..... - ... _- ... ........ -- '0 10 .., 1:: Utilities_-----4------- .e --_ ... u ::: .... ii5 70 100 200 300 500 800 1,500 (140) (200) (400) (600) (1,000) (1,600) (3,000) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Source: Chenery' and Syrquin (1975, figure 5). 18 ECONOMIC GROWTH AND STRUCTURAl, CHANGE The main difficulty in describing these changes is the lack of any internationally available indicator of the capital stock by sector. Since attempts to measure the properties of production functions by in- direct methods have not yet achieved satisfactory results, it is only possible to give a very partial account of this aspect of the transforma- tion, and this is one of the major weaknesses in existing models of development. If all economic sectors had the same production function and faced the same factor prices, the patterns of factor use would follow the pattern of output already described. The observed divergence between changes in employment and in output reflect diffcrences in both production functions and the supply of labor and capital to each sector. The patterns of transformation in output and employment are compared in figures 1-4 and 1-5. They show that the shift of labor out of primary production lags considerably behind the transformation of output. The available country studies suggest that this lag is the result of the greater concentration of physical and human capital in manu- facturing and overhead facilities. As a consequence employment in pri- mary sectors exceeds industrial employment up to the income level of $800 even though value added in industry surpasses that in primary activities at an income level of $400. A crude measure of the lag in labor productivity in agriculture can be derived by comparing its share in output to its share in employ- ment.14 A comparison of the relative productivity in primary and other sectors is given in figure 1-6. In the early stages, primary labor produc- tivity grows more slowly and falls from 75 percent of the average to about 50 percent at an income of $600; it then rises more rapidly in the cross-country regressions. This phenomenon has been studied in detail by Kuznets (1971), who concludes that it must reflect the low total productivity of all factors in agriculture rather than merely the substitution of labor for capital. The slow rise in average agricultural output per person employed is more a reflection of the rapid growth of population and the lack of 14, This measure is defined as: (Vp/V) ..;- (L./L) = (V./L.) ..;- (V/L). where (V /L) is average labor productivity for the economy, Since mining is typi- cally less than 10 percent of primary output, this measure is essentially an esti- mate of the productivity of labor in agriculture relative to the national average, as suggested by Kuznets (1971). DIMENSIONS OF THE TRANSITION 19 Figure 1-5. Transformation of Employment ,, 70 Population = 10 million persons '., e= Y at midpoint '. Primary ., u .... 60 .E C; 50 ,.0 ~ (l) ,.0 ..... '0 40 (l) tlO ..... '" [i ~ 30 u E: i':! ~'" .... 20 Services 0 ,.0 '" ...:I 10 " " Industry 70 100 zoo 300 500 800 I, SOD (140) (200) (400) (600) (1,000) (1,600) (3,000) Per eapita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Source: Chenery and Syrquin (1975, figure 8). alternative employment opportunities than of the inherent charac- teristics of agricultural production functions. In countries without surplus labor and with fairly equal access of all sectors to capital and technical knowledge, this lag has not been observed. 15 Related processes A number of socioeconomic processes are related more or less di- rectly to the transformation of factor use and in turn produce some 15. For example, in the earlier history of New ZeaJand, Canada, or Argentina, where land was not a constraint. 20 ECONOMIC GROWTH AND STRUCTURAL CHANGE Figure 1-6. Productivity of Labor by Sector 2.0 r - - - - - - - - - - . . , . . - - - - - - - - - - - - , ::'..'..,~. Services 1.5 c- t; l Industry ; : - ~ " " - 'lL .. ~ ____ :5 1.0 1---.-.-.-------I:---p-ri-m-a-ry----.-.-.-,.-'-'-.--l . .. 0.5 - I I I I · I I 70 100 200 300 500 800 1,500 (140) (200) (400) (600) (1,000) (1,600) (3,000) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Source: Chenery and Syrquin (1975, figure 9). repercussions on resource allocation. Among the most closely con- nected is urbanization, which is caused by the increase of industry and other modern activities as well as by the incentive to migrate from rural areas. 16 The urban population typically exceeds the rural popu- lation above an income level of $500,17 Development policy in the latter part of the transition is increasingly dominated by the invest- ment and employment problems created by rapid urban growth. Industrialization and rapid urbanization typically lead to an urban bias in resource allocation, a concentration of income growth in the modern sectors of the economy, and a worsening of its relative dis- tribution. The initial hypotheses about the nature and causes of this phenomenon were formulated by Lewis (1954), Kuznets (1955, 1963), and Myrdal (1957). The inequality of distribution tends to grow until the benefits of increased investment and technological progress are spread more widely and economic dualism is reduced. 16. Migration is analyzed as an aspect of dualism by Harris and Todaro (1968); an evaluation of recent evidence is givcn by Yap (1975). 17. Chenery and Syrquin (1975, pagc 55). ALTERNATIVE PATTERNS OF SPECIALIZATION 21 Worsening income distribution, although a normal feature of post- war development patterns, is by no means inevitable. Unlike the other development processes discussed so far, cases such as Yugoslavia, Tai- wan,18 and Korea 19 exist in which it has not occurred to a significant extent. 20 Statistical analyses show that a number of structural features- higher education, lower population growth, less dualism-are asso- ciated with more equal income distribution. 21 Although a number of these relations have been analyzed in country studies, a comprehen- sive theoretical framework that brings together the major components has not yet been formulated. The final dimension of the transition to be noted here is the demo- graphic transition. On the grounds of universality and basic relations to other development processes, this dimension should be incorporated in any statement of thc agenda for development theory and policy. 1ue reduction in mortality and fertility rates is intimately connected to the increase in per capita income and other aspects of moderniza- tion. As in the case of income distribution, there is nothing inevitable about the long lag between mortality and fertility reduction that characterizes most developing countries.22 Alternative Patterns of Specialization The average patterns of development just described can be used as a basis for identifying alternative sequences that characterize coun- tries having different resource endowments and development strate- gies. This has been done by subdividing the total sample into more homogeneous groups using indexes of size and export patterns. The results will be presented in two steps: (a) the alternative patterns 18. Throughout the volume, the Republie of China is referred to as Taiwan. 19. Throughout the volume, referenees to Korea are to the Republic of Korea, otherwise known as South Korea. 20. Experience with income distribution is discussed further in chapter II. 21. See Adelman and Morris (1973), Chenery and Syrquin (1975), and Ahlu- walia (1976). Some of the policy implications of these relations are explored by Chenery, Ahluwalia, Bell, Duloy, and Jolly (1974). 22. A survey of population policy in relation to other aspeets of economic devel· opmcnt appears in King and associates (1974). Cross·country regressions in the present framework appear in Chenery and Syrquin (1975, page 57). 22 ECONOMIC GROWTH AND STRUCTURAL CHANGE estimated for the principal groups of countries; and (b) a suggested typology of development strategies. 23 Sources of variation In principle, countries could be classified either on the basis of direct indicators of differences at the factor level (such as surplus labor and supplies of natural resources) or on less direct measures of the effects of factor differences on commodity production and trade. In practice, the greater abundance of data at the commodity level provides a much more promising point of departure than do factor supplies, since it becomes possiblc to include a much wider range of countries. A second reason for preferring this approach is that it can be more directly linked to government policies, which have their pri- mary effect on commodity production and trade. Both theoretical and statistical analyses have focused on explain- ing differences in the timing of various structural changes. This em- phasis derives from the observation that there is greater structural variation among countries during thc transition than there is within the extreme groups of traditional and developed economies. The vari- ation in the transformation at the commodity level can be traced to the degree of openness of the economy and the extent of specializa- tion in primary exports. The small, open economy with a high level of primary exports specializes in primary production and lags in the de- velopment of industry. The opposite extreme is found in the large, relatively elosed economy lacking a resource base for exports, which must industrialize relatively early to expand its income.24 To measure the differences in timing of the structural changes asso- ciated with these and other factors, it is necessary to subdivide the sample of countries into groups having more uniform resource endow- ments and trading conditions. In the absence of satisfactory direct measures of natural resource endowments, the difference between a given country's actual export pattern and the average pattern pre- dicted for its size and income level is used as an indirect measure. On this basis the trade orientation of each country can be defined; the 23. This section is based largely on Chenery and Taylor (1968) and Chenery (19i3). 24. These phenomena are much more prononnced when industry is disaggre- gated by sector, as in Chenery and Taylor (1968) and chapter 3 of this volume. ALTERNATIVE PATTERNS OF SPECIALIZATION 23 resulting patterns of specialization are described as (a) primary ori- ented, (b) balanced, or (c) industry oriented.25 Another factor that has been shown to have a substantial effect on the timing of the transformation is the size of the economy. In the standard regression equation (1.1), market size is measured by popu- lation size and per capita income. The pure "scale effect" is deter- mined by the regression coefficient for population.26 The effects of scale on the profitability of domestic investment are greater at low income levels, where economies of scale in production are larger in relation to market size. The availability of a large domestic market also affects the development strategy of large countries, which tends to reinforce the purely economic effect of scale OIl the timing of the transition. 27 The indexes of trade orientation and scale can be combined to di- vide the sample according to a two-way dichotomy: large or small, pri- mary or industry oriented. Since the sample of large countries is too limited to justify a statistical breakdown of this category, separate regressions have been calculated for only three groups: large (L); small, primary oriented (SP); and small, industry oriented (SM) .28 The principal dimensions of the transition have been estimated separately for these three groups, each of which contains from twenty- five to thirty countries and from 400 to 500 observations on the prin- cipal variables. A statistical comparison of the resulting estimates at the income level of $200, which is the point of maximum difference among the three patterns, is given in table 1-2. This comparison shows that the estimates of most structural features are improved by using 25. The index of trade orientation is defined as: TO = [(E. - Em) / E] [(Ep - Em) / Ej, where Ep, Em, and E are determined from the pooled regres- sions described above (pages 8, II). A classification of the trade patterns of eighty- six countries on this basis for 1965 is given by Chenery and Syrquin (197 S, tables 10 through 13). A comparison of a similar classification to one based on direct measures of resource endowments (arable land and mineral resources) is given in Chenery (1964). 26. Total GNP (yN) is a less useful measure of market size than are its indio vidual components, since increasing the level of income has a different effect than does increasing population. The latter will be identified as a pure scale effect. 27. The differences between policies of large and small countries arc discussed in chapter 7. 28. The intermediate group (Small Balanced) was not used in this regression analysis for statistical reasons, but it is retained in the typology of development strategies discussed on pages 36-40. Table 1-2. Alternative Patterns of Specialization (Per capita GNP = $200) Estimated value at y = .$200 Standard error of estimate Total Total Item sample' Large h SP" SIW sample' Largeb SP' SM' Accumulation Investment 0.188 0.201 0.154 0.195 0.050 0.050 0.048 0.040 Saving 0.171 0.181 0.163 0.094 0.050 0.050 0.048 0.040 Governmcnt revenue 0.181 0.158 0.198 0.181 0.050 0.053 0.047 0.046 Transformation of demand Private consumption 0.686 0.708 0.686 0.746 0.068 0.070 0.071 0.052 Transformation of trade N Total exports 0.218 0.123 0.239 0.266 0.115 0.052 0.08] 0.154 4>- Primary exports 0.136 0.063 0.203 0.098 0.060 0.028 0.074 0.039 Manufactured exports 0.034 0.028 0.016 0.088 0.064 0.033 0.014 0.093 Capital inflow _d _d _d _d 0.016 0.020 -0.009 0.101 Trade bias _d _d _d _d 0.47 0.28 0.78 0.04 Trade orientation 0 -0.19 0.31 -0.43 _d _d _d - d Transformation of production Primary share 0.327 0.320 0.358 0.235 0.079 0.057 0.086 0.067 Industrv share 0.215 0.250 0.177 0.216 0.057 0.056 0.043 0.063 Utilities and services 0.458 0.430 0.465 _e _e _e _e 0.549 Production bias 0.112 0.070 0.181 0.019 _d _d _d _d Production orientation 0 -0.042 0.069 -0.093 _d _d _d _d Sources: Chenery and Syrquin (1975, tables 3, 14, 15,1'2, S5, S7, S8, S9, and Sll). c. Population of 5 million. a. Population of 10 million. d. Not applicable. b. Population of 40 million. e. Not estimated. ALTERNATIVE PATTERNS OF SPECIALIZATION 25 Figure 1-7. Alternative Patterns of Specialization (a) SP (b) L (c) Sf.,,! v .. ,, ,, " ,. , , ; ,; ,, ; .5 ! 1 p '" 1 , Ep j 100 : :g (200) V p ; " \D r-- 0- ...... v. /// 1,...' ; ,,1. F : /,--'" j....- ... F"I : : 200 500 200 500 200 800 (400) (1,000) (400) (1.000) (400) (1,600) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) SOUTce: Chenery and Syrquin (1975, tables 14 and 15). more homogeneous groups instead of the full sample. The standard errors of estimate for the nine measures of the transformation shown are significantly reduced in nearly half of the cases. 29 Stratification of the sample also allows for differences in the timing of the various de- velopment processes and thus gives a better picture of their relations to each other. The main differences in the transformation of production and trade among the three patterns are shown in figure 1-7. Although similar 29. The rationale for stratifying the sample and the tests of homogeneity (F tests) of alternative subdivisions are discussed in Chenery and Syrquin (1975, pages 162-67). 26 ECONOMIC GROWTH AND STRUCTURAL CHA.."'GE changes take place in each, they differ greatly in their timing. I shall therefore use two benchmarks to compare them: first, the point (P) at which industrial production exceeds primary production and, sec- ond, the point (E) at which nonprimary exceeds primary exports. (Since no major differences have been identified in the patterns of demand for these three groups, this component of the transformation is not shown.) The characteristic pattern of large countries is distinguished from the all-country norms by a low level of primary exports, for which earlier development of industry compensates. Because of the lower level of total trade, the shift in the pattern of production is more de- pendent on changes in internal demand and takes place relatively early in the transition. There is a substantial difference between the two patterns for small countries. Primary exports in the SM group are only half the levels of the SP category. This is offset by a higher capital inflow and earlier industrialization. 111e effect of industrial specialization is to advance the shift to industry (point P) in the SM group from the average in- come level of $300 to $220. Conversely, primary specialization in the SP group delays this shift to an income of about $500. Despite the substantial differences in market size and resources that characterize these three groups of countries, the implied rate of indus- trial growth is shown to be remarkably similar. In relation to the average pattern described in the first section of this chapter, the large country leads by a decade and the small primary exporter lags by a similar period (assuming a gro\vth in per capita GNP of 2 percent a year). As will be shown in chapter 3, however, the variation in timing of individual sectors of industry is much greater. A typology of development strategies The three cross-country patterns of the large economy, the small, primary oriented economy, and the small, industry oriented economy suggest alternative sequences of structural change that stem from their size and natural resources. But these patterns can also be considerably modified by government policies. I shall therefore classify the coun- tries in each group according to indexes that also reflect policy differences. In this way a typology of development strategies ean be ALTERNATIVE PATTERNS OF SPECIALIZATION 27 established that is related to the policy choices of a country as well as to its natural endowments. 3o The distinctions to be made are derived from a number of studies of the effects of trade and aid policies on the pattern of resource allo- cation. On the trade side the principal distinction is between outward- looking policies, which favor or at least do not penalize exports, and inward-looking policies, which are designed to favor import substitu- tion in industryP The effect of outward-looking policies can be iden- tified either by a normal or relatively high share of exports in GNP compared with the predicted value for a country or by a rise in this share over time. Similarly, effects of inward-looking policies are indi- cated by a relatively low or declining share of exports. The policies used to promote import-substituting industrialization, furthermore, make manufactured exports unprofitable and result in a continuation of primary specialization, as well as a low level of trade. Inflows of external capital modify these patterns somewhat; capital inflows are not needed by successful primary exporters, who more often have a net outflow from servicing existing debt. For other coun- tries, borrowing provides a temporary alternative to further import substitution or part of a liberalization package for countries trying to shift from inward to outward policies. In either case a successful aid- supported strategy is usually characterized by a relatively low but increasing share of exports in GNP. This analysis suggests several indexes that can be used to identify the aggregate effects of trade and aid policies. Those used here are: (a) the relative export share (EfE); (b) the rate of growth of exports relative to GNP; (c) the index of trade orientation defined above (TO); (d) an index of production orientation (PO); and (e) the share of external capital in GDP (F /Y). All except (d) have been discussed. The production orientation index measures whether the share of manufacturing is higher or lower than that predicted by the standard regression equation for the in- 30. The term strategy is used for convenience to encompass both of these as- pects, but it does not imply that governments necessarily make deliberate choices among policy alternatives. 31. See Little, Scitovsky, and Scott (1970); Balassa (1970, 1971b); and Diaz Alejandro (1974, 1975). 28 ECONOMIC GROWTH AND STRUCTURAL CHANGE come and size of a given country.32 ll1e index of production orienta- tion (PO) is used to distinguish the effects of primary specialization (where PO is positive) from import substitution (wbich produces negative or zero values of PO). A typology of development strategies based on the deviations of each country from the average patterns of production and exports is given in table 1-3 for 1965. This typology includes almost all develop- ing countries that are far enough into the transition to identify the strategy being followed and that have the data required for classifica- tion. 33 With few exceptions the transitional countries can be readily classified into one of the four basic strategies indicated in the table, although the dividing line between import substitution (II-A) and more balanced development (II-B) is fairly arbitrary.34 But since the purpose of the typology is to help in identifying typical sequences of structural changes and the policies associated with them, the classifi- cation of marginal countries has little effect on the analysis. 32. The concept, analogous to export orientation, is defined as PO = (Vp - V m) CV. -'Vm), where the second term is the predicted difference between primary and industrial value added. This definition has not been normalized for the variation of total commodity production in GNP, which is quite limited. 33. This table is a slightly revised version of a similar classification in Chenery and Syrquin (1975, table 16), and takes account of more recent country studies. Several eountries at the earliest stage of the transition (Uganda, Sudan, and Sierra Leone) have been omitted; Nigeria. Algeria, and Indonesia have been added since their development patterns are now established. Most communist countries are omitted for lack of comparable data. 34. The classifieation is based on the following values of the structural indexes: I. Primary specialization a. Primary-oriented exports (TO greater than 0.10) b. Primary-oriented production (PO greater than 0.07) c. Export level usnally above normal II-A. Import substitution a. Primary export orientation (TO greater than 0.10) b. Low total exports (exports below 0.7S of normal levels) c. Production not primary oriented (PO less than +0.07) II-B. Balanced production and trade a. Normal export orientation (TO between -0.10 and +0.10) b. Normal production orientation (PO between -0.07 and +0.07) III. Industrial specialiwtion a. Industrial export orientation (TO less than -0.10) b. Industrial production orientation (PO less than -0.07) Indexes that do not fit the criteria established for each pattern are shown with an asterisk. In ambiguous cases, the classification was based on the policies followed in the mid-1960s. TRANSITIONAL COUNTRIES 29 Of the four development strategies identified here, the first two- primary specialization and import substitution-have been adequately described in the development literature, but the other two have re- ceived less attention. I have tried to identify almost all transitional countries with one of the four strategies to determine the typical characteristics of each. Transitional Countries How is this typology of development strategies used? To illustrate, I shall indicate some of the characteristic features of the transition for four representative countries in each category. This overview is de- signed to provide a background for the study of individual countries and particular features of structural change in the following chapters. Of central importance is the sequence of structural changes in pro- duction, trade, and capital inflows and the relation of these changes to government policies. The sixteen countries that illustrate this discussion are among the more successful examples of the early and middle phases of each of the four strategies as indicated by the achievement of moderate to high growth and a sustainable pattern of structural change.35 Graphs for each country of the principal indexes of structural change at the commodity level illustrate the magnitude and sequence of changes since 1950. TIlese graphs (figures 1-8 to 1-11) are designed to show the relations among changes in production, trade, and capital inflows. To facilitate comparison among conn tries, the transition is divided into earlier and later phases. In the average pattern described in the first section of this chapter, the early phase runs from an income level of $100 to about $300 and the later phase from $300 to $1,500 (in 1964 prices). In relation to the balance-of-payments constraint, the early phase is characterized by emphasis on primary exports, "easy" import substitution, and the availability of external aid on soft terms. In the later phase the trade balance requires a shift to nonprimary exports, "second stage" import substitution, and external borrowing on harder terms. 35. The representative countries are listed in table 1-3. I havc omitted some of the morc extreme examples, such as Saudi Arabia and Singapore, but have included Israel because it is analyzed in detail in chapter 8. 30 ECONOMIC GROWTH AND STRUCTURAL CHANGE Table 1-3. A Typology of Development Strategies, 1965 GNP Exports per per capita capita (annual (annual average average GNP growth growth Rela- Popu- per rate rate five lalion capita 1960- 1960- export Strategy (N)" 1965" 75)" 75)0 lever I. Primary specialization Tanzania 12 67 3.0 1.0 1.52 Nigeria 49 88 3.4 9.5 1.60 Indonesia 84 105 2.4 9.0 0.73 Bolivia 4 124 2.5 6.0 0.95* Sri Lanka 11 142 2.0 -1.0 n.a. Ivory Coast 4 179 3.5 2.7 1.37 Zambia 4 179 2.0 1.2 2.36 Algeria 12 202 1.8 IJ n.a. Dominican Republic 4 215 3.4 2.6 0.61 Iran 218 8.1 7.9 1.39 Iraq 8 249 3.3 1.6 n.a. Malaysia 9 258 4.0 3.1 1.86 Saudi Arabia 7 271 6.6 11.2 2.30 Nicaragua 2 330 2.4 4.8 0.99* Venezuela 9 830 2.2 -1.0 1.09 Total population" 243 II-A. Import substitution India 481 84 l.3 0.5 n.a. Ghana 8 156 -0.2 -2.5 0.70 Ecuador 5 195 3.4 7.9 0.73 Brazil 81 216 4.3 5.4 0.94 Colombia 18 228 2.7 0.5 0.68 Turkev 31 244 4.0 5.l 0.44 Chile 9 419 1.3 1.3 0.53 Mexico 43 434 3.2 0.9 0.73 Uruguay 3 498 0.5 2.0 0.64 Argentina 22 787 3.1 1.8 0.48 Total population" 700 II-B. Balanced development Thailand 31 llO 4.6 5.4 1.40* Philippines 32 l49 2.5 1.3 1.21 TRANSITIONAL COUNTRIES 31 Share manufac- tured exports in manufac- Pro- Trade turing duction orien- pro- orien- Capital tation d duction C taliond inflow· Strategy I. Primary specialization 0.05* 0.18 0.06* -0.038 Tanzania 0.34 0.02 0.15 0.021 Nigeria 0.35 0.01 0.19 0.005 Indonesia 0.19 0.01 -0.06 0.060 Bolivia n.a. n.a. n.a. n.a. Sri Lanka -0.04* 0.24 0.11 -0.034 Ivorv Coast 0.36 0.02 0.16 -0.170 Zan{bia n.a. n.a. 0.03 0.050 Algeria Dominican 0.22 0.04 -0.04 0.023 Republic 0.48 0.06 0.22 -0.059 Iran n.a. n.a. 0.34 -0.159 Iraq 0.31 0.28 0.20 -0.046 Malaysia 0.47 0.00 0.47 -0.425 Saudi Arabia 0.12 0.12 0.12 0.024 Nicaragua 0.90 0.00 0.31 -0.096 Venezuela II-A. Import substitution n.a. 0.08 n.a. 0.021 India 0.03 n.a. n.a. 0.077 Ghana 0.34 0.01 0.01 -0.002 Ecuador 0.54 0.05 -0.03 -0.026 Brazil 0.29 0.03 0.04 -0.010 Colombia 0.30 0.02 0.12* 0.014 Turkey 0.32 0.06 -0.08 -0.008 Chile 0.35 0.03 -0.02 0.001 Mexico 0.28 0.03 -0.13 -0.060 Uruguay 0.67 0.02 -0.01 -0.014 Argentina II-B. Balanced development -0.06 0.18 -0.03 0.013 Thailand 0.13* 0.06 -0.05 0.009 Philippines (Table continues on the following pages) 32 ECONOMIC GROWTH AND STRUCTURAL CHANGE Table 1-3 (continued) GNP Exports per per capita capita (annual (annual average average GNP growth growth Rela- Popu- per rate rate tive lation capita 1960- 1960- export Strategy (N)' 1965" 75)" 75t leveld II-B. Balanced development (continued) Svria 5 174 2.2 1.3 0.89 Morocco 13 179 1.9 1.8 0.89 EI Salvador 3 241 1.8 2.4 1.03 Guatemala 4 278 2.4 5.2 0.65* Peru 12 289 2.7 -1.4 0.72 Costa Rica 2 361 3.4 7.7 0.78 Jamaica 2 420 3.6 3.1 1.24 South Africa 18 552 2.3 1.7 1.46* Spain 32 572 5.7 16.8 0.73 Greece 9 585 6.6 11.0 0.34 Ireland 3 815 3.2 6.4 1.07 Total population" 165 III. I ndustry specialization Kenya 10 96 3.2 2.3 1.72 Egypt 29 138 1.5 0.4 1.37 Taiwan 12 201 6.3 19.3 0.78 Yugoslavia 20 415 5.5 15.5 1.30 Hong Kong 4 512 6.5 7.3 2.34 Singapore 2 522 7.6 7.7 n.a. HighF Pakistan 114 84 3.3 1.8 0.88 South Korea 28 123 7.1 28.6 0.64 Tunisia 4 198 4.1 6.2 0.82 Portugal 9 361 6.9 7.0 0.99 Lebanon 2 446 3.1' 6.7 0.67 Israel 3 1,126 5.2 8.1 0.62 Total population" 237 * Features that deviate from the criteria for each pattern. n.a. :\fot available. a. Chenery and Syrquin (1975, tables 10 to B). b. \Vorld Bank Atlas: Population, Per Capita Product, and Growth Rates (Washington, D.C., 1977). TRANSITIONAL COUNTRIES 33 Share manufac- tured exports in manufac- Pro- Trade turing duction onen- pro- onen- Capital tation" duction C tationd inflow' Strategy II-B. Balanced development (continued) -0.03 0.08 -0.07 0.004 Syria -0.03 0.16 0.02 -0.015 Morocco -0.01 0.22 -0.01 0.024 El Salvador 0.12* 0.14 0.07 0.028 Guatemala -0.06 0.24 0.01 0.014 Peru -0.D1 0.13 -0.00 0.104 Costa Rica -0.06 0.12 -0.07 0.020 Jamaica 0.06 0.13 0.05 0.000 South Africa -0.05 0.08 0.02 0.039 Spain -0.03 0.07 0.13* 0.127 Greece -0.04 n.a. n.a. 0.089 Ireland III. Industry specialization -0.21 0.23 -0.l2 -O.OlD Kenva -0.10 0.10 -0.16 0.025 Egypt -0.57 0.35 -0.11 0.033 Taiwan -0.56 0.30 -0.09 0.007 Yugoslavia -1.03 1.35 -0.34 0.004 Hong Kong n.a. n.a. n.a. 0.039 Singapore HighF -0.31 0.12 -0.00* 0.057 Pakistan -0.69 0.23 -0.00* 0.076 South Korea -0.30 0.14 -0.12 0.143 Tunisia -0.70 0.32 -0.17 0.050 Portugal -0.31 0.04 -0.06* 0.208 Lebanon -0.28 0.21 -0.10 0.130 Israel c. Chcncry and Syrquin data base, extended through 1975, from IBRD Data Bank, April/May 1978. d. Chenery and Syrquin (1975, table 16). c. Sums may not add due to rounding. f. 1963-73. 34 ECONOMIC GROWTH AND STRUCTURAL CHANGE In the average pattern industry reaches the level of primary output at about $300. The bulk of the increase in output and employment comes from the primary sectors in the earlier phase and later from industry and services. Each strategy can be characterized in relation to this average sequence as accelerating the advent of later-stage activi- ties in some respects and retarding them in others. Primary specialization Virtually all countries start to develop by specializing in exports based directly on their natural resources; only later do they develop exportable manufactured goods and services. 36 The strategy of con- tinued primary specialization maintains this pattern during at least the first half of the transition to take advantage of favorable export possibilities. Except for three oil exporters-Iran, Indonesia, and Nigeria-the fifteen countries listed in table 1-3 as having followed this strategy in the 1960s are all small. Only very small countries Or those with very favorable mineral exports have maintained a signifi- cant primary specialization in the later phase of the transition. ll1e strategy of primary specialization originates more in the re- source endowments of a country than in deliberate policy choices. In virtually all cases the initial development of primary exports as well as of the supporting infrastructure was based largely on foreign invest- ment. After the initial investment, debt service and transfer of profits usually lead to a net outflow of capital in countries following this pattern. Primary specialization can also be characterized as a strategy of deferred industrialization. This description implies that industry will be developed after income levels and investment rates have already risen as a consequence of the growth of primary production, as in the cases of Sweden, Denmark, Canada, or Australia in an earlier period. In this sequence higher wage rates inhibit development of the most labor-intensive exports, but relatively high investment rates permit the country to develop other types of manufactured exports (or services) when it becomes necessary to augment its primary exports. The desirability of continuing to rely on primary exports as a source of foreign exchange depends largely on the supply of the underlying natural resources, the growth of world demand, and the need to dis- 36. III the UNCTAD "A" classification of exports followed here, the initial process- ing of raw materials is treated as a primary export. TRANSITIONAL COUNTRIES 35 tribute the benefits of growth more widely. In the period following World \Var II, world demand has been relatively buoyant for petro- leum and most minerals and sluggish for many agricultural products. The relative growth rates of the countries in group I have been signifi- eantly affected by these differences in export markets. The four examples of primary specialization shown in figure 1-8 illustrate some of these characteristic features and pose some of the main policy issues. Iran has reached quite a high level of GNP through specializing in mineral exports. Venezuela (which is not shown) fol- lowed a similar strategy a decade earlier. Both have developed manu- facturing for their domestic markets on the basis of the high savings and investment rates made possible by their oil exports. Their indus- trial sectors must be highly protected, however, so long as the exchange rate is based on their existing exports. In Venezuela, the limits to the domestic market have already produced a slowdown in growth despite ample supplies of capital and foreign exchange. Malaysia and the Ivory Coast are among the most successful exam- ples of development based mainly on agricultural exports in the early phases. 111ey have followed generally outward-looking trade policies that have stimulated new primary exports and have also allowed manufacturing to grow in response to domestic demand without ex- cessive protection. This strategy has led to a broadly based pattern of primary specialization that has been less vulnerable to fluctuations in the prices of individual commodities as compared with countries that are more narrowly specialized. A major drawback to the strategy of primary specialization is the tendency for its benefits to be concentrated in a narrow segment of the economy unless the government intervenes to spread them more widely. Sri Lanka is perhaps the most notable case of a government that has used a substantial portion of the revenues generated largely by taxing primary exports to subsidize the consumption of the poor. As indicated in chapter] 1, this policy has led to a considerable im- provement in income distribution, literacy, and life expectancy, but it has been at the cost of a diminishing rate of growth of GNP and grow- ing unemployment. 'nlere have been three major difficulties with the strategy of con- tinued primary specialization: (a) initial concentration of growth in a few sectors; (b) continued dependence on foreign investment; and ( c) problems in developing manufactured exports after wages and in- comes have risen. Figure 1-8, Examples of Primary Specialization, 1950-75 Sri Lanka Ivory Coast Malaysia ran )i ~ ~ ~. : I ~ v Ep o£J !1 __ ~ v Ep ~E. til 100 .~(200) J V. 1,- V, ~ V' v. , V. VV .. ~ '"0 I., '·Vm I # " · Em+E, #': I : :,1 I E L/l 'r. "II : "E~,+E, I :, \0 PI/ 1 E +E, E. I t ,/ l' I t ~ ' : Itt 0- W ...... t ,. VmA /1: 1 · -" I .. t , 0\ .a t Vm 1 i' ... ~ V..A t " ! ! #' ~ : I : I '0 '"0 10 I-- , F I,/ vi )' ;:) "'"" \0 0- -< · ,. I /. Em+E, ./ I ,, /~: : : ~ .r; '0. e ... f)1 I I I/ ... 1960 Eo+E, I tt: '/ .1975 1/ t I I 100 100 300 300 :;00 800 (200) (600) (600) (600) (1,600) Per capita GNP in 1964 U.s. dollars (1976 dollars in parentheses) Source; World Bank data files. TRANSITIONAL COUNTRIES 37 Since most of the countries that have faced these problems in the past have shifted to a more balanced development strategy, these problems are taken up below. It is worth noting here, however, that the policy of late industrialization has led to quite rapid industrial growth in several of the SP countries. Malaysia and the Ivory Coast in particular have reached higher levels of industrial output than a num- ber of similar countries that focused on industrialization at an earlier stage without developing an export base to support continued growth in GNP. Balanced development Category II includes twenty-three countries that are not markedly specialized in either primary production or manufacturing. Starting from a pattern of primary specialization, thcse countries have become less specialized for one of two reasons: either their resource base or external demand was inadequate to sustain continued expansion of primary exports, or they chose to industrialize to avoid some of the undesirable features of continued primary specialization. 'DIe countrics following a deliberate strategy of import substitution (group II-A) can be distinguished from the remainder (group II-B) by the policies they have used to stimulate industry and by the extent to which the structure of production and trade has been distorted. The import-substituting industrialization syndrome has been well de- scribed by Little, Scitovsky, and Scott (1970) and Bruton (1970).37 The main features of the syndrome are high protection for domestic industry (rates of effective protection usually in excess of 50 percent and sometimes over 100 percent), overvalued exchange rates, and a consequent lack of incentive to export manufactured goods. Over time these policies lead to relatively low levels of exports, diversion of resources from agriculture, and ultimately a slowdown in the growth of industry and of GNP as the possibilities for import substitu- tion are progressively exhausted. Because of this market limit, the strategy of inward-looking development usually succeeds in elimi- nating the specialization in primary production but not in achieving industrial specialization or manufactured exports. In contrast to the distortions produced by prolonged protection, 37. Industrialization policies of most of the countries in group II-A have been described in the case studies that form the basis for these two sUIveys. 38 ECONOMIC GROWTH AND STRUCTURAL CHANGE the typical country in group II-B shifted away from primary specializa- tion with much less reduction in the openness of the economy or in export incentives. The difference hetween the two groups is one of degree, since all countries make use of protection to some extent to offset infant industry limitations. In the later stage of the transition the clearest distinction between the two lies in the development of manufactured exports, which follows after the shift from inward- looking policies to export promotion. Since import substitution is a more effective policy for large coun- tries because of their larger domestic markets, they make up the hulk of group II-A. TIle characteristics of import-substituting industrializa- tion (lSI) will therefore he illustrated mainly by the experience of large countries. Although all of the countries in group II-A suffered considerably from the economic distortions that characterize the lSI strategy, the four illustrated here-Brazil, Turkey, Colombia, and Mexico-have modified this policy sufficiently so that the rate of growth since 1960 has been fairly rapid. Figure 1-9 shows the characteristic reduction in growth of primary output and exports and very late development of manufactured ex- ports in all four cases. The effect of shifting to a more outward-looking trade policy in the mid-1960s is shown hy the subsequent rise in manu- factured exports. For Turkey and Mexico, however, the principal factor offsetting the trade bottleneck has been the growth of service earnings (tourism and emigrant remittances) .38 The other six countries in the import-substitution group in table 1-3 failed for long periods to develop a feasible strategy for the trans- formation of their economy and consequently their growth has heen relatively low. 89 Despite the limited success of many countries that have tried to accelerate industrialization through protection, the rapid growth of several major countries modifies somewhat the general condemnation of this strategy. For example, it is not clear to what extent Brazil's recent success in exporting manufactured goods was dependent on its 38. As shown by Little, Scitovsky. and Scott (1970), rVlexico has had a lower degree of effective protection and may be considered a borderline case between groups II·A and II-B. 39. Ecuador has recently been rescued by oil exports. Most of the other coun- tries in group II-A have moved away from the lSI strategy to some extent since 1970. Figure 1-9. Examples of Import Substitution, 1950-75 '"" 'i3 Brazil I Turkey Colombia Mexico V -B 1Vm ,II · , .~ I:: ~ 1. ill00 I- ,/( 'V : / " .S (200) ::; ~,' : ! v~,~vm · · V.M· ,. P 1. : , V,,::'-'V. '" ~ .' p jv . , ,, "::: : ;. 1 : : V' i ~ ~ "0 vi ;:) v. i · . '" ~ 1 F 1 :: I, r V",.' EvyjEp I I : ,,. ~.t<' ,. F ;0 r-- If J.: 0' V. jlEm+E, Em+E,- ~E. W '<> .... '" ~ ..g 10 I- (20) Epi !i~Em+E' " YlE .. XS:;" , Em+E, ,.·! t I! ": : : : 1 , I I J p E.,+E,. : F .. I CIl ~ 'V' /'" I t ,i I'l'/! : l~ \ ,f 'I ;0 0' .r:l : , ~ 11 t ~, : I .. E~,+E, ~ ~ I· 1950 '6. ~ .... F-\ I:Ii \,1 Fl I i I· .& 1960 197') '" p... \1 i I I 100 300 300 100 300 300 800 (200) (600) (600) (200) (600) (600) (1,600) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Source: \Vorld Bank data files. 40 ECONOMIC GROWTH AND STRUCTURAL CHANGE earlier policy of inward-looking industrialization. io In the long run the most serious consequences of the lSI strategy are likely to stem from the inappropriate technological choices and unequal pattern of income distribution that it leads to, which have so far proved harder to change than the distorting effects on exports. Because of the vvidespread preference of large countries for inward- looking policies in the past, the examples of balanced transformation under more outward-looking policies are mainly smaner countries.41 Of the four examples shown in figure 1-10, Costa Rica best illustrates the effects of outward-looking policies in maintaining primary exports at a fairly high level while facilitating the shift to manufactured ex- ports. Like many other countries in this group, Costa Riea has been exporting more than 10 percent of its manufactured output, in eon- trast to the countries in group II-A, where typical export proportions were less than 5 percent. 42 The promotion of manufactured exports makes it possible to sustain total export growth despite limited in- creases in primary exports. The two main features distinguishing groups II-A and II-B, therefore, are the share of manufactures exported and the level to which total exports are allowed to decline before a shift takes place to more outward-looking policies. Industrial specialization The strategy of early industrial specialization for export markets is an alternative response to a limited resource base. This strategy differs from industrialization for the domestic market in that it requires the development of industries in which the country will be able to com- pete on the world market once production is established. In successful cases this has meant a shift away from inward-looking policies after an initial period of protection of a decade or so. 40. General appraisals of the Latin American experience with trade policies are given by Diaz Alejandro (1975) and Baer (1972). There are no successes with lSI among the smaller countries in Latin America or elsewhere to compare to the four larger countries cited here. 41. Of the four large countries in group H-B, Thailand and the Philippines are at an early stage in which thc pattern of the transformation is not well established. During the 1960s, Spain was rapidly completing the structural transformation after shifting to an outward-looking policy. South Africa is in several respects a special case. 42. The measure used is the ratio of manufactured exports to value added in manufacturing, which is shown in table 1-3. Figure 1-10. Examples of Balanced Development, 1950-75 ~ '-' ,.c; Thailand II Philippines II Pew Costa Rica '1:: ~ '" ! VmJv p _ V·~"Ep 0. ~ .; (200) ..!!! "0 "0 100 V. /.V .V P rl / V ' v.1 · V.~Vm .: V ,.-1 · l{/,· L /,.F ..1 / I 4:- ...... c.fi ::i '<0 Ie-- 0"- ~ V: ,£ ,, , {J.El' , m ,.E",+E, V:.//· V E· " i Vmii 1 E m +E .. 1'. E B p Vm Em+E'r F i tAl 1 ~' II i'l .... '" »ttl , I . ,_, .. Em+E, - :f1 .g c.fi 10 (20) ., ,,, ,:I .. F I ::i / I ~ '<0 0"- E m +E : I' I.J[~'" · 1950 ~ , .a .0. I .A 1960 '" u .... '-' · 1975 I A.. 300 300 100 800 (600) (600) (200) (1,600) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Source: World Hank data files. 42 ECONOMIC GROWTH AND STRUCTURAL CHANGE The potential advantages of this strategy are those predicted by the theory of international trade: a better and earlier selection of new export sectors, no discrimination between domestic markets and ex- ports, and more efficient use of labor and capital. The way these prineiples have worked out in several cases of outward-looking indus- trialization has been discussed by Kecsing (1967), Balassa (1970, 1971a), and Fei and Ranis (1975). However, there is not yet enough empirical evidence to determine the point at which reducing protec- tion and exposing domestic industry to international competition will result in benefits from export expansion that outweigh the eosts of dismantling ineffieient activities. By 1970 the strategy of early industrial specialization had been suc- cessfully pursued by less than a dozen transitional countries, although several others have made periodic moves in this direction. 43 Four of the most successful examples of this strategy-Korea, Taiwan, Yugo- slavia, and Israel-are shown graphieally in figure 1-11.44 Eaeh of these countries started from a disrupted economy after \Vorld War II followed by a decade or so in whieh industry was developed on the basis of import substitution. A substantial inflow of external resources was available to support rapid growth without the necessity of increas- ing exports immediately. In the early or middle I%Os each country shifted from an inward to an outward orientation and actively pro- motcd exports. lbc success of this strategy is shown in figure 1-11 by the rapid in- crease in manufactured exports, the accelerated growth in manufac- turing output and GNP, and diminishing dependence on inflows of external capitaL Other countries in group III also reached high rates of growth once the basis for manufactured exports became well estab- lished. As a result the nine countries with established industrial spe- cialization account for some 70 percent of the exports of manufac- 43. Although countries in table 1-3 are classified on the basis of their export patterns, the strategy of industrial specialization cannot be considered well estab, Iished until manufactured exports reach at least 10 percent of the value added in manufacturing (corresponding to 4 to 5 percent of manufacturing output). In 1973, this was the case for all countries in group III (table 1·3) except Lebanon. Several balanced countries in the late stages of the transition (Ireland, Spain, and Greece) had also reached this degree of specialization. 44. I omit the extreme cases of Hong Kong, Singapore, and Puerto Rico from the discllssion because their experience has less relevance to the more typical countries. -0"'·"'" ~ ... # ~'--r'<'r--~ I _. -~·I ~r-- .~. 500 Korea ! Taiwan Yugoslavia Israel .. Vm (1,000) ~ : ,,-' ! F " ~Em '" . /,,' j +E, ~ 200 .g (400) ~V m ." /1 U') ~ /'\ // ::i \0 l""- s:; 50 Vp .:" :,'! ; ",+E, : .Jf-ff'/'m+E, , '; V, V",.' i ; ,'- · V ' i:". y~: i p~m' ~ (l00) , .F I: V ' ,,,: V,: I I ~ (l~Jr v.fj/~:~' V"::~E' E'+:~~E' ..!:l! , I: ... I: ' I Es :g " I I: : I 4:- w I E, !l 0. F" : ' · i l\i, · I":' F ; i F."'~ \ !:l : i J' \ I I : i l l . . ~ · . · 1 I l' 1 E +E .. '\ l' ~ ! 1 · 1950 Ii: ....> \ I 1 ~ 1960 I I 1 .1975 \r I~Ol! 1 100 300 1 (~~g) ~-:3:~0::- ~I (~,500 (600)Per capita GN . (600) (200) 300 100 (200) 800 i 8 _ _1....._LI (200) 500"" Wodd 8'nkd,,, " " pm 1964 U.S. dol"', (1976 doll'" m p""nth,,,,) . (1,600J (600J (1 ,230) ,000J 44 ECONOMIC GROWTH AND STRUCTURAL CHANGE tures from all developing countries even though they only contain about 5 percent of the LDC population. 45 Since only a small number of countries possessing rather special social and political characteristics have so far succeeded in the strategy of early industrial specialization, it is difficult to identify its essential requirements with any accuracy. These requirements include entre- preneurship, the quality of the labor force, and aceeptance of wage rates that are not so high as to nullify the country's comparative advantage in labor-intensive industries. 46 Although at the outset in- dustrial wages typically have been relatively low by international standards, they have risen in proportion to the rise of productivity as capital has accumulated and surplus labor has been absorbed. 47 Main- tenance of this strategy therefore requires eountries to shift continu- ously to new types of manufactured exports as their comparative advantage changes. 48 Implications This survey illustrates the range of experience that underlies the principal development patterns identified in the second section of the chapter. Further analysis of individual countries is essential to relate the observed structural changes to the policies that produced them. The two approaches are complementary in that cross-country studies determine average patterns of structural change and the extent of intercountry variation, while country studies permit a more systematie assessment of poliey differences. Each of the four strategies for transforming the structure of pro- duction and trade has produced its share of successes and failures in the period following \\Torld War II. Success in sustaining relatively high rates of growth has depended more on an ability to modify trade and investment policies in the light of the results achieved than on the initial choice of strategy. As will be shown in chapter 10, changes 45. This calculation covers group III with the omission of Kenya, Egypt, and Pakistan. 46. So far the requirements for this strategy have been found in the Far East and 'Vfediterranean countries but have not yet emerged in South Asia, Sub-Saharan Africa, or Latin America. 47. See Fei and Ranis (1975) on Korea and Taiwan, Pack (1971) on Israel, and Dubey (1975) on Yugoslavia. 48. This aspect of structural change is discussed at greater length in Chenery and Keesing (1979). IMPLICATIONS 45 in development policies during the 19605 produced a significant im- provement in the gro~1h of a number of countries compared with their previous performance. Although rapid growth can be achieved with quite different alloca- tion policies, each of the main strategies encounters a somewhat dif- ferent set of problems. In the first two approaches-primary special- ization and import-substituting industrialization-policy and resources tend to be focused on a limited segment of the economy, which accen- tuates the tendency to develop a dualistic structure in both agri- culture and manufacturing. ~Iost of the cases of worsening income distribution that are identified in chapter 11 are countries following one of these strategies. The other two strategies-balanced develop- ment and industrial specialization-tend to rely more heavily on market forces and require an outward-looking export policy that makes better use of comparative advantage in labor-intensive exports. Apart from the central1y planned economies, most countries with good or im- proving distributions of income fall in these two groups although their extcrnal policies provide only a partial explanation of this association. In conclusion, to offset the impression of rigid "laws" that is some- times given by statistical estimates of development patterns, I would stress the variation in the timing of different aspects of the transition that has been observed in countries follo\ving different strategies. At the commodity level, while the shift from primary production to industry takes place on the average at an income level of about $300 per capita, it can be as early as $150 in the industry specialization and import substitution strategies and as late as $1,000 or more when countries start with primary specialization. At a given income level, however, the variation in primary production contributes more to this difference in timing than does the variation in the level of industry. At the factor level the variation in timing of the transition is at least as great although it has not been stressed here for lack of comparable information. In a few countries in which growth is widely diffused- Korea and Taiwan, for example-the ending of the phase of surplus labor has come shortly after the shift from primary production to industry. 'This is by no means the case with import-substituting indus- trialization, where dualism and surplus labor tend to persist much 10nger.49 49. Since the measures of dualism and surplus labor are only available for Japan, Taiwan, Korea, and a few other countries, this statement is admittedly a matter of speculation. Chapter 2 Models of the Transition THE GROWING BODY OF COMPARATIVE STUDIES of developing countries provides a broad base for theoretical analysis and policy. However, most existing theory was formulated before the main features of the transition from a traditional to a developed economy had been de- scribed in any detail, and many assumptions underlying these theories are quite different from the typical structural relations that are now observed. To make the best use of the existing empirical material, therefore, it is necessary to reshape and extend earlier models so as to make them more relevant to the processes of development described in chapter 1. Two main approaches to theoretical analysis are currently used by development economists. The earlier approach typically added spe- cial assumptions to neoclassical or classical theories, such as the sur- plus labor assumption of dual-economy models. The resulting models are still quite abstract and amenable to general mathematical analysis. To maintain this level of generality, the economy under study is rarely divided into more than two sectors, and the identification of these two sectors with their empirical counterparts is largely illustrative. The more recent trend in development theory is to abandon the search for completely general results in order to work with models whose properties can be derived by statistical estimation. This ap- proach recognizes the need to disaggregate an economy into several sectors to reflect differences in production and demand functions as well as in trading possibilities. This formulation has been called, somewhat misleadingly, a "planning model," even though it has been used more for analytical purposes than for actual planning. The basic distinction is between empirically based, or inductive, models and more abstract formulations, in which the functioning of the economy is deduced from elementary assumptions about economic 46 ALTERNATIVE MODES OF ANALYSIS 47 behavior. The former lead to hypotheses that can be tested under dif- ferent conditions, while the latter are more difficult either to validate or to disprove. In establishing a framework for policy analysis, it is not so much a question of choosing between these two approaches as of trying to combine the best features of both. Planning models typically consist of linear functions that produce detailed policy conclusions but do not allow for much substitution in production or demand. Neoclas- sical models have the opposite characteristics: excessive flexibility as compared to observed behavior and very general policy conclusions. The papers in this volume try to develop an intermediate form of analysis that is aimed at discovering the characteristic properties of observable processes and their implications for development policy. A typical chapter takes up certain long-term changcs in the economy and develops a model to explain them and to deduce a set of policy implications. Although the preparation of a development program for a single country might include many of these aspects, generalizations about development policy can be more readily derived from simpler and more specialized models. Alternative Modes of Analysis In developing a class of models to analyze the characteristic prob- lems of the transition, I have started from empirical studies of the kind illustrated in chapter 1. To discover the policy implications of these phenomena, it is first necessary to specify a model structure that can reproduce the main features observed in typical economies. This aim can be achieved through the choice of technological and behavioral relations to be included in the model and through the institutional limits placed on the values that can be taken by its policy variables. In proceeding from an analysis of observed behavior to a determi- nation of policy, it is necessary to specify the initial conditions from which the economy starts and the ways in which changes in policy variables will affect future economic performance. Most development policy can be characterized as "second best," in the sense that it ap- plies to situations in which institutional limits prevent the achieve- ment of the optimum results defined by neoclassical theory. A basic eonclusion of the "Theory of Second Best" is that policy implications 48 MODELS OF THE TRANSITION derived as corollaries of the unconstrained optimum do not neces- sarily apply to the constrained situation.1 This conclusion makes it impossible to accept the neoclassical model of general equilibrium as a sufficient basis for development policy without first examining the consequences of significant departures from its underlying assumptions. In developing alternative modes of analysis, however, the neoclassical model of growth and resource allo- cation does provide a useful point of departure, since it is the only system that has been worked out in detail. I shall discuss in turn the need for specifying additional objectives and constraints, the result- ing possibilities of disequilibrium, and the characteristics of the alter- native approaches that have been formulated to deal with the "devcl- opment case." Objectives and constraints TIle conventional policy objective of maximizing consumption or welfare over time can be broken down into several components to fit the policy needs of developing countries. First, it is useful to distin- guish between more immediate and more remote development objec- tives: (a) the growth of output or consumption over a medium-term planning period of, say, a decade; and (b) the creation of a structure of investment, production, and trade that will facilitate further growth. Relative to the properties of formal growth models, the latter objective corresponds to "getting on the turnpike" -that is, establish- ing an economic structure that will maximize future progress toward specified consumption and distribution objectives. Several examples are noted in chapters 1 and 10 of countries that have been willing to sacrifice near-term growth to develop more viable structures of pro- duction or international trade for the future, and of others that have pushed rapid growth at the expense of needed structural changes. There are clearly tradeoffs between these two sets of objectives. A second differentiation of the general growth objective involves some breakdown of income recipients by level of income, function, or region. In the formulation of policy models, these distributional 1. The discussion of second-best theory was initiated by Lipsey and Lancaster (1956,1959), McManus (1959), and Mishan (1962); its principal application to development policy has been in thc field of trade. ALTERNATIVE MODES OF ANALYSIS 49 dimensions can be considered either as aspects of the welfare func- tion to be maximized or as constraints that need to be satisfied (for example, minimum income growth for given poverty groups or re- gions). These constraints serve to rule out growth processes the benefits of which are too heavily concentrated in one segment of the population. The constraints that need to be incorporated into a model of devel- opment comprise whatever factors limit the achievement of these three aspects of welfare maximization: (a) aggregate income growth in period T; (b) structural change needed for further growth; and ( c) the desired distribution of its benefits, however defined. For exam- ple, a given input needs to be identified explicitly whenever the possi- bility of substituting other commodities or factors for it-either directly or indirectly-is limited. High substitutability thus makes it legitimate to treat all tradable goods as a single commodity for many purposes. Commodities and factors that are commonly distinguished in development models because of limited substitution possibilities include capital stocks, skilled labor, agricultural output (which is not completely tradable), and overhead facilities. The limits to the supply of these commodities and factors change as resources are devoted to increasing them. It requires time as well as resources, however, to transform unskilled into skilled labor, to redi- rect investment from agriculture to manufacturing, to adopt new technologies, and to move workers from rural to urban locations. This time requirement is recognized in concepts such as absorptive ca- pacity or learning by doing. Although they cannot be measured with great precision, such limits to flexibility and change are important features of models that try to explain development phenomena. Possibilities of disequilibrium The need to develop specialized models for the analysis of develop- ment policy is closely related to the possibilities of persistent disequi- librium in commodity or factor markets in the course of the transition. In the advanced countries, the identification of a realistic set of cir- cumstances that could cause persistent unemployment of both labor and capital led to the formulation of the Keynesian model for short- term analysis. 111e usefulness of this model for policy analysis stems from thc fact that it focuses attention on the source of the disequi- 50 MODELS OF THE TRANSITION librium conditions and the differences In economic behavior and policy responses that they produce. Several of the characteristics discussed in chapter 1 may either pre- vent a developing economy from reaching equilibrium over extended periods or cause economic equilibrium to be politically or institu- tionally unsustainable. In addition to time lags, the main characteris- tics that may lead to persistent disequilibria are the properties of the demand and production functions that form the basis of the neoclas- sical system. In general terms, high elasticities of substitution in de- mand and production facilitate the adjustment of the system to shifts in internal and external demand and the uneven growth of factor supplies, while low elasticities make the adjustment more difficult. Three sets of substitution mechanisms exist in an open economy: direct substitution among components of domestic demand, direct substitution among factors of production, and indirect substitution among commodities and factors by way of international trade. To a considerable extent, these mechanisms are alternatives, so that less substitution in demand can be offset by more substitution among primary factors or in international trade. 2 However, some of the ad- justment mechanisms operate only with very long time lags and it becomes an empirical question whether a significant disequilibrium is likely to develop in commodity or factor markets while the adjust- ment is taking place. Two general types of disequilibrium have been identified in a num- ber of developing economies and have been investigated in a variety of models. These are (a) the persistence of undemsed (surplus) labor, and (b) the persistence of a "trade limit" to increased output that leads to llndemtilization of both labor and capital in some parts of the economy. Both of these phenomena can be interpreted either as failures of the substitution mechanisms to operate fast enough in re- sponse to market forces or as the result of misguided policies that prevent these forces from operating as effectively as they otherwise might. As in the Keynesian case, there has been extensive controversy over the logical necessity for either of these forms of disequilibrium to per- sist. The demonstration, however, that under idealized assumptions of perfect foresight and instantaneous adjustments there would be no 2. These possibilities are examined in chapter 4. ALTERNATIVE MODES OF ANALYSIS 51 trade limits to growth and no surplus labor-either Keynesian or structural--does not reduce the importance of these phenomena for policymakers. On the contrary, it can be argued that it is by con- structing models that explain these and other departures from equi- librium that policies to offset them can be developed and their occur- rence reduced. TIle methodological position taken in the following studies is that disequilibrium should be treated as an observable phe- nomenon; the design of models should allow for its existence rather than exclude it by assumption. Inductive versus deductive models Like Keynesian economics, development economics is built around observed departures from the neoclassical system. Once admitting the possibility of persistent disequilibrium, one cannot then rely on neo- classical assumptions to deduce policy guidance without support from empirical analysis. Since no alternative paradigm of equal generality is available, however, the neoclassical system is commonly used as a starting point for models that elaborate the consequences of em- pirically based assumptions. To analyze the disequilibrium phenomena suggested above, it is useful to' disaggregate the economy into at least two sectors. At this level of abstraction, specific assumptions can be introduced in the form of properties of demand and production functions, and the solu- tions to the model can usually be expressed in a general analytical form. Although this formulation is valuable in explaining the char- acteristics of a given phenomenon, it does not contain the variables and constraints that are needed to explorc policy alternatives in any detail. \Vhen the objective is to analyze a particular economy rather than to explore a particular phenomenon, model formulation can be guided by the availability of data, and there are significant advantages to fur- ther disaggregation. An inductive or planning model is designed to bring out the constraints on resource allocation in a given country and the effects of specified changes in policy. The conclusions of such models may not be widely applicable, however, because they are de- rived from a particular set of initial conditions and a given specifica- tion of structural relations. To reach more general conclusions, the estimates derived for a particular country can be replaced by param- eters that are representative of a class of developing countries. It is 52 MODELS OF THE TRANSITION also useful to simplify those elements of the model that do not cen- trally affect the phenomcna being studied.:! Although the deductive approach of abstract theory and the induc- tive approach represented by planning models have different meth- odological origins, there has been considerable convergence in their concepts and analytical techniques in recent years. The introduction of greater empirical content into theory has blurred the distinction between theoretical models and their planning counterparts. There is a growing body of development phenomena that has been studied in both analytical modes, each enriching the other. Much of this work has been done in the context of two-sector models, as in the case of trade theory. There is little gain from further disaggregation unless empirical estimates of demand and production functions are also in- troduced. This convergence of deductive and inductive methods will be illustrated by tracing the evolution of the analysis of several devel- opment phenomena. The logical subdivision of an economy into two sectors must be based on one primary criterion although several secondary distinctions can be made empirically. Three primary distinctions or dichotomies have been widely used in constructing two-sector models of develop- ment: (a) the use of output for consumption or for investment; (b) the characterization of technology as modern (capital- using) or traditional; and ( c) the role of commodities in external trade, tradable or non tradable. Although the first dichotomy is commonly used as a basis for growth theory in a closed economy, it is less relevant for developing econ- omies because machinery and equipment are supplied largely from imports. Since there is no cmpirical counterpart to an investment goods sector in this case, this distinction has more recently been in- corporated into the third type of model. The subdivision of the economy into modern or capital-using and traditional or subsistence sectors, which was initiated by Lewis (1954), 3. This sequence is illustrated in chapters 5 and 6 of part two and chapters 8 and 9 of part three. ALTERNATIVE MODES OF ANALYSIS 53 has been extremely fruitful because it has also made it possible to associate differences in products (industry versus agriculture) and in savings behavior with the basic dichotomy in a plausible way. Lewis proposed the dual economy model to describe the phase of develop- ment during which labor is moving from the subsistence to the capi- talist sector, "to elaborate a different framework for those countries which the neoclassical (and Keynesian) assumptions do not fit." His assumptions have been formalized and extended in different ways by a number of authors, notably Fei and Ranis (1964) and Kelley, Williamson, and Cheetham (1972).4 The latter introduce a num- ber of empirically based assumptions as to the properties of demand and production functions, so that the implications of various aspects of the dual structure emerge more clearly in a series of dynamic simulations. Two-sector development models for an open economy that incor- porate the distinction between tradable and non tradable commodities were initially derived from interindustry planning models in which tradable commodities are disaggregated to a considerable extent. In- stead of investment (and growth) being limited by the domestic production of investment goods, it is limited by the supply of foreign exchange received either from exports or from external borrowing. 5 This formulation shifts the focus of policy from the artificial issue of balance between the production of investment and consumption goods to the basic question for an open economy of the balance be- tween tradable and nontradable goods, since exports from any sector can be exchanged for imports of investment goods. It is a characteris- tic property of these models that when the supply of essential imports (raw materials as well as investment goods) cannot be increased as rapidly as the available supply of savings would require, investment is 4. Fei and Ranis (1964) stress the surplus labor aspects of the Lewis system, while Kelley, Williamson, and Cheetham (1972) use a dual economy model with differences in savings behavior but neoclassical production functions. In more re- cent work, Fei, Ranis, and Kuo (1979) have used the dual economy framework as a basis for explaining ineome distribution in Taiwan. 5. Early statements of the idea of a separate trade limit to growth are found in Prebisch (1950), Chenery and Clark (1953), Chenery (1955), and Little (1960). Arrow (1954) developed the fonnal properties of the trade-limited model used by Chenery and Clark. This kind of model is discussed in detail in part three of this volume. 54 :MODELS OF THE TRANSITION determined by the "trade limit" rather than by the supply of savings. 6 Since thesc initial formulations, there has been a parallel develop- ment of models of trade-limited growth using both inductive planning models and more abstract neoclassical and Keynesian theory. In less aggregated nonlinear planning models, the trade limit appears in the form of a high shadow price of foreign exchange and a low value for additional savings, but the basic implications of the model are not significantly affected! At a more general level of analysis, similar re- sults are demonstrated by Findlay (1971) and Bruno (1976). Bruno provides a full-scale reconciliation of trade and growth theory in a two-sector framework, drawing extensively on his planning models for Israel. The examples of the evolution of dual-economy and trade-limited models of development illustrate the possibilities for a fruitful inter- action between the deductive and inductive approaches to theory.s Much of the actual analysis of the dual economy has been conducted in the more abstract deductive mode, probably because of the greater difficulty in dividing all productive activities into two categories. rOle solution to this problem will require further disaggregation to more homogeneous groups of producers, as has been proposed by several authors.u r~is will certainly be true if account is to be taken of both trade and production distinctions. In shifting from comparative static to dynamic models, the advan- tages of simulations based on numerical analysis as compared with analytical solutions have become generally recognized. As has been stressed in chapter 1, the developing countries are in a transitional phase of their growth, which is better regarded as a process of adapta- tion and adjustment rather than as a steady state phenomenon. The focus on long-term cquilibrium conditions that is characteristic of 6. The description of this formulation as a "two·gap" model derives from the role of external borrowing in raising both the import and the savings limits. 7. See Tendulkar (1971) and chapter 9 of this volume. 8. A similar progression can bc observed in the analysis of the Keynesian case, in which earlier general formulations of multiplier'accelerator interactions have been replaced by econometric models that capture the effects of lags and the inter- action among eeonomic sectors in a more realistie way. This experience also shows, however, that the analysis of simple models is a valuable step toward more realistic formulations. 9. See, for example, Reynolds (1969). DEVELOPMENT PHENOMENA 55 general equilibrium theory is not consistent with the nature of such transitional phenomena. Instead, development theory should be more concerned with the general characteristics of adjustment processes over periods of ten to twenty years. Several approaches to this type of analysis are proposed in this volume. Development Phenomena The subject matter of development economics has evolved from two main types of inquiry: explanation of phenomena observed in developing countries and exploration of properties of theoretical models. These two sources have opposite defects as bases for policy formulation. Analysis based only on the experience of individual countries restricts the choice of policy to what has already been tried. However, analysis that ignores much of past history-which is char- acteristic of both neoclassical theory and planning models-tends to omit important policy constraints. Some consolidation and extension of these approaches is therefore needed to define a set of phenomena that should be taken into account in development theory and policy. Older branches of economics illustrate the effects that both theo- retical and empirical developments have had on the definition of the subject matter as well as the nature of the conclusions. 10 Development economics shows a number of parallels to business cycle theories at a time when the agenda derived from deductive models was being re- placed by empirically based statements of the problems of income stabilization. Earlier descriptions of cyclical behavior had revealed certain regularities that became the subject of theoretical analysis. At the same time the specification of theoretical mechanisms, such as the cobweb theory and the multiplier-accelerator interaction, led to a search for their empirical counterparts and the subsequent refinement of these hypotheses. As governments began to intervene more extensively to regulate 10. Kuhn (1962) stresses the effect of the prevailing analytical framework or paradigm on the definition of problems for research as well as on the evaluation of the results. In economics the dominance of the neoclassical paradigm has tended to shift the focus of theoretical work away from the problems of empirical origin to the properties of abstract logical systems whose empirical relevance is often of only secondary interest. 56 MODELS OF THE TRANSITION the cyclical bchavior of the economy, the agenda of this branch of economics expanded from the analysis of observed phenomena to assessing the effects of various policy instruments. The current phase of research is characterized by detailed simulations of the expected effects of alternative policy packages and refinement of the under- lying models in the light of observed results. Drawing on this analogy, development phenomena that are derived from theoretical models can be identified with the set of properties that produce them, in the way that a Keynesian multiplier-accelerator model produces a type of cyclical behavior that is determined by the parameters of the system. The turning points of the Lewis-Fei-Ranis models of a dual economy are of this nature, in that they can be de- scribed by the interaction among a set of relations in the model. In this case, empirical observations have also served to evaluate com- peting hypotheses and to refine the underlying specifications. Development phenomena that are derived from observation have been elaborated in a different sequence. TIley lead initially to a search for a plausible type of interaction that might explain them, then to consideration of alternative hypotheses, and finally to attempts to discriminate among these hypotheses in empirical models. 'This se- quence is illustrated by Kuznets's (1955) hypothesis-based On inter- country comparisons-that the size distribution of income worsens in the early stages of development and later improves. This observation has stimulated a substantial body of both theoretical and empirical research. Table 2-1 lists six phenomena that illustrate the sources of the agenda of development economies. Although by no means exhaustive, this list includes some of the empirical foundations of development theory together with a selection of the theoretical formulations that have been based on them. The items in the list range from well- established empirical phenomena, for which several plausible explana- tions have been advanced, to phenomena based mainly on the proper- ties of models, for which verification is being sought in more detailed empirical work. The first two phenomena, development patterns and worsening income distribution, stern from the work of Kuznets and others that has already been discussed in chapter 1.11 Early theoretical analysis 11. See pages 6-20. DEVELOPMENT PHENOMENA 57 Table 2-1. Examples of Development Pheoomena Phenomenon Empirical evidence Theoretical formulations Development Uniform changes in de- Theories of balanced patterns mand, production, growth trade, employment Nurkse (1953,1957) Clark (1940) Lewis (1955) Kuznets (1957) Taylor (1969) Chenery (1960) Worsening "U-shaped curve" Models of distribution income Kuznets (1955) Bacha-Taylor (1976) distribution Adelman-Morris Williamson (1976) (1973) Ahluwalia (1976) Eeonomic Country studies of Japan, Dual-economy theories dualism Taiwan, Korea, Lewis (1954) India, Philippines, Fei-Ranis (1964) Brazil, Kenya, and Kelley-Williamson- others Cheetham (1972) Trade-limited Country studies of Israel, Two-gap models growth Pakistan, India, Chenery-Bruno (1962) Greece, Argentina, McKinnon (1964) Colombia, Turkey, Trade and growth models and others Findlay (1970) Bruno (1976) Scale effects Engineering evidence Unbalanced growth Haldi (1960) Rosenstein-Rodan Cross-country evidenee (1943) Chenery-Taylor (1968) Scitovsky (1959) Chenery (1959) Phases of Turning points develop- Lewis (1954) ment Takeoff Rostow (1956) Phases Chenery-Strout (1966) Taylor (1971) of the relative growth rates of different economic sectors is found in Nurkse's theory of "balanced growth" (1953,1957), which focuses on the need to balance domestic demand with domestic supply because of limited trading possibilities. More recent theoretical work- 58 MODELS OF THE TRANSITION discussed in chapters 3 and 4-tries to explain both the observed uniformities as well as some of the differences in patterns of produc- tion and demand. Kuznets's hypothesis of worsening and then improving income dis- tribution (the "U-shaped curve" hypothesis) has been verified as a general phenomenon of developing countries from cross-section anal- ysis but is not characteristic of all countries. 12 A variety of explana- tions has been advanced to explain this general tendency, and these are now being tested in country studies. TIle formal models proposed so far have a strong normative orientation and have not yet been widely applied to historical analysisY:l 1be next two development phenomena-dualism and trade limited growth-have been identified through the continuing interaction of theory and observation. As indicated above, the two types of model are similar in positing mechanisms that prevent the full use of labor and capital at all times. Empirical verification of these two phe- nomena and the evaluation of competing theoretical formulations has taken the form of examining the observable corollaries of each. The relative constancy of real wages (or elastic supply of unskilled labor) that is the central feature of dual economy models has been verified in a sufficient number of eountries-Japan, Korea, Taiwan, Brazil, India, and others-to be regarded as a widespread if not universal phenomenon. But the choice between the classical and neoclassi- cal variants of the dual economy model requires further empirical research .14 'The theoretical origins of trade pessimism and the phenomena of trade-limited growth have been indicated in the previous sections. Empirical verification has concentrated On two aspects: (a) the iden- tification of countries and periods in whieh the trade limit has been important; and (b) the properties of the underlying relations (export demand, import substitution, external capital supply) that lead to a 12. A summary and evaluation of the most recent evidence is given in Ahluwalia (1976) and in chapter 11. 13. Comprehensive country models are available for Korea (Adelman and Rob- inson, 1978), Brazil (Taylor and Lysy, forthcoming), and Turkey (Dervis and Robinson, 1978). 14. Discussions of alternative fonnulations and some of the evidence supporting them are given in Jorgensen (1966); Lewis (1972); and Kelley, Williamson, and Cheetham (1972). DEVELOPMENT PHENOMENA 59 persistent trade limit. Case studies have shown this limit to be par- ticularly important in countries that have to shift rapidly from de- pendence on primary exports to nontraditional alternatives when the former experience slow growth, when the terms of trade have turned against countries dependent on trade, or when the flow of external capital is curtailed. These findings are discussed at length in part three. Development phenomena associated with economies of scale form a separate category in table 2-1 because they require a drastic depar- ture from neoclassical assumptions. The empirical evidence as to the relative importance of economies of scale for smaller countries is well established. l11eoretical formulations incorporating scale economies have been proposed by Rosenstein-Rodan (1943), Scitovsky (1959), and Chenery (1959) ,15 all of which imply unbalanced growth of different sectors of the economy to adjust to the lumpiness of the required investments. Since market forces may not lead to optimal investment patterns under these conditions, empirical testing of these models is very difficulU 6 The last type of phenomenon listed in table 2-1-development phases or regimes-is the most speculative and least susceptible to empirical verification. Three examples are listed of theories that lead to a sequence of phases. Rostow's (1956) sequence, focusing on the concept of a takeoff, was derived from historical evidence, but his conclusions have been widely disputed (Kuznets, 1963a). The prin- cipal change of phase in the Lewis-Fei-Ranis dual-economy models- from surplus labor to neoclassical conditions-has been identified fairly precisely in Japan and Taiwan and remains a plausible hypothe- sis. The phases in the Chenery-Bruno-Strout trade-limited models are not alleged to follow a particular sequence, since the causes of the trade limit are of shorter duration and may recurP Despite limited success in identifying phases of development that have widespread validity, the concept has considerable appeal as a way of linking the phenomena of transitional growth to those of de- veloped countries. The various sources of disequilibrium that underlie most development phenomena-limited substitution, scale econo- mies, shifts in trade composition, lack of information, and so on- 15. The last of these is reprinted as chapter 5 of this volume. 16. A more general dynamic analysis of scale effects appears in chapter 6. 17. Empirical verification of this model is discussed in chapter 10. 60 MODELS OF THE TRANSITION tend to become less important in mature economies, making the neo- classical framework more acceptable for analysis and policy. With the partial exception of the dual economy, all of these phe- nomena are taken up in the following chapters. In most cases, the problem is given a concrete form by using structural relations taken from a given country or group of countries. An Approach to Policy Economic analysts have responded to the identification of develop- ment phenomena by trying to incorporate them in planning models of increasing complexity. These efforts have provided considerable insight into the interrelations among different policy instruments but have not yet gained wide acceptance because of their requirements for extensive data. Even when these problems can be overcome, the complexity of the analysis hampers generalization and overall policy judgments. A second response has been to incorporate several of the develop- ment phenomena identified above-notably surplus labor and scarce savings and foreign exchange-as initial conditions in a neoclassical model, as in the Little-Mirrlees (1974) and Squire-van der Tak (1975) approaches to project evaluation. This procedure preserves most of the merits of the neoclassical approach to policy, since it only requires the empirical determination of a few critical parameters (notably shadow prices for labor and foreign exchange and the assumed specification of the social welfare function). Project evaluation is by no means a com- plete basis for development policy, however; it can only be applied in connection with a macroeconomic analysis of the underlying de- velopment phenomena. In the present volume, an attempt is made to use the techniques of development planning as a basis for generalizations about develop- ment phenomena. In each case this attempt is made after a more detailed planning model has been applied to one or more countries; these studies are used as a guide to the aggregation and simplification of structural relations. Tl1e exploration of development phenomena based on these simplified models provides an intermediate mode of policy analysis between the rather arid generalizations that can be derived from neoclassical models and the overly specific results of an optimum solution to a planning model for a particular country. AN APPROACH TO POLICY 61 Applications of simplified planning models Simplified planning models can serve several purposes: (a) as a basis for reconciling inductive and deductive modes of analysis; (b) as a means of generalizing tbe results of more complete planning models of individual countries; and (c) as a means of exploring alter- native development strategies for a particular country before a more detailed analysis is undertaken. The first of these functions was indi- cated above,18 and has played an important role in the evolution of planning models over the past twenty years by focusing on the differ- ences in underlying assumptions that lead to conflicts between neo- classical policy and the prescriptions of less flexible planning models. Generalization of the results of planning models is an important step in their acceptance as a basis for policy. The construction of mul- tisectoral models for a number of developing countries has shown that the simple structural hypotheses formulated in the 1950s do not pro- vide an adequate basis for policy, but they have not yet led to alter- native formulations of general concepts to take their place. 19 The most desirable general procedure for the formulation of simple models for policy analysis is an inductive one. This approach assumes that a fairly detailed model has been estimated to analyze one or more of the development phenomena described in table 2-1 for a particular country. Such a model can be specialized for the analysis of a par- ticular phenomenon by aggregating those relations that are not critical to the analysis and varying the parameters that affect the results. In some cases the simpler model can be thought of as a reduced form of the original, arrived at by specifying a number of parameters. 20 It is sometimes possible to estimate the reduced form for a number of countries, which leads to generalizations that would not be feasible with the original version. The construction of models that recognize the possibility of various kinds of disequilibrium opens up a new agenda for policy analysis: ( 1) Since disequilibrium usually represents an inefficient use of resources, what are the means by which it can be avoided? 18. See pages 51-55. 19. A survey of the current state of the art of multisectoral planning is given in Blitzer, Clark, and Taylor (1975). The characteristics of the earlier structuralist approaches to policy are discussed in Chencry (1975). 20. TItis procedure is illustrated in Chencry and Bruno (1962), which appears as chapter 8 of this volume. 62 MODELS OF THE TRANSITION (2) Recognizing that disequilibrium is an observable feature of many developing countries, are there characteristic sets of poli- cies appropriate to different kinds of disequilibrium? (3) Since there may be a tradeoff between short-term growth and long-term restructuring of the economy to promote better income distribution or more sustainable growth in the future, what measures of economic performance are appropriate to tran- sitional economies? These are the types of general questions that can be fruitfully inves- tigated with simplified planning models using data that are repre- sentative of a given situation or type of developing country. In recent years the disequilibrium phenomena associated with dualism, scale effects, and trade limits have been explored to some extent in this way. The following pages contain a few general observations on these topics that are subsequently elaborated in the context of specific problems. Adjustment mechanisms 111e poliey measures that are adopted to avoid disequilibrium and thus achieve more rapid development can be characterized as adjust- ment mechanisms. Tncre is little need to consider such mechanisms in a strictly neoclassical world because the possibility of disequilibrium rarely arises (as when external shocks are introduced). For example, the effectiveness of devaluation in eliminating disequilibrium in the balance of payments has been shown to depend on the elasticities of demand and supply for imports and exports; low elasticities may limit the effectiveness of this policy instrument and require supple- mentary adjustment mechanisms. This observation applies to a wide range of adjustments to both internal and external disequilibria. The flexibility of the neoclassical model stems from several sets of assumptions: sufficiently high elasticities of substitution, shiftable resources, perfect foresight, and no lags in adjustment processes. As these assumptions are modified by substituting empirical estimates of price elasticities and observable limits to rates of change of various economic magnitudes (for example, labor mobility, reduction in con- sumption, and increase in investment or exports), flexibility is dimin- ished and the possibilities of disequilibrium increased. Some indica- tion of the average rates of change in these magnitudes as well as the "best" or most rapid adjustments observed appear in chapters I and 10. Although these empirical modifications to the assumptions of neo- AN APPROACH TO POLICY 63 classical theory may be too pessimistic in some cases, it is more useful to formulate policy models that cover typical developing countries than to limit the analysis to the few that most closely resemble the neoclassical case. It is equally unreasonable, however, to discount the effectiveness of the neoclassical mechanisms because they do not nec- essarily work in all cases; such an error characterizes planning models that ignore substitution among commodities and factors. As con- ceived of here, adjustment mechanisms are needed to supplement market forces when disequilibrium already exists or when the system cannot adjust as rapidly as would be required to meet social objectives. In classifying adjustment mechanisms, it is useful to start from the distinetion between tradable and nontradable commodities and factors that has been stressed earlier. Trade has the effect of replacing the supply limitations for individual commodities by a single limit on the supply of foreign exchange. The most important adjustment mechanisms for most tradable commodities are thus those affecting international trade and capital flows. External borrowing increases flexibility by making it possible to delay the development of exports by five or ten years in countries where growth of produetion for the domestic market can proceed more rapidly. For non tradable commoditics and factors it is nccessary to consider other adjustment mechanisms. Commodities for which the elasticity of substitution is low-as in the case of food or transportation-have often become bottlenecks because of the long lead times involved in increasing output plus the limited possibilities for augmenting domes- tic supplies by imports.21 For such sectors the extra market adjustment mechanisms include building capacity ahead of demand (in the case of lumpy investments) and labor training and institutional changes designed to increase the absorptive capacity for increased investment (particularly in agriculture). \Vhen these measures prove inadequate, intervention in the market through the rationing of imports, food, and other essentials is the commonest means of minimizing the wel- fare effects of disequilibrium. When the possibilities of indireet substitution through trade and borrowing are added to these direct adjustment measures, the poten- tial for aVOiding disequilibrium is greatly increased. Some of the 21. Except in small countries with high ratios of exports to GNP, food is at best quasi-tradable, since the volume of food required greatly exceeds the volume of exports and food shortages can only be partially offset by increased imports. 64 MODELS OF THE TRANSITION prices implied by equilibrium, however-sueh as low wage rates or high food prices-may be politically and socially unacceptable. In such a case the best solution that is attainable may still include some forms of disequilibrium. 22 Economic performance and structural change In the neoclassical economy, an increase in the gross national prod- uet is a relatively good proxy for the growth of economic welfare. Although it needs to be modified for the distribution of benefits by income groups and discounted over time, GNP still provides a useful basis for evaluating alternative policies and even for comparing coun- tries whose income distribution is not too different. In most developing countries, where structural change and the possi- bilities of disequilibrium are more important, the use of GNP to evalu- ate performance needs to be reconsidered. If, as indicated above, large changes in both the productive and distributional structure of the economy are needed in the course of the transition, progress toward these objectives must be allowed for in assessing performance, along with the growth of output. Although an allowance for structural change may not make much difference over a period of fifty years, it can have a substantial effect on the judgment of development over one or two decades. This general proposition can be clarified by examples taken from the discussion of the transition in chapter 1. If country A is well endowed with exportable natura1 resources, a rapid increase in GNP from low income levels is relatively easy because export revenues in- crease savings and plentiful imports reduce the need for changes in the productive structure. However, the later stages of the transition have typieally been quite diffieult for such countries because the need to change the structure of savings, produetion, and trade has not been anticipated. The early progress of such a country is thus exaggerated if we look only at the increase in its per capita GNP. Conversely, if country B lacks natural resouree endowments it must bring about changes in production and trade at an earlier stage, but is then closer to the conditions needed for sustained growth at a lower income level than the first type of country. It is therefore quite possible for country B (Korea or India) to be considerably more "developed" at an esti- 22. The effects of such limits are analyzed in chapter 4. AN APPROACH TO POLICY 65 mated income level of $150 or $200-in the sense of moving toward completion of the transition-than country A (Zambia or Papua New Guinea) at a higher level of income. 23 Similar observations apply to policies associated with improving income distribution. Measures that increase the consumption of the poor without increasing their earning power contribute less to devel- opment in the long run than do policies that increase their produc- tivity and future earnings. It is clear, therefore, that the trade-offs among the three elements of the social welfare function suggested above-level of GNP, development capacity, and distribution-are em- pirically significant. So far little has been done to consolidate these elements into a usable measure of economic perfonnance. Simulation of the effects of alternative policies over several decades provides one approach to this problem. Several of the chapters in this volume follow this procedure in assessing policies to deal with econ- omies of scale, the trade limit, and the reduction of poverty. In addi- tion to its use in evaluating alternative policies within a country, this approach can also be applied to intercountry assessment of the effects of alternative trade and aid policies. 23. The conversion of income figures on the basis of exchange rates may also bias the estimated income levels of primary exporters upward as compared "ith a purchasing-power comparison. Part Two ............................................. ........................... -_ ................ _..... . Internal Structure ALTHOUGH THE ALLOCATION OF RESOURCES is influenced simultaneously by internal and external forces, it is useful to make an initial separa- tion of these factors to study the effects of individual development phenomena. In analyzing the internal structure, therefore, changes in exports and capital inflows are taken as given and foreign exchange essentially is treated as a separate factor of production. This simplifi- cation thus focuses on the problems of allocating labor, capital, and foreign exchange within constraints of given demand functions and technology. In part three this procedure will be reversed: the specification of the internal structure of production will be simplified and the focus will shift to foreign exchange earning, saving, and borrowing over time. There are also onc or two studies in each section that relax this arbitrary simplification to take account of the effects of relative prices on both internal and external allocation decisions. The general problem addressed in part two is the changing alloca- tion of production, foreign exchange, inveshnent, and labor by sector in the course of the transition. In addition to the total supply of each resource, constraints to their allocation are provided by the nature of internal and external demand and by the technological relations specified. TIle phenomena studied include the effects of the changing composition of internal and external demand, the limits to substitu- tion behveen capital and labor, the consequences of economies of scale, and the indirect effects of interdependence among economic sectors. The simplest analytical framework that encompasses this type of problem is an open input-output model combined with various specifications of production and demand functions. Since these studies explore observed development phenomena in relatively simple formulations, I have disaggregated the economy only to the extent 67 68 INTRODUCTION TO PART TWO required by the particular problem and supportable by available data. I have also used various analytical approaches, ranging from simple simulations to integer programming, depending on the nature of the phenomena. Chapter 3, The Process of Industrialization, investigates relations among the patterns of demand, trade, and production that have been briefly described in chapter 1. To analyze the changing composition of industry, manufacturing is disaggregated to the two-digit level of the Standard Industrial Classification to allow for sectoral differences in demand, trade, and production. The model is estimated from intercountry data and explains first the average variation in the pattern of production as incomes rise. Different patterns of trade and capital inflows are then specified to study the effects of the alternative development strategies identified in chapter l. Chapter 4, Substitution and Structural Change, addresses the more general problem of optimal changes in the structure of production and factor use when allowance is made for variation in factor supplies and relative prices. Since the empirical basis for this analysis is more limited, the input-output model is aggregated to four sectors. The model is restated in a neoclassical form by specifying demand, pro- duction, and import functions for each sector and by determining prices endogenously. The central problem is the ability of the economy to substitute labor for capital, either directly or indirectly, under realistic assumptions about trading possibilities and thc elastic- ities of substitution in production and demand. Alternative assump- tions are made about the elasticity of substitution and the supply of external capital to compare the neoclassical formulation to surplus- labor and trade-limited models. The effects of economies of seale on the optimal allocation of capital and foreign exchange over time are studied in chapters 5 and 6. This problem was initially posed by Rosenstein-Rodan (1943), but until recently it has been analyzed only in rather intuitive terms. Chapter 5, The Interdependence of Investment Decisions, formulates a partial equilibrium model that shows the quantitative significances of economies of scale in determining related investment decisions in the mining, steel, and metal working sectors. Economies of Scale and Investment over Time, chapter 6, gives a more general dynamic formulation that identifies optimal investment patterns that charac- terize economies with different relative supplies of capital and external resources. This chapter shows Rosenstein-Rodan's "big push" to be INTRODUCTION TO PART TWO 69 an extreme form of a general phenomenon of unbalanced growth over time. l11e central theme of all four studies in part two is the extent to which resource allocation decisions are constrained by the nature of technology and demand to follow similar patterns to increase total output and welfare. At low income levels, choice is limited by the inelasticity of domestic demand and by economies of scale that make small increment, to capacity inefficient. Increased flexibility can be secured by making efficient use of foreign exchange and by choosing sectors in which labor can be substituted for capital. All of these papers grow out of research into the development prob- lems of specific countries. Chapter 3 originated as an attempt to generalize the conclusions of a study concerning the pattern of Japan's development (Chenery, Shishido, and Watanabe, 1962). Instead of repeating this analysis for other countries, the model was reestimated using intercountry data so that more typical development patterns could be simulated. A similar procedure characterizes chapter 5, in which a model originally based on Chilean data was generalized by using estimates of economies of scale in steel production for a number of producers in Latin America. The other two studies in part two abstract further from the original country perception of the problem to introduce additional elements of a general equilibrium system. Chapter 4 simplifies the production structure of chapter 3 and conccntrates on thc direct and indirect substitution possibilities inherent in a price endogenous formulation. Similarly, chapter 6 simplifies the technology of chapter 5, but extends the time horizon across several decades to explore optimal investment patterns in a more complete dynamic system. Chapter 3 The Process of Industrialization SUCCESSFUL DEVELOPMENT IN VIRTUALLY ALL COUNTRIES has been char- acterized by an increase in the share of manufacturing in total output. This structural change is both a cause and an effect of rising ineome. Industrialization results from the interplay of rising demand for manufactured goods, ehanging factor proportions, trade policies, and technological advance. Although some of these factors are quite simi- lar among countries, others vary with resource endowments and the development strategies adopted. Because of the complexity of these interactions, they have been studied in detail only in individual countries. TIle comparative anal- ysis of chapter 1, however, indicates that many aspects of develop- ment patterns are quite similar among countries and suggests that the main source of variation stems from differences in resources, scale, and trade policies. TIle alternative development strategies identified in chapter 1 are based on this hypothesiS. This chapter develops a model to simulate the evolution of the structure of production with rising income and to study the sources of the observed differences in development patterns. The model uses an interindustry framework in which domestic demand, trading pat- terns, and technology are treated as functions of income levels and resource endowments. An analysis of factor use appears in the follow- ing chapter. To isolate the effects of differences in trade patterns and capital inflows, demand and technology are assumed to vary only with the level of income. This assumption makes it possible to simulate the alternative sequences of industrial expansion that correspond to the 70 A MODEL OF INDUSTRIALIZATION 71 typical patterns of trade and capital inflow identified in chapter l. In addition to explaining the uniform elements in observed patterns of industrial growth, this procedure also explains the sources of indus- trialization and clarifies some of the issues raised in earlier theoretical controversies over the need for "balanced growth." A Model of Industrialization ll1c methodology for the study of industrialization and other devel- opment phenomena that is employed in this chapter has evolved through the interaction of model formulation and empirical testing. Since the choice of assumptions to be retained depends on the results of earlier experimentation, a brief account of the evolution of this approach may be useful. Background and methodology The antecedents to this study cover four phases: (a) a definition of the role played by industrialization in dcvelopment in a formal inter- industry model; (b) an analysis of intercountry variations in the principal structural elements; (c) the usc of interindustry models of individual countries to explain the role of industrialization; and (d) a generalization of the results through policy simulations. My initial analysis of the process of industrialization was an out- growth of a study of southern Italy (Chenery, 1953, 1955). This study used an optimizing model in \vhich the choice between import substitution and export expansion of the major tradable commodities could be analyzed under varying assumptions about factor prices and capital inflows. In this case, as in subsequent experiments for other countries, the variation in the optimum investment program was shown to take place largely in the industrial sector. Tbe main factors affecting the pattern of industrialization were the supply of primary exports, the shadow price of labor, and the extent of external bor- rowing. T11ese elements subsequently provided a basis for the typology of development patterns used in this book. Generalization of the findings of country-based models was based on studies of the uniformity of the underlying relations. These were examined in intercountry studies of demand (Houthakker, 1957), 72 THE PROCESS OF INDUSTRIALIZATION employment (Kuznets, 1957), interindustry flows (Chenery and Watanabe, 1958), and patterns of production and imports (Chenery, 1960). Tbe latter study specified a relation between the growth of each sector of industry and the growth of per capita income, based on the substantial uniformity observed among countries in demand elasticities and input-output coefficients. Estimates of these relations provided a starting point for formulating the sector growth equations used here. Since these early comparative studies were based on pure cross- section analysis, the validity of their use for describing changes over time was a matter of speculation (Kuznets, 1966). The relation be- tween time-series and cross-section estimates of production patterns was examined by Chenery and Taylor (1968) and-for these and other processes-by Chenery and SyTquin (1975). Although in a few cases the differences between the two sets of estimates are sufficiently large to make the cross-section results misleading, this is not the Case for either consumer demand or industrial production, where the two sets of results are quite similar. As indicated in chapter 1, these find- ings tend to justify the use of cross-section estimates as a means of describing the "stylized facts" of development. Refinement of the industrialization model has come from its appli- cation to a number of countries. The initial attempt to describe the role of industrialization in a developing country in these terms was carried out for Japan over the period 1914 to 1954 (Chenery, Shishido, and Watanabe, 1962). Others have applied similar models to Israel (Bruno, 1966), Korea (\Vestphal and Kim, 1977), Taiwan (Kuo, 1979), and Turkey (Ce1asun, 1978). As a result of this work, a sys- tematic treatment of the change in input coefficients has been incor- porated in the present version of the model. The last step has been to establish a typology of development strategies that forms a basis for analyzing the effects of different initial conditions and policies. Country samples based on the typology de- scribed in chapter 1 serve as a basis for specifying some of the relations in the model. These alternative specifications make it possible to study the sources of the differences in industrial patterns in more detaiL This evolution has led to the formulation of different models of structural change at the commodity and at the factor level. The first of these is an input-output model that explains the interrelations A MODEL OF INDUSTRIALIZATION 73 among demand, trade, production, and employment; it can be applied either for historical analysis or to simulate alternative production pat- terns. The second is a price endogenous model \vith neoclassical spec- ifications, which is used to analyze the possibilities for substitution among both commodities and factors. 1 The simulation model of this chapter is designed to answer ques- tions related to the transformation of resource allocation in the course of the transition: \Vhat changes in the composition of output are implied by typical patterns of growth of domestic dcmand, exports, and imports? What effects do differences in the size of domestic mar- kets or the inflow of external capital have on output patterns? How do typical inward-looking and outward-looking trade policies affect the sectoral allocation of resources? Answers to these questions require a sufficient disaggregation of tradable commodities to capture the main effects of changes in comparative advantage and market size. 2 Nontraded goods can be treated in a more aggregated fashion, since they are subject to less variation as a result of alternative policies. In the empirical specification of the model, these desiderata must be reconciled with limitations on data availability. There is a great deal more information on both a cross-country and time-series basis on the components of the gross national product and other aggregates than there is on demand and supply by industrial sector. To take ad- vantage of the greater availability of aggregate data, the model has been specified in two stages: (a) a set of relations that determine the variation in the major aggregates with the level of income, and (b) a set of relations that specify the sectoral breakdown of each aggregate. Solutions for the levels of production and value added are determined as functions of the elements determined in the second stage. There are three exogenous variables in the model: the level of per capita income, the population of the country, and the inflow of capi- tal. These were shown in chapter 1 to be major determinants of 1. In an earlier attempt to synthesize these relations (Chenery, 1969a), all of these aspects were included in a single model. But the greater availability of data on commodity flows led to a more detailed econometric specification of the input- output model and a more aggregated and illustrative specification of the price endogenous model. The latter appears in chapter 4 of this volume. 2. Phenomena such as dualism that require analysis at the factor level are con- sidered in chapter 4. 74 THE PROCESS OF INDUSTRIALIZATION changes in the eeonomic structure. s These three variables are used to estimate typical patterns of growth of the major aggregates over the income range of $100 to $1,500 (in 1964 dollars)--or $200 to $3,000 in 1976 prices-which is taken to represent the transition from an underdeveloped to a developed economic structure. The formal structure of the model consists of five sets of equations that are solved in sequence to determine the pattern of output and factor use corresponding to any combination of exogenous variables. The first step is specification of the aggregate composition of final demand and trade as a function of the three exogenous variables. The second is disaggregation of each of the five aggregate components of demand and trade into twenty-three commodity groups, which are then consolidated into a single vector of net final demands. The third set of equations provides for incorporation of systematic changes in input-output relations. The fourth step is the calculation of output levels by applying the updated input coefficients to the net final de- mands from step two. The last step is determination of value added (or, alternatively, the amounts of labor and capital used) in each sector. The following subsections discuss the general nature of these relations, which are summarized in table 3-1. Parameter estimates are given in Chenery and Syrquin (1979).4 Because the model focuses on the effects of changes in demand, trade, and technology, it constitutes only a partial explanation of the process of industrialization. Factor accumulation and production func- tions need to be added to complete the system. Aggregate relations The relations of the main components of GNP to the three exog- enous variables have been estimated by Chenery and Syrquin (1975) for the period 1950 to 1970, using a sample of ninety-three countries. 3. There is no significant association between population and income level. AI· though the capital inflow is related to both of these variables, it is treated as ex· ogenous because it is greatly affected by national and international policies. 4. The present version of the model has been reestimated in collaboration with Moises Syrquin. The fonnulation of the changes in input coefficients with income level is due to him. The model is used to explain average patterns of industrializa- tion and to compare country experience in Chenery and Syrquin (1979). The technical features of the model are described in Syrquin and Elkington (1978). A MODEL OF INDUSTRIALIZATION 75 'These aggregate relations are taken as the starting point for the more detailed specification of patterns of demand and trade in step two of the formal model. Instead of determining the consumption of food directly as a function of income, for example, consumption is first es- timated as a share of income and then food as a share of consumption. Aggregate functions are estimated for the five principal components of the gross domestic product (Y): consumption (C), investment (I), government expenditure (G), exports (E), and imports (M). Each aggregate then serves as a control total for its sector compo- nents. Aggregate variables are related by the accounting identity for gross domestic product: (3.1 ) Since the model is estimated from cross-country data, all variables are defined in per capita terms. Chenery and Syrquin (1975, p. 38) have estimated each compo- nent as a function of GDP per capita (y), population size (N), and the level of the capital inflow (F).5 Because the shares are interre- lated, the same semilog nonlinear equation was fitted for each: (3.2 ) Sk-a f31 1n Y+f32(lny)2 Yl In N + Y2 (In Np + F, where Sk is the share of the given component in GDP.6 Thc estimates of aggregate demand are shown in figure 1-2 for a country of 10 million. As income rises from $100 to $1,500 per capita, the share of consumption falls from 72 percent to 60 percent of GDP. The major changes in the composition of demand stem from the decline of food consumption from 36 percent to 15 percent of GDP; nonfood consumption, gross investment, and government expenditure all increase. TIle shares of exports and imports (see figure 3-1) show 5. This formulation has been discussed in chapter 1. In this simulation model, total GDP (Y) and per capita GDP (y) are identical. In country applications, Y=yN. 6. The estimates apply to 1970. Since the sample covers the 1950-70 period, time trends are allowed for by the use of dummy variables for each five·year pe. riod. The functional form adopted and the use of the same sample assures that the sum of the shares will add up exactly to unity. 76 THE PROCESS OF INDUSTRIALIZATION Table 3-1. Summary of Simulation Procedure Variables Stage Operation Exogenous' Endogenous I. Determination of aggregate y, N,F C,I,G shares of demand and E,M trade in CDP Ep , Em, E. II. Disaggregation to 23 sectors C* C. of final demand and trade 1*, G* Ii, G i E* '" E* m, E* 8 Ej,Mi M* y,N,F III. Updating of input-output A2 matrix IV. Calculation of production (D*i + T*j) Xi and value-added levels by sector y Vi V. Determination of capital and labor inputs by X*i K;,L; sector· a. Exogenous variables marked with an asterisk are determined in previous stages. b. Adding-up in stage I is implied by the functional form and uniformity of the sample. Control totals are imposed on stage II by adjusting components propor- tionately. Totals in stage IV are properties of the input-output modeL c. This operation is not reported in the present simulations. only small increases with income although they are greatly influenced by the size of the country.7 The elements of domestic demand are not affected significantly by population size or capital inflow. Sector relations The second step in specifying the model consists of disaggregating each of the five aggregates in equation (3.1) into twenty-three com- 7. The figures for CNP and its components are measured at factor cost in U.S. dollars of 1964. A factor of 2.0 can be used to convert to dollars of 1976. A MODEL OF INDUSTRIALIZATION 77 Function Control totals" (2) $11; a + f31 In y + f32 (In yl Y= C +I +G +E - M + Yl In N + Y2 (In N)2 + f F F = M E E = Ep + Em + E. (3) log C i = a + b i log C + b 2 (Iog Cl ~Ci C Ii = di I, G t gi G ~Ii = I (4) log Xi f30 + f3, log Y + f32 log N ~Gj =G for i not in manufacturing ~Mi=M ~Ei = E k , k = p, 111, s. ifk (see text) (9) Xi = !,rij (D j + Tj) !'(n + T i) Y (10) Vi = ViXi ~Vi V Y (11) L,t L/ + 1/ (Xit (12) K/ K;Q + k.,t (Xit modity groups.s Although the simulations are carried out at this level of detail, the results will be aggregated to twelve sectors to simplify the discussion. 8. Estimates based on data available for the 1950-64 period are given in Chen- ery (1969a). Although a complete reestimation of the model did not prove to be necessary, we have modified some of the more significant relations on the basis of more complete intercountry data. The initial estimates were based primarily on a set of fifteen input-output tables, supplemented by demand and export funetions for individual commodity groups. Further details are given in Chenery and \Vata- nabe (1965), Chenery and Ginsberg (1969), and Syrquin and Elkington (1978). 78 THE PROCESS OF INDUSTRIALIZATION Figure 3-1. Structure of Trade 30r-------------~------------------------~ ~ 25 " ...... o --r-- ----- -- - \ Manufactured exports (E m) BU 10 _ Primary exports (Ep) Services exports (E,) \ - ~ 5 .L-------- ................ -_ ....... 300 1,500 (600) (3,000 ) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) The basic equation used to determine the composition of private consumption 9 is of the form: (3.3 ) where C represents total consumption expenditure per capita. The squared term was added to allow for sectors in which budget studies (Lluch, Powell, and Williams, 1977) have shown that the expenditure elasticity changes as incomes rise. For investment and government expenditure, the sectoral compo- sition is assumed to remain constant. The initial distribution was de- rived primarily from input-output sources. Exports are disaggregated in two steps. Estimates by Chenery and Syrquin (1975) are used to break down the total into three categories 9. The disaggregated domestic demand functions were largely estimated on the basis of input-output data, supplemented by intercountry comparisons of budget studies (Houthakker, 1957) and national aggregates. A MODEL OF INDUSTRIALIZATION 79 -primary (Ep), manufactured (Em), and services (Es)-using equa- tion (3.2). These results are shown in figure 3-1 for a country of 10 million population. Each category is then decomposed by commodity using equation (3.2).10 The components are then adjusted to equal total values of Em, and E·. Imports have also been estimated using equation (3.2) for manu- factured goods and equation (3.4) otherwise. In summary, step two results in about eighty nonzero relations that determine the net final demand for each sector. The results for a country with a per capita income of $400 and a population of 10 million are shown in table 3-2, along with the changes and income elasticities above and below that level. The dominant effect on the composition of consumption is the fall in the share of foodstuffs, accompanied by a shift from unpro- cessed to processed foods. The shares of the other principal necessities, clothing and textiles, remain fairly constant, while those of durables and other less essential manufactures increase substantially. The rising share of investment in GNP also increases the demand for construction and equipment. The share of services in total demand remains fairly constant. Taken together, these changes can be described as the de- mand response to rising income. The changes in the average trade pattern shown in figure 3-1 and· table 3-3 are somewhat more pronounced than the changes in domes- tic demand.ll As income increases, the share of manufactured exports rises from about 8 percent at $100 to more than 40 percent of total exports at a per capita income of $1,500. Although the main improve- ment in the trade balance comes from primary sectors below $400 per capita income, they are replaced by light manufactures in the range $400 to $800. Taken together, the changes in the composition of domestic de- mand and trade both produce a relative decline in the demand for primary products and a relative rise in thc demand for manufactured goods. Subsequent analysis will show the way in which these shifts are translated into changes in patterns of production and factor usc. 10. A simpler form is used for sectors other than manufacturing: (3.4) log E. =: fJo + fJl log y + fJ. log N. 11. Alternative trade patterns are shown in figure 3·3. Table 3-2. Demand Response to Rising Income: Base Case Final demand (y = $400) Change in D I tlcome elasticity' Sector C G I D 200-4D0 400-800 Primary Agriculture 32.1 0.64 0 32.7 757 8.10 0.380 0.319 Mining 0.9 0.64 0 1.5 1.03 3.64 1.674 1.777 Total prilllary 33.0 1.28 0 34.2 8.60 11.74 0.418 0.416 Light industry Food 43.2 3.06 0 46.3 2259 40.13 0.967 0.901 Textiles and clothing 27.1 0.98 0 28.0 14.14 26.14 1.014 0.951 \Vood, paper, and printing 8.9 1.45 0 10.4 6.19 15.66 1.305 1.325 Total light industry 79.2 5.49 0 84.7 42.92 81.93 1.020 0.976 Heavy industry 00 Chemicals, rubber, and 0 petroleum 7.6 0.70 0 8.3 4.74 11.09 1.221 1.224 Nonmetallic minerals 0.9 0 0 0.9 0.36 0.56 0.795 0.730 Basic metals 4.3 0.29 0 4.6 3.14 9.07 1.636 1.565 Machinery 1.5 0.64 28.38 305 17.08 37.00 1.184 1.146 Total heavy industry 14.3 1.63 28.38 44.3 25.32 57.72 1.223 1.203 Nontradables Construction 0 1.21 46.25 47.5 26.36 56.07 1.169 1.125 Utilities 22.8 3.07 0.37 26.3 13.77 26.80 1.070 1.014 Services m.2 48.66 4.16 168.0 84.44 167.37 1.008 0.997 Total nontradables 138.0 52.94 50.78 241.3 124.57 250.24 1.044 1.025 Total 264.4 61.30 79.18 404.9 201.41 401.63 Note: Based on simulation nm with a population of 10 million and capital inflow at the value predicted for each in· corne leveL a. Elasticities with respect to per capita GNP were computed by dividing log differences. Table 3-3. Trade Response to Rising Income: Base Case y = $400 Change $200-$400 Change $400-$800 Trade Trade Trade Sector Exports Imports balance Exports Imports balance Exports Imports balance Tradables Primary Agriculture 44.35 15.02 29.33 35.20 11.42 23.78 22.65 20.74 1.90 Mining 2.35 4.22 -1.87 1.54 3.31 -1.77 2.73 12.43 -9.70 Total primary 46.70 19.24 27.46 36.74 14.73 22.01 25.38 33.l7 -7.80 Light industry Food 4.20 7.89 -3.69 3.42 5.83 -2.41 11.11 6.42 4.69 Textiles and clothing 5.92 5.81 0.12 2.99 4.52 -1.52 7.31 4.34 2.96 Wood, paper, and printing 4.l2 3.44 0.69 3.l7 2.70 0.48 9.66 3.54 6.l2 Total light industry 14.24 17.l4 -2.88 9.58 13.05 -3.45 28.08 14.30 13.77 00 Heavy industry ....... Chemicals, rubber, and petroleum 2.32 15.13 -12.80 1.76 11.09 -9.32 5.41 13.50 -8.10 Nonmetallic minerals 0.68 1.34 -0.66 0.42 1.01 -0.58 1.72 0.84 0.88 Basic metals 7.86 5.84 2.02 6.84 4.04 2.80 1l.73 6.52 5.21 Machinery 0.76 25.55 -24.80 0.59 18.44 -17.86 9.55 26.01 -16.46 Total heavy industry 11.62 47.86 -36.24 9.61 34.58 -24.96 28.41 46.87 -18.47 N ontradables Construction 0 0 0 0 0 0 0 0 0 Utilities 6.68 4.80 1.88 5.08 3.55 1.53 8.41 5.39 3.02 Services 17.25 12.34 4.91 13.13 9.l3 4.00 21.68 13.82 7.86 Total nontradables 23.93 17.l4 6.79 18.21 12.68 5.53 30.09 19.21 10.88 Total 96.50 101.36 -4.86 74.l6 75.01 -0.85 111.96 113.60 -1.64 Note: Computed from a simulation run with a population of 10 million and capital inflow at the value predicted for each income level. Totals may not add due to rounding. 82 THE PROCESS OF INDUSTRIALIZATION I nterindustry relations Even at low income levels, substantial amounts of all commodities produced are used as inputs by other sectors of production, and this share rises with income. An analvsis of industrialization must there- fore pay as much attention to interindustry relations as to changes in final demand. The simulation model is based on an initial input-output table and a way of incorporating systematic changes in input coefficients asso- ciated with changes in the income leveJ.12 Several studies of interin- dustry relations have identified systematic changes in input-output coefficients over time. They have shown a tendency for similar changes to affect all uses of a given commodity (Carter, 1970), and for increased intermediate use to be positively correlated with the growth of output (Chenery, Shishido, and \Vatanabe, 1962; Vaccara and Simon, 1968). To capture the spirit of these results, an increase in input coefficients is assumed for relatively fast growing products (for example, energy or steel) and a decrease for slower growing prod- ucts (such as agricultural products) .13 The procedure for incorpo- rating teehnological change is given in Chenery and Syrquin (1979); it resembles the RAS technique in that an original A matrix is updated by pre- and postmultiplying it by diagonal matrices that capture sub- stitution effects and ehanges in value-added ratios. 12. The model began as a modified version of the 1951 Japanese coefficient matrix in which revisions had been made to make it representative of semi- industrial countries at an income level of $400. The Japanese matrix was adjusted by Watanabe (1961) on the basis of comparisons of input coefficients in fifteen countries. The revision procedure used international prices where they were notably different from the Japanese prices of imported goods and average ratios of value added to output in each sector. The input·output matrix that results from this set of adjustments was taken as representative of a country at an income level of $400 per capita. 13. This procedure was developed by Syrquin (1976). \\'bile only illustrative, it produces results that are generally consistent in direction and magnitude with the studies of Japan and Taiwan cited above. In addition to the row change, columns are scaled up Of down to keep the ratios of value added to gross output at their original levels, with one exception. For agriculture, the value added ratio is assumed to deerease as development progresses, in accordance with cross-country estimates. See Syrquin and Elkington (1978). A MODEL OF INDUSTRIALIZATION 83 Solutions The solutions for sector output and value added (or factor use) are detennined in the final steps, in which the exogenous variables are taken from the first three steps. SECTOR OUTPUT. The basic equation for the open Leontief system determines output in each sector of the economy: (3.5) i = 1, ... ,23 where Xi is total production in sector i, Wi is intermediate use, D. is domestic final demand, and T. is net trade. The latter three are defined as: (3.6) Wi = !,a;jXj , j (3.7) (3.8) In (3.6) ail is the input of commodity i per unit of output of com- modity ; and is taken as fixed for each income level. Substituting (3.6) into (3.5) and solving for Xi gives sector produc- tion levels as a function of domestic demand and net trade. (3.9) Xi = !,rij(D + Tj), j j i = 1, ... , 23 where rij is an element in the inverse Leontief matrix [I - A]-I. Value added is then obtained as: (3.10) i = 1, ... ,23 where the value added ratio Vi is assumed to be constant for all sectors except agriculture. Table 3-4 gives the solution to the interindustry model at an income level of $400, based on the normal demand and trade vectors of tables 3-2 and 3-3. The first two columns (Wi and D i ) indicate the relative importance of intennediate and final domestic demand. Subsequent analysis will show that the growth of intermediate demand plays a major part in explaining the changing structure of production. FACTOR USE. Because of variation in capital-labor ratios and produc- tivity changes, the increases in investment and employment by sector Table 3-4. Total Demand and Supply at Income Level of $400: Base Case Domestic Intermediate Final Value supply share demand demand Net trade Production added Xi/eX. + Sector (Wi) (D;) (T;) (Xi) (Vi) (u,) Tradables Primary Agriculture 44.7 32.7 29.4 106.7 72.8 0.877 Mining 13.5 1.5 -1.8 13.1 9.8 0.757 Total primary 58.2 34.2 27.6 119.8 82.6 0.862 Light industry Food 34.5 46.3 -3.7 77.1 23.6 0.907 Textiles and clothing 20.4 28.0 0.1 48.5 23.1 0.892 00 Wood, paper, and printing 26.5 10.4 0.7 37.6 16.5 0.915 ""'" Total light industry 81.4 84.7 -2.9 163.2 63.2 0.904 Heavy industry Chemicals, rubber, and petroleum 22.5 8.3 -12.8 18.1 9.2 0.545 Nonmetallic minerals 8.3 0.9 -0.7 8.5 5.9 0.867 Basic metals 25.0 4.6 2.0 31.7 12.2 0.845 Machinery 8.5 30.5 -24.8 14.2 73 0.357 Total heavy industry 64.3 44.3 36.3 72.5 34.6 0.603 Non tradables Construction 10.1 47.5 0 57.5 26.5 1.000 Utilities 20.0 26.3 1.9 48.2 33.7 0.909 Services 46.5 168.0 4.9 219.4 159.6 0.947 Total nontradables 76.6 241.8 6.8 325.1 219.8 0.950 Total 280.5 404.9 -4.9 680.5 400.0 0.870 A MODEL OF INDUSTRIALIZATION 85 are quite different from the pattern of increase in production. In coun- try applications of this model factor use has been determined from the following type of input functions: (3.11 ) L/ =L/ + II (X/- Xi") and (3.12 ) K/ = Kt° kit (X/ - X,"), where 1/ and k/ are incremental capital and labor coefficients and measurements are made for changes in output from a base year (t=o). Since no representative estimates of these coefficients from cross- country data are yet available, the present simulations stop with the estimation of value added by sector. An analysis of factor use based on more illustrative data is given in chapter 4. The five sets of funetions just described constitute a multisector model of produetion and trade which translates an increase in national product into increases in value added and factor use. Table 3-1 sum- marizes the relations among these functions and shows how solutions to each successive stage provide values of the exogenous variables for the next. Although the model analyzes the demand side of the econ- omy in some detail, it does not include the mechanics of income gen- eration, investment, and consumption; typical consumption and investment patterns are assumed without attempting to explain how they are generated. Validation The model discussed above is designed to explain some of the causes of industrialization and to determine the effects of alternative trade policies and resource constraints. 'Ine next step is to simulate the average pattern of industrialization and to examine the validity of the results. This solution will then be used as a base case to which alternative patterns of resource allocation can be compared. Development patterns are simulated by calculating successive solu- tions in which the level of per capita income (y) is increased from $100 to 51,500 (in 1964 prices). Population size (N) is held constant at approximately its median (10 million), and capital inflow is given its predicted value in relation to y and N. The income range cor- responds to the transition from an underdeveloped to a developed 86 THE PROCESS OF INDUSTRIALIZATION economy.14 The aggregate results for the base case are set out in figure 3-2 and compared to those predicted from regression equations for the same sample of countries. Sector results appear in table 3-5, using re- gressions for a more limited sample for comparison. In interpreting these results, two features of the estimation proce- dures should be kept in mind. First, since the model was calibrated for the middle of the income range, most of the structural relations are more valid for income levels from $200 to $800 than for either extreme. (This is also true of the industry regressions in table 3-5.) Second, since the model focuses on the relations between demand and production, no allowance has been made for systematic changes in the ratio of value added to output except in the case of agriculture. For these reasons, the model can be thought of as simulating changes in production rather than absolute levels of value added. Figure 3-2 shows that the simulation reproduces the aggregate changes in the structure of production with reasonable accuracy. In absolute terms the regression estimates of value added are about 10 percent higher in primary sectors and about 10 percent lower in in- dustry. But the changes in the three principal aggregates with income, as measured by their growth elasticities in the ranges of $200 to $400 and $400 to $800, agree within about 5 percent with the regression estimates. A similar comparison for the main industrial sectors also appears in table 3-5. Here the growth elasticities generally agree within about 15 percent, although in several sectors the absolute differences are larger. Part of the difference is due to the fact that the regression estimates are based on industrial censuses, which omit handicraft pro- ducers, and therefore understate production at low levels of income. In summary, the solution to this simulation model with average values of the exogenous variables produces patterns of structural change that are sufficiently close to the estimates determined from cross-country regressions to justify the use of the model to explain the observed patterns-and also to give support to the interpretation of 14. Chcnery and Syrquin (1975, chapter 1) show that this range covers approx- imately 90 percent of the observed change in the structure of demand, trade, and production. If income levels are measured at purchasing power equivalents rather than from exchange rates-as discussed by Kravis, Heston, and Summers (1978)- the increase in per capita incomes over this range is reduced from 15 to less than 10. PATTERNS OF SPECIALIZATION 87 Figure 3-2. Structure of Production (population = 10 million) 50 r:. I, Part A: Simulated ~ Services " ..... 0 I ~ ---1. -----~=-=-~-----/ ~ 40 .... '" t:: <)J ;:; ... 0. 30 ~ Industry _--+------ .- \----------- ........ ..... ~ o 2 _ _ -utilities 1 10 _ - - - - - - - .-J.---_ _ ~ J ",.-.=::-:-------------- - Ci5 o - ------!I Heavy industry 1~00~----~2dO~0--~3±00~--~~----~~------~I,~500 (200) (400) (600) (3,000) 50r---------------~~~~--~--------------~ Part B: Observed Services ...... o -- ~ ------ --- .-- ,. ............ ." ..... "' ....... --- --- --- .§ Industry > --- --//~ itj 20 \ ----- 1 ]0 ······ ' I Uti,\" __ -- o ---------- (------- 100 200 300 1,500 (200) (400) (600) (3,000 ) Per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Table 3-5. Comparison of Growth Elasticities from Model to Regression Results Growth elasticities Value added: regression model at Lower income range Higher income range y equal to ($200-$400) ($400-$800) co co 200 400 800 Regression Model (4) (5) Regression Sectors (1 ) (2) (3) (4) (5) (6) (7) Aggregate' Primary 1.14 1.l0 1.09 0.48 0.524 0.92 0.45 0.468 0.96 Industry 0.86 0.89 0.89 1.36 1.309 1.04 1.26 1.265 1.00 Utilities 0.96 1.01 1.13 1.22 1.164 1.05 1.21 1.044 1.16 Services 0.99 1.03 l.05 1.10 1.043 1.05 1.02 0.995 1.03 Industryb Light industry Food 0.92 0.90 0.85 1.038 1.068 0.97 0.950 1.031 0.92 Clothing 0.39 0.41 0.45 1.378 1.309 1.05 1.206 1.093 Textiles 0.85 0.75 0.67 1.057 1.239 0.85 0.871 1.031 Leather products 1.84 1.89 1.92 1.120 1.084 1.03 1.000 0.973 1.03 Lumber and wood 1.56 1.34 1.16 1.372 1.589 0.86 1.306 1516 0.86 Papcr and printing 0.65 0.69 0.74 1.519 1.435 1.06 1.481 1.373 1.08 Manufacturing n.e.s. 0.42 0.48 0.45 1.573 1.361 1.16 1.308 1.396 0.94 Total light industry 0.79 0.76 0.73 1.173 1.230 0.95 1.085 1.149 0.94 Heavy industry Rubber products 1.24 0.88 0.66 U62 1.659 0.70 0.956 1.380 0.69 Chemicals and petroleum 1.74 1.52 1.08 1.308 1.503 0.87 U85 1.679 0.71 Nonmetallic minerals 0.97 0.90 0.80 1.285 1.392 0.92 1.134 1.311 0.86 Metal products 1.13 1.13 1.06 1.702 1.706 1.00 1.511 1.609 0.94 00 Total heavy '" industry 1.23 1.14 0.99 1.500 1.607 0.93 1.358 1.557 0.87 Note: Elasticity of sectoral per capita output with respect to per capita income computed by dividing log differences. a. Regression values from Chenery and Syrquin (1975 b. Regression values from Prakash (1978). 90 THE PROCESS OF INDUSTRIALIZATION these patterns as representing average changes over time.Hi There are larger discrepancies at the twenty-three-sector level due to the smaller samples used in the disaggregated estimates. Analysis of these differ- ences must await further work at the country and sector level. Patterns of Specialization There are two starting points for the analysis of industrialization, one stressing supply conditions and the other demand constraints. The theory of international trade focuses on differences in resource endowments and changing factor proportions. This theory explains how the composition of trade and production follows from a given combination of factor supplies, but it does not imply any particular pattern of industrialization over time. The effects of demand limitations, both domestic and foreign, are stressed in the theory of balanced growth proposed by l\' urkse (1959). He assumes that price and income elasticities of demand for exports limit the scope for the operation of comparative advantage and con- cludes that the pattern of growth of domestic demand will be the principal determinant of the pattern of output expansion. 16 Although each of these theories furnishes a useful point of depar- ture for further analysis, each omits important elements stressed by the otherP Since they have conflicting implications-one stressing the similarity of demands and the other the diversity of resourees- the observed uniformities in resource allocation can only be analyzed in an empirical framework that incorporates both sets of factors. The industrialization model will now be used to interpret observed 15. This conclusion was less true of the 1969 estimates of the model, which omitted technological change and did not use control totals for the same sample of countries for all the aggregate variables. 16. The following quotation summarizes Nurkse's basic proposition: "Each industry must advance along an expansion path detennined by the income elas- ticity of consumer demand for its product. This simple idea must be the starting point in any expansion of production for domestic markets in the less developed countries, insofar as external demand conditions do not favor the traditional pat- tern of 'growth through trade'." See Nurkse (1961, page 251). 17. Tile effects of these two starting points on actual development policy are discussed in chapter 7. PATTERNS OF SPECIALIZATION 91 patterns of change in outpUt. 18 Although there is not yet an empirical version of the principle of comparative advantage that links resource cndowments and trade patterns in a satisfactory way, it is possible to infcr this linkage from the trading patterns of different groups of countries. 19 Starting from typical patterns of trade and capital inflows, the model determines the sectoral growth rates that are necessary to satisfy the specified demands. Tbe results show the relations between differences in international specialization and corresponding differ- ences in production patterns. Typical patterns Chapter 2 outlined several ways to generalize about development phenomena. Since such phenomena cannot be adequately represented by the properties of general models, it is customary either to compare the experience of several countries or to simulate alternatives for a· given country. The procedure adopted here combines elements of both approaches: countries are first grouped into the three main develop- ment patterns identified in chapter 1 and then the aggregate relations in the model are estimated for each group. As a result, each type of country is characterized by a separate set of structural relations. The typology used for this purpose categorizes countries according to their size and pattern of specialization in international trade. In small countries, trade is a large proportion of GNP and the effects of specialization are more pronounced. Although trade patterns also vary considerably in large countries, the domestic market is relatively more important. Large countries are treated as a single group for purposes of estimation because of their small number and greater homoge- neity. Tbe separate specifications of the model are therefore based on a classification of all countries into the three groups described in chapter 120 : 18. Although this chapter does not investigate the optimality of the observed patterns of specialization, it does provide a basis for the optimizing models of later chapters. 19. For the manufacturing sectors, Balassa (1977) has recently applied an em· pirical version of the factor proportions explanation of changing comparative advantage to developing countries with promising results. 20, The criteria used in classifying countries include production as well as trade indexes. A list of all conn tries in each group is given in Chenery and Syrquin (1975, table 2). 92 THE PROCESS OF INDUSTRIALIZATION Large countries (L), those with populations greater than 15 million in 1960 (or 20 million in 1970); Small, primary-oriented countries (SP) , those relatively specialized in primary production and exports; and Small, industry-oriented countries (SM), those relatively specialized in the production and export of manufactured goods. The principal advantage of using estimates for country groups in- stead of comparing countries directly is to bring out their common features and thus avoid the peculiarities of individual cases. In explor- ing the empirical content of concepts such as "primary specialization" and "balanced growth," the extreme cases are also valuable in indi- cating the possible range of variation. To preserve this virtue, I have specified an extreme form of each of the three typical patterns, based on the characteristics of representative countries. This procedure gives two sets of values for the exogenous variables for each of the three specifications of the model, or six different cases for comparison. Rep- resentative examples of each case appear in table 3_6.21 Since export patterns account for much of the difference among these patterns, they are considered first. The regression estimates of primary and manufactured export levels for each of the three country types are plotted in figure 3-3, along with the three extreme variants and the average pattern. Since there is a general similarity between cross-section and time-series estimates of structural relations, the former will be taken as illustrative of the latter. 22 The three typical export patterns (Ll, SPl, and SMl) show as much variation as might be expected from trade theory. At the middle income level of $400, the composition of exports ranges from 5 percent manufactured goods in the SPl pattern to more than 50 percent in the other two patterns. The median small country of 5 million persons has a value of commodity exports at each income level that is twice that of the median large country of 40 million. The substantial capital in- flow that characterizes the early stages of the SM pattern is reflected in 21. The examples listed are typical of the three main groups; they do omit the smaller subgroups, such as small balanced (Costa Rica and Peru, for example) and large primary·oriented (such as Iran). 22. See Chenery and Syrquin (1975, chapter 5). In recent years there has been a trend toward higher levels of manufactured exports as a result of policy changes that would modify these relations somewhat. Figure 3-3. Comparison of Observed Export Patterns (1964 U.S. dollars, 1976 dollars in parentheses) 300 ~I- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - , (600) 100 / . (200) ./ / .~ "". "' ..... // \Q W 's. '" u .... OJ Po. ..... V> 15 Po. ----- sPL--/ .._ ............ ; ~~\,k .Sj/~;.7 ·.5-.. . . ; ~ / /'''-- ..c::A Income levels 100 (200) 0 ~ OJ 10 )j;;-a'" Sv-.},-- ,.L:. 200 (400) · >. .... (20) 0-- ______ /" ,/ 400 (800) · '" 8 _ - .,/ 800 (1,600) · ~ 0- -----,,--...-:':-- ...- ........ D-""'"" ,1. " v,' 1"500 (3000) t::.. Ll. Large countries (N = 40 million) ------ .. '--' L2. Very large countries (N = 140 million) SP 1. Small, primary·oriented countries SM 1. Small, industry-oriented countries 1 SP2. Small countries with extreme primary SM2. Small countries with high capital inflow ( 2) specialization 0.1 1 10 100 1,000 (0.2) (2) (20) (200) (2,000) Manufactured exports per capita, Em 3-6. Examples of Trade Patterns, 1965 Per capita Popula- GNP tion (Y) Capital Export Primary Manufactured (N) (1964 UB. inflow share exports exports Country (millions) dollars) (F) (ElY) (EpIY) (EmIY) Large (Ll) Thailand 31 110 0.Ql3 0.18 0.12 0.03 Egypt 29 138 0.025 0.19 0.09 0.03 \Q Philippines 32 149 0.009 0.16 0.10 0.01 '""" Yugoslavia 20 415 0.007 0.22 0.05 0.12 Spain 32 572 0.039 0.11 0.02 0.02 Italy 52 989 -0.026 0.17 0.02 0.09 Semiclosed (L2) India 481 84 0.021 0.04 0.02 0.02 Brazil 81 216 -0.026 Turkev 31 244 0.014 0.06 0.04 0 Mexico 43 434 -0.001 Argentina 22 787 -0.014 0.08 0.07 0.01 Primary-oriented (SP1) Sri Lanka 11 142 -0.004 0.26 0.23 0 Ivory Coast 4 179 -0.034 0.32 0.22 0.03 Ecuador 5 195 -0.002 0.17 0.16 0 Guatemala 4 278 0.028 0.17 0.12 0.02 Chile 9 419 -0.008 0.14 0.10 0.02 New Zealand 3 1,806 0.026 0.22 0.19 0.01 Extreme primary specialization (SP2) Zambia 4 179 -0.170 0.57 0.55 0 Malaysia 9 258 -0.046 0.46 0.38 0.04 Saudi Arabia 7 271 -0.425 0.58 0.53 0 Venezuela 9 830 -0.096 0.31 0.29 0 Industry-oriented (SM 1) Taiwan 12 201 0.033 0.18 0.07 0.09 '-0 Portugal 9 361 0.050 0.26 0.03 0.12 Vi Greece 9 585 0.126 0.09 0.03 0.02 Austria 7 1,052 0.010 0.26 0.02 0.15 Finland 5 1,486 0.Dl9 0.21 0.02 0.16 High capital inflow (SM2) Tunisia 4 198 0.143 0.20 0.09 0.03 Lebanon 2 446 0.208 0.19 0.03 0.01 Puerto Rico 3 936 0.211 0.44 n.a. n.a. Israel 3 1,126 0.130 0.19 0.04 0.08 Source: Chenery and Syrquin (1975, tables 10, 11, 12, 13, and S.3). n.a. Not available. 96 THE PROCESS OF INDUSTRIALIZATION relatively low export levels; the converse-substantial capital outflow and high exports-is true of the SP pattern. The trade patterns of the three extreme cases were developed on the basis of the country examples shown in table 3-6. These cases have the following characteristics: 'Ine SP2 case illustrates the extreme primary specialization that is exemplified by oil and mineral producers.23 TIle L2 case illustrates the large, semic10sed economy, which has low ex- port levels resulting from a combination of large size and inward- looking trade policies. (This case is defined by export and import levels that are roughly half those of Ll.24) The extreme examples of this pattern in 1965 were India, Indonesia, Pakistan, Brazil, Turkey, and Argentina. The SM2 case illustrates the effects of high capital inflow of the magnitude observed in Tunisia, Lebanon, Puerto Rico, and Israel. The regression equations for the SM group of countries show that higher capital inflow is associated with lower primary ex- ports but has little relation to the level of manufactured exports. A preview of the effects of this wide variation in trade patterns on the aggregate structure of production is provided by figure 3-4. There is considerably less specialization in production than in trade because of the similarity of demand patterns in all three groups. There is also some tendency for production patterns to converge at higher income levels as capital inflows and thc effects of primary specialization be- come less important. The following subsections analyze these results in more detail. Since the simulations start from observed trade patterns, they determine the corresponding allocation of resources throughout the economy but do not specify the policies that underlie the observed patterus. Industrialization in large countries The main distinguishing characteristic of the development patterns of large countries is a relatively low level of international trade. The larger the country, and the more inward-looking its policies, the closer it gets to the polar case of a closed economy. The typical large-country pattern has exports of about 12 percent of GNP throughout the transi- 23. Since these countries also have a high capital outflow, the latter character- istic has been taken as the defining property of SP2 in simulating this case. 24. In modeling this case, I have assumed a population level of 140 million, which produces this result without changing the structure of the model. Figure 3-4. Simulated Patterns of Tradable Output (1964 U.S. dol1ars, 1976 dollars in parentheses) Ll. Large countries (N = 40 million) L2. Very large countries (N = 140 m SPI. Small, primary-oriented countries SPI SP2. Small countries with extreme primary speciali:wtion SP2 SMI. indu~try-oriented countries :> SM2. countries with high capital inflow 13 "0 .--L2 ~ ./-;.,.t:.Ll ,,*""'/ 15 .,.-:"./ -0 1 NT '-l e;. " .5 A:: Income levels 100 (200) 0 200 (400) · 400 (800) .. 800 (1,600) · 1,500 (3,000) t;. 800 1,500 (1,600) (3,000) Manufacturing value added V m 98 THE PROCESS OF I~DUSTRIALIZATION tion; the hypothetical quasi-closed economy (L2) has about 6 percent. The range of variation is illustrated in table 3-6. The solutions for cases L1 and L2 at an income level of $400, and a comparison to the average pattern, appear in table 3-7. The typical trade adjustment of large countries results primarily from the fact that their domestic markets are larger, their natural resources are more diversified, and their internal transport costs are higher than those of smal1er countries. All of thcse differences lead to a shift from external to internal markets and from foreign to do- mestic sources of supply. Such economic forces are augmented to some degree by virtually all governments through deliberate policies of import substitution in the earlier periods of industrialization. Al- though these policies are somewhat less extreme today, they were quite pronounced in the 1960s, the period for wl1ich these trade esti- mates were made. 25 The two main features of the trade responses shown in table 3-7 are the lower level of primary exports (29 percent of the normal level in case L1 and 6 percent in case L2), which is largely offset by a reduc- tion in manufactured imports. In the semiclosed economy (L2) lower manufactured exports are offset by a further reduction in manufac- tured imports. The cffects of these trade patterns on the levels of value added in the two cases are shown in the second section of table 3-7. In case Ll the major effect is to reduce primary output by 20 percent (or 4 per- cent of GNP) from the average pattern and to increase heavy industry by 50 percent (also 4 percent of GNP). The grea ter economies of scale and capital requirements of heavy industry favor its earlier introduc- tion in larger economies. 26 Light industry is relatively insensitive to the differences in trading patterns at an income level of $400 because import substitution has been largely completed by this time. The results for the semicIosed economy (L2) represent a continua- tion of these trends. There is a further reduction in primary exports and production, and to a lesser extent in manufactured exports. To 25. Virtually all large countries have emphasized import substitution to a con- siderable extent and the shift toward more export-oriented policies should not be exaggerated. Among large developing countries, only Korea and Yugoslavia now export more than 10 percent of their total manufactured output, and most others are still below 5 percent. 26. The optimal timing of such investments is discussed in chapter 6. PATTERNS OF SPECIALIZATION 99 offset the reduction in foreign exchange availability, there is further import substitution in heavy industry. But the effect on the aggregate structure of production over the course of the transition is relatively small, as shown in figure 3-4. The main effects would appear in the inefficient use of capital and labor that is required by excessive import substitution. Industrialization in small countries The factors that lead to limited trade and balanced growth in large countries produce the opposite effects in smaller countries. The latter are characterized by less diversified resources and smaller markets, which increase the benefits from external trade. Foreign capital is also more available to most small countries and it plays a larger role in financing investment and imports than it does in large countries. All of these factors make it more efficient for small countries to concentrate on the production of a smaner range of tradable goods and to import a larger share of the total. But there is great diversity in the patterns of specialization that are being successfully imple- mented. Although the traditional specialization in primary exports is still the most common among the poorer countries, a growing number have moved away from the more extreme forms of primary specializa- tion and a few have abandoned it entirely at relatively low income levels. The more successful members of the latter group have relied on substantial inflows of external capital during a period of a decade or so in which they were able to develop an industrial base for export growth. Since changes in the comparative advantage of developing coun- tries are determined by the relative abundance of natural resources and increases in skilled labor and capital as well as by the size of the domestic market, it is only possible to draw general inferences about the optimal commodity composition of exports and imports at a given level of income. 27 TIle procedure followed here is to determine the average composition of trade for the SP and SM groups, which have been classified on the basis of their overall pattcrn of specialization. Despite the crudeness of this procedure, there is little doubt about the 27. Changes in comparative advantage with rising incomes associated with changing factor proportions are analyzed by Balassa (1977). Table 3-7. Solutions for Large Countries: Income Level of $400 Case L1: Large Case L2: Semiclosed Case 1: (Index: value (Index: value Average of case L1 of case L2 pattern -+- value of -+- value of Item (Value) (Value) case 1) (Value) case 1) Population (millions) lO 40 140 Capital inflow (percentage of GDP) 1.22 1.62 1.33 1.61 1.32 ...... Trade response c c Primary exports (Ep) 51.20 14.83 0.29 3.08 0.06 Primary imports (Mp) 19.24 11.78 0.61 7.83 0.41 Manufactured exports (E.,,) 21.37 18.22 0.85 Manufactured imports (Mm) 65.00 31.82 0.49 15.54 0.24 Service exports (E.) 23.93 12.95 0.54 4.21 0.l8 Service imports (M.) 17.14 8.87 0.52 4.75 0.28 46.00 0.48 21.70 0.22 101.36 52.46 0.52 28.12 0.28 Simulated value added, bv sector Primary tradables Agriculture 72.79 53.70 0.74 49.91 0.69 Mining 9.79 11.87 1.21 12.33 1.26 Total primary 82.58 65.57 0.79 62.24 Light industry tradables Food 23.57 23.75 l.01 23.97 1.02 Textiles and clothing 23.13 25.13 1.09 24.96 1.08 Wood, paper, and printing 16.45 17.74 1.08 18.42 1.12 Total light industry 63.15 66.62 1.05 67.35 1.07 Heavy industry tradables Chemicals, rubber, and petroleum 9.15 12.68 1.39 14.31 1.56 Nonmetallic minerals 5.93 7.27 1.23 7.36 1.24 Basic metals 12.19 13.56 1.11 14.08 1.16 ..... J 7.25 18.32 22.32 3.08 c ...... Total heavy industry 34.52 51.83 1.50 58.07 1.68 Nontradables Construction 26.53 29.44 1.11 28.94 1.09 Utilities 33.63 33.34 0.99 32.65 0.97 Services 159.60 153.20 0.96 150.75 0.94 Total nontradables 219.76 215.98 0.98 212.34 0.97 Total 400.00 400.00 400.00 Source: Model simulations. Note: Totals may not add because of 102 THE PROCESS OF INDUSTRIALIZATION dominant effect of favorable natural resources on the export pattern of the SP countries. The cross-country regressions also determine the variation in capital flows associated with each pattern and their relations to the compo- sition of imports and exports. In the traditional pattern of primary specialization, the initial investment of foreign capital has been one of the main causes of the early growth of output, which in most cases took place before \VorId \Var II. In the postwar period most SP coun- tries have had a net capital outflow to service these investments. For the smaller group of SM countries that have begun to specialize in manufactured exports, however, there has been a continuing net inflow of capital, largely from public sources, in the postwar period. PRIMARY SPECIALIZATION. For the average SP country, table 3-8 shows that, at an income level of $400, primary exports are about 50 percent greater than in the normal case and primary output is 25 per- cent higher. Apart from food processing, almost all manufacturing is lower than the average because of lower manufactured exports and smaller intermediate demand. 28 These effects are all accentuated in the more extreme case of primary specialization (SP2), which cor- responds to countries like Zambia or Malaysia in the 1960s. As indicated in chapter 1, the pattern of primary specialization can also be described as delayed industrialization. The delay is greater in most sectors of heavy industry. The Chenery-Taylor (1968) regression estimates show longer lags than the simulations, suggesting that the full effects of specialization have not been captured by the present estimates of the trade vectors. To continue to grow beyond the middle income levels, most coun- tries need to diversify their exports because in few cases (apart from oil exporters such as Venezuela and Iran) is the resource base suffi- cient to maintain the pattern of primary specialization. A number of countries (for example, Argentina, Chile, and Uruguay) have experi- enced difficulty in shifting from primary specialization to a more bal- anced export pattern and have undergone prolonged periods of rather inefficient import substitution during which foreign exchange became a major limitation to further growth. T11is lag in transforming the export pattern is so common that it may be considered a typical fea- 28. A third factor leading to lower production levels of both manufactured goods and non traded goods is the export surplus, which is offset by a smaller share of output going to domestic use. PATTERNS OF SPECIALIZATION 103 ture of the SP pattern of development although it can be avoided by suitable policies. 29 INDUSTRIAL SPECIALIZATION. The SM pattern of early industrializa- tion is largely a postwar phenomenon that appears so far in only a dozen or so countries. 30 These countries are characterized by limited natural resources, adequate skilled labor, and access to substantial amounts of external capital from public or private sources. In the typical SM pattern, the inflow of capital replaces the foreign exchange normally earned from primary exports for a period sufficient for indus- try to become established and to develop the ability to export. Even in the most successful cases (Israel, Taiwan, Singapore, and Korea), a substantial capital inflow has been required for more than a decade. The sectoral effects simulated for this pattern of specialization at an income level of $400 are shown in table 3-9. Early industrialization has been concentrated in the light industries,31 a phenomenon that is confirmed by studies of the individual countries concerned. During the period of high capital inflow, both primary output and heavy industry are considerably below the nonn. The case of very high capital inflow (SM2) could also be described as specialization in nontradable production, which is a necessary corol- lary of the high import surplus. The reduction in the inflow of capital that characterizes the second phase of this development strategy re- quires a rapid growth of industry, as will be shown by the analysis of changes over time in the next section. In summary, in the SM pattern foreign borrowing and industrial specialization enable a country to expand its imports and GNP growth more rapidly than would be permitted by the growth of its traditional exports. The possibilities inherent in this mechanism are shown by the fact that earnings from primary exports (apart from petroleum) from all developing countries have grown at less than 4 percent a year in real terms for the past twenty years, although manufactured exports have grown at more than 12 percent. In the SM2 case, the capital in- 29. Most of the countries that are classed as following an import-substitution strategy in 1965 by Chenery and Syrquin (1975, table 16) would have been classed as specialized in primary exports at an earlier date. 30. These countries are listed in table 1-3. 3l. Notable among these industries are textiles and clothing. Analysis of recent trends in these countries by Chenery and Keesing (1979) shows growing special- ization in the skill-intensive branches of metalworking and in finished products in general. Table 3-8. Solutions for Small Primary-oriented Countries; Income Level of $400 Case SP2: Small, Case SP1: Small, primary-oriented primary-oriented varying F Case 1: (Index: value (Index: value Average of case SPI of case SP2 pattern value of ...;- value of Item (Value) (Value) case 1) (Value) case 1) Population (millions) 10 5 5 ...... Capital inflow (percentage of GDP) 1.22 -1.60 -6.03 0 -1>0. Trade response Primary exports (Ep) 51.20 78.42 1.53 97.15 1.90 Primary imports (Mp) 19.24 15.16 0.79 14.68 0.76 Manufactured exports (Em) 21.37 4.33 0.20 3.48 0.16 Manufactured imports (Mm) 65.00 56.50 0.87 54.53 0.84 Service exports (Es) 23.93 9.91 0.4 ] 6.79 0.28 Service imports (M.) 17.14 14.59 0.85 14.09 0.82 Total exports (E) 96..50 92.65 0.96 107.42 1.11 Total imports (M) 101.36 86.24 0.85 83.30 0.82 Simulated value added, by sector Primary tradables Agriculture 72.79 84.21 1.16 91.47 1.26 Mining 9.79 20.18 2.06 25.34 2.59 Total primary 82.58 104.39 1.26 116.81 Light industry tradables Food 23.57 23.89 1.01 23.28 0.99 Textiles and clothing 23.13 20.13 0.87 19.16 0.83 Wood, paper, and printing 16.45 14.35 0.87 13.85 0.84 Total light industry 63.15 0.92 56.29 0.89 Heavy industry tradables Chemieals, rubber, and petroleum 9.15 9.69 1.06 9.80 1.07 Nonmetallie minerals 5.93 5.15 0.87 4.95 0.83 Basic metals 12.19 8.73 0.72 8.64 0.71 7.25 7.09 0.98 6.52 0.90 ..... Total heavy industry 34.52 30.66 0.89 29.91 0.87 <:::> \rt Nontradables Construction 26.53 23.23 0.88 22.36 0.84 Utilities 33.63 30.81 0.92 29.19 0.87 Serviees 159.60 152.56 0.96 145.43 0.91 219.76 206.60 0.94 196.98 0.90 Total 400.00 Source: Model simulations. Note: Totals may not add because of rounding. I 1 ~ Table 3-9. Solutions for Small Industry-oriented Countries: Income Level of $400 1 Case SM 1: Small, industry oriented Case SM2: Small, high capital inflow 1 Case 1: Average (Index: value of case SM1 (Index: value of case SM2 pattern --:- value of --:- value of I , ....... Cl Item Population (millions) Capital inflow (percentage of GDP) (Value) 10 1.22 (Value) 5 8.56 case 1) 7.04 (Value) 5 17.21 case 1) 14.16 0\ Trade response Primary exports (Ep) 51.20 27.72 0.54 11.33 0.22 Primary imports (Mp) 19.24 28.19 1.47 33.30 1.73 Man ufactured exports (Em) 21.37 49.01 2.29 36.36 1.70 Manufactured imports (Mm) 65.00 103.63 1.59 121.64 1.87 Service exports (Es) 23.93 47.72 1.99 69.96 2.92 Service imports (Ms) 17.14 26.83 1.57 31.53 1.84 Total exports (E) 96.50 124.45 1.29 117.65 1.22 Total imports (M) 101.36 158.67 1.57 186.48 1.84 Simulated value added, by sector Primary tradables Agriculture 72.79 54.84 0.75 40.59 0.56 Mining 9.79 5.25 0.54 3.40 0.35 Total primary 82.58 60.09 0.73 43.99 0.53 Light industry tradables Food 23.57 22.11 0.94 22.71 0.96 Textiles and clothing 23.13 34.09 1.47 29.19 1.26 Wood, paper, and printing 16.45 19.41 1.18 18.11 Total light industry 63.15 75.61 1.20 70.01 1.11 Heavy industry tradables Chemicals, rubber, and petroleum 9.15 5.18 0.57 2.83 0.31 Nonmetallic minerals 5.93 5.23 0.88 5.16 0.87 Basic metals 12.19 8.98 0.74 7.50 0.62 Machinery 7.25 3.97 0.55 0.77 0.11 ...... Total heavy industry 34.52 23.36 0.68 16.26 0.47 0 '-l Nontradables Construction 26.53 27.51 1.04 28.74 1.08 Utilities 33.63 38.17 1.14 43.61 1.30 Services 159.60 175.25 1.10 197.39 1.24 Total nontradables 219.76 240.93 1.10 269.74 1.23 Total 400.00 400.00 400.00 Source: Model simulations. Note: Totals may not add because of rounding. 108 THE PROCESS OF INDUSTRIALIZATION flow is sufficient to avoid the necessity of developing primary exports at all, since foreign borrowing covers the gap between import needs and the earnings from manufactured exports. 32 Sources of Industrialization Although there are general similarities in the rise of industry in all the patterns identified, it has been shown above that there are also important differences among them in timing and sectoral composi- tion. Since the simulation exercises of the preceding section have suc- ceeded in explaining most of these differences as stemming from the changing structure of demand and trade, they can be used as a basis for comparing the relative importance of these elements in different sectors and at different points in the transition. The methodology developed for this purpose is based on an alge- braic decomposition of the growth of each sector into four factors which are identified as: (a) domestic demand effects, (b) export ex- pansion effects, (c) import substitution effects, and (d) effects of technological change. Since there are several plausible ways of carry- ing out such a decomposition, it is first necessary to choose a con- sistent set of concepts for use in the present model. The method se- lected provides logically consistent definitions of concepts such as import substitution and balanced growth. There are several advantages to this approach for policy analysis. Toe first advantage is in the provision of a quantitative framework for comparing the effects of different development strategies, both over time and among countries. The second is in the determination of the relative importance attached to elements of speciali7.ation and balance that are present to somc degree in all development strategies; this approach shows how they differ among productive sectors. Decomposition of structural change Relatively stable relations in an economic or social system are com- monly described as its structure. \Vithout a formal model of the un- derlying relations, any observed change in the composition of demand or other economic aggregate ean be defined as a structural change. The formulation of even a rudimentary model of the underlying 32. This mechanism is analyzed in detail in part three of this volume. SOURCES OF INDUSTRIALIZATION 109 processes makes it possible to narrow this definition by distinguishing between (a) changes in composition that are predicted by the model with constant parameters and (b) those that result from changes in the structural parameters.S3 The model as explained so far needs to be extended slightly to distinguish between the two kinds of changes. Ideally, the functions in the model should represent stable relations that can be identified with economic behavior. The specifications of domestic final demand and intermediate demand can be derived from general equilibrium models by assuming constant prices and are thus adequate for this purpose. The sectoral import functions, however, need to be restated in such a way that they can reflect the effects of import substitution instead of being merely a function of income. In a general equilibrium system with endogenous prices, the levels of exports, imports, and production are determined simultaneously. In addition to the relations contained in the present simulation model, a more complete specification would include production costs for each commodity and demand functions for exports. If it were possi- ble to disaggregate each industrial sector into twenty or thirty homo- geneous groups of commodities, a linear programming solution would determine the products to be produced and imported and in this way relate the aggregate import proportion for the sector to the balance of payments constraint.34 If each productive sector is disaggregated in this way, a useful dis- tinction can be made between undertaking new activities and expand- ing old activities. The former can be considered import substitution and identified as a change in the productive structure. 35 \\Then all of the commodities making up a given industry are aggregated, the share of demand that is supplied from imports will remain constant so long 33. This distinction between the structure of the economy and the structure of the model is discussed by Machlup (1963). With a more complete specification of the underlying processes, a larger proportion of the observed change can be explained by the model and a smaller proportion is attributable to changes in its structural parameters. Except where noted, I shall use the term structural change to refer to the structure of the economy. 34. The properties of models that derive optimal trade patterns from this kind of assumption are discussed in Chenery and Kretschmer (1956), Bruno (1967), and Weisskopf (1971), using data for southern Italy, Israel, and India, respectively. 35. \Vith optimal resource allocation, import-substituting activities are only undertaken if the value of foreign exchange increases or the cost of production is reduced. Expansion of existing activities takes placc at constant prices and costs in response to growing demand. 11 0 THE PROCESS OF INDUSTRIALIZATION as no new activities are undertaken and the demand for eaeh com- modity expands at the same rate. The introduction of new activities to produce commodities formerly imported, which is what is custom- arily meant by import substitution, can be measured by an increase in the ratio of domestic production to total supply in a given sector. 11lis definition provides the basis for the present decomposition of sources of industrialization.36 By defining domestic production as equal to exports plus a constant fraction, Hi, of total domestic demand in each sector, the basic equa- tions of the open Leontief model can be rewritten as: (3.13) Xi = Hi (Wi + D;) (3.14) Mi - m, (Wi + D;), where 111t is the proportion of domestic demand supplied by imports and Hi mi = L The solution to the Leontief system appearing above as equation (3.9) then can be rewritten as: (3.15) where the coefficients rlj are the elements in the domestic inverse and t is an index of either income level or time.:17 Based on equation (3.15), the problem of decomposing the sources of growth of each sector can be defined as explaining the increase in sectoral output, Xl, in terms of changes in two sets of exogenous variables (D/ and E/) and two sets of structural parameters (u/ and 1',/). The most direct approach to this problem is to express the increase in output by the increment between two time periods (or income levels): AX = X2 - Xl. Substituting in equation (3.1 3) gives the following expression for the increment in sectoral output: (3.16) 36. The main alternative is to define imports in relation to total GNP rather than to sector demand, as was done in the study of Japan by Chenery, Shishido, and Watanabe (1962) and of Norway by Balassa (1977). Although this defini- tion has some advantages for the analysis of trade, it provides less insight into the sources of industrialization. A comparison of methods of decomposition is given by Syrquin (1976). 37. The input-output table separatcs each input into a domestic and an im- ported component; the domestic coefficient is equal to the domestic component divided by the level of output. In the simulation model, the samc ratio of domestic supply to demand is assumed in each use, so that iiii = Ula,!. SOURCES OF INDUSTRIALIZATION 111 This formulation will be called a direct decomposition because the increase in intermediate demand is not traced to the changes in the four exogenous variables. Solving the Leontief system as in equation (3.15) gives the cor- responding total decomposition formula 88 : (3.17) aX. = ~ i'i/ ul aDj 1 + ~ Till1EI J + ~ T;,/ (DI + WI) .. -50 ( -100) Per capita GNP 120 THE PROCESS OF INDUSTRIALIZATION each sector from proportional growth over the same interval. To facili- tate analysis of the timing of structural change, the deviations apply to the whole period rather than to a constant increment of income. Comparing the two figures, it is clear that demand contributes much more to the explanation of the increase in output by sector than it does to the analysis of its changing composition. In the primary sectors figure 3-5 (part a) shows that increased ex- ports and increased domestic demand contribute equally to output growth in the first two periods. In later periods increasing dependence on primary imports largely offsets the growth of exports. Figure 3-5 (part b) shows that the two trade effects provide the main explanation of the continued decline in the primary share of output in the later periods. Jnlight industry there is moderate growth (from $42 to $52) in the increment for each $100 of GNP, which is concentrated in the nonfood sectors. Since domestic demand for food lags behind the growth of GNP, the rising share of light industry is due primarily to the growth of exports. 43 The rise in output of heavy industry from $16 to $48 offsets the de- cline in the increments of primary output in the average pattern. Domestic demand is less important as a source of growth in this sec- tor, and import substitution more important, than in the other sectors. THE L PATTERN. The sources of growth and structural change in large countries are shown in figure 3-6 (parts a and b). In all sectors they differ from the average pattern in having smaller effects of import substitution and export expansion. Conversely, the increments to output produced by domestic demand are somewhat larger because of the smaner initial proportions of imports. The result is a pattern that resembles Nurkse's concept of "balanced growth" as described above, in which increases in domestic demand provide 65 percent or more of the explanation of the growth of each sector (see part a of figure 3-6). The analysis of the changing composition of output in part b of figure 3-6 shows that changes in trade still play an important role in large countries. Even though exports constitute only 10 to 12 percent 43. Since the increment in food processing remains virtually constant at $20 for each $100 increase in GNP, it is combined with other light industry. But food processing is shown separately in the analySis of factor use in the following section. SOURCES OF INDUSTRIALIZATION 121 of GNP, their changing composition adds substantially to the growth of both heavy and light industry. Import substitution takes place early in the transition in large countries and is of relatively little significance in later periods. Al- though there is considerable variation in the timing of import sub- stitution at a less aggregated level, petroleum refining is the only sector in which it accounts for more than 10 percent of the increase in output by period lII.44 THE SP PATTERN. TIle sources of growth and structural change in small primary-oriented countries are in many respects the opposite of those in large countries. Starting from a higher base, primary exports continue to provide the main source of the increase in exports and in primary output through much of the transition. As a result, indus- trialization takes place later and is due more to import substitution than to export expansion. Thcse differences are brought out by a comparison of figure 3-7's parts a and b to the average patterns of figure 3-5. TIlere is consider- ably less growth of light industry (apart from food processing) be- cause the SP countries do not typically follow policies that favor the export of labor-intensive manufactured goods. Import substitution is also somewhat delayed in light manufacturing and even more so in heavy industry. Because industrialization is deferred, howcver, import substitution and the expansion of the demand for intermediate goods become more important in the later periods and rates of industrial growth are often quite high. In the latter part of the transition the growth of primary exports no longer offsets the limited growth of manufactured exports in the typical SP pattern. As was pointed out in chapter 1, this has often caused a slowdown in overall growth in SP countries but has been overcome in more successful cases by appropriate ehanges in policy. THE SM PATJERN. 'The small, manufacturing-oriented countries re- semble other small countries in their high dependence on trade but are closer to the large countries in the overall composition of their exports. In the majority of cases they are also highly dependent on an (text continues on page 130) 44. By contrast, import substitution is important in chemicals (16 percent), petroleum (42 percent), machinery (14 percent), and transport equipment (19 percent) during this period in the average pattern. 3-6 a). Sources Growth bv Sector: Country Pattern 50 ( 100) ~ OJ Primary (agriculture and mining) Light industry (including food) ~ ':5 t:: ~ 40 80 78 '" ( "'- ,5 C/O ~ '" ~ 30 (60) '" \D I r'~ ry. ~ 20 87 '- N ... '" (40) N ~ 0 '" "'" ~ ~ \D 0- ~ n o 0. '" u ,... "-' p.. I II III IV I II III IV -10 Per capita GNP Per capita GNP Total increment: I, 22.3; II, 16.3; III, 11.8; IV, 9.7. Total increment: I, 44.7; II, 46.4; III, 47.2; IV, 48.0. 50 (100) '" i'A Heavy industry (including machinery) Cj) ..c:: I:: 40 ~Demand Cj) .... co P- .5 .... '" .2i! 67 o Exports '0 "0 'Cl D Import substitution 1- ~ Technological change - 0- .... '" 20 .2i! (40) Note: Numbers on thc bars indicate percentages of total '0 change. "0 ..... N cri W :) 10 ""1- (20) v::> 0- o M Cj) ~ I II III IV -10 (-20 100-200 400-800 800-1,500 (200-400) (800-1,600) (1,600-3,000) Per capita GNP Total increment: I, 26.8; II, 34.7; III, 41.9; IV, 48.5. Figure 3-6 (part b). Sources of Deviations from Proportional Growth: Large Country Pattern 40 \80)1 Primary (agriculture and mining) Light industry (including food) 30 ,t (60) Total deviation 97 :: I -173 co ~ (40) II -29.4 '" ~ 11l -47.4 51 "0 10 IV -56.1 (-~. 24 39 'D t--, ...... N - 0- ~ ~ ::::; 0 ~ .g U1 -10 45 ;::J (-20) "'l- Total deviation 'D ..... 0- til -20 (-40) I 40 37 I II 10.6 14.1 III 17.1 'a til u -30 IV 21.1 ..... u p... (-60) I I II I III IV I II III IV -40 I ( -80) 100-200 800-1,500 (200-400) 1,600-3,000) Per capita GNP Per capita GNP Heavy industry mcmam2 machinery) 29 r:s;] Demand .:: DExports ... '" "'" (3 'V \D D Import substitution 1"'- s: ~Technological change ... if; Note: Numbers on the bars indicate percentages of total "'" (3 'V deviation. ...... -10 t-...I 0 \rI :::> (-20) "I- Total deviation \D 0- I 15.9 ,...; -20 II 31.B III 60.3 IV 78.6 ... OJ P.. Per capita GNP Figure 3-7 (part a). Sources of Output Growth by Sector: Small Primary Pattern Primary (agriculture and mining) Light industry (including food) 40 (80) 78 75 .S 30 i::: ..!!! (60) '0 "'0 \C) I 58 !C- 6'j ..... N - O' .... '" 20 0\ '" '0 '"Cl <.n 10 ;::i (20) ..". \C! ..... 0' .;3 0 'a '" (.) .... "-' ~ -10 Per capita CNP Per capita CNP Total increment: 1,37.8; II, 32.1; III, 25.5; IV, 19.1. Total increment: T, 40.1; II, 42.5; III, 44.2; IV, 46.0. ~ ~ v 40 Heavy industry (including machinery) ..., ..c (80) ;::: v .... '~ " ~Demand .S 69 .... 30 DExports '" .Z:1 (60) "0 Dlmport substitution '" ID t- ~Teehnological change O'- ...... 20 .... (40) '" .Z:1 Note: Numbers on the bars indicatc percentages of total "0 change. ..... tv '" rIi '-l ~ "t- ID 0'- ...... n o 's. 13 .... v 1=4 I II III IV (-20) ' -10 .......... '_,..'" _~_'."" .~~ I ~~~ .I -~~ Per capita GNP Total increment: 1,13.9; II, 22.0; Ill, 31.6; IV, 43.3. Figure 3-7 (part b). Sources of Deviations from Proportional Growth: Small Primary Pattern 50 .~ (100 )1 Primary (agriculture and mining) Light industry (including food) '" C ..... r--. 0' N ~ 00 '" ~ ,.!!l a "0 v5 00 ::i ~ ~ T 37T6 Total deviation 49 >C (-40) -'0 I 11.9 :::: -30 II 16.8 2 III 23.3 is.. (-60) IV 33.2 '" v .... -40 '" P.. (-80) I I II III I IIV I I II III IV -50 ( -100) , ~ I )vvv- /,vvv J Per capita GNP Per capita GNP 41 \. ~~v II Heavy industry (including machinery) I\. 't I ~ E:;3Demand .::: DExports '" D Import substitution ~ '" ::g (40l I 55 'V 'D r~ ~Techllological change 2:: '" ~ Note; Numbers on the bars indicate percentages of total r ...!!! '- '0 s deviation. N \0 'V '" L ~ i -20 (-40) Total deviation I 93 II 25.5 -30 III 63.9 (-60) lV 137.7 .... v -40 0..( I II III IV -50 Per capita GNP 130 THE PROCESS OF INDUSTRIALIZATION inflow of external capital in the early stages of the transition. The result is a unique development pattern whose properties are less easily predictable from general economic reasoning than are those of the other two patterns. There are several significant effects of the early dependence on ex- ternal capital. The total output of tradable goods is lower and non- tradables correspondingly greater. To permit a reduction in the relative importance of capital inflows, the trade adjustments are greater and demand effects considerably less in both types of industry. In success- ful countries this trade adjustment leads to very high growth of indus- try in the middle and later stages of the transition. In comparison to the SP pattern, there is a shift from primary prod- ucts to light manufactures as the main source of export expansion, which is shown in parts a and b of figure 3-8 in the relatively high dependence on export growth and above normal increases in total output. Tbere is less difference between the two sman-country pat- terns in heavy industry because both are affected by the small domes- tic market in the first two periods. In later periods, however, the SM countries become exporters of heavy industrial products to a much greater extent than the SP countries, as shown in figure 3-8 (part b). The need to develop light manufactured exports has led the SM countries to follow more outward looking policies early in the transi- tion. When these have been achieved, they have made the later stages easier than in the case of most SP countries, which typically industrial- ize behind protective barriers that inhibit the subsequent growth of exports. Changes in factor use The preceding discussion has been conducted at the commodity level to analyze the relations between trade and production. To high- light the implications for resource allocation, it is necessary to restate the results in terms of value added, which can then be translated into increases in factor use. 111e methodology for this extension appeared in the first section of this chapter. The levels of value added in each sector are derived from value-added coefficients, which are assumed to remain constant in all sectors except agriculture in the cross-country version of the model. The decomposition of value added follows directly from the decom- SOURCES OF INDUSTRIALIZATION 131 position of output; the relative importance of each source of growth is the same in each sector except agriculture. This translation from output to value added has the advantage of decomposing an increment in GNP by sector, thus permitting compari- sons to be made among the increases in value added and factor use by sector in each period. Such a breakdown of an increment in GNP of $200 is shown in table 3-11 for period II (pages 138-39). The non- traded goods-social overhead and services-are included to give a complete decomposition of marginal resource allocation. The sources of the changing importance of each sector over time are compared in figure 3-9. This figure shows the composition of an increase in GNP of $100 in each period. Thc decline in the increment in value added in primary sectors is offset by increases in light indus- try, heavy industry, and social overhead facilities. In each case shifts in exports (and to a lesser extent, import substitution) augment the effects of shifts in demand patterns. 45 The extent to which the incremental allocation of resources differs among the three patterns is shown in table 3-11 for period IU6 The effects of specialization can be highlighted by comparing the percent- age of the increment in GNP allocated to eaeh sector to the average allocation. In the sectors in which each specializes, the SP pattern allocates an additional 6 percent of GNP to primary production (com- pared with 13 percent in the average pattern); the SM pattern allo- cates an additional 5 percent of GNP to light manufacturing (com- pared with 12 percent on the average) and 4 percent more to services; and the L pattern allocates an additional 5 percent to heavy industry (compared with 12 percent in the average pattern) . The sources of these differences have already been discussed for the tradable goods; the translation into value added merely makes them comparable across sectors. The larger share of non traded goods in the SM pattern is a reflection of the high capital inflow and the greater supply of traded commodities from imports. This share is re- duced as the import surplus declines in later periods. (text continues on page 140) 45. Food processing is shown separately from other light industry because in this case the decline in incremental demand is offset by a rise in incremental ex· ports, leading to a constant increment in value added. Technological change is omitted since it has been specified on a largely hypothetical basis. 46. This table supplements the analysis in tables 3·7 through 3-9, which com· pare the output composition of the major patterns at an income level of $400. Figure 3-8 (part a). Sources of Output Growth by Sector: Small Manufacturing Pattern 50 (100) '" ~ Primary (agriculture and mining) Light industry (including food) OJ ,.,::: "E 40 ~ (80) '" P< .8 W ~ '" 30 :g (60) '-0 t·- ..... s:; W N 20 '" '"' ..!::! :g en ;:) "'¢" '-0 0' ...... .... OJ P... I II III IV ~10 ) Per capita GNP Per capita GNP Total increment: I, 21.5; II, 13.9; III, 8.3; IV, 5.3. Total increment: I, 49.6; II, 53.5; III, 55.0; IV, 54.6. Heavy industry (including machinery) ~Demand .5 o Exports ~ ..!ll '"0 30 (60) o Import substitution '" 'D r- ~ Technological .... '" 20 ~ Note: Numbers on the bars indicate percentages of total ..!ll (40) change. ..... :g w <:n w 10 ::) (20) "" 'D .... '" o ... v >l-. -10 ) Total increment: 1,5.6; II, 17.4; 1ll, 31.0; IV, 48.0. Figure 3-8 (part b) . Sources of Deviations from Proportional Growth: Small Manufacturing Pattern 90 (180) ~ Primary (agriculture and mining) Light industry (including food) 1! -;: 70 1::: '0- "" (140) .::: ~ ~ Total deviation Total deviation "0 50 "U (100) I -16.6 I 18.9 \0 II -31.9 II 26.7 I .... w ""'" - I'-- 0- ~ 30 III IV -54.2 -68.9 III 32.7 IV 25.8 t ~ "0 "U (60) 89 ~ J ;:J 10 I -or (20) II III IV i \0 0- 0 2 '5. -10 '" u ~ (-20) v 0.. 32'44 -30 (-60) 800-1,500 (1,600-3,000 ) Per capita GNP Per capita GNP 90 180) 149 Heavy industry 70 ~Demand ( 140) .:: DExports Total deviation 29 ... ~ I -0.9 ~ '" 50 II 22.6 D Import substitution "0 (100) III 77.4 52 ~ Technological \0 r-- ~ N 186.0 n 30 ...::: '0 ~ ~ (60) 271 I II l!JJ Note: Numbers on the bars indicate percentages of total deviation . .... W "0 VJ Vt 10 ::i ( 'T" '-0 ~ ...... 0 n '15. -20 0: u (-40) .... '" P- -30 I IT I III I IV (-60) \ "'''''''IJ I"VV! \ IVV Per caPIta GNP Figure 3-9. Sources of Growth in Value Added: Average Pattern Food Demand Exports ....... w 0\ Import Suhstitutionl -5 o 15 -5 o 15 -5 o 15 (-10) (30)( -10) (30)( -10) (30) Value added for each $100 increase in per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) ~ 100-200 200-400 400-800 R)<] 800-1,500 ~ (200-400) D (400-800) D (800-1,600) b::::ld (1,600-3,000) Heavy industry Social overhead Demand " 'w'"' '-I Import Suhstitution 5 o 15 -5 o 5 15 -s o 15 10) (30) ( 10) (10) (30) ( (30) Value added for eaeh $100 increase ill per capita GNP in 1964 U.S. dollars (1976 dollars in parentheses) Table 3-11. Incremental Decomposition of Value Added, $200-$400 AV = $200 Export Import Techno- Domestic expan- substitu- logical Change Total demand sion tion change in VA Sector ~V (D) (E) (IS) (TC) coefficient Average pattern Primary 25.2 19.5 18.1 -3.2 -2.0 -7.3 '- w Food 12.3 10.4 15 0.1 0.3 00 Light industry (excluding food) 23.9 15.1 5.4 1.2 2.1 Heavy industry 23.2 12.4 5.6 3.4 1.8 Tradables 84.6 57.4 30.6 1.S 2.2 Utilities and services lIS.4 97.6 16.4 -0.2 1.6 Total 200.0 15S.0 47.0 1.3 3.8 Small primary Primary 38.6 19.8 29.4 -LO -8.3 Food 12.6 10.5 1.4 0.3 0.4 Light industry (excluding food) 20.l 14.9 1.9 1.6 1.6 Heavy industry 21.1 12.1 2.2 4.8 2.1 Tradables 92.4 56.8 34.9 5.7 -8.3 Utilities and services 107.7 96.0 8.6 1.1 2.0 Total 200.0 152.8 43.5 6.8 4.8 -8.3 Large Primary 18.3 23.4 3.9 -1.5 -2.0 -5.5 Food 12.0 11.0 0.7 0.1 0.2 Light industry (excluding food) 24.5 17.3 4.5 0.8 2.0 Heavy industry 33.3 22.4 4.9 3.3 2.8 Tradables 88.1 74.1 14.0 2.7 3.0 -5.5 Utilities and services 112.0 102.5 8.1 0.4 1.0 Total 200.0 176.6 22.1 3.1 4.0 -5.5 Small manufacturing Primary 14.5 16.1 11.6 -5.2 -2.4 -5.7 Food 10.9 10.3 0.7 -0.1 -0.0 Light industry (excluding food) 33.3 13.9 15.4 2.1 1.9 '-' w Heavy industry 16.8 6.7 5.0 4.1 '-0 Tradables 75.5 47.0 32.7 0.9 0.5 -5.7 Utilities and services 124.6 27.7 -0.8 0.2 Total 200.0 144.5 60.4 0.1 0.7 -5.7 Note: Components may not add to totals, due to rounding. 140 THE PROCESS OF INDUSTRIALIZATION The final and potentially most interesting step in the analysis of structural change is to translate these shifts in the composition of value added into increased employment and capital requirements. So far this type of analysis has only been done for a few countries and the results do not yet lend themselves to generalization in a cross-country mode1. 47 Nevertheless, a few observations can be made to indicate the nature of the available results. In analyzing the allocation of investment over time, the sectors having relatively high capital-output ratios-notably social overhead facilities and heavy industry-become much more important. Al- though figure 3-9 shows that in the middle of the transition they each contribute about 25 percent to the increase in GNP, their com- bined share of investment requirements is well over 50 percent. Con- versely, these sectors contribute relatively little to employment. The sources of employment growth must be found in the factors that lead to increases in agriculture, light industry, and services. 48 Overview This chapter has developed an empirical interindustry model of the process of industrialization based on elements of trade and develop- ment theory. The main purpose of the model is to explain the inter- relations among changes in domestic demand, international trade, capital flows, and the structure of production. Although the model has evolved from the study of industrialization in individual countries, the present version is estimated from cross-section data for three groups of countries distinguished by their size and trade patterns. The analysis focuses on the changes in resource allocation that cor- respond to different patterns of trade and capital inflow. An archetype of each kind of country is constructed from separate estimates of the aggregate equations in the model for each country group. The results of the simulations are validated by comparing the predicted changes 47. A decomposition of labor use is given for Japan in Chenery, Shishido, and Watanabe (1962). 48. One of the principal obstacles to estimating these relations from produc· tion functions is the widespread existence of surplus labor, mainly in agriculture and services. SOURCES OF INDUSTRIALIZATION 141 in the structure of production to those observed in the cross-country regressions. Several empirical hypotheses underlie these results: The first is a generalized Engel hypothesis that rising income produces a relatively uniform change in the composition of domestic demand, character- ized by a substantial decline in the share of food consumption and corresponding rise in manufactures. The second hypothesis is that patterns of international trade vary systematically with the level of income, but are also substantially affected by country size, natural resources, and trade policies. The third is that the main differences in the observed patterns of production are associated with differences in trade patterns and hence focused in sectors producing tradable goods. The first two hypotheses have been explored elsewhere and are taken as the starting point for the analysis. The third is generally sup- ported by the simulations of the consequences of alternative trading patterns. These results cannot be considered a test of the hypothesis as formulated here since there is no available alternative to which they may be compared. 49 The testing and refinement of basic concepts is a major purpose of empirical research. This chapter has taken several concepts that have been loosely defined in the development literature and adapted them to a specific interindustry framework. Although there are other pos- sible interpretations of such general notions as balanced growth and import substitution, the present formulation shows the value of a eonsistent set of definitions and makes possible an assessment of the relative importanee of the several eauses of struetural ehange. The results of the formulations in this chapter can also be com- pared to insights gained from purely deduetive models that are typi- eally based on one of two polar cases, the closed economy or the completely open, "small country" case. The closed economy that is usually assumed in optimal growth theory finds its closest empirical eounterpart in my "semiclosed" or "very large eountry" case. Although internal fadors-domestic demand and technological change-ac- count for 60 to 70 percent of the shifts in resource allocation in the 49. The "naive" hypotheses that all sectors expand in proportion or that elasticities are the only sources of structural change are empirically uninteresting. 142 THE PROCESS OF INDUSTRIALIZATION simulations of this case, the external factors are sufficiently important to modify the policy conclusions from such an analysis very substan- tially.50 Similarly, the simulations of the two small-country prototypes show that, although changing comparative advantage-as reflected in import substitution and nonproportional expansion of exports- is the major factor in the explanation of structural change for these cases, the balanced grO\vth elements are quite significant as well. ll1ese results provide a basis for applying a similar methodology to the study of industrialization in individual countries. In addition to the conceptual framework developed here, the cross-country model can be used to identify the peculiarities of individual country expe- rience by comparison to what would be predicted from a simulation using the exogenous variables for the country concerned. rfbis pro- cedure highlights the ways in which initial conditions and ceo nomic policies of each country modify the typical patterns analyzed here. 51 50. The policy implications of this observation are explored further in chapters 6 and 7. 51. This methodology is being used in a comparative analysis of eight countries under the \Vorld Bank research project by Chenery, Kubo, Robinson, Syrquin, and Westphal, "Sources of Industrial Growth and Structural Chauge," the results of which are to appear in monograph form. Some initial findings appear in Chenery and Syrquin (1979) and Kubo and Robinson (1979). Chapter 4 Substitution and Structural Change with William J. Raduchel THE ABILITY OF A DEVELOPING ECONOMY to adapt to rapid structural change depends largely on elasticities of substitution in demand, pro- duction, and trade. On the one hand, to the extent that production requires inputs of commodities and factors in fairly fixed proportions -as was assumed in chapter 3-growth is more likely to be impeded by shortages of specific inputs than by a general scarcity of resources. On the other hand, if commodities and factors are highly substi- tutable in satisfying human wants-as implied by neoclassical theory -one need have no great concern for the supply of anyone of them. Disagreement over the extent of potential substitution therefore lies at the heart of some of the major issues of development policy. Differences in substitution assumptions are closely related to the time period considered. Neoclassical theory has been mainly con- cerned with conditions of long-term equilibrium in which physical and human capital can be transformed into whatever specific form is most appropriate. Development policy, however, must deal with shorter planning horizons of five to ten years. The extent of substitu- tion that can take place within this period depends on factors such as This chapter has been adapted from Chenery and Raduehel (1971). 'I11e second and third sections are reproduced with little change. The fourth section extends the analysis to consider the relations between surplus labor and capital inflows. The solutions were carried out by Timur Kuran, with assistance from Sherman Robinson and Alexander Meeraus. 143 144 SUBSTITUTION AND STRUCTURAL CHANGE the magnitude of the additions to the stock of capital, the mobility of labor among sectors, and the rate of assimilation of new agricultural techniques. The degree of substitution that should be assumed for policy purposes is thus somewhere between the complete flexibility of the long run and the technological rigidity that characterizes most planning models. Proposals for avoiding the bottlenecks that hamper development have tended to concentrate on a few specific aspects of this problem: increasing flexibility through trade, more appropriate (labor-intensive) technology, or changing the composition of demand so as to use less capital or imports. The present chapter was conceived as a way to consider these possibilities simultaneously so as to bring out their interrelations. For this purpose we have devised a methodology that is more illustrative than the detailed sectoral analysis of chapter 3 but that retains its empirical origins in cross-country estimates of the main parameters. The analysis consists of four parts. The first decomposes the aggre- gate transformation of productive factors into social welfare into four sets of relations that can be studied empirically. These are taken as a basis for a price-endogenous general equilibrium model whose pa- rameters can be specified from cross-country data (pages 146-55). The relative importance of direct and indirect means of substituting labor for capital is then discussed (pages 155-65). Finally, the possi- bilities of surplus labor and the extent to which it can be reduced by capital inflows are taken up (pages 165-71). Substitution among Commodities and Factors General equilibrium theory postulates a series of transformations by which factors of production are combined to produce social wel- fare. We can distinguish four types that are important to the empiri- cal analysis of resource allocation: (a) transformation of generalized factors of production (capi. tal, labor, natural resources) into specific factors, such as capital goods, skilled labor, and irrigated lands 1 ; 1. In the case of capital this transformation into a specific fonn is the basis for the phenomenon of "embodiment." It is obvious that a similar process of special· ization in a particular use and type of technology also applies to land and labor although the possibilities of conversion may be greater. SUBSTITUTION AMONG COMMODITIES AND FACTORS 145 (b) transformation of specific factor services and raw ma- terials (plus other inputs) into finished products; ( c) transformation of exported commodities into imports by means of international trade; and (d) transformation of commodities and services into "social welfare" in accordance with individual and social preferences. Each type of transformation allows for some degree of substitution among inputs in producing a given level of output. Direct substitution among factors occurs in the first two types, which include all activities of physical production. 2 Types (c) and (d) involve substitution among commodities. These transformations provide indirect means of substituting among capital, labor, and land by varying the compo- sition of trade or final demand within speCified constraints. Planning models customarily allow for some substitution through trade but tend to exclude the other possibilities. Apart from trade, each transformation function is stated in the form of a vector of in- puts "required" to produce a given level of output-of skilled labor, steel, food, or "welfare." Linear programming models of resource allocation are normally formulated on those assumptions. Indirect substitution through international trade is limited by the range of variation in factor proportions available to produce tradable com- modities and by the elasticity of demand for exports. Empirical studies based on this type of model typically show a rather restricted scope for substituting labor for capita}!! The degree of substitution that should be included in more real- istic planning models depends to a large extent on the time period considered. Since the possible variation in output levels and input proportions is determined largely by the new capacity installed, the extent of direct substitution among factors is limited by the amount of new investment. Fixed-coefficient planning models are most appro- priate to periods of less than five years, in which indirect substitution through changes in demand or trade may wen outweigh the direct substitution possibilities. For a more general analysis of substitution 2. In the illustrative specification of the model in the following section, these two types of substitution are combined, but in a dynamic and less aggregated version they should be kept separate. 3. See Chenery and Kretschmer (1956), Bruno (1966), and Eckaus and Parikh (1968). 146 SUBSTITUTION AND STRUCTURAL CHANGE we shall consider a "medium term" of ten years, in which the capital stock in a developing economy may be augmented by 50 percent to 100 percent, and a "long term" in which only tastes and technology are specified. 4 Evaluating possibilities for substitution requires estimates of the effects of relative prices of commodities and factors on levels of de- mand, trade, and factor use. Disaggregating the economy into more homogeneous sectors or groups of commodities should make it possible to get more accurate estimates of these effects. Up to now the study of production functions at the two-digit level used in chapter 3 has not produced estimates that are sufficiently robust to warrant this degree of disaggregation,5 and we have therefore adopted a more illus- trative four-sector subdivision of the economy. Our analysis takes as a point of departure a set of projeetions that exclude substitution in production and demand and only allow for trade. This base projection is similar to the average pattern given in chapter 3 for an economy inereasing from an income level of $200 to $300. From this starting point the following questions are considered: (a) At a given level of GNP, what are the effects of the several types of substitution on the composition of demand, trade, and production? (b) How much increase in employment can result from fea- sible reductions in the relative price of labor? ( c) How much increase in welfare can be produced by the optimal reallocation of commodities and faetors? ( d) What is the effect on employment and social welfare of an increased inflow of capital? A Planning Model with Substitution Models that are currently used for planning and policy purposes usually allow for substitution only by means of international trade. 6 Some additional flexibility may be introduced on an ad hoc basis by 4. It is only in the long term that the distinction between general and specific factors can be ignored. 5. Morawetz (1976) provides a comparison of time series and cross-section estimates of manufacturing production functions and stresses their variability at the two-digit level of disaggregation. 6. The formulation and use of this type of model is discussed in Chenery (1971) and Blitzer, Clark, and Taylor (1975). A PLANNING MODEL WITH SUBSTITUTION 147 varying the composition of demand or other constraints in successive solutions. Substitution through trade is limited either by fixing exports exogenously or by assuming that increased exports can only take place at declining prices.7 'The bases for a more general treatment of substitution in an em- pirical interindustry framework are suggested by Chenery and Uzawa ( 1958) and Johansen (1960). The former show how imports, exports, and final demand can be made functions of equilibrium prices in a computable programming model, while the latter demonstrates the use of Cobb-Douglas production functions in each sector. We shall extend the Chenery-Uzawa model to include CES (constant elasticity of substitution) production functions for capital and labor as well as a more satisfactory set of demand relations of the type used by Johansen. TIle interindustry model The formulation of the model follows Chenery and Uzawa (1958). The model consists of a conventional input-output core for com- modity production to which have been added nonlinear demand functions, import and export functions, and production functions for direct factor use. We assume m desired commodities (final demands) and several primary factors-inputs available from outside the system. In the following illustration there are three primary inputs-labor, capital, and foreign exchange-and fOUI commodities. A summary of the notation follows. VARIABLES Xj output of sector; Vi value added in sector; Pi shadow prices of commodity i PK , PL , PF shadow prices of capital, labor, foreign exchange, respectively; 7r = P K / PI, level of exports from sector i level of imports of commodity i total supply of capital use of capital in sector j total supply of labor use of labor in sector j 7. This treatment of exports is discussed by Chenery and Kretschmer (1956) and Bruno (1966). 148 SUBSTITUTION AND STRUCTURAL CHANGE Y. final demand for commodity i R total resource use D maximum foreign trade deficit Y total domestic demand PARAMETERS FOR EACH SECTOR a.;j input coefficient of commodity i into sector i c efficiency parameter in the production function 8 distribution parameter in the production function (J' elasticity of substitution between capital and labor () price elasticity of final demand k capital coefficient (K/X) labor cocfficient (L/X) v value added per unit of output g eost of unit import of commodity in foreign exchange h price of unit export of commodity in foreign exchange ex. slope of export demand function t; slope of import substitution function The general problem of resource allocation can be formulated as either the attainment of a given level of welfare with a minimum use of scarce resources or the maximization of the welfare achievable from given resources. The minimizing formulation is more convenient for comparing the effects of different types of substitution because other elements are held constant. This formulation is presented in model I and used to study the ways in which labor can be substituted for capital. The maximizing formulation of model II is needed to analyze the effects of increasing the inflow of external capital and is presented in the final section. In model I the objective is to minimize R, the total use of labor, capital, and external resources valued at predetermined prices: (4.1 ) From the national accounts identities, we can also write8 : (4.la) v} = (PKk j = PLl j ) X j KjPK LlL + = VjX} , R = !Vj PFD. 8. The value added for each unit of output, VJ, is only a constant for a given set of factor prices. A PLANNING MODEL WITH SUBSTITUTION 149 Since D is held constant for the solutions to model I, the objective becomes to minimize the use of domestic resources. We require that production plus imports in each sector be sufficient to meet final and intermediate demands plus exports: (4.2) Foreign trade is restricted by requiring that the trade deficit should not exceed a predetermined amount, 15: (4.3 ) All quantities must, of course, be nonnegative. Equations (4.1) through (4.3) are all familiar as standard elements of an economywide linear programming modeL9 We now introduce some new features. Although the seetors are linked by a fixed input- output matrix, we permit direct substitution between capital and labor in the production of value added for each sector.l0 This can be done without grcatly complicating the computation if the production func- tion specified is homogeneous and does not have increasing returns. So long as factor intensities depend only on relative factor prices, the several factors can be treated as one for a specified set of prices. This formulation permits us to use the substitution principle of Samuelson (1951) and others as a basis for the optimizing procedure. Production functions To have a general production function for which some empirical estimates are available, we assume that output in each sector is related to capital and labor inputs according to aCES function l l ; (4.4) x= c[IlK-P 9. Such models are illustrated by San dee (1959), Eckaus and Parikh (1968), Bruno (1967), Weisskopf (1971), and Tendulkar (1971). 10. This procedure was suggested by Johansen (1960), who used Cobb-Douglas functions for each sector. Note that the model can easily be generalized to permit alternative production activities in the linear programming manner. But this would require engineering data that are generally unavailable. II. This is the form given by Arrow, Chenery, Minhas, and Solow (1961). Specifications of this function used in our model are illustrated in figure 4-1. 150 SUBSTITUTION AND STRUCTURAL CHANGE For fixed relative factor prices we can derive the following cost- minimizing capital-output and labor-output ratios: (4.5) k K = ~_ (~ - 0 (1 0 ) y ] 1/1' 1 X C (4.6) where y (PK~~)-P/l+P PJ~ 0 We emphasize that we are treating capital as one factor capable of use in all sectors, which is strictly valid only in the long run. Foreign trade Domestic resources can be transformed into foreign exchange through either exports or substitution for imports. These possibilities will be described by an export revenue function and an import substi- tution function. Taken together they describe the net trading possi- bilities for each sector as a function of the cost of earning or saving foreign exchange. Following Chenery and Kretschmer (1956) we assume that the average revenue in foreign exchange earned for each unit of exports in sector; (h j ) is a declining function of the amount exported: (4.7) It will pay to expand exports in each sector until the marginal revenue earned is equal to the marginal cost of production valued in shadow prices (Pj/P / ). An import substitution function can be constructed along similar lines, since the average amount of foreign exchange saved in each sector through domestic production depends on the level of imports. As illustrated in \Veisskopf's (1971) detailed programming analysis of import substitution in India, the rupee cost of replacing a dollar's worth of imports in the metal products industry varies according to the particular commodity produced. If only a small amount of foreign exchange is allocated to imports of metal products, it should be used for the commodity with the highest local production cost-that is, A PLANNING MODEL WITH SUBSTITUTION 151 in which the import price at a given exchange rate is relatively the lowest. If imports of metal products are increased, the average foreign exchange cost (gj), of replacing a rupee of domestic production ",ill rise as we move up the scale of comparative advantage. The possibili- ties for import substitution in sector i can therefore be represented by assuming that gj is an increasing function of M j · Assuming a linear function gives: (4.8) Equation (4.16) gives a condition for an optimal solution parallel to that for exports: imports of commodity i should be increased until the marginal savings of domestic resources is equal to the opportunity cost of imports (Pi/P t ). In this formulation, export expansion and import substitution are treated symmetrically. 'Ine overall substitution possibilities provided by foreign trade can therefore be shown by combining the solutions for the optimal levels of imports and exports into a composite net trade function which shows both E j and M j as a function of P) Pt. Such functions are derived in equations (4.15) and (4.16) and illus- trated in figure 4-2. Final demand The other form of substitution included in this model is in the composition of final demand. Any set of demand functions can be used as long as the quantities consumed depend only upon prices and income. We assume that market prices are proportional to shadow prices and that the income and own-price elasticities of demand are constant. Our demand functions can be stated as: (4.9) where A is a factor of proportionality relating market and shadow prices. Since we wish to hold some measure of welfare constant, A will be defined as a price deflator such that a Laspeyre's index of welfare remains constant: (4.9a) IP/Yj = constant, where Pi" are the base year market prices for which A = 1. That defi- nition makes demands a function of all commodity prices. 152 SUBSTITUTION AND STRUCTURAL CHANGE Our programming problem can be restated as follows. For fixed P K and PL , minimize IvjXj subject to the following COllstraints 12 : (4.10) (4.11 ) The solution involves minimizing the following Lagrangian form with respect to X, M, E, P, and P : F (4.12 ) A( X,M,E,P,PF ) - I VjXj I Pj (Xj Mj - Ej Yj - ~ t aj,Xi ) PF(D I gjM j +I hjE}). A general exposition of this type of model is given in Chenery and Uzawa (1958). It is easy to see that the first-order conditions for an extremum of this form readily yield the following equations, which are familiar from linear programming solutions: (4.13 ) where P [Pj] and v [Vj] represent the inputs of the composite fac- tor; and, (4.14) X=(I-A)-l(Y E-M). Combined with the non negativity requirements (E j > 0, M j >0) and (M j < Y j E j ), the first-order conditions also yield equations for the determination of exports and imports: (4.15) Ej ¥i 1=:'PI) if> 0; otherwise E J O. (4.16) Mj _ ""j ~;:!Pf) if > 0 and < (Yj + Ej ). J M j = 0 if the first condition is violated and ¥J + E j - 0 if the second condition is violated. The initial solutions to this model were attained from an algorithm based on an iterative procedure suggested by Chenery and Uzawa (1958) .13 In the present version both model I and modeln have been solved by the Generalized Reduced Gradient Method, a more power- 12. The inequalities of equations (4.2) and (4.3) are replaced by equalities, since it is known in advance that both sets of constraints will be binding. 13. This algorithm was developed by Raduchel (1970). A PLANNING MODEL WITH SUBSTITUTION 153 ful algorithm which is well suited to this type of nonlinear program- ming problemY Estimation Illustrative estimates of the parameters in this model are given in table 4-1.1" Alternative specifications are given for low, medium, and high elasticities in trade, demand, and production.16 In general, all elasticities are likely to be higher over longer periods as the capital stock, trade patterns, and tastes become adjusted to price changes. Our estimates are intended merely to illustrate realistic orders of mag- nitude. Wherever possible they are derived from econometric esti- mates of comparable functions, but no attempt has been made to determine consistent estimates of the model as a whole. The starting point for our example is the illustrative four-sector model of Chenery and Uzawa (1958) P The seetors differ signifi- cantly in their demand and trade relations and in their role in the development process. Sector 1 includes most of the manufactured goods that are typically produced and exported at low levels of in- come, while sector 3 includes the heavy industrial products (machin- ery, metals, chemicals) that are typically imported by underdeveloped coun tries. The parameters in the production functions (u and 8) for the long-term case (2b) are based on the sector values given in Arrow, Chenery, Minhas, and Solow (1961), which were derived from a com- parison of capital and labor use in Japan and the United States. Lower values of the elasticity of substitution were set at arbitrary fractions of these long-term estimates to show the effects of substitutability over shorter periods (the 8's are adjusted to maintain comparability in the basic solutions) . 'I'he parameters in the demand functions for the medium term are based on vVeisskopf's (1971) intercountry estimates. Is The "high" 14. A further discussion of the properties of the models and the nature of the solutions is given in a technical appendix to this chapter; see Kuran (1979). 15. These parameters are unchanged from the 1971 version of this chapter. 16. These specifications will be combined in different ways to analyze each type of substitution separately. 17. The sector classification and input structure for this model were in turn derived by aggregating the fourteen-sector programming model given in Chenery and Kretschmer (1956). \Ve have kept the same commodity input coefficients and export functions but modified the remainder. 18_ Johansen (1960) uses a similar formulation. His estimates of the own- pricc elasticities for Norway on a twenty-sector basis also average about -0.5. 154 SUBSTITUTION AND STRUCTURAL CHANGE Table 4-1. Illustrative Estimates of Parameters Sector (1) (2) (3) (4) Light Food and Heavy Parameter industry agriculture industry Services Production parameters' All alj 0 0 0 0 d2j OJ 0 0 0 as; 0.2 0.1 0 0 d4j 0.2 0.3 OJ 0 Low (J" 0 0 0 0 P oa 8 0.915 0.944 2.600 0.800 c 3.83 3.24 4.0 1.80 (J" 0 0 0 0 P Ob 8 0.276 1.034 2.60 0.77 c 2.551 3.39 4.00 1.77 Medium (J" 0.11 0.29 0.2 0.05 P,a 8 0.326 0.443 0.991 0.00798 c 3.97 3.33 1.67 1.84 (J" 0.22 0.58 0.4 0.1 P'b 8 0.41 0.47 0.92 0.08 c 3.99 3.33 l.8 1.89 High (J" 0.45 1.15 0.4 0.2 P 2a 8 0.456 0.483 0.917 0.23 c 4.0 3.33 1.8 1.92 P 2b (J" 0.93 U5 0.8 0.4 8 0.484 0.483 0.769 0.344 c 4.0 3.33 1.96 1.96 Demand parameters b Low y' 100 230 220 450 Do () 0 0 0 0 Medium y. 100 230 220 450 Dl (J -0.674 -0.246 -0.587 -0.352 High y. 100 230 220 450 D2 (J -1 -1 -1 1 Trade parameters c Low 0 0 0 0 , a To y 0 0 0 0 0 0 0 0 SUBSTITUTION BETWEEN CAPITAL AND LABOR 155 Table 4-1 (continued) Sector (1) (2) (3) (4) Light Food and Heavy Parameter industry agriculture industry Services Medium a 0.005 0.001 0.01 0 Tl y 1.0 1.1 1.0 0 ~ 0.005 0.0157 0.00178 0 High a 0.0025 0.0005 0.00178 0 T2 y 1.0 1.1 1.0 0 ~ 0.005 0.0157 0.00178 0 a. aij intermediate commodity requirements. " elasticity of substitution. l! distribution parameter. c = efficiency parameter. b. yo = final demand at base point. (J = price elasticity. c. a = slope of export earnings function. 'Y = intercept. ~ slope of import earnings function. fL intercept 1.0 for all sectors. value of (J was arbitrarily set at -1.0 for all commodities to indicate an upper limit to substitution in demand. 'TIle trade parameters were chosen to yield a realistic variation in the pattern of trade as P KIP L varies. Imports and exports are less than 10 percent of GNP in most solutions, which is typical of larger coun- tries. The import substitution functions are arbitrary, although they could be derived for particular countries from a disaggregated model. Import and export functions for each sector can be combined as shown in figure 4-2 to dctermine the net trade in each commodity group as a function of the ratio of its shadow price to the price of foreign exchange. The production, demand, and trade functions are illustrated in figures 4-1 through 4-3. lbey will be used in conjunction with the solutions in tables 4-2 and 4-3, below, to show the variation in each sector that results from the three types of substitution. Substitution between Capital and Labor Several kinds of disequilibria arise in the course of development; these are variously attributed to limited possibilities for substitution 156 SUBSTITUTION AND STRUCTURAL CHANGE Figure 4-1. Long-run Production Functions (Pita): Unit Value Added K ,3 \ 1.0 0.8 "2 '5. U'" 0.6 - - - Heavy industry (3 ) 0.4 ~---- -------SerVlces (4) -- -- "-. . 0.2 --- Food and agriculture (2) Light industry (1) o L Labor or to inappropriate price policies. The most pervasive are foreign ex- change scarcities and labor surpluses. Since rising unemployment has become a serious problem even in many developing countries with substantial rates of growth, we will focus initially on questions of labor-capital substitution and employment policy.19 Direct substitution between capital and labor appears to be of greater significance for employment than indirect substitution under most of our assumptions. We therefore illustrate the direct mecha- nism first and then take up the possibilities for substitution by way of demand and trade when direct substitution is restricted. 19. A more disaggregated form of the present model would be necessary to discuss the effects of substitution on the trade bottleneck with any degree of realism. SUBSTITUTION BETWEEN CAPITAL AND LABOR 157 Figure 4-2. Net Trade Functions (T.) Exports 300 Light industry \ Food and \ agriculture \ \ 100 \ \ \ 1.25 1.5 1.75 2.0 0~--4----+----+---~~~~--+----r~~~ 0.25 0.5 0.75 10:--; -...:::::-- - - - P;/ I . - _ · Food and -:- - . '- - _ ~nculture (2) 100 Heavy industry (3) 200 300 Light industry (1) Imports Direct factor substitution The possibilities of substituting labor for capital are determined at a given level of GNP by varying the relative factor prices over a spe- cified range and computing a series of optimal solutions. 20 Figure 4-4 gives a set of isoquants for the whole economy that are derived for an increase in GNP of 1,000. They show the extent of substitution that results from each of the five assumptions about the parameters in the production functions, assuming in each case the medium-term values for demand parameters (Dl) and trade parameters (T1) . Tl1e overall elasticity of substitution is shown more directly in fig- 20. In each case the price of foreign exchange is varied in such a way as to keep the capital inflow D at zero. The GNP is defined by equation (4.9) and set at 1,000. All base prices (P/) were set equal to 1.0 by adiusting the efficiency pa· rameters CJ. 158 SUBSTITUTION AND STRUCTURAL CHANGE Figure 4-3. Demand Functions (D 1 ): Unadjusted log Y -- , ~~ , -........... Food and .. , ", "'::::,_ _ agncu I ture (2) '::-- , '"", . - .. Services (4) ' . Heavy , .industry (3) log P P = 1.0 log P ure 4-5, in which factor proportions are plotted against relative factor prices for the same set of solutions. The slope of each curve at a given point gives the elastieity of substitution, which varies from 0.64 in the long-run case (2b) down to 0.21 in case lao The overall elasticity is a weighted average of the elasticities in the four sectors, but in the present example it is fairly constant. To apply these results to a case of potential surplus labor, we as- sume that the economy will have an additional 500 units of capital and 750 units of labor over a planning period of ten years. Solution to the programming model given by specification T 1 D 1 Poa yields the re- sults shown in solution 2 in table 4-2: total output of 1,000 and use of labor and capital of about 500 units each. 21 Figure 4-4 shows that, under the other production assumptions, 750 units of labor could be fully employed and capital reduced by between 10 percent and 28 percent at the same level of output if there were no constraint on rela- tive factor prices. 21. Solution 1 corresponding to PK/Pr, = 1.0 is taken as the basic OT market solution for reasons discussed below. SUBSTITUTION BETWEEN CAPITAL AND LABOR 159 Figure 4-4. lsoquants for GNP = 1,000 P'b Assumptions 1,000 Trade = Tl \P,. , Demand D,= Production = Po« to P,. 900 \\ \\, 800 \ \P'b \ \\ -. '" 's 700 \\:,\ \ "\\ PIa '" U \ ~\\ 600 ~\\\ , KIL 0.67 sao 400 .. - ..... - -------P·· it, 400 I I 500 600 I I 700 I 800 900 I - - --j __ P2b 1,000 Labor (L) \Vhether an economy can achieve some of the combinations of labor and capital indicated by these aggregate production functions depends on the government's ability either to subsidize labor incomes and tax profits or to use extra-market allocation methods. In the pres- ent example we will assume that it is not feasible to reduce the return to a unit of labor below 25 percent of the return to a unit of capital because of the government's limited ability to subsidize wages. Toe effect of such a limit on the attainable factor proportions is shown by solutions 2 to 6 in table 4-2, in which the ratio of PK/PL is set at 4.0. 'The minimum wage limit is indicated in figures 4-4 and 4-5 by a line connecting these solutions for each model specification. The portion of each isoquant in figure 4-4 to the right of this line is not attainable because the labor supply is not available at a lower wage. Since labor would receive only about 20 percent of the total 160 SUBSTITUTION AND STRUCTURAL CHANGE Figure 4- 5. Variation in Capital Intensity with Factor Prices 70~------------------------------------~ 60 50 40 P,a 30 P2~ P2a P,. \ Assumptions 25 \ \, \ \ Trade = T, 20 \ \ \ \ Demand D, = 15 \ \, \ Production Poa to p,,, \ \ .\ \ \ \ 10 9 8 \ \. \ \\ 7 \.. \ 6 5 \ " \ \ __________\I2.~,~-~4-\12---Minimum wage 4 \ " \ \ (PdPK 0.25) = OJ .~ 3 \ ... \ ~ .... .2 2.5 2 '\b,\\ u JS 1.5 \\\\ '" \\\ -.5 rn 1.0 1 ~ 0.9 0.8 0.7 0.6 0.5 \'\ \\\~ 0.4 \ \ \\\ OJ \\ \ '. \ . 0.25 \ ' \ \ 0.2 \\ \ \ " \ \ 0.15 \ \p,:.., \ pPfJI 'Pla' "p.., \ ' P2h Capital·labor ratio (log K/L) product at this relative wage, a realistic limit may be more restrictive than we have assumed. With the assumed wage limit it is possible to achieve the full use of labor only under the higher assumptions as to direct substitution, P2a and P2b. In the latter case, expansion of output from 1,000 to SUBSTITUTION BETWEEN CAPITAL AND LABOR 161 about 1,220 (from point 7 to point 8 in figure 4-4) would be feasible by making full use of capital and labor. Indirect factor substitution Indirect factor substitution takes place through the effects of chang- ing commodity prices on both foreign trade and domestic demand. Detailed programming models for several countries suggest that a variation in overall capital-labor ratios of 10 percent to 20 percent might be achieved by way of trade alone through a fourfold variation in the relative price of capital and labor.22 A good part of this variation is lost in our model through aggregation, which has the effect of reduc- ing the range of variation in commodity prices and its resulting effects on trade. To secure a more realistic change in commodity prices as factor prices change, we have therefore altered tbe specification of the production function for sector 1 to make it less capital intensive.23 1bc effects of substitution through trade alone can be seen in table 4-3 by comparing the trade-only solution to the basic solution. \Vith substitution only through trade, both the capital and labor coefficients and the final demand are held constant. Raising the relative price of capital from 1.0 to its limit of 4.0 has the effect of increasing the cost of producing the most capital-intensive commodity (3) by 25 percent and lowering the price of commodity 1 by 13 percent. As a result there is increasing trade and greater specialization, with a net reduc- tion in the capital-labor ratio of 5 percent from the initial solution. 111e effects of also allowing maximum substitution in demand (D 2 ) are shown in solution (c) of table 4-3, in which all other elements of the model remain the same. There is a drop of 20 percent in consump- tion of commodity 3, which has almost as great an effect as the reduc- tion in capital use due to international trade. Demand and produc- tion rise by a corresponding amount for commodities 1 and 4 in re- sponse to their lower prices. The net effect of introducing demand elasticities is thus a further reduction in the capital-labor ratio from 0.95 to 0.90. 22. See, for example, Chenery and Kretschmer (1956); Bruno (1966); Weiss· kopf (1971); and Tendulkar (1971). 23. This change is shown in tables 4·} and 4-3 as specification (P ob ) and is illustrated in figure 4-1. A small adjustment was also made in sector 2 to maintain the total use of both capital and labor at 500 in the basic solution. The overall effects of this change in the case of direct substitution are negligible. Table 4-2. Effects of Direct Substitution in Production, Medium Demand and Trade: Model I Labor Capital coefli- coefli- Final Net Ou.t- cient dent demand trade put Price Labor Capital Solution Model Factor prices Sector 1i kt Yi T; Xi Pi Li K, 1 T,D,Poa l'K/P[, 1.0 1 0.261 0.239 100 -6 94 1.00 25 23 K/L = 1.06 2 0.309 0.291 230 20 260 1.00 80 76 ...... Pp = 0.947 0\ 3 0.250 0.650 220 -16 249 1.00 62 162 1'-' 4 0.556 0.444 450 572 1.00 318 254 Total 485 514 2 T,D,l'oa PK/P1J 4.0 1 0.261 0.239 101 -0.2 101 0.981 26 24 K/L 1.005 2 0.309 0.291 233 63 306 0.956 94 89 PI<' = 2.548 3 0.250 0.650 199 59 191 1.187 48 124 4 0.556 0.444 467 598 0.899 332 266 Total 501 503 3 T,D,P1a PK/PL = 4.0 1 0,290 0.230 101 1.2 100 0.983 29 23 K/L = 0.841 2 0.394 0.247 234 68.4 312 0.935 123 77 l'p 2.464 3 0.319 0.617 199 -62.4 188 1.187 60 116 4 0.584 0.428 466 598 0.905 349 256 Total 561 472 4 T1D1P 1b PK/P L = 4.0 1 00313 0.212 101 99 0.981 31 21 K/L = 0.712 2 0.484 0.201 235 73.4 318 0.916 154 64 Pp 2.378 3 0.384 0584 199 -65.6 185 1.185 71 108 4 0.607 0.415 465 598 0.914 363 248 Total 619 441 5 T 1D 1 P2a PK/PL = 4.0 1 0.374 0.182 102 -2.6 99 0.970 37 18 K/L = 0.549 2 0.645 0.120 238 92.8 341 0.863 220 41 Pp = 2.288 3 0.386 0.579 197 -79.5 171 1.211 66 99 4 0.661 0.394 463 602 0.923 398 237 Total 721 396 6 T 1 D1P2b PK/PL = 4.0 1 0.490 0.125 103 0.2 104 0.951 51 13 '- K/L = 0.418 2 0.642 0.121 237 82.5 330 0.891 212 40 0'1 Pp = 2.178 w 3 0.574 0.494 197 -74.4 176 1.206 101 87 4 0.760 0.336 463 600 0.923 456 202 Total 820 343 Note: Solution numbers correspond to points in figures 4-4 and 4-5. Totals may not add due to rounding. Table 4-3. Effects of Indirect Substitution through Trade and Demand Labor Capital coeffi- coeffi- Final Net cient cient demand trade Output Price Labor Capital Model Factor prices Sector Ii k. Yi Xi Pi ~ (a) Basic solution T 2 D oPo /) = 1.0 Pl(/P L 1 0.392 0.108 100 -7 93 1.00 36 10 K/L == 1.0 2 0.295 0.305 230 26 265 1.00 78 81 PF 0.933 3 0.250 0.650 220 -20 245 1.00 61 159 4 0.565 0.435 450 573 1.00 324 249 ...... Total' 1,000 499 499 0\ (b) Trade '"'" T 2 D oPob PK/PL = 4.0 1 0.392 0.108 100 22 122 0.871 48 13 K/L = 0.945 2 0.295 0.305 230 61 303 1.017 89 93 Pp 2.421 3 0.250 0.650 220 -76 199 1.246 50 129 4 0.565 0.435 450 585 0.933 331 255 Total' 1,000 518 489 Demand and trade T 2 D 2 PoQ PK/PI~ = 4.0 I 0.392 0.108 115 22 137 0.871 54 15 K/L 0.900 2 0.295 0.305 226 61 301 1.017 89 92 PI<' 2.424 3 0.250 0.650 177 -76 158 1.246 39 103 4 0.565 0.435 483 616 0.933 348 268 Total' 1,000 530 477 a. Totals may not add due to rounding. SURPLUS LABOR AND EXTERNAL CAPITAL 165 Although we have assumed relatively high elasticities for both de- mand and trade in this illustration, the total amount of indirect sub- stitution between capital and labor is only 10 percent, somewhat less than the lowest assumption for direct substitution in production. Without further empirical evidence, it is impossible to say whether or not these results represent realistic orders of magnitude. \V11en all three types of substitution are included in the same model, the separate effect of each is reduced. This results primarily because of the smaller variation in relative prices of commodities. When direct substitution is ruled out, as in table 4-3, price variation and indirect substitution become much larger. The same effect is shown by comparing solutions I and 2 in table 4-2. Surplus Labor and External Capital The preceding section illustrates conditions under which surplus labor may persist evcn with substantial opportunities to substitute labor for capital, both directly and by changing the composition of demand and trade. Relaxation of the minimum-wage assumption would lead to higher employment, but it would not add greatly to total production, which is constrained by the supply of capital. Al- though we have omitted the low-productivity (informal or tradi- tional) activities in which "surplus" labor is actually employed, their inclusion would not significantly affect these results. \Ve now consider the consequences of using an inflow of external capital to reduce the imbalance between the supplies of labor and capital. In effect this adds a fourth type of substitution to the model by increasing external resources in the near term and repaying them in a subsequent period. Although analysis of the longer-term conse- quences of a capital inflow requires a dynamic model (such as that of chapter 9), it is useful to consider the medium-term effect of addi- tional capital in the context of the present model. Both the low productivity of labor and the high value of external capital in transitional economies are reflections of limited substitu- tion possibilities. These phenomena are reduced or eliminated when elasticities of substitution become sufficiently high. The following analysis focuses On this aspect of the more general problem of struc- tural disequilibrium. Our objective is to determine the effects on optimal resource allo- 166 SUBSTITUTION AND STRUCTURAL CHANGE cation and total output of adding to the stock of capital through external borrowing. For each level of capital inflow and total invest- ment we determine the maximum GNP that can be produced, retaining the same constraints on the supply of labor and the minimum wage. The deficit in the balance of trade in the terminal year is increased to be consistent \vith the required inflow of capital over a decade. For this purpose, we use a maximizing version of the programming model (model II) instead of the minimizing version discussed up to now. Although more complicated to solve, this form enables us to specify all of the constraints and to derive the relative factor prices corresponding to each solution. The objective function to be maximized in model II is total de- mand in base year prices 24 : (4.17) Since the model was calibrated to make all base year prices (P/) equal to 1.0, total demand is simply Y = .IYj · For a given import surplus, GNP is also maximized, since it is conventionally defined as total de- mand minus the import surplus. A full statement of model II is given in the technical appendix. 25 One should note that the shadow prices produced as a dual of the optimal solution in model II are not necessarily equal to the "market" prices which are solved endogenously as part of the constraint system. Although empirically the two sets of prices do not differ by much, we have not attempted to achieve full consistency between them. 26 Un- less stated explicitly, prices reported below are the endogenously solved "market" prices. To bring out the effects of varying degrees of substitution on the relations among external capital, surplus labor, and GNP growth, we will carry out several identical optimization experiments with three specifications of the model. Since the previous section has shown that 24. In model I, by contrast, aggregate welfare is held constant. The measure of welfare chosen is real GDP valued at base year prices, which is Laspeyre's index: (4.9a) y constant. The parameter A in equation (4.9) serves to normalize the demand functions given the prices so that total demand (or GDP) equals the fixed level. 25. See Kuran (1979) 26. See Negishi (1962) and Dixon (1975) for a more complete discussion of these issues. SURPLUS LABOR AND EXTERNAL CAPITAL 167 direct substitution is much more important in this model than in- direct substitution, we vary only the production parameters. For this purpose we will use three of the specifications examined above, which will be identified as: L, low substitution, defined as the combina- tion TIDl PQQ ; 1\1, medium substitution, defined as the combina- tion T1D1P 1a ; and H, high substitution, defined as the combination T 1 D t P 2a · As shown in figure 4-4, the constraints were chosen in such a way that surplus labor exists initially in the first two specifications but not in the high-substitution case. Solutions for each model for a level of total demand (and GNP) of 1,000 have already been given in table 4-2 and illustrated in points 2, 3, and 5 of figures 4-4 and 4-5. Our analysis will compare the effects of increasing capital inHows under surplus-labor and full employment conditions. Given the illus- trative nature of the model specifications, we comment only on the major differences among the simulations. 21 Table 4-4 gives selected solutions to the three models for total demand, value added, factor use, and related variables. The first solu- tion in each case gives the maximum output obtainable with the 500 units of capital specified in the previous section. As already indicated, only the high-substitution model permits full employment of 750 units of labor within this capital constraint. Solutions 2 and 3 for each model show the effect of adding 20 per- cent (100 units) and 40 percent (200 units) to the increment to the capital stock through external borrowing. The trade deficit in the terminal year is increased to 50 and 100 to reHect a steady increase in the inflow of capital over the ten-year period. rOle use of common constraints facilitates comparison of the models under surplus-labor and full employment conditions. In addition, solutions M* and L * indicate the points at which each model shifts from surplus labor to full employment (the Lewis-Fei-Ranis commercialization points) .28 TIle behavior of the model under surplus labor conditions is illus- trated by specification L. So long as there is surplus labor, the wage- rental ratio remains at the minimum of 0.25 and commodity prices in the economy are constant in real terms. The real prices of labor and 27. Further details of the solutions to the maximizing model are given in Kuran (1979). 28. See Fei and Ranis (1964, pages 208-10). More accurately, this is the point at which disguised unemployment disappears and wages are determined by com- petitive forces. 168 SUBSTITUTION AND STRUCTURAL CHANGE Table 4-4. Effects of Increasing External Capital Factor Use Factor Prices Shadow Real price of Total Trade Real rental foreign Solution" demand Labor Capital deficit wage" ratel> exchange L1 996 507 500 0 0.392 1.566 1.003 L2 1,231 596 600 50 0.392 1.566 0.947 L3 1,461 683 700 100 0.392 1.566 0.899 L* 1,697 750 823 162.5 0.392 1.566 0.619 MI 1,059 605 500 0.102 1.606 0.978 M2 M* 1,302 1,483 707 750 600 693 ° 50 96.5 0.402 0.402 1.606 1.606 0.923 0.714 M3 1,518 750 700 100 0.532 1.466 0.872 HI 1,175 750 500 0 0.530 1.529 0.941 H2 1,359 750 600 50 0.668 1.349 0.900 H3 1,526 750 700 100 0.820 1.174 0.843 a. L solutions assume low substitution in production; !'vi solutions assume medium substi· tution; H solutions assume high substitution. Solutions are numbered to correspond to points in figure 4·6. capital therefore remain constant as does the capital-labor ratio in each sector. 29 This represents the capital-widening phase described by Fei and Ranis (1964). An inflow of external capital has a maximum effect on employment and output under these conditions. This is shown graphically in figure 4-6, which shows the increase in output, and in figure 4-7, which shows the decline in surplus labor. The productivity of external capital is indicated by the slope of the output-capital curves. Under full employment conditions, an inflow of capital raises the productivity of labor quite substantially but has somewhat less effect on GNP than in the surplus labor cases. \\Then the capital stock is increased from 500 to 600 units with high substitution (that is, H 1 to H2), the increment in GNP is only 77 percent as great as in the surplus labor specification (L1 to L2). The difference between Land H be- 29. Real prices are determined by dividing the shadow prices of factors by the cost·of·jiving index. The latter is an average of commodity prices weighted by base· year demands. SURPLUS LABOR AND EXTERNAL CAPITAL 169 Trade Real Value Value Real Real Wage labor Real Share of of capital value rental in- capital of imports exports inflow" added ratio come income labor 94.5 94.5 14.3 982 0.25 199 783 0.203 119.2 69.2 57.5 1,173 0.25 234 940 0.199 147.9 47.9 97.3 1,364 0.25 268 1,096 0.196 162.5 0 114.0 1,583 0.25 294 1,289 0.l86 106.7 106.7 13.0 1,046 0.25 243 803 0.232 135.0 85.0 54.7 1,248 0.25 284 964 0.228 96.5 0 81.0 1,402 0.25 289 1,113 0.206 106.6 6.6 93.2 1,425 0.36 399 1,026 0.280 71.9 71.9 12.4 1,162 0.35 398 764 0.342 78.4 28.4 49.0 1,310 0.50 501 809 0.382 100.0 0 89.4 1,437 0.70 615 822 0.428 h. See footnote 29. c. Capital inflow is equal to total demand minus real value added; it is valued in domestic prices aud differs from the trade deficit by the inclusion of implicit tariffs and subsidies. comes greater with further increases in external capital as a result of diminishing returns to capital. The more realistic medium specification illustrates both phases. Since it assumes a moderate degree of substitution in each sector, the capital-output ratios are lower than in specification L and the increase in GNP greater in the surplus labor phase. Once full employment is reached at point M*, the productivity of capital begins to decline but is still relatively high. The distribution of income between labor and capital that is implied by the factor prices of the solutions is shown in table 4-4. Although a substantial fraction of the return to capital may in some countries accrue to the state and be redistributed to low-income groups, the effect of surplus labor on wages and income distribution has been shown to be a significant factor in explaining worsening income distribution. 30 If we take labor income as a proxy for the income of lower-income 30. Some of the available evidence from country studies is reviewed in chapter II. 170 SUBSTITUTION AND STRUCTURAL CHANGE Figure 4-6. Output Effects of Capital Inflow 1,600 L* 1,500 1,400 1,300 "0 the amount of commodity i used in sector i being dijX j · Production activities (X,) require inputs from other sectors in the model and also from outside the system (inputs 8 through 10). Import activities (Mj ) require an input of foreign exchange in the amount indicated by the coefficient in line 7 for each unit of product supplied. Imports are possible for commodities 1, 2, 3, and 5. (For all activities, the com- modity supplied is indicated by the subscript.) (c) The first eight equations in the model constitute a set of restric- 180 THE INTERDEPENDENCE OF INVESTMENT DECISIONS Table 5-1. The Model of Production Activities' Equation Commodity Ml Xl M2 X2 Ma (5.1 ) Metal products' 1.0 1.0 (5.2 ) Iron and steel -0.22 1.0 1.0 (5.3 ) Iron are -O.OB I.e (5.4 ) Electric power -0.01 -0.02 ( 5.5) Coal -0.10 (5.6) Transport -0.01 -0.02 (5.7a) Foreign exchange (a}-0.B5 (a)-L2 -1.1 (5.7b) (b)-0.BI5 (b)-LOS (5.8) Other inputs -0.17 -0.09 (5.9) Labor -0.7 -0.2 (5.10 ) Capital -0.7 -2.7 a. All input coefficients are measured in value for each unit of output except labor, which is in man-years. The value units for outside demand (and hence for labor) are arbitrary. b. "Metal products" refers to machinery, vehicles, and other products in the 36, 37, and 38 categories of the International Standard Industrial Classification. tions on the possible levels of production, imports, and exports. Each equation is formed by multiplying each coefficient in the row by the corresponding activity level (Xj, Mj, or E j ) and setting the total equal to the demand outside the system. For example, the equation for iron and steel, reads: (5.2 ) 0.05Xs - 0.OlX4 O.OIX, 0.02X 6 = 1,000. The total supply is given by (M2 X z ); the total use of iron and steel in the sectors within the system is 2.jd 2j X j ; and the "outside" use in the rest of the economy is 1,000. Outside demands are assumed only for the first two commodities, since the existence of outside demands for the remainder does not affect the nature of the solution.7 The 7. Production or import levels ealculated for the remaining eommodities can be considered as increases above a given level, which is not affected by the invest- ment ehoices in the seetors being analyzed. THE MODEL 181 Activities' Outside Given Xa X4 M5 X5 Xs E7 Xs demand prices 1,000 -0.05 -0.01 -0.01 -0.02 1,000 1.0 0 -0.02 1.0 -0.03 0 -0.25 1.0 1.0 -0.07 0 -0.50 -0.20 l.0 0 1.0 l.0 0 -0.10 -0.08 -0.17 -0.10 1.0 0 -0.3 -0.4 -0.7 -1.0 -1.0 1.5 -0.5 -0.7 -2.5 -2.2 1.5 l.0 equation for foreign exchange supplied and demanded has a similar form: (5.7a) (For reasons explained below, alternative assumptions a and b \\'111 be made as to the magnitude of the coefficients specifying the cost of imports of commodities 1 and 2.) (d) The use of primary factors, capital and labor, is shown in equa- tions (5.9) and (5.10). For these unproduced inputs, no restriction is placed on supply since only a small range of variation will be con- sidered. Instead, prices are assumed to be given by conditions in the rest of the economy, with the price of capital arbitrarily set at 1.0. At a later stage, economies of scale will be introduced by making average capital and labor coefficients a declining function of the level of out- put. (The activity analysis model can readily handle supply limitations in a more general case.) A solution to a programming model such as this consists of a set 182 THE INTERDEPENDENCE OF INVESTMENT DECISIONS of nonnegative activity levels which satisfy equations (5.1) through ( 5.8). In linear programming, it is necessary to consider only basic solutions-those which have only as many positive activity levels as there are equations. A similar rule holds when there are economies of scale, so that almost all of the solutions with which I shall be con- cerned are basic solutions. In the present model, a basic solution will contain one activity having a positive output for each commodity restriction. TIle total number of such combinations is two to the fourth power or sixteen, in the present example, since there are alter- native sources for four commodities. s There are only six different solu- tions arising from these possible combinations, however; they are given in table 5-2. The coefficients used in table 5-1 are intended to be realistic, but to avoid local peculiarities I have not used the actual data of any one country. The selection of sectors and of the input-output data (apart from steel) was based on a comparison of interindustry struc- ture in four countries. 9 Latin America data were used for the steel industry,I° and Japanese labor and capital coefficients for the remain- ing sectors.u Import prices and export costs are hypothetical. The proportions of external demand for steel and metal products are ini- tially fixed at arbitrary levels, but the effect of varying them is consid- ered explicitly in a later section. Prices and external economies Since I wish to isolate the effects of interdependence in production, I assume that income is inereasing and investment is taking place at a given rate in the economy as a whole. The level of income at any time determines the specified demands for steel and metal products, while the investment opportunities in the remainder of the economy fix the marginal productivity of investment and the opportunity cost of labor. Prices of "other inputs" are determined by labor and capital costs, 8. Although it would be quite simple to include alternative techniques of pro- duction in each sector, this was not necessary in the present case because exports and imports provide an alternative to local production. Since the more important choice in these sectors is between imports and the most efficient production tech· nique, I have limited the possibilities to these two. 9. Chenery and Watanabe (1958). 10. UN Economic Commission for Latin America (1954). 11. From Watanabe (1961). Table 5-2. Basic Solutions to the Model Factor use Activity level Capi- Solution Ml Xl M2 X2 Ms X3 Xi Ms X" Xa E' 1 Xg Labor" tal" Total 1,000 1,000 0 1,865 187 2,052 4,383 7,461 '- 00 ° 1. 2 1,000 1,000 1,220 1,001 ° 20 ° ° 80 10° 1063 113 80 1,285 20 1,009 10 301 196 2,296 1,425 4,035 5,332 7,479 7,470 IN 3. 1,000 1,007 25 83 815 206 1,357 5,231 7,267 4. 1,000 1,221 98 35 133 34 241 313 1,532 5,195 7,493 5. 1,000 1,229 98 41 141 112 324 1,446 5,065 7,234 Note: The total of sixteen possible combinations of imports and production in sectors 1, 2, 3, 5 is reduced to six because when steel is imported, demands for ore and coal are zero and solutions 0 and I each represent four possible bases; I have omitted the possibility of importing ore and producing coal or vice versa because it does not arise under my assumptions as to the extent of coordination. a. Exports under import assumption b. b. Labor and capital in case I. 184 THE INTERDEPENDENCE OF INVESTMENT DECISIONS since the internal structures of the industries producing them are omitted from the model. Except as specified, prices will be assumed to satisfy the conditions of marginal cost pricing.J2 An initial position is assumed in which each commodity is either produced or imported, which is a basic solution. I also assume a price of labor of 1.5, which is its opportunity cost. TIle commodity prices and the price of foreign exchange can then be cal- culated from the condition that price equals marginal cost. (In case I, with no economies of scale, marginal cost is also average cost.) TIlis calculation involves solving eight simultaneous equations, one for each activity, of the following form: (5.11 ) 1 ... 8) where dij is the (marginal) input coefficient for input 1 in activity;. Prices as thus defined are the same as the shadow or equilibrium prices of a programming system for the linear case except for the exogenous labor input, whose price is givenP TIle relative prices with which the economy starts are determined by the source of supply-from domestic production or imports-of each commodity and the cost of securing foreign exchange. I will assume that initially all of the commodities that can be imported- metal products, steel, iron ore, and coal-are imported. 14 Once the price of forcign exchange is determined, their prices will be given. TIle price of "other inputs," P s, is readily computed from the price of labor and capital to be 3.0. By substituting P s - P lO into equation ( 5.11) for the export activity, the price of foreign exchange is deter- mined to be 4.0. The price of each imported commodity is then found by multiplying its eost in foreign exchange by the price (opportunity cost) of foreign exchange. The same procedure is followed in later solutions. 12. For simplicity, I ignore differences in the durability of capital and risk among sectors and assume that the gross rate of return required by investors in each sector is the same. Variations in these factors could readily be introduced but would serve no useful purpose in the present context. 13. The general formulation used in activity analysis and the economic inter- pretation of shadow prices are discussed in Dorfman, Samuelson, and Solow (1958). 14. Except for iron ore, for which there is no demand if steel is not produced. THE MODEL 185 TIle prices of the domestically produced commodities in this and subsequent solutions must be determined simultaneously, since each sector (except the first) sells to one or more of the others. To facilitate this solution, the matrix of coefficients has been arranged in order of maximum triangularity-that is, the elements above the diagonal in the solution are reduced to a minimum.l~ The set of initial prices is shown in the first row of table 5-3. Starting from this initial position, in which demands for commodi- ties 1 and 2 are supplied through imports, I shall measure the effects of various factors which would render investment in one or both sectors profitable. In each case, a calculation will be made of the amount of investment that will take place under alternative assump- tions as to investors' reactions. 16 The social efficiency of these reactions will be measured by the reduction in the total cost of supplying the given demands. Total cost in tum equals capital plus labor used, with labor valued at its opportunity cost of 1.S. The alternative assumptions about investors' behavior can be de- fined as: (a) individual reactions-investment will take place in sectors that yield profits greater than Or equal to the existing marginal productivity of capital (taken as 1.0) at present prices and in amounts determined by present demands for the commodity; and (b) co- ordinated reactions-investment will take place in sectors and in amounts that together will supply the outside demands at thc mini- mum total cost. (For the linear case I, this assumption can be stated in terms of individual profitability at futurc prices.) 111ese assumptions represent the extreme range between no fore- sight and socially optimum decisions on investmcnt. TIley are not intended as descriptions of the actual behavior of unplanned and planned economies, but as a basis for cstimates of the maximum dif- ference in performance. The second part of thc assumption about individual reactions will not be applied to sectors in which domestic production is already es- tablished, because no investment could take place in a power-using 15. The solutions for both prices and quantities were made with the Gauss- Seidel method of iteration. The method as applied to input-output systems is explained in Evans (1956). 16. The maintenance of the original pattern of supply also requires investment in exports, but it will keep the original prices unchanged. 186 THE INTERDEPENDENCE OF INVESTMENT DECISIONS industry, for example, without an expansion of power production. The amount of induced investment required in sectors where production is already established will therefore be assumed to take place even with individual reactions. This assumption is necessary in order that a given set of alternatives be feasible-that is, that they satisfy all the restrictions. The cost of each set of alternatives can be measured in one of two ways. The first is to calculate the production reqHired in each sector and from this result to determine the amount of labor and capital needed throughout the economy. lbis has been done for the six basic solutions in table 5-2. In the initial situation (solution 0) the outside demands are supplied from imports, and production takes place only in sectors 7 and 8, exports and "other inputs." The total cost of this production is 2,052 units of labor and 4,383 units of capital. Using the assumed opportunity cost of 1.5 for labor, the total cost of this alternative is 7,461. TIle total cost of the other alternatives is calcu- lated in the same way. The use of prices provides a second method of calculating the total cost of each set of alternatives, which is more interesting from an economic point of view. Each price represents the total capital used directly and indirectly to produce a unit of net output. In the initial situation of example B in table 5-3, the prices of commodities 1 and 2 are shown as 3.26 and 4.20. Multiplying the outside demands by these prices and adding gives 7,460 as the total cost as before (except for rounding). This result corresponds to the dual solution of a linear programming system. 17 Under the assumptions made, any profitable investment will neces- sarily reduce the total cost of factors required to supply the given final demands. The difference between the amounts required under the two assumptions about investors' reactions provides a measure of the quantitative importance of the interdependence of investment deci- sions. In considering some types of policy, it is desirable to allocate this difference to individual sectors. \Vhen this is done, we will have a partial measure of the type of external economies that Rosenstein- Rodan had in mind: the difference between social and private profit- ability resulting from the recognition of interdependence. 17. The calculation of the price solution is explained further in the appendix to this chapter. Table 5-3. Effects of Coordi1U1tion on Profits and Prices: Case I Solu- Profitability of investment" Prices Investment assump- tion Total tionb N 2 3 5 2 3 4 6 7 cost" Example Initial posi- tion a 0 3.40 4.80 4.40 5.04 4.00 4.43 4.00 Individual in- vestment in -0.01 , 2 a 2 +0.58 - 3.40 4.22 4.40 5.04 4.00 4.43 4.00 7,620 Coordination of 1 and 2 a 4 +0.12 +0.59 - 3.28 4.21 4.40 5.03 4.00 4.42 4.00 7,490 Example B: "- Initial posi- 00 'l tion b 0 3.26 4.20 4.40 5.04 4.00 4.43 4.00 7,460 Individual in- vestment in exports b 0 -0.01 -om +0.70 +1.42 3.26 4.20 4.40 5.04 4.00 4.43 4.00 7,460 Coordination of 2, 3, 5 b 3 -0.01 +0.20 +0.70 +1.42 3.26 4.00 3.70 4.68 2.58 4.32 4.00 7,260 Example C: Coordination of 1, 2, 3, 5 a 5 +0.17 +0.79 +0.70 +1.42 3.23 4.00 3.70 4.68 2.58 4.32 4.00 b 5 +0.03 +0,20 +0.70 +1.42 a. The necessary expansion in sectors 4 and 6 is assumed in aU cases. b. Import assumptions: (a) m 1 = 0.85, m 2 1.2; (b) m 1 = 0.815, m 2 = 1.05. c. Profitability for each unit of output at input prices indicated, with output price that of the initial position. d. Total cost equals (1000 PI + 1000 c. No investment in the sector. 188 THE INTERDEPENDENCE OF INVESTMENT DECISIONS Measurement of the Effects of Interdependence Investment will take place in an industry only if its marginal pro- ductivity is greater than or equal to that assumed for the rest of the economy.18 Assuming that investments that were profitable in the past have already been made, one of the following conditions must obtain if investment in steel or metalworking is now to become profitable: (a) technology must be available that is more efficient at ex- isting prices of inputs and outputs than that already in use; (b) the cost of one of the exogenous inputs must fall; (c) the cost of obtaining foreign exchange must rise (that is, there should be a change affecting activity E7 in table 5-1); ( d) planning must be coordinated, provided unused resources exist in one of the supplying sectors; or ( e) demand must be expanded. These factors may produce investment in one or several sectors, and the results mayor may not be different as between individual invest- ment decisions and coordinated decisions. In this section I shall try to identify the more important types of external effects-cases in which the results of the two assumptions are different-and to measure their quantitative significance in the example I have chosen. It will be useful to separate the case in which there are internal economics of scale from that with constant costs throughout, since it is often asserted that external economies are merely the result of inter- nal economies elsewhere in the economy. I shall take up the constant cost assumption first as case I, since it is analytically simpler although empirically less important. To make the analysis easier to follow, I shall use the same set of interindustry data throughout (equations 5.1-5.6). 111is permits the use of the basic solutions given for sectors I through 8 of table 5-2 for both cases. Economies of scale are assumed to affect only the labor and capital coefficients, which are outside of the interindustry system. For each case, the same three examples, illustrating some of the 18. I assume no difference among plants in an industry except for the scale effects discussed under case II below. MEASUREMENT OF THE EFFECTS OF INTERDEPENDENCE 189 initiating factors listed above, will be worked out and the magnitude of the external effects measured. Case I: external effects with constant costs The principal effects of interdependence in production can be classi- fied as effects on users and effects on suppliers. 19 Each type will first be illustrated separately by assuming only partial coordination among in- vestment decisions, and then their combined effect will be shown by assuming complete coordination. EXAMPLE A: EF.l"ECTS ON USERS. External effects of investment in in- dustry Y on users of commodity Y can be illustrated by Scitovsky's example of the effects of an innovation in Y. Assume that it is now profitable to invest in the production of stcel, which had previously been imported. In example A of table 5-3, the import price of steel has been taken as 1.2 units of foreign exchange to illustrate this possi- bility. The initial prices, as calculated above, are given in the first line for all commodities. The import price of metal products has been set just under the cost of production with current prices, however, so that without coordination investment will take place only in steel. The effect on steel cost and the total cost are shown in the second line. If coordination takes place, however, investment in sector 1 will also be profitable because of the lower cost of steel, and a larger investment in sector 2 will be needed to supply the increased demand. (This is the situation envisaged by Scitovsky in the example previously cited.) No change in the source of supply of other inputs is assumed, so the reduction in the total cost of supplying the given demands-from 7,620 to 7,490-can be attributed entirely to the coordination of in- vestment in steel and the steel-using industries. Others of the initiating factors listed above-a fall in the cost of labor or capital, or a rise in the cost of securing imports-can produce the same effect. It should be emphasized that, under the assumption of constant costs, the difference between uncoordinated and coordinated invest- ment in this case is only one of time. Once the investment in steel has 19. I have abstracted from less direct effects, such as the use of common factors of production, by assuming an elastic supply of exogenous inputs at constant costs over the relevant range of demand. All produced inputs are therefore available at constant cost. 190 THE INTERDEPENDENCE OF INVESTMENT DECISIONS been made, it will be profitable to invest in metalworking unless the priee of steel is kept above its east. Furthermore, if both seetors are unprofitable at the intial prices of imports, coordination will not make them profitable. Both of these conclusions will be changed in succeed- ing examples. EXAMPLE B: EFFECTS ON SUPPLIERS. Vilhen expansion in one sector increases demands for inputs, no external economies are created if the price of the inputs reflects the opportunity costs of the factors used to produce them. \Vhen increased demand in one sector leads to a de- mand for immobile factors which have no alternative uses, or for commodities produced from immobile factors, however, the situation may be different. In the present example iron are and coal illustrate this possibility. TIle market for them is limited by transport costs, particularly in areas where transport facilities are not well developed. 20 If an industry that can use them locally is established, however, they may be much cheapcr than would the imported material, as has been assumed here. Example B in table 5-3 illustrates the effect of investment in the steel industry on the profitability of investment in coal and ore sup- plies. If on the one hand it is assumed that there is no market for these commodities outside the region because of high transport costs,21 additional investment in these seetors will not take place un- less steel production (or some other use) is established locally. On the other hand, steel production on the basis of imported materials is un- profitable. In this ease, there will be no investment in any sector with- out coordination of all three. With coordination, the cost of suppl;ing the existing demand for steel is reduced from 4,200 to 4,000. A similar external effect on suppliers may be produced by a reduc- tion in the cost of a factor whieh affects profits in all seetors. Assume that the opportunity east of labor is reduced (for example, by using its calculated value instead of market cost) from 1.5 units of capital to 1.0. (A reduction in the supply cost of capital or a rise in the cost of foreign exchange would have a similar effeet.) Investment in both sectors 1 and 2 will then become profitable to individual investors, but 20. The case becomes more significant when one makes the more realistic as- sumption of economies of scale in transportation. 21. An example is provided by the coal and iron ore deposits in Colombia, which cannot be economically transported to the coast. MEASUREMENT OF THE EFFECTS OF INTERDEPENDENCE 191 the amount of investment undertaken in sector 2 will be 18 percent less than if there is coordination, since the demand from sector 1 will not be taken into account initially. The difference in the cost of sup- plying the given demands is 130 in this case. 22 As in example A, coor- dination will affect only the timing of investment because it will be profitable to expand investment in sector 2 when the demand for steel from sector 1 becomes apparent. EXAMPLE C: EFFECTS ON BOTH SUPPLIERS AND USERS. Example C in table 5-3 shows the effects of complete coordination of both the sup- pliers and the users of the steel industry. In this case, all commodities are produced locally and nothing is imported. As compared to partial coordination in example A, prices drop in both sectors I and 2, and there is a reduction of 260 in the total cost of supplying the given de- mands. As compared to partial coordination in examplc B, only the price of metal products drops and the saving is much smaner. A comparison of example C with the initial assumption of example B shows the maximum external economies that can be attributed to coordination alone in the present example when there are no econo- mies of scale. In the initial position, no investment is profitable by itself, and demand would continue to be supplied from imports at a cost of 7,460. Coordinated invcstment in all sectors reduces the cost of supplying the same demands to 7,230. l11e saving is attributable to the existence of local resources of coal and iron ore which can be economically exploited with coordination but not otherwise. Case II: external eRects with economies of scale The introduction of economies of scale not only makes external ef- fects more important in the examples given previously, but it makes possible some types that do not exist in the case of constant costs. These will be taken up after the empirical basis for introducing econ- omies of scale has been discussed. THE NATURE OF ECONOMIES OF SCALE. Despite the theoretical im- portance of economies of scale, their quantitative significance has been 22. The calculation is not shown because the mechanism is similar to that in example A. 192 THE INTERDEPENDENCE OF INVESTMENT DECISIONS investigated only in a limited number of industries. The available evidence suggests that the economies of producing a larger volume of output occur mainly in the direct use of capital and labor and in in- puts (maintenance and overhead costs of various kinds) related to them. The quantities of materials needed to produce a given com- modity seem to vary little with output unless the increase in scale makes possible the use of a different type of process. The value of such materials may fall with increases in the amount purchased, due to internal economies in other sectors, but this results from the work- ing of the model itself. Most studies of scale effects apply to plants or processes rather than to whole industries. 'I11e determination of cost variation for a whole industry must take into account location factors and market struc- ture-for example, whether the increase will come from one plant or several, from new plants, the expansion of old, and so forth. In the present example, these problems are important mainly in the metal- working sector because in steel and its suppliers it can be assumed that the increase in output will come from a single source. Steel is the sector in which economies of scale are of greatest impor- Table 5-4. Economies of Scale in Steel Production Capacity of plant" Decrease fTOm 50 Cost for each tonh 50 250 500 1,000 to 1,000 Raw materials 33.84 31.26 31.26 25.68 8.16 Maintenance and miscellaneous 20.59 11.11 10.57 9.83 10.76 Capital charges 122.93 101.20 87.10 85.05 37.88 Labor cost 32.00 15.20 8.57 6.60 25.40 Total cost 209.36 158.77 137.50 127.16 82.20 Total investment for each ton 492 405 348 340 152 Note: This table is adapted from data in UN Economic Commission for Latin America (1954, pp. 112-16). a. Capacity in 1,000 tons of finished steel a year. b. The costs (in dollars) are taken from engineering calculations for hypothetical integrated plants of different sizes located in the eastern United States. Labor costs are taken here at 50 percent of U.S. costs and charges for depreciation and profit at 25' percent of capital invested to reflect Latin American conditions. (These are not the capital charges used in the original study, which are unrealistically low.) Data for iron, steel, and finishing stages have been consolidated. MEASUREMENT OF THE EFFECTS OF INTERDEPENDENCE 193 tance in the present case. Table 5-4 summarizes some of the results of a detailed study by the UN Economic Commission for Latin America of economies of scale in the steel industry based on design data for plants of various sizes. Production cost in the small cst plant consid- ered is 65 percent higher than in the largest, and this does not exhaust the economies that are possible with a smaller range of products. 23 To use these data in the interindustry analysis, I have based the in- put coefficients of table 5-1 on a plant of 250,000 tons, which is typical of Latin American steel production. Economies of scale will be as- sumed only in the use of capital and labor, which account for perhaps 90 pcrcent of the cost reduction shown in table 5-4 if the price of inputs is kept constant. 24 For this and the other sectors of the model, capital and labor will be treated as a single input, which is represented by a linear equation 25 ; (5.12) where f. is total use cost of capital and labor at the prices assumed, fi is a constant, and 'Yi is the long-run marginal eost of labor and capital. A linear function fits the data for steel production eosts quite well above 250,000 tons. At the representative size of plant chosen, the ratio of marginal to average cost of labor and capital inputs is about 0.67. This ratio has been used to determine the input function for sector 2 in table 5-5, with the constant term fixed so as to equate total cost in cases I and II at the initial demand of 1,000. The input functions for the remaining sectors were established on the samc principles but on a hypothetical basis. 26 It would be hard to 23. The plants are designed to produce 80 percent of the range of steel products typically demanded in Latin American countries. The remainder would not nor- mally be economical and would be imported. See table 5·8. 24. All of the reduction shown in the cost of raw materials is due to economies of scale in the transport sector. 25. Capital and labor inputs in chemical process industries (which include metallurgy) have been found to confonn quite well to a relation of the following fonn; f to (X/Xa) ..p. Wessel and Chilton (1952, 1953) give engineering data for chemical plants in which the value of if; averages about 0.6 for capital and 0.2 for labor. if; is the ratio of marginal to average cost. Equation (5.12) may be regarded as a linear approximation to this function which holds over a specified range. 26. The greatest economy of scale is assumed in transportation, the least in mining. 194 THE INTERDEPENDENCE OF INVESTMENT DECISIONS Table Input Functions for Labor and Capital: Case II Input function Total cost equal Sector for combined input' to case I at" MC/AC' 1 f, = 500 + 1.25 X, X, = 1,000 0.71 2 f2 = 1,000 + 2.00 X 2 X2 1,000 0.67 f3 = 0.80 x'3 0.84 3 4 5 f4 = 2.5 X, fo 1.1 X, Xa X. = x'5 = 0 ° 0.66 0.85 6 fs 1.75 Xs X6 = 0 0.49 7 fr = 3.7 X7 All values 1.00 a. Expressing labor in eapital units at the ratio of 1.5 : 1.0. b. That is, output at which total cost for labor and capital are the same in both case I and case II. c. Marginal cost in case II divided by average eost in case I. specify a typical situation for mining, transport, and power production without assuming a specific location. No economies or diseconomies of scale have been assumed in ex- ports. Total cost has been assumed to be equal for each sector in cases I and II at the values of exogenous demand for each commodity, which are those that the individual investor takes into account.27 The form of the analysis is such that it can readily be adapted to use cost studies of the kind usually prepared in connection with investment programs. A COMPARISON OF CASES I AND II. When economies of scale exist in supplying sectors, external economies will be larger because the in- crease in demand from the using sector will make possible cheaper production by the supplier. I shall now analyze the same examples as before with the assumptions just made as to input functions for capi- tal and labor in order to show the significance of introducing econo- mies of scale. Table 5-6 has been constructed by applying the alternative input functions of cases I and II to the basic quantity solutions given in 27. This assumption results in a constant term of 500 in sector 1 and of 0 in the other sectors. It might have been more realistic to assume a constant tCl1n in sectors 3 and 5, where no production exists in the initial position, but the differ- ence in result would be small. 5-6. Direct Factor UBe by Sector and External Economies: Cases I and II Difference from External Sectors Econo- Solu- Direct factor use in sectorb Total initial position econo- coordi- mies of tion factor miesin nated" scale N 1 2 3 4 5 6 7 usee Case I Case II sector 2 Initial Position 0 0 0 0 0 0 0 7,460 7,460 Example A 2 I 2 0 3,272 0 80 0 81 4,038 7,471 +11 '- 1,2 I 4 2,260 3,992 0 138 0 140 961 7,491 +31 '-0 I.rt 1,2 II 4 2,260 3,772 0 95 0 78 961 7,166 -294 0.030 Example B 2,3, 5 I 3 0 3,295 100 100 174 337 3,260 7,266 -194 0.020 2,3,5 II 3 0 88 69 151 188 3,260 7,043 -417 0.046 Example C 1,2,3,5 I 5 2,260 4,019 123 162 217 453 0 7,234 -226 0.023 II 5 2,260 3,790 108 III 189 0 6,710 -750 0.097 a. Import assumption b is used for all three b. Direct factor use in the sector includes the use of "other inputs." c. Figures vary slightly from those given in tables 5·2 and 5-3 because of rounding. 196 THE INTERDEPENDENCE OF INVESTMENT DECISIONS table 5_2.28 Unlike case I, a coordinated investment will be profitable with the given economies of scale without any innovation, and so I have assumed the import prices of example B throughout. The use of factors in each sector shows that in example A most of the difference between the two cases comes from the larger output in sector 2, but in example B the difference is almost entirely due to the other supply- ing sectors. The difference is most pronounced in example C, where investment in all sectors is coordinated, because in this case the econo- mies of scale in the supplying sectors (particularly transportation) reach substantial proportions. The allocation ot external economies by sector The savings in factor use that result from coordination pertain to the whole set of investments, rather than to anyone of them, because all are necessary to the result. If all investments in the economy were centrally planned, there would be no need to try to allocate this type of external economies because the determination of the optimum in- tegrated plan would be sufficient. In economies in which not all invest- ment is under government control, however, the question arises as to how external economies can be taken into account in policies designed to improve the efficiency of individual investment decisions. This problem arises concretely in less developed countries in the attempt to establish investment priorities for the economy as a whole or within certain areas, such as manufacturing. These priorities are intended to guide the allocation of loan funds or foreign exchange, or to be used as a basis for other measures by which the government in- fluences investment. The rationale for allocating the benefits of coordination to one sec- tor or another must derive from the institutional setting. In the ex- treme case, it may be assumed that if one investment is made and its output sold at the price of the optimal solution, then investment in supplying industries and using industries \\ill follow if the return on these investments is equal to the marginal productivity of capital else- 28. For case I, this calculation serves as a check on the rcsults of the price solu- tion given in table 5-3. For case II, marginal cost prices can also be used to deter- mine total factor use, but the quantity solution is the more convenient. The price solution for this case is given in the appendix to this chapter. \Vith economies of scale, the optimal solution will be a basic solution, as it was in the linear case. MEASUREMENT OF THE EFFECTS OF INTERDEPENDENCE 197 where in the economy. In the present example, investment in the steel industry may have this effect, and, in fact, many governments under- take direct investment in thc steel industry partly to stimulate invest- ment in related industries. I shall therefore calculate the effect that taking account of external economies has on the profitability of invest- ment in steel. External economies may be thought of either as an addition to the value produced by a plant or as a reduction in its cost. Since I have assumed given demands in this chapter, it is more convenient to take the latter approach. 29 The savings due to the coordination of invest- ment shown in table 5-6 can be taken as being equivalent to a reduc- tion in the investment required in the steel industry. The profitability of investment is increased thereby, and the difference provides a measure of external economies in sector 2. In example IB the calcu- lation would be as follows 30 ; Market Shadow prices prices Price of output 1.050 4.20 Cost of inputs 0.379 1.517 Profit margin 0.671 2.683 Capital per unit of output 2.700 Capital saved per unit of output (194/1 ,000) -0.194 Adjusted capital per unit of output 2.506 Profitability: original 0.248 adjusted 0.268 External economy 0.020 29. In Chenery (1953), I suggested that external economies be treated as an addition to the value added in production, although I did not indicate how they might be measured. If this approach were adopted, the additional value would be the additional production achievable with the factors saved. 30. In each case I have used the scale of investment without coordination (1,000) as the basis of comparison. Input costs are derived by taking the price of capital equal to its marginal productivity (0.25) in exports. 198 THE INTERDEPENDENCE OF INVESTMENT DECISIONS The concept is somewhat more complicated when there are econo- mies of scale in steel production because the average productivity of investment in the plant is less than the marginal productivity of addi- tional increments of capacity. This phenomenon will lead to some overbuilding of capacity for the given demand at any moment in time, a factor which I ignore. Here I have calculated the average produc- tivity of investment in the same way for case II as for case I, with the results shown in the last column of table 5-6. The external economies measured in this way add some 8 to 10 per- cent to the productivity of investment in the steel industry in case I, and up to 40 percent with coordination of all sectors in case II. If pos- sible economies in the supply of exogenous inputs of goods and ser- vices (which account for about 15 percent of all costs) were taken into account, the figure would be somewhat higher. The inclusion of external economies of this magnitude might make the difference be- tween an unpromising steel project and one which should be included in a development program. Size of the market TIle effect of small markets combined 'h1th economies of scale in production is one of the main explanations given for the lack of growth of poor countries. 31 In analyzing external economies, I have so far taken the size of the market as given. In case II, the demand for each commodity was taken to be just below the size which would make investment attractive to an individual entrepreneur. I shall now abandon this assumption and determine the difference that coordi- nation makes in the minimum scale at which investment becomes profitable. I shall consider first sectors 1 and 2 separately and then the optimum pattern of investment for the whole complex. INVESTMENT IN INDIVIDUAL SECTORS. TIle analysis of investment in individual sectors will compare the effect of variation in exogenous demand on the cost of supply under four assumptions: (a) imports; (b) uncoordinated investment in one sector; (c) coordinated invest- ment in supplying sectors 3-6; and (d) coordinated investment in all sectors. The cost of supplying demands of any size in each sector under these assumptions is given by the equations in table 5-7. 31. See Nurkse (1953). MEASUREMENT OF THE EFFECTS OF INTERDEPENDENCE 199 Table 5-7. Cost of Supplying in Each Sector under Various Assumptions Assumption Sector 1 Sector 2 (a) Imports S, 3.26Y, 5. = 4.2Y. (b) Uncoordinated investment 5, = 500 + Z.78Y, S. = 1,000 + 3.22Y. ( c) Coordination of sectors 3-6 5, = 500 + Z.739Y, 52 = 1,000 + Z.782Yz (d) Coordination of 5, = 500 + 2.4Z9Y, 5. = 1,000 + 2.782Y. aU seetors + (1,000 1.4I8Y.) + (500 0.831Y, ) where Y2 < 1,040 where Y 1 < 1,040 Y, > 600 SOUTce: See appendix to this chapter. The differences among the last three equations arise as follows: (b) Uncoordinated investment takes account of economies of scale in the givcn sector but assumes average cost of inputs as given by the initial prices of table 5-3 (example B); ( c) Coordination of sectors 3-6 takes the cost of these inputs as determined from the input functions of table 5-5; ( d) Complete coordination makes the same assumptions as (c) and, in addition, determines the net cost of supplying the given sector -that is, I or 2-when the demand in the other is held constant. If uncoordinated investment in sector 1 is unprofitable but investment in sector 2 makes it profitable, the investment in sector 2 is credited with the difference between the cost of supplying commodity 1 from domestic production and from imports. These functions, therefore, are valid only over the range in which uncoordinatcd investment in the other sector is not profitable. 32 A comparison of the several alternatives is given for each sector in figures 5-1 and 5-2.:33 In each case, I have assumed a single value of production in the other scctor (the general case is taken up in the next subsection). In sector 1, imports are profitable up to a demand of 1,040 at the assumed value of Y2, and investment in sector 1 alone 32. In sector 2 there is no saving over assumption (c) and no justification for investment in I unless Y, is greater than 600-that is, unless the sccond term is negative. In sector 1, the last term can be either positive or negative. H. In sector I, curve (c) is omitted because it is only slightly below curve (b). 200 THE INTERDEPENDENCE OF INVESTMENT DECISIONS Figure . Minimum Scale of Investment in Sector 1: Metal Products (Y. 300) .e: 4,000 ~ a ::: o u ..... o ....., 8 '" ~ 2,000 1.040 1,640 Outside demand (Yl) gives the lowest cost supply from this point to a demand of 1,640. Thereafter, investment in both sectors is profitable with coordination. In sector 2, coordination of the supplying sectors lowers the minimum scale at which investment becomes profitable from 1,040 to 700. Coor- dination of investment in the using sector lowers it still further at the assumed demand in sector 1. TIIE PAT1ERN OF INVESTMENT OVER TIME. The preceding analysis ean be generalized to shed some light on the optimal pattern of invest- ment over time. '\Then investment is completely coordinated, there are four alternative combinations of production and imports that may be most efficient at different combinations of demand: (a) imports of both 1 and 2; (b) production of 1, imports of 2; (c) imports of 1, pro- duction of 2; and (d) production of both 1 and 2. MEASUREMENT OF THE EFFECTS OF INTERDEPENDENCE 201 Figure 5-2. Minimum Scale of Investment in Sector 2: Steel Production (y, = 1,000) (a ~ (b) h (c) (d) 4,000 '0 ~ 8 -;;; '0 2,000 t-< 4,0 700 1,040 Outside demand (Yo) In the analysis of sector 1 above, it was shown that it would be efficient to proceed from A to B to D as demand for metal products expands. I now wish to find all possible efficient expansion paths for this set of production functions. 'The four alternatives can be expressed as combinations of the sup- ply functions in table 5-7, as follows 34 : Alternative Sector 1 Sector 2 A a a B c a C a c D d d 34. The combined cost function for alternative D is shown in the appendix to = this chapter to be; S (500 + 2.429 Y1 ) + (1,000 + 2.782 Y,). 202 THE INTERDEPENDENCE OF INVESTMENT DECISIONS Figure 5-3. Effects of Demand on Optimal Investment 1,500 AreaD (Produce 1 and 2) AreaB (Produce 1, Import 2) AreaC (Import 1, Produce 2) 700 1,000 1,500 Exogenous demand for steel (Y2) TIle areas in which each alternative is most efficient can be de- lineated by solving each pair of equations simultaneously to find the boundary at which the two alternatives have equal cost. The results of this analysis are given in figure 5-3. Figure 5-3 shows that anyone of the three sequences from A to D may be the most efficient, depending on the proportions of demand for the two commodities. The possibilities are illustrated by the three expansion paths shown. Path a represents the ratio Yt/ Y2 3.0 and gives the sequence A-B-D as in the example in figures 5-1 and 5-2. Path f3 has the ratio 1.0 and leads directly from A to D, while path y has the ratio 0.5 and leads to A-C-D. Although this model is much too simplified to permit a direct appli- MEASUREMENT OF THE EFl"ECTS OF INTERDEPENDENCE 203 cation of the results, it indicates at least that optimal industrial devel- opment may take different paths. To take only Latin American exam- ples, it may be suggested that, roughly speaking, Peru and Venezuela recently have been in area A, Argentina in area B, Chile in area C, and Brazil and Mexico in area D, as indicated by the data in tables 5-8 and 5-9 on production and imports. AU of the countries mentioned Table 5-8. Production and Imports of Crude Steel (1,000 metric tons a year) 1955 Imports Production Imports as per- centage Country 1946 1950 1955 1946 1950 1955 of total Argentina 170 250 250 440 870 1,450 85 Brazil 230 820 1,160 430 400 400 26 Chile 30 70 340 100 80 90 14 Colombia 0 10 40 120 190 300 88 Mexico 270 390 730 270 370 340 32 Peru 0 0 0 190 100 Venezuela 50 340 700 860 93 Source: UN Economic Commission for Latin America (1954) . All steel- producing countries in Latin America are included. Table 5-9. Relative Costs of Production of Hypothetical Steel Plants Relative Import Country Capacity cost' prices· Argentina 850 92 115 Brazil 716 85 llO Chile 230 82 III Colombia 250 76 108 Mexico 430 83 108 Peru 150 90 110 Venezuela 300 94 106 United States 1,000 72 SOUTce: UN Economic Commission for Latin America (1954, p. 51). Capacity is based on domestic markets in 1950. a. Based on a 250,OOO·ton plant in the United States as 100. The figures do not include profit on the investment. b. Based on the Pittsburgh price plus freight in 1948. 204 THE INTERDEPENDENCE OF INVESTMENT DECISIONS have local sources of iron ore, but the scale at which steel production becomes profitable varies with costs of production and imports.s5 Evaluation TIle main purpose of this analysis has been to develop a model that could measure some of the principal effects of interdependence on in- vestment decisions. In evaluating the results, it is necessary to see first whether the simplifications made in setting up the model change the nature of the conclusions. I then point out some of the theoretical and practical implications of the results achieved. Theoretical formulation The effects of the simplifying assumptions are the following: THE OMISSION OF INCOME EFFECTS. In taking as given the final de- mands for the commodities studied, I have calculated the saving in factor use resulting from the coordination of investment rather than the increase in income generated. Factor use is measured in invest- ment units, and so the increase in income achievable with the factors saved can be determined by multiplying the total by the marginal productivity of capital (0.25 in export prices). In example II C, an increase in output of 188 (9.9 percent of the outside demands) could be produced with the factors saved. The argumcnt for including additional income effects as part of external economies has not been stated very clearly. If market prices of labor or other resources exceed their opportunity costs, this fact is a source of difference between private and social benefit, but it is not necessarily a result of the interdependence of productive activities. If opportunity costs rather than market prices are used for labor and other resources, the method used here will include this effect in the calculated return to capita1.36 TIle income effects that have not been included are those stemming 35. Steel production has been established with government help in all these countries, but the timing mayor may not have been optimal. 36. This difference may, of course, prevent individual investors from reaching the social1y desirable decisions. EVALUATION 205 from economies of scale in sectors outside the model. An increase in total demand will lower the average cost of production in each sector where scale economies are significant and will give rise to a real saving for the economy. These effects can be measured only in a model cov- ering the whole economy in which resources rather than demands would be taken as given. AGGLOMERATION EFFECTS. Agglomeration effects derive from the physical propinquity of different types of production. In part they consist of the Marshallian type of external economies-creation of a pool of skilled labor, common services, and so forth-which can result from the expansion of one or several industries in one place. In part they are due to a reduction in the physical quantities of certain inputs required-particularly transport and storage. To take account of these factors adequately would require a regional model, but some of them can be included in the prices at which imports are assumed to be a substitute for domestic (or local) production. For example, it is esti- mated that a metalworking plant in Latin America that relies On imported steel has to kecp a stock equal to six to nine months of con- sumption to allow for longer delivery times, while a plant using local supplies needs only a three-month supply.37 'Ine cost of carrying the larger stock should be charged to the import activity. (In the present example, an extra six-month supply of steel increases the capital co- efficient by 0.1 and lowers profits in metalworking by 12 percent.) PARTIAL VERSUS GENERAL EQUILIBRIUM. The model used here is a truncated interindustry system in which the exogenous inputs used in the sectors studied were assumed to be available at constant prices. As the total inputs employed in the omitted sectors represent only about 15 percent of the total factor requirement, the use of a more complete model is not likely to change the calculation of labor and capital requirements significantly. Since I have assumed a given out- side demand, the use of labor is changed only to the extent that pro- 37. One suggestion of the importance of having a local supply of steel is the spurt in metalworking production that has accompanied the establishment of steel production in each Latin American country. See UN Economic Commission for Latin America (1954, pp. 59-67). It is hard to evaluate this experience, however, because the countries concerned suffered from balance of payments difficulties and imports were periodically restricted. 206 THE INTERDEPENDENCE OF INVESTMENT DECISIONS duction in the sectors analyzed is more or less labor intensive in the aggregate than production in the export sectors for which it substi- tutes. It has been assumed here that the industrial complex would use less labor and more capital than the primary production for export which it replaces. In a complete model, the accuracy of the initial assumption as to the relative value of labor and capital could be tested in a complete solution, but it cannot be said a priori whether the external economies calculated with equilibrium prices would be greater or smaller. SUBSTITUTIO~. Substitution in production and consumption is omitted from the present model although substitution in production could be allowed for by using alternative activities with varying input coefficients. Any substitution effects resulting from the price changes produced by investment coordination would add to the total econo- mies achieved. CHANGES OVER TIME. The preceding analysis has concentrated on one future period and ignored the fact that both demands and the expected efficiency of production with new techniques will change over time. In a more complete model, the discounted sum of values and costs in each alternative would have to be compared in order to determine the optimum timing for investment in each. Economies of scale cause plants to be built in advance of the growth of demand,38 and hence the more complete analysis would lower the break-even point between domestic production and imports in both the indi- vidual and coordinated cases. It would probably be lowered more in the coordinated case, particularly if coordination justifies the use of a lower discount rate. A second dynamic phenomenon, the fact that the efficiency of new plants that must train their labor force may be expected to increase over a considerable period, is usually allowed for in static analysis by assuming "normal" operating conditions at some time in the future. Tbis simplification would not lead to a difference between the indi- vidual and coordinated decisions unless different discount rates are used. 38. This phenomenon has been investigated in Chenery (19.52) and is analyzed further in chapter 6 of this volume. EVALUATION 207 Each of these simplifying assumptions is thus seen to be either neutral or to have the effect of understating the magnitude of the external economies resulting from coordination. Of the factors men- tioned, agglomeration effects are probably the most important, as well as the hardest to measure. The other four could be included in the present type of model if it were desirable to do so. DYNAMIC AND STATIC EXTERNAL ECONOMIES. 'The analysis of the ef- fects of investment necessarily assumes that a certain amount of net in- vestment takes place in a growing economy. The cases studied have traced the difference between the response of individual investors to various factors that might make investment profitable-innovation, a change in relative factor prices, unused resources, or the growth of de- mand-and the optimal response of the whole economy. Since these differences are due mainly to the fact that present prices do not provide accurate guides to the optimal allocation of investment resources, it seems appropriate to call them dynamic external economies. If the further adjustments envisioned by static equilibrium theory take place, some but not all of these differences will disappear. In the examples of case I, the external economies resulting from innovation or from a fall in the price of one of the inputs are purely dynamic phenomena. If there is nO further increase in demand, indi- vidual investors win eventually arrive at the same result as coordinated investment, since it is assumed that they are not barred by indivisibil- ities, ignorance, or other market imperfections. 'This does not mean that the economies of coordination are illusory even in this case, how- ever. If a certain amount of investment is being made each year, the increase in output will be higher with coordination than without it, and the rate of growth will be greater.39 The adjustment to a condition of static equilibrium may therefore never be made. In the remaining example of case I, in which coordinated invest- ment is profitable because of unused resources, the elimination of growth does not produce any tendency for individual investors to arrive at the same result as coordination because the reason for the existence of unused resources is the absence of complementary invest- ment in several interrelated sectors. Neither is there any tendency for 39. This case is studied in the context of a programming model for the whole economy in Chenery (1955), where a difference in growth rates of 10 to 20 per- cent is suggested by the difference in marginal productivity of investment. 208 THE INTERDEPENDENCE OF INVESTMENT DECISIONS the difference to be eliminated in the examples of case II, where there are economies of scale. Tne diffcrence between the effects of the initial assumptions of growth (increasing factor supplies) and stationary equilibrium (given factor supplies) is shown in F1eming's criticism (1955) of the use made of external economies by Rosenstein-Rodan and Nurkse. On static equilibrium assumptions, potential external economies from a given investment are largely offset by rising factor costs. \\'hen there is positive net investment and a growing labor force, the question is one of alternative uses of these additional factors, and thcre is no a priori assumption that a eoordinated program that realizes external economies will utilize more of any factor than the alternative invest- ments that would take place without coordination. In fact, it is quite possible that coordination will use less of both capital and labor to obtain the same result as uncoordinated investment. Although I have been analyzing an essentially dynamic phenom- enon, the method used has been that of comparative statics-com- paring the results of alternative behavioral assumptions at a given moment in the future. The method is adequate for the assumption of perfect coordination, and also for the case where there is no tendency to depart from the existing pattern of production and prices, since in both cases the expectations of investors are fulfilled. An explicitly dynamic model would be needed to trace out intermediate cases when the expectations of investors are not fulfilled. I am, therefore, not able to estimate the extent to which uncoordinated investment falls short of the ideal, except in the case where it leads to a perpetuation of the existing sources of supply. All of my comparisons (except exam- ple I A) were made ,vith this limiting case of an unchanging produc- tion pattern. Importance of external economies If my assumptions about economies of scale in the principal sectors are at all realistic, it can be concluded from the analysis in table 5-6 that the economies of coordination are likely to be substantial in the case studied; furthermore, the external economies may be significantly greater if account is taken of the factors omitted from the analysis. I can only speeulate as to the probable importance of external econo- mies in other parts of the economy. Irrigation, for example, is similar to the steel-metalworking case in several respects: the existence of EVALUATION 209 large economics of scale in the supply of inputs (dam building, for example), the significance of the cost of the commodity produced (water) to agricultural processes using it, and the existence of immo- bile resources (dam sites and arid land) without alternative uses. Other overhead facilities-transportation and power-resemble the steel mill in having large economies of scale, although the cost of their product is usually a smaller fraction of the cost of production of users. The most distinguishing feature of overhead faeilities, however, is that their services must be supplied locally, and imports do not provide an alternative source. The case for planned investment in such facilities is, therefore, particularly strong, but the variety of uses which they can serve may make it less important to influence the decisions of indi- vidual investors in using industries. Among manufacturing industries, the example studied is perhaps the most important case. Interrelated chemical process industries, such as petroleum refining and petrochemicals, may provide examples in which the economies of coordination are comparable, particularly when the intermediate products are not readily salable. One may perhaps conclude from this kind of observation that dy- namic external economies are sufficiently important to affect the opti- mal pattern of development throughout the transitional period from a primary-producing economy to one with well-developed overhead facilities and diversified industry. '111e effect of recognizing external economies is to make it more desirable to undertake interrelated ac- tivities together On an adequate scale than to increase production on all fronts simultaneously. Tile existence of dynamic external economies has sometimes been used as an argument for the necessity of a large spurt in investment to get a process of cumulative growth under way. In a closed economy with economies of scale this might be true, but when a large propor- tion of manufactured goods is imported, emphasis can be placed first On one group of investments and then on another. In any case, it is doubtful that democratic governments have much leeway in picking the level of investment, and the more realistic problem is to make the best use of what is available. Implications for development policy The most important policy question raised by the preceding anal- ysis is the extent to which the government has to intervene in order to 210 THE INTERDEPENDENCE OF INVESTMENT DECISIONS secure the bencfits of coordination. l11is is a very large subject, on which I have only a few scattered comments to offer. I will take up three types of mechanism for coordination: (a) integration under private control; (b) the Lange-Lerner system of centrally administered prices; and (c) direct control of investment. PRIVATE INTEGRATION. The main form of private coordination is the integration of several investments under a single ownership or control. It is likely to take place where the external economies are substantial and the sectors involved are not too numerous. TIle exploitation of natural resources provides a common example; mining, specialized transport, and primary processing are often developed together in order to produce a salable commodity when the domestic processing industry does not already exist. From the public point of view, the drawback to private coordi- nation is the large amount of capital required for an integrated investment, which is often not available to a single firm in an under- developed country, and which, if made, leads to a monopolistic posi- tion because of the difficulties of entry. These arguments do not apply so strongly to foreign investment for export, where such integration is very common. \Vith private integration of investment, some of the benefits of coordination are likely to be lost because the capacity of the auxiliary facilities (machine shops, power, transport, and so forth) win usually be designed to satisfy the needs of the integrated firm only rather than to serve other potential users. TIle investment which might be profit- able in other sectors, therefore, may not take place. Private coordination is likely to occur only when it is institutionally feasible to capture a substantial part of the external economies through integration, price discrimination, or otherwise. In the steel complex, the integration of the sectors supplying the steel mill is quite common in less developed countries (although limited in the United States by the antitrust laws), but the integration of steel production and metalworking is less common because of the diversity of products; it is also less desirable socially because the monopoly problem would be made much worse. INDIRECT COORDINATION THROUGH ADMINISTERED PRICES. TIle discus- sion of the Lange-Lerner system of administratively controlled prices and decentralized production decisions has been concerned with the EVALUATION 211 ability of such a system to maintain an efficient level of production in each industry.4o Here the question is whether or not a correct calcu- lation of future equilibrium prices would lead individual investors to the optimal investment decisions. In my case I, '.vith constant returns to scale, the shadow prices of the optimal solution would lead to the right choice of investments, but not necessarily to the right magnitudes. 41 To determine the proper prices, the government would also have to calculate the corresponding quantities and publication of these estimates might furnish adequate guides to the probable demand for various products. Tbis is one of the main functions of a development program. The actual path by which the economy would move from its initial position to a future equilibrium would have to be explored in a dynamic model, but it would appear that, where eeonomies of scale are not too great, prices could serve as the main instrument of coordination, unless the lags in private responses in critical sectors were too long. (The administra- tive problems raised by this procedure are serious but will not be explored here.) DIRECT COORDINATION OF INVESTMENT. Although marginal cost pric- ing (combined with a subsidy or other method of covering total cost) leads to the optimum scale of use of an existing capital good when there are economies of scale-as in the classical railroad examples-it is not adequate to produce the optimum amount of investment in new facilities. To secure the optimal choice of investments in the ex- amples of case II, the total cost of various alternatives must be com- pared. As in the short-run analysis, the marginal conditions determine the optimal seale of output, but the total cost calculation determines whether the investment is desirable at all. 42 The policy implication of this result is that the magnitude of the initial investment as well as the price of output may have to be con- trolled in order to secure all the external economies. 43 If the optimum plant could be built by suceessive expansions, there would be less argu- 40. The argument and the exceptions to it are summarized in Bator (1958). 41. See Dorfman, Samuelson, and Solow (1958, pp. 61-63). 42. This problem is treated in chapter 6. 43. The question of the optimal scale of output and the feasibility of various fOnTIS of marginal-cost pricing are discussed for the case of irrigation by Margolis (1957). 212 THE INTERDEPENDENCE OF INVESTMENT DECISIONS ment for detennining the optimum scale in advance, but in the range of output where economies of scale are most important-whether in dam construction, power plants, or stee] mills-this is unlikely to be the case, Given, then, that some control of the magnitude of investment may be needed, it may be possible to limit it to a few key sectors in an industrial complex. In my example, the construction of the opti- mum size steel mill, power plant, and transport facilities would make profitable to private investors the optimal investment in coal, are production, and metalworking. Of course, if any of these products is monopolized, the price will be higher and the quantity of output lower than the optimum. It is particularly important to prevent monopoly pricing of inputs to other industries because, unless the effects of de- creasing costs are externalized, the investment in other sectors may not take place. The sectors in which correct initial investments are critical to secur- ing thc optimal results are those having the most significant economies of scale and those for which imports do not provide a substitute. These properties are combined in most overhead facilities, where the case for government ownership or control has long been recognized. As be- tween steel and metalworking in the present example, establishment of either one might make the other profitable, but the greater economy of scale and monopoly position of steel argue for its selection.44 111e preceding discussion should not be taken as an endorsement of indiscriminate building of steel mills or other basic industries through government intervention. What has been shown is that the benefits to the economy of such investments may be understated by their ex- pected profitability to an individual investor, and that coordinated planning may tip the balance in their favor. In the absence of any measure of the quantitative significance of external economies, how- ever, the benefits of such investments may easily be overestimated by the governments of less developed countries. TIle main purpose of this chapter has been to present a framework for objective comparison of the alternatives. Given elastic demand and supply and a some- what lower capital coefficient in the export sector of the example used here, the rational policy would be to increase exports not:withstand- 44. The economy of scale in some sectors of metalworking (for example, auto- mobile production) may be equally great and justify government action. APPENDIX. THE CALCULATION OF PRICES AND TOTAL COSTS 213 ing the existence of unused mineral resources and potential external economies. Appendix. The Calculation of Prices and Total Costs Two types of prices are used in the solutions given in table 5-3. Present prices are those to which the individual investor responds in deciding whether to invest. Future prices are those which satisfy the conditions of marginal cost pricing given by equation (5.1 1) after a given set of investments has been made. Present prices are also as- sumed to satisfy equation (5.11) for the sectors included in the model, although the exogenous inputs mayor may not have prices represent- ing their opportunity costs. Future prices are therefore equal to pres- ent prices if the source of supply is unchanged, as in the second line under example B. Case I: no economies of scale Tbe determination of the optimum pattern of investment with co- ordination of all sectors is (in case I) a problem in linear program- ming. It can be stated as follows: to minimize total cost of production, as measured by the prices of the exogenous inputs, subject to the restrictions in equations (5.1) through (5.8) of table 5-1. The dual variables (Uj) or shadow prices of this solution are identical with the future prices and may be defined a5 45 : (5.13 ) (f 1, ... , 7) where Cj is the cost of the exogenous inputs in each activity. 111C solution to equation (5.13) can be determined in a number of ways, including the iterative procedure suggested above and the use of the inverse of the basis. The second method is given by the follow- ing equation: (5.14 ) Uj = l: ri,iCi , i 45. See Dorfman, Samuelson, and Solow (1958, chapter 7). The equation is the same as (5.1) with Cj substituted for the cost of exogenous imports, since ali = 1.0 by assumption. 214 THE INTERDEPENDENCE OF INVESTMENT DECISIONS where Tij is the element in row i and column i of the inverse matrix. For the optimum solution, in which all commodities are domestically produced, the first two columns in the inverse and the calculation of the corresponding shadow prices are shmvn in table 5-lD. As indicated earlier, the total cost of supplying the exogenous de- mand is given by: (5.15) Case II: economies of scale With decreasing average costs, the dual variables can still be defined by equation (5.1 3) with Cj taken as marginal cost, since I have as- sumed constant marginal cost over the relevant range of output. The total cost of production must include all constant terms in the input functions (5.12) of table 5-5, however. The equation for total cost of supply then becomes: (5.16) where the constant terms it apply to all sectors supplied from domestic production. The values of Ul and U2 for case II are also computed in table 5-10. Substituting them in equation (5.16) gives the total cost with coordination as: (5.16a) S = 1,500 + 2.429 Y1 + 2.782 Yz . Table 5-10. Calculation of Dual Variables in Cases I and II Case I Case II Columns in Sector Sector Sector Sector Rows inverse matrix 1 2 1 2 (i) T' l ri2 c, CiTil Ciri. C; Ciril cir'2 I LOOO 0 2.26 2.260 0 1.76 1.760 0 2 0.222 1.007 3.27 0.726 3.293 2.27 0.504 2.286 3 0.Dl8 0.080 1.25 0.022 0.100 1.10 0.020 0.088 4 0.016 0.025 3.99 0.062 0.101 2.74 0.043 0.069 5 0.028 0.113 1.54 0.043 0.174 1.34 0.038 0.151 6 0.029 0.083 4.06 0.Il7 0.337 2.26 0.065 0.188 Dual variables (Uj) 3.231 4.005 2.429 2.782 APPENDIX. THE CALCULATION OF PRICES AND TOTAL COSTS 215 To derive the equations in table 5-7, I assume the outside demands in one sector to be constant and determine the cost of supplying vari- ous levels of demand in the other sector. The overhead costs are allo- cated by holding the cost of supplying the other (fixed) demand constant at the cost of imports, which is subtracted from the total supply cost. The equations in table 5-7 can therefore be derived from equations (5.16) as follows: (5.17) The calculation of the dual variables Uj from equation (5-13) for each assumption is shown in table 5-11. Assumptions The assumptions underlying the supply functions in table 5-7 are as follows: ( a) Imports: the supply functions are merely the cost of im- ports. (b) Uncoordinated investment: the values Ui are taken from table 5-3 (example B, initial position); Cj is the marginal direct cost from table 5-5. ( c) Coordination of sectors 3-6: the u; are those of the coor- dinated solution for sectors 3-6 and the same as (b) for sector 2. ( d) Coord ina tion of all sectors: the values of u j for each sector are those computed in table 5-10. In this case, where there Table 5-11. Calculation of Dual Variables for Equation (5.17) Sector 1 Sector 2 Input cost Input cost Inputs Case (b) (c) (d) Case (b) (c) + (d) 2 .92 .92 .61 3-6 .10 .06 .06 .95 .51 7 .51 .51 .51 .27 .27 Cj 1.25 1.25 1.25 2.00 2.00 Uj 2.78 2.74 2.43 3.22 2.78 Note: Each entry is the corresponding (a'Ju,) from equation (5.13). 216 THE INTERDEPENDENCE OF INVESTMENT DECISIONS is domestic production in both sectors 1 and 2, equations (5.17) become: (5.l7a) 51 5 - 4.2 Y2 = 1,500 + 2.429 Y1 - 1.428 Y 2 , 52 =5 - 3.26 Y 1 = 1,500 + 2.782 Y2 - 0.831 Yl' These equations apply only over the range for which imports are the economical alternative without coordination and in which domestic production in the other sector would be profitable with coordination. Chapter 6 Economies of Scale and Investment over Time with Larry E. Westphal THE ALLOCATION AND COORDINATION OF INVESTMENT is a central fea- hue of developmcnt policy. Among the various factors making coor- dination desirable, economies of scale have particular significance. In formal terms, economics of scale lead to the breakdown of marginal rules of allocation, which are implicitly followcd in a decentralized market economy. 11ms, lack of coordination can result in failure to allocate investment in accord with a country's comparative advantage, as demonstrated in chapter 5. But the significance of economies of scale extends beyond the identification of the optimal mix of activities to the determination of optimal scales of investment, where dynamic eonsiderations assume critical importance. 'Ine issues that arise in coordinating investment across sectors and over time in the presence of increasing returns to scale are discussed in general terms in the next section. The succeeding section narrows the discussion to focus on several elements of the problem of invest- ment choice that have traditionally been of greatest coneern to deve1- This chapter is a revision of Chenery .and Westphal (1969) that takes account of recent work in this field. We are indebted to David Kendrick, Stephen Marglin, and Paul Roberts for helpful discussions on the formulation of the problem and computational problems. Andrew Szasz assisted in thc computations. 217 218 ECONOMIES OF SCALE AND INVESTMENT OVER TIME opment economists, namely, the effects of limited prospects for exports and foreign resource inflows. TIle earlier approaehes are refor- mulated in such a way that formal optimizing procedures can be ap- plied to a numerically specified general equilibrium model. A series of experiments is then undertaken using an illustrative model to de- termine the characteristic features of the optimal investment patterns. The results obtained generalize the conclusions of the static analysis in chapter 5. Statement of the Problem For tradables produeed under inereasing returns, an important question coneerns the scale at which import substitution should be undertaken. TIle minimum efficient scale of import substitution is determined by the plant size at which unit production cost is equal to the c.i.f. (cost, insurance, and freight) import price, converted at the appropriate shadow exchange rate. But this size of plant typically faUs well below that at which increasing returns are exhausted. Cor- respondingly, the optimal size of a plant is generally greater than its minimum efficient size. The same holds true for successive invest- ments. In turn, regardless of whether the output is tradable, the opti- mal size of a plant is by no means necessarily equal-or even close- to the size at which economies of scale disappear (assuming there is such a size). Evidently, then, the optimal size of a plant must be de- termined in light of specific circumstances. The effects of economies of scale on the pattern of investment over time have been analyzed in a partial equilibrium framework by Chenery (1952) and Manne (1967). The growth of demand and the supply cost of imports are taken as given; the optimal size of a plant is determined to be a function of the degree of increasing returns, the rate of growth of demand, the relative cost of imports, and the interest rate. Applications of this model demonstrate that successive invest- ments in an activity characterized by economies of scale should be separated by several years or more beeause of the low volume of the annual growth of demand relative to plant sizes that achieve sufficient scale economies. TIle growth of production capacity thus takes plaee as a succession of large, discrete jumps, typically separated by intervals of several years. In the context of partial equilibrium analysis, two modes of accom- STATEMENT OF THE PROBLEM 219 modation to the discontinuity over time are possible. 1 At one extreme, production may grow continuously in pace with domestic demand even though capacity grows discontinuously. T1le accommodation here is achieved through overinvestment and entails the gradual elimination of excess capacity during intervals between successive investments. At the other extreme, production may grow discontinu- ously in step with capacity. Here the accommodation is achieved by temporarily satisfying the growth of demand with imports and entails periodic import replacement. Only the former mode is relevant in the case of non tradable outputs. In turn, for tradable outputs, assuming that the relative cost of imports is not prohibitive, the optimal policy lies between these extremes-alternation between phases of gradually falling excess capacity and of temporarily rising imports during inter- vals between plant construction. The use of temporary imports to delay investment permits larger plants to be built and thereby affords greater reali7..ation of increasing returns; overinvestment achieves the same benefit but involves a different tradeoff. An extension of the analysis suggests that transitory excess capacity need not be optimal. For tradables, temporary production for export may provide a profitable way to maintain continuous full use of capacity, even if it requires selling at a lower price abroad than on the domestic market. \Vbat matters in this respect is simply whether the marginal f.o.b. (free on board) export price is greater than marginal production cost. In turn, for tradables and nontradables alike, the growth of domestic demand may be adjusted to correspond more nearly to the uneven growth of capacity for domestic supply. Indeed, there is likely to be pressure in this direction. \Vhen there is excess capacity, the shadow price of the output is equal to the recurrent (that is, variable) cost of production, which does not include any charge for capacity use. TIle shadow price must be higher than the recurrent cost of production under full use of capacity when the rental on capacity is determined by the marginal 1. Throughout the discussion in this chapter, it is assumed that the country in question constitutes a single geographic market. This assumption does not in any way bias the principal results of Ollf analysis, but simply permits us to omit the complications introduced by spatial disaggregation. Manne (1967) provides a com- prehensive discussion of the spatial disaggregation of the partial equilibrium model to include trade among market areas within a country. As he shows, plant location becomes an important consideration once internal transportation costs are incor- porated into the analysis. 220 ECONOMIES OF SCALE AND INVESTMENT OVER TIME value of the output in its least valuable domestic use. 2 The result is a form of peak load pricing behavior, which should shift the growth of domestic demand forward toward points at which new plants come on line and thus reduce the extent of transitory excess capacity.3 Other considerations aside, the discontinuous growth of production to match the growth of supply capacity is desirable since it means more rapid realization of the gains from economies of scale. If up- stream intermediate inputs are required for production, however, addi- tional costs must be incurred to increase their supply. In turn, if the output of the activity in question is an intermediate input to produc- tion downstream, additional investment is required to increase do- mestic demand. Temporary inflows of foreign resources could be used to accomplish an identical match between production and capacity without affecting unrelated activities elsewhere in the economy, but limitations on the volume or terms of foreign resource inflows are likely to preclude this mode of accommodation. Thus, further assum- ing that the domestic savings rate can not be adjusted to accommo- date fluctuations over time in the rate of investment, tradeoffs over time among the scales of the related activities and between these and the scales of unrelated activities clsewhere in the economy must then be considered. TIIC lumpiness of individual investments in activities having econo- mies of scale is an important factor in assessing this tradeoff. Invest- ments in industries such as steel and electric power are extremely costly in relative terms. Where foreign resource inflows are limited, simply to construct one plant of warranted size may displace a sub- stantial amount of investment e1sewhere. 4 Simultaneously to under- take all of the related investment (some of which may also be lumpy) required to attain initial full use of capacity is likely either to sacrifice 2. For tradables, the shadow price must always lie between the f.o.b. price of the first unit exported and the c.i.f. price of the last unit imported if all of domes· tic demand were satisfied through imports. 3. Note, however, that recurrent cost may rise abruptly and then fall gradually during intervals between plant construction as a result of start-up problems and leaming.by·doing in new plants as well as of induced changes in the prices of spe- cialized inputs required for production. This pattern would reduce the scope for adjustment of demand growth in response to intertemporal price differentials. 4. Moreover, operation of the plant may equally increase production costs else· where due to limited supplies of domestic resources, such as skilled labor, unless there are offsetting external resource inflows. FORMULATING THE MODEL 221 economies of scale or to restrict investment in unrelated activities more than is desirable. The lumpiness of investment also plays a major role in determining the interval between successive rounds of investment in related sets of activities. l11ese considerations imply another limitation of partial equilibrium analysis, which assumes that the shadow prices of investment resources and other inputs over time are unaffected by the timing of investment in activities characterized by economies of scale. 5 Formulating the Model Formal general equilibrium analysis has only dealt in a very limited way \vith the factors that must be considered when contemplating investment over time under increasing returns. Development theorists considered these issues at the inception of "modern" development economics, however, when a variety of special theories and intuitive suggestions regarding optimal development strategy were put forward. Among the best known are the Rosenstein-Rodan theory of the big push (1943, 1961), the Nurkse-Lewis theories of balanced growth (Nurkse, 1953, 1961; Lewis, 1955), Leibenstein's (1957) concept of critical minimum effort, Hirschman's (1958) theory of unbalanced growth, and the attempts of authors such as Fleming (1955), Lipton (1962), Scitovsky (1959), Streeten (1959), and Sutcliffe (1964) to reconcile some of the conflicting conclusions. Central to the debate concerning these theories have been the effects of (a) limited possibilities for the export of manufactured products and (b) limits on the use of external resources to secure an initial spurt of investment and growth. Discussion has served to clarify the issues, but it has not yet provided an adequate basis for empirical analysis. Our purpose in the remainder of this chapter is therefore to formulate the problem of investment choice in general equilibrium terms to investigate the effects of these constraints in a computable numerical model. The issues cannot be resolved without a more precise description of the economic structure and a statement of policy objectives. This specification requires a comprehensive formulation in which welfare is 5. See Westphal (1971). 222 ECONOMIES OF SCALE AND INVESTMENT OVER TIME maximized subject to constraints on demand and factor use. \Vith allowance for economies of scale, our formulation of the problem will follow the modified input-output approach to development program- ming that is exemplified in the models of Chenery and Kretschmer (1956), Manne (1963), Bruno (1966), and Eckaus and Parikh (1968).6 These programming models dctermine the optimal pattern of investment when the composition of domestic final demand de- pends only on per capita income and the main choices are between domestic production and imports. The problem of vertical and hori- zontal interdependence is thus posed in an empirical context and simple form-that is, with fixed input coefficients and marginal con- sumption propensities. Our general model has the following structural characteristics: (a) linear production activities with economies of scale in the use of capital and constant coefficients for other inputs; (b) two scarce factors-domestic investment resources and foreign exchange; ( c) import and export activities for traded commodities; ( d) domestic consumption of each commodity as a function of income; and (e) a welfare function that depends upon consumption over time, terminal reSOlUces, and the amount of terminal debt. In turn, to capture the determinants of the investment patterns dis- cussed in the literature cited above, we include additional elements in the specific model used here. These are: (f) an extreme form of limited export possibilitics for manu- factured goods, namely no exports of these products; and (g) severely limited access to foreign loans. We also assume that it would not be in the country's comparative ad- vantage to import primary products. The principal patterns previously discussed may be identified as: (a) Rosenstein-Rodan-Nurkse balanced growth, which is characterized by simultaneous investment in many sectors and a large capital inflow (big push) in the early periods; 6. The principal differences among these analyses lie in the degree of disaggre- gation, the treatment of investment and other resource limitations, and the nnm- ber of time periods considered. FORMULATING THE MODEL 223 (b) SCitovsky-Streeten unbalanced growth, in which there is an alternation of investment among sectors with imports filling the gaps between supply and demand 7 ; and ( c) specialization according to existing comparative advan- tage, normally in the export of traditional primary products and import of manufactured goods. Since the arguments for balanced growth are based on the limited possibilities of expanding exports, we have found it useful to define the problem as that of jointly allocating investible resources and foreign exchange over time. The lumpiness of efficient investments makes it desirable if not necessary to borrow and invest irregularly, so that the balance of payments becomcs as important in determining the optimal investment pattern as the balance between savings and investment. There are several tradeoffs to be considered in a framework of sig- nificant scale economies and two scarce factors: (a) the cost of borrO\ving versus the gains from exploiting economies of scale; (b) the cost of deferring investment versus the gains from larger plants; ( c) the loss of present consumption versus the creation of greater future capacity; and (d) the loss of current commodity output versus the gain from exploiting scale economies in overhead (nontradable) services. As noted in the previous section, optimal allocation of investment in- volvcs balancing all of these factors; it rarely produces the simple pat- terns implied by partial equilibrium analyses. Previous studies of Chenery (1959) 8 and Haldi (1960) used static interindustry models to explore some of the effects of economies of scale on investment patterns. TIlese attempts were severely limited by the lack of an efficient method of solving programming problems in- cluding economies of scale, short of enumerating all of the feasible solutions. TIle development of mixed integer programming algorithms 7. We refer to Streeten's (1959, pages 176-77) analysis of the effects of econo- mies of scale, in which he cites an example taken from Scitovsky. Hirschman's (195'8) concept of unbalanced growth is based on psychological reactions that cannot readily be included in this type of analysis. 8. Reproduced as chapter 5 of this volume. 224 ECONOMIES OF SCALE AND INVESTMENT OVER TIME makes it feasible to work with models containing a greater number of activities characterized by economies of scale. To exploit these possi- bilities, we have designed a four-sector model containing scale econo- mies in two industries which is solved for ten time periods. After an extensive process of trial and error, we have developed a computable model that includes most of the significant features of previous studies. Activities in the model The first cycle of investment and production activity and the pa- rameter values for the basic model are represented in activity analysis form in table 6-1 for the convenience of readers wishing to study the details of the model's formulation. The magnitudes of the most im- portant parameters are intended to be realistic although the degree of aggregation makes the results of illustrative value only. Production and trade activity and consumption levels in period 0 (the initial period) are derived from specified initial capital endow- ments under the assumption of no excess capacity. From these activity levels the initial endowments of savings and foreign exchange avail- able for investment are determined. Consequently, the investment variables of period 0 are the first set of activities in the dynamic model. The four production sectors in the model are intended to represent the manufacturing complex and the export sector of a dual economy. The four sectors include industries that produce finished, interme- dia te, and primary goods (sectors 1 through 3 respectively) as well as a sector providing overhead facilities (sector 4). \Vith one exception, the rest of the economy-that is, handicraft industry and traditional, primary production not included in sector 3-is excluded from the modeJ.9 The production of finished and intermediate goods requires inputs from sectors outside the model. The auxiliary investment (AI) activities thus show the cost of the increased capacity (outside the four sectors included) necessary to provide the other intermediate inputs required for production in sectors 1 and 2.1 0 9. As an alternative interpretation, the four sectors in the model might be understood to represent a highly integrated set of sectors within an economy. One stich set is the complex analyzed in chapter 5. 10. The importance of "auxiliary investment" in this model is the role it plays in matching the pattern of investment to the supply of investible resources over time. Overinvestment in a sector to realize economies of scale need not be accom- panied by overinvestment to provide necessary upstream intcrnlediate inputs. FORMULATING THE MODEL 225 The intersection of the production and trade activities with the production and primary resource constraints of period 1 in table 6-1 gives the standard input-output portrayal of economic activity. Due to our hypothetical economy's stage of development, it is assumed that finished and intermediate goods cannot be exported profitably and primary products are not imported. The output of the overhead sector is not tradable. Production in sectors 1 and 2 also requires non- competitive imports of intermediate inputs, as shown in the foreign exchange constraint (FE row) for period 1. Domestic savings are generated by production in each sector, a formulation that avoids the necessity of measuring total income sepa- rately. We assume that the traditional sector of the economy does not supply net savings and that public and private savings are rigidly linked to production, as specified by the entries in the domestic savings constraint (DS row) in the production columns (Xl through X4 ),u The consumption activity (eN), which enters the welfare function, assumes rigid complementarity in the growth of demand for the four goods.!2 The consumption activity is stated in terms of the growth of consumption above its level in the initial period (T - 0),13 We turn now to the investment variables of period O. A one- period gestation lag from investment to usable capacity is assumed for each sector, so that investments in period 0 come on line in period 1 (see the capacity constraints). Economies of scale enter the model in the investment cost functions of sectors 2 and 4 (activities F 2, 12 and F4 , 14). Thesc cost functions are characterized by a fixed charge (Fj ), incurred if investment takes place, and a variable charge (Ij), which depends on the level of investment. The resulting cost function ex- hibits constant marginal cost and declining average cost.14 The fixed charge constraints assure that the fixed charge will be incurred if capacity is built through thc operation of the variable charge activity. Investment in sectors 1 and 3 and auxiliary investment take place 11. This formulation allows for differences in tax and savings rates among sec- tors, but we have not used this possibility in our numerical experiments. 12. Lack of substitution in final consumption is the source of horizontal inter- dependence assumed by Nurkse and other balanced growth theorists. 13. Thus initial consumption levels appear as exogenous demands on the right· hand side of the production constraints. 14. The two average cost functions are shown in figure 6·5. Table 6-1. Tableau of the First Cycle of Activity Investment Constraints Equation number' 1,° F,o 1.° I 30 FO . 1,° Al,· AI.· T=O Primary resourees Foreign exchange (FE) 1 -0.20 -19.0 -0.15 -0.60 -57.6 -1.28 -0.10 -0.05 Domestic savings (DS) 2 -0.53 -42.0 -0.45 -7.00 -108.0 -2,40 -0,27 -0.15 T=l Capacity N N Finished goods 4-1 1.0 0\ Intermediate goods 4-2 1.0 Primary goods 4-3 1.0 Overhead facilities 4-4 1.0 Auxiliary capacity (1 ) 5-1 1.0 Auxiliary capacity (2) 5-2 1.0 Production Finished goods 6 Intermediate goods 7 Primary goods 8 9 I Overhead facilities Primary resources Foreign exchange (FE) 12 I Domestic savings (DS) 13 ~ 1 Fixed charge eonstraints Intermediate goods +1.0 t 1 Overhead facilities 200,0 -1.0 +1.0 I 200.0 -1.0 ! Welfare Activities T:l Transfer of primary inputs Production ConstTaints Equation number' X,' X, · T 0 Primary resources Foreign exchange (FE) 1 1.0 -1.0 -1.0 Domestic savings (DS) 2 1.0 0.8 T==l Capacity Finished goods 4-1 -l.0 Intermediate goods 4-2 -1.0 Primary goods 4-3 -1.0 Overhead facilities 4-4 -1.0 Auxiliary capacity (1) 5-1 -1.0 N Auxiliary capacity (2) 5-2 -1.0 N ......,.. Production Finished goods 6 1.0 Intermediate goods 7 -0.40 1.0 Primary goods 8 -0.12 -0.48 1.0 Overhead facilities 9 -0.10 -0.21 -0.35 1.0 Primary resources Foreign exchange (FE) 12 -1.07 1.0 -0.D4 -0.06 Domestic savings (DS) 13 -1.07 0.2 0.2 0.2 0.2 Fixcd charge constraints Intermediate g Overhead facilities Welfare (tableau continues on the following page) Table 6.1 (continued) Activities T=l Equation Righthand Constraints number' M' 1 M2' E3' CN' side T 0 Primary resources Foreign exchange (FE) 1 ~ -40.0 Domestic savings (DS) 2 :;"" -82.3 T= 1 Capacity Finished goods 4-1 -50.0 N Intermediate goods 4-2 -75.0 N co Primary goods 4-3 ~ -160.0 Overhead facilities 4-4 :;"" -126.7 Auxiliary capacity (1 ) 5-1 :;"" -50.0 Auxiliary capacity (2) 5-2 :;"" -75.0 Production Finished goods 6 1.0 -0.45 90.0 Intermediate goods 7 1.0 -0.32 75.0 Primary goods 8 -l.0 -0.04 ll.5 Overhead facilities 9 -0.19 50.0 Primary resources Foreign exchange (FE) 12 -l.0 -l.0 1.0 :;"" a Domestic savings (DS) 13 ~ b Fixed charge constraints Intermediate goods = lor 0 ~o Overhead facilities = lorO ~O Welfare 1.0 Note: So that a positive entry in the matrix an have been mult ~te t he expressions requirements resp(~ctively: n = 1 pertains to fixed charge a:....:. l: fJJ F/+ E . values of there parameters arc given ill the intCTficction of f2J 1/+:Z , "Ii AI/_R 1 UF}:'"'l+ M,st. l: til! F/ + Jo! ~ . Ut) 1/ + )=1 , J! AI/ -- BI 0.8.1\19, N N \0 230 ECONOMIES OF SCALE AND INVESTMENT OVER TIME at constant average cost. Investment in capacity in any sector requires two factors, foreign exchange (FE) and domestic savings (DS). TIle domestic savings requirement associated with an investment activity gives the total resource cost of investment, while the foreign exchange requirement expresses its import component separately. The produc- tive mechanism whereby savings are translated into capacity through investment is not detailed in the modeP5 The column of activities headed "Transfer of primary inputs" pro- vides for the transfer of foreign exchange and domestic savings both between periods and within the period. Without activities providing for transfer over time of investible resources, investment in a given period would be limited to the investible resources generated in the period, and this in turn would severely limit the extent to which scale economies can be realized through the concentration of investment in one period. In the real world, intertemporal transfer activities in- clude changes in foreign exchange reserves and stocks of commodities, external borrowing, and changes in the proportion of income that is saved. \Ve include two such activities, borrowing and changes in reserves. The debt or borrowing activity (D) secures additional resources from outside the economy. In accord with the conceptual framework of national income accounting, increasing debt yields equal amounts of both foreign exchange and domestic savings in the current period at the cost of repaying both resources in a later period. 16 The UFE activity (representing reserve changes) transfers unused foreign exchange from one period to the next. We neglect stockpiling of commodities, so unused domestic savings cannot be similarly trans- ferred without being combined with imports as completed investment projects. A third transfer activity (MS) converts foreign exchange into do- mestic resources for investment. The additional resources imported may be thought of as investment goods-such as cement-that are normally produced domestically. The latter aspect has led us to 15. Justification for this simplification used is found in the desire not to con- found the analysis by incorporating the effects of investment on the composition of demand for the output of the various sectors. 16. The net foreign capital inflow within period T is equal to DT (l.07) D T -> [D-' 0]. The assumed real interest rate is 7 percent for each period. FORMULATING THE MODEL 231 assume that 1 unit of foreign exchange can replace only 0.8 units of domestic investment resources. l1 The full model consists of ten cycles of activity like that shown in table 6-1; it begins with investment in period 0 and ends with pro- duction, trade, and consumption in period 10. In some experiments the model has also been computed for shorter periods, which only necessitates a change in the valuation of terminal resources. Constraints on the system Here we give a brief verbal description of the principal constraints on the system. 18 In the first period the total use of foreign exchange on investment account and the amount of savings to be invested are limited to the initial endowment of investible resources plus borrow- ing. Investment in each succeeding period is similarly constrained by the production, trade, and transfer activities of that period. The debt limit is formulated in such a way that the debt of the previous period must be repaid along with interest charges, either through renewed debt or from savings and foreign exchange generated by current production. The limited availability of foreign borrowing to a less developed country is one of the central constraints on investment choice, a fact that has been stressed in criticisms of the balanced growth theories. W c specify this limit in the form of a ceiling On the allowable debt in each period. In our experiments, the debt limit will be varied para- metrically to show its effect on the optimal pattem of investment during the plan period. Since labor is a free good in the model, production is limited only by available capacity. In sectors 1 and 2, production is further limited by the auxiliary capacity needed in sectors outside the model to pro- vide intermediate inputs. Since resources devoted to primary produc- tion are not readily transferable to other sectors, we require that the 17. The use of imports to substitute for domestic investment resources through the operation of the MS activity also has the effect of simultaneously yielding an increase in total investment over the sum of the domestic savings generated by production in sectors 1 through 4 plus net foreign borrowing. 18. An appendix containing a formal algebraic statement of the model is avail· able from the World Bank. See the preface to this volume. 232 ECONOMIES OF SCALE AND INVESTMENT OVER TIME production level in sector 3 does not fall. TIle consumption of each product, both on final and intermediate account, is equal to produc- tion plus imports (less exports) in each period. TIle final constraint is a means of incorporating a nonlinear element of diminishing marginal utility in the objective function. Consump- tion growth (that is, CNT - CNT-l) is constrained" to be at least eleven units in each period. \Vithout this constraint, optimization often leads to the conccntration of consumption growth in a short interval. The welfare function The welfare function that is maximized is the discounted value over an infinite horizon of the consumption growth generated by activity during the plan period. 19 Postterminal consumption is included in the welfare function by placing values on terminal capacity and on in- vestible resources in excess of initial endowments. Thus, the incre- ment in consumption in each period over its preplan value (that is, CNT) is given a weight in the objective function equal to 1.0 divided by a discount factor. For the terminal period, the composite of capacity necessary to produce a unit of consumption is given the value of one unit of con- sumption sustained forever after. As a result, terminal excess capacity is not ruled out as being necessarily suboptimal. In turn, since opera- tion of the model terminates with production and consumption in the final period, it is necessary to insure sufficient foreign exchange and savings for postplan growth. 111e economy is required to provide these resources for postplan invesbnent in amounts at least equal to their initial endowments. Value is given in the objective function (in terms of the contribution to postterminal consumption) to the "excess" of foreign exchange and domestic savings over the initial endowments that is passed into the postterminal future. External debt is allowed in the terminal year, but its level cannot be increased above that of the previous year. loe cost of terminal debt in the objective function is the opportunity cost in terms of post- terminal consumption of the foreign exchange and domestic savings that are required for repayment. (There is complete repayment of debt by the terminal period only in the case of model I.) 19. The appendix (see footnote 18) provides complete details of the valuation of tenninal magnitudes. FORMULATING THE MODEL 233 In summary, the function to be maximized is the sum of three ele- ments: (a) the discounted value of increased consumption during the plan period; (b) the value of capacity in the terminal period in excess of that initially available; and (c) the value of terminal foreign ex- change and domestic savings in excess of initial endowments, less the cost of terminal debt. All activities in the welfare function are valued in terms of composite units of the growth of consumption over its preplan level, with discounting to the initial period using a constant discount rate. TIle discount rate used in the experiments reported below is 5 percent for each period. Solving the model Static general equilibrium analysis teaches that a two-step proce- dure is required to evaluate investment projects having economies of scale.20 First, the plant's size must be such that the demand for its output, when priced at marginal cost, is just satisfied. Second, the sur· plus afforded purchasers of the output by virtue of their buying at a price equal to marginal cost must be compared with the loss equal to total production cost at that scale less total receipts from sales at mar- ginal cost. A plant of the scale so determined should be established only if the surplus exceeds the loss; otherwise, no plant should be built. A dynamic analysis requires finding the optimal scale and timing of successive investments, which is, of course, vastly more complicated than a static evaluation. Nonetheless, the same underlying principles apply. First, marginal conditions continue to determine the shadow price of the output period by period. Second, the optimal pattem of investment-that is, the timing of successive investments-is that which maximizes the discounted value of surplus minus loss cumu- lated over aU periods. These principles imply a straightforward method of solution. \Vith respect to the sectors having economies of scale, the number of possi· ble alternative timings of successive investments under our model is 2 raised to the power of 20 (equal to 1,048,576), since there are two scale-economy sectors and ten time periods. 21 Associated with each 20. See lIotelling (1938) and Dort (1958). 21. Note that these alternatives range from investment in both sectors in every period to no investment in either sector in any period. Many of these alternatives, however, may not be feasible. 234 ECO:"lOMIES OF SCALE A:"ID I:"IVESTMENT OVER TIME alternative is a linear programming problem, obtained by setting each fixed-charge variable equal to either 0 or 1 in accordance with the specified timing of successive investments. In principle, the optimal investment pattern could be found by solving the linear programming problem associated with each time pattern to detcrmine that which gives the highest welfare value. Given a particular timing of successive investments in the scale· economy sectors, the optimal scales of investment in these sectors as well as the optimal levels of all other activities-including the optimal timing and scales of investment in activities not subject to increasing returns-are detem1ined by marginal conditions, albeit in the general equilibrium context of linear programming. As a result, the dual varia- abIes (or prices) associated with the solutions arc similar to shadow prices of linear programming in that they represent marginal value products, given the "basis" and the pattern of fixed charges incurred. It is valid to usc prices to test the profitability of othcr activities for a given timing of investment in the scale economy sectors. But, unless the particular timing is known to be optimal, nothing can be inferred thereby regarding profitability in the optimal solution. Correspond- ingly, the prices from the optimal solution cannot be used to judge the profitability of an activity not already in the model, irrespective of whether that activity is characterized by constant or increasing re- turns. ll1e model must be solved again whenever a new activity is introduced. In particular, the question of whether or not to invest at a particular point in a sector with scale economies can only be deter- mined by comparing the welfare values of solutions in which that in- vestment does and does not take place. The straightforward method of solution described above, while useful for pedagogic purposes, is prohibitively expensive to apply in practice. Various algorithms exist that circumvent the need for com- plete enumeration. In solving the model in 1966, we used the Land and Doig (1960) algorithm, which at the time was one of the more sophisticated methods of solving mixed integer programming prob- lems. 22 Nonetheless, it did not prove practical to solve the ten-period model directly.2:{ We therefore adopted the following recursive pro- cedure. 22. This algorithm is discussed in greater detail in the original version of this chapter, Chenery and Westphal (1969). 23. The Land and Doig algorithm has been superseded by more powerful ap· proaches, so that today it would be practical to solve the ten·period model directly. INVESTMENT PATTERNS 235 First, a four-period model having the same specification as the ten- period model (except that the terminal valuations differ) was solved for the optimal pattern of fixed charges. Next, a seven-period model with the integer variables (the Fl's) of the first four periods set at the values of the optimal four-period solution was solved for the opti- mal pattern of fixed charges in the last three periods. This process was then repeated in going on to the ten-period solution. If the adjustment for finite horizon bias embodied in our terminal valuations were correct, then the best ten-period solution obtained in this fashion would be the optimal solution. It is not known whether our adjustment corrects properly for the bias, however; in all proba- bility, it does not. Thus, a number of alternative solutions to the ten- period model were computed, on an ad hoc basis, using knowledge gained from the shorter planning period models to set the integer variables at different values. Since little improvement was achieved, we are confident that our "best" solutions are at least close to globally optimal. In the next section we report the welfare values for all of the solu- tions to the ten-period model that were generated. As implied by the discussion above, each of these solutions results from a process of suboptimization; it is optimal given the pattern of fixed charges in- curred (that is, given the timing of successive investments in the seale-economy sectors). Investment Patterns To derive some broad conclusions from our experiments in opti- mization, it is first necessary to identify some general properties of the solutions. 24 Since a complete description of a ten-period solution requires stating the values of several hundred variables, we have sought fundamental elements that determine the dominant features of the pattern of resource allocation. For most purposes a solution is ade- quately described by the yearly amounts of capacity, production, and imports in each sector. The pattern of investment is dominated by the 24. Solution here means the optimal solution given a particular timing of sue, cessive investments in the scale economy sectors. The best of these solutions for a given set of parameter specifications will be referred to as the "best" solution to the corresponding model. 236 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Table 6-2. Classification of Investment Patterns Number of power plants Number of (sector 4) steel plants (sector 2) 1 2 3 0 AI' A2 A3 1 Bl B2 B3 2 Cl C2 C3 a. Solutions corresponding to this pattern were not obtained. size and timing of eapacity increases in the scale-economy sectors. The simplest characterization of an investment pattern, therefore, is by the number and timing of plants built in sectors 2 and 4 during the ten-year period. 25 To lend greater realism to the following discussion, we shall henceforth identify intermediate goods (sector 2) with steel and overhead facilities (sector 4) with electric power. A typology of the investment patterns observed among the solutions that were generated is given in table 6-2. Tbere may be zero, one, or two steel plants built over the ten periods; these alternatives are desig- nated A, B, and C. Similarly, there may be one, two, or three invest- ments in electric power, which are designated 1, 2, and 3. Thus, for example, a combination of one steel and two power investments is identified as pattern B2; this will be shown to be the best pattern. l11e basic specification of the model assumes a debt limit of 75 in every period, and will be identified as model 1. Alternative specifications raise this limit to 150 (model II) and 300 (model III). A complete listing of the solutions that were enumerated in the search for the best solution to each model is given in table 6-3. These solutions cover eight out of the nine possible patterns shown in table 6-2 and provide the basis for the following analysis of the characteristics of optimal investment and production programs and the effects of changing the availability of foreign resources. The optimal pattern The features typical of an optimal investment and production pro- gram will be shown by considering the best solutions to models I and 25. Our cost function implies that capacity is increased only by building new plants. INVESTMENT PATTERNS 237 II. Their assumptions differ only in the maximum external debt that is allowed in each period, which is 75 units in model I and 150 units in model II. Model II is taken as representative of developing countries in the 1960s, in which the net deficit on current account typically amounted to about a quarter of gross investment. 26 111e increase in debt of 150 units in model II finances about a quarter of investment in the first five years under the best solution. Model III (discussed below) is designed to show the effects of an equal increase in debt in the seeond five years. The solution to each model is depicted graphically in a set of fig- ures, numbered to correspond to the model's designation. \Vithin each set, part A gives the optimal levels of capacity, production, and tradc in each of the sectors; part B, the sources of financing for invest- ment; part C, the uses of investment by sectors 2 and 4; and, part D, the sources and uses of foreign exchange. A separate chart, figure 6-4, compares the growth of consumption over its preplan value in the best solutions to the models. The principal features of the best solutions to models I and II are the same. Because capacity is fully used in period 0, investment is first undertaken in electric power, which is necessary for increased pro- duction in all other sectors. A steel plant is built in period 2 and a second power plant in period 6. These three plants aecount for about 40 percent of total investment in these solutions. There is temporary overinvestment in both sectors, with excess capacity remaining for several periods after plants come on line. 27 To adjust to these lumpy investments, there is an inflow of foreign capital in periods 0, 2, and 6, and investment in other sectors is cur- tailed. In fact, the low debt ceiling assumed in model I causes other investment to be virtually eliminated in these three periods. The in- vestment and borrowing patterns are dominated by the timing of power plant construction, which takes up to 30 percent of total invest- ment resources and is concentrated in two periods. Scale economies make it optimal to build plants large enough to take care of the 26. This is the median value found by Chenery and Strout (1966), which appears as chapter 10 of this volume, for a sample of thirty-one countries. 27. Excess capacity in the terminal period is an aberration caused by the finite horizon and is thus not considered to be a significant feature of the pattern of investment. 238 ECONOMIES OF SCALE AND INVESTMENT OVER TIME growth of demand for the fonowing five years, even when it is neces~ sary to defer almost all other investment to do so. Investment in steel takes place only once; the increase in capacity in both models is equal to about twelve years' growth of total demand. 111e availability of imports makes it efficient to postpone construction in order to build a larger plant with lower unit costs. T11e inclusion of vertical and horizontal interdependence in our model brings out a further feature not previously noted: it is efficient in model I (and to a lesser extent in model II) to reduce steel produc~ tion in periods 5 and 6 to postpone the construction of added power capacity so that a more efficient power plant can be built in period 6. In this way, the import of steel indirectly postpones the requirement to expand power production and makes possible a more economical plant. The expansion of primary production and exports during these periods provides a more economical alternative to increasing power and steel production, even though long~run comparative advantage favors steel when a larger plant can be built. As a further reflection of the importance of realizing greater econo~ mies of scale, consumption increases are held to a minimum until Table 6- 3. Welfare Values for the Solutions Obtained Timing of investment in sector: Investment 2 4 pattern (steel) (power) Model I Model II }'vlodel III A2 0,5 (1) 3,950 0,7 (2) 3,266 A3 0,4,7 (3) 2,572 Bl 2 0 (4) 3,553 ( 5) 4,291 B2 1 0,5 (6) 5,068 1 0,7 (7) 4,130 2 0,4 (8) 3,265 2 0,5 (9) 3,405 ( 10) 5,512 2 0,6 (11 ) 3,436* (12) 4,165* ( B) 5,538* 2 0,7 (14 ) 3,382 (15) 4,157 (16 ) 5,511 2 0,8 (17) 4,080 (18 ) 5,398 3 0,5 (19) 5,008 3 0,7 (20) 4,117 INVESTMENT PATTERNS 239 Table 6-3 (continued) Timing of investment in sector: I Ilvestment 2 4 pattern (steel) (power) Model I Model II Model III B3 1 0,4,7 (21) 3,254 2 0,3,7 (22) 3,148 2 0,4,6 (23) 3,234 2 0,4,7 (24) 3,342' (25) 3,711 2 0,5,6 (26) 5,034 2 0,5,7 (27) 3,295 2 0,7,8 (28) 3,746 3 0,4,7 (29) 3,255 5 0,4,7 (30) 3,221 Cl 2,7 (31 ) 3,601 2,8 ° ° (32 ) 3,644 C2 1,3 0,5 (33) 4,912 2,6 0,7 (34) 5,471 2,7 0,7 (35) 4,068 (36) 5,474 2,7 0,8 (37) 5,372 2,8 0,7 (38) 5,438 C3 1,3 1,4,7 (39) 3,114 2,6 0,4,7 (40) 3,179 2,7 0,4,7 (41) 3,165 Note: Each value pertains to the optimal solution for the timing of investmeuts in sectors 2 and 4. Solntion nnmbers are given in parentheses. The best solution for each model is starred. a. The best solution for pattern B3, model I. after the seeond power plant has been built (see figure 6-4). Once the bottleneck has been removed, there is a rapid increase in the use of power. In period 7, production in all sectors expands, final consump- tion rises more rapidly, and primary production is diverted from ex- ports to the production of intermediate and finished goods. Because of the evident effects of the power shortage in both models, one might suppose that advancing the construction of the second power plant by a year or two would be advantageous. This tums out not to be true, as may be seen from the welfare values given in table (text continues on page 245) 240 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Figure 6-1. Best Solution for Model I Part A. Levels of Capacity, Production, and Trade Sector 1: Finished goods 300~--~--~~~--~--~--~--~--~--~--' c c ~ ~ 200~--+---~--~----+---4----+----~--+----'~~ u .~ ~ o "'C ..... o Capacity, production (K"X,) o 5 6 7 8 9 10 Period Sector 2: Intermediate goods 300 >- u .: ... 200 v ~ Capacity (K2) u - - .~ ~It(/¥/~V~ ~ v E 0 "'C ..... 0 ..... '" '2 ;::J 100 Total supply (X2+M2)r~ J ~~ Ex;e:;" ca~aCltY;(-: ~ ~i,";i ~o~~~ fproduction (X2) ~ ><' o 2 3 4 5 6 7 8 9 10 Period INVESTMENT PATTERNS 241 Sector 3: Primary products 300 --- G !:l .... , (J ~- '...." ... 200 . u p ~ :::1- u ~ ... E -0 "0 0 "" :!::: C 300 ::> Period INVESTMENT PATTERNS 245 6-3. The smaller plant that would have to be built would hamper growth and raise costs in later years. Ine general pattern of investment revealed by these solutions is much closer to the Scitovsky-Streeten concept of alternating invest- ments in different sectors than it is to any of the versions of balanced growth. Primary exports are essential to this pattern. In the short run it is efficient to increase investment in primary production, even though long-run comparative advantage favors import substitution in the two manufacturing sectors. The latter factor is outweighed by the increased flexibility made possible by imports during the transitional period when capital is relatively scarce. The effects of alternating investments in sectors having scale econ- omies on the financing of total investment and imports are shown in the figures B, C, and D for each model. Irregular investment requires the prior payment of debt (and building up of exchange reserves) to finance the bulge in investment and imports of investment goods in period 6 by again increasing debt and reducing reserves. The need to balance investment over time also leads to investment in finished goods (in model I) in advance of the increase in demand, even though there are no economies of scale in this sector. \Vithout this further adaptation to the lumpiness of investment in other sec- tors, the increase in total consumption would have to be deferred even longer. Effects of varying external resources It is apparent from the preceding discussion that external resources are exceedingly important to the realization of economies of scale, because they make possible periodic bulges in total investment consid- erably in excess of domestic savings. The cost assigned to external debt in the terminal year causes complete debt repayment in the best solu- tion to model r; terminal debt is only eighteen units in model II. The marked increase in capacity in all sectors between models I and II is therefore due almost entirely to temporary borrowing, which makes it possible to construct larger and more efficient plants. Moreover, it is not an increase in the net inflow of external resources that is responsible for the larger increase in capacity, but rather greater flexibility in its timing. The undiscounted total net foreign capital inflow over the plan period is greater in the best solution to model I, because of higher (text continues on page 251) 246 ECONOMIES OF SCAI,E AND INVESTMENT OVER TIME Figure 6-2. Best Solution for Model II Part A. Levels of Capacity, Production, and Trade Sector 1: Finished goods 300 , I I i I c s:: ~ :; 200 X, +M, u ~ rJ S Capacity, production /~ IK"Xl 0 'V Total supply (K"X,) _ i ..... 0 (X,+Ml) i .... 100 ~ ~ 'S Impo ~ i i I I o 2 3 4 5 6 7 8 9 10 Period Sector 2: Intermediate goods 300~~--~--~~-r---r---r---r--'---~--' o 2 3 4 6 7 8 9 10 Period INVESTMENT PATTERNS 247 Sector 3: Primary products 300 » u c:: ... u .Sl ., /'" '" v ~ S -- --- ", , 0 '"0 ti ...... 100 "' ...... -= --- '-, ..... .,..- Exports (Ea) r--" "," ",,,,"- - ·S ;:::J ." --"'; o 2 3 4 5 6 7 8 9 10 Period Sector 4: Overhead facilities 300 >, ~P ~/- u c:: ~ ) e 200 .Sl Capacity (K.) I' ~ f-o"""" ~ ~~~' v S 0 '"0 Production (X.) ti .... '" 100 ·S ;:::J o 2 3 4 5 6 7 8 9 10 Period 248 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Part B. Sources of Financing for Investment 400~--~--~--~--~--~--~--~--~~ i 300 1\ ~ot" "",l" 200 ~ ! ! (f Domestic saving / /~ \ f..-- -- >. u c:: ~ 100 I, \ \ \-rr- \ ./ V\-Lc -i~\ ../ ; I \ \ :::l \ u I \ '\ ,+,BMS)\ 0 \ E~= imp~t. (. +l I \ '" I , ". \'''' \ v E 0 I" ' , V \ "\ \, '1j ..... 0 ... ~ ""', ,If ' \ - .. --- ", ; "',,\, 2 "I "a ::; 0 \ \ / ; - '- ' -, , ' I I I ~-- ----- " "'"'\ I I I \ \\ I~, I \ Net foreign capital inflow \ [D'" - (1.07)DT-lj \ ! I \ \ -100 , ~, '\ \ \ ~9 -200~__~~~~~4~~~1'__~__~__8__~ o 23 __ 67 Period INVESTMENT PATTERNS 249 Part C. Uses of Investment Period 250 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Part D. Sources and Uses of Foreign Exchange 200~--r---T---~--~--~--~--~--~--~ t' ::: . u C ;:: .... :::l 200 u .S1 ..., ~ S 0 "0 ..... 0 V> ."!:: c ;;:J o 2 3 4 5' 6 7 8 9 10 Period INVESTMENT PATTERNS 253 Sector .3: Primary products 300 /" ;;., /" '" \::: \l) Capacity, production (K.,Xa) /" \:: ::l 200 ~ u ,./ ~l1 ~ E 0 "0 '0 100 - ---r--. "- ,- --- \ ,,' " " Exports (En) Xl '2 ;:; " \ \, ,/' o I 2 3 4 5 6 7 8 9 10 Period Sector 4: Overhead facilities 300 -...- W ~,...::: >- u \::: g 200 ) = u Capacity (K.) If .~ ~ Wk~' ~ \l) E 0 Production (X.) "0 '0 100 ..... '" '2 ;:::J o 2 3 4 . 6 7 8 9 10 Period 254 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Part B. Sources of Financing for Investment 500,.---,...--,...--r-__r---,---,--..,.--..,.---, 1\ Tot;l sources 400i~-+_~--+--+--1----tI-'\r-/-f+--+---I 300~--~--~~~~~_4--~-~\+_--+_--; ,J Domestic saving 1\ ; !\ \ _ - - \ / ...--~ '\\ -!0 : ;r-" .-f--I\ _f-f-lt1\ r-' rr 100h--t~1=-:j;,. ""==-+----++---++-H-\-..l~-----if--__I \ I Excess imports \ (. 8MS) I,' ; , '. \ " \ , \ ~ _____ V \\ " \ /.-- , 1 \\ '". ...... .. i '~ "",, ........... :i.......... . . . . . . /1 \/ OI~~\~/~~~'+---~l~~~'-4--~i~-~-~-~-~ ! \! .--- --- Net foreign capital inflow \ , [DT _ (1.07)DMj \ I -IOOI~--~~~~r_~r-~\~~~-~--_4--_; -200~~~~___+--_+__~--~--~--t-~ o 23456789 Period INVESTMENT PATTERNS 255 Part C. Uses of Investment 500r---r---r---r---~--~--~--~--~--~ ~ ., ~ 300 t:: ::l v .~ v a 0 .... "V 0 ~ '" "Z ~ Period 256 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Part D. Sources and Uses of Foreign Exchange 300r---r---r---r---~--~--~--~--~--~ ..... o tl 400r--r--r--"---"'-'-""--~--"T""---r--~ '2 !:l o 2 3 4 5 6 7 8 9 Period INVESTMENT PATTERNS 257 Figure 6-4. Growth of Consumption over Preplan Value in the Best Solutions 350 Model III / 300 V / V 250 >. u Model V t:: ~ ::: 200 V (J ,,/ .~ i[l S 0 1/ v/ 'U ..... 0 MOdel}! 150 !l oS ;::J V 100 / V 50 / :/ V /'" /'" V o 2 3 4 6 7 8 9 10 Period 258 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Figure 6-5. Effect of Plant Size on Unit Cost of Investment (Points indicate plants constructed in best solution to each model) 8 7 1\1(1) 6 5 \ Il(1) 4 I 1lI(1) 1(2) !'-,.. Il(2) . . . . .r--- -.!!!.(2) 3 \\ Sector 4 (/>Ower) 2 \ "-i'- I--.. I Sector 2 (steel) II m- o 20 40 60 80 100 120 140 160 180 200 Plant size Note: Roman numbers indicate the particular model; arabic numbers indicate the se- quence of plants if more than one is constructed in the best solution. INVESTMENT PATTERNS 259 lble 6-4. Effects of External Resources Ane Model Model Change Model Change Imber Item I II from I III from I Increase in terminal con- sumption over initial value 28 40 +12 54 +26 2 Total welfare (W) 3,436 4,165 +729 5,538 +2,102 3 Total invest- ment over ten periods 1,237 1,398 +161 1,758 +521 4 Total net capital inflow over ten periods' -27 -66 -39 182 +209 5 Total savings over ten periods (line 3 - line 4) 1,264 1,464 +200 1,540 +276 6 Average productivity of investment (line 2 -;- line 3) 2.78 2.98 3.15 7 Marginal productivity of investment 4.53 4.03 Sources of im- provement over model I 8 Reduction in in- vestment cost" 102 185 9 Increased savings 200 294 10 Increased capital inflow -39 209 11 Total (lines 8 + 9 + 10) 263 688 12 Productivity (~W -;- line II ) 2.77 3.05 \late: Numerical values were derived from the best solution to eaeh modeL t. The total uudiscounted value over ten periods, net of repayment and interest charges: o[DT -(1.07) Dr-,]. l. The difference between actual investment cost and the cost of building the same size nts at the average unit costs of model 1. 260 ECONOMIES OF SCALE AND INVESTMENT OVER TIME The increased capital inflow in model III produces a gain in welfare of 60 percent above model 1. Cost savings from larger plants and in- creased savings from more rapid growth again account for the major portion of the total. Comparing models II and III, it can be seen that a net increase in capital inflow amounting to roughly 10 percent of total investment is highly productive because of these indirect effects. Further increases would be less productive because the size of invest- ments in model III exploits most of the available economies of scale. Deviations from the optimal pattern Since perfect coordination of investment cannot be achieved in either a planned or a market economy, it is important to determine the effects on total welfare of departures from optimality in the size and timing of investments. As these effects have been investigated by Manne and others (1967) in a partial equilibrium context, we will coneentrate on the adjustments that are necessary in the pattern as a whole. A detailed examination of one alternative solution iHustrates the adjustments that can be made and their net cost. l11e best pattern for all three specifications of our model requires sufficient foresight to accept periodic power shortages to build larger plants. A plausible "real world" alternative would be to keep up with demand at the cost of building smaller and more expensive plants. Such an investment policy is illustrated in figure 6-6, where the invest- ment cycle in electricity is cut from six years to three to four years and three plants are built instead of two. Optimization of model I under these restrictions gives the investment and production patterns shown in the figure. Keeping up with the demand for power leads to higher steel and power production over the period and greater power capacity at the end. TIle cost of this policy is reflected in smaller power plants that cost 15 percent more for each unit of capacity and hence reduce the investment possible in finished and primary goods production. The welfare cost to the economy of this single change in investment policy is 3 percent of the total value of model I, as shown in table 6_3. 29 This 29. The best solution to model! for pattern B3 is number 24, which has a value of 3,342 and for which the levels of capacity, production, trade, and consumption are shown in figure 6-6. Solution number 11. which has a value of 3,H6, is the overall best for model 1. INVESTMENT PATTERNS 261 pattern requires less advance planning of investment, however, and is less dependent on steel imports. Table 6-3 gives a number of other examples of the cost of changes in the investment pattern or of departures from the best timing. Tim- ing of investment is more critical in model I because limited external resources reduce the possibilities for adjustments in other sectors. The investment patterns that are most closely competitive to the best (B2) are B3 for model I and C2 for models II and III. Continued specializa- tion in primary exports instead of import substitution in steel (pat- tern A) is quite costly in all models under the given assumptions, although the construction of the steel plant can be postponed two or three years without great loss (as shown by solution number 30). Apart from the ten-period solutions given in table 6-3, we have calculated the effects of a number of different specifications of the parameters in the model for shorter periods. Among the changes tested were higher discount rates, greater returns from exporting, and possible exports of manufactured goods. Since the results were essen- tially those that would be expected, they will not be reproduced. Implications tor development planning These experiments bring out an aspect of development plans that is only beginning to be adequately appreciated: their function in bal- ancing investment over time. Scitovsky reached a very similar conclu- sion in trying to reconcile the arguments for balanced and unbalanced growth: ... [S]everal-year investment plans, of which one hears so much nowadays, may be regarded as plans for unbalanced growth, extending to several years so as to restore balance by the end of the period for which the plans are made. In the interim, im- balance manifests itself by the completion of productive ca- pacity before the demand for its full utilization has arisen, or by the creation of consumers' or producers' demand before the capacity to fill this demand is completed. The temporary excess capacity may have to be accepted in most cases as an inevitable cost of (temporarily) unbalanced growth for the sake of secur- ing economies of scale; the temporary excess demand may be filled by imports, which is one reason why the availability of foreign loans or foreign exchange is so strategic a factor in invest- ment planning. Such dependence on foreign trade, however, is 262 ECONOMIES OF SCALE AND INVESTMENT OVER TIME Figure 6-6. Model I, Pattern B3 (Solution Number 24): Levels of Capacity, Production, and Trade Sector 1· Finished goods 300 e- .., I:l " 200 " :::l u X,+M, Kl'X~ .!:l .., -:;; e - 0 '"0 0 '" 100 Total supply (X,+Ml) Capacity, production ( "X,) .-:- , / ~ "'E :.:> o 1 2 3 5 6 7 8 9 10 Period Sector 2: Intermediate goods 300r--~-....,....-""-"";;"'-r--r-~-....,....-""---r----. e- ::: I!:: " 200 :::l u U .';: .., '" e - 0 '"0 0 '" =a :.:> o 2 3 5 6 7 8 9 10 Period INVESTMENT PATTERNS 263 Sector :3: Primary products 300 G I:: ~ .... _ /V ::l 200 u .\:1 Capacity, production (K.,X.) V" ~ OJ S 0 "0 ..... -- -- ..... 100 Exports (E.) ....... ·S 0 '" ." ",,,'- ..... -.... f- ....... ; - - "" .... ", f--- ;:J o 2 3 4 5 6 7 8 9 10 Period Sector 4: Overhead facilities 300 G c:: 'V" ~~ OJ .... .... ::l u p" ~ ,OJ Capacity (K.) ~ l/.~ E 0 "0 ..... W4:- J ~ J>roduction (X,) 0 .~'" 100 I:: ;:J o 2 3 4 , - 6 7 8 9 10 Period 264 ECONOMIES OF SCALE AND INVESTMENT OVER TlME very different from that which accompanies unbalanced growth concentrated on industries with a comparative advantage. For one thing, this is a temporary dependence, while that [i.e., the other] is permanent; for another, the dependence here is primarily on foreign import supplies, there on foreign export markets. 30 Our analysis should not, however, be used uncritically to support Scitovsky's characterization in terms of the dependence on import supplies as opposed to export markets. We have confined our detailed investigation to a specific set of structural characteristics. In particu- lar, along \vith a continuing need for imports of investment goods, we have assumed extreme pessimism regarding the prospects for steel exports. 111is affects the optimal choice of activity levels with respect to both the short and the long run, as may be seen by analyzing the cause of the increase in excess capacity in stecl production that is ex- perienced in thc best solution to each model before the second power plant comes on line. At first impression, the rise in excess capacity is very puzzling indeed, for it takes place simultaneously with an increase in steel imports. But this apparently anomalous behavior is dictated by the necessity to repay past foreign borrowing and build up foreign exchange reserves in anticipation of the construction of the second power plant. Once it is fully used, capacity in the first power plant places a severe con- straint on the generation of the foreign exchange that is needed. The direct plus indirect requirement for electric power for each unit of foreign exchange saved through steel production exceeds the total requirement for each unit of foreign exchange earned by primary ex- ports. 31 In the short run, therefore, the generation of marginal foreign exchange is maximized by exporting primary products to import steel. This procedure releases electric power that may be used to increase primary exports, which in turn generates additional foreign exchange. Suppose now that the input-output coefficients in the model were changed to reverse the relative use of electric power for each unit of produetion in sectors 2 and 3. The economy would no longer experi- ence a rise in excess capacity for steel production occurring simul- 30. Scitovsky (1959, pages 213-14). 31. Electric power requirements are respectively 0.40 [equal to 0.21 + (0.48 + 0.06) X 0.35] and 0.35 for each unit of production in sectors 2 and 3. INVESTMENT PATTERNS 265 taneously with increased steel imports. Further supposing that exports of steel were possible, this activity would replace the growth of pri- mary exports during the periods before the second power plant comes on line.32 But it does not necessarily follow that there would be a re- duction in economywide excess capacity, for the result could simply be to transfer excess capacity from steel to primary production. Two factors are responsible for this possibility. rThe first is the lumpiness of investment in the scale-economy sectors, which requires that investments in other sectors take place in the intervening periods between the construction of successive steel and power plants. The second is the continuing need to use foreign exchange in investment, which requires exports even when the economy's long-run comparative advantage favors import substitution in sectors 1 and 2 and continued domestic production of primary goods. Together these factors may dictate short-run excess capacity in a sector exhibiting constant re- turns, as was found in sector 1 immediately before the construction of the second power plant in the best solution to model I. A short-run comparative advantage in exporting a product produced under increasing returns therefore can be consistent with a long-run comparative advantage in exporting some other product (perhaps produced under constant returns), and vice versa. Moreover, tem- porary dependence on trade to adjust to lumpy investment may in- volve exports as well as imports of products produced under economies of scale. The optimal pattern of investment over time thus depends on a number of interrelated factors, with conditions in world markets and the precise degree of lumpiness playing important roles along with the underlying production relations. It should also be recognized that "lumpiness" is a relative concept, which concerns the magnitude of a particular investment project in relation to its use of available investment resources as wen as in rela- tion to the markets for its inputs and output ( s ). The larger a project is in relative terms, the more far-reaching and greater the effects of its implementation will be. Correspondingly, the interdependencies asso- ciated with a given project will be quite different depending upon the economy in which it is implemented. For example, there is a big 32. Export and import prices for steel need not be equal. Thus, we are here additionally assuming that the total use of electric power for each unit of foreign exchange revenue from exports of steel is less than that from exports of primary products. 266 ECONOMIES OF SCALE AND INVESTMENT OVER TIME difference between building the first large-scale integrated steel mill in Korea, the fourth or fifth in India, and the fiftieth in the United States. '\\'hat must be planned at the economywide or regional level in some countries may be equally well planned at the sector or enter- prise level in other countries. Similar investments thus imply the need for coordination at different levels in economies of different sizes and at different stages of development. Conclusions This chapter has attempted to give greater precision to the discus- sion of alternative investment patterns by restating several of the principal hypotheses in programming terms. We have shown that under realistic assumptions about the nature of horizontal and vertical interdependence, the timing of investment in scale-economy sectors has a substantial effect on investment timing in other sectors and hence on the whole investment pattern. The optimal pattern balances the gains from larger plants against the costs of deferring investment in other sectors and the resulting loss of growth in income and savings. The general characteristics of this pattern include both the type of alternation among sectors envisaged by Scitovsky and Streeten and the exploitation of an integrated spurt in investment foreseen by Rosenstein-Roda n. The use of a comprehensive optimizing model has also provided some new perspectives on the characteristics of optimal patterns of al- location. In a rapidly growing economy, the timing of investment is often more important than the choice of sectors along static, compara- tive advantage lines. To exploit economies of scale in some sectors, it is necessary to limit investment in others, sometimes by using small- scale techniques that will prove to be inefficient at a later date. In our model, primary production performs this function; it is expanded whenever necessary to secure imports of manufactured goods, but these imports are periodically replaced with lower cost domestic production. The concept of optimal overcapacity was derived by Chenery (1952) from a partial equilibrium analysis; explicit consideration of the effects of interdependence leads to an analogous concept of opti- mal shortage. The reduction in unit cost from building a bigger power plant next year must be weighed against the cost of lost production CONCLUSIONS 267 this year in sectors that use power. 'These opportunity costs can only be determined in a more comprehensive analysis. Our optimal power cycle consists of three to four years of excess capacity and one to two years of power shortages, the latter being reflected in overcapacity in the sectors that use power. Even though the choice between domestic production and imports is central to the present model, comparative advantage is hard to de- fine or measure in a system containing economies of scale and limited investment resources. In the scale-economy sectors, the critical ques- tions are those of plant size and the timing at which import substi- tution becomes profitable. There is no possibility of a timeless ranking of projects along partial equilibrium lines because the need to accom- modate lumpy investments causes relative profitability to vary year by year. Although this phenomenon is exaggerated by considering only part of the whole economy, as we have done, it is of considerable em- pirical importance in less developed countries. Finally, our results emphasize the importance of jointly planning the allocation of investment and foreign exchange over time to allow for alternating production and trade. The theoretical and practical advantages of this broader approach to development planning appear to be substantial as compared to more traditional methods of al1o- cating each scarce factor separately.33 33. Methods of applying the general approach outlined here are diseussed by Westphal (1975), who demonstrates some of the advantages of mixed-integer programming as a planning technique. Part Three External Structure THE ROLE OF EXTERNAL TRADE AND CAPITAL FLOWS in the process of development is the general problem addressed in part three. The term external structure is used to characterize those aspects of resource allocation that stem from the supply and demand for foreign ex- change. In relation to part two, emphasis is shifted from the internal constraints on resource allocation to the external constraints. \Vben it is recognized that increasing exports and replacing imports usually require investments in physical and human capital or the development of agriculture and natural resources, it is evident that for a planning period of five to ten years the supply of foreign ex- change becomes in effect a separate factor of production. As in the case of skilled labor, this supply can be increased over time by the application of labor and capital, but in any given period there are definite limits to total imports. The corresponding concept of "trade- limited growth" and its relation to other limits (such as investment and skills) provide the starting point for the analysis of changes in the external structure. Recognition of the possibility of an external limit to growth is a critical element in analyzing the role of external capital flows. \\Then imports are identified as a separate factor of production, external capital is seen to perform a dual role by adding to both investable resources and foreign exchange. The elaboration of this concept and application of the resulting models to the study of development alternatives in different kinds of countries are the central problems treated in part three. A general framework for these studies is provided by the assessment of Comparative Advantage and Development Policy in chapter 7. This chapter discusses the conflicts that may result from taking either trade theory or development theory as a point of departure for policy 269 270 INTRODUCTION TO PART THREE analysis. The use of a linear programming framework is suggested as a way of resolving some of the conceptual conflicts and of linking project evaluation to economywide analysis of resource allocation. A survey of the planning procedures used by developing countries high- lights the relation between a country's size and its tendency to use an inward-looking or outward-looking approach to policy. Chapter 8, Development Alternatives in an Open Economy, develops a planning model for Israel in whieh imports are treated as a third factor of production along with capital and labor. In con- sidering a country whose productive structure has been adjusted to a continuing inflow of external resources, it is evident that a reduction in this flow cannot be offset merely by limiting consumption and raising domestic saving. Also required is increased production of traded goods through import substitution or export expansion. Although a disaggregatcd interindustry analysis is used to determine the necessary changcs in thc external structure, the results are sum- marized in two reduced-form equations that indicate the relations among the levels of exports, imports, saving, and investment and the increase in GNP. Since the inflow of capital is an element in each equation, this formulation has come to be known as the two-gap model of the relation between capital inflows and development. Analysis based on the structural parameters of the Israeli economy indicates the conditions under which the trade gap or the savings gap is likely to be more restrictive and provides a measure of the marginal productivity of external borrowing in each situation. Chapter 9, Optimal Patterns of Growth and Aid, gives a dynamic analysis of the relation between changes in flows of external capital and the corresponding changes required in the structure of domestic production and external trade. Pakistan was chosen as a basis for the analysis because it is broadly representative of the conditions under which poor countries are attempting to accelerate their growth. 1be main questions posed are: (a) \\/hat is the optimal inflow of capital over a period of twenty years under various assumptions about the cost of borrowing? (b) What should be the allocation of resources between traded and nontraded goods over time to make the most effective use of external resources? (c) \Vhat are the aggregate benefits to the economy and the marginal productivity of external borrowing? The analysis of chapter 9 compares optimal development strategies under alternative assumptions about national objectives and policy INTRODUCTION TO PART THREE 271 constraints, using a linear programming model to determine the best allocation of internal and external resources. The chapter also shows the effects of several forms of rationing by aid donors. Subsequent studies of Colombia (Vanek, 1964), Greece (Adelman and Chenery~ 1966), Israel (Bruno, 1967), India (Tendulkar, 1971), and Turkey (Dervis and Robinson, 1978) have investigated the rela- tions between internal and external constraints on development policy in a variety of country settings. All of these studies demonstrate the importance of an integrated analysis of internal and external resource allocation. Country studies such as these provide the basis for the more general exploration of the role of external capital in develop- ment that is undertaken in chapter 10. The contrast between the policy implications of trade-limited models of the type developed here and more orthodox, neoclassical models has led to considerable controversy.1 As was pointed out in chapter 2, institutional or technological constraints on economic behavior may prevent the achievement of the optimum allocation determined by neoclassical theory. The essays in part three can therefore be considered as an exploration of a type of "second bcst" situation that has been widely identified in empirical studies of developing countries. 1. Since much of this debate concerns the simplified two-gap models that are used in the llulticollntry projections of chapter 10, it is discussed in the postscript to that chapter. Chapter 7 Comparative Advantage and Development Policy IN THE GREAT REVIVAL OF INTEREST in economic development that has marked the period following World \Var II attention has centered on two main questions: first, what determines the overall rate of economic advance?; second, what is the optimal allocation of given resources to promote growth? Analysis of the growth rate relied mainly on Keynes- ian tools and produced a multiplicity of aggregate growth models. The second question, however, reopens more ancient economic issues, and their analysis must start from the classical and neoclassical solu- tions. Only recently have the two types of discussion tended to come together in the more comprehensive framework of general equilibrium analysis. In the field of resource allocation, controversy centers around the implications of the classical principle of comparative advantage, ac- cording to which growth is promoted by specialization. The defenders of this principle draw their inspiration from David Ricardo, J. S. Mill, and Alfred Marshall, while the lines of attack stem from Friedrich List, J. A. Schum peter, A. A. Young, and J. H. Williams. T11e chief criticism is that comparative advantage is esscntially a static conccpt that ignores a variety of dynamic elements. This issue is of great practical importance to the governments of developing countries, most of which take an active part in allocating investment funds and other scarce resources. The main purpose of the discussion has therefore been to discover workable principles for the This chapter was prepared in 1960 as a survey article for the American Eco- nomic Review. I am indebted to Moses Abramovitz, Bela Balassa, Lawrence Krause, and Bernard Haley for helpful comments. 272 CONFLICTS BETWEEN TRADE THEORY AND GROWTH THEORY 273 fonnulation of development policy. The classical approach derives these principles from international trade theory, while its critics base their analysis on theories of growth and development. Elements of a dynamic, general equilibrium theory are needed to resolve the differ- ences between the two approaches. The more general analysis is of very limited value, however, unless its empirical implications can be ascertained. This chapter discusses the analysis of resource allocation in less developed economies from three points of view. The first section (pages 272-81) tries to ascertain the extent to which the allocation principles derived from trade theory and from growth theory can be reconciled with each other without losing their operational signifi- cance. The second section (pages 281-99) comparcs various ap- proaches to the measurement of optimal resource allocation in terms of their logical consistency and their applicability to different condi- tions. The third section (pages 299-306) examines some of the prac- tical procedures followed in setting investment policy in developing countries in the light of the earlier discussion. Finally, some of the important theoretical issues are reexamined to indieate their practical significance. Conflicts between Trade Theory and Growth Theory TIle principal contradictions between comparative advantage and other principles of resource allocation derive from their different orientation and assumptions. The classical analysis focuses on long-run tendencies and equilibrium conditions, while modern theories of growth are concerned with the interaction among producing and con- suming units in a dynamic systcm. Since both approaches are familiar, I shall only try to identify the differences in assumptions and emphasis that lead to different policy conclusions. Implications ot comparative advantage tor resource allocation The modern version of the comparative cost doctrine is essentially a simplified form of static general equilibrium theory.! The optimum 1. See Haberler (1950). An excellent discussion and synthesis of the several versions of trade theory is given by Caves (1960). The terms comparative ad- vantage and comparative cost are used interchangeably in most discussions. 274 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY pattern of production and trade for a country is determined from a comparison of the opportunity cost of producing a given commodity 'Nith the price at which the commodity can be imported or exported. In equilibrium, no commodity is produced that eould be imported at lower cost, and exports are expanded until marginal revenue equals marginal eost. Under the assumptions of full employment and perfect competition, the opportunity cost of a commodity, which is the value of the factors used to producc it in their best alternative employment, is equal to its market value. Market prices of factors and commodities can therefore be used to determine comparative advantage under eom- petitive eonditions. Long-term ehanges are not ignored, but they are assumed to be refleeted in eurrent market prices. 1be Heekscher-Ohlin version of the comparative cost doctrine has been widely recommcnded as a basis for development policy because it provides a measure of comparative advantage that does not depend on the existence of perfect competition and initial equilibrium. This version states that a country ,vill benefit from trade by producing commodities that use more of its relatively abundant factors of pro- duction. TIle country will export these commodities and import com- modities using more of its relatively scarce factors unless its pattern of domestic demand happens to be biased toward commodities using domestic factors. The critical assumptions in this analysis are that factors of production are comparable among countries and that pro- duction functions are the same. These assumptions are not required by classical trade theory. The applicability of the comparative cost doctrine to conditions in developing countries has been reexamined by Viner and its validity has been reaffirmed with some modifications. Viner criticizes the Heckscher-Ohlin version because its assumption of comparable factors does not allow for observable differences in their quality.2 In his answer to critics of the comparative cost approach, however, Viner admits the necessity of interpreting comparative advantage in a dynamic set- ting in which the efficiency of production may change over time, ex- ternal economies may exist, and the market prices of commodities and factors may differ from their opportunity cost.:1 As Nurkse points ont, these modifications rob the original doctrine of much of its practical 2. Viner (1952, page 16). 3. Viner (1958). CONFLICTS BETWEEN TRADE THEORY AND GROWTH THEORY 275 value. 4 It is now necessary to have an explicit analysis of the growth process itself before it is possible to determine, even theoretically, where comparative advantage lies; market prices and current oppor- tunity costs are no longer sufficient. Implications at growth and development theory tor resource allocation Theories of growth and development are concerned with the inter- actions over time among producers, consumers, and investors in re- lated sectors of the economy. In the writings of such economists as Rosenstein-Rodan (1943), Lewis (1954), Nurkse (1953), MyrdaI (1957), Rostow (1956), Dobb (1960), and Hirschman (1958b), there is much more emphasis on the scquence of expansion of production and factor use by sector than On the eonditions of general equilibrium. Growth theory either ignores comparative advantage and the possi- bilities of trade completely, or considers mainly the dynamic aspects, such as the stimulus that an increase in exports provides to the de- velopment of related sectors, or the function of imports as a carrier of new products and advanced technology. \Vith this different point of view, gro\vth theorists often suggest investment criteria that are quite contradictory to those derived from considerations of compara- tive advantage. The conflicts between these two approaches to resource allocation may be traced either to differences in assumptions or to the inclusion of factors in one theory that are omitted from the other. Growth theory contains at least four basic assumptions about developing economies that differ strongly from those underlying the comparative cost doctrine: (a) factor prices do not necessarily reflect opportunity costs with any accuracy; (b) the quantity and quality of factors of production may change substantially over time, in part as a result of the production process itself; (c) economies of scale relative to the size of existing markets are important in a number of sectors of pro- duction; and (d) complementarity among commodities is dominant in both producer and conSumer demand. Some of the implications of these factors were developed by Rosen- stein-Rodan (1943) and Nurkse (1953) as arguments for balanced 4. Nurkse (1959, page 76). 276 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY growth, by which is meant simultaneous expansion of a number of sectors of production. 5 Assuming an elastic supply of either capital or labor, these authors show that investment will be more profitable in related sectors, because of horizontal and vertical interdependence, than in the same sectors considered separately. Market forces will not necessarily lead to optimal investment decisions because present prices do not reflect the cost and demand conditions that will exist in the future. TIlis effect of investment in one sector on the profitability of investment in another sector, via increased demand or reduced costs, has been called by Scitovsky (1954) a "dynamic external economy." The imputation of these economies to the originating sectors may seriously affect the estimate of comparative advantage. If we assume fixed investment resources instead of an clastic sup- ply, the same set of factors provide an argument for concentrated or unbalanced growth. 6 To achieve economies of scale in one sector, it may be necessary to devote a large fraction of the available invest- ment funds to that sector and to supply increased requirements in other sectors from imports (or to curtail them temporarily). The optimal pattern of investment will then be One which concentrates first on one sector and then on another, with balance being ap- proached only in the long run. Streeten (1959) developed further dynamic arguments for unbalanced gro\Vth from the fact that tech- nological progress may be more rapid if increases in production are concentrated in a few sectors, while Hirschman (1958b) argues for imbalance to economize on entrepreneurial ability. Tne historical significance of the balanced growth argument has been examined by Gerschenkron (1952), Rostow (1956), and Ohlin (1959), in the context of nineteenth-century industrial development in Europe. They show that vertical interdependence has been impor- tant in stimulating the growth of related industrial sectors, although the nature and origin of these complexes differ from country to coun- try. In one ease they may be related to exports, in another to expan- sion for the domestic market. The importance of interdependence among producers emerges fairly clearly from these historical studies. The net effect of the discussion of dynamic interdependence and balanced versus unbalanced growth is to destroy the presumption that 5. The tenn balanced growth has been given a variety of meanings, but the idea of simultaneous expansion OIl several fronts is eommon to all of them. 6. See Scitovsky (1959) and Sheahan (1958). CONFLICTS BETWEEN TRADE THEORY AND GROWTH THEORY 277 perfect competition, even if it could be achieved, would lead to the optimum allocation of resources over time. Since the doctrine of com- parative advantage in its conventional form is a corollary of gcneral equilibrium theory, the theoretical qualifications that apply to the latter also apply to the former. If, then, the doctrine of comparative advantage is to be useful for development policy, the essential ele- ments of the growth analysis must bc combined with it. Dynamic modifications of comparative advantage Traditional trade theory does not exclude changes in the supply of factors and other data over time, but it does insist that under perfect competition the effects of such changes will be reflected in thc market mechanism. If, on the other hand, we take comparative advantage as a principle of planning rather than as a result of market forces, we can include any foreseeable exogenous changes in technology, tastes, or other data without going beyond the framework of comparative statics. Some of the modifications suggested by growth theory are dynamic in a more essential way, in that a particular change depends not only on the passage of time but on other variables in the system. For exam- ple, the rate of increase in the productivity of labor in an industry may depend on an increasing level of production in that industry. Some of these dynamic elements can also be analyzed by methods of compara- tive statics if our purpose is only to choose among alternative courses of action. The four assumptions of growth theory discussed in the preceding subsection lead to the following requirements for the analytical frame- work to be used in determining comparative advantage in a growing economy7: (a) recognition of the possibility of structural disequi- librium in factor markets; (b) the inclusion of indirect (market and nonmarket) effects of expanding a given type of production; (c) si- multaneous determination of levels of consumption, imports, and production in related sectors over time when decreasing costs result 7. Some of these criticisms of static analysis were made years ago by \Villiams (1929), and a number of the elements were, of course, recognized by the classical economists themselves. I am not concerned with explicit criticism of the elassical analysis, but with the possibility of reconciling it with growth theory. 278 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY from the expansion of output; and (d) allowance for variation in the demand for exports and other data over time. These changes destroy the simplicity of the neoclassical system, in which allocation decisions can be based on a partial analysis because adjustments in the rest of the economy are reflected in equilibrium market prices. In the dynamic analysis, it may not be possible to state that a country has a comparative advantage in producing steel with- out specifying also the levels of production of iron ore, coal, and metal- working over time. In short, we are forced to compare alternative pat- terns of growth rather than separate sectors, and we cannot expect to find simple generalizations of the Heckscher-Ohlin type concerning the characteristics of individual lines of production. Since there is no well-developed body of theory concerning the for- mal properties of the system just outlined,8 I shall only try to indicate in a general way the modifications that some of these elements of growth theory will produce in the analysis of comparative advantage. FACTOR COSTS. It is generally agreed that costs of labor and capital in developing countries do not reflect their opportunity costs \vith any accuracy because of market imperfections, but there is \Vide dis- agreement as to the extent of the typical discrepancies. Some types of labor may be overvalued while particular skills are undervalued. Fac- tor costs may also change markedly over time as a result of economic development, so that an advantage based on cheap labor may prove quite limited in duration. As Lewis (1954) and Hagen (1958) have shown, the effects on comparative advantage of correcting for disequi- librium factor prices are often very substantial. (The effects of dis- equilibrium in factor markets are discussed further on pages 281-99.) EXPORT MARKETS. Two of the main arguments against the trade pat- tern produced by market forces concern (a) the fluctuating nature and (b) the low income and price elasticities of the demand for pri- mary products. The existence of cyclical fluctuation is wen established, but the income and price elasticities vary considerably among primary commodities. Their net effect on the terms of trade of primary pro- 8. In his survey of trade theory, Caves (1960) shows that attempts to introduce dynamic elements have been concerned mainly with particular aspects and have led not to new principles, but rather to extensions of statie results. CONF'LICTS BETWEEN TRADE THEORY AND GROWTH THEORY 279 ducers over time is a matter of dispute,9 These characteristics are often used as an argument for reducing specialization in underdevel- oped countries and for expanding industry for local consumption rather than expanding primary exports. lO These factors can be admitted without seriously modifying the prin- ciple of comparative advantage. 1be market value of the stream of export earnings should be reduced to reflect the drawbacks to the economy resulting from its variable characteristics, and this social value should be used in comparing investment in primary exports to other alternatives. \Vhen export demand has a low elasticity, marginal revenue should be used in place of average revenue. Since it is quite likely that the market evaluation of the attractiveness of an invest- ment in exports will differ from this social evaluation, some form of government intervention may be warranted. It is wrong, however, to conclude from this analysis that continued specialization in primary exports may not be the best policy because even the corrected re- turn on cxports may be greater than that on alternative investments. The supply of foreign investment may also be greater for export production. PRODUCTIVITY CHANGE. The possibility of rising efficiency as labor and management acquire increasing experience in actual production has long been recognized and forms the basis for the infant industry argument.n This argument has been generalized to include the effects of increasing production in any industry on the supply of skilled labor and management available to other industries. Since manufacturing is thought to have more important training effects than primary produc- tion,12 the fact that improvements in factor supply are not reflected in the market mechanism may introduce a bias against manufactur- ing. The empirical basis for this argument has been questioned by several economists, who assert that there is often as much scope for technological improvement in agriculture as in industryY~ Without trying to settle the empirical question that has been raised, it may be concluded that productivity change is an important factor and there- 9. See Viner (1958). 10. See Singer (1950) and Prebisch (1959). 11. See Williams (1929). 12. See Myint (1958) and Prebisch (1959). 13. See Viner (1952) and Schultz (1956). 280 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY fore that comparative advantage should be measured over time. It cannot be said, however, that allowance for this factor will always favor manufacturing. DYNAMIC EXTERNAL ECONOMIES. As indicated above, dynamic exter- nal economies are received by an industry from cost reductions or demand increases in other sectors. Cost reductions may result from economies of scale, productivity increases, or new technology. The cus- tomary analysis of comparative advantage sector by sector would re- quire that the cost reduction from simultaneously developing related sectors be allocated separately to each. If a group of investments ,\Till only bc profitable when they are undertaken together, however, com- parative advantage can only be determined for alternative combina- tions of investments. As shown in chapter 5 of this book, not only do market prices fail to produce the best investment allocation in this situation, but any structure of equilibrium priccs may also be an inade- quate guide in the presence of economies of scale. There is considerable evidence that external economies are more important in the industrial sectors than in primary production because of internal economics of scale, training effects, and high demand elas- ticities. lneir omission from the market mechanism is therefore likely to bias resource allocation against manufacturing. rThe quantitative significance of this factor is very hard to determine, however, since it involves simultaneous changes in a number of seetors. UNCERTAINTY AND FLEXIBILITY. The limited ability of polieymakers to foresee changes in demand and supply conditions puts a premium on flexibility in the choice of a development strategy. This factor not only argues against specialization in one or two export commodities but it also favors the development of a diversified economic structure which will enable the economy to shift to new types of exports or import substitutes when changing trade conditions may require them. Kindleberger (1956) saw this factor as the main explanation for his finding that the terms of trade have favored developed countries al- though they have not favored countries exporting manufactured goods in generaJ.14 The argument is similar to that of Stigler (1946) con- cerning the optimum choice of techniques in a manufacturing plant. 14. This argument is also discussed by Caves (1960, pp. 264-66) . THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATION 281 The optimum design for a changing market is likely to differ from the optimum under static conditions because in the former case the proper criterion is lowest-cost production for varying operating levels and with changes in product design. Similarly optimum development policy should result in a pattern of resource allocation that allows for unforeseen changes in supply and demand conditions even at the cost of some loss of short-term efficiency. The Measurement of Optimum Resource Allocation The development of an adequate theory is only the first step in for- mulating economic policy. To reach practical conclusions, it is also necessary to specify tbe environment in which the policymaker functions. Relevant aspects of a particular society include its general objectives, the policy instruments to be considered, and the informa- tion available. Tbe theory must then be combined with these ele- ments in such a way as to yield guides to action or "decision rules" for particular situations. Although the growing science of operations research is concerned with the development of decision rules for business and military oper- ations, less progress has been made in developing an operational ap- proach to long-run economic policy. Tinbergen (1956) and Frisch (1958) have outlined a general framework for policy analysis, but it has had relatively little impact on the discussion of the development of underdeveloped countries. In this field the failure to specify ade- quately the decision making environment and to distinguish between decision rules and the corollaries of pure theory has led to considerable confusion. Since the information needed for overall economic analysis is avail- able to a very limited extent in developing countries, there has been a considerable effort to derive decision rules or investment cri- teria that can be based on partial analysis. I shan group the various suggestions into three categories: (a) factor-intensity criteria; (b) productivity criteria; and (c) programming criteria based on account- ing prices. Although these various approaches often lead to contradic- tory results, each has some merit as a form of decision rule if properly qualified. In general, the theoretically more valid formulations require more information and must be replaced by cruder approximations when adequate data are not available. Since a major part of the litera- 282 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY ture in the development field has been devoted to the discussion of investment criteria, it is important to identify the sources of conflict among them and to specify the circumstances under which each may be approximately correct. In economic theory, capital and labor are assumed to be separately allocated in single units to different uses. In national planning, how- ever, it is more convenient to consider the decision to instaH a given productive process or plant, representing the allocation of a group of inputs in specified quantities, as the basic choice. Investment criteria are customarily formulated for projects of this sort, since they form the basis for the decisions of planning authorities. This procedure rec- ogIlizes that very small productive uIlits are uneconomical, and it per- mits a consideration of different scales of output. TIle choice of tech- niques can be considered as a choice among projects producing the same output from different input combinations. In this way the allo- cation procedure can be divided into two steps: the choice of the best technique for a given type of product, and the decision whether to produce the commodity at all. The principle of comparative advan- tage is more directly relevant to the second type of choice, but the two cannot be separated entirely. Factor-intensitv criteria TIle simplest approach to any allocation problem is to concentrate on the scarcest resource. Since this is often capital in underdeveloped countries, it seems reasonable to choose the technique that uses the least capital to produce a given output. 111e same logic is applied to the choice of sectors of production: an underdeveloped country is ad- vised to produce and export commodities that use relatively less capital per unit of output and to import items requiring more capital. Statements of this type occur in many economic writings of the 1950s. Buchanan (1945) was among the first to state this criterion for invest- ment in underdeveloped countries and to base policy recommenda- tions upon it. The "minimum capital-output ratio" criterion is only valid under the following restrictive conditions 15 : (a) Either capital is the only 15. A rigorous analysis of the validity of marginal and average factor-output ratios as indicators of optimum allocation in a two· factor system is given by Bator (1957) . THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATION 283 scarce factor in the system, or other inputs are so abundant relative to capital that the latter is the dominant element in determining cost differences. (b) Either the same output is produced by each invest- ment alternative, or the market values used to compare the different products coincide with their social values. (c) Production takes place under constant costs. 'Tne use of the capital-output ratio theoretically requires a measure- ment of the total capital used in producing a given commodity, includ- ing the capital used in producing all materials and services purchased. Alternatively, the indirect use of capital can be allowed for by deduct- ing the east of purchased inputs from the value of output and express- ing the eriterion as the ratio of capital to value added. This procedure requires the further assumption that market prices correetly reflect the use of capital in the rest of the economy. A closely related allocation criterion is the capital intensity: the ratio of capital to labor. This test is derived directly from the Heckscher-Ohlin version of the comparative cost doctrine. If the same production functions exist in all countries and if capital is scarce rela- tive to labor in the developing countries, comparative advantage in the latter ean be identified by low capital-labor ratios. This ap- proaeh does not assume that labor has zero opportunity cost, as does use of the capital-output ratio, but only that the ratio of labor cost to capital cost is lower than in the country's trading partners. To allow for differences in the quality of labor among countries, it is sometimes suggested that the assessment of relative labor cost should be made for labor units of equal efficiency-for example, the labor required in each eountry to perform a given type of operation with the same capital goods and organization. A prineipal criticism of the use of both these ratios is that they ignore the existence of other factors of production, such as natural resources. If either labor or natural resources has a significant oppor- tunity cost, the capital-output measure must be replaced by the more general marginal productivity of capital criterion, which is discussed in the next subsection. To judge comparative advantage by the capital-labor ratio is to as· sume either that this ratio will be the same for the same industry in all countries, or that capital is equally substitutable for labor in produc- ing all the commodities traded. Deviations from these assumptions, along with the omission of other inputs and variations in efficiency by 284 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY sector, make the capital-labor criterion a very crude approximation indeed to a proper estimate of comparative advantage. Marginal productivity criteria A more comprehensive allocation criterion is the social marginal product of a given unit of resources in a given use. \\There the factor- intensity criteria are at best only correlated with the increase in na- tional income produccd by a project, the productivity criteria try to measure the increase. The marginal productivity test is in turn less general than the over-all programming approach, because it is based on a partial equilibrium analysis that is only valid for relatively sman changes in the economic structure. 16 The several forms of marginal productivity criterion that have been proposed differ in the assumptions made about the social welfare func- tion and in the extent to which allowance is made for the indirect effects of a given allocation. All versions are alike in assuming that the government controls, directly or indirectly, a certain fraction of the investable resources of the country and wishes to allocate them in such a way as to maximize future welfare. Since the productivity criteria are usually applied to investment projects rather than to single units of capital, they are marginal only in the sense that a project normally constitutes a small fraction of the total capital invested in a given year. For very large projects a break- down into smaller units would be more appropriate. TIlE STATIC SMP CRI'IERION.As proposed by Kahn (19;1), the social marginal product (SMP) is a general equilibrium concept which is conventionally defined as the net contribution of a marginal unit (project) to the national productP The related decision rule is to rank investment projects by their SMP and to go down the list until the funds to be allocated are exhausted. Alternatively, any project having an SMP above a given level can be approved. Kahn used the SMP criterion to show the fallacies in thc factor- 16. Surveys of marginal productivity and other investment criteria are given by Castellino (1959), Vaidyanathan (1956), and the United Nations (19591). 17. To be more accurate, cost and output streams should be discounted to the present, but I shall not be concerned with differences in the time pattern of output of different projects. THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATION 285 intensity measures that had been advocated by Buchanan (1945), Polak (1943), and other writers. He pointed out that: 'The existence of a particular natural resource, specialized skills, particular climatic conditions, or the importance of a particular product or service may make the SMP of capital higher in a line which is more capital inten- sive than in another which is less 50."18 He also argues that even when there is substantial rural unemployment, a considerable amount of capital and other inputs are required to transport, train, and house the workers who are to be employed elsewhere. Kahn's arguments against the simple capital-intensity criteria appear to have been generally accepted, although he admits that a lower capital-output ratio may be a useful guide when other information is lacking. Some years ago I suggested modifications in the SMP criterion to allow for artificial elements in the price system (tariffs, subsidies, and so forth) and to provide for the evaluation of labor and foreign ex- change at opportunity cost rather than at market value.19 Further allowances for the difference between market price and social value can be made by estimating the benefits to be provided to other sec- tors in the form of external economies, and by including overhead costs in the estimate of the cost of labor. All of these elements are included in Eckstein's synthesis and extension of the productivity approach. 20 The SMP criterion is entirely consistent ,vith the general program- ming approach discussed below, which derives opportunity costs from an explicit analysis of total factor use. In the absence of such an over- all analysis, the corrections suggested for the calculation of the pro- ductivity of investment are likely to be quite approximate. There is no logieal conflict between the results of the SMP analysis and the dictates of comparative advantage because each is a corollary of a general equilibrium solution over a given time period. THE MARGINAL REIl\;-VESTMENT CRITERION.A sharp criticism of the SMP criterion was made by Galenson and Leibenstein (1955), who 18. Kahn (1951, p. 40). 19. See Chenery (1953). 20. Eckstein (1957) pointed out that the assumption of capital rationing im- plies a social judgment as to both the amount of investment in the current period and the discount to be applied to future outputs, since the market rate of interest is rejected for both purposes. 286 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY challenged some of its basic premises. They would substitute a dif- ferent social welfare function in which the aim is to maximize per capita income at some time in the distant future rather than to maxi- mize a discounted stream of income over time. They also assume severe restrictions on the policy instruments available to the govern- ment, and in particular deny its ability to affect the rate of saving by fiscal measures. Under these assumptions, it is necessary to take ac- count of the division of income resulting from a project between profits and wages, since savings from the former are higher. To maximize the total output at some distant future time, Galen- son and Leibenstein easily show that the most "productive" project is not necessarily the one that maximizes national income in the near future but the one that leads to the highest savings. Sincc it is assumed that neither voluntary saving nor taxes can be extracted from wages, the most productive project will be the one with the highest profit rate per unit of capital invested. 21 The assumption that profits are saved and reinvested leads to the marginal reinvestment quotient as a decision-rule to be applied in place of the SMP. Galenson and Leibenstein push their argument one step further and identify the most profitable projcct as the one with the highest capital-labor ratio. 1bis result leads them to the paradoxical conclu- sion that the factor-intensity rule should be reversed: countries should prefer the most capital-intensive rather than the least capital- intensive techniques to promote savings and future growth. This conclusion involves an implicit assumption about the nature of pro- duction functions: that increasing the capital intensity will neces- sarily raise the average return to capital in each sector of production. This assumption is obviously not true in general and is not necessarily true of existing productive techniques. 1be savings effect of a given project should therefore be measured directly and not assumed to vary in proportion to the capital-labor ratio. Galenson and Leibenstein have been widely criticized for their extreme assumptions, in particular for the use of a social welfare func- tion in which the starvation of half the population in the near future would appear to be a matter of indifference and for the assumption that limitations on fiscal policy make a lower income preferable to a 21. I omit the possibility of an effect on population growth, which leads Galenson and Leibenstein to state the criterion on a per capita basis. THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATION 287 much higher one if the former has a higher savings component.22 Their analysis has nevertheless been useful in emphasizing that other effects of an investment beside its immediate contribution to the na- tional product should be included in the productivity criterion.23 THE MARGINAL GROWTH CONTRIBUTION. Eckstein (1957) successfully reconciled the conflict between the Kahn-Chenery SMP approach and the Galenson-Leibenstein reinvestment approach, and in so doing he has provided a considerable generalization of each. First, he assumed that the social objective is to maximize the present value of the future consumption stream. With a zero discount rate, this objective ap- proximates the long-term income objective of Galenson and Leiben- stein, while with a high discount of future consumption it leads to the maximization of income in the short term. Second, Eckstein assumed that there is a different savings (reinvestment) coefficient associated with each project, but he allowed for any savings rate out of wages and profits. From these assumptions, he derived a measure of the mar- ginal growth contribution of a given project that consists of two parts: (a) an efficiency term, consisting of the present value of the consump- tion stream; and (b) a growth term, consisting of the additional COn- sumption to be achieved by reinvesting savings. The relative importance of the two terms depends largely on the rate of discount that is applied to future consumption. Even with a low rate of discount, the significance of the second term depends on how much variation there is in the fraction of income saved among different projects. If the savings ratio is not related to the form of in- come genera ted, then, as Bator (1957) showed, there is no conflict between maximizing income in the short run and in the longer run. Eckstein's formula provides for all possib1e intermediate assumptions between the two extreme views of the determinants of savings. 24 In principle, one might include other indirect dynamic effects, such 22. See, for example, Bator (1957), Eckstein (1957), Hirschman (1958a), and Neisser (1956). 23. Leibenstein (1958) restated in more restrained form his arguments for in· c1uding labor training, savings, population growth, and other indirect effects in a comprehensive productivity measure. 24. Sen (1957) independently formulated a more general investment criterion that is very similar to Eckstein's, in which the SMP and reinvestment criteria are shown to be limiting cases. 288 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY as the value of the labor training provided, in the measurement of the total productivity of a given project. There is a danger of double counting if partial-equilibrium analysis is extended too far, however, and most indirect effects can be more readily evaluated in the more general programming framework considered below. Programming criteria and accounting prices The allocation rules discussed up to now are based on the existing economic structure and are strictly applicable only for relatively small changes in it. Although it may in many instances be necessary to rely primarily on these marginal criteria for lack of data on the rest of the economy, it is important to have some way of testing larger changes and of evaluating the errors that are introduced by the marginal pro- cedure. Furthermore, without a more comprehensive analysis it is impossible to reconcile fully the conflicting policy implications of comparative advantage and growth theory. The difficulties of partial analysis increase with the number of modifications that have to be applied to market prices to arrive at social value. Both the factor-intensity ratios and the partial produc- tivity measures assume that there is one principal restriction on the system, the scarcity of capitaL Those measures do not allow for the fact that in allocating capital according to anyone of these rules some other restriction on the system, such as the supply of foreign exchange, of skilled labor, or of a particular commodity, may be exceeded. The programming approach to resource allocation begins with the problem of balancing supply and demand for different commodities and factors of production. Until recently, practical programming methods have been more concerned with ensuring the consistency of a given allocation of resources with certain targets than with testing the efficiency with which resources are used. Historically speaking, the programming approach is thus the operational counterpart of the theory of balanced grov,th, from which much of its conceptual frame- work is derived. One of the earliest attempts at formulating a comprehensive devel- opment program for an underdeveloped area was Mandelbaum's illus- trative model for southeastern Europe. 25 He started, as have many 25. See Mandelbaum (1945). THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATIO~ 289 subsequent programs, from an estimate of the increase in national in- come required to absorb a prospective increment in the labor forcc. The allocation of capital and labor was made initially from demand estimates and by analogy to the structure of more advanced countries. The principle of comparative advantage was only introduced in- tuitively in modifying the initial projection. The main test of resource allocation is the balance of demand and supply for each sector and factor of production. The development of mathematical programming methods makes it possible to carry out this type of analysis in a much more precise way. In several countries, consistent development programs have been for- mulated by using input-output analysis, as in the studies of the Eco- nomic Commission for Latin America (UN, 1957, 1959a, and 1960). It was only with the development of linear programming, however, that it became possible to reconcile the consistency criteria and the produc- tivity criteria in a systematic way. A link between the test of consistency (feasibility) in resource allo- cation and the test of productivity (efficiency) is provided by a con- sideration of the price implications of a given allocation. Assume that a set of production levels has been worked out so as to be consistent with the available supplies of labor, capital and natural resources, given the structure of consumer demand and the country's trading possibilities. These sector production and trade levels constitute a "feasible program." Any such program implies a unique set of com- modity and factor prices if the economy is in equilibrium. If pro- duction activities are assumed to operate at constant costs, linear programming provides a method of calculating the shadow prices cor- responding to the equilibrium conditions, in which the price of each commodity is equal to its cost of production. 26 Prices are determined by the solution to the follO\ving set of simultaneous equations, one for each production activity included in the program: (7.1 ) (I=1, ... ,n) where aij is the input or output of commodity or factor i by activity i, and Pi is the shadow price of commodity or factor i. The input coeffi- cients may be measured at existing prices or in other convenient units. 26. The assumptions of linear programming and methods of finding solutions to programming models have been discussed in a number of publications, such as Dorfman, Samuelson, and Solow (1958). 290 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY In an open economy, activities of importing and exporting are also included in the system, and the price solution contains the equilibrium price of foreign exchange. An example of this calculation is given in table 7-1, which will be explained shortly. The use of shadow or accounting prices in evaluating investment projects was suggested by Tinbergen (1955 and 1958), Frisch (1958 and 1959), and Chenery (1955 and 1958). Although Tinbergen did not use a linear programming framework, his accounting prices for factors had the same meaning as shadow prices: the opportunity cost implied by a given resource al1ocation. 27 He suggested computing the costs associated with a project by using accounting prices; any project that showed a positive net return over cost (including capital cost) should be approved. This test is equivalent to the SMP criterion, as shown below. The general linear programming problem is to maximize the value of a linear objective function subject to linear constraints. In develop- ment programs, the principal constraints are that the demands for commodities and factors should not exceed their supplies; the function to be maximized is usually taken as the national income. Alternatively, the objective may be the aehievement of a given increase in output at minimum cost in investment (including foreign investment). Other social objectives, such as a minimum employment level or a specified degree of regional balance, can be included as additional restrictions on the program. The instrument variables can also be constrained to fall within specified limits, as in the models of Frisch. 28 To illustrate the meaning and use of shadow prices in evaluating investment projects, I shall take up a very simplified programming 27. Tinbergen defines accounting prices as those "that would prevail if (i) the investment pattern under discussion were actually carried out, and (ii) equilibrium existed on the markets iust mentioned" [that is, labor, capital, and foreign ex- change markets]; see Tinbergen (1958, p. 39). The relation between accounting and shadow prices is discussed in Chenery (1958) and Qayum (1960). 28. Frisch was one of thc strongest advocates of the usc of linear programming for development planning. He wrote: "In the beginning of 1959, during my work as a United Nations expert in Cairo, I was confronted with the problem of work- ing out a methodology for optimal investment programming in a rapidly expanding underdeveloped country. I have always believed-and my Cairo experiences have confirmed it-that such a method must be formulated in terms which ultimately make the problem amenable to linear programming. Otherwise one is practically certain to be taken by surprise afterwards in unexpected balance of payments diffi- culties and other troubles"; see Frisch (1959, p. 1). THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATION 291 model that is worked out in more detail in chapter 5. The truncated system given in table 7·1 covers only a small part of the economy, but it will serve to illustrate the way in which interdependence influences investment decisions and the effect of having more than one scarce factor. The model contains four production activities (Xt,X2 ,XS,X",) and three import activities (Ml,M2,Ma). Each activity is represented in table 7-1 by a column of coefficients, a,j, showing the amount of input ( -) or output (+) of commodity i when the activity is operated at unit level. (1bese coefficients are the boldface figures in columns 1 to 7.) The net output is taken as unity in all cases. The production activity Xl, for example, represents the production of one unit of metal products from 0.22 units of iron and steel, 0.20 units of "other inputs," 0.70 units of labor, and 0.70 units of capital. The import activity Ml provides an alternative way of supplying a unit of metal products by an expenditure (input) of 0.85 units of foreign exchange. A similar choice is provided between X2 and M2 (iron and steel) and between Xs and Mg (iron ore). The fourth production activity shows the resources used in the marginal export sector to provide a unit of foreign exchange. In a complete programming model, the amounts of all commodities required for final use at a given level of income would be entered as restrictions on the solution. Similarly, the amounts of available capital and labor of different types would be specified. In this limited illustra- tion, the problem is to supply requirements of 1,000 each for metal products and iron and steel at minimum cost. Iron ore and foreign exchange are therefore taken to be intermediate goods having no net outside demand. "Other inputs," labor, and capital are supplied from outside the model at prices reflecting their opportunity costs in the rest of the economy. The main difference in principle between this submodel and a complete programming system is that the prices of only the first four commodities are determined in the model in the present case, while in general all prices are so determined. The four restrictions in the model consist of equations stating that the supply of each of the first four inputs must be equal to the speci- fied demand 29 : 29. I omit the possibility of overfulfilling demands, since there are no joint products in the present case. 292 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY Table 7-1. Evaluation of Production and I mport Activities by Accounting Prices Production activities Commodities and Xl X2 X3 X4 factors (1) (2) (3) (4) Metal products 1.00 (3.41 ) Iron and steel -0.22 1.00 ( -0.89) (4.03 ) Iron ore -0.08 1.00 ( -0.25) (3.12) Foreign exchange 1.00 (4.01 ) Other inputs -0.20 -0.25 -0.70 -0.10 ( -0.62) (-0.78) (-2.17) (-0.31 ) Labor -0.70 -0.20 -0.30 -1.00 (-1.05) ( -0.30) (-0.45) ( -1.50) Capital -0.70 -2.70 -0.50 -2.20 ( -0.70) ( -2.70) ( -0.50) ( -2.20) Social profitabilityb Trial a -0.59 -0.41 +0.25 -1.00 Trial b -0.03 +0.37 +1.23 0 Trial c +0.15 0 0 0 Trial d 0 -0.03 0 0 Production and im- port levels Trial a 0 0 0 2,050 Trial b 0 1,000 80 850 Trial c 1,000 1,220 98 0 Trial d 1,000 0 0 1,464 Note: This table is based on table 5·1. Prices satisfy equation (7.1) except for P. in trial 1. Figures in parentheses are (a'i P.) for trial c. a. No restriction. b. Calculated from equation (7.4). THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATION 293 Import activities Accounting prices --- Trial Trial Trial Trial Restric- Ml M2 M3 a b c d tions (5) (6) (7) (8) (9) (10) (11) (12) 1.00 (3.41 ) 2.55 3.42 3.41 2.26 1,000 1.00 (4.03 ) 3.60 4.82 4.03 3.50 1,000 1.00 (3.12 ) 3.30 4.42 3.12 2.19 0 -0.85 -1.20 -1.10 (- 3.41) (-4.81 ) (-4.41 ) 3.00 4.02 4.01 2.92 0 3.00 3.20 3.10 2.20 1.50 1.50 1.50 0.50 1.00 1.00 1.00 1.00 - · 0 a 0 a 0 0 0 -0.78 -1.29 -0.22 0 -1.02 1,000 1,000 o 1,000 o o 1,220 ° ° ° o o 294 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY (7.2) Xl + Ml 1,000 -0.22Xl X2 + M2 1,000 X4 - -0.08X2 + X3 + M3 = O.85M! - 1.20M2 - 1.10 M3 = ° ° 'The objective is to minimize the amount of capital required to sup- ply the given final demands, with the use of labor and "other inputs" valued at their opportunity costs in terms of capital. This is the same as supplying each commodity at minimum unit cost, since the amount of each to be supplied is fixed. A feasible solution to the model contains either a production or an import activity for each of the three commodities plus the export ac- tivity for foreign exchange. The corresponding activity levels can be determined from equations (7.2) and are shown at the bottom of table 7-1. The amounts of the outside factors (Fi)-labor, capital, and "other inputs"-required by each solution can then be determined from the following equations: (7.3) Other inputs: F5 = 0.20Xl + 0.25X2 + 0.70Xa + O.lOX 4 Labor: F6 0.70Xl + 0.20X2 OJOX + 1.00X 3 4 Capital: F7 O.70Xl 2.70X2 0.50Xg 2.20X4 The programming model thus contains two types of equations: price equations of the type of (7.1), and equations for the supply and demand of commodities and outside factors, (7.2) and (7.3). As out- lined in Chenery (1958), the general procedure for solving a program- ming model of this type involves three steps: (a) finding a feasible program or set of activity levels that satisfies the supply-demand re- strictions; (b) calculating the shadow prices associated with the given program; and (c) using these prices to determine whether any im- provement in the initial program is possible. This procedure is re- peated as long as any further improvements can be made. The programming criterion used to compare projects or activities is the social profitability of each as measured from the shadow prices. Any profitable activity should be included in the program. It is the re- calculation of prices that distinguishes this procedure from the partial programming approach suggested by Tinbergen. In either case, how- ever, the test of social profitability of activity i can be expressed as: (7.4 ) II j - I ajjP;. i By definition, the activities that were used in determining the shadow prices will have a profitability of zero. The optimum solution is iden- THE MEASUREMENT OF OPTIMUM RESOURCE ALLOCATION 295 tified by the condition that all other activities have zero or negative profitability. Some idea of the type of adjustment that results from moving from partial toward general equilibrium analysis may be given by determin- ing solutions to the model in table 7-1 under four different procedures: (a) the use of market prices; (b) correcting for the overvaluation of foreign exchange; (c) finding the optimum solution for the submodel alone; and (d) finding the optimum solution for the submodel with changes in the opportunity costs of labor and other inputs determined from a general programming model. 'The accounting prices corre- sponding to each assumption are shown in columns 8 to 11 of table 7-1. The calculation of social profitability of each activity, given the accounting prices, is illustrated in the table for trial c by giving cost and revenue figures in parentheses in columns 1 to 7. For trial a, assume that market prices are based on the cost of im- porting and are determined by setting profits on the import activities equal to zero, with a given foreign exchange cost of 3.00. The exchange rate is assumed to be overvalued, so that the price of foreign exchange is less than the cost of securing it through expanded exports. At these market prices, only activity Xa (iron ore) is profitable, but there is no domestic demand for iron ore unless steel is also produced (the export price is lower than that of imports because of transport costs). The use of market prices therefore leads to imports of steel and metal products, since the opportunity cost of expanding exports is not taken into ac- count. The corresponding activity levels are shown at the bottom of the table. Assume now that trial b corrects for the existing structural dis- equilibrium by setting the price of foreign exchange equal to its oppor- tunity cost of 4.02 as determined from the export activity X 4 · Allow- ance is also made for a rise in the aecounting price of "other inputs," some of which are imported. A new set of accounting prices for com- modities 1 through 3 is determined from the cost of imports. Substi- tuting these prices into equation (7.4) shows that X 2 and Xa are both profitable (7r2 = 0.37, 'Ira 1.23). Investment should therefore take place in steel, iron are, and exports on this test. In trial c, to find the optimum solution to the submodel by linear programming, we can start from trial b and recalculate the shadow prices from the activities that are included: X 2,Xa,X4 ,M1 · The four shadow prices P 1 to P 4 are determined by applying equation (7.1), taking the prices of the outside inputs (P 5, P6 , and Pd as given. The 296 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY elimination of excess profits from the prices of iron ore and steel low- ers the cost of producing metal products, providing an example of pecuniary external economies. Instead of a loss, activity Xl now shows a profit of 0.15 and should be substituted for the import activity M 1 · \Vith the original prices for labor and capital, the optimum solution to the submodel is therefore to produce all three commodities and import nothing, since all import activities are unprofitable. If in trial d a similar analysis is carried out for the economy as a whole, it is likely that the initial estimate of the opportunity cost of labor (equal to its market price) will be revised. Assume that the shadow price of labor (equal to its marginal product in the rest of the economy) is only a third of its market price, or 0.5 units of capital. This lower labor cost will reduce the costs of production in different activities in proportion to their use of labor. Since exports are cheap- ened more than steel production by this calculation, it now becomes socially profitable to import steel and produce metal products. TIle optimality of this solution is sho\'\l1:1 by the prices in trial d, in which there is a loss of -0.03 on X g· The optimum quantity solution is shown at the bottom of the table. Valuing other inputs and labor at their accounting prices, trial d has a capital cost of 5,760, compared to 8,200, 7,470, and 7,290 in trials a, b, and c. Ine programming approach of trials c and d adds two elements to the analysis of accounting prices. The first is the inclusion of repercus- sions on input prices from investment in supplying sectors. This is one of the main types of dynamic external economies that are omitted from partial analysis. This element is much more significant when there are economies of scale. The second element is the revision of the initial estimate of the opportunity costs of labor, capital, and foreign exchange. This revision is determined by the relation between supply and demand for these factors and thus takes into account the require- ments of feasibility.30 Tne profitability criterion (sometimes called the simplex criterion) that is used in linear programming is logically equivalent to the SMP 30. An example in which these successive adjustments are calculated in detail is given in Chenery (1958). Frisch (1959) outlined a computational procedure for handling large numbers of investment projects without going beyond the capacity of simple calculating equipment. THE MEASUREMENT OF OPTIMU!vI RESOURCE ALI.DCATION 297 test if the same prices are used in both. The two can be put in a com- parable form as follows: (7.4a) Social profit on activity j: (7.5) SMP of investment in activity i: where -kj is used for the capital input coefficient instead of d7j. An activity having a positive social profit in equation (7.4a) will have an SMP of greater than 1.0 in (7.5), and the same projects would be ac- cepted by either test. If the prices used are not the equilibrium prices, however, the project rankings by the two formulae will not necessarily be the same. Although the example given here contained only one technique of production for each commodity, linear programming methods readily encompass alternative techniques. In a trial application of linear pro- gramming to Indian planning, Sandee (1959) includes three alterna- tive ways of increasing agricultural output-increased use of fertilizer, irrigation, and extension services-which are substitutes over a limited range. The four alternative techniques for producing textiles cited by Galenson and Leibenstein (1955) could also be more properly evalu- ated in a programming model in which the cost variation associated with their different requirements for materials, maintenance, and skilled labor could be included. But it is only necessary to include alternative techniques in a programming model when the choice between them depends on the outcome of the solution. Probably in most cases the range of shadow prices can be foreseen accurately enough to determine in advance which technique is more efficient for a given country. The initial assumption can always be verified after the analysis has been completed by using the resulting prices. Linear programming can be extended to include many of the indi- rect effects of investment that are suggested by growth theory. The production of trained labor, the effect on savings, or other indirect benefits can be considered as joint outputs whose value can be speci- fied in the objective function. Similarly, indirect costs of production, such as the provision of housing to urban workers, can be included as additional inputs. The shadow prices computed from such an ex- 298 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY panded system will therefore reflect nonmarket as well as market inter- dependence to the extent that it can be specified in quantitative form. In formal terms, it is also quite easy to extend the programming model in time and to compute future prices for commodities and factors. 'Ine measurement of social profitability could then be made against a pattern of changing future prices. Given the degree of uncer- tainty attached to all fuhlre economic magnitudes, however, this is not likely to be a very useful procedure beyond the customary five-year planning period except in the most general terms. It would, however, be desirable to cstimate the change in the equilibrium prices of foreign exchange and labor over a longer period of time, since these are the most important variables in choosing among investment projects. Investment criteria and comparative advantage The linear programming approach provides a convenient link to the principle of comparative advantage because the optimal pattern of trade is determined simultaneously \\1th the optimum allocation of investment. TIle model is considerably more general than that of mar- ket equilibrium because it allows for different social objectives and takes account of costs and benefits other than those entering the market. The limitations to the programming model are of two sorts: the form of the restrictions that are specified, and the omission of relations that cannot be expressed in quantitative form. Thc introduction of inelastic demands or increasing costs does not create any more theoretical difficulty in a programming model than in the corresponding general equilibrium system, although the compu- tational aspects of such models have not been widely explored. TI1e accounting prices perform the same function as guidcs to proper allo- cation, but the test of social profitability must be applied in marginal rather than average terms. In development programs, this modification is particularly important in the case of exports, where the price elas- ticity of demand is often rather low.Hi As Nurkse (1959) points out, marginal comparative advantage for the developing countries may for this reason be quite different from that inferred from the average costs and prices of primary exports. The existence of increasing returns creates the same problem for the 31. A programming model including this feature is given in Chenery (1955). COMPARATIVE ADVANTAGE AND BALANCE 299 programming model as it does for equilibrium theory. Marginal-cost pricing is not sufficient to determine whether an investment should be undertaken, and the total cost of alternative solutions must also be considered. Although practical methods of solving programming models containing decreasing costs now exist, they do not give allo- cation criteria that rely only on accounting prices. It is approximately correct to say that beyond a certain output level country A has a comparative advantage in the production of steel, but the precise determination of the break-even point depends on the level of output in other sectors also. 32 The most serious theoretical qualification to the principle of com- parative advantage comes from the type of nonquantitative interde- pendence among sectors that is assumed by Hirschman (19 58b ). If, as he supposes, one growth sequenee is more effective than another be- cause it economizes on decisionmaking ability or provides a greater incentive to political action, a set of criteria having little or nothing to do \vith comparative advantage is implied. The empirical significance of these psychological and sociological factors remains to be estab- lished, but they lead to a conflict that cannot be resolved in economic terms. When the practical limitations on information and analysis are rec- ognized, the possibilities of conflict between comparative advantage and growth theory are greatly increased, and \Viles (1956) suggests that marginal efficiency calculations may be less important. An aver- sion to risk-taking may be a valid reason for limiting the extent of speCialization in the export of primary products beyond the amount that would be optimum in the light of more accurate information. An inability to measure the extent of economies of scale, labor training, and other sources of external economies also makes possible a con- tinuing disagreement as to their magnitude. Comparative Advantage and Balance in Development Programs The inconsistent procedures that governments employ in formu- lating development policies are probably the most important source of 32. The nature of solutions to this type of problem is considered in ehapter 5. In this situation of decreasing average cost, the programming model may provide a greater improvement over the solution using partial criteria. 300 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY conflict between the dictates of comparative advantage and of growth theory. Official pronouncements on development policy usually allege that both types of criteria have been (or should be) used in drawing up the program that is put forward, but the procedure followed in reconciling conflicts between the two is rarely made explicit. Since the analytical basis of most development programs is quite limited, it is important to look into the procedure that is actually used to discover sources of bias. Development programs must simultaneously confront two sets of problems. In the short run, progress is hampered by structural dis- equilibrium in factor markets and in the demand and supply of particular commodities. 'Ibis disequilibrium is reflected in the balance- of-payments difficulties that beset most low-income countries as they try to accelerate the process of development. In the longer run, the choice among sectors becomes increasingly important because the pattern of growth in each period will depend on the choices made previously. Development programs that are influenced mainly by the existing structural disequilibrium therefore tend to stress the need for greater balance between domestic demand and supply, while those that take a longer view tend to pay more attention to comparative advantage. Although the procedures actually followed cannot be ascertained with any accuracy by an outside observer, these two aspects can be identified from characteristic elements in the analysis. The balanced growth approach is generally associated with target-setting in impor- tant sectors, stress on the avoidance of bottlenecks, and attempts to equate the supply and demand of labor, capital, and the more impor- tant commodities. The extreme cases of this type of procedure are found in the communist countries. Less extreme examples, in which some attention is paid to comparative advantage, are the procedures of the Indian Planning Commission and the UN Economic Com- mission for Latin America (ECLA). Characteristic elements of the comparative advantage approach are attempts to measure the relative efficiency of different types of pro- duction, the weighing of balance-of-payments improvements against other benefits to the economy (by means of accounting prices or other- wise), and usually a greater emphasis on partial analysis than on over- all projections. Examples that will be cited are Puerto Rico, the Philippines, and Israel. COMPARATIVE ADVANTAGE AND BALANCE 301 Procedures emphasizing domestic balance The planning procedures developed in the USSR and applied with some modification in other communist countries represent in extreme form the use of balance as a criterion for resource allocation and the virtually complete omission of any test of comparative advantagc. As revcaled in studies by Montias (1959) and Balassa (1959), the main tool of Soviet-type planning is a very detailed system of material bal- ances specified in quantitative terms. Policy objectives are translated into production targets in which priority is given to heavy industry and other sectors that are expected to contribute to further growth ("leading links"). Prices are used mainly as rationing devices and have no necessary connection with production costs. The cumbersome cal- culations involved in arriving at balance of supply and demand for a large number of commodities limit the alternatives that can be tried out, so the main effort is to find a feasible program.33 The question of comparative advantage scarcely arises in the USSR because of its size and diversified resources, although similar problems arise in connection with the choice of production techniques. When the Soviet planning system was transplanted to the satellite countries, however, it ran into difficulties because of its inability to determine the advantages to be secured from trade. According to Balassa, the idea of comparative advantage did not exist in Hungarian develop- ment policy (at least until recently) although trade has a high ratio to GNP.34 Exports are determined by import needs, and the insti- tutional structure is such as to encourage exporters to meet targets for exports \vithout regard to production costs. Since prices do not re- flect resource use, it is impOSSible to determine where comparative ad- vantage lies and to what extent the trade pattern deviates from the optimum. Despite their violation of most short-term welfare considerations, the success of Soviet planning methods in producing a rapid rise in the national product makes them attractive to many underdeveloped countries. In India, for example, Mahalanobis's plan-frame for the second five-year plan drew heavily on Soviet methodology. He starts from the assumption that the rate of investment is determined by 33. See Montias (1959). 34. See BaJassa (1959, p. 264). 302 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY the level of domestic production of capital goods: "As the capacity to manufacture both heavy and light machinery and other capital goods increases, the capacity to invest (by using home-produced capital goods) would also increase steadily, and India would become more and more independent of the import of foreign machinery and capital goods."35 His analysis implies that export possibilities are so limited that they can be ignored, so that the composition of demand is lim- ited by the composition of domestic output. To raise the level of investment, Mahalanobis concludes that investment in industries producing capital goods should be increased from less than 10 percent to 30 to 35 percent of total investment in the second five-year plan. As Komiya (1959) has shown, Mahalanobis's approach to devel- opment ignores price and demand considerations completely. The targets for the four sectors in his model appear to be based mainly On the goal of creating heavy industry, which is assumed to be the key to future growth. Criteria of efficiency and comparative advantage are cntircly omitted from his analysis. Although there are traces of the Mahalanobis approach in the sec- ond and third five-year plans formulated by the Indian Planning Com- mission, the final results were much less extreme. One basic problem was that exports were expected to rise only half as fast as national in- come between tbe first and third plan periods, while demand for the goods initially imported tended to rise much more rapidly. 111e in- elastic demand for traditional Indian exports means that a consider- able proportion of investment must be devoted to commodities that are presently imported. \Vithin this category, the principles of com- parative advantage should apply. In actuality, the emphasis shifted somewhat from heavy industry in the second plan to agriculture in the third. In the latter document, increasing self-sufficiency in basic indus- trial commodities-steel, petroleum, and machinery, for example- was listed as a high-priority objective, but so was the maximum devel- opment of agriculture. 36 \m1ether the resulting targets were con- sistent with comparative advantage is not considered in the published analysis. 37 35. Mahalanobis (1955, p. 18). 36. See Government of India Planning Commission (1960). 37. On the basis of a simplified linear programming model, Sandee found that "up to 1970 more effectivc ways to employ capital for development exist than highly capital intensive steel· making," suggesting that an analysis of comparative advantage would indicate more reliance on imports. The nonmarkct benefits of production are omitted from his analysis, however. See Sandee (1959, p. 25). COMPARATIVE ADVANTAGE AND BALANCE 303 The balance-of-payments difficulties of many Latin American coun- tries have also been a major factor in shaping the programming pro- cedure developed by the Economic Commission for Latin America (UN, 1955). l11is approach has been applied in considerable detail in studies of Colombia, Argentina, and Peru.38 One basic conclusion of these studies is that the growth of exports will be much slower than the growth of demand for goods that are currently imported. Invest- ment therefore has to be heavily oriented toward import substitution, and the equality of supply and demand must be tested on a commod- ity basis to avoid balance-of-payments difficulties. In the three cases mentioned, this balancing process was carried out by means of an input-output analysis in which imported goods are distinguished from domestic products in each category. In principle, comparative advantage can be used in the ECLA pro- cedure as a basis for the choice of import substitutes, but this has apparently been done only to a limited degree. Since the main em- phasis is on balance, there is a danger that the initial assumptions as to levels of exports will not be reexamined after the extent of import sub- stitution required by a given program has been determined. l11e result may be a considerably lower productivity of investment in import sub- stitutes than in exports if the two are not systematically compared. The drawbacks to this procedure are more serious in small countries like Colombia and Peru than in a large country like India, in which imports supply a smaller fraction of the total demand for commodities. Procedures emphasizing comparative advantage Among countries having development programs, procedures that stress comparative advantage are less common than those emphasizing balance. Practically all policy statements list among their priority criteria factors presumably leading to comparative advantage, but there is little evidence concerning how they are applied in drawing up programs. 'Ibe development procedures of the government of Puerto Rico come as close to being a pure application of comparative advantage as Soviet procedures are of principles of balanced growth. Unlike many low-income countries, Puerto Rico has an elastic demand for its ex- ports to the U.S. market and can attract U.S. eapital for profitable 38. See United Nations (1957, 1960, and 1959). 304 COMPARATIVE ADVANTACE AND DEVELOPMENT POLICY investments. The government's policy has been to give tax remission for ten years and to provide overhead facilities, labor training, and other inducements to industries that will benefit the island's econ- omy. In deciding which industries to promote, the Economic Dcvel- opment Authority has studied the long-term comparative advantage of a large number of alternative projects, since comparative advantage will lead to both satisfactory profits and maximum income. Low-cost labor (even with allowance for differences in productivity) has been the main element in comparative advantage, since most industrial materials must be imported. Allowance is also made for external econ- omies in industries that will supply inputs to other sectors.39 Under this policy, the growth of per capita income was as rapid (nearly .5 percent annually) and the development of industry as marked (from 19 to 25 percent of GNP) over the years 1948 to 1958 as in any country following a deliberate policy of balanced growth. The planning procedure depends very largely on the particular rela- tion of Puerto Rico to the United States and its sman size. 111ese factors make it unnecessary to worry about the elasticity of demand for exports or the dangers of dependence on foreign sources for essen- tial imports, which so preoccupy the Indian and Latin American plan- ners. With reliable export and import markets, domestic balance is not a problem. Since the assumptions of the classical model are not approached so closely in most underdeveloped countries as in Puerto Rico, the calculation of comparative advantage usually departs farther from the market evaluation. In a more typical case the Philippine National Economic Council outlined a procedure for applying the SMP for- mula under Philippine conditions.40 This analysis starts from the market evaluation of the profitability of an investment and adds corrections for the project's effect on the balance of payments, its use of domestic materials, and its use of domestic labor, each with a suit- able weight. 'This procedurc may be justified by comparison to the linear programming criterion of social profit. In principle the proper correction to private profit is obtained by giving each a value equal to 39. The Puerto Rican experience is discussed by Baer (1959); the evaluation procedures are described in mimeographed reports of the Economic Development Authority. 40. See Philippines National Economic Council (1957). COMPARATIVE ADVANTAGE AND BALANCE 305 the difference between its shadow price and its market price. 41 In the Philippines, this would mean a bonus for labor and a penalty for for- eign exchange use (or a bonus for foreign exchange saving). Higgins showed that the weights assigned in the Philippines tend to exag- gerate these effects.42 The use of the same weight for all domestic materials may lead to serious error, since not all are overvalued by market prices. The government of Israel has developed one of the most systematic procedures for measuring comparative advantage as a basis for allo- cating investment funds and foreign exchange. In effect, the Ministry of Finance evaluates projects on the basis of accounting prices for foreign exchange and capital, taking into account the indirect use of foreign exchange in sectors supplying inputs such as power or indus- trial materials. The calculation is summed up as the cost in domestic resources of a dollar earned or saved, and it is applied equally to ex- ports and to import substitutes. 'The calculation of domestic value added is also made by exporters as a basis for export subsidies.43 In al- locating the government's development budget, priority is given to projects whose domestic cost of earning or saving foreign exchange is less than the current estimate of its accounting price. This procedure can also be rationalized by means of the linear programming criterion of social profitability. Instead of measuring the value derived per unit of investment with accounting prices for foreign exchange and labor, as in the SMP formula, the cost per unit of foreign exchange acquired is computed using an accounting price for capital. \Vhen the same shadow prices are used, all three measures give the same result. Although it is dangerous to generalize from the limited evidence on development policies that is available, there appears to be some rela- tion between the type of proeedure adopted and the characteristics of the economy in a number of the cases examined. Small countries are forced to pay more attention to comparative advantage because they 41. The social profit, II h may be expressed as: (7.4b) = II J llj + l ai l .1P., where ll} is private profit per unit of output calculated at market prices and ~, is the difference between the market price and shadow price of commodity i. The elements toP, may be regarded as weights attached to each input or output coefficien t. 42. See Higgins (1959, pp. 654-62). 43. See Bank ofIsrael (1960, p. 23). 306 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY cannot hope to produce the whole range of manufactures and primary products, while large countries may be tempted to follow more autar- chic policies.44 TIle importance given to balanced growth also depends to a large extent on the country's recent experience with its export markets and the state of its foreign exchange reserves and borrowing capacity. Puerto Rico and Isracl can both count on substantial capital inflows which make it unneccssary for them to approach balanced trade in the near future, while India has much less leeway. Conclusions ll1is chapter has considered development policy from the stand- point of economic theory, as a problem in operations research, and as it is actually carried on by governments. Much of the confusion in the field stcms from a failure to distinguish these different levels of anal- ysis. ll1eorists are prone to suggest decision rules that omit some of the relevant institutional limits, while economists who have been working in particular areas often arrive at conclusions that do not fit other cases. As in other fields of economics, most of the disagreement can be traced to implicit differences in assumptions. There are a number of contradictions between the implications of trade theory and growth theory. To make the two theories consistent, it is necessary to discard the assumption of equilibrium in factor mar- kets, to allow for changes in the quantity and quality of factors of production over time, and to take account of internal and external economies of scale. Although under these assumptions market forccs do not necessarily lead to efficient resource allocation, a pattern of production and trade can be determined that maximizes income over time. The commodities to be produced and traded cannot be deter- mined by a simple ranking procedure along the lines of classical com- parative advantage because of the interdependence among sectors. At best, it may be possible to say, for example, that a country has a comparative advantage in steel production for a specified set of pro- duction levels in supplying and using sectors. In advanced countries, this qualification may be unimportant, but in the less developed ones it is crucial in a number of industries. 44. Japan is one exeeption to this generalization, partly due to its dependence on imported raw materials. CONCLUSIONS 307 Much of the attack on the use of comparative advantage is based on its omission of various nonmarket elements. It is assumed that the inclusion of the latter favors the development of industry, and special benefits are often attributed to capital goods and heavy industry. The intangible benefits stemming from trade in the form of new products, improved technology, and technical assistance tend to be overlooked in this discussion. Although I support the crities who wish to include more of growth theory in determining the desirability of specializa- tion, I doubt that this extension \-vill favor balanced growth to the extent that they suppose. loe other main theoretical attack on comparative advantage is aimed at its supposed support for continued specialization in primary exports. Granting the low elasticity of demand for many primary products, it is wrong to conclude that comparative advantage is thereby superseded by principles of balanced growth. TIle increasing shortage of foreign exchange makes it even more important to econo- mize on its use and to seek efficient ways for increasing its supply. loe eomparison of domestic to foreign sources of supply that is implied by comparative advantage is no less relevant to this situation than to the case in which investment is more evenly divided between exports and import substitutes. The aspects of growth theory which do not seem to be reconcilable with the notion of comparative advantage are the sociological and po- litical effects of choosing one production pattern instead of another. While the concept of opportunity cost can be extended to include a number of nonmarket phenomena, such as labor training and over- head facilities, it can hardly be stretched to cover differences in fertil- ity rates or political attitudes. So far as I can see, in the present state of knowledge of social phenomena, considerations such as these may be used to modify the results of economic analysis but cannot be di- rectly incorporated into it. At the level of operations research, the search for simple decision rules for investment in low-income countries seems to have been use- ful mainly in exposing the fallacies in some of the common rules of thumb. One can specify conditions under which ratios such as the capital intensity or the effect on the balance of payments would be a valid indicator of the desirability of an investment, but the apparent gain in simplicity is offset by the danger of applying the test in inap- propriate circumstances. A more fruitful approach to partial equi- librium analysis is provided by the use of accounting prices to compute 308 COMPARATIVE ADVANTAGE AND DEVELOPMENT POLICY the social profitability of a given use of resources. This method allows simultaneously for several overvalued or undervalued inputs, and it can inc1ude whatever elements of general equilibrium analysis are available. Sinee market forces cannot be relied on to balance supply and demand under conditions of initial disequilibrium and accelerated growth, a principal concern of development policy is to ensure the consistency of production levels 'with commodity demands and factor supplies. The technique of linear programming is designed to com- bine the test of eonsistency with the test of the social profitability of a given resource use. Although it cannot be applied very extensively in developing countries as yet, the programming methodology serves as a guide to improved practical measures. To most economists, a survey of the procedures actually followed in designing development policy would probably suggest that balance is overemphasized and that the potential gains from trade are often neglected. TIlis emphasis may be partly justified by the greater un- certainties attached to trade and by an aversion to risk that is greater than seems warranted to the outside observer. Better understanding of the working of the developing economies and better information for planning are needed to redress the balance and enable countries to secure the potential gains from trade without conflict with measures for domestic development. Chapter 8 ..... -.----- ..... --- ........... __ ............... . Development Alternatives in an Open Economy: The Case of Israel with Michael Bruno CURRENT GROWTH MODELS have serious deficiencies as a basis for devel- opment policy. In focusing on the savings-investment relationship and the possibilities of substitution between capital and labor, they exclude questions of equal concern to policymakers, such as the chang- ing structure of demand, the role of foreign trade, and the allocation of resources. As a result, formal growth theory fails to clarify the rela- tions among the several instruments of development policy, which should be one of its major functions. As Tinbergen (1956) has taught us, a policy model should contain variables reflecting the economic goals of the SOciety and the main This chapter is based on an analysis of development alternatives that was pre- pared in the Bank of Israel and the Ministry of Finance and submitted to the Israeli Government in 1959. David Kochav, Zvi Sussman, and Carmella Moneta participated in the original analysis. The model has now been made more explicit, but most of the assumptions and structural estimates of the earlier analysis have been retained. We have stated the alternatives in a more precise form than was possible in practice, however, and our evaluation of them is entirely our own. 309 310 ALTERNATIVES IN AN OPEN ECONOMY instruments of government policy, and it should specify the more important structural relations connecting them. The degree of com- plexity required of a policy model for a developing economy is sug- gested by the following list of typical objectives, instruments and structural limitations. Objectives: maximum income, full employment. Instruments: capital imports, tax policy, trade policy, investment allocation. Limitations: composition of demand, balance of payments, labor supply and requirements, capital supply and requirements. The omission of sevcral of these instruments and limitations may make the use of simpler models seriously misleading. The practical significance of these additional elements may be bettcr appreciated from the study of actual cases than from general argument. \Vith this aim in mind, this chapter analyzes the main development alternatives in Israel in the next few years and shows the interrelations among the main instruments of development policy. In the course of the study we shall make use of an aggregate model that includes the variables and limitations listed above. 'TIle structural relations are characteristic of developing economics in which trade and capital imports playa significant role, and the model is therefore thougbt to be applicable to a considerable range of countries. 1 This approach also lcads to a measure of the productivity of foreign assis- tance which can provide a basis for intercountry comparisons. Limits to Growth in Israel Growth models proposed for advanced countries determine the level of income from the following elements: (a) the existing factor supply (labor, capital stock); (b) the rate of population increase; ( c) the rate of savings; and ( d) the efficiency of factor use and its change over time. 1. A similar model was used by Chenery and Goldberger (1963) in a study of development alternatives in Argentina. LIMITS TO GROWTH IN ISRAEL 311 In the analysis of less developed countries, three additional elements must be considered: (e) the inflow of foreign resources (excess of imports over exports) ; (f) the present and future composition of demand; and (g) the ability to plan and carry out development activitics (investment, technical assistance, and so forth). This second group of factors is relatively unimportant for the ad- vanced countries, where they are usually omitted.2 For the less devel- oped countries, however, they often provide the principal limits to growth. We shall indicate briefly the relative importance of these ele- ments to Israel's growth over the past decade to provide a background for the analysis of future development possibilities.3 Gross national product in Israel increased at an average annual rate of nearly 11 percent in the 1950-59 decade. Until 1952 there was mas- sive immigration, but by 1955 the increased labor force had been absorbed into fairly full employment. lbe second half of the decade was a period of steady growth and of increasingly stable prices, despite the temporary effects of the Sinai campaign in 1956. Most of our estimates of the structural characteristics of the economy will be based on this five-year period. Input and output data for the whole decade and for the second half separately are given in table 8-1. Before 1953, growth was rapid but very uneven. It was limited by the chaotic social and economic conditions accompanying massive immigration into Israel and by the society's ability to carry out devel- opment projects. There was no overall shortage of capital or labor, and it is doubtful that greater availability of external resources would have permitted a significant increase in domestic output. The situation changed between 1954 and 1957, when productive capacity increased more rapidly as a result of previous investments. During this period the composition of demand and the supply of foreign exchange seem to have been the effective limits to growth, and severe cxchange controls were maintained. Despite continuing inflation, excess capacity developed in a number of consumer goods 2. In Europe, during the recovery period after World War II, factors (e) and (f) assumed an importance comparable to that in less developed countries today. 3. Detailed analyses of growth in Israel over this period are given by Patin kin (1960) and Gaathon (1961). 312 ALTERNATIVES IN AN OPEN ECONOMY sectors, while the growth of exports, although rapid, was limited by high costs and the difficulty of penetrating new markets. Since 1958, an average growth rate of GNP of 10 percent has been maintained with less strain on the economy, prices have been sta- bilized and import restrictions have been reduced. Since full employ- ment and a substantial annual rate of increase in consumption (more than 4 percent per capita) have been achieved, a reduction in the amount of foreign borrowing is thought to be more important than an increased rate of growth. The control of inflation, measures to in- crease exports, and the allocation of public and private investment are all aimed at this result. Since the government is expected to con- tinue to finance a capital inflow of the present magnitude for some time after German reparations and other donations are reduced, one of the principal problems of future policy will be to determine the optimum level of borrowing. In designing a model to analyze future growth possibilities, several prospective changes from the past growth pattern must be recognized. Table 8-1. Israel's Economic Growth, 1950-59 (Millions of Israeli Pounds at 1958 prices) Annual rates of increase (percentage) 1959- 1950- 1955- Supply and use of resources 1950 1955 60' 59/60 59/60 Supply of resources Gross national product ],561 2,683 4,010 10.6 10.6 Imports 667 808 1,085 5.6 7.8 Total supply 2,228 3,491 5,095 9.3 10.0 Use of resources Private consumption 1,095 1,962 2,860 10.7 11.1 Government consumption 328 552 750 9.2 8.1 Gross investment 710 695 940 3.4 8.0 Replacement (50) (48) ( 100) Exports 96 282 545 21.3 18.6 Total use of resources 2,228 3,491 5,095 9.3 10.0 Financing of gross investment Gross domestic savings 138 169 400 III 25.4 Import surplus 572 526 540 Gross savings as percentage of gross investment 20 24 43 LIMITS TO GROWTH IN ISRAEL 313 Table 8-1 (continued) Annual rates of increase (percentage) 1959- 1950-1955- Supply and use of resources 1950 1955 60' 59/6059/60 Factor inputs Employed labor (thousands) 399 584 683b 6.2 4.0 Capital stock" (nondwelling) 1,462 3,139 4,920b 14.4 11.9 Capital-labor ratio" 3.66 5.38 7.20b 7.8 7.6 Productivity increased Output/employed labor 4.7 4.7 Output/ capital -2.8 -2.7 Output/ (labor plus capital) 2.1 2.2 Population and labor force' Population (thousands) 1,267 1,750 2,080 5.6 4.2 Labor force (thousands) 427 623 727 6.0 3.7 Per capita resources GNP per capita 1,233 1,533 1,930 4.8 5.2 Consumption per capita 864 1,121 1,370 4.8 4.2 Sources; Bank of Israel and Gaathon (1961). The period covered is nine and one-half years. a. Provisional figures for the fiscal year 1959/60 on which the subsequent anal- ysis is based. Later estimates differ slightly. h. For 1959. e. At 1957 prices. d. From Gaathon (1961). Output here refers to net domestic product at factor cost. e. From the Central Bureau of Statistics. They stem in large part from the declining importance of external re- sources in financing investment and imports. The import surplus has declined from 85 to 50 percent of imports and from 80 to 57 percent of gross investment over the past decade, and it will have to continue this decline in the next few years. 4 In the past growth has taken place with little change in the composition of the national product, to which agriculture contributes 11 percent, industry, mining, and con- struction 29 percent, and transport and services about 60 percent. The 4. Based on the official exchange rate of 1£1.8 to US$l.OO; the proportion would be higher if the effective rate of 2.4 were used. 314 ALTERNATIVES IN AN OPEN ECONOMY high proportion of serviees is the counterpart of the large import sur- plus, which provides a substantial fraction of the total commodity supply. Tne eomposition of demand and the limits to the domestic supply of particular commodities will become increasingly important as the economy is forced to become more self-sufficient. Similarly, the low level of domestic savings is likely to become an effective limit to future growth, which it has not been in the past. TIle Israeli experience of the past ten years reemphasizes the impor- tance to rapid growth of capital accumulation, an importancc that has tended to be obscured by the study of periods of slower growth in countries such as the United States. Gaathon's study of the relative importance of increased factor supplies and rising productivity shows that the increase in total factor productivity is about 2 percent per year, not much higher than that in the United States and other west- ern countries." TI1e high rate of growth comes mainly from the high level of capital formation, which has doubled the capital-labor ratio in the past decade. Capital formation has therefore accounted for some 60 percent of the 5 percent annual increase in per capita output in Israel, in contrast to 10 percent of a 2 percent annual increase in the United States.6 A Model of Development Alternatives T11e functions of a policy model are to determine consistent sets of economic policies and to facilitate the choice among them. The na- ture and results of development policies are represented by the vari- ables in the model. TIle model must also include in some form the principal elements affecting the rate of growth, which were discussed in the preceding section. Requirements of a policy model Tinbergen (1956) and TIleil (1958) have distinguished several types of variables in policy models: (a) predetermined or exogenous 5. See the entries for "productivity increase" in table 8·1. Gaathon (1961) fol- lows the procedure used in Abramovitz's (1956) study of the United States, in which the increase in total factor inputs is found by weighting the growth of labor and capital by their approximate shares in total income (0.67 and 0.3 3 in Israel) . On this basis, the total factor supply grew at 9 percent for thc decade, derived from a 6 percent growth of labor use and a 14 pcrcent growth of capital. 6. This finding is confirmed by Bruno's subsequent study (1968). A MODEL OF DEVELOPMENT ALTERNATIVES 315 variables; (b) instrument variables (that is, those subject to goveru- ment control); (c) objective variables (that is, those reflecting the aims of policy); and (d) other endogenous variablcs (that is, those that are irrelevant for policy analysis). In Tinbergen's approach the objectivc variables are taken as given; the problem is to find the best combination of values for the instrument variables. Theil's approach is more general. He assumes that the social welfare depends on values of both instrument and objective variables. The optimum program is that which produces a maximum of welfare consistent with the con- straints on the system. In the absence of a quantifiable welfare func- tion, however, the practical problem of setting values for the objective variables still remains. Since a single optimum program cannot be determined by eco- nomic analysis alone when there are several objective variables, our approach will be to establish a set of alternative programs that in- cludes the feasible degree of variation in all of the relevant variables. In cases where the government has established a fixed policy goal, such as full employment or a specified defense expenditure, we follow Tinbergen in setting this as a fixed objective of the program. Other- wise we follow Tbeil in considering a range of values for the objective variables. In this second category are eonsumption, the total produc- tive capacity of the economy, and the foreign debt. These together are assumed to determine the social welfare. The nature of the welfare function is considered further on pages 337-39. To determine the range of feasible programs we include two types of controlled variables. The first are instruments of government policy, as defined by Tinbergen; they are subject to more or less direct control by the government, as in the case of the exchange rate or the level of foreign borrowing. The second type may vary within limits set by institutional factors, but mayor may not be directly influenced by government policy_ loe institutional limits must be included in deter- mining realistic programs, however. These four categories are not mutually exclusive; an instrument variable can also be an objective or be subject to institutional limits, for example. Several combinations are shown in the following classi- fication of the policy variables that will be included in the model (see page 316). The institutional1imits provide a crude substitute for a more complete welfare function, since they can be used to exclude values of any variable that are clearly in conflict with welfare maxi- 316 ALTERNATIVES IN AN OPEN ECONOMY Objectives Policy Institu- instru- tional Policy variables Fixed Variable ments limits Gross national product (V) X Private consumption (C) X Public consumption (G) X Forcign capital inflow (F) X X X Unemployment rate (u) X Savings rate (s) X X Exchange rate (r) X Rate of increasc in labor productivity (l) X mization in the particular society.7 They can also be used to anow for uncertainty as to the nature of a particular structural relation, as in the determinants of increased productivity. Since many factors affect growth, some of which can only be ade- quately rcpresented in a multisectoral model, it is useful to divide the analysis into two parts. To start with, an aggregate model can be used to determine the main development alternatives. The most promising of these can then be subjected to a more detailed analysis, which is not feasible until the range of possibilities has been narrowed dowu. The detailed results can in turn be used to revise the initial estimates of the aggregate model. The main problem in designing an aggregate model for this pur- pose is to identify in advance the factors that may prove to be effec- tive limits to growth. When a particular restriction-such as the composition of demand-is omitted from a model, it implics that whatever changes take place in this element will not significantly affect the parameters in the model. In some cases it will be necessary to subject this assumption to a quantitative test in order to determine its validity. rThe model can be built up in this way by adding those relations that prove to have a significant effect. The idea of separate and conflicting limits to growth is a basic element of Harrod's (1939) pioneering work. Although he is pri- marily concerned with the cyclical aspects of differences between the 7. Instead of treating the rate of unemployment as a fixed objective, one might assign institutional limits to it. A MODEL OF DEVELOPMENT ALTERNATIVES 317 limits set by the supply of capital and the supply of labor, his rela- tions can be reinterpreted as a simple policy model of development alternatives. As is shown in equations (8.13) and (8.14), the Harrod model thus interpreted contains two equations, corresponding to the supply-demand balances for capital and labor. If all the parameters are fixed the maximum rate of growth will be determined by one of the two equations, and either labor or capital will be in excess supply.8 In a policy model, however, some of the parameters become variablcs, and these equations determine the value of any two variables, as, for example, the savings rate and growth rate at full employment. It has been argued above that there is a third general limitation to growth on a par with thc two considered by Harrod: the balance of payments. As a policy problem, the balance-of-payments limitation is quite similar to the savings-investment limitation. An increase in the rate of growth often requires a change in the structure of income use in order to reduce the proportion going to consumption and hence to increase savings. Such an increase may also require a change in the structure of production to reduce the ratio of imports to total output. It is not clear a priori, either in Israel or in other countries, which of these structural relations is morc likely to limit growth or which is harder to change. The parallelism between the two is completed by the fact that a foreign capital inflow plays a dual role in adding to both investment and foreign exchange resources. Statement ot the model The model of development alternatives proposed here incorporates these three limits. They are described by means of ten endogenous variables, including the two variable objectives (V and C). The vari- ables used in the model are as follows 9 : ENDOGENOUS (UNCONTROLLED ) VARIABLES Vt Cross national producpo Ct Private consllmption 10 8. In Harrod's terminology the labor·determined solution gives the "natural" rate of growth and the capital-determined solution the "warranted" rate. = 9. The subscript t refers to the year. In the base year t 0, and in thc final year = of the planning period t n. 10. V, C, G, F, and u are also obiective variables. 318 AL TERNATIVES IN AN OPEN ECONOl\IY It Total investment net of replaeement Rt Replacement Et Exports of goods and services Mt Imports of goods and services St Gross domestic savings Kt Total capital stock Nt Labor supply Lt Labor demand INSTRUMENT AND CONTROLLED VARIABLES Gt Government current expenditure 10 Pt Foreign capital inflow1o u Unemployment rate 10 (Nt Lt/N t ) s Marginal propensity to save (AS! !:,.V) r Effective exchange rate I Annual increase in labor productivity EXOGENOUS (PREDETERMINED) VARIABLES Initial values of all variables t Time P ei Export price in sector i 1<0 Initial unused capital stock K" Final unused capital stock In its initial form, the model consists of twelve equations, of which seven describe the structure of the economy, three specify the resource limitations, and two are definitional. The model is later reduced to four equations in eight variables by eliminating the eight irrelevant endogenous variables-all except V and C. A development program can then be specified by assigning values to four of the variables and determining the values of the remaining four from the model. Since it is assumed that decisions on development strategy can be based on the values of the policy variables at the end of a planning period of n years, solutions are only needed for a single period. The equations are presented first in general form; estimates of the param- eters are then given for the Israeli economy for a five-year planning period of 1959/60 to 1964/65. The aggregate model will be supple- mented by an interindustry analysis in order to take account of the composition of demand and of supply limitations in estimating the parameters. A MODEL OF DEVELOPMENT ALTERNATIVES 319 THE AGGREGATE PRODUCTION FUNCTION. Although the Harrod.. Domar model has been criticized for its omission of substitution be- tween labor and capital, it is generally recognized that substitution can only take place over a period of time and depends to a large extent on the installation of new equipment. \Ve shall therefore treat the labor-capital ratio as a function of time, but shall assume that both inputs are required in fixed proportions at any moment. For simplicity, a trend will be associated only with the labor input, which is con- sistent with the experience of the past decade. \Vith complementary inputs, output is limited by whichever one is exhausted first. In Israel, as in most of the less developed economies, this factor is more likely to be capital. We therefore write the pro- duction function as dependent on the stock of capital and the effec- tiveness of its use: where f3 represents the average product for each unit of increase in the capital stock. The second term in this equation allows for the possi- bility of reducing the level of excess capacity, K, which is a significant factor in Israel and elsewhere. More important than the direct substitution between labor and capital is the effect of a change in the composition of output on capi- tal requirements. This will be allowed for by estimating f3 from an interindustry analysis so that it becomes a weighted average of the output-capital coefficients in each sector. Any substantial departure from the assumed composition of the increase in output will therefore require a recalculation of f3.H LABOR DEMAND. The demand for labor depends on the level of out- put and the increase in average labor productivity: (8.2) where "\0 is the average labor input for each unit of output at the be- ginning of the period and 1 is the annual rate at which it decreases. The estimate of 1 should also take account of the anticipated compo- 11. The coefficient Ii similarly depends on the sectors in which excess capacity is reduced. 320 ALTERNATIVES IN AN OPEN ECONOMY sition of output. As estimated statistically from past trends, 1 includes effects of both substitution and technological change. Since we assume complementarity between capital and labor, equa- tion (8.2) may be restated as a production function by solving for V t if labor should become the factor limiting growth. IMPORT DEMAND. Tne demand for imports depends on the five com- ponents of total demand: (8.3) Mt p./Ct + p.g"G + pl(It + R + p./E t t) t· Each import coefficient p'r is derived from a solution to an input- output model containing specified proportions between domestic supplies and imports in each sector. The coefficient p'r therefore rep- resents the total imports required directly and indirectly for each unit of each type of demand. The future import proportions are derived from the anticipated effective exchange rate, r; use of a higher ex- change rate will result in additional import substitution and a fall in the import coefficients.12 REPLACEMENT. The replacement of capital depends on the age and estimated life of the capital stock in different sectors of the economy. In aggregate form, this relation may be indicated by: (8.4 ) although in practice R t is estimated on a sector basis. As Damar (1957) has emphasized, the replacement of worn-out equipment is substantially less than the conventional allo'wance for depreciation in a growing economy. We therefore define net invest- ment It as gross investment less replacement. A part of net invest- ment is covered by the excess of depreciation allowances over actual replacement. SAVINGS. Gross domestic savings depends on the level of per capita income, its functional distribution, and the government's tax policy. In the absence of an adcquatc basis for estimating the separate effects 12. Import substitution as a result of a higher exchange rate may also produce a lower productivity of capital, but this can only be taken account of explicitly in more detailed models. A MODEL OF DEVELOPMENT ALTERNATIVES 321 of these factors, an aggregate relation of the following form will be assumed: (8.5 ) S" So + s(V" - Yo), where the marginal propensity to save out of increased GNP (s) is taken as an instrument variable. It represents the combined effects of tax policy, changes in income distribution, and other policy mea- sures that affect savings. LABOR SUPPLY. The supply of labor is determined from the natural increase of the existing population plus net immigration. For sim- plicity, the combined result is assumed to take an exponential form: (8.6) EXPORTS. The level of total exports is the sum of the individual commodities and services exported; each is assumed to depend on the effective exchange rate and on foreign prices: (8.7) 'The time period is also assumed to affect achievable export levels be- cause of the time needed to penetrate new markets and to establish export organizations for new products. SAVINGS-IlI.'VESTMENT EQUILIBRIUM. (8.8) BALANCE-OE'-PA YMENTS EQUILIBRIUM. (8.9) where 1\111 E t , and F t are all measured at constant domestic prices,13 13. The foreign capital inflow is normally given in current-value dollar terms. If we denote the latter by F'. we have F. rF,' IP" where r is the exchange rate and Pt is an implicit price index defined by the expression p, = (pmM. p.E.j/(M,- E,), pm and p. are the respective import and export prices (indexes) on foreign markets. In the Israeli case import prices are assumed constant and export prices are assumed to fall, Hence P, > 1 and F t < rF t '. 322 ALTERNATIVES IN AN OPEN ECONOMY EMPLOYMENT EQUILIBRIUM. (8.10 ) where the proportion unemployed (u) is an objective variable. TOTAL NET CAPITAL FORMATION. t=,,-l (8.10 II t = (K" Ko). t=o To express the model in terms of initial and final-year values only, an approximation of the form, (8.11a) may be substituted for equation (8.1 1 ); p depends on the rate of growth of investment and the length of the planning period. Some assumption of this sort is needed to make In determinate. GROSS NATIONAL PRODUCT. (8.12 ) REDUCED FORM OF THE :MODEL. The reduced form of a policy model is a set of equations involving the policy variables only, with all the irrelevant endogenous variables eliminated.14 In the present case we eliminate eight variables and eight equations, so that the model is reduced to the following four equations in eight variables (the terms in square brackets are constants) : [.IYo(l+ )nj(1 u) y (8.1 3) v" Ao (1--l)" (8.14 ) v _ [p//3Vo + So - R,,] sVo + Fn n- p//3-s (8.15) V (1- P-e}E" + (1- P-c}F" + (Pc pg)Gn n - pc + (pi pc)p/ /3 + [(pi Pc)(p//3Vo - Rn)] - ftc + (pi ftc)p//3 14. The concept is discussed by Theil (1958, chapter 7), under somewhat different assumptions. A :MODEL OF DEVELOPMENT ALTERNATIVES 323 ( 8.16) C G= (1 s)V,,+ (s-so)V o where So = So Vo and Vo = p(Ro - K,,) + Vo Equations (8.13), (8.14), and (8.15) correspond to the three equi- librium conditions for labor, capital, and foreign exchange. When no limit is placed On F the labor-force equation (8.13) provides the ulti- mate limit to growth because the other two equations can be satisfied by increased foreign borrowing. With a given F, there are three sep- arate limits. Equation (8.14) corresponds to the Harrod-Domar equation, as can be seen by assuming a one-year period, no excess capacity, and equal marginal and average (net) savings rates. The equation then beeomes: (8.l4a) ,,- 6.V F + (3 V or j3s (8.l4b) Vo + 13Ft> V" 1-{3S TIle meaning of the balance-of-payments limit can be clarified by assuming that all of the import coefficients Pi are equal and again tak- ing a one-year period. TIle result is: (8.15a) Estimates for Israel in 1964/65 The following are our estimates of the parameters and predeter- mined variables in the above equations for the planning period 1959/ 60 to 1964/65. 15 P(Ko Kn) = 210 Vo = 4,010 No 727 f3 = 0.364 p 1::,/10 + It + 12 + Is + I. 0.221 y = 0.034 15. The principal sources of data for these estimates are Bank of Israel (1960). Bruno (1964), Chenery (1955), and Patinkin (1960). 324 ALTERNATIVES IN AN OPEN ECONOMY Ao = 0.164 JLo = = 0.15, JLo 0.25, /Li 0.375, JL. 0.40, for r 2.5 /Lc = 0.14, JLo 0.24, JLi = 0.31, JL. = 0.39, for r = 3.0 JLo 0.13, JLo = 0.22, JL. 0.30, JL. = 0.38, for r = 3.5 So 0.10 The estimates of f3 and JL are taken from solutions to the input- output model described in Bruno (1964), assuming a composition for the increase in demand that reflects the expected rise in income level. Tbe degree of import substitution was estimated under three different assumptions as to the effective exchange rate (that is, the rate used for planning purposes) as explained below. Estimates of R,.2 and AO are based on Gaathon's (1961) detailed analysis of labor and capital in- puts by sector 16 By using these values in equations (8.1) to (8.7) we obtain the fol- lowing structural equations of the model for Israel 17 : (8.1' ) V" 4,010 + 210 + 0.364 (K" - Ko), (8.1a' ) V" 4,220 + 1.6451" ( 8.2') L" = 0.172(1 l)5V.. (8.3' ) M" = O.l3C" + 0.22G" + 0.30(1" + R,,) + 0.38E" for r 3.5 (8.4') R" 210 (8.5' ) S.. = + 400 s(V" 4,010) ( 8.6') N" = 727(1.034)° 860 (8.7' ) En 1,000 for r 2.5 E" = 1,150 for r = 3.0 En = 1,400 for r 3.5 The four equations in the reduced form of the model are: FULL-EMPLOYMENT EQUILIBRIUM (8.13') V" 4,990(1 u)/(1-l)5 SAVINGS-INVESTMENT EQUILIBRIUM (8.14') V" (2,760+F,,-4,OlOs)/(0.608 s) 16. Rn is determined from investments made before the beginning of the period. 17. Values are in millions of Israeli pounds at 1958 prices. Labor is in thousands of workers. THE RANGE OF POLICY CHOICE 325 BALANCE-OF-PAYMENTS EQUILIBRIUM (8.15') Vn 3.73Fn - 0.38G" + 5,440 for r 3.5, En 1,400 TOTAL CONSUMPTION (8.16') en Gn - (1- S)Vn + (8 - 0.10) 4,010 The Range of Policy Choice The model is now in a convenient form to show the range of feasible combinations of instrument variables that can be considered by a policymaker. \Ve define a feasible program as a set of values for the policy variables for which: (a) equations (8.13), (8.14), (8.15), and (8.16) are satisfied; and (b) no controlled variable falls outside a predetermined range. In deciding on the range of values to consider for the controlled variables, we have taken account of the following factors: (a) the like- lihood of political support for any large change from present condi- tions, such as a change in income distribution or a reduction in the rate of increase in consumption; (b) some of the economic implica- tions not included in the aggregate model, such as the increased pro- duction in individual industries and the specific skills and resources required; and (c) the administrative feasibility of various measures, such as limits to consumption and stimulation of exports. \Ve have generally assumed three values for the controlled vari- ables: (a) a minimum, representing a pessimistic assessment of future possibilities; (b) an intermediate value, usually based on past trends or a specific forecast; (c) a maximum or most optimistic value, be- yond which thc likelihood of further increase is too small to be con- sidered for planning purposes. As indicated earlier, the meaning of these limits varies with the nature of the controlled variable. Govern- ment expenditure and the unemployment level are taken as fixed objectives, so only one value is estimated. The savings rate and pro- ductivity increase are determined by social and institutional as well as economic factors; the range here reflects uncertainty as to the ability to achieve structural changes as well as the effect of economic changes that are taking place. The exchange rate and foreign borrowing, on the other hand, are pure instrument variables whose limits are set by welfare considerations. 326 ALTERNATIVES IN AN OPEN ECONOMY TIle limits assumed (for controlled variables) are as fo11ows 18 : Effective exchange rate (Israeli Growth in pounds Foreign Marginal labor GOl'CTIlmcnt Assump- to US. capital savings productivit), Unemplo),- current tion dollars) inflow rate (percentage) ment level expenditure a 2.5 240 0.165 3 0.05 1,010 b 3.0 330 0.25 4 0.05 1,010 c 3.5 480 0.30 5 0.05 1,010 lbese estimates are largely a matter of judgment, and we shall give nO detailed justification of them. A few of the relevant factors are noted below. TIle exchange rate (T) is used in the sector analysis as a basis for estimating the extent of import substitution and exports; it thus ap- pears only indirectly in the aggregate model. The minimum value of 2.5 pounds per dollar corresponds to the effective rate in 1959/60. 'D1e maximum rate of 3.5 was established on the basis of the exchange rate at which it would be profitable to export or substitute for imports in principal sectors. At higher exchange rates such a large increase in output would be required in some sectors that other factors, such as the ability of exporters to expand into new markets, would act to ef- fectively limit further inereases in exports and import substitution. 11l TIle maximum export forecast takes these factors into account. In presenting our results, we shall use only the highest value of the ex- change rate (r = 3.5) in determining import substitution, since it proved to be the optimum. The range of exports will be retained, how- ever, since it depends on other faetors beside the exchange rate. The limits to the level of foreign borrowing are based on estimates 18. In the three assumptions, exports corresponding to the different effective exchange rates are En 1,000, E. = 1,150, and E< =1,400. Foreign capital in· flow in dollars is $150 million under assumption a, $200 million under assumption b, and $285 million under assumption c (using the official rate of 1£ 1.8 = US$l.OO) . 19. As shown in Chenery (1955), the optimum exchange rate for planning purposes can be formally determined from a programming model in which de- mand elasticities for exports as well as capital and labor inpnts are specified. Although it was not possible to carry out this calculation in detail, the rationale of this procedure was followed here. THE RANGE OF POLICY CHOICE 327 of the country's ability to secure credit. The desirability of borrowing the maximum amount is considered on pages 337-38. TIle savings rate is derived from the fraction of the increase in GNP that has gone into government revenue and private savings in the past five years (31 percent). 111e minimum future level (Sa = 0.165) al- lows for the expected level of government expenditure, but keeps the total constant. TIle maximum limit (so 0.30) is based on the expe- rience of other countries, but it represents an extremely optimistic assumption for Israel, which has a relatively equal distribution of income. Labor productivity is included as a controlled variable because it is necessary to consider several possibilities as to its future rate of in- crease. Our intermediate assumption is that the increase will be 4 percent a year, which is somewhat less than in the past. A maximum of 5 percent and a minimum of 3 percent are assumed. A growth in labor productivity of less than 3 percent would not be consistent with the required reduction in the balance of payments deficit. Full employment is taken as a fixed objective, so only one value of u is estimated. The effective full-employment level is taken to be 5 per- cent unemployed (approximately the 1959 level) to anow for the composition of the Israeli labor force. Higher levels of unemployment will be considered in several trial solutions, however. Government current expenditure is also a fixed objective, based on the need to maintain the defense establishment and an optimum level of welfare expenditures. "\Ve now wish to find the set of development programs that satisfy the equations of the reduced model and fall within the predetermined limits for the controlled variables. 20 The analysis can be shown most simply in two-dimensional geometry, taking V and one controlled variable as axes and plotting the curves that result from setting each of the other controlled variables at its maximum and minimum value. 21 Three such graphs are shown in figures 8-1 through 8-3, in 20. All programs satisfy equation (8.16), which will only be used in the section headed "The Choice of Policy" in the analysis of welfare. 21. It is desirable to have V as one of the axes, since it is the main determinant of social welfare, and the welfare effect of varying eaeh restriction is thus shown direetly. Sand E are used as axes instead of the instrument variables sand r which determine them, because the boundaries are linear in the former variables. The fourth possible graph, that of l against V, has been omitted because the assumption of a given level of unemployment rules out the possibility of independent variation. Figure 8-1. Limits on S-V Axes l. l~ 1,000 800 W (;1 N t::i 00 .;; 600 '" CJ) 400 200 1 I F. 0 Gross national product Figure 8-2, Limits on E-V Axes l. I. Ie F. 1,400l-- Ec w .... V> .... N -.0 ! 1,20°1 Eb Fo 1,0001- E. 800 Gross national Figure 8-3. Limits on F-V Axes 1. I t.. 600 500 w 400 w <::> ::: s,. .!2lJ '" 300 & 200 100 , .- - I . 10,/( ...... ~- Sa Gross national product THE RANGE OF POLICY CHOICE 331 which the system is plotted on the S-V, E-V, and F-V axes. Each graph shows four sets of boundaries corresponding to the productivity limits (1), the savings limits (s), the export limits (E), and the bor- rowing limits (F). The intermediate values of Eo, Fa, and 10 are also shown. l11e set of all feasible programs can now be determined quite read- ily. Starting from any point inside all of the boundaries, such as point 7, we proceed in any direction until we hit a maximum or a minimum boundary and then move along this boundary until we reach another, which again causes a change in direction. Starting from point 1 in each figure, we can trace out the limits 1-2-3-4-5-~1 in this way. They outline the shaded feasible area in each figure. The algebraic counterpart of this procedure is to find a point of in- tersection such as 1, at which two variables (s and F) are at one of their limits and the other two (1 and E) are within theirs. We can then solve in turn for the other intersections and among them pick out those that bound the feasible area. 22 The solutions for the six vertices are given by points I through 6 in table 8-2. TIle economie implications of the various restrictions can be shown by starting from the improbable assumption (for Israel) that there will be no inflow of foreign capital five years hence; exports are taken at their middle value of 1,150 and I at its lower value of 0.03. The corresponding solution is shown as point 10 in table 8-2 and in each figure. At this point the balance of payments is the limiting factor and the savings limit is redundant. The shortage of imports would anow GNP to rise only 9 percent above its 1959 level, which is well below the 18 percent increase that would be permitted by the availability of sav- ings under the most pessimistic assumption, shown at point 9. If we now allow foreign capital to increase, keeping the same ex- port level, we can move from point 10 along En in each figure until GNP reaches 5,290 at point 8, which is the intersection with the savings boundary Sa. At this point the fact that investment requirements have been rising more rapidly than import needs makes the minimum sav- ings level the more restrictive factor. For further increases in V, the 22. Instead of this trial-and-error procedure, the problem can be put in mathe- matical programming form by specifying an objective function. The technique of parametric programming would then enable us to trace out the feasible area by varying the constraints on the system. In the present case the graphical approach is quite adequate, however. Table 8-2. Selected Solutions to the Model Sav- Supply Resource use ings Controlled variables Point V+F V M G C l+R E S s F r u Base year 5,095 4,010 1,085 750 2,860 940 545 ,.uu ytU a w Solution 1 6,290 5,810 1,490 1,010 4,100 1,180 ],010 700 0.165 480 c 0.04 0.05 w N 2 6,610 6,130 1,610 1,010 4,230 1,370 1,130 890 0.231 480 c 0.05 0.05 3 6,460 6,130 1,670 1,010 4,080 1,370 1,340 1,040 0.300 330 c 0.05 4 6,050 5,810 1,590 1,010 3,860 1,180 1,350 940 0.297 240 c 0.04 0.05 5 5,760 5,520 1,480 1,010 3,750 1,000 1,240 760 0.238 240 c 0.03 0.05 6 5,870 5,520 1,440 1,010 3,860 1,000 1,090 650 0.165 350 c 0.03 7 6,140 5,810 1,550 ],010 3,950 1,180 1,220 850 0.248 330 c 0.04 0.05 8 5,540 5,290 1,400 1,010 3,670 860 1,150 610 0.165 250 c 0.03 0.095 9 4,730 4,730 1,290 1,010 3,200 520 1,290 520 0.165 0 c 0.03 0.19 10 4,360 4,360 ] ,150 1,010 3,060 290 1,150 290 0.420 0 c 0.03 0.25 THE RANGE OF POLICY CHOICE 333 main function of external resources is to provide capital rather than foreign exchange. As shown in figure 8-3, the increase needed in F for each unit of increase in V is higher as we move along Sa than along E b · At point 1 we reach a level of GNP of 5,810, which is the maximum that can be achieved without an increase in the minimum savings rate or an excess of foreign capital over its assumed limit. At point 1 the increase in labor productivity is at the intermediate value of 0.04. The segments 1-2-3 of the boundary in each figure show the effect of raising savings from its minimum level to its maximum level. At point 2 it is not possible to raise GNP any further because the maxi- mum increase in labor productivity is reached. Further savings reduce the amount of foreign borrowing required for the same level of GNP and hence increase the export requirements at thc same time. This process is stopped by the maximum savings boundary So; figurc 8-3 shows that the maximum export level Eo would permit only a small further reduction in F. It will be shown on pages 337-39 that under different assumptions as to the properties of the social welfare fune- tion the optimum program lies between point 2 and point 3 for the optimistic productivity assumption (or between 1 and 4 for the inter- mediate assumption). The segments 4-5-6-1 of the boundary have practical significance only if the attainable productivity increase turns out to be less than the intermediate assumption of 4 percent. Point 5 shows the mini- mum increase in income that must be achieved to maintain full employment at the lowest rate of productivity increase considered realistic. The segment 5-6 gives other possible combinations of 5, F, and E that would yield this income and employment level. Figure 8-3 reveals very clearly the dual role that foreign capital plays in supplying both savings and foreign exchange. Assume, as we did initially, that Sa and Eo are the maximum values for the savings rate and exports. When there is no capital inflow growth was shown to be more severely limited by the shortage of foreign exchange than by the potential level of savings.23 Since import requirements increase less rapidly than investment requirements as income increases,24 the two limits become equal1y restrictive at the intersection of Sa and EI> 23. The difference would be even greater if thc lower valuc of import substitu· tion implied by Eb and Tb = 3.0 were used. 24. The slope of the savings-investment curve is dF/dV = 0.44; the slope of the import-export curve is dF / dV 0.27. 334 ALTERNATIVES IN AN OPEN ECONOMY at point 8. At higher growth rates the shortage of savings is more restrictive than the supply of foreign exchange at the present savings rate of Sa. If the effective maximum for exports were given by E" (which represents nearly a doubling of E in five years), foreign ex- change would remain the more restrictive factor until a growth rate of nearly 8 percent is reached. In Israel in 1960, therefore, the balance of payments seemed as likely to dctermine the rate of growth as the savings-investment limit. Bottlenecks and Sector Limitations An aggregate model contains an optimistic bias in its presentation of development alternatives. Although the magnitude of the shift in resources required to satisfy the balance-of-payments limitation may seem to be relatively small, the total may conceal obstacles to increas- ing particular kinds of output. For example, if instead of saying that under solution 7 above, 38 percent of the increase in output must take place in the export sectors, we say that exports of manufactured goods must triple in five years, the specific nature of the resource shift becomes clearer. The possibilities for achicving such an increase can be evaluated realistically only in terms of specific commodities, how- ever. Similar comments apply to possible changes in labor and capital productivity, which must also be judged on a sector basis. Some fornl of interindustry analysis is required to determine the sector effects of the development programs that have becn outlined above in aggregate terms. We employ an input-output model of the type used in studies of Italy, the Nctherlands, Argentina, and other countries for this purpose.25 \Ve have, however, gone one step further in using the results of the interindustry analysis to determine alterna- tive parameters for the aggregate model, as in the three estimates of the import equation for different values of the exchange rate. Apart from modifying the parameters in the aggregate model, sector calculations are needed as a basis for judging other and more specific factors that may limit growth. Organizational ability, for example, can hardly be discussed in general terms, but it is possible to estimate limits to the rates at which new land can be brought into cultivation, 25. This use of input-output analysis is discussed in Chenery and Clark (1959, chapter 11). BOTTLENECKS AND SECTOR Ll~IITATIONS 335 new types of crops can be grown, or export organizations can be extended to new products and countries. 'These factors can be in- troduccd into a formal model only as limits of varying degrees of probability to the carrying out of particular types of development activity. In the case of Israel, these limits were evaluated initially on the ba- sis of a twenty-sector input-output model. The use of this model was discussed by Bnmo. 26 To illustrate the nature of the results, we pre- sent in table 8-3 part of a detailed comparison of two of the most desirable programs. Program A is based on forecasts of import sub- stitution and exports using the intermediate assumption of 3.0 as to the future value of foreign exchange. The requirement for foreign capital determined from the solutions corresponds approximately to our upper limit. Program B was derived by raising the effective ex- change rate sufficiently to reduce the foreign capital inflow to the intermediate level ($200 million) by means of additional import sub- stitution and exports; it corresponds to point 7 in figures 8-1 and 8·2. The two programs assume a 4 percent annual increase in labor pro- ductivity and the same GNP. Both programs require a substantial shift of resources from agricul- ture and services to mining and manufacturing. 111is change in the composition of output is necessitated by limited domestic and foreign demand for agricultural products, the required reduction in the im- port surplus and the consequent need to expand the production of minerals and manufactured goods, particularly those that substitute for imports or can bc exported. 111e difference in sector growth rates between the two programs is not very large, but it becomes increasingly significant the more finely we disaggregate industries. 111e same is true of exports of manufac- tured goods, in which the growth rate of the previous five years must be doubled to make up for limited possibilities in other sectors. 111e sector composition of output also affects the evaluation of the growth in productivity that is likely to be achieved. Both in agricul- ture and industry, productivity growth has been low for several years after the establishment of new types of production. A larger amount of import substitution, involving a greater proportion of new ventures, 26. As indicated in Bruno (1964), a more detailed interindustry model has been constructed, but this model was not available at the time the present study was carried ou t. 336 ALTERNATIVES IN AN OPEN ECONOMY Table 8-3. Comparison of Alternative Development Programs 1964/65 Alternative Alternative 1959/60 A B Controlled variables Exchange rate (r) 2.5 3.0 3.5 Exports (E) 545 1,130 1,220 Marginal savings rate (s) 0.14 0.15 0.25 Foreign capital inflow (F) 540 490 330 Increase in labor productivity (l) 0.047 0.04 0.04 Annual rate of increase in production" Agriculture 13 7.2 7.0 Mining } 18.3 20.3 Man ufacturing 11 12.1 13.0 (a) Chemicals (including petroleum and rubber) (15.5 ) (18.3 ) (b) Base metals (17.5 ) (21.8 ) (c) Machinery ( 14.0) (14.5) Construction 6 2.8 2.8 Power 10 10.7 10.4 Transportation 9 9.9 9.9 Services 10 7.5 7.3 Total GNP 10 8.5 8.5 Annual increase in exports' 18.6 16.3 17.8 Agriculture (183) (9.2) (9.2 ) Mining and manufacturing (13.0) (243) (263 ) Services and transport (21.8 ) (11.0 ) (11.6 ) a. Growth rates for 1959/60 pertain to 1955-58; for 1964/65 to 1959-64. is therefore likely to lower the productivity increases achievable in the near future and to raise the capital requirements per unit of output. The difficulties in increasing exports and adjusting to the changed composition of domestic demand were a major factor in our conclu- sion that program B represents something like the maximum increase in GNP achievable with a capital inflow of $200 million. With a higher capital inflow, and consequently with less serious changes in the com- position of output, it is more likely that something close to our upper productivity limit could be accomplished. THE. CHOICE OF POLICY 337 The Choice of Policy As suggested by 111eil,21 formal welfare analysis may throw some light on the optimum choice of policies, although the possibilities for determining the properties of the social-welfare function are very lim- ited. In the case of development policy, welfare analysis may, for example, be of some help in the determination of the optimum level of foreign borrowing. We assume the social welfare to be a function of three variables: ( 8.16) W W[(C+G),F,K], in which for simplicity we do not distinguish between public and pri- vate consumption. Instead of the capital stock we can substitute GNP, since with fixed capital coefficients, one implies the other. The productive possibilities of the economy are summarized in equations (8.13), (8.14), and (8.15). Assuming values for the remain- ing controlled variables (Sa = 0.165, Eb 1,150, 1 = 0.04), we can determine total consumption as a function of foreign borrowing. The production possibility curve P-P in figure 8-4 is constructed from equation (8.16) and the data in table 8-2 on these assumptions. The shape of the curve derives from the fact that increased foreign re- sources have a high productivity at low levels of F because up to point 8 growth is limited by the import restriction. Between point 8 and point 1 the increase in F permits a further increase in investment and GNP, but beyond point 1 GNP is constant because full employment of labor has been reached and the increase in F goes only to increase con- sumption. The three segments of the P-P curve therefore correspond to the three equations of the model, each of which is binding over a given range. To find the optimum program, we need to specify some properties of the function W. Taking consumption over time as the measure of welfare,28 we can (in principle) measure the welfare loss due to an increase in foreign borrowing as the sum of interest costs, the effects of reduced future borrowing power and the loss in future consump- 27. See Theil (1958, chapter 7). 28. The omission of the capital stock understates the advantages of higher pro- duction, but beyond point 1 there is no change in V or K. 338 ALTERNATIVES IN AN OPEN ECONOMY Figure 8-4. Welfare Analysis 5,400 /W / p / LA 5,200 ~\ \0 Labor limit + \0 /' ./) - '" .~ .: 5,000 \V- /' 6 'C . alnta Ii" . Imlt .2 4,800 E "''"" '" '" 0 u 4,600 'E 0 E- 'Balance of payments limit 4,400 10 Increase in foreign debt (Total increase = 2.5 Fn.) tion due to the need to develop an export surplus. 1l1ese elements suggest an increase in the unit cost of borrowing to the economy as F increases. A constant welfare curve W - W having these properties is indicated in figure 8-4. In the case of Israel it is almost certain that the optimum program falls somewhere on the segment 1-8, and prob- ably in the full employment range 1-6. Levels of borrowing above point 1 would produce less rise in consumption, while below point 8 the productivity of borrowing is very high, and certainly greater than its social cost. Even if this analysis could be carried out more accurately, a num- ber of qualifications would have to be added to allow for other vari- ables. A high import surplus makes it unnecessary to raise savings and THE PRODUCTIVITY OF FOREIGN ASSISTANCE 339 exports in the short run, and may inhibit the structural changes re- quired in the longer run. Since so little is known about the determi- nants of saving, it might be more prudent to finance as much invest- ment as possible from domestic sources to bring about the needed rise in s, even though the rate of increase in consumption would be re- duced somewhat as a result. A similar argument can be applied to the development of exports, where it takes a considerable time to establish markets for new products. The maximization of consumption over a limited period of time is therefore not an adequate goal of policy, even as a first approximation. The Productivity of Foreign Assistance In our study of development alternatives we have also provided a measurement of the value of foreign assistance, which is of consider- able interest in itself. The model shows the increase in GNP and in total consumption that results from varying quantities of external resources. The consumption frontier P-P in figure 8-4 can equally well be thought of as a measure of the productivity of foreign aid. The corresponding path in figure 8-3, showing the increase in GNP achiev- able with varying amounts of F, is an alternative measure. To explore this aspect of the model, we define the marginal produc- tivity of foreign aid as the increase in GNP in the year n (elY,,) that is achievable from a unit increase in external resources distributed over the planning period. The increase in aid is thus defined as: dA !: = ~ dF t . For marginal increases in investment and aid in period n, with no change in period 0, the relation p dln/!dl t in equation (8.l1a) becomes p 2/n. The same relation holds for dFn/dA = p = 2jn. Differentiating equations (8.13), (8.14), and (8.15) with respect to A and substituting these values gives the fonowing measures of the productivity of aid when each of the boundary conditions applies. Full-employment boundary Since foreign borrowing does not affect equation (8.1 3), the deriva- tive of V with respect to A is zero for increases in aid beyond point I in figures 8-3 and 8-4. All of the increase therefore goes into consump- tion. This point corresponds to the concept of the "capacity to ab- 340 ALTERNATIVES IN AN OPEN ECONOMY sorb" that has been suggested by various writers as an upper limit to foreign assistance. Savings-investment boundary From equation (8.14), we derive the following equation for the productivity of aid: dV" {3 ) dF" {3 (8.17 ) dA ( P - {3s dA {3s 1---- P If there is a zero propensity to save (s) the marginal productivity of aid is equal to the marginal productivity of investment, {3. As s in- creases, the effect of a unit of aid on GNP rises because the increase in domestic savings permits the increment in investment to exceed the amount of aid. For the maximum estimate of the marginal propensity to save in Israel, s 0.30, the productivity of aid reaches 0.50 or 38 percent above its productivity with no saving. Equation (8.17) thus gives a precise expression to the effect of savings out of increased in- come, whose importance to accelerated growth has been stressed by Rosenstein-Rodan (1961a) and other writers. The ioreign-exchange boundary Countries for which foreign exchange is the binding 1imit to growth will normally have the highest productivity of external aid. The equa- tion for the productivity of aid that is derived by differentiating equa- tion (8.15) is: (8.18 ) dV" (1 Pe)P dA - }-to + (f./-i - }-tc)p! {3 In the case of Israel the propensities to import given on page 324 produce values of the productivity of aid that range from 0.85 to 1.10 as the exchange rate is increased from 2.5 to 3.5. A comparison of equations (8.17) and (8.18) shows that with plausible values of the parameters the productivity of aid will range from 0.2 to 0.6 or so when domestic savings are the limit to growth and from 0.4 to over l.0 when the balance of payments is the limiting factor. Low productivity is associated with high capital-output ratios, low savings and high import propensities. It is only in cases where marginal imports are a very large fraction of the increase in GNP (0.40 THE PRODUCTIVITY OF FOREIGN ASSISTANCE 341 or more) that the productivity of aid along the foreign-exchange boundary is likely to drop below that along the savings boundary. The distinction between these two cases ean be very important in assessing the effect of foreign assistance. It accounts for the high pro- ductivity of aid under the European Recovery Program, where foreign exchange was clearly the faetor limiting expansion in most countries. A similar situation may oecur in less developed countries, such as India, Israel, or Argentina, in which for various reasons foreign ex- change has become more of a bottleneck than savings. Measurement of the productivity of aid is, of course, only one part of a rational allocation criterion; the other is the donor's evaluation of the relative desirability of an increase in output as between coun- tries. Although economics cannot contribute much to the latter question, an improvement in the measurement of productivity by con- sidering alternative development programs seems a realistic possibility. Chapter 9 Optimal Patterns of Growth and Aid: The Case of Pakistan with Arthur MacEwan ONE OF THE PRINCIPAL MEANS for poor countries to accelerate their development is by using external resources to supply additional im- ports and to finance a higher level of investment. Although this policy offers substantial benefits, it also requires that the structure of the economy be adapted to accommodate the expected resource inflow over a substantial period of time. For this reason, the extent of re- liance on external capital-public and private-becomes one of the critical elements of development strategy. TIIere has been relatively little theoretical study of the benefits and costs of using a controlled inflow of resources to promote develop- ment. Formal growth models typically either ignore this variable or take it as fixed. In the formulation of development programs by plan- ning organizations, the projected inflow of aid and private capital is determined largely on a historical and political basis rather than through a systematic evaluation of alternatives. This is true in Paki- stan as well as in most countries receiving foreign assistance. l11is chapter explores the properties of optimal growth strategies in which the total amount and time pattern of the resource inflow can be varied within limits.l TIle problem is studied both from the point L The linear programming model used in this study was suggested by Robert Dorfman, who took the lead in its initial formulation. It is derived from the projection model used ill chapters 8 and 10. \Ve are indebted to Robert Dorfman, Alan Strout, and Joel Bergsman for helpful comments. 342 THE ANALYTICAL FRAMEWORK 343 of view of the borrowing country trying to make the best use of its domestic and foreign resources and from that of the lender trying to assess the productivity of additional amounts of public assistance in different recipient countries. 111ese different viewpoints are reflected in alternative forms of the objective function that is maximized to d~termine the optimal policy. While the formulation of the problem is designed to bring out its general features, the resulting programming model is applied to the planning situation described in the Pakistan Perspective Plan for 1965-85. Apart from the variables affecting the inflow of external capital-which is taken as given in the Pakistan plan-we have taken most of our other assumptions from the plan to isolate the effects of variation in external resources. The Analytical Framework The problem of optimal growth paths over time has been studied only under assumptions that are rather far from those describing the policy choices facing underdeveloped countries. Tl1e main weaknesses of existing growth models are that they (a) assume closed economies, (b) focus mainly on the allocation of resources between investment goods and consumer goods production, (e) ignore some of the cen- tral constraints on policy, and (d) study long-run equilibrium condi- tions rather than developments over a relevant planning period.2 There is therefore little carryover to the problem at hand of the spccific results so far achieved by formal analysis. The principal con- tribution of growth models is to show the importance of formulating an explicit welfare function and of relating alternative strategies to both the parameters in this function and the restrictions placed on the system. GrowtlI models and planning models Tbe analytical framework used here is largely derived from detailed empirical models of open economies that are dependent on external 2. Despite these differences, we have gotten considerable insight into the present problem from Goodwin (1961), Uzawa (1964), and Stoleru (1965). 344 OPTIMAL PATTERNS OF GROWTH AND AID assistance. 3 Thesc studies use models in which import choices and al- ternative levels of external capital are explicitly considered. Multisec- toral analyses are uscd in most of thcm to dcrive relations among capital inflow, import requirements, savings rates, investment alloca- tion, and overall growth for the planning pcriod considered. In these disaggregated open models, the balance-of-payments limit replaces the capacity to produce investment goods as a general factor limiting growth. The inflow of external capital plays the dual role of raising both this specific resource limit and the savings limit on thc rate of investment. Although most of the planning models cited above include some elements of optimization for a five- or ten-year period, they do not consider the pattern of capital inflow over a long enough period to show the welfare implications of alternative strategies of aid and growth. They do suggest, however, that there are some common characteristics of optimal patterns of aid and growth. The model de- veloped for the present analysis is designed to e~plore this possibility more systematically. It tries to relate the features of optimal growth patterns to the development policies and objectives of aid recipients and the assistance policies of the donors. The empirical background Pakistan was chosen for this study because it received substantial external aid and it had an explicitly formulated twenty-year plan for future growth. It is also broadly representative of the initial condi- tions from which many poor countries are attempting to start a pro- cess of accelerated growth. The typical features most relevant to the present analysis, taken from the comparative analysis of thirty-one developing countries in chapter 10, are summarized below. Compared with the medians of this sample, Pakistan in 1965 had lower per capita income, savings and investment rates, and growth of GNP, but the im- provement in its development performance in previous years was significantly better than the average. 4 3. Primarily the studies of Chenery and Kretschmer (1956) of southern Italy, Sandee (1959) and Bergsman and Manne (1966) of India, Chenery and Bruno (chapter 8 of this volume) of Israel, Adelman and Chenery (1966) of Greece, Manne (1966) of Mexico, and Tims (1965) of Pakistan. 4. The initial conditions and measures of prior performance in Pakistan are given on pages 355-76. The subsequent separation of Bangladesh (then East Pakistan) from West Pakistan makes these projections of only methodological interest. THE ANALYTICAL FRAMEWORK 345 The following observations provide a basis for both the design of the model and the range of values for the policy variables over which it will be tested. (a) External resources-three-quarters of which are classified as public assistance-normally finance 20 to 30 percent of both investment and imports in underdcveloped countries and a higher proportion of the increases in these elements in typical cases of rapid development. (b) There is substantial evidence of a limit to the ability of developing countries to transform large increases in external resources into productive investment. The most convenient measure of this absorptive capacity limit is the rate of increase in investment that a country can achieve on a sustained basis. Rates of 15 to 20 percent a year have been observed in anum· ber of countries, but there has been no case of a higher value ovcr any substantial period. ( c) Shortages of imported investment goods and raw ma- terials provide a limit to growth in a number of countries. In contrast, although the capacity to produce the nonimportable components of investment is a potential bottleneck, it is more easily avoided and rarely observed. ( d) Gross marginal savings rates are significantly above the initial average rates of about 12 percent of GNP in the thirty-one country sample; they reach 30 percent in the upper quartile of countries. But there are no observed cases of marginal rates ap- proaching 50 percent or more which are implied by most theo- retical analyses of the "optimal" rate of savings.5 ( e) The availability of external capital permits an economy to grow at the limit corresponding to its ability to increase its capital stock rather than at the lower rate implied by its ability to increase domestic savings. A period of accelerating growth in which investment, savings, and external assistance all increase is therefore to be expected; it is observed in a number of countries. (f) Under present institutional arrangements for the transfer of resources from advanced to developing countries, the amount available is rationed among claimants whose total demands sub- stantially exceed the supply. Since supply conditions vary greatly 5. See, for example, Goodwin (1961) and Stoleru (1965), who have derived illustrative paths of optimal savings for underdeveloped countries from a variety of assumed welfare functions. 346 OPTIMAL PATTERNS OF GROWTH AND AID among recipients, however, different formulations of the re- striction On external capital may be appropriate for different coun tries. Elements of a two-sector model of an open economy 1bcse empirical observations require a substantial reformulation of conventional aggregate growth models. It remains to be seen whether some of the qualitative results of two-sector closed-economy analysis can be carried over to the open-economy case. With the addition of the choice of the capital inflow over time, however, the optimizing problem can be put in similar terms. In both cases, we are primarily interested in the general behavior of the principal variables that de- scribe a growth pattern or development strategy over time. The em- pirical studies summarized above suggest the following characteristics for a two-sector model of an open economy. SECTOR BREAKDOW::-<. The basis for disaggregating the economy is crucial because of the limits that it imposes on the possibilities for future growth and of the way in which it reflects the role of the capital inflow. Disaggregation into two sectors should show the capacity of an open economy to transform domestic resources into foreign ex- change, which can then be used to fill the gaps between the composi- tion of demand and the composition of supply. Although the foreign exchange bottleneck cannot be identified with a particular industrial category, the need to allocate capital and labor to increasing its supply is quite similar to the allocation of resources to the production of investment goods in a closed economy. In our model, a category of "trade-improving" produetion will be identified, which produces either increased exports or substitutes for goods presently imported. \Vhcther the corresponding commodity is cotton, steel, or maehincry is irrclevant.6 A two-sector model that embodies this distinction could be derived from solutions to an interindustry model in which the input structure and composition of final demand is fixed. In this way, the inputs of capital and imported goods required for an expansion of output with the existing economic structure could be determined. Possibilities for 6. A similar conc;eptual problem arises in identifying investment goods in a two-sector model, which can only be solved empirically by means of an inter- industry analysis. THE MODEL 347 import substitution or introduction of new exports could then be de- scribed by additional activities as in Chenery and Kretschmer (1956); the possibilities of transforming capital and labor into foreign ex- change ("trade improvement") could be determined by an optimiz- ing procedure. In a multisectoral analysis the result would be a rise in the incremental capital cost as the output of the trade improve- ment sector rises, reflecting the operation of the principle of compara- tive advantage. (For Pakistan, we will represent this input function by a constant incremental capital eoefficient, since we cannot estimate the function directly. ) SCARCE FACTORS. Instead of capital and labor, the scarce factors relevant to our analysis are capital and foreign exchange. The ration- ing of external capital means that its supply must either be taken as given or valued at an opportunity cost reflecting its scarcity. The tranS- formation of unskilled into skilled labor can be treated as part of the investment process, however, and total labor supply is not likely to be a limiting factor within the period relevant for the analysis. 7 POLICY OBJECTIVES AND RESTRICTIONS. Within the limitations of two- sector analysis, it is desirable to ineorporate restrictions that reflect both the limited flexibility of economic systems and the political 1im~ its to feasible policy changes. For example, any significant reduction in per capita consumption (which occurs in many so-called optimal growth paths) should probably be ruled out as politically infeasible. The introduction of such constraints makes the conclusions more rea1istic, a1though the resu1ts are less susceptible to generalization in the form of simple decision rules. The Model l11e problem of determining an optimum pattern of aid and growth over time will now be stated in linear programming form. The objec- tive is to maximize a social welfare function, incorporating benefits ( consumption) and costs (capital inflow) for each period of time. 7. The unemployment rate in Pakistan was estimated in the Pakistan Plan (Gov- ernment of Pakistan, 1964) at 20 percent and the growth of population at 2.6 percent. In wuntries having less unemployment, a more explieit treatment of the potential labor limitations might be needed, as in chapter 8. 348 OPTIMAL PATTERNS OF GROWTH AND AID The constraints are the policy goals and the definitional, structural, and behavioral relations for each time period. Yariables and parame- ters are defined below. The variable and parameter values used in the basic solution are given in tables 9-1 and 9-2. VARIABLES Y = gross national product yl = production for import substitution and export expansion yo all other production I = total gross investment ]I = investment in import substitution and export expansion 10 all other investmcnt S = gross savings F = net capital inflow M demand for traditional imports E = traditional exportsS C = consumption PARAMETERS Y = cost of foreign capital (exogenously specified) i rate of discount p = postplan growth rate r = rate of discount on postplan consumption 8 weight on postplan consumption "7 = weight for terminal year income incorporating discount procedure for future consumption e exogenous rate of growth of traditional exports kl capital-output ratio for import substitution and export expansion ko capital-output ratio for other production a marginal savings rate mo = marginal import rate on income ml - marginal import rate on investment fJ = maximum feasible rate of growth of investment p = minimum allowable rate of growth of consumption T terminal year of the plan T- n year in which aid must cease 8. Traditional imports and traditional exports mean imports that would be required and exports that could be sold were the structure of the economy to remain unchanged from the base year. THE MODEL 349 Social welfare function We wish to maximize the general welfare function where This function has three parts: (a) the discounted sum of consump" tion prior to the terminal year of the plan; (b) an indicator of the discounted value of consumption in all ycars posterior to the plan, with a variable weight!!; and (c) the discounted sum of total capital inflow with a weight, y, representing the price of foreign capital, which varies according to the supply conditions for the country concerned. By varying y and certain policy constraints, it is possible to simulate a wide range of supply conditions. If no policy constraints affecting supply conditions were added, the supply of foreign capital would be assumed to be infinitely elastic at the price y. 111is assumption of infinite elasticity is modified in the two alter- native forms of the model in order to yield a more realistic statement of the scarcity of foreign capital: · In the "Basic Solution" defined below we have imposed the con- dition that foreign aid must terminate in a given year (T n) prior to the terminal year of the analysis (T).lO In this case the supply of foreign capital remains perfectly elastic at the price y prior to (T - n), but for years after (T - n) the economy must be self-sufficient. · In a second alternative form l l we assume that the total quantity of discounted aid received during the plan cannot exceed a given amount. 9 . We put a weight, 11, on postplan consumption and use a weight of unity for the present value of plan period consumption so that the numeraire of the shadow prices that the solution yields will be the value of consumption in year 1. 10. See inequality (9.14) below. II. See inequality (9.16) below. 350 OPTIMAL PATTERNS OF GROWTH AND AID The results obtained by solving the model using these different specifications of the supply conditions are discusscd on pages 355-76. TIle question arises as to whether the welfare function is formulated from the point of view of a recipient or a donor. The answer is that it can represent views of either recipients or donors, as well as a variety of views within each group. Different welfare assumptions are repre- sented by the values given to the parameters in the objective function. For example, a country having a high preference for improvement of living standards during the period of the plan, compared with concern for living standards in the distant future, would give a relatively low weight to postterminal consumption. This view implies a low value of 8 or a plan discount rate (i) that is low relative to the postplan dis- count rate (T). The higher rate in later periods can also reflect a judg- ment as to the diminishing marginal utility of added consumption. A second example is the donor or planning authority that desires the recipient country to become self-sufficient by the end of the plan period. In this case conditions in the short run are not of primary con- cern though certain minimum standards must be met. This view can be represented in the basic model by a high value of S. The donor would not view supply conditions as given, but would use the model to help in establishing supply conditions. Our treatment of postplan consumption in the welfare function assumes that after period T the economy will proceed along a path of self-sustaining growth and that a constant portion (1 - ex.) of in- come will be consumed. An estimate of the self-sustaining rate of growth (p) can thus be determined. 12 Our use of discount rates in the welfare function is based upon the standard time preference arguments. \Ve allow for a higher discount rate in later years, which can be justified in terms of diminishing mar- ginal utility of rising per capita income. As time passes there is a cor- responding rise in per capita consumption and the marginal utility of consumption declines. Is (TIle discontinuity of year T is chosen for convenience but does not significantly affect the conclusions.) 12. As t becomes large, the average savings rate approaches the marginal savings rate, and the aggregate capital-output ratio approaches a (constant) weighted average of the two sectoral capital-output ratios in inequalities (9.7) and (9.8). The ratio of the average savings rate to the aggregate capital-output ratio yields the long-run rate of self-sustaining growth. 13. This argument is made by Goodwin (1961) in determining the optimal savings rate. THE MODEL 351 Definitional equations GNP is the sum of the net output of the two sectors: regular produc- tion and production for trade improvement, (9.1) . Gross investment is similarly the sum of investment in the two sectors: (9.2) Investment is equal to domestic savings plus net foreign capital: (9.3 ) TIle trade gap is determined by the excess of the demand for "tradi- tional" imports over the sales of "traditional" exports, less the output of the trade improvement sector. 14 111e trade gap must be filled by a net flow of external resources, F t : (9.4) This definition of the trade gap leads to a formulation of the na- tional income equality which shows trade improving production as a reduction in the trade gap: (9.5) Traditional exports are assumed to grow at an exogenously determined rate: (9.6) 111ese exports can be produced at the capital-output ratio of regular production. Structural and behavioral constraints Since labor is assumed to be in surplus, production in each sector is limited only by capital in that sector and by the supply of imports. 14. As explained above "traditional" imports and exports mean imports that would be required and exports that could be sold if the structure of the economy were to remain unchanged from the base year. 352 OPTIMAL PATTERNS OF GROWTH AND AID The structure of the economy in the base year is the basis for defining the limit to regular production: t-1 (9.7) VtoLVo+l/ko ~Il. (I Production for trade improvement requires a higher capital-output ratio and, by definition, investment in this sector begins only after the plan has commenced: t~~l (9.8) Vl L I/kl ! Il. 1 The aggregate capital-output ratio is a weighted average of the capital- output ratios of the two sectors; it changes over time as the distribu- tion of invcstment between the two sectors ehanges. In the period of self-sustaining growth the proportion of trade improvement in- vestment asymptotically approaches a limit of about 25 percent of total investment. In the basic solution of the model, the economy is forced to self-sustaining growth after t = 20. Maximum savings in any year is a function of base year savings and the increase of income since the base year: (9.9) As V t becomes large the average savings rate will approach the mar- ginal savings rate, cx. The marginal savings rate can be viewed as partially a behavioral constraint and partially an instrument of policy. \Vithin certain limits the government could institute policies that would affect cx. vVithin the model presented here, however, the mar- ginal savings rate is taken as given. The requirement for goods traditionally imported is a function of base year imports and the increases from the base year in income and investment: (9.10 ) Although the marginal import rates ean be affected by policy deci- sions, within the present model they are taken as technical parameters. The relatively high value of the marginal import rate on investment (ml) produces some of the pressure of rapid growth upon the trade gap. THE MODEL 353 The observed limits to the ability of an underdeveloped country to absorb increases in the supply of capital are incorporated in the model by placing an upper limit U~) on the rate of growth of investment: (9.11 ) Although an underdeveloped country may be able to raise its absorp- tive capacity in time, it is in the early years of the plan-when little could be done to raise the absorptive capacity-that the upper bound on growth of investment is of greatest importance. 15 It is also necessary for technical reasons to place a lower bound on the growth of investment. To prevent unrealistic declines in invest- ment-which the model would otherwise yield-we have included the following constraint: (9.12 ) Policy constraints l11e welfare function largely defines the policy goals of the nation. However, certain goals can only be formulated in terms of absolute targets and must therefore be stated as constraints of the modcl. One such goal is the undesirability of allowing per capita con- sumption to decline. lois can be prevented by the inclusion of a constraint requiring total consumption to grow at least as rapidly as population: (9.13 ) Another policy goal which it is necessary to formulate as a con- straint is the requirement that capital inflow be terminated by some predetermined year: (9.14 ) Ft L. 0 for t =T- n, T. The significance of this modification of the aid supply condition was pointed out above in the discussion of the welfare function. 15. The absolute limit on absorptive capacity is somewhat arbitrary. This limit implies that no further investment can take place because of shortages of comple- mentary inputs. It would probably be more realistic to assume that above this limit further investment can be carried out but only at higher capital-output ratios and with longer time lags. This more realistic assumption could be incorporated into our linear model by using step functions. 354 OPTIMAL PATTERNS OF GROWTH AND AID Alternative forms of the model As formulated above, the model allows the foreign assistance supply conditions to be specified in two forms: either as the price of foreign capital (y) or as the terminal date for capital inflow (T n). An alteration of the model allows a third method of specifying supply conditions. In this third form we place an upper limit on the total quantity of aid received over the plan and specify neither a price of foreign aid nor a tcrmination date. That is, we add the constraint, ! ,~~F_t_~ T (9.15 ) L. F. t=l TIle three forms of the model will be discussed on pages 355-76. It will be shown that equivalent results can be obtained from each form. For example, if a price is specified, a termination date and a total quantity of aid will be endogenously determined. We can therefore summarize the three forms of the model as follows: Termination Form Price (y) date (T n) Total aid (F) I speCified determined determined 2 determined specified determined 3 determined determined specified It is, of course, possible to combine two of these forms although only one of them will turn out to be effective. Tl1is was done in the basic solution (see the next section of this chapter) where both a minimum price of aid and a maximum termination date were specified. The solution then determines which limit is controlling. Limitation of the pattern ot aid As explained below, foreign assistance is typically rationed on an annual basis by the donors. To reflect this supply limitation in our model, we will compute a set of solutions in which capital inflow GROWTH ALTERNATIVES FOR PAKISTAN 355 cannot exceed a given ratio to GNP. TIlis results in adding the follow- ing limit to the model: (9.16) In the experiments discussed below q = 0.05. OTHER LIMITS. In developing the basic model, alternative forms of some of the structural relations were investigated. The most im- portant of these was the use of separate upper and lower bounds on the rate of growth of investment in each sector. This procedure is based on the rationale that production and trade improvement are actually two distinct types of investment since trade improvement requires the construction of new plants and the development of new industries. Although this assumption prevents the rapid shifting be- tween one form of investment and another, it does not significantly alter the qualitative form of the results. It was therefore omitted from the final form of the model. Growth Alternatives for Pakistan 111e Pakistan Planning Commission made two twenty-year projec- tions or "perspective plans" as a basis for its Third Five-Year Plan for 1965-70.16 In both of these projections the net inflow of external resources is assumed to decline steadily and to approach zero by 1985. Little reason is given for this assumption apart from the desire to become independent of foreign assistance. Its effect on other objec- tives of the plan, such as the terminal year income, is not discussed. To isolate the effects of varying amounts of external assistance, we start from the planning situation described by the objectives and con- straints of the Pakistan plan. The plan document and other analyses of the Pakistan economy are used to determine plausible values for the parameters in our model and possible variations in them. We have made no attempt, however, to incorporate all of the economic and political considerations that affect the preparation of a develop- 16. Government of Pakistan (1964, 1965). 356 OPTIMAL PATTERNS OF GROWTH AND AID ment program. Our results were not designed as a critique of the plan but to suggest the possibilities for more effective development strate- gies if assistance policies could be modified. Our procedure is as foHows. \Ve first detennine an optimum solu- tion to the model in its original fonn based on welfare objectives and performance characteristics similar to those in the Pakistan plan. This basic solution provides a point of departure for several sets of experi- ments. The first is designed to show the welfare effects of supplying assistance under conditions that more elosely approximate existing arrangements. The second set of experiments shows the effects of development performance on aid requirements. In both cases, we have assumed a range of values for the external capital inflow to show the increases in consumption and income made possible by increasing aid. Taken together, these experiments bring out the interrelations between development strategy and foreign assistance policy and suggest the advantages of greater coordination between the two. The basic solution The development of the model described in the preceding section required a period of experimentation. It was necessary to determine a satisfactory form of the model in which (a) the postulated objective function led to a rate of growth of national output similar to that taken as the objective of the Pakistan long-term plan and (b) im- plausible fluctuations in consumption and investment were eliminated. lne end product of these experiments is contained in relations (9.1) through (9.14) above. The result of maximizing the welfare function subject to these fourteen constraints (for each time period) will be called the basic solution. The initial conditions and structural parameters assumed in the basic solution are given in tables 9·1 and 9-2, which also present the corresponding values from the two versions of the Pakistan Perspec- tive Plan wherever they are availableP 'Ine welfare function pa- rameter values used in the basic solution are given on the next page. 17 _ \Ve started from the preliminary version of the Pakistan five-year plan for 1965-70 (Government of Pakistan, 1964) and twenty-year perspective (version 1 in table 9.1) and made some revisions after the final plan (version 2) became available (Government of Pakistan, 1965). GROWTH ALTERNATIVES FOR PAKISTAN 357 Table 9-1. Base Year Data (millions of 1965 rupees) Model values Pakistan plan values Base vear t 3 Version 1 Version 2 Variable 1962 1965 1965 1965 Fo Foreign aid 1,183 1,956 2,750 3,690 So Savings 3,381 4,620 4,200 4,710 10 Investment 4,564 6,586 6,950 8,400 Mo Imports 3,743 4,920 5,700 6,990 Eo Exports 2,559 2,954 2,950 3,050 Vo National income 37,380 42,539 44,000 45,540 Co Consumption 33,999 37,919 39,800 40,830 Sources: Version 1: Government of Pakistan (1964). Version 2: Government of Pakistan (1965). Note: Model values are averages derived from a time trend for the years 1957- 62, which were thought to be more representative than the actual data for 1962. NONSTRUCTURAL PARAMETERS IN TIlE BASIC SOLUTION i = 0.08 (rate of discount during plan period) r = 0.10 (rate of discount on postplan consumption) p = 0.073 (postplan rate of growth) y = 2 (cost of foreign capital) "f} = 3.4 (defined in the text) o 1 (relative valuation of postplan consumption) T = 23 (terminal year of plan) T - n = 20 (year in which aid must cease) 111e growth of national output in the basic solution is shown in figure 9-1 and table 9-3 to be approximately midway between the two versions of the perspective plan and therefore adequately representa- tive of Pakistan's objectives. The time paths of the other variables in the basic solution are shown in figures 9-2 through 9-5 and tables 9-4 and 9-5. Since the solution to the model does not distinguish between that part of trade-improving investment which is import substitution and that part which is export expansion, we have made an arbitrary distribution of trade improvement output for illustrative purposes. 18 18. Trade improvemeut production was allocated to export expansion so long as the rate of growth of exports did not exceed 6 percent, which was the export forecast in version 1. When this level was reached, the remainder was allocated to import substitution. Table 9-2. Value of Structural Parameters Pakistan Pakistan Pakistan Pakistan Pakistan plan plan plan plan plan 1965-85 1965-70 1970-75 1975-80 1980-85 Model Ver- Ver- Ver- Ver- Ver- Ver- Ver- Ver- Ver- Ver- w 1965- sion slOn slOn sion sion sion sion sion sion sian \J1 00 Parameter 85 1 2 1 2 1 2 1 2 1 2 a Marginal savings rate 0.24 0.286 0.25 0.23 0.22 0.26 0.25 0.30 0.28 0.31 0.25 mo Marginal import rate on income 0.10 0.072 0.06 n.a. 0.12 n.a. 0.09 n.a. 0.06 n.a. 0.04 m l Marginal import rate on investment 0.35 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. ko Incremental capital-output ratio, regular production 3.0 3.6 2.9 3.5 2.9 3.5 2.9 3.6 2.9 3.7 3.0 kl Incremental capital-output ratio, trade improvement 4.5 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. r Rate of population growth, in percent 2.5 l1.a. 2.6 2.6 2.7 2.7 2.8 2.6 2.6 2.2 2.1 f3 Maximum rate of growth of investment 0.13 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. e Rate of growth w of exports, in \.rt -..0 percent 4.9 6.0 7.9 6.0 9.5 6.0 6.0 8.6 6.0 4.9 Sources: Version 1: Government of Pakistan (1964). Version 2: Government of Pakistan (1965). n.a. = not available. Note: In the model: Tno, m., and kG were estimated from time trends for 1957-62; e and a were modified to reflect improved performance in 1963 and 1964. 360 OPTIMAL PATTERNS OF GROWTH AND AID Table 9-3. Growth Rates and Significant Ratios for the Basic Solution of the Model and the Two Versions of the Pakistan Plan Percent rate of Years Plan growth Vn/Vo I"/V,, S"/V,, Fn/V" 1"1/1,, Basic Solution 1965-70 III 5.9 1.33 0.21 0.14 0.07 0.05 1970-75 IV 7.7 1.45 0.27 0.17 0.10 0.07 1975-80 v 8.0 1.47 0.23 0.19 0.03 0.39 1980-85 VI 6.3 1.36 0.21 0.21 0 0.21 Pakistan Plan, Version 1 1965-70 III 5.4 1.30 0.19 0.13 0.06 1970-75 IV 5.9 1.33 0.20 0.16 0.04 1975-80 v 6.7 1.38 0.22 0.20 0.02 1980-85 VI 6.8 1.39 0.24 0.23 0.01 Pakistan Plan, Version 2 1965-70 III 6.7 1.38 0.20 0.14 0.07 1970-75 IV 7.3 1.43 0.21 0.17 0.04 1975-80 V 7.5 1.44 0.22 0.20 0.02 1980-85 VI 7.5 1.44 0.23 0.22 0.01 Source: Same as table 9-1. In this table the subscript n refers to the final year of the particular plan and the subscript 0 refers to the first year of the particular plan. THE PATTERN OF INVESTMENT AND CAPITAL INFLOW. An examination of the binding constraints and their shadow prices shows that the twenty-three-year period of the basic solution can be divided into three subperiods or "regimes." Each regime may be identified by the set of constraints binding it. Since some are binding throughout (the limits on capacity, savings, and trade) the regimes can be described in terms of those that change. 'This gives the combinations in the basic solution that are displayed in the informal table at the top of page 361. In the first regime19 the economy grows at the maximum rate per- mitted by the absorptive capacity limit on total investment, with only a small fraction allotted to import substitution. Since investment rises 19. This regime corresponds to phase 1 of the model described in chapter 10. GROWTH ALTERNATIVES FOR PAKISTAN 361 Distinguishing Regime Description constraint Period I Maximum investment Upper bound on 1963-76 and growth rate of growth of investment (9.11) II Trade improvement Lower bound on 1977-81 rate of growth of investment (9.12) III Balanced growth No foreign 1982-85 capital (9.14) Figure 9-1. Growth of GNP in Pakistan's Perspective Plans and in Two Solutions to the Model 40 (200) P-2 -------Pakistan Plan, version 1 (P-l) / - - - Pakistan Plan, version 2 (P-2) --Basic solution to the model (A-20) - - - Limited aid pattern solution to the model (B-20) 30 (150) '" ~ . '"0 ...., Q) C;; 140 530 ;:; :: .5 0 § ~ Q) ;0 f-. 120 500 C;; '0 f-. Total aid (F) GROWTH ALTERNATIVES FOR PAKISTAN 371 TIle decline in the marginal productivity or value of aid results from the fact that as its quantity is increased, the use of external resources for investment purposes has to be postponed because of the limitation on absorptive capacity. This postponement on the one hand reduces the amount of additional consumption and saving achieved per unit of additional aid during the plan period. On the other hand, since the value of future aid is discounted at 8 percent, a dollar of aid today is equal in present value to $4.70 twenty years from now. For this reason, there is no decline in the marginal produc- tivity of total discounted aid as measured by its effect on the terminal year income although there would be a decline with a lower discount rate. TIle two componcnts of the welfare function are given separately in table 9-6 and figure 9-6 (part b) to show these two effects. For any aid total, the marginal product in figure 9-6 (part a) is equal to the sum of the marginal effects of aid on total consumption and terminal income with V 7' given its appropriate weight. The development sequence represented by the three regimes of the basic solution is unaffected by changes in the total amount of aid. As the total is reduced, the length of each of the first two regimes is shortened as indicated in table 9-6. The effect on the optimal time path of aid is shown in figure 9-5. For example, solution A-14 shows the effect of reducing the total aid by 50 percent from the basic solution and consequcntly shortening the period of aid from twenty to fourteen years. It is significant for assistance strategy that the optimal paths of all the variables are unaffected in regime I by an earlier tCrD1ination date. TI1crefore a change in the total aid anticipated necd not affect plan- ning during this period. ANNUAL RATIONING OF AID. 'l11C procedures by which public capital is currently supplicd to developing countrics result iT) a system of rationing in which thcre tends to be an absolute ceiling on the amount of aid furnished to any country in anyone year. This ceiling can be represented in our model by limiting the annual inflow to a pre- determined fraction of GNP. We shall analyze the effects of such a limit in Pakistan by assuming a maximum of 5 percent of GJ't Ratio of foreign capital inflow to GNP in year t(Ft/Vt ) E Rate of growth of exports Since the basic model is designed to explain the functions of aid and to evaluate current performance of developing countries, it is useful to have in mind the typical values of the principal parameters. Table 10-1 gives the upper quartile, median, and lower quartile values of each parameter for a sample of thirty-one countries during the 1957-62 period. The sample covers most of the underdeveloped world, and the median values are quite close to the aggregate UN estimates for all underdeveloped countries. 9 The median capital-output ratio 9. The UN estimates investment at 16 percent of GNP in 1960 and a growth of GNP of 4.4 percent for the previous decade. See United Nations (1963, pages 19, 37). 388 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-L Distribution of Parameter Values, Thirty-one-Country Sample Upper Lower Parameter Symbol quartile Median quartile -~- ..... Highest five years in recent past Compound growth rate of gross investmcnt f3 0.19 0.14 0.10 Relations during 1957--62 Compound growth ratc of gross investment 0.12 0.07 om Incremental capital-output ratio (assuming one-year lag) k 2.78 3.52 4.72 Compound growth rate of GNP r 0.062 0.046 0.034 Ratio of gross invcstment to GNP in 1962 (after time-trcnd fitting) Io/Vo 0.20 0.17 0.14 Ratio of net foreign capital inflow to GNP in 1962 (after time-trend fitting) 4>0 0.07 0.04 0.01 Ratio of gross national saving to GNP in 1962 (after time-trend fitting) ao 0.16 0.12 0.09 Marginal national saving ratio (change in saving -+- change in GNP) a' 0.26 0.19 0.02 Ratio of gross imports of goods and services to GNP in 1962 (after time-trend fitting) jLo 0.16 0.20 0.39 Marginal import ratio (change in gross imports of goods and , services change in GNP) po 0.01 0.20 0.46 Compound growth rate of exports of goods and services 0.080 0.051 0.021 Change in gold and convertible foreign currency reserves, December 1956 to December 1962 -+- change in GNP 1957-62, GNP first converted to 1962 U.S. dollars' (after time-trend fitting) , 0.101 -0.001 -0.065 p a_ Excludes Trinidad and Tobago and Mauritius because of lack of data. AID AND THE TRANSITION TO SUSTAINED GROWTH 389 (3.5) and saving rate (0.12) suggest that without external assistance the typical growth rate of underdeveloped countries in this period would have been about 3.4 percent or less than 1 percent per capita. INVESTMENT-LIMITED GROWTH. Our hypothesis of an economy with limited flexibility suggests the use of a programming model in which growth proceeds at the highest rate permitted by the most limiting factor.lo We assume to start with that the balance of payments does not become the limiting factor. A process by which self-sustaining growthll can be attained by using aid to fill the temporary gap be- tween investment ability and saving ability can be derived from the following description of the economic structure: DEFINITIONS (10.1 ) (10.2) CAPACITY LIMIT (10.3) ABILITY TO INVEST (10.4 ) SAVING LIMIT (l0.5) TARGET GROWTH RATE (10.6) 10. A more complete statement of this model in linear programming form, given in chapter 9, considers the implications of the present analysis for the optimal planning of development. 1 L This concept will be defined as growth at a given rate with capital inflow limited to a specified ratio to GNP which can be sustained without concessional financing. 390 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT The capacity limit (10.3) is based on the Harrod-Damar assump- tion that a specified amount of investment is needed to increase out- put. The assumption of a linear capital-output function is a matter of convenience. A similar formulation can be derived from more general production functions of the Cobb-Douglas type if there are not sig- nificant changes in the relative costs of labor and capital. Since in most cases the period of transition is one in which the total supply of labor is not a significant limitation, it is plausible to approximate the aggregate production function in this wayP The limit on the ability to invest (10.4) is introduced to rcflect the widely hcld view that absorptive capacity for additional investment in any period is limited by the supply of complementary inputs, which can only bc increased as a result of the development process. \Ve refer to the parameter f3 as the skill limit, reflectiug the skill formation required of managers, skilled labor, and civil servants in order to in- crease productive investment. ls The highest observed value for the skill limit over any recent five-year period is about 20 percent a year, but few countries have sustained a growth of investment of over 10 percent for as long as ten years. The saving limit (10.5) is designed to include not only the marginal propensity to save but the government's ability to increase total saving by changes in the tax structure and by other policies. For this reason, we make the saving limit a function of total GNP (and hence of timc) rather than of per capita income. 14 The target growth rate (10.6) reflects the almost universal practice in deve10ping countries of summing up the principal goal of develop- 12. The introduction of a nonlinear relation between capital and output would not materially affect the conclusions of our analysis. Intercountry regression analyses suggest that there is a reduction in the capital·output ratio at higher growth rates but little relation to per capita income. Efforts to estimate more general production functions from time series in underdeveloped countries have been inconclusive be· cause of the limited data available. B. In the original model for Israel in chapter 8, the skill limit was associated with labor only, but in the more typical underdeveloped country the managerial aspect is at least as important. 14. Fei and Paauw (1965) have used a similar model to analyze aid require- ments for the case in which investment resources provide the limit to growth (our phase II). 111ey have adapted the Rosenstein-Rodan (1961a) model by assuming that per capita saving is a constant fraction of the increment in per capita income. There has been no empirical test of the relative merits of this alternative specifica· tion of the savings function compared to ours, but they yield similar results when the rates of growth of per capita income and population do not vary greatly. AID AND THE TRANSITION TO SUSTAINED GROWTH 391 ment in a given rate of increase in GNP. In the present context, it also reflects the fact that foreign assistance is limited and is unlikely to be available to finance growth rates much above 6 to 7 percent even if they were attainable. Since the average terms on external loans are based largely on the country's future economic prospects, this puts a limit on the total amount which it can afford to borrow. For all these reasons, either a target growth rate or some other reflection of the fact that investment cannot indefinitely exceed saving must be in- cluded in the modeJ.15 To complete the system, we need some minimal assumptions as to the objectives of the recipient country and the conditions under which aid is provided. \Ve assume that aid is sufficiently limited-or expen- sive-to make the recipient unwilling or unable to increase aid merely to increase consumption without also securing some rise in GNP. Second, we assume that the country tries to maximize consumption until the target growth rate is attained. These assumptions lead to a determinate pattern of growth whose welfare implications will be examined below. The model of investment-limited growth contains six restrictions and five variables. Under the assumptions made, there is no incentive to build excess capacity or to increase aid by reducing saving. In- equalities (10.3) and (10.5) therefore become equalities. The increase in GNP will be limited first by the ability to invest and then by the target growth rate if the investment rate reaches the level (kr) re- quired to sustain it. \Ve shall denote the first period as phase I, which is described by equations (10.1) to (10.5). In phase II, inequality (10.6) becomes effective and replaces inequality (10.4) as a restriction on the system. i6 The growth path and aid requirements over time can be described by solving for V t and F t in each phase and determining the point at which the economy passes from phase I to phase II. Phase I is characterized by a constant growth in investment at the annual rate of 11 and by an accelerating growth rate of GNP. From 15. Alternative formulations are discussed on pages 404-09. 16. As shown in chapter 9, this result can be derived in more formal terms by maximizing a welfare function having the characteristics indicated subject to the given restrictions. Each phase is then defined by the restrictions which are binding, which have positive shadow prices. This linear programming formulation is quite useful if we replace the assumption of a target growth rate by a more complicated set of limits, but it is unnecessary with the simplified assumptions made here. 392 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT equations (10.3) and (10.4) it can be determined that the increment in investment in each period is a constant ratio (13k) to the increment in GNP. Solving the system for the level of capital inflow, gives: (10.7) Ft Fo + (13k - at) (V t - Yo), where Po = 10 - So. 1bis equation shows that the increment in ex- ternal capital (Ft - Fo) finances the difference between the incre- ment in investment and the increment in saving. Without increased capital inflow, a country having the median values of k and a' (table 10-1) would have a growth of investment of about 5 percent a year. To achieve a growth of investment of 10 percent would require that nearly half of the increased investment during phase I be financed by external capitalP This formulation can be interpreted in terms of Harrod's original suggestion of different growth rates corresponding to the growth of the labor force (the "natural" rate) and the potential saving limit (the "warranted" rate). \Ve have replaced the natural rate with a skill- determined rate based on the ability to invest. External assistance fills the gap between investment and saving, permitting the higher rate to be reached. Phase I ends in year m when investment reaches a level adequate to sustain the target rate of growth: (l0.8) Substituting this value for investment in the equations for phase I gives the value of GNP in the terminal year: ((3-ro) (10.9) Vm=V o ({3 rm ) If, for example, investment grows at 10 percent a year, the hypo- thetical median country could increase its investment rate from 12 percent of GNP with no aid to the 21 percent needed to support a 6 percent growth target in a period of eleven years. 17. With k = 3.5 and a.' = 0.19, growth of investment at 10 percent would require M 0.3 5 A V of which 0.19 A V (54 percent) would be financed by in- creased savings and 0.16 AV (46 percent) by increased capital inflow. AID AND THE TRANSITION TO SUSTAINED GROWTH 393 The time to complete phase I can be determined by solving for m in the following equation: r (f3-To) (10.10) To (f3-r) where and Phase II in our model corresponds to the process of aid and growth discussed by Rosenstein-Rodan (1961a, 1963). GNP and investment rise at a constant rate with external assistance determined by the difference between kr and at. Solving the system for the rate of growth yields a modified form of the Harrod-Damar equation: at M) in phase II or a loss of potential saving, (S < 5) in phase III. A better coordinated development policy would attempt to reduce the required capital inflow by substituting investment for imports-or vice versa-to equate the two gaps ex ante over the long run. 32 If we assume efficient resource allocation, the equilibrium exchange rate-reflecting the opportunity cost of earning or saving foreign ex- change at the margin-will be a function of the amount of inflow of 3 L A generalization of these relations is given in the next subsection. 32. Given the durability of capital, it may not be efficient to equate the two in the short run, especially when there is a significant degree of disequilibrium to start with. AID AND THE TRANSITION TO SUSTAINED GROWTH 405 external resources, F. Under eeteris paribus assumptions, a reduction in F normally implies an increase in the value of foreign exehange as marginal activities of import substitution or additional exports are undertaken. Since capital is the only scarce domestic resource in our model, we assume that a higher capital coefficient is needed to reduce the import requirements of model 1 and conversely that a saving in capital will result when imports increase above the minimum level.38 These assumptions form the basis for model 2, in which an import- substitution activity is added to model }.34 Investment in import sub- stitution {Im} requires imported capital goods and a greater amount of capital (bk) for each unit of imports replaeed by domestic production. The net reduction in import requirements at time t (Mm) is given by: 1 t-l (10.17) Mm = - ~ 1m - aIm t bk 0 t t, where a is the import content of 1m above the average for the eeonomy and b is greater than 1. Equation (10.13) of the basic model will then be replaced by: (10.13') The capacity limit, equation (10.3), must also be modified to allow for the lower productivity of capital in import substitution. Model 2 wi1l bc used to estimate the minimmll capital inflow needed to achieve a given level of GNP, first in the Pakistan example and later in comprehensive projections. For this purpose we assume (a) that total import substitution (positive or negative) is sufficient to eliminate the difference betwecn the two structural gaps over the period 1962-75, and (b) that this type of investment increases linearly throughout the period. Solutions calculatcd for Pakistan for varying 33. The efficient reallocation of resources to accord with variations in the capital inflow is analyzed in detail in Chenery (1955), wh ich provides the basis for the aggregate formulation given here. \Ve have approximated the diminishing marginal productivity in import substitution by a single incremental ratio. H. The same argument can be made for export expansion, using the marginal revenue product of additional investment to allow for the inelasticity of export demand. For convenience, we assume only import substitution here. 406 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-5. Effects of Imf)ort Substitution Policy on Capital Inflow, Pakistan, 1962-7; (all figures expressed as ratios to 1962 GNP) Alternative growth targets 1. Target 1975 GNP 1.468 1.665 1.886 2.133 2.410 2. (GNP compound growth rate) (0,03) (0.04 ) (0,05) (0.06) (0.07) 3. Cumulative exports (both models) 1.33 1.33 1.33 1.33 1.33 Modell (cumulative values) 4. GNP 17.09 18.29 19.60 21.02 22.55 5. Investment 1.54 2.19 2.94 3.78 4.74 6. Savings: (potential) ( 2.00) (2.29) (2,60 ) (2.94 ) (3.31 ) 7. Savings: realized 1.16 1.70 2.31 2.94 3.31 8. Imports: (potential) (1.71 ) (1.S3 ) ( 1.96) (2.10) ( 2.26) 9. Imports: realized 1.71 1.83 1.96 2.17 2.76 10. Excess consump- tion 0.S4 0.59 0.29 0 0 11. Exces~ imports 0 0 0 0.07 0.50 12. Net capital inflow 0.38 0.50 0.63 0.84 1.42 13. (Dominant phase) (1II ) (1II ) (III ) (11) (II) Model 2 (cumulative values), 14. Gl\'P 17.09 18.29 19.60 21.02 22.55 15. Investment 1.81 2.38 3,03 3.76 4,58 16. (Percentage of invest· ment in import substitution) (44) (24) (9) ( -2) ( -10) 17. Savings 2.00 2.29 2.60 2.94 3.31 18. Imports 1.14 1.43 1.76 2,15 2.60 19. Net capital inflows -0.20 0,09 0.43 0.S2 1.26 \Velfare effects 20. Consumption, model 1 15.93 16.60 17.29 IS.D7 19.24 21. Consumption, model 2 15.09 16.00 16.99 18.07 19.24 22. Change in consumption (line 21 20) -0.84 -0.59 -0.29 0 0 23. Change in capital inflow (line 19 - 12) -0.57 -0.40 -0.20 -0.02 -0,}6 24. (Ratio, line 22 -7 23) 1.47 1.47 1.47 0 a Note: This table assumes no constraints on growth of investment or GNP. This means that a country could invest sufficient capital in each year to attain the GNP growth rate given in line 2. Actual 1962 investment was sufficient for an initial growth rate of about 0.04, a. The basis for model 2 is given in the text. The formulas for calculation are detailed in Chenery and Strout (1965, annex B). AID AND THE TRANSITION TO SUSTAINED GROWTH 407 growth targets are shown in table 10-5.35 Cumulative values of the two gaps in model 1 and of the single gap in model 2 are plotted in figure 10_2.36 Figure 10-2 shows that at a growth rate of GNP of 5.2 percent, the cumulative values of the two resource gaps in model 1 are the same and equal to the total requirement for foreign capital in model 2. At lower growth rates, the trade gap predominates in model 1 and the difference between the two curves represents an excess of consump- tion. In model 2, about a third of this excess is utilized to finance the additional investment needed for import substitution; the remaining two-thirds is eliminated by reducing the capital inflow.:n At growth rates above 5.2 percent, phase II predominates in the basic model and the possibilities of reducing the capital inflow through (negative) import substitution are less. On our assumptions, substituting imports for investment would produce a reduction in the dominant saving gap of about a third of the difference between the two gaps in model 1. A more efficient method of reducing the cumulative capital inflow in cases where the saving gap exceeds the trade gap is to accelerate the rate of growth at the beginning of the period instead of maintaining a constant growth rate. 1bis early growth has the effect of increasing total saving as well as total imports, but the net reduction in the capital inflow is greater than with model 2.38 The practical scope for 35. We have used a value of b of 1.5, which implies that additional import substitutes would become profitable at an average exchange rate 50 percent. higher than the present effective rate. This value was judged to be the upper limit to the cost of import substitution or increased exports in the amounts needed to reconcile the two gaps. A comparable value was found in the investment programming model for southern Italy (Chenery, 1955). 'I1le additional import content of in- vestment (a) is taken as 0.25, corresponding to a total import content of 035. 36. For model 1, we have assumed that the total capital inflow would be equal to the larger of the cumulative resource gaps because of the possibilities of adjust- ment through stock change, variations in imports, building ahead of demand, and so forth. Taking the larger resource gap on an annual basis would give somewhat higher totals for modell in table 10-5 and figure 10·2. 37. For example, at a growth rate of 4 percent excess consumption of 0.59 in model 1 is converted into increased investment of 0.19 and reduced imports and capital inflow of 0.40 in model 2. The same proportions hold at other growth rates below 5.2 percent. 38. Tne theoretical aspects of the variable growth mechanism are more fully explored in Chenery and Strout (1965). There it is shown that slowing down growth to save on external capital in phase III is a very inefficient alternative to import substitution. In the present example it results in a loss of $11 of consump- tion for each dollar of capital inflow saved. 408 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Figure 10-2. Gap Equaliz.ation through Additional Import Substitution (Model 2): Pakistan, 1962-75 Implicit GNP growth rate, 1962-75 1.4 Model 2 -- 1.2 > :'- 1.0 -- .,:; ~ ~ ~ C>. '" () '" 0.8 c:: .!!/l 0.6 -\!:! 0 ... .... c:: 0.4 ... :E -3 E 0.2 :::> U hf---t""""'*--t-Phase 1I+-----<"i predominates in Model I 0 -0.2 -0.4 Final GNP as a ratio to initial GNP (Vn/V Q ), where n = 13 years raising the growth rate in most countries is limited by absorptive capacity constraints, however. A more comprehensive analysis of the efficiency of alternative growth patterns would require us to abandon the assumption of a given target growth rate and to determine the growth target and the pattern of capital inflow from the objectives of the economy and the AID AND THE TRANSITION TO SUSTAINED GROWTH 409 limits to the use of various policy instruments. This approach has been used in the linear programming analysis of optimal growth patterns in chapter 9. The results confirm our assumption that it is efficient to eliminate the ex ante difference between the two resource gaps to the extent feasible. The main features of the growth pattern determined from the basic model also characterize the optimal solu- tions to the more general planning mode}.39 The productivity of external resources TIle productivity of an increment in external resources supplied to a developing country can be measured by the corresponding increase in consumption or total income that it makes possible. The value of ex- ternal resources depends on the extent to which they facilitate the fuller lIse of domestic factors. In our models, it is possible to measure the effect of increasing the supplies of investmcnt funds and foreign exchange, but we have no estimate of the possibilities for raising the skill limits to growth. Figure 10-2 provides one measure of the marginal productivity of external resources in Pakistan over the range of growth rates indi- cated. 40 Under the assumptions of modell, the productivity of aid is much higher in phase III, when the balance of payments is the factor limiting growth. This result is stated in more general terms in the following formulas for the derivative of the terminal year income with respect to the total capital inflow from the equations in table 10-2. For phase II: d(V t + 1) 1 (10.18) d(!, k a'/ 39. The principal differences between model I and thc linear programming solutions to Pakistan are (a) a continuation of phase r beyond the point at which the target rate is achieved to maximize the benefits of accelerated growth; (b) reo placement of phases II and III by a single regime in which the capital inflow is re- duced to 0 with the two gaps kept equal by import substitution, as in model 2 above. 40. Since figure 10-2 is designed to illustrate the effects of import substitution, we have omitted the complicating element of absorptive capacity, which would raise the total capital inflow required for higher growth rates and put an absolute ceiling all the maximum growth achievable of about 6.3 percent by 1975 in the Pakistan example. Discounting the total capital inflow at 8 percent would reduce the total value by 30 to 40 percent and raise its marginal productivity. 410 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT For phase III: d(V t + t ) 1 (10.19) m d(!, F't ) where Values of y for re1cvant time periods and growth rates are: ~.-~.-~.-~.--. r t 4 t =:. 9 t =:. 14 0.03 1.8 4.1 5.9 0.05 1.8 3.8 55 0.07 1.7 3.5 4.9 0.10 1.7 3.4 4.4 These formulas give the follmving values for the increase in terminal year income for each dollar of increase in eumulative assistance for Pakistan and for the median parameter values of table 10-1 (assuming r 0.05 and t 14): Productivity Pakistan lHedian values Phase II 0.44 0.35 Phase III 1.14 0.91 For periods under twenty years, there is a pronounced tendency for the two productivity curves to have the relative slopes indicated for Pakistan, with phase II predominating at high rates of growth.H As the length of time increases, the productivity of assistance in phase II rises because of the additional saving generated, while the productivity in phase III falls. Under the more optimal policies assumed in model 2, there is a single productivity curve with a slope closer to that of phase II in model 1.42 41. This result was also obtained in chapter 8 for Israel and by McKinnon (1964) for more specialized assumptions. An estimate of the productivity of aid to Greece is given in Adelman and Chenery (1966). 42. TIle marginal productivity curve derived from the linear programming model of c1lapter 9 is similar to that for model 2. AID AND THE TRANSITION TO SUSTAINED GROWTH 411 For long-term development policy, it is more useful to consider the total assistance required to complete the transition to self-sustaining growth in relation to the country's performanee. This can be done by varying the parameters for phase II of model 1 over the range of values observed in table 10-1. TIle results are given graphically in figure 10-3, which shows the total undiscounted capital inflow required to produce Figure 10-3. Total Capital Inflow Required to Reach Self-sustaining Growth ~ k == 4.0 $. H = Median "'- k = 3.5 performance .... N 4.0 = U U p/Jer quartile p.." performance Z ~ " ~ .S ...., 0 k 0 '';:: 3.0 '" .... '" '" <.> ;> '';:: .E :::l S ~ 8 ..... 2.0 '" c: r). Thereafter, the in- vestment rate must be adequate to sustain the target GNP growth rate (IIV kr). (b) Saving criteria. The marginal savings rate must be greater than the target investment rate (a' > kr) unless the average rate of saving is already above this level. (c) Trade criteria. Either export growth must exceed the growth target for c:,>p or the marginal import ratio (P/) must be substantially less than the initial average ratio. TIle complete statement of the trade criterion is given by equation (10.16). These criteria have been used to classify the thirty-one countries into the four main groups shown in table 10-6. The classification is based on the saving and trade performance needed to achieve a self- sustaining growth rate of 5 percent.4 7 As to the investment criteria, all countries except Burma have shown an absorptive capacity for invest- ment of greater than 5 percent in the recent past, but the five countries indicated bv an asterisk do not meet either investment criterion for the 1957-62 period. The twelve countries in group A satisfy the criteria for approaching or maintaining self-sufficiency and nine of them have already achieved growth rates of 5 percent or more. Half of this group is heavily de- pendent on external capital for its continued growth, while most of the others have favorable exports and little or no net capital inflow. Of the nineteen countries that fail to satisfv one or both criteria for approaching self-sustaining growth, failure on the trade side seems to be at least as important as deficiencies in saving and investment. More detailed studies suggest that a number of these countries-India, Greece, Turkey, Chile, Colombia, Costa Rica, Bolivia, Guatemala, and Honduras-have recently shown symptoms associated with phase 47. The parameters should reflect underlying structural characteristics rather than "realized" values in this historical period. The high marginal import ratios for Chile, Nigeria, and Turkey, for example, may reflect phase II slackness in the foreign trade constraint. The true structural import ratios may be enough lower to move Chile from group D to C and Nigeria from B to A. Use of parameters for the 1953-63 period would move Mexico from group C to A. PROSPECTS FOR THE TRANSITION 41 ., III in our model: import shortages, substantial excess capacity, and in some cases falling saving rates. 48 One of the most suggestive features of this grouping of countries is the predominant role played by exports. Ten of the twelve countries in group A have export growth rates of 6 percent or more and hence could eventually reach self-sustaining growth of 5 percent even if the ratio of imports to GNP remained constant. Conversely, one of the most significant aspects of the unsatisfactory performance of countries in group D is the stagnation of their exports, which has typically led to increased requirements for external capital and falling saving rates. 'There is almost no example of a country which has for a long period sustained a growth rate substantially higher than its growth of exports through continuing import substitution. In the past Brazil, Colombia, Turkey, and India have done so for considerable periods, but each has run into severe balance-of-payments difficulties before changing its export policies. lois comparative assessment also tends to dispel the notion that performance as measured here is necessarily associated with the initial income level. In this period, at least, there is little correlation between initial income levels and either marginal saving rates or balance-of- payments performance. Projections of future growth Since less developed countries vary widely in their ability to mobilize their O~T1 resources and to utilize external resources, estimates of fu- hue assistance requirements for groups of countries are not very useful. \Ve have therefore made a series of projections for each of fifty countries to explore the range of future growth possibilities and corresponding assistance requirements. Although the projection for any single country is fairly crude, this approach has the great ad- vantage of taking into account absorptive capacity, import require- ments, and other limitations that can only be judged on a country basis. Our analysis is designed to explore the possibilities for accelerating 48. Aggregate evidence is given in Chenery and Strout (1965); examples of more detailed analyses of the trade gap are found in Adelman and Chenery (1966), Mamalakis and Reynolds (1965), Vanek (1964), and Williamson (1965). Table 10-6. Indicators of Growth, 1957-62 Capital Saving Growth inflow Investment performance performance Trade performance in GNP Number' Country Fo/Vo kf IO/VD ao a' Eo/Mo f ",,'I[to r A. Countries meeting both saving and trade criteria b 42 Burma'" 0 0.205 0.16 -0.003 0.17 0.21 1.01 0.021 -0.91 0.046 6 Israel 0.20 0.154 0.31 0.10 0.11 0.15 0.51 0.194 1.16 0.103 7 Jordan 0.24 0.068 0.17 0.19 -0.07 0.09 0.43 0.080 0.70 0.111 45 Korea * 0.10 0.172 0.12 0.001 0.03 0.27 0.42 0.165 0.39 0.040 49 Malaya -0.04 0.116 0.18 0.18 0.22 0.26 1.08 0.059 1.31 0.062 8 Pakistan 0.04 0.117 0.12 0.15 0.09 0.25 0.64 0.075 1.97 0.041 25 Panama 0.06 0.156 0.18 0.10 0.12 037 0.85 0.100 1.31 0.051 27 Peru -0.01 0.155 0.20 0.03 0.21 031 1.04 0.143 0.86 0.073 46 Philippines 0.02 0.139 0.14 0.05 0.12 0.30 0.90 0.046 0.06 0.050 4>- ..... 43 Taiwan 0.07 0.134 0.22 0.13 0.15 0.29 0.65 0.083 0.90 0.074 0\ 47 Thailand 0.01 0.106 0.17 0.10 0.16 0.22 0.93 0.077 0.82 0.080 28 Trinidad -Tobago 0.1 0 0.217 0.31 0.05 0.22 0.11 0.88 0.107 1.25 0.078 B. Countries meeting saving criterion only" 11 Argentina 0.Q3 0.533 0.24 0.09 0.21 0.83 0.80 0.043 2.66 0.019 13 Brazil 0.D3 0.132 0.19 0.09 0.15 0.19 0.74 -0.023 0.D3 0.067 3 Greece 0.06 0.151 0.21 0.12 0.15 0.26 0.67 0.051 1.05 0.060 21 Honduras'" -0.01 0.203 0.13 om 0.13 0.25 1.03 0.028 -0.18 0.033 4 India 0.02 0.145 0.14 0.07 0.12 0.20 0.68 0.014 0.07 0.048 36 Nigeria 0.05 0.185 0.14 0.09 0.09 0.19 0.76 0.059 1.98 0.033 C. Countries meeting trade criterion onlyb 5 Iran 0.01 0.177 0.15 0.01 0.14 0.11 0.95 0.080 1.04 0.049 23 Mexico 0.01 0.153 0.14 0.02 0.13 0.11 0.91 0.051 0.54 0.050 29 Venezuela -0.08 0.326 0.19 -0.08 0.27 -0.26 1.25 -0.065 -3.41 0.043 D. Countries meeting neither trade nor saving criterion 12 Bolivia* 0.Q7 0.216 0.11 -0.02 0.04 -0.16 0.62 -0.005 0.34 0.029 15 Chile 0.06 0.138 0.13 0.12 0.07 0.10 0.73 0.061 2.12 0.D38 16 Colombia 0.04 0.208 0.20 0.04 0.16 -0.12 0.76 -0.031 1.21 0.050 17 Costa Rica 0.05 0.236 0.16 -0.01 0.11 -0.10 0.83 0.016 0.60 0.039 20 Guatemala * 0.02 0.176 0.10 -0.05 0.08 -0.03 0.86 0.014 -0.64 0.036 34 Liberia 0.56 0.390 0.67 0.57 0.11 0.21 0.50 0.033 2.85 0.046 50 Mauritius 0.09 0.249 0.19 0.08 0.10 -0.39 0.81 -0.010 1.37 0.034 26 Paraguay 0.03 0.318 0.16 0.01 0.13 0.08 0.81 0.025 0.69 0.026 40 Tunisia 0.18 0.245 0.26 0.27 0.08 -0.84 0.59 -0.086 2.59 0.034 9 Turkey 0.03 0.240 0.15 0.04 0.12 -0.02 0.72 0.050 2.95 0.D30 Source: Chenery and Strout (1966, table A·I ) . Symbols: = GNP growth rate T .... -!>. = ratio of investment to GNP needed for 5 percent GNP growth rate k'f "I Iu/Vo = investment/GNP ratio in 1962 i annual growth rate of investment = 1962 ratio of saving to GNP Cto a' = marginal saving/GNP ratio E./M" = 1962 ratio of exports to imports e = export growth rate 1" /1'0 ratio of marginal to average import/GNP coefficients FoIVo = ratio of capital inflow to GNP in 1962 * Do not meet either set of investment criteria. a. Country numbers correspond to those in table 10·7. b. Criteria: (a) Saving criteria: IX':='" kf, or 1X0 kr, where f 0.05 (b) Trade criteria: 1" (Eo/Mo)(l+e)P-l - ~ , for some p ~ 50 years where f = 0.05 1'0 (1 +n p - 1 (c) Minimnm investment criteria: The initial investment rate of the countries marked by an asterisk is insufficient to maintain a percent GNP growth rate, even if the capital·output ratio were to fall to 3.0 (i.e. la/V, < 0.15), and the rate of increase in investment is insufficient to ever achieve a 5 percent GNP growth rate (i 0.05). < 418 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT growth through a combination of improved country performance and additional external resources. \\1 e therefore specify a considerable range of performance possibilities, based on the preceding survey of historical performance. The range of values chosen for each parameter is designed to show the extent to which the performance variables affect the country's growth and its aid requirements. METHODOLOGY. 'The mcthodology to be used follows closely that used for Pakistan on pages 399-404. A similar range of variation in performance has been specified for each of the fifty countries in the sample. Principal attention has been given to the twenty-five countries having the largest effect on assistance requirements. As a starting point, we estimated the six parameters in model 1 from the historical performance in each country, modified in some cases by the experience of similar countries. 'f11e average of the resulting target growth rates for all countries (projected to 1975) is 4.4 percent, ap- proximately the same as the recent past. 49 To evaluate the possibilities for accelerated growth, we divided the six policy parameters into three groups: the growth limits for invest- ment and GNP (f3 and f); internal performanee factors (k, (x', and p.'); and export growth (£). Starting from the historical estimates, we speeified two sets of more optimistic assumptions for growth limits and internal performance factors and one alternative set of export projections.50 These alternative sets of parameter values are shown in table 10-7. The possible combinations of the sets of values for the parameters provide a basis for eighteen projections for eaeh of the fifty countries. In judging the range of possible performanee for each country, we took into account its historical performance, its development plan, the observed performance of other countries and some aspects of political performance. \Ve relied heavily on the development programs of the major eountries in making the intermediate or "plan" estimates of both growth targets and internal performance. Plan targets and per- 49. These historically based estimates are shown in table 10-7. They differ from the parameters for 1957-62 in Chenery and Strout (1966, table A-I) primarily in the elimination of abnormal or biased values that need not persist with reasonable policies-for example, falling export and saving rates, abnormally high capital coefficients, and so forth. In large part, these abnormal values represent the effects of disequilibrium conditions on our estimates. The revisions reduce the estimates of aid requirements. 50. Details are given in Chenery and Strout (1965). PROSPECTS FOR THE TRANSITION 419 formance are defined here as those aehievable with moderate improve- ments in development policies in relation to past experience. The most optimistic (upper-limit) estimates assumed that almost all eountries could attain the median observed value of the marginal saving rate (0.20) and could limit the marginal import coefficient to the normal value derived from intercountry comparisons. Our notion of the upper limit implies a probability of perhaps one in four that the given target growth and performance could be at- tained. For all countries, the average of the plan growth targets through 1975 turns out to be 5.2 percent and the average of the upper- limit targets is about 6 percent. TIle "plan" estimates range from 3 percent to 9 percent, with a heavy concentration between 5 percent and 7 percent51 To explore the range of growth possibilities systematically, we have adopted the same degree of optimism for all countries in each trial calculation. Projections on this basis are designed to reveal the range of possibilities that is interesting for poliey purposes rather than to forecast the most probable course of development in each country. 111e projections were made from year to year according to the formulas of the appropriate phase in model 1.52 Cumulative results for the eighteen combinations of growth targets, country performance, and exports are given in table 10-9 and regional projeetions for 1970 and 1975 in table 10-10. The projections in table 10-9 include measures of excess consump- tion and excess imports, which show the extent to which aid require- ments eould be reduced through policies designed to equalize the two resource gaps. Since the empirical possibilities for such policies eannot be ascertained without detailed studies of each country, we shall apply the overall factors derived on pages 404-09 to estimate the reduction in capital inflow that might be achieved in this way. THE PHASES OF GROWTII. The projection of growth paths under alternative assumptions provides a more general evaluation of the relative importance of the two resource limitations than does our 51. Whatever the validity of our subjective judgments about the possibility of improved performance, this procedure has seemed preferable to a more mechanical approach to testing the sensitivity of the results to various types of change. Our principal conclusions are not greatly affected by differences in judgment about the possibilities for individual countries. 52. Machine computations involve a test in each year to determine the appro- priate growth phase and set of equations to apply for the next year. 420 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-7. V dlue of Parameters Used in Pro;ections Target growth rate Maximum rate of of GNP (f) gmwth of investment (H) is- (U)pper No. Country torical (P)lan Limit H P U Near East 2 Cyprus 0.009 0.030 0.050 0.060 0.060 0.070 3 Greece 0.060 0.065 0.070 0.100 0.l00 0.l00 5 Iran 0.044 0.055 0.065 0.060 0.070 0.090 6 Israel 0.090 0.090 0.100 0.120 0.120 0.150 7 Jordan 0.056 0.056 0.080 0.160 0.160 0.160 9 Turkey 0.053 0.060 0.070 0.080 0.080 0.090 10 U.A.R: 0.045 0.055 0.060 0.050 0.070 0.080 South Asia 1 Cevlon b 0.042 0.050 0.060 0.095 0.100 0.100 4 India 0.043 0.053 0.065 0.100 0.100 0.100 8 Pakistan' 0.045 0.053 0.060 0.130 0.130 0.130 Latin America 11 Argentina 0.031 0.043 0.055 0.150 0.150 0.150 12 Bolivia 0.033 0.045 0.056 0.060 0.080 O.OSC 13 Brazil 0.055 0.055 0.070 0.080 0.080 0.08C 14 British Guiana 3 0.029 0.040 0.050 0.100 0.100 OJOC 15 Chile 0.035 0.050 0.055 0.060 0.080 O.lOC 16 Colombia 0.050 0.061 0.070 0.060 0.080 OJOC 17 Costa Rica 0.055 0.060 0.069 0.060 0.080 0.10( 18 Ecuador 0.042 0.050 0.055 0.060 0.080 0.08( 19 EI Salvador 0.050 0.060 0.065 0.060 0.100 O.lO( 20 Guatemala 0.040 0.050 0.055 0.060 0.080 O.OSC 21 Honduras 0.037 0.045 0.050 0.064 0.070 O.OSC 22 Jamaica 0.040 0.045 0.055 0.060 0.080 O.lO( 23 Mexico 0.050 0.060 0.070 0.067 0.080 O.lO( 24 Nicaragua 0.042 0.050 0.055 0.063 0.080 O.OS( 25 Panama 0.050 0.050 0.060 0.144 0.140 O.l4( 26 Paraguay 0.020 0.030 0.040 0.060 0.140 O.l4! a. Including Syria. b. Now Sri Lanka. c. Including Bangladesh. d. Now Guyana. PROSPECTS :FOR THE TRANSITION 421 Table 10-7 (continued) Target erowth rate Ma.ximum rate of of GNP (1) growth of investment (H) is- (U)pper No. Country torical (P)lan limit H P U 27 Peru 0.055 0.055 0.070 0.100 0.100 0.100 2& Trinidad and Tobago 0.050 0.060 0.088 0.090 0.100 0.100 29 Venezuela 0.045 0.060 0.070 0.080 0.080 0.100 Africa 30 Algeria 0.020 0.035 0.050 0.060 0.060 0.060 31 Ethiopia 0.045 0.045 0.050 0.150 0.150 0.150 32 Ghana 0.045 0.055 0.060 0.098 0.098 0.098 33 Kenya 0.017 0.035 0.050 0.060 0.060 0.060 34 Liberia 0.057 0.060 0.060 0.150 0.150 0.150 50 Mauritius 0.034 0.034 0.034 0.083 0.080 0.080 35 Morocco 0.028 0.040 0.060 0.0.50 0.060 0.070 36 Nigeria 0.040 0.045 0.050 0.082 0.082 0.082 37 Rhodesia- Nyasaland 0.043 0.040 0.045 0.060 0.060 0.060 38 Sudan 0.051 0.055 0.055 0.140 0.140 0.140 39 Tanganyika" 0.042 0.050 0.056 0.060 0.060 0.080 W Tunisia 0.041 0.050 0.060 0.150 0.150 0.150 U Uganda 0.017 0.040 0.050 0.060 0.060 0.080 Far East .2 Burma 0.032 0.040 0.050 0.060 0.060 0.060 H Indonesia 0.010 0.030 0.040 0.010 0.035 0.050 .5 Korea, Republic of 0.043 0.050 0.060 0.050 0.060 0.080 ~9 Malaya! 0.040 0.050 0.060 0.119 0.120 0.120 ~ Philippines 0.050 0.055 0.060 0.051 0.060 0.070 f3 Taiwan 0.060 0.070 0.080 0.133 0.133 0.133 '7 Thailand 0.050 0.060 0.065 0.091 0.091 0.091 ·8 Vietnam (South) 0.029 0.035 0.040 0.060 0.060 0.060 e. Now Tanzania excluding Zanzibar. £. Now Malaysia excluding Sabah and Sarawak. (table continues on the following pages) 422 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-7 (continued) -- ..- - Incremental aggregate Marginal gross capital-output savings ratio ratio No. Country A B C A B C Near East 2 Cyprus 5.00 4.00 3.50 0.140 0.190 0.230 3 Greece 3.10 3.10 3.10 0.230 0.230 0.250 5 Iran 3.70 3.70 3.50 0.140 0.150 0.250 6 Israel 3.19 3.00 3.00 0.220 0.300 0.300 7 Jordan 3.37 3.37 3.37 0.200 0.200 0.250 9 Turkey 2.91 2.91 2.91 0.200 0.256 0.256 10 U.A.R: 2.68 2.68 2.68 0.150 0.170 0.200 South Asia I Cevlon b 3.24 3.24 3.24 0.110 0.150 0.200 4 India 3.20 3.20 3.20 0.180 0.2lO 0.250 8 Pakistan" 3.00 3.00 3.00 0.160 0.240 0.240 Latin America 11 Argentina 7.21 5.30 4.30 0.220 0.220 0.250 12 Bolivia 4.00 4.00 4.00 0.100 0.150 0.200 13 Brazil 2.90 2.90 2.50 0.270 0.270 0.280 14 British Guiana d 5.00 5.00 5.00 0.200 0.250 0.330 15 Chile 3.40 3.40 3.00 0.120 0.160 0.200 16 Colombia 4.80 4.80 4.80 0.200 0.260 0.300 17 Costa Rica 3.27 3.27 3.27 0.130 0.200 0.250 18 Ecuador 3.74 3.74 3.74 0.140 0.200 0.240 19 EI Salvador 2.50 2.50 2.50 0.110 0.180 0.210 20 Guatemala 4.67 3.50 3.50 0.150 0.200 0.250 21 Honduras 3.90 3.50 3.50 0.120 0.150 0.200 22 Jamaica 4.00 3.50 3.50 0.160 0.180 0.200 23 Mexico 2.52 2.52 2.52 0.170 0.170 0.22C 24 Nicaragua 3.72 3.72 3.72 0.150 0.200 0.22C 25 Panama 2.50 2.50 2.50 0.120 0.200 0.20C 26 Paraguay 5.00 4.00 4.00 0.130 0.130 O.l5C a. Including Syria. h. Now Sri Lanka. e. Including Bangladesh. d. Now Guyana. PROSPECTS FOR THE TRANSITION 423 Table 10-7 (continued) I ncremental aggregate Marginal gross capital-output savings ratio ratio No. Country A B C A B C 27 Peru 4.94 4.94 4.77 0.285 0.285 0.185 28 Trinidad and Tobago 3.65 3.65 3.65 0.200 0.250 0.250 29 Venezuela 3.64 3.64 3.64 0.290 0.290 0.290 Africa 30 Algeria 3.30 3.30 3.30 0.060 0.100 0.200 31 Ethiopia 2.50 2.50 2.50 0.140 0.170 0.200 32 Ghana 3.70 3.50 3.20 0.130 0.150 0.200 33 Kenya 5.00 4.00 4.00 0.120 0.150 0.200 34 Liberia 5.00 5.00 5.00 0.110 0.150 0.200 50 Mauritius 4.97 4.97 4.97 0.080 0.080 0.080 35 Morocco 5.00 4.00 3.50 0.130 0.150 0.200 36 Kigeria 3.80 3.80 3.80 0.090 0.110 0.200 37 Rhodesia- Nyasaland 5.00 5.00 5.00 0.160 0.180 0.200 38 Sudan 2.50 2.50 2.50 0.110 0.150 0.200 39 Tanganyika· 2.93 2.93 2.93 0.110 0.150 0.200 40 Tunisia 4.62 4.00 4.00 0.150 0.200 0.250 41 Uganda 5.00 4.00 4.00 0.110 0.150 0.200 Far East 42 Burma 5.00 4.00 4.00 0.160 0.180 0.200 44 Indonesia 2.75 2.75 2.75 0.050 0.100 0.150 45 Korea, Republic of 3.27 3.27 3.27 0.100 0.150 0.200 49 Malaya' 2.52 2.50 2.50 0.190 0.200 0.200 46 Philippines 2.58 2.58 2.58 0.260 0.260 0.260 43 Taiwan 2.62 2.62 2.62 0.210 0.210 0.250 47 Thailand 2.50 2.50 2.50 0.250 0.250 0.250 48 Vietnam (South) 3.69 3.70 3.70 0.000 0.100 0.150 e. Now Tanzania excluding Zanzibar. f. Now Malaysia excluding Sabah and Sarawak. (table continues on the following pages) 424 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-7 (continued) -- .. Annual growth rates Marginal . -.. ~ import ratio Exports Popu- lation No. Country (a) (b) (c) 1 2 (1963) --~.-.- .. Near East 2 Cyprus 0.470 0.470 OAI0 0.0088 0.0116 0.017 3 Greece 0.190 0.190 0.180 0.0544 0.0712 0.009 5 Iran 0.232 0.232 0.180 0.0544 0.0712 0.025 6 Israel 0.400 0.400 0.300 0.1122 0.1468 0.035 7 Jordan 0.370 0.370 0.330 0.0571 0.0748 0.027 9 Turkey 0.110 0.170 0.110 0.0306 0.0400 0.029 10 U.A.R.· 0.200 0.200 0.150 0.0136 0.0178 0.025 South Asia 1 Cevlon b 0.220 0.220 0.190 0.0177 0.0231 0.028 4 India 0.070 0.070 0.050 0.0204 0.0267 0.024 8 Pakistan· 0.100 0.150 0.100 0.0374 0.0489 0.026 IJatin America 11 Argentina 0.070 0.170 0.020 0.0286 0.0374 0.017 12 Bolivia 0.220 0.220 0.220 0.0068 0.0089 0.023 13 Brazil 0.090 0.090 0.070 0.0286 0.0374 0.031 If British Guiana d 0.470 0.470 0.470 0.0544 0.0712 0.030 15 Chile 0.120 0.190 0.120 0.0190 0.0249 0.023 16 Colombia 0.200 0.230 0.170 0.0272 0.0356 0.029 17 Costa Rica 0.280 0.280 0.280 0.0354 0.0463 0.039 18 Ecuador 0.206 0.206 0.206 0.0340 0.0445 0.030 19 El Salvador 0.268 0.268 0.210 0.0374 0.0489 0.027 20 Guatemala 0.149 0.149 0.149 0.0340 0.0445 0.030 21 Honduras 0.195 0.195 0.195 0.0190 0.0249 0.030 22 Jamaica 0.206 0.206 0.206 0.0340 0.0445 0.013 23 Mexico 0.110 O.llO 0.100 0.0537 0.0703 0.031 24 Nicaragua 0.281 0.281 0.281 0.0340 0.0445 0.034 25 Panama 0.385 0.385 0.350 0.0143 0.0187 0.030 26 Paraguay 0.249 0.249 0.249 0.0054 0.0071 0.022 a. Including SYTia. b. Now Sri Lanka. c. Including Bangladesh. d. Now Guyana. PROSPECTS FOR THE TRANSITION 425 Table 10-7 (continued) Annual growth rates Marginal import ratio Exports Popu- lation No. Country 1 2 (1963) - .. ---~-.----~ (a) _ (b) ._.... _ .... (c) .... ~-~ .. 27 Peru 0.240 0.240 0.200 0.0524 0.0685 0.023 28 Trinidad and Tobago 0.700 0.500 0.300 0.0952 0.1246 0.030 29 Venezuela 0.314 0.180 0.160 0.02II 0.0276 0.034 Africa 30 Algeria 0.200 0.200 0.090 0.0272 0.0356 0.025 31 Ethiopia 0.116 0.116 0.116 0.0544 0.0712 0.014 32 Ghana 0.220 0.220 0.220 0.0156 0.0205 0.025 33 Kenya 0.266 0.266 0.040 0.0340 0.0445 0.030 34 Liberia 0.573 0.573 0.573 0.0422 0.0552 0.015 50 Mauritius 0.456 0.456 0.456 0.0000 0.0000 0.032 35 Morocco 0.150 0.150 0.150 0.0204 0.0267 0.027 36 Nigeria 0.280 0.280 0.280 0.0544 0.0712 0.020 37 Rhodesia- Nyasaland 0.513 0.513 0.220 0.0755 0.0988 0.028 38 Sudan 0.220 0.220 0.190 0.0612 0.0801 0.028 39 Tanganyika" 0.188 0.188 0.100 0.0333 0.0436 0.020 40 Tunisia 0.260 0.190 0.190 0.0340 0.0445 0.021 H Uganda 0.168 0.168 0.090 0.0272 0.0356 0.025 Far East 42 Burma 0.177 0.177 0.177 0.0782 0.1023 0.022 44 Indonesia 0.070 0.070 0.070 0.0109 0.0142 0.023 45 Korea, Republic of 0.240 0.260 0.180 0.0578 0.0756 0.029 19 Malaya' 0.419 0.419 0.419 0.0211 0.0276 0.032 k5 Philippines 0.170 0.170 0.170 0.0313 0.0409 0.032 B Taiwan 0.205 0.190 0.190 0.0544 0.0712 0.029 ~7 Thailand 0.160 0.160 0.150 0.0462 0.0605 0.031 ~8 Vietnam (South) 0.217 0.217 0.217 0.0252 0.0329 0.028 ... --~ ... e. Now Tanzania excluding Zanzibar. f. Now Malaysia excluding Sabah and Sarawak. 426 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table lO-8. Proportion of Countries with Foreign Capital Requirements Determined by Investment-Saving Gaps (percent) I ntcrnal performance characteristics Historical Plan Upper limit GNP growth Low High Low High Low High targets exports exports exports exports exports exports 1965 Historical 28 40 22 24 18 24 Plan 52 62 32 46 34 44 Upper Limit 72 80 54 70 48 58 1975 Historical 32 40 20 34 ]8 24 Plan 38 58 24 36 18 30 Upper Limit 50 68 30 48 22 40 Source: USAID, Office of Program Coordination, "Z3-Y car Projections" of Sep- tember 16, 1964, for modell, fifty-country sample. attempted identification of these limits in current situations. Table lO-8 shows the proportion of the fifty countries in which the saving- investment gap was the limiting factor-and hence the determinant of capital inflow-in each of the eighteen trial projections. The most striking result of this tabulation is the predominance of the trade limit; it is more important than the saving limit in 1975 in fifteen of the eighteen sets of alternatives. This breakdown shows the quantitative significance of three factors that have been discussed previously in general terms. (a) At higher growth rates the saving limit tends to become more important, for reasons analyzed on pages 404-12. Under most as- sumptions about the other parameters, a rise in the growth rate from the historical average of 4.4 percent to the upper-limit average of about 6 percent increases the number of countries in which the saving limit is controlling by 50 percent or more. (b) The saving limit is increasingly dominated over time by the trade limit under historical conditions of internal performancc. This result points to the need for more import substitution unless export prospects can be drastically improved. (c) A 40-percent increase in the assumed rates of growth of exports (from the low to the high assumptions) removes the trade limit in only four to six of the fifty countries under most assumptions. Un- PROSPECTS FOR THE TRANSITION 427 realistically large increases in exports would be required to reduce the importance of the balance of payments limitation greatly by 1975. Development performance and assistance needs The projections in table 10-9 are designed to show the way in which assistance needs vary with the export possibilities and internal performance of the developing countries. For this purpose the 900 separate country projections have been aggregated using the same degree of optimism as to exports and internal performance for each country. To summarize the results graphically, the 18 aggregate solu- tions of table 10-9 are plotted in figure 10-4, giving three points on Figure 10-4. Foreign Capital Requirements of Fifty Developing Countries: Alternative Development Patterns, 1962-75 (billions of 1962 U.S. dollars) Average GNP growth rate, 1962'-7') 0.050 200 150 .s Country a performance '" u Q assumptions l>O .§ "Historical" ..8 t' 100 "Plan" Q ~ "Upper limit" - - IJower export assumption (3.8 percent annual growth through 1975) - Higher export assumption (4.8 percent 50 annual growth through 1970; 5.2 per· cent through 1975) Gross national product in 1975 428 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-9. Aggregate Projections for 1962-75 (billions of 1962 U.S. dollars; cumulative values include the years 1962 through 1975) Historical country performance His- Upper- torical Plan limit Line growth growth growth No. targets targets targets 1 1975 Gross National Product 297 327 354 2 (Implicit GNP growth rate) (0.044 ) (0.051 ) (0.058) Cumulative values of variables 3 Exports: low growth (3.8 percent a year) 441 441 441 4 Exports: high growth (5.2 percent a year) 480 480 480 5 Gross national product 3,186 3,356 3,485 6 Gross investment 476 591 703 7 National savings (potential) (475) (502) (526) 8 National savings, realized 365 435 491 9 Imports (potential) (533) (561) (581) lO Imports, realized 552 596 652 II Unrealized savings (line 7 minus 8) llO 67 35 12 Excess imports (line 10 minus 9) 19 35 71 13 Total unrealized savings and excess imports (line 11 plus 12) 129 101 106 14 Net capital inflow: low exports HI 156 212 15 Net capital inflow: high exports 90 138 201 Capital inflow, excluding countries with net capital outflow': 16 Low exports 120 165 220 17 High exports 105 150 211 18 Consumption: low exports 2,821 2,920 2,995 19 Consumption: high exports 2,800 2,903 2,984 Source: USAID, Office of Program Coordination, machine listings of September 16, 1964 Note: See table 10-7 for values of parameters used. a. Foreign resource flows are measured on a net basis. In any particular year most nel flows are capital inflows, but some countries {for example, Venezuela, Malaya, and Burma: may have estimated potential capital outflows under the assumptions made. The alternativ( net capital estimate shown here excludes these potential capital outflows. PROSPECTS )"OR THE TRANSITION 429 Plan country Upper-limit country performance performance His- Upper- His- Upper- torical Plan limit torical Plan limit growth growth growth growth growth growth targets targets targets targets targets targets 297 328 356 298 329 360 (0.044 ) (0.052) (0.058) (0.044 ) (0.052) (0.059) 441 441 441 441 441 441 480 480 480 480 480 480 3,188 3,363 3,502 3,195 3,373 3,522 461 572 684 448 557 670 (495) (528) (558) (521) (562) (598) 353 430 .505 364 451 538 (533) ( 562) (583) (509) (541) (560) 548 582 620 525 547 573 142 98 53 157 111 60 15 20 36 16 6 23 157 118 89 173 117 82 108 142 179 84 106 133 83 119 164 60 84 116 106 152 187 94 117 141 100 131 173 85 99 125 2,835 2,933 2,997 2,831 2,922 2,984 2,811 2,909 2,981 2,807 2,899 2,967 430 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-10. Regional Projections, 1970 and 1975 (billions of 1962 U.S. dollars) 1962 1970' Targets and Histori· Upper performance cal Plan limit Exports Low Low High High Near East (7 countries)" GNP 20.94 32.05 33.22 33.22 31.48 (Percentage of total GNP) (78) Investment 3.76 5.56 6.27 6.27 7.27 Savings 2.51 3.26 3.45 3.92 5.18 Imports 4.61 7.28 7.80 8.02 7.76 Exports 3.34 4.98 4.98 5.67 5.67 Foreign resources' 1.26 2.30 2.82 2.35 2.09 South Asia (3 countries) GNP 46.22 64.83 69.52 69.52 73.45 (Percentage of total GNP) (97) Investment 7.57 8.89 11.66 11.66 14.93 Savings 6.46 6.85 9.08 9.26 12.83 Imports 3.73 5.20 5.74 5.74 5.45 Exports 2.63 3.16 3.16 3.35 3.35 Foreign resources' 1.10 2.04 2.58 2.39 2.10 I.atin America (19 countries) GNP 62.64 88.60 93.44 93.44 98.56 (Percentage of total GNP) (95) Investment 11.05 15.04 17.61 17.61 20.19 Savings 10.26 13.28 14.95 15.82 18.39 Imports 11.00 15.23 16.12 16.48 16.50 Exports 10.20 13.46 13.46 14.69 14.69 Foreign resources' 1.55* 2.21 * 3.27* 2.58* 2.27 Africa (13 countries) GNP 17.04 22.28 23.58 23.58 24.56 (Percentage of total GNP) (73) Investment 2.69 2.97 3.61 3.61 4.25 Savings 1.58 1.65 2.07 2.20 2.85 Imports 5.49 7.42 7.63 8.16 8.12 Exports 4.38 6.08 6.08 6.75 6.75 Foreign resources' 1.11 1.36* 1.53 1.40 1.4( PROSPECTS FOR THE TRANSITION 431 ~~ ... -~ ..- -..- -..- - . Annual growth, 1962-75' 1975" (percentage) Histori- Upper Histori- Upper cal Plan limit cal Plan limit .. __. - Low Low High High Low Low Low High High 42.12 44.92 44.92 48.55 5.5 6.0 6.7 7.46 8.70 8.70 10.63 5.4 6.7 8.3 4.35 4.48 5.49 7.73 9.71 10.82 11.53 11.23 6.60 6.60 8.32 8.32 5.4 5.4 7.3 7.3 3.11 4.22 3.21 2.90 7.2 9.7 7.5 6.6 80.13 89.96 89.96 100.20 4.3 5.3 6.1 10.99 15.08 15.08 20.37 2.9 5.4 7.9 8.12 11.12 11.47 17.27 6.42 7.52 7.52 7.02 3.56 3.56 3.91 3.91 1.0 1.0 3.1 3.1 2.86 3.96 3.61 3.11 7.6 10.4 9.6 8.3 11.01 121.56 121.56 134.42 4.5 5.2 6.1 18.85 23.13 23.l3 28.49 4.2 5.8 7.5 16.l3 18.55 20.24 25.92 18.92 20.79 21.74 21.40 16.19 16.19 18.83 18.83 3.6 3.6 4.8 4.8 2.87* 5.00* 3.80* 3.23* 4.9 9.4 7.l 5.6 26.56 29.11 29.11 31.36 3.5 4.2 4.8 3.63 4.55 4.55 5.59 2.3 4.1 5.8 2.13 2.68 2.94 4.07 9.09 9.45 10.68 10.59 7.59 7.59 9.07 9.07 4.3 4.3 5.8 5.8 1.53* 1.86* 1.61* 1.58* 2.5 4.0 2.9 2.8 (table continues on the following page) 432 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Table 10-10 (continued) (billions of 1962 U.S. dollars) 1962 1970' Targets and Histori- Upper performance cal Plan limit Exports Low Low High High Far East (8 countries) GNP 23.68 30.92 33.53 33.53 34.68 (Percentage of total GNP) (86) Investment 3.11 3.11 4.28 4.28 5.01 Savings 2.26 1.80 2.65 3.00 3.57 Imports 4.80 6.41 6.73 6.82 6.98 Exports 3.95 5.10 5.10 5.54 5.54 Foreign resources" 0.95* 1.31 1.64* 1.31 * 1,45' 50-country total GNP 170.52 238.68 253.31 253.31 262.72 (Percentage of all LDC'S GNP)d (89) Investment 28.18 35.59 43.45 43.45 51.65 Savings 23.07 26.84 32.20 34.20 42.82 Impocts 29.63 41.54 44.02 45.22 44.81 Exports 24.50 32.81 32.81 36.00 36.00 Foreign resources" 5.97* 9.22* 11.84* 10.03* 9.31 a. See table 10-9 for sources and concepts. Column headings identify both targets an performance standards. b. Excludes the oil-producing countries except for Iran. Includes Greece, Turkey, and th U.A.R. PROSPECTS FOR THE TRANSITION 433 Annual growth, 1962-75" 1975" (percentage) Histon- Upper Histori- Upper cal Plan limit cal Plan limit Low Low High High Low Low Low High High 36.91 42.07 42.07 44.97 3.5 4.5 5.1 3.86 5.50 5.50 6.85 1.7 4.5 63 2.14 3.10 3.79 4.76 7.78 8.46 8.70 9.08 6.06 6.06 6.99 6.99 3.3 33 4.5 4.5 1.7l* 2.41* 1.78* 2.11 * 4.6 7.4 4.9 6.3 96.73 327.62 327.62 359.50 4.4 5.2 5.9 44.79 56.96 56.96 71.93 3.6 5.6 7.5 32.87 39.93 49.93 59.75 2.8 4.3 5.1 7.6 51.92 57.04 60.17 59.32 4.4 5.2 5.6 5.5 40.00 40.00 47.12 47.12 3.8 3.8 5.2 5.2 -_ l2.08* 17.45* .... _ _ ... 14.01* ]2.93* 5.6 7.8 6.6 5.9 c. The figures shown here exclude potential capital outflows (see table 10·9, note a); gional totals are denoted by an asterisk (*) if they include such a ease. d. Excludes Cuba, Mainland China, and North Korea. 434 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT each of the six curves. Curve P2, for example, shows the increase in cumulative capital inflow from $100 billion ($7.7 billion a year) needed to sustain an average growth rate of 4.4 percent to $173 billion ($13 billion a year) to sustain a growth rate of 5.8 percent, assuming plan performance and high exports. The graphic prescntation facilitates analysis of the productivity of extcrnal assistance under alternative assumptions. Along curve P2 an increase of a billion dollars of GNP in 1975 can be secured for a cumula- tive aid input of $1.2-$1.3 billion between 1962 and 1975.53 The pro- ductivity of aid is significantly lower with historical performance and significantly higher with upper-limit performance, as shown by the relative slopes of the curvCS. 54 Variation in export optimism affects the level of total aid but not its marginal productivity. The effects of individual elements can be isolated in table 10-9. Starting from the central estimate of $131 bi1lion in capital imports for plan growth, high exports, and plan performance, we can identify the following effects of ehangcs in different sets of policy variables: (a) A decrease in export growth from 5.2 percent to 3.8 percent causes a reduction of exports of $39 billion and an increase of total capital inflow of $21 billion. (b) An increase in internal performance to the upper limit (with a constant growth rate) causes a reduction of capital inflow by $32 billion. (c) A fall in the growth rate to 4.4 percent (with no change in in- ternal performance) causes a reduction in external capital require- ments of $31 billion and of consumption by $98 billion. The relative importance of these changes varies with the starting point and depends largely on which of the growth limits predominates. At the upper limit growth rates, where the saving constraint is more important, the effect of increasing exports on aid requirements is less.55 53. An alternative calculation would show a cumulative addition to GNP over the period of nearly five dollars for each dollar of capital inflow. There is some decline in aid productivity at higher growth rates due to the shifting of countries from phase III to phase II. 54. The marginal productivity of aid in the three high export cases is 0.54 for historical performance, 0.81 for plan performance, and 1.54 for upper-limit per· formance. Corresponding values derived above from equations (10.18) and (10.19) using median observed values of the parameters were 0.35 for phase II and 0.91 for phase III. 55. This effect is more pronounced at low growth rates if we do not exclude countries having capital exports. PROSPECTS FOR THE TRANSITION 435 At plan growth rates, about half of any increase in exports is reflected in a reduced need for external resources in the model 1 solutions, since the external requirements of countries in phase II are not affected. Perhaps the most notable feature of this analysis is the sensitivity of aid requirements to variations in internal performance. At historical growth rates, the maximum reduction due to improved performance is about 20 percent, but at the 6 percent growth rate, upper-limit per- formance would reduce external capital needs by 40 percent. Put in other terms, the capital inflow required to sustain 4.4 percent growth with historical performance would sustain 5.4 percent growth if all countries could achieve the upper-limit standards. The main cause of the greater sensitivity at higher growth rates is the greater importance of saving out of increased income as GNP grows. 111is sensitivity would be even more pronounced if we assumed that saving depends on per capita rather than total income levels. To compare our results to other estimates, we can state them in terms of the net capital inflow in 1970 and the implied increase in external assistance between 1962 and 1970. Omitting the less likely combinations of assumptions, the indicated range of capital require- ments in 1970 is from $10-$17 billion,56 corresponding to rates of growth of external capital of 3 percent to 10 percent from its $7.4 billion value in 1962.57 Tl1is range compares to the UN estimate for 1970 of $20 billion and to Balassa's (1964) range of $9-$12 billion. Our estimates have the advantage of making explicit assumptions as to country performance and of showing how the total depends on them. The possibility of further reduction in assistance needs through better development policies is indicated in table 10-9 by the magni- tude of the excess imports for countries in phase II and unrealized saving for countries in phase III. \Vith the moderate improvement of performance that is represented by the plan growth targets and plan 56. The principal combinations for 1970 are given in table 10-10 and the full range for 1975 in table 1O·1l. 57. We have lIsed a factor of 1.25 to convert our sample results to the require· ments of all less developed countries. The 1962 figure of $7.4 billion is based on balance of payments figures in United Nations (1964, tables 1 and 11), and is lower than the OEeD estimate of $8.5 billion of capital inflow in the same year. It includes eapital flows to Turkey and Greece, and excludes Puerto Rico and $0.7 billion of capital outflows from major oil exporters. The discrepancies between the UN and OEeD estimates are discussed in United Nations (1965, annex, pages 6-8). 436 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT performance, there would be $98 billion of unrealized saving and $20 billion of excess imports. The predominant need is to convert the unrealized saving into additional investment which will substitute for imports or increase exports. The theoretical limits to the possibilities for reducing aid require- ments in this way are shown in table 10-10. As explained on pages 404- 09, the efficiency of the import substitution mechanism in converting surplus saving into a reduction in capital requirements may be on the order of 50 percent to 65 percent under plausible assumptions. More massive import substitution would raise the marginal capital co- efficient for thc additional production and thus lower the possibilities for efficient reductions in external capital. To illustrate the extent to which further import substitution or addi- tional exports affects assistance requirements, we have made a set of projections with model 2 on the assumption that not more than 10 percent of total investment in each country could be devoted to this purpose. The results are shown in table 10-11. l11ey suggest that opti- Table 10-11. Comparison of Modell and Model 2 Projections of 1975 Capital Inflow (billions of 1962 U.S. dollars) Internal performance characteristics GNP Historical Plan Upper limit growth targets Low High Low High Low High exports exports exports exports exports exports A1odell-Projections Historical 12.1 10.0 ll.8 9.0 9.4 7.1 Plan 18.7 16.2 17.4 14.0 12.0 9.2 Upper Limit 26.l 24.2 22.5 19.6 15.9 12.9 Model 2-Optimal adjustment' Historical 7.9 6.1 7.7 5.4 5.5 3.5 Plan 11.1 10.3 10.3 6.9 5.5 2.9 Upper Limit 20.8 20.8 14.1 13.7 7.6 6.0 .. -~.-~. .. ~- .. -~ .. ~ Source: USAlD, Office of Program Coordination, "23-year Projections," machine listings of September 16, 1964. Note: External capital requirements exclude negative flows (eapital outflows) from countries estimated to be net potential capital exporters by 1975. a. Estimated by assuming conversion of "excess imports" (table 10-9) to addi- tional import substituting investment in amounts not exeeeding 10 percent of total investment estimated for equivalent model I development alternative. The 1975 external capital "savings" under this assumption range from $3.4 to $9.5 billions. INTERNATIONAL ASSISTANCE POLICIES 437 mum planning for structural change might reduce requirements for external capital in 1975 by a third or more at plan growth rates com- pared to the more rigid trade assumptions of model I. Some of the regional implications of the projections are brought out in table 10-10. The regional growth rates corresponding to the average plan target of 5.2 percent vary from 4.2 percent for the sample of African countries to 6.0 percent in the Near East. South Asia shows the most rapid increase in capital inflow relative to its growth in GNP, reflecting the relatively high absorptive capacity assumed and rela- tively low initial savings rates. Perhaps more significant than the actual estimates is the demonstration that the allocation of external assis- tance in accordance with comparable standards of performance would be likcly to result in substantial shifts in the regional distribution of foreign assistance. International Assistance Policies The analysis in this chapter has shown the conditions under which external assistance may make possible a substantial acceleration in the process of economic development. Our focus has been on the inter- relations among external resource requirements and the development policies of recipient countries. Analysis of these relations leads to several principles of general applicability to international assistance policy. 'The central questions for assistance policy are the measurement of the effectiveness of external assistance, the policies that recipient countries should follow to make best use of external resources, and the basis for allocating assistance among countries. This section sum- marizes the main implications of our analysis for each of these ques- tions and adds some qualitative elements that have been omitted from the formal analysis. The effectiveness of assistance In the short run the effectiveness of external resources depends on their use to relieve shortages of skills, saving, and imported commod- ities. The productivity of additional amounts of assistance over short periods can be measured by the increase in output resulting from the fuller use of domestic resources which they make possible. 438 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT Over longer periods, the use that is made of the initial increase in output becomes more important. Even if the short-run productivity of aid is high, the economy may continue to be dependent on external assistance indefinitely unless the additional output is allocated so as to increase saving and reduce the trade gap. Over the whole period of the transition to self-sustaining growth, the use that is made of the succes- sive increments in GNP is likely to be more important than the effi- ciency with which extemal assistance was used in the first instance. This point was demonstrated in the discussions of figure 10-3, which showed the dependence of total aid requirements on the marginal saving rate. To emphasize this point, assume that the productivity of investment in the first five years of the upper-limit development se- quence outlined above for Pakistan had been one-third lower, requir- ing a correspondingly larger amount of investment and extemal aid to achieve the same increase in GNP. The effect would be to increase the total aid required over the seventeen-year period to achieve self- sufficiency by some 45 percent This, however, is less than the effect on aid requirements of a reduction in the marginal saving rate from 0.24 to 0.22. The critical elements in the development sequence are getting the initial increase in the rate of growth, channeling the increments in income into increased saving, and allocating investment so as to avoid balance-of-payments bottlenecks. These long-run aspects are 1ikely to be considerably more important than the efficiency \vith which ex- ternal capital is used in the short run.58 TIle long-run effectiveness of assistance is also likely to be increased by supporting as high a growth rate as the economy can achieve with- out a substantial deterioration in the efficiency of use of capital. 'I11is conclusion was derived in Chenery and Strout (1965) and is elabo- rated in chapter 9 of this volume. There are also several factors omitted from the formal models that argue for more rapid growth: (a) the fact that a smaller portion of the increase in GNP is offset by population growth; (b) the gain in politieal stability and govern- mental effectiveness that is likely to result; (c) the greater likelihood of being able to raise marginal saving rates and export growth when GNP is growing more rapidly 59; and (d) the greater likelihood of 58. This conclusion is demonstrated in the evaluation of the effectiveness of aid to Greece by Adelman and Chenery (1966). 59. The advantages of more rapid growth with constant per capita marginal saving rates are demonstrated by Fei and Paauw (1965). INTERNATIONAL ASSISTANCE POLICIES 439 attracting foreign private investment to finance the needs for external capital. While the last three factors cannot be measured with any accuraey, they appear to have been important in most countries that are success- fully completing the transition, such as Israel, Greece, Taiwan, Mexico, and the Philippines. These examples support the theoretical conclusion that the achievement of a high rate of growth, even if it has to be initially supported by large amounts of external capital, is likely to be the most important element in the long-term effectiveness of assistance. The substantial increases in internal saving ratios that have been achieved in a decade of strong growth-from 7 percent to 12 percent in the Philippines, 11 percent to 16 percent in Taiwan, 6 percent to 14 percent in Greece, and -9 percent to 12 percent in Israel-demonstrate the speed with which aid-sustained growth can be transformed into self-sustained growth once rapid development has taken hold. GO Policies for recipient countries Although the receipt of external assistance may greatly reduce the time required for a country to achieve a satisfactory rate of growth, dependence on substantial amounts of external resources creates some special policy problems. One lesson from the preceding analysis is that the focus of policy should vary according to the principal limitations to growth. Just as optimal countercyclical policy implies different responses in different phases of the business cycle, optimal growth policy requires different "self-help" measures in different phases of the transition. In phase 1, where the growth rate is below a reasonable target rate, the focus of policy should be on increasing output, implying an in- crease in the quality and quantity of both physical capital and human resource inputs. Our statistical comparisons suggest that a rate of growth of investment of 10 percent to 12 percent is a reasonable target for countries whose initial investment level is substantially below the required leveL Phase I can be completed by most countries in a decade if this increase in investment is accompanied by sufficient improvement in skills and organization to make effective use of the 60. In 1979 one could add Korea, Tunisia, Colombia, Costa Rica, and others to this Jist. 440 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT additional capital that becomes available. Although it is probably more important in this phase to focus on securing increases in pro- duction and income, a start must also be made on raising taxes and saving if international financing is to be justified by performance. As phase I is completed, the rate of increase in investment can be allowed to fall toward a feasible target rate of GNP growth, which is unlikely to be more than 6 percent to 7 percent. The focus of develop- ment policy should then he increasingly on (a) bringing about the changes in the productive structure needed to prevent further increases in the balance of payments deficit, and (b) channeling an adequate fraction of increased income into saving. Although theoretical discus- sion has tended to stress the second requirement, the first appears to have been more difficult in practice for many countries. Since sub- stantial import substitution is required just to prevent the ratio of imports to GNP from rising, export growth at least equal to the target growth of GNP is likely to be necessary to reduce external aid. As the focus of development policy changes, the instruments of policy must change accordingly. Somewhat paradoxically, successful performance in phase I, which would justify a substantial and rising flow of foreign assistance, may make success in phase III more difficult. If investment and other allocation decisions are based on the exchange rate that is appropriate for a substantial flow of aid, they are not likely to induce sufficient import substitution or increased exports to make possible a future reduction in the capital inflow. Planning should be based on the higher equilibrium exchange rate that would be appropri- ate to a declining flow of aid for the necessary changes in the produc- tive structure to be brought about in time. It is the need for rapid structural change which sets the lower limit to the time required to complete the transition to self-sustaining growth. Even though the simplified model underlying figure 10-3 suggests the possibility of completing this transition in less than twenty years starting from typical Asian or African conditions, it is very unlikely that any such country can meet all the requirements of skill formation, institution building, investment allocation, and so forth in less than one generation. Policies for donor countries Donors are concerned with criteria for the al1ocation of aid among recipients and the means for controlling its use. Allocation and con- INTERNATIONAL ASSISTANCE POLICIES 441 trol policies are complicated by the mixture of objectives that motivate international assistance, the most important of which are (a) the eco- nomic and social development of the recipient, (b) the maintenance of political stability in countries having special ties to the donor, and (c) export promotion. This mixture of motives has led to a complex system of aid administration in all countries. The predominant basis for development loans is the individual in- vestment project, for which external financing is provided to procure capital goods from the donor country. Loans not limited to equipment for specific projects are provided to a few selected eountries against the balance-of-payments needs of development programs. S1 Substantial but declining amounts of grants are also furnished for budgetary support of ex-colonies and other dependent areas. Our analysis suggests some directions in which improvements can be sought in the existing methods of supporting economic develop- ment, which is the objective on which all parties agree. We consider first methods of transferring resources to individual countries and then allocation of assistance among countries. THE TRANSFER OF ASSISTANCE. Any system for transferring resources must include: (a) a basis for determining the amount of the transfer, (b) specification of the form of resources to be furnished, and (c) a basis for controlling their use. On all these counts the project system has the virtue of simplicity. It also provides for detailed evaluation of the investments that are directly financed from external aid-which may be 10 percent or so of total investment-and for increasing their productivity through technical review. Although the project system has much to commend it when the main focus is on increasing the country's ability to invest, it becomes increasingly inappropriate as the development process gets undcr way. As the rate of growth increases, we have shown that the effectiveness of aid depends more on the use that is made of the additional output than on the efficiency with which a limited fraction of investment is carried out. Furthermore, an attempt to finance thc amount of ex- ternal resources needed during the peak period of an optimal growth path-which may imply aid equal to 30 percent to 40 percent of total investment-by the project mechanism alone may greatly lower the efficiency of use of total resources. Limiting the form of assistance to 61. This procedure is often called program lending. 442 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT the machinery and equipment needed by substantial investment projects is likely either to lower the rate of growth or to distort the pattern of investment. In these circumstances, assistance is more effective if the range of commodities supplied is broadened to permit the recipient's pattern of investment and production to evolve in accordance with the principle of comparative advantage. G2 vVhile domestic supply can- and indeed must-lag behind demand in some sectors to accommodate the needed resource transfer, the country should also be preparing to balance its international accounts by the end of a specified transitional period. Since donors fear that uncontrolled imports may be wasted in in- creased consumption \\.~thout the restraints imposed by the project mechanism, an alternative means of control is needed. Part of the solution lies in relating the amount of aid supplied to the reCipient's effectiveness in increasing the rate of domestic saving, so that the added resources will necessarily increase saving and investment as income grows. As development planning and statistics on overall per- formance improve, this type of "program approach" is becoming increasingly feasiblc, both from the point of view of determining the amounts of assistance needed and of assessing the results.s3 The strongest argument for the program approach arises for coun- tries in phasc III, where the balance of payments is the main factor limiting growth and there is typically excess capacity in a number of productive sectors. In this situation, the highcst priority use of imports is for raw materials and spare parts to make more effective use of exist- ing capacity; project priorities should give primary weight to import substitution and increased exports. Donor controls should be primarily concerned with the efficient use of total foreign exchange resources, which can only be assessed adequately in the framework of a develop- ment program. ALLOCA1.10N OF ASSISTANCE. If the objectives of the donor countries could be expressed as some function of the growth of each recipient, it 62. This observation applies to aid in the form of agricultural commodities as well as to aid in the form of machinery or any other specified goods. 63. In the 1960s the U.S. government used the program approach in India, Pakistan, Turkey, Tunisia, Chile, Colombia, and Brazil. See USAID (1963) and (1965). The program approach has been adopted more widely by other donors since then. POSTSCRIPT 443 would be possible to allocate aid primarily on the basis of expected development performance. The varying political objectives of the donors complicate the problem because each would give somewhat different weights to a unit of increase in income as among recipients. Even with this limitation, however, there may be considerable scope for reallocating a given amount of aid or for selective increases in in- dividual country totals in accordance with criteria of self-help. The predominant project approach now in use favors countries whose project preparation is relatively efficient. Other qualities that are equally important to successful development-tax collection, pri- vate thriftiness, small-scale investment activity, export promotion-are ignored in focusing on this one among many aspects of better resource use. 64 Where fairly reliable statistics are available, an alternative procedure would be to establish minimum overall performance standards for each country and to share the aid burden among interested donors through a consortium or other coordinating mechanism. For example, a country starting in phase I might have as its principal performance criteria: (a) growth of investment at 10 percent per year at a mini- mum standard of productivity, and (b) the maintenance of a mar- ginal saving rate of 0.20 (or alternatively a specified marginal tax rate). There would be little possibility to waste aid on these terms, since the required increase in savings would finance a large proportion of total investment. Appropriate overall standards for saving rates and balance-of-payments policies for countries in phase II and phase III could also be established without great difficulty. A country main- taining high standards-say, a marginal savings rate of 0.25 and a marginal capital-output ratio of less than 3.3-could safely be allotted whatever amount of aid it requested in the knowledge that the larger the amount of aid utilized, the higher would be its growth rate, and the more rapid its approach to self-suffieiency. Postscript There have been two assessments of the projections made in this chapter since its original publication. Morawetz (1977, pages 21-22) 64. It is perhaps more than coincidence that most of the striking successes in development through aid-Greece, Israel, Korea, and Taiwan, for example--were financed largely on a non project basis. 444 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT compares the actual growth of forty-five of the fifty countries to that predicted for the period 1962 to 1975. He finds that twenty countries exceeded the plan projections by mOre than 0.5 percent, thirteen equalled the projected rates, and twelve fell below them. Morawetz concludes that the actual outcome was about half way between the medium and upper limit projections when equal weight is given to each observation. Chenery and Carter (1973, 1976) evaluate the overall performance of the two-gap model and analyze the relations among trade, capital flows, and growth. 'They consider only tpe first projection pcriod (1962-70) and limit the sample of countries to thirty-seven that are representative of the performance of the international system.65 For this period the average growth of the sample countries was 5.3 percent, approximately the same as that projected. The analysis focuses on the modifications needed in the two-gap model to improve its usefulness and on thc sources of variation in country performance. This postscript summarizes some of the conclusions that emerge from the Chenery- Carter assessment. Analytical framework Most discussion of the framework of two-gap analysis has focused on its relevance to developing countries and only indirectly on its usefulness to the suppliers of capital. The main criticisms of this approach fall under two headings: (a) a neoclassical critique, which denies the validity of the trade limit concept, and (b) skepticism that an increase in external resources is likely to lead to higher investment rather than a rise in consumption. The neoclassical case is effectively presented by Bruton (1969), who argues that the trade limit is unlikely to occur in a well-managed economy and, furthermore, that basing aid policy on this hypothesis provides the wrong incentives to developing countries. There is no issue between us on the first question: model 2 of the present chapter -as well as the studies of Israel and Pakistan in chapters 8 and 9-a11 demonstrate that a trade gap in excess of the savings gap will not appear jf resources are optimally allocated over time. But this norma- tive argument does not reduce the probability that a dominant 65. For this purpose, ten of the fourteen countries with populations below 5 million are omitted as are three abnormal cases (Algeria, Vietnam, and Rhodesia). POSTSCRIPT 445 trade limit will actually occur with some frequency as a result of either ineffective national policies or unanticipated changes in external markets. 66 The extent to which countries typically use external resources to increase investment or consumption is also an empirical question. Econometric studies by a number of authors have shown the need to modify the savings function used in this chapter (equation 10.5), to allow for this choice. 67 Tbe more general hypothesis is that inflows of external resources have several origins and will normally lead to in- creases in both investment and consumption. Since saving is con- ventionally measured as the difference between investment and the capital inflow (equation 10.2), an increase in consumption with given GNP must be reflected in a fall in saving. 68 This hypothesis can be incorporated by adding a term to the savings function (10.5): where a" is the proportion of the capital inflow that is used for con- sumption. The original form thus becomes the extreme case in which all of the external flow is invested and a" O. Since St is defined as the savings that would result from given gov- ernment policies in periods when this constraint is binding, the parameters in the savings function can only be properly estimated from a sample of observations for which this assumption is valid. Of the authors who have analyzed this relation, only \Veisskop£ (1972) obtained a separate estimate of a" (0.23) for a group of countries for which this assumption could be made. A number of other studies that did not attempt to make sueh a separation produced estimates of a" in the range of 0.45 to 0.55 for the negative effect of 66. Bruton's other main objection to the two·gap approach is that a shift to less import·intensive forms of investment can eliminate the "gap between the gaps." As was pointed out in my reply (1969b), however, using all the potential saving for less productive investment is not equivalent to getting rid of the trade limit nor does it reduce the desirability of external borrowing. 67. See, for example, Strout (1969). Chenery and Eckstein (1970), Griffin and Enos (1970), Landau (1971), Singh (1972), Papanek (1972), and Weisskopf (1972). A survey of these and other studies of the relations between capital inflow and savings is given by Mikesell and Zinser (1973). 68. 1bis relation is discussed at length by Papanek (1972). 446 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT capital inflow on measured savings. 69 These results are consistent with the two-gap hypothesis, which implies that actual savings will be less than potential savings when either the ability to invest [equation (10.4)], or the trade limit is binding. In analyzing the role of capital inflows in development policy from the viewpoint of either the borrower or the lender, a" should be treated as a policy variable. An effective development policy is one that channels a high proportion of external resources into investment when savings constitute the binding constraint. The projections in this chapter used the optimistic assumption that a" = O. Actual ex- perience wiII be compared to the projections made on this basis in the following subsection. 70 Overall results An overview of the structural characteristics of the Chenery-Strout projections is given in table 10-12, which COmpares average values of the basic parameters projected for 1962-70 under the intermediate (plan) assumptions to the observed values as well as to the historical values for 1957-62. Of the internal parameters, the incremental capital-output ratio (leoR) fell somewhat more than "vas anticipated, while the mean marginal savings rate (0.18) was between the ex ante and ex post rates projected (0.196 and 0.140). For the external parameters, the average growth of exports and the marginal propensity to import turned out to be slightly higher than projected. The vari- ance around these means was generally underestimated, however- as it is in most projections-as shown by a comparison of the uppcr and lower quartile values for each distribution. These uDweightcd averages can be quite misleading in assessing the overall performance of the international system because they ignore both the size of countries and their initial levels of income. To remedy these defects, table 10-13 compares the aggregate results for three groups of countries in terms of GNP growth and the sources of foreign exchange. 69. See Strout (1969) and Chencry and Carter (1976, page 308). A com- parison of time-series and cross-country estimates is given in Chenery and Syrquin (1975). Differing interpretations of the evidence for individual countrics are given by Griffin and Enos (1970) and Papanek (1972). 70. Since the effects of the past capital inflow are incorporated in statistical estimates of iX', it is only changes in the relative importance of F that will affect predicted savings. POSTSCRIPT 447 Table 10-12. Summary of Structural Parameters and Growth Rates (37 ·country sample) Upper Lower ~"-'--'--"-~'--'--"-" quartile Median Mean .. _ quartile ... _ - _ . Historical estimates (1957-62)" ICOR 4.310 3.270 3.600 2.830 Marginal savings ratio 0.205 0.150 0.162 0.115 Marginal import ratio 0.250 0.205 0.202 0.135 Rate of growth of exports (percentage) 7.120 4.450 5.125 3.425 Rate of growth of GDP (percentage) 5.050 4.400 4381 4.000 Chenery-Strout "plan" estimates (1962-70), lCOR 3.770 3.270 3340 2.720 Marginal savings rate (ex post) 0.235 0.200 0.140 0.150 Marginal savings Ia te (ex ante) 0.235 0.200 0.196 0.150 Marginal import rate (ex post) 0.331 0.200 0.251 0.260 Marginal import rate (ex ante) 0.236 0.190 0.204 0.164- Rate of growth of exports (percentage) 7.120 4.450 5.080 3.160 Rate of growth of imports (percentage) 6.470 4.770 5.270 3.720 Rate of growth of CDP (percentage) 6.000 5300 5.290 4.750 Actual values (1960-70)" ICOR 3.800 3.000 3.250 2.450 Marginal savings rate 0.245 0.212 0.180 0.100 Marginal import rate 0.332 0.228 0.214 0.078 Rate of growth of exports (percentage) 8.090 5.370 5.140 2.640 Rate of growth of imports (percentage) 8.910 6.030 5.820 3.100 Rate of growth of CDP (percentage) 6.450 5.100 5.360 3.900 a. Computed from table 10·7. h. Chenery and Carter (1976, table 3), based on intermediate (plan) assump' tions in table 10-7. The middle-income countries (group A) were relatively successful in expanding their exports and thereby reduced their need for external capital compared to what had been projected. As a result, the inflow of capital was less than half the anticipated requirement, while the growth of GNP fell only slightly short of that projected. India has been treated as a separate category because of its size and the significant differences in its experience from that of other poor countries. For the poor countries in group B, the capital supplied was 83 percent of what was projected, while GNP growth was 10 percent higher. For India, however, the capital supplied was only 55 percent of that calculated to be necessary to support GNP growth of 5.3 percent, and actual growth of GNP was only 3.2 percent. As indieated below, Table 10·13. Aggregate Growth, 1960-70 of 1970 U.S. dollars) Total GNP GNP growth Popula- 1970 (percentage) Capital inflow tion (millions) 1960 Projected Actual Projected Actual Projected Actual Group A 321 74 135 129 6.1 5.7 30 13 Sixteen countries with per ca pi ta income more than "' "'"" OQ $190 Group B 606 43 69 4.8 5.3 24 20 Twenty countries with per capita income less than $190 Groupe India 538 35 59 50 5.3 3.5 12 6 Total 1,465 1 263 251 5.6 5.1 66 40 Source: Chenery and Carter (1973, p. 463). Country detail is given in table 10·14. POSTSCRIPT 449 this is the principal case in which slow growth can be attributed to a substantial extent to the limited supply of external capital. The transition The growth sequence implied by the Chenery-Strout projections has three main features: (a) acceleration of growth to an acceptable level (assumed to be at least 5.5 percent); (b) use of external capital to supplement both savings and exports; and (c) eventual reduction in dependence on external capital to sustain satisfactory growth. Al- though it is not possible to verify the functioning of this mechanism by conventional econometric tests because of thc interaction of the several limiting factors, table 10-14 presents some elements of country performance that facilitate less formal judgments. In the 1962-70 period, twenty-five out of the thirty-seven countries in the sample accelerated their growth, and almost all of these sus- tained satisfactory rates through 1975. The rate of success was much greater among the middle-income countries, three-quarters of which achieved growth in excess of 5.5 percent through 1975. By contrast less than half of the poor countries managed to do so. This difference was somewhat more pronounced than had been anticipated. 71 Two systematic differences in the external factors -contributed to these results. Although primary exports grew about as expected (3 percent for 1962-70), manufactured exports grew much more rapidly (15 percent compared to the 6 percent projected). More middle- income countries were able to benefit from the favorable develop- ment of manufactured exports, but these exports also provided the basis for spectacular growth in Taiwan and Korea. The statistical relations between real growth of GNP and real growth of exports and imports for the present sample are given by the follow- ing regression equations (t ratios in parentheses) : G y = 3.9 0.24 G e and (11.7) (6.9) Gy 3.3 0.29Gm R2 = 0.63, (9.4) (8.0) 71. In the plan projections, eleven of the sixteen middle-income countries and nine of the twenty-one poor countries were expected to achieve at least 5.5 percent growth through 1975. 1be actual results were twelve middle-income and eight poor countries (including two beneficiaries of the rise in oil prices). Table 10 -14. Comparison of External Aspects of Growth Projections F inanc! ng of imports (1960) Growth of GNP (Share) Per Popula- Histori- Plan" capita tion cal Actual Actual rate Capital GNP (millions) (191)7- (1960- (1962- (1962- Primary Other inflow 62) 70) 71» 75) exports exports (F) (2) (3) (4) (5) (6) (7) (8) (9) A. Middle-income countries 1960 GNP per capita above $190 -I:>. 1. Israel 843 3 9.0 7.9 8.3 9.0 0.13 0.40 0.47 \.ro 0 2. Venezuela 752 10 4.5 5.8 5.7 6.0 1.61 0.00 -0.61 3. Argentina 681 23 3.1 4.0 4.5 4.3 0.81 0.09 4. Greece 417 9 6.0 7.3 7.2 6.5 0.34 0.05 0.61 5. Jamaica 388 2 4.0 5.1 5.7 4.5 0.87 0.12 0.01 6. Chile 371 10 3.5 3.9 3.0 5.0 0.74 0.08 7. Mexico 352 51 5.0 7.2 6.6 6.0 0.77 0.13 0.10 8. Costa Rica 340 2 5.5 6.5 6.3 6.0 0.78 0.04 0.18 9. Guatemala 253 5 4.0 5.1 5.9 5.0 0.81 0.06 0.13 10. Peru 247 14 5.5 4.5 4.7 5.5 0.80 0.33 -0.13 II. Colombia 221 22 5.0 4.9 5.8 6.1 0.99 0.01 0.00 12. Turkey 217 35 5.3 6.4 6.6 6.0 0.49 0.06 0.45 13. El Salvador 210 4 5.0 5.4 4.9 6.0 0.74 0.08 0.18 Malavsia 208 11 4.0 6.2 6.5 5.0 1.22 0.08 -0.30 15. Brazil 193 93 5.5 5.3 7.8 5.5 0.73 0.09 0.18 16. Iran 192 29 4.4 8.3 10.8 5.5 1.21 0.07 -0.28 Subtotal 321 B. Low-income countries 1960 GNP per capita below $190 Ecuador 182 6 4.2 5.1 6.4 5.0 1.00 0.02 -0.02 18. Morocco 167 15 2.8 3.9 4.5 4.0 0.82 0.22 -0.04 19. Jordan 160 2 5.6 6.4 2.9 5.6 0.25 0.02 0.73 20. Ghana 158 9 4.5 2.2 2.4 5.5 0.77 0.06 0.17 21. Tunisia 156 5 4.1 3.5 6.8 5.0 0.51 0.26 0.23 22. Philippines l49 37 5.0 5.9 5.6 5.5 1.03 0.05 -0.08 23. Taiwan 147 l4 6.0 10.0 9.4 7.0 0.52 0.08 0.40 24. Sri Lanka 131 13 4.2 3.9 4.2 5.0 0.90 0.01 25. Egypt 129 33 4.5 4.2 3.7 5.5 0.92 0.09 -0.01 26. Thailand 111 36 5.0 8.0 7.7 6.0 0.72 0.20 27. Korea, Republic of 104 32 4.3 9.4 10.1 5.0 0.22 0.04 0.74 28. Kenya 101 11 1.7 6.7 6.7 3.5 0.96 0.18 -0.14 -I:L 29. Indonesia 89 116 1.0 3.0 5.3 3.0 1.05 0.Ql -0.06 "" "- 30. Uganda 89 10 1.7 5.1 3.9 4.0 1.10 0.08 -0.10 31. Sudan 88 16 5.1 3.9 5.8 5.5 1.04 0.00 32. Pakistan 71 130 4.5 5.1 5.3 0.45 0.26 0.29 33. Nigeria 70 55 4.0 3.0 6.4 4.5 0.71 0.01 0.27 34. Tanzania 65 13 4.2 6.1 5.2 5.0 1.01 0.07 -0.08 35. Burma 56 28 3.2 2.7 2.9 4.0 0.91 0.04 0.05 36. Ethiopia 45 25 4.5 5.0 4.2 4.5 0.85 0.03 0.12 Subtotal 606 C. India 83 538 4.3 3.5 3.4 5.3 0.35 0.28 0.37 Total 1,465 (table continues on the following pages) Table 10-14 (continued) Growth of Cumulative capital inflow exports exports Primary factured Profected Actual (predicted) (actual) exports exports Imports (1962-70) (1962-70) (10) (11) ( 12) (13) (14) (15) (16) A. Middle-income countries 1960 GNP per capita above $190 1. Israel 14.7 15.4 15.3 17.8 12.5 3,657 4,488 2. Venezuela 2.8 4.3 4.7 72.6 8.6 -6,111 3,007 -l>.. 3. Argentina 3.7 2.6 1.1 10.1 0.3 4,565 -1,197 I.rt N 4. Greece 7.1 8.6 3.6 22.8 6.9 2,635 5,218 5. Jamaica 4.5 5.4 5.1 6.l 6.6 197 225 6. Chile 2.5 6.2 6.7 -0.3 2.5 2,884 -113 7. Mexico 7.0 4.7 3.6 12.2 5.5 670 424 8. Costa Rica 4.6 9.7 -4.3 26.7 11.4 362 343 9. Guatemala 4.5 10.4 7.0 28.0 9.2 836 185 10. Peru 6.9 2.2 2.3 3.1 1.6 1,015 -167 II. Colombia 3.6 3.8 1.8 28.8 3.9 4,409 411 12. Turkey 4.0 7.7 13.0 13.3 6.3 3,587 1,820 13. El Salvador 4.9 7.5 3.8 22.5 7.9 185 I 14. Malaysia 2.8 5.4 4.1 13.7 5.7 -112 -921 15. Brazil 3.7 6.2 7.6 14.4 3.8 4,936 114 16. Iran 7.l 10.7 10.8 6.3 13.3 -14 -3,070 Subtotal 29,939 13,270b D. Low-zncome countnes 1960 GNP per capita below $190 17. Ecuador 4.5 2.0 1.0 -1.4 4.8 393 166 18. Morocco 2.7 2.7 2.1 2.3 2.8 1,422 265 19. Jordan 7.5 8.3 8.4 19.3 7.5 1,041 1,076 20. Ghana 2.1 -1.3 -4.4 15.7 -4.2 1,311 234 2l. Tunisia 4.5 5.5 6.3 3.0 3.4 1,062 1,553 22. Philippines 4.1 9.2 6.8 18.3 9.9 1,041 774 23. Taiwan 7.1 20.7 2.9 46.6 15.5 1,435 411 24. Sri Lanka 2.3 -0.7 1.0 4.8 0.7 531 394 25. Egypt 1.8 0.2 -2.9 12.6 1.9 3,539 1,697 26. Thailand 6.1 7.9 9.1 0.4 9.9 192 1,024 27. Korea, Republic of 7.6 27.1 11.2 53.3 17.8 2,769 4,026 28. Kenya 4.5 2.4 2.6 2.9 6.7 110 75 29. Indonesia 1.4 5.0 -3.2 43.5 10.1 2,508 3,431 ..j:,. 30. Uganda 3.6 4.5 4.6 6.3 6.0 362 209 Vi IN 3l. Sudan 8.0 2.6 3.5 -4.6 3.4 255 245 32. Pakistan 4.9 2.7 -3.6 13.9 5.2 3,998 3,478 33. Nigeria 7.1 7.6 6.1 15.9 3.6 2,210 854 34. Tanzania 4.4 5.6 3.4 4.1 6.4 402 -51 35. Burma 10.2 -11.7 -12.2 -3.0 -7.6 -116 89 36. Ethiopia 7.1 5.6 5.7 -6.9 6.5 218 208 Subtotal 24,799 20,207" C. India 2.7 2.0 0.2 3.9 -0.4 1l,457 6,313 Total 66,195" 39,791" Source: Chenery and Carter (1976, tables 4, 5, and 8). Column (5) from Morawetz (1977) and World Bank data tapes. Note: Totals may not add due to rounding. a. Median projections of Chenery and Strout. h. Excluding outflows. 454 FOREIGN ASSISTANCE AND ECONOMIC DEVELOPMENT where G y , G e, and G m are the growth rates of GNP, exports, and imports. Table 10-14 suggests that a shortfall from projected exports was a significant factor in the slow growth of Peru, Ghana, Sri Lanka, Sudan, and Burma. Conversely, accelerated export growth contributed to high GNP growth in Greece, Malaysia, Iran, Taiwan, Korea, and Indonesia. The causal nature of the export relation is supported by Michae1y (1977). The second main difference arises from the conditions under which external capital has been supplied. The volume of capital available on concessional terms has been considerably less than anticipated, and India has been the country primarily affected. Of the shortfall in total capital flows to the poor countries-$l0 billion over the 1962- 70 period-more than half was in the projected flow to India. In summary, although overall progress toward achieving self- sustaining growth has been as substantial as that projected for 1962-75, such growth has proven considerably harder for the poor countries.72 The rapid expansion of trade has been of more benefit to the middle-income countries that were able to develop manufactured exports, and to the exporters of oil and a few other minerals, but less favorable to the bulk of the poorer countries. 73 It should also be recognized that the growth potential of poor countries is somewhat lower so long as most of their production comes from agriculture, where possibilities for expansion are rarely as great as 5 percent. The importance of the trade limit to growth appeared to have diminished in the export boom of the early 19705, but this situation was reversed after 1974 as a result of the rise in the price of petroleum and the slowdown in wor1d growth. Since then the trade limit has again become widespread and exerts a major influence on develop- ment policy. To sustain growth, many developing countries have had to change the structure of their external trade and also to increase their external borrowing to levels exceeding those of the 19608. The international system of anocating both concessional and non- concessional capital has favored the middle-income countries through- 72. Chenery and Carter (1976) discuss some of the more notable examples of retarded growth. For example, the failure to diversify out of stagnant primary ex· ports was a major factor in countries snch as Ghana and Sri Lanka. 73. In a number of cases it was the failure of developing countries to take ad- vantage of export markets, because of policies that discriminated against exports, that was primarily responsible for this result. POSTSCRIPT 455 out this period. Edelman and Chenery (1977) show that only half of the total commitments of concessionalloans and grants were made to poor countries during the 1967-74 period, and that countries with per capita incomes under $200 (in 1970 dollars) received only half as much on a per capita basis as did those in the income bracket of $200 to $500.74 Although donor countries have stated their intentions to favor the poorer countries, progress in this direction has been slow. The distribution of private capital is considerably more skewed, since it is based on export performance and other determinants of credit- worthiness. This combination of internal and external factors has produced the major weakness in the postwar pattern of development: the failure of the poor to share equitably in the benefits of rapid world growth. The dimensions of this problem and the extent to which it can be reduced by alternative policy measures constitute the subject matter of the following chapter. 74. These figures are based on the grant equivalent of loans to a110w for differ- ences in terms. Omission of India-which receives only half the poor country average-reduces but does not eliminate this discrepancy. Chapter II Growth and Poverty in Developing Countries with MontekS. Ahluwalia and Nicholas G. Carter ALTHOUGH THE OUTPUT OF THE WORI.D ECONOMY has expanded at an unprecedented rate in the past quarter century, the benefits of growth have reached the world's poor only to a very limited degree. This phenomenon is not the result of any failure of developing countries as a group to share in the general economic expansion. Tbeir income per capita rose by almost 3 percent a year over this period-consider- ably faster than in the past. The failure lies in the distributional pattern of past growth, which has left the poorest groups largely outside the sphere of economic expansion and material improvements. There are two aspects to this phenomenon. First, the impressive record of the Third World as a whole conceals the fact that most of the poorest countries, containing the principal concentrations of the world's poor, have experienced lesser increases in income. Second, and equally important, there is mounting evidence that the growth pro- cesses under way in most developing countries are such that incomes of the poorer groups increase more slowly than the average. This chapter attempts to synthesize and extend the comparative studies of growth and distribution in developing countries by Chenery and Carter (1973, 1977), Ahluwalia and Chenery (1974), and Ahluwalia (1976). Statistical support was provided by Hazel Elkington and Tamar Katz. We are indebted to Bela Balassa, Clive Bell, John H. Duloy, T. N. Srinivasan, Paul Streeten, Lance Taylor, and Jeffrey Williamson for valuable comments. 456 THE DIMENSIONS OF GLOBAL POVERTY 457 International debate has centered around the design of structural changes to offset these trends. Proponents of a New International Economic Order consider the principal policy objective to be the acceleration of growth in developing countries, with special conces- sions to the poorest among them. Others give greater weight to policies to improve the internal distribution of income, including direct measures to satisfy the basic needs of the poorest groups. These issues have been discussed so far in largely qualitative terms with little attempt to translate global targets for the eradication of poverty into more specific strategies whose feasibility can be examined. This chapter suggests a quantitative framework for such an analysis and derives some preliminary conclusions from it. Although there is not yet an adequate statistical basis for a formal analysis of the central relations involved, there has been considerable progress in the past few years in several areas: (a) the definition and measurement of the incidence of poverty, using both physical and monetary indexes; (b) securing internationally comparable data on income levels, based on purchasing power comparisons; and (c) measurement of the distribu- tion of income and consumption within developing countries. Our study is in three parts. First is the estimation of the extent of absolute poverty in developing countrie~ and of the relations between income distribution and rising levels of outpnt. Second is an analysis of past trends in growth and poverty in a representative group of countries and of the implications of projecting these trends on the basis of present policies. Third is a consideration of possible improve- ments on this performance through accelerating income growth, improving its distribution, and reducing fertility. vVe conclude with a comparison of alteruative approaches to poverty reduction and their implications for national and international action. Despite the tenta- tive nature of some of the underlying assumptions, our analysis demonstrates that a combination of several approaches and of national and international action is more likely to succeed in reducing poverty than exclusive reliance on anyone of them. The Dimensions of Global Poverty This section attempts to evaluate the scale of poverty in the de- veloping world and the available evidence on the effect of growth on poverty. The analysis is based on a sample of thirty-six countries, 458 GROWTH AND POVERTY IN DEVELOPING COUNTRIES which are listed in table 11-1. 'The sample is broadly representative of developing countries with mixed or market-oriented economies. These countries span the wide range of income levels observed in the de- veloping world and reflect its distribution by broad geographic regions. Together the countries in our sample account for about 80 percent of the population of the developing world, excluding China. l Defining absolute poverty The first step in measuring the scale of poverty is to establish a common poverty line to be applied across countries. Such a definition is necessarily arbitrary. Attempts to define absolute poverty in terms of some objectively determinable minimum level of consumption that is neccssary for "continucd survival" do not escape this problem, since the notion of continued survival is undefiued. At the very least we would need to specify survival through some given life expectancy in a given environment. Present levels of life expectancy in most poor countries are quite low and do not provide a basis for defining minimum requirements. Increases in life expectancy will require higher levels of real consumption, including not only better food in- take but also a better general environment for health and nutrition. Not only is the notion of a biologically determined absolute poverty level imprecise, it is in any ease wrong to think that poverty should be defined solely by biological requirements. Ultimately, concepts such as poverty lines are operationally meaningful only when they acquire some social reality, that is, when a sufficient social consensus insists that a particular level of living represents an objective claiming a high social priority. Once we recognize that acceptability by con- temporary social standards is a fundamental requirement, it follows that poverty lines used in national policy debates will vary among countries, reflecting differences in levels of economic, social, and 1. The principal limitation on the size of the sample is the availability of data on income distribution. Indeed, of the thirty-six countries in our sample, fairly reliable distribution data were available only for twenty-five. For the remainder we have used estimates of the distribution of income based on cross-country com- parisons. We have resorted to this procedure only in cases where inclusion of the country was very desirable either because of its size or to ensure adequate geo· graphical representation. THE DIMENSIONS OF GLOBAL POVERTY 459 political development. By the same token, these standards will also change over time. For these reasons any effort to define a poverty line to be applied across countries and over time must be approached with caution. \Ve have concluded, however, that with all its limitations this measure can still provide a useful basis for international policy. For this purpose, it is less important that the poverty line correspond to some objective criteria for minimal levels than that the absolute level chosen be con- servative and consistently applied. We have based our definition on the poverty lines that have been used in India, which is the largest and one of the best studied developing countries. There is an extensive literature on the measurement of poverty in India, and a variety of poverty lines have been proposed, some of which have received official sanction. The most widely used poverty line is defined by the total consumption expenditure needed to ensure a daily supply of 2,150 calories per person, given the observed expendi- ture patterns of the Indian population.2 Estimates of the extent of poverty by this standard vary from year to year, but most estimates range between 40 and 50 percent of the total population. For our study we have adopted an intermediate position, setting the poverty level to be applied across countries as the income per head accruing ap- proximately to the 45th percentile of the Indian population. Applica- tion of this essentially South Asian standard across all developing countries yields estimates of poverty that are conservative in the sense of understating the extent of the problem by standards appropriate for richer countries. Having chosen a poverty line, the ncxt step is to apply it in such a way to ensure comparability across countries. The use of official ex- change rates to define equivalent levels of expenditure in different countrics does not ensure equivalent levels of real purchasing power. \Ve have attempted to overcome this problem by using "equivalent purchasing power conversion ratios" estimated by Kravis and associ- ates from data conected by the United Nations International COl11- 2. It should be emphasized that poverty lines defined by consumption expendi- ture ignore the fact that there is very considerable variation in caloric intake achieved at any given level of expenditure. In any case, the underlying specification of a single caloric norm is itself questionable. Nutritionists have shown that there is very considerable variation in caloric requirements even for the same individual over time. Table ll-l. Sample Panel: Per Capita GNP, Population, and Poverty Percentage of population in 1975 GNP per capita" poverty in 1975 using Kravis at official 1975 using Kravis using official adjustment exchange population adjustment exchange Country' factors rates (millions) factors rates Group A (income less than 350 rcp dollars·) 1 Bangladesh 200 72 80.7 64 60 2 Ethiopia 213 81 27.3 68 62 3 Burma 88 30.9 65 56 -I>. 4 Indonesia 280 90 130.0 59 62 0'1 Q 5 Uganda 280 115 11.5 55 45 6 ZaIre 281 105 20.6 53 49 7 Sudan 281 112 18.1 54 47 8 Tanzania 297 118 14.8 51 46 9 Pakistan 299 121 73.0 43 34 10 India 300 102 599.4 46 46 Subtotal 284 99 1,006.3 51 49 Group B (income of 350-750 rcp dollars) 11 Kenya 413 168 13.4 55 48 12 Nigeria 433 176 75.3 35 27 13 Philippines 469 182 42.5 33 29 14 Sri Lanka 471 185 14.1 14 10 15 Senegal 550 227 4.3 35 29 16 Egypt 561 238 37.2 20 14 17 Thailand 584 237 41.6 32 23 Hi Ghana 628 255 9.8 25 19 19 Morocco 643 266 17.3 16 20 Ivory Coast 695 325 5.9 25 14 Subtotal 511 209 261.4 31 24 Group C (income more than 750 ICP dollars) 21 Korea, Republic of 797 325 34.1 8 6 22 Chile 798 386 10.6 11 9 23 Zambia 798 363 4.9 10 7 24 Colombia 851 352 24.8 19 14 25 Turkey 914 379 39.7 14 11 26 Tunisia 992 425 5.7 10 9 Malaysia 1,006 471 12.2 12 8 28 Taiwan 1,075 499 16.1 5 4 29 Guatemala 1,128 497 5.5 10 9 ..j),. 30 Brazil 1,136 509 106.8 15 8 .... 0\ 31 Peru 1,183 503 15.3 18 32 Iran 1,257 572 33.9 13 8 33 Mexico 1,429 758 59.6 14 10 34 Yugoslavia 1,701 828 21.3 5 4 35 Argentina 2,094 1,097 24.9 5 3 36 Venezuela 2,286 1,288 12.2 9 5 Subtotal 1,220 577 427.6 13 8 Total 555 237 1,695.3 38 35 Sources: GNP and population from World Bank data files; Kravis adjustment factors from Kravis, Heston, and Summers (1978b). a. Countries are ordered by 1975 GNP per capita adjusted bv the Kravis factor. b. In 1970 U.S. dollars. c. See footnote 3 in this chapter. 462 GROWTH AND POVERTY IN DEVELOPING COUNTRIES parison Project (ICP).3 Using these ratios, we can convert the per capita GNP levels in each country into GNP per capita measured in donars of 1970 U.S. prices-hereafter ealled ICP dollars. The resulting estimates are shown in table 11-1. Our poverty line is easily calculated given the income distribution for India for 1975 and its estimated level of per capita GNP in ICP dollars. vVe have chosen a poverty line of 200 ICP dollars-the level of the 46th percentile-which is then applied to the income distribution and per capita GNP data for other countries to estimate the extent of poverty in each case. 4 This income-based approach to defining poverty makes no explicit allowance for the achievement of minimum levels for essential public services such as health, education, access to clean water, and sanita- tion. 'I'hese are fundamental elements in a more complete definition of poverty that are of crucial importance in designing a balanccd program of poverty alleviation, but they remain outside the present analysis. The extent of poverty in developing countries The procedure just described enables us to estimate the extent of poverty in each country using an income level that reflects comparable levels of purchasing power. These estimates are reported in the fourth column of table 11-1. For purposes of comparison, we have also estimated the extent of poverty in our sample without the conversion ratios. In this case, we measure per capita GNP for each country in U.S. 3. The International Comparison Project has been a joint responsibility of the United Nations Statistical Office, the World Bank, and the International Com- parison Unit of the University of Pennsylvania. Two volumes of results have been published; see Kravis, Kenessey, Heston, and Summers (1975), which includes detailed estimates for India, and Kravis, Heston, and Snmmers (1978a). The con- version factors used here were estimated by Kravis, Heston, and Summers (1978b). These ratios are called "Kravis factors" in this ehapter; the resulting unit of value is identified as an ICP dollar. Other methods of estimating conversion factors from the rcp data are under study. 4. Although the Indian estimates of poverty are based on a consumption stan- dard applied to the distribution of population across consumption levels, in our study we have defined the poverty line by the income per head observed at the 46th percentile of the Indian population. The use of per capita GNP instead of per capita personal income (which is clearly more appropriate) is dictated by the absence of data on the personal income component of GNP. Since the proportion of personal income in GNP declines at higher levels of development, our procedure probably understates the extent of poverty. THE DIMENSIONS OF GLOBAL POVERTY 463 dollars by converting at official exchange rates, calculate the income level of the 46th percentile in India, and apply this level to the data for all other countries. These estimates are shown in the last column of table II-I. Since in each case the poverty line is based on the income of the same percentile of the Indian population, the difference between the two estimates lies in the extent to which poverty in other countries is altered relative to India. In general, we find that the use of purchasing power ratios reduces the differences between the incidence of poverty in middle and highcr income countrics compared to the low income countries. 'Inc use of ICP dollars also raises the estimates of poverty relative to India in the low income countries, This rise reflects the fact that the Kravis pur- chasing powcr ratios suggest that GNP levels in both groups of countries are overstated relative to India. 5 The major features of global poverty as rcvealed in the cstimates based on purchasing power ratios correspond broadly to other esti- mates. 6 Almost 40 percent of the population of the developing coun- tries live in absolute poverty defined by income levels that are in- sufficient to provide adequate nutrition by South Asian standards. The bulk of the poor are in the poorest countries: in South Asia, Indonesia, and Sub-Saharan Africa. 'These countries account for two- thirds of the developing world's total population and well over three- fourths of the population in poverty. 1nc incidence of poverty is 60 percent or more in countries having the lowest levels of real GNP. Although the incidence of poverty is much lower for the middle income developing countries in our sample, our estimate of poverty in this group of countries increases from 24 to 31 percent when purchasing power ratios are used to determine GNP. There is a similar increase in the high-income group from 8 to 13 percent. 5. Although the use of the Kravis conversion ratios is clearly a step in the right direction, it also raises questions that we have not addressed. For example, it is likely that the purchasing power ratios vary for different income groups within a country. Since an important element underlying this correction is the relative un- dervaluation of services in low income countries, and since services are dispropor· tionately consumed by the rich, it may be that official exchange rates understate incomes of the rich more than of the poor. By applying a single average correction factor to GNP per capita we ignore this problem. A similar argument can be made for distinguishing between couversion rates relevant for different regions and for urban and rural areas. 6. See, for example, Chenery, Ahluwalia, Bell, Duloy, and Jolly (1974, chap- ter 1). 464 GROWTH AND POVERTY IN DEVELOPING COUNTRIES It is interesting to compare our estimates of absolute poverty to those reported for selected Latin American countries in a recent joint study by the Economic Commission for Latin America and the W orId Bank. 7 This study estimates a much higher incidence of poverty in Latin America-around 40 percent for the region as a whole-but this results from the adoption of poverty lines that are significantly higher than those derived from South Asia. For example, the food budget was geared to a higher minimum nutritional level and was constrained to ensure some minimum consumption of higher value foods (meat, fruit, eggs, and milk).8 The food budget thus obtained was used to define two different poverty lines: a "destitution line," defined as income equal to the food budget, and a "poverty line," defined as income equal to twice the food budget to allow for nonfood ex- penditures. Estimates of the extent of poverty in the Latin American countries in our sample are broadly in line with Altimir's estimates of the extent of destitution (about 19 percent of the population of Latin America). Furthermore, projcctions using either method show little prospective decline in absolute poverty with present trends. Poverty and growth: a review of evidence The extent of poverty in any country depends upon two factors: the average level of income and the degree of inequality in its distribu- tion. Although the estimates of income growth are relatively good, we have little reliable information on how the distribution of income has changed over time. Systematic time series data based on reliable sources and using comparable concepts are simply not available. At most there is a handful of countries for which we have observations for two or more years spanning a decade or so. In the absence of time series data for individual countries any assessment of changes in the distribution of income accompanying development in the past must be based on what can be inferred from cross-country data. This evidence has been extensively studied in recent years and a brief summary of the findings is presented below. 7. See Altimir (1979). 8. The calorie requirement used is also higher-ranging between 2,260 and 2,350 calories per person per day-and there is also a specified minimum protein requirement varying between 40 to 43 grams per person. Furthermore, the food budgets were calculated to provide explicit allowance for a minimum consumption of vegetables and fruit to provide minerals and vitamins in a balanced diet. THE DIMENSIONS OF GLOBAL POVERTY 465 The central theme in the continuing debate on trends in income distribution is whether development in the past has been accompanied by such an increase in inequality that the poor have benefited rela- tively litt1e from overall growth. Much of this debate has its origin in the classical contributions of Kuznets (1955, 1963), who hypothesized that the process of development was likely to be accompanied by a substantial increase in inequality, which would reverse itself only at a relatively advanced stage. Kuznets's original speculation was based on fragmentary historical data for the now developed countries, but in its later development, especially at the hands of subsequent con- tributors, the investigation of this hypothesis has relied almost en- tirely upon cross-country evidence. A number of studies-Adelman and Morris (1973), Paukert (1973), Chenery and Syrquin (1975), and Ahluwalia (1976) -using different and progressively more reliable sets of cross-country data have reported confirmation of the hypothesis to some degree.!! 1be average pattern discerned in the data is one of significant increase in inequality as income levels rise from the least developed to about U.S. $600 per capita in 1975 prices. 1o lbe extent of the increase in relative inequality reported by different authors varies substantially. At one extreme, Adelman and Morris (1973) have argued that the cross-country data suggest that economic growth will be accompanied by a process of prolonged absolute im- poverishment for large sections of the population. Others, such as Ahluwalia (1976), have argued that although the crosHountry evi- dence points to increasing inequality in the early stages, this docs not completely offset the effect of growth. Income levels of the poorer quintiles arc likely to rise-but much more slowly than the average. rIbe limited time series evidence provides some support for Ahlu- walia's conclusion. There are a number of countries for which esti- mates are available of the distribution of ineome (or consumption) at two points in time spanning about ten years in eaeh case. Although many of these countries appear to show some decline in the shares of the poorer quintiles over time, in no casc is this decline in shares 9. It must be emphasized that given the limitations of cross-country data, this evidence is at best persuasive. For a skeptical view, see Papanek (1978). 10. For the typical developing country, $600 in 1975 prices and at official ex- change rates is approximately equal to 800 rep dollars at 1970 prices. This is the estimated turning point for the share of the lowest 60 percent. As noted by Ahluwalia (1976), the turning point for the lower 40 percent and the lowest 20 percent occurs at successively higher levels of per capita GNP. 466 GROWTH AND POVERTY IN DEVELOPING COUNTRIES sufficiently steep to offset the recorded growth in mean incomes.H Some of this evidence is discussed below. A simplified representation of the Kuznets effect is given in figure II-I, which plots the per capita income of the top 40 percent of the population against that of the bottom 60 percent. Lines of constant per capita income appear as downward sloping straight lines. Ahlu- walia's estimate of the Kuznets curve in these dimensions appears as a eurve \\1th maximum inequality in the vicinity of 800 ICP dollars (1970 prices). Between the income levels of 200 and 800 dollars the share of the lower 60 percent declines from 32 percent to 23 percent of the national income. A country that followed this average relation would have about 80 percent of the increment accruing to the upper 40 percent of its citizens and quite modest increases for the remaining groupsP T11is average relation can be compared with the observed movement of individual countries over specified periods of time, as shown by the arrows in figure II-I. These observations relate to a relatively short time period-typically ten years-and the data are often not strictly comparable for a given country. Nevertheless, the broad picture of intertemporal movements is generally consistent with the average cross-country path indicated by the Kuznets curve. The underlying ob- servations for each country do not show any case of a decline in the per capita income of the lowest quintile. It is important to emphasize that the average cross-country relation should not be interpreted as an iron law. Individual countries that are able to establish the preconditions for a more egalitarian distribution of income and to stimulate growth in such a policy environment, as illustrated by Yugoslavia, Taiwan, and Korea, may well be able to avoid or moderate the phase of increasing inequality. Bnt there are a number of reasons why such a pattern is likely to emerge with a continuation of past policies, especially in the nonsocialist countries II. It should be noted that here we are referring to changes over time for the economy as a whole. This does not rule out the possibility of increased impoverish- ment in a part of the economy, for example, the rural are-as in general, or a par- ticularly depressed region. See, for example, Griffin and Khan (1978) for an exposition of the view that growth has been accompanied by increasing rural poverty. 12. Ahluwalia's (1976) estimates from which these aggregate measures are derived were made for quintiles, which show that most of the change in shares is concentrated in the upper 20 percent and lower 40 percent of the population. THE DIMENSIONS OF GLOBAL POVERTY 467 Figure 11-l. The Kuznets Curve with Country Observations Kuznets Curve 3,000 I I , I Mexico , Trend lines ,, I 1964-74 Taiwan ,, 1963-73 Yugoslavia ,, I 2,500 1963-73 Sri Lanka I 1965-76 Korea 1961-71 Costa Rica ~ .., 1953/5-63/5 India ;.... 1961-71 Philippines 1963-73 Turkey l:: 2,000 1964-74 Colombia 1960-70 Brazil [ 1963-75 1961-71 Mexico Peru o 'f"" 0. .B CIl -:S Y = 1,500 o<;) E 8 Country observations .5 1958 BA Burma .8 1968 CH Chile 'a (:l 1,000 . 1,000 1964/5 EG 1971 IND Egypt Indonesia ., KE' 1970 1969 1970 NC KE MA Ivory Coast Kenya Malaysia TA: 1970/71 PAK Pakistan 500 .' .PAK 1960 SEN Senegal IND /India Y= 500 1969 TA Tanzania 1962 TH Thailand ·.:BA 1970 TUN Tunisia t'TH 1959 ZA Zambia o 1,500 Pcr capita income of the bottom 60 percent (YH) (1970 ICP dollars) Note: 0 turning point of the Kuznets Curve. characterized by sharp inequalities in the initial distribution of pro- ductive wealth (including land) . For one thing, development typically involves a shift of population from the low income, slower growing agricultural sector to the high income, faster growing modern sector. This process, which is central to the dual economy theories of Lewis 468 GROWTH AND POVERTY IN DEVELOPING COUNTRIES (1954) and Fei and Ranis (1964), can be shown to generate a phase of widening inequality.13 This is especial1y true when the growth of the modern sector takes an increasingly capital intensive form, as in Mcxico and Brazil, with incomes per person employed rising relatively rapidly but with a limited increase in employment. It is less true of the more labor intensive form, illustrated hy Taiwan and Korea, which is characterized by high rates of absorption of labor by the modern sector and a more rapid approach to full employment. Policy clearly has an important role in determining which form predominates. lbere are several other factors that contribute to widening in- equality. Economic growth is likely to produce a more rapid rise in the demand for skilled labor compared with unskilled labor, leading to widening inequality in the early stages when the supply of skilled labor expands relatively slowly. These disequalizing factors often are exacerbated by an institutional and policy framework that is biased in favor of the modern, urban sectors of the economy, leading to an excessive flow of resources to these sectors and increasing the incen- tives for capital intensive production. Combining the available evidencc with these a priori considerations, we conclude that the most likely outcome associated with economic growth in poor countries is some increase in inequality. Tbe projec- tions discussed below adopt this assumption in the Base Case but depart from it in considering the effects of improved distributional policies. The use of the Kuznets curve in projections also implies that the distribution of income will improve in countries with a per capita income above 800 Iep dollars without specifying the effort required to redirect government policies. Needless to say, we cannot assume that this improvement will take place automatically. 'Ibe low inequality observed in the deve10ped countries today is as much the result of institutional evolution resulting from particular historical and political factors as of their level of development. It has been argued by Bacha (1977) that the observed reduction in inequality in the developed countries over the first half of this century arose from social and political changes following World War I that are not likely to be B. See Robinson (1976) and Ahluwalia (1976). Bell (1979) gives a more formal demonstration of the possibility of U-shaped cnrves relating to the wage share in simple dual economy models. CONSEQUENCES OF EXISTING POLICIES 469 replicated in countries approaching industrial maturity today.14 We note that of the countries in group C (table ll-l), all of which are past the turning point estimated from cross-country data, only Taiwan shows some evidence of experiencing the second phase of the Kuznets curve. However, although our projections may be overoptimistic about future developments in countries approaching industrial maturity, this assumption docs not affect our projections of global poverty. Consequences of Existing Policies Our analysis focuses on three aspects of development in eaeh country in the sample: (a) the growth of aggregate income; (b) the growth of population; and (c) changes in the distribution of income by deciles. Ibese measures can be combined to analyze the evolution of relative and absolute poverty over time for individual countries or groups of countries. The base projection The conceptual basis for our analysis is a disaggregation of income growth in each country into a separate growth pattern for each income class (decile), expressed in Iep dollars of constant purchasing power. When applied to our sample of thirty-six countries, this procedure yields a time series of per capita income for 360 population units. These can be aggregated to determine the numbers of people below a given poverty line in groups of countries as well as measures of relative inequality ",-ithin groups. This procedure will be applied both in analyzing past trends in individual countries and in detennining the distributional consequences of projected growth. Despite the variable quality of the data on income distribution, it is 14. The most important of these developments was the strengthening of or- ganized labor and its subsequent political role in developing a welfare state. This argument should not be overstated, however, since these institutional changes were themselves based on an underlying economic transformation in the role of labor arising from the phenomenon of labor scarcity and the greater role of human skills in the production process. Similar processes can be expected to occur in the future, although in different social and political contexts, and they are likely to strengthen tendencies toward greater equality, although perhaps not so soon as predicted by the cross·country estimates. 470 GROWTH AND POVERTY IN DEVELOPING COUNTRIES useful to retain individual countries as units of observation because of both the substantial difference in their initial conditions and the necessity of defining the scope for policy changes on a country basis. The analysis focuses on the results for groups of countries, since ag- gregation reduces the effects of errors in country data. The procedure indicated above is carried out in four steps: (a) estimation of the income level of cach country (Y/) for the past (1960-75) and projcction of this level for the future (1975-2000) ; (b) estimation of population (N/) by country for the same periods; (c) estimation of income shares by deciles (Di/) for each country and hence the level of income for each decile group; and (d) determination of the number of people (N/) below the absolute poverty liue in each year. Tbe results of step (c) can be used to compute measures of inequality for any country or group of countries. GROWTH IN INCOME AND POPULATION. 111e present study was de- signed to determine the distributional consequences of existing coun- try projections of GNP and population. 111ese projections have been madc by the World Bank in the context of a global analysis of inter- national trade and capital flows. They provide a point of departure (Base Case) from which to consider changes in internal and extemal policies. The Base Case incorporates changes in GNP growth expected to occur with some improvement in existing policies as well as changes in population growth that can be anticipated from existing demo- graphic trends. Table 11-2 gives the growth in population and GNP determined on this basis for the 1975-2000 periodYi Since our main concern is with the incidence of poverty, the as- sumptions made for the poorer countries are more important than those for the rest of the panel. In the past the economies of the poorer countries have grown more slowly because of the greater weight of 15. The growth rates are based on projections to 1985 or 1990 that underlie recent World Bank studies of the world economy (1977, 1979). These rates lie between the Base Case projection for 1980 to 1990 of the latter (5.6 percent) and the more optimistic projection (6.6 percent) . CONSEQUENCES OF EXISTING POLICIES 471 the agricultural sector (whose growth is limited by both demand and technology), lower rates of savings, and other structural factors. In addition, the international environment has been somewhat less favorable to the growth of poor countries and domestic policy failures have probably been more pronouneed.16 For the future some acceleration of growth is anticipated in several of the largest poor countries-India, Indonesia, and Bangladesh-so that the average of the group is expeeted to increase from 3.8 percent to 4.7 percent. Although only a limited acceleration of aggregate growth is projected for the other developing countries, this will be augmented by a fan in the rate of population growth as they move further into the demographic transition. CHANGES IN DISTRIBUTION. The specification of plausible changes in distribution raises two main issues. If existing policies continue, what changes in distribution can be expected as a result of the growth processes now under way? Conversely, if a government takes stronger measures to improve income distribution, what is likely to happen to growth? Although there is no consensus among economists on either of these questions, we shall make the following assumptions for the present analysis. In the first place, since time series data will not support a separate analysis of each country, we initially assume average behavior for countries having a given initial income level and distribution. More specifically, we assume that the cross-section estimates of the Kuznets relation discussed above are representative of the behavior of mixed economies over time in the absence of effective government action to alter them. ]11is leads to a worsening of income distribution for coun- tries below a per capita income of 800 ICP dollars. \Ve also assume that there is a tendency for income distribution to improve above this level, although the historical causes of this improvement have been as much political as economic. To translate these assumptions into a projection procedure, we construct a separate Kuznets curve for each country so the distribution estimated for 1975 (or any other year) can be projected to higher or 16. The relation between growth and income level is analyzed by Robinson (1971) and Chenery, Elkington, and Sims (1970). Chenery and Carter (1973) give an evaluation of the effects of internal and external factors on past growth for a sample of countries similar to the present one. Table Il-2. Sample Panel: Indexes of Growth and Distribution (Base Case) Number of people GNP Population Share of lowest in poverty growth rates growth rates 40 percent (millions)" 1975 Proiection 1975 Projection Country 1960-75 1975-2000 1960-75 1975-2000 estimate for 2000 estimate for 2000 Group A (income less than 350 ICP dollars) I Bangladesh 2.4 4.6 2.7 2.5 20.1 17.4 52 56 2 Ethiopia 4.3 4.1 2.1 2.6 16.8 15.0 19 25 3 Burma 3.2 2.5 2.2 2.1 15.7 15.2 20 29 -4 '-l 4 Indonesia 5.2 5.5 2.1 1.7 ]6.1 12.7 76 30 N 5 Uganda 4.0 3.2 2.9 2.8 14.4 14.0 6 12 6 Zaire 4.3 4.8 2.6 2.7 14.6 12.7 11 13 7 Sudan 3.0 6.0 2.8 2.8 14.5 12.0 10 8 8 Tanzania 6.8 5.4 2.9 2.9 14.3 12.3 8 9 9 Pakistan 5.6 5.2 3.2 2.7 16.5 14.5 32 26 10 India 3.6 4.5 2.3 1.9 17.0 14.6 277 167 Subtotal 3.8 4.7 2.4 2.1 16.7 13.9 510 375 Group B (income of 350-750 ICP dollars) 11 Kenya 7.0 5.9 3.2 3.5 8.9 7.7 7 11 12 Nigeria 7.1 5.2 2.6 2.9 13.0 11.8 27 30 13 Philippines 5.6 7.3 3.0 2.4 11.6 10.3 14 6 14 Sri Lanka 4.2 3.8 2.4 1.7 19.3 18.2 2 2 15 Senegal 1.5 4.0 2.1 2.4 9.6 8.9 1 2 16 Egypt 4.2 6.1 2.5 1.8 13.9 13.5 7 5 -. I 18 Ghana 2.7 2.l 2.6 2.9 11.2 11.9 2 6 19 Morocco 4.4 6.2 2.7 2.8 13.3 10.9 4 2 20 Ivorv Coast 7.7 5.8 3.3 2.9 10.4 10.4 1 1 Subtotal 5.5 5.8 2.7 2.6 12.0 10.0 81 70 Group C (income more than 750 ICP dollars) 21 Korea, Republic of 9.3 8.1 2.1 1.6 16.9 19.1 3 1 22 Chile 2.3 6.0 2.2 1.5 13.1 14.3 1 1 23 Zambia 3.4 4.9 2.9 3.3 13.0 12.9 0 1 24 Colombia 5.6 7.4 3.1 1.8 9.9 11.5 5 2 25 Turkev 6.4 6.3 2.2 2.1 9.3 10.4 6 4 26 Tunisia 6.1 7.5 2.5 1.9 11.1 13.3 1 0 27 Malavsia 6.7 6.7 2.8 1.8 11.1 13.3 1 1 4- 'l w 28 Taiwan 9.1 6.2 2.8 1.7 22.3 24.4 1 0 29 Guatemala 6.1 6.0 2.5 2.7 11.3 12.4 1 1 30 Brazil 7.9 2.9 2.6 9.1 11.9 16 7 31 Peru 5.7 6.3 2.8 2.5 7.3 8.8 3 2 32 Iran 9.5 7.2 3.1 2.4 8.2 11.0 5 2 33 Mexico 6.8 6.8 3.4 3.0 8.2 10.8 8 6 34 Yugoslavia 5.8 6.1 1.0 0.7 18.8 23.9 1 0 35 Argentina 4.0 4.5 1.5 1.0 15.1 18.5 1 1 36 Venezuela 5.8 6.8 3.5 2.9 8.5 12.9 1 1 Subtotal 6.4 6.9 2.6 2.2 9.9 10.0 54 30 Total 5.4 6.2 2.5 2.2 9.8 6.5 644 475 Source: GDP growth adapted from World Bank (1977). a. Totals may not add due to rounding. 474 GROWTH AND POVERTY IN DEVELOPING COUNTRIES lower levels of income. I7 The country-specific curves differ in their starting points but have the same curvature. The effect for representa- tive countries above or below the average distribution is shown by the solid lines in figure 11-2. The Base Case projection for a country such as India, which is close to the average to start with, would follow the Kuznets curve quite closely. Other countries, such as Brazil O[ Korea, retain their relative positions above or below the average distribution and in this sense are assumed to run "parallel" to the Kuznets curve. IS Although this is a highly stylized interpretation of the existing evidence, it is more plausible than assuming that there is no effect of eeonomie development and industrialization on distribution, which is the only obvious alternative. 19 111e question of the effect of changing distribution on growth arises later when we take up alternative distributional policies. We then assume that there is a reduction in growth proportional to the fall in the share of income (and saving) of the upper income group. The effects of this assumption are illustrated by the second set of projec- tions in figure 11-2, which are discussed below. By way of summary, the relations among the several determinants of poverty are shown schematically in figure 11-3. 1be links between internal and external policies and the extent of poverty in each country are traced through three intermediatc variables: population (N), per capita GDP (y), and income distribution (D). Projections of popula- tion growth and GDP are taken from other studies and become ex- ogenous variables in the Base Case. The distribution of income by decile (D,j) is generated as a function of the initial distribution and the average variation with per capita income. In the Base Case simulation there is no feedback from the distribu- tion of income to the rate of growth. But in simulations where we depart from the distributional behavior of the Base Case we anow for a feedback from income distribution to GDP growth. Other possible feedbacks, such as those from GDP and its distribution to population growth, are not analyzed explicitly and are indicated by dashed lines. 17. The formula for this function assumes that the two parameters in the quadratic function describing the Knznets curve apply to each country. See Ahlu- walia, Carter, and Chenery (1979, appendix). 18. TIle actual projections are made for each decile and are aggregated here to illustrate the overall effects. 19. We have, however, made caleulations based on the latter assumption to show the net effect of the Kuznets hypothesis. See Chenery and Carter (1977). CONSEQUENCES OF EXISTING POLICIES 475 Figure ] 1-2. Alternative Simulations, 1975-2000 Brazil 7,000 ,,Kuznets Curve ,,, ,i i 6,000 1: 8 Ii) 0- ~ 4,000 0- .8 Il) -:5 ..... o Il) § u 3,000 Y = 3,000 .5 .~ 0- 15 v .... '""' 2,000 Y = 2,000 · · · ,, :India 1,000 --Base Case Y = 1,000 - - - 45 percent to l'"dC bottom 60 percent j ° Per capita income of the bottom 60 percent (Y B) ( 1970 ICP dollars) 476 GROWTH AND POVERTY IN DEVELOPING COUNTRIES Figure 11-3. Schematic Representation of Causal Relations Underlying Simulation Models population f---------..-{ policy External environment and policies (GEeD growth, trade. aid) Numbers in poverty (N,) Internal environ· ment and policies (investment levels, human skilL~, economic efficiency, prior distribution of assets, policies to help poorer groups) - Indicates relations explicity modeled. - - - Indicates relations taken acconnt of by other studies but exogenous to the model. - - - Indicates relations that should be modeled but cannot be taken account of with our present state of knowledge. A mathematieal statement of the simulation model is given in Ahlu- walia, Carter, and Chenery (1979). Trends in inequality and poverty lne basic projections are given in table 11-3 for the three country groups and for the poorest 40 percent of the people in each group. lne trends for the 1960-75 period are simulated in the same way for purposes of comparison. These results bring out the main links be- tween overall growth and the changes in inequality and absolute poverty. INEQUALITY. The sources of growing inequality in the developing world as a whole can be summarized by means of the Theil index, CONSEQUENCES OF EXISTING POLICIES 477 which permits a decomposition of inequality into two elements: within and among countries. 20 This index and the proportion due to inequality among countries are shown below for past and future periods: 1960 1975 2000 Theil index 0.57 0.67 0.77 Proportion due to variation among eountries (pereentage) 32 37 50 There is a substantial inerease in overall inequality and in the pro- portion due to variation among countries. 'This increase in inequality can be examined in terms of the interaction of two factors: (a) the intereot1ntry lag between the countries of group A and all developing countries, and (b) the within-group lag in the growth of the poor within these countries in relation to the average. TIlUS, the growth of income of the poorest 40 percent in group A countries (who account for about 80 percent of the world's poor in our Base year) can be written as: C a Cpa (ll.l) C X C C a = 7JP"C , where Cpa is the growth of incomc of the poorest 40 percent in group A, C a is the growth of group A, C is the growth of all develop- ing eountries, and 7Jpa measures the total lag in the form of an ineome elasticity. Table 11-4 gives measures for these ratios for groups A and B. In the past the lag in the growth of poor countries (7Ja) was a more important faetor than the lag in growth of thc poor within these countries (7Jp) .21 For the future, we project higher growth rates for 20. The Theil index is defined as zy. In(ydni), where y. and n. are the shares of income and population of a given decile unit in the total income or population of the group of countries. This index varies between 0 in the case where all income shares are equal to population shares, and log lin}, where n; is the population share of the smallest group. Note that inequality is at a maximum when all income accrues to the smallest possible group. 21. If the two factors were measured for individual countries rather than for group A as a whole, the difference would be even greater. Table 11-3. Growth and Poverty in Developing Countries Projections for 2000 b 1960 Historical Country groups estimates 1975" trend Base Case C Per capita income of the total -/:. populations (Iep dollars '1 00 in 1970 prices) An developing countries 367 555 (2.8) 1,446 (3.9) 1,462 (4.0) income countries 233 284 (1.3) 464 (2.0) 536 (2.6) Middle income countries 337 1 (2.8) 1,179 (3.4) 1,189 (3.4) High income countries 714 1,220 (3.6) 3,829 (4.7) 3,724 (4.6) Per capita income of lowest 40 percent of the population (Iep dollars in 1970 prices) All developing coun tries 109 136 (1.5) 205 (1.7) 236 (2.2) Low income countries 102 118 (1.0) 159 (l.2) 186 (1.8) Middle income countries 104 1 (2.6) 300 (2.7) 288 (2.6) High income countries 174 301 (3.7) 961 (4.8) 1,114 (5.1) Income shares of the lowest 40 percent of the population (percentages) An developing countries 11.9 9.8 5.7 6.4 Low income countries 17.5 16.7 13.7 13.9 Middle income countries 12.4 12.0 10.1 9.7 High income countries 9.7 9.9 10.0 12.0 Number of poor (millions) All developing countries 597 644 589 475 Low income countries 438 510 493 375 Middle income countries 86 81 56 70 High income countries 72 54 40 30 -!:>. '-1 \Q Percentage of population in poverty All developing coun tries 50.9 38.0 20.2 16.3 Low income countries 61.7 50.7 29.5 22.4 Middle income countries 49.2 31.0 11.4 14.2 High income countries 24.9 12.6 5.4 4.0 a. Figures in parentheses are annual growth rates between 1960 and 1975. b. Figures in parentheses are anllual growth rates between 1975 and 2000. e. Growth rates adapted from World Bank (1977). 480 GROWTH AND POVERTY IN DEVELOPING COUNTRIES Table 11-4. Growth Lags within and among Country Groups Inter- Within Total Country group group elasticity group Period" G pi Gi 'YJi 'YJp 'YJpi A. Low I 1.0 1.3 0.46 0.77 0.35 II 1.8 2.6 0.65 0.69 0.45 B. Middle I 2.6 2.8 1.00 0.93 0.93 II 2.6 3.4 0.85 0.76 0.65 Total I 1.5 2.8 0.54 0.54 II 2.2 4.0 0.55 0.55 a. Period I, 1960-75, period II, 1975-2000; "11 = GdG, "1. Gp./G ·. several of the poor countries-notably India and Bangladcsh-which ",ill raise the intergroup ratio from 0.46 to 0.65. The worsening of internal distribution will be accentuated by more rapid growth, how- ever, so that the two factors become of roughly equal importance in explaining the projected lag in the income growth of the poor (0.45). For the middle income group, on the other hand, both lags are pro- jected to increase. As a result no improvement in the total elasticity (0.55) is projected for developing countries as a group with existing policies. POVERTY. Tbe lag in the growth of income of the poorest 40 percent in low-income countries is obviously at the core of the problem of world poverty. Trends in absolute poverty are shown in the last two sections of table 11-3 and in figure 11-4. In 1975 the numbers of the poor were still increasing in almost all of the countries in group A but had started to decline in most countries in groups Band C. For the future a continued rise is projected in a number of the poor countries, but that rise is offset by a reversal of this trend in Indonesia, India, and Pakistan. The net effect is a decline in absolute poverty of about 25 percent between 1975 and 2000. 22 In overall terms the numher of absolute poor in all developing countries (except for the centrally planned economies) would be on the order of 600 million people in the year 2000 when allowance is 22. The less optimistic projection in table 11-3, based on historical trends in each country, shows only a modest reduction in poverty in 2000 to the level existing in 1960. CONSEQUENCES OF EXISTING POLICIES 481 made for countries omitted from our sample. This figure defines the magnitude of the problem to be addressed in devising policies that will lead to a more rapid reduction in poverty. In relative terms these projections represent impressive progress in reducing poverty-from about 50 percent of the population of developing countries in 1960 to 16 percent in 2000-but they fall considerably short of the results that might be achieved with more effective policies. Variation in country experience Although we have usable timc series data for only twelve countries, they include a considerable variety of experience with both growth and distribution. Table 11-5 contains the statistical data underlying the graphic analysis of these countries in figure 11-1. This table also shows the growth rates of per capita income of the lower 60 percent and its relation to the country average ('YJp). The countries in table 11-5 have been classified into three groups on the basis of two criteria: the income share of the lower 60 percent in the latest year and the share of the increase in income going to this group over the period. The five countries in group I have a terminal share ranging from 28 to 39 percent and incremental shares above 30 percent. At the other extreme, the three countries in group III have terminal shares in the range of 18 to 21 percent and incremental shares of only 16 to 18 percent. Distribution in the former group is conSiderably better than the Kuznets curve in figure 11-1 and in the latter it is considerably worse. The four countries in group II are close to the average relation. To determine the scope for improvement in distributional perform- ance, it is useful to focus on the experience of group I. Taiwan, Yugo- slavia, and Korea have maintained relatively good distribution along with rapid growth over the period indicated. They have in common a relatively equal distribution of assets at the beginning of the period observed, largely as a result of political changes following World War II. The Taiwan-Korea development strategy included substantial initial land reforms, great emphasis on education, and an overall strategy favoring labor-intensive expansion in the nonagricultural sec- tors, especially labor-intensive manufactured exports. 23 In Yugoslavia, 23. TIle experience of Taiwan is analyzed by Fei, Ranis, and Kuo (1979); for Korea, see Hasan, Rao, and others (1979). Table 11-5. Changes in Income Distribution in Selected Countries (in 1970 ICP dollars) Increments in Share of bottom income per capita 60 percent Growth rate Income Income elasticity Period per of of capita Top Bottom Bottom bottom observa- in initial 40 60 Incre- 60 60 Country tion year Total pcrcent percent Initial Final mental Total percent percent 1. Good performance Taiwan 1964-74 562 508 758 341 0.369 0.385 0.395 6.6 7.1 1.1 ~ Yugoslavia 1963-73 1,003 518 822 316 0.357 0.360 0.365 4.2 4.3 1.0 00 N Sri Lanka 1963-73 388 84 58 101 0.274 0.354 0.513 2.0 4.6 2.3 Korea, Republic of 1965-76 362 540 938 275 0.349 0.323 0.311 8.7 7.9 0.9 Costa Rica 1961-71 825 311 459 212 0.237 0.284 0.336 3.2 5.1 1.6 II. Intermediate performance India 1954-64 226 58 113 21 0.310 0.292 0.258 2.3 1.6 0.7 Philippines 1961-71 336 83 155 35 0.247 0.248 0.250 2.2 2.3 1.0 Turkey 1963-73 566 243 417 128 0.208 0.240 0.279 3.6 5.1 1.4 Colombia 1964--74 648 232 422 106 0.190 0.212 0.240 3.1 4.3 1.4 III. Poor performance Brazil 1960-70 615 214 490 31 0.248 0.206 0.155 3.1 1.2 0.4 Mexico 1963-75 974 446 944 114 0.217 0.197 0.180 3.2 2.4 0.8 Peru 1961-71 834 212 435 63 0.179 0.179 0.179 2.3 2.3 1.0 CONSEQUENCES OF EXISTING POLICIES 483 the socialization of the means of production combined with large transfers of income from richer to poorer regions have been the principal factors favoring egalitarian growth.24 These three countries have provided the largest absolute increases to the income of the poor, with incremental shares accruing to the lower 60 percent ranging be- tween 0.31 and O.4D. By contrast, Sri Lanka provides an example of a sustained policy of improving distribution through income transfers, without the parallel achievement of high growth.25 Tne incremental share of the lower group in Sri Lanka exceeded 50 percent, but there was also a reduction in resources available for investment that contributed to lower growth and rising unemployment. In summary, this range of country experience suggests the following conclusions that provide a background for our projections: (a) Differences in distributional policies have been at least as important to poverty alleviation as differences in aggregate growth rates. (b) A marginal share for the lowest 60 percent of income recipients of about 40 percent is as high as has been observed in countries in which growth has been sustained at reasonable levels. 26 (c) Substantial improvements in income distribution have taken place under a variety of policies. The market mechanism has been the principal instrument in Taiwan, which was not impeded by highly unequal distribution of land and other assets to start with. Income transfers for investment or consumption have been important in Yugoslavia and Sri Lanka. (d) The time series evidence snpports the cross-section results as far as the worsening phase of inequality is concerned. There is no documented case of a country that has avoided the initial worsening in income distribution that is implied by uneven sec- 24. Elements of the Yugoslav strategy are discussed by Sehrenk, Ardalan, and EI Tatawy (1979). 25. The effeets of these policies are analyzed by Isenman (forthcoming). 26. The only higher increment in our sample is Sri Lanka (51 percent), where high redistribution through the budget impeded the growth of the poor as well as the rich and has since been modified. A similar modification of strongly redistribu· tive policies in favor of growth has recently taken place in Tanzania and Cuba although information is not available as to their effects. 484 GROWTH AND POVERTY IN DEVELOPING COUNTRIES toral growth; Taiwan has come the closest to maintaining the relatively equal shares that are typical of the poorest countries.27 (e) Although there is little time series evidence for develop- ing countries at higher income levels, the theoretical case for im- provement as a result of the automatic working of economic forces is not strong. Mexico and Brazil illustrate the likelihood of continued worsening of income distribution well above the level of 800 Iep dollars in the absence of effective policies to counter- aet this tendency.28 The Scope for Improvement The prospects for 2000 described by the Base Case can scarcely be regardcd as satisfactory either by the countries concerned or by the world community. These possibilities are a far cry from targcts sueh as abolishing absolute poverty by the end of this eentury. In this section we shall examine a realistic range of possibilities for further reducing poverty through a combination of accelerating GNP growth, improving the distribution of income, and reducing population growth. Alternative policies Tbe first step is to establish what improvements in performance are feasible in each of these dimensions on the basis of country experi- ence. 29 We then examine the cffect that such improvements might have on global poverty by 2000. The effects of each type of policy will be simulated separately and then in various combinations. ACCELERATING GROWTH. We have seen that the GNP growth rates used for the Base Case projection already imply a significant accelera- tion in growth of the low income countries. The studies upon which 27. Fei, Ranis, and Kuo (1979) show that there was some worsening in distri" bution up to 1968 and some improvement since then. 28. The Brazilian experience has been widely debated. See, for example, Baeha and Taylor (fortheoming), Langoni (1973), and Fishlow (1972). 29. We examined but rejected the alternative of deriving an optimum degree of redistribution by assuming a welfare function and a hypothetical growth tradeoff, since it is not possible to gct plausible estimates of the latter. THE SCOPE FOR IMPROVEMENT 485 these projections are based conclude that an acceleration of this order can be achieved largely through domestic policy changes aimed at increasing domestic savings and efficiency in resource use without a major change in the international environment. Higher levels of concessional assistance and improvements in in- ternational markets would make higher growth rates possible for these countries. To establish an optimistic range for this improvement, we assume that the growth rates for the countries in group A could be increased by 1 percentage point for the period 1980 to 2000. This increase corresponds to the optimistic alternative used in current pro- jections of the World Bank (1979). An increase in growth in the poor countries from 4.7 to 5.7 percent will require an incrcase in foreign exchange availability of about the same magnitude. The international policies required to achieve such an increase include substantial trade liberalization, particularly in the products that can be exported by the poorer countries, and an increase of some 20 percent in concessionallending to the poor countries. Al- though this increase would imply that a larger share of GNP be devoted to official development assistance (ODA) by the OECD countries-the present level is 0.35 percent-it would be substantially less than the international target of 0.70 percent if it could be concentrated in the poorest countries.30 IMPROVING DISTRIBUTION. The experience with policies affecting in- come distribution has been discussed in the previous section. The best results were obtained by Taiwan, Yugoslavia, and Korea, in which between 30 percent and 40 percent of the increment in income went to the bottom 60 percent of the population and rapid growth was sustained. These countries now have distributions that compare favorably with the industrialized countries.:u To estimate an upper limit to the possibility for redistributing in- come without substantially disrupting growth from this experience, we assume that 45 percent of the increment of G!' will go to the bottom 60 pereent. This amount is higher than the ineremental share observed in any developing eountry except Sri Lanka, where growth 30. Fuller discussions of the possibilities of increasing growth of the poor coun- tries and the implications for trade and aid policies are given by the World Bank (1978,1979) . 31. See Ahluwalia (1976). 486 GROWTH AND POVERTY IN DEVELOPING COUNTRIES was substantially reduced; it corresponds to an incremental sharc of about 25 percent to the lowest 40 percent, which is as high as the average share in almost any conntry.32 Although this assumption may be technically feasible for any country, it is barely conceivable for all developing countries. The requirements of such a strategy have been discussed extensively both in general terms and for particular countries.a3 Many elements of distributional policies remain untested and speculative, but there is substantial agreement that the benefits of growth accruing to the poor can be increased through policies that (a) increase "linkage" of the poor to the faster growing segments of the economy so as to in- crease the flow of indirect benefits; and (b) provide much greatcr direct support to productive activities upon which the poor are heavily dependent and which have a potential for efficient expansion. Some of the elements of such a strategy serve both to increase GNP and to improve its distribution. Policies intended to remove incentives for excessive use of capital in individual sectors and thus help to in- crease employment are obvious examples. But there are also policies which may have adverse effects on GNP growth, at least in the short run. Diversion of investment resources into activities aimed at im- proving the productivity of the poor may involve such costs in the initial stages. In many countries an adequate flow of the benefits of growth to the poor can only be ensured if steps are also taken to correct the highly skewed distribution of productive assets, especially agricultural land. \Ve conclude that on both theoretical and empirical grounds some loss of growth can be anticipated if a target of distributing 45 percent of the increment of GNP to the lowest 60 percent is achieved. The most familiar argument for the growth loss is the decline in saving that is likely if income is shifted from the rich to the poor, since private saving, which is done mainly by upper income groups, is likely to fall. There are also adverse effects on ineentives to domestie private in- vestors arising from the adoption of radieal distributional policies. To allow for these adverse effects we have assumed a growth loss arising from the implementation of the distributional objective. Specifically, we postulate that the rate of growth of GNP \hill fan in proportion to 32. Ibid. 33. See CheneI)', Ahluwalia, Bell, Duloy, and Jolly (1974) and the country studies cited above. THE SCOPE FOR IMPROVEMENT 487 the decline in the income share of the richest decile compared with its share in the Base Case.34 REDUCING POPULATION GROWTH. The third element affecting the scale of global poverty is the rate of growth of population. Tbe scope for further improvements in this area over the next two decades above what is assumed in the Base Case should not be exaggerated. There are a number of factors contributing to rapid population growth in developing couutries, not least the decline in mortality initially gener- ated by the control of communicable diseases, which can be expected to continue in the future with a sustained rise in living standards. An offsetting decline in fertility has begun in many countries, the rate being determined by a large number of socioeconomic changes. Popu- lation control programs can facilitate this process when other condi- tions are favorable, but they have not been shown to accelerate it to any great degree. For the present analysis, we assume that the feasible scope for re- ducing population growth in each country is given by the difference in growth rates between the Medium Variant in the United Nations (1975) population projections and the Low Variant, modified for more recent information on fertility. \Ve have accordingly reduced the population growth rates of the Base Case in the same proportion.lI5 The results for groups of developing countries are shown below. "Re- duced Population Gro\\1h" implies roughly a 10 percent decline in the rate of increase in each group of countries to 2000. This limited decline reflects the substantiallcad time required for population con- trol policies to take effect. Our alternative assumptions for population growth rates are: Reduced population Country group Base Case growth ----------------------------------------------------- A. Low income countries 2.13 I.S5 B. Middle income countries 2.41 2.22 C. High income countries 2.27 2.06 All developing countries 2.21 1.97 34. For a detailed statement of the tradeoff mechanism, see Ahluwalia, Carter, and Chenery (1979). 35. Base Case population growth rates come from background material for World Bank (1978). 488 GROWTH AND POVERTY IN DEVELOPING COUNTRIES A1ternative projections The effectiveness of these three kinds of policies in alleviating poverty will be demonstrated by simulating their effects to 2000 and comparing them to the Base Case. The analysis is designed to bring out the relative effectiveness of each approach for the main gronps of countries and to define feasible objectives for international action. Although we have examined a large number of policy mixes, the principal results can be summarized by considering first the three policies in isolation and then in two combinations. 111ese polieies are set out in table 11-6 in the order of their effectiveness in reducing poverty. Since differences in performance within the country groups add little to the general conclusions, we assume that each policy ap- plies equally to all countries in a groUp.36 We shall use two measures of poverty alleviation in evaluating the results: the number of people below the absolute poverty line and the per capita income of the relatively poor, defined by the lowest 40 percent. OPTION A: REDUCED POPULATION GROWTII. Since the Base Case is relatively optimistic about the possibilities of lowering fertility, the scope for further reductions in population growth in this century is limited. Given the young age structure in developing countries, even a dramatic dccline in total fertility rates would not lead to a much more rapid decline in population growth until after 2000. Simulations based on the lower UN estimates reduee absolute poverty by about 15 percent in 2000 compared with the Base Case and increase the per capita income of the bottom 40 percent by 10 percent. As shown be- low, however, the impact of population policy is greater in conjunc- tion with other mcasures. OPTION B: ACCELERATED GROWTH OF POOR COUNTRIES. TIlis option illustrates the principal contribution that policies of the advanced countries make to alleviating world poverty. A 1 percent increase in 36. The differences in initial conditions produce a considerable variation in country behavior, however. as is illustrated in the poverty projections in figure U·S below. THE SCOPE FOR IMPROVEMENT 489 growth of poor countries would eliminate in large part the growth lag between group A and the other two groups. The result is to reduce world poverty by 30 percent in 2000 compared with the Base projection. OPTION C: INCREMENTAL REDISTRIBUTION. This option assumes that all developing countries adopt effective redistributive policies which ensure an incremental share of 45 percent of increased income for the lowest 60 percent with a moderate loss in per capita growth (from 4.0 to 3.5 percent overall). In the low-income countries this would have about the same effect on reducing poverty and raising the in- comes of the poor as did option B. Further reductions in poverty would Come in the middle- and high-income groups. On balance, the results suggest that focusing on improvements in distribution may be as effective in redueing world poverty as the accelerated growth option. OPTION D: REDISTRIBUTION PLUS ACCELERATED GROWTH. Although the results of combining options Band C are not entirely additive, the two policies together produce a substantial improvement over either one in isolation. The combination of accelerated growth of poor countries and a larger incremental share to the poor within each country completely eliminates the growth lag between the poor and the rest of the popnlation in developing conntries. As a result, al- though the total GNP of developing countries would be 5 percent less in 2000, the per capita income of the bottom 40 percent would be 45 percent higher. A tradeoff of this magnitude would clearly be de- sirable on most social welfare functions. OPTION E: MAXIMUM IMPROVEMENT. The greatest improvement on our assumptions results from combining all three policies, which is equivalent to adding lower population growth to option D. The most significant effect is on the number of absolute poor, which is reduced by a further 15 percent from option D. The main interest in this re- sult, which compounds three optimistic sets of assumptions, is to show that the elimination of absolute poverty by the year 2000 is not a credible policy objective. Even though the number of poor is re- duced to a third of its present level, more than 200 million remain below the poverty line. Table 11-6. Alternative Scenarios for 2000 Accelerate GNP 45 percent growth of incremental Reduce low income share population countries to lowest Options Options Country groups growth 1 percent 60 percent BandC A, B, andC Option A B C D E Per capita income of the total populations (Iep dollars ""'- '-0 in 1970 prices) Q All developing countries 1,557 (4.2) 1,545 (4.2) 1,326 (3.5) 1,384 (3.7) 1,470 (3.9) Low income countries 573 (2.8) 680 (3.6) 488 (2.2) 590 (3.0) 626 (3.2) Middle income countries 1,254 (3.7) 1,189 (3.4) 1,042 (2.9) 1,042 (2.9) 1,093 (3.1) High income countries 3,961 (4.8) 3,724 (4.6) 3,394 (4.2) 3,394 (4.2) 3,608 (4.4) Per capita income of lowest 40 percent of the population (ICP dollars in 1970 prices) An developing countries 251 (2.4) 287 (3.0) 284 (3.0) 342 (3.8) 367 (4.0) Low income countries 197 (2.0) 230 (2.8) 232 (2.7) 290 (3.7) 312 (3.9) Middle income countries 308 (2.8) 288 (2.6) 374 (3.6) 374 (3.6) 407 (4.0) High income countries ],208 (5.7) 1,114 (5.4) 1,352 (6.2) 1,352 (6.2) 1,450 (6.5) Income shares of the lowest 40 percent of the population (percentages) All developing countries 6.5 7.4 8.6 9.9 Low income countries 13.8 13.9 19,0 19.7 Middle income countries 9,8 9.7 14.4 14.4 14.9 High incomc countries 12.2 12.0 15.9 15.9 16.1 Number of poor (millions) All developing countries 407 335 305 263 221 Low income countries 315 235 232 ]90 157 income countries 6] 70 44 44 38 High income countries 31 30 29 29 26 Percentage of population in """ \0 >- poverty All developing countries 14.9 ll.5 10.5 9.0 Low income countries 20.1 14.0 13.9 11.4 Middle income countries 13.1 14.2 8.9 8.9 8.1 High income countries 4.4 4.0 3.9 3.9 3.7 Note: Figures in parentheses are percentage annual growth rates between 1970 and 2000. 492 GROWTH AND POVERTY IN DEVELOPING COUNTRIES Policy implications To bring out the policy implications of these simulations, we shall restate the results in a more general form, first from the standpoint of individual countries and then for international policy. Finally, we return to the question of defining policy objectives in more realistic terms. Figure 11-4 shows the effects of both raising income and improving its distribution on the level of absolute poverty in a representative country.1l7 This figure shows that at low income levels there is rela- tively little difference in the poverty level between the average (Kuznets curvc) and the best observed distribution. In the middle income range, however, the proportion of the population in poverty is much more sensitive to changes in distribution. The figure illustrates alternative combinations of increased per capita income and improved distribution that would lead to any desired reduction in povcrty. In exploring policy alternatives for individual countries, it is also necessary to take account of the rate of growth that can be generated with the existing economic structure. On the one hand, a poor coun- try, such as Indonesia, that can move quite rapidly parallel to the Kuznets curve may make more progress in reducing poverty than a slowly growing country with better distribution, such as India or Bangladesh, as is illustrated in figure 11-5. On the other hand, in the typical middle income countries, which tend to have more rapid growth and less equal distributions, improved distribution is often more effective in reducing poverty than is accelerated growth. \Vhat- ever the starting point, however, it should not be necessary to wait until per capita income rises above 1,000 rep dollars to reduce absolute poverty below 10 percent. Turning to the relation between national and international policies, we have designed figure 11-6 to illustrate the possible tradeoffs in- volved. International trade and aid policies have their most substantial effect on grmvth and exert relatively little influence on internal dis- 37. Since the figure is computed for a Lorenz curve based on a lognormal dis- tribution, it needs to be modified to apply to the actual distribution of any given country. Thc country scatter is plotted from poverty shares, shown in table 11-1, which are not entirely consistent with the income shares shown because of differ- ences in indi. .idual distributional patterns. Figure 11-4. E ffeets on Poverly of I mprovitzg Distribution 60 r · Indonesia Kenya · · Country observations in 1975 >. Philippines t · Senegal .lonetary Fund Staff Papers 14 (Novem- bcr): 478-97. MachI up, F. 1963. Essays in Economic Semantics. Englewood Cliffs, N.J.: Prentice-Hall. Mahalanobis, P. C. 1955. "The Approach of Operational Research to Planning in India." Sankhya 16 (December): 3-131. Mamalakis, M., and Reynolds, C. 1965. Essays on the Chilean Economy. Homewood, Ill.: Irwin. Mandclbaum, K. 1945. The Industrialization of Backward Areas. Oxford: Blackwell. Manne, A. S. 1963. "Key Sectors in the Mexican Economy." In Studies in Process Analysis, ed. A. S. Manne and H. M. Markowitz. New York: John \Viley and Sons. - - - . 1966. "Key Sectors in the Mexican Economy, 1962-72." In The Theory and Design of Economic Development, ed. I. Adelman and E. 111Orbecke. Baltimore: Johns Hopkins Press. Manne, A. S., and others. 1967. Investment for Capacity Expansion: Size, Location, and Time-Phasing. London: Allen and Unwin. )08 REFERENCES Margolis, J. 1957. "Welfare Criteria, Pricing, and Decentralization of a Public Service." Quarterly Tournal of Economics 71 (August): 448-63. McKinnon, R. I. 1964. "Foreign Exchange Constraints in Economic Development and Efficient Aid Allocation." Economic TournaI 74 (June): 388-409. McManus, M. 1959. "Comments on the General Theory of Second Best." Review of Economic Studies 26 (June): 209-24. Michaely, M. 1977. "Exports and Growth: An Empirical Investigation." TournaI of Development Economics 4 (January): 49-53. Mikesell, R. F., and Zinser, J. E. 1973. "The Nature of the Savings Function in Developing Countries: A Survey of the Theoretical and Empirical Literature." TournaI of Economic Literature 11 (March): 1-26. Millikan, M. 1966. "Comment." In The Theory and Design of Economic Development, ed. I. Adelman and E. Thorbeeke. Baltimore: Johns Hopkins Press. Mishan, E. J. 1962. "Second Thoughts on Second Best." Oxford Economic Papers n.S. 14 (October): 205-17. Montias, J. M. 1959. "Planning with Material Balances in Soviet-type Economies." American Economic Review 49 (December): 963-85. Morawetz, D. 1976. "Elasticities of Substitution in Industry: What Do we Learn from Econometric Estimates?" World Development 4 (January): 11-15. - - - . 1977. Twenty-five Years of Economic Development, 19.50 to 1975. Baltimore: Johns Hopkins University Press. Musgrave, R. A. 1969. Fiscal Systems. New Haven: Yale University Press. Myint, H. 1958. "The Classical Theory of International Trade and the Underdeveloped Countries." Economic TournaI 68 (June): 317-37. Myrdal, G. 1957. Economic Theory and Underdeveloped Regions. London: Duckworth. Negishi, T. 1962. "On the Stability of a Competitive Economy: A Survey Article." Econometrica 30 (October): 635-69. Neisser, H. 1956. "Investment Criteria, Productivity, and Economic Development." Quarterly Journal of Economics 70 (November): 644-46. Netherlands Central Planning Bureau. 1956. Scope and Methods of the Central Planning Bureau. The Hague. Nurkse, R. 1953. Problems of Capital Formation in Underdeveloped Countries. Oxford: Blackwell. - - - . 1957. "Balanced and Unbalanced Growth." In Lectures in Economic Development. Istanbul: Faculty of Economics; Ankara: Faculty of Political Sciences. REFERENCES 509 ---.1959. Patterns of Trade and Development. Stockholm: Almquist and \Viksell. - - - . 1961. "Balanced and Unbalanced Growth." In Equilibrium and Growth in the World Economy, ed. G. Haberler and R. M. Stern. Cambridge, '\1ass.: Harvard University Press. Ohlin, P. G. 1959. "Balanced Economic Growth in History." Papers and Proceedings of the American Economic Association 49 (May): 338-53. Oort, C. J. 1958. Decreasing Costs as a Problem of Welfare Economics. Amsterdam: Drukkerij Holland NV. Organization for Economic Cooperation and Development. 1965. Devel- opment Assistance Efforts and Policies: 1965 Review. Paris: OEeD for the Development Assistance Committee. Pack, H. 1971. Structural Change and Economic Policy in Israel. New Haven: Yale University Press. Papenek, G. F. 1972. "The Effeet of Aid and Other Resource Transfers on Savings and Growth in the Developed Countries." Economic Journal 82 (September): 934-50. ---.1978. "Economic Growth, Income Distribution, and the Political Process in Less Developed Countries." In Income Distribution and Economic Inequality, ed. Z. Grilichcs and others. Frankfurt: Campus Verlag. Patinkin, D. 1960. The Israel Economy: The First Decade. Jerusalem: The Falk Project for Economic Research in Israel. Paukert, F. 1973. "Income Distribution at Different Levels of Develop- ment: A Survey of Evidence." International Labour Review 108 (August-September): 97-125. Philippines National Economic Council. 1957. The Five-Year Economic and Social Development Program for Fiscal Years 1957-1961. Manila. Polak, J. J. 1943. "Balance of Payments Problems of Countries Re- constructing with the Help of Foreign Loans." Quarterly Journal of Economics 57 (February) : 208-40. Prakash, V. 1978. "Patterns of Industrial Development - Basic Tables and Diagrams'," Development Economics Department memorandum (July). Washington, D.C.: World Bank. Prebisch, R. 1950. The Economic Development of Latin America and its Principal Problems. New York: United Nations Economic Com- mission for Latin America. - - - . 1959. "Commercial Policy in the Underdeveloped Countries." Papers and Proceedings of the American Economic Association 49 (May): 251-73. Qayum, A. 1960. Theory and Policy of Accounting Prices. Amsterdam: North-Holland. 510 REFERENCES Raduchel, W. J. 1970. "Technical Supplement to 'Substitution in Plan- ning Models'." Cambridge, Mass.: Harvard Univcrsity Project for Quantitative Research in Economic Development. Ranis, G., and Fei, J. C. H. ]961. "A Theory of Economic Develop- ment." American Economic Review 51 (September): 533-65. Reynolds, L. G. ]969. "Economic Development with Surplus Labor: Some Complications." Oxford Economic Papers 21 (March): 89-103. Robinson, S. 1971. "Sources of Growth in Less Developed Countries: A Cross-section Study." Quarterly Journal of Economics 8) (August): 391-408. - - - . 1976. "A Note on the U Hypothesis Relating Income Inequality and Economic Development." American Economic Review 66 (June): 437-40. Rosenstein-Rodan, P. ] 943. "Problems of Industrialization in Eastern and Southeastern Europe." Economic Journal 53 (June-September): 202-11. - - - . 1961a. "International Aid for Underdeveloped Countries. Review of Economics and Statistics 43 (May): lO7-38. - - - . 1961b. "Notes on the Theory of the 'Big Push'." In Economic Development for Latin America, ed. H. S. Ellis and H. C. vVallich. London: Macmillan. 1963. "Determining the Need for and Planning the Use of External Resources." In Organization, Planning, and Programming for Economic Development, vol. 8 of the U.S. Papers Prepared for the United Nations Conference on the Application of Science and Tech- nology for the Benefit of the Less Developed Areas. \Vashington, D.C.: U.S. Government Printing Office. Rostow, vV. \V. 1956. "The Take-off into Self-sustained Growth." Economic Journal 66 UvIarch} : 2 :;-48. Saito, K. \V. 1977. "The Determinants of Savings Behavior: A Survey of the Evidence." Development Economics Department memorandum (March). \Vashington, D.C.: \Vorld Bank. Samuelson, P. A. 1951. "Abstract of Theorem concerning Substitutability in Open Leontief Motels." In Activity. Analysis of Production and Allocation, ed. T. C. Koopmans. New York: \Viley. Sandee, J. 1959. A Long-Term Planning Model for India. U.N. Report no. TAA/IND/22. New York: United Nations. - - - . 1960. A Demonstration Planning Model for India. New York: Asia Publishing House. Schrenk, M., Ardalan, C., and EI Tatawy, N. A. 1979. Yugoslavia: Self-management Socialism and the Challenges of Development. Balti- more: Johns Hopkins University Press. REFERENCES 511 Schultz, T. W. 19')6. "Latin American Economic Policy Lessons." Papers and Proceedings of the American Economic Association 46 (May): 425-32. Scitovsky, T. 19')4. "Two Concepts of External Economics." Journal of Political Economy 62 (April): 143-')1. ---.1959. "Growth Balanced or Unbalanced." In The Allocation of Economic Resources, ed. M. Abramovitz and others. Stanford: Stanford University Press. Sen, A. K. 1957. "Some Notes on the Choice of Capital Intensity in Development Planning." Quarterly Journal of Economics 71 (Novem- ber): 561-84. Sheahan, J. 1958. "International Specialization and the Concept of Balanced Growth." Quarterly Journal of Economics 72 (May): 183-97. Singer, H. \V. 1949. "Economic Progress in Underdeveloped Countries." Social Research 16 (March): 1-11. - - - . 1950. "The Distribution of Gains between Investing and Bor- rowing Countries." Papers and Proceedings of the American Economic Association 40 (May): 473-85. Singh, S. K. 1972. "TIlC Determinants of Aggregate Savings." \Vorld Bank Staff \Vorking Papers, no. 127 (March). 'Washington, D.C.: World Bank. Squire, L., and van cler Tak, H. G. 1975. Economic Analysis of Projects. Baltimore: Johns Hopkins University Press. Stigler, G. 1946. "Production and Distribution in the Short Run." In Readings in the Theory of Income Distribution, American Economic Association. Philadelphia: Blakiston. Stoleru, L G. 1965'. "An Optimal Policy for Economic Growth." Econometrica 33 (April): 321-48. Strectcn, P. 1959. "Unbalanced Growth." Oxford Economic Papers 11 (June): 167-90. Strout, A. 1M. 1969. "Korea's Use of Foreign and Domestic Resources: A Cross-Country Comparison." In Practical Approaches to Develop- ment Planning: Korea's Second Five-Year Plan, ed. Irma Adelman. Baltimore: Johns Hopkins Press. Sutcliffc, R. B. 1964. "Balanced and Unbalanced Growth." Quarterly Journal of Economics 78 (November): 621-40. Syrquin, M. 1976. "Sources of Industrial Growth and Change: An Alter- native Measure." Paper read at the European Meeting of thc Econo- metric Society, August, at Helsinki, Finland. Syrquin, M " and Elkington, H. 1978. "'I11e Data Base for the Simulation Model. Dcvelopmcnt Economics Department memorandum (Novem- ber). Washington, D.C.: World Bank. 512 REFERENCES Taylor, L. J. 1969. "Development Patterns: A Simulation Study." Quarterly Journal of Economics 83 (May): 220-41. 1971. "Investment Timing in Two-Gap Models." In Studies in Development Planning, ed. H. B. Chenery. Cambridge, Mass.: Harvard University Press. Taylor, L. J., and Lysy, F. J. Forthcoming. "Income Distribution Simula- tions, 1959-79." In Models of Growth and Distribution in Brazil, ed. E. Bacha and L. J. Taylor. New York: Oxford University Press. Tendulkar, S. D. 1971. "Interaction between Domestic and Foreign Resources in Economic Growth: Some Experiments for India." In Studies in Development Planning, cd. H. B. Chencry. Cambridge, Mass.: Harvard University Press. Theil, H. 1958. Economic Forecasts and Policy. Amsterdam: North- Holland. Tims, \Y. 1965. "Growth Model for the Pakistan Economy: Macro- Economic Projections for Pakistan's Third Plan" (March). Karachi: Planning Commission, Government of Pakistan. Tinbcrgcn, J. 1955. "The Relevance of Theoretical Criteria in the Selec- tion of Investment Plans." In Investment Criteria and Economic Growth, ed. M. Millikan. Cambridge: Cambridge University Press. - - - . 1956. Economic Policy: Principles and Design. Amsterdam: North-Holland. - - - . 1958. The Design of Development. Baltimore: Johns Hopkins Press. United Nations. 1954. A Study of the Iron and Steel Industry in Latin America, II, G.3, vol. 1. New York: UN Economic Commission for Latin America. - - - . 1955. Analyses and Projections of Economic Development. New York: UN Department of Economic and Social Affairs. - - - . 1957. Analyses and Pro;ections of Economic Development, III: The Economic Development of Colombia. Geneva: UN Depart~ent of Economic and Social Affairs. - - - . 19 59a. Analyses and Projections of Economic Development, VI: The Industrial Development of Peru. Mexico City: UN Department of Economic and Social Affairs. - - - . 1959b. Manual of Economic Development Projects. New York. - - - . 1960. Analyses and Pro;ections of Economic Development, V: The Economic Development of Argentina. Mexico City: UN Depart- ment of Economic and Social Affairs. ---.1963. World Economic Survey, part 1. New York. - - - . 1964. International Flow of Long-Term Capital and Official Donations, 1960-1962 (E/3917/Rev.1). New York: UN Department of Economic and Social Affairs. REFERENCES 513 - - - . 1965. "Accelerated Flow of Capital and Technical Assistance to the Developing Countries: Measurement of the Flow of Long-Term Capital and Official Donations to Developing Countries; Concepts and Methodology." In Report of the Secretary-General (A/5732, February). New York. - - - . 1975. Single-Year Population Estimates and Projections for Maior Areas, Regions, and Countries, 1950-2000 (ESA/P /WP.56, October). New York: UN Department of Social and Economic Affairs. U.S. Agency for International Dcvelopment. 1963. Principles of Foreign Economic Assistance. Washington, D.C.: USAID Program Coordination Staff. - - - . 1965. Proposed Mutual Defense and Development Programs for FY1966. 'Vashington, D.C.: USAID Program Coordination Staff. Uzawa, H. 1964. "Optimal Growth in a Two-Sector Model of Capital Accumulation." Review of Economic Studies 31 (January): 1-24. Vaccara, B. N., and Simon, N. 1968. Factors Affecting the Postwar Indus- trial Composition of Real Product: The Industrial Composition of Income and Product. Studics in Income and Wealth, no. 32. New York: Columbia University Press. Vaidvanathan, A. 1956. "A Survey of the Literature on Investment C;iteria and Development of Underdeveloped Countries." Indian Economic Journal 4 (October): 122-44. Vanek, J. 1964. "A Case Study in Development Analysis: Future Foreign Resource Requirements of Colombia." Washington, D.C.: U.S. Agency for International Development. Viner, J. 1952. International Trade and Economic Development. Glcn- coe, Ill.: Free Press. - - - . 1958. "Stability and Progress: The Poorer COllntries' Prob1em" (with a comment by R. Nurkse). In Stability and Progress in the World Economy, ed. D. Hague. London: Macmillan. \Vatanabe, T. 1961. "A Test of the Constancy of Input-Output Co- efficicnts among Countries." International Economic Review 2 (Sep- tember): 340-50. Weisskopf, T. E. 1971. "Alternative Patterns of Import Substitution in India." In Studies in Development Planning, ed. H. B. Chenery. Cambridge, Mass.: Harvard University Press. - - - . 1972. "The Impact of Foreign Capital Inflow on Domestic Savings in Underdeveloped Countries." Journal of International Eco- nomics 2 (February): 25-38. Wessel, II. E. 1952, 1953. "New Graph Correlates Operating Labor Data for Chemical Processes." In Data and Methods of Cost Estima- tion (a collection of reprints from Chemical Engineering) . 514 REFERENCES Westphal, L. E. 1971. "An Intertemporal Planning Model Featuring Economies of Scale." In Studies in Development Planning, ed. H. B. Chenery. Cambridge, Mass.: Harvard University Press. - - - . 1975. "Planning with Economies of Scale." In Economy-\Vide Models and Development Planning, ed. C. R. Blitzer, P. B. Clark, and L. J. Taylor. London: Oxford University Press. \Vestphal, L. E., and Kim, K. S. 1977. "Industrial Policy and Develop- ment in Korea." \VorId Bank Staff \Vorking Papers, no. 263 (August). \Vashington, D.C.: \Vorld Bank. \Viles, P. 1956. "Growth versus Choice." Economic Journal 66 (June): 244-55. Williams, J. H. 1929. "The Theory of International Trade Reconsidered." Economic Journal 39 (June): 195-209. [Reprinted in Readings in the Theory of International Trade, American Economic Association. Philadelphia: Blakiston.] Williamson, J. G. 1965. "Projected Aid Requirements for Turkey, 1960- 1975." USAID Discussion Paper, no. 10 (December). \Vashington, D.C.: U.S. Agency for Intemational Development. ---.1976. "Sources of American Inequalitv, 1896-1948." Review of Economics and Statistics 58 (November): 387-97. \Vodd Bank. 1976. \Vorld Tables 1976. Baltimore: Johns Hopkins Uni- versity Press. - - - . 1977. "Prospects for Developing Countries, 1978-1985." \Vash- ington, D.C. - - - . 1978. World Development Report, 1978. New York: Oxford University Press. - - - . 1979. \Vorid Development Report, 1979. New York: Oxford University Press. Yap, L. 1975. "Internal Migration in Less Developed Countries: A Survey of the Literature." \Vorld Bank Staff \Vorking Papers, no. 215 (Sep- tember). \Vashington, D.C.: vVorld Bank. Index Absorptive capacity limit, of invest- Bator, F. M., 282n, 287 ment, 414; economic growth and, Big push theory, 221, 251 361, 389; measure of, 345, 413; as Bolivia, 414 measure of development, 374; raising Brazil: balance of payments in, 415; of, 353 comprehensive model for, 58n; dual- Accounting prices, 288-98, 307 ism in, 58; import substitution in, 38; Accumulation: definition of, 9; mea- income distribution in, 468, 474,483; sures of, 10-11; resource allocation manufactured exports of, 38; pro- and, 15 gram approach to U.S. aid for, 442n; Adjustment mechanisms, 50, 62-M trade pattern of, 96 Agglomeration effects, 205, 207 Bruno, Michael, 54, 59, 61n, 309 Agriculture: average output per person Buchanan, N. S., 282, 285 employed in, 18-19; labor surplus in, Burma, 454 140n; taxation of, 9; during transi- Business cycle theory, 55 tion, 5; value added in, 82n Ahluwalia, '\Iontek S., 456 Algeria,28n Canada, 19n, 34 Argentina, 19n, 203; foreign exchange Capital; accumulation of, 5; allocation in, HI; input-output model for, 334; of, 68, 282-83; human, 5, 9; replace- trade pattern of, 96, 102 ment, 320; by sector, 18; shortage of, Australia, 34 347; substitution between labor and, 67, 68, 155-68. See also Capital inflow; External capital Balanced growth theory, 71; defini- Capital inflow, 4, 8; cost of debt from, tion of, 275-76; demand limitation 232; for desired GNP level, 405, 407, effects and, 90, 92; investment and, 411-12; in development pattern sim- 221, 222, 261, 264; and structural ulation, 85; in development policy, change, 108, 141, 142 29, 446; and employment, 146; Balance of payments: 300; develop- growth patterns and, 73, 74, 91; for ment loans and, 441; disequilibrium industrialization, 103, 130; invest- in, 300; effect on investment pattern ment effect on, 218, 220, 222, 231, of, 223; equilibrium in, 321, 324; limitation to growth from, 317, 323, 245, 251; limits on, 345, 355,360- 344 62; productivity of, 260; saving-in- Balance of trade, deficit in, 149, 167 vestment gap to determine, 395-96; Balassa, Bela A., 301 under surplus-labor and full employ- Bangladesh, income distribution in, 478, ment conditions, 167-71; reserves 493 and, 230; social welfare and, 146, 515 516 INDEX Capital inflow (continued) Demand: components of, 16; disaggre· 260. See also External capital; Ex· gation of, 74; Engel's law and, 7; ternal resources final, 151-53; food, 7, 16; as func- Capital·labor ratio, 161, 283-84 tion of income level, 70, 75, 79; as Capital-output ratio, 140, 150; criterion source of industrialization, 108, Ill, for minimum, 282-83; in optimal 113; substitution in, 50, 143, 146; growth policy model, 352 transformation of, 5, 7, 11, 16. See Carter, Nicholas G., 456 also Demand functions Caves, R. E., 273n, 278n, 280n Demand functions, 70, 77n, 78n, 153- CES. See Constant elasticity of substitu- 54 tion Demographic transition, 7, 21 Chile; cconomies of scale in, 2,69, 203; Denmark,34 program approach to U.S. aid for, Devaluation, 62 44 2n; self-sustained growth criteria Developing conntries, 1; capital flow to, applied to, 414; trade pattern of, 102 382; development objective of, 48; Cobb-Douglas production function, development performance of, 412- 147, 149n, 390 15, 427-37; effect of international Cobweb theory, 55 economic system on, 379-81; extent Colombia, 190n, 203, 271; balance of of poverty in, 462-64; increasing dif· payments in, 4l5; import substitution ferentiation among, 3; projection of in, 38; program approach to U.S. aid future growth in, 41 special con- for, 442n; self-sustained growth cri- cessions to poorest, 457 teria applied to, 414 Development loans, 441 Combined trade limit, 395-96 Development models. See Interindustry Commodities, in transformation analy- model; Planning models; Program- sis: classification of, 14; demand func- ming models; Two-sector model of tions of, II; variations in supply of, development 16 Development patterns, 11, 14; countnes Comparative advantage principle: in classified by, 91; differences among, development policies, 299-306; econ' 115, ]20-21, 130; sources of differ- omies of scale and, 238, 245, 265; ences in, 70. See also Export pattern; limitation to, 299; requirements for Production pattern; Trade patterns; analysis of, 277-78; resource alloca- Specialization pattern tion and, 272, 277-81 Development phenomena, 3; definition Comparative cost doctrine, 274 of, 2, 55; development patterns as, Constant clasticity of substitution (CES), 56; development phases as, 59; dual- 147, 149 ism as, 58, 62; and economies of Consumption: aggregate functions of, scale, 59, 62; models to analyze, 49, constraint on growth of, 232; 52, 60; trade-limited growth as, 58, food, 75, sectoral composition of, 62; worsening income distribution 78; in welfare function, 225, 322 as, 56 Coordination of investment. See Invest Development policy, 1, 2, 6; balanced ment, coordination of growth approach to, 300; compara- Costa Rica: balanced development in, tive advantage approach to, 299-306; 40, 92n; self·sustained growth criteria constraints on, 48-49; import sub· applied to, 414 stitution and, 27, 28, 29; investment Critical minimum effort theory, 22] coordination and, 173-74, 217,261, 264-66; inward-looking, 37, 40, Decomposition of structural change, 73, 96; models for alternative, 314- 108-11,113-15 15,317-25; outward·looking, 27, 40, INDEX 517 42, 73; range of choice in, 325-34; optimal resource allocation to pro- social objectives of, 290-98, 337, mote, 272, 275-76; phases of, 419- 349-50; trade and aid, 27, 65; urban· 26; projection of future, 415-27; ization and, 20. See also Accumula- trade-limited, 50, 51, 58, 269, 387, tion; Development models; Develop- 394-96,454 ment process; Development strategies; - self-sustained, 383-85, 387, 389, 396; Transformation criteria for measuring progress to- Development process, estimation of: ward, 413-14; exports role in, 415, logistic curve for, 8 449, 454. See also Balanced growth; regression equation for, 8; classifica- Growth models; Growth theory tion of commodities in, 14; cross· Economies of scale, 2, 4, 67; across country, 10, 16, 18, 21n, 44,72,102, sectors analysis of, 217, 224-25, 233, 140; timc series, 10, 16, 72 237; demand growth rate and, 218; Development strategies, 4; balanced, 28, development phenomena associated 37-38, 40, 45; flexibility in choice with, 59; effect on investment alloca· of, 280; import substitution as, 28, tion by, 217, 223; effect of limited 29, 37, 45; industrial specialization capital inflow on, 218, 220, 222,231, as, 23, 26, 28, 40,42,44,48; optimal 245, 261, 266; effect of limited ex- growth, 221, 270, 342-43, 376-77; port prospects on, 218, 222; excess primary specialization, 23, 26, 28, 29, capacity from, 220, 261, 264; in ex- 34-37, 45; problems of, 45; typology ports, 194; internal, 188; long·run of, 6, 26-27, 72 comparative advantage under, 238, Disaggregation, 14n, 67; demand, 74; 245,265; nature of, 191·-92: optimal export, 78--79; income growth, 469; shortage from, 238-39, 266; optimal manufacturing, 68; sector, 76-77, overcapacity from, 266; over time, 109, 146; trade, 73, 74 217, 218, 233; plant size and, 218, Disequilibrium, 2, 48; adjustment 233; relative import costs and, 218; mechanisms to avoid, 62-64; in bal- shadow prices under, 219, 234; short- ance of payments, 300; labor·capital run comparative advantage under, substitution and, 155-65, 172; sources 265; in steel industry, 192-93; in of, 49, 50, 54, 59-60; unemployment tradables production, 218 and,156 with investment coordination, 199- Dobb, M., 275 200,217; cost of, 214-15; evaluation Dual economy: models of, 46, 53, 54, of, 233-35'; mechanisms from, 210- 56, 58, 59; theories of, 20, 45 12; tradeoffs in, 223. See also Exter- Dynamic external economy, 276 nal economies; Scale effect Ecuador, 38n Education, income distribution and, 21 Eckstein, 0., 285n, 287 Effective protection, 37, 38 Economic dualism. See Dual economy Embodiment, phenomenon of, I44n Economic growth, 1, 2; from domestic Employment: capital inflow effect on, resources use, 384; elements affecting 146; equilibrium in, 322; ful1, 167- rate of, 31O-11; external resources to 71, 274; primary sector, 34; trans- promote, 342-43, 376-77; GNP in- formation in, 5, 18 crease as indication of, 64; and in- Energy. See Power production come distribution, 468, 470-71; Engel's law, 7,114,141 internal performance for, 427-37; in- Equilibrium: competitive, 174-75; vestment·limited, 386, 389-93; limits namic, 174, 207-08; partial, 205-06, to, 269, 316-18, 342, 345-46, 385- 218-19,223,307; static, 207-08, 273 87, 419; neoclassical model for, 48; Exchange rate: equilibrium, 404; to 518 INDEX Exchange rate (continued) Factors of production: costs of, 278; estimate import-substitution and ex- effcct of disequilibrium on, 278; di- ports, 326; limits on, 325-26, 334; rect, 157-61; indirect, 161-65; sub- in model of development alterna- stitution among, 144-46. See also tives, 314, 318, 325. See also Foreign Factor use exchange Factor use: changes in, 130-31; coor- Export orientation, 28n dination of investmcnt and saving in, Export pattern, 92, 96 196-97, 204; external economies and, Exports: aggregate functions of, 75; 196-98; interindustry model for, 83, agricultural, 35; commodity supply 85, 14749. See also Factors of pro- and, 16; economies of scale in, 194; duction; Valuc added effects on investmcnt of limited pros- Fei, John C. H., 42,53,59 pects for, 218, factors influenc- Fertility rate, 21, 380 ing level of, 321; and industrializa- Food: consumption of, 75, 79; demand tion, 108, 111; labor-intensive, 34; for, 7, 16; processing of, BIn manufactured, 27, 38, 40, 42, 44, Foreign assistance: allocation of, 442- 96; mineral, 34; in model for pro- 43; benefits and costs of, 342-43; duction and investment analysis, 179, donor couutry policies for, 440-43; 184; primary, 14,22, 34,92,96, 103, effectiveness of, 437-39; marginal 238, 278-79; projection for develop value of, 364, measure of value ing country, 427-37; sectoral com- of, 339-41, models of open economy position of, 78-79; in semic10sed dependent on, 343-44; model for role economy, 98; for sustained economic of, 386-89; opportunity cost of, 365; growth, 415, 449, 454; traditional, productivity of, 377, 434, 437, 438; 351 program approach to, 442; rationing External debt: cost of, 232; direct effects of, 354, 375, 377; recipient country of, 251; indirect effects of, 251; limits policies for, 439-40; technology on, 220, 222, 231,245,251, 325, 326 transfer through, 382 External capital, 2, 3, 269; adjustment Foreign exchange: allocation of, 67, 68, mechanism for, 63; effect on resources 224; from capital inflow, 317; cost of limit by, 344; GDP share of, 27; for securing, 184; economics of scale ef- industrialization, 42, 71; self-sustain- fect on, 68, 224; foreign aid and, ing growth requirements for, 396-97; 340-41; import substitution and, surplus labor and, 165-71; variation 150-51; investment effect on, 230, in supply of, 369-74. See also Capi- 231, 261, 264; limitation on growth tal inflow; External debt by, 340-41; in model for production External economies, applied to invest- and investment analysis, 179, 181, ment, 174-77; allocation by sector 184; opportunity cost of, 184; primary of, 196-98; dynamic, 207-08, 276, exports as source of, 34; resource importance of, 208-09; by in- allocation and, 269; shortage of, 307, dustry, 190; innovation and, 177, 347; transformation of domestic re- prices and, 182-86; static, 207-08 sources into, 346. See also Exchange External resources: economic growth and, 385; investment and imports rate financed by, 345; limitations on in- Foreign resources inflow. See Capital flow of, 218, 220, 345, 354-55; pro- inflow ductivity of, 409-12; systems for Frisch, R., 281, 290, 296n transferring, 441-42. See also Capi- Full employment: capital inflow tal inflow; External capital; Foreign 167-71; effective level of, 327; equi- assistance librium in, 324; foreign aid and, 339- INDEX 519 40; in Israel, 311, 338; opportunity Hirschman, A. 0., 275, 299 cost under, 274 Honduras, 414 Galenson, W., 285-87, 297 Hong Kong, 42n GDP. See Gross domestic product Human capital, 5, 9 General equilibrium, 48, 55, 109, 144; limitations of, 173 Generalized reduced gradient method, Iep. See United Nations International 153 Comparison Project Gerschenkron, A., 276 Import-substituting industrialization Ghana, 454 (lSI), 27, 45, 98; characteristics of, GNP. See Gross national product 37-38; unequal income distribution Government consumption, 16 from, 40. See also Import substitution Government expenditures: aggregate Imports: aggregate functions of, 75; functions of, 75; and investment in commodity supply and, 16; demand human capital, 9; sectoral composi- for, 320; marginal rate for, 352; sec- tion of, 78 toral composition of, 79; temporary, Greece, 42n, 271; development per- 219, 238; traditional, 351n. See also formance of, 439; exports role in Import-substituting industrialization; economic growth of, 454; self-sus- Import substitution tained growth criteria applied to, 414 Import substitution, 28; easy, 29; effect Gross domestic product (GDP), 27, 75 on productivity of, 335-36; exchange Gross national product (GNP): capital rate to estimate, 326; and foreign inflow for desired level of, 405, 407, exchange, 150-51; and foreign trade, 41 438; in development policy 150-51, 155; and industrialization, model, 322; as element of welfare 108, no, 111,141, 142; investment function, 65; for evaluating economic in, 405; plant size and, 218; second performance, 64; exports growth rela- stage, 29; timing of, 121. See also tion to, 27; and market size, Import-substituting industrialization relations between sources and uses of, Income, per capita: demand response 14; sectoral allocation of, 131; target to, 70,75,141; in development pat- growth rate for, 391-92,439-40 tern simulation, 85, 86; functions of Growth models: for advanced countries, levels of, 70; growth patterns by, 73, 310; deficiencies in, 309, 343; for 74; market size and, mortality Israel, 310-14; for less-developed and fertility rates and, 21 countries, 311 Income distribution: economic growth Growth theory: dynamic external econ- and, 470-71, 474; egalitarian, 466; omies under, export markets as element of welfare function, 65; and, 278-79; factor costs under, 278; factors contributing to more equal, productivity change under, 279; and 21; import substitution and, 40; pol- resource allocation, 275-76 icies to improve, 457; and population Guatemala, 414 growth, 21, 470 Income inequality: factors contribnting to, 20-21, 467-68; Kuznets hypothe- Harrod-Domar model, 316-17, 319, sis on, 56, 58, 465, 466, 468, 471, 323, 386,390,393 492; trends in, 476-80 Heavy industry, 98; delayed industrial- Income redistribution, 48,·86; incre- ization in, 102; growth of, 120; im- mental, 490-91; tradeoff between port substitution and, 114 growth and, 494-95 Heckscher-Olin theory, 274, 278 Incremental capital-output ratio (leoR), Higgins, B., 305 446 520 INDEX India: balance of payments in, 415; on, 231-32; external economies con- capital flow to, 447, 454; dcvelop- cept applied to, 174-77; external ef- ment program of, 302; dualism in, fects of, 189-91; in import substitu- 58; external eapital role in, 271; for- tion, 405; limits on, 344, 360-64, eign exchange in, 341; import sub- 385-86, 389-91; lumpiness of, 220- stitution in, 150; income level for 21, 223, 237, 245, 265; in metal- growth in, 64; measurement of pov- working, 176-77, 190; over time, erty in, 459; program approach to 200, 206-07, 218; private integration U.S. aid to, 442n; role of external of, 210; primary specialization and, eapital in, 271; self-sustained growth 34, 35; productivity of, 198, 199- criteria applied to, 414; trade pattern 200; profitability of, 175, 185, 186, of,96 196, 197, 276; reinvestment approach Indian Planning Commission, 300 to, 285-87; scale of, 233-34; by sec- Indonesia, 28n; primary specialization tor, 78,186,199-200,217,224,245; in, 34; prospects for reducing poverty social marginal product approach to, in, 493; role of exports in economic 284-85; in steel produetion,. 176-77, growth of, 454; trade pattern of, 96 189-90, 197-98, 238-39; for sus- Industrialization: alternative patterns tained economic growth, 414; tim- of, 9; changes in factor use for, 130- ing of, 233-34, 245, 251, 261, 266; 31; decomposition of structural uncoordinated decisions on, 177, 199 change and, 71,108-11,113-15; de- - allocation of: economies of scale ferred, 34, 37, 102; differences in effeet on, 217-23; marginal rules of, trade patterns and, 115, 120-21, 130; 217; optimal, 223, 276 in Japan, 2, 72, 82n; through protec- - coordination of: and development tion, 38; in small countries, 99, 102; policy, 173--74, 217; direct, 211-12; and urban bias in resource allocation, with economies of scale, 199-200, 20. See also Import-substituting in- 214-15, 217, 223, 233-35; mecha- dustrialization; Model of industrial- nisms of, 210-12; in metalworking ization and steel industries, 177; saving in Industry_ See Heavy industry; Light in- factor use from, 196-97, 204; un- dustry coordinated investment compared Innovation, external economies from. with, 189-90, 199 176-77,207 - interdependence of decisions for, Input-output model. See Interindustry 177-78, 220, 265; horizontal versus model vertieal, 222, 238, 266, 276-77; op- Interest rate, and investment pattern, timal shortage from, 266. See also 218 Absorptive capacity limit of invest- Interindustry model, 67, 70, 72-73, ment; Investment, coordination of; to determine sector effects of develop- Investment criteria; Investment pat- ment program, 334; for production tern; Overinvestment and investment, 182, 188, 204, 224- Investment criteria: and comparative 31; with substitution, 147-49 advantage, 298-99; factor-intensity, International economic system, effect 281, 282-84; productivity, 284-88; on less developed eountries, 379-81 programming, 288-89 Investment, 1, 5, 10, 16; through ad- Investment-limited growth, 386, 389- ministered prices, 21 0-11; aggregate 93 functions of, 75; balancing of, 245; Investment pattern: alternative, 245, capital inflow effect on, 218, 220, 266; balance of payments effect on, 345; comparative advantage principle 223; cost of changing, 261; factors applied to, 238, 245, 265; constraints determining, 218; implications for INDEX 521 development planning from, 261-67; Labor: changing allocation of, 67; cost optimal, 218, 222, 223, 236-39, 245; of, 278; demand for, 319-20; factors for power plants, 260; typology of, determining supply of, 321; oppor- 236 tunity cost of, 182, 184, 190; sub- Iran, 92n; economic growth of, 454; stitution between capital and, 67, 68, primary specialization in, 34, 35, 102 155-65; supply of trained, 279, 288, Ireland,42n 299, 307; supply of unskilled, 5, 58. Irrigation, external economies in, 208 See also Labor productivity; Labor IS!. See Import-substituting industrial- surplus ization Labor-output ratio, 150 Israel: capital formation in, 314; capital Labor productivity: in agriculture, 18; inflow to, 312; development perfor- definition of, 18n; demand for labor mance of, 439; development program dependent on, 319; limits on, 327 of, 270, 305, 306; foreign exchange Labor surplus, 44, 45, 51; and dis- in, 340, 341; full employment in, equilibrium, 50, 158; external capital 311, 338; GNP composition in, 313; and, 144, 165-71; minimum wage GNP growth in, 311, 312; import and, 159; in service sector, 140n surplus of, 313; industrial specializa- Lange-Lerner system of centrally ad- tion in, 42, 103; inflation in, 311, ministered prices, 210-11 312; limits to growth in, 310-14, Laspeyre index of welfare, 151 335; planning model for, 54, 318; Latin America: balance of payments productivity in, 311; role of external in, 303; development programming capital in, 2, 271, 312; trade pattern for, 303; metalworking industry in, of,96 205; steel industry in, 193, 203-04 Italy, 71, 334 Lebanon,42n,96 Ivory Coast, 35, 37 Leibenstein, H., 285-87, 297 Leontief, \Vassily, 83, 110-H, 178 Less developed countries. See Develop- ing countries Japan, 45n, 69, 140n, 182; dualism in, Lewis, W. Arthur, 5, 52-53, 56-57, 67, . 58, 59; sources of industrialization 173, 275 in, 2, 72, 82n Light industry, 98; early industrializa- tion in, 103; growth of, 120 Limited structural flexibility, 386, 389 Kahn, A. E., 284-85 Linear programming model, 149, 182; to Keynesian model, 49,50, 51, 54n, 56 determine optimum aid and growth Kindleberger, C. P., 280 pattern, 347--49; for resource alloca- Korea, 45n; comprehensive model for, tion, 270, 290-98, 308. See also 58n; dualism in, 58; income distribu- Programming model tion in, 21, 466, 468, 474, 482; in- come level for growth in, 64; indus- trialization model applied to, 72; MacEwan, Arthur, 342 industrial specialization in, 42, 103; Mahalanobis, P. C., 301-02 role of exports in eeonomic growth Malaysia: income distribution in, 35; of, 449,454; surplus labor in, 45 industrial output of, 37; primary Kravis, Irving B., purchasing power specialization in, 35, 102; role of ex- ratios estimated by, 459, 462n, 463 ports in economic growth of, 454 Kuznets, Simon, 5, 18, 20, 59; income l'vlandelbaum, K., 288 worsening hypothesis of, 56, 58, 465, Manufacturing: disaggregation of, 68; 466,468,471,492 resource allocation to, 279-80 522 INDEX Marginal cost pricing, 184, 299 New International Economic Order, Marginal productivity: of capital, 204; 457 criteria for, 284-87 New Zealand, 19n Marginal saving rate, 345, 352, 374, Nigeria, 28n, 34 375 Nontradables, economies of scale in, Market: classification of interrelations 219 in, 178n; investment and size of, 198; Nnrkse, R., 57,90, 173,274, 298 measurement of size of, 23 Nutrition, poverty and, 464 Metalworking industry: demand in, 182, 201; investment in, 176-77, 190; in Latin America, 205 Mexico: import substitution in, 38; in· Ohlin, P. G., 276. See also Heckscher- come distribution in, 468, 483; ser· Ohlin thcory vice earnings in, 38 Open economies, 343; model design Minerals: exports of, 34; taxation of for, 344-45; two-sector model for, mining of, 9; world demand for, 35 346-47 Model of industrialization: aggregate Opportunity cost: of foreign aid, 365; relations in, 73, 74-76; cross·country under full employment, 274; of estimates in, 72, 73, 75; domestic labor, 182, 184, 190; of overhead demand in, 75-76; interindustry rela· facilities, 307 tions in, 70, 82; methodology for, Overhead facilities: external economies 71-74; sector relations in, 76-79; in, 178, 209; opportunity cost of, 307 solutions of, 83, 85; time series esti· Overinvestment, 219, 237; excess ca- mates in, 72, 73; validation of, 85- pacity from, 261, 264 86,90 Models of development. See Interin· dustry model; Planning models; Pro- gramming models; Two-sector model Pakistan: capital inflow to, 405-07; of development development performance of, 344, Montias, J. M., 301 438; national output in, 357; opti· Morawetz, David, 443-44 mal growth strategy for, 342, 376- Mortality rate, 21 77; productivity of external resources Multiplier-accelerator interactions. 54n, in, 409-10; program approach to 55, 56 U.S. aid for, 442; rationing of aid to, !vlyrdal, G., 275 371, 375; role of external capital in, 2, 270, 344-45; trade pattern of, 96; transition to sustained growth in, 397,402-04 National income accounting, 230 Pakistan Perspective Plan, 402; develop- Natural resources: endowment measure ment program modcl applied to, 343, for, 22; exports based on, 34; and 375 specialization, 8; and structural Pakistan Planning Commission, third change, 6, 64; and trade patterns, 99, five year plan, 355-56 102; transformation of use of, II Neoclassical models, 45, 51; approach Papua New Guinea, 65 to, 46; flexibility of, 62-63; of gen- Partial equilibrium model, 205-06 eral equilibrium, 48; of growth and Peru, 92n, 454 resources allocation, 48 Petroleum, world demand for, 35 Net capital formation, 314, 322 Philippine National Economic Council, Netherlands, 334 304 INDEX 523 Philippines, 40n; development perfor- primary, 238-39, 279; by sector, mance of, 439; development program 83; structural changes in, 29; sub of,304 stitlltion in, 50, 143, 146; trade-im- Planning models, 6, 7; approach of, proving, 346, 351, 355. See also 47; deductive, 52, 54; with final de- Production pattern; Transformation mand, 151-53; incorporation of de- of production velopment phenomena in, 60; multi- Production orientation index, 27-28 sector, 61, 85; policy analysis from, 61; Production pattern, 72, 83, 91 simplified, 61; production functions Productive capacity, 7,9 in, 149-50; with substitution, 146-55 Profits: from investment, I 185, - with economies of scale, 221-31; 186, 196, 197, 276; savings from, 9; mixed integer programming in, 223- during transition, 5 24, 234; specification in, 222, 225; Programming model: for prodnction structural characteristics of, 222. See and investment analysis, 178-86; for also Interindustry model resource allocation, 288-98 Polak, J. J., 285 Protection. See Effective protection Population, 8; in development pattern Puerto Rico, 42n; developmcnt pro- simulation, growth patterns by, gram of, 303-04, 306; trade pattern 73, 74; income distribution and of,96 growth of, 21,470,474; market sizc Purchasing power equivalent, 86n, and, 23; rural versus urban, 20 114n; conversion ratios for, 459, 463 Poverty, 380; absolute, 457, 458-59; accelerated growth to reducc, 484 85, 490; extent of, 462-64; income- based approach to, 462; measuring Raduche1, William J., 143 scale of, 458-59; and nutrition, 464; Reserves, capital inflow and, 230 policy alternatives for reducing, 493- Resource allocation, 7, 15; changes in 95; redistribution of income to im- composition of, 113, 140, 141; com- prove, 485-86, 490-91; reduced pop- parative advantage implica bons for, ulation growth and, 486-87, 490 272,273-75,277; constraints on, 51, Power production: investment in, 185- 69; under coordinated investment 86, 237-39, 260; overcapacity in, decisions, 177; development versus 237,267 growth implications for, 275-76; Price endogenous model, 73, 144 factor-intensity criteria to measure, Prices: accounting, 288-98, 307; ad- 281, 282-84; factor substitution and, ministered, 21 0-11; cffect of invest- 144-45, l48; growth theory implica- ment coordination on, 213-16; en- tions for, import substitu- dogenous, 109; equilibrium, 184; and tion and, 10911; investment decisions external economies, 182-86; monop- relating to, 173; neoclassical model oly, 212. See also Shadow prices for, 48; productivity criteria to mea- Primary exports: based on foreign in- sure, 281, 284-88; programming cri- vestment, 34; definition of, 14; and teria to measure, 281, 288-98; sec- industrialization, 71; specialization toral, 73; socioeconomic processes in, 22 and, 19-20; trade theory implications Production: changing allocation of, 67; for, 277; urban bias in. 20; comparative advantage in, 279-80; under uncoordinated investment deci- criteria for marginal, 284,287; effects sions, 177 of interdependcnce in, 188, 189-91; Rosenstein-Rodan, Paul, 59, 174, functions, 74, 149-50, 153; in infant 186,275 industry, 279; model for, 179-80; Rostow, W. W., 275, 276 524 INDEX Saudi Arabia, 29n differences in, 16; industrial, 17, 23, Saving-investment gap, 395, 404, 407, 26,40,42, 44, 103; primary, 26, 28, 444 34, 92, 96, 99, 102; typical, 91·92, Saving limit, 385, 390,395 96. See also Trade patterns Savings, 5, 9; capital inflow and, 230, Sri Lanka: income distribution in, 35, 445; constraint on, 225; factors in- 482; primary specialization in, 35; fluencing, 320-21; marginal rates for, role of exports in economic growth 345, 352, 374, 375; limits on, 327; of, 454; taxation in, 35 to measure accumulation, 10; for sus- Steel production: demand for, 182; tained economic growth, 414; in ecouomies of scale in, 192-93; excess two-sector model, 53, 54 capacity in, 264-65; external econ- Savings-investment equilibrium, 321, omies in, 19D, 191; investment in, 324; foreign aid and, 340 176-77,189-90,197-98,238·39; in Scalc effect, 23, 57, 59 Latin America, 182, 193, 203-04; School enrollment, 10 monopoly in, 212; in production Scitovsky, Tibor, 37, 59, 175,189,261, analysis model, 179, 180 264,276 Stigler, G., 280 Second-best theory, 47-48 Strout, Alan M., 382 Sectors: allocation of resources by, 73; Structural change: decomposition of, cost of production in, 186; econ- 108-11, 113-15; definition of, 108- omies of scale across, 217, 224-25, 09; economic growth and, 64; gov- 233, 237; growth elasticity by, 86; ernment policies and, 29; indexes of, heavy industry, ll5, 120; industrial- 29; management of, 1, 2; timing of, ization by, 108; investment by, 78, 22,23,120 186, 199-200, 217, 224, 245; light Substitution, 4; capital-labor, 155-56; industry, ll5, 120; output by, 83; among commodities and factors, primary product, 115, 120; relations 144-45; in consumption, 206; in de- among, 76-79; value added by, 74, mand, 50, 143, 146; direct factor, 82,86, 130-31 157-61; elasticity of, 63, 143-44, Semiclosed economy, 96, 98--99, 141 157-58; indirect factor, 161-65; Service earnings, 38 mechanisms for, 50; in production, Shadow prices: under economies of SO, 143, 146, 206; in trade, 50, 143, scale, 211,213,219,234; to evaluate 146 investment projects, 290-98; market Sudan, 28n, 454 prices versus, 166 Surplus labor. See Labor surplus Sierra Leone, 28n Sweden, 34 Singapore, 29n, 42n; industrial special- ization in, 103 SkiIllimit, 390 Taiwan, 45n; development performance Social marginal product (SMP), 284-- of, 439; dualism in, 53n, 58, 59; in- 86; Philippines and, 304 come distribution in, 21, 53n, 466, Socioeconomic processes, 7, 8n, 19-20 468, 482; industrialization model ap- Social welfare: capital inflow effect on, plied to, 72; industrial specialization 146; transformation of productive in, 42, 103; role of exports in eco- factors into, 144-45 nomic growth of, 449, 454; surplus South Africa, 40n labor in, 45 Soviet Union, 301 Target growth rate, 389-90,439-40 Spain, 40n, 42n Taxation, 9, 10 Specialization pattern, 90-91: balanced, Technological change, 69, 131 n; and 23, 28n, 90; effect on transition of industrialization, 82, 108, 1I1; and INDEX 525 resource allocation, 141; response to components of, 17; by sector, 17-18; income level by, 70 socioeconomic processes and, 19 Technology transfer, 382 Transformation of production, 5, 7, Thailand, 40n 14, 16, 18; from industry share, 34; Theil, Henri, 314,476-77 from primary share, 26, 34; from Timing, of transition, 22,23,45 service share, 34; social welfare from, Tinbergen, Jan, 281, 290, 308; 314~15 144-45 Tradable commodities, 7], 131; econ- Transformation of trade, 7, 14, 16; omies of scale in production of, 218, through manufactured exports, 35, 219; sectors producing, 115; shadow 40; through primary exports, 22, 26 prices of, 220n; temporary produc- Transition to developed economy, 1, 5, tion for export of, 219 7; definition of, 6; diversity versus Trade, 5; disaggregation of, 74; import uniformity in, 6-7; by middle-income substitution and, 150~51, 155; re- countries, 449, 454~55; phases of, sponse to rising income by, 70, 79; 29; regression analysis for, 8, 14, 23; and sectoral resource allocation, 73; with self·sustained growth, 383~85, as substitution mechanism, 50; struc- 396,413-15; specialization effect on, tural changes in, 29; substitution in, 16 50, 143, 146; for sustained economic Transportation: economies of scale in, growth, 414. See also Trade gap; 193n, 196; external economies in, Trade-improving production: Trade· 178,190,209 limited growth; Trade patterns; Tunisia, 96, 442n Trade theory Turkey: balance of payments in, 415; Trade and aid policies, 27, 65 comprehensive model for, 58n; im· Trade gap, 351, 352, 395,404,407,444 port substitution in, 38; industrializa· Trade-improving production, 346, 351, tion model applied to, 72; program 355 approach to U.S. aid for, 442n; role Trade-limited growth, 50, 51, 58, 269, of external capital in, 271; self·sus- 387, 394-96 tained growth criteria applied to, 414; Trade-limited models, 54, 59 service earnings in, 38; trade pattern Trade orientation, index of, 22, 23; of,96 definition of, 23n; to identify effects Two-gap model, 54n, 317, 386~97, of trade and aid policies, 27 444-46 Trade patterns: alloeation of GNP by, Two-sector model of development, 52- 131; in, 79; classification of, 54; alternative forms of, 354; applied 23, 91; in heavy versus light indus- to Pakistan, 356-69; characteristics try, 98; industrializatiou model for, of, 346-47; constraints in, 348, 351~ 70, 90~9l; large country, 23, 26,91-- 53, 360-64; design of, 345-46; ob- 92, 96, 98-99, 120~21, 131; natural jective of, 247~48 resources and, 99, 102; of primary specialization, 92, 96, 99, 102; of semic10sed economy, 96, 98~99, 141; small industry-oriented, 23, 92, 99, Uganda, 28n 121, 130, 131; small primary-oriented, Unbalanced growth theory, 221, 23,92, 121, 130, 131 261, 264, 276 Trade theory, and resource allocation, Unemployment, 156 275-76, 277 United Nations Economic Commission Transformation of demand, 5, 7, Il, 16 for Latin America, 193, 289, 300; Transformation of employment, 5, 18 development programming by, 303; Transformation of factor use, 7, 14; poverty study by, 464n 526 INDEX United Nations International Compar- as, 65; in development models, 315, ison Project (rep), 459, 462; dollars 337; GNP level as, 65; income dis- of constant purchasing power of, 469 tribution as, 65; in optimal growth Urbanization, 7, 19-20 policy model, 349-50, 353, 356-57 Uruguay, 102 \Vestphal, Larry E., 21 7 USSR. See Soviet Union Williams, J. E., 277n World Bank, 470 Value added: by sector, 74, 82, 86, 130-31; trade patterns and, 79,98 Venezuela, 35, 102 Yugoslavia: income distribution in, 21 Viner, J., 274 466, 482, 483; industrial specializa- tion in, 42 Wages, 44, 159 Welfare function, 49; consumption in, 225, 232, 350; development capacity Zambia, 65, 102 THE WORLD BANK