Economic Models for Policy Making : Principles and Designs Revisited [1 ed.] 9781136220883, 9780415509046

Citation preview

Economic Models for Policy Making

Over the past decades, many different kinds of models have been developed that have been of use to policy makers, but until now the different approaches have not been brought together with a view to enhancing the systematic unification and evaluation of these models. This new volume aims to fill this gap by bringing together four decades’ worth of work by S. I. Cohen on economic modelling for policy making. Work on older models has been rewritten and brought fully up to date, and these older models have therefore been brought back to the fore, both to assess how they influenced more recent models and to see how they could be used today. The focus of the book is on models for development policies in developing economies, but there are some chapters that relate to economic policies in transition and developed economies. The policy areas covered are of typical interest in developing and transition economies. They include those relating to trade liberalisation reforms, sustainable development, industrial development, agrarian reform, growth and distribution, human resource development and education, public goods, and income transfers. Each chapter contains a brief assessment of the empirical literature on the economic effects of the policy measures discussed in the chapter. The book presents a platform of economic modelling that can serve as a refresher for practising professionals, as well as a reference companion for graduates engaging in economic modelling and policy preparations. S. I. Cohen is Emeritus Professor of Economics at the Erasmus University Rotterdam, the Netherlands. He founded the Foundation for Economic Research Rotterdam and is a regular advisor to UN, WB, and EU agencies on development issues and in field missions.

Routledge Frontiers of Political Economy

1

Equilibrium Versus Understanding Towards the rehumanization of economics within social theory Mark Addleson

2

Evolution, Order and Complexity Edited by Elias L. Khalil and Kenneth E. Boulding

3

Interactions in Political Economy Malvern after ten years Edited by Steven Pressman

4

The End of Economics Michael Perelman

5

Probability in Economics Omar F. Hamouda and Robin Rowley

6

Capital Controversy, Post Keynesian Economics and the History of Economics Essays in honour of Geoff Harcourt, volume one Edited by Philip Arestis, Gabriel Palma and Malcolm Sawyer

7

Markets, Unemployment and Economic Policy Essays in honour of Geoff Harcourt, volume two Edited by Philip Arestis, Gabriel Palma and Malcolm Sawyer

8

Social Economy The logic of capitalist development Clark Everling

  9 New Keynesian Economics/Post Keynesian Alternatives Edited by Roy J. Rotheim 10 The Representative Agent in Macroeconomics James E. Hartley 11 Borderlands of Economics Essays in honour of Daniel R. Fusfeld Edited by Nahid Aslanbeigui and Young Back Choi 12 Value, Distribution and Capital Essays in honour of Pierangelo Garegnani Edited by Gary Mongiovi and Fabio Petri 13 The Economics of Science Methodology and epistemology as if economics really mattered James R. Wible 14 Competitiveness, Localised Learning and Regional Development Specialisation and prosperity in small open economies Peter Maskell, Heikki Eskelinen, Ingjaldur Hannibalsson, Anders Malmberg and Eirik Vatne 15 Labour Market Theory A constructive reassessment Ben J. Fine

16 Women and European Employment Jill Rubery, Mark Smith, Colette Fagan, Damian Grimshaw 17 Explorations in Economic Methodology From Lakatos to empirical philosophy of science Roger Backhouse 18 Subjectivity in Political Economy Essays on wanting and choosing David P. Levine 19 The Political Economy of Middle East Peace The impact of competing trade agendas Edited by J.W. Wright, Jnr 20 The Active Consumer Novelty and surprise in consumer choice Edited by Marina Bianchi 21 Subjectivism and Economic Analysis Essays in memory of Ludwig Lachmann Edited by Roger Koppl and Gary Mongiovi 22 Themes in Post-Keynesian Economics Essays in honour of Geoff Harcourt, volume three Edited by Claudio Sardoni and Peter Kriesler 23 The Dynamics of Technological Knowledge Cristiano Antonelli 24 The Political Economy of Diet, Health and Food Policy Ben J. Fine

25 The End of Finance Capital market inflation, financial derivatives and pension fund capitalism Jan Toporowski 26 Political Economy and the New Capitalism Edited by Jan Toporowski 27 Growth Theory A philosophical perspective Patricia Northover 28 The Political Economy of the Small Firm Edited by Charlie Dannreuther 29 Hahn and Economic Methodology Edited by Thomas Boylan and Paschal F O’Gorman 30 Gender, Growth and Trade The miracle economies of the postwar years David Kucera 31 Normative Political Economy Subjective freedom, the market and the state David Levine 32 Economist with a Public Purpose Essays in honour of John Kenneth Galbraith Edited by Michael Keaney 33 Involuntary Unemployment The elusive quest for a theory Michel De Vroey 34 The Fundamental Institutions of Capitalism Ernesto Screpanti 35 Transcending Transaction The search for self-generating markets Alan Shipman

36 Power in Business and the State An historical analysis of its concentration Frank Bealey 37 Editing Economics Essays in honour of Mark Perlman Hank Lim, Ungsuh K. Park and Geoff Harcourt 38 Money, Macroeconomics and Keynes Essays in honour of Victoria Chick, volume one Philip Arestis, Meghnad Desai and Sheila Dow 39 Methodology, Microeconomics and Keynes Essays in honour of Victoria Chick, volume two Philip Arestis, Meghnad Desai and Sheila Dow 40 Market Drive and Governance Reexamining the rules for economic and commercial contest Ralf Boscheck 41 The Value of Marx Political economy for contemporary capitalism Alfredo Saad-Filho 42 Issues in Positive Political Economy S Mansoob Murshed 43 The Enigma of Globalisation A journey to a new stage of capitalism Robert Went 44 The Market Equilibrium, stability, mythology S N Afriat 45 The Political Economy of Rule Evasion and Policy Reform Jim Leitzel

46 Unpaid Work and the Economy Edited by Antonella Picchio 47 Distributional Justice Theory and measurement Hilde Bojer 48 Cognitive Developments in Economics Edited by Salvatore Rizzello 49 Social Foundations of Markets, Money and Credit Costas Lapavitsas 50 Rethinking Capitalist Development Essays on the economics of Josef Steindl Edited by Tracy Mott and Nina Shapiro 51 An Evolutionary Approach to Social Welfare Christian Sartorius 52 Kalecki’s Economics Today Edited by Zdzislaw L. Sadowski and Adam Szeworski 53 Fiscal Policy from Reagan to Blair The left reers right Ravi K. Roy and Arthur T. Denzau 54 The Cognitive Mechanics of Economic Development and Institutional Change Bertin Martens 55 Individualism and the Social Order The social element in liberal thought Charles R. McCann Jnr. 56 Affirmative Action in the United States and India A comparative perspective Thomas E. Weisskopf 57 Global Political Economy and the Wealth of Nations Performance, institutions, problems and policies Edited by Phillip Anthony O’Hara

58 Structural Economics Thijs ten Raa 59 Macroeconomic Theory and Economic Policy Essays in honour of Jean-Paul Fitoussi Edited by K. Vela Velupillai 60 The Struggle Over Work The “end of work” and employment alternatives in post-industrial societies Shaun Wilson 61 The Political Economy of Global Sporting Organisations John Forster and Nigel Pope

69 Ordinary Choices Individuals, incommensurability, and democracy Robert Urquhart 70 Labour Theory of Value Peter C. Dooley 71 Capitalism Victor D. Lippit 72 Macroeconomic Foundations of Macroeconomics Alvaro Cencini 73 Marx for the 21st Century Edited by Hiroshi Uchida

62 The Flawed Foundations of General Equilibrium Theory Critical essays on economic theory Frank Ackerman and Alejandro Nadal

74 Growth and Development in the Global Political Economy Social structures of accumulation and modes of regulation Phillip Anthony O’Hara

63 Uncertainty in Economic Theory Essays in honor of David Schmeidler’s 65th birthday Edited by Itzhak Gilboa

75 The New Economy and Macroeconomic Stability A neomodern perspective drawing on the complexity approach and Keynesian economics Teodoro Dario Togati

64 The New Institutional Economics of Corruption Edited by Johann Graf Lambsdorff, Markus Taube and Matthias Schramm 65 The Price Index and its Extension A chapter in economic measurement S.N. Afriat 66 Reduction, Rationality and Game Theory in Marxian Economics Bruce Philp

76 The Future of Social Security Policy Women, work and a citizens’ basic income Ailsa McKay 77 Clinton and Blair The political economy of the Third Way Flavio Romano

67 Culture and Politics in Economic Development Volker Bornschier

78 Marxian Reproduction Schema Money and aggregate demand in a capitalist economy A. B. Trigg

68 Modern Applications of Austrian Thought Edited by Jürgen G. Backhaus

79 The Core Theory in Economics Problems and solutions Lester G. Telser

80 Economics, Ethics and the Market Introduction and applications Johan J. Graafland 81 Social Costs and Public Action in Modern Capitalism Essays inspired by Karl William Kapp’s theory of social costs Edited by Wolfram Elsner, Pietro Frigato and Paolo Ramazzotti 82 Globalization and the Myths of Free Trade History, theory and empirical evidence Edited by Anwar Shaikh 83 Equilibrium in Economics: Scope and Limits Edited by Valeria Mosini 84 Globalization State of the art and perspectives Edited by Stefan A. Schirm 85 Neoliberalism National and regional experiments with global ideas Edited by Ravi K. Roy, Arthur T. Denzau, Thomas D. Willett 86 Post-Keynesian Macroeconomics Economics Essays in honour of Ingrid Rima Edited by Mathew Forstater, Gary Mongiovi and Steven Pressman 87 Consumer Capitalism Anastasios S. Korkotsides 88 Remapping Gender in the New Global Order Edited Marjorie Griffin Cohen and Janine Brodie 89 Hayek and Natural Law Erik Angner 90 Race and Economic Opportunity in the Twenty-First Century Edited by Marlene Kim

  91 Renaissance in Behavioural Economics Harvey Leibenstein’s impact on contemporary economic analysis Edited by Roger Frantz   92 Human Ecology Economics A new framework for global sustainability Edited by Roy E. Allen   93 Imagining Economics Otherwise Encounters with identity/difference Nitasha Kaul   94 Reigniting the Labor Movement Restoring means to ends in a democratic labor movement Gerald Friedman   95 The Spatial Model of Politics Norman Schofield   96 The Economics of American Judaism Carmel Ullman Chiswick   97 Critical Political Economy Christian Arnsperger   98 Culture and Economic Explanation Economics in the US and Japan Donald W. Katzner   99 Feminism, Economics and Utopia Time travelling through paradigms Karin Schönpflug 100 Risk in International Finance Vikash Yadav 101 Economic Policy and Performance in Industrial Democracies Party governments, central banks and the fiscal–monetary policy mix Takayuki Sakamoto 102 Advances on Income Inequality and Concentration Measures Edited by Gianni Betti and Achille Lemmi

103 Economic Representations Academic and everyday Edited by David F. Ruccio 104 Mathematical Economics and the Dynamics of Capitalism Goodwin’s legacy continued Edited by Peter Flaschel and Michael Landesmann 105 The Keynesian Multiplier Edited by Claude Gnos and LouisPhilippe Rochon 106 Money, Enterprise and Income Distribution Towards a macroeconomic theory of capitalism John Smithin 107 Fiscal Decentralization and Local Public Finance in Japan Nobuki Mochida 108 The ‘Uncertain’ Foundations of Post-Keynesian Economics Essays in exploration Stephen P. Dunn 109 Karl Marx’s Grundrisse Foundations of the critique of political economy 150 years later Edited by Marcello Musto 110 Economics and the Price Index S.N. Afriat and Carlo Milana 111 Sublime Economy On the intersection of art and economics Edited by Jack Amariglio, Joseph W. Childers and Stephen E. Cullenberg 112 Popper, Hayek and the Open Society Calvin Hayes 113 The Political Economy of Work David Spencer

114 Institutional Economics Bernard Chavance 115 Religion, Economics and Demography The effects of religion on education, work, and the family Evelyn L. Lehrer 116 Economics, Rational Choice and Normative Philosophy Edited by Thomas A. Boylan and Ruvin Gekker 117 Economics Versus Human Rights Manuel Couret Branco 118 Hayek Versus Marx and Today’s Challenges Eric Aarons 119 Work Time Regulation as Sustainable Full Employment Policy Robert LaJeunesse 120 Equilibrium, Welfare and Uncertainty Mukul Majumdar 121 Capitalism, Institutions and Economic Development Michael Heller 122 Economic Pluralism Robert Garnett, Erik Olsen and Martha Starr 123 Dialectics of Class Struggle in the Global Economy Clark Everling 124 Political Economy and Globalization Richard Westra 125 Full-Spectrum Economics Toward an inclusive and emancipatory social science Christian Arnsperger

126 Computable, Constructive and Behavioural Economic Dynamics Essays in honour of Kumaraswamy (Vela) Velupillai Stefano Zambelli

136 The Moral Rhetoric of Political Economy Justice and modern economic thought Paul Turpin

127 Monetary Macrodynamics Toichiro Asada, Carl Chiarella, Peter Flaschel and Reiner Franke

137 Macroeconomic Regimes in Western Industrial Countries Hansjörg Herr and Milka Kazandziska

128 Rationality and Explanation in Economics Maurice Lagueux 129 The Market, Happiness and Solidarity A Christian perspective Johan J. Graafland 130 Economic Complexity and Equilibrium Illusion Essays on market instability and macro vitality Ping Chen 131 Economic Theory and Social Change Problems and revisions Hasse Ekstedt and Angelo Fusari 132 The Practices of Happiness Political economy, religion and wellbeing Edited by John Atherton, Elaine Graham and Ian Steedman 133 The Measurement of Individual Well-Being and Group Inequalities Essays in memory of Z. M. Berrebi Edited by Joseph Deutsch and Jacques Silber 134 Wage Policy, Income Distribution, and Democratic Theory Oren M. Levin-Waldman 135 The Political Economy of Bureaucracy Steven O. Richardson

138 Business Ethics and the Austrian Tradition in Economics Hardy Bouillon 139 Inequality and Power The economics of class Eric A. Schutz 140 Capital as a Social Kind Definitions and transformations in the critique of political economy Howard Engelskirchen 141 Happiness, Ethics and Economics Johannes Hirata 142 Capital, Exploitation and Economic Crisis John Weeks 143 The Global Economic Crisis New perspectives on the critique of economic theory and policy Edited by Emiliano Brancaccio and Giuseppe Fontana 144 Economics and Diversity Carlo D’Ippoliti 145 Political Economy of Human Rights Rights, realities and realization Bas de Gaay Fortman 146 Robinson Crusoe’s Economic Man Construction and deconstruction Edited by Ulla Grapard and Gillian Hewitson

147 Freedom and Happiness in Economic Thought and Philosophy From clash to reconciliation Edited by Ragip Ege and Herrade Igersheim 148 Political Economy After Economics David Laibman 149 Reconstructing Keynesian Macroeconomics Volume 1 Partial perspectives Carl Chiarella, Peter Flaschel and Willi Semmler 150 Institutional Economics and National Competitiveness Edited by Young Back Choi 151 Capitalist Diversity and Diversity within Capitalism Edited by Geoffrey T. Wood and Christel Lane 152 The Consumer, Credit and Neoliberalism Governing the modern economy Christopher Payne 153 Order and Control in American Socio-Economic Thought U.S. social scientists and progressive-era reform Charles McCann 154 The Irreconcilable Inconsistencies of Neoclassical Macroeconomics A false paradigm John Weeks

155 The Political Economy of Putin’s Russia Pekka Sutela 156 Facts, Values and Objectivity in Economics José Castro Caldas and Vítor Neves 157 Economic Growth and the High Wage Economy Choices, constraints and opportunities in the market economy Morris Altman 158 Social Costs Today Institutional analyses of the present crises Edited by Wolfram Elsner, Pietro Frigato and Paolo Ramazzotti 159 Economics, Sustainability and Democracy Economics in the era of climate change Christopher Nobbs 160 Organizations, Individualism and Economic Theory Maria Brouwer 161 Economic Models for Policy Making Principles and designs revisited S. I. Cohen

Economic Models for Policy Making

Principles and designs revisited

S. I. Cohen

First published 2013 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Simultaneously published in the USA and Canada by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 2013 S. I. Cohen The right of S. I. Cohen to be identified as the author of this work has been asserted by him in accordance with the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Cohen, S. I. Economic models for policy making : principles and designs revisited / by S.I. Cohen.   p. cm.   Includes bibliographical references and index.   1. Developing countries--Economic policy. 2. Economic policy. I. Title.  HC59.7.C588 2012     330.01--dc23  2012009526 ISBN: 978-0-415-50904-6 (hbk) ISBN: 978-0-203-09701-4 (ebk) Typeset in Times New Roman by Bookcraft Ltd, Stroud, Gloucestershire

To Els To Sophie, Yasmine, Nelson and Midas

Contents

List of figures xix List of tables xxi Preface xxv 1 Introduction 1  Focus on economic models for exploring policies 2  Outline: method and content 3  Economy-wide policy models 4  Partial models 5  Concluding remarks

1 1 3 4 15 18

2

Some essentials in economy-wide policy models 1 Introduction 2  CEM models: focus on market clearance via quantities 3  CGE models: focus on market clearance via prices 4  SAM models: relationship to the CEM and CGE models 5  Review table and concluding remarks

21 21 24 31 41 42

3

Socio-political regimes and economic development: exploratory models on agrarian reform in India and Chile 1 Introduction 2  Main features of the model 3  The modelling framework 4  Selected results from application to India 5  Application to Chile 6  Concluding remarks

44 44 44 45 53 58 62

4

Social economic development goals in economy-wide policy models: an application to Korea 1 Background 2  A unifying approach towards social economic development goals 3  Specification of the model 4 Application 5  Analytical versus policy uses: breakdown of policy making 6  Concluding remarks

64 64 65 69 79 84 89

xvi  Contents 5

Growth and distribution in SAM models: various applications 1 Introduction 2  Tabulation and construction of the social accounting matrix 3  The SAM as an economy-wide model 4  Output and income SAM multipliers: results for ten countries 5  Decomposition of SAM multipliers into transfer, open-, and    closed-loop effects 6  Identification of gainers and losers in SAM multipliers 7  Strategic choices for growth with redistribution 8  Discussion of scope and limitations

91 91 92 95 98 102 105 109 111

6

Simplified statics and dynamics in the CGE model: parameterisation and simulations for Indonesia 1 Background 2  The static CGE model 3  Parameterisation of the CGE model 4  Static policy simulations 5  The dynamic model 6  Dynamic policy simulations 7  Concluding remarks

114 114 116 123 127 128 131 133

7

Growth with redistribution through liberalisation with restructuring: a CGE policy model of Nepal 1 Introduction 2  Key features 3   Model specification 4  Application and policy simulations 5  The dynamic model with a restructured economy 6  Concluding remarks

136 136 137 140 147 150 156

8

Sustained development of land resources: a policy model for Sudan 1 Introduction 2  The model 3  Estimation and baseline forecasts 4  Policy simulations: benefits and costs 5  Concluding remarks

159 159 160 174 178 184

9

Simulation results of SAM models for transiting economies: Russia falls and China rises 1  Comparative analysis of economic systems 2  Salient differences in economic performance: Russia and China 3  The SAMS of Russia and China 4  SAM multipliers in Russia and China 5  Gainers and losers in Russia and China 6  Summary and conclusions

185 185 189 191 192 195 196

Contents  xvii 10 Transiting from fixed- to flexible-price regimes: SAM-CGE models of Poland and Hungary 1 Introduction 2  The fixed-price SAM model 3  The flexible-price CGE model 4  Results of applied policy simulations to Poland 5  Results of the simulations for Hungary 6  Concluding remarks

202 202 203 204 211 216 220

11 Public spending multipliers in extended SAM models for a developed economy 223 1 Introduction 223 2  Multiplier analysis in a first SAM for the Netherlands 224 3  Changes in SAM multiplier results over ten years 228 4  Extension: incorporation of regional subdivisions in the SAM model 230 5  Extension: urbanisation levels 237 6  Concluding remarks 239 12 Fiscal policy simulations in adapted CGE models: the Netherlands 1 Introduction 2  The basic model 3  The elaborate CGE model 4  Structuralist CGE model 5  Concluding remarks 13 Normed planning of human resource development: a roadmap model for Ethiopia 1 Background 2  The targeted select group of countries 3  Long-range targeting model for HRD 4  Roadmap results and transition paths to destinations 5  Additional refinements 6  Matching of the labour market in the short and medium terms 7  The roadmap as part of a sustained development trajectory:    anticipated economy-wide imbalances 8  Summary and conclusions

240 240 243 250 252 257 258 258 259 259 262 272 274 275 278

14 Labour market imbalances and adjustments: forecast model with RAS component 284 1 Background 284 2  Aggregate demand and supply at the sector level 285 3  Demand and supply by occupation and education: forecast model 287 4  Labour market adjustment: RAS iterations 289 5 Applications 291 6  Earnings imbalances: human capital versus job competition 298 7 Application 300 8  Concluding remarks 302

xviii  Contents 15 Privatisation decisions during transition: a CBA model applied to Poland 1 Introduction 2  Time horizon and notations used in the model 3  Transaction values for foreign buyer and seller government 4  Expected sales and profits 5  Costs: direct, associated, and replacement investment costs 6  Impact on government revenue 7  Empirical results 8  Concluding remarks

303 303 304 306 308 309 311 312 314

16 Economic policy solutions to social queuing problems: a random sampling model 1 Background 2  Quality adjusted life years (QALY) 3  Linking QALY to earnings 4  The model 5  Quantification 6  Findings and discussion 7  Additional random sampling and policy simulations 8  Concluding remarks

316 316 317 319 320 323 327 329 330

17 Modelling convergence in economic growth between rich and poor countries 1 Introduction 2  The convergence hypothesis: supply-side theory and evidence 3  The convergence hypothesis: demand-side theory and evidence 4  Empirical results 5 Demonstration 6  More convergence through transfer mechanisms

333 333 334 335 339 342 344

18 Modelling of distinctly behaving economic systems: theory and applications 346 1 Introduction 346 2  Behavioural types and behavioural settings 347 3  Prototypes of dominant behaviours in economic systems 350 4  The start and the long-range development of economic systems 353 5  Empirical validation 357 6  On the future outlook for economic systems 361 7  Concluding remarks 365 Notes 366 References 381 Index 389

List of figures

  3.1 Causal ordering in the landlord-leader model 52   3.2 Causal ordering in the peasant-leader model 52   4.1 Korea model solutions: indices of selected aim variables. GDP index takes 1982 = 100 82   4.2 Korea: recurring multiplier effect of public current plus associated capital expenditure 83   4.3 Causal ordering of the analytical model 85   4.4 Causal ordering of the planning model 87   5.1 Circular flow 92   8.1 Sudan: trade-off between economic growth and degradation loss, 1990–2000 183   8.2 Sudan: degradation loss: difference between policy and baseline, 2000; billion Sudanese pounds 183   8.3 Sudan: GDP and green GDP: difference between policy and baseline, 2000; billion Sudanese pounds 183   9.1 The analytical framework for comparing performances 186 12.1 Government budget balance (GBD) as percentage of GDP, and growth of GDP: the Netherlands, 1970–2010 241 13.1 Ethiopia: anticipated annual growth rates of main variables of the education system 2005–30 according to the roadmap, percentages 270 13.2 Ethiopia: roadmap gaps in various HRD positions, benchmark 2005 271 13.3 Ethiopia: anticipated imbalances economy wide elsewhere 277 13.4 Ethiopia: regression diagrams for four equations 283 14.1 Pakistan: educational returns under human capital and job competition 301 15.1 Main phases in a privatisation venture 304 16.1 QALY 318 16.2 Earnings functions by age group and attained education 326 16.3 Performance under more allocations to health. Appraisal situation I: W = L 331 16.4 Performance under more allocations to health. Appraisal situation II: W = QL 332

xx  List of figures 17.1 17.2 18.1 18.1 18.2 18.3 18.4 18.5

Relationship between the exogenous share in national income X/Y and income per capita Y/N 337 Relationship between multipliers m and exogenous share in national income X/Y 338 (a, b, c) Configurations of three socio-economic systems: HIM, FIM, SIM 351 (d) A sketch of the multi-poles system (MPM) 352 Positioning of economies along axis of dominant systemic interactions 358 Importance of household settings: country groups ranking of importance of family 359 Pro-firm and pro-state attitudes 359 Displacement tendencies with significant effects on the future dominance of economic systems 364

List of tables

2.1 Notations 24 2.2 Causal ordering of the CEM model 29 2.3 Causal ordering of the CGE model 34 2.4 Review summary of the economy-wide models in the book: some essentials 43 3.1 Notations 46 3.2 Basic run parameters, and changes therein under different policy simulations, India 55 3.3 Basic run projections and policy simulations for year t = 20, India 56 3.4 Initial year t = 0, basic run projections (BRP), and policy simulations for year t = 20, Chile 61 4.1 List of aim variables 67 4.2 List of instrument variables 67 4.3 Notations 70 4.4 Alternative strategies of fixing income targets by social groups: deviations from analytical solutions for Korea, 1980 88 4.5 Predictive performance of the analytical and planning forms, Korea 89 5.1 Specimen of SAM 94 5.2 Notations 96 5.3 The causal ordering of the SAM model 97 5.4 Selected output and income multipliers 98 5.5 Output multipliers Ms,jj’ and income multiplier Ms,hj’ following an exogenous spending injection of one unit in activity sectors, ten countries 100 5.6 Output multiplier Ms,jh’ and income multiplier Ms,hh’ following an exogenous income transfer of unit to household groups, ten countries 101 5.7 Breakdown of SAM multipliers MS into transfer effects (Leontief multiplier) M1 and closed-loop effects M3 , ten countries 104 5.8 Gainers and losers index, GLI, of multiplier effects of sector injections, ten countries 107 5.9 Impacts of spending injections and income transfers on growth and distribution: Indonesia 110 6.1 Notations 117

xxii  List of tables   6.2 Coefficients relating to sectoral production functions; eqs. 1, 2, Indonesia 125   6.3 Coefficients of factor income distribution to institutions, ωqh , πh; eq. 7, Indonesia 125   6.4 Coefficients of income transfers between institutions, ρhh’ ; eq. 7, Indonesia 126   6.5 Coefficients of institution’s expenditure on commodity consumption γhj ; eq. 9, Indonesia 126   6.6 Additional notations 129   6.7 (First) social accounting matrix for Indonesia, 1975 135  7.1 Notations 140   7.2 Nepal: selected results from four policy simulations in the static model; percentage change of simulation results over base values 1996 151   7.3 Newly introduced endogenous variables and coefficients 153   7.4 Nepal: selected results from four policy simulations in the dynamic model; percentage change of simulation results over base values, 2006 157  8.1 Notations 163   8.2 Sudan: annual growth rates of main variables under unchanging and varying relative prices, 1990–2000 175   8.3 Sudan: green GDP and degradation loss by sector of activity under unchanging (U) and varying relative prices (V), 1990–2000 178   8.4 Sudan: percentage change of variables between policy runs and base run in year 2000 179   9.1 China and Russia: GDP and indices of GDP 190   9.2 Output of major goods in China and Russia 190   9.3 SAM multipliers of Russia 1990 193   9.4 SAM multipliers of China 1989 193  9.5 GLIs Russia 1990 197  9.6 GLIs China 1989 198   9.7 Social accounting matrices of Russia 1990, in billion roubles 200   9.8 Social accounting matrices of China 1989, in billion yuan 201 10.1 Notations 205 10.2 Poland: results of policy simulations (percentage change from base) 213 10.3 Poland: other results of simulations (percentage change from base) 213 10.4 Hungary: results of simulations (percentage change from base) 217 10.5 Hungary: other results of simulations (percentage change from base) 217 10.6 Growth and distribution discrepancies derived from policy runs on price flex and price fix regimes (CGE − SAM) 220 11.1 Multipliers of sectoral spending injections Ms,jj’ and Ms,hj’ , distribution of multiplier effects and computations of gainers and losers index, GLI, Netherlands, 1978 226 11.2 Multipliers of household income transfers Ms,hh’ and Ms,jh’ , distribution of multiplier effects and computations of gainers and losers index, GLI, Netherlands, 1978 227 11.3 Changes in gainers and losers over time 229

List of tables  xxiii 11.4 Multiplier effects of sectoral spending injections, Ms,hj’ and Ms,jj’ , and regional income transfers, Ms,hh’ and Ms,jh’, for the two periods of 1981 and 1985, Netherlands, and computations of the gainers and losers index, GLI 233 11.5 Decomposition of changes in sector output and regional income between 1981 and 1985 in terms of changes in SAM multipliers and changes in exogenous variables, Netherlands 235 11.6 Urbanisation levels: gainers and losers over time, and decomposition of growth into endogenous and exogenous factors 238 12.1 Review of the CGE model specifications in the basic and adapted forms 243 12.2 Notations 244 12.3 Basic CGE model Netherlands: policy simulations of reductions in indirect tax and direct tax 251 12.4 Elaborate CGE model Netherlands: policy simulations of reductions in indirect tax and direct tax 253 12.5 Restructured CGE model Netherlands: policy simulation of reductions in indirect tax and direct tax 256 12.6 Netherlands: income distribution effects of fiscal policy simulations 256 13.1 Notations 261 13.2 HRD positions: gaps between required levels and actual levels for Ethiopia in 2005 263 13.3 Ethiopia: projected population by age groups and total 264 13.4 Ethiopia: projected GDP and GDP per capita 264 13.5 Ethiopia: roadmap structure in terms of the three educational levels (primary: secondary: tertiary) 266 13.6 Ethiopia: future development paths of enrolments, teachers, and expenditure by educational level 268 13.7 Regional indicators (2006/07) and Regional Index of Educational Development 273 13.8 Estimated equations and detail assessment of estimated variables 280 13.9 Roadmap transition path 281 14.1 Colombia, Korea, Pakistan: annual growth rates of GDP, employment, and labour productivity in modern and non-modern sectors, and labour force for the periods 1970–80 and 1980–90 286 14.2 Occupations: forecasts of periodical growth rates per annum of demand and supply, imbalance rates, and unemployment rates after labour market adjustment for 1980, 1990 294 14.3 Education: forecasts of periodical growth rates per annum of demand and supply, imbalance rates, and unemployment rates after labour market adjustment for 1980, 1990 295 14.4 Intensity of forecasted labour market adjustment as measured by Theil’s coefficient 296 14.5 Row and column multiplicators for 1970–80 and 1980–90 297 14.6 Predicted major adjustments in the educational/occupational mix for 1980–90, Colombia (C), Korea (K), Pakistan (P) 298

xxiv  List of tables 14.7 Pakistan: results of different frameworks for estimation of rates of return to education 301 15.1 Notations 305 15.2 Economic and technical characteristics of selected Polish industries 313 15.3 Poland: main characteristics of the privatising enterprise and privatisation transactions 313 15.4 Poland: structure of discounted buyer costs, percentage 314 16.1 Notations 321 16.2 A representative human resource matrix by age and education 323 16.3 The representative population of 1000 patients by age, education, and disease types 324 16.4 Values of κgd 324 16.5 Values of φgd and Qged 325 16.6 Estimates of α and β coefficients 327 16.7 Average W and standard deviation for alternative policies obtained from sets of 20 simulations per policy 328 16.8 Ratios of relative performance of alternative policies 328 17.1 SAM features and GNP per capita of sixteen countries 340 17.2 Regression results of equations (4) and (5) 341 17.3 Selected simulations: initial runs for rich and poor countries 343 17.4 Selected simulations: alternative runs assuming different income levels 343 18.1 Indicators reflecting degree of firm intensity in FIM-related countries 360 18.2 Structure, conduct, and performance in transiting economies reflecting differentiated dominance of state settings 361 18.3 Future outlook of major countries as reflected by the Index of Interactive Influence based on country shares in the world totals with respect to population and GDP 364

Preface

The vast expansion in economic models for policy making has turned the field almost beyond systemic comprehension. In such a mushrooming world of economic models it is increasingly difficult to oversee the whole and deduct the cores. Deduction is a vital task in the development and application of scientific inquiry. Once in a while there is a need for a consolidation round in which related models are bundled and studied in concert and in retrospect. This book is such a consolidation round for works done by the same author on policy models in the past four decades. By bringing together in this book various policy modelling applications by the same author, integrating them in retrospect, adapting and rewriting them to highlight standards; and by relating these modelling applications via cross references to policy models of others, this work seeks to humbly contribute to the systemic unification of policy modelling. Furthermore, useful and handy modelling devices, methods and approaches that were developed in the contexts of past policy research, when systematised and simplified by leaving out hanging details, can benefit younger model builders in their modelling applications. Moreover, many of the policy problems of the past in developing and transiting economies are still with us today, albeit in modified shapes and in many more countries. The book contains ten economy-wide policy models and six partial models. The economy-wide models cover combined econometric multisector models, social accounting matrix (SAM) models and computable general equilibrium (CGE) models. They model and analyse policies of agrarian reform, basic needs, growth with redistribution, fiscal policy, trade liberalisation and sustainable development; and are applied to Chile, Colombia, India, Indonesia, Korea, Nepal, Pakistan and Sudan, among others. Two models study transiting performance of Russia and China, and regime transition towards price-driven economic systems in Poland and Hungary. And two more investigate dynamics, regional, and fiscal policy in SAM and CGE models for the Netherlands. The partial models include a normed planning model for human resource development for Ethiopia; a forecasting model of labour imbalances combined with an iterative fitting model to simulate labour market adjustments in Colombia, Korea and Pakistan; a model of privatisation transactions in transiting context, Poland; a model of economising solutions for resolving social queues combined with a random sampling model to investigate probabilities of effective economising

xxvi  Preface solutions in the Netherlands; modelling and measuring convergence in economic growth between rich and poor countries; and the modelling of distinctive behavioural types in socio-economic systems. The sixteen models are preceded by an introductory chapter on substantive contents, and an introductory chapter on modelling methods. We have sought to strike a workable balance between content and method in the presentation and analysis of each model; while placing details on content, methods, estimation and elaborations in endnotes. Taken together, the sixteen models reveal a particular approach towards policy modelling that has its origin in Tinbergen and Havelmo, and has gained support from Lucas, Krugman and many others. The approach perceives policy models as experimental designs: a kind of applied laboratories in which restructured designs and fitting policies can be demonstrated, tested and recommended for decision making and for changing policy. The approach extends and perceives policy models as creative designs to resolve policy problems that are not easily solvable by conventional means, and in conventional policy frameworks (that is, restricted time horizons, restricted instruments and unchanging foundations). The approach is open to experimenting with ‘innovative architects’, while respecting as the starting point the status of the policy model as econometrically tested ‘matching device’. It goes without saying that any element of experimental design incorporated in the policy model must observe that it is scientifically principled, functionally useful and operationally effective. Some details and selected results on several models included in the book were published in refereed journals, but all chapters have been rewritten. The renewal and adaptation of the modelling applications to fit within the focus of the book required that many applications had to be reformulated, throwing out some details, extending on others and rerunning some simulations. In their original forms, the applications reflected the changing modes of notations over the years. Besides, as some of the models were jointly developed with collaborators, the notations used for basically the same intended variables, coefficients and indices tended to vary between the models. An attempt was made to harmonise the specification of equations, and facilitate uniformity in the notations used throughout the book. As regards collaborative work, I wish to acknowledge my indebtedness and thanks to Sanjaya Acharya, Eisa Abdel Galil and Rini Braber for joint work in models presented in Chapters 7, 8 and 11, respectively. My thanks go also to Fred Lafeber, Hans Tuyl and Marco van Kessel for collaborative works done in Chapters 5, 11 and 16, respectively. The consolidation venture that we are presenting is more than a renovation of recipes to fit to tastes of the day. The final product required concentrated thought across a wide collection of models and applications, finding common denominators, some rerunning of policy simulations, retabulation of results and rewriting. It took close to two years to get the work done. During this period I was fortunate to be freed from the load of incidental preoccupations and social obligations. My deep thanks go to Els, who carried the load and was able to manage with my virtual absence. I am highly indebted to her understanding attitude and

Preface  xxvii cooperative spirit. I would also like to mention the personal support which I enjoyed from other family members: from Bram and Elles, and from Bas and Angele. In the two years of work on the book I was able to count on the secretarial, printing and computational support of the Department of General Economics at Erasmus School of Economics, for which I am thankful. In particular, my appreciation goes to Jany, Milky and Thea, who were quick and effective in providing the required support and servicing. S. I. Cohen Rotterdam, 2012

1 Introduction

1  Focus on economic models for exploring policies This book has some resemblance to renovations of recipes, which existed for years ago, to fit them to changing expectations and tastes of the day. There is the attractive result here that the renovations, recipes, and ingredients fit with each other, have together a common history, and are as topical today as they were then. While most of the content of the book comprises adaptations of research work done over some four decades, each chapter has been practically rewritten so that they reinforce each other and form a coherent whole that would be of special interest and immediate relevance to today’s economic modellers, policy analysts, development advisors, and the teaching profession in the related areas. In this first chapter, we introduce the lines along which the renovation venture is focused, formulated, and implemented, and the reasons why it is worthwhile to undertake this renewal attempt. We shall also discuss the outline of the book and the linkages between the chapters, and end with concluding remarks. To start with, it is important to underline where our focus will be within the various categories of economic models. There are pure models for formulating theories; and a wide range of applied models for testing theories, explaining events, measuring relations, forecasting variables, planning development, policy analysis, for decision making, and for teaching purposes as well. Our focus is on applied models for policy analysis and policy making, which can simultaneously serve for teaching purposes. Most of the models are economy-wide models applied to developing economies, with a few on transiting and developed economies. The book also includes several partial models. Analytical frameworks that qualify for what one would call today economic models date back from the nineteenth century, such as the theoretically oriented demand and supply schedules of Marshall; and even before that, in the eighteenth century, there was the empirically oriented tableau economique of Quesnay. But, the modern use of the term economic models was first introduced by econometricians of the 1930s, and became the recognised medium of economic sciences among theoretically and empirically oriented economists from about the 1950s onwards. It has become conventional to see economic models as being either theoretical or empirical: the first dealing with theory and the second dealing with applications.1 There is also a tendency, not altogether correct, to conceive the first

2  Introduction as being more suited for studying micro aspects of economic processes, while the second as being more suited for macro aspects. The two categories, theoretical and empirical, have different epistemologies, roles, characteristics, and limitations. However, there are cases of overlapping, complementarity and reinforcement between the two groups. There are no also strict rules for developing and using either of them. The result is a fluent and shadowy landscape of models, thus creating significant challenges and difficulties in the systematisation, transparency, and evaluation of the developed models and their use. Generally speaking, occurrence of these challenges and difficulties is less frequently encountered among the theoretical/mathematical models as compared to the empirical/applied models. There are various reasons for these differences. In epistemic terms, pure economic models are conceived as idealised frameworks for isolating, analysing and formulating crucial economic relationships. In contrast, applied economic models are built constructions of conceived worlds that necessarily contain additional properties and approximations that the modeller introduces to assure some degree of coherence between the presumed theory and the available data. It is often not possible simply to confront theory with data, since the data contain many more things than the theory. In applied modelling it is necessary to reformulate and reconstruct the model to suit the particular situation. For instance, Morgan and Knuuttila (2012) observe that as properties are added and attributed to the modelled economies and their behaviour, the model may start to look like an experimental construction rather than an idealised representation of the actual system. According to Lucas (1980), the fictional, artificial, and simulated elements that the modeller may introduce can allow for applied laboratories in which policies—which would be prohibitively expensive to experiment with in actual economies—can be demonstrated and tested out more quickly and effectively. Much earlier, Haavelmo (1944, 1964) and Tinbergen (1956, 1963) thought along similar lines by postulating that applied models should not be treated as matching devices, but as experimental designs. Today, many prominent economists2 agree that the way applied model construction takes place is more of an intuitive and creative activity than one of abiding to strict and verifiable rules that fit theory to data. Given the nature and purposes of applied modelling assignments, it can be expected that there are no generally agreed upon scientific rules for economic modelling that the profession abides with in executing these assignments. Each modeller, applying his or her artistic and creative skills, can excel by innovating along newer avenues; and if fortunate enough, he or she would end by successfully demonstrating and validating the construction and deriving valuable findings. These endeavours have significantly increased our understandings of policy problems and how to manage them. But there is a cost. The accumulation over the years of mixtures of applied economic policy models that treat hundreds of different problems and policies in different contexts makes the field of policy modelling almost beyond systematic comprehension. Add to this that with each slight difference or variation in assumptions, formulation, closure, measurement, and specifications of the simulated policy, there will always be differences

Introduction  3 in results and findings. These differences are read but are seldom subjected to further studies. It is as in many areas of art, when every singer comes up with their own songs. There is little or no effort made to explore common backgrounds. In such a mushrooming world of economic models for policy making, there is a genuine need for consolidation rounds: bundling related models and studying them in concert. Related models can be bundled in terms of their treatment of the same problem, or models that use similar data, or models constructed by the same author. By bringing together in this book various related economic models for policy making by the same author, integrating them in retrospect, and performing appropriate adaptations to highlight standards and commonalities among the models, it is hoped that the bundle contributes in some modest ways to enhancing the systematic unification and evaluation of policy modelling. There is more reason for taking up this retrospective renovation. It is typical of applied policy modelling studies that their results are read and appreciated at the time of their publication and dissemination, but the technical and research details, devices, and approaches that were developed in the process are forgotten thereafter. It is usual that most of the readers’ attention goes to the policy findings, while the technical details are often lost and overshadowed by the policy topic. The technical details may be of interest to a relatively smaller number of readers, and the interest may be of a timely nature. As a result, useful modelling devices, methods, and approaches that were developed in the contexts of past policy research tend to vanish. New applications have to develop their own devices, methods, and approaches; they could have benefited from previous works, if appropriately bundled and systematised. Moreover, it is also true that many of the policy problems of the past are still with us today, albeit in modified shapes and in many more countries. It is hoped that when the modelling details and policy problems are brought together and integrated, as this book will do, there can be a fruitful use of the retrospective outcomes, as they prove to be adaptable and applicable to more policy applications.3 Some past works and lines of thought can be only effectively linked in retrospect, thus making the linking of works an essential task in scientific research.

2  Outline: method and content The selected policy models in this book and their arrangement are the outcome of two considerations: the primary consideration relates to modelling methods; the secondary consideration relates to policy content: aspects of the setting, substance, and topic of the particular models. Regarding modelling methods, the next chapter will discuss technical issues in policy modelling. In this chapter we comment on the policy content. Chapter 2, as was just mentioned, will deal at some length with methodological issues. On methods, it is sufficient to mention at this point that the book is divided into two parts: we deal in the first part with ten economy-wide policy models and in the second part with six partial or theme models. The economywide policy models belong to three model categories: the combined econometric multisector model, which we dub as CEM model, and which was the mainstream

4  Introduction in the sixties and seventies; the social accounting matrix corresponding models, which followed immediately, in short SAM (social accounting matrix) models; and the extensive family of computable general equilibrium (CGE) models, which has become the mainstream for the last three decades. The ten chapters in the first part follow this evolution in the economy-wide policy modelling; there is more on this and related issues in the next chapter. The second part contains six models relating to various policy themes, they are partial models in the sense that they are not economy-wide. Methodologically, they follow different modelling approaches to be reviewed in due course. It seems natural that the primary consideration in the organisation of the book goes to methods. That settled, we can now introduce aspects relating to content of the various models. First, there are the ten economy-wide policy models. These models can be grouped in terms of the country context to which they apply: thus, models are grouped into those for developing, transiting, and developed economies. Among the ten chapters, six are devoted to developing economies. These are followed by two chapters on transiting economies, and two chapters on developed economies. Second, within each of these country contexts the various models are arranged more or less following their periodic occurrence. This helps in appreciating the changing interest of development policy modelling over time in response to changing realities. Furthermore, a periodic ordering would give insight into the progressive refinement of modelling methods, approaches, and content over the recent past. In this sense, there is a positive correlation between method and content. In this first chapter, we introduce the backgrounds to the various models of each chapter and their relations to each other. It is unavoidable in this introduction not to mention circumstantial remarks, since the circumstances at the different times have played their role in the particular specifications and applications of the models. To keep track of the main lines, such circumstantial remarks will be shifted as much as possible to endnotes.

3  Economy-wide policy models 3.1  Development context Models applied to developing countries form the majority of the book. They represent alternative responses in the later sixties, seventies, and eighties to the felt disappointment with the development planning effort in raising the welfare of the populations at large in developing countries. In the opening decades of development economics, mainly in the 1950s and 1960s, most development planners, economists, and statisticians identified economic development solely with economic growth. The central development objective was the growth of the GDP, and sometimes the GDP per capita. A superficial view, but very prevalent among development economists at the time was that where domestic and foreign savings are insufficient to promote investment and growth, then infusions of foreign aid (that is capital and exchange), if large enough, are by themselves sufficient to induce economic growth and economic development. Next to macro econometric

Introduction  5 models that forecasted economic growth, and multisector models that advised on industrial development and foreign trade, it was very convenient at the time to construct and apply so-called gap models to developing countries. These gap models calculated for individual developing countries the kind and height of the gaps that constrain economic growth, and the required infusions and composition of foreign aid to reach predetermined targets of GDP growth. These and related models formed the technical framework for the advisory work of the United Nations Committee for Development Planning (UN/CDP), which acted as the platform for launching and revising various plans for the First, Second, and Third Development Decades, and made recommendations on targeted economic growth and required foreign aid for the developing world.4 To the surprise of many observers, as data were gathered and analysed on actual performances in developing countries a new picture emerged. It became recognised that in spite of attaining some reasonable growth rates in their GDP, most developing countries were showing increasing income inequalities, surges in the population falling behind poverty lines, accompanied with increasing underemployment and unemployment, and deteriorating food and other living conditions. With the exception of a few lucky countries in East Asia and the Far East that were successful in combining growth with redistribution, the analysed data for the developing world revealed the contrary. The new picture brought a revision in interests and insights by economists and planners dealing with development problems. Four different types of reactions emerged as the response to disappointments in the achieved progress in social welfare. The responses were critical, to varying degrees, of the central role of the GDP in development planning, the models, policies, practices, and expectations of development planning. The four types of reactions can be characterised as follows. 1 The pessimistic perspective saw development planning and development policy as meaningless under the ‘soft state’. The postulates were that the skewed distribution of socio-political power hindered state power and economic development, and that radical reforms in the societal structure were necessary if development policy was to achieve its goal of a speedy and balanced social and economic development. 2 The social development perspective was more positive in tone. It saw the fault lying in the planning framework’s misplaced focus on GDP. The perspective opened the way for reformulating development models and plans in terms of incomes of social groups, employment and basic needs, and redirecting policy making accordingly. 3 The redistribution with growth perspective brought structure in the picture, and sounded more realistic in tone. It emphasised that there are dualities and linkages in the economy between rich and poor population groups, and between high-productivity and low-productivity activities that determine development structures, growth, distribution, trickle-down effects, and moving-up mechanisms. The advice was to focus on economy-wide modelling of dualities and linkages, and use them in designing and implementing

6  Introduction

4

packages of development strategies aimed at redistribution with growth. Methodologically, this perspective is associated with the popularity of the SAM serving as a helpful statistical tool in highlighting dualities and linkages in the economy-wide circular flow. The free markets perspective took off and became the mainstream response for the time being. It is argued that in the longer run, market-oriented economies develop more rapidly in terms of both growth and equality than planning-oriented economies. Consequently, development policy should be more reliant on the free operation of the market forces of supply and demand and market-determined prices. Where state intervention is necessary due to market failure, state intervention should complement, and not be a substitute for market forces. Methodologically, CGE models came just in time to allow for the realisation of the free-markets perspective in policy modelling. The flexibility of CGE models allowed for extending these models to treat new policy areas such as liberalisation reforms, environmental sustainability, energy resources, fiscal policies, and so on.

The four responses are well represented by the models in Chapters 3 to 8, which were developed and applied consecutively to various developing countries. Chapter 3 represents the pessimistic perspective response. It presents a model that stylises socio-political power in a feudal-oriented system, that is rural India of the 1960s, and in a reform-inclined regime, that is Chile under Allende; and it explores policy alternatives under the two contrasting settings. Chapter 4 represents the social development perspective. It builds up a social development and a basic needs approach and incorporates supporting variables in a combined macro econometric and inter-industry model, and applies the model to Korea. Chapter 5 represents the distribution and growth perspective with its emphasis on linkages and leakages in the circular flow between social groups, firms, government, and sector activities. The SAM is converted in a model that allows tracing of these linkages and leakages. The applications cover various developing countries. Subsequently, we include three chapters or models with a focus on market forces as effective drivers of the development process; these would fall into the free-markets perspective. These models specify the economy along the lines of CGE models. One model displays a simplified transformation of the SAM into a CGE model, and reflects on outcomes of policy simulations. The working of the model in its static and dynamic forms is demonstrated for Indonesia. Another model refines the CGE model to incorporate liberalisation measures with economic restructuring that combine to generate outcomes with higher growth and greater equality. The application is for Nepal. Another model operates within the CGE modelling framework, and incorporates environmental degradation and counteractive policies that assure sustainable growth. The application is to Sudan. Some introductory remarks will be made on each chapter and its related model. Chapter 3 presents in some way a pessimistic response to disappointments with the development performances of the sixties. The qualification pessimistic has to do with discussions reflecting opposing opinions on the development prospects

Introduction  7 among members of the UN/CDP. Jan Tinbergen, who acted as chairman, held an optimistic view, seeing governments searching genuinely for effective development policies and playing prominent roles in promoting development. In contrast, Gunnar Myrdal stood for the opposite view, that government in many developing countries represented the ‘soft state’, and that the course of the economy was dictated by powerful interest groups of landowners and other proprietors. Myrdal’s voluminous book The Asian Drama presented a pessimistic and a rebuffing perspective in which development planning, plans, and targets did not matter, and in which the benefits of economic development went primarily to the wealthy and feudal landlords.5 When Tinbergen was awarded the Nobel Prize in 1969, he promised to devote the proceeds of the award to fund research on the neglected topic of agrarian reform.6 The funded research culminated in a book on the modelling of agrarian structures and agrarian reform, followed by several published articles;7 these are the topic of the chapter. The basic model specifies four principal actors: landlords, peasants, the nonagricultural sector, and the government. Peasants and landlords are assumed to behave differently according to their own separate institutional attitudes, production function, savings and consumption patterns, tax and interest rates, and so forth. The novelty of the approach lies in adapting the basic model to applied models that represent different configurations of the socio-political power structure, that is, configurations in which the sole dominant actor is the landlords, or the peasants, or the non-agriculturalists. In the applied model only one actor is the leader and he has a decisive role in determining his desired goals, while the other actors are followers. For example, in a feudal society, the landlords are assumed to determine the future course of their welfare variables while other actors are followers. Government policy is supposed to operate within the prescribed sociopolitical structural constraints. In adapting the basic model to fit, and be applied, to different socio-political situations, the procedure followed was, first, to formulate a basic model that is underdetermined with 24 unknown variables and 22 equations, and second, to move towards an applied model, where two variables regarded as crucial in defining a particular political structure were exogenously fixed by the leading actor, thus yielding a determinate model. Which two variables are to be specified exogenously will depend on the particular actor who moves the economy, that is, the ‘leader’ group and their preferred welfare variables. The remaining actors are ‘follower’ groups. The chapter will apply the model to a socio-political system characterised by a landlord-leader configuration, that is, rural India of the 1960s, and will explore policy alternatives for the ‘soft state’ in the Indian context. The chapter will present also an application to Chile under the Allende regime where the proposed confiscatory reforms corresponded closely with a peasant platform. The applied model predicted grave costs from the proposed reforms for all other actors in Chile, and not least for the influential non-agricultural sector, suggesting that substantive opposition to the proposed reforms was inevitable. (The model was run and its alarming results were obtained in the year preceding the physical liquidation of the Allende regime by opposition forces.)

8  Introduction The role of government in these models depends on the particular sociopolitical agrarian structure described and simulated. The applied models contain instruments under varying degrees of government control: confiscation, and fiscal and pricing measures. The values of these instruments cannot influence the exogenously set variables of the dominant leading group but can influence other variables of the leading group and all variables of follower groups. The degree of intervention and the choice of the means would depend on the political colour of the government. Reflecting in retrospect, this chapter can be described as a modelling exploration that represents a cross fertilisation between the socio-political and economic realities that Myrdal has so vividly observed and analysed, and the renowned modelling approaches and positivist attitude towards economic policy that Tinbergen developed and applied. Chapter 4 is the response to the social development perspective. In the later sixties, the United Nations Research Institute for Social Development (UNRISD), Geneva, took the lead in criticising GDP as a development goal; and instead proposed, measured, and compared for many developing countries various basic needs indicators. Soon after, other UN agencies came up with their own lists of development indicators. But there was no work done to incorporate the proposed indicators in an economy-wide policy model. Our attempt at the time to fill this gap is in Chapter 4. It contains an applied model that supplements the development goal of the aggregate GDP by income and employment variables belonging to social groups and other variables on the satisfaction of basic needs; and it extends the scope of policy instruments accordingly.8 For the better-off groups, income is a sufficient indicator of wellbeing. For poverty groups, who are not in a position to advance their incomes sufficiently to attain basic needs, non-monetary indicators relating to nutrition, housing, health, and education are considered as targeted aim variables. The model considers also conventional macro and sector instruments, social group related instrument variables such as taxes, transfers, subsidies, and spending on public and merit goods. Another main feature relates to the formation and effects of the incorporated basic needs variables. The formation depends on the parts of income spent on the private consumption of these components, government spending, and propensities that apply for converting both types of spending in realised levels of basic needs. Regarding the effects of realised levels of wellbeing, the model integrates what various studies have already established, namely, that greater satisfaction of basic needs increases labour quality and labour productivity; and, in turn, can thus contribute to enhancing economic growth. This is done via specifying the labour input in the production function in a particular sector in terms of full capacity units. As wellbeing increases the worker’s capabilities are enhanced towards full capacity. One central question in development planning at the time was how to proceed with policy making when there are so many aim variables to be targeted and instrument variables to be solved. In answering this question we revert to the causal ordering of the modelled system, due to Simon (1953). It turns out that the prefixing of disposable incomes by social groups is the most logical first move in

Introduction  9 the policy-making process, followed by basic needs and employment targets. The GDP variable appears down the list following causal ordering. Primarily for reasons of data availability the Republic of Korea was selected as a test case for the model. The model was estimated on the basis of data for the sixties and employed to simulate development up to the early eighties. Quite interestingly, the results for Korea showed positive and coherent performances for the introduced socio-economic goal variables and the GDP. Predictions made around 1970, towards the next 12 years, happened to coincide very closely with the realised ex post values years later. The relation between the research results, government policy and economic performance is another story.9 Next, Chapter 5 represents what can be called the ‘redistribution with growth’ perspective. The term is due to Hollis Chenery. Under leadership of Chenery, acting as Vice President, Development Policy, World Bank, the world platform for development policy started shifting from the UN/CDP to the World Bank. In 1974 an important milestone in development economics was reached with the publication of Chenery et al. (1974). The title of their book was Redistribution with Growth, and their general theme was that distributional objectives should become an integral part of development strategy, should be expressed in terms of growth of income and consumption of different socio-economic groups, and be structurally related to their endowments. They surveyed existing multisector planning models and found them to be inadequate for formulating development strategies in those terms. They concluded that what is needed is a full circular flow model that provides a compact treatment of the determination of both the growth and distribution of income in different groups. They stressed also the importance of incorporating the dualistic nature of production and income generation in developing countries, that is, the formal and informal segments of economic activities; and the links between the segments and the socio-economic groups. Although these structural aspects were also incorporated in our preceding models in Chapters 2 and 3, what is particular for the redistribution with growth perspective is the focus on the trade-off between the two goals of growth and redistribution, and the recommendation for using a full circular flow modelling framework that highlights the linkages and interdependencies between different household groups, their factors of production, and the employment of these factors in economic activities. There is only one unified statistical framework that is capable of (a) collapsing the economy-wide circular flow in one compact shot, (b) is ideally flexible in linking production activities to production factors, income, and expenditure of various actors, (c) is generally accessible with available data, and (d) being squarely designed, and under simplified assumptions, it can be converted into an economy-wide model that covers various dimensions of development strategies. This is the social accounting matrix (SAM) first developed as an integrated system of national accounts by Stone, and brought into life again via a large number of applications and refinements carried by many contributors, including Pyatt, Thorbecke, Adelman, Robinson, Cohen, and many more. A few SAM applications are selected for Chapter 5. After briefly treating the conversion of the matrix into a model, the first application reviews multiplier

10  Introduction results relating to the trade-off between growth and equality for ten developing countries. The second application decomposes the multipliers into various effects. The third application uses the SAM multipliers to identify gainers and losers. A fourth application highlights the significance of dual structures in production and earnings in producing interdependent patterns of growth and distribution. Next are the models in Chapters 6, 7, and 8, which operate along the free markets perspective. They have in common a period that was dominated by a neoliberal outlook. In policy making, and in policy modelling, already from the late seventies onwards, there was a gradual shift from state planning proper to reliance on market mechanisms. The standpoint was that—with the exception of a few East Asian countries with high development performances—past government interventions in developing countries have caused incorrect prices, market distortions, limited growth, and regressive redistribution. The remedy sought was to let free-market forces determine the right prices. The free market perspective was strengthened further in the nineties by the collapse of the communist regimes and the rise of the Washington Consensus. Since about the year 2000, the neoliberal outlook has undergone modifications and refinements to accommodate liberalisation reforms to both market and state failures, and to include new areas of development policy such as transparent governance, environmental sustainability, and directed pro-poor economic growth. In terms of policy modelling, the CGE model fitted most with the free market perspective and its aftermath. Methodologically, it was natural that construction and use of the SAM preceded the extensive use of CGE models, since computationally some SAM is required as a benchmark for applying the CGE model. In substantive terms, the CGE model is a more helpful and flexible tool than the SAM model, as it is better able to calibrate price and quantity adjustments in factor and product markets, and thus placing the trade-off between growth and equality in a wider scope. Chapter 6 is primarily on transformation of the SAM into a CGE model. This early application, carried out for Indonesia, is of a demonstrative nature; and it is by far the first CGE for Indonesia. The chapter contains a static CGE model with a given supply of capital by sector and given total labour. The static model is modified and extended later towards a dynamic version of the CGE model, in which the supply of capital by sector and various skill types of labour are made endogenous. Policy simulations pertaining to upward shifts in the efficiency parameter in the production functions of the industrial sector and of the services sector are then simulated in both the static and dynamic versions, and the results compared. These simulations are meant to highlight the operation of the CGE model under the contrasting assumptions of the static and dynamic versions, and to emphasise the importance of the particular specifications in generating differentiated outcomes. For example, the simulated productivity gains in the industrial sector, which can be initially interpreted as a positive development, would turn out to have detrimental effects for wage earners who would experience a fall in their remuneration rate now that less labour is required but labour supply is given in the static CGE model. The fall in their income reduces aggregate demand further and cuts into the prospects for economic growth. By contrast, the flexibility of

Introduction  11 the demand and supply of factors in the dynamic version of the CGE model is able to push the results in other directions. Chapter 7 contains a CGE model that examines the growth and distribution impacts of import liberalisation and structural reforms. Many developing economies have undergone trade liberalisation in the context of the structural adjustment programme of the IMF and the World Bank during the last couple of decades. Most empirical studies find that trade reforms are accompanied by productivity growth, technological advancement, falling mark-ups, and a reshuffling of resources towards more efficient firms. However, the review of the substantial empirical research does not lead to a robust conclusion on the distributional impact of trade liberalisation. Moreover, the combinations of complex but partial phenomena, choice of variables, countries and periods, and data inadequacies and approximations have rendered the empirical works highly heterogeneous and thus defying conclusive findings. The highly diversified results on reforms in so many countries acted as motivation for undertaking this research. The chapter formulates and applies a CGE model for Nepal, making use of the SAM and other Nepalese data.10 The model simulated and analysed several liberalisation and stabilisation reforms. Simulation results showed that the liberalisation reforms were growthenhancing, but that the rich benefited more than the poor. The next stage of policy modelling was to expand and modify the static model to a dynamic model along the lines of the previous chapter, and envisage a restructured future economy that allows for growth with redistribution during a transformational period of ten years. The formulated structural changes were designed to turn the growth impact of the liberalisation reforms into a pro-poor growth impact. The structural changes simulated included improvements in efficiency parameters, reorganisation of investment patterns along with reallocation of factors of production by both household group and activity type. The simulations of trade liberalisation and other reforms were then applied to the dynamic and restructured specification of the CGE model, and the pro-poor growth effects were evaluated. The analysis suggests that a better development performance is possible when some particular phasing of the various reform measures is followed. The timing of the policy reform combinations matters, thus. Chapter 8, with title ‘Sustained development of land resources: a policy model for Sudan’ contributes to policy modelling in the area of environmental economics.11 There are four adverse interdependencies commonly acknowledged in agricultural activities. First, the inaccessibility of poor farmers to modern technical knowledge and information leads to misuse of natural resources. Second, farm-gate prices in most developing countries are far below their world market levels. This discourages farmers’ incentives towards soil conservation and encourages soil depletion. Third, lack of well-defined private property rights over natural resources lead to overexploitation and degradation of these resources. Fourth, pressured by their poverty, poor farmers adopt short-term survival strategies, and overuse land resources, thus giving environmental protection a low priority. The model, which shares common features with CGE models, incorporates these four interdependencies.

12  Introduction The model is used to address two sets of questions. First, what are the future prospects of a green gross domestic product, and is there a need in the medium and long terms for a genuine concern over resource degradation, or not? Second, which combinations of policies are most effective in their contribution to a sustainable and balanced development in terms of growth and distribution? The alternative policies treated relate to human capital, price incentives, property rights, and poverty reduction. In the African context, the results suggest that while the prospects of environment friendly economic development—a rising green GDP—are weak in the medium run, under certain structural conditions there is a range of effective policies that resolve conflicts between economic growth, fair distribution, and resource degradation; thus they contribute to a rising green GDP along with poverty reduction. 3.2  Transiting economy context Transition economics started coming into being around 1990. The regime crises in the Soviet Union and its European allies, and the breakup of the Soviet Union, opened the door for a new area of applied economic policy. Understandably, the new leaders in these countries looked westwards to the European Union and the United States for a helping hand to restructure their polity and economy. In practical terms, rightly or wrongly, there was only one alternative economic-political system available for adoption. This is the market-based economic system prevalent in the United States and the European Union.12 Western countries responded enthusiastically, displayed their systems of national institutions for adoption, and mobilised technical assistance towards that end. International agencies such as the IMF, the World Bank, other UN agencies, and the European Union mobilised their resources and gave advice and aid in transmitting these countries from state economies to market economies. Subsequently, from 1990 onwards, there was a surge in demand for economists who were willing and able to contribute towards the economic transition. This new demand for transition economists was especially attractive for economists working on developing countries, who were experiencing, already since the eighties, the negative effects of the fading out of the golden times of development economics.13 Many academic economists and economic policy advisors working on developing economies joined the caravan of transition economies.14 The modelling studies included in Chapters 9, 10, and 15 represent some of the activities undertaken on transition economics in the period 1990–2000. Chapter 9 reports on multiplier analysis of comparable SAMs for Russia and China. The benchmark is around 1990, which constitutes a crucial year in their transition to more mixed market-state economies. Russia’s GDP grew between 1979 and 1989 by 43.2 per cent and then decreased between 1989 and 1997 by about 60 per cent. China’s GDP has been increasing at an annual average rate of 9.5 per cent since 1979, giving a total increase between 1979 and 1997 of more than four times. The relative sizes of the two economies have reversed position in historically unmatched terms during less than two decades. Contrasting performances in growth and distribution between these two major countries have been

Introduction  13 persistent for a long period since the 1960s, suggesting that the differences in the structures and mechanisms behind these performance trends are enduring, and can be subjected fruitfully to a static comparative systemic analysis along the lines of the SAM framework. The model derived from the SAM simulates the multiplier effects of two types of policy instruments: (a) the effects of demand injections in sectors of activity on the output growth of activities, that is growth multipliers; and on the income distribution on household groups, that is income multipliers; and (b) the effects of income transfers to household groups on the output growth of activities and the income distribution on household groups. Growth multipliers in China are found to be higher than in Russia, reflecting more intensive and relatively equally spread circular flow interactions. Income multipliers are found to be less regressive in China than in Russia, which reflects stronger trickle-down effects and weaker leakage-up effects in the income and expenditure patterns of rich and poor household groups in China as compared to Russia. Chapter 10 makes handy use of the characteristic differences between SAM and CGE models to highlight the systemic differences between a planningoriented economic system that is driven by quantity adjustments, and a marketoriented economic system that is driven by price adjustments. The SAM and CGE models can be seen as the opposite poles between the central planning model and the free market model. The impact multipliers in a fixed-price SAM model assume unchanged relative prices so that all impacts go into quantity changes. In contrast, a free-market economy is commonly modelled as a CGE model. The rules for market clearance in a CGE model are different from those in a fixed-price SAM model. In the CGE model, producers maximise their profits and consumers maximise their utility in markets in which the demand for and supply of products and factors are cleared at flexible equilibrium prices. By switching from the SAM to the CGE model, the latter is able to replicate a free market situation with endogenous prices. The chapter simulates alternative policies of demand allocations on sectors, and income transfers on households. The simulations are applied to the fixedprice SAM model, that is, plan regime; and to the flexible-price CGE model to give alternative results that apply for the market regime. In this chapter, different results of the simulated alternative policies under the two models are analysed for Poland and Hungary, and for two different base periods. 3.3  Developed economy contexts All models above were applied to developing and transiting economies. SAM and CGE models are equally suited to deal with policy analysis for developed economies; and this is the focus of Chapters 11 and 12. Chapter 11 demonstrates the ability of the SAM model to investigate various issues relevant for a developed economy, that is the Netherlands. The applications were implemented in the context of an EU research grant on comparing and integrating national accounts systems in European countries. Other countries for which SAMs were constructed and analysed were Germany, Italy, Spain, Hungary, and Poland. In

14  Introduction this context it was possible to construct for the Netherlands a ten-year series of SAMs starting from 1978 up to and including 1987. The series was supplemented later with SAMs for 1995 and 2000. The chapter will study the SAM multipliers for an initial year, examine how they change over later years, and identify segments that were gainers and losers over a period of ten years. Available data for the Netherlands allowed for breaking up the population by income deciles groups, and for the regionalisation of the SAM into four geographical areas (North, East, South, and West); and the decomposition of regional economic performance over more years. Another extension of the SAM treats changes in urbanisation patterns. SAM multiplier analysis applied to a developed economy for more years gives some support to the turnaround hypothesis that future growth is conditioned by a weakening of (internal) multiplier effects and an increased dependence of the economy on (external) exogenous variables, that is, spending and transfers by government and rest of the world. This hypothesis gains validity under alternative extensions of the SAM, and for different periods. Chapter 12 treats a few aspects of modelling fiscal policy in a CGE model in the context of the Netherlands. Successful application of fiscal policy in times of economic crises and low growth is a very challenging job, and the number of countries renowned for a successful management of the fiscal budget is limited, the Netherlands being one of them. The CGE model in this chapter, built in the mid-eighties,15 was meant to address the fiscal policy problems faced in 1980–3, when the budget deficit recorded its highest share in the GDP, at 6 per cent, and economic growth was stagnant at 0 per cent (or falling at –1 per cent). In deciding on inclusion of this model in the book, the central questions to be answered were whether fiscal policy modelling methods of the mid-eighties are of any value today, and whether the policy content of three decades ago has any relevance for today’s discourses. With slight updating of the specifications of the model, and subsequent revisions in results, the answers were affirmative. The chapter formulates three forms of the CGE model and runs various fiscal policy simulations. The three CGE model forms are labelled basic, elaborate, and structuralist CGE models. In the basic form price elasticities of consumption, exports, and imports are set at zero, making the model less responsive to price changes. In the elaborate form, these elasticities are set at positive values. The structuralist CGE model specifies sticky wages, causing unemployment, and determining thus that the government paid unemployment benefits. Social security payments are also realistically modelled as sticky, in the sense that they are coupled to the consumer price index. Two revenue-reducing policy simulations with equivalent incidence on the government budget are considered: (a) reduction of indirect tax in the services sector, and (b) instituting a wage subsidy to lighten the wage bill that business firms pay. To allow for a fair appraisal, each simulation is set at 1 per cent of the government revenue. Depending on model mechanisms, the end result of the revenue reduction will differ and can be above or below the 1 per cent loss in government revenue. This end result can be one of the criteria in appraising policy effectiveness.

Introduction  15

4  Partial models The book contains six partial models that are treated in Chapters 13 to 18. Methodologically, each model follows a different approach. We introduce these chapters below. Chapter 13 formulates a normed planning model (NPM), where the development norms (in this case, the parameters of human resource development in the longterm perspective) are derived from cross-country regressions on best-performing countries. These norms are then applied to Ethiopia with the object of formulating a roadmap for the development of their educational system. The work done has its origin in a request for technical assistance in 2007 by the United Nations Development Programme (UNDP) and the Ethiopian Ministry of Education (EMOE). The assignment was to assist Ethiopia with formulation of a roadmap for long-term development of its educational system along lines that could lift Ethiopia’s rank among developing countries in the context of UNDP’s Human Development Index. Drawing a roadmap for human resource development, HRD, and education for the coming 25 years for a vast country like Ethiopia cannot sensibly rely on such methods as manpower forecasts or educational returns and market signalling that are more valid for the short- or medium-term, and for wellcircumscribed country outlooks and data. Furthermore, given the rapid paces of technological change and global competition, and the subsequent changing sector mix in the national economy, these methods may not be valid even for the medium term. A roadmap with a horizon of 25 years, starting from the base year of 2005, can only be drawn by making judgmental use of the past experience of selected countries that are known to have performed best in HRD, and whose realised structure of the education sector can be projected backwards so as to apply to Ethiopia. If the predicted paths of the select group of best-performing countries are followed along with associated institutions and policies, it is most likely that similar successful performances would occur in Ethiopia. The first task in the assignment was thus to list the select group of bestperforming countries. The second task was to develop a model of HRD and educational development, and estimate it on the basis of data from the select group. The third task was to use the HRD and educational development patterns of the model to outline the Ethiopia roadmap. The fourth task was to plan an adjustment period in which the current situation (that is enrolments, teachers, schools, costs, and the breakup of their financing into government and private resources; each of these for the primary, secondary, and tertiary levels of education) becomes adapted to the long-term balanced path that is predicted by the roadmap. Chapter 14 models the other end of human resources: utilisation. It formulates a forecasting model of labour imbalances in terms of demand and supply for occupational types and educational levels, and applies an iterative fitting model (known as the RAS algorithm or RAS method) to simulate labour market adjustments and clearance of the imbalances in the labour market. Economic theory has focused more on market equilibrium than on market imbalances. Although underlying causes behind market imbalances are known,

16  Introduction there is relatively less knowledge and analysis on processes of balancing demand and supply in particular markets in the real world. Formulating the demand for manpower skills in terms of demanded occupational types, and formulating the supply of manpower skills in terms of supplied educational levels, there is no reason why a priori the expectations on the occupational demand side would match the expectations on the educational supply side. The RAS-type iteration applied allows these demands and supplies to meet, adjust to each other, and end up in the matched mix of occupational types against educational levels. The model is applied to Colombia, Korea, and Pakistan. A further check is carried out to investigate how far the labour market imbalances are reflected in corresponding gaps in labour earnings. Chapter 15 presents a model that was developed and applied in the context of the privatisation drive in transiting economies. It models privatisation transactions along approaches of cost benefit analysis (CBA) as seen from buyers and sellers.16 The buyer’s price depends on the discounted present value of future sales, additional investment costs which have to be made, and desirability of meeting competitive returns on investment. The seller’s price is based on government’s calculations of gains and losses from the privatisation decision. Since the transaction may take place only if it satisfies both sides, buyer’s offers are set against the seller’s; the gaps between the two are evaluated for different sectors. The model includes a possibility of describing the process of negotiations by incorporating values of several exogenous parameters which can be managed by the buyer and the seller in order to satisfy both offers. The model is quantified and applied to several sectors of the Polish economy. Empirical results show how these sectors rank in attractiveness to buyer and seller. Simulations are applied also to vary values of several exogenous parameters that are managed by buyer and seller, with the objective of evaluating how these simulations would facilitate a takeover agreement between buyer and seller. Chapter 16 models benefits and costs of alternative ways of resolving social queues, and uses a random sampling model, RSM, to investigate the economic effectiveness of the alternatives. The model is concerned with showing that the formulation of economic criteria and their application to facilitate resolving social queues can bring about higher social welfare levels than, and can be superior to, current practices of allocating the limited public service to the many demanders of the service. The model studies the efficiency and equity effects of discrimination in medical treatments requiring surgical operations. It tries to answer questions on whether a discriminatory approach based on earnings income would increase the capacity and productivity of the health care sector, and result in the shortening or elimination of social queues. The model makes use of the Quality Adjusted Life Years (QALY) approach, as an indication of the effectiveness of medical attention in terms of gained life years in good health after treatment. In this way medical authorities can be guided as to which activities to undertake within their budgetary resources that contribute most to a healthy community. The model introduces the income effects and links them to the QALY approach. Selection of patients to be treated is based on productivity, as represented by the prospective earnings income of the queuing patients. Random sampling

Introduction  17 simulations (that is, the Monte Carlo method) show that patient selection based on maximising discounted lifetime earnings would release resources that can be used for treating more patients and thus reduce waiting time, and reduce the overall cost pressures. The results will show a convergence between efficiency and fairness goals. Although the application is for the health sector in the Netherlands, the problem of social queues is not unique to health services, or developed countries. Social queues can be observed in specific types of education, subsidised housing, unemployment retraining, poverty alleviation, and for many public goods in predominantly centrally planned economies. Chapters 17 and 18 are somewhat different to the rest. The model in Chapter 17 belongs to descriptive analytical models where a specific hypothesis is formulated to explain stylised facts. The hypothesis is then measured, analysed, and its policy implications studied. The hypothesis attempts to explain stylised facts on catching-up tendencies in economic growth between poor and rich countries. Taking all rich versus all poor countries together, various analyses of the statistical material have shown that there is a slight catching-up tendency. Economic theorising on these tendencies and empirical testing emphasised supply factors. We offer a demand-side model based on the SAM. We model the relationship between (a) growth multipliers, and (b) the exogenously assumed shares of government expenditure and foreign trade in the GDP. This is done for a selection of 16 rich and poor countries. The empirical results support the hypothesis of conditional convergence. The model predicts, after adjusting for peculiarities of economic systems, higher economic growth for countries at lower as compared to higher levels of income per capita, which is indicative of a convergent tendency. The main cause behind this convergent tendency is the ability of a poor country to increase significantly the shares in the economy of government expenditure and foreign demand; and to reap growth benefits from the associated changes in the compositional patterns of goods and services. In contrast, rich countries are, in relative terms, near satiation with respect to the shares of government and exports in the GDP. The conditional results are obtained after adjusting for system-country peculiarities. The chapter examines policy implications of this conditionality. The models in Chapter 18 are best described as conceptual analytical models. Several hypotheses are formulated to approximate distinctive behavioural types in socio-economic systems. The hypotheses are conceptual and are not readily testable, but their empirical validity can be demonstrated from comparative data across countries that represent differing socio-economic systems. We formulate an analytical framework that considers the roles of changing distributions of agents on behavioural settings and of their interactive behaviour across settings. The framework emphasises the location and interaction of agents in distinct behavioural settings as the clue for understanding how agents, and the economic system they form, become aligned with a particular behavioural setting, take over the typical behavioural type that associates with that behavioural setting, and spread it to other settings via various channels.

18  Introduction These processes lead in the long run to the formation of distinct economic systems with distinct behaviours. The focus is on three behavioural settings: household, firm, and state. These are primarily driven by social, economic, and political behavioural motives, respectively; and they result in three distinct systems, denoted by HIM, FIM, and SIM. The analytical framework models the evolution of the household intensive economic system (HIM), and its development into either the firm intensive economic system (FIM) or the state intensive economic system (SIM). The framework discusses a fourth economic system that has been named multipolar system (MPM). The empirical part of the chapter positions countries worldwide along the four types of economic systems, validates the positioning via various indicators, and carries the analysis further towards studying the future outlook of economic systems and global governance.

5  Concluding remarks We started by stating that the field of policy modelling has flourished and mushroomed over the recent past to an extent that makes it increasingly difficult to oversee the whole and deduct the cores. Deduction is a vital task in the development and application of scientific inquiry. Once in a while there is a need for a consolidation round in which related models are bundled and studied in retrospect. This book is such a consolidation round for works of the same author on policy models in the last four decades. Taken together, these models reveal a particular approach towards policy modelling that has its origin in Tinbergen and Haavelmo, and has gained support from Lucas, Krugman, and many others. The approach perceives policy models as experimental designs: a kind of applied laboratory in which restructured designs and fitting policies can be demonstrated, tested, and recommended for decision making and for changing policy. The approach extends and perceives policy models as creative designs to resolve policy problems that are not easily solvable by conventional means. The approach sees a promising perspective in the incorporation of ‘innovative architects’ in policy models; with due respect to the benchmark status of the policy model as an econometrically tested ‘matching device’. Of course, it is understood: an experimental design-oriented model must observe that it is scientifically principled, functionally useful, and operationally effective. Some of the lessons and deductions gained from this consolidation venture are briefly summarised. First, the economy is subordinate to the socio-political regime that overrules it. Explicitly, or implicitly, specification of the socio-political regime is the primary closure rule in a policy model. In some contexts it is necessary to specify this primary closure explicitly. Chapter 3 contains models of agrarian reforms applied to India and Chile that highlight this crucial issue, demonstrate policy limits for subordinate agents, and detect policy opportunities as well. Second, there is a wide range of secondary closure specifications available in policy modelling. We find that experimenting with alternative secondary closures and investigating their consequences is vital for gaining insight into

Introduction  19 how results are generated, and using this insight for recommending policy. Alternative secondary closure specifications recur in most of the economywide models presented. This is related to analytical versus planning forms of the model (applied to Korea, Chapter 4), static versus the dynamic (Indonesia, Chapter 6), fixed foreign exchange rate and unknown foreign capital flow versus its opposites (Nepal, Chapter 7), changing versus stable relative prices (Sudan, Chapter 8), fixed versus flexible prices in planning and in market-oriented transitions (Poland and Hungary, Chapter 10), and alternative clearance of product market balances and the savings-investment balance in CGE models. Third, although it is natural that specification of the model precedes its solution, experimental design in policy models demands a reverse process as well, whereby incorporating desired and removing undesired solutions would require respecification of the model. This may go beyond inserting changes in parameter values, and can involve restructuring of the model and its causal ordering. This is particularly demonstrated in the CGE model that combines liberalisation with redistribution (Nepal, Chapter 7), in the CGE model on constraining fiscal-policy imbalances (Netherlands, Chapter 12), and in modelling the resolution of social queues along economic criteria (Netherlands, Chapter 16). Fourth, in some modelling contexts, a confrontational approach that shows the benefits and costs of opposite parties is most effective in understanding the conflict situation and resolving it. This is most apparent in the modelling of privatisation as viewed from buying bidder and selling government (Poland, Chapter 15); but is also apparent in the modelling of labour market imbalances among demanders and suppliers (Colombia, Korea, Pakistan, Chapter 14); and to varying extents when applied to social groups in Chapters 3 and 4. Fifth, a useful concept, such as the productivity effect of higher wellbeing, can be operationalised by designing it as an attainment index, whereby the effect matures as the observed attainment reaches its maximum. This is fruitfully used in the application to Korea. Attained sustainable productivity use and depletion of land resources is another concept that was designed accordingly (Sudan, Chapter 8). A related concept is that of the Index of Interactive Influence that is indicative of the relative dominance of interacting agents, groups, or countries (Chapter 18), though this index cannot attain its maximum for an agent. Sixth, in some contexts, combination, linking or feedbacks of different types of models can be helpful in representing the complex realities. This is demonstrated by combining forecasting with iterative fitting models—RAS methods— in simulating labour market adjustments and checking the outcomes against an analysis of differential rates of return (Pakistan, Chapter 14). Other examples are: a cost benefit model and a random sampling model are fruitfully combined to investigate probabilities of social and economic gains (Netherlands, Chapter 16); the combination SAM and CGE models for investigating transition is another example (Poland and Hungary, Chapter 10); and alternative decompositions and designs of the SAM model generate important feedback results (Chapters 5, 9 and 11). Seventh, the socio-political context, in which the economy functions, matters a lot; but autonomous economic forces matter too, in shaping structure, conduct,

20  Introduction and performance. The relative weights of the ‘socio-political’ and the ‘economic’ may change over time, and some believe in favouring the latter, with significant policy implications. For the time being, exploring models of distinct socio-political and economic systems, demonstrating their applicability, and investigating their implications, are some first contributed steps in connecting the otherwise separated disciplines (Chapter 18). Eight, it was just stated that economic forces matter in shaping structure, conduct, and performance. The SAM, when standardised and applied to more countries, would reflect on advantages that some countries have above others, and on the underlying forces behind them. (This was done for ten countries in Chapter 5, and for China and Russia in Chapter 9.) Furthermore, the same modelled economic forces are able to explain convergence in economic growth between rich and poor countries (Chapter 17). Ninth, since economic forces matter in determining performance, normed planning models (NPMs) for a particular sector, that are parameterised on the basis of the best-performing countries can be a helpful tool in drawing long-term roadmaps for that particular sector or development theme in a lagging country searching for the highest performance. (This is demonstrated in a HRD roadmap for Ethiopia, Chapter 13.) Tenth, and finally, the time horizon in experimental-design policy models is longer than in the short-term policy models. The modeller has a lesser degree of freedom in the latter case, which obliges him to conceive the model as a matching device. Creative designs need more years to implement. The exact formulation of time units in the long term is less compelling in most of the presented models, which can be viewed as an advantage. If the trajectory is followed, the outcome will be realised a couple of periods earlier or later. (This is implicit in the NPM, Chapter 13, the long-term models of Chapters 3 and 8, and the SAM models applied to comparative statics of various countries, chapters 5, 9, and 17.) A final note is in place before concluding. The renewal and adaptation of earlier research applications to the stated focus required, for many applications, the rewriting of the respective chapters, throwing out some details, extending on others, and rerunning some simulations. In their original forms, the applications reflected the changing modes of notations over the years. Besides, as some of the models were jointly developed with collaborators, the notations used for basically the same intended variables, coefficients, and indices tended to vary between the models. To facilitate uniformity in the notations used throughout the book, an attempt was made to harmonise the specification of equations, and to use throughout the same notations for variables that are more or less similarly defined.17

2 Some essentials in economy-wide policy models

1 Introduction In this chapter we mainly review a few essentials in the construction and application of economy-wide policy models. The methodological review will attempt to highlight the positioning of the various models towards each other and to policy modelling in general. To limit our discussion to essentials, it is important to delineate what we mean by an economy-wide (policy) model. While there are many ways in which economy-wide models can be classified and assessed, and there are no common opinions on these, nevertheless an economy-wide model is immediately recognised when it contains four types of accounting balances that specify equality of supply with demand in factor markets, product markets, the financial market, and in foreign payments. To assure ex post equilibrium in all four types of balances, adjustors are needed to equalise supply and demand. For example, the demand and (or) the supply sides of a factor or a product market need to contain variables that adjust themselves so that the demand and supply sides are equalised. Factor remuneration and employment levels are such adjustors in the factor market. Product prices and inventory changes are typical for the product markets. Several questions may be raised on interpretations of the above statements. What do the four accounting balances stand for, and what is special about them? Should all accounting balances be present in the model to make it an economywide model? Should the components and mechanisms of both the demand side and the supply side be specified explicitly? To what extent should the adjusting variables be explicitly specified? What is special about the four accounting balances? The factor market balances state that for each specified factor of production (one or more labour types, capital, and so forth), demand for it equals its supply. Adjustor variables, required to assure ex post equality of demand and supply, can be utilisation rates, remuneration rates, and the like. The three other accounting balances are special in the sense that they are three sets of interdependent accounting balances that are part and parcel of a consistent system of national accounts. For example: a The commodity/activity/sector/product balances equalise the supply of a product j (production plus imports) to the demand for that good (intermediate

22  Some essentials

b c

deliveries, consumption deliveries, investment deliveries, inventory changes, and export deliveries); for each product j there is a separate j balance. The financial flow balance, also known as the savings-investment balance equalises supply of investment funding (private, government, and foreign savings) to the demand for investment. Foreign payments balance equalises the supply of foreign exchange (that is, exports plus foreign capital inflow) to its demand (that is, imports). Because of the interdependence of the product balances of j × (a), (b), and (c), one of the balances is derivable from the others and can be omitted from the model. If the model is correctly specified, its solutions will automatically generate the one balance left out.

Should all accounting balances be present in the model to make it an economywide model? In principle, this should be the case. A model which focuses on the product markets, and does not somehow connect (explicitly or implicitly) the factor accounts with the product accounts is feasible and can be very useful; but, in our view, it is not an economy-wide model. Incorporation of the cross-dependence of factor and product market balances, the financial market balance, and the foreign payments balance are basic to economy-wide models. Specification of the cross-dependence can be explicit or implicit. Should the components and mechanisms of both the demand side and the supply side be explicitly specified? To what extent should the adjusting variables be explicitly specified? These are difficult questions to answer. Laying any strict boundaries on these issues will ignore a large number of past and present-day economy-wide modelling exercises. Many models are focused on and elaborate on the demand side, while others focus on the supply side. In many cases, the adjusting variables behind the equalisation of demand with supply are not explicated, or discussed; however, the implied adjustors can be uncovered through further analysis of the model or, equally possible, adjustors cannot be found because of ambiguities and shortcomings in the model. Since ambiguities and shortcomings cannot be identified before examining the model, it is necessary to be more flexible in drawing the boundaries of economy-wide modelling with regard to the presence and absence of adjusting variables. Within the broad domain of economy-wide policy models as viewed above, and in the context of the next ten economy-wide policy models to be displayed in this book, we tend to classify these models into three groups. 1

The first group can be called combined econometric multisector, (CEM) models. 2 The second group comprises computable general equilibrium (CGE) models. 3 The third group relates to models that have as background the social accounting matrix, SAM models; these have commonalities with the other two groups of models, but do not fit into either. SAM models lie somewhere between the CEM and CGE models in terms of both methodology and their time of appearance in the modelling literature. This is also the reason why in the coming chapters we start with CEM models, end with CGE models,

Some essentials  23 and place the SAM models in between. Chapters 3 and 4 treat CEM models. Chapters 6, 7, 8, 10, and 12 are applications of CGE models. In between are Chapters 5, 9, and 11 which are applications of SAM models. This chapter examines a stepwise approach towards the specification and closure of the three types of models. This will be helpful in positioning various chapters of the book in a broader perspective. Compared to each other, the CEM models combine behavioural equations of remuneration rates, supply response, investment, consumption, and so forth that are econometrically estimated from time series; with inter-industrial equations based on an input-output table for the base year. In these models, equalisation of demand and supply, market clearance, and adjusting variables are dominated by quantity variables, with a subordinate role for price variables, most of which are considered to remain fixed in relative terms. In contrast, the CGE model, in its static form, is calibrated from the benchmark dataset for a particular year, and employs (operationalises) some basic assumptions of general equilibrium theory (though some deviations from the theory creep into the operationalisation process). Furthermore, the links between the factor and product markets are more compactly formulated. The major difference is the greater reliance of the CGE on price variables in clearing markets, these being the drivers of quantity adjustments. The SAM, which converts the economy-wide circular flow into a matrix of transactions between economic agents, can be appropriately transformed into a matrix of coefficients, restructured and inverted to form a SAM model that shows how endogenous variables are affected by changes in policy-controlled variables. As in the case of CEM and CGE models, SAM models are extendable in various directions, but to a lesser extent. SAM models share with most CEM models the common feature of clearance of the factor and product markets via quantity variables. At the other end, SAM has been a valuable instrument in defining and advancing economy-wide models; in particular, the SAM for a particular year operates as the benchmark dataset that the CGE model employs for calibration. An analytical review of the evolution of and links between the CEM, SAM, and CGE model types, which complements the general picture we display, and which is highly recommended to read, can be found in Robinson et al. (1999). We shall in this chapter focus on the general picture, and minimise here the use of references to specific models. In the following chapters there will be sufficient opportunities to refer to published models that connect with the topics of the respective chapters. This chapter is organised as follows. Section 2 deals with the specification, structure, and properties of the CEM models. Section 3 takes up the same equations and reformulates them so as to include price variables in the framework of CGE models. The section then extends step by step the further elaboration of the CGE model in various directions. Section 4 examines the SAM briefly, and highlights its intermediate position between the CEM and CGE models. Finally, Section 5 concludes with a review table of basic features of the models.

24  Some essentials

2  CEM models: focus on market clearance via quantities 2.1  Model specification The CEM model, Box 1, makes use of the list of notations in Table 2.1. The list will be supplemented by additional notations as they are introduced in later sections. The equation numbers of the CEM model will be preceded by Roman number I, those of the CGE model by II. Furthermore, it is noted that throughout the book t is the index used to denote year t. Table 2.1 Notations Index j denotes activities (sectors of production), h denotes institutions (households and firms), g is for government, and r is for the rest of the world Endogenous variables: Value of total consumption expenditure by activity j Cj Ej Value of exports of activity j FCF Foreign capital flow expressed in foreign currency (in some CGE models FCF will be made exogenous) Value of government budget deficit GBD Value of installed investment (see endnote 2) I ICHj Value of inventory change in sector j KDj , KSj Demand for capital in activity j, and supply of capital in activity j LDj , LS Demand for labour in activity j, and total supply of labour Value of total competitive imports Mj Remuneration rate of capital in activity j KRj Remuneration rate of labour in activity j LRj Vj Value of the gross product of activity j (that is, gross value added of sector j) Xj Value of the gross output of activity j Disposable income of households h. Index h is extendable to include the Yh category of firms Zh Value of total incomings of households h. Index h is extendable to include the category of firms Value of total revenue of government Zg Exogenous variables: Cgj Government consumption expenditure on activity j FXR Foreign exchange rate (in some CGE models FXR will be made endogenous) Government expenditure on installed investment Ig Transfer payments from government g to household group h (extendable to firms) Tgh Coefficients αjo Calibration constant in the production function of activity j αjj′ Input/output delivery share between activities j αrj Non-competitive import share in the output of activity j βj Labour elasticity of production relating to activity j Rate of capital depreciation δ

Some essentials  25 εjo , εj γhj ıj κoj μrj , μ j θjc , θcj πjh , πjg ρhhʹ σ τhg τjg ωjh Λj Ωj Φ

Exports proportion in the output of activity j in benchmark year, and price elasticity of exported products of activity j Consumption proportion spent by households h on final consumption products of activity j Delivery share of activity j in the total delivery of installed investment Installed investment destined for use in sector j as a proportion of total installed investment. Sj= koj =1 Competitive imports proportion in the final consumption (and investment) delivered by activity j in benchmark year, and price elasticity of imported products of activity j Conversion shares from activities j to commodities c; and from commodities c to activities j Profit distribution share of activity j that goes to households h (and firms); and to government g Transfer proportion of incomes between households h (extendable to firms) Rate of growth of labour supply Rate of direct tax paid by households h to government g. The h index is extendable to include firms Rate of indirect taxes on output of activity j collected by government g Earnings distribution share of labour in sector j that goes to households h Functional operators for regressed employment functions, specified by sector j Functional operators for regressed wage functions, specified by sector j Functional operator for regressed private investment function

The first six sets of equations are on factors of production and the factor market. It is usual that economy-wide multisector models would specify the determination of the gross product in sector j as the combined results of two levels of production technologies. The first level is that of the production function technology. This can be formulated along the lines of Harrod-Domar, Cobb-Douglas, or CES production functions. (We shall keep to Cobb-Douglas, since this is the formulation followed in chapters that display the CEM model and in most chapters on CGE models. In one of these we go for the CES formulation).1 The second level is the input-output technology (that is, Leontief technology). Eq. I.1 represents the first level. It shows a Cobb-Douglas production function where the sector product Vj is determined by two factors of production: one homogeneous labour type that is employed, or demanded, denoted by LDj (later on more skill types of labour are introduced), and capital in use or in demand, denoted by KDj and their respective elasticities. The labour elasticity of production is βj and the capital elasticity of production is 1 − βj. Because of the assumption of constant economies of scale in the Cobb-Douglas production function, the two elasticities add to 1, while αjo is a calibration coefficient. Eq. I.2 represents the second level of production technology. Employing Leontief technology, domestic and imported intermediate deliveries and indirect taxes are deducted from the gross output to give the value added, all of these per activity j.

26  Some essentials Eq. I.3 is a regressed equation that specifies the average wage rate in each sector as a function of labour productivity, the unemployment rate, and unspecified exogenous variables. Eq. I.4 specifies profits by sector as the residue: the value added less wage payments. (Dividing profits by the capital stock gives profitability, or returns to capital for each sector separately.) Eq. I.5 is a set of two equations specifying the labour balance. Eq. I.5.1 specifies a technical-behavioural regressed function of demand for labour by sector, LDj , as dependent on production, wages, and other exogenous factors outside the model variables. Eq. I.5.2 specifies a projected total supply of labour, LS. The difference between total demand for all sectors and the labour supply is unemployment. Eq. I.6 is a set of two equations specifying the capital balance. Eq. I.6.1 equalises the capital in use KDj to capital supply KSj. Eq. I.6.2 states that capital supply by sector is reduced annually by depreciation, and augmented annually via a destination proportion, by (fixed) investment, or what can be better named installed investment. To simplify things, full utilisation of the capital stock is assumed. In the CEM model, wages and profits are structurally determined. This is in contrast with a CGE model based on neoclassical theory, in which each factor earns its marginal productivity. Since the exogenous values of labour and capital supply are given, eq. I.1 to eq. I.6 together are directly solvable to give per sector, where applicable, the six unknowns of gross output, value added, employed labour and capital, and wage and capital returns. We can proceed now to the specification of product markets and financial flows in several steps. In the second block of equations, eq. I.7 distributes factor incomes on earning household groups and firms h. Their incomes are supplemented by government transfers (and eventually foreign transfers, but we ignore these for the moment). Eq. I.8 shows disposable income after deduction of direct taxes. Eq. I.9 shows government revenues, which include rental income from public enterprises, and direct and indirect taxes. Eq. I.10 calculates the government budget deficit as revenues less public spending on consumption, transfers, and investment. The third block of equations specifies demand for the product markets. Eq. I.11 distributes private and government consumption demand on products of sector j via simple consumption proportions. Eq. I.12 specifies demand for installed investment,2 which is based on an econometrically specified investment function of variables encountered above, such as expected growth of output (approximated by past output growth), average profitability, government investment, and unspecified exogenous variables. As indicated, it is also conceivable in the context of a planning setting to fix the path of all required investment exogenously, I = I*. Eq. I.13 gives the specification of exports in a simplest form. Eq. I.14 does the same for competitive imports, which are taken as a proportion of consumption and investment. This simple model assumes a fixed foreign exchange rate. Finally, the fourth block contains the three national accounts balances which bind the whole model together in one accounting system. The first balance, eq. I.15, is a set representing product market balances. The equations state that supply

Some essentials  27 of products for sector j (consisting of output and competitive imports) equals the demand for them, consisting of intermediate goods, consumption, investment, inventory changes, and exports. In the final analysis, it is the inventory changes by sector ICHj that carry the burden of adjustment in equalising supply and demand in the product markets. The second balance, eq. I.16, is the financial market balance, also generally known as the savings investment balance: private savings plus government savings plus foreign savings (that is, foreign capital flow in foreign currency, FCF, converted to national currency via a given foreign exchange rate, FXR), are equal to gross capital formation (that is, investment plus inventory changes). In the final analysis, it is foreign capital inflow FCF which carries the burden of adjusting the financial flows of savings with investment. The third balance, eq. I.A.1 is the foreign payments balance. Payments spent on imports are equal to payments received for exports and foreign capital inflow. This accounting balance is not part of the model, and that is why it is numbered otherwise as an additional equation.

Box 1  CEM model: focus on market clearance via quantities Production, factor use and factor supply Vj = αjo ΠLDj βj KDj (1 − βj ) Xj = [1 / (1 − Σj′ α j′ j−αrj −τ jg  )] Vj LRj = Ωj [((Vj / LDj ), (LS − Σj LDj  ) / LS), exogenous variables] KRj . KSj = Vj−(LRj .LDj ) LDj = Λ j [Vj , LRj , exogenous variables] LS = LS t−1 (1 + σ ) KDj = KSj KSj = (1 − δ) KSj , t − 1 + ĸoj I t − 1

j (I.1) j (I.2) j (I.3) j (I.4) (I.5.1) (I.5.2) j (I.6.1) j (I.6.2)

Income: private and government Zh = Σj ωjh (LRj . LDj ) + Σj πjh (KRj . KDj  ) + Σh′ rh′h Zh′+ Tgh Y h = (1 − τhg ) Zh Zg = Σj πjg KRj . KDj + Σh τhg Y h + Σj τjg Xj Σj Cgj + ΣhTgh + Ig= Zg + GBDg

h (I.7) h (I.8) (I.9) (I.10)

Consumption, investment, exports, and imports Cj = Σh γhj Y h + Cgj I = Φ (Σj Vj, t,…, t − 1 , average KRj , exogenous variables) + Ig; or I = I* Ej = εjo Xj Mj = µrj Σj Cj + µri ιj I

j (I.11) (I.12) j (I.13) j (I.14)

National accounting balances: Product markets, financial flows, foreign payments j (I.15) Xj + Mj = Σj′ ajj′ Xj’ + Cj + ιj I + ICHj + Ej (I.16) (Σh Y h − Σh Σj γhj Y h ) + (Zg −Σj Cgj −Σh Tgh ) + FXR . FCF = I + Σj ICHj (I.A1) Σj Ej + FXR . FCF = Σj Mj + Σj αrj Xj

28  Some essentials The model is determinate and consists of 18 sets of equations (16 numbered equations plus two not separately numbered equations, these are eqs. I.5.2 and I.6.2) in 18 sets of unknowns. When these 18 equations and 18 unknowns are solved, they will automatically reproduce the foreign payments balance. In a circumscribed and interdependent accounting system of n rows, knowledge of n − 1 rows is sufficient to solve the system. The nth row is not required and is left out. We choose here to leave out the foreign payments balance of eq. I.A.1, but the same results can be obtained if instead of this balance, one would choose to leave out eq. I.16, or one of the j balances in eq. I.15. Deleting one of the balances is interpretable in terms of Walras Law that states that in general equilibrium, specification of market clearance for n − 1 markets is sufficient, and the nth market clearance can be dropped, this being automatically guaranteed. In principle, Walras Law applies for a state of general equilibrium, irrespective of whether the general equilibrium has come through price adjustors, quantity adjustors, or a combination of both. In this sense, the CEM model is as much a general equilibrium model as is the CGE model. The major difference lies in the specification and choice of adjustors. As was already stated, in the CEM model the adjustors are mainly quantity variables, while in CGE model the adjustors are price variables, which in turn, drive quantity adjustments. 2.2  Structure and clearance mechanisms in CEM model An ex post equilibrium state in a specific market is a situation where supply and demand are equal and where some adjustors carry the burden of market clearance. There can be more than one mechanism. Since there are four ex post equilibrium states (referring to factor markets, product markets, financial market, and foreign payments), and these markets are moreover linked to each other, a discussion of a clearance mechanism is a discussion of closure rules, and this can become ambiguous and may end nowhere if it is done in general terms, and without reference to one specific model. Add to this that any such specific model has to be scrutinised and its structure dissected in ways that can isolate and systematise the ordering of the different mechanisms that hold for the different equilibrium states, before a fruitful discussion of closures can be made. These conditions are not observed in discussions on model closures, with the result that there is much overlapping and less clarity. The discussion on model closures in this context dates back to Sen (1963), where he emphasised that some adjustor from the factor, product, or financial markets is required to satisfy the necessary ex post equality between savings and investment. Since then, discussions and applications have produced many overlapping options which make it increasingly difficult for many modellers to make the right choices in model specification. It is easy to quote some ten or more clearance mechanisms that were proposed to settle the savingsinvestment equality; most of them were made without reference to a specific model, and are noncommittal.3 Note that options and associated overlapping multiply in numbers when the model contains four different ex post equilibrium conditions. Choosing the right angles in specifying closures and market clearance, and identifying equilibrium adjustors is central for making the right choices in

Some essentials  29 designing the model. This is illustrated below. First, let us take the CEM model specified in Box 1. It focuses on quantity variables and does not consider price or money variables. Within that focus, it can be directly checked, and confirmed, that the model equalises supply with demand in the four ‘markets’. These relate to equality of supply and demand of labour and capital in the factor market: LD, LS, KD, KS, in eqs. I.5 and eq. I.6; the equality of supply of Xj and demand for Xj in the j product market balances, eqs. I.15; and the equality of total savings to total investment in the savings-investment balance representing the financial market, eq. I.16. The same applies to the foreign payments balance in eq. I.A1. As explained earlier this balance is meant for retrospective verification. Second, rules of causal ordering due to Simon (1953)4 can be applied to sketch the structure of the model. This will reveal its mechanisms. Causal ordering shows which variable at which order is responsible for the solution of other variables in the same and higher orders, and equally important, which variables are positioned as adjustors within which orders. This is shown in Table 2.2. The ordered structure of the model will throw light on domination and subordination of markets to each other, their clearance mechanisms, adjustors, and implications for variables and equations positioned at higher orders. In Table 2.2, the first order is the factor market. The sector variables of gross output Xj, capital use and capital supply KDj and KSj, labour use and labour Table 2.2 Causal ordering of the CEM model Eq.

Vj KDj KSj LRj LS LDj KRj Xj Zh Y h Zg GBDg Cj I

 1

X X

 3

X

X

 5.1 X

X

 5.2

X

X

 6.2

= 0

X

= 0

X

= 0 = LS

X

= 0 = KS

X

 4

X X

 2

X

 7

X

X

X

= 0 X

X

X

X

X

= 0 = Tgh

X

 8

X X

 9

Prefixed

X

X

 6.1

X

X

X

X

10

= 0 X

= 0

X X

11 12

Ej Mj ICHj FCF

X X

13

X

16 Note (a) = Ʃj Cgj + Ʃh Tgh

= Cgj = Ig , Vj

X X

X

14 15

= (a) X

X X X

= 0 X

X X X X X

X

X

= 0 X X

= 0 X

= ƩhTgh FXR

30  Some essentials supply LDj and LS, and hence unemployment and labour wage LRj are simultaneously determined in the first order. With the exceptions of KSj and LS, which are practically predetermined, each of the other variables carries part of the adjustment. The next variables, for which solutions are obtained, are ordered in accordance with their earliest occurrence. So, after the first order, come solutions for capital returns KRj and for the value added Vj. Next are several orders involving variables and equations of the institutions and their demand for products. The structure is simple, because each subsequent unknown is diagonally ordered and solved in a single equation. The last two orders in the model involving eqs. 15 and 16 are methodologically significant. They represent the product market balances, and the financial market balance, respectively. In the case of the product market balances, the table shows that the mechanism for equalisation of the supply with the demand for each sector is inventory changes at each sector, ICHj. It seems natural that inventory changes should be the product market adjustors in a quantity-focused model. Speaking in general terms, in an uncoordinated world, there are supplies of output produced and brought in the product markets. These supplies are confronted with uncoordinated demands for intermediate, consumption, investment, and export goods. As these supplies and demands do not balance, their equalisation necessarily means that inventory changes have to carry the ex post equilibrium adjustments in the product markets. Finally, in the last order, in eq. 16, supply of total savings meets demand for total investment (this is partly installed investment and partly the adjusting variables of inventory changes that were just found in a lower order). In an uncoordinated world, savings and investments do not meet. If both ends are to be equalised within the particular specifications of the model, and given the fixed exchange rate FXR, it is natural that the adjusting variable in the financial market would be foreign capital flow, FCF. In general, if the dissection of the causal ordering indicates flaws in terms of theoretical foundations or empirical verification, then these are sufficient reasons for further analysis and redesigns. If the analysis is satisfactory, the modeller can proceed to a further elaboration and (or) application of the model. Before proceeding further, a few remarks can be made on the position of several variables in this model. Take government budget deficit, GBDg, specially marked in the table. Once solved GBDg does not recur again, and thus does not influence other variables in the model. This raises the suggestion that it can be skipped from the model. But GBDg is incorporated in the model because there will be further elaborations on constraining its level so as to achieve a stable fiscal regime; in such a case, the variable becomes more integrated in the whole model. Another variable, foreign capital flow FCF, shares the same apparent redundant feature. In this case, verification of the systemic consistency of all solutions of the model requires feeding the solved values of FCF, together with those for exports and imports in the foreign payments balance, eq. I.A.1. Besides, the foreign exchange rate is assumed initially to be fixed; if it is made flexible, that may require making FCF unknown and that would affect the model in various ways. Apart from the above considerations, for most policy simulations that are usually carried

Some essentials  31 out, the obtained results for GBDg and FCF are major considerations in policy assessments. Finally, in the context of many developing countries, it may not be feasible to specify and measure reliably a behavioural-technical investment as in eq. I.12. It is then usual to fix the path of required investment exogenously, which is justifiable in a planning setting and where public investment is a major share of installed investment. Thus, eq. I.12 is specified as I = I*. This is a difference in detail and has no effect on the structure of the model, as is displayed in Table 2.2.

3  CGE models: focus on market clearance via prices 3.1  Specification, structure, and solution of the static model In highlighting the common and different elements in the CEM and CGE models, the same numbering of equations in both models is followed. Eqs. 1 to 6 refer to the factor markets. Eqs. 7 to 10 refer to income formation of institutions. Eqs. 11 to 14 describe the demand of these institutions for consumption, investment, imported, and exported goods. Eqs. 15, 16 and A1, are the same accounting balances common for all economy-wide models. Equations in the CGE model will be preceded by Roman number II, to distinguish them from the CEM model where I is used. In the production sphere, sector production is described along the lines of the Cobb-Douglas production function, or alternatively a CES production function. Both allow for substitutions between various labour skill types and capital. Labour and capital are paid their marginal productivities. Different labour skills are incorporated. Because for a specific labour skill type labour is mobile between the sectors, this results in a competitive uniform wage rate per skill type. Installed capital is assumed immobile, resulting in differentiated returns per sector. By definition, the static CGE model does not need equations to determine factor supply over time (eqs. 5.1 and 6.1 are dropped and this simplifies things). Later on, when the model is extended to a dynamic version, such equations are reincorporated but in different forms. Clearance of the factor markets in the CGE (that is, equalisation of factor supply with demand) is fully a matter of adjustable factor prices in the CGE model, which in turn drives factors to reallocate among activities. This applies for mobile labour and mobile capital. It will not apply to installed capital that is assumed not to be usable in activities other than where it was initially installed. The market mechanisms described are typically neoclassical, compared to the structural mechanisms applying to the CEM model. Of course, structural assumptions can be introduced in the CGE model, as will be done in Chapter 12. For the moment, the CGE model will replicate (non-structured) freely operating markets. Institution incomes, eqs. 7 to 10, are the same as in the CGE and CEM models. The spending behaviour for demanded products (goods, or commodities) is different in the two models. The CGE incorporates product prices, XPj, and where there are substitution possibilities regarding demand for alternative products, these prices will be functional in allocating the demand on competitive

32  Some essentials products. In eq. 11, private consumption of products j is dependent on their relative prices and on household incomes, and on the applicable price and income elasticities; the values of these elasticities add up to unity, in conformity with the assumed Cobb-Douglas functional relationship. In eq. 11.1, newly incorporated, the prices are taken relative to one of the products. The simple formulation followed here is to take the product of activity 1 as the numeraire, XP1. Now, that the price of this product is given, the model will have XPj price variables in j − 1 to solve. Next, is eq. 12 specifying investment demand; this equation undergoes no change. (We shall elaborate further on this equation in the next section.) Eqs. 13 and 14 treat foreign trade. In this CGE model, and for the moment, the foreign exchange rate, FXR, is taken as exogenous, though this status would be modified in alternative specifications later. Eq. 13 specifies quantity exported, (Ej / XPj). This is assumed in the first instance to keep its given benchmark quantity level, εjo Xj / XPj , but it can vary from that level, depending on the differential between offered domestic price, XPj , and the foreign price converted to own currency, (FXR . epj ), and on the export price elasticity, εj. In eq. 14, a similar formulation is followed for consumption of competitive imports as for exports: the benchmark quantity level is the import proportion for the benchmark µ jo multiplied by consumption and investment; but it can vary depending on relative domestic and import prices, and import price elasticity µ j. When the export and import price elasticities are zero, these two CGE equations return back to their CEM form. Next are the accounting balances. Eq. 15 in the CGE model, that is, the j set of product market balances, shows two major differences from the CEM model. The first difference relates to the intermediate deliveries that are expressed nominally in the prices of the recipient sectors, but are converted into quantities and reexpressed in the prices of the delivering sector. The second difference is the biggest and is the defining difference between the two models. Having in mind Table 2.2 on causal ordering, there are now in the CGE model XPj prices, j − 1 in number, which have to be solved in eq. 15. They are the main adjustors in the equality of supply with demand of the j products. The adjustor roles of quantity inventory changes, ICHj, in the CEM model are taken over by price adjusting variables XPj, counting j − 1 in number. Because the set of product market balances is j in number and the endogenous prices are j − 1 in number, there is one degree of freedom that can be utilised as an unknown to make the system determinate. Modellers have differed on the specific closure of this one degree of freedom. We propose to argue here for one specific closure, to be denoted by closure A, which we shall employ in some CGE models in this book; but we shall also describe alternative option B in the next section, and make use of it in some other chapters. Our proposed closure denoted as closure A gives a role for inventory changes as an adjustor in the product markets. (There is closure B that will be discussed in a later section.) Our closure A distinguishes between micro-meso adjustors for the individual sector products (these are the prices XPj which are j − 1 in number, and a macro adjustor for the whole economy, which is the macro value of inventory changes, denoted by ICH. This aggregate is distributed on the sector products via observed proportions υj which usually correspond with the relative size of sector output. Sector-specific product market gaps are viewed to be cleared

Some essentials  33 Box 2  CGE model: focus on market clearance via prices Production, factor use and factor supply Vj = αjo Π LDqj βqj KDj (1 − Σ q βqj ) Xj = [1 / {(1 − Σj′ αj′ j XPj′ / XPj ) − α rj − τjg ]Vj LRq . LDqj = βqjVj KRj . KSj = (1 − Σq βqj )Vj Σj LDqj = LSq KDj = KSjo

j (II.1) j (II.2) q, j (II.3) j (II.4) q (II.5) j (II.6)

Income private and government h (II.7) Zh = Σj ωhj (LRj . LDj ) + Σj πhj(KRj . KDj ) + Σ h′ ρh′h Zh′+ Tgh h (II.8) Y h = (1 − τhg  )Zh Zg = Σj πjg KRj . KDj + Σh τhg Y h + Σj τjg Xj (II.9) (II.10) Σj Cgj+ Σh Tgh + Ig= Zg + GBDg Consumption, numeraire; investment, exports, and imports γ γ j (II.11.1) Cj = Σh [γ hjo ΠXPj hj { γ'hjY h } (1 − Σ j hj )] + Cgj (II.11.2) XP1 = 1 I = Φ (Σj V j, t , … , t − 1 , average KRj , exogenous variables) + Ig; or I = I* (II.12) j (II.13) Ej / XPj = [εjo Xj / XPj,] [(FXR . epj) / XPj )]εj j (II.14) Mj / XPj = [µrj Cj )] [(XPi / (FXR . mpj )] μ j + μri ιj Ij National accounting balances: Sector balances, savings investment, foreign payments j (II.15) Xj + Mj = Σjʹ (αjj′ XPj / XPj′) Xj′ + Cj + ιj I + υj ICH + Ej [Σh Y h − (Σj Cj − Σj Cgj )] + [Zg − Σj Cgj − Σh Tgh ] + FXR . FCF = I + ICH (II.16) (II.A1) Σj Ej + FXR . FCF = Σj Mj + Σj αrj Xj

by the respective sector prices, XPj. Cyclical supply and demand gaps for the macro economy are viewed to be represented by the aggregate inventory change, ICH, and to be proportionately distributed on the sectors following their relative weights in total output; see eq. II.15. The savings investment balance, eq. II.16, raises again the issue of choosing closures. The variables of domestic savings and investment are already solved as the modeller lands in the savings investment balance. What is left to solve are the variables of FCF and FXR. One of these two variables should by necessity be endogenous, and the other exogenous. The case of FCF endogenous and FXR exogenous can be denoted as closure C. The opposite case where FXR is endogenous and FCF is exogenous is denoted as closure D. Both are fully legitimate in a deductive analysis for policy making, as will be shown in the chapters to come. To continue the exposition we shall choose here closure C for presentation purposes, which is also convenient, as it can be seen that the CEM and CGE models are closest to each other when the C rather than the D specification is followed. Furthermore, in terms of causal ordering, it will be demonstrated in the next section that the structure of the model is simpler under closure C than under closure D.

34  Some essentials Finally, the foreign payments balance eq. II.A.1 is the additional nth equation left out, being automatically generated from other accounting balances; this is same for the CEM and CGE models. The causal ordering structure of the CGE model is displayed in Table 2.3. The first order of the CGE and CEM models is about the same, though the CGE model is somewhat simpler because it is static and has fewer equations that determine factor supplies over time. But within the first order, the interactions between the unknowns, to be solved simultaneously, is more subtle in the CGE than in the CEM model, as prices and quantities are solved together. Proceeding downwards the causal ordering of eqs. 2, 4, and 7 to 10, is the same in both models. As can be expected, the big difference between the two models occurs in the product markets with the introduction of prices XPj in the CGE model. The causal order that stands for clearance of the product markets in the CGE model contains five equation sets that should be solved simultaneously as indicated in Table 2.3. The last causal order standing for the savings investment balance, where FXR is given and FCF is the adjustor, that is closure D, gives the same tail structure in both models.

Table 2.3 Causal ordering of the CGE model Eq.

Vj KDj LRq LDqj KRj Xj Zh Y h Zg GBDg Cj I

 1  3  5  6  4  2  7  8  9 10 11.1 11.2 12 13 14 15 16

X X X X X X X X X

X

X X X X X

X

X

X X

X X X X X

X X X

X X

Ej Mj ICH XPj XP1 FCF

X

X

X

X X

X X X

X X X

X X X X

X

Note: (a) = Σj Cgj + Σh Tgh + Ig..(b) = Σj Cgj − Σh Tgh

X X X X X X X X X

X X

X

Prefixed = = = = = = = = = = = = = = = = =

0 0 LS KS 0 0 Tgh 0 0 (a) Cgj 1 Ig , Vj FXR FXR 0 Σh Tgh FXR

Some essentials  35 3.2  Adaptations to alternative closure specifications The CGE model as described in Box 3 can undergo modifications and elaborations when subjected to policy simulations. We shall discuss several adaptations in this section, and their aligned closures. Closures, clearance, and adjustors in product market balances: It was stated in the previous section that because the set of product market balances is j in number and the endogenous prices are j − 1 in number, there is one degree of freedom that can be utilised as an unknown to make the system determinate. Modellers have differed on the specific closure of this one degree of freedom. We have placed this one degree of freedom with the aggregate inventory change ICH, as a macro adjustor of demand supply product imbalances for the economy as a whole. We called this closure A. In contrast, closure B, which is equally plausible, is done in two steps. The first step is to wipe out fully the role of the inventory change variables, ICHj, as adjustors in the product markets on the ground that they can be positive or negative for different products and that it is complex to operate within the CGE model. To get rid of the role of inventory changes as adjustors, they are often specified as fixed proportions of the gross output of the sectors. In this way, inventory changes are eliminated as adjustors. Alternatively, insert an aggregate inventory change ICH that can be fixed as exogenous, and distribute it proportionately on sectors j, or fix ICH at zero, or omit it altogether; consider implicitly that inventory change is part of I, and redefine I extensively to be the whole gross capital formation, (that is installed investment plus inventory change). In any one of these variations, the role of inventory change as an influential variable is eliminated altogether from the model. The second step is to throw out as well the equation of the investment demand altogether, that is eq. 12, so that the model does not contain any autonomous investment, from neither the private nor the government parts; and, instead, turn the variable of installed investment, I, into an adjustor of supply and demand in the product markets, to supplement the j − 1 price variable and thus reach the j number of unknowns required to close the model. Closure B assigns the roles of adjustors to sector prices and investment. It can be traced that the causal ordering of the CGE model undergoes only little modification under closure B, when compared to closure A. The major difference is that the role of inventory change as adjustor as closure A is taken over by investment in closure B. Furthermore, the specification of the savings-investment accounting balance remains the same. If FXR is fixed, then FCF becomes the adjustor of the financial market balance, which is the same in Tables 2.2 and 2.3. There are several problems with the approach in closure B. (a) Erasing the role of quantity adjustments in inventory changes in balancing product markets distorts the real world, since cyclical macro inventory changes are acknowledged macro-driven adjustors, next to commodity prices, which represent micro-driven adjustors. Regarding operationalisation difficulties of incorporating endogenous ICH, these are solvable along the lines we have proposed. (b) Deleting the equation which specifies demand for investment, eq. 12, may encroach on theory foundations. While it can be reasoned that inclusion of a behavioural equation for

36  Some essentials investment demand may not fit in an CGE model that is initialised for one benchmark year, the investment demand can still be inserted as an exogenous value for the benchmark year, as indicated in the equation, I = I*. A causal ordering analysis of the B closure (of using investment as an adjustor of the product markets) boils to saying that installed capital and the residual difference between goods produced and demanded (that is, inventory changes) are heaped together to give the addition to capital supply (production capacity); it is well known that inventory changes cannot constitute production capacity. (c) While for the benchmark year it is conceivable that producers had the right knowledge for combining the realised production of installed investment goods versus inventory change goods, this knowledge may not be available if policy simulations bring about other mixes of investment and non-investment goods. (d) The same argument of absence of knowledge on the stipulated mix will hold for buyers of installed investment goods versus inventory-change goods. The two closures are summarised in Box 3. Chapters 6, 7, and 8 will follow closure A, which is richer in terms of scope. In contrast, Chapters 10 and 12 will follow closure B. These selections are most suited for the country applications at hand. (In general, the A closure associates with the Johansen closure— see Endnote 3, which attributes a decisive role to an autonomously determined investment, implying a subordinate role for savings. In contrast, the B closure is a neoclassical closure where savings dominate over investments. The A closure would fit more in the development context where investment decisions are, for a great part, governmental. The B closure can be more appropriate for the developed country context, though not necessarily.) Closures, clearance, and adjustors in the savings investment balance Two CGE models in this book, in Chapters 6 and 8, are applications to Indonesia and the Sudan; they specify fixed foreign exchange rates, FXR. In these models, foreign capital flow FCF is endogenous. In Chapter 7, a CGE model of Nepal, the model starts similarly with fixed FXR and endogenous FCF, but switches in other policy simulations to a flexible FXR. Making the FXR endogenous required making FCF endogenous. In policy making, the deductive shift between opposite closures is helpful in highlighting the mechanisms of the CGE model, and in probing the feasible area for combining policy measures to obtain the best policy outcomes. In

Box 3  Closures for the product market balances Closure A: Specify eq. II.12 to determine installed investment I. Distribute IHC proportionately on sectors j. Equilibrium adjustors in (A): XPj for j − 1, and IHC. Closure B: Eliminate altogether the role of inventory changes in the model in one or the other way. Omit eq. II.12 which autonomously specifies installed investment, I, so that I becomes endogenous Equilibrium adjustors in (B): XPj for j − 1, and I.

Some essentials  37 Chapter 12, which is an application to the Netherlands, the CGE model is specified from the start in terms of an endogenous FXR and an exogenous FCF. The closure of an exogenous FXR and an endogenous FCF was called option C. We denote the opposite closure as option D. Both are displayed in the boxes below. The causal ordering of the CGE model changes appreciably when the modeller shifts from option C to D. Under closure C, the product market balances preceded the savings-investment balance in the solution of the model, see Table 2.3. In closure D, the two balances have to be solved simultaneously in addition to the foreign trade equations where the unknowns of exports and imports are specified as dependent on the unknown FXR. The highest (last) causal order under closure D comprises, therefore, three equation sets counting j in number (eqs. II.13, II.14, and II.15), and one single eq. II.16. Within this order, the following unknown variables are solved simultaneously: XPj for j − 1 product markets, ICH, Ej , Mj , and FXR, see Box 4. There is one more interesting closure in the savings investment balance that can be mentioned which is especially applicable in the early settings of economic transition and related recessionary situations. We shall call this option E; it will be the subject of further elaboration in Chapter 10. The closure was applied to Hungary and Poland. For both countries in the early 1990s, stabilisation of the economy and making some gains in economic growth was only achievable if there was a heavy dependence on FCF. The applied closure breaks up the FCF into two components. One part, denoted by FCFGBD, is meant to cover the government budget deficit GBD. The other part is directed to support the savings gap with respect to private sector investment, and is denoted by FCFPSI. The model is then supplemented by an additional equation that makes FCFGBD and FCFPSI

Box 4  Closures for the savings investment balance Closure C: Prefixed: FXR. Endogenous: FCF Equilibrium adjustor: FCF in eq. II.16 Closure D: Convert status of FXR into endogenous and FCF into exogenous Equilibrium adjustors: The causal order consisting of eqs. II. 13 to 16 is to be solved simultaneously to give solutions to an equal number of unknowns: Ej, Mj, XPj  for j − 1 sectors, ICH, FXR Closure E: Break down FCF into two endogenous foreign capital inflows: one going to the public sector and one to the private sector. Introduce the following three additional equations: FCF = FCFGBD + FCFPSI FCFGBD = ϕg/p FCFPSI FCFGBD = GBD

II.A.2 II.A.3 II.A.4

The causal order consisting of eqs. II.10, II.13, 14, 15, 16, and II.A.2, 3, 4 is to be solved simultaneously to give solutions to an equal number of equilibrium adjustors: Ej , Mj , XPj , FXR, GBD, FCF, FCFGBD and FCFPSI.

38  Some essentials dependent on each other via a ratio ϕg/p. The closure is able to solve for unknown values for both flows in correspondence with the needs of the government and private sectors, maintaining at the same time a flexible foreign exchange rate, FXR. In closure E it is possible to achieve at the same time an endogenous status for FXR and FCF, which turns to be an attractive formulation. Causal ordering requires simultaneous solution of the four sets of equations enumerated above in closure D, plus three additional equations introduced in the model, see Box 4. This model will be applied in Chapter 10. Constraining the government budget deficit The specification so far of the GBD is open-ended. Various governments are pressured to lay down upper limits on the deficit. Since GBD is almost always a government budget deficit and seldom a surplus, it is sometimes recommendable to constrain the budget deficit and limit it to a fixed proportion of the GDP. This is especially necessary as in the CGE model so far; FCF is endogenous and fills the whole gap of the budget deficit, which would bring limitations on the use of foreign capital by the private sector. Besides, the CGE model does not contain a monetary subsystem that can adjust to inflationary pressure if government overspends and the budget deficit overshoots. GBD can be constrained by introducing eq. III.A.5 where νg is the permissible share of the government budget deficit in the GDP, as displayed in closure F. The addition of one equation to the model requires seeking an unknown to keep the model determinate. This is accomplished in several steps. Let total government current expenditure (this is the sum of all consumption Cgj and all transfers Tgh, call it GCX) become endogenous and carry the adjustment burden. The GCX can be distributed in assigned proportions to the various items. The closure is summarised in Box 5. For a detailed application, see the model in Chapter 12 on The Netherlands. Incorporation of this refinement does not modify the causal ordering of the CGE model in any substantial way. It is also possible to multiply the direct tax rates by a newly introduced endogenous variable standing for tax scaling tariff. Such a variable has been denoted by TXS, and is incorporated in one of the policy simulations applied in the CGE model in Chapter 12 for the Netherlands.

Box 5  Constraining the budget deficit Closure F (III.A.2) Introduce new equation GBD = νg Σj Vj Introduce new endogenous variable for Government Current Expenditure denoted by GCX. Replace Cgj by γgj GCX, and Tgh by γgh GCX, whereby Σg (γgj + γgh  ) = 1. (III.10.1) Rewrite eq. II.10 to become (γgj + γgh  ) GCX + Ig= Zg + GBDg

Some essentials  39 3.3  Elaborations and refinements We abstained from specifying several refinements of the product market to keep the model simple and to maximise oversight. A few refinements can be mentioned now. None of these change the causal ordering of the CGE model so far described; and they do not involve changes in model closures. Eq. II.11 in Box 3 will undergo the elaborations displayed in Box 6. So far, the activity products produced by sector j were assumed to be equivalent to the commodities that are traded in the product market. The two sorts of goods can be distinguished now, denoting the latter with c. The distinction between activities j and commodities c would require adding to the XPj prices another set of prices for the commodities, denoted by SPc. The SPc is a composite supply price for commodity c formed by applying the shares of domestically produced goods and imported goods in the supply of commodity c, thus θjc and μrc, to the respective prices of domestic goods and imports, XPj and (FXR . mpc ), respectively. The equation for the SPc can be specified as in eq. III.11.1. There are four consequent elaborations for the CGE model, which are displayed in eqs. III.11.2, III.11.3, III.11.4, and a revision of eq. III.14 in Box 6. First, in eq. III.11.2 the private consumption function is rewritten in terms of the commodities c and the newly introduced commodity composite prices SPc. Second, in eq. III.11.3 the nominal private consumption demand in commodities c is converted into quantities, distributed on sectors and revalued in terms of the sector prices. When government consumption by sector is added the final consumption by sector is obtained. Third, the availability of prices of consumption commodities opens the possibility of constructing a consumer price index, CPI, which is a weighted sum of the SPc prices; the weights are the proportions of c in the whole of consumption. The CPI can be used as numeraire which is fixed at 1, CPI = 1, as shown in eq. III.11.4. This would replace the previously specified numeraire of XP1 = 1. Even though one can compute the CPI and use it for assessment of general price increases, it is not obligatory to use the CPI as numeraire. For instance, in Chapter 7 we use XP1 = 1 as numeraire, and also compute the CPI for analytical purposes. In Chapters 11 and 12, the CPI is employed as numeraire. Fourth, the imported proportions of final demand can be directly linked with the equation for imports, eq. III.14. The already introduced conversion rates θjc are handy in converting the imports of commodities to activities and, eventually, for activating a role for import price elasticities at both the commodity and activity levels.

Box 6  Elaborations of the CGE model SPc = Σj θjc XPj + μrc (FXR . mpc  ) c (III.11.1) c (III.11.2) Cc = Σh [γhco Πhc SPc γhc {γ'hc (1 − σh  ) Y h } 1 − Σcγhc] j (III.11.3) Cj = XPj Σc θcj (Σh Cc / SPc  ) + Cgj (III.11.4) CPI = Σc ζc SPc = 1 j (III.14) Mj / XPj = Σc θcj [(μrc Σh Chc / SPc  ) {XPj / Σcθ cj (FXR . mpc  )}μ j + μri ɩj I

40  Some essentials The above elaborations can be applied to investment goods as well. Fortunately, the simplification is often made that there is only one homogeneous good with one investment price index, this being formed as a weighted average of the prices of the investment goods which are domestically produced and those which are imported. This elaboration would require two handlings: (a) an equation similar to eq. III.11.1 is required to determine the investment composite price, which can be denoted by IP; (b) the other handling required is to convert nominal investment, I, from its implicit expression in IP prices, into its expression in terms of the price index of the delivering sector XPj, along the same lines as in eq. III.11.3. Besides expressing imports, consumption, and investment goods in terms of commodity c, exports too can be classified accordingly. Another elaboration is in reference to two sources of government revenue that were not mentioned so far: these are import duties and excise taxes. When these are introduced, eq. II.9 has to be increased with these additional revenues.5 3.4  Dynamic model The extension of the static to the dynamic model involves adding x sets of equations. In correspondence with these there should be x newly introduced sets of endogenous variables. The extensions relate to making endogenous the supply of the labour factors of production, and capital. These extensions will be thoroughly treated and applied in Chapter 6 (Indonesia CGE model), and revisited and applied again with some modifications in Chapter 7 (Nepal CGE model). It is sufficient to mention here the main contours of the extensions. While in the static model the total labour force is fixed, in the dynamic model the total labour force grows annually. The distribution of this labour force, on various skills, changes endogenously over the years. Taking the case of two skills (high and low) the dynamic model postulates that the skills ratio of high to low skills, RHL, after n years in the future, is determined by the relative remuneration rates of the two skills in the current year, RLq, next to other influencing factors such as the relative costs of acquiring these skills, and non-economic factors. The choice of the prospective worker (student) in the current year between offering himself (or herself) to high-skilled versus low-skilled employment depends on the relative benefit-cost ratios associated with the skill types. Once earners make their choice as to what skill level they will pursue, it takes several years, say (n) years, before the effects of the choices are realised in a changed structure of the labour supplies. The (n) years may vary between four and ten years, depending on the educational system and the classification range of high and low skills. The change in the skill ratio for the current year t, from the past year t − 1, is assumed to follow a distributed lag pattern in terms of its value in the future year of t + n and in the past year t − 1, This is done via specifying distributed lag adjustment coefficient(s). The prospective skill ratio for the coming years would gradually change under adaptive expectations and distributed lags. The supply of capital in the dynamic model is extended as follows. In the case where installed capital in a particular sector cannot move to other sectors—this

Some essentials  41 is the situation of immobile capital—the total investment solved annually from the static model is distributed over using sectors in proportions which are a function of the ratio of capital profitability to average capital profitability, that is KRj / Average of KRj. Sectors which show high profitability in period t − 1 will increase their share in total investment in period t, in line with an inter-sectoral investment-adjustment elasticity. Sectors with higher profitability would thus experience relatively greater increases in their capital stock and in their production than other sectors. The mobile situation is one where installed capital is reemployable in alternative sectors and its distribution over the sectors in any one year is realised in a way that assures one rental rate for capital in all sectors. The dynamic formulation for the mobile situation is simpler than for the immobile situation. Total capital is augmented each year by the total investment that is solved by the static model.

4  SAM models: relationship to the CEM and CGE models As Chapters 5, 9, and 11 will discuss the SAM model and its applications in detail, it is sufficient here to comment briefly on derivation of the SAM model, and refer to some comparative features between the SAM and the CEM and CGE models. The social accounting matrix, converts the economy-wide circular flow into a matrix of transactions between economic agents. As the SAM is a square matrix, when appropriately manipulated, it can be operated as an economy-wide model. To transform the social accounting matrix into an economy-wide model requires performing several steps. Assuming proportional relationships for the cells in terms of their column totals, a SAM coefficient matrix that relates variables to each other is obtained, call it AS. This is similar to the Leontief matrix of interindustry production technology relationships, AL, but is more comprehensive in coverage. By separating the variables in the SAM into endogenous vector v and an exogenous vector e the SAM model can be written as v = ASv + e. Inversion of the SAM coefficient matrix would give v = (I − AS)−1e = MSe, where MS is the SAM multiplier matrix. Although the structure of the SAM model will be shown to correspond closely with those of the CEM and CGE models, in Chapter 5, there are important differences. First, the SAM model cannot specify explicitly the factor market balances because it can register money flows only, and not factor inputs as labour numbers or capital stock. That is why the SAM’s solutions are considered to be determined by demand forces. The assumption is that supply forces adjust. Notwithstanding, the final outcome of the factor market handlings in the form of factor payments are explicitly incorporated in the SAM. In general terms, even though the factor market cannot be modelled in the SAM, and there is no explicit supply side of production, the incorporation of the factor outcomes means that the SAM takes account indirectly of the factor market operations, and it accommodates to these in implicit ways. Because of the implicit nature of the accommodation, the two versions of the clearance of the factor markets discussed under the CEM and

42  Some essentials CGE models are both plausible under the SAM. For example, in the CEM model, remuneration rates of the production factors are structurally determined; this can be perfectly accommodated in the factor outcomes incorporated in the SAM. In contrast, in the CGE, the remuneration rates are equal to their marginal productivities, following (mainstream) neoclassical theory; this too is implicated in the factor outcomes incorporated in the SAM. In the above senses, the SAM is flexible in its interpretation of the mechanisms of the factor market, and has been conceptually useful in unifying the distance between structural and neoclassical positions. The SAM can be described, in some sense, as the bridge crossing between the CEM model and the CGE model, both in terms of methodological grounds and of historical occurrence. Especially in the context of the SAM-CGE modelling, the availability of the SAM is a prerequisite for the calibration of the CGE model. Parameterisation will be treated in some detail in Chapter 6.

5  Review table and concluding remarks To repeat, the basic difference between the three types of economy-wide policy models—the CEM, SAM, and CGE models—is that while ex post market equilibrium is achieved predominantly via quantities in CEM and SAM, the dominant adjustors and drivers towards ex post equilibrium are relative prices in the CGE model. This makes the CGE more flexible and adaptable than the quantitydriven models. But there are many country cases in which relative prices are stable, and where a quantity model can be still defended as a valid approach to modelling the economy. One of the aims of this chapter, and the book as a whole, is to organise and synthesise applications of economy-wide policy models as much as possible along uniform lines—even though the models belong to different prototypes, were applied in different country contexts, and contain many less important details that we have left out—to assure more focus. In Table 2.4 we give (a) a review summary of the ten economy-wide models in the coming chapters; (b) how they fit into the three prototypes of CEM, SAM, and CGE models; (c) their specifications of production functions, factor market balances, product market balances, and financial market balances; and (d) the primary adjustors in each of these balances. The table is helpful in positioning forthcoming models within the expanding panorama of economy-wide models. The table can serve the same purpose with respect to highlighting the essentials of related models by other authors.

Some essentials  43 Table 2.4 Review summary of the economy-wide models in the book: some essentials Chapter model

Production function

Closure specifications and balance adjustors in market clearance: Factor market balances

Product market balances

Financial market

Developing and transition economy contexts: CEM models 3. Socio-political regime models: India, Chile 4. Social-economic development model: Korea

H-D: Nj, Lj, Kj Inputoutput C-D: Lqj, Kj Input-output

Foreign trade in agriculture and non-agriculture and investment I(a) Factor returns unemployment

Closure A Adjustors: ICHj in j balances

Closure C Adjustor: FCF

Exogenous: Cgj, Tgh , Ij, Ej Adjustors: Xj, Y h a.o. Exogenous: Cgj, Tgh , Ij, Ej Adjustors: Xj, Y h a.o.

Closure C Adjustor: FCF

6. Simplified statics and C-D: Lqj, Kj Factor returns dynamics: Indonesia Input-output LR, KR

Closure A Adjustors: XPj for (j − 1), ICH

Closure C Adjustor: FCF

7. Liberalisation restructuring: Nepal

C-D: Lqj, Kj Factor returns Input-output LR, KR

Closure A Adjustors: XPc for (c − 1), ICH

Closure C Adjustor: FCF Closure D Adjustors: FXR a.o.

8. Sustained development: Sudan

agr= H-D: N, K ind = C-D: L. K input-output C-D: Lj, Kj Input-output

Factor returns NR, LR, KR

Closure A Closure C Adjustors: Adjustor: FCF agriculture: Eagr\ industry: XPind

Factor returns LR, KR

Closure B Adjusters: XPj for (j − 1), I

Closure E Adjustors: FXR a.o.

Factor returns (not specified)

Exogenous: Cgj, Tgh, Ej Adjustors: Xj, Y h a.o. Closure B Adjustors:XPj for (j − 1), I

Exogenous: FCF Adjustor I

SAM models 5. Growth and Input-output Factor returns distribution: developing (not specified) countries 9. Transition performance: China Russia

Input-output Factor returns (not specified)

Closure C Adjustor: FCF

CGE models

10. Flexible prices in transition: Poland, Hungary

Developed economy context 11. Spending Input-output multipliers in extended SAM models: Netherlands 12. Fiscal policy in CES: Lj, Kj adapted CGE models: Input-output Netherlands

Factor returns LR, KR

Closure D Adjustors: FXR a.o. Closure F Adjustors: budget items

Notes Abbreviations: H-D = Harrod-Domar; C-D = Cobb-Douglas; a.o.= among others (a) For the handling of the adjustor in this model, see Chapter 3.

3 Socio-political regimes and economic development Exploratory models on agrarian reform in India and Chile

1 Introduction The background to this chapter is of special interest. In the sixties and early seventies UN Committee for Development Planning, chaired by Jan Tinbergen, held an optimistic view about governments playing prominent roles in promoting development. The view was criticised by Gunnar Myrdal who saw government in the development context as ‘soft state’, with powerful landowners obstructing development.1 Appreciating the views of Myrdal, when Tinbergen was awarded the Nobel Prize in 1969, he devoted the award proceeds to fund research on agrarian reform. The funded research culminated in a book on the modelling of agrarian structures and agrarian reform, and several published articles.2 This chapter starts from an economy-wide model in the tradition of the combined econometric multisector (CEM) model (see Chapter 2). Even though econometric estimation is restricted to regressing saving and consumption functions, while estimates of other coefficients were obtained from various statistical sources, the model is multisectoral and is best classified as a CEM model. The model is adapted to incorporate socio-political structures, and is used to draw boundaries for feasible economic policy making in the given socio-political setting. This is done for two case studies: India and Chile. The chapter is organised as follows. Section 2 describes the main features of the model. Section 3 specifies the basic model and explains its adaption to different socio-political systems, and displays the leftover feasible areas for conventional economic policy making. Section 4 adapts and applies the model to what we have called a ‘landlord-leader’ socio-political structure of India. Section 5 adapts the model to a ‘leftist regime’ and applies it to the agrarian reforms proposed in the period 1966–1972 under President Allende and his predecessors in Chile. Section 6 concludes.

2  Main features of the model In spite of many analytical and policy studies done on agrarian reform there is little work on addressing the problem within an economy-wide development model. One reason lies in the argument that implementation of agrarian reform is not so much a matter of policy modelling as it is of political decision. The argument is not valid, however, since all government actions require political decisions, and many of them are commonly modelled. Starting from scratch, the

Socio-political regime models  45 model distinguishes between four actors: landlords, peasants, the non-agricultural sector, and government. The main features of the actors and their interrelationships are summarised below. Peasants and landlords are assumed to behave differently according to their own separate institutional attitudes, production function, savings and consumption patterns, tax and interest rates, and so forth. Several links are modelled between peasants and landlords: (i) transfer of land assets between the two groups, (ii) financial lending by landlords to peasants, (iii) crop sharing between the two groups in the case of tenancy, and (iv) a functioning of the landlord as employer and the peasant as a hired worker with paid farm employment. When a transfer of land occurs from one group to another, it is assumed that the transferred land acquires the characteristic production function and value of the new owner group and ceases to function with the parameters of the previous owner group. Several links are modelled between the agriculture and non-agriculture sectors: (i) net financial assets of agriculture are invested in non-agriculture; and (ii) since the model is formulated within an input-output framework, we note that intermediate goods are delivered to all sectors, while capital formation goods are delivered solely by the non-agricultural sector, and final consumption goods demand for each sector’s goods depends on the income created in the concerned sector and the other sectors. In applying the model, the procedure followed to adapt it to different sociopolitical situations is, first, to formulate a basic model that is under determinate (in this case, the model contains 24 unknowns and 22 equations), and second, with preset values of two variables regarded as crucial in defining a particular political structure, yielding thus a determinate model. Exactly which two variables to specify exogenously will depend on the particular leading actor who moves the economy and his preferences. The remaining actors and their variables are followers. The government functions explicitly as a collector of a lump-sum land tax which can be redirected to investment, and can within limits adjust credit and wage policy. But the variation of any of the controllable instruments happens within the prescribed socio-political closure of the model, which can be dominated by the interests of (a) the landlords group, or (b) the peasants group, or (c) the non-agricultural population.

3  The modelling framework 3.1  Specification of the model The model makes use of the notations in Table 3.1. All variables refer to the end of the year, and variables in monetary units are expressed in constant prices. The model is summarised in four blocks, Boxes 1 to 4. Eqs. 1 to 7 refer to the peasant group; eqs. 8 to 14 refer to the landlord group; eqs. 15 and 6 concern the agricultural sector as a whole; eqs. 17 to 19 deal with the non-agricultural sector; and eqs. 20 to 22 specify national accounting balances, typical of economy-wide models.

Table 3.1  Notations Indices: h = Household groups and associated sectors of production, wherein 1 = peasant group, 2 = landlord group, 3 = non-agriculture. All agriculture is denoted by subscript (1+2). Endogenous variables E( 1 + 2 ) , Exports less imports of the agricultural sectors (1 + 2), and non-agriculture (3) E3 Fh Value of financial assets possessed by group h = 1, 2 I31 , I32 , Gross investment purchased by group or sector h, and delivered by nonI33 agriculture, h = 3 Nh Value of land assets by group, h = 1, 2 Price per acre of land possessed by group h, h = 1, 2 NPh NQh Land acreage possessed by group h, h = I, 2 Savings by group, h = 1, 2, 3 Sh Wh Value of total wealth by group, h = 1, 2 Gross production by group (sector), h = 1, 2, 3 Xh Yh Disposable income by group, h = 1, 2, 3 Exogenous variables Minimum of basic needs income per person for the poor farmers population BNY1 group 1 Foreign capital flow expressed in national currency FCF NQ(1 + 2) Total agricultural land acreage Total population P Coefficients αhh′ Input-output coefficients of intermediate deliveries from sector h to sector h′, h = 1, 2, 3 Discrepancy proportion between owner’s valuation and market value of land, δh h = 1, 2 ϕ23 Proportion of financial assets of landlords group (2) invested in nonagriculture (3) ϕ21 Proportion of financial assets of landlord group (2) lent to peasants group (1) γhao, γha Intercept, and propensity to consume agricultural goods in total consumption of group h, h = 1, 2, 3. The corresponding coefficients for non-agricultural goods are (1 − γhao ) and (1 − γha ) ιh Rate of discount used by a group for discounting its future income, h = 1, 2 ι12 , ι32 Rate of interest applying to loans made by rich farmers (2) to poor farmers (1); and to financial assets held by rich farmers (2) in the non-agricultural sector (3), respectively κ3 , κ32 Investment per unit of land in farmland 1 and 2 that is required to sustain productivity; investment is delivered by the non-agriculture sector 3 κgh Publicly financed investment per acre in land belonging to farmer group h, h = 1, 2 κ3 Incremental capital-output ratio the non-agricultural sector, 3 Proportion of the peasant population in the total population ν1 Initial production per acre, and its growth rate, h = 1, 2 πho , πh Proportion of land rental (includes capital rental) in gross output, h = 1, 2 ρh

Socio-political regime models  47 ρhh′ σho ,σh, , σ′h τh ωh ω21

Proportion of land rent of sector h which is transferred to sector h, h = 1, 2 Intercept, propensity to save from income, and propensity to save from wealth, h = 1, 2, 3. This savings function holds for landlords. The savings equation for peasant households and for non-agriculture are specified otherwise Land tax per acre, h = 1, 2 Labour earnings share in gross production, h = 1, 2 Wage payments received by peasants h = 1 as a proportion of labour earnings of landlords, h = 2

Box 1 contains equations for the peasants and landlords groups. Eq. 1 formulates gross production, X1, as a function of the number of acres that peasants possess, NQ1, initial productivity per acre, π10, and an exponential growth of productivity per acre, π1. Eq. 2 relates investment requirements I1 to production growth from actually possessed land. This reflects a Harrod-Domar production function, where investment in capital is directly related to growth of production, that is, π10 and π1. Eq. 3 determines the price of land, NP1, in several steps. First, an owner’s valuation of an acre of owned land is defined as the future return to that acre (land rental × future productivity levels), that is, ρ1 π10 (1 + π1 )t, discounted by the peasant group’s own discount rate ι1 and is obtained as the sum of a geometric series to give present value, thus ρ1 π10 (1 + π1 )t (1 + ι1 ) / (ι1 − π1 ). Second, the owner’s valuation differs from the market price for the same acre due to different discount rates that sellers and buyers may have, cultural attachment of farmers to own land, liquidity needs, indivisibility of offered land, non-transparent market information, and the like. To obtain the market value, the above expression is multiplied by a discrepancy ratio of market to own valuation, denoted by δ1. Third, when a constant land tax per acre is levied, τ1, the price of the taxed land is reduced; following a geometric series the present value of the reduction is −τ1 (1 + ι1 ) / ι1. The three steps combined give the price of a unit of land. Eq. 4 specifies the value of land possessed, N1, which is simply quantity multiplied by price. Eq. 5 defines wealth, W1, to consist of two assets, namely value of land and all that goes with it, such as capital and inventory, N1, and financial assets, F1. In the case of the peasants group, the financial assets consist of held cash and negative borrowings from the landlords group which usually surpass the cash held; so that financial assets can be expected to be negative.3 Eq. 6 defines income to consist of value added on own land, wages earned from hiring out labour to group 2, crop-sharing returns earned from tenancy arrangements with group 2, less interest paid on borrowed loans from the landlords group, less land taxes. There can also be some income from non-agricultural activities but, for simplification, this is assumed to be absent. Eq. 7 formulates savings as a residual of income after meeting basic needs of the population group, expressed as a basic needs income per capita BNY1

48  Socio-political regime models multiplied by the peasant population, ν1P. It is usual that savings are negative, which pushes peasant households to borrow from the landlords group. Eqs. 8 to 14 refer to the landlords group, indexed as 2. This a set of equations similar to those of the peasants group, indexed as 1, except for the following differences. Landlords do not hire out their own labour to peasants, and they do not function as tenants in land possessed by peasants. The contrary is the case, so that part of group 2’s value added is transferred to group 1. In contrast, the landlords group makes loans to the peasants group and receives rent in return. They also have property in non-agriculture and receive rent in return from it. The financial assets of landlords consist of cash held, loans made to peasant farmers, and property in non-agriculture. The loans and the property are specified as proportions of their total financial assets. The cash held is not explicitly specified but can be derived as residual.4 The savings behaviour of the landlord group is formulated otherwise in eq. 14. Next to the positive dependence of savings on income, the landlord is assumed to target a desired level of wealth. The closer is the realised to the desired wealth, the less there is an incentive to save, giving thus a negative relationship.

Box 1  Equations applying to the peasants group and landlords group Peasants group: output, investment, land price, land assets, financial assets, income, savings X1 = π10 (1 + π1 )t NQ1 I31 = κ1 π1o π1 NQ1 NP1 = {δ1 ρ1 π10 (1 + π1 )t (1 + ι1 )} / (ι1 − π1 ) −ι1 (1 + ι1 )/ι1 N1 = NP1 . NQ1 W1 = N1 + F1 Y1 = (ω1 + ρ1 ) X1 + (ω21 ω2 + ρ21 ρ2 ) X2 − ι12 ϕ21F2 − τ1 NQ1 S1 = Y1 − BNY1 ν1 P

(1) (2) (3) (4) (5) (6) (7)

Landlords group: output, investment, land price, land assets, financial assets, income, savings (8) X2 = π20 (1 + π2 )t NQ2 (9) I32 = κ2 π2o π2 NQ2 (10) NP2 = {χ2 β2 π2o (1 + π2 )t (1 + ι2 )} / (ι2 − π2 ) − τ2(1 + ι2 ) / ι2 (11) N2= NP2 . NQ2 (12) W2 = N2 + F2 (13) Y2 = [(1 − ω21 ) ω2 + (1 − ρ21  ) ρ2] X2 + ι12 ϕ21F2 + ι32 ϕ23F2 − τ2 NQ2 (14) S2 = σ2o + σ2 Y2 − σ′2 W2

Box 2 specifies equations applicable for the whole agricultural sector. Eq. 15 sums land quantities owned by peasants and landlords to give total agricultural land, NQ (1 + 2), which is fixed in acreage. Eq. 16 defines financial assets in agriculture as a whole, done in three steps.

Socio-political regime models  49 First, financial assets of peasants in year t is that in year t − 1, plus savings, less capital investment on land, plus (less) the amount collected (spent) in the sale (purchase) of transferred land: F1 = F1, t − 1 + S1 − I1 + [land sales and/or purchases]. Second, a similar equation holds for landlords. F2 = F2, t – 1 + S2 − I2 + [land sales and/or purchases]. Finally, since the sum of transferred acreage is zero and land is transferred at a common price, the two expressions can be equalised to give the outcome for eq. 16.

Box 2  Agriculture: land distribution, financial assets, value added NQI + NQ2 = NQ(1 + 2) (15) F1 − F1, t − 1 + F2 − F2, t − 1 = S1 + S2 − I1 − I2 (16)

Box 3 contains equations relating to non-agriculture. Eq. 17 is a Harrod-Domar production function5 where investment determines the change in gross production, via the incremental capital-output ratio k3 . Eq. 18 specifies Leontief input-output technology to derive the value added from the gross output. Reference can be made to Chapter 2 where a common feature of an economy-wide model was stated as combining a production function technology with input-output technology in the specification of production by activity. Eqs. 17 and 18 combine these two levels of specifying production for non-agriculture. The same has also been followed for agriculture.6 To obtain in eq. 18 the income of the rest of the economy (that is non-agriculture), adjust the gross value added to two items: deduct interest payments on financial assets owned in non-agriculture by the landlords group; and add land-tax revenue collected by government (the latter being part of the rest of the economy). Eq. 19 is the savings function for the rest of the economy.

Box 3  Non-agriculture (X3 − X 3, t − 1 ) = k3 (I33, t − 1 ) (17) (18) Y3 = (1 − Σh αh3 ) X3 − ι32 ϕ23 F2 + τ1 NQ1 + τ2 NQ2 S3 = Σ30 + Σ3 Y3 (19)

Box 4 contains national accounting balances. Eqs. 20 and 21 specify the product balances for agriculture and non-agriculture. In these balances, we employ h to denote agricultural sectors 1 and 2, and hʹ to cover agricultural sectors 1 and 2 and the non-agricultural sector 3. In eq. 20 agricultural production for

50  Socio-political regime models peasants and landlords is aggregated to give the total supply, which is equal to the total demand for agricultural goods consisting of intermediate deliveries, final consumption specified via consumption functions for spending agent groups, and net exports; noting that agriculture does not deliver investment goods. Eq. 21 is the equilibrium balance of supply and demand for the non-agricultural products. The difference between the two balances lies in the fact that non-agriculture delivers all investment goods in the economy. These consist of investment made by the two farming groups in agriculture and by non-agriculture. Government, indicated by subscript g, is additionally engaged in public investment in the farming lands via the terms κg1 NQ1 + κg2 NQ2. Eq. 22 is the savings investment balance showing the supply of domestic savings plus foreign savings (taking the form of foreign capital flow FCF) equal to the investment demand.7 Eq. A1 is the foreign payments balance that equalises net exports (exports less imports) to FCF. This balance is automatically realised when the other national accounting balances in eqs. 20, 21, 22 are realised, and is not counted as an independent equation of the model; hence its numbering as A1.8

Box 4  National accounting balances X1 + X2 = Σh Σhʹ αhhʹ X hʹ + {Σh [γha,o + γha ( Y h − Sh )]} + E(1 + 2) X3 = Σhʹ α3hʹ Xhʹ + {Σh(− γha,o+(1 − γha )(Yh − Sh )) + I31 + I32+ + I33 } + κg1 NQ1   + κg2 NQ2 + E3 S1 + S2 + S3 + FCF = I31 + I32 + I33 + κg1 NQ1 + κg2 NQ2 E (1 + 2) + E 3 = FCF

(20) (21) (22) (A1)

The basic model contains 24 endogenous variables, specified in Table 3.1, in 22 equations. The exogenous variables are total farm land NQ(1 + 2), population P, basic needs income BNY1, and foreign capital flow FCF. The model is under determined and possesses two degrees of freedom. The next section will examine how the socio-polity determines these two degrees of freedom and the closure of the model. 3.2  Adaptation of the basic model to different socio-political contexts The solutions of the model are directed towards the fulfilment of predetermined targets of one of the main actors. This represents the case where one actor has absolute control of the socio-political structure and is the sole decision-maker in the country while the other actors are passive. For example, in a feudal society, the landlords are assumed to determine the future course of their welfare variables while other actors are followers. The model is adaptable to represent different socio-political structures, that is, cases in which the respective dominant actors are peasants, non-agriculturalists, and the state.

Socio-political regime models  51 In terms of adapting the model to different socio-political situations, the procedure followed was, first, to formulate a basic model that is under determined with 24 unknown variables and 22 equations, and second, assume that the two variables regarded as crucial in defining a particular political structure can be exogenously set, yielding a determinate model. Exactly what two variables to specify exogenously will depend on the particular actor who moves the economy, the ‘leader’, and his or her preferences, the ‘leader’s welfare variables’. The remaining actors are ‘followers’. The approach is to hypothesise different exogenous variables for different socio-political systems. In a political system characterised by a feudal agrarian structure, what is called here a landlord-leader model, it can be realistically assumed that the sole movers and decision-makers in the economy is the landlords group; their two target variables may be described by landlord’s wealth, W2 , and landlord’s income, Y2. When both variables are fixed, a determinate model is obtained. To take a different example, in a socio-political system in which peasants are powerful enough to be the leaders, the predetermined variables can be such representative target variables as possession of more land, NQ1, and reduction or acquittal of financial liabilities, (−F1). This is a different determinate model than the previous one. Another example is that in which the mover of the economy is the non-agricultural sector and its managers. As a country becomes more modernised and urbanised, the emergence of a strong urban-oriented leadership can be expected. This leadership usually consists of the intelligentsia, the industrialists, labour unions, and the military. If the urban sector acts as leader, then one of these subgroups, or a combination of them, via the government or in spite of the government, will dominate political developments. Predetermined variables that would represent the welfare of the urban sector should differ in accordance with the distribution of power within the urban sector. One applicable target variable is the income of the non-agricultural sector Y3. Another target variable is non-agricultural wealth, which, although not explicitly specified in the model, can be represented by setting the agricultural financial assets (F1 + F2 ) at lower values, thereby increasing the financial assets of the non-agricultural sector. In the above examples the socio-political system assigns the model closure: the combination of exogenous and endogenous variables. It is a matter of further research and judgement to identify the type of socio-political system of particular countries. Once identified, each socio-political system produces a different causal structure. Causal ordering was discussed in Chapter 2; see also Simon (1953).9 The causal ordering of the landlord-leader and of the peasant-leader regimes are displayed in Figures 3.1 and 3.2. In Figure 3.1, the predetermined target variables Y2 and W2 are shaded. They are fed into eqs. 8, 11, 12, 13, shown bracketed, which are simultaneously solved to give values of F2 , Y2 , X2 , and the important variable of landlord, land, NQ2. Once known, its counterpart variable, peasant land, NQ2, becomes known and the solution of the model proceeds further by simple substitutions from lower to higher orders along a diagonal pattern. The peasant-leader model, Figure 3.2,

52  Socio-political regime models Eq.no. 10, 3

Y2 , W2 F2

14, 8, 11, 12, 13, 17

NP2 NP1 NQ N2 NQ2

9, 15,18

I 32

4 2,1,19 6 14, 7 16, 5

Y3

NQ1 N1

Y 2 , W2 S2 Y2

X2

I 31 X 1

S3

Y1 S1 F1

W1

20 22

I 33

21

E3

E(1 + 2) FCF

Figure 3.1 Causal ordering in the landlord-leader model

NQ1

Eq.no. 10, 3 4, 2, 1

NP2

NP1 NQ

11,9, 8

N1 NQ2

15, 5

N2

I 32

I 31 W1

X1 F1

X2 Y1

6 12, 13, 14, 16, 7, 17

W2

Y 2 S2

F2

X3

S2 Y3 S3

18 19

E(1 + 2)

20 22 21

I 33

FCF

E3

Figure 3.2 Causal ordering in the peasant-leader model

has a simpler structure. The predetermined target variables are shaded: NQ1 and F1. These allow for quick solutions for the peasant variables. Again, once NQ2 becomes known, the landlord variables are solved in a simultaneous system of four equations, marked in bracket. In both cases, the sector product balances of agriculture and non-agriculture, and foreign payments balance active at the higher orders, take decisive roles in determining the trade balances by sector, E(1 + 2), and E3, and capital formation, I33. We consider now the role of government. Depending on the particular sociopolitical power structure, the model considers various agrarian policies as being under different degrees of government control. These policies can be (radical) direct reforms as in the case of land confiscation and redistribution and debt cancellation, or can take the form of various indirect policy measures. Direct reforms cannot be realistically simulated in a landlord-leader setting where landlords have targeted variables that defy direct reform. But confiscating

Socio-political regime models  53 and redistributing land and debt acquittal are plausible in a peasant-friendly socio-political setting; and can be simulated by fixing Nh and Fh. Indirect policy measures apply in principle to all settings, irrespective of the socio-political power distribution. The model is able to consider and run the following policies. Yield rates, πho and πh, can be raised by productivity promoting measures. Credit policy can be pursued towards adjusting discount and interest rates: ιh, ι12, and ι32. 3 Land-tax policy can be entered via changes in tax rates: τ1 and τ2. 4 Tenure and wage regulation policies can be entered via changes in cropsharing arrangements and remuneration levels: ρhh′ and ωhh′. Furthermore, the framework allows for additional policies, such as new land settlements that can be introduced through certain extensions in the model.

1 2

The degree of intervention and the choice of the means would depend on the socio-political power setting and the political colour of the government. This can be illustrated by a few examples. If in a landlord-leader society the government is also dominated by landlords, there is little chance that any agrarian reform favouring peasants would take place. However, when the composition of government is not biased in favour of landlords, government is usually able to pursue indirect policy measures. In substantive terms, in a landlord-leader setting the government tacitly accepts the setting and works within that realm. When the socio-political power structure is more favourable to peasants, and the government has also the same inclination, the scope for policy making is wider and can include direct reform as well as indirect policy measures.

4  Selected results from application to India Our classification of the Indian socio-polity in the early sixties as a landlordleader type is essentially in agreement with Bardhan (1974). Application of the landlord-leader soft-state model to India is consistent with thorough analysis made by authorities on India, for example, Thorner (1956), Myrdal (1968), and Joshi (1975). For the substantive discussion and justification of the characterisation of India as a landlord-leader society, see Cohen (1978).10 The initial year chosen for applying the model is 1961. Estimation of the model is derived from data of the seventies. Table 3.2 gives values for a few parameters on which we focus. Among these are the predetermined growth targets for wealth and income of the landlord group, w2 and y2 , which are assumed to continue their past trends of a growth rate of w2 = y2 = π2 = 2.6 per cent per annum, reflecting conservatism of traditional landlords to maintain wealth and income growth in line with the growth of land productivity. The model is solved consecutively to give the initial values for 1961 (t = 0) and twenty following years. Results for the main variables are found in Table 3.3. The results reflect the path of the Indian economy fairly well, giving an annual projected growth rate of GDP of about 3.5 per cent over the whole 20 years against an observed rate of about 3.3 per cent.

54  Socio-political regime models One important result of these basic projections is the almost unchanged distribution of possessed land between landlords and peasants. In the projected 20 years only 3 million acres (ma) shift hands from landlords to peasants of a total land of 330 million acres. A second important result relates to the distribution of financial assets, which become increasingly concentrated in the hands of landlords at the cost of increasing the debts of peasants. As a result of the above shifts in land and financial assets, the distribution of wealth among landlords and peasants, that formed a ratio of 3:1 in the initial year, kept to about the same ratio of 3:1 in year 20. The ratio for income shifted from 2.2:1 to 2.4:1, respectively. The non-agricultural production and income are realistically projected to grow at higher rates than those for agriculture. The outcome regarding land distribution, NQ2 and NQ1, is very significant in determining other results. Substituting eqs. 10 to 13 into each other gives NQ2 in terms of the future targeted growth rates of w2 and y2, and a large host of parameters that include ω21, ω2, ρ21, ρ2, ι12, ϕ21, ι32, ϕ23, τ2, χ2, β2, π2o, π, ι2. Substituting values of these except π2 gives a solution for NQ2 as in eq. A.2. Deducting this from total land as in eq. 15, gives NQ1. NQ2 = [54900(1 + y2 )t − 80204(1 + w2 )t] / 2.75 − 98.51 (1 + π2 )t (A.2) This equation indicates that under a common growth rate of targeted wealth and income that is equal to π2, and if the term 2.75 would have been zero, then NQ2 would remain constant at its initial year’s value of 264 million acres. Because of the term 2.75, there is a slight annual reduction but at a diminishing rate, so that NQ2 stabilises after 20 years at 261 million acres. Next to basic run projections, it is conceivable to simulate an alternative future in which a modern landlord leadership predominates, and whose wealth and income aspirations are more ambitious than the traditional, less bounded to agriculture, and who would aim at y2 > w2 > π2. It can be seen from eq. A2 that this would lead to decreasing values of NQ2, as modern landlords are better off substituting financial assets for land assets. Under y2 = w2 > π2 or w2 > y2 > π2, representing a commercialisation wave, the opposite would occur making it more attractive for modern landlords to increase their land possession. We shall maintain here the basic run projections where traditional landlords predominate. Within the boundaries of the projections from the basic simulation, the question is how big is the scope for policy making that can raise the welfare of the subordinate group, without touching the predetermined targets of the dominant group? Notwithstanding the softness of the state, can it still pick successful policies at the margin? A few policy simulations applied to the Indian model will deal with these questions. We have stated before that the direct reform measures of deowning and redistributing land are excluded in the landlordleader model. But there are ample possibilities for indirect policy measures that can be simulated: (1) promotion of levels of land productivity with special emphasis on land of peasants, (2) provision of more credit facilities at lower interest rates, in particular to peasants, (3) fixation of higher taxes, and (4) tenure reform in the form of increases in the agricultural wage rate and in the peasants’ share in crop-sharing arrangements. The model contains parameters

Socio-political regime models  55 Table 3.2 Basic run parameters, and changes therein under different policy simulations, India Parameters

Values in basic run projections

1 Higher productivity

2 Credit advances

3 Higher taxes

4 Tenure reform

Leader targets Y2 = Y2,0 (1 + y2 )t y2 = 0.0259

W2 = W2,0 (1 + w2 )t w2 = 0.0259 Parameters















< 0.3>

< 0.27>















FCF

−5.1 (1.0 − 0.1)t



−5.1 (1.0 − 0.1)t

on each of these that can be adjusted, the model rerun, and the effects of each adjustment studied. Table 3.2 shows relevant parameters and how they were adjusted. Table 3.3 gives the results of each of the four policy simulations, and these can be compared with the basic run projections in the first and second columns, all for year t = 20. One remark is still to be made before looking at the results. First, to permit an unbiased appraisal of the effectiveness of alternative policies, we aimed at stimulating relative changes in the policy measures that are equivalent in terms of the public effort needed in implementing them. A common adjustment in the policy parameters of 15 per cent was followed. When one policy involves adjustment in more parameters, the sum of the adjustments should also amount to 15 per cent. The first policy raises farm productivity. A policy can be formulated in a ‘discriminate’ form (for example, public investment in lands possessed by landlords, κg2, can be cut back, and thus reduce their productivity growth π2), or in a ‘neutral’ form (both κg1 and κg2 are raised, and thus the productivity growth of both farmlands is raised: π1 and π2; with the rises higher for group 1 than 2). We ran simulations for both policies. Results of the discriminate policy are inferior to the neutral one. For example, wealth and income of the landlords is unaffected by counterproductive measures to their land productivity; instead, they will shift the incidence of the discriminate policy to peasants; and they will buy more land to meet their targets, leaving less land for peasants. Besides, the earnings income of peasants from work they perform in landlords’ farms would fall with falling land productivity. In contrast, under the neutral policy, rich farmers have higher produce yields on possessed land and need less land to maintain their targets, and thus they sell some land to peasants who will now benefit twice from having

56  Socio-political regime models higher yields and more land. We ran for each of the four policy measures both the discriminate and neutral forms. All results show that the discriminate policy is detrimental to all parties concerned, and is always deficient to the neutral one. As discriminate policies are self-defeating, the results in Table 3.3 will focus on the neutral policies. The credit policy employs part of the available foreign capital inflow, FCF, to finance credit advances to all farmers, giving relatively more to poor than to rich farmers. This easing of credit reduces the discount rates of peasants Table 3.3 Basic run projections and policy simulations for year t = 20, India t=0 Variablesa

t = 20

Policy simulations, values in t = 20

Notation Initial year

Basic run 1 Higher 2 Credit 3 Higher 4 Tenure Combined projections productivity advances taxes reform policiesb

Peasants Land owned Price per acre Fin. assets Wealth Output Income

NQ1 NP1 F1 W1 X1 Y1

65 1911 −3 122 23 25

68 3346 −26 203 42 38

82 3571 −43 251 54 43

105 3675 −82 303 65 42

69 3357 −25 205 43 39

84 3346 −56 225 52 38

156 3940 −74 541 129 54

Landlords Land owned

NQ2

264

261

247

225

261

246

174

1222

2058

2135

2177

2064

2080

2775

Price per acre NP2 Fin. assets

F2

42

70

80

118

69

97

125

Wealth

W2

365

608

608

608

608

608

608

Output

X2

61

100

97

86

102

94

54

Income

Y2

55

92

92

92

92

92

92

Other variables Agr. output X1 + X2

84

143

151

151

145

146

182

F1 + F2

39

45

37

37

44

41

51

Output nonagr X3

126

281

281

282

283

283

723

Inc. nonagr

Y3

126

166

169

170

168

170

455

Exports agr.

E(1 + 2)

12

4

9

9

5

6

Exp. nonagr Nat. income Inc. dist. Rank eff.

E3 ΣY Y1 /ΣY

−17 150 0.1648

−5 296 0.1292

−10 303 0.1409 III

−10 303 0.1380 II

−6 299 0.1295 V

−7 300 0.1270 IV

Agr. assets

601 0.0906 I

Notes a Land in million acres; value variables in billion rupees. Shaded rows are predetermined. b The policy simulations have an immediate impact on NQ1 instantly in the initial year, which is not realistic and results in exaggerating the long-run result. While NQ1 in the initial year t = 0 amounted to 65 million acres, the immediate average impact of the four policy simulations was to raise it to 82 million acres. To neutralise this bias, the terminal value of NQ1 is adjusted downwards by the factor 65/82; and the outcome carried forward throughout other variables where applicable.

Socio-political regime models  57 and landlords by 10 per cent and 5 per cent, respectively (sum=15 per cent, see preceding remark).The lower discount rates stimulate landlords to sell land to peasants, thus increasing the latter’s output and income. Besides, under lower interest rates, peasants pay less on their borrowings. As land moves from landlords to peasants, where land productivity is higher, agricultural output is raised. Output of non-agriculture is unaffected but income increases as liabilities fall. The third policy simulated is raising the land tax by 15 per cent; to maintain the same income net of the land tax, the farmer is assumed to raise productivity per acre by the amount of the tax. Being only a small change in productivity, it is assumed that this is made possible by raising the labour input. Results of the tax policy do not differ much from the basic projections. In many policy discussions that took place in the development context, some great expectations were built around land taxes, but the simulations here show poor results in this direction. The fourth policy simulates changes in the tenancy regime by which the shares of peasant income in the wage bill and land rental of farmland possessed by landlords are increased by 15 per cent. The results are confined to some redistribution of land, with the feedback effects to the rest of the economy limited, when compared with the basic run projections. When results of the four policies are assessed against such criteria as reaching the highest levels of land redistribution, rural welfare, agricultural production, agricultural income and non-agricultural income, and reaching the lowest financial liabilities for the non-agricultural sector (as the net financial assets of farming groups, F1 + F2 , are held in the non-agriculture sector, these assets are also the latter’s liability), we reach the following conclusion. This is that policies of credit advances perform best and then, in a descending order, are followed by productivity measures, tenancy measures, and higher taxation, all of them compared with the basic projection, which ends as last. The positive effects of the credit policy are somewhat exaggerated, since the policy leads to an immediate redistribution of land in the initial year 0, due to the newly stipulated discount and interest rates, which is not realistic. In practice, it will take a number of years before realisation. We correct for this bias in the next paragraph. One interesting question that can be raised concerns the relative performance of a combination of indirect reform measures. Would such a combined run give a better or a worse performance than the best of the individual runs? And if so, why might this be so? A combined simulation was carried out. This starts from the input of the basic run projections and incorporates into it the combined adaptations of the four policies shown in Table 3.2. The results obtained for the combined policies, shown in Table 3.3, final column, outrank in a remarkable way all of the other results found earlier. For example, the combined simulation predicts (above the base run) an additional transfer of land from rich to poor amounting to 88 million acres; when the sum of the additional transfers achieved by each indirect measure separately accounts for only 68 million acres. This better performance is due to the multiplicative effects of combining independent measures that are non-linearly related. Rates of taxes, discount, interest, yield, and yield growth are nonlinearly related. These rates also determine the prices per acre, which reach their highest levels under the combined simulation.

58  Socio-political regime models These effects enhance the prospects for increasing the wealth and income of rich farmers. To keep to their wealth and income targets the rich farmers require less land than they possess. Land is sold to the poor farmer, therefore, which explains the simulated rise in F2 as against a fall in FI. The higher agricultural production and income—via interindustrial deliveries and consumption demand—increases non-agricultural production and income, which are further pushed up by rising net exports and investment. Under the combined simulation, the national income would grow at an annual average rate of 7.2 per cent; all other runs gave a figure of about 3.5 per cent. GDP growth rates achieved by India in the seventies and eighties were somewhere between the two rates. This is not meant to suggest that if the proposed policy simulations were followed that a higher growth rate, towards 7.2 per cent, would have been achievable. Even though the model applied is an economy-wide model, its focus is specific, and it abstains from detailed treatment of savings, investment, consumption, and trade; which are main determinants of GDP growth. It is interesting also to point out that the share of the income of the peasants group in the GDP, which is a measure of income distribution, is highest under the combined policy at 9 per cent.

5  Application to Chile The Chilean socio-polity has been for a long time, and is still, a sensitive mixture of the landlord-leader, peasant-leader, and industrialist-leader types discussed earlier; the balance of power has shifted from the first to the second and the third over a period of three decades, and is still under pressure. Chilean governments (Allessandri 1957–1964, Frei 1964–1970, Allende 1970–1973, Pinochet 1973– 1990) played important roles in either actively promoting these shifts in power or in simply giving legality and shape to them. Although there is a long history of sporadic conflict between landlords and peasants on farms all over Chile, it has never yielded a sustained organised movement among peasants. However, during Frei’s regime rural workers organised when the state offered them patronage and protection. During Allende’s period, the peasant movement was able to multiply its power. Rural unions’ membership increased from 53,000 in 1967 to 140,000 in 1970 and 218,000 in 1972. This period also saw an upsurge in rural strikes, farm seizures, and land expropriation by peasants, the convening of the first Peasant Assembly in 1971, and the drawing of official demand declarations by peasant leaders. By 1973, the peasantry had become too independent to be managed in an orderly manner, but that year was also its climax year of power. The military overthrew the Allende government and installed the Pinochet regime. The agrarian problem has drawn vast attention from politicians and laymen, and greatly influenced Chilean events. Yet, Chile is highly urbanised (urbanisation rate = 90 per cent), and has a non-agricultural sector accounting for more than 90 per cent of GDP. A priori, it can be reasoned that agrarian solutions that are unacceptable to the industrial population have little chance of realisation in a society with this kind of socio-political balance. There is substantial evidence

Socio-political regime models  59 that the movement of the peasantry was looked upon with distrust by the industrial population (and industrial leaders). The political obsession with the agrarian problem derives as much from catching the political support of the peasantry votes as it does from attempts to reform the economically lagging tenure system, and its low agricultural productivity. The socio-political structure of Chile in the 1960s and early 1970s can be conceived in our analytical framework as a mixture between the peasant-leader model and the industry-leader model. The peasant-leader model, it is recalled, is the one in which peasants’ welfare variables are considered predetermined, especially land property, NQ1, and financial debt, F1. To allow for the fact that the industrial leadership is not a follower at any cost we incorporate boundary conditions on certain variables; these represent minimum solutions that are acceptable to the industrial leadership. Data organisation and parameter estimates of the model were calculated from various sources and made consistent for initial year 1965. The distinction between small (peasant) and large farmers (rich) was made along the lines of CIDA (1966). This resulted in a population distribution in the ratio of 92 per cent to 8 per cent, as against a land distribution ratio of 4 per cent to 96 per cent, which is extremely skew worldwide. Distribution of agricultural output is less skewed due to higher productivity of peasant farms: π10 > π20. A determinate solution of the model requires fixation of target variables NQ1 and F1. Although problematic, we took as representative of NQ1 the actual transfer of land to peasants that occurred in Chile between 1965 and 1972, as being an indication of the will of the peasantry. This is a period in which peasants had recourse to farm seizure and thereby expedited, and perhaps determined, the confiscation of land at a tempo that was quicker than that anticipated by the state. The land transfer from rich to poor is targeted at 0.09 million hectares per year, which forms only a meagre 0.003 of land held by landlords. For the approximate fixation of the future values of peasantry financial debt F1 we have assumed that the peasantry would feel satisfied with elimination of the debt within half a generation. This is a period that present peasant leaders would outlive. This assumption would reduce debts by 25 million escudos per year, noting that total debts amounted to 365 million escudos in 1966. The model was solved for 1966 and 20 consecutive years to give basic run projections (BRP). The projected annual growth of the GDP over 20 years, 4.5 per cent, is higher than the actual growth rate reached in Chile: 4.1 per cent for 1965–70, and 2.0 per cent for 1970–75. Although the model overestimates growth prospects, the projections do reproduce the structural relationships that determined Chilean development. The projections show that in spite of targeted transfer of land and the reduction of the liabilities, the income share of peasants deteriorates during the 20 years. There is an improvement in the income share of rich farmers and a very substantial increase in the rich farmers’ wealth due to an increasing price for their land and increased financial assets, due to the sale of some land, and to accumulated savings from higher income. The projections indicate a development impasse. It is important to analyse the implications of these projections for the industrial leadership, who may have the last word. The assumed target for land transfer

60  Socio-political regime models leads to a high accumulation of financial assets by landlords F2. Peasants too are targeted to cut their financial liabilities, that is, to increase their financial assets F1. As a result, the non-agricultural sector, including the public sector, becomes increasingly in debt to rich farmers. This financial debt is projected to reach 40 per cent of non-agricultural income after 20 years. These are outstanding debts that the non-agricultural sector has to repay. In other words, the incidence of the targeted NQ1 and F1 would fall on the non-agricultural earners. In addition, interest payments on this debt cut into the income of non-agricultural earners. The targeted NQ1 and F1 can be shown to affect the interests of the industrial sector unfavourably in other ways too. Prosperity in the industrial sector requires a strongly growing agricultural output with an export surplus that can be used, among other things, to import investment goods for industrial growth. The projections indicate that, in spite of the higher agricultural output, the demand for agricultural consumption goods over exceeds its output with the result that net food imports (exports less imports) would rise from a value of E(1 + 2) = −2,541 in the initial year to −3,912 in year 10 and −7,918 in year 20, thereby using up both the net industrial trade balance, E3, and the foreign capital inflow, FCF. The conclusion that can be drawn at this stage is that although in the BRP the welfare of peasants rises, this is small when compared to the gain of rich farmers, and more important, they have an unfavourable effect on the relative position of the otherwise politically very strong industrial sector, as shown above. This would seem to be sufficient reason why industrial leadership would not allow these specific projections to materialise. If the industrial leadership can be assumed to set up boundary conditions on such variables of interest to them as (F1 + F), Y3 , and E(1 + 2) , the obtained projections would have most likely violated these boundaries. Results signal an accentuating impasse. No doubt the significant changes in socio-political power which occurred in Chile in 1973 and thereafter are just too complex to be clarified or predicted by means of an exploratory model simulating the Frei and Allende regimes, but this much is evident: the simulations suggest that the stipulated land reforms would have implications that are politically unacceptable by the industrial leadership, and would sooner rather than later be cut short, obstructed, or modified. Given the above impasse, the object of the subsequent alternative policy simulations that we ran was to propose, adapt, and study in several steps more moderate strategies of agrarian reform that have a greater chance of achieving a future redistribution of wealth and income from rich to peasant farmers, without undermining industrial interests. With the complexity of Chilean politics at the time, finding ways and means of securing this delicate balance appeared to be the central development issue. We report here on five policy simulations that were run, among others. Policy A is a less ambitious land redistribution and debt acquittal along the 1967 Agrarian Reform Law (ARL). By comparison, the target NQ1 in year 20 in ARL = 0.6 (BRP), while the target annual fall in F1 in ARL = 0.3 BRP. Policy B adds to policy A credit advances, thus reducing ι1, ι12, ι2, ι32. Policy C adds to policy B public investment to increase farm productivity, thus raising π1, π2, κ1, κ1.

Socio-political regime models  61 Policy D adds to policy C a harmonised tax at a higher level: τ1 = τ2. Actual situation discriminated against poor farmers, and favoured rich farmers τ1 > τ2 Policy E adds to policy D higher agricultural wages, via raising ω21. Results in Table 3.4 show a persisting stalemate under policy simulations A (moderate land reform) and B (credit advances): failing output and rising inequality in agriculture, and a high financial liability of non-agriculture to agriculture. This indicates a helpless role for national policy in producing a coherent development pattern that satisfies the expectations of all actors. Among all policy runs, the combination of C (public investment to raise agricultural productivity) with E (higher agricultural wage rates) is the most promising policy package. Policy D (tax policy) does not improve on C or D. The combination of higher productivity and higher wages reinforce each other. The proposed package gives the highest levels obtained for income of poor peasants, as their share in GDP doubles compared to BRP. Agricultural output, which is a very strategic variable in the Chilean context, is also highest under this simulated package. Results show that non-agricultural income appears not to be very sensitive to alternative strategies. On the other hand, the findings show that the financial liabilities of the non-agricultural sector—another strategic variable in the Chilean context— reach their lowest levels under this simulated package. The combination of higher agricultural productivity and higher wages for poor farmers appears to be the most favourable scenario for solving the Chilean agrarian development impasse. Besides, this scenario may need the least effort to achieve in view of long-term free labour market forces which tend to raise wage rates with rises in labour productivity.

Table 3.4 Initial year t = 0, basic run projections (BRP), and policy simulations for year t = 20, Chile t=0

BRP t = 20 Policy A Policy B Policy C Policy D Policy E

Y1

  783  2231

 1537

 1647

 3378

 3367

 5234

Y2

  915  2716

 2801

 2850

 3774

 3287

 1028

Y3

16301 38187

38352

37715

36633

 36391

36848

X(1 + 2)

 2416  4637

 4190

 4515

 8247

 8165

 8145

X3

30981 75784

75363

74185

71496

71203

70444

E(1 + 2)

−2541 −7198

−7416

−7283

−3947

−3934

−4258

 2707  4564

 4782

 4649

 1313

 1300

 1625

Financial F1 + F2   235  8177 liability non-agr.

 7676

 8328

 6165

 4638

 2547

17999 43134 ΣhY h Yt /ΣhY h 0.044 0.052

42690 0.036

42212 0.039

43785 0.077

43585 0.077

43110 0.121

Income poor farmers Income rich farmers Income nonagriculture Output agriculture Output nonagriculture Exports agr.

Exports non-agr. E3

GDP Share poor in GDP

62  Socio-political regime models

6  Concluding remarks Incorporation of the socio-political structure in economic models introduces more realism in evaluating policy making. The approach followed here to incorporate and specify the socio-polity is realistic, simple, and flexible. Six remarks can be added as ways of elaborating on the approach. First, division of the total population into groups with common interests and incorporating their relative strengths brings more realism to policy making. The relative strengths of interest groups determine who leads and who follows, and subsequently, the socio-political profile of the nation. Different socio-political profiles of power distribution represent different model closures, yield different model structures, and show different solution sets over time. When predicted performances are compared with observed ones in a particular economy it is possible to reveal the socio-political structure of that economy. In this respect, it is interesting to quote a supporting statement made later in Taylor (1990), Chapter 2: ‘Setting (model) closure is impossible unless class structures and economic power relationships have already been defined’. Second, the procedure for constructing first an under determinate model consisting of economic relations with an n number of degrees of freedom; and second, making the model determinate by fixing n unknowns belonging to the leader is a handy simplification of a complex problem. Third, translation of the objectives and desires of a leader group into numerical targets is problematic. Lacking substantive quantitative studies of this kind, we were inclined to use some approximate procedures for India and Chile. The setting of precise targets is nowhere required, however. In fact, it can be even useful to inflate the targets of the leader group and study the effectiveness of measures under extreme conditions. There is perhaps another reason why the targets of the leader groups can be better when overrated: for the actual implementation of most government measures, the cooperation of the leader group is necessary, that is, in the Indian context; and, to assure this cooperation, a premium on the targets of the leader group may prove to be necessary. Fourth, the impact of one and the same government measure may vary under different socio-political leaderships. However, if the impact of certain measures is found to be constant (or ranks consistently high in terms of effectiveness), irrespective of the socio-political system in which the measures function, this would indicate that such measures may form a policy core in all situations. The approach can then be employed as a simulation to find the more basic policy measures. In such a case, difficulties that may be encountered in identifying the real leaders, or in fixing the predetermined values for their welfare variables, become secondary. Fifth, the approach permits testing the feasibility of reformist policies in controversial regimes. If, in spite of functioning within the boundaries set by the leader group(s), reformist policies are still not feasible, this can be considered as an indication that there is something wrong with the foundation, in which case a deeper change in the foundation is necessary; this was shown for Chile.

Socio-political regime models  63 Sixth, the particular model closure demonstrated here can be called a fixed closure, as it represents a situation where the dominant actor group fixes its target(s). It is possible to adapt the model with flexible closures. For instance, each actor group can be modelled to maximise its preferred targets within a bounded range, with greater weights for the more dominant actors. The government can be modelled as the arbiter in reaching a bargained solution. The government may then maximise a common aim whose benefits go to all groups—national product—subject to technical and behavioural constraints of the model and the bargaining constraints. In closing, it can be stated that few governments (and political parties) are likely to sponsor models that explicitly incorporate power structures. The utility of the type of models presented here lies more in their exploratory signalling of reform deadlocks and the exploratory search for feasible policies.

4 Social economic development goals in economy-wide policy models An application to Korea

1 Background In the early decades of development economics, mainly in the fifties and sixties, economic development was analysed and planned in terms of economic growth. The central development objective was the growth of GDP, and sometimes GDP per capita. Empirical data on the actual performances in most developing countries gave a different picture, showing increasing income inequalities, more population groups falling behind poverty lines, increasing underemployment and unemployment, and deteriorating food and other living conditions. The new picture brought a revision in interests and insights by economists and planners dealing with development problems. One of the reactions that emerged was the social economic development perspective, which saw the misplaced focus on GDP as the fault in the planning framework, and they went for reformulating development models and plans in terms of broader social economic development goals. In spite of the accelerated interest in a social economic development perspective there were no attempts made to put up an operational framework that would unify new goals in the development planning models that were commonly used at the time. There was a flood of definitions and concepts on basic needs, social indicators, social aims, and so on, but most of them were dispersed and not accessible in economic models. Our contribution in this area was to broaden the scope of policy modelling and planning practices towards the social economic development perspective. This chapter formulates and applies a social economic development model that supplements GDP growth with income and employment variables belonging to social groups, and indictors on attainment of wellbeing, and it extends the scope of policy instruments and other relationships accordingly. Primarily for reasons of data availability, the Republic of Korea was selected as a test case. The model was estimated on the basis of data for the sixties and was employed to simulate development up to the early eighties. Interestingly, the empirical applications around 1970 showed positive and coherent future performances for Korea on its GDP, distribution, employment, and basic needs attainment for the seventies and early eighties; these outcomes were also confirmed by the realised development performance of Korea years later. The chapter is organised as follows. Section 2 is devoted to laying out an operational framework for specifying an integrated system of social economic

Social economic development models  65 development goals. Section 3 specifies the model. Section 4 displays some detail on estimation of the model, the projected results, and it examines multiplier properties of the model. Section 5 shifts the model from its analytical form to a planning form. The analytical form of the model treats aim variables as unknown and instruments as given. The complete planning form is usually described as the reverse, with aim variables taking the form of fixed targets and instrument variables becoming unknown. The shift from analysis to planning raises questions on decomposition of the model for planning purposes, and the structural properties of the outcome. The section deals with these questions, and demonstrates its application via a few simulations, and adds some elaborations. Section 6 ends with concluding remarks.

2  A unifying approach towards social economic development goals In identifying and selecting goal variables in the context of social economic development policy it is essential to distinguish between social groups of common interest, followed by listing the goal variables for each group separately, which can be of a monetary nature (that is, income) or of a non-monetary nature (that is wellbeing). The literature on wellbeing indicators is long and overlapping. The United Nations Research Institute for Social Development (UNRISD) was one of the starters.1 It implemented in 1966–8 a research programme on identifying and quantifying indicators of basic needs relevant for developing countries. Soon after, various agencies of the United Nations came up with their own basic needs indicators. Of course, as is well known, the interest in quantifying wellbeing indicators has a long history. One of the first articles to quantify basic needs indicators at the international level was by Bennett (1948). The concepts and measurements of wellbeing have undergone substantive refinements since then in terms of depth and width, with a most comprehensive system in Sen (1982). In between Bennett and Sen, there have been tens of contributions on the topic, and the contributions of UNRISD and other UN agencies are but a few of them. In recent years, tens of foundations worldwide have developed various quality of life and wellbeing indicators. The integration of any set of these social indicators in economic models is a different story. We shall review the various steps required and comment on how we approached them. The first step is to incorporate some division of the total population into social groups with more or less homogeneous interests, and whose wellbeing levels can be traced in the available statistics. The outcome is subordinate to available data on consumption, expenditure, employment, and labour supply on such social groups. In most developing countries, data do not go beyond a disaggregation into household groups headed by wage earners, salary earners, and the large group of employers, the self-employed and family workers2; with possible subdivisions into rural and urban areas. The data for Korea allow for the distinction between these three social groups, the term household groups is also used interchangeably; they will be denoted by index h; h = 1, 2, 3, respectively.

66  Social economic development models The next step is to specify and (or) select the goal variables and wellbeing indicators thought to be of relevance. On this issue we were guided by two survey reports at the time: United Nations (1966) and United Nations (1971). These were particularly helpful in justifying in quasi-objective ways the selection; otherwise, it would have been a subjective choice. The surveys that were conducted among planning bureaus of 48 developing countries listed income, employment, and four basic needs: food, housing, health, and education, next to macroeconomic monitoring indicators. To start with, the income goal with the highest score in the surveys was GDP growth, which we shall denote by WG. Our orientation would focus as much on the disposable income per capita received by each household group, denoted by WY h where h = 1, 2, 3. It is obvious, as WG can be derived from the weighted sum of the growth WY h, and population growth, any predetermined double fixation of WG and WY h has distributive implications. If planning is done in terms of the WY h’s, then the aggregate WG follows. The goal with the next highest score in the survey of planning bureaus was a minimal unemployment rate; this can be expressed in the form of employment rates that relate employment to labour supply, to be denoted by WMh where h = 1, 2. Since the unemployed are those who are seeking work for pay, that is, potential hired employees, this aim variable would by definition apply to the hired workforce only, which are only the groups of wage earners and salary earners; it excludes unemployment status for the group of employers and the self-employed. This group may suffer from underemployment, but that should show up in a low income, which is taken care of WY h. The income and employment goals in the UN surveys were directly followed by satisfaction of basic needs with respect to nutrition, shelter, health, and education. As the focus is on the attainment of such basic needs, it can be justifiably assumed that for all households above the poverty line the satisfaction of these basic needs is not a problem. Basic needs attainment variables are thus especially relevant for population groups with the lowest incomes. If the highest concentration of nonfulfilment of basic needs can be assumed to occur among wage earners,3 then it follows that such a basic needs variable as nutritional intake should be solely of relevance in reference to the lowest income group, that is, wage earners. For example, the model will specify an aim variable of average daily intake of calories per capita for group h = 1, denoted by WC1, and not for other groups. Similarly, dwelling levels—expressed as (higher) average numbers of rooms per capita—are relevant as an aim variable only if they refer to group 1, for example, WD1. For education we take the primary enrolment ratio WN. Any increase in this ratio refers to the eligible school-going population for the country as a whole, and would tend to benefit primarily the children of the lowest income group. With respect to health needs, we shall work with the national survival rate as an aim variable, WH, partly because available data do not make possible the division of health benefits by social groups, but mainly because the characteristic external effects which accompany health levels in the whole nation make it more desirable to formulate this aim for the whole nation.4 We shall also make use of a composite index of basic needs attainment consisting of the four indicators to be denoted by BNA1, which would apply for the lowest income group. Table 4.1 contains a summary.

Social economic development models  67 Table 4.1 List of aim variables Wage earners household group, h = 1

Salary earners, h=2

Employers etc., h=3

Whole nation

Disp. Inc. p.c. Disp. Inc. p.c. economic growth Disposable income per WY2 WY3 WG capita WY1 Employment rate Employment rate WM1 WM2 Calorie intake per capita WC1 Dwelling space per capita WD1 Health level for lowest income household group, approximated by survival rate WH Educational level for lowest income household group, approximated by primary school enrolment rate WN Composite index, basic needs attainment, BNA1, consisting of WC1 , WD1 , WH and WN; specified in eq. 30

The next step is obvious. Introduction of these aim variables requires the incorporation of suitable instrument variables that could enhance their values. The model contains a large number of policy instruments controllable by government. The analysis will concentrate on budget items: these are various tax revenue items and government spending in the form of current allocations related to the distinguished aim variables, among others; income transfers to social groups; and public investment by sector of activity j, as indicated in Table 4.2. The next steps are to deal with such questions as how the values of these aim variables are generated and how to incorporate and measure their productivity effects in an economy-wide model. On the formation of aim variables, the levels Table 4.2 List of instrument variables Wage earners, h=1

Salary earners, Employers h=2 etc., h = 3

Direct tax rate τ1g Income transfers Tg1 Food subsidies Cg1 Rent subsidies Cg2

Direct tax rate τ2g Income transfers Tg2

Whole nation

Direct tax Indirect tax rates by sector j, τjg rate τ3g Foreign trade taxes, τjgm , τjge Income transfers Tg3 Current expenditure on health, Cg3 Current expenditure on education, Cg4 Current expenditure on economic services, Cg5 Current expenditure on other social services, Cg6 Public investment by sector j, Igj

68  Social economic development models of income among the social groups determine how much is consumed privately of each good or service. This private consumption, together with public allocations, determines final demand by sector, and in turn production by sector, and demand for and supply of labour by skill type and social group. As a result, labour productivity rates and employment rates are obtained. These two rates play dominant roles in the formation of the factor earnings accruing to each social group. This completes the circular flow. Regarding any productivity effects of better living conditions, the model attempts to integrate what various studies have already established, namely, that higher levels of WC1, WD1, WH, and WN, up to a certain degree, increase the capacity of labour. The productivity effect of higher wellbeing is long established in economic thinking.5 There are many empirical studies for different countries that have tested and confirmed the productivity effect. Studies on the effects of higher nutrition show that the higher energy produced allows a labourer to work near his full capacity. The housing effect is supported by empirical evidence on the positive influence on labour productivity from rehousing projects. Higher health levels save on work time lost due to morbidity, and while the quality of work improves, life time and economic activity increase. Furthermore, the productivity effects of more education are very well established. The combined effects of the above factors are usually considered to be more crucial at low income levels, but may not hold at higher incomes. How do we incorporate the productivity effect of higher wellbeing in an economy-wide policy model? This is done in three rounds. First, in a Cobb-Douglas production function at the sector level with factors of production capital and labour, the productivity effect will be more applicable to low skill and not high skill. Second, assume for the benchmark year that employed labour of low skill, indexed as 1, in activity j, thus LD1j, is functioning at below its full working capability due to shortfalls in wellbeing attainment from some desired level; denote wellbeing attainment for this labour type by the general index introduced earlier BNA1, the highest attainment level is at BNA1* = 1.0, and thus BNA1 < BNA1*. Define full capability labour in year t as labour multiplied by the basic needs attainment index, (BNA1 . LD1j ). Because for the benchmark the term (BNA1 . LD1j ) is less than LD1j , the calibration coefficient for the benchmark will be raised. Third, as the wellbeing attainment rises in later years to its maximum level, the impacted labour recovers its full capacity, and contributes to a larger production level. Note the productivity effect of higher wellbeing which is conditional on which labour is impacted, and on what is included in the wellbeing indicators. Of course, once BNA1* = 1.0 is reached, new concepts of what constitutes the next stage of wellbeing can be formulated and incorporated for a subsequent round of enhancing functioning capabilities.

Social economic development models  69

3  Specification of the model 3.1 Introduction The model that we use belongs to the category of combined econometric multisector (CEM) models that were discussed in Chapter 2, and it shares their points of strength and shortcomings. The potential use of combined models in development planning has been suggested by a number of authors.6 Published works on combined models remain very much in their infancy, however.7 Among the benefits are flexibilities in extending and reformulating the model in various forms to suit policy making, and econometric verification of the formulated relationships. The main shortcoming is that most of the adjustments in the balancing of factor and product markets take place via quantity changes, under unchanging relative prices. The original model contained many sectors. To increase focus we shall consider only two sectors producing goods (these are agriculture and extended industry), and two sectors producing services (these are commercial, and social services). Because of the inclusion of income and employment formation in the model, it follows that the model should incorporate the demand for and supply of manpower by skill types.8 Particularly for the specification of manpower supply by skill types, we had to elaborate on the educational system. In contradistinction to the demand for and supply of manpower by skill types, there are the demand for and supply of labour by wage and salary earners’ households groups. The classifications by skill level and earning type are closely related. As was mentioned above, the model generates the ten aim variables in Table 4.1, which are specified to be partly influenced by the instrument variables in Table 4.2. Regarding the productivity effects of higher wellbeing, the model integrates what various studies have established, namely, that basic needs attainment, BNA1, increases labour quality and labour productivity; and this in turn, can contribute to enhancing economic growth. The model does not consider BNA1 effects on demographic transitions and economic growth, which are known to be substantial. One central question in development planning at the time was how to proceed with policy making when there are so many aim variables to be targeted and instrument variables to be solved. The model reverts to the ordering structures of Simon (1953) to answer the question. Practically speaking, the analytical model is a simultaneous system which has to be solved in one shot. Conversion of the analytical model into a planning model simplifies the system substantially. It turns out that the prefixing of income variables by social groups is the most logical first move in the policy-making process; and when accompanied by basic needs and employment targets, the planning model would presume a significantly simplified diagonal structure that can readily be split into stages of policymaking. The notations followed, in Table 4.3, are in conformity with other chapters. Unless otherwise stated, all variables are measured at the end of a period of six years. The model solves for three six-year periods: 1968, 1974, and 1980. Monetary variables are expressed in constant prices of 1965, in billion won (thousand million). Demographic variables are in 1000 persons.9

Table 4.3 Notations Indices: j = sectors of activity (whereby 1 = agriculture, 2 = industry, construction, transport combined; 3 = commercial services, 4 = social services). c = categories of consumption (1 = food, 2 = housing, 3 = health, 4 = education 5 = other commodities, 6 = other services). e = educational levels ( i = primary, ii = secondary, iii = higher); q = labour skills (1 = low, 2 = high); h = social groups defined by head of household (1 = wage earners, 2 = salary earners, 3 = employers and self-employed). Subscript g is to identify government, and r for the rest of world. Endogenous variables Cj Total private and government consumption expenditure on activity j Chc Consumption expenditure by household group h on commodity c, h = 1, 2, 3; c = 1,…, 6 Ej Exports of sector j Foreign capital flow expressed in foreign currency FCF Government budget deficit GBD Total installed investment I Private demand for installed investment Ip Inventory change in sector j . j = 1, 2; agriculture and industry etc. ICHj Capital utilised in sector j . j = 1, 2; agriculture and industry etc. KDj KSj Capital supply in sector j . j = 1, 2; agriculture and industry etc. Labour demand for skill q in sector j; q = 1, 2; j = 1,…, 4 LDqj LDh Employed number of workers belonging to social group h; h = 1, 2 Mj Competitive imports of sector j Enrolment total in educational level e NTe NEe Enrolment of new school entrants in first year in educational level e LRh Remuneration rate of wage earners and salary earners, h = 1, 2 LSq , LSh Labour supply by skill q and by household group h, h = 1, 2 Vj Gross product of activity j, that is, value added of activity j Gross output of activity j Xj Yh Disposable income of household group h, h = 1, 2, 3 Zh Factor and other income receipts of social group h, h = 1, 2 Factor and other income receipts of social group 3 Z3 Government revenue Zg The following endogenous variables are aim variables GDP growth rate WG WY h Income per capita specified by social group, h = 1, 2, 3 WMh Employment rates of wage and salary earners, h = 1, 2 Food and nutrition aim variable is daily calorie intake per capita for the WC1 lowest income group, that is, wage earners and related dependents Housing aim variable is dwelling space (rooms) per capita for the lowest WD1 income population group, that is, wage earners and related dependents Health aim variable, that is, survival rate applying for the whole population WH WN BNA1

Education aim variable is enrolment ratio for primary education Basic needs attainment index for (low-skilled) wage earners, maximum = 1.0

Exogenous variables Government consumption expenditure on good c Cgc Foreign exchange rate FXR Igj Government investment expenditure in sector j Total population P Transfers from government to institutions Tgh Transfers from rest of world to household group h and to government g Trh , Trg Coefficients αjj′ Input-output delivery coefficients from sector j to sector j′ Calibration coefficient in Cobb-Douglas production functions for the αjo sectors j, j = 1 (agriculture), j = 2 (industry etc.) αrj Non-competitive imports share in the output of sector j Labour elasticities of production by skill and sector. Capital elasticity = βqj 1 − Σj βqj Intercept and slope coefficient in a regressed costing function relating χeo , χe enrolments to spending by educational level: e = primary, secondary, higher χte Expenditure share of educational level e in the total gross spending on education δj Depreciation rate of capital in sector j . j = 1, 2 Export growth rate for sector j . j = 1, 2, 3, 4 εj ϕ1,1,o , ϕ1,1 Intercept and slope coefficient in a regressed costing function of total expenditure on food per capita per unit of nutritional intake, made dependent on income per capita. Equation applies for wage earners, first subscript 1; and for expenditure on food, second subscript 1 ϕ1,2,o , ϕ1, 2 Intercept and slope coefficient in a regressed costing function of expenditure on housing per capita per unit of housing space, made dependent on income per capita. Equation applies for wage earners, first subscript 1; and for expenditure on housing, second subscript 2 Intercept and slope coefficient in a regressed costing function of health ϕ3,o , ϕ3 expenditure per capita per unit of survival rate, taken as dependent on income per capita. Equation applies for the whole nation, and for expenditure on health, denoted by subscript 3 Intercept and slope coefficient in regressed household consumption γhco , γhc functions relating consumption on commodity c to disposable income of social group h; c = 1,…, 6; h = 1, 2 , 3 Intercept and slope coefficient in a regressed labour supply function of ηqho ,ηqh wage earners and salary earners relating to skill q; h = 1, 2; q = 1, 2 Intercept and slope coefficients in a regressed private investment function ιo, ι, ιʹ, relating private investment demand to GDP, profit rate, public investment, ιʹʹ, ιʹʹʹ and other exogenous variables, respectively ιj Investment delivery share by activity j in the total delivery of installed investment. Sector of industry etc. ( j = 2) delivers all investment, thus: ι2 = 1. All other ιj = 0 φjc, φgjc Distribution coefficients of private consumption expenditure by good c to sector j, and same for government consumption expenditure. C = 1,…, 6; j = 1,…, 4 κpj Installed investment destined for use in sector j as a proportion of total installed investment. Σj = κoj continued overleaf

72  Social economic development models κj λeo, λe, λʹe λqo , λeq , λʹeq μ jc , μ ji νqjo , νqj π1 , πe ρho, ρh, ρʹh ρg σh τhg, τjg τjgm , τjge ϖj

Incremental capital output ratio for sector j. See endnote 10, on relation of κpj to κj Survival rate of enrolled students from previous period; and two flow operators for the newly entering students within the educational system approximated in terms of current period t and previous period six years earlier, t − 6. Survival rate of labour force from previous period; and two flow operators for students passing through and exiting from the educational system, and eventually joining the labour force; approximated in terms of current period t and previous period six years earlier, t − 6. Competitive imports shares in consumption and in investment by sector j. Because sector 2 is the only one delivering investment goods, it follows that μφi is zero for sectors 1, 3, 4 Intercept and slope coefficient in an economically defined labour requirements function giving number of workers of skill q required for producing gross value added by sector j Population belonging to wage earners in the total population, and proportion of the age group eligible to primary education in the total population Intercept and slope coefficients in a regressed remuneration rate function that relates average labour earnings to productivity and unemployment rate; h = 1, 2 Factor income share in the GDP due to government Savings rate of household group h Rates of direct taxes on income of household group h, and indirect taxes on gross output of sector j Rates of import tariff and export duty collected by government Utilisation rate of capital in sector j . j = 1, 2

3.2  Specification of the model The first block, in Box 1, of the model relates to factors of production and the factor market. Eq. 1 specifies Cobb-Douglas production functions by sector j (for agriculture and for industry) with three factors of production: capital KDj, high-skilled labour LD2j, and low-skilled labour measured in fully productive labour units, where LD1j is the number of demanded (employed) low-skilled workers, and BNA1 is the basic needs attainment level of low-skilled workers. When BNA1 is at 100 per cent, the worker performs at his full productivity level. The term (LD1j . BNA1) attempts to measure the workforce in fully productive units. A workforce LD1 of 5 million with BNA1 at a level of 90 per cent is equivalent to 4.5 million workers with a 100 per cent productive capability. Higher levels of nutrition, housing, health, and general education lead to a higher BNA1 and a higher labour quality, as will be formulated in eq. 30, and has the same effect on production as increased labour inputs. Consequently, substitution possibilities can occur between labour quantity and quality.10 In this equation, βqj are labour elasticities of production, 1 − Σqβqj is the capital elasticity, and αjo is a calibration coefficient.

Social economic development models  73 Eq. 2 incorporates the second level of production technology typical of economy wide models that is, a Leontief input-output system. Gross value added by sector, Vj , is derived as a proportion of the gross output by sector, Kj , after deduction of domestic and imported intermediate deliveries, and indirect taxes. The equation holds for all sectors. Eq. 3 states that the utilised capital in the non-service sector j is equal to the installed supply of capital. The utilisation rate ϖj is given data. Eq. 4 specifies the supply of capital in the non-services sector j as that of the previous year less the depreciation at rate δj, plus the share of the destination sector in newly installed private investment, plus government investment destined to the sector. The destination shares of private investment, κpj , when summed over all sectors add to 1. The values of these shares take account of the relative requirements for investment goods among the destination sectors; hence, they may be periodically updated.11 Eq. 5 gives the demand for manpower skills q by sector j. The quantity demanded would depend on technical rates of manpower inputs per unit produced, and the sector product X. The closest approximations available to estimate these functions are the input rates implicit in development plans (the Korean Second Five Year Plan). We assumed the technical requirements to be a linear function of the plan requirements with values of νqjo and νqj to be found from regressions on plan data. To account for the fact that the potential effects of higher labour quality have still to be realised, the potential is multiplied by realisation coefficients which are calibrated by regressions on observed data. Eq. 6 specifies the supply of manpower skills, as depending partly on survival and retirement from the previous stock, and partly on various flows of students who leave the school system and enter the labour force. These flows are the result of combined rates of student transitions, graduation, and labour participation applied to new enrolment of students to various educational levels in the past. These yearly flows are summed and approximated in terms of two periods, t and t − 6. The final form of this equation falls in three terms.12 Eq. 7 specifies the number of new enrolments as the number of total enrolment less the outflow of students leaving the educational system. This outflow is the result of past new enrolment less mortality, dropping out, and graduation. The same approximations in the previous equation are followed here, so that the specification of the equation is done in terms of periods t and t − 6, and falls also in three terms, (see previous note). Eq. 8 makes total enrolments by educational level dependent on total spending on education, which is consumption category 4, by households and by government, Σh Ch4 + Cg4. The spending is distributed on the three educational levels via proportions χte. The obtained financial resources available to a specific educational level undergo a regressed costing function of costs per student to determine total enrolment at each level. Eqs. 6, 7, and 8 are very closely linked. In eq. 8 spending determined total enrolment places at each educational level, NTe. Deducting from these the present stock of enrolled students and school leavers gives the new enrolments NEe in eq. 7. When these go through the educational system and after several years graduate, many of them will participate in the labour force, hence eq. 6.

74  Social economic development models Eq. 9 converts the demand for and supply of manpower by skill types into demand and supply by social groups, LDh and LSh , respectively. The following conventions facilitate the conversion. Commonly, unemployment data are collected from compulsory insurance schemes, trade unions, or employment offices. In this context, persons in unemployment are defined as workers whose employment contracts have been terminated, temporarily suspended, or are to be started for the first time, and who are seeking paid work at the prevailing market wage rates. As such, the state of involuntary unemployment, often defined as simply unemployment, occurs exclusively among employees and potential employees. Because of a lack of appropriate data at the time, it was assumed that unemployment among wage earners would correspond to unemployment among the lower skills, and similarly for the case of salary earners and the higher skills. Given our previously mentioned classification of occupations into low and high skills, these correspond very closely with the wage and salary modes of payment. Eq. 10 counts on the above considerations and consistently follows the assumption that the bulk of the supply of manpower with lower skills can be classified as wage earners, while that of higher skills is mainly salary earners. The conversions allow for regresses estimates. Box 1  Production functions and factor markets equations Cobb-Douglas production functions for the sectors of agriculture and industry j = 1, 2 (1) Xj = αjo Π[(BNA1 . LD1j )β1j . LD2jβ2j KDj(1 − β1j − β2j)] Leontief technology-based value added by sector Vj = (1 − Σ j′ α j’ j−α rj − τjg ) Xj

j = 1,..,4 (2)

Capital supply KDj = ϖj KSj KSj = (1 − δj ) KSj , t − 1 + κ pj Ip + Igj Labour demand by skill type LDqj = νqjo + νqj Vj

j = 1,2 (3) j = 1, 2 (4) q = 1, 2, j = 1,..,4 (5)

Labour supply by skill type LSq = (λqo LSq,t − 6 + Σe λeq NEe + Σe λʹeq NEe,t − 6

q = 1, 2 (6)

Enrolments in educational levels NTe = λe0 NTe, t − 6 + λe NEe + λʹe NEe, t − 6

e = i, ii, iii (7)

Financing and costing of total enrolments NTe = χeo + χe [χte (Σh Ch4 + Cg4 )]

e = i, ii, iii (8)

Employment, unemployment, labour force, remuneration rates by earning group q = 1, 2 (9) LSh − LDh = LSq − Σj LDqj Labour supply of wage employees and salary employees LSh = ηqho + ηqh LSq

q = 1, 2 (10)

Remuneration rates of wage employees and salary employees LRh = ρoh + ρh (Σj Vj / LDh ) − ρʹh (1 − WMh )

h = 1, 2 (11)

Social economic development models  75 Eq. 11 gives a regressed explanation for the remuneration rates for wage and salary earners. After testing alternative formulations two main factors appeared to be the most relevant for explaining the variances in labour earnings, LRh. These are the labour productivity rate, and the gap between supply and demand forces of the labour market. The unemployment rate (that is 1 − employment rate WMh ) reflects these labour market forces. The next block of equations, in Box 2, reviews private and government income, and their demand for consumption goods, investment goods, exports, and imports. Eq. 12 specifies income received by household groups, Zh , headed by wage earners and salary earners, h = 1, 2. Additional sources of income are transfers from government and rest of world. Eq. 13 specifies income received by the group of self-employed and employers, h = 3, as the total GDP less factor incomes received by wage earners, salary earners, and government. Private profit is thus defined as the residual: the value added less other factor payments. An additional source of income to the group is foreign transfers. Eq. 14 specifies disposable incomes by social group, Y h , as incomes received less taxes paid. Eq. 15 specifies consumption of commodity c by social group h, Chc , within consumption functions that are dependent on disposable income. Eq. 16 specifies government revenue Zg as consisting of factor incomes, direct and indirect taxes, and import and export duties. Eq. 17 equalises government current expenditure on consumption and transfers plus government investment expenditure (on the right-hand side) to government revenue and government budget deficit, GBD, (on the left-hand side). Eq. 18 gives the private and government demand for consumption goods converted to the sector level. Commodities are distributed on sectors via converter coefficients. Eq. 19 formulates the private demand for installed investment goods. There are a few examples of multisector models which include a function for private investment demand.13 In general, more reality is built into the model when a behavioural equation for investment demand is incorporated. In short-term econometric models it is common practice. Empirically, private investment is found to be explained by expectations on economic growth, profit rates, and interest rates. Additional explaining variables can be enumerated such as complementary versus competitive effects of government investment, liquid reserves, and foreign commitments to invest domestically. The scope of the model allows for incorporating a few effects, but not all. To account for economic growth, lagged variables of GDP were tested, giving current GDP as the best fit. One candidate for an explaining variable is profit income divided by employed capital, meant to approximate the profit rate. Another variable is public investment, Σj Igj, which can be expected to show a positive complementary effect in the development context.14 Eq. 20 adds private investment demand to government investment to give total installed investment. Eq. 21 specifies, in a simple way, exports by sectors along lines of Chapter 2. Eq. 22 specifies competitive imports as proportions of consumption and investment, following Chapter 2.

76  Social economic development models Box 2  Household income, consumption, investment, government accounts Receipts of household groups of wage and salary earners Zh = LRh . LDh + Tgh + Trh Income of the group of employers and self-employed Z3 = Σj Vj − Z1 − Z2 − ρ g Σj Vj + Tr3

(13)

Disposable incomes by social group Y h = (1 − τhg ) Zh Private consumption by social group h of commodity c Chc = γhco + γ hc (1 − σh ) Y h

h = 1, 2 (12)

h = 1, 2, 3 (14) h = 1, 2, 3 c = 1,..,6 (15)

Current revenue of government Zg = ρg Σj Vj + Σh τhg Zh + Στjg Xj + Σj τjgm Mj + Σjτjge Ej + Trg

(16)

Government expenditure and budget deficit Σc Cgc + Σh Tgh + Σj Igj = Zg + GBD

(17)

Consumption demand converted to sector level j Cj = Σc φ jc Σh Chc + φ gjc Cgc

j = 1,..,4 (18)

Private investment demand Ip = [ιo +ι Σj Vj,t + ιʹ [Z3/(KD1+KD2 )] +ι′ʹΣj Igj + ι''' (unspecified exogenous)] (19) Total investment demand I = Σj Ipj + Σj Igj

(20)

Export demand Ej = E j,t − 1 (1 + εj )

j=1,..,4 (21)

Competitive imports Mj = μ jc Σj Cj + μ ji I

j=1,..,4 (22)

The third block, in Box 3, lays down conventional aim variables, and what has become known as basic needs aim variables and their formation. Together, they are eight sets of variables, denoted by the letter W and an additional letter indicative of the aim. To start with, the first is the GDP growth rate, denoted by WG, in eq. 23. Next are the per capita disposable income by social group h, denoted by WY h in eq. 24. Next are the employment rates for wage earners, and for salary earners are denoted by WMh in eq. 25. Then there are the basic needs aim variables for the lowest income group that relate to the satisfaction of food and nutritional calorie intake, WC1, dwelling needs, WD1, health needs, WH, and education needs, WN. There is also the basic needs attainment index, BNA1, which combines the above. As was stated earlier, the basic needs indicators are relevant in the case of population groups who are not able to satisfy these needs, considering the level of income they receive. In terms of the population groups considered in the model,

Social economic development models  77 this would apply most for the group of wage earners. Hence, the subscript 1 is attached to the nutrition and housing aim variables, WC1 and WD1. In the cases of health and education, it is difficult to identify these indicators at the group level. The average level reached for the whole population is taken to be representative for what is reached at the wage-earner level. Eq. 26 specifies the formation of nutritional levels. The derivation of calories from a certain amount of food expenditure would require knowledge of the average cost per calorie. For example, realised calories per capita = (consumption expenditure on food per capita) / (average cost per calorie); or written otherwise, average cost per calorie = consumption expenditure on food per capita / realised calories per capita. The model is concerned with the nutritional conditions of the lowest income group, h = 1. Their consumption expenditure on food, which consists of a private component, C1,1 , plus public allocations of food directed to this group, Cg,1 . Consumption expenditure on food per capita per annum is, therefore, (C1.1 + Cg.1 ) / π1 P. The realised calorie intake per capita for the group is denoted by WC1. Dividing the first by the second term gives the average annual cost per calorie per day, which can be assumed to be a linear function of the general level of living of the group, represented by the income per capita of the group, thus (Y1 / π1P). Usually the unit cost tends to increase rapidly as the general level of living approaches higher levels, owing to the gradual shift in the pattern of food composition from cheap food articles to more expensive food articles with the same calorie content. As a result, the equation can be written for estimation purposes in the form of eq. 26, in which ϕ1,1.o and ϕ1,1 are the parameters which require estimation by regression. Eq. 27 specifies formation of housing levels similarly to the above formulation. A specification for the supply of rooms per capita for the lowest income group can be written, wherein the numerator is the rent expenditure per capita and the denominator stands for the average rent per room (assumed a function of income per capita). Eq. 28 follows preceding formulations to define the survival rate (which is 1  − mortality rate) for the whole nation in terms of a numerator (the health expenditure per person) and a denominator (the average health costs of avoiding one death). Although the equation fitted with regressed data, its predictions of the mortality rate were at variance with those from established demographic projections; besides, the link of the equation with the model turned out to be empirically weak. Hence, it was decided to reformulate eq. 28 into eq. 28.1, giving WH = λ, as eq. (28.1), where λ is the exogenously given survival rate for the whole nation and thus employ given projections of the mortality rate, which is 1 – λ. This proved to be more functional also, since the model required values of mortality rates for manpower with different skill types and pupils of different ages, and these were simply set at levels proportional to the mortality rate for the whole population. Eq. 29 is the enrolment rate defined as enrolment in primary education as a percentage of the eligible child population. Derivation of enrolment levels was specified previously in eq. 8. Eq. 30 is a basic needs attainment index for the wage earning or low-skilled group, BNA1. The index is close to the Human Development Index developed

78  Social economic development models by the UNDP, and often used in assessing country rankings. The attained values of the four basic needs indicators are divided by their requirement levels, and the outcome is averaged. The specification of the index chooses to exclude substitution possibilities between the four indicators, but otherwise is also justified. The first equation of the model assigns a role for BNA1 in influencing the labour input equivalent of low-skill workers. Labour works at its full capacity when BNA1 attains the value of 100 per cent; that is, basic needs are fully satisfied. To repeat, in demonstration of the argument, a workforce of 5 million with BNA1 at a level of 90 per cent is equivalent to a workforce of 4.5 million with a 100 per cent productive capability. In eq. 30 the following is observed: where WC1 / wc > 1, the value is reduced to 1, similarly for WD1 / wd, WH / ws, and WN / wn.

Box 3  Wellbeing variables (aim variables) GDP growth WG = (Σj V j,t / Σj V j,t − 1 ) −1

(23)

Disposable income per capita WY h = Y h / πh P

h = 1, 2, 3 (24)

Employment rates WMh = LDh / LSh

h = 1, 2 (25)

Food: calorie intake (C 1,1+ C g1 ) / (π1 P . WC1 ) = ϕ1,1,0+ ϕ1, 1 (Y1 / v1 P)

(26)

Housing: dwelling level (C1,2 + Cg,2 ) / (π1 P . WD1 ) = ϕ1,2,0 + ϕ1,2 (Y1 / v1 P)

(27)

Health: Survival rate ( = 1 − mortality rate) (Σh Ch3 + Cg3 ) / (P . WH) = ϕ3,0 + ϕ3 (V / P)

(28)

Education: enrolment rate in primary education WN = NTe / πe P

e = i (primary education) (29)

Basic needs attainment index BNA1 = (¼) (WC1 / wc + WD1 / wd + WH / ws + WN / wn)

(30)

The fourth block, in Box 4, consists of national accounts balances referring to product market balances, savings investment balance, and foreign payments balance. Eq. 31 equalises the supply of goods to their demand. Supply consists of gross output by sector, Xj , and non-competitive imports, Mj . Demand consists of intermediate deliveries, consumption, investment, inventory change, and exports. Two notes need to be added. The only sector that delivers investment goods is the industry sector ( j = 2). Thus, ι2 = 1, and ιj = 0 for other sectors. Inventory changes, ICHj , applies to the non-service sectors. The product-market balances for the services sectors have ICHj = 0.

Social economic development models  79 Eq. 32 equalises total savings to total gross capital formation in what can be called the financial market balance. Total savings consist of private savings, government savings, and foreign savings, that is foreign capital flow converted in national currency, FXR . FCF. Total savings are equal to installed investment and inventory change. With its 32 sets of equations, the model contains the equivalent number of 32 sets of endogenous variables. It is a determinate model. There is the additional equation of the foreign payments balance, but this is automatically obtained when the model of 32 equations is solved and, as such, it stands outside the model. Hence, it is numbered eq. A1. The equation is useful in a check on the consistency of solutions of the model. It was explained in Chapter 2 that when one of the three balances is left out, and the remaining balances are substituted into each other and solved, they will automatically reproduce the balance which was left out. This is also the rule that guides the construction of a social accounting matrix and is common usage in guaranteeing the consistency of the interdependent accounts. So, to maintain the model as a square matrix we choose to leave out the foreign payments balance, but the same results are obtained if instead of skipping the foreign payments balance the modeller skips the investment-savings balance. Deleting one of the balances is interpretable in terms of Walras Law that states that in general equilibrium, the modeller may impose market clearing for n − 1 markets and drop the nth market, this being automatically guaranteed, it would thus be redundant to specify it. In principle, the same rule applies for a state of general equilibrium, irrespective of whether the general equilibrium has come through price adjustors, quantity adjustors or a combination of both. The structure and closure of the model will be reviewed in Figures 4.2 and 4.3 below.

Box 4  Product market balances Sector product balances Xj + Mj = Σ αjj′ Xj′ + Cj+ ιj I + ICHj + Ej

j = 1,.,4 (31)

Financial market balance (savings investment balance) (Σh σh Y h ) + (Zg − Σc Cgc − Σh Tgh ) + FXR . FCF = I + Σj ICHj

(32)

Foreign payments balance Σjαrj Xj + Mj + Σh Trh + Trg = Σj Ej + FXR . FCF

(A1)

4 Application 4.1  Estimates and projections The model contains coefficients whose estimation falls into two categories: technical or institutional coefficients that are directly available from published material; and behavioural coefficients estimated via regressions. The basic data for the estimation of the first category are Korean single-year data, especially the

80  Social economic development models input-output tables of 1963 and 1966, and estimates of Korean capital and inventory in 1968. Production function parameters for agriculture and industry (eq. 1) are from various publications of the Korea Development Institute.15 Initial labour input rates by skill and sector (eq. 5), are derived from the Korean Second Five Year Plan. We applied these rates to the observed value added by sector for the observation period to give the ‘plan requirements’. These were regressed against observed employment to generate calibration coefficients that were used to adjust the initial planning input rates to their actual values.16 Data on the progress of cohorts of pupils at various educational levels, participation in the labour force, and mortality are from various Korean ministerial sources. Estimates of all transition rates λs in eqs. 6 and 7 assure consistency for the observation period.17 Estimation of behavioural parameters was done by applying OLS to observations for 1960–71. The data used are Korean time series of household surveys, labour surveys, and national accounts. All estimates selected are econometrically viable and are in conformity with expectations based on economic theory. A few remarks follow on the obtained estimates. Eq. 8, which specifies costing coefficients, fitted better for primary and secondary education than for higher education. In terms of relative current costs per student, the three levels are related in the ratio 1.0:3.0:14.2, respectively. Eq. 10, which specifies conversion coefficients of manpower supply by skill types into social groups, fitted very well with the data.18 Several regressions of eq. 11, which explains separately wage and salary rates for low and high-skilled labour, were tested and the best fits selected. The estimates allow calculating the elasticity of the remuneration rate by earner type LRh, with respect to average productivity, ΣjVj / LDh , calculated at mean values. The elasticities are above unity, and are found to be higher for wage earners than that for salary earners, which may indicate that economic growth in Korea favoured wage earners relatively more than salary earners. Calculated elasticities of remuneration to unemployment are −0.933 and −0.177 for wage earners and salary earners, respectively. These results emphasise the greater dependence of the wage rate on the employment situation; salary formation is more autonomous.19 Next, are 18 well-fitted consumption functions covering three household groups and six consumption categories, eq. 15. The equations fitted slightly better for wage and salary earning groups than for the employers and related group; this is expected, given the latter’s heterogeneity, and the fact that their consumption and income data were obtained as residuals after deducting figures for other groups (obtained from surveys of households and labour) from national aggregates (available from national accounts). The estimates are in general conformity with well-known changing patterns of consumption with rising incomes. Calculated consumption elasticities at mean values show, for the three groups, elasticities lower than 1 for food and housing, higher than 1 for health and education and about 1 for commercial and other social services. In contrast, eq. 19, specifying the private investment function, performed as the poorest of all equations and had to be slightly modified to produce reasonable estimates.20 Finally, there are the regressed coefficients relating to the formation of nutrition, housing, and health. Eq. 26 explains the variance in unit costs of daily

Social economic development models  81 calories by the annual disposable income per capita. The regressed data fitted well, allowing for the calculation of an elasticity of calorie cost with respect to income per capita, evaluated at mean values, at the value of 0.606. This implies that a 1 per cent increase in income per capita adds 0.6 per cent to calorie cost; thus, leaving less income for more calorie consumption.21 Eq. 27, which relates to housing levels, gave a well-performing regression as well. The calculated elasticity of the cost of room per capita with respect to income per capita, evaluated at mean values, amounts to 0.5891.22 The elasticities are practically equal for both nutrition and housing. Eq. 28, which relates to health, performed well too. But the calculated elasticity of survival cost with respect to income per capita came out at 1.0393, implying that growth in income per capita accelerates the growth in the costs of avoiding death. This reflects the general tendency for costs of medical care to rise more than incomes. As was stated earlier, because the equation is not able to connect functionally with the rest of the model, we chose to replace eq. 28 with eq. 28.1, which takes over given mortality rates from the demographic projections instead of predicting them endogenously. Estimated on the basis of data for 1960–72, the model is solved for three sixyear periods: 1966, 1974 and 1980. Comparisons between projected solutions and observed values for the sixties show that the employed model represented the Korean economy fairly well. With respect to the post-observation periods, the predictions for GDP and its composition for 1974 and 1980 were in a range of values close to those actually observed. More insight on the validation of the outcomes can be gained from an evaluation of structural ratios between the solved variables. The investment share in GDP increases remarkably—from 0.24 to 0.34 and to 0.39 over three periods—which is consistent with the high growth of GDP at 9.0 per cent. Regarding financing of investment, the ratio of domestic savings to foreign capital inflow increases from 0.57 to 0.74 to 0.94, indicating increasing self-sufficiency. The general labour coefficient that divides total employment by total product undergoes remarkable reductions in each period— from 8.4 to 5.1 and to 2.7 in 1968, 1974, and 1980, respectively—reflecting the remarkable growth in labour productivity. Concurrently, labour input shifts substantially to the high skilled and away from the employer and self-employed category. Employment rates, WMh, develop gradually towards fuller employment indicating that by 1980 unemployment is reduced to 5 per cent for wage earners and to 2 per cent for salary earners. The combined employment rate for both groups reaches 0.96 in 1980, see Figure 4.1.23 The remarkable growth in labour productivity, together with a favourable labour market, raised earnings appreciably. Results show that with respect to the disposable income per capita WY h , while in 1962 the group of employers, self-employed and family workers had the highest value, from 1968 onwards salary earners were the best off. Wage earners were the least off. The income gap between the two earner groups appears to widen, but that between earner groups and group 3 is shown to shrink. For 1968 and 1980, the Gini index of income concentration derived from the income forecasts of the model and population projections indicates a general progress towards greater equality, as shown in Figure 4.1. The Korean combination of high economic growth (WG = 9.6 per

82  Social economic development models cent), with fair distribution tendencies (Gini index falling from 0.48 to 0.46); is often quoted as exemplary in the development context. Regarding the progress of general living conditions for the (lowest income) wage earner group, the indicator of calorie consumption per day per person, WC1 is the first to reach its minimum requirements level, that is 2290, followed by the enrolment rate WN and the survival rate WH. The housing indicator, WD1, is the most remote from its ‘satisfaction point’.24 The progress of the composite index of the general living conditions, BNA1, with special reference to wage earners, is shown in Figure 4.1. This index is found to progress gradually from its observed level of 64 per cent in 1968 to a predicted level of 92 per cent in 1980, and at this rate to reach its maximum of 100 per cent in a couple more years. The improvement amounts to 28 percentage points over a period of 12 years. It can be asked: what is the relative contribution of the productivity effect to the economic growth in the 12 years? Applying this 28 per cent to the low-skill labour supply elasticities in the agriculture and industry production functions (0.24 and 0.34), and weighting the outcomes by the GDP shares of agriculture and industry (0.32 and 0.48 in 1971) gives a contribution of 6.44 per cent over the 28 years, or an average annual contribution of 0.54 per cent, whereas the solved average growth rate of the GDP amounted to about 9.0 per cent. The share of the contribution of the productivity effect in GDP growth can thus be roughly estimated at about 6 per cent: that is 0.54 per cent to 9.0 per cent. The contribution is a once and for all effect that ceases as BNA1 reaches its satiation level in the beginning of the 1980s. 4.2  Multiplier properties To bring more policy practices into the analytical framework, the conventional multiplier analysis was extended in two respects. First, while the model generates impact multipliers that measure the immediate effect of a unit increase in an exogenous variable on endogenous variables, and generates interim multipliers that give the effect of a sustained change in the exogenous variable in period t on endogenous variables in later periods, there is another notion of the multiplier that is applicable in this model. When a government makes plans in period t for 1 GDP

0.8 0.6

GINI

0.4

WM

0.2 0 1966

BNA1 1968

1970

1972

1974

1976

1978

1980

1982

Figure 4.1 Korea model solutions: indices of selected aim variables. GDP index takes 1982 = 100

Social economic development models  83 a specific addition to public expenditure that is for food and housing subsidies, health and educational allocations, or income transfers; or makes a revision in tax rates, these measures are not only sustained in later periods t’, but often tend to recur in later periods t’, especially when there is a commitment to maintain the extended services. Assuming that the magnitude of the manipulated addition in a particular exogenous variable in each period is the same, one obtains what can be called recurring multipliers, which were simulated. Second, in the real world, government allocations to consumption and investment are closely tied. Often, an increase in public consumption Cgc implies a certain proportionate increase in public investment Igj. This complementarity relationship makes it essential to evaluate the combined multipliers of Cgc and Igj. For example, the ratios of Igj to Cgc in Korea in the observation years were stable, and amounted to zero for food and housing subsidies; they were 1.0, 1.4, 2.0, and 3.4 for spending on health, primary, secondary, and higher education; and 3.0 and 3.3 for spending on other goods and other services, respectively. For brevity, we report briefly on findings on the recurring multiplier effects of current and associated capital spending bundles on the distinguished budget spending categories: food, housing, health, education, other goods, and other services. We focus on two recurring multiplier effects: one is on gross domestic product, GDP, that is ΣjVj, and is plotted along the x-axis in Figure 4.2; the other is on the income of the poorest population group, Y1, which is plotted along the y-axis in Figure 4.2. When the two variables are considered together, they would reflect on both growth and distribution. The upper right quarter of the figure is where positive growth is combined with positive redistribution. The joints indicate transition of the multiplier effect from one six-year period to a later period. As can be directly seen, there are trade-offs between immediate and later returns; the immediate returns in the first period are not rewarding in terms of both growth and distribution, but accumulated returns over a couple of periods are positive in both terms and tend to accelerate. The sluggish effect on GDP in the first period tends to diminish, somewhat, the income accruing to wage earners Y1. Both effects change sign as GDP increases in later periods. X axis = ∆GDP

Y axis = ∆ Y1

40

30

food or housing subsidies

25

health

35

20 education

15

10 5 0 –5

–5

5

10

15

20

25

other goods & services

Figure 4.2 Korea: recurring multiplier effect of public current plus associated capital expenditure

84  Social economic development models The results show that the multiplier as defined here is insignificant for food and housing subsidies, and barely shows up in the figure. Health spending contributes to both aim variables. Education spending contributes somewhat more. Allocations to other goods and other services perform better. The main mechanism behind these effects is that capital appears to be scarcer than labour in determining economic growth, while skill upgrading counts too. In this context, public current allocations that are associated with higher capital expenditures would, over time, increase the availability of investment goods and their upgrade leading to higher economic growth. If food and housing subsidies were accompanied by agricultural and housing investment projects, their multipliers would have been positive and significant. The results can explain, to some extent, why impatient governments often show little readiness to wait for longer periods before reaping the rewards of public spending. Another mechanism, already mentioned, is the productivity effect of higher levels of BNA1; this is less active in the Korean context as the relevant satiation level of BNA1 was practically reached.

5  Analytical versus policy uses: breakdown of policy making 5.1  Structure and functioning of the analytical model In conformity with accepted uses, the analytical form of the model treats aim variables as unknowns and instruments as given. The complete planning form is usually described as the reverse, with all aim variables taking the form of fixed targets and instrument variables becoming unknowns. In this section we assess both forms. Figure 4.3 gives the causal ordering of the analytical form, showing for each equation number the related endogenous variables, and the controllable policy instruments that can be employed to influence solutions. Simon (1953) defines the general rule for establishing causal ordering: endogenous variables of the first order are those which are influenced by parameters (and predetermined variables) and (or) by each other, mutually. Endogenous variables of the second order are those which are influenced only by parameters (and predetermined variables), and (or) endogenous variables of the first and second orders, and so on. A model containing one order is a complete simultaneous system in which, in principle, any controllable policy instrument or data can influence all unknowns. (For discussion of the causal ordering of prototype CEM and CGE models, see Chapter 2.) All the 32 equations and variables are solved within one order, except those at the very end of the figure: these are eqs. 16, 17, and 32 that relate to Zg, GBD, and FCF.25 The very high simultaneity in the model is due to various interdependencies that exist between the sectors that produce goods, that is agriculture and extended industry, marked by G, and the sectors that produce services, marked by S. Production technology is modelled differently in the two categories, whereby production in the G sectors combines the two levels of production function and input-output technologies; the S sectors produce their services along input-output technologies only. The types of sectors interact at the input-output level, and compete for labour and investment inputs. While

Social economic development models  85 Solution structure

Equations 3jG, 5jG

KDjG XjG

1jG, 8 2G, 2S, 7

VjG

23, 11

NTe

VjS

5jS , 6 10, 29 9,25

NE

τjg

LDqjS

VjS

GDP = ƩVj

Instruments

LD qjG

LSq

LDh

GDP

WN

(Zh, Zg)

19,14

Yh

Ip

15,24 Cj

I Mj

21,22

Igj, Cgc

Chc Ej WC1

26–7-8 XjS XjG KSjG

Tgh Igj, τjg

, WYh

Chc

20,18

32

WMh

LRh

WG

12,13

31 30 16 4 17

Cg4

LSh

ICHjG

WD1

WH

Cgc

BNA1 τhg, τjg, τgE, τgM

Zg GBD

Cgc , Tgh, I gj FCF

Figure 4.3 Causal ordering of the analytical model

the formation of labour supply is dependent on educational spending (placed at the right-hand side in the figure), this spending is only known if incomes are determined. Capital formation is delivered and primarily purchased by the goods sectors (on the left-hand side), but this can only be known if outputs and incomes are determined. If the two sector categories would fall into lower and higher causal orders, the structure of the model would have been diagonal and simpler; but it is otherwise. Furthermore, the productivity effect of basic needs attainment brings more simultaneity to the model, since they enter the production function in the very first equation, but the solution of BNA1 can only be obtained at the end of the circular flow, and after realisation of consumption. The dashed arrows in the figure showing an upward direction highlight the high degree of interdependence and simultaneity. 5.2  The planning form: a decomposed structure In redesigning the model towards its planning form (which comes to fixing aim variables and solving for suitably selected policy instruments) the question presents itself as to whether there exist meaningful causal orderings that can tell which specific aim variables should be fixed first, and which others should follow later. Granted that for the moment that there are no political preferences that dictate the form in which the planning model should be shaped, the planning advisor may search, on purely technical grounds, for the logically simplest structure. By putting side to side the matrix structures of a number of conceivable

86  Social economic development models alternative planning forms, then that which has the simplest structure can serve as a starting point. It logically follows also that after starting with the simplest structure one should extend it later to more complex structures. In this way, the analytical form of the model is adapted in a step-wise manner to the fixation of targets for more aim variables of a more complex nature until the complete planning form is approached. For our particular purpose, we shall consider that as we move to forms whose mathematical structures fall into a larger number of orders, we approach a simpler structure. By shuffling rows and columns we can triangulate the particular structures of alternative forms and observe which structure is simpler. Application of this procedure shows that first: the planning form which fixes targets for disposable incomes per capita, WY h, and uses direct tax rates and (or) income transfers as unknowns, has a simpler structure than the analytical form. Figure 4.4 displays the case where income transfers are the adjustors, but the same can be done via direct taxes or a combination of them. As was mentioned earlier, fixing Y h for all groups is equivalent to fixing GDP, and choosing a targeted distribution of income. In the formulation of distribution policy, various alternatives are supposed to be evaluated in the light of their effects, not only on the welfare of the immediate beneficiaries, but also for the whole economy. In the next section we shall examine some applications of alternative distributions. After this first planning form, the most logical step is that in a second round such supplementary aims as higher nutrition, housing, education and so on, for the lowest income social group, can be fixed; in particular, WC1, WD1, (WH is already predetermined), and WN. Public allocations towards these ends are the instruments to be solved for: these are Cg1, Cg2, Cg3, and Cg4, respectively. This simplifies the structure of the model appreciably, since the labour productivity effect of the targeted higher quality of life can be immediately fed in the production functions in eq. 1 (see the upper left-hand side of Figure 4.4), while the allocations to education Cg4, allows determining straightforward NTe, NEe, LSq, which are further converted into LSh (see the upper right-hand side of Figure 4.4). In a third round, with labour supply known, fixation of employment targets implies that distribution of production on the sectors of activity has to adjust itself so as to absorb the labour supply consistent with the employment targets; this occurs in equations in the upper shaded order in Figure 4.4. Once production levels of all sectors and labour inputs are solved, the main adjustors in the final analysis become: (a) distribution of installed investment on the goods-producing sectors, leading us to solve for the unknown policy instruments of public investment, Igj, in sectors j = 1, 2 (see the lower left-hand side of Figure 4.4); and (b) distribution of inventory changes ICHj among the product balances in the goodsproducing sectors, j = 1, 2; shown in the lower shaded order. This third round can be described as the complete planning form, since it incorporates fixed targets for all target variables (note that the target variables of GDP and BNA1 follow indirectly from the others). Contrasting figures 4.3 and 4.4 shows the planning model to be fully decomposable in clearly defined diagonal orders, which increases insight and control over the model and its outcomes. The crux of the analytical model fell into one

Social economic development models  87 Equation number 14

Yh Chc

15, 29 30,26–7–8 7

BNA 1

Cg1 ; Cg2 ; Cg3 ;

MjS

EjS LD

1jG , 11 GDP= ΣVj 15 3, 19, 23

KD jG KS n

4, 12,13

Igj

21,22 26–7–8

I EjG

WN

NEe

WC1, WD1

LS h WMh

LDh

5, 31 jS , 9 2

4, 31,30

Cg4

NTe

LSq

CjS

18jS, 6 10 22jS, 21j S 8

Fixed aims WYh

Solution structure

XjG

XjS

VjG

VjS

LD q

GDP Ip

ΣjLD qj

LR h

WG

WMh Zh

Tgh XjG

XjS

MjG

Zg

ICH jG

16

FCF

17

GBD

Figure 4.4 Causal ordering of the planning model

order that had to be solved simultaneously. The planning model falls into some 13 subsequent orders that are solved one after another. There are two more orders common to both models; these are at the bottom, where FCF and GBD are determined. The planning model is not only more helpful than the analytical model in increasing analytical insight and management control, but policy makers would prefer it too. As policy makers are equally interested in aim variables and in the achievement of a balanced government budget, they like to see how the budget is shaped by targeted variables. The planning model does not specify all policy instruments as unknowns. It contains additional degrees of freedom (instruments that are fixed a priori, and which can be modified to suit a balanced budget). These are tax rates τhg , τjg , τjge , τjgm , and current expenditures Cg5 , Cg6 , and capital expenditure in the services sectors, Igj , j = 3, 4. How do the above forms treat clearance of the product and financial market balances? This comes to specification of closure rules. In both forms, clearance of the product market balances happens via inventory changes in the goodsproducing sectors and, as such, the closure specification employed can be described as affiliated to closure A, (see Chapter 2). With respect to the savingsinvestment balance, as domestic savings and investments are determined in equations that preceded, and as the foreign exchange rate, FXR, is fixed, it is foreign capital flow, FCF, which is the primary adjustor. This closure was denoted as

88  Social economic development models closure C in Chapter 2. However, because all of the equations in the analytical model are involved in a simultaneous solution of the unknowns, specific adjustors cannot be assigned per equation. However, under the planning form of the model, the solution structure of the model is more diagonal, which allows for the above assignment of adjustor status to particular variables. 5.3  Some illustrations The first planning form where incomes of social groups are fixed and taxes are made unknown can be planned via alternative strategies of income redistribution. Results of three alternative strategies are shown in Table 4.4. A first strategy, I, stipulated increases in the disposable income of wage earners by 10 billion won in 1968, 1974, and 1980, all to occur without affecting other incomes. This led to lower GDP at the end of the third period, in 1980, involves a reduction of −15.68 billion won, a higher requirement of foreign capital inflow +16.94 billion won, and a higher budgetary deficit + 25.68 billion won. In addition, this strategy raises unemployment for both manpower types. A second strategy, II, applied income transfers from salary earners to wage earners, at the same rates and periods. This strategy reduces the above strains on GDP, FCF, GBD, and unemployment rates. A third strategy, III, applied the income transfers from the employers and related group to wage earners, at the same rates and periods. This strategy brought about the best performance with respect to the four macro variables. Historically, too, in the process of development a trend is often observed towards a larger income share for wage and salary earners, at the cost of employers and related. When the aim is to raise basic needs attainment, there is the choice between indirectly raising income (via reduced taxes or via transfer payments) versus provisions in kind. Eqs. 26 and 27 allow reflection on the effectiveness of cash transfers versus provisions in kind. Besides the obvious fact that not all the cash transfers would be spent on the targeted food or targeted housing, there is the second and more important aspect that cost is a function of income; and increases in income are bound to increase the unit cost of the basic needs and diminish the amounts purchasable. As a result, direct cash transfers are shown to be less effective than provisions in kind in raising standards of living.

Table 4.4 Alternative strategies of fixing income targets by social groups: deviations from analytical solutions for Korea, 1980

GDP FCF

Fixed income strategies

Fixed income strategies

I

II

III

I

II

III

−15.68  16.94

−6.57  4.44

−1. 48  0.83

 25.68 −83

 6.57   −35

 1.48    −7

GBD LDl + LD2

Note: GDP, FCF, and GBD in billion won, LD in thousand persons

Social economic development models  89 5.4  Other elaborations The empirical investigation of the model in both its analytical and planning forms allows us to make a few observations of a general nature. In a comparison between the multiplier effects of (the common) exogenous variables in both the analytical and planning forms it is found that for most unknowns the effects can be substantially different between the two forms. This testifies to the high sensitivity of the reduced form to the particular specification of the model structure. It is also worthwhile to investigate and compare the predictive performances of the analytical and planning forms in other respects. Tinbergen (1970) mentions that highly ordered model structures would give solutions which are more reliable than the outcome of model structures with fewer orders. This is not always supported, as the comparisons in Table 4.5 show a prediction advantage for the analytical over the planning form. The results are consistent with the nature of the aim variables and instrument variables. In the analytical model, the unknown aim variables of disposable income have a bigger base, so that an absolute error in the unknown becomes a relatively small one. But in the planning model, where income taxes are the instrument variables, and which have a smaller base, an absolute error in their prediction can become, relatively speaking, a very high one. In general, in most economic models aim variables carry higher values than do instrument variables.

6  Concluding remarks The big advance made in defining, measuring, applying and comparing wellbeing indicators stands in contrast with little work achieved in incorporating such wellbeing indicators in economy-wide policy models. This is to some extent understandable in view of difficulties encountered in formulating and verifying relationships that link wellbeing indicators with conventional variables in an economic model. In this chapter, we attempted to incorporate basic needs indicators, next to other development goals, in an economy-wide model, and with some success. Several lessons can be summarised. First, because the number of indicators that can be integrated is by nature quite limited, a systematic and consensus-oriented approach needs to be followed in choosing and (or) minimising the selection of indicators. Our approach was to stick to income per capita Table 4.5 Predictive performance of the analytical and planning forms, Korea Analytical (1) (2) (1) / (2) Observed Solution Y1 Y2 Y3

0.33 0.1288 0.4941

0.2973 0.1287 0.5187

0.9009 0.9992 1.0502

Policy (3) Observed

(4) Solution

(3) / (4)

τ1g τ2g τ3g

−0.0454  0.0058  0.0884

wrong sign 1.21 0.5

0.0082 0.0070 0.0443

90  Social economic development models goals of household groups and supplement these with wellbeing indicators in case income is around or below the poverty line. Second, there is a mixed encounter with the formation of wellbeing indicators. Easy and handy shortcuts worked well in areas of food, housing, and education, but there is less success with concise specifications for health indicators. It is also interesting to note the empirical verification of an increasing cost per wellbeing unit with higher income per capita. Third, incorporation of the productivity effect of higher wellbeing is best and simplest done through defining the affected (number of) worker units in terms of capability worker units, whereby full capability is reached when basic needs are attained. It was possible to estimate the contribution of this productivity effect at about 6 per cent of the total GDP growth. The contribution is a once and for all effect that ceases as the wellbeing indicators reach their satiation level, that is the case in Korea in the early eighties. Fourth, when there are so many aim variables to be targeted and instrument variables to be solved, there is the central question in policy making on how to proceed with policy making, that is which aim variables should be planned first (or treated first), and which others should logically follow? Causal ordering of the modelled system, due to Simon (1953), is very useful in decomposing the analytical model in its planning forms. It turns out that the prefixing of disposable incomes by social groups is the most logical first move in the policy-making process, followed by basic needs, and employment targets. The GDP variable appears down the list following causal ordering. It is understood, of course, that the conversion of the analytical model into its planning forms (where goal variables are fixed) does not imply that central-planning machinery is assumed. Working with planning forms is a device in policy making that increases insight into and control over the mechanisms of the model.

5 Growth and distribution in SAM models Various applications

1 Introduction Special reference was made in the introductory chapter to the publication of the Chenery et al. (1974) book, Redistribution with Growth, where they surveyed existing multisector planning models and found them to be inadequate for formulating development strategies. They concluded that a basic reformulation is required, and that what is needed is a compact model that provides a unified treatment of the determination of growth and distribution among different social groups: effective assessment of development strategies requires working with a structured economywide model that depicts the whole circular flow of the economy, with explicit divisions into production activities, production factors, possessing institutions (that is, households, firms, government), and the rest of the world. In terms of descriptive statistics, there is only one unified statistical framework that is capable of collapsing the economy-wide circular flow into a compact framework, includes all national accounting balances required for modelling equilibrium, and that allows for an appraisal of growth and distribution, next to other development issues. This unified statistical framework is the social accounting matrix (SAM), first conceptually developed as an integrated system of national accounts by Stone, and brought into life again via a large number of applications and refinements carried out by many contributors, including Pyatt, Thorbecke, Adelman, Robinson, and Cohen. The idea of a SAM is traceable to Quesnay’s Tableau Economique in 1758. The idea was revived some 200 years later. Hicks coined the term social accounting in 1942; the realisation of a SAM was the work of Stone (1947); and it was associates of Stone, working in the context of developing countries, who 30 years later presented the first comprehensive publication of a SAM for a developing country, cf. Pyatt and Roe (1977). In the two to three decades following the introduction of the SAM, there has been a remarkable shift of interest from the input-output matrix and multisector models to the SAM as a basic framework for economic modelling. Practically speaking, all developing countries have now some reference available on a constructed SAM, and many have used the SAM as a benchmark for calibrating a relating computable general equilibrium (CGE) model. In this chapter, a few SAM applications are selected from the many SAM multiplier analyses applied to developing countries. After briefly introducing the

92  Growth and distribution in SAM models basics of the SAM and its conversion into a circular flow model, various applications review multiplier results relating to the trade-offs between growth and equality for ten developing countries. One last application goes further and highlights the significance of dual structures in growth and distribution, making use of SAM multipliers. This is illustrated for one of the ten developing countries: Indonesia.

2  Tabulation and construction of the social accounting matrix By way of introduction, it can be stated that the SAM is compiled according to the same accounting principles as input-output tables: each transaction is recorded twice, so that any ingoing in one account must be balanced by an outgoing in another account. However, the SAM contains a complete list of transactions describing income and expenditure, and production flows among sectors, factors of production, and groups of households. These transactions are usually grouped into several sets of accounts belonging to various economic agents, as will be elaborated later. The SAM itself is nothing more or less than the transformation of the circular flow among agents in the national economy, as depicted in Figure 5.1, into a matrix of transactions between the various agents, as displayed in Table 5.1. In the first line of the figure, production activities employ factors of production and pay them. Second, institutions (including households, firms, and government) earn their incomes from supplying factors of production. Third, institutions spend their incomes on consumption of commodities and on savings that are turned into demand for capital goods. Fourth, production activities receive demands for consumption and capital goods and supply them back to the market. To produce these goods, the production activities employ factors of production and remunerate them in the factor market, thus closing the circle with the first line in the figure.

Payment to factors (Factors account) Incomes of institutions: i.e. households, firms, government

transfers

Consumption by commodity (Wants account) Production by sector : Activities Figure 5.1 Circular flow

spending exports, imports

G O V E R N M E N T

R E S T of W O R L D

Growth and distribution in SAM models  93 Separated on the right-hand side are transactions of government and rest of the world (ROW). At the upper bound, both government and ROW engage in income transfers to households. At the lower bound, government spending goes to sector production via recurrent and investment expenditures, while ROW links with activities via exports and imports. In correspondence with the circular flow, the SAM contains the following account categories: 1 Factors of production accounts; 2 Institution accounts belonging to different socio-economic household groups and to firms. Note that the main difference in accounting transactions between households and firms is that households spend on consumption while firms do not; 3 Institution account belonging to government; 4 Commodities accounts consist of consumption spending of the institutions on goods and services;1 5 Activities accounts consist of payments received and paid for in and out deliveries among the sectors; 6 The capital account consists of capital transactions aggregated over all institutions and activities; 7 The rest of world account. It is worthwhile to mention that the SAM accounts can be alternatively ordered. In the past the author, along with many others, constructed SAMs that started with the commodities account. Others start with the activities account. In retrospect, it is more fluent to start with the factors account, and proceed further with institutions and activities, and close up with the capital account, as is proposed here, and followed throughout the book. This has the advantage of achieving closer correspondence of the SAM with the CEM and CGE models, which start with equations relating to the factor market and specify further equations following the above order, and end with the three national accounting balances. The circular flow transactions are presented in a matrix form in Table 5.1. The factors account usually distinguishes between remuneration of capital and several types of labour. Value added paid out to the factors of production labour (wages) and capital (profits) comes from production activities that employ these factors. In the household institutions accounts, the rows show how they receive their incomes from factor remunerations and transfers and, in the column, how they pay out some of their incomings as direct taxes and transfer payments and spending on consumption; the rest goes to savings. The government institutions account is the counterpart of the above. The commodities account includes expenditure on consumption categories such as food, housing, clothing, health, education, transport, and other goods and services, which can also be classified straightaway in terms of activities. The activities account shows in the row the money receipts of the producing sectors from the sale of consumption goods, investment goods, intermediate goods, and exports. Column-wise, the sales revenue of the producing sectors goes in earnings to factors of production

Households, firms h Government

Commodities

Activities j

Capital

Rest of world ROW

2

3

4

5

6

7

Totals

Factors k

1

Factor income = payments

Factor income

Factors k

Table 5.1 Specimen of SAM

Government consumption

Transfers

Government

Private expenditure = pr. income

Government expenditure

Private savings Government savings

Private consumption

Direct taxes

Households firms h

Consumption spending

Competitive imports

Domestic consumption

Commodities

Domestic investment

Capital

gross input = gross output

Total investment

Non-competitive Competitive imports imports

Domestic intermediate

Indirect taxes

Value added earned

Activities j

Payments by ROW

Foreign capital flow

Exports

Transfers

Transfers

Total savings = total investment Receipts of = payments by ROW Grand total

Government income = exp. Consumption expenditure Gross output

Factors payments Private incomes

Rest of Totals world ROW

Growth and distribution in SAM models  95 employed, and in costs for intermediate deliveries used. This fifth row/column represents product-market balances, which is central for economy-wide models and accounting systems. All of these five accounts can be called current accounts in contrast to the sixth account, which is the capital account. This account equalises all domestic and foreign savings received to investment spending. The sixth row/column is the savings-investment balance, which was discussed thoroughly in Chapter 2. Finally, the rest of the world account shows imports to be matched with exports and foreign capital flow. This seventh row/column is the foreign payments balance. As was observed in previous chapters, this balance (account) follows automatically from the whole system, as other accounting balances are fulfilled. The construction of the SAM is usually done in two steps. In a first step, the published estimates of the national accounts are sufficient to construct an aggregate matrix of incomings and outgoings for the accounts. In the second step, the accounts undergo disaggregation. The required data for disaggregation the SAM of include: (a) the household budget surveys for breaking up the household account into specific household groups, and specifying their incomes by source and their expenditures by type of consumption product; and (b) the input-output table for disaggregating the production activities by producing sectors and converting data for transforming the product classification into the sector classification. It can happen that the data are not available for a couple of cells in the disaggregated SAM or they are less reliable; in such cases, sound judgements on the initial values are required. The definitive values are obtained by applying the RAS method to the whole matrix, which simultaneously assures consistency of the grand totals of the rows and columns. Table 5.1 summarises a specimen of SAM with the seven account types: factors of production k, households and firms h, government g, commodities j, activities j, capital, and rest of world.

3  The SAM as an economy-wide model Making use of notations in Table 5.2, and taking cell entries as proportions of their column totals, and applying substitutions for relevant cells, the SAM can be displayed as a SAM model in seven equations, as in Box 1. Eq. 1 corresponds with row 1 of the SAM, where factor earnings are derived from value added coefficients ϖjk. The value added is gross production after the deduction of received domestic intermediate inputs via αjj' , imported intermediate inputs via μrj , and indirect taxes via tax rates τjg. Eq. 2 corresponds with row 2. It specifies incomes of institutions as consisting of proportions ρ of factor payments and received transfers. Eq. 3 corresponds with row 3. Government income consists of direct and indirect taxes and transfers. Eq. 4 corresponds with row 4. It shows disposable incomes of private agents after deduction of direct taxes (via τhg) and distributes it on consumption expenditure by sector (via budget proportions γhj). Government consumption spending on sector j is separately specified.

96  Growth and distribution in SAM models Table 5.2 Notations Indices: k = factors, h = household groups (and firms), j = commodities and activities; furthermore, g for government and r for rest of the world Endogenous variables Cj Private consumption expenditure on sector j Foreign capital flow FCF Factor income of factor type k, referring to labour of different skills and capital Vk Gross production of sector j Xj Received incomes of household groups (and firms) h Zh Revenue of government Zg Exogenous variables Cgj Public consumption expenditure on sector j Exports of sector j Ej Investment delivered by sector j Ij Tgh , Trh Transfers from government and rest of the world to institutions

Eq. 5 corresponds with row 5. The equation gives the product balance by sector j. The supply of product or good j consists of the gross output plus imports of competitive goods (these are the imported parts of consumption and investment expressed via the import coefficient μ j) is equal to the demand for product j consisting of intermediate deliveries, consumption, investment, and exports. Investment delivery by sector j is a portion ιj of total investment. Eq. 6 corresponds with the capital account in row 6, in contrast with the above accounts that treated current transactions. The equation specifies the financial-market balance that equates private, government, and foreign savings to total investment. Eq. A.1 corresponds with row 7. Supply of foreign exchange payments received by the country (that is, exports and foreign capital inflow) is used to pay for the demand for foreign exchange (imports of competitive and non-competitive intermediate goods). As was stated in previous chapters, this balance is automatically obtained from the six equations above, and does not count as a separate equation; hence it is denoted by A.1. When n account balances are joined to each other in a consistent whole, the nth balance will follow from the others. The system is solved without the nth balance, though it serves as a consistency check. The structure of the SAM model can be traced in the causal ordering form of the model in Table 5.3. In these equations the endogenous variables and related coefficients are placed on the left-hand side, and the exogenous variables are placed on the right-hand side. As our focus is on development policy, the endogenous variables in this model would include factor remuneration, household income, consumption, and production. In the context of assessing development strategy it is best to take as exogenous variables the outlays of government, investment, and ROW.2 Eqs. 1, 2, 4, and 5 form the first and decisive order where the four major endogenous variables are solved: Xj, Vj, Zh, and Cj. In the next orders the budget revenue financing and foreign capital flow are solved.

Growth and distribution in SAM models  97 Box 1 The SAM model V k = Σ jϖjk (1 − Σj'α j’ j − α rj − τ jg )Xj k (1) h (2) Zh = Σk ρkh V k + Tgh + Trh (3) Zg = Σh τhg Y h + Σj τjg Xj + Trg Cj = Σh γhj (1 − τgh ) Zh + Cgj (4) j (5) Xj + μrj [Cj + ιj (I + ICH)] = Σj′ ajj′ Xj′ + Cj + ιj (I + ICH) + Ej (6) Σh (1 − Σj γhj ) (1 − τgh ) Zh + (Zg − Σj Cgj − Σh Tgh ) + FCF = (I + ICH) (A1) Σj Ej + FCF = Σj μrj [Cj + ιj (I + ICH)] + Σj αrj Xj

This structure of the SAM is equivalent to that of the CEM and CGE models in Chapter 2, Tables 2.2 and 2.3, respectively. In all three models, government consumption, transfers, and investment are set exogenously. Exports E are exogenously and (or) semi-autonomously determined in these models. The six endogenous variables of Cj, V k, Xj, Zh, Zg and FCF are endogenous in all three models. The crucial differences are in the adjustment mechanisms. The SAM shares product-market clearance adjustors with the CEM. Both rely on quantity adjustors in the clearance of the product-market balances and the financial-market balance. But while quantity adjustors clear markets in the SAM and CEM, it is price adjustors which drive market clearance in the CGE. A broader discussion of differences between SAM, CEM, and CGE models is in Chapter 2. More generally formulated, let the endogenous variables on the left side be denoted by vector v, exogenous variables on the right side by vector e, and the SAM matrix of coefficients by As; then the system can be described generally by v − Asv = e. Inverting the SAM matrix of coefficients and solving gives eq. A.2, where Ms stands for the SAM matrix of multipliers. Table 5.3 The causal ordering of the SAM model Eq Endogenous variables Factor Income of payment household Vk groups Zh 1

−V k

2 4

ΣkρkhV k

5 3 6

Pre-determined variables Consumption Output of sectors Xj by sectors Cj

Σ jϖjk (1 − Σj′ αj′j − αrj − τjg ) Xj

−Zh −Σh γhj(1 − τhg) Cj Zh −(1 + μ rj ) Cj (1 − Σj′ αjj′ ) Xj′ –Σh τhg Zh Σh (1 − Σj γhj ) (1 − τgh) ) Zh

−Σj τ jg Xj

Zg

FCF

Spending Transfers by by gov. government and ROW and ROW = 0 = = Cgj

Tgh + Trh

= ιj (1 − μrj ) (I + ICH) + Ej −Zg = Trg Zg FCF = Σj Cgj + Σh Tgh I + ICH)

98  Growth and distribution in SAM models v = (I – As )–1 e = Ms e

(A.2)

Depending on the disaggregation of the SAM, Ms can be a very large matrix. For most analytical purposes a small selection of specific multipliers is sufficient. The chapter will focus on the impacts of exogenous spending injections in sectors j (that is, Cgj, Ij, Ej) and of exogenous income transfers to household groups h (that is, Tgh, Trh). The impact will be traced on such endogenous variables as the output by sector Xj and the income by household group Y h. As such, the analysis can be restricted to four types of multipliers. These are the multipliers showing impacts by two types of exogenous impulses: spending and transfers, and on two endogenous variables: sector output and household incomes, Table 5.4.

Table 5.4 Selected output and income multipliers Impact of exogenous impulse on endogenous variables

Spending injection in sector j’

Income transfer to household group h’

Output of sector j, Xj Income of household group h, Y h

Ms,jj Ms,hj′

Ms,jh Ms,hh

There are several uses of these multipliers. For years directly before and after the year of computation, the SAM multipliers can be employed to show the economic effects of an exogenous spending stimulus for a particular sector emanating from government allocations, investment, or exports. The two economic effects to be examined are the output effects (that is, the growth of output and its distribution on sectors), and the income effects (that is, generation and distribution of income on household groups). In a similar way, one can assess output and income effects of income transfers to household groups from government or abroad.

4  Output and income SAM multipliers: results for ten countries The comparative analysis of this section considers SAMs for Colombia, Korea, Pakistan, and Suriname, constructed by the same team of investigators, and for six other countries: Egypt, Kenya, India, Indonesia, Iran, and Sri Lanka, from various published sources.3 The SAMs of these countries are very diversified in content. However, appropriate aggregation and modifications brought the various SAMs to a common base. In the process, we necessarily exclude some valuable information on some countries, but enhance drawing more general conclusions on the relationships between growth and distribution. Foremost is the specific disaggregation of the activity and household accounts. The classification of activities had to be limited to four large groups of sectors: agriculture, mining, industry, and services. Distinguishing more sectors would reduce the uniformity and comparability of the ten SAMs reported here. The classification of households emphasises dualities in the location of population in urban and rural areas, and the differentiation within urban and rural groups by level of

Growth and distribution in SAM models  99 income earned. This differentiation is done by distinguishing urban households into three groups: employers, employees, and self-employed; and breaking down rural households into three groups: large landowners, medium landowners, and small landowners or landless households. For a couple of countries, a seventh residual group was incorporated so as to accommodate classifications which did not fit the standardised six categories. The SAMs of Egypt and India form exceptions to the above, as they follow personal-income distributions in distinguishing between household groups with different income levels. The focus in the applications reported here obviously lies with the contributions of the SAM framework to deliberations on development policies. Accordingly, the analysis will start with (a) a review of results for output and income multipliers linked to sector injections and household transfers, (b) decomposition of these multipliers into transfer, open-loop, and closed-loop effects, (c) analysis of the growth and distributive built-in bias in the structure of the economy as reflected by the SAM; this will allow identifying gainers and losers, and (d) the compact use of SAM simulations to highlight strategic choices for growth with redistribution. Table 5.5 shows output multipliers for spending injections in sector activities. The multiplier reported here is Msjj′. Most of the countries fall in a middle range: Colombia, Korea, Iran, Indonesia, Sri Lanka, and Kenya, scoring 4.6, 3.6, 3.6, 3.4, 3.1, and 2.9, respectively.4 In general, output multipliers are highest for impulses in the agricultural sector, followed by the services sector, which is ranked second in seven out of the ten countries. Industry ranks between third and fourth. Mining does not show a common ranking among the ten countries. The five Asian countries of Iran, Pakistan, India, Sri Lanka, and Indonesia show a particularly homogeneous pattern of sector ranking regarding their output multipliers. Turning to income multipliers, Ms,hj' , it is observed from Table 5.5 that for all countries together, the size of the income multipliers are about half the output multipliers, 0.53 on average. For individual countries it is noted that Egypt, Pakistan, Colombia, Kenya, and Korea extract a lower income proportion from output, around 0.45, while Sri Lanka, Indonesia, and India extract a higher proportion, around 0.64. In the latter countries, it takes less of an output effort to produce a unit of household income. Table 5.6 summarises results regarding multiplier effects of income transfers to household groups. The impact of income transfers on output and income are denoted by Ms,jh, and Ms,hh', respectively. Compared with the previous table, in all cases average output multipliers of income transfers are lower than output multipliers of sector injections, but income effects from income transfers are higher than income effects from sector injections. For most of the ten countries, it is found that income transfers to the poorer household groups—mostly in rural areas—generates the highest output and income multipliers. Injections into richer households—mostly urban—have lower output and income multipliers. These results suggest that a redistribution of incomes from urban to rural household groups may increase growth and equality.

Table 5.5 Output multipliers Ms,jj′ and income multiplier Ms,hj′ following an exogenous spending injection of one unit in activity sectors, ten countries Country: Variable

Multiplier Agriculture Mining

Industry

Services

Average

Colombia 1970 Output Income

Ms,jj′ Ms,hj′

4.447 2.146

4.952 2.512

4.287 1.651

4.678 2.230

4.591 2.135

Surinam 1979 Output Income

Ms,jj′ Ms,hj′

1.967 1.153

1.871 0.777

1.849 0.816

1.911 1.220

1.900 0.992

Egypt 1976 Output Income

Ms,jj′ Ms,hj′

2.897 1.570

1.721 0.854

2.537 0.893

2.997 1.473

2.538 1.073

Kenya 1976 Output Income

Ms,jj′ Ms,hj’

2.872 1.867

3.099 1.127

2.541 0.910

2.934 1.501

2.861 1.351

India 1968–9 Output Income

Ms,jj’ Ms,hj′

6.910 4.427

5.185 3.384

5.977 3.328

5.711 3.599

5.946 3.684

Indonesia 1975 Output Income

Ms,jj′ Ms,hj’

4.176 2.567

na na

2.414 1.395

3.579 2.395

3.390 2.119

Iran 1970 Output Income

Ms,jj’ Ms,hj′

4.358 2.584

3.025 1.470

3.576 1.752

3.595 2.085

3.639 1.973

Korea 1973 Output Income

Ms,jj′ Ms,hj′

3.514 1.883

3.994 2.080

3.198 1.144

3.766 1.809

3.618 1.729

Pakistan 1978–9 Output Income

Ms,jj′ Ms,hj’

9.457 4.430

8.911 4.116

8.227 3.534

8.483 4.033

8.770 4.028

Sri Lanka 1970 Output Income

Ms,jj′ Ms,hj′

3.334 2.289

na na

2.929 1.714

3.007 2.163

3.090 2.055

Note: na: not available

Table 5.6 Output multiplier Ms,jh′ and income multiplier Ms,hh′ following an exogenous income transfer of unit to household groups, ten countries Country: variable

Multiplier Urban

Rural

Poor Middle Rich

Poor Middle Rich

Other Average groups All

Colombia Output Ms,jh′ Income Ms,hh′

4.117 4.417 2.795 2.913

4.014 4.809 4.356 2.756 2.965 2.863

3.108 3.958 2.341 2.728

4.083 2.766

Surinam Output Income

Ms,jh′ Ms,hh′

1.539 1.869

1.147 1.45 1.669 1.839

1.261 1.735

1.349 1.785

Egypt Output Income

Ms,jh′ Ms,hh′

2.593 2.464 2.124 2.083

1.662 2.108 1.582 1.759 1.901 1.692

1.482 1.651

1.982 1.87

Kenya Output Income

Ms,jh′ Ms,hh′

2.463 2.088 2.353 2.224

1.309 2.507 2.31 1.743 2.339 2.36

2.4 2.088 2.299 2.071

2.195 2.198

India Output Income

Ms,jh′ Ms,hh′

6.288 4.932

Indonesia Output Ms,jh′ Income Ms,hh′

3.146 3.011

4.433 3.752 2.727 3.916 2.761 3.44

5.361 4.342 3.366 3.413 3.104 3.142

3.314 3.092

Iran Output Income

Ms,jh′ Ms,hh′

2.568 2.414

3.459 2.916

Korea Output Income

Ms,jh′ Ms,hh′

3.206 3.388 2.475 2.557

3.146 3.041 2.798 2.445 2.387 2.773

2.515 2.143

3.016 2.38

Pakistan Output Income

  Ms,jh′ Ms,hh′

7.525 7.698 4.421 4.496

7.536 8.561 8.123 4.435 4.875 4.579

7.378 7.579 4.359 4.433

7.774 4.52

Sri Lanka Output Ms,jh′ Income Ms,hh′

1.962 2.323

2.33 2.378

2.305 2.689

2.265 2.53

3.014 2.665

102  Growth and distribution in SAM models

5  Decomposition of SAM multipliers into transfer, open-, and closed-loop effects The decomposition of the aggregate multipliers into meaningful effects will be the focus of this section. Recalling eq. A.2, the aggregate multiplier matrix MS in this equation can be decomposed into three multiplier matrices M1, M2, M3, as in eq. A2.1. See Box 2 for the derivation of the decomposition, v = ASv + e = (I − AS )−1e = MS e = M3 M2 M1 e (A2.1) M1 , which is known as the transfer multiplier, captures intra effects resulting from transfers that happen between one variable and other variables belonging to the same endogenous account (that is, effects of an impulse in a production activity on other production activities, or in other words, the Leontief multipliers). M2 , known as the open-loop effects, captures the effects of one endogenous account on other endogenous accounts (that is from production, to factor income, to household income). M3 , standing for the closed-loop effects, ensures that an effect from a specific account is subjected to circular flows through all endogenous accounts (that is from household income, to product consumption, to production activities, to factor income, and then to household income, and so on again and again through all four types of endogenous accounts). The understanding of the terms direct and indirect effects in the traditional output-input context may cause confusion in the SAM context. The following can be stated on terminology: the transfer effects are the Leontief effects, and they include both the direct and indirect effects in the limited sense and scope of the activities accounts. Similarly, the open-loop effects can be interpreted to include direct and indirect effects. However, the closed-loop effects are exclusively indirect effects, but then in the broader sense of the whole SAM framework. We shall limit the presentation here to the decomposed impacts of exogenous demand injections in activities on activities, Ms,jj , and comment briefly on the decomposition of other impacts. The decomposition of Ms,jj’ in terms of the three effects is shown in eq. A3, Ms,jj’ = M3,jj * M2,jj * M1,jj’ (A3) aggregate = closed * open * transfer. Since M2,jj does not engage other accounts than its own account of activities j, the open-loop effects are not applicable here, M2,jj being an identity matrix. The aggregate multiplier falls thus into the transfer effects (which are the Leontief multipliers), M1,jj′ , and the closed-loop effects, M3,jj. Table 5.7 displays the aggregate effects of Ms,jj’ next to the transfer effects, M1,jj’ , the latter representing the Leontief multipliers. The difference between the two effects is the closed-loop effects. With the exception of Surinam, in all other countries the closed-loop effects overshadow the transfer effects by large margins. For the ten sampled countries as a whole, the transfer or Leontief effects form only about one-third, while the closed-loop effects form two-thirds of the SAM multipliers. While the transfer effects, that is the Leontief multipliers, point to manufacturing as the key

Growth and distribution in SAM models  103 Box 2  Decomposition of SAM multipliers The aggregate multiplier matrix can be decomposed into three multiplier matrices M1, M2, and M3, as in the equation below. Ml gives transfer effects; M2 captures the open-loop effect, which is the interaction among and between the endogenous accounts; M3 is the closed-loop effects that ensure that the circular flow of income is completed among endogenous accounts. v = Av + e = (I − A)−1e = MSe = M3 M2 M1e (i) The formal derivation of the decomposed multipliers proceeds by separating matrix à from A, provided that à is of the same size as A, and that (I − Ã)−1 exists. v = Av + e = (A − Ã)v + Ãv + e = (I − Ã) (A − Ã)v + (I − Ã)−1e = A* v + (I − Ã)−1e (ii) Here, (I − Ã) refers to the transfer multiplier, M1. Derivation of M2 and M3 proceeds further as in equations (iii) to (v). Both sides of equation (ii) can be multiplied by A*, substituting for A*v from eq. (ii) and rearranging terms to give: v = A*2 v + (I + A*) (I + Ã)−1e (iii) The same manipulation can be repeated with A*2 up to A*k , so that in general: v = A*k v + (I + A* + A*2 + … + A*(k − 1)) (I − Ã)−1e (iv) For any positive value for k it is true then that: v = (I − A*k)−1 (I + A* + A*2 + … + A*(k − 1)) (I − Ã)−1e (v) Here then, (I − A*k)−1 is identified with M3 . (I + A* + A*2 + … + A*(k − 1)) is identified with M2 and, as was just mentioned, (I − Ã)−1 refers to M1. The multipliers can also be rearranged in an additive form. The multiplicative decomposition can be rearranged as done by Stone (1978), into four additive components, namely, the initial injection I and the net contributions of the transfer effect T, open-loop effect 0, and closed-loop effect C, as follows: Ms = I + (M1 − I) + (M2 − 1)M1 + (M3 − I)M2M1 = I + T + O + C (vi)

sector with the highest multiplier effects and the foremost contributor to growth of production activity from an additionally allocated unit of spending, these conclusions are turned upside down when the closed-loop effects are incorporated; this results in SAM multipliers that are higher for agriculture, mining, and services than for manufacturing. Many development policy recommendations on key sectors for economic growth and employment creation that were based on inter-industry analysis, typical of the fifties and sixties, lost their validity within the SAM framework. The results should not be interpreted to mean that if the ten countries had expanded in the past relatively more in agriculture than in manufacturing that they would have necessarily achieved a higher overall growth; realisation of the multiplier potentials requires significant exogenous expansion of spending both domestically and in the ROW in agriculture versus manufacturing; this

104  Growth and distribution in SAM models Table 5.7 Breakdown of SAM multipliers MS into transfer effects (Leontief multiplier) M1 and closed-loop effects M3 , ten countries

Colombia SAM multiplier Transfer effect (Leontief multiplier) Surinam SAM multiplier Transfer effect (Leontief multiplier) Egypt SAM multiplier Transfer effect (Leontief multiplier) Kenya SAM multiplier Transfer effect (Leontief multiplier) India SAM multiplier Transfer effect (Leontief multiplier) Indonesia SAM multiplier Transfer effect (Leontief multiplier) Iran SAM multiplier Transfer effect (Leontief multiplier) Korea SAM multiplier Transfer effect (Leontief multiplier) Pakistan SAM multiplier Transfer effect (Leontief multiplier) Sri Lanka SAM multiplier Transfer effect (Leontief multiplier) Average all countries SAM multiplier Transfer effect (Leontief multiplier) Closed-loop effect Ratio closed/transfer

Agriculture

Mining

Industry

Services

4.447 1.478 1.968 1.218 2.898 1.321 2.871 1.160 6.910 1.350 4.176 1.584 4.357 1.396 3.515 1.388 9.463 1.865 3.335 1.378 4.39 1.41 2.98 2.11

4.953 1.499 1.871 1.385 1.722 1.348 3.099 2.118 5.184 1.240 na na 3.025 1.454 3.995 1.451 8.924 1.888 na na 3.28 1.55 1.73 1.12

4.287 1.985 1.849 1.316 2.537 1.575 2.540 1.755 5.977 1.936 2.414 1.345 3.576 1.668 3.198 1.847 8.110 2.187 2.929 1.524 4.68 1.71 2.96 1.73

4.679 1.558 1.912 1.121 2.997 1.353 2.934 1.615 5.711 1.411 3.579 1.240 3.594 1.344 3.767 1.619 8.610 1.671 3.007 1.193 4.08 1.41 2.67 1.89

Note: M3 is the difference between MS and M1 ; na = not available

is a point for discussion. Besides, both the SAM and the input-output matrix emphasise the demand side of the economy and do not consider limits on the supply side. It is now easy to consider and comprehend the decomposition of the multiplier effects of sector-demand injections on household incomes, eq. A4, Ms,hj’ = M3,hh * M2,hj * M1, jj’

(A4)

In eq. A.4 a production injection creates M1,jj’ transfer effects (that is, a Leontief multiplier), carries these through M2,hj open-loop effects, thereby creating household incomes, after which these household incomes are engaged in a circular flow through the whole system and back to incomes—that is the closed-loop effect of M3,hh. The latter are behind the phenomenon that income-redistribution injections, after rotating through the circular flow of the SAM, have a tendency to vanish and reproduce the initial situation. The expenditure patterns of the

Growth and distribution in SAM models  105 different household groups tend to benefit sectors, earnings, and their households incomes in the same ways. Turning to multipliers of exogenous institutional income transfers to institutions, their decomposition can be made along the lines of eq. A5. Ms,hh’ = M3,hh * M2,hh * M1,hh’

(A5)

The decomposition is simple. Because of a lack of data on transfers among household groups, the transfer effect may not be fully traced and also, because the open-loop effect is not relevant in this context, the implication is that the closed-loop effect is about equal to the aggregate multiplier. Finally, the decomposition of the multiplier effects of household-income transfer injections on sector production is displayed in eq. A6, Ms,jh′ = M3,jj * M2,jh * M1,hh

(A6)

Given that M1,hh′ is not applicable, the transfer injections will be passed directly to sector activities through M2,jh and from there subjected to a circular flow through M3,jj.

6  Identification of gainers and losers in SAM multipliers Besides studying the level of a multiplier and the decomposable mechanisms that determine them, it is important to uncover the inherent structural distributive bias of the multiplier effects with respect to the constituents of an account. Which constituents are gainers and which are losers? The structural pattern of gainers and losers determines future development. To identify the structural pattern, a gainers and losers index for the sector to sector effects, GLIjj’ , is defined as follows: 5 GLIjj’ = [(Ms,jj’ − δjj› ) / (Σj Ms,jj’ − 1)] / [Output j,o / Σj Output j,o ] (A7) where j and j’ denote receiving and injecting sectors, and δjj›, (the Kronecker symbol, equals 1 if j = j’ and 0 in other cases) is subtracted to obtain the share of j in the multiplier effects of j’ on all sectors after deducting the initial injection. The result is divided by the actual output share of sector j in year 0, that is, the year to which the SAM refers. For values of GLIjj’ > 1, < 1, and = 1, there are positive, negative, and neutral redistributive effects. For instance, values of GLIjj’ = 1 mean that sector injections would exactly reproduce the sector dispersion pattern of the base year. Similarly, a gainer and loser index, GLIhj’ , for the multiplier effects of sector injections j’ on the income of institutional group h can be calculated. Values of GLIhj’ = 1 mean that sector injections would reproduce exactly the share group i had in total income in the base year. A higher value would mean an increase in its income share; a negative value would mean a decrease, GLIhj’ = [(Ms,hj’ ) / (Σh Ms,hj’ ) ] / [Income h,o / Σh Income h,o ]

(A8)

The two indices are calculated for the ten countries in Table 5.8. As regards the dispersion pattern of the output multipliers, in eight of the ten countries

106  Growth and distribution in SAM models agriculture gains from an injection in either agriculture or industry scoring GLI above unity. The results suggest that the value of GLI in favour of agriculture tends to diminish as a country develops. The growth bias changes in favour of mining, industry, and services. As regards the dispersion pattern of the income multipliers, the results show that for the ten countries demand impulses in the agricultural sector bring about a more equal distribution of income. Considering the countries individually, the results show the spending injections in agriculture to be most progressive in Colombia, India, Iran, Korea, and Pakistan and least progressive in Sri Lanka and Surinam. These results are partly due to open-loop relationships that strongly link sector activities in agriculture to particular household groups, such as the rural poor in India, Iran, Korea, and Pakistan. By contrast, there are relatively weaker correspondences between activities, factors, and households in the other countries. The results are also due to closed-loop effects that shift relatively fewer resources from poorer to richer groups than the other way around. This reflects more selforiented consumption patterns among poorer household groups in India, Iran, Korea, and Pakistan compared to the other countries, where rich and poor show more uniform consumption patterns. These results show that as far as spending injections in agriculture are concerned, progressive redistribution and higher growth are not in conflict with each other. Results indicate the presence of degrees of freedom in selecting balanced socio-economic development policies, in spite of the existence of countervailing processes which cause parts of redistribution and growth potential to vanish. Although agricultural multipliers favour both growth and equality, and multipliers of the other sectors are more inequality oriented in relative terms, the weight which is given to agriculture in the exogenous impulses during the course of development is usually much lower than for mining, manufacturing, or services. This allocation pattern of sector injections tends to make the income distribution more skew. This tendency is not characteristic of the multiplier system, but it is the outcome of the pattern of sector spending stimulus. This sector allocation pattern is likely the result of: (1) the exogenously determined limited opportunities for expanding agricultural demand (that is, it is more difficult to find foreign markets for agricultural exports than for non-agricultural exports); and (2) more restrictive production constraints in the case of agriculture as compared to industry, mining, or services. The same analysis of GLI is applicable to income transfers as well. In correspondence with the formula of GLI for sector injections, two types of GLI for household transfers can be formulated, giving GLIjh’ and GLIhh’. GLIhh’ = [(Ms,hh’ − δhh› ) / (Σh Ms,hh’ − 1)] / [Income h,o / Σh Income h,o ] (A9) GLIjh’ = [(Ms,jh’ ) / (ΣjMs,jh’ ) ] / [Output j,o / Σj Output j,o ] (A10) The results—which correspond with earlier results discussed above and need not be shown again—confirm the general tendency of obtaining a high GLI for the poorer household groups, especially the rural ones, which is an indication

Table 5.8 Gainers and losers index, GLI, of multiplier effects of sector injections, ten countries Actual dispersion in per cent

Multiplier dispersion in per cent

GLI

Average

Agriculture

Industry Average Agriculture Industry

20.52 1.42 40.92 37.14 62.09 37.90

19.15 1.15 44.30 35.40 62.89 37.11

21.62 0.93 43.66 33.79 59.64 40.36

20.34 1.45 44.12 34.09 63.74 36.25

0.93 0.81 1.08 0.95 1.10 0.98

1.05 0.65 1.07 0.91 0.96 1.06

0.99 1.02 1.08 0.92 1.03 0.96

7.48 29.23 18.08 45.22 57.47 42.53

13.81 6.46 25.00 54.73 57.63 42.38

14.03 0.07 27.38 58.52 56.61 43.39

25.57 0.06 25.47 48.90 57.83 42.18

1.85 0.22 1.38 1.21 1.00 1.00

1.88 0.00 1.51 1.29 0.99 1.02

3.42 0.00 1.41 1.08 1.01 0.99

21.43 5.39 39.24 33.95 66.39 33.61

27.17 9.32 36.20 27.31 56.39 43.61

35.03 5.07 37.41 22.49 25.77 74.24

30.19 4.50 38.14 27.17 63.59 36.40

1.27 1.73 0.92 0.80 0.85 1.30

1.63 0.94 0.95 0.66 0.39 2.21

1.41 0.83 0.97 0.80 0.96 1.08

Agriculture 23.50 Mining 0.46

41.91 34.13 37.62 62.36

20.21 1.96 46.05 31.78 40.68 59.32

28.60 0.35 38.99 32.07 30.01 69.98

22.22 0.91 49.98 26.89 43.10 56.90

0.86 4.26 1.10 0.93 1.08 0.95

1.22 0.76 0.93 0.94 0.80 1.12

0.95 1.98 1.19 0.79 1.15 0.91

37.02 2.20 35.84 24.94 60.40 39.60

43.58 2.11 27.35 26.95 56.40 43.60

48.68 1.79 25.07 24.45 45.58 54.42

40.52 2.70 30.22 26.56 57.38 42.62

1.18 0.96 0.76 1.08 0.93 1.10

1.31 0.81 0.70 0.98 0.75 1.37

1.09 1.23 0.84 1.06 0.95 1.08

Colombia Agriculture Mining Industry Services Urban Rural Surinam Agriculture Mining Industry Services Urban Rural Egypt Agriculture Mining Industry Services Urban Rural

Kenya

Industry Services Urban Rural India Agriculture Mining Industry Services Urban Rural

continued overleaf

Actual dispersion in per cent

Multiplier dispersion in per cent

GLI

Average

Agriculture

Industry Average Agriculture Industry

66.58 na 17.01 16.41 28.96 71.04

73.07 na 12.53 14.40 23.17 76.83

56.05 na 27.66 16.29 33.76 66.25

1.84 na 0.50 0.55 0.84 1.08

2.02 na 0.37 0.48 0.67 1.17

1.55 na 0.81 0.54 0.98 1.01

33.99 35.79 74.93 25.07

27.67 12.11 31.94 28.28 69.26 30.74

32.71 7.80 31.90 27.59 54.77 45.23

27.72 9.82 32.67 29.79 73.38 26.62

1.23 1.56 0.94 0.79 0.92 1.23

1.46 1.01 0.94 0.77 0.73 1.80

1.23 1.27 0.96 0.83 0.98 1.06

14.25 1.02 47.65 37.08 67.41 32.58

19.50 1.01 46.58 32.91 66.80 33.20

23.17 0.84 45.14 30.85 61.06 38.94

16.70 1.42 51.17 30.71 66.76 33.26

1.37 0.99 0.98 0.89 0.99 1.02

1.63 0.82 0.95 0.83 0.91 1.20

1.17 1.39 1.07 0.83 0.99 1.02

25.38 2.45 43.17 29.00 31.34 68.66

27.19 2.27 41.27 29.26 28.49 71.51

25.76 2.86 43.61 27.77 32.37 67.63

1.01 2.58 0.99 0.96 0.93 1.04

1.08 2.39 0.94 0.97 0.84 1.08

1.02 3.01 1.00 0.92 0.96 1.02

37.31 na 30.13 32.55 26.63 73.37

43.11 na 26.15 30.74 21.64 78.37

32.60 na 35.04 32.36 27.77 72.23

1.14 na 0.94 0.92 0.95 1.02

1.32 na 0.81 0.87 0.77 1.09

1.00 na 1.09 0.92 0.99 1.00

Indonesia Agriculture Mining Industry Services Urban Rural

36.14 na 33.94 29.92 34.46 65.53

Iran Agriculture 22.46 Mining 7.75 Industry Services Urban Rural Korea Agriculture Mining Industry Services Urban Rural

Pakistan Agriculture 25.24 Mining 0.95 Industry Services Urban Rural

43.73 30.08 33.88 66.11

Sri Lanka Agriculture 32.64 Mining na Industry Services Urban Rural

32.13 35.23 28.08 71.93

Growth and distribution in SAM models  109 of the existence of progressive inclinations in the mechanisms of income distribution. The fact that the actual distribution of income in the ten countries is worse than what the income multipliers indicate must be the result of regressive patterns of the actual exogenous impulses.

7  Strategic choices for growth with redistribution It was demonstrated above that the SAM is a framework that allows for an assessment, in simplified ways, of the trade-offs between economic growth and income equality under different combinations of policy instruments. The assessment gains more rigour when the SAM classifications of the economic activity sectors (where economic growth applies) and the SAM classification of household groups (where income distribution is the focus) are done in ways that reflect and maximise the strength of linkages between the two classifications. More specifically, in the process of economic development there are the high-productivity sectors that lead the economy into higher economic growth, and there are the low-productivity sectors that lag behind. The first tends to correspond with the formal or modern sectors; the second contains what are generally known as traditional and informal sectors. Households whose earnings are coupled to the more productive sectors become better off, while households whose incomes come from the work they do in the low-productivity sectors will get relatively less income. This relative independence of the two segments is the essence of the growth-equality relationship in the Kuznets Curve in the early processes of economic development. But there are also dependence relationships between the two segments that are dubbed as ‘trickle down’ and ‘evaporation upwards’ which mitigate the growth–equality relationship. The simplest way to highlight this mix of independence and dependence is by reclassifying the SAM so as to distinguish between (better-off) households working in the modern and formal sectors (the more productive sectors, (MPS)), and (low-income) households working in the traditional and informal sectors (the less-productive sectors (LPS)). An assessment of the growth and distribution problem along these lines of classification would illuminate the basic linkages that characterise the problem. This section will develop the SAM model to give insights to the magnitudes involved, taking the case of one of the larger developing countries considered, Indonesia. The validity of the results for other developing countries as a whole and for current years is not in doubt, as witnessed by more recent studies.6 To increase our focus we regrouped activities into a formal or modern sector that depicts the MPS, and an informal or traditional sector that depicts the LPS. Household groups were regrouped into the richer household group, RHG, and the poorer household group, PHG. Other accounts in the SAM were left untouched. Table 5.9, applied to Indonesia, shows that an injection of a unit, of say a million rupiahs, in the LPS gives a potential increase in total output of 4.9, two-thirds of which benefits the LPS itself, and one-third benefits the MPS. The injection generates a total income of 2.3 units, but only a minor share of this income, about 23 per cent, goes to poorer households, PHG, and 77

110  Growth and distribution in SAM models per cent goes to richer households, RHG. This distribution of the multiplier effects is practically the same as the actual income distribution between the two household groups. The GLI is 1.0. If the injection is done in the MPS, the total output and income effects would be less: 3.6 and 1.6 respectively. How are these effects distributed over the sectors and households? The LPS will benefit relatively more than the MPS, but the relative distribution of benefits on the household groups will not be affected. Although these results show that the relative distribution of incomes between poor and rich households is neutral with respect to expansion of either the MPS or LPS, the absolute levels of the income multipliers are higher in the case of LPS injections than MPS injections, and this would imply that more households are likely to cross the poverty line upwards. The policy implications are clear. First, the stimulus to LPS brings higher economic growth than does the stimulus for MPS. The fact that in the real world the opposite is happening must be due to various exogenous injection stimulus that favour MPS over LPS. Second, while stimulus of MPS benefits PHG, stimulus of the LPS benefits PHG more. Supporting the less-productive segment is thus more able to reduce poverty, but the relative income distribution remains unaffected under both cases. Table 5.9 also gives the output and income effects of income transfers to households. Government or ROW can make the transfers. Results show that Table 5.9 Impacts of spending injections and income transfers on growth and distribution: Indonesia Actual dispersion per cent (1975)

Impact of spending Impact of income transfer injection in sectors (a) to household groups(b) LPS

HPS

Poorer PHG Richer RHG

Output m Multiplier   4.9   3.6   4.2 Percentage distribution of the output (multipliers) on sectors: LPS  36.1  65.8  53.6  62.0 HPS  63.9  34.2  46.4  38.0 All sectors 100.0 100.0 100.0 100.0 Income multiplier   2.3   1.6   3.0 Percentage distribution of the income (multipliers) on household groups: Poorer, PHG  23.5  22.6  23.7  49.5 Richer, RHG  76.5  77.4  76.3  41.5 All households 100.0 100.0 100.0 100.0

  3.9  39.3  60.7 100.0   2.8  15.3  84.7 100.0

Source: Biro Pusat Statistik Indonesia (1982). (a) The low productivity sector, LPS, is approximated by combining the analytical results of traditional agriculture and small-scale manufacturing to represent the informal sector, and taking the rest of the economy to represent the high-productivity sectors, HPS (including among others commercial agriculture and large-scale manufacturing). (b) Poorer household groups, PHS, includes rural landless farm workers and urban lower-income earners. Richer household groups, RHG, includes land owners and upper-income earners in rural and urban areas.

Growth and distribution in SAM models  111 a transfer to the PHG creates more output and income than does a transfer to RHG. The reasons lie in the fact that poorer households consume and spend most of the incomes they receive and, given their numbers, these incomes reenter the circular flow more extensively and intensively than in the case of the richer households. There is a greater leakage in the latter case. The results show also that, in contrast to the policy of sector injections, the actual income distribution between the two household groups will be significantly affected by the direction of the income transfer. The PHG in Indonesia of the eighties appear to end up withholding back in their own group about 50 per cent of the income transfer, which represents a higher share than their actual income distribution of about 24 per cent. Therefore, the ‘evaporation upwards’ effect appears to be relatively weak. In similar ways, the results show too that the ‘trickle down’ effect is also low, with only 15 per cent of a simulated transfer to RHG trickling down to PHG, the substantial portion, of 85 per cent, remaining with RHG. Of course, while it is possible that these magnitudes may be approaching each other over time, the gap is too wide to be reversed in a couple of decades. Finally, the results validate the policy conclusion that for achieving economic growth—with poverty reduction and, possibly, a more equitable income distribution—such instruments as transfers to poor households are more effective than are those of injections in production activities.

8  Discussion of scope and limitations In some sense, any consistent economy-wide model must be congruent with some accounting matrix that frames the model. The two models in the previous chapters also had social accounting matrices that came close in specification to the SAM of this chapter, even though they were not explicated. The important difference between the previous models and the SAM type of model here is that by specifying a compact SAM, the resulting model is kept compact and, nevertheless, can reflect on a wide range of issues in development policy. The SAM framework has more contributions. First, the SAM serves as a helpful tool in setting up an integrated statistical system of accounts for differentiated agents in the economy. The advantage of forcing national statistics into a SAM is that the statistician can discover inconsistencies and gaps, which went unnoticed before. Second, SAM facilitates the initialisation of corresponding economywide models and, in particular, CGE models. Third, once a SAM is available it can be used to give a quantitative diagnosis of the structure of the whole economy. Fourth, the availability of SAMs for different countries and for more years allows for the investigation of economic mechanisms behind superior and inferior performances of the countries. Any analytical framework has its limitations along with its contributions. Several limitations of the SAM, and counterpart arguments, are in place. First, cell entries of the SAM are amounts that are products of prices × quantities. However, quantities and prices are not explicitly disentangled. In the fixedprice multiplier model, supplied amounts are supposed to adjust to demanded amounts. They will, but if there is restricted capacity the result is inflation.

112  Growth and distribution in SAM models This may require a revision downward in the real sizes of multipliers. Of course, if the size of the injection is relatively small, which is usually the case, the fixed-price multiplier results can still be seen to represent realisable quantity effects. It is also feasible to check in a simple way within the SAM framework whether the capacity limits will be violated or not. The supply side can simply be modelled as a relationship between the investment rate and economic growth via an incremental capital-output ratio, k, as in K/X = k (∆X/X). From the SAM, we obtain multiplier effects for K and X. If division of the multiplier effects of K by those of X gives values equal to or above K/X for the base period, this implies that the SAM solves for sufficient investment to meet the projected capacity increase. Second, it is commonly perceived that a SAM-inverted model belongs to the prototype of demand-oriented models. However, under general equilibrium conditions the SAM can be seen as a representation of both the demand and the supply side as well, which is why in CGE the SAM is directly used to initiate CGE models. Third, the SAM model describes an endogenous economy with fixed relative prices and complementarities—based on production and consumption functions. Producers and consumers are assumed to face fixed prices, and in their pursuit of profit and utility maximisation, respectively, adjustment takes the form of changes in quantities supplied and demanded. With regard to the assumption of producers and consumers facing given relative prices, this is common practice in short-run models. Moreover, even in the longer run, having in mind the broad categories of sectors and products in the SAM, analysts can draw on empirical evidence over long periods which supports indefinite shifts in relative prices between such broad categories, cf. Bleaney and Greenaway (1993). Fourth, there is the shortcoming that the SAM is a static model and its reflections are valid only for the year for which it is observed. Since more SAMs can be quantified for more years the static shortcomings may need to be assessed in relative terms. Fifth, it is noted that the coefficient matrix in the SAM model, As, is a matrix of fixed average proportions. Linear relationships are assumed throughout: constant shares of factor remunerations in total output, of household incomes in the various factor payments, of commodities in household expenditure, and of sectors in commodity production. Compared to averages, observed marginal coefficients are better, since they incorporate income and scale effects, but they can be disputed as their estimated values may carry other than income effects, which is inconsistent with the SAM framework. While consumption propensities can be calculated sensibly as marginal instead of average values, the problem is severe for input-output coefficients and factor-earnings coefficients, and other coefficients in the model which do not usually depict stable marginal propensities. Taking a proportion of the coefficients as marginal and the rest as average propensities introduces an estimation bias. Under the circumstances, many analysts of cross-country data would prefer assuming and using uniform fixed coefficients systems over incomparable specifications for individual countries.

Growth and distribution in SAM models  113 Sixth, the size of the multiplier depends to some degree on the level of aggregation. This argument is less relevant in the context of a uniform aggregation for the compared countries. The differences in multipliers due to alternative aggregations were tested for the ten countries studied. The differences do not go beyond the 8 per cent for individual countries, which is a moderate margin of uncertainty. Last, and not least, evaluation of the multipliers of the SAM model cannot be done in isolation from the closure rules applied. The size of the multipliers depends on the choice of the exogenous and endogenous variables, which, in turn, depends on the problem studied. However, for the short-term horizon, there is a tacitly accepted rationale among most economic modellers to consider various types of government spending and variables of the ROW as exogenous. This rationale was postulated in Koopmans and Montias (1971), and is particularly relevant in the early phases of economic development where decisions on investment are dominated by government.

6 Simplified statics and dynamics in the CGE model Parameterisation and simulations for Indonesia

1 Background This chapter demonstrates how a simple static computable general equilibrium (CGE) model of the Indonesian economy can be constructed and parameterised using the social accounting matrix (SAM) as a database. It will also convert the constant labour supply and capital stock in the static model to dynamic specifications allowing formulation and solution of a dynamic version of the CGE model. This being the first chapter in the book that applies CGE models, it is natural to open with a short review of the development of the CGE model as a major tool for analysis and policy making. As is common for all well-established analytical tools, the CGE model has also gone through various phases before it became institutionalised: discovery, launching, extensive applications, and more recently, retrospective evaluations. The discovery of the CGE as an operational model that can be readily calibrated for the economy-wide circular flow can be traced back to Johansen (1960), where he presented a multisector general equilibrium model for the Norwegian economy. At the time, econometric models based on time-series data and sometimes combined with multisector models (that is the CEM model) formed the mainstream. But Johansen’s work was not immediately noted; it was only later, in the mid-seventies and the eighties, that the launching phase of the CGE model took place. A limited number of pioneers, mostly advising in the context of developing countries, together with the active support of the World Bank, set the scene for the first generation of CGE models. There is a recurring controversy on who was first, second, or third in leading the launching phase. The launching can also be perceived as a collective effort at mapping a barely discovered territory at the time. Each contribution added something that was missing in the map. The following list of contributors is illustrative of the collective effort. Making use of neoclassical theoretical foundations, Adelman and Robinson (1978) provided perhaps the first CGE models for a developing country, Korea. A somewhat different direction followed in structural thinking—which lays more emphasis on socio-political institutions—also had its impact on the launching phase of CGE models. The structuralist approach is apparent in Lysy and Taylor (1977), who developed and applied their first CGE model for an analysis of income distribution in Brazil.

Simple statics and dynamics in the CGE model  115 At about the same time, the World Bank initiated country studies on structural adjustment, trade, and distribution that made use of CGE models with elements from both neoclassical and structural thought. The work by Dervis et al. (1982) is in part an outcome of the World Bank initiative. In the next few years, tens of CGE models for developing countries followed. For a survey see Decaluwe and Martens (1988). The launching phase of the CGE model was a success story. Other factors that helped in elevating the CGE model to the mainstream were the preceding experience gathered in constructing SAMs, essential for calibrating the CGE models, the advances made in quick computation, the recognition that endogenous prices are essential in economy-wide models, and the flexibility of the CGE modelling framework in incorporating modifications and extensions. The contributions mentioned above are incomplete without due mention of works by Scarf (1969) and Scarf and Shoven (1984), which aimed at developing the links between the theoretical framework of general equilibrium, applied general equilibrium models, and the empirical CGE models as we know them today. The extensive application phase that started from around 1990, and which has continued expanding ever since, is overwhelming. CGE models proved to be very useful tools for the analysis of many policy areas including tax reforms, international trade, economic integration, environment policy, resource allocation, income distribution, regional development, financial policy, and so on. A compact review of the applied extensions of CGE models is impossible due to the rapid and extensive expansion of CGE modelling and applications. However, as the next chapters will treat CGE models that focus on some of the just mentioned policy areas, it is natural that we shall review related works. In recent years, but not exclusively, there were retrospective evaluations of the CGE model research programme, as well as efforts to link them with other modelling directions; see for instance, Robinson et al. (1999) and Taylor (1990). In the meantime, the CGE modelling activity has become a worldwide industry with its teaching curricula, software, data banks, graduate dissertations, and advisory teams. The CGE model paradigm has finally become part of the academic establishment. This advance has not gone unnoticed by rebels on the sidelines. Some critical views are justifiable; others are not sufficiently well-founded. McKitrick (1998) and Mitra-Khan (2008) are two such examples. The CGE model in this chapter dates back to the year 1980 and belongs to the launching phase of CGE models. The other CGE models in the book are more recent and would belong to the extension phase. Does it make sense to pay attention to a demonstration model that was developed three decades ago in view of the many advances made in CGE modelling since then? Probably not for some, and it is likely that it is useful for others. In particular, there are some presentation features of the CGE model that were already there—and others that we have added and revised—which highlight alternative formulations of the model and specifications of its working mechanisms. There are elaborations on the links between the (SAM) and CGE models, not only with regard to empirical calibrations, but also regarding shared structures and mechanisms. The CGE model is also reformulated at first in a more elaborate form in correspondence

116  Simple statics and dynamics in the CGE model with contemporary specifications, and in a process of deduction we arrive at the simplified form which was applied then. The static CGE model is also extended to incorporate dynamic paths for labour skill-formation, and capital accumulation. The policy simulations that we apply relate to raising efficiency in alternative production activities, which are important ingredients of economic development, but are often neglected in policy discussions. The choice made to apply the model to Indonesia goes back to circumstances associated with the launching phase of SAMs and CGE models. Various Dutch institutions collaborated with the Biro Pusat Statistik Indonesia in constructing the first SAM for Indonesia dated 1975, made available in 1979, and published officially in BPS (1982). The availability of the SAM for Indonesia was followed by a rush among several economists working on Indonesia to fit CGE models to the SAM. This focus on calibration will be apparent in the next section. But, as was stated earlier, the calibration aspect is only one of several aspects that are treated in this chapter.1 The chapter is organised as follows. In Section 2, we specify the static CGE model and examine its structure. In Section 3, we estimate the CGE parameters from the available SAM for Indonesia at the time. In Section 4, the model is used to run policy simulations that investigate effects of policy measures to enhance productivity in alternative economic activities. The analysis is done with the object of highlighting the mechanisms of the CGE model. In Section 5, several dynamic aspects will be introduced in the model. In Section 6, the policy simulations are repeated in the context of the dynamic model. Finally, in Section 7, concluding remarks are made.

2  The static CGE model In this section the core of the general equilibrium model is presented. It is a general purpose model, which contains only a minimal number of equations necessary to describe the economic process. It has the advantage that all parameters needed are derived from the SAM. For more specialised applications the model needs to be elaborated, in which case supplementary data are required to estimate additional parameters. Because we are primarily interested in domestic transactions we kept the relations with the rest of the world as simple as possible. Foreign prices are assumed to be equal to domestic prices to keep the model homogeneous in all prices. The foreign exchange rate is fixed. For an elaborate treatment of foreign trade quantity and price variables see this in the next chapter, and in Chapters 10 and 12. The 18 equation sets of the model can be grouped in four blocks. The first block, Box 1, comprises eqs. 1 to 6, which together form the factor market. In eq. 1, sector output is determined from a two-level production structure, and is derived in several steps of sub-equations, ending up in eq. 1. To start with, at the lower level, value added in volume terms, or quantity terms, VQ j , is expressed as the product of a Cobb-Douglas production function with (demanded) employed factors of labour LDqj , and capital KDj , with βqj standing for production elasticities relating to labour type q and activity j, and (1 − Σq βqj ) denoting

Table 6.1 Notations Indices: The index j represents sectors, h stands for institutions which can be a specific household group, firms or government, q stands for types of labour, and r is used to denote actors abroad, that is, rest of the world. Endogenous variables Chj Value of consumption expenditure of institution h on commodities od activity j Value of total consumption expenditure by activity j Cj Ej Value of exports of activity j Foreign capital flow expressed in foreign currency FCF LDqj Demand for labour of skill qualifications q in activity j KDj Demand for capital in activity j GBD Value of government budget deficit Value of installed investment I ICH Value of inventory change Value of total competitive imports Mj Remuneration rate of capital in activity j KRj LRq Remuneration rate of labour of qualification type q Value of the gross product of activity j (that is gross value added) Vj Xj Value of the gross output of activity j XPj Price index of the gross output of activity j, whereby XP1(agriculture) is taken as numeraire Yh Disposable income of households h. Index h is extendable to include the category of firms Zh Value of total incomings of households h. Index h is extendable to include the category of firms Value of total revenue of government Zg Exogenous variables Cgj Government consumption expenditure on activity j Foreign exchange rate FXR Government expenditure on installed investment Ig Private sector demand for installed investment Ipsi Value of capital stock in activity j KSj LSq Labour supply of qualification type q Tgh Government transfer payments to household group g, extendable to firms Coefficients and data ajo Calibration constant in the production function of activity j αjj’ Input-output delivery coefficients between activities j αρj Non-competitive imports share in the output of activity j in benchmark year βqj Labour elasticities of production relating to labour type q and activity j εjo Quantity of exports as proportion of the quantity of output of activity j in benchmark year εj Price elasticity of demand for exports of activity j γhj Consumption budget shares of spending institution h on products of activity j continued overleaf

118  Simple statics and dynamics in the CGE model Table 6.1 Notations continued ιj μrj μj πjh , πjg ϖjr ρhh; τhg τjg υj ωqh

Delivery share of installed investment by activity j in the total installed investment Competitive imports proportion in domestic supply of products of activity j Price elasticity of imports of products of activity j Profit distribution shares of activity j to households h; to government g; and rest of world r Transfer rates of households to households h. The h index is extendable to include firms Rate of direct tax paid by households h to government g; the h index is extendable to include firms Rate of indirect taxes on output of activity j collected by government g Inventory change of activity j as proportion in total inventory change Earnings distribution shares from labour type q to households h

production elasticities relating to capital in activity j. Thus, smooth substitution possibilities are assumed in the Cobb-Douglas production function, expressed by VQ j = αjo LDj βqj KDj (1 − Σqβqj). At the higher level, sector value added in quantity terms, VQ j, is a fixed proportion of the sector output in quantity terms, XQ j, after the deduction of domestic and imported intermediate input coefficients and indirect tax rates, following Leontief, thus: VQ j = (1 − Σj (αj′j − αrj − τjg ) XQ j. By definition, nominal gross output, Xj , equals output quantity, XQ j , times the output price, XPj; thus: Xj = XPj . XQ j. Substitution of the three sub-equations into each other gives the final form of eq. 1, displaying the generation of sector gross output in terms of the sector price and parameters of the two-level production structures. The further elaboration and solution of the model make no more use of the two specific variables of XQ j and VQ j. For the sake of simplification, these two variables are purposely omitted from the list of notations. The outcome of the above handlings is eq. 1. In eq. 2, making further use of Leontief coefficients, the value added by sector, Vj, is found by subtracting the cost of domestic intermediate deliveries, imported intermediate deliveries, and indirect taxes from gross output by sector, appropriately priced. In eq. 3, producers are assumed to maximise profits under perfect competition in product and labour markets; as a result, in the short run they will hire employees until wages equal the value of their marginal product. Eq. 4 shows the remuneration of capital as the remainder of nominal value added after labour is paid its share. Because the volume of capital is fixed in the short run, the remuneration rates of capital may differ among sectors. Eq. 5 states that the demand for (employment of) labour by skill level is equal to the supply of labour by the corresponding skill type. The latter is exogenously fixed in the short run. At full employment, wages carry the burden of adjustment.

Simple statics and dynamics in the CGE model  119 Eq. 6 states that the demand for (employed) capital is equal to its supply for each activity level j. The latter is given in the short run. Here, too, remuneration of capital in each activity adjusts to the particular demand and supply of capital in the relevant sector.

Box 1  Production function, factor market balances Xj = XPj {1 / (1 − Σj (αj′j − α rj − τjg )} [αjo Π LDqj βqj KDj (1 − Σq βqj)] Vj = (1 − Σj αj′j XPj′ / XPj ) Xj − α rj Xj − τjg Xj LRq . LDqj = βqj Vj KRj . KSj = (1 − Σqβqj ) Vj Σj LD qj = LSq KDj = KSj

j (1) j (2) q.j (3) j (4) q (5) j (6)

The next block, in Box 2, presents the formation of the institutional accounts of household groups (and firms)2 denoted by index h, and government, indicated by g. In eq. 7 the total income of recipient household groups (and firms) is specified. Factor remunerations are distributed over household groups h via wage and profit shares ωhq and πh. Income transfers paid to and received from other household groups are fixed proportions, ρhh′ , of the income of the paying party. The equation incorporates the net results of these transfers. Furthermore, net income transfers are received from government, Τgh. (There are also net transfers received from rest of world, but these are not reported in the current SAM of Indonesia; they are left out.) In eq. (8) direct taxes are subtracted from the incomings of households and firms to arrive at disposable income. Eqs. 9 and 10 specify consumption expenditure by private institutions and government. This can be done in great detail, but under some assumptions it can be compressed in an abridged form. Suppose that the household group h spends their consumption expenditure on consumer commodities denoted by c, which is the usual case. In eq. (i), this consumption by h of c, denoted by Chc, is dependent on the relative prices of commodities, c, denoted by SPc, and on the disposable income of household group h, Y h, and on the applicable own and cross-price elasticities and income elasticities. The sum of the elasticities adds up to unity, in conformity with the Cobb-Douglas functional relationship that is also used above in the production sphere. Hence we have the specification of the income elasticity as (1 − Σc γhc ). The prices are taken relative to one of the commodities or a package of commodities c. In this equation, γhc are the start budget proportions spent on c, γ'hc are the price elasticities, and γhco is a calibration coefficient. In eq. (ii), the realised consumptions classified by c are converted into consumption by activity j, making use of converter coefficients θjc, and dividing by SPc to obtain quantities that are again converted into value in terms of activity prices XPj. Government consumption by sector j is added to give consumption demand by sector.

120  Simple statics and dynamics in the CGE model Chc = [γhco Πhc SPc γhc { γhc Y h } 1 − Σc γhc] Cj = XPj Σc θjc (Σh Chc / SPc ) + Cgj

h. c (i) j (ii)

The first equation can be cut down in size if it is assumed, for the sake of demonstration, that the price elasticities equal zero, so that the income elasticity would equal 1.0, and the distribution of spending on commodity c will be consistent with the initial budget shares γhc. The price effect disappears and Chc becomes proportionately dependent on Y h. Furthermore, both equations are greatly simplified if it is assumed that the classification of commodities c corresponds exactly with activities j, which is common in many SAMs that lack data to do it otherwise—the SAM of Indonesia is one such case. When θjc takes the value of 1 in the diagonal and zero otherwise, SPc and XPj disappear from the equation. Substitution of these assumptions into the two equations results in the simplified form of eqs. 9 and 10 which will be kept in this model. Eq. 11 specifies the demand for installed investment3 by the private sector and government. As the possibility of specifying an investment function for private investors was not feasible, the formulation had to choose to fix the whole investment level autonomously at its known value for the benchmark year. This is sufficient for solving the static model for the base year. A forecast of I is required for future years when the model is run in its dynamic version. The forecast will be based on semi-autonomous grounds, as will be elaborated in Section 4.

Box 2  Income and spending of private sector and government Zh =Σqωqh LRq . LSq + Σj Πjh KRj . KSj + Σ h′ ρhh′ Zh′ + Tgh Y h = (1 − τhg ) Zh Chj = γhjY h Cj = Σh Chj + Cgj Ij = τj (Ipsi + Ig ) Zg = Σj πjg KRj . KSj + Σhτhg Y h + Σj τjg Xj Σj Cgj + Σh Tgh + Ιg = Ζg + GBD

h = 6 groups, firms (7) h = 6 groups, firms (8) h.j (9) j (10) j (11) (12) (13)

In eq. 12, government revenue consists of factor income from ownership of public enterprises, plus direct taxes from households and firms, plus indirect taxes. (There are also net transfers from abroad but the current SAM of Indonesia does not show them; we leave them out.) In eq. 13, government current expenditure consisting of consumption and transfers, plus government fixed expenditure, that is investment, Ig , are put in balance with government revenue, Zg , and a government budget deficit, GBDg. The third block, in Box 3, relates to demand, supply, and price formation in the product markets. Preliminary equations are specified in Roman numbers, after which simplifications are applied to the preliminary equations with the object of highlighting the role of price formation in the CGE model, and reducing the model to a minimum number of equations. The model still remains a CGE model,

Simple statics and dynamics in the CGE model  121 but with less reliance on prices in clearing the product markets. In eq. (iii), the quantity of exports of activity j is initially specified as the quantity of exports for the benchmark year εjo (Xj /XPj) and is further made dependent on relative prices of domestic production XPj , a fixed foreign exchange rate and foreign exports prices, FXR . epj, and a price elasticity εj. Ej / XPj = εjo (Xj /XPj ) (FXR . epj / XPj )εj

j (iii)

In eq. (iv), the quantity of competitive imports is initially specified as a proportion μrj of the quantity of consumption and investment, and is further made dependent on relative domestic and foreign prices, fixed FXR, and a price elasticity, μ j. Mj / XPj = (μrj (Cj + Ij) / SPj ) (XPj / FXR . mpj )μ j

j (vi)

In eq. (v), the composite price of sector j is a weighted average of domestic and import prices. SPj = μrj (FXR . mpj  ) + (1 − μrj ) XPj

j (v)

The first basic simplification in this block is to assume that FXR is fixed and prices of export j, import j, and activity j are equal; this is feasible and is realised in various situations, that is eq. (vi) epj = mpj = XPj

j (vi)

When this assumption of equal prices is substituted in the above equations, the simplification would eliminate the effects of relative prices and the foreign trade elasticities in influencing foreign trade streams and clearance of the product markets. The supply composite price will be equalised to the sector’s product price, making the supply composite price a redundant variable in this simplified model. Although the model keeps all features of the CGE model, the effect of the simplification is to reduce the role of prices and increase the role of quantities in balancing the supply and demand of the product market. The equations are reduced to take the form of eqs. 14 and 15, in Box 3. It is noted that the foreign trade equations do not make explicit a specification for the foreign exchange rate. Assuming a fixed foreign exchange rate is done on purpose. This is a second basic simplification that is representative and consistent with situations of fixed foreign exchange regimes. If the model had focused on foreign trade in an open economy, with free operations with the rest of the world, the specification of an endogenous foreign exchange rate would be the obvious choice. Most open economy CGE models specify an endogenous foreign exchange rate, but not necessarily, as will be elaborated in chapters 8, 9, 10, and 12. Finally, because the model can only determine relative prices, a price normalisation rule is required. It would have been possible to make use of a consumer price index (CPI) as numeraire, but the required data on commodities were not separable from those on activities. This being the case at the time, we chose the agricultural price for numeraire, as specified in eq. 16. The agricultural price level can be considered a good approximation for consumer prices in situations

122  Simple statics and dynamics in the CGE model where food commodities take up the largest share in consumption expenditure. All price and nominal variables are thus defined relative to the price in agriculture. (Despite choosing the agricultural price as numeraire, we compute later on, when we analyse results of policy simulations, a CPI for an approximate evaluation of changes in real welfare, cf. Section 5.)

Box 3 Foreign trade and product markets Ej = εjo Xj / XPj Mj = μrj (Cj + Ij ) XP1 = 1.0

j

(14) (15) (16)

The fourth block, in Box 4, consists of national accounting balances. Eq. 17 gives product market balances that assure equilibrium in the product markets. For each activity j: the supply side consisting of the produced output, Xj, supplemented by competitive imports, Mj, is equal to the demand side, consisting of delivery of intermediate goods appropriately converted to the applicable price index of the delivering activity, XPj, final consumption goods delivered by j, the part of investment goods delivered by j, inventory change occurring in j, and exports of j. Which variables are the adjustors of the product balances? These must be j in number. Domestic prices of the activities, XPj, in total (j – 1) in number, carry the burden of adjustment between the supply and demand by sector, together with the variable of inventory change for the economy as a whole, ICH. (See also the discussion on adjustors in Chapter 2, Sections 2 and 3.) In the equalisation of supply and demand in the product markets, XPj represents the adjustments at the specific product level, while inventory change, ICH, represents adjustment at the macro level of the economy. Eq. 18 specifies the savings investment balance, which can also be called the financial market balance. It states that total savings (the sum of private savings, government savings, and foreign savings) is equal to total investment (that is fixed installed investment plus inventory change). While private savings are positive, government savings are usually negative reflecting a budget deficit. Net foreign savings are net foreign capital flows converted to national currency, FXR . FCF. The gap between total investment and domestic savings is filled by foreign capital flow, FCF; which takes the role of adjustor. Eq. A1 is the foreign payments balance. It equalises outgoing foreign payments (these are competitive imports and imported intermediate inputs) to incoming foreign payment (these are exports, and foreign capital inflow, FXR . FCF). The model, consisting of equation sets 1 to 18, is a determinate model that solves for all endogenous variables as enumerated in the list of notations, also counting 18 sets of unknowns. The solutions so obtained will reproduce the consistency of the foreign payments balance automatically, hence denoting this accounting balance as an additional equation that is laid outside the model.

Simple statics and dynamics in the CGE model  123 Box 4 National accounting balances Sector product balances Xj + Mj = Σ j (α jj’ XPj / XP j′ ) Xj′ + Cj + Ij + υj ICH + Ej

j (17)

Savings investment balance Σh (Y h − Σj Chj ) + (Zg − Σj Cgj − Σh Tgh ) + FXR . FCF = I + ICH (18) Foreign payments balance Σj Μj + Σj arj Xj + Σj Πjr KRj . KSj = Σj Ej + FXR . FCF (A1)

If the structure of this CGE model is displayed following the scheme of causal ordering in Simon (1953),4 then it will be seen that it corresponds almost exactly with the structure of the basic CGE model described earlier in Chapter 2, Table 2.3. The adjustors in the factor markets are the remuneration rates of the factors of production. The adjustors in the product markets consisting of j activities are the price levels XPj of j − 1 activities (noting that one of the activities serves as numeraire), and the inventory change for the economy as a whole. This closure in the clearance of the product markets was denoted as closure A. The adjustor in the savings investment balance is foreign capital flow, FCF, which was denoted as closure C.

3  Parameterisation of the CGE model A SAM is a prerequisite for calibration of the CGE model. The SAM is a statistical presentation of the circular flow and of transactions between agents, which the CGE model specifies in equation form. The object of this section is to demonstrate the basic links between the primary form of CGE models and the SAM, and the use of the SAM for calibrating the CGE model. A supporting description of the ways in which a CGE model and the SAM are linked to each other is in Robinson et al. (1999). The 1975 SAM of Indonesia, included in Annex Table 6.7 (at the end of this chapter), distinguishes between three factors of production, six household groups, firms and government, and a capital account, activities account (commodities correspond with those in the four sectors of activities, namely: agriculture, industry, trade, and services). Furthermore, there is the rest of the world (ROW) account. The CGE model is organised on the same accounts of the SAM, and the CGE equations follow the same order as the rows of the SAM. The factor accounts come first, institution accounts are next, followed by institutions’ spending on consumption (commodities account) and savings (capital account). Next are the activities and the ROW accounts. Described otherwise, the last three rows of the SAM are the three national accounting identities that correspond with eqs. 17, 18, and A1 in the CGE model. In both the SAM and the CGE model, the row of the so-called ROW, and eq. A1 standing for the foreign payments balance, represent the same entries. Neither needs to be counted in

124  Simple statics and dynamics in the CGE model the system, because the entries in them are automatically fulfilled if (a) all other rows and columns in the SAM are correctly inserted, and (b) similarly for the CGE model, if n − 1 connected markets are solved, clearance of the nth market is self-evident. What was stated in Chapter 2 can be repeated here. Both the CGE and the SAM incorporate the three equilibrium balances that are part and parcel of economy-wide models. One is the product market equilibrium (for commodities or activities) as represented in the CGE model by the set of equations 17, and by the activities account in the SAM, rows 14 to 17. Second is the savings investment balance representing the financial market equilibrium, eq. 18 in the CGE model, which corresponds with the capital account in the SAM, row 17. Third, is the foreign payments equilibrium in eq. A1 in the CGE model and row 18 in the SAM. The parameters of the CGE model are derived from benchmark values of nominal variables in the SAM under the assumption that the benchmark SAM represents the ex post equilibrium situation. Furthermore, because price variables in the CGE model are expressed as indices, and they take the value of 1 for the benchmark, the values of variables and their relationships in the SAM are interpretable in quantity (real) terms. The estimates of the parameters should be such that calibration of the CGE model reproduces the same SAM. The SAM of Indonesia is in Annex Table 6.7 at the end of the chapter. The first block of the model contains input-output coefficients and factor elasticities of production. The coefficients of input-output deliveries, non-competitive imports, and indirect taxes: αjj′, a rj, and τjg, respectively, are calculated simply by dividing the appropriate cells in the SAM by their column totals, giving their estimates in Table 6.2. The labour elasticity βqj is calculated by dividing the factor payments,5 LR1, LD1j, by the value added, Vj. Furthermore, by defining sector prices as indices, as described above, Vj / XPj becomes equal to Vj, which is readily obtainable from the SAM, and is thus inserted in eq. 1. From the assumption that capital supply, KS, is equal to capital in use, KD, the benchmark values of KSj , are estimated at 10,541 for agriculture, 13,552 for mining, industry, and construction, 6894 for trade and transport and 3112 for services. When all these data are inserted in eq. 1 the calibration coefficients, αj , for the four sectors are obtained as shown in Table 6.3. The resulting calibration constants are close to 1 for agriculture and trade and transport, nearly 3 for services, and slightly above 0.5 for industry. Turning to the second block of the model, the coefficients that distribute factor income over households, ωhq and πh in eq. 7, are calculated by dividing the applicable cells in the factor account columns of the SAM by the column totals. As can be expected, the estimates in Table 6.3 show that low-skilled labour income accrues mainly to low-income households. High-skilled labour income goes mainly to urban upper-income groups. Recipients of capital income are mainly landowners and firms. Transfer coefficients between institutions, τhh’, defined as the ratio of outgoing transfers to income of the transferring institution, are calculated by dividing the respective income transfers by their respective column totals. Transfers are

Simple statics and dynamics in the CGE model  125 Table 6.2 Coefficients relating to sectoral production functions; eqs. 1, 2, Indonesia Coefficients SAM columns rows

18. agriculture

19. industry

20. trade

21. services

Production function level coefficients 1. low-skill earnings β1j 2. high-skill earnings β2j 1 − β1j − β2j 3. profits Calibrated coefficient αjo

0.4353 0.0045 0.5602 0.9154

0.1738 0.0203 0.8059 0.5635

0.4088 0.0550 0.5362 1.1791

0.1526 0.4389 0.4085 2.8154

Input output level coefficients 12. Gov.-indirect taxes τjg 14. Agriculture α1j 15. Industry α2j 16. Trade α3j 17. Services α4j 18. Rest of world μj

0.0097 0.2867 0.0227 0.1509 0.0062 0.0255

0.0179 0.0370 0.1709 0.0961 0.0124 0.1593

0.0099 0.0729 0.0719 0.0181 0.0458 0.0637

0.0033 0.0100 0.0577 0.0148 0.0177 0.0334

Table 6.3 Coefficients of factor income distribution to institutions, ωqh , πh; eq. 7, Indonesia SAM columns rows

Low-skill earnings ωih

High-skill earnings ω2h

Profits ϖh

  4  Rural farm owners   5  Rural farm workers   6  Rural non-farm upper   7  Rural non-farm lower   8  Urban upper income   9  Urban lower income 10 Firms 11  Government general

0.3908 0.1057 0.0492 0.1979 0.0773 0.1790 0.0000 0.0000

0.0574 0.0301 0.2361 0.0847 0.5127 0.0790 0.0000 0.0000

0.2433 0.0162 0.0104 0.0437 0.0403 0.0805 0.4705 0.0073

dominated by flows of funds from firms to households and government, for which transfer rates are 0.122 and 0.377, respectively; Table 6.4. Coefficients of institutions’ expenditures on consumption by activity, γhj, generally known as budget shares, are calculated by dividing consumption expenditures for the various commodities in the institutions columns by disposable income (1 − τhg) Zh. For households, the largest share of income is spent on agricultural products. Government spends nearly half of its income on services, Table 6.5. The Indonesian data at the time did not classify commodities differently from sectors. As can be noted in the SAM, Table 6.7, sectors and commodities correspond to each other. The product market can thus be formulated throughout in

126  Simple statics and dynamics in the CGE model Table 6.4 Coefficients of income transfers between institutions, ρhh′ ; eq. 7, Indonesia SAM columns rows

R farm R farm R nonowners workers farm upper

R non- Urban Urban Firms farm upper lower lower

  4  Rural farm owners   5  Rural farm workers   6  Rural non-farm upper   7  Rural non-farm lower   8  Urban upper income   9  Urban lower income 10 Firms 11  Government general

0 0 0 0.012 0 0.001 0 0.011

0 0 0 0.009 0 0.001 0 0.004

0 0 0 0.003 0 0.002 0 0.003

0 0 0 0.022 0 0.003 0 0.008

0 0 0 0.002 0 0.043 0 0.024

0 0 0 0.001 0 0.028 0 0.016

0 0 0.023 0.009 0.122 0.011 0.008 0.377

Table 6.5 Coefficients of institutions’ expenditure on commodity consumption γhj ; eq. 9, Indonesia SAM columns rows

R farm R farm R nonowners workers farm upper

R non- Urban Urban Firms farm upper lower lower

13 Agricultural 14 Industrial 15 Trade 16 Services

0.665 0.096 0.074 0.074

0.634 0.141 0.115 0.094

0.766 0.112 0.100 0.087

0.557 0.169 0.098 0.097

0.341 0.224 0.141 0.149

0.434 0.187 0.185 0.138

0.000 0.000 0.000 0.000

terms of the produce of activities j. It was stated earlier in the description of eqs. 9 and 10 that the model is substantially simplified by working with one set of prices for the activities; that holds also for commodities, instead of having two separate sets of prices for commodities and sectors, which is more usual in CGE models. Among the government parameters in eq. 12, direct taxes are computable from the intersection of rows 4 to 10 with column 11 in the SAM. Indirect taxes are computable from the intersection of rows 13 to 16 with column 12 in the SAM. Other data on exogenous government consumption, Cgj , and transfers, Tgh , are directly obtainable from column 11 in the SAM, Table 6.7. Foreign trade parameters in eqs. 14 and 15, relating to exports, εjo , and imports, μrj, are derived from the ‘Rest of the World’ column and row. The coefficients allocating investment goods to sectors of origin, ι, which show up in eq. 11, are also simplified, since the sector of industry ( j = 4) is the sole deliverer of investment goods, meaning that ι4 = 1 and ι1 = ι2 = ι3 = 0. Finally, the SAM for the benchmark year does not specify the inventory changes ICH; these are summed together with the installed investment I. Since

Simple statics and dynamics in the CGE model  127 the model is calibrated for the benchmark year with the given value of I the solution of the model for the base year for IHC is, by definition, also zero. However, this need not be so for years beyond the base year, as then the model will be solved in its dynamic version. As an aggregate adjustor of supply with demand in the future, IHC can take positive or negative values. To allow for distribution of the IHC on the sectors, we defined υj as the shares of the output of sector j in the total output of all sectors, and we fixed their estimates from the benchmark data.

4  Static policy simulations In this section we demonstrate the functioning of the 18 equation model by simulating two supply impulses to economic growth. The results will be reviewed with the object of highlighting the mechanisms of the model, and the need for it to be cautiously interpreted as being demonstrative and specific to the simplified data base employed for Indonesia. Besides, it is important to take into account the generality of the assumptions and the absence of qualitative structural characteristics of the Indonesian economy in the CGE model. The two simulations are: (1) a 10 per cent productivity increase in industry and construction; and (2) a 10 per cent productivity increase in services. Simulation 1: 10 per cent productivity increase in industry and construction The efficiency parameter in the production function in industry is increased by 10 per cent. The result is an 8.78 per cent increase in the volume of output. However, because demand is inelastic,6 the favourable volume effect is reversed by a −10.01 per cent price decrease, resulting in a reduction of value added in industry in relation to other sectors. So all factors employed in the sector with the increased productivity undergo a decline in their income. The increased productivity makes more than 5 per cent of the labour employed in industry redundant, while the owners of capital see their remuneration reduced by 5.35 per cent because of the lower value of the marginal product of capital. Because a relatively large share of the national capital stock is installed in industry, total remuneration of capital falls with the reduced value added in industry. The wage rate of high-skilled labour increases slightly, relative to the low-skilled wage rate (0.42 per cent versus 0.21 per cent) because very little high-skilled labour is employed in industry. The change in factor incomes is not large enough to cause a change in income distribution: there is a small fall in income for all groups, varying from −0.65 per cent for landowners to −0.14 per cent for farm workers. The fall in nominal income does not mean that welfare has decreased, however. We have computed a CPI, which is a weighted average of sector price levels, making use of sector shares in total GDP. This CPI has decreased by more than the fall in nominal income. Thus, it can be concluded that purchasing power has increased for all household groups. The welfare increase of the different groups is confirmed by a +2.5 per cent increase in real GDP (the average increase in the volume of value added).

128  Simple statics and dynamics in the CGE model Simulation 2: 10 per cent productivity increase in services As in the previous experiment, the efficiency parameter of the production function is increased by 10 per cent. The resulting production increase in the service sector is +9.17 per cent, which is only slightly higher than the increase in industry production in the previous simulation. However, in this simulation there are hardly any spillover effects to other sectors in which the increase in production does not exceed 0.2 per cent. This is due partly to the fact that in Indonesia the service sector is rather small, and partly due to the lack of backward linkages. The resulting increase in real GDP is less than half of the increase in the industry experiment: +1.19 per cent. Again, as a result of inelastic demand, the value added of the sector with the productivity increase falls relative to the sectors without productivity increases (−2.52 per cent), although the fall is less than in the previous simulation. Because the majority of high-skilled workers are employed in the service sector, the decrease in value added is detrimental for their relative wage rate, which decreases by −0.21 per cent. The decrease in the remuneration for the high-skilled workers leads to a worsening of the income position for non-farm upper-income groups (both rural and urban: −0.58 per cent and −0.45 per cent, respectively). These groups account for a large proportion of high-skilled workers. But as the CPI has decreased by even more, this would allow an increase in real GDP to reach all households.

5  The dynamic model Incorporation of dynamics in the CGE models had proceeded already in the 1980s, once the static CGE model was established as a useful tool for economic analysis and policy making. Economists worked first at extending the static model to a sequence of static equilibria with the object of tracing price changes and quantity reallocations following exogenous shocks and (or) parameter modifications in each sequence. In a later stage, the modelling of dynamics was improved by considering inter-temporal optimisation and forward-looking behaviour, resulting in two different approaches: the recursive equilibrium approach and the full dynamic approach. A classification of dynamic CGE models along these lines is in Dixon and Parmenter (1996). The recursive equilibrium approach assumes that the model is solved for a sequence of static equilibria. The equilibria are connected through accumulation in the factors of production. The approach does not incorporate inter-temporal aspects of decision making. One of the early attempts to introduce dynamic elements into a CGE model for the US economy in Ballard et al. (1985) assumed that consumers have myopic expectations regarding future prices, especially regarding the future rate of return to capital. An example of recursive dynamics with adaptive expectations for the Australian economy is in Dixon and Rimmer (2002). The full dynamic approach derives consumers’ behaviour and producers’ behaviour, as in Knudsen et al. (1998), and Bovenberg (1985), respectively, from both intra- and inter-temporal optimisation. These models are based upon the perfect foresight hypothesis and

Simple statics and dynamics in the CGE model  129 describe the transition path to the new equilibrium point. The models are exploratory and generally require a great computational effort because all of the equations defined over the whole time horizon are solved simultaneously. Their scope is usually short of economy-wide policy models. The dynamic features that we apply in this chapter would fall under the recursive-sequence equilibrium approach. The basic steps to make the model dynamic lie in updating the variables of labour and capital availabilities that so far were kept fixed. The updating can be based on behavioural relationships or on extrapolated growth. Here, this will be done mostly on the basis of behavioural relationships and partly on extrapolated growth. For the extensions, additional notations are required as shown in Table 6.6. The extension of the static to the dynamic model involves adding seven sets of equations to the static model of 18 sets, thus resulting in a total of 25 equation sets. In correspondence with these there are seven newly introduced sets of endogenous variables, as listed in the notations above. These extensions bring the model back into its determinate form, and allow its solution via simple substitutions and matrix inversion. The extensions are summarised in Box 5. Eq. 19 states that the total labour force, consisting of low and high skills, grows annually. (The growth rate that applied to Indonesia was λ′ = 0.025.) Eq.20 distributes labour over the two skill types for the long term along endogenous lines. The skills ratio of high to low skills, RHL, grows over time, influenced by what is generally acknowledged as social demand (or sometimes Table 6.6 Additional notations Newly introduced endogenous variables AKR Idst,j IDSj KSj RHLt + n LS1 LS2

Average profitability of investment aggregated over all sectors, that is, weighted average Nominal investment in destination sector j Investment in destination sector j expressed as proportion in total investment, whereby the following condition is observed Σj = IDSj = 1 Capital stock of sector j Ratio of high-skilled labour (q = 2) to low-skilled labour (q = 1), in a future prospective year t + n Supply of low-skilled labour (q = 1) Supply of high-skilled labour (q = 2)

Newly introduced coefficients Depreciation rate of capital goods δ Educational cost incurred and earnings forgone in acquiring skill ηq Elasticity of inter-sector investment shifts to profitability differentials κ Growth rate of the total labour force λ′ Calibration coefficient in the equation of the skill ratio λo Elasticity of skill upgrading in response to wage differentials λ Distributed lag adjustment for realizing long-term change in the labour ratio φ between low and high skill

130  Simple statics and dynamics in the CGE model called non-economically motivated behaviour), and economically motivated behaviour. Given the close association between the two factors, and especially with regard to occupational education, it is functional to focus on the economically motivated behaviour, which can give a reasonable approximation of the effects of the combined behaviour. The workers’ choice between offering himself (or herself) to high-skilled versus low-skilled employment depends on the relative benefit-cost ratios associated with the two skill types. Benefits are the remuneration of labour, LR 2 , and LR1, respectively. Costs are indicated by η2 and η1, respectively; and they include the education costs incurred in acquiring skill q, and foregone earnings during the formal acquisition of the higher skill.7 The incurred costs of η2 and η1 are interesting policy parameters—they can be varied, and their long-term effects on the structure of the labour force can be evaluated. To keep things simple, the application assumes that the cost of higher education in Indonesia is four times the cost of lower education, that is, η2 = 4η1. Once earners make their choice as to what skill level they will pursue, it takes several years, say (n) years, before the effects of the choices are realised in a changed structure of the labour supplies. The (n) years may vary from about 4 to 10 years, depending on the educational system and the classification range of high and low skills. The ratio of higher to lower labour skills, RHLt + n, can be written for the long-term period t + n as in the equation below, which starts with the benchmark for the ratio at (LS2 / LS1 )o, and allows for some growth due to social demand expressed by λ′, and goes further to specify its dynamics in terms of net relative returns to both skills, with λ as the elasticity of skill upgrading due to differentials in skill returns, thus: RHLt + n = (1 + λ′)(LS2 / LS1 )0 {(LR2,t / η2 )/(LR1,t / η1 )} λ For estimation purposes the equation can be edited as in eq. 20, in Box 5, where the calibration coefficient λo can have a value equal to or higher than the benchmark for the skill ratio. There is the estimation problem here that the period (n) cannot be specified a priori, since behind it are many lags and overlapping variables of the educational and demographic systems that do not permit a short cut. The value of (n) can only be specified along empiricist lines. It is clear also that when the period (n) is longer, then RHLt + n will tend to be higher, thus leading to a higher value of the skill-upgrading elasticity. In principle, the equation can be regressed for various fixations of the (n) years, and the best-fitting regression for λo and λ can be selected, and in this way determine the appropriate number of years (n). This may require many data that are usually not available. An alternative approach is to postulate that the elasticity λ assumes the value of unity, and iterate from available data series the gestation years (n) that correspond with the postulate of λ = 1. (The application for Indonesia produced a value of (n) close to 6 years.) Finally, in eq. 21, the change in the skill ratio for the current year t, as from the past year t − 1, is assumed to follow a distributed lag pattern in terms of its value in the future year of t + n and in the past year t − 1. This is done via a distributed lag adjustment coefficient, φ. In this equation, the proportion φ is definable as t − (t − 1) / (t + n) − (t − 1), giving a value of (1/n + 1). (With n = 6 years in Indonesia, the value of φ is fixed at 0.14.)

Simple statics and dynamics in the CGE model  131 Turning now to the capital stocks, in the static model these were fixed for each sector and were not allowed to reallocate between the sectors. The dynamic model allows these stocks to grow and to reallocate in search of the highest profitability. For updating the capital stock, and for making its distribution on the sectors endogenous, use is made of four equations: eqs. 22 to 25. Eq. 22 distributes newly made investment expenditure over the using sectors in proportions which are a function of the ratio of sector profitability to average profitability KRj / AKR. Sectors which show high profitability in period t – 1 will increase their share in total investment in period t, in line with an inter-sector investment adjustment elasticity denoted by κ, that is set equal to 1. Sectors with higher profitability would thus experience relatively greater increases in their capital stock and in their production than other sectors. The share of investment destined to sector j in total investment, IDSj, is defined in eq. 22, holding ΣjIDSj =1. The benchmark investment shares, IDSj, t – 1, are approximated by taking the sector shares in the total operating surplus: KRj . KSj / ΣjKRj . KSj . (The approximation resulted in the following investment shares: 0.3091 for agriculture, 0.3974 for industry and construction, 0.2022 for trade, and 0.0913 for services.) Eq. 23 defines the average profitability for all sectors. Eq. 24 distributes total investment expenditure over sectors of destination to give the nominal investment in destination sector j, denoted by Idst, j. This is obtained by multiplying the newly introduced share variable by total investment. Eq. 25 subtracts depreciation from the capital stock, and adds new investment to it. Note that the investment outlays have to be divided by the price of investment goods, XPj, to determine the investment in physical units. Box 5  Newly introduced equations LS1,t + LS2,t = [LS 1,t − 1 + LS 2,t − 1 ] (1 + λ′ ) (19) RHLt + n = (LS2 / LS1 )t − 1 {(LR2,t / η2 ) / (LR1,t / η1 )} / {(LR2,t − 1 / η2 ) / (LR1,t − 1 / η1 )} (20) (LS2 / LS1 )t = (LS2 / LS1 ) t − 1 + φ [RHL t + n−(LS2 / LS1 ) t − 1 ] (21) j (22) IDSj,t = IDSj,t − 1 {(KRj,t / AKRt ) / (KRj,t − 1 / AKRt − 1 )} κ (23) AKR = Σ j (KRj . KSj ) / Σ j KSj j (24) Idst, j = IDSj . I j (25) KSj = KSj,t − 1 (1 − δ j ) + Idst, j,t − 1 / XPj′

6  Dynamic policy simulations To set the dynamic model in motion, the paths of the following exogenous variables need to be specified, Cgj , Tgh , Trh , Trg , Ig , Ipsi (see the list of notations in Section 2.3). Government consumption expenditure and transfers were projected based on past trends. In the case of Ig and Ipsi , which together add up to installed investment I, we considered initially an annual growth rate of GDP at 5 per cent, and making use of an average incremental capital-output ratio, at 2.2, predetermined the required value of I for each year between 1975 and 1980. The preliminary

132  Simple statics and dynamics in the CGE model runs of the model showed a lower growth of GDP than was initially considered, by about 0.5 per cent. This justified resetting the predetermined required values of (I = Ipsi + Ig) at a lower level. The dynamic model was applied to simulate the time path from 1975 to 1980. The time path is the sequence of temporary equilibria under unchanged policies. To get an idea of changes that would take place over time in the base time path, we will first discuss the predicted rates of growth between 1979 and 1980. The growth of GDP would amount to 4.40 per cent per annum. The largest contribution to this growth would come from the industry sector scoring a real valueadded growth of 5.37 per cent. The better performance of industry is mainly the result of the increased demand for investment goods (+6.73 per cent), of which industry is the only supplier. This would lead to higher profit rates for industry and a faster accumulation of capital in this sector (+5.89 per cent). As a result of increased supply, the price level of industry would tend to decline, reaching a reduction of −0.27 per cent in 1980. The distribution of income would change, although only slightly, in favour of the higher-income groups, which can be traced back to the deteriorating position of low-skilled labour in the factor market. The sectors that employ mainly lowskilled workers, such as agriculture and trade, would grow at a slower rate than other sectors: 4.12 per cent and 3.78 per cent respectively. We also applied simulations of the static model described in Section 3 to the dynamic model, starting with the year 1975. A few reflections can be made on the results with the object of highlighting the working mechanisms of the model. In the first policy simulations, 10 per cent productivity increases in industry and construction were applied. Substantial differences are found between the results of the dynamic and static simulations. The effect on GDP is more than twice as large in the dynamic simulation: 6.1 per cent as compared to 2.5 per cent in the static simulation. This can be explained by the decline in the price index of the industrial sector by 9.14 per cent, which makes it possible to buy more capital goods at a given level of savings. Therefore, capital is accumulated faster than in the base time path. This effect is reinforced by the fact that an increased inflow of foreign savings pushes up total investment expenditure by 5.59 per cent. The performance of the industry sector would improve in the dynamic as compared to the static simulation. This can be attributed to both demand and supply factors. Demand for the sector’s output would increase, resulting from the increase in total investment. On the other hand, the supply of the sector would be reduced, because the low profit rate would lead to a decrease in investment in this sector. Thus capital accumulation would be lower than in other sectors and so is the growth of output. Combined with the inelasticity of demand, the reduced supply would have a favourable effect on value added. However, although the decrease in value added would be reduced in comparison with the static simulation, value added generated in industry and construction would still be lower than in other sectors. So, also in the dynamic simulation the effect of an increase in industrial productivity would lead to a reduced importance of industry and construction in the domestic economy.

Simple statics and dynamics in the CGE model  133 The second policy simulation stipulated a 10 per cent productivity increase in services. The results of the static simulation—which can take into account the effects for 1975 only—are similar to those of the dynamic simulation. The largest difference appears in the predicted decline of the wage rate of the highskilled labour. Because potential suppliers of high-skilled labour would be discouraged by the lower wage, they would tend to quit the education system and would prefer to offer themselves as low-skilled labour. As a result, the supply of high-skilled labour would drop by −0.83 per cent. This, together with an expected diminished demand, would result in a reduction in the remuneration rate by −0.74 per cent.

7  Concluding remarks The chapter demonstrated how a static CGE model is harmonised with and parametrised from the SAM. The static CGE model is expandable to a dynamic version that allows factors of production to grow over time and reallocate among alternative activities, moved by incentives of relative returns in these activities. The dynamic model developed short cuts for approximating time lags between responding to incentives and the realisation of factor reallocations. Policy simulations were run and their results analysed with the objective of tracing the working mechanisms of the models. The first policy simulation envisaged a 10 per cent productivity increase in the sectors of industry and construction. The second policy simulation applied a 10 per cent productivity increase in the services sector. Although results of the static and dynamic simulations differ in magnitude, both predict that a productivity increase makes the innovating sector worse off in comparison with non-innovating sectors. This outcome occurs in cases of (a) low demand elasticities, that is, constant budget shares in eq. 9; and (b) non-sensitive quantities of exports and imports to relative price changes as specified in eqs. 14 and 15. While the simplifications introduced in the model are functional in highlighting and comprehending the structure of the model and its mechanisms, these same simplifications may reduce empirically the utility of the model for policy simulations. Notwithstanding advantages and shortcomings of the simplifications introduced, policies directed solely to rising productivities in the various sectors (and thus saving on labour use) may fail to sustain demand if they are not accompanied by measures to stimulate the reemployment of the laid-off labour. The US experience serves as a good example. (Productivity throughout 2009 rose 3.5 per cent, the best showing in six years and a reflection of companies’ ability to produce more with fewer workers. Still, many economists contend that the big productivity gains are actually harming the economy’s prospects for a sustainable rebound. If companies stop slashing their work forces and rehire laid-off workers, that will boost incomes and give households the support they need to increase consumer spending, which accounts for 70 per cent of economic activity.)

134  Simple statics and dynamics in the CGE model The results suggest that policy making which relies solely on the supply side (that is, enhancing productivity, as was simulated here, is an example) may fail to accomplish sustained development if they are not accompanied by policies that handle the demand side as well. By analogy, it is also logical to expect that policies focusing solely on the demand side would not accomplish sustained development if policies for tackling the supply side are left out. The next two chapters will show greater policy gains when instruments from the demand and supply sides are appropriately phased and combined.

  61   32  251   90  545   84

711 615

467  68  61  53 642

346 105  61  60

 49

 332

2422  351  268  268

  5

 534  351  221  233

 698  301  298  222   62

1412 1684 1682 3810

 882  196  160  131

  22  229   88 1669

   6   40   26 1448

  2

−39

  88   13    3    2   33  467    2   73   47   43

 14

1970  131   84  354   46   2  326  652    5   1 3810   59   39   2

4163 1063 8097 3731

1627  440  205  824  322  745

Source: Biro Pusat Statistik Indonesia (1982)

Total

14 Agriculture 15 Industry 16 Trade 17 Services 18  Rest of world

Activities

Institutions 4  Rural farm owner 5  Rural farm worker 6  R non-farm higher 7  R non-farm lower 8  Urban high 9  Urban low 10 Firms 11 Government 12  Indirect taxes 13  Capital account

Factor earnings 1  low skill 2  high skill 3  profits

277

1902 277

   2  218  194  769   −30

 567

  73   12   14   31   24   28

5 6 7 8 9 10 11 12 RF RNF RNF Urb Urb Firms Gov Ind. work high low high low ment tax

Institutions

1 2 3 4 Low High Profit RF skill skill own

Factor earnings

Annex Table 6.7  (First) social accounting matrix for Indonesia, 1975

3410

3410

8968

2571  204 1353   56  229

  87

1945   20 2503

7886

 292 1348  758   98 1256

 141

 694   81 3218

4254

 310  306   77  195  271

  42

1248  168 1637

2096

  21  121   31   37   70

   7

 276  794  739

2569

 423  907  772   −26

 493

8968 7886 4254 2096 2569

3731  615  642 1412 1684 1682 3810 1902  277 3410

4163 1063 8097

ROW Total

13 14 15 16 17 18 Capital Agriculture Industry Trade Services acc

Activities

7 Growth with redistribution through liberalisation with restructuring A CGE policy model of Nepal

1 Introduction Many developing economies have undergone trade liberalisation in the context of the structural adjustment programme of the IMF and the World Bank during the last couple of decades. The motivation is straightforward: if a regulated closedlike economy is to benefit from growth opportunities offered by the increasingly globalising world, then the economy should open up and implement trade liberalisation and structural adjustments. These actions are also considered necessary for obtaining membership in the World Trade Organization (WTO). In general, early voluntary actions on unilateral liberalisation are considered to be preferable over postponed and often compulsory actions that involve protracted and complicated bilateral and multilateral negotiations. There are various empirical studies on whether the predicted gains from trade liberalisation have materialised in countries that applied them. Most of the studies find that trade reform is accompanied by productivity growth, technological advancement, falling mark-ups, and a reshuffling of resources towards more efficient firms, among other things. In contrast, there are detrimental effects that have drawn equal attention, such as higher unemployment among workers displaced from less competitive industries and substantial rises in wages for the more competitive industries, resulting in a widening of income inequalities. Moreover, there is also some evidence that the impact on employment growth is small and adjustment costs are high. An elaborate list of references to these studies is found in Acharya (2011), and need not be repeated here. As the focus in this chapter is modelling the growth and distribution effects of trade liberalisation and structural adjustment, a few of the many models that dealt with this issue can be mentioned as a means for positioning our model. For example, in case of South America, Ianchovichina et al. (2001) modelled the effects of price changes on poverty and income distribution following the tariff reform of Mexico as a salient feature of its membership of the North American Free Trade Agreement (NAFTA). The impact on household welfare was found positive. Moreover, when non-homothetic individual preferences were modelled, import liberalisation had a good prospect of benefiting the poor more than the rich. Models for South Asian countries, Bandara and Yu (2003), reveal that small economies benefit more from unilateral liberalisation than multilateral

Liberalisation, restructuring: CGE model  137 preferential trading agreements. But these benefits are not sustained for the long run. In the African context, Chitiga et al. (2005) applied a CGE model for Zimbabwe showing that trade liberalisation benefits unskilled labour employed intensively in agriculture, mining, and services, while most of the manufacturing sectors shrink, leading to a fall in demand for skilled labour and capital. Poverty is reduced more in urban than in rural areas, but the income distribution pattern remains stable. In the Arab context, reference can be made to Lofgren et al. (1999), which is another dichotomy CGE model applied to Morocco that studies the effect of liberalisation on rural and urban incomes. The review of the models for individual countries that are related to our model is more extensive in Acharya (2011). It suffices here to state the shared conclusion, which is that the review does not lead to a robust conclusion regarding the distributional impact of unilateral import liberalisation. Even if the liberalisation could generate pro-poor growth, the sustainability of this impact is questionable. Moreover, the combinations of modelling details, differing model structures, choice of variables, and overlapping data make the whole significantly incomparable. Furthermore, the sustainability of the pro-poor growth following unilateral trade openness must have the fulfilment of some initial conditions, which are not laid down and are rarely explored in the reviewed models. The present modelling effort responds to the above mentioned gaps and it attempts to contribute some approaches towards resolving them. There are two objectives of the modelling effort. The first objective is to measure the impacts of various unilateral trade liberalisation measures to a small open economy of Nepal in a general equilibrium framework. Second, as the impact of liberalisation is found to be ineffective in promoting pro-poor growth, further modelling focuses on reformulations, tests, and redesigns of the economy’s structure in ways that could generate economic growth that is higher and pro-poor. The outline of the chapter is as follows. Section 2 formulates a general equilibrium model that is highly representative of the Nepalese economy. Section 3 reports on the calibration of the model and the simulation results of various unilateral trade liberalisation policies. Section 4 presents specific modes of economic restructuring that are hypothesised to be conducive to the pro-poor growth. The formulation of the modes is based on a thorough understanding of how this representative model of the economy works in generating its results. Section 5 repeats the simulated liberalisation policies analysis but under the restructured economy. Section 6 concludes.

2  Key features The CGE model of this chapter shares some associations with two other specific CGE models, and it is functional to signal them, mention the similarities, and emphasise the differences. One model is that built in the International Food Policy Research Institute (IFPRI), and described in Lofgren et al. (2001). The common feature that we share with this model is the detailed specification of the supply and demand for commodities in the product markets versus activities in the production sphere, whereby foreign prices of imported and exported

138  Liberalisation, restructuring: CGE model commodities are comprehensively incorporated in determining composite commodity prices. Another relevant model in the present context is by Cockburn (2001) who conducted a CGE analysis on trade liberalisation and poverty in Nepal, though his study relates to regional poverty, while ours define poverty in terms of socio-economic groups at the national level. There are minor and major differences between our model and the models of Lofgren and Cockburn. Besides differences in the specification of model closures, government budget balances, product market balances, and numeraire, the major difference is that we evaluate a wider combination of trade liberalisation and structural adjustment measures than are present in the above models. Furthermore, and more importantly, while the quoted models are static, we supplement the static version with a dynamic version that is particularly designed to meet the policy modelling tasks that are derived from the analysis of shortfalls in the policy simulation results of the static version. Because poverty reduction requires more time to achieve, the extension of the static CGE model to a dynamic CGE model is the logical step to undertake. Before entering into the specification of notations and equations, some main features of the model can be summarised. The model contains the four characteristic accounting balances in economy-wide models, which link the various parts of the model in a systemic whole, and characterise its equilibrium conditions. These are the factor market balances, product market balances, investment-savings balance, and the foreign payments balance. As the model specifies the government explicitly, there is the additional government budget balance that equalises its outgoing spending to its incomings, though inclusion of this budget balance is not a prerequisite condition in economy-wide models. A few comments will be made on each of these accounting balances. The factor market balances consider three factors of production: manpower with lower and with higher skills, and capital, employed in various activities. The remuneration rates of these factors, and their allocation among the sector activities, are fully the result of interactions between demands for them and their supplies. All three factors are movable from one activity to another. The model contains four activities denoted by j (agriculture, industry, commercial services, and other services). The product market balances are similarly broken down into four commodities denoted by c. Although there is a high correspondence between j and c, the correspondence is not diagonal. The model uses different types of prices for commodities. There are the price levels of exports and imports by commodity c (that is EPc and MPc , these are based on given world market prices quoted in foreign currency, converted to national currency via the foreign exchange rate, and adjusted to levies collected by government). Then there are price variables at the domestic production levels, XPc , composite prices of commodities appropriately weighted by the shares of domestic productions and imports, SPc , and domestic supply prices of domestic production inclusive of indirect taxes, DPc . In a general equilibrium model only relative prices can be established. In specifying these relative prices the model uses as numeraire one of the DPc commodities, namely that of industrial goods, which is an important normed base in Nepal, this is denoted by DP c = 2(industrial). Next to c – 1 commodity

Liberalisation, restructuring: CGE model  139 prices, which are adjustors in clearing the product market, there is the aggregate inventory change, IHC, which acts also as adjustor in balancing the product markets. This closure specification in the clearance of the product markets was denoted as closure A in Chapter 2. (Closure B would eliminate the role of inventory change, IHC, in the model and replace it by bringing in installed investment, I, as adjustor, see Chapter 2.) The closure specification for the savings investment balance is formulated along the lines displayed in Chapter 2. The model determines domestic savings ahead of installed investment which is autonomously determined by the private and government investors. The adjustment mechanism for balancing the financial market of savings and investment is one of two alternatives: it is either the foreign savings (this is foreign capital flow, denoted by FCF, which takes the role of the adjustor, keeping the foreign exchange rate fixed; or the adjustor is the FXR, which is the unknown, while FCF is prefixed. The status dependence between the variables of FCF and FXR was discussed earlier in Chapter 2, where the two closures were respectively denoted as closures C and D, and are relevant in this chapter as well. For a deductive study of the effects of alternative policy simulations, the model initially considers FCF as endogenous and the foreign exchange rate, FXR, as exogenous, thus following closure C; and we end up with the policy simulation where FXR is unknown and FCF is prefixed, thus following closure D. Finally, in correspondence with the underlying SAM, the CGE model distinguishes between three institution accounts: households, firms, and the government. To keep it simple, households and firms are denoted by the same index h; so there is no need for separate equations for firms. It will be taken for granted for the model as a whole, therefore, that when reference is made to h, this would cover various household groups, and firms; with the understanding that contrary to the household groups whose factor income consists of wages and profits, the firms’ income consists of profits only; and contrary to household groups who spend on consumption, the firms’ income is fully saved and invested. Otherwise, both households and firms pay taxes on incomes, and may be involved in transfers. Among the households, four groups are distinguished. This subdivision of households is based on various socio-economic and locational characteristics applicable to Nepal. The first two groups, Urban Residing Households (URH) and Rural Upper households (RUH), are rich households whereas Rural Small Households (RSH) and Rural Landless Households (RLH) are poor, with the latter being the poorest. Government enters the model with equations and variables for revenue, consumption expenditure, transfer payments, investment spending, and a budget deficit. The model applies detailed distribution of factor incomes by household, factor, and activity types. These three cross-links allow linking changes in the production of activities to changes in factor remunerations and household incomes and highlight the impact of activity and factor reallocations on household incomes, and the consequences thereof for the poverty situation. This is one of the aspects often overlooked in many CGE models.

140  Liberalisation, restructuring: CGE model

3  Model specification The model contains 28 equation sets in 28 sets of unknown variable that can be described to fall into four blocks. Table 7.1 summarises all indices used, and defines variables and coefficients.

Table 7.1 Notations Indices: c = commodity categories (1 = agricultural, 2 = industrial, 3 = commercial 4 = services). h = household groups (1 = URH, 2 = RUH, 3 = RSH, 4 = RLH), 5 = firms. j = activity sectors (agriculture, industry, commercial services, other services) q = labour skills (1 = low-skilled labour, 2 = high-skilled labour)). Furthermore, use is made of the subscripts g and r to denote government, and rest of the world, respectively. Endogenous variables: price, quantity, and value variables of activities and commodities DPc Domestic price of domestic output by commodity c, with particular reference to DPc(industrial), which serves as numeraire, equal to 1.0 EPc Export price in domestic currency by commodity c Import price in domestic currency by commodity c MPc Composite supply price by commodity c SPc XPc Producer price by commodity c Output price by activity j XPj Domestic output quantity sold domestically by commodity c DQc Exports quantity by commodity c EQc Imports quantity by commodity c MQc Composite supply quantity by commodity c SQc Domestic output quantity by commodity c XQc Gross output quantity by activity j XQ j Gross output of activity j Xj Value added of activity j (value) Vj

Endogenous variables: Others Consumption by household h of commodity c Chc Total consumption by commodity c Cc FCF Foreign capital flow expressed in domestic currency. FCF is converted in an exogenous variable in an alternative specification of model closure GBDg Government budget deficit Installed investment I Inventory change ICH Installed investment deliveries by commodity c Ic Demand for capital in activity j KDj Rental rate of capital KR Demand for labour of type q in activity j LDqj Remuneration rate of labour factor q LRq Disposable income of household h, applies also for firms Yh Factor income and received transfers of household group h, applies also to firms Zh Government revenue Zg

Exogenous variables Export price in foreign currency by commodity c epc Import price in foreign currency by commodity c mpc FXR Foreign exchange rate (domestic currency per unit of foreign currency). FXR is converted in an endogenous variable in an alternative specification of model closure Cgc Government consumption expenditure on commodity c Government expenditure on installed investment Ig Supply of labour factor production factor q LSq Supply of capital stock in the whole economy KS Tgh Government transfer payments to household group h, extended to include firms Transfer from rest of world r to household h, extended to include firms Trh Transfer from rest of world r to government g Trg

Coefficients Calibration parameter in production function of activity j αjo Intermediate delivered quantity of c as input per unit of activity j αc j Calibration parameter for output transformation (CET) function αco Share parameter for output transformation (CET) function αc Exponent for output transformation (CET) function (–1 < α c’ < ∞) αc’ Non-competitive import share in the output of activity j αrj Labour elasticity of production relating to labour type q in activity j βqj Calibration parameter for composite supply (Armington) function χ co Share parameter for composite supply (Armington) function χc Exponent for composite supply (Armington) function (–1 < χc′ < ∞) χc′ Calibration parameter for exports by commodity c εco Export price elasticity by commodity c εc Consumption price elasticity of commodity c by household group h γhc Consumption budget share of commodity c in the consumption of household γhc# group h Calibration parameter in consumption functions γhco ιo , ιk , ιg Private investment function coefficients, whereby ιo is calibrated, ιk is elasticity of investment with respect to rental rate of capital, and ιg is with respect to public investment ιc Distribution share of total investment on commodity c Calibration parameter for imports by commodity c μco Import price elasticity coefficient by commodity c μc θc j , θjc Converted units of commodity c per unit of activity j, and otherwise πjh , πjg Profit distribution shares from activity j that go to household h (and firms), and to government g Transfer payments (net) from household h′ to h, as proportion of the income of ρh′h the paying group h′ σh Savings proportion in the disposable income of household group h Sales tax rate by commodity c collected by government g τcg τcg,imp Import tariff rate by commodity c collected by government g Export tax rate by commodity c collected by government g τcg,exp Direct tax rate on income of household group h collected by government g τhg Wages distribution shares from activity j for labour of skill type q received by ωqjh household group h υc Inventory change in commodity c as proportion of the aggregate inventory change

142  Liberalisation, restructuring: CGE model The first block, Box 1, comprises the factor market equations. Eq.1 combines the Cobb-Douglas production function and Leontief input-output coefficients in determining the volume of gross output per sector, Xj . XPj , as functions of the employed labour skills and capital, and technical coefficients of input-output deliveries. The details of deriving this equation are found in Chapter 6. Eq. 2, making further use of Leontief input-output coefficients, the value added by sector, Vj , is found by deducting from the gross output the appropriately repriced value of the composite commodities c used in the production of activity j, and value of intermediate imports per unit of activity j. Eq. 3 assumes producers maximising profits under perfect competition in product and labour markets. As a result, in the short run they will hire employees of skill q until the labour remuneration rate for that skill, LR q, equals the value of the marginal product of that labour skill. Eq. 4 states that the remuneration rate of capital, KR, is equal to the value of its marginal product. Capital is assumed to be mobile between sectors, and thus, capital returns are the same for all sectors. Eq. 5 fixes the supply of labour by skill type, and equalises this supply with the demand for that labour skill. The remuneration rate of labour by skill type carries the burden of adjusting demand to supply in the labour market. Eq. 6 fixes the supply of capital and equalises this supply to the demand in all activity levels. Here too, the remuneration rate for capital across activity carries the burden of adjusting demand to supply of capital for all activities. This is the case of the mobile situation for installed capital; see Chapter 2, which allows capital to shift from one activity to another in pursuit of higher rental differentials. The second block of equations relates to income and spending of households and government. In eq. 7, total income of recipient household groups (and firms) is specified as consisting of factor remunerations distributed over household groups h via wage shares ω that are detailed by household group h, skill q, and sector j, and profit shares π detailed by household group and sector, thus ωhqj and πhj. (Incorporating earnings details by sector of activity is an important element for policy making; this is often missed in many CGE models due to the aggregative nature of the SAMs behind them which do not link household earnings to the sector where the income is earned. For example, simultaneous earning links of activity and factor to household were not explicated in the previous chapter.) Next

Box 1  Factor market: production functions, factor use, and factor remuneration Xj = XPj {1 / (1 − Σc αcj − μ j } [αjo Π LDqj βq . KDj (1 − Σqβqj) ] V j = {1 − (Σj αcj SPc / XPj ) − αrj } Xj LRq = βqj Vj / LDq j KR = (1 − Σq βqj ) Vj / KDj Σj LDqj = LSq Σj KDj = KSj

j (1) j (2) q. j (3) j (4) q (5) j (6)

Liberalisation, restructuring: CGE model  143 to factor incomes there can be income transfers paid to and received from other household groups which are assumed to be fixed proportions, ρh′h , of the income of the paying party. The equation incorporates the net results of these transfers. Furthermore, there are income transfers received from government, Τgh , and net income transfers from rest of world, Trh. The conversion of these foreign transfers from foreign currency to national currency via the foreign exchange rate, FXR, is explicitly incorporated in the equation. In eq. 8, direct taxes are subtracted from the incomings of households to arrive at disposable income. In eq. 9, consumers save a fixed proportion of their disposable income and spend the rest on consumption that is, (1 − σh ) Y h. This is their spending budget constraint. The consumption function assumes that the consumer maximises a Cobb-Douglas utility function, given the spending budget constraint. Accordingly, the distribution of the consumption expenditure on commodities is determined initially by the start budget shares γhc# , and adjusted subsequently to a price effect (via own and cross-price elasticities, γhc , in interaction with the composite prices of competing commodities, SPc ), and to an income effect (via the expenditure elasticity which is (1 − Σc γhc ), in interaction with the expenditure level). (The Cobb-Douglas utility function requires that the sum of the elasticities involved is equal to 1, hence the specification of the expenditure elasticity as (1  −  Σc γhc ). In the specific case where price elasticities equal zero and the expenditure elasticity equals 1.0, the distribution of spending on commodity c will be consistent with the initial budget shares, γhc#.) In eq. 10, total consumption expenditure for commodity c is the sum of private consumption by all households of commodity c, and government consumption expenditure on commodity c, Cgc. In eq. 11 installed investments1 are broken down into public investment, taken to be exogenous, and private investment, which is specified as a behavioural function in terms of two main factors. One factor is the profit rate (private investment increases as the rental rate of capital increases via elasticity ιk. The other factor is the complementary character and crowding-in effect of public investment leading to more private investment via elasticity ιg. In this equation, ιo serves as a calibration coefficient.2 In eq. 12 installed investment deliveries classified by commodity c, that is Ic , is a proportion ιc of the total installed investment, I. Because there is only one commodity, denoted as industrial, which constitutes investment goods, we have ιc = 1.0 for the industrial commodity c = 2, and zero for all others. In eq. 13, government revenue is defined to consist of factor income from ownership of public enterprises; plus direct taxes from households (and firms); plus indirect taxes on domestic commodities, import duties, and export duties; plus net transfers relating to households (and firms); and net transfers from abroad converted to national currency. In eq. 14, the left-hand side shows government expenditure consisting of current expenditure (consumption and transfers: ΣcCgc + ΣhΤgh), plus fixed investment Ig , while the right-hand side shows how they are financed; namely, by government revenue, Zg , and a budget deficit, GBDg.

144  Liberalisation, restructuring: CGE model Box 2  Factor incomes, disposable incomes, private consumption, investment, government Zh = Σq Σj ωq jh LRq . LDqj + Σj πjh KRj . KDj + {h = 6 groups, firms} (7)    Σ h′ ρh′h Zh′ + Τgh + FXR . Trh {h = 6 groups, firms} (8) Y h = (1 − τhg ) Zh h.c (9) Chc = γhco Πhc SPc γhc {γhc# (1 − σh ) Y h } 1 − Σ c γhc c (10) Cc = Σh Chc + Cgc (11) I − Ig = ιo KRιk Igιg c (12) Ic = ιc I Zg = ΣjπjgKRj . KDj + ΣhτhgY h + Σc τcg DPc . DQc + Σc τcg,imp FXR . mpc ΜQc + Σc τ cg,exp FXR . epc ΕQc+ FXR . Trg (13)    (14) Σc Cgc+ Σh Τgh + Ig = Zg + GBDg

The third block is in Box 3. The equations here relate to demand, supply, and prices in the product market. The price system of the model is elaborate, primarily as a result of the assumed quality differences between commodities of different origins and destinations (exports, imports, and domestic outputs used domestically). A distinction is made between the quantity and price level of the supply composite of commodities, SQc and SPc; the quantity and price level of the domestic output of commodities that is sold domestically, DQc and DPc; and the quantity and price level of the domestic output (that is sold domestically and is exported internationally), XQc and XPc. Composite price variables SPc are important, among others, in determining the distribution of consumption on commodity c (see eq. 9), and clearing the product market (see eq. 27). The domestic price variables DPc are crucial in determining national contra foreign price competitiveness and the size of imports and exports, as will be incorporated in eqs. 17 and 21. Composite supply (Armington function): Eqs. 15 and 16 specify the composite supply. In eq. 15, for each commodity, absorption (that is total domestic spending on the commodity) is the sum of spending on domestic output, including an upward adjustment for the sales tax, and imports. All purchasers buying composite quantity SQc pay the composite price, SPc. Prices of both domestically produced goods and imported goods are expressed in domestic currency, and they are both adjusted with their respective taxes. Eq. 16 treats the imperfect substitutability between imports and domestic output. This is captured by a CES (constant elasticity of substitution) aggregation function in which the composite commodity that is supplied domestically is ‘produced’ by domestic and imported commodities, and they enter in this function as ‘inputs’. This means demander preferences over imports and domestic output can be expressed as a CES function, along the lines proposed by Armington, the originator of the idea. The restriction on the value of χc′ , (−1 < χc’ < ∞), assures that the corresponding isoquant is convex to the origin, in terms of a production economics equivalent to a diminishing technical rate of substitution.

Liberalisation, restructuring: CGE model  145 Import ratio and price: Eq. 17, which applies to imported commodities, specifies how the mix between the quantity of imports and the domestic output for commodity c is realised. It states that the domestic price / import price ratio determines the import / domestic demand ratio, subject to applicable elasticities. Eq. 18 gives a simple definition of the price of imported commodity c as the world market price converted to domestic currency and increased with the import tariff. Domestic output (CET function): Turning now to domestic output details, eq. 19 defines the value of domestic output of commodity c, at producer prices, as the sum of the value of domestic output sold domestically and the export value expressed in domestic currency. The imperfect transformability between domestic output for exports and domestic sales is treated in eq. 20 via a constant elasticity of transformation (CET) function that captures the transformation. The CET function is identical to a CES function, except for negative elasticities of substitution. The isoquant corresponding to the output transformation function will be concave to the origin given the restriction imposed on the value of αc′ (−1  0 while τ2 = τ3 = 0. Land tenure security, LTSj, is a third factor. LTSj is defined as the ratio of actual lease period to the relatively secure one, which is generally viewed to be 40 years. In the absence of land tenure security, farmers adopt short-term cultivation practices, and neglect fallow periods, crop rotation, soil conservation, and improvements. All of these exhaust and degrade the land. A greater LTSj leads to a greater ASPh; and this effect is captured by elasticity λj. The LTS factor is especially relevant in mechanised agriculture where the actual lease term is between 15 and 25 years; hence, we focus on λ2 > 0, while λ1 = λ3 = 0. Satisfaction of basic needs is a fourth factor. If poor farmers earn enough to satisfy their basic needs, then they are motivated to exploit the land in more productive ways. The gap between what they earn (their value added from the land they cultivate, V3 , plus supplementary wages they receive for work they deliver in irrigated and mechanised agriculture, expressed as ψ31 X1 and ψ32 X2) and the income they need to fulfil their basic needs (denoted by YBNj and which is externally normed) is crucial for enhancing productive efforts. It is proposed to simulate the positive effect of a narrowing of the income gap on the attained sustainable productivity via elasticity νi. This factor is particularly valid for poor farmers, who are primarily in the subsistence sector,4 so that ν3 > 0, while for the other sector νj = 0. The above stylised facts for irrigated, mechanised, and subsistence agriculture result in the specifications of ASPj for the three sectors as in eqs. 1.1, 1.2, and 1.3. As stated earlier, the equations assume Cobb-Douglas specifications, so that φ1 + τ1 = 1, φ2 + λ2 = 1, and φ3 + ν3 = 1. Specification of eq. 1.4 for biomass forestry ( j = 4) requires elaboration. The attained sustainable productivity index, ASP4 is expressed as the proportion of last year’s actual stock of biomass trees in cubic meters per feddan of forestry land, BTS4,t − 1, to the technical norm β4,norm for sustainable development. As ASP4

Sustained development in CGE model  167 reaches the value of 1, forestry recovers back to its optimal level of sustainable productivity, for example: ASP4 = BTS4/β4,norm. Specification of eq. 1.5 for the livestock sector is similar to that of forestry. The attained sustainable productivity index for livestock, ASP5 , is the ratio of a first term representing last year’s actual vegetation per livestock unit, measured in grass available in metric tons per feddan, VGT5,t − 1, and divided by the average number of livestock unit per feddan, LVS5,t − 1, to a second term representing a normed vegetation per livestock unit for sustained development, θ5,norm, which is technically given by livestock experts. Attaining a value of ASP5 = 1 is an ideal and optimal state. Specifications of BTS4, VGT5, and LVS5 are done in separate equations. BTS4, which is the actual biomass trees in any one year, is specified in eq. 2 as the level of the previous year plus a natural growth based on an autonomous growth rate β4, as well as its responsive accommodation to progress made in the sustainable productivity index ASP4 / ASP4,0 via elasticity β′4 , plus voluntary biomass tree planting per feddan, BTP4, less tree wood cutting per feddan, which is equivalent to the produce yield volume of the forestry activity, YLD4. The absence of conservation knowledge and a private property rights system in forestry has often resulted in BTP4 = 0; while wood cutting, YLD, which represents the output of the sector, has become a prosperous activity but with possibly negative environmental consequences. The three variables of BTS4, BTP4, and YLD4 are expressed in biomass trees per unit of land, that is, per feddan. Vegetation state, VGT5, is specified in eq. 3. Its annual improvement is influenced by exogenous variations in rainfall, ΔRFL, and by the controllable extension of water wells, ΔWLS, via appropriate elasticities θ and θ'5. The vegetation level is, as a whole, dependent on the change in the attained sustainable productivity index (ASP5 / ASP5,0 ) via elasticity θ''5. The constant economies of scale inherent in the Cobb-Douglas function require that θ5 + θ'5 + θ''5 = 1. The current livestock per feddan, LVS5, is determined in eq. 4 by the previous livestock per feddan plus an autonomous net growth rate, ϖ5, as well as its responsive accommodation to progress made in the sustainable productivity index ASP5 / ASP5,0 via elasticity ϖ'5, plus artificial insemination, ANS5 , that remained at zero in the past; less average livestock slaughter per feddan, YLD5, that represents output of the livestock sector in physical terms. Attention can be turned now to equation set 5 which determines the annual physical yield, or volume of produce per unit of land of activity j, YLDj. Eq. 5 is detailed for each of the three agricultural activities as follows. Current yield is the product of the yield of the previous year, and an annual growth rate, πj , which follows improvements in production technology that are partially exogenous, but that are also dependent on the attained sustainable productivity. In this equation the attained sustainable productivity index, ASPj, is the accommodating variable that can be controlled policy-wise to influence the growth of the yield. Furthermore, the realisation of any produce requires current and fixed inputs, such as intermediate deliveries and investment. The two inputs will be incorporated in related equations for agriculture that assume Leontief and Harrod-Domar technologies, respectively. The actual yields in physical terms per feddan in the cases of forestry (that is wood cutting) and livestock (livestock slaughtered) are determined by a combination of

168  Sustained development in CGE model technical, market, behavioural, and policy considerations, giving eqs. 5.4 and 5.5. The technical consideration lies in the fact that the yield per feddan in physical terms, whether it is wood cutting or livestock slaughter, YLDj , in any current year needs to be standardised in terms of its equivalent quality in relation to the base year. This is done by adjusting the physical yield to the variation in the sustainable productivity index ASPj / ASPj,0. The market considerations lie in converting the adjusted physical yield in revenue terms. This is done by multiplying the adjusted physical yields by the produce price for the base year pj,0, the world relative price index specific to the products of the respective activity, XPj, and their respective cultivated land quantities NQ j. For a small country like Sudan the world relative prices of AFL commodities are exogenously given: relative to the irrigated sector that serves as numeraire. The behavioural consideration lies in the urge of forestry and grazing households to engage, respectively, in wood cutting and livestock slaughter, and sell these products up to the point at which they can earn a desirable income level that more or less satisfies the basic needs of their growing population. This level is fixed exogenously at YBNj, as was proposed for the agricultural population. It can also be expressed in a simplified form as the previously attained income of the social group plus an exogenously stipulated annual addition. We shall keep to the general formulation of YBNj, without further specification. Finally, policy considerations enter if government would allocate transfers and (or) provisions in kind, denoted by Tgj, to the populations associated with forestry and livestock activities such that their gaps between required income and earned incomes are reduced. Poor people, in desperation to secure basic life needs, may be forced to destroy the environment. Therefore, if government can help them through a transfer payment to partially close their poverty gap, then this may preserve the environment.5 Box 1  Attained sustainable productivity and yield in agricultural activities Farming activities ϕ ASP1 = χ1 (FKN1, t − 1 ) 1 (1 − τ 1g )τ1 ϕ ASP2 = χ2 (FKN2, t − 1 ) 2 (LTS2, t − 1 )λ2 ϕ ASP3 = χ3 (FKN3, t − 1 ) 3 {(V3 + ψ31 X1 + ψ32 X2 + Tg3 ) / YBN3 }ν3

(1.1) (1.2) (1.3)

Forestry and livestock (1.4) ASP4 = BTS4,t − 1 / β4,norm (1.5) ASP5 = (VGT5,t − 1 / LVS5,t − 1 ) / θ5,norm β' (2) BTS4 = BTS4,t − 1 (1 + β4 ) (ASP4 / ASP4,0 ) 4 + BTP4 − YLD4 ; BTP4 = 0 θ θ' θ'' VGT5 = VGT5,t − 1 (ΔRFL5 ) 5ΔWLS5 ) 5 (ASP5 / ASP5,0 ) 5 (3) ϖ (4) LVS5 = LVS5,t − 1 (1 + ϖ5 ) (ASP5 / ASP5,0 ) '5 + ANS5 − YLD5 ; ANS5 = 0 Yield in the farming activities YLDj = YLDj,t − 1 (1 + πj ) ASPj Forestry and livestock [pj,0 XPj . YLDj (ASPj / ASPj,0 )] NQ j + Tgj = YBNj

j=1, 2, 3 (5.1, 5.2, 5.3) j = 4, 5 (5.4, 5.5)

Sustained development in CGE model  169 The second block, in Box 2, refers to eqs. 6 to 9 to be written generally for all five land use activities. Eq. 6 gives the value per feddan for activity j, NPj , as the rental share of the land factor in the yield per feddan, ρj, valued at world prices pj0 XPj and discounted with a constant discount rate Δ. Eq. 7 is a set of four equations. They specify the mechanism behind the distribution of types of land other than the predetermined irrigated land. The allocation is fully governed by market forces. Changes in land allocation between type j and mechanised agriculture (sector 2) from a base year are formulated as a function of relative changes in last year’s land value as compared to the base year. Hence, μj is the land to value elasticity of substitution between sector j = 2 (mechanised land) and sectors j = 3, 4, 5, formulated in terms of change in terms of land quantity to change in land value. Practically speaking, land conversions proceeded in the past from forestry, to subsistence, and mechanised agriculture, but mostly to the latter, making it logical to take land values of mechanised agriculture as numeraire for the purpose of clearance of the land market. To account for incurred costs ηj in the conversion of land i from one use to another, and for the changing states of sustainable productivity ASPj of the concerned land j, the term ηj ASPj is deducted from the value of the concerned land j as per last year and the base year. Furthermore, equation set 7 contains the accounting identity stating the fixed sum of land for all sectors, NQ. Eq. 8 gives the value of agricultural output, Xj, of activity j as the price of the product in the base year pj,o weighted by the relative price index for the current year XPj, the numeraire being the price of the produce of irrigated agriculture. The price expression is then multiplied by both the actual yield per feddan, YLDj and the quantity of land under cultivation in that sector, NQ j. Eq. 9 specifies the investment goods required by destination activity, and adds these to give total investment in the aggregated (all) agriculture sector, Iagr. The equation relates investment requirements in each AFL activity to production growth of the activity via sector incremental capital-output ratios, denoted by κj. Now that output expansions of the land resource sectors are known from eqs. 1 to 6, the necessary investments that are complementary for these expansions are specified and summed in eq. 9. Box 2  Other equations on agriculture, forestry, and livestock Value per unit of land NPj = (ρj pj0 . XPj . YLDj ) / δ

j =1,..,5 (6)

Distribution of land on agricultural activities NQ j′ / NQ j = (NQ j’,o / NQ j,o) [(NPj’,t − 1 − η j′ ASPj’,t − 1 ) (NPj,o ) / j = 3, 4, 5 (7.3, 7.4, 7.5)   (NPj’,o − η j′ ASPj’,0 ) (NPj,t − 1 )]μ j (7. total) NQ = NQ1 + NQ2 + NQ3 + NQ4 + NQ5 Value of gross output by agricultural activity X j = p j,o . XPj . YLDj . NQ j Investment requirements of all agriculture (Iagr,t + Iagr, t − 1 )/2 = Σ j κj (Xj − Xj, t − 1 )

h = 1,..,5 (8) (9)

170  Sustained development in CGE model The third block of equations, in Box 3, consists of equations that specify the non-agriculture sector, denoted by j = 6, and connects it to the agriculture sectors j = 1,..,5 within an inter-industry framework. Eqs. 10, 11, and 12 present a simplified modelling of the output of the non-agriculture sector. In eq. 10, output quantity follows a Cobb-Douglas production function, with labour and capital as factors of production. Eq. 11 specifies the growth of labour supply. Eq. 12 specifies the growth of capital formation. The annual non-agriculture investment added to the capital stock of non-agriculture is total investment less investment destined for agriculture. Eq. 13 is generally written for agriculture and non-agriculture. Value added is obtained by deducting three items from the output: domestic intermediate input deliveries, imported intermediate deliveries, and indirect taxes. Domestic intermediate deliveries are quantity-based input-output coefficients αj'j, that are adjusted to relative price indices of receiving and delivering sectors XPj′ / XPj. Shares of non-competitive imports are αrj, and rates of indirect taxes are τjg. Eq. 14 specifies demand for consumption goods, Cj. These are based on CobbDouglas-like consumption demand functions in terms of the gross domestic product ΣjVj and relative price indexes XPj. Eq. 15 specifies a behavioural function for private investment Ipsi. One determining factor is the profit rate. Because this is not generated by the model the price index of non-agriculture is used as proxy. A higher price raises cost and can discourage investment, but the higher price can also be an incentive for greater investment. In Sudan the larger part of investment is imported and is thus not negatively affected by the domestic price. The net result is a positive price effect that can be represented by elasticity ιp. The other factor is the crowding-in effect of public investment leading to more private investment via elasticity ιg. With given values of the two elasticities, ιo is obtained as calibration coefficient.6 Eq. 16 defines installed investment,7 I, as the sum of private investment, Ipsi, and government investment, Ig. Eq. 17 specifies foreign trade at the sector level. As stated earlier it is assumed that the agricultural sectors are able to export their products as long as the prices offered are equal to the international prices of these products. We assume thus that net exports Ej for sectors j = 1,…,5 are endogenous. The export-import balance for the non-agriculture sector, j = 6, is formulated along the lines of Chapter 2. Quantity of net exports of the sector is dependent on the output quantity of the sector (the first term), and the relative price of the sector to the given world price, and the export price elasticity ε6 (the second term). Eq. 18 specifies the positions of government revenue and spending. Government, by imposing price levies on irrigated agriculture, appropriates the difference between the world price for the irrigated product and that which it offers to farmers. The revenue that the government appropriates in this way is used to finance several spending items relating to agriculture; these are cost of policies relating to transfers to poor populations in subsistence cultivation and forestry Σj Tgj; cost of policies for education and training of farmers, Σjφj Δ FΚΝj; and there is a subsidy to farmers if exportable agricultural surplus exceeds an upper limit, ΣjΣj (Ej – Ej ,max )8. Furthermore, government agricultural savings, GSA, are used for financing government consumption and investment in the

Sustained development in CGE model  171 whole economy. In Sudan, the incidence of total government revenue has been for many years roughly evenly split between agriculture and non-agriculture. The contribution of non-agriculture to government revenue and spending is specified as the indirect tax on output, τ6g X6 , on the right-hand side.

Box 3  Equations relating to non-agriculture and the whole economy Output of the non-agriculture sector (10) X6 / XP6 = {1 / (1 − Σj αj›6 − αr6 − τ6g )} [α6o Π LD6 β6 KD6 (1 − β6] LD6,t = LD6,t − 1 (1 + λ6 ) (11) (12) KD6,t = (1 − δ 6 ) KD6,t − 1 + (I − I agr ) Whole economy: value added, consumption, investment, foreign trade, government j = 1,…,6 (13) Vj = (1 − Σj αj›j XPj’ / XPj ) Xj − αrj Xj − τjg Xj j = 1,…,6 (14) Cj = γjo Π(XPj )γ'j (γj Σj Vj )1 − Σjγ'j ι ι (15) Ipsi = ιo (XPnon agriculture) p (Ig ) g (16) I = Ipsi + Ig (17) E6 / XP6 = (ε6,o X6 / XP6 ) (XP6 / ep6 )ε6 Σj (τjg . pj XPj . YLDj . NQ j ) − Σj Tg j − Σj φj ∆ FKN − Σj σj (Ej − Ej,max ) + GSA = τ6g X6 (18)

The fourth block, in Box 4, displays the three national accounting balances, and a few additional equations that are of interest for policy making and resource management. In eq. 19 the sector product balances by delivering sector state such that total supply equals total demand of goods by the delivering sector j. The supply is the output of the sector. The demand comprises intermediate deliveries adjusted to the price indices of the receiving and delivering sector, consumption, investment, and net exports. The installed investment is specified by delivering sector. Since only the non-agriculture sector delivers capital goods, we have ι6 = 1, and ιj = 0 for the other sectors. The equalisation of supply and demand in the product markets runs differently for the agricultural sectors and for non-agriculture. The supply of agricultural sectors, Xj for j = 1, 2, 3, 4, 5 is already determined from the agricultural blocks in Boxes 1 and 2. The supply of non-agriculture goods, X6, is solved from eqs. 10, 11, and 12 in Box 3. On the demand side, the intermediate deliveries, consumption and investment, along with net exports of non-agriculture, are all solved for and can be substituted in eq. 19. The result of the above discussion is that the six j balances are left with only the following adjustor variables to be determined: net exports of the five agricultural activities, Ej for j = 1 to 5, and the price level of non-agriculture, XP6 , which is expressed in terms of the numeraire (that is, the given prices of the irrigated activity products). Looking for adjustors in the agricultural product balances, these can be only net exports, Ej, which were specified earlier as endogenous. Because they are offered and traded at given exogenous international prices, and since the size of the country’s exports in the total world imports is marginal, it is very realistic to

172  Sustained development in CGE model take Ej for the AFL sectors as endogenous. This makes net exports the natural adjustors in the AFL market balances. In the case of the product balance of non-agriculture the natural adjustor is its output price, XP6. Inventory change ICH is incorporated in eq. 19, but has a subordinate role. It is insignificant in agriculture and mostly at the zero level, so that all inventory change occurs in non-agriculture, making υ6 = 1. The aggregate level of ICH is fixed at its level for the benchmark year, implying that its relative significance diminishes as the solutions of the model show economic growth for coming years. Eq. 20, the savings investment balance, states the equality between supply of financial funding (private savings, government savings, and foreign savings) and demand, consisting of investment and inventory changes. Since the foreign exchange rate FXR is fixed, the adjustor in this financial market balance is foreign savings, that is, FCF expressed in foreign currency. The model, with its 20 sets of equations, is mathematically complete at this point. When the accounting balances of sector product balances and savings-investment balance are substituted in each other they automatically give the foreign payments balance of eq. A1. This states the equality of foreign trade (exports less imports, or net exports ΣjEj, plus non-competitive imports Σjα rjXj) to foreign capital inflow, FXR . FCF.9 This equation is placed outside the model as it is an implied equation, generally highlighted by Walras Law.10 The model of 20 sets of equations counts in total 54 single equations. The model counts an equal number of 54 endogenous variables. There are 28 endogenous variables that relate directly to the five AFL sectors (ASPj , YLDj , NPj , Xj; four NQ j variables relating to sectors 2, 3, 4, 5; and four variables of BTS4 , VGT5 , LVS5 , and required investment of all agriculture Iagr). Furthermore, there are the remaining 26 endogenous variables that directly or indirectly relate to the whole economy consisting of three sets of variables in j = 6 sectors (they include Vj , Cj , and Ej ) and eight individual variables of output, and labour and capital use in non-agriculture, X6 , LD6 , and KD6 , government savings, GSA, private installed investment, Ipsi , total installed investment I, and the two adjustor variables of the output price of non-agriculture, XP6 , and FCF. Counted otherwise, the model consists of 20 sets of endogenous variables derivable from 20 equation sets. It is important to underline that the formulation of the model into is in two parts: (a) AFL sectors, consisting of eqs. 1 to 9, with its 9 sets of endogenous variables; and (b) the ROE with the remaining 11 equation sets, eqs. 10 to 20, which contains 11 sets of endogenous variables. This divisional structure contributes substantially to the modelling of a simplified structure. Eqs. 1 to 9 is an autonomous system and is solved without access to the equations or variables of the ROE part. But solution of the ROE part requires the prior solutions for the AFL sectors. There are three additional equations meant to emphasise some policy considerations and can be seen as a linked tail to the model. Eq. A2 underscores the effects of environmental degradation by estimating the degradation cost DCSj to be incurred so as to reclaim foregone sustainable productivity, this being defined

Sustained development in CGE model  173 in relation to the desirable value of ASPj = 1. Reclamation cost per feddan has been generally estimated for the average piece of land of average sustainable productivity at about 10 per cent of the foregone yield per feddan as a result of resource degradation, that is ζ j = 0.1. Foregone yield per feddan is the lost rental product of the land factor, ρj YLDj, valued at the world price of the produced commodity, pjXPj ; the degree of loss being dependent on degradation level, (1 – ASPj ). If ASPj is at its ideal level of 1, then the reclamation cost becomes zero. The degradation loss per feddan is then multiplied by NQ j to give total degradation loss.11 The indicator of land degradation costs, Σj DCSj , is useful in highlighting the environmental loss of economic growth since it gives the welfare loss due to natural resources misuse; and it can tell how simulated agricultural policies will reduce the cost of environmental degradation. A supplementary indicator in eq. A3 is GRN, for the green GDP, obtained by deducting degradation cost from GDP. Alternatively, eq. A4 computes the ratio of the change in GDP to the ratio of change in ΣjDCSj , which gives a relative measure of the trade-off between growth and depletion, denoted by TRD, or broadly interpreted it would give the created income per added unit of degradation. Results from the absolute measure, that is GRN, and the relative measure, that is the trade-off ratio, do not necessarily point in the same direction. The performance of degradation indicator DCSj is decomposable into three elements: changes in yield per feddan, YLDj, changes in sustainable productivity, ASPj, and changes in the quantity of land exploited, NQ j, although it is clear from the model that all three components are interrelated. Higher ASPj enhances YLDj, and results in shifting land from more-costly to less-costly degrading sectors.

Box 4 National accounting balances and additional equations Product market balances Xj = Σ j (αjj' XPj / XPj’ ) Xj′ + Cj + ιj I + νj ICH + Ej

j = 1,..., 6 (19)

Financial market balance (Savings investment balance) Σj (Vj − Cj ) + GSA + FXR . FCF = I + ICH

(20)

Foreign payments balance Σj Ej + Σj αrj Xj = FXR. FCF

(A1)

Effects of land degradation DCSj = ζ j (pj XPj ρj YLDj ) (1 − ASPj ) NQ j Green GDP GRN = Σj Vj − Σi DCSi

j = 1,…,5 (A2) (A3)

Trade-off economic growth and land degradation TRD = Δ Σj Vj / Δ Σj DCSj (A4)

174  Sustained development in CGE model

3  Estimation and baseline forecasts The base year of the model is 1990. A social accounting matrix (SAM) of Sudan for 1990, constructed in the context of the model, provided the required data for most variables and coefficient estimates. The crop mix plays an important part in the present study. Crop data were aggregated following the product mix per sector to give main variables by sectors. This applies to the crucial variables of land cultivated, land productivity, agricultural production, land values, and the sectoral time series of price indices.12 Additional available data for other main variables allowed a calibration for 1990–95, which was used to estimate missing values of some parameters. Various studies undertaken by the Ministry of Agriculture in Sudan allowed estimation of elasticities relating to the sustainable productivity indices. Once the calibration was done, the entire model was solved annually for ten years over the period 1991–2000. In these solutions, the exogenous variables were set to grow at their past growth of 1980–90.13 The model is solved to reproduce the observed values in the year 1990, and solutions for the subsequent ten years. Within the product markets, we make use of two price closure scenarios for the AFL sectors: the situation under unchanging relative prices (U), and the situation under exogenously changing relative prices, that is varying relative prices (V), as projected from world market tendencies. It is likely that the actual relative prices in the future may be somewhere between the two scenarios. Though, because the numerical differences in relative prices between the two scenarios happen to be minor, the impact is minimal.14 Even though the outcomes are not significantly different, comparing the growth and environmental results under unchanging and changing relative prices can give insight into the relevant aspects of world market price tendencies, as to whether these tendencies would favour growth or the environment. Results are shown in Table 8.2 for the closure, with unchanging relative prices (U) in the 5th column and for changing or varying relative prices (V) in the 6th column.15 We consider first the (U) closure. In the irrigated sector, since neither sustainable productivity, ASP1, nor irrigated land, NQ1, undergo changes, other variables of the irrigated sector are shown to grow annually at 4.9 per cent, being fully dependent on the same projected growth of the yield, YLD1. Results show that the mechanised and subsistence sectors will increase their land at the cost of forestry and livestock due to the relatively higher value of land in agriculture, but that relatively more land will shift to subsistence than mechanised. This is because as the income of subsistence farmers grows (indicated by Y3,poor in the table), their cultivation practices improve slightly and their sustainable productivity, ASP3, is raised, while the mechanised sector undergoes no change in ASP2, thus allowing the subsistence yield, YLD3, to grow faster than the mechanised, ASP2, making possible a greater appreciation of the value of land for subsistence, NP3, than mechanised, NP2. Growth of production and income will thus be higher in subsistence than mechanised, but growth in degradation costs is about the same, indicating less degradation per unit of growth in subsistence compared with mechanised. Taking the three sectors of agriculture together, agricultural

Table 8.2 Sudan: annual growth rates of main variables under unchanging and varying relative prices, 1990–2000 Measurement b Base year 1990

Unchanging Varying relative relative prices (U) prices (V)

Index Metric ton Ls M. feddan M. Ls M. Ls

0.700 0.625 56241 5.0 23647 843

0.000 4.900 4.900 0.000 4.900 4.900

0.000 4.900 4.900 0.000 4.900 4.900

Mechanised agriculture Sust. productivity ASP2 Yield YLD2 Land value NP2 Cultivated land NQ2 GDP mechanised V2 Degradation costs DCS2

Index Metric ton Ls M. feddan M. Ls M. Ls

0.463 0.220 20971 7.0 11781 788

0.000 4.900 4.900 1.875 6.867 6.867

0.000 4.900 4.199 1.887 6.880 6.880

Subsistence agriculture Sust. productivity ASP3 Yield YLD3 Land value NP3 Cultivated land NQ3 GDP subsistence V3a Income poor Y3,poor Degradation costs DCS3

Index Metric ton Ls M. feddan M. Ls M. Ls M. Ls

0.393 0.095 2156 9.0 5408 13690 117

0.788 5.727 5.727 2.549 8.422 6.624 7.834

0.757 5.694 4.887 2.446 8.280 6.560 7.717

Forestry ASP4 BTS4 YLD4 NP4 NQ4 V4 DCS4

Sust. productivity Stock/feddan Output/feddan Land value Woodland GDP forestry Degradation costs

Index Cubic meters Cubic meters Ls M. feddan M. Ls M. Ls

0.580 37.1 0.104 1765 152 13104 314

−2.912 −3.320 5.271 −3.320 −0.169 5.093 8.322

−2.913 −3.321 5.271 −3.782 −0.163 5.099 8.329

Livestock ASP5 LVS5 VGT5 YLD5 NP5 NQ5

Sust. productivity Stock/feddan Vegetation/feddan Output/feddan Land value Grazing land

Index Ls Metric ton Livestock Ls M. feddan

0.82 0.293 1.067 0.033 2056 77

−0.890 −1.258 −0.713 1.237 0.046 −0.176

−0.891 −1.257 −0.713 1.232 0.046 −0.171

Symbol

Description

Irrigated agriculture ASP1 YLD1 NP1 NQ1 V1 DCS1

Sustainable productivity Yield Land value Cultivated land GDP irrigated Degradation costs

continued overleaf

176  Sustained development in CGE model Symbol

Description

Measurement b Base year 1990

Unchanging Varying relative relative prices (U) prices (V)

V5 DCS5

GDP livestock Degradation costs

M. Ls M. Ls

7762 169

1.060 4.615

1.059 4.611

−10831 1.0 196508 130955 22631 192660 2234

8.187 2.345 5.131 5.131 5.116 5.181 6.284

8.200 7.161 5.077 5.077 5.003 5.139 6.283

Non-agriculture and whole economy Net exports M. Ls E6 Price index Index XP6 Output M. Ls X6 GDP non-agriculture M. Ls Y6 Investment M. Ls I GDP all sectors M. Ls Σi Vi M. Ls Σi DCSi Degradation costs all

Notes a As was defined in eq. 1.3, the income of subsistence farmers is Y3,poor = V3 + ψ31 X1 + ψ32 X2 + Tg3 b M.= million, Ls = Sudanese pound

output and agricultural value added grow annually by 6.0 per cent, while agricultural degradation loss grows slightly faster at 6.3 per cent. Regarding the forestry sector, continued wood cutting under the pressure of meeting basic needs compels sustainable productivity, ASP4, to fall annually by 2.9 per cent, leading to a greater fall in stock of trees biomass, BTS4, by 3.3 per cent. There is a similar fall in land value, NP4, triggering the shift of forestry land to agriculture that was already noted above. Forestry production and income grow annually at 5.1 per cent, while degradation costs grow annually at the high level of 8.3 per cent. Regarding livestock sector, both livestock, LVS5, and vegetation yield, VGT5, fall, leading to deterioration in sustainable productivity, ASP5, land value, NP5, hardly changes, but some grazing land, NQ5, nevertheless shifts to agriculture. Production and income grow annually at 1.1 per cent, with degradation costs growing at 4.6 per cent. In the non-agriculture sector, output, X6 , and income, Y6 , grow annually by about 5.1 per cent, and the price index, XP6 , by 2.3 per cent. For the economy as a whole, investment I and GDP grow annually by 5.2 per cent. We consider now the model solutions with changing or varying relative prices, the (V) closure. This can be argued to be the more complete closure since it captures substitution effects due to both internal and external stimuli, and thus gives a more comprehensive treatment of land reallocation between the sectors. This closure incorporates future changes in relative prices in reallocating land. The observed post-benchmark trends of world prices for 1990–2000, appear to favour forestry and livestock on mechanised and subsistence. This period is characterised by relative world prices that are environment friendly. The result for the Sudan is a moderation in the shift of land away from forestry and livestock to mechanised and subsistence, when compared with the solutions under unchanging prices (U). The outcome is a lesser degradation, but, also, a lower

Sustained development in CGE model  177 growth. The improved environment is demonstrated to occur at the cost of less economic growth. In agriculture, the irrigated sector remains unaffected. Performance of the mechanised improves and that of subsistence worsens, reflecting the slight positive bias of world prices to mechanised over subsistence. The reduced performance in subsistence reduces incomes and drives down sustainable productivity, land yield, and land value, making it less attractive for subsistence to attract more land. Results for other variables are about the same for both closures. Although degradation increases at a higher rate than income, deducting degradation costs from the GDP gives positive and increasing values of what was termed ‘green GDP’. This is demonstrated in Table 8.3. For the whole economy, GDP and degradation costs grow at slightly lower rates in the changing compared to the unchanging price regime. The reduction in GDP growth is 0.042 per cent per annum (that is, 5.181 to 5.139), while the reduction in degradation growth is 0.001 per cent per annum (that is 6.284 to 6.283). As was stated above, results shows that the world pattern of changing relative prices at the turn of the century can be considered environment friendly, this being at the cost of a reduction in GDP growth. In relative terms, the loss in economic growth can be assessed to be more significant in magnitude than the gain in environment, if equal weights are assigned to a percentage change in growth and environment. Several conclusions can be drawn. First, the model predicts land shifts from forestry and livestock uses to mechanised and subsistence, because of land values under cultivation rising at a greater rate caused by relatively higher sustainable productivities, ASPj , and, to a lesser extent, land yields, YLDj , in agriculture. Second, production and income would increase but so will degradation cost as land transfers from forestry and livestock to mechanised and subsistence agriculture, the latter being more resources-depleting sectors. The average degradation cost per unit of land in mechanised and subsistence agriculture, the sectors to which land is reallocated, exceeds that of the forestry and livestock sectors, the sectors from which land is shifted. In the base year, the average degradation cost per unit of land was 112.6 Sudanese pound (Ls) for mechanised, 13.1 Ls for subsistence, 2.1 Ls for forestry and 2.2 Ls for livestock. In spite of the association of higher growth with higher degradation costs, there is a net increase in the green GDP when degradation costs are deducted from income, as shown in Table 8.3. This tendency is substantiated by empirical evidence from the Sudan and other developing countries at a similar stage of economic development. Third, the model shows slightly lower growth in sector output coupled with lower degradation costs under changing versus unchanging relative prices. Past trends in relative prices at the world level thus favour sustainable productivity in land use at the cost of more economic growth in the Sudan. However, the net effect, if this can be represented by the green GDP, that is GRN, or by TRD, is less favourable under the changing than under the unchanging relative prices.

178  Sustained development in CGE model Table 8.3  Sudan: green GDP and degradation loss by sector of activity under unchanging (U) and varying relative prices (V), 1990 –2000 Sector

Irrigated Mechanised Subsistence Forestry Livestock Economy

Green CDP

Degradation costs / GDP

1990 Base Valuea

2000 Unchanging prices (U) % Δpa

2000 Varying prices (V) %Δ pa

1990 Base

2000 Unchanging prices (U) %Δ pa

2000 Varying prices (V) %Δ pa

 22804.3  10993.4   5291.1  12789.1   7592.5 190426.0

4.900 6.867 8.435 5.001 0.966 5.167

4.900 6.880 8.292 5.007 0.965 5.125

0.036 0.067 0.022 0.024 0.022 0.012

 0.000  0.000 −0.542  3.072  3.518  1.049

 0.000  0.000 −0.520  3.073  3.515  1.088

Note a Values are measured in million Sudanese pounds (Ls)

4  Policy simulations: benefits and costs The economy-wide model with changing relative prices is used to appraise the following policies: i

Educating and training of farmers in the three agricultural sectors (human capital). ii Reducing government tax on the irrigated sector (price incentives). iii Defining property rights in mechanised and forestry sectors (property rights security). iv Making transfers to the poor population in subsistence and forestry sectors (basic needs). Each policy has costs and benefits. The policy of human capital involves costs of educating and training farmers. The price incentives policy is a tax reduction that diminishes government savings. The policy of basic needs involves transfers to the poor population that reduce government savings. The costs of these three policies are taken up in eq. 18. Implementing the policy of secure property rights would involve some cost, but this is relatively insignificant, and is ignored. In all policies, the size of the effort is being put at a comparable level so as to make an unbiased comparative appraisal of policies. The percentage change between policy run results and base values for the year 2000 are presented in Table 8.4.

Table 8.4 Sudan: percentage change of variables between policy runs and base run in year 2000 Symbol Description

Measurement Baseline 2000 Value

Human Capital Δ Value (%)

Price Incentives Δ Value (%)

Property Rights Δ Value (%)

Basic Needs ΔValue (%)

0.7 1.009 90742 5.0 38155 1361

 5.4

10.2

 0.0

 0.0

 5.4

10.2

 0.0

 0.0

 5.4

10.2

 0.0

 0.0

 0.0

 0.0

 0.0

 0.0

 5.4

10.2

 0.0

 0.0

−7.9

−16.0

 0.0

 0.0

0.463 0.355 31642 8.4 22917 1533

 8.2

 0.0

20.2

 0.0

 8.2

 0.0

20.2

 0.0

 8.2

 0.0

20.2

 0.0

−4.1

−2.2

14.9

−8.3

 3.7

−2.2

38.1

−8.3

−3.6

−2.2

14.1

−8.3

0.424 0.165 3475 11.5 11983 25844 248

22.1

 4.1

 2.9

15.6

22.1

 4.1

 2.9

15.6

22.1

 4.1

 2.9

15.6

 8.0

 1.8

−1.6

 6.8

31.9

 6.0

 1.2

23.4

17.5

 6.8

 4.9

25.8

10.4

 2.8

−0.9

 9.2

0.432 26.5 0.173 1201 149.8 21548 701

 0.0

 0.0

 0.2

 1.0

 0.0

 0.0

 0.3

 1.2

−0.2

−0.6

−0.4

−10.0

 0.0

 0.0

 0.3

 1.2

−0.3

 0.0

−0.5

 0.0

−0.4

−0.6

−0.9

−10.0

−0.5

−0.6

−1.1

−10.6

0.749 0.258 0.993 0.038 2047 75.7

 0.8

−0.2

 1.0

 0.5

−0.3

 0.0

−0.4

−0.1

 0.7

−0.1

 0.8

 0.4

 0.7

−0.2

 0.6

 0.4

 0.0

 0.0

 1.0

 0.0

−0.3

 0.0

−0.5

 0.0

Irrigated agriculture ASP1

YLD1 NP1

NQ1 V1

DCS1

Sus. productivity

Index

Yield

Metric ton

Land value

Ls

Cultivated land

M. feddan

GDP irrigated

M. Ls

Degradation costs M. Ls

Mechanized agriculture ASP2

YLD2 NP2

NQ2 V2

DCS2

Sust. productivity Index Yield

Metric ton

Land value

Ls

Cultivated land

M. feddan

GDP mechanised

M. Ls

Degradation costs M. Ls

Subsistence agriculture ASP3

YLD3 NP3

NQ3 V3

Y3,poor

DCS3

Sust. productivity Index Yield

Metric ton

Land value

Ls

Cultivated land

M. feddan

GDP mechanised

M. Ls

Income poor

M. Ls

Degradation costs M. Ls

Forestry ASP4 BTS4

YLD4 NP4

NQ4 V4

DCS4

Sust. productivity Ratio Stock/feddan

Cubic meters

Output/feddan

Cubic meters

Land value

Ls

Woodland

M. feddan

GDP forestry

M. Ls

Degradation costs M. Ls

Livestock ASP5 LVS5

VGT5

YLD5 NP5

NQ5

Sust. productivity Ratio Stock/feddan

Ls

Vegetat/feddan

Metric ton

Output/feddan

Ls

Land value

Ls

Grazing land

M. feddans

continued overleaf

180  Sustained development in CGE model Symbol Description

Measurement Baseline 2000 Value

V5

M. Ls

DCS5

GDP livestock

Degradation costs M. Ls

Non agriculture and whole economy Net exports M. Ls E6 P6 X6 Y6

Price index

Index

Output

M. Ls

GDP non-agr.

M. Ls

I

Investment

M. Ls

Σι Yi

GDP all sectors

M. Ls

Σι DCSi Degradation cost Gov

Gov. revenue

M. LS M. LS

Human Capital Δ Value (%)

Price Incentives Δ Value (%)

Property Rights Δ Value (%)

Basic Needs ΔValue (%)

8625 266

 0.4

−0.2

 1.0

  0.3

−1.8

 0.3

−2.4

  −1.2

−23820 2.0 322451 214766 36876 317993   4109  17023

32.4

26.0

47.1

  1.3

−2.4

−0.2

−2.2

  −16.3

−1.0

−1.4

−2.1

  1.0

−1.0

−1.4

−2.1

  1.0

 0.1

 0.0

 0.1

  0.4

 1.421

 0.287

 1.314

  0.280

−3.507

−6.024

 4.858

  −4.444

−20.1

−25.5

−19.8

−43.7

Simulation I: human capital The human capital policy focuses on eqs. 1.1 to 1.3. Proportions of educated and trained farmers in each of the irrigated, mechanised, and subsistence sectors are adjusted upward by 30 per cent. The impact is that the sustainable productivity indices rise in all agricultural sectors, but mostly in subsistence where performance is above base forecasts by 22.1 per cent. Subsistence farmers, now earning higher incomes, enhance their sustainable productivity further. Yield and value of land in various sectors increase unevenly, leading to land reallocation from mechanised, forestry, and livestock sectors to subsistence. Agricultural GDP grows significantly while degradation loss in agriculture falls. The costs of this policy are a fall in government savings by −20.1 per cent as a result of costs of training farmers and subsidies on exports that have exceeded prescribed limits. These costs are lower than those involved in alternative policies. Greater exports of agricultural goods are compensated by greater imports of non-agricultural goods in a model of balance of payments equilibrium and constant foreign exchange rates. This reduces the growth of the non-agricultural GDP by 1 per cent, as seen in Table 8.4. Forestry GDP, which is dependent on intermediate deliveries to non-agriculture, falls by −0.4 per cent. Attained sustainable productivity, ASP4, is unaffected while degradation costs DCS4 fall. In contrast, livestock GDP rises by 0.4 per cent due to enhanced final consumption demand for livestock products resulting from an elevated growth in overall GDP. The decline of livestock per feddan relieves the pressure on grazing land, and, as a result, the sustainable productivity index, ASP5, improves while degradation costs, DCS5, fall. In summary, the policy of investing in human capital, achieves a greater economic growth on the base forecast of 1.4 per cent, and a mitigation of degradation of 3.5 per cent. The policy contributes significantly to the reduction of the trade-off between growth and environment.

Sustained development in CGE model  181 Simulation II: price incentives This policy intervention focuses on the irrigated sector in eq. 1.1. The government tax rate on the price of irrigated product is reduced from 0.30 to 0.21, that is, by 30 per cent. Attained sustainable productivity rises by 10.2 per cent. As result, yield and land value rise at the same rate. Degradation cost falls by 16 per cent. As subsistence farmers get additional earnings from working in the irrigated sector their income, Y3,poor , increases by 6.8 per cent. This causes the attained sustainable productivity in subsistence agriculture to rise by 4.1 per cent, as do yield and land value, leading to a reallocation of land from mechanised to subsistence agriculture. The policy results in a net decline in degradation loss. The costs of this policy consist of a decline in government revenue by 25.5 per cent due to forgone tax income, and the cost of subsidising agricultural exports that have exceeded their upper limit. The GDP of the non-agricultural sector is negatively affected along the same lines as in the previous policy simulation and would fall by 1.4 per cent. This causes a further fall in the output of forestry accompanied by a reduction in degradation loss. The livestock output is also diminished due to less slaughter and the slow economic growth. These effects result in overstocking, a fall in the grazing land sustainable productivity index, and a rise in the degradation loss. The overall impact of this policy is that the GDP of the whole economy grows by 0.3 per cent while total degradation cost falls by 6 per cent. The previous policy of human capital thus performs better than the policy of price incentives. Simulation III: property rights The policy intervention here focuses on eqs. 1.2 and 1.4. The measure of land tenure security in the mechanised sector, denoted by LTS2, is increased from 0.50 to 0.65, that is, a 30 per cent upward adjustment in lease term. The impact is that the sustainable productivity index in the mechanised sector increases by 20.2 per cent, as do yield and value of land. Consequently, land moves from subsistence, forestry, and livestock activities to mechanised agriculture, where the value of land is rising faster relative to other land-using sectors. With the movement of land, degradation loss increases in mechanised agriculture since it is the sector with the highest degradation cost per unit of GDP. In subsistence agriculture, the productivity index increases by 2.9 per cent, as do yield and value of land. At the same time, subsistence poor farmers income, Y3,poor , is boosted by growth in mechanised agriculture. Degradation loss decreases as a result of land moving out of the sector to mechanised agriculture. Government revenue falls by 19.8 per cent, since it incurs the cost of subsidising agricultural exports that have exceeded their upper limit. The balance of payments adjustments lead to a negative growth in non-agriculture of −2.1 per cent. We have introduced a similar property rights effect in forestry, not reported in the model specification. This allowed the stock of tree biomass per feddan to increase and, as a result, the sustainable productivity index rose while degradation loss declined. In the livestock sector, due to a higher consumption of meat

182  Sustained development in CGE model caused by higher GDP growth, the sustainable productivity index rises, causing vegetation yield to rise and degradation loss to fall. GDP of the whole economy rises by 1.3 per cent, but degradation costs rise by 4.8 per cent. This is explained by the movement of land to mechanised agriculture, which has the highest degradation cost per unit of GDP among the land-using sectors. This policy has a more positive impact on achieving economic growth than that of price incentives, but it has higher degradation costs. The trade-off is intensified. Simulation IV: basic needs This policy intervention focuses on eqs.1.3 and 5.4 for the subsistence and forestry sectors, respectively. Income of the poor population in these two sectors is adjusted upwards by 30 per cent through a direct government transfer, Tg. The impact on subsistence agriculture is that the attained sustainable productivity, yield, and price of land rise by 15.6 per cent. Land is moved from mechanised to subsistence agriculture. Degradation costs fall in mechanised agriculture where land is moving out, but increases in the subsistence sector where land is moving in. Government revenue falls by −43.7 per cent as a result of making transfers to poor farmers. In a similar way, with an income transfer, forestry workers are less inclined to cut trees, so the tree stock biomass per feddan increases, attained sustainable productivity increases, and degradation costs decline. The sectoral income from cutting wood trees falls by the amount of the government transfer, that is, −10 per cent. GDP increases by 0.3 per cent, while total degradation loss falls −4.4 per cent. The policy of making direct government transfers to the poor population in the subsistence and forestry sectors brings about more economic growth with less degradation. Population dependent on the livestock sector is also better off.16 However, the transfer costs involved are highest compared to other policies. The four policy simulations compared The relative performance of the four policies in relation to the base forecasts is demonstrated graphically in Figures 8.1 and 8.2. Visualize in Figure 8.1 a diagonal line with an inclination of 45 degrees. Policies that lie most below this diagonal line and are most stretched to the right contribute more to economic growth and to a reduction in degradation loss, and, therefore, mitigate the tradeoff between growth and environment. Policies that lie above the diagonal line thus perform least in resolving the trade-off. All the simulated policies, except that of property rights security, contribute to economic growth as well as reduce degradation loss, when measured from the baseline. When the relative effectiveness of policy is evaluated in terms of this trade-off, the most effective policies across the 1990s are those of strengthening human capital and price incentives, with human capital ‘better’ in GDP, and price incentives ‘better’ on environment. These are followed by the policy of basic needs. The policy of property rights, which lies above the 45-degree line, tends to intensify the trade-off by promoting economic growth at the cost of more degradation loss.

4.2343 3.9843 3.7343 3.4843 3.2343 2.9843 2.7343 2.4843 2.2343 192.6603

212.6603

Baseline

232.6603

I. Human capital

252.6603 272.6603 GDP (Billion LS) II. Price incentives

292.6603

III. Property rights

312.6603 IV. Basic needs

Figure 8.1 Sudan: trade-off between economic growth and degradation loss, 1990–2000

0.300 0.200 0.100 0.000 I

−0.100

II

III

IV

−0.200 −0.300

Figure 8.2 Sudan: degradation loss: difference between policy and baseline, 2000; billion Sudanese pounds

5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0

GDP Green GDP

I

II

III

IV

Figure 8.3 Sudan: GDP and green GDP: difference between policy and baseline, 2000; billion Sudanese pounds

184  Sustained development in CGE model In Figures 8.2 and 8.3, the policy numbering (that is I, II, III, IV) is the same as that shown in the legend at the bottom of Figure 8.1. Figure 8.2 shows the degradation loss due to policy intervention measured from the baseline for the year 2000. As shown, the policies of human capital (I), price incentives (II), and basic needs (IV) reduce degradation costs compared to the baseline of the year 2000. The policy of property rights security (III) increases degradation costs. Figure 8.3 shows the effect of policy intervention on GDP and green GDP (that is GDP after deduction of degradation loss) as measured from the baseline for year 2000. The policy of human capital, that is (I), produces the highest green GDP. Price incentives (II) and basic needs (IV) show positive but moderate changes in green GDP. The real surprise here is in the policy of property rights security (III) which shows a strong positive performance in green GDP, reversing the impressions obtained from Figure 8.1. When deterioration of the environment is monetised and is allowed to be traded against the gains in this policy, it climbs to second place in terms of green GDP.

5  Concluding remarks Basically, traditional uses of land, that is subsistence cultivation, forestry, and grazing activities, have lower degradation costs per unit of output relative to the modern land uses, that is, irrigated and mechanised. Economic development involves a shift of land from the first to the second. Therefore, reallocation of land from the traditional uses to the modern uses leads to more production, but at the same time it involves more loss in terms of resource depletion. Fortunately, there are policies that can mitigate this trade-off between economic growth and resource degradation. The policy of investment in human capital and training programmes appear to be the most effective in combating the trade-off and resolving the conflict between growth and environment. This policy is followed by price incentives and income transfers to maintain basic needs. The policy for promoting greater security in property rights favours growth at the cost of environment, but scores high in terms of green GDP. The specification that takes the economy-wide model and incorporates the effects of changing relative prices based on world market prices trends represents a workable assessment of the trade-off between growth and environment. The observed trends in relative prices at the world level at the turn of the century (1990–2000) seem to favour environment, but the effect is very marginal. The cost to economic growth of the environment-friendly relative prices is relatively greater, however. The net effect is a lower green GDP in the Sudan. Given the present distribution of land use in the country, the results suggest that there is no need to worry about reaching a critical ecological balance between growth and environment any time soon.

9 Simulation results of SAM models for transiting economies Russia falls and China rises

1  Comparative analysis of economic systems This chapter is devoted to developing and applying the framework of the social accounting matrix (SAM) as an analytical tool in evaluating the performance of economic systems. We shall begin in this section by laying out basic elements of an evaluation framework, reviewing alternative analytical frameworks, and finishing up by choosing the SAM as a comprehensive and an easily workable model for evaluating comparable economies. Once this is done, the chapter will apply a comparative analysis of the economic performances of the Russian and Chinese economies in the midst of their economic transitions. Comparative assessments of economic systems distinguish between several concepts that are often brought together, as is done in Figure 9.1. They are: (1) Economic system (meaning the internal structures); (2) Outside forces (such as environment and the state); (3) Outcome of economic activity; (4) Criteria for judging the performance of economic systems; and (5) Assessment of performance as being the result of outcome and criteria. There is no one commonly accepted view on what comprises an economic system. Most economists though adopt the conventional view that the economic system encompasses four complementary internal structures, consisting of the agent structure comprising the profile of the agents and their preferences, the institutional structure that includes rules guiding the behaviour of agents, the information structure serving to inform agents about the environment in general and the actions chosen or contemplated by other agents making interrelated decisions, and the technological structure defining transformation boundaries within which agents transform value added. These topics are discussed at more length in the final chapter of this book. The outside forces having an impact on the economic system are often taken to be the environment and the state. External influences exerted by other countries, and random factors such as the weather and natural disasters, are part of the environment. The environment is relatively less controllable than is the state. State policies obviously affect economic activity. They are not independent of the economic system, but they are not fully endogenous either. The economic system and the external forces work together to determine the outcomes of economic activity, which are commonly considered to be the

186  SAM models for transiting economies production of goods and services, their distribution among economic agents, and their subsequent consumption by these agents. The outcomes are not static, since they are supposed also to include provisions for future production, distribution, and consumption. To make an assessment of performance of an economic system, performance criteria need to be specified. A distinction can be made between prevailing norms and the norms of those making comparisons. The prevailing norm represents the norm or preference function of those decision makers who actually resolve the economic decisions. The norm of the analyst making the comparison represents the norm adopted by a given student of economic systems for the purposes of comparison. Thus, the analyst could focus only on production efficiency or the growth rate of GNP in a comparative study of different economies. The point is that no discussion of a system or of an economy’s performance can proceed in an orderly manner unless the norm used to evaluate performance is made explicit. Performance criteria from the viewpoint of the one making the comparison can be based upon conceptual notions of static efficiency of production, consumption, and distribution (mainly based on Pareto optimality conditions), and dynamic efficiency (including inter-temporal allocation between consumption and investment, as well as inter-temporal distribution). Performance criteria can also take the form of operational indicators such as economic growth, factor productivity, income equality, poverty reduction, consumer satisfaction, stability, freedom, and so forth. While conceptual notions are essential as a tool of analysis and for theoretical understanding, they are not easily convertible into quantifiable indicators that can be applied for individual countries and that represent specific economic systems. Operational indicators are easier to apply. We shall focus on two operational indicators: growth and equity. The diagram in Figure 9.1 can be formalised in an equation form as is done in eq. 1. Economic order discussions are often formulated in terms of the following equation. See, for instance Koopmans and Montias (1971).

External environment

Internal environment

Economic System

State

Policies

Economic Outcome

Performance criteria

Assessment of economic performance

Figure 9.1 The analytical framework for comparing performances

SAM models for transiting economies  187 Outcome = function of (system, external environment, government policies)

(1)

This equation is valid for the medium to the long run. In the longer perspective, there will be more interactions among the four sets of variables. For example, the outcome over many periods can influence the form of the state and its policies in the future. Environment also has cumulative effects on the state and its policies. Further, the system is not independent of the outcome in the very long run. The quantitative analysis of the economic performance of a specific economic system always has to be approximated in terms of real economies that are aligned to that specific economic system. The set of national economies can be denoted by Y and the particular national economy observed is y, and there are many national economies y,…, y'. The economic system type can be denoted by M, and there are several of them, m,…, m’. The economic outcomes are expressed in indicators forming an outcome set V, consisting of elements v,…, v'. In the present context, two V variables are of special interest; these are economic growth and income equality. Variables of the external environment and government policies, both to be assumed exogenous, will be denoted by E, and they include e,…, e'. The economic outcome of a specific country, Vy , is thus explained in terms of three factors: the identifiable economic system to which the country belongs, My , the external environment and government policy variables circumventing the country, Ey . Figure 9.1 and eq. 1 can now be expressed more concretely as in eq. 2 that specifies that outcome for observed countries is dependent on the systems to which they belong, and variables of the external environment, and government policies. Vy = function of (My , Ey )

(2)

Economists have applied the analytical framework of Figure 9.1, the general equation, eq. 1, and the quantifiable equation, eq. 2, to different themes and at different levels of aggregation when comparing countries. With the object of elaborating a general systems approach, we propose to modify eqs. 1 and 2 towards eqs. 3 and 4, respectively. The refinement introduced is consistent with the reduced-form solution of a system of equations. Outcome = (system multipliers) (environment, policies) Vy = (system multipliers of My ) (Ey )

(3) (4)

What we have in eq. 4 is a more specific multiplicative relationship between the inverse of the system of equations—the system multipliers—and the exogenous variables that stand for the external environment and government policies. One may extend the analogy with the modelling jargon and denote the exogenous variables relating to the external environment as uncontrollable and those relating to government policy as controllable. The solution of the economic model, read the economy outcome, is the result of a multiplication of the reducedform multipliers of the economy in question by the vector of exogenous variables facing that economy.

188  SAM models for transiting economies Practically speaking, almost all comparative holistic studies of economic systems are guided in their application by eq. 2. The empirical literature is very scanty as regards guidance by eq. 4, since that will require comparing system multipliers of two countries. Nevertheless, eq. 4 is the more comprehensive formulation, and is the one chosen for this chapter. It will be demonstrated that it is more useful, and more easily accessible, to work with eq. 4. For applying a comparative multiplier approach, a number of issues have to be resolved. Choosing a general system approach requires specifying a concrete model for a concrete economy. Recall eq. 4 above, giving the reduced-form solution of an economic system. This can simply be written as: Vy = My (Ey ) (5) where Vy denotes outcome, My is for system multipliers, Ey is for the exogenous variables; all relate to a country y that can be identified with a specific economic system, My. The question now is along which line of specification should the economic system be modelled in the first place? Broadly speaking, there are three prototypes of models which can connect V, M, and E to each other: (1) models that emphasise the supply side, production functions, and production capacity; (2) models that emphasise the demand side, purchasing power, and expenditure patterns; and (3) general equilibrium models that equally emphasise demand and supply forces in the clearance of factor and product markets. General equilibrium models are more comprehensive than supply-side or demand-side models, and can be usefully employed for a comparative analysis of economic systems that share competitive rules of market clearance, but they lose applicability in cases of economic systems where market clearance is restricted or highly structured. Among the remaining two alternatives—of a supply model and a demand model—there are more advantages in working with demand rather than supply models. Demand models give outcomes for both performance criteria: growth and equity. Supply models only treat growth. Demand models can be easily supplemented by supply constraints, while it is more difficult to satisfy the contrary situation. Measurement, testing, and predictions in demand models are more robust than in the case of production functions. Given the above considerations, the choice here is for the demand-oriented model as an analytical framework. Since national accounts supplemented by industry, household, and government statistics can be conveniently integrated in the form of a SAM, and because the form of the SAM is that of a general economic system, it is therefore very well suited as an analytical framework. After appropriate manipulations, already dealt with in Chapter 4, the inverse of the coefficient matrix of the SAM would give the SAM multipliers MS in eq. 5, above. Using a comparative analytical framework for compared countries, the SAM can be helpful in deriving normalised patterns and indicating prospective patterns for economies in transition. The SAM multiplier analysis can give insight as to which economic mechanisms needs to be strengthened to push a specific economy closer to its prospective patterns. This will be demonstrated for the contrasting performances of Russia and China for a crucial year in their economic transitions.

SAM models for transiting economies  189 The use of SAM as a model that generates multipliers can be demonstrated from a very simple example. Take the simplest Keynesian model, which contains an equation relating consumption to income via a propensity to consume, and an equation defining income as consumption plus exogenous investment. This is thus a model of two equations in two endogenous variables of consumption and income. The model can be written as a square matrix which is then inverted to give a Keynesian multiplier showing the impact of a change in investment on income. Similarly, in an input-output analysis an endogenous vector of economic activities, v, can be predicted from a Leontief matrix of input-output coefficients, AL , and a vector of exogenous final demand, e. That is, v = AL v + e = (I − AL)−1e = ML e, where ML is the Leontief multiplier matrix. The SAM is also a square matrix but it is larger in content as it covers the whole circular flow, economy wide. Being a square matrix, the SAM can be operated as a model of the economy. By appropriate manipulations of this square matrix, it is possible to derive SAM multipliers that are more comprehensive than are those of Keynes and Leontief together. To transform the SAM into an economy-wide model requires performing several steps. Assuming proportional relationships for the cells in terms of their column totals, a SAM coefficient matrix that relates variables to each other is obtained, call it AS. This is similar to AL but is more comprehensive in coverage. By separating the variables in the SAM into an endogenous vector v and an exogenous vector e the SAM model can be written as v = ASv + e. Inversion of the SAM coefficient matrix would give eq. 6, where MS is the SAM multiplier matrix. v = (I − AS )−1e = MS e

(6)

How is the chapter further organised? Section 2 presents salient data on what can be called China rises and Russia falls. Section 3 introduces the SAMs of both countries. Section 4 runs a multiplier analysis for Russia and China with a focus on economic growth potential, income equality prospects, and comparative policy implications. Section 5 extends the analysis towards identifying gainers and losers in Russia and China; Section 6 concludes.

2  Salient differences in economic performance: Russia and China Even though the comparative analysis for Russia and China undertaken here is based on the SAM benchmark for one year only, around 1990, the results obtained show consistency and durability that are supported by contrasting trends in the two countries over some earlier decades, and during transition, and after. Before dealing with the SAM analysis, some comparative data on the fall of Russia and rise of China are displayed. Russia’s GDP—measured at constant prices of 2000 in US dollars—grew between 1979 and 1989 by 43.2 per cent and then decreased between 1989 and 1997 by about 46 per cent, then gradually recovered to reach a level in 2006 that is roughly twice the level of 1979, Table 9.1 row 4. In contrast, since the start of the economic reforms in China, dating from 1979, China’s GDP increased constantly and became 2.5 times as much by 1989. The growth continued at

190  SAM models for transiting economies Table 9.1 China and Russia: GDP and indices of GDP

China Russia China Russia China/Russia

Constant prices 2000 US dollars billion Constant prices 2000 US dollars billion (GDP for 1979 = 100) (GDP for 1979 = 100) GDP China/ GDP Russia

1979

1989

2000

2006

170 278 100.0 100.0   0.6

428 398 251.8 143.2   1.1

1198  382  704.7  137.4   3.1

2092  544 1230.6  195.7    3.8

Source: World Bank at http://devdata.worldbank.org/query/

Table 9.2 Output of major goods in China and Russia Item

Electric energy (bln kwh) Steel (mln tons) Synthetic fibres (mln tons) Mineral fertilisers (mln tons) Tractors (000) Television sets (mln) Washing machines (mln) Textile fabrics (bln m) Grain (mln tons) Meat (mln tons) Unweighted average

China

Russia

1990

1995

% change 1990

1995

% change

620  63.5   1.6  18.7  39  26.8   6.6  18.8 355.0  25.1

1000   94.0    2.9   24.5   63   34.7    9.4   21.0  417.0   42.0

61 48 81 31 62 29 42 12 17 67 45

862  51.3   0.2   7.5  21   0.98   1.3   1.7  63.5   3.4

−20 −43 −66 −53 −90 −79 −76 −80 −41 −65 −61

1082   89.6    0.6   16.0  214    4.7    5.4    8.4  107.8    9.8

Sources: Russian Statistical Yearbook, several years; Economics of Transition, vol. 4, No. 1, 1996, p.289; China Statistical Yearbook, several years; World Economic and Social Survey, 1994, United Nations, p.259.

annual rates of around 9 and 10 per cent, raising the GDP level to about 12 times its level of 1979, Table 9.1 row 3. In 1979 China’s GDP was about 60 per cent of that of Russia. Ten years later in 1989 they were about equal. The relative sizes of the two economies have since then reversed position in historically unmatched terms during less than two decades. In 2000, the GDP ratio of China to Russia was 3.1; in 2006 the gap widened to 3.8, Table 9.1 row 5. Most of the increase in the gap between Russia and China occurred during the period of the short transition in Russia. The contrast between the recession downfalls in Russia and the growth drive in China and the extent of their, respectively, positive and negative performances can be comprehended from a comparison of the output (in physical terms) of some ten major goods, as is shown in Table 9.2. A non-weighted average gives an increase of 45 per cent over five years for China, and a decrease of 61 per cent for the same years for Russia.

SAM models for transiting economies  191 The contrast in the economic performances of these two major countries has persisted for a long time and shows constancy, even in the periods of reform, suggesting that the differences in the structures and mechanisms behind these trends are endurable and can fruitfully be subjected to comparative systemic analysis. The SAM framework for comparative analysis is tested below and is found to be suitable for this context.

3  The SAMS of Russia and China The aggregate SAM for Russia is constructed from the published estimates of the national accounts for 1990. These accounts have been disaggregated further into five factors, five household groups classified by income ranges, firms, government, aggregate capital account, four commodities, and three production activities, and the ROW, together resulting in a SAM of 21 rows by 21 columns, as found in Annex Table 9.7 at the end of the chapter. The required data for disaggregating the SAM included: (a) the household budget survey for breaking up the household account into the five income groups, specifying their incomes by source, and their expenditures by type of product; (b) the input-output table for disaggregating the production activities; and (c) an initial converter table for transforming a product classification into a sector classification. The household budget survey provided distributional structures of receipts and expenditures by household groups. These are multiplied by the number of households and applied to the aggregate household account to give the disaggregated detail. Data in the input-output table are aggregated to suit the classification into three sectors. The absolute values thus obtained fit directly within the disaggregated SAM. Regarding the converter matrix between products and sectors, this was constructed in a preliminary way from the codes of the household budget survey and the input-output table, and later subjected to several adjustments to ensure consistency of the grand totals of its rows and columns, solved by applying the RAS method. For China, we have constructed a comparable aggregate SAM and a disaggregated 19 rows by 19 columns SAM for 1989, displayed in Annex Table 9.8 at the end of the chapter. It consists of about the same accounts as for Russia, except that there is less disaggregation in the factor and household accounts. Transforming the SAM into an economy-wide model and computing SAM multipliers is done in two steps. First, accounts of the SAM are separated into exogenous and endogenous and then regrouped so that expenditures and receipts of the endogenous accounts are placed on the left side of the equality signs, while those of exogenous accounts would fall in the right side of the endogenous accounts, see Chapter 5. We follow here an established convention for basically centrally planned economic systems that assumes the expenditure accounts of capital, government, and ROW as exogenous. For the Russian SAM of 21 rows by 21 columns, the separation of the three exogenous accounts leaves over an endogenous component constituting 18 variables. The endogenous accounts form an 18 by 18 sub-matrix containing all of the flows from endogenous to endogenous

192  SAM models for transiting economies accounts. For the Chinese SAM of 19 rows by 19 columns, the separation of the three exogenous accounts results into an endogenous component of 16 variables. Second, we assume proportional relationships for the sub-matrix cells in terms of their column totals, so that a SAM coefficient matrix, AS , is obtained. Inversion of this matrix gives the SAM multipliers MS , showing how the endogenous variables will respond to a unit change in the exogenous variables.

4  SAM multipliers in Russia and China Given the above sizes of the SAMs, the sizes of the multiplier matrices are pretty large and for analytical purposes a selection of multipliers is necessary. This chapter follows a selection of multipliers that was also the focus in Chapter 5, namely, multiplier effects of exogenous spending injections in sector activity j' and of exogenous income transfers to the household group h', on the output of activity sector j, and on the income of household group h. Together, we thus distinguish four multipliers to be studied. Table 9.3 gives the multiplier effects of exogenous spending injections in sector activities j' on output variables MS,jj' and income variables MS,hj'. The results show that for Russia, a spending injection in the sectors, on average, of say one billion roubles (br) has a multiplier effect on output of 2.81 br, and a multiplier effect on income of 0.62 br. On average, an income transfer to household groups of 1.0 leads to a combination of an output multiplier of 2.09, with an income multiplier of 1.40. The corresponding results for China in Table 9.4 show that spending injections lead to output and income multipliers of 3.26 and 1.20, while income transfers lead to output and income multipliers of 2.84 and 1.66, respectively. China’s performance is higher than Russia’s in all four respects. In general, the size of the multipliers of an inverted matrix is relatively high if the inverted SAM coefficient matrix, that is, the endogenous part, represents a large share of the economy, and correspondingly, if the exogenous part represents a small share. Multipliers are relatively low if the endogenous share is small and the exogenous share is large, as this exogenous share is not ploughed back into the economy. The exogenous share in the SAM, consisting of investment, government, and ROW, will generally depend on the economic system, the development level, and the size of the country. The shares of investment and government are expected to be greater in planning-oriented economies, especially among those with a larger defence budget. Economic theory also predicts a greater share of investment, government, and ROW at more advanced levels of economic development. Countries with a larger population tend to have lower shares of transactions with ROW, other things remaining the same. Given the above, it is not surprising that the exogenous share as defined here is higher in Russia than in China. This is also evident from the SAMs that show a higher exogenous share for Russia than China, with respectively 19.6 and 14.7 per cent of the economy considered exogenous. The corresponding endogenous shares are 80.4 and 85.3 per cent for Russia and China, respectively, implying a lesser circular flow in Russia than in China. As a result, the SAM multiplier should be expected to be smaller in Russia than in China, as is shown in the tables.

SAM models for transiting economies  193 Table 9.3 SAM multipliers of Russia 1990 Effects

Multipliers of spending injections into sectors Agriculture

Total output 2.95 Total income 0.76

Industry

Services Average Highest/ Lowest

3.05 0.52

2.44 0.59

2.81 0.62

1.25 1.46

Multipliers of income transfers to household groups 400 2.26 1.43

2.15 1.41

2.08 1.39

Average Highest/ Lowest 2.09 1.61 1.40 1.14

1.52 1.29

Table 9.4 SAM multipliers of China 1989 Effects

Total output Total income

Multipliers of spending injections into sectors Agriculture Industry

Services

Average

Highest/ Lowest

3.70 1.66

3.00 1.06

3.26 1.20

1.23 1.87

3.07 0.89

Multipliers of income transfers to household groups

Total output Total income

Rural farm

Rural Urban Urban self- Average non-farm employees employed

Highest/ Lowest

3.30 1.77

3.08 1.72

1.78 1. 24

3.12 1.73

1.85 1.43

2.84 1.66

However, there remains one question requiring an answer. Which of the two countries is more successful in generating more output, and more income, per 1 percentage point of the endogenous share? It can be calculated, on average, that a spending injection in Russia gives an output multiplier of 2.81 for an endogenous share of 80.4 per cent, implying an output multiplier of 0.035 for each endogenous percentage point. China’s performance is greater in this respect, (that is 3.26 / 85.3 = 0.038), as it is able to achieve from an equivalent spending injection an output multiplier of 0.038 for each endogenous percentage point. The difference amounts to a positive edge of about 10 per cent, (that is 0.038 / 0.035). This edge can be interpreted as a more efficient use of the circular flow of the economy. Why does China surpass Russia in the efficient use of the circular flow? Three related reasons can be given. First, a better-knit economy, in the sense of having more extensive and intensive transactions between its agents is characterised by

194  SAM models for transiting economies more transactions and a more complete SAM. The more that the SAM cells are filled, the greater is the multiplier effect. The extreme situation of an autonomous sector that produces and supplies exclusively for its own employed-labour households, and who buy exclusively from this sector, will show very low multipliers. Centrally planned economies tend to save on transactions and emphasise autonomy. This stands in contrast to free market economies which propagate more transactions and a higher turnover of circulating funds. While both Russia and China share features of a command economy, there is general agreement in the empirical literature that the Russian system has been less forthcoming than the Chinese system has in creating multichannels for the flow of goods and services, and in creating a more flexible framework for resolving imbalances. Second, there may be specific flows with higher multiplier effects than others, but in the longer run an economy which manifests little structural change will tend to show less variation in the multiplier effects of alternative injections. In contrast, a rapidly moving economy undergoes frequent structural changes and the variation in multipliers is bound to be greater. The introduction of a new activity or flow extends the circular flow by that activity and links with other activities resulting in a widening of the variation between multipliers, as well as higher overall multipliers. Russia has been less successful than China in modernising its economy and extending the circular flow; this is associated with a lower range of multiplier values in Russia as compared to China. We tried to demonstrate these tendencies for Russia and China in Tables 9.3 and 9.4. The highest and lowest ratios of output multipliers show about the same value for Russia (1.25) and China (1.23). However, the disparity in the income multipliers for Russia is lower (1.46) than for China (1.87), suggesting more replication of the status quo in Russia as compared to China. A related remark is with respect to the size of the multipliers from spending injections in alternative sectors. Comparative SAM multiplier results for developing countries (see Chapter 5) and also for European countries (see Chapter 11) show that spending injections in the sectors of services and agriculture have the highest output and income multipliers, while a spending injection in industry lags behind in its multiplier effects. These results are due to the high expenditure on food and services and greater earnings flows in services as compared to industry. While the results for China are in general agreement with the findings for other countries, those for Russia differ as they show a predominance of the output multiplier for spending injections in industry. These results are due to a restricted circular flow in Russia that tends to downgrade the role played by the sectors of agriculture and services and foregoes, as a result, potentially higher multiplier effects. For assessment purposes, the income multiplier is a more relevant concept than the output multiplier for two reasons: (1) earned income is closer to the efficiency notion of value added than gross output, and (2) earned income by household groups is a better indicator of social welfare than gross output. Assessment of the income multipliers of Russia and China leads to two remarks. First, the average income multiplier from demand injection for Russia is found to be 0.62 that is achieved at an endogenous share of 80.4 per cent, implying a multiplier of

SAM models for transiting economies  195 0.0077 for each (one) endogenous percentage point. Applying this norm to China should result in an income multiplier of 0.66 as compared to the SAM income multiplier of 1.20, which is a significant edge. A similar calculation for Russia on the basis of China would give a normalised income multiplier for Russia of 1.3 as compared to the SAM income multiplier of only 0.62. The conclusion is that China fares better than Russia does in aggregate income multiplier effects as well. Furthermore, the ratio of income to output multipliers is 0.22 and 0.37 in Russia and China, respectively, supporting the hypothesis of greater leakages of value added and (or) a lower efficiency in factor use in Russia as compared to China. Attention can now be directed to the multiplier effects of income transfers to household groups. These are generally consistent with the preceding results for spending injections, showing output and income multipliers of income transfers to household groups to be lower in Russia, 2.09 and 1.40 respectively, than in China where they reach 2.84 and 1.66, respectively, even though the performance edge tends to be narrowed to the extent that the income: output ratio of income transfers is reversed in favour of Russia, reaching 0.67, as compared to China of 0.58. In the light of the preceding results this outcome can be interpreted to mean that transfer payments to household groups in Russia occur in an economy with a relatively less intensive and extensive circular flow, and with more emphasis on the direct rather than indirect effects, and hence converting transfer payments directly into higher income rather than higher output effects. The Chinese economy, in contrast, allows the transfer payment to be turned over more intensively and extensively permitting more output and more income.

5  Gainers and losers in Russia and China Besides comparing the levels of the multipliers, it is also important to study the dispersion of the multiplier effects over the respective sectors and households and to assess the underlying structural bias that determines gainers and losers in these two countries. For these we have developed the gainers and losers index, GLI, already introduced in Chapter 4. There are four GLI, corresponding with the four multiplier effects that are the focus of the analysis. The dispersion impact of a spending injection in sector j' on the output of each activity sector j, is denoted by GLIjj' , and on the income of each household group h is denoted by GLIhj'. In correspondence with these, there are two types of GLI following an income transfer to household group h'. These are gainers and losers indices among the activity sectors experiencing an impact, GLIjh' , and gainers and losers indices among the household groups GLIhh' that experience an impact. The formulas for the four indices are briefly displayed in Box 1 in eqs. 7 to 10; see also Chapter 5. Values of 1 are neutral, in the sense that the multiplier effect reproduces the same relative position in the base year of the account experiencing an impact. Values above 1 identify gainers, and below 1 identify losers. Table 9.5 shows all four indexes for Russia. Results show that the effect of exogenous spending in sectors rewards the agricultural sector positively, the value of GLI being between 1.06 and 1.89. In contrast, the results show a negative

196  SAM models for transiting economies Box 1 Gainers and losers index, GLI GLIjj′ = [(Ms,jj’ − δjj′ ) / (Σj Ms,jj’ − 1)] / [Outputj,o / Σj Outputj,o ] GLIhj’ = [(Ms,hj’) / (Σh Ms,hj’)] / [Incomeh,o / Σh Income h,o ] GLIhh’ = [(Ms,hh’ − δhh′ ) / (Σh Ms,hh’ − 1)] / [Incomeh,o / Σh Income h,o ] GLIjh’ = [(Ms,jh’ ) / (ΣjMs,jh’)] / [Outputj,o / Σj Outputj,o ]

(7) (8) (9) (10)

growth bias for the services sectors of GLI between 0.69 and 0.74, and a uniform replication of the share of industry with GLI being 1.0. Considering the effects of the same exogenous spending in sectors on income distribution among receiving household groups, the results show injections in the various sectors to have the same effect of high repressiveness. The poorest household group comes off badly with GLI around 0.7. Most benefits go to the richest groups, which are calculated to score GLI of 1.05 or more. Next we may consider the GLI of exogenous income transfers to household groups. The pattern is the same as was found for spending injections. Among the sectors, the share of agriculture ends up better off, services worse off, and industry is unaffected. Among the household groups the poorest are disfavoured, GLI = 0.7, while the richest are favoured with GLI = 1.05. Nevertheless, that the actual income distribution in Russia shows more equality than the SAM multipliers demonstrate is due to the positive effect of annually repeated initial injections to the poorest household groups. The results for China can be now reviewed from Table 9.6. On average, spending injections in China favour both industry and agriculture, but industry more so than agriculture; this is in contrast with Russia which does the opposite. Neither country favours services. Income redistribution effects are uniform for alternative spending injections. The effects favour rural households GLI > 1.0, and disfavour urban households GLI < 1.0, and to the extent that the poorest population lives in rural areas the multiplier effects can be interpreted to promote more income equality. It was observed before that the results for Russia show the opposite GLI effects as regards income distribution. These results reveal sector earning and household expenditure patterns and mechanisms, which redistribute income towards the richer groups in Russia as opposed to a redistribution towards poorer groups in China.

6  Summary and conclusions The use of the SAM as a framework for the comparative analysis of systemic differences in economic performance serves as an important tool in outlining and checking the differences, and gives valuable insight into the patterns and mechanisms that cause the differences in performance. The SAM was applied to analyse the size and the structure of multiplier effects of demand injections into sectors, and of income transfers to households. We focused on multiplier effects relating to economic growth and income equality. The growth and equity

SAM models for transiting economies  197 Table 9.5 GLIs Russia 1990 Gainers and losers

Exogenous spending injections into activity sectors Agriculture Industry Services

GLIjj′ gainers and losers by sector Agriculture 1.89 Industry 0.99 Services 0.74

1.31 1.10 0.74

1.06 1.17 0.69

GLIhj′ gainers and losers by household groups Hh. 400 roubles pm 1.06

0.74 0.93 0.99 1.01 1.05

0.73 0.92 0.99 1.01 1.05

Exogenous income transfers to household groups 400

1.41 1.16 0.61

1.37 1.16 0.62

1.35 1.16 0.62

1.31 1.15 0.66

GLIhh′ gainers and losers per household groups Hh. 400 roubles pm 1.05 1.05

0.72 0.92 0.98 1.01 1.05

0.72 0.92 0.98 1.01 1.05

0.72 0.92 0.98 1.01 1.05

GLIjh′ gainers and losers by sector Agriculture 1.50 Industry 1.15 Services 0.60

multipliers and their decomposition point to a better performance of China than Russia with respect to both growth and equity. These findings increase understandings of the widening gap in economic performance between these two countries during the last two decades. The comparative analysis for Russia and China was part of a research programme on comparative SAMs that included, in addition, four West European countries (Germany, Italy, the Netherlands, and Spain), and two East European countries in the transition phase (Hungary and Poland); cf. Cohen (2009). It is worthwhile to summarise the conclusions of these applications, and their correspondence with the results for Russia and China. The differences found in growth and equity performance between the West European (WE) and East European (EE) countries are explainable partly by differences in their levels of economic development and partly by differences in their economic systems. When these two effects are isolated, it appears that the WE economies appear to be more

198  SAM models for transiting economies Table 9.6 GLIs China 1989 Gainers and losers

Exogenous spending injections into activity sectors Agriculture

Industry

Services

GLIjj′ gainers and losers by sector Agriculture 1.30 Industry 1.03 Services 0.71

0.91 1.17 0.70

0.97 1.15 0.70

GLIhj′ gainers and losers by household group Rural farm 1.37 Rural non-farm 1.37 Urban employee 0.46 Urban self-employed 0.47

0.91 0.91 1.13 1.15

0.76 0.76 1.35 1.38

Exogenous income transfers to households groups Rural farm

Rural Urban Urban selfnon-farm employees employed

GLIjh′ gainers and losers by sector Agriculture 1.45 Industry 0.93 Services 0.82

1.46 0.95 0.76

1.45 0.95 0.78

1.44 0.96 0.76

GLIhh′ gainers and losers by household group Rural farm 1.17 Rural non-farm 1.17 Urban employee 0.76 Urban self-employed 0.77

1.18 1.18 0.74 0.75

1.17 1.17 0.75 0.76

1.17 1.17 0.75 0.76

efficient than EE in their use of inputs, and are able to generate higher growth multipliers relative to their exogenously held portions of their respective economies. It was found that there is a growth bias favouring primary sectors in EE economies, while the size and share of multipliers of the services sectors were relatively low. These features are also shared by Russia. For bringing Russia on track with prospective patterns, the SAM analysis suggests that exogenous forces (government expenditure and exports) should be geared to the advantage of services and away from the past bias towards agriculture and industry. The mechanisms of SAM multipliers also need to be adapted through appropriate investments in favour of services. As far as distribution bias is concerned, all transition economies used to be (before transition) less regressive in primary income distribution than are the selected West European countries, and these tendencies are reflected in the SAM multiplier analysis that shows more regressive income distribution mechanisms

SAM models for transiting economies  199 in WE than in the EE economies. During and after transition, income inequality increased appreciably in EE; the same was noted for Russia, and for China, though to a lesser degree. The role of the government budget in planning-oriented economies before transition was geared towards resource allocations in a fixed factor-remuneration system; and, understandably, there was little need for using the government budget as an instrument for income redistribution. In contrast, in market-oriented systems the government budget is geared more towards the task of balancing income distribution. Given the rise in income inequality with transition and, by WE standards, the role of government policy in transition countries is likely to change towards the WE standards. That is, reduced government intervention in the factor market and in the formation of primary incomes, and a more significant role for the state as a redistributors of secondary income transfers to correct for the regressive mechanisms of primary income formation. The results of this chapter show that the required policy actions for redistributing incomes via secondary transfers appear to be greater in Russia than in China.

  3.5   3.0   4.5   4.2  19.5 404.3  49.2

e

h

i

j

30.1  2.5 23.9  5.8

34.2  3.0 32.1  7.4

32.8  3.0 33.3  7.3

105.2  11.9 111.7  36.1

Food

6 Activities

 47.0  47.9  41.0   8.0   7.1 288.8

 130.1  622.0  269.5   49.0   63.9 1589.5

  2.8 212.7  68.8  47.0  13.7 948.8

  −4.7   23.1  11.6

 41.7   0.7  288.8 188.9 173.7 1589.5 184.3   4.5  948.8 −16.0  435.8  20.9  162.9 435.8 162.9 6815.3

 236.1   23.0  220.5   61.6

  66.0   72.9   94.1   96.9  459.5  496.9  395.4

 150.3   16.1   55.9  488.8

 14.0   48.3  88.0  13.4    1.3   1.4  53.0   2.9  32.4  230.2 226.2

8 Total ROW

 455.5

Service

7 Cap

 29.7  152.1 273.7

Alco- Non- Service Agric. Min. hol food +ind

5 Commodities

 66.5 143.2 20.5 180.6  307.6  11.5 61.6  2.7  6.7 11.5 13.9 146.9 281.1 −115.0  26.4  2.5  28.4 66.0 72.9 94.1 96.9 459.5 496.9  395.4 236.1 23.0 220.5 61.6

33.8  2.6 19.5  5.0

  30.2   22.8   25.1   23.9  100.8

3 4 Firms Gov.

 2.4  2.5  3.0  2.8 13.1 92.6  2.4  3.9  5.9  6.6  47.7 99.4

g

Source: Cohen (2002b). The ordering of accounts is done here otherwise

55.9 488.8

 0.9  4.0  5.7  6.5 38.8

d

f

c

a

b

2 Households

1 Factors

Social accounting matrices of Russia 1990, in billion roubles

1 Factors a Wage enterprises b  Soc. sec. c  collective farms d  private farms e Others 2 Households  2.7 f  400 rbl. 3 Firms 150.3 4 Government 5 Commodities Food Alcohol Non-food Services 6 Activities Agriculture Mining and industry Services 7 Capital 8 Rest of world 455.5 150.3 16.1 Total

Annex Table 9.7

2 Households

488.0

483.1   4.9

20.6

 5.1

 14.1

241.9

 6.2  1.7  2.0  1.1  0.5

123.8  31.2  35.9  19.8  17.1

 4.0 40.1  7.8 15.6  5.2  3.4

201.7 77.7

  1.8  3.4

159.8  25.9  52  17  44.6

 2.2

Source: Cohen (2002b). The ordering of accounts is done here otherwise

378.8

301.1  77.7

Urban Rural Profits Urban Urban Rural Rural labour labour employ self farm non emp farm

1 Factors

230.9  6.7  29.9 46.6  69.1 13.3 329.9 66.6

128.9 483.1 258.8

483.1

6 Activities

105.5 43.1 65.6

 10.6  9.8  73.9 32.3 13.1  21.0 10.8 42.7

Serv.

 28.7 134.6   3.4   5.2  12.6 604.1

 101.6  752.8  173.8   20.0  185.9 1782.9

 33.7 295.9  60.4  20.0   7.0 821.4

  5.3  162.5  79.9

 14.9  113.9 129.5 324.0   43.5  11.3  75.4  228.9 183.7

Food Cloth Misc Educ Serv Agr. Ind.

5 Commodities

 15.5 110.2

  4.2

3 4 Firms Gov.

Social accounting matrices of China 1989, in billion yuan

1 Factors Urban labour Rural labour Profits 2 Households Urban employee 237.7 Urban self-emp  20.6 Rural farm Rural non-farm 3 Firms 4 Government 5 Commodities Food Clothing Miscellaneous Education Services 6 Activities Agriculture Industry Services 7 Capital 8 Rest of world 258.3 Total

Annex Table 9.8

 329.9   66.6  105.5   43.1   65.6

 241.9   20.6  301.1   77.7  483.1  258.8

 258.3  378.8  488.0

8 Total ROW

 95.4  71.2  604.1 242.3  51.3 1782.9 243.7  54.3  821.4  28.7  581.4  205.5 581.4 205.5 7114.3

7 Cap

10 Transiting from fixed- to flexible-price regimes SAM-CGE models of Poland and Hungary

1 Introduction The proper approach to the transformation of the planning-oriented economies to market-oriented economies is a controversial subject. Some economists have used the term shock therapy to emphasise the necessity of a change in mentality among consumers and producers. Others advocate a gradual approach to economic restructuring. Some support a combination of the two approaches. There are momentarily no operational frameworks that are able to appraise such choices against each other, or are able to monitor consequences of the choice made. This is rendered most difficult in situations of hectic price changes and unstable relations that preclude the possibility of a reliable aggregation of data and empirical modelling during transition. Less ambitious, but nevertheless helpful in reflecting on performance aspects of the transformation process, is a stylised comparative static analysis of the economy under the pre-transition planning system and under the transiting market system. The social accounting matrix, SAM, model can be seen as a baseline measurement of the general equilibrium interactions in the economy for a particular year, under the basic assumption that prices are fixed and quantities adjust, which is the typical of the pre-transition state in centrally planned economies. At the other end is the flexible price computable general equilibrium (CGE) model that represents the market-based system. These two models, the price fix SAM model and the price flex CGE models are very handy in replicating a pretransition centrally planned situation and a transitional phase to the free-market situation, respectively. Even though the two versions refer to the same threshold year of 1990, the two models operate differently. By running one and the same policy simulation in both versions, it is possible to detect the locations, directions, and sizes of the discrepancies between the two models. Furthermore, by applying sensitivity analysis, one can appraise under which structural changes the discrepancies can be more effectively resolved. The impact multipliers in a fixed-price SAM model assume relative prices unchanged so that all adjustments and impacts go into quantity changes. As such, this version of the SAM can be seen to represent a truly centrally planned economy in which prices are fixed by the state and quantities carry the burden of the adjustment. The impact multipliers of an imposed injection—to be fully

Transition: fixed- to flexible-price models  203 realised in quantity changes—assume the availability of sufficient production capacity. On the other hand, a free market economy is commonly modelled as a CGE. As is well known, the rules of the game in a CGE model are different from those in a fixed-price model. In the CGE model producers maximise their profits and consumers maximise their utility in markets in which the demand for and supply of products and factors are cleared at flexible equilibrium prices. The SAM versus CGE can thus be seen as the opposite poles between the central planning model and free market model. As the CGE model that we use is static the economy is forced to operate within its given factors of production in 1990. As such, the CGE represents the difficult period of adopting market rules in a short-run transition period. In the medium and the long runs, factors of production are expanded, allowing reaping the economic benefits of more efficient allocations of resources on activities that are guided by flexible relative prices. The two economies in transition serving as case studies are Poland and Hungary. SAM and CGE models are calibrated for both countries for the transition threshold year of 1990.1 Two policy simulations are run on both models and the results analysed. Among the objectives of the analysis are (a) identifying in which directions, and by which magnitudes, the economy and impacts of policy will be oriented under the CGE model; and (b) in which ways do the transitional impacts differ in the contexts of Poland and Hungary; or in other words, which of the two countries is more prepared and can benefit more by switching from the price fix to the price flex regime. The two policy simulations will be denoted by S and H. The first policy simulation is aimed at shifting the production structure towards services, to be called experiment S, where government demand for services is increased by 1 per cent of total government expenditure. The measure can be interpreted as been taken to reduce the bias against the underdeveloped service sector inherited from the old planning system. The second policy simulation is aimed at changing the income distribution in favour of the poorest households, to be called experiment H. The same amount is transferred as welfare payments to the lowest decile household group, which can be interpreted as a measure to reduce the adverse effects of transition for the poorest households. The chapter is organised as follows: In Section 2 the fixed-price SAM model and its applicability to centrally planned economies will be discussed. In Section 3 the CGE model is presented. Two policy simulations involving sector injections and income transfers are performed with both models for the two countries. Section 4 examines the results for the Polish economy. Section 5 does the same for the Hungarian economy. Section 6 adds concluding remarks.

2  The fixed-price SAM model Although the SAM goes beyond the Leontief model regarding content, as it incorporates the whole circular flow, nevertheless, the fixed-price SAM model shares some common features with the Leontief input-output model, which is well anchored in the planning procedures of the former Soviet Union, and in

204  Transition: fixed- to flexible-price models the so-called ‘material balances’ planning procedures that were widely used in centrally planned systems. There are five premises that facilitated that work of planners in centrally planned economies, and these five premises are all maintained in the fixed price SAM model and the Leontief model; thus making these models appropriate proxies for mimicking the planning environment.2 The five premises are: (1) planners aim at achieving consistency between resources and uses, (2) planners know all the relevant technological coefficients, (3) all technologies are characterised by fixed input proportions and constant returns to scale, (4) primary resources (capital stock, labour force) are outside the model, and (5) prices are fixed. As the fixed price SAM model accommodates these five premises, it is well suited in representing the functioning of a planned economy. In contrast, the flexible prices CGE model to be displayed in the next section is most suited for replicating a market-oriented economy; notwithstanding that, the CGE model is parameterised from the SAM tables.

3  The flexible-price CGE model The CGE model portrays an unconstrained free market economy, to which the transition economies of Poland and Hungary have moved in 1990, and thereafter. All markets distinguished in the model, namely, the labour and capital markets, five product markets, the financial market in its simplified form of the savingsinvestment balance, and the foreign exchange market, are fully competitive. Demand and supply are equalised at all markets by instantaneous adjustment of all prices. Both the SAM and the CGE models break production into five sectoral activities indexed by j, contain the two factors of production of labour and capital, classify households into ten income decile groups indexed by h, include firms, government, and the rest of the world. The CGE model goes further by introducing a set of commodities indexed by c. The sectoral activities j use and produce commodities c. Producers employ labour and capital subject to technological production constraints, taking all prices as given, with the objective of maximising profit. Household groups, h, are suppliers of labour and capital and are demanders of commodity c. They maximise utility subject to their budget constraint, again taking all prices as given. Government levies taxes and premiums, and it redistributes income over households. Government spends also on activities. The rest of the world is engaged in trade and financial transfers with the domestic economy under freely moving foreign exchange rates. The equilibrium described by the model is of a static nature without inter-temporal effects. The equilibrium solutions can be altered by the government through changes in its expenditures and tax rates; these are exogenous. This section will specify the CGE model making use of notations in Table 10.1. The model is displayed in four boxes. Box 1 contains six equation sets. Sectoral output is determined from a two-level production structure, and is derived in several steps of sub-equations, ending up in eq. 1. To start with, at the lower level, value added in volume terms, or quantity terms, VQ j, is expressed as the product

Table 10.1  Notations Indices: c = commodities; j = sectors; h = household groups by income deciles (and firms). Furthermore, government is denoted by subscript g and rest of world r Endogenous variables C hc Value of consumption expenditure by household h on commodity c Cj Value of total consumption expenditure by activity j Composite price of consumption goods by type of commodity c SPc SPi Composite price index of the investment good i Demand for labour in activity j LDj KDj Demand for capital in activity j Value of exports of sector j Ej Foreign exchange rate (national currency units per 1 US dollar) FXR FCFGBD Foreign capital flow due to finance government budget deficit, in US dollar FCFPSI Foreign capital flow due to finance private sector investment, in US dollar Government budget deficit GBD Value of total investment I Ij Value of investment by activity j Value of competitive imports expressed at sector level j Mj Remuneration rate of labour LR KRj Remuneration rate of capital by activity j Vj Value added (gross domestic product) of activity j Xj Value of gross output of activity j Price index of gross output of activity j XPj Yh Disposable income of household group h. Applies also to firms Zh Incomings of household group h. Applies also to firms Revenue of government Zg Exogenous variables Price index of exports epj Price of imported consumption goods by commodity c in US dollars mpc mpi Price of the imported investment good i in US dollars Consumer price index, taken as numeraire CPI Cgj Value of consumption expenditure by government on activity j Investment spending by government Ig Supply of capital in activity j KSj Supply of labour LS Transfers from rest of the world to households (and firms), and government Trh , Trg Transfers from government to households (and firms) Tgh Coefficients αjo Calibration coefficient for production function of sector j αjj’ Input output intermediate delivery shares from sector j to j′ αrj Non-competitive import share in the output of sector j βj Labour elasticities of production relating to all labour in sector j continued overleaf

206  Transition: fixed- to flexible-price models Table 10.1  Notations continued εjo εj γhc γhc# γhco ϕg/p μrc , μri πh , πg , π r θjc , θji θcj , θij ρh′h τjg τhg τeg , τ mg υj ωh ζc

Exports quantity as a proportion of output quantity of sector j, measured for the benchmark year Export price elasticity of the world market demand for domestic products of sector j Consumption price elasticity of commodity c by household group h Consumption budget share of commodity c in the consumption of household group h Calibration parameter in consumption functions Ratio of public sector to private sector access to foreign capital flow, that is, FCFGBD / FCFPRS Quantity shares of competitive imports in the composition of consumption commodity c, and in the composition of investment goods i Profit distribution shares received by households h, including firms, government g, and rest of world r Quantity shares of domestically produced consumption goods c by sector j, and of domestically produced investment goods i by sector j Conversion coefficients from consumption commodity c to sector j, and from investment good i to sector j Transfer payments received by household h from household h′ as proportion of the latter’s income Where applicable h is extendable to include firms Indirect tax rate paid by sector j to government g Direct tax rate paid by households h to government g, the index h is extended to cover firms Export duty rate, and import tariff rate Inventory changes as a proportion of the output of sector j Wage distribution shares received by household h Weights in the CPI are proportions of products in aggregate household expenditure

of a Cobb-Douglas production function with employed factors of labour LD j, and capital DKj, and βj standing for labour elasticities of production in activity j. Thus, smooth substitution possibilities exist at this level: Q j = αj ΠLDjβj KSj(1 − βj ). At the higher level, sectoral value added in quantity terms, VQ j, is a fixed proportion of the sectoral output in quantity terms, XQ j, after deduction of domestic and imported intermediate inputs coefficients and indirect tax rates, following Leontief, thus: VQ j = (1 − Σj (αj′j − αrj − τjg) XQ j. By definition nominal gross output, Xj, equals output quantity, XQ j, times output price, XPj; thus: Xj = XPj . XQ j. Substitution of the three sub-equations in each other gives the final form of eq. 1, displaying the generation of sectoral gross output, in terms of the sectoral price, Leontief coefficients and Cobb-Douglas production function parameters. Further elaboration and solution of the model makes no more use of the two specific variables of XQ j and VQ j. For the sake of simplification these two variables are left out from notations in Table 10.1, which is done on purpose. The result is eq. 1.

Transition: fixed- to flexible-price models  207 In eq. 2, making further use of Leontief coefficients, the value added by sector, Vj, is found by subtracting the cost of domestic intermediate deliveries, imported intermediate deliveries, and indirect taxes from the gross output by sector, appropriately priced. In eq. 3, each sector is assumed to consist of many similar firms that all maximise profits in perfectly competitive product and labour markets. This means that wages and prices are given for the individual firm, who acts as if they are one large price-taking firm. Following first order conditions for profit maximisation, firms will hire labour LDj until the wage rate, LR, equals the value of its marginal product. In eq. 4, the remuneration of capital by activity, KRj, is the residual nominal value added after labour is paid its share. Because the volume of capital is fixed and given for each activity, remuneration rates KRj have no opportunity to adjust to be equal to the value of the marginal product of capital. Consequently, remuneration rates of capital may differ among sectors. In eq. 5, the total demand for labour in all activities, ΣjLDj, is equal to the exogenous supply, LS. In eq. 6, demand equals supply of capital by activity j, which is given for each activity j separately.

Box 1  Production functions and factor markets Xj = XPj [1 / {1 − Σj (αj′j − αrj − τjg )}] [αj Π LDj βj KDj(1 − βj )] Vj = (1 − Σj αj′j XPj′ /XPj ) Xj − αrj Xj − τjg Xj LR . LDj = βj Vj KRj . KDj = (1 − βj ) Vj Σj LDj = LS KDj = KSj

j (1) j (2) j (3) j (4) (5) j (6)

Box 2 relates to incomes of households and government. Eq. 7 specifies the incomings of household group h. Total income, Zh, consists of factor remunerations of labour and capital, expressed in fixed proportions, ωh and πh, of the total factor remunerations, respectively, plus three types of received transfers: net transfers from other households and firms, net transfers from government, and net transfers from the rest of world converted in national currency. The equation applies in similar form to firms. In this sense, h, is extended to represent firms as well. In eq. 8, disposable income of households, Y h, is defined as taxable income minus direct taxes and social security contributions paid to government. Both social security contributions and direct taxes together form a fixed proportion, τhg, of taxable income. The equation applies in similar form to firms. In this sense, h is extended to represent firms as well. In eq. 9, government revenue is defined to consist of factor income from ownership of public enterprises, plus direct taxes from households and firms,

208  Transition: fixed- to flexible-price models plus indirect taxes, plus net transfers from households (and firms), and net transfers from abroad, converted to national currency. The spending and transfer items in this budget equation are exogenous. In eq. 10, the government budget deficit, GBD, is defined as government expenditure minus government revenue. Government expenditure is the sum of current expenditures on the various sectors, Cgj, and government investment, Ig. The financing of the GBD in the transition period is specific for the transitional situation faced and is treated in a later equation, eq. 21.

Box 2  Incomes and spending of institutions: households (and firms) and government Zh = ωh LR . LS + πh Σj KRj . KSj + Σh′ ρh′h Zh′ + Tgh + FXR . Trh h (7) h (8) Y h = (1 − τ hg ) Zh Zg = πg Σj KRj .KDj + Σhτhg Yh + Σj τjg Xj + Σj τmg Mj + Σj τeg Ej + Σh Tgh + FXR . Trg (9) (10) GBD = Σj Cgj +Σh Tgh + Ig − Ζg

Box 3 treats equations that specify quantities and prices in the product markets. In eq. 11, consumption of households h is distinguished by commodity c, Chc. Consumption behaviour for each household group h takes the form of a Cobb-Douglas utility function, incorporating a price effect and an income effect per commodity c. The price effect is via the relative composite price SPc and price elasticity γhc. The income effect is via the applicable household disposable income Y h and an income elasticity that together with the price elasticities adds to unity, thus, 1 − Σc γhc. In this equation, γhc# denote the budget shares of commodity c, and γhco stand for calibration coefficients. We start first with equations for exports and imports. In eq. 12, commodity categories are composed of domestic goods, distinguished by sector of origin, and imported goods, of which the volume shares θjc and μrc in the composite are fixed. Thus, no substitution is possible between domestic and imported goods. The composite price index of consumed commodity c, SPc , is a weighted average of the two component prices. The weights are the respective shares of the domestic and imported components in the composite, whereby Σj θjc + μrc = 1. In eq. 13, the consumer price index, fixed at the value of 1.0, is taken as numeraire. The weights are the proportions of the respective products in aggregate household expenditure. This specification implies that changes in household income can be interpreted as changes in real household income. In eq. 14, total consumption expenditure on domestic goods by activity sector j, Cj , is defined as the sum of private consumption and government consumption by sector j. With regard to private consumption, this is done in several steps. The value of consumption expenditures by household groups h on commodity c, Chc , is summed over all households. The outcomes are divided by the commodity composite price index SPc to obtain quantities. These resulting quantities are

Transition: fixed- to flexible-price models  209 distributed on the activity sectors making use of the coefficients θcj , whereby θ  cj = θjc / (1 − μrc ). The obtained private consumption quantities by sector are then multiplied by the price indices of the respective sectors XPj to obtain the value of private consumption by sector. Regarding government consumption, these are exogenously given as Cgj. In eq. 15, the model specifies a composite of the investment goods consisting of domestic investment goods, again distinguished by sector of origin, and imported investment goods. The price index of the composite investment good SPi is a weighted average of domestic output prices XPj and the price index of a composite of imported investment goods, mpi , converted via exchange rate FXR. The weights are the respective shares of the components in the composite τj and τr , whereby Σj θji + μri = 1. In eq. 16, nominal investment expenditure by sector of origin Ij can now be specified as total investment quantity, distributed on the sectors, and multiplied by the relating sector prices. The distributional shares are θij , whereby θij = θji / (1 − μri ). Eq. 17 specifies the value of exports by sector j. Domestic products are assumed to compete on the world market with goods produced abroad. The product heterogeneity implies that the world market demand for domestic products depends partly on domestic prices, XPj, the foreign exchange rate, FXR, and the world price for exports, epj (which is the weighted average of the fixed prices of competitors on the world market), and the export price elasticity for world demand of domestic products of sector j, εj. In this equation, the quantity exported, Ej / XPj, keeps its benchmark level (εjo Ej / XPj ) but can vary from that level, depending on the differential between the foreign price converted to own currency, (FXR . epj ) and the offered domestic price, XPj. If it is assumed that the export price elasticity is equal to 1, the following is obtained Ej = (εjo Xj / XPj) . (FXR . epj ). If furthermore, XPj / FXR is equal to epj, export value is reduced to Ej = εjo Xj. Eq. 18 specifies the value of competitive imports at the sector level j. Competitive imports relate to the imported part of consumption expenditure by commodity c. The value of these imports is derived by dividing total household consumption of commodity c by the composite price index to obtain quantities. These are then multiplied by the previously mentioned import proportions in real terms, μrc. The outcome is then distributed on the sectors via θcj, and multiplied by sector prices XPj to give nominal values of competitive imports by sector, Mj. Box 3  Product markets Chc = γhco Πhc SPc γhc {γhc# (1 − Σh ) Y h } 1 − Σcγhc SPc = Σj θjc XPj + μrc (1 + τ mg ) (FXR . mpc) CPI = Σc ζc SPc = 1 Cj = XPj Σc θcj (Σ h Chc / SPc ) + Cgj SPi = Σj θji XPj + μri (FXR . mpi ) Ij = XPj θij (I / SPi ) Ej / XPj = εjo (Xj / XPj ) . (FXR . epj ) / XPj)εj Mj = XPj [Σcθcj μrc (Σh Chc / SPc )] + XP j [Σiθij μri (I / SPi )]

h. c (11) c (12) (13) j (14) i (15) j (16) j (17) j (18)

210  Transition: fixed- to flexible-price models The fourth block, Box 4, displays national accounting balances and related equations. In eq.19, the sectoral balances for the product markets are displayed. The supply of goods on the left-hand side consisting of output and imports equals the demand for these goods in terms of intermediate deliveries, consumption, investment, change in stocks, and exports. The clearance of the product markets follows the closure specification B described in Chapter 2, where sector price levels for j − 1 sectors, and installed investment are the specified adjustors. Eq. 20 is the savings investment balance. The formulation of this balance is adapted to the special situation of transiting economies. Total domestic savings are the sum of (a) savings by the private sector defined as disposable income minus consumption expenditure of households and firms, denoted by (ΣhY h − ΣhΣcChc ), and (b) and savings by government defined as revenues less spending. The latter takes usually the form of dissavings, since spending exceeds revenue, giving thus a government budget deficit, GBD. How are the savings investment gaps of the private and public sectors financed? We break up net foreign savings into two foreign capital flow streams expressed in foreign currency and converted to national currency via foreign exchange rate FXR: (a) one is attracted and used by the private sector (mainly firms), denoted by FCFPSI, (b) the other constitutes international borrowing and grants by the government for the partial financing of the government budget, denoted by FCFGBD. The combination of transitional difficulties and restructuring reforms dictate that the government concerned would aim at financing its budget via foreign borrowing and grants, with the support of international financial bodies. It follows that the (b) part of net government savings or budget deficit, cancels out against the (b) part of foreign capital flow that is drawn and used by government; thus, both the (b) parts can eventually be left out from the equation. In eq. 21, the constraining of the government budget deficit takes a specific form given the transitional environment and the recessionary circumstances at the time. The economies of the two countries modelled did not have the resources and capacities to undertake and succeed in structural reforms simultaneously and to control the budget deficit within noninflationary bounds. Injections from foreign capital inflows were necessary to achieve the transitional goals. The model assumes that foreign capital flow in the forms of borrowings and grants will be available for financing the government budget deficit. The related variable is denoted by the lengthier letters FCFGBD, to be converted in national currency via FXR. This component of foreign capital is targeted to equal government budget deficit, GBD. In eq. 22, the empirical relationship between the two components of foreign capital flow—the part that goes to the private sector and the part that goes to the public sector for budget financing—have significant economic meanings and policy implications. The relationship itself is given little attention in economywide models. The incorporation of a basic relationship between FCFPSI and FCFGBD will allow for an interdependent distribution between the two types of foreign capital flow in a CGE modelling framework. Although the coefficient which links both types, ϕg/p, is the result of a complex interplay of national and international influences and may move erratically between years and among

Transition: fixed- to flexible-price models  211 countries, it tended to be high during the early years of transition when the burden of structural adjustments were carried by governments and necessitated more foreign capital flows under public sector management. As the market economy institutes itself more forcefully the coefficient tends to diminish. The introduction of ϕg/p in CGE models expands the range of policy simulations and gives a greater insight into policy making. The specifications of eqs. 21 and 22 were reviewed in Chapter 2 and were denoted as closure E. They allow for an endogenous determination of both the foreign exchange rate and the foreign capital flow. This happens to be an attractive combination in policy modelling. The model is complete at eq. 22, consisting of 22 sets of equations against 22 sets of endogenous variables. Solutions of the unknowns will automatically result in the foreign payments balance, which is formally kept outside the model, and denoted as eq. A1. In general, the rule holds that when there are n interconnected accounting balances forming a coherent accounting system, solutions obtained from any combination of n − 1 balances, and thus leaving out one balance, are guaranteed to fit into that balance which is left out. This general rule corresponds with Walras Law meaning that, in an economy with n markets, it is sufficient to solve n ‒ 1 simultaneous equations for market clearing. As stated, the foreign payments balance is left outside the model. This eq. A1 shows on the left-hand side the capital account balance consisting of the foreign capital flow to the private sector, FCFPSI, and the foreign capital flow destined to finance the government budget deficit, FCFGBD. The right-hand side gives the current account balance, which consists of imports of consumption goods, imports of investment goods, and imports of intermediate goods, less exports, factor incomes to the rest of world, and net transfers by households (and firms) and government, from and to the rest of world, converted to national currency where applicable, that is if the net transfer is an incoming. Box 4 National accounting balances and related equations Xj + Mj = Σj XPj αjj′ Xj′ / XPj′ + Cj + Ij + υj Xj + Ej I + Σjυj Xj = (Σh Y h − Σh ΣcChc )+ (Zg − Σj Cgj − Σh Tgh ) +    FXR (FCFPSI + FCFGBD) FXR . FCFGBD = GBD FCFGBD = ϕg/p FCFPSI FXR. (FCFPSI + FCFGBD) = Σj Mj + Σj αrj Xj − [Σj Ej + π r Σj KRj KDj +    FXR (Σh Trh + Trg )]

j (19) (20) (21) (22) (A1)

4  Results of applied policy simulations to Poland 4.1 Overview To demonstrate the different consequences of government policy under the two opposite regimes of central planning (fixed prices, formulated along the lines of the SAM model) and free market (flexible prices, formulated along the lines of

212  Transition: fixed- to flexible-price models CGE model), the two policy measures S and H referred to earlier are simulated under both price regimes. The exogenous impulse in both simulations amounts to 1 per cent of total government expenditure. In the case of Poland, this was equal to 1.144 thousand billion zlotys in 1990. The results of the simulations are presented in Tables 10.2 and 10.3. The results will be discussed for each simulation (S and H) and each price regime (fixed-price SAM model and flexible-price CGE model), separately. 4.2  Simulation of an increase in government spending on services As can be expected, in the SAM model, additional expenditures of government in services lead to increased output of services, by 0.67 per cent, Table 10.2, row 5. Output of other activities adjusts complementarily according to input-output ties, and increases by about 0.27 per cent. Construction, which has the strongest forward linkage, increases most by 0.36 per cent. The spending injection results in a GDP growth of 0.40 per cent. Factor incomes grow with increasing output at approximately the same rate of 0.40 per cent. Labour incomes are affected slightly more than capital, because of the shift in the branch output composition towards labour intensive branches. With rising factor incomes, institutions increased their income: households by 0.31 per cent, and firms by 0.39 per cent. Among households, the higher income deciles gain more than the lower, as the rich earn primary factor incomes while the poor are more dependent on secondary income transfers (such transfers constituted 85.4 per cent of the total income of the first decile group). As far as outlays of institutions are concerned, savings increased more than consumption (that is, 0.40 per cent versus 0.31 per cent). This is explainable in terms of higher income growth of firms than households, and higher propensity to save of firms than households (viz., 53.6 per cent versus 21.8 per cent). In the CGE model, additional government spending in services raises nominal output of services but by much more, 1.57 per cent. Because total factor supplies are given, part of the increase in demand disappears into a rise in the price index (0.92 per cent, in Table 10.3 row 5, column 3). As a result, the real increase in the output of services is much smaller at 0.65 per cent. Moreover, as factor supplies are given, the rise in output of services can now be realised only through reallocation of factor supplies from other activities; this is restricted to the allocation of labour, because capital is immobile. The mechanism goes along the following lines: The government budget deficit rises through the increase in spending, which is financed by foreign capital flow. It has been generally assumed at the eve of the transition that if the countries in question are to pass through the transition phase successfully, this is only possible if sufficient foreign capital flow is available to support the required transitional restructuring adjustments to the new price regime. There is thus an increase in the supply of foreign currency. This induces an appreciation of the zloty, the foreign exchange rate FXR has to decrease by –2.69 per cent to restore equilibrium on the balance of payments. The appreciation diminishes the competitive position of Polish exporters, and exports decrease. So, exports carry the burden

Table 10.2 Poland: results of policy simulations (percentage change from base) Accounts

SAM model

CGE model

CGE-SAM

Experiment Experiment Experiment Experiment Experiment Experiment S H S H S H Activities Agriculture

0.27

0.42

−0.18

 0.08

−0.45

−0.34

Light industry

0.28

0.40

−0.31

 0.00

−0.59

−0.40

Heavy industry

0.25

0.27

−0.86

−0.57

−1.11

−0.84

Construction

0.36

0.37

 0.94

 0.78

 0.58

 0.41

Services

0.67

0.32

 1.57

 0.79

 0.90

 0.47

Average

0.36

0.34

 0.05

 0.08

−0.31

−0.26

Factors Labour

0.41

0.34

 0.65

 0.46

 0.24

 0.12

Capital

0.39

0.33

 0.52

 0.39

 0.13

 0.06

Institutions 1st decile

0.06

5.46

 0.09

 5.47

 0.03

 0.01

2nd decile

0.16

0.14

 0.26

 0.18

 0.10

 0.04

3rd−9th decile

 −

 −

   −

   −

   −

   −

10th decile

0.38

0.32

 0.57

 0.41

 0.19

 0.09

Average

0.31

0.59

 0.47

 0.67

 0.16

 0.08

Firms account

0.39

0.33

 0.52

 0.39

 0.13

 0.06

Capital account

0.40

0.38

 0.76

 0.59

 0.36

 0.21

Macro indicators GDP

0.55

0.47

 0.57

 0.42

 0.02

−0.05

Investment

0.58

0.55

 0.76

 0.59

 0.18

 0.04

Domestic savings 0.49

0.47

 0.04

 0.00

−0.45

−0.47

Ind. consumption 0.43

0.90

 0.46

 0.73

 0.03

−0.17

Imports

0.53

−2.54

−1.77

−2.98

−2.30

0.44

Table 10.3 Poland: other results of simulations (percentage change from base) CGE model

CGE−SAM

Value added prices a Output prices

Real output

Real output

Activities

  S

  H

  S

  H

  S

  H

  S

  H

Agriculture Light industry Heavy industry Construction Services

 0.48  0.30 −0.24  1.50  1.39

 0.41  0.40 −0.14  1.17  0.78

 0.17  0.08 −0.37  0.50  0.92

 0.17  0.04 −0.24  0.42  0.51

−0.36 −0.23 −0.49  0.43  0.65

−0.10 −0.04 −0.33  0.36  0.28

−0.63 −0.51 −0.74  0.07 −0.02

−0.52 −0.44 −0.60  0.01 −0.04

Note a Value added prices, denoted by VPj, are derived from output prices, XPj, and input-output coefficients as follows: VPj = XPj − Σj αjj XPj’ − (FXR . Σj mpj) μ j − τjg XPj

214  Transition: fixed- to flexible-price models of the macroeconomic adjustment and have to give in to make the extra government spending possible. On the sector level it is heavy industry, which is to a large extent reliant on exports, that suffers most, with real output decreasing by −0.49 per cent, Table 10.3. The decrease in the output levels of agriculture and light industry is less pronounced, as both sectors have a more diversified demand structure and are less vulnerable to changes in the pattern of final demand. In contrast, output of construction is favoured, and is predicted to increase by 0.43 per cent, caused by the boost in investment expenditure that amounts to 0.76 per cent. The results imply that investment will rise at a rate equal to the increase in total savings, that is, 0.76 per cent, which is for a significant part due to access to foreign savings, (FCFGDP + FCFPSI). An important effect of the greater foreign capital flow (FCFGDP + FCFPSI) and the appreciation of the FXR is the strong improvement in the terms of trade, which is manifested by high value added prices, see column 1, Table 10.3. This leads—in spite of fixed resources and limited growth in a static CGE model— to an eventual recovered increase in the total value added, or GDP, namely by 0.57 per cent, which is higher than in the pre-transition price fixed regime of the SAM model. The rise in factor prices results in an increase in household income averaging 0.47 per cent (the rise in household income is smaller than the rise in factor payments, because part of household income stems from fixed income transfers). Examining the income distribution, this is shown to become more unequal, and also when compared with the pre-transition price fixed regime of the SAM model. The above predicted tendencies were also for a large part actually manifested during the early years of the transition. 4.3  Simulation of an income transfer by government to the first income decile household group This is the second policy simulation run on both models. For the SAM model, results show that the additional transfer to the lowest income decile raises the income of this decile by 5.46 per cent in 1990. Incomes of other decile groups grow only a little because of the low share of the first decile group in the total of household incomes (only 6.l per cent), and the absence of direct transfers among household groups. Eventually, all household incomes grow by only 0.59 per cent, on average. Increased incomes together with an altered decile distribution generate increased consumption but with a changed pattern that favours light industry, 0.40 per cent, and agriculture, 0.42 per cent. Increases in output raise factor incomes by 0.34 per cent, with labour and capital growing at approximately the same rate, as agriculture is relatively labour-intensive while light industry is relatively capital intensive. The income transfer to the household group with the lowest income results in a moderate growth, GDP rises by 0.47 per cent. As can be expected, the income transfer results in an income distribution that is appreciably more progressive. In the CGE model, additional government transfers to the lowest decile household group results in same increases for incomes of those households, that is 5.47 per cent. Other households’ incomes also increased, which are the consequence of

Transition: fixed- to flexible-price models  215 the improvement in the terms of trade resulting from the appreciation of the zloty. As in the previous experiment, the appreciation of FXR is caused by an increased inflow of foreign capital to finance the government budget deficit. It should be remarked here that these mechanisms are incorporated in the model and that the appreciation and accompanying improvement in terms of trade do not arise from the fact that the government implements the injection. A transfer by the rest of the world to the poorest households would have given almost exactly the same results. In fact, in the final analysis, what is simulated in the early transition, is a foreign income transfer from abroad that is implemented through the national government. Consumption demand in the H simulation is predicted to increase more than in the S simulation, which is due to the immediate impact of the transfer on incomes and consumption in the H simulation. The large increase in consumption favours the output of agriculture and light industry. However, this advantage is not enough to offset the decrease in these activities’ export earnings due to the FXR appreciation, so that real production in agriculture and light industry decreases by: −0.10 per cent and −0.04 per cent, respectively. Heavy industry undergoes the largest decline in production, resulting from its strong dependence on exports and the FXR. In contrast, production levels in construction rise by 0.36 per cent due to higher investment. Services expand by 0.28 per cent via increases in consumption and least dependence on diminished exports. The summed result for the GDP in the transitional price flex CGE model is an increase by 0.42 per cent (it was 0.47 per cent in the pre-transition price fixed SAM model). Regarding income distribution, the income transfer policy simulation favours the poorest decile equally in both models and thus result in a more progressive income distribution, though the final outcome in the CGE model is less so than in the SAM model, as is indicated by the income share of the tenth decile group that rises by 0.41 per cent in the CGE compared to a lower rise of 0.32 per cent in the SAM model. 4.4  Flexible versus fixed prices: a comparison The generated aggregate changes in the GDP due to policy simulations do not vary much under the CGE model when compared with the pre-transition fixed price regime of the SAM model. The CGE model simulates changes in the GDP of +0.02 per cent and −0.05 per cent for the two policy simulations, columns CGE-SAM, Table 10.2. This is because the total resources are given in the transition-year static CGE model. Because total resources are fixed, price changes in the CGE model cannot generate a rise in production for all sectors, or a rise in income for all actors in the system; they force redistribution of production on sectors and redistribution of income between household groups. The applications indicated that the CGE model reallocates output and resources (labour) away from heavy industry to construction due to appreciation of the FXR and the higher investment, respectively. As for the income, this is redistributed in the CGE model more in favour of the higher income deciles, showing the transitional flexible price regime (CGE model) to be regressive when compared to the pretransition fixed price regime (SAM model).

216  Transition: fixed- to flexible-price models One and the other suggest that in Poland the comparative advantage of transiting from a planned to a market cleared economy is unattractive in the short-run transition. But, what was modelled is a short-run static transition CGE model with the given available resources in the transition year. In the medium term, (a) resources would increase, and (b) be reallocated efficiently following price differentials, thus allowing for higher economic growth in both the (a) and (b) respects. Of course, in both models the ROW pays for the increase in incomes. Foreign capital inflows FCFGBD and FCFPSI are endogenous. The favourable impact of the increase in foreign borrowing only holds for the short term. In the long run, debts have to be amortised and interest has to be paid. The same argument holds under the fixed-price regime, since foreign capital inflow is needed to compensate for deficits in the savings-investment balance and foreign payments balance.

5  Results of the simulations for Hungary 5.1 Overview The SAM and CGE models were also calibrated for Hungary for 1990. Because the format of the Hungarian SAM deviates from the format of the Polish SAM, some minor adjustments in the CGE model were necessary.3 The two policy simulations undertaken with the Polish model were repeated for Hungary. In the first experiment government’s spending on services was increased by 1 per cent of total government expenditure, while in the second experiment income transfers to the poorest household group were increased by the same amount. The impulse was equal to 9.065 billion forints in 1990. The simulation results are in Tables 10.4 and 10.5. 5.2  Simulation of an increase in government spending on services In the Hungary SAM model, the increase in government spending on services equal to 1 per cent of the total government income elevates the nominal production value of services by a high percentage of 1.23 per cent, row 5 column 1, Table 10.4. Construction shows the second largest output growth, +0.52 per cent, heavy industry shows the smallest increase in output, namely 0.34 per cent. The direction of these results in Hungary is the same as in Poland, but the size and dispersion of the impacts are greater in Hungary. So, in Hungary the injection leads to a larger sectoral dispersion than in Poland. This is explainable by the larger government sector in Hungary which implies that the impulse will be a higher percentage of benchmark production than in Poland. As a result, the average rise in production value (0.62 per cent) is also higher than in Poland (0.36 per cent). Correspondingly, GDP increases by 0.71 per cent, compared to Poland at 0.55 per cent. The resulting income redistribution favours upper deciles, and these can be calculated to show that the regressive effect is lower in Hungary than in Poland; see shortened comparative Table 10.6.

Table 10.4 Hungary: results of simulations (percentage change from base) SAM model Accounts

CGE model

CGE−SAM

Experiment Experiment Experiment Experiment Experiment Experiment S H S H S H

Activities Agriculture Light industry Heavy industry Construction Services Average

0.46 0.41 0.34 0.52 1.23 0.62

0.55 0.49 0.29 0.39 0.50 0.45

−0.18 −0.76 −1.12 0.27 1.49 −0.09

0.16 −0.21 −0.61 0.24 0.62 0.01

−0.64 −1.17 −1.45 −0.25 0.26 −0.71

−0.39 −0.70 −0.90 −0.15 0.11 −0.44

Factors Labour Capital

0.72 0.71

0.46 0.45

0.89 0.79

0.56 0.52

0.18 0.09

0.10 0.06

Institutions 1st decile 0.39 2nd decile 0.44 3rd−9th decile    − 10th decile 0.64 Average 0.54 Firms account 0.71 Capital account 0.46

13.95 0.28    − 0.41 0.99 0.45 0.41

0.49 0.54    − 0.79 0.67 0.79 −0.19

14.00 0.34    − 0.50 1.07 0.52 0.00

0.09 0.10    − 0.15 0.13 0.09 −0.66

0.05 0.06    − 0.09 0.08 0.06 −0.41

0.46 0.41 −0.21

0.87 −0.19 −0.48

0.55 0.00 −0.40

0.15 −0.66 −0.34

0.09 −0.41 −0.19

1.05

0.67

1.12

0.12

0.07

0.49

−2.88

−1.64

−3.36

−2.13

Macro-indicators GDP 0.71 Investment 0.46 Domestic −0.14 savings Industry 0.54 consumption Imports 0.48

Table 10.5 Hungary: other results of simulations (percentage change from base) CGE model

CGE−SAM

Value added pricesa Output prices

Real output

Real output

Activities

 S

 H

  S

  H

  S

  H

  S

  H

Agriculture Light industry Heavy industry Construction Services

0.84 0.58 0.47 0.89 1.24

0.57 0.52 0.34 0.57 0.67

 0.18 −0.33 −0.29  0.28  0.65

 0.15 −0.14 −0.16  0.19  0.35

−0.36 −0.44 −0.83 −0.01  0.83

 0.01 −0.06 −0.45 −0.05  0.26

−0.82 −0.85 −1.17 −0.53 −0.40

−0.54 −0.43 −0.74 −0.34 −0.24

Note a Value added prices, denoted by VPj , are derived from output prices, XPj  , and input-output coefficients as follows: VPj = XP − Σj αj′j XPj′ − (FXR . Σj mpj) μ j − τjg XPj

218  Transition: fixed- to flexible-price models In Polish simulations, in which the distribution of income has become less equal over time, the distribution of income was fairly stable over time in the Hungarian simulations. In the Hungary CGE model, the macro adjustment to the simulated policy goes along the same lines described for Poland, but the measured impact is greater. The increase in government spending leads to a decline in the government budget surplus and external finance via foreign capital inflow leading to an appreciation of the forint (FXR changes by −2.69 per cent) and pulling down exports. Heavy industry suffers the most because of its strong dependence on declining exports, the decline in real output amounts to −0.83 per cent, agriculture and light industry undergo declines in real output (−0.36 per cent and −0.44 per cent); however, for light industry, nominal output decreases even more, because the output price also has decreased. The increase in the service sector is large at 0.83 per cent, because it is the sector which receives the injection and is also most immune to FXR appreciation and exports. All these figures are much larger and more dispersed for Hungary than in the Polish simulations. However, there is a difference: in Hungary the output of the construction sector hardly changes in real terms, while in Poland it was an important source of growth with an increase in real terms of 0.43 per cent, row 4, Table 10.2. This is because contrary to the Polish situation, there is no boost in investment expenditure in Hungary. The reason for this is the high exogenous inflow of foreign capital, and the fall in the exchange rate would reduce the domestic value of these foreign savings, and this offsets the income-related increase in domestic savings. The outcome is a small decline in investment expenditure. The net effect on the GDP is the highest registered so far, at +0.87 per cent. Average household income increases by 0.67 per cent. Also, for CGE simulations the distribution of income becomes more unequal. 5.3  Simulation of an income transfer by government to the first income decile household group In the SAM model, the simulated transfer raises the income of the poorest household group by 13.95 per cent. Again, this is much larger than the increase that was found in the Polish simulations. Also, the average increase in household income, which amounts to 0.99 per cent, is larger than in Poland. This is due to the larger size of the simulated transfer in Hungary. The income transfer causes a large increase in food consumption that favours agricultural production. Factors of production (labour) move to agriculture from other sectors. The sectors of light industry and services lag slightly behind. Construction experiences only a moderate increase in its output, because the growth in investment stays behind the growth in consumption. Again, heavy industry is the worst performing sector resulting from its strong reliance on falling exports (due to appreciation of domestic currency). As for the GDP, it increases by 0.46 per cent. This policy simulation being an income transfer, the simulation affects distribution progressively. In the CGE model, the transfer increases the income of the lowest decile by 14.0 per cent. Average household income increases by 1.07 per cent. Average

Transition: fixed- to flexible-price models  219 consumption increases by 1.12 per cent, with food increasing most. On the sectoral level, output of agriculture is positively influenced by the increase in food consumption and negatively by the fall in exports and the fall in output of light industry, to which it has strong forward linkages. Both effects outweigh each other, and ultimately there is hardly any change in output. Heavy industry suffers the most; real output declines by −0.45 per cent, due to falling exports. In nominal terms, the decline is reinforced by a fall in the output price. For the same reason real output of light industry falls slightly. Services, which are strongly linked to consumption, and are not affected by falling exports, benefit most. Average factor income, or GDP, grows by 0.55 per cent. Without additional government transfers to households, the income of households lags behind labour income because the non-labour components of household income are either constant (in the case of transfer income) or grow slower than labour income (in the case of capital income). Because the lower income deciles receive a high percentage of their income from the government (in Hungary, the poorest income decile receives almost half of its income from the government), higher labour income generates income mainly for the richer income deciles. 5.4  Flexible versus fixed prices: a comparison The simulations show that in real terms all activities perform worse under a flexible price regime. The worst performance is in heavy industry with the discrepancy (CGE – SAM) in real output between −0.74 and −1.17 percentage points, depending on type of stimulus, see Table 10.5. Other sectors show also negative discrepancies. The lower performances are attributed to the high capital inflow, appreciated currency, and falling exports, though all these are circumscribed by the nature of the static CGE model which holds factors of production at their levels in the base transition year. Despite the constraint by the given factors of production, the GDP, household income, and expenditure are predicted to be at higher levels in the CGE simulations than in the SAM simulations. The GDP discrepancy (CGE – SAM) is 0.015 per cent and 0.09 per cent in the two policy simulations. These are also higher for Hungary than for Poland, suggesting that the shift from the plan setting of the fixed prices regime to the market-oriented flexible prices regime is more attractive and rewarding for Hungary than for Poland, see comparative results in Table 10.6. The difference in the predicted GDP performance between the two countries lies for a great part in the greater foreign capital that is predicted to flow to Hungary than to Poland. The actual foreign capital in flow per capita during the six transition years of 1989–95 was also higher for Hungary than for Poland. As for the distribution of income, the CGE model shows consistently more regressive distributions than the SAM for the policy simulations run. There appears to be some trade-off between the GDP and the distribution discrepancy (CGE – SAM) in the policy simulations for Hungary. This stands in contrast to Poland, with an absent GDP discrepancy (practically zero), and a predicted distribution discrepancy between the two price regimes.

220  Transition: fixed- to flexible-price models Table 10.6 brings together comparative results from Tables 10.2 and 10.4 for Poland and Hungary on the discrepancies (CGE – SAM) for the GDP and a simple measure of income distribution, namely the ratio of the income increases that go to the first decile and the tenth decile. The results show that the shift from the price fix to the price flex regime is more attractive and rewarding for Hungary than for Poland. This also suggests that it may be easier and more forthcoming to institute the shift towards freely operating markets and flexible prices in Hungary than in Poland. It can be generally reasoned that because Hungary started its transition towards market clearance earlier than Poland, and institutionalised markets more gradually than Poland, this is an advantage for Hungary over Poland which should show up in greater advantages following the flexible price CGE model than the fixed price SAM model. This comes out clearly in the results. Shifting from SAM to CGE implies more GDP growth for Hungary than for Poland. Also, in terms of income distribution, the regressive tendencies of the shift are less in Hungary than in Poland. The results can be interpreted to mean that there are more incentives in Hungary than in Poland to pursue the shift for all agents. As the market institutions are strengthened, it becomes more economically rewarding to strengthen price mechanisms and operate through them. Some caution can be rightly placed regarding the above. The size of impacts from the shift in the price regimes that is greater in Hungary than in Poland is in part due to the size of the simulated policy measures, which are higher in Hungary than in Poland. A 1 per cent of the government budget in Hungary is bigger in US dollars than the same in Poland. However, one can argue also that what matters is the relative shift from a planned to a market-oriented system, and the 1 per cent base is the right one to consider.

6  Concluding remarks There are valid arguments for interpreting the results of the SAM model as describing a stylised centrally planned economy with fixed prices, and those of the static CGE model as reflecting a transitional base year when free market mechanisms with flexible prices are initially set. In the early years(s) of transition, the national income level can only be maintained, or eventually be increased,

Table 10.6 Growth and distribution discrepancies derived from policy runs on price flex and price fix regimes [CGE – SAM] Poland

Hungary

Δ In percentage points

Injection simulation

Transfer simulation

Injection simulation

Transfer simulation

ΔGDP Δ1st decile/Δ10th decile

0.02 0.03 / 0.19

−0.05   0.01 / 0.09

0.15 0.09 /0.15

0.09 0.05 / 0.09

Sources: Tables 10.2 and 10.4

Transition: fixed- to flexible-price models  221 when government spending measures that are necessary for the restructuring of the economy are supported by sufficient inflows of foreign capital. This is assumed in the formulation of both the SAM and CGE models. The policy simulation results from the static CGE model compared to the SAM model show that government injections lead to about the same levels of production in Poland. The discrepancies between the two models are higher for Hungary and indicate higher growth effects in favour of the flexible price regime of the CGE model. In a static CGE model the economy adjusts to exogenous shocks by both quantity and price movements within the given total endowments of labour and capital resources of the base transition year. Price changes can only generate a redistribution of production factors, production, and income over the different actors in the economy. Generally speaking, the CGE applications show reallocations of labour and output away from the heavy industry sector in favour of services, and often construction; agriculture and light industry maintain their positions. The causes behind these reallocations were examined. Regarding reallocation of income among household groups, the tendency is that the CGE model generates more inequality; however, these are less in the case of Hungary then Poland. The results show that the advantage of switching from a fixed to a flexible price regime in terms of both growth and distribution appears to be greater in Hungary than in Poland. Since the 1990s, both Poland and Hungary have reoriented themselves appreciably towards a market economy. The representation of these centrally planned economies by means of a fixed price regime as in the SAM model has lost relevance. Furthermore, an instantaneous shift of regimes from planning to market will bring with it changes in technology, returns, and so forth, which are not fully reflected in the specification of a CGE model based on data which have not yet undergone the influence of the regime changes. Notwithstanding this, the applications did illuminate new dependencies that emerge in the transitional phase of the flexible price regime, made some predictions that have actually occurred during transitions, and demonstrated important differential results and mechanisms between Poland and Hungary that may have their origin in the different timing and approaches followed by the two countries when they first started moving from the plan-oriented to the market-oriented regime. With the object of examining the sensitivity of the results obtained to the rapidly changing situation in both countries, we investigated as to how far the obtained results for 1990 deviate from results for pre-transition years. We constructed SAMs for Poland (1987), and Hungary (1988), and calibrated the CGE model for these two years as well. The following is a summary of the deviations. For the earlier years there were greater household income effects in the flexible-price model than the fixed. For 1990 the difference in impact on household income between both price regimes is less pronounced. This holds for both experiments in both countries. The smaller difference in 1990 can be explained by the smaller predicted change in the foreign exchange rate in 1990. This has to do with the relative size of exports and imports. The situation of a country with higher levels of exports facilitates the adjustment process towards reaching equilibrium in the foreign payments balance, because the adjustment takes place

222  Transition: fixed- to flexible-price models mainly through changes in exports, as imports are non-competitive and thus do not react directly to changes in the foreign exchange rate. The results suggest that the level of exports relative to imports plays an important role in determining the effect of the transition on household incomes. In both countries, levels of exports were higher in 1990 than in the previous years, allowing for lower appreciations in the FXR.

11 Public spending multipliers in extended SAM models for a developed economy

1 Introduction This book has displayed so far models applied to developing and transiting countries. SAM and CGE models are equally suited to deal with policy analysis for developed economies; and this is the focus of the current and next chapter. Application of these models to a developed economy can contribute to more insight into policy making. There are various reasons supporting this ambitious statement. In terms of policy making, the SAM is specially suited for weighing the trade-off between growth and distribution in the context of a growing share of the non-earning population, and the necessity of institutional transfers to mend the situation, which are major concerns in the welfare state. The CGE model, whose basic premise is reliance on market mechanisms for clearance of demand and supply, is also more likely to be applicable in a developed than in developing or transiting economies. Because the database in developed economies can be more far reaching in terms of more years and greater disaggregations than in most developing and transiting economies, application of the SAM models to developed economies makes it possible to expand the scope of the modelling applications, and to test more rigorously the validity of the obtained modelling results. Of course, several formulations of the SAM models need to be specified differently when applied to an industrialised economy. Foremost among these is the model closure; in particular, the subdivision of variables between endogenous and exogenous variables. For example, the SAM applications of chapters 5 and 9 considered gross capital formation as exogenous, which is supported by the fact that investment in developing and transiting economies belongs for a great part to the public domain and is often subjected to government decisions. In a developed economy, investment is better treated as an endogenous variable. In passing it can be added that the outcomes from application of the SAM in different settings add insight into the working of the models and highlight the relative nature of the analytical results and their policy implications. This chapter will demonstrate the relevance of the SAM for various useful applications for a developed economy, that is, the Netherlands. The chapter is organised as follows. Sections 2 and 3 will study the SAM multipliers, and examine how they change over a period of the ten years for which the SAMs were .

224  Extended SAM models for a developed economy constructed and applied.1 Section 4 will extend the analysis towards a regionalised SAM into four geographical areas generally known in the Netherlands by North, East, South, and West; and its use in the decomposition of economic performance over a couple of periods. Section 5 will briefly examine refinements relating to changes in urbanisation patterns over time. The final section concludes.

2  Multiplier analysis in a first SAM for the Netherlands The availability of a series of SAMs for the Netherlands for ten consecutive years (see note 1) allows for investigating the dynamic properties of the SAM. To simplify the presentation of results, we choose to discuss the structural properties of the SAM multipliers for the first year of the series, namely 1978, and to analyse subsequently the changing pattern of these structural properties as they appear in later years with intervals of three years each: thus 1978, 1981, 1984, and 1987. First, a few words on statistical sources and specifications. All ten SAMs were exclusively constructed from estimates of the national accounts as found in the publications of the Dutch Central Bureau of Statistics. The SAMs are disaggregated into 8 production activities, 7 products, 2 factors, 10 household groups classified by income deciles, firms, government, social insurance, an aggregate capital account, and a rest of world account, together resulting in a matrix of 32 rows by 32 columns. The required data for carrying out the disaggregation included: (a) household budget surveys (b) the input-output tables, and (c) an initial converter table for transforming a products classification into a sectoral classification; this initial converter matrix was constructed in a preliminary way from codes of household budget surveys, the input-output table and the industrial classification, and later subjected to several adjustments to assure consistency of the grand totals of its rows and columns. The adjustments were obtained by applying the RAS method for various years. All of the SAMs produced reflect the fact that the household budget surveys do not report on income transfers between household groups, which is a limitation. For all of the computed SAMs one and the same allocation of variables between exogenous and endogenous was followed. The assumption was that government and rest of world were exogenous. Because social insurance is predominantly under state control, this account is taken as exogenous as well. But in contrast with the SAM applications to developing and transiting countries where investment decisions are dominated by government, and where investment was assumed exogenous, the SAM applications for the Netherlands classified investment as an endogenous variable, being predominantly driven by market forces. As a result, the endogenous accounts formed a smaller matrix of 29 rows by 29 columns. After expressing the transactions in the latter matrix as average propensities of their corresponding column totals, the resulting coefficient matrix was then inverted to give the customary SAM multipliers, MS. This is a multiplier matrix of 29 by 29. For reasons of focus and space, only a few parts of this matrix are selected for comment. These relate to multipliers of spending injections to

Extended SAM models for a developed economy  225 sectors, and multipliers of income transfers to households. The attention in both multipliers goes to the effects on the variables of sectoral output and household income. Following the same notations as in the rest of the book, Table 11.1 gives the multiplier effects of spending injections in specific sectors on the sectoral output, that is, MS,jj′, and on household income, that is, MS,hj′. The specific sectors selected for the purpose of presentation are agriculture and industry. The results show that a spending injection of one unit, say a million dollars, or euros, or Dutch guilders at the time,2 has a multiplier effect which ranges between 2.21 and 1.81 units of output following an injected unit to either of the sectors of agriculture and industry, respectively. The average for all eight injected sectors is about 1.96 output units. How are the incomes of households affected? In the case of the average multiplier for all eight injected sectors, the income of households increases by 0.723. This gives an average income to output ratio of 0.723 / 1.956 = 0.37. This ratio can be compared with the corresponding income to output ratio in the SAM analysis for developing countries, see Chapter 5. The average ratio for ten developing countries is 0.53. In its generation of household income, the economy of the Netherlands has a greater dependence on a larger intermediate goods delivery system, compared to the developing countries. Besides, as corporate firms earn a greater part of the capital income, there is relatively less income left for households in the developed than in the developing economy. SAM applications to other European countries and to richer countries in general, show that the income to output ratio tends to be lower in richer countries as compared to developing countries (see Chapter 17 for elaborations). The implication is that in richer countries more units of output have to be produced to generate one unit of income. What does the SAM model for the Netherlands predict regarding gainers and losers among sector activities and among household groups? As explained in Chapter 5, a gainers and losers index, GLI, indicates whether a particular sector in the multiplier analysis (that is, following sectoral spending injections or household income transfers) is able to increase its share of the total output above its actual share in the observed year, in which case it is a gainer. When the sector’s share in the multiplier analysis is lower than the one observed, the sector is a loser. Similarly, GLI with reference to the incomes received by household groups indicates which household groups are gainers or losers in the circular flow. Results in Table 11.1 show that that when all sectors are equally injected the losing sectors are services, industry, and agriculture, with GLI < 1, and that there is a highly positive growth bias towards the sectors of public utilities, construction, and banking, with values of GLI significantly above 1.0. The trade sector is a modest gainer, with GLI around 1.1, while the sector of transport tends to hold its relative position with GLI around 1.0. The results also show interesting inter-sectoral reallocation effects such as the case of construction, which is a loser following injections in agriculture and industry, but a substantial gainer following injections to other sectors, turning the sector of construction into a net gainer as well if all sectors are injected. Also interesting to note is the big boost which banking receives from injections to industry. These inter-sectoral redistribution effects are understandable in view of the concentrated backward and forward linkages between these sectors.

226  Extended SAM models for a developed economy Table 11.1 Multipliers of sectoral spending injections MS,jj′ and MS,hj′ , distribution of multiplier effects and computations of gainers and losers index, GLI, Netherlands, 1978 Multiplier effects on

Output multiplier Size Distribution per cent Agriculture Industry (and mining) Public utilities Construction Trade Transport Banking Services Income multiplier Size Distribution per cent 1st decile 2nd decile 3rd decile 4th decile 5th decile 6th decile 7th decile 8th decile 9th decile 10th decile

Actual Multiplier shares following shares sectoral spending injections in selected sectors

GLI = columns 2, 3, 4 divided by column 1 that gives actual sharesa

1978

Agriculture Industry Av. 8 sectors

Agriculture Industry Av. 8 sectors

  1.808 100.0   3.15  42.99   5.06   4.61  17.23   5.07  12.08   9.8

  1.956 100.00   4.41  32.73   4.99  17.86  14.78   6.03  10.45   8.76

GLIjj′

100.00   4.78  38.27   2.72   9.17  13.34   5.46   7.76  18.50

MS,jj′   2.212 100.00   9.04  48.6   4.83   3.06  14.68   3.83   8.08   7.87

100.00   3.45   5.27   6.35   7.26   8.05   9.25  10.77  12.41  15.24  21.94

MS,hj′   0.669 100.00   1.00   2.00   5.31   7.07   8.42   9.79  11.58  13.17  15.92  25.75

  0.572 100.00   0.80   1.83   5.19   7.00   8.53   9.70  11.72  13.38  16.15  25.69

  0.723 100.00   0.86   1.89   5.23   7.02   8.50   9.73  11.67  13.31  16.08  25.71

1.89 1.27 1.78 0.33 1.10 0.70 1.04 0.43

0.66 1.12 1.86 0.50 1.29 0.93 1.56 0.53

0.92 0.86 1.83 1.95 1.11 1.10 1.35 0.47

0.23 0.35 0.82 0.96 1.06 1.05 1.09 1.08 1.06 1.17

0.25 0.36 0.82 0.97 1.06 1.05 1.08 1.07 1.05 1.17

GLIhj′

0.29 0.38 0.84 0.97 1.05 1.06 1.08 1.06 1.04 1.17

Notes a GLI is defined as follows: GLIjj′ = [(MS,jj − δjj′ ) / (Σj MS,jj′ − 1)] / [Output j,o / Σj Output j,o] GLIhj′ = [(MS,hj′) / (Σh MS,hj′)] / [Incomeh,o / Σh Income h,o]

Considering the effects of overall sectoral spending injections on income distribution, the results show that any sectoral injection is extremely regressive. The households representing lower income deciles experience GLI much below unity, that is, the first and second deciles scoring 0.25 and 0.36, respectively; the middle groups close to 1.0; and the richer groups in the seventh, eighth, ninth and tenth income deciles scoring GLI > 1.0. For injections in the individual sectors, the impact of an injection in agriculture appears to be slightly less regressive than an injection in industry as evident from a predicted smaller loss in the relative positions of the first and second income decile groups.

Extended SAM models for a developed economy  227 Next is Table 11.2 which treats the output and income multiplier effects of income transfers to household groups, that is, MS, jh′ and MS,hh′, respectively. For illustration, the table displays the effects of transfers to the poorest first decile and the richest tenth decile, and as well the case of equal transfers to all ten deciles. The effects of a unit transfer lead to an increment in output averaging 1.17; this can be compared to the case of spending injections where the average was about 1.96. However, it is transfer injections that lead to higher income multipliers, reaching 1.40 units, and not spending injections that bring only about 0.72 units. The table shows, furthermore, that transfer injections to the lowest, and to lower deciles in general, have higher multiplier effects for all households and, Table 11.2 Multipliers of household income transfers MS,hh′ and MS,jh′ , distribution of multiplier effects and computations of gainers and losers index, GLI, Netherlands, 1978 Multiplier effects on

Output multiplier Size Distribution per cent Agriculture Industry(and mining) Public utilities Construction Trade Transport Banking Services Income multiplier Size Distribution per cent 1st decile 2nd decile 3rd decile 4th decile 5th decile 6th decile 7th decile 8th decile 9th decile 10th decile

Actual Multiplier shares following shares income transfers to selected household groups

GLI = columns, 2, 3, 4 divided by column giving actual shares in SAMa

1978

1st decile

1st decile

10th decile

Average all 10 deciles

100.00   4.78  38.27   2.72   9.17  13.34   5.46   7.76  18.50

MS,jh′ 1.348 100.00   4.22  28.91   5.39   2.61  23.57   5.48 15.19  14.63

  0.789 100.00   4.01  28.47   4.73   2.49  23.25   5.86  14.13  17.06

  1.172 100.00   3.37  25.53   5.41  12.85  20.49   5.47  12.89  14.00

100.00   3.45   5.27   6.35   7.26   8.05   9.25  10.77  12.41  15.24  21.94

MS,hh′   1.497 100.00   0.86   1.88   5.23   7.02   8.50   9.73  11.68  13.31  16.08  25.71

  1.296 100.00   0.85   1.87   5.23   7.01   8.50   9.59  11.68  13.32  16.09  25.71

  1.398 100.00   0.85   1.88   5.23   7.02   8.50   9.69  11.68  13.32  16.08  25.71

10th decile

Average all 10 deciles

0.84 0.74 1.74 0.27 1.74 1.07 1.82 0.92

0.71 0.67 1.99 1.40 1.54 1.00 1.66 0.76

0.25 0.36 0.82 0.97 1.06 1.04 1.08 1.07 1.06 1.17

0.25 0.36 0.82 0.97 1.06 1.05 1.08 1.07 1.06 1.17

GLIjh′

0.88 0.76 1.98 0.28 1.77 1.00 1.96 0.79 GLIhh′

0.25 0.36 0.82 0.97 1.06 1.05 1.08 1.07 1.06 1.17

Note a GLI is defined as follows: GLIhh′ = [(MS,hh′ − δhh′) / (Σh MS,hh′ − 1)] / [Incomeh,o / Σh Income h,o] GLIjh′ = (MS,jh′) / (ΣjMS,jh’ )] / [Output j,o / Σj Outputj,o]

228  Extended SAM models for a developed economy therefore, more national growth and income than the case of transfer injections to the highest or upper deciles. Table 11.2 analyses the relative effects of exogenous income transfers to selected household groups on each household group and each sector. The table shows that alternative transfers would influence the relative distribution of output on sectors in the same way (a loser bias towards the sectors of agriculture, industry, and services with GLI < 1, a gainer bias towards the sectors of public utilities, construction, banking, and trade with GLI > 1, and a stable position for transport at GLI around 1). It is noted that the gaining and losing sectors following income transfers are about the same as those of sector injections. As regards how the impact of transfers is distributed on the incomes of decile groups, what happens is that although an initial transfer to a certain decile increases, at first, this decile’s income absolutely and relatively, the effect tends to vanish when the circular flow of the SAM is completed (from income to expenditure, to production, to earnings and income, and back again and again). The inherent distributive properties of the economy of the Netherlands are shown to be generally regressive. This is apparent from both Tables 11.1 and 11.2. That, nevertheless, the actual income distribution in the Netherlands shows more equality than the GLI of the SAM multipliers is due to the progressive effect of annually repeated income transfers from middle and upper income deciles and their transmission to the lowest income deciles.

3  Changes in SAM multiplier results over ten years Availability of more SAMs for the Netherlands allows for an analysis of changes in sizes and structures of multiplier effects. Comparative results are shown in an abridged form in Table 11.3 over a period of ten years. The presentation is done with three-year intervals, giving years 1978, 1981, 1984 and 1987. It may be objected that the various SAMs over ten years are not strictly comparable in real terms, since they are expressed in current prices of their respective years. It is noted however that during the ten years, the Paasche price index values increased for GDP, for the consumption, and for the investment expenditure, respectively by 3.6 per cent, 3.9 per cent, and 2.8 per cent, on average, per year. The price values by sector showed more variation. Given the narrow range within which prices of various categories have moved between 1978 and 1987, a comparative analysis of the level of multipliers may not be handicapped if it is in current prices. In so far as the main concern of the analysis is a comparison of the growth and distribution patterns of the multiplier effects between two periods, it is legitimate and practical to treat each period in terms of its own set of prices. Notwithstanding this, mention can be made of an exploratory study by the author to compute the SAM for the years 1981, 1984, and 1987 in constant prices.3 Admittedly, such conversions can be complicated and controversial since alternative procedural options can be pursued in applying the conversions and obtaining SAMs in constant prices; nevertheless, the exploratory study showed little variation in relative prices, which supports the validity and practicality of an analysis of SAMs in current prices.

Extended SAM models for a developed economy  229 The comparative results in Table 11.3 show mixed changes in the size of output and income multipliers over the ten years. In general, a rise in SAM multipliers over time reflects a relative increase of the endogenous component of the SAM model and a relative reduction of the exogenous component. The opposite holds in the case of falling SAM multipliers, implying a reduced endogenous component, a weakening of the circular flow, and an increased influence of exogenous factors in determining growth and distribution. The years 1978–84 examined for the Netherlands coincided with a period of fiscal reforms and cuts in the governmnent budget and a sluggish growth in the world economy, and thus, a weakened exogenous component. Later years, 1984–87, showed a strengthening of the exogenous component at the cost of a weakening endogenous component. This later tendency seems to hold under normal conditions.

Table 11.3 Changes in gainers and losers over time Affected entities

Output multiplier

Size Gainers and losers Agriculture Industry Public utilities Construction Trade Transport Banking Services Income multipliers

Size Gainers and losers

1st decile 2nd decile 3rd decile 4th decile 5th decile 6th decile 7th decile 8th decile 9th decile 10th decile

Multipliers and GLI following spending injections in all sectors (average)

Multipliers and GLI following income transfers to all household groups (average)

1978

1978

1981

1984

1987

MS,jj′ 1.96 GLIjj′ 0.92 0.86 1.83 1.95 1.11 1.10 1.35 0.47

2.26

2.30

2.35

0.86 0.85 1.47 2.12 1.18 1.00 1.44 0.49

0.79 0.85 1.46 2.35 1.15 0.97 1.44 0.49

MS,hj′ 0.72

0.77

GLI hj′ 0.25 0.36 0.82 0.97 1.06 1.05 1.08 1.07 1.05 1.17

0.29 0.26 0.62 0.89 1.00 1.05 1.02 1.11 1.15 1.23

1981

1984

1987

1.38

1.38

1.38

0.72 0.83 1.41 2.24 1.16 1.01 1.38 0.50

MS,jh′ 1.17 GLIjh′ 0.71 0.67 1.99 1.40 1.54 1.00 1.66 0.76

0.69 0.66 1.63 1.51 1.60 0.92 1.81 0.76

0.66 0.66 1.68 1.67 1.52 0.91 1.87 0.78

0.62 0.66 1.68 1.42 1.47 0.95 1.76 0.77

0.78

0.88

MS,hh′ 1.40

1.47

1.47

1.53

0.29 0.33 0.66 1.01 0.90 1.05 1.06 1.07 1.13 1.21

0.22 0.39 0.57 0.82 0.93 1.05 1.03 1.12 1.12 1.23

GLI hh′ 0.25 0.36 0.82 0.97 1.06 1.05 1.08 1.07 1.06 1.17

0.27 0.25 0.61 0.89 1.00 1.05 1.02 1.11 1.16 1.22

0.27 0.32 0.66 1.01 0.89 1.06 1.06 1.07 1.13 1.21

0.21 0.38 0.57 0.81 0.93 1.05 1.03 1.13 1.12 1.23

230  Extended SAM models for a developed economy The changing structure of the multiplier effects can be reviewed. First, the results of sector injections on gainers and losers in sectoral output that were found for 1978 are prolonged for the ten-year period (upper left-hand side of the table). Although there are small variations from a year to another, the results are generally stable with the exception of the sector of agriculture which is predicted, according to the SAM and GLI analyses, to experience a further worsening in its share of total output. For example, the GLI of agriculture decreases from 0.92 to 0.86, to 0.79, and to 0.76 in the four years. Otherwise. The other sectors maintain more or less their GLI values over the ten years. Second, while the above tendencies generally hold for the impact of income transfers on the relative positions of sectoral output (upper right-hand side of the table), there are subtle differences. Because a large portion of income transfers is spent by consumers on demand for services, the previously assessed GLI for services (around 0.49) recovers somewhat (becomes around 0.77). Similarly, other sectors that benefit relatively more from income transfers and consumer spending are public utilities, trade, and banking, and thus their GLIs tend to increase. The relative gains in these sectors is at the cost of relative losses in the the sectoral output of the sectors of agriculture, industry, construction, and transport. This is valid for practically the whole period of ten years. Third, the results of sector injections on household incomes (lower left-hand side of the table) show that the regressive structure of the economy that was already noted for 1978 is of an enduring nature over the ten years. Furthermore, the multiplier distribution suggests that there is a strengthening of the regressive tendencies as shown by a falling GLI for the first, third, fourth and fifth income deciles; and a fluctuating but generally increasing GLI for the second, seventh, eighth, ninth and tenth income deciles. There is only one income decile with a constant GLI over the ten years; this is the sixth decile (GLI = 1.05). Fourth, and last, the above-mentioned changes towards a more regressive income distribution following spending injections are noted also for the observed years when the stimuli are income transfers (lower right-hand side of the table). It can be very interesting to use the analytical framework of this chapter to determine the changing patterns of GLIs among sectors and household groups for more recent years, and to lay out the SAM-related properties behind these changes. Such an endeavour would require initiating a research effort for constructing a new series of annual SAMs, as has been reported upon in this chapter. The analytical framework can be replicated.

4  Extension: incorporation of regional subdivisions in the SAM model It was possible to reproduce additional versions of the SAM where household groups were classified by geographical region, or by urbanisation level, or by demographic composition.4 These refinements allowed extending the SAM analysis appreciably into new directions. In this section we elaborate on the regionalised SAM; the next section elaborates briefly on the urbanisation extensions.

Extended SAM models for a developed economy  231 The questions addressed are: how can internal structuring forces and external intervening forces working together in the economy-wide circular flow explain patterns of regional development? And when and where are the external intervening forces more significant than the internal structuring forces? But first, we briefly review how these questions are commonly investigated without the aid of economy-wide models. Most studies on the economics of regional differences in the Netherlands focus on, respectively, the regional growth and distribution of firms and households, their interactions and impact on the regional labour markets, and the evaluation and formulation of regional public policy. Although the Netherlands falls into 12 provinces, most analysis aggregates the provinces to the four regions of North, East, West, and South. Most studies agree that the West—which is economically the most developed region—has been growing lately at lesser rates than the East and South, and that the North lags behind but that there are indications that it is catching up with the rest. Kemper and Pellenbarg (1993) studied the spatial dynamics of firms in the Netherlands; they concluded that in spite of a greater mobility of firms in the West, on balance there was a movement of firms towards the East and the South. Meester (1994) studied location preferences of firms in the Netherlands giving ground to expectations that economic activity in the West tends to decrease and the attraction of the East and the South as regions for the establishment of new firms tends to increase. As a result, a reduction of regional differences takes place. Priemus et al. (1995) studied household preferences (that is, environmental quality, social contacts, costs, and so forth), showing more value for northern, eastern, and southern regions at the expense of the West. At the same time, the studies point to a reduction in the regional differences of population growth in recent years. Government policy on the housing market (that is, location of new building) was found to be significant in the moving decisions of households. Congestion in the West opened the way for a regional de-concentration of population and residence. Regional public policy in the seventies and eighties targeted the northern part of the country, which has been traditionally the main problem area. In the late eighties a new regional policy memorandum was released which postulated a shift from equity and solidarity to efficiency. In the nineties and later, regions were urged to utilise their possibilities and become responsible for their own economy. Attention has shifted from focusing on specific regions to focusing on specific sectors, cf. Delft (1995). The studies mentioned above approach the problem of regional development using primarily frameworks that lay emphasis on the spatial settlement of business and households. They pay less attention to the economy-wide linkages which couple activities, households, and public policy to each other in the different regions. The SAM model avoids this shortcoming by modelling the economywide circular flow as a general equilibrium framework. This allows for treating economy-wide linkage effects and exploring growth and distribution tendencies of changing regional patterns over an intermediate period of time. The actual regional configuration at any one time can be considered to be the result of an interaction between internally structuring forces and externally intervening forces. Accordingly, the SAM was redone and was disaggregated so as to distinguish

232  Extended SAM models for a developed economy between the four regions. The SAM multiplier analysis proceeds further along the same lines that were stated earlier. Multipliers of the SAM are analysed to show the impacts of sectoral spending injections and of regional income transfers on the growth and distribution of sectoral output and on the growth and distribution of regional incomes. This is done for the four regions for two different periods, namely 1981 and 1985, giving Table 11.4. There are four findings. The first conclusion is the noted tendency towards generating multiplier effects that are lower in a later period (see the ‘size’ rows in Table 11.4). The average output and income multipliers of spending injections are diminished. This applies also for the output multiplier when the policy impulse is an income transfer to household groups (but the income multiplier remains the same). If this tendency would continue to recur as more periods with normal economic activity are introduced to the analysis, that would suggest that the circular flow is losing steam in generating economic growth, or in maintaining the same growth levels of past years. The implication is that economic growth and income generation in future years, in the Netherlands, may become increasingly more dependent on exogenous impulses from government and the rest of the world. The second conclusion is a noted tendency towards a more balanced regional development. The North region improves its GLI from a level of 0.95 in the first period to 1.01 in the second period; South improves its GLI from 1 to 1.03; East is unaffected and keeps the same GLI at about 1.0. The loser is West with GLI in 1981 at 1.01, and going down to 0.98 in 1985. Given an initial situation characterised by a richer West and poorer North and South, the multiplier tendencies suggest that the circular flow is changing and tends to redistribute incomes towards a more equal distribution among the four regions. This stands in contrast to the regressive multiplier tendencies in income distribution at the income decile level which was noted in the previous section. The contrasting multiplier tendencies in income equality are harmonised and are explainable by the growing concentration of populations of regressively affected lower income deciles in the richest region, the West region, thus, bringing the West region more on par with the other less wealthy regions, and showing progressively oriented regional income multipliers. Third, the regional changes are accompanied by the previously noted pattern of gainers and losers with regard to economic activities. This is now seen to hold for more periods studied. Some losing sectors in the static SAM are seen to lose more when dynamic SAMs are applied: this is the case for agriculture, where the GLI, which is disadvantaged in the first period at 0.86 and 0.67 (depending on the policy impulse), deteriorates further to 0.77 and 0.62. Another loser sector that shares the same pattern with agriculture is light industry. The opposite holds for the sectors of mining, utilities, construction, and banking, whose advantaged positions tend to grow; for instance, spending injections push the GLI of mining from 1.44 to 1.62 over time; while income transfers are specially advantageous to the sectors of utilities, construction, and banking, pushing their GLI from 1.59 to 1.66 for utilities, from 1.65 to 1.71 for construction, and from 1.79 to 1.84 for banking. In between there are sectors that experience losses in their favoured positions (such as trade), and other sectors that are promoted (these are heavy industry and services).

Extended SAM models for a developed economy  233 Table 11. 4 Multiplier effects of sectoral spending injections, MS,hj′ and MS,jj′ , and regional income transfers, MS,hh′ and MS,jh′ , for the two periods of 1981 and 1985, Netherlands, and computations of the gainers and losers index, GLI Affected variable

Output multiplier Size Distribution % 1 Agriculture 2 Mining 3  Light industry 4  Heavy industry 5  Public utilities 6 Construction 7 Trade 8 Transport 9 Banking 10 Services Income multiplier Size Distribution % North East West South

Actual Multiplier analysis for 1981 1981 Average of Average of all sectoral all regional spending income injections transfers

100.00   4.81   4.40  24.44  12.16   3.53   8.03  12.11   5.63   7.77  17.76

100.00  11.69  19.99  42.98  25.34

MS,jj′ 2.273 100.00   4.16   6.35  18.49   9.24   5.07  16.51  14.46   5.63  11.32   8.78 MS,hj′ 0.772 100.00  11.16  19.93  43.58  25.33

GLIjj′ MS,jh′ 1.34 100.00 0.86   3.23 1.44   2.81 0.76  17.53 0.59  11.65 1.49   3.40 2.06  13.26 1.19  19.20 1.00   5.15 1.46  13.88 0.49  13.34 GLIhj′ MS,hh′ 1.448 100.00 0.95  11.13 1.00  19.96 1.01  43.54 1.00  25.37

Actual Multiplier analysis for 1985 1985 Average of Average of all sectoral all regional spending income injections transfers

GLIjh′

0.67 0.64 0.72 0.53 1.59 1.65 1.59 0.91 1.79 0.75 GLIhh′

0.95 1.00 1.01 1.00

100.00   4.73   5.09  24.71  12.16   3.53  6.77  12.32   5.51   8.38  16.79

100.00   9.72  20.96  44.43  24.89

MS,jj′ 2.269 100.00   3.66   8.25  16.84  10.07   5.27  15.29  14.21   5.43  12.35   8.63 MS,hj′ 0.769 100.00   9.84  20.97  43.72  25.47

GLIjj′ MS,jh′ 1.310 100.00 0.77   2.94 1.62   3.83 0.68  16.84 0.83   6.66 1.49   5.86 2.26  11.59 1.15  18.49 0.98   5.04 1.47  15.42 0.51  13.32 GLIhj′ MS,hh′ 1.448 100.00 1.01   9.81 1.00  20.83 0.98  43.74 1.02  25.62

GLIjh′

0.62 0.75 0.68 0.55 1.66 1.71 1.50 0.91 1.84 0.79 GLIhh′’

1.01 0.99 0.98 1.03

Fourth, the sector of transport occupies a special position in the Netherlands. It maintains about the same levels of GLI in the two periods no matter which policy incentive is stimulated and, implicitly, it is also unaffected by changes in the regional distribution. The sector is highly flexible and is able to accommodate to different economic outlooks; this may be facilitated by the compact size of the country and the extensive presence of transport activities in all four regions. The two SAMs for 1981 and 1985 allow for a decomposition of the performance of the economy over time into the part that is due to changes in SAM multipliers and the part due to changes in exogenous variables. Recalling the reduced form of the SAM multiplier, as shown in eq. 1, this gives the vector of endogenous variables ν as a function of SAM multipliers, MS , and the exogenous vector, e. v = AS v + e = (I − AS )−1e = MS e

(1)

234  Extended SAM models for a developed economy Rewriting this equation for the two periods of 1981 and 1985 and subtracting gives the change in the endogenous variables, ∆v, as shown in eq. 2.

∆v = v85 − v81 = MS,85 e85 − MS,81 e81 = MS,85e85 − MS,85 e81 + MS,85e81 − MS,81e81   = MS,85 ∆e + ∆MSe81 (2) The change in endogenous variables, ∆v, is decomposable in two terms: a change in exogenous variables (at constant SAM multipliers), and a change in SAM multipliers (at constant exogenous variables). Table 11.5 applies eq. 2 to the regional SAMs of 1981 and 1985. The table decomposes the changes in the endogenous variables of sector outputs, and regional incomes for the country as a whole and for the regional subdivisions, into the two terms of eq. 2. Results show that sector output and household income of the country as whole, increased respectively by 18.78 per cent and 17.81 per cent, in 1981–85. This is basically due to the exogenous change of 19.8 per cent for total output and of 18.7 per cent for total income, together with a minor decrease in the percentage change of the multiplier, 1.0 per cent and 0.8 per cent, respectively. Realised economic growth of sector activities can now be explained in terms of multiplier changes and exogenous changes. Realised growth of banking (28 per cent) is remarkably high when contrasted with that of construction (almost zero growth). The endogenous circular flow has favoured banking as reflected in positive multiplier changes for banking (+8.9 per cent). In contrast, the highest negative multiplier change is recorded for construction (−9.7 per cent). Among other gainers in terms of multiplier changes is public utilities (+3.8 per cent), and among losers in terms of multiplier changes is agriculture (−7.2 per cent). The exogenous changes are positive for all sectors and have varied between 9.8 per cent for construction and 24.2 per cent for industry. The table shows that while the relative strength of stimulus to growth does differ across sectors, growth in sector activities is consistently more dependent on the growth of exogenous stimulus than the growth in endogenous circular flow mechanisms, that is, increases in SAM multipliers. These results, which have many policy ramifications, are clear and state that as far as determinants of economic growth are concerned, the exogenous stimulus dominates over changes in the endogenous circular flow; thus making the economy more dependent on external than internal factors. The table elaborates on the income changes that took place in the regions. Household groups living in the East had the highest growth rate of income amounting to 23.6 per cent, as compared to the national average of 17.8 per cent. The performance of the East is explainable in terms of a positive change in the multiplier effect over time of 2.0 per cent as compared to negative changes in the multiplier effects for the other regions. There is also a positive change in the exogenous effect of 22.0 per cent in favour of the East, as compared to a national average of 19.0 per cent. Hence, both the endogenous and exogenous forces have worked in favour of the eastern region during the period considered. The North did worst, with a decline in income of −2.1 per cent. In the North, during the eighties, economic activity was relatively low, and the region suffered from a high degree of unemployment. The decline of −2.1 per cent that is

−756 14521 28874 12288 54927

Regional income North East West South Total

−3068 1243 −688 −79 −2593

−2270 3960 −6575 2219 854 −5077 −2570 −779 4512 −839 −564 2312 13728 29562 12367 57520

7502 6809 38649 16051 4337 5131 19050 6786 9803 5131 129250 872 3775 8422 3610 16680

452 571 2463 149 901 −429 2401 735 2013 6727 15982

(4)

11316 3635 8924 1633 12876

371 541 2263 1823 802 3749 2674 733 2246 2001 17204

(5)

2756 5869 12216 7124 27964

6681 5697 33923 14078 2634 1811 1975 5319 5544 6402 96064

Notes Column (7) = Column (1) / value of variable in 1981; for the other columns, (8) = ((2)/(1)) × (7); (9) = ((3)/(1)) × (7)

5233 10769 32075 18270 5191 54 6480 6007 14315 14292 122685

(3)

(6)

−2.10 23.55 21.78 15.72 17.81

16.64 37.44 20.09 24.01 23.36 0.10 20.83 16.33 28.21 12.32 18.78

(7)

(2)

Rest of the world

(1)

Government Social security

−8.51 2.02 −0.52 −0.10 −0.84

−7.22 13.77 −4.12 2.92 3.84 −9.68 −3.25 −2.12 8.89 −0.72 −1.00

(8)

Multiplier change

Overall change

Of which

Due to multiplier changes

Overall

Due to exogenous changes

Changes in percentages

Changes in millions of current Dutch guilders, 1981−85

Sectoral output Agriculture Mining Light industry Heavy industry Public utilities Construction Trade Transport Banking Services Total

Receiving endogenous account

6.41 21.54 22.30 15.83 18.65

23.86 23.67 24.21 21.09 19.52 9.78 24.07 18.44 19.32 13.04 19.78

(9)

Exogenous change

Table 11.5 Decomposition of changes in sector output and regional income between 1981 and 1985 in terms of changes in SAM multipliers and changes in exogenous variables, Netherlands

236  Extended SAM models for a developed economy explainable in terms of a deterioration in the multiplier value of −8.5 per cent, and a modest increase of the exogenous stimulus of 6.4 per cent. Apparently this policy did not have the expected success. The endogenous circular flow located in the North worked against northern growth; besides, the exogenous contribution to northern growth, at 6.0 per cent, is less than that for national growth, at 19.0 per cent. Since then, things have changed for the better—following government incentives towards improving the attractiveness and the income position of the North—by rewarding settlement premiums to industries and shifting parts of the public administration to the North, among other policies. Table 11.5 allows also for a decomposition of the exogenous stimulus by source into government, social security, and rest of the world. What are the consequences of the exogenous stimuli by source on regional growth in the period 1981–85? All three exogenous impulses are particularly beneficial for households in the eastern, western, and southern parts of the Netherlands, with the highest benefits in the case of impulses from the rest of the world, followed by government, and social security (this applies for absolute amounts and for percentages). But in the case of the North, it can be seen that social security payments are dominant as the mechanism for income generation. The same conclusion that was drawn on determinants of sector growth applies generally for determinants of regional incomes as well. Growth in the exogenous stimulus component tends to overshadow the growth in the endogenous circular flow mechanisms component (that is, increases in SAM multipliers) in the generation of regional income. This makes regional performance more dependent on external factors (government and rest of world) than internal factors (circular flow mechanisms). As was stated earlier, the decomposition in this chapter ignored the effects of changing relative prices over time. Converting input-output matrices to constant prices of one basic year, not to mention the more difficult task of converting extended SAMs, is a shortcoming that has been hardly resolved in the literature. To what extent should the obtained results be qualified in view of their limited application to current prices? Some clues are possible from the following trials that we have applied. Eq. 2 can be alternatively reformulated to give a decomposition of performance starting from a different base (year), for instance taking 1985 as the base instead of 1981, (or switching between 1978 and 1988), giving thus eq. 3

∆v = ∆MSe85 + MS,81 ∆e

(3)

This was applied, too. It is comforting to note that the obtained results under the 1985 base do not differ in significant ways from those where 1981 is the base. The similarity of results in using eqs. 2 and 3 offers some evidence that selection of a base year, 1981, or 1985, for decomposition of past performance does not significantly influence the obtained results. The implication is that for the case at hand an eventual correction for relative changes in prices may not change the results significantly. It can be added that eqs. 2 or 3 can be formulated more generally as in eq. 4.

∆v = ∆MS e81 + MS,81 ∆ e + ∆ MS ∆e

(4)

Extended SAM models for a developed economy  237

5  Extension: urbanisation levels The next extension of the SAM is in the direction of urbanisation levels. In seeking a simplified way of looking at trends and changes in urban development, urban dynamics is usually described to consist of a four-stage cycle: (1) Industrialisation occurs with major growth for the manufacturing sector; (2) the service sector and transport facilities grow; (3) increased appreciation for quality of life and the rise of what can be called an environmental sector; and (4) rise of the information sector. The cycle is marked by spatial concentration at first, followed by a deconcentration later. The predictions of this urban dynamic model are accordingly described as structurally determined. The actual urban configuration at any one time is considered to be the result of an interaction between internally structuring forces and externally intervening forces. Urbanisation studies for the Netherlands5 suggest that the Dutch urban system is in stage (3) of the cycle. This stage is marked by spatial deconcentration, which means a shift of people and jobs from cores (central towns) to rings (surrounding suburban municipalities). Some features of this stage are a rapid rise in energy prices and a contraction of average family size, while public transport, town renovation, and spatial planning are given more weight in government policy. The Dutch urban system saw a deconcentration of population that peaked in the early seventies and has since then declined. There has been also a movement towards stage (4): a more balanced development of central towns and suburban areas (revitalisation or reurbanisation policy), a widespread computerisation of society, increased attention to small-scale industry and a structural increase of leisure time. Urban dynamics can be placed in relation to economy-wide changes that shape urban locations. The SAM is well suited for investigating causal relationships that connect urbanisation patterns to economy-wide changes and policies. For this purpose, two SAMs, for 1981 and 1985, were constructed in which household groups were classified into six urbanisation levels: (a, b) Rural and urban municipalities, the division line being respectively more and less than 20 per cent of the labour force active in agriculture; (c) Dormitory towns with at least 30 per cent non-residents; (d) Small towns with 10,000 to 30,000 inhabitants; (e) Medium-sized towns with 50,000 to 100,000 inhabitants; (f) Large towns with more than 100,000 inhabitants, that is, cities. Other SAM specifications and multipliers are the same as displayed earlier. A brief selection of the SAM analysis on urbanisation will be reviewed with the object of demonstrating the wider scope of issues that the SAM framework can address. For 1981, the income multiplier effect of an income transfer of one unit to cities was highest, at 1.47 units, and was lowest when the recipient is a dormitory town, at 1.41 units. Table 11.6 assembles computation of GLI following either an overall spending injection or an overall income transfer to urbanisation levels, which shows that rural and urban municipalities, dormitory, and small towns are gainers, scoring GLI > 1, while cities loose most, with GLI at 0.87. Another loser is medium-sized towns with GLI at 0.95. While the same GLI ranking among urbanisation levels

238  Extended SAM models for a developed economy is maintained in the results for 1985, a tendency is observed towards convergence of gainers and losers, with the exception of cities which end up losing more, with GLI at 0.84. The internal mechanisms of the SAM appear to give a sound explanation for the factually observed tendencies for deconcentration and retreat of cities. It is also interesting to observe that built-in relative loss for cities in the SAM multipliers corresponds with a similar built-in relative loss for the sector of services compared to other sectors (see previous sections). This correspondence is supported by an appreciable demand for services in large towns. The demand for services compared to agriculture or industry is more dependent on the prosperity of the large towns than other urbanisation levels. A decomposition analysis of economic performance over 1981–85, was done along the same lines of eq. 2, results are shown in Table 11.6, right-hand side. It is sufficient to indicate the cases of the best and worst performers. Rural municipalities come out as having the highest growth in income, amounting to 83.2 per cent as compared to a national average of 17.8 per cent. This is the result of a positive change in the multiplier effect over time of 37.4 per cent and an even higher positive change in the exogenous effect of 45.9 per cent as compared to the national averages of −0.8 per cent and 18.6 per cent respectively. Large towns, together with medium-sized towns, had the lowest performance in income growth during the period. Large towns had a nominal growth of 0.23 per cent. This is due to a deterioration of the multiplier component of −11.7 per cent and an increase in the exogenous stimulus of 11.9 per cent. The results confirm the previously stated conclusion that changes in the exogenous component of the economy (assumed here to be government and the rest of the world) outweigh changes that occur in the endogenous circular flow mechanisms (changes in multipliers) in determining economic performance. This hypothesis will be the subject of more discussions in Chapter 17.

Table 11.6 Urbanisation levels: gainers and losers over time, and decomposition of growth into endogenous and exogenous factors Urbanisation level(UL)

Rural municipalities Urban municipalities Dormitory towns Small towns Medium-sized towns Large towns Total

Gainers and losers index GLI

Decomposition of income growth 1981−85

Following spending injection

Income growth percentage

Due to changes in multipliers component percentage

Due to changes in exogenous component percentage

83.24 13.31 45.54 41.52 −0.03  0.23 17.81

37.36 −4.77 10.14 18.80 −8.20 −11.71 −0.82

45.88 18.08 35.40 22.72  8.17 11.94 18.62

Following income transfers

1981 1985 1981

1985

1.14 1.08 1.16 1.02 0.95 0.88

1.10 1.07 1.07 1.09 0.98 0.84

1.09 1.06 1.06 1.09 0.99 0.85

1.15 1.09 1.16 1.02 0.94 0.87

Extended SAM models for a developed economy  239 When the exogenous stimulus is decomposed by source, not shown in the table, it was found that the exogenous stimuli of the government are particularly beneficial for large towns which are clearly associated with public employment and public services. The exogenous stimuli of social security was found to favour dormitory towns, rural municipalities, and urbanised rural municipalities, where, presumably, old-age pensioners live; the exogenous stimuli of the rest of the world favour large towns, urbanised rural municipalities, and middle-sized towns.

6  Concluding remarks The SAM forms an appropriate framework for integrating various statistics on diverging areas, linking these areas to economic activities in an economy-wide model, and investigating causal relationships. Among the interesting results found is that externally intervening forces gain in strength as compared to the internally structuring forces in determining the income growth by receiving income deciles, regions, or urban levels, and the output growth of production sectors. Income-earning divisions that are more associated with gaining sectors of economic activities are also gainers. The gaining sectors are mining, banking, and public utilities; losing sectors are agriculture, light industry, construction, and services. It needs emphasising that as the multiplier effects are assumed to be fully realised in quantity terms and without leakage into price increases, the SAM results are most meaningful in the short to medium periods in an economy in which inflationary tendencies are modest and relative prices are stable. Other extensions show further that for the periods studied, large towns and the services sector had in common a negative growth bias in terms of both internal and external influencing forces. The SAM model suggests further that a deconcentration of cities towards towns, rural municipalities, dormitory towns, and small towns goes simultaneously with enhanced growth for the sectors of mining, banking, and public utilities at the cost of light industry, construction, and services. Results of a multiperiod SAM model applied to a developed economy suggest a possible turnaround hypothesis that future growth is conditioned by a weakening of—internal—multiplier effects and an increased dependence of the economy on—external—exogenous variables, that is, spending and transfers by government and the rest of the world. This hypothesis is validated under alternative specifications and extensions of the SAM, and for different periods.

12 Fiscal policy simulations in adapted CGE models The Netherlands

1 Introduction Fiscal policy is the use of government spending and government taxation to influence desirable outcomes of the economy, with due consideration for maintaining a sustainable balance between expenditures and revenues. Successful application of fiscal policy in times of economic crises and low growth is a very challenging job, and the number of countries renowned for a successful management of the fiscal budget is limited. This chapter treats a few aspects of modelling fiscal policy in a CGE model in the context of the Netherlands. A few remarks may be made on the background of this model and its positioning with respect to the fiscal policy experience of the Netherlands and the later advancements made in the art and science of fiscal policy modelling. The Netherlands experience with the use of fiscal policy is often quoted as an example of the better performing experiences, also renamed as the ‘Dutch miracle’. This is generally true for the Netherlands in the last few decades when compared to other countries. But the performance was not always uphill throughout the years. When economic progress and the budget balance are assessed over the last 40 years, the Netherlands shows that there were a few ups and downs, but recovery was smooth. This is clear from Figure 12.1. Budget stability in the Netherlands was mainly due to the structural deficit ceiling, introduced as a budget rule in 1961. The fiscal policy rule suffered under the rising public expenditure, reaching a striking level of over 60 per cent of GDP, and the rule became no more workable at the time of the first oil crisis, and was abandoned in 1974. Spending increased and revenue fell, and when the second oil crisis occurred the deficit reached its highest level relative to GDP in 1981, at 6 per cent. Strict cuts in spending and expanding globalisation brought the budget once again in balance together with moderate growth. The discipline was reinforced from 1995 into the next 12 years by pursuing policies towards meeting the EMU fiscal policy rule of 3 per cent of GDP.1 The recession of 2009 and the Euro crises of 2011 brought economic prospects and deficit financing back to a situation similar to that faced in the early eighties. The growth and budget difficulties of the period of 1980–3, and the subsequent success in resolving them, are recurring again in 2010–11, 30 years later.

Fiscal policy in adapted CGE models  241 % GDP

1970

1975

percent government budget balance

1980

1985

1990

1995

2000

economic growth

2005

2010

Figure 12.1 Government budget balance (GBD) as percentage of GDP, and growth of GDP: the Netherlands, 1970–2010 Source: CPB (2010)

The CGE model in this chapter, built and applied in the mid-eighties, was meant to address the fiscal policy problems faced in 1980–3, when the budget deficit recorded its highest share in GDP, at 6 per cent, and economic growth around 0 per cent, or in negative territory −1 per cent. In particular, the model was calibrated for 1981 and it introduced wage and transfer rigidities with budget consequences, incorporated built-in fiscal rules to curb spending and reduce the budget deficit, and carried additional fiscal policy simulations with the objects of targeting higher growth, lower unemployment, and restraining inequalities. In deciding on inclusion of this model in the book, the central questions to be answered were: (a) whether fiscal policy modelling methods of the mid-eighties are of any value today, and (b) whether the policy content of three decades ago has any relevance for today’s discourses. With respect to (a), there have been great advances in modelling fiscal policy since then. Already in 1998, the Central Planning Bureau (CPB) came up with a comprehensive CGE model—named MIMIC—incorporating a wide range of fiscal policy aspects, cf. Graafland and Mooij (1998). They employed an applied general equilibrium model containing behaviour of firms, households, and the functioning of the labour market with attention given to human capital formation and labour use in the shadow economy. The public sector is largely exogenous. The MIMIC was used to analyse the impact of several tax policies. Another advance in fiscal policy modelling is the model of the European Union—Directorate-General for Economic and Financial Affairs (DSGE), Ratto et al. (2006). The model studies effects of fiscal stabilisation policies

242  Fiscal policy in adapted CGE models in the euro zone. The authors incorporate nominal rigidities in the labour and product markets, look at fiscal stabilisation via government consumption, investment, transfers and wage taxes, and study the empirical evidence for systematic countercyclical fiscal policy. Recently, under leadership of the IMF, the scope of fiscal policy modelling was extended to the global economy; cf. Kumhof et al. (2010). Notwithstanding these major advances in fiscal policy modelling, we support the view that adaptations of the CGE model to rigid market contexts still have a long way to go, and that our early attempt is still valid in several contexts. On (b), the content of fiscal policy today in most of the European Union is about reducing deficits and pushing growth, and this was also the same problem faced by our case study for the early eighties. As such, the fiscal problems of the early eighties in the Netherlands are again valid three decades later. In both (a) and (b) respects, our view is that the current model is not far removed from the current situation; and that it is worthwhile to include it and take its approach and results into account. With the above background in mind, the chapter will specify various forms of the adapted CGE models and run several policy simulations. Two revenue reducing policy simulations are considered: (a) reduction of indirect tax in the services sector by 4.4 per cent, and (b) instituting a wage subsidy equal to 0.46 per cent of the wage bill. The next section will explain how these measures are specified. The guiding principle is that each policy simulation should have the same cost for the government revenue, which was set at 1 per cent of the government revenue.2 This allows for a fair appraisal among the policy simulations. The specific heights of the two measures, that is, 4.4 per cent and 0.46 per cent are determined by the guiding principle. Of course, depending on the model mechanisms, the end result of the revenue reduction will differ for the policy simulations, and can be above or below the 1 per cent loss in government revenue. This end result can be a criterion in appraising policy effectiveness. The policy simulations will be applied to various forms of the adapted CGE models. These forms are labelled basic, elaborate, and restructured CGE models. The main features of each form are tabulated in Table 12.1. The main differences are the following. In the basic form we shall set price elasticities of consumption, exports, and imports at zero, making the model less responsive to price changes. In the elaborate form, these elasticities are set at positive values, making the model fully responsive in its price effects. The structuralist CGE model chooses sticky wages, causing unemployment, and determining thus government payments of unemployment benefits. Social security payments are also realistically modelled as sticky in the sense that they are coupled to the consumer price index. The structuralist CGE model incorporates other policy premises as well that take distance from neoclassical assumptions. The next three sections will specify these three model forms, and apply and analyse the fiscal policy simulations described above.

Fiscal policy in adapted CGE models  243 Table 12.1 Review of the CGE model specifications in the basic and adapted forms Model specification

Basic CGE Model

Elaborate CGE Model

Structuralist CGE Model

Production functions

Cobb Douglas Input-output Full employment: adjustors are wage rates Primary incomes determined along neoclassical lines Tax and transfer rates fixed budget shares

CES Input-output Sticky wages: adjustors are (un)employment

Closure B (see Ch.2) Benchmark value Benchmark value Adjustors: XPj for j − 1, I FCF given Adjustor: FXR

CES Input-output Full employment: adjustors are wage rates Primary incomes determined along neoclassical lines Taxes and transfer rates fixed Income and price elasticities Closure B Price elasticity Price elasticity Adjustors: XPj for j − 1, I FCF given Adjustor: FXR

Government consumption Government transfers

Directly dependent on revenue Directly dependent on revenue

Directly dependent on revenue Directly dependent on revenue

Government investment Government revenues Government budget deficit GBD

given

given

Tax rates given

Tax rates given

Endogenous (openended)

Endogenous (openended)

Factor markets Primary incomes Disposable incomes Consumption Investment Exports Imports Product markets Financial market

Primary incomes for a part structurally determined Tax and transfer rates flexible Income and price elasticities Closure B Price elasticity Price elasticity Adjustors: XPj for j − 1, I FCF given Adjustor: FXR

Controlled exogenous – Unemployment benefits determined by number unemployed, CPI – Social security transfers given, coupled to CPI given Direct tax rate adaptor endogenous Constrained as proportion of GDP

2  The basic model 2.1  Specification The specification of the model in this chapter is done in a compressed form; many details are left out so that attention can be focused on fiscal policy. For example, the specification of the model will be restricted to five sector activities only, denoted by j, and commodities c are aligned to sectors j. The original work contained more of each. Household groups, denoted by h, consist of ten income deciles, which are commonly available in the SAMs for the Netherlands. The index h applies for firms as well.

Table 12.2 Notations Indices: the following indexes are employed. Sectors are denoted by j, whereby 1 = agriculture; 2 = light industry; 3 = heavy industry; 4 = construction; 5 = services. Commodity c corresponds closely with sector j Household groups (and firms) are denoted by h, whereby 1, … ,10 stand for ten income deciles, and the 11th group is for firms. Furthermore, the subscripts g and r stand for government and rest of world Endogenous variables Chc Value of consumption expenditure by household h on commodity c Cgj Value of government consumption expenditure spent on sector j Value of consumption expenditure by activity j Cj LDj Demand for labour in activity j Value of exports at sector level j Ej FXR Foreign exchange rate: national currency units per one US dollar GBD Government budget deficit Value of total investment defined to include installed investment and I inventorychange Ij Value of gross capital formation by activity j Value of competitive imports at sector level j Mj Remuneration rate of labour LR Demand for capital KDj Remuneration rate of capital by activity j KRj SPc Composite price index of consumption goods by type of commodity c components and foreign SPi Composite price index of the investment good i Tgh Value of government transfer expenditure to households (and firms) h Value added (gross domestic product) of activity j Vj Xj Value of gross output of activity j XPj Price index of gross output of activity j Yh Disposable income of households h. The variable is extendable to firms as well. Zh Incomings of households h. The variable is extendable to firms as well Revenue of government Zg Exogenous variables epj Price level of exports by activity j in US dollars mpc Price level of imported consumption good c in US dollars mpi Price level of imported investment good i in US dollars Consumer price index, taken as numeraire CPI FCF Foreign capital flow expressed in foreign currency Investment expenditure of government Ig Supply of capital in activity j KSj Supply of labour LS Trh, Trg Net transfers from rest of world to households (and firms), and government Coefficients αjj′ Input output intermediate delivery coefficients from sector j to j′ αjo Calibration coefficient for production function of sector j

Fiscal policy in adapted CGE models  245 αrj βj εjo

Noncompetitive import share in the output of sector j Labour elasticity of production relating to production activity j Exports quantity as a proportion of output quantity of sector j, measured for the benchmark year Export price elasticity of the world market demand for domestic products of εj sector j γhco Calibration coefficient in consumption function of household h commodity c γhc# Consumption proportions spent by household h of commodity c Cross-price elasticity of consumption for commodity c by household h γhc γgj Share of government consumption expenditure on sector j Proportion of government revenue earmarked for government consumption γg expenditure μrc , μri Competitive import share in consumption good c and investment good i , respectively μj Competitive import price elasticity applicable for goods of activity j πjh , πφ, Profit distribution shares generated in activity j that is received by household h (extendable to firms), government, g, and rest of world, r. πjr θjc , θji Quantity shares of domestically produced consumption goods c by sector j and of domestically produced investment goods i by sector j, respectively θcj , θij Conversion coefficients from consumption commodity c to sector j, and from investment good i to sector j, respectively ρhh′ Transfer payment between household h (and firms), as proportion of their incomes ρgh Government transfers to household h (and firms), as proportion of government earmarked transfers Government earmarked transfers, as proportion of government revenue ρg revenue τhg Direct tax rate paid by households h to government g; the index h is extended to cover firms τjg Indirect tax rate paid by sector j to government g ωjh , Wage distribution shares generated in activity j that goes to households, h, and ωjr rest of world, r ζc Weights in the consumer price index are shares of commodity c in total household consumption

Description of the model can be brief, since most of its equations were encountered in previous chapters, and more generally in Chapter 2. The first block of six equations refers to the production function and the factor market. The CGE model for the Netherlands was initially formulated and applied for various labour types, as are most of the economy-wide models in this book. Later adaptations of the CGE model for the Netherlands included structural constraints that worked with the whole labour force as one labour type. To simplify the presentation in this chapter, we choose to present the labour force in the model in terms of one labour type throughout the whole chapter. Eq. 1 considers the Cobb-Douglas production function, but this will be elaborated towards a constant elasticity of substitution (CES) production function in a later section. Eqs. 2 to 6 solve for the remuneration of labour and capital under full utilisation of the supplies of labour and capital.

246  Fiscal policy in adapted CGE models Box 1 Production functions and factor markets Gross output by sector (production functions) Xj = XPj {1 / (1 − Σj (αj′j − αrj − τjg )} [αjo ΠLDjβj . KDj (1 − βj) ]

.j (1)

Value added by sector Vj = (1 − Σj αj′j XPj′ / XPj ) Xj − αrj Xj − τjg Xj

j (2)

Remuneration and use of labour LR . LDj = βj Vj

j (3)

Remuneration and use of capital KRj . KDj = Vj − LR . LDj

j (4)

Labour demand equals labour supply by aggregated skill q Σj LDj = LS

(5)

Capital demand equals capital supply by sector j KDj = KSj j (6)

The second block, eqs. 7 to 12, specifies the incomes received by household groups h (and firms), and government. Eq. 7 specifies factor incomes of household groups and firms. Eq. 8 gives disposable income after taxes and transfers. Eq. 9 shows government revenue consisting of factor income from public enterprises, direct and indirect taxes, and net transfers from abroad. Government current spending is on transfers, Tgh , and on consumption, Cgj , that are coupled proportionately to government revenue as done in eqs. 10 and 11, and are thus endogenously specified. Investment spending is exogenous Ig. All government spending less government revenue returns a government budget gap, which is usually a deficit, in eq. 12. These specifications will undergo modifications in the restructured CGE model.

Box 2 Institutions accounts Primary income of recipient institutions (taxable income) h = 6 groups, firms (7) Zh = Σj ωjh LR . LDj + Σj πjh KRj . KDj Disposable income (after taxes and transfers) Y h = (1 − τhg )Zh + Σ h′ ρhh′ Zh′ + Τgh + FXR . Trh Government revenue Zg = Σj πjg KRj . KDj + Σhτhg Zh + Σj τjg Xj + FXR . Trg Government spending and budget deficit Tgh = ρgh ρg Zg Cgj = γ gj γg Zg [Σj Cgj + Σh Tgh ] + Ig = Zg + GBD

h (8) (9) (10) (11) (12)

Fiscal policy in adapted CGE models  247 The third block, eqs. 13 to 20, specifies the composition of consumption, investment, exports, and imports in ways equivalent to models in previous chapters, with special reference to Chapter 2. In eq. 13, consumption functions generating Chc include the income effects via Y h and the price effects via composite price levels SPc and own and cross price elasticities. In eq. 14, derivation of SPc is specified in terms of the domestically produced shares and the import share, multiplied by their respective price levels. Eq. 15 specifies the consumer price index as a weighted average of SPc. Eq. 16 allocates the private consumption by commodity c, and government consumption expenditures to sectors of activity j. Eq. 17 specifies the composite price for investment goods, SPi. Eq. 18 allocates investment goods i to the sectors of origin j. The foreign trade equations, 19 and 20, specify quantities of exports and competitive imports. The realised exports can exceed the quantity share of exports in output in the benchmark if the domestic export price is below the world export price, subject to an export price elasticity. The quantity of competitive imports is modelled along the same lines. The competitive import share in consumption and investment for the benchmark year can be exceeded or diminished, depending on relative domestic and import prices and the import price elasticity. (Note that the competitive import share μrc combines with domestically produced supply shares θjc in defining SPc in eq. 14.) The fourth block, Box 4, displays national accounting balances. In eq. 21, supply of goods consisting of output and imports equals the demand for these goods in

Box 3 Product markets Private commodity consumption γ Chc = γhco Π SPc hc {γhc# Y h } (1 − Σ c γhc ) Prices of composite commodities SPc = Σj θjc XPj + μrc (FXR . mpc ); whereby Σj θjc + μrc = 1 Consumer price index CPI = Σc ζc SPc Private and government commodity consumption allocated to sectors Cj = XPj Σc θcj (Σ h Chc / SPc ) + Cgj ; whereby θcj= θjc / (1 − μrc )

h.c (13) c (14) (15) j (16)

Price of gross fixed capital formation goods SPi = Σj θji XPj + μ ri (FXR . mpi); whereby Σj θjc′+ μrc′ = 1

(17)

Gross fixed capital formation allocated to sectors Ij = XPj Σi θij (I / SPi ); whereby θc′j = θjc′ / (1 − μrc′ )

(18)

Exports by sector Ej / XPj = εjo (Xj /XPj ) {(FXR . epj) / XPj }εj

j (19)

Competitive imports allocated to sectors Mj / XPj = Σ cθcj [(μrc Σh Chc / SPc ) {XPj / Σcθcj (FXR . mpc )} μj + Σi θij μri I / SPi

j (20)

248  Fiscal policy in adapted CGE models terms of intermediate deliveries, consumption, investment including inventory change, and exports. In eq. 22, the savings investment balance is specified. The model is complete at eq. 22, consisting of 22 sets of equations against 22 sets of endogenous variables. Solutions of the unknowns will automatically result in the foreign payments balance, which is formally kept outside the model, and denoted as eq. A1. Following Walras Law, in an economy with n markets, it is sufficient to solve n − 1 simultaneous equations for market clearing.3

Box 4 National accounting balances Product market balances Xj + Mj = Σ j (α jj′ XPj / XP j′ ) Xj′ + Cj + I j + Ej

j (21)

Savings investment balance (Σh Y h − Σh Σc Chc ) + (Zg − Σj Cg−Σh Tgh ) + FXR . FCF = I

(22)

Foreign payments balance Σj Mj + Σj αrj Xj + Σjωjr LR LDj + Σj Π jr KRj KDj = Σj Ej + FXR    (Σh Trh + Trg ) + FXR . FCF

(A1)

The clearance of the product balances occurs via sector prices and investment which carry the burden of adjustment (that is, closure B, see Chapter 2). The clearance of the savings investment balance is done via the foreign exchange rate, exports, and imports, as foreign capital flow is assumed given (that is, closure D, see Chapter 2). Because the two accounting balances are to be solved simultaneously along with other equations pertaining to exports and imports, specific adjustors cannot be assigned to the specific balances. The closure is a joint B plus D closure. This closure is generally described as neoclassical, emphasising that the value of the supply side of savings (domestic savings are generated by the model, and net foreign savings, that is foreign capital flow, are prefixed exogenously) precedes and determines investment. 2.2  Applications with zero price elasticities: calibration and results of policy simulations The parameters of the CGE model are derived from the 1981 benchmark values of nominal variables in the SAM under the assumption that the benchmark SAM represents the ex post equilibrium situation. Furthermore, because price variables in the CGE model are expressed as indices, and they take the value of 1 for the benchmark, the values of variables and their relationships in the SAM are interpretable in quantity (volume or real) terms. The SAM cannot give data on price elasticities relating to the product markets. These have to be obtained from various sources or calculated from supplementary data. The reliability of such estimates is often a matter of discussion. It should

Fiscal policy in adapted CGE models  249 be determined in which ways the model results are affected by the values chosen for these elasticities. Application of sensitivity analysis that employs alternative estimates is a helpful tool in exploring the margins within which the values of these elasticities should lie. With this objective in mind, we shall eliminate the effects of price elasticities relating to the product markets in the basic version of the model, and run several policy simulations and examine their results. We shall in the next section, that contains various elaborations, assign the price elasticities positive values based on various sources, and rerun the same policy simulations and examine and compare the results. This will allow identification of the depth of the price effects in the product markets in the CGE model, and the margins within which their estimates can be considered as robust. There are three types of price elasticities for the product markets in the model that are treated along the above lines. In eq. 13, cross price elasticities of consumption, γhc, will be put at zero, so that the equation becomes simplified to Chc = γhc Y h. In eq. 19, the export price elasticity, εj, will be initially set at the value of 1.0, and for XPj = EPj . FXR, the export of sector j will take the benchmark value Ej = εjo Xj. Furthermore, in eq. 20, the import price elasticity will be initially set at zero, which will make the quantity of competitive imports solely dependent on the quantities of consumption and investment. The same effect is obtainable if the relative price differential between domestically produced goods and imported goods remains constant. The basic CGE model is now ready for running the two fiscal policy simulations on the revenue side that were mentioned in the introduction. The results are in Table 12.3. The first simulation is the reduction of the indirect tax rate on the services sector, that is τjg, j = 5, by −4.44 per cent. The instrument occurs in eqs. 1, 2 and 9. Via eqs. 1 and 2, it becomes more attractive to supply more services; the wage rate is pushed up, and thus increasing disposable income, private consumption and investment; this last favours the construction sector which grows the most. Employment is withdrawn from other sectors reducing their production. Via eq. 9, the reduction in the indirect tax reduces government revenue. This works in the opposite direction. The incidence of lower government spending strikes services most, so that higher supply of services via eq. 1 is nullified by lower demand for services, resulting in the fall of the price index of services, and nominal production by −0.13 per cent as compared to the benchmark. The real production and value added of the services sector barely change. Construction is up, other sectors are down. Total effect on gross production and GDP is zero. Disposable income is up by 0.19 per cent, but government revenue is down by −0.85 per cent (note that this is less than the −1.00 per cent initial input of the policy simulation, the difference is recuperated). The net effect on the volume of production and value added is zero. How is the income distribution affected? The primary income distribution is unaffected. But the secondary income distribution is worsened as shown by the higher value of Theil’s coefficient. This happens because with the lower government revenue, there are less transfers going to households in the poorer income deciles. All things taken together, the returns of the policy are not inviting because of supply and demand forces

250  Fiscal policy in adapted CGE models working in opposite directions. Furthermore, the increased private consumption and investment lead to more competitive imports and a shortfall in the foreign payments balance, leading to an appreciation of foreign exchange currencies with respect to the national currency, that is, depreciation of own currency by 13.0 per cent. The second simulation is the incorporation of a wage (or employment) subsidy that is equal to 0.46 per cent of the wage bill. This is specified by adding the term Σj 0.46 (LR . LDj ) to eq. 3, which allows the sector firms to have access to more workers and (or) to raise wages; and by inserting the term −Σj 0.46 (LR . LDj ) in eq. 9, which reduces government revenue by the size of the subsidy. The results show that the supply enhancement effects via eq. 3 are smaller than the demand dampening effects via eq. 9. The reduced government revenue in eq. 9 leads to reduced spending and reduced demand for the labour-intensive sector, that is services, whose price index falls by −0.12 per cent, employment falling by −0.1 per cent, and output by −0.19 per cent. Agriculture is the highest gainer via positive inter-industry relations with the other sectors, but also the light and heavy industries are gainers. These shifts in production are accompanied with labour reallocations between the sectors that are much greater in the current than the previous policy, making it possible to realise a 0.31 per cent rise in the wage rate, and yet a fall in the value added of −0.13 per cent. Moreover, the three gaining sectors increase their exports, but their intermediate imports rise even more. In spite of the rise in exports and the decline in competitive imports, the intermediate imports weigh more, bringing the foreign payments balance into negative territory, and resulting in a greater depreciation of the national currency than in the previous policy. The rise in disposable income is lower and the decline in government income is higher than in the previous policy simulation. Theil’s coefficient of secondary income distribution is at about the same level for both simulations. The current policy is, by far, even less inviting to implement than the previous policy.

3  The elaborate CGE model 3.1 Elaborations This section specifies two types of elaborations that were applied to the basic CGE model. The elaborated model will then be used to rerun the policy simulations. The analysis of the results will give some idea about the sensitivity of the model solutions to the elaborate form of the model. The first elaboration is the replacement of Cobb-Douglas production function in eq. 1 by a CES production function, as in eq. 1′. This is a more generalised function. Although there are a number of variants of the CES production function, Arrow et al. (1961) is the standard specification that applies, where ϕ is the efficiency parameter, δ is the distribution parameter, and σ is the substitution parameter that determines the constant degree of substitutability of capital and labour.

Fiscal policy in adapted CGE models  251 Table 12.3 Basic CGE model Netherlands: policy simulations of reductions in indirect tax and direct tax

Wage rate LR Total employment Σj LDj Highest sector empl LDj Lowest sector empl LDj Total wages Σj LR . LDj Total profits Σj KRj . KDj Total value added Σj Vj Disp. private income Σh Y h Government revenue Zg Theil coef of primary income distribution Theil coef of secondary income distribution

Reduce indirect tax for services

Reduce wage tax (same as wage subsidy)

 0.32  0.00 Cn = 0.12 Ag = −0.13  0.32  0.30  0.31  0.19 −0.85  0.00

 0.31  0.00 Ag = 0.41 Sv = −0.10 −0.16 −0.08 −0.13  0.09 −1.00  0.01

 0.31

 0.29

Total output price (XPj ) Total output value ΣjXj Highest sector Xj Lowest sector Xj Highest composite SPc Lowest composite SPc Total consumpt. Σj Cj Investment I Total exports Σj Ej Total comp. imports Σ j Mj Foreign exchange rate FXR

Reduce indirect tax for services

Reduce wage tax (same as wage subsidy)

 0.07  0.07 Cn = 0.34 Sv = −0.13 Cn = 0.25 Sv = −0.13  0.18  0.29 −0.01  0.05

−0.01  0.01 Ag = 0.20 Sv = −0.19 Ag = 0.18 Sv = −0.12  0.08 −0.02  0.20 −0.17

 0.13

 0.25

Note Values are simulated results less benchmark values, expressed in percentage points. Abbreviations for sector j (and commodity c): Ag = Agriculture, Li = Light industry, Hi = Heavy industry, Cn = Construction, Sv = Services

Xj = XPj {1 / (1 − Σj (αj′j − μ j − τjg )} ϕj [(1 − δj ) KD1 − σj + δj LDqj − σj] − 1/ σj j (1′) The values of the substitution parameters employed are from Kuipers et al. (1980), having about the same sector classification as in this model. The substitution elasticities for all sectors, except agriculture, are smaller in the CES compared to the Cobb-Douglas production function. The lower substitution possibilities in the CES—as will be apparent from the repeated policy simulations below—show up in smaller changes in quantity variables in association with larger changes in price variables. The second type of elaboration is meant to activate the price effects in the model by assigning positive values to various price elasticities. The price elasticities occur in three equations. The own and cross price elasticities in the private consumption functions of household groups, γhc, appear in eq. 13. Given the Cobb-Douglas form of the function the income elasticity in this equation is 1 − Σc γhc. These elasticities for about the same categories of commodity c and household groups h are borrowed from Nieuwenhuis (1985, p.23). As for the export price elasticity, εjr, in eq. 19 we borrowed values for these elasticities from two macro models for the Netherlands that give values in the range between 1.56 and 2.0, which are substantially higher than the initial value of

252  Fiscal policy in adapted CGE models 1.0. The same sources were used for pegging values of the import price elasticities, μ j , in eq. 20.4 3.2  Applications with positive price elasticities: results of policy simulations While the introduced lower substitution possibilities in the CES tend to reduce the shifts in factors between activities, the introduced positive values of various price elasticities tend to increase the reallocation of demand in the product markets. The joint interaction of both elaborations is borne in the results of the policy simulations, Table 12.4. Compared to the benchmark values, the policy simulations with the elaborate model gives lower mutations than in the case of the basic model. This indicates that the dampening effects of the CES elaborations are more influential than are the reallocative effects of price elasticities of consumption, exports, and imports. Otherwise, all the simulation results are in the same direction. Because the policy simulations raise the wage rate, further rises occur in income, consumption, investment, and imports. But as the policy simulations reduce taxes, this ends in reduced government spending, has a negative effect on the price and output levels of the services sector, and would shift resources from this sector to other sectors. As the other sectors are exporters and importers the foreign trade balance is affected. The net effect is a depreciation of the FXR. There is a relative improvement in the performance of the wage subsidy policy over the reduced indirect tax policy under the elaborate model, due particularly to the CES impact of lesser factor substitution and lesser shifts of labour between sectors. Values of total output, private income, government revenue, consumption, and investment improve, while the depreciation of FXR is less.

4  Structuralist CGE model 4.1 Modifications Even though the Netherlands is generally recognised as a free market economy, and the CGE model described so far reflects fully this recognition, the real world of the factor market differs from the premises of the free market. The wage rate in the Netherlands is not equal to its marginal productivity, but is the result of periodic negotiations between labour and employers’ organisations, with some participation of government circles. The wage rate tends to be sticky in the short run and mainly coupled to changes in the consumer price index, CPI. As the wage rate ceases to be the factor market adjustor, unemployment becomes the adjustor. As the unemployed labour receives unemployment benefits, a direct link is created between the factor market balance and the government fiscal balance. Accordingly, the CGE model requires restructuring to allow the endogenous determination of unemployment benefit transfers in the fiscal balance. This makes the model more realistic and also opens the possibility of treating other government transfers and consumption expenditure in more realistic ways. Apart from unemployment benefits, the other component of transfers is social

Fiscal policy in adapted CGE models  253 Table 12.4 Elaborate CGE model Netherlands: policy simulations of reductions in indirect tax and direct tax

Wage rate LR Total employment Σj LDj Highest sector empl. LDj Lowest sector empl. LDj Total wages Σj LR . LDj Total profits Σj KRj KDj Total value added Σj Vj Disp. private income Σh Y h Government revenue Zg Theil coef of primary income distribution Theil coef of secondary income distribution

Reduce indirect tax for services

Reduce wage tax (same as wage subsidy)

 0.28  0.00 Cn = 0.07 Ag = −0.05  0.29  0.27  0.28  0.15

 0.38  0.00 Ag = 0.35 Sv = −0.10 −0.07 −0.01 −0.05  0.16

−0.88  0.00

−0.94  0.00

 0.31

 0.29

Reduce indirect tax for services

Reduce wage tax (same as wage subsidy)

Total output price (XPj ) Total output value ΣjXj Highest sector Xj Lowest sector Xj Highest composite SPc Lowest composite SPc Total consumption Σj Cj Investment I

 0.07  0.06 Cn = 0.30 Sv = −0.16 Cn = 0.26 Sv= −0.16  0.14  0.26

0.01 0.03 Ag = 0.18 Sv = −0.16 Ag = 0.13 Sv = −0.09 0.15 0.07

Total exports Σj Ej Total comp. imports Σ j Mj Foreign exchange rate FXR

 0.02 −0.08

0.13 0.01

 0.20

0.14

Notes Values are simulated results less benchmark values, expressed in percentage points. Abbreviations for sector j (and commodity c): Ag = Agriculture, Li = Light industry, Hi = Heavy industry, Cn = Construction, Sv = Services

security transfers. The net transfers here (payments out less collected premiums) are determined by demographic factors, welfare considerations, and a coupling of the payments out to the CPI. As for the government consumption expenditure on sectors, Cgj, these tend to be sticky too, and are better formulated as exogenously given in real terms, and converted to nominal values based on the applicable sector prices. The above mentioned adaptations lead to reformulating eqs. 3 and 5 into 3′ and 5′; coupling thus the wage rate to its benchmark value, λo, and the CPI in eq. 3′; and instituting labour unemployment, LU, as the adjusting variable in the equalisation of demand to supply of labour in eq. 5′. The unemployment percentage in 1981 was at the very high level of 10.4 per cent. LR = λo CPI j

Σj LDj + LU = LS

(3′) (5′)

The other reformulations necessary are with regard to the specification of government transfers and consumption expenditure in eqs. 11 and 12. Government transfers to households Tgh are now specified as the sum of (a) unemployment benefits coupled to the unemployment numbers, and (b) social security payments that are demographically given in real terms and coupled to the CPI in nominal terms, as reformulated in eq. 10′. Regarding (a), the height of

254  Fiscal policy in adapted CGE models unemployment benefits is a minimum proportion λm of the average wage rate LR, while the distribution shares of the unemployment numbers on the income decile household groups, λgh, tend to be stable and are borrowed from Fulpen et al. (1985), p.103. Regarding (b), the height of social security payments in real terms to an income decile household group is dependent on exogenous factors such as the demographic composition of the group, income level, and welfare norms, assumed here given at Tsho, which is obtainable after deducting the above mentioned unemployment benefits from total transfer payments as found in the SAM benchmark. This given social security transfer is projected to increase linearly by a demographically given proportion ρgho for the forthcoming years t, and is coupled to the CPI to give nominal values. Furthermore, government consumption expenditure by sector j, Cgj, is specified exogenously starting from the benchmark level in real terms Cgjo / PCj , and projecting volume Cgjo forwards following pronounced government policy targets at the time. The target is that the real spending is to fall linearly by a small proportion γgo for some forthcoming years t, as in eq. 11′; this is in contrast to its previous endogenous coupling to government finance. Although the official target of the Dutch government in 1981–2 was to reduce consumption expenditure by 1 per cent for each coming year t, our simulation went for 0.5 per cent only for the year examined. Ex post, the average reduction actually achieved by government over the period 1982–90 was even less, at about 0.3 per cent per year.5 Tgh = (λm LR) λgh LU + Tsho (1 + ρgho t) CPI

(10′)

Cgj = PCj [Cgjo (1 − γgot)]

(11′)

So far, modifications applied were reformulations of the same equations of the elaborated CGE model, namely eqs. 3′, 5′, 10′ and 11′. As a result, the value that government budget deficit can take, GBD in eq. 12, has become completely open ended. While government revenue is dependent on taxes, and the levels of income and economic activities, government transfers and consumption expenditures go their own way depending on unemployment, the CPI, and exogenous determinants. The modifications have made both ends become independent of each other. For a healthy management of the fiscal balance it is essential to constrain GBD. This is especially relevant in the context of models that are restricted to the real economy, and do not contain monetary variables. To avoid the situation that an open-ended GBD can result in an inflationary pressure which is not accountable in the current model, the GBD needs to be constrained by a fiscal rule. In fact this reverts more or less back to the official policy practice of the sixties, and in anticipation of what was targeted towards the turn of the century and ahead of the realisation of the EMU. The fiscal rule is introduced by adding an equation that specifies that the GBD should amount to a policy given proportion, νg, of the GDP, as in eq. 23 below. GBD = νg Σj Vj

(23)

To make the model determinate the introduction of eq. 23 has to be compensated by introduction of an unknown variable. The policy simulation introduces

Fiscal policy in adapted CGE models  255 the direct tax rate adapter, TAX, as the endogenous variable which adjusts the fiscal balance so that it conforms to the newly added budget restriction. The direct tax rates τhg remain unchanged but they are multiplied by the endogenous variable (1 + TAX). The modifications occur in two equations only. In eq. 8, the first term becomes (1 − (1 + TAX)τhg )Zh, so that households will pay an adjusted direct tax. Similarly, in eq. 9, the second term becomes Σh (1 + TAX) τhg Zh, so that the government collects adjusted direct taxes. 4.2  Results of policy simulations Table 12.5 gives the percentage change of the simulated result relative to benchmark values. Total output increases by 0.18 per cent and 0.26 per cent in the two policies. Total employment increases by 0.32 per cent and 0.42 per cent. The employment effects could have been higher if not for the following considerations that limit the effects. First, the low substitution possibilities of labour to capital in the CES production functions, to which we have referred before, tend to dampen the absorption of labour, as the marginal productivity of labour diminishes quickly with increasing output. (We did run the structural model with Cobb-Douglas production functions as well; the employment effects of the policy simulations were about 30 per cent higher.) Second, a rise in domestic output is accompanied by a relatively higher rise in imports than in exports due to the globally oriented character of the Dutch economy; this depreciates the FXR by 0.32 per cent in the two simulations, thus diminishing the value of the marginal productivity of labour and reducing returns for employing more labour. Sector-wise, while the output of construction increases most (+0.42 per cent) in the indirect tax policy, it is heavy industry that gains most (+0.47 per cent) in the wage subsidy policy. Heavy industry is advantaged because of its relatively higher substitution possibilities and its greater export response. In both simulations, the services gains the least given its large dependence on government finance. It is important to note, however, that the service sector is not negatively affected as was the case in the basic and elaborate models. In the structural model, the government is able to recuperate more revenue back than it originally laid down in initiating the policy simulation. Government revenue Zg is shown to increase by 0.11 and 0.13 per cent in the two policies, while the initial cost of the policy simulation amounted to −1.0 per cent of the budget. This is made possible via the endogenous tax rate adjustor, TAX, which comes out at +0.23 per cent and +0.11 per cent in the two policy simulations. The required TAX rise is sufficiently moderate to allow total private disposable incomes to increase by 0.13 per cent and 0.15 per cent in the two simulations, and push consumption and investment to higher levels. The moderate rise in TAX needed to maintain the fiscal rule (GBD = νg Σj Vj ) is due to: (a) the rise in output and incomes creates more tax revenue, irrespective of TAX; (b) the rise in employment reduces the payment of unemployment benefits by government and reduces the burden of balancing the budget deficit with higher TAX; and (c) the reduced price level of the services goods by −0.22 per cent and −0.18 per

256  Fiscal policy in adapted CGE models cent reduces nominal government current expenditure by similar percentages and thus reduces the need for fiscal adjustments. The impact so far described makes both policy simulations attractive to follow, with the wage-subsidy policy performing better than the indirect tax reduction policy. However, in both policies the disposable incomes of the lowest four income deciles fall, and the other income deciles gain. Theil’s coefficient increases by about half a percent (0.40 per cent and 0.53 per cent in the two simulations). The shift in the income distribution is highlighted in Table 12.6. The relative loss of the lower deciles is due to (a) the fact that they benefit less from economic growth since social security payments and related transfers form the larger part of their income, and (b) because unemployment and unemployment benefits concentrate

Table 12.5 Restructured CGE model Netherlands: policy simulation of reductions in indirect tax and direct tax Reduce indirect tax for services Wage rate LR Total employment Σj LDj Highest sector empl LDj Lowest sector empl LDj Total wages Σj LR . LDj Total profits Σj KRj . KDj Total value added Σj Vj Disp. private inc. Σh Y h Government revenue Z g Theil coef of primary income distribution Theil coef of secondary income distribution Total unemployment Σq LUq

Reduce wage tax (same as wage subsidy)

 0.00  0.00  0.32  0.42 Ag = 0.53 Ag = 1.16 Li = 0.21 Sv = 0.33  0.31  0.04  0.55  0.36  0.40  0.11  0.13  0.15  0.11  0.13 Not calculated  0.40

 0.53

−2.73

−3.65

Total output price (XPj) Total output value ΣjXj Highest sector Xj Lowest sector Xj Highest composite SPc Lowest composite SPc Total consumption Σj Cj Investment I Total exports Σj Ej Total comp. imports Σ j Mj Foreign exchange rate FXR Direct tax scaling index TXS

Reduce indirect tax for services

Reduce wage tax (same as wage subsidy)

 0.05  0.23 Cn = 0.42 Sv = 0.02 Ag = 0.29 Sv = −0.22  0.12  0.39  0.22 −0.26

 0.00  0.26 Hi = 0.47 Sv = 0.06 Ag = 0.27 Sv = −0.18  0.14  0.29  0.41 −0.23

 0.32

 0.33

 0.23

 0.11

Notes Values are simulated results less benchmark values, expressed in percentage points. Abbreviations for sector j (and commodity c): Ag = Agriculture, Li = Light industry, Hi = Heavy industry, Cn = Construction, Sv = Services

Table 12.6 Netherlands: Income distribution effects of fiscal policy simulations Policy simulation

1st

2nd

3rd

4th

5th

6th

7th

8th

9th

10th

Indirect tax reduction −0.03 −0.11 −0.01 −0.01 0.11 0.17 0.17 0.19 0.21 0.22 Wage subsidy −0.06 −0.016 −0.04 −0.05 0.11 0.19 0.19 0.21 0.24 0.27

Fiscal policy in adapted CGE models  257 in the lowest deciles, a reduction in unemployment allows them to receive wages but at the same time reduces unemployment benefits.

5  Concluding remarks The scope of fiscal policy in this chapter was exclusively limited to modelling and assessing economic performance in the short (and medium term). Although the treated problem of constraining the budget deficit without sacrificing economic growth is conceived as a short- to medium-term one, the problem does not seem to be resolved once and for all. Even if the problem has been solved in the eighties or nineties, unfavourable internal and (or) external circumstances in later years throw the problem open once again. In some sense, the treated fiscal policy problems become of a long-term interest. But this is not fiscal policy for the long term. We did not touch on modelling the prospects of sustainable economic welfare and sustainable public finance formulas for the long term, which are far reaching and pose much greater challenges than the treated problem. A few alarming remarks are well placed in this context. Demographic factors (that is, low fertility, greater longevity, and retirement of the baby-boom generations) can be expected to lead to a doubling of the ratio of retirees to people of working age. An ageing population drives up the claims on pensions and medical care, while a reduced workforce diminishes the base from which these claims need to be financed. A balanced budget of the public sector that meets these conflicting forces is a critical challenge in the long run.

13 Normed planning of human resource development A roadmap model for Ethiopia

1 Background UNDP’s Human Development Index, HDI, ranked Ethiopia in 2005 as number 169 out of 177 countries.1 In 1990 Ethiopia was ranked as number 111 out of 130 countries, implying that the country was relatively better off in the past. Lifting human resource development (HRD) in Ethiopia to levels comparable with middle-level countries is a long-term undertaking and may require a couple of decades or more of concentrated and consistent long-range perspective investment under governmental direction. This comes to drawing up an HRD– education sector roadmap for 20 to 30 years, which is what this chapter will present. The chapter is based on an advisory assignment in 2009 on behalf of United Nations Industrial Development Organization (UNIDO) to Ministry of Education (MOE), Ethiopia. Drawing from the broader national development goals and the international experience of the better performers, the aim of the policy model in this chapter is to outline a roadmap for the next two to three decades. In particular, what should be the educational mix at large in terms of enrolments, teachers, and so forth at the various educational levels? What will be the financial implications? Which features within the education system need to be highlighted? Which other factors outside the education system will be critical and need to be monitored and observed? Drawing an HRD roadmap for 20 to 30 years cannot rely on methods of manpower forecasts or educational returns and market signalling which are valid for the short and medium term. Given the pace of technology and the rapidly changing sector mix in the economy, and global competition, these methods may not even be valid for the medium term. The policy modelling of a 25 years roadmap with base year 2005, can only be drawn by making judgmental use of the past experience of selected countries that are known to have performed better in HRD, and whose realised structure of the education sector can be projected backwards so as to apply to Ethiopia. If the predicted paths of the select group of countries are followed along with associated institutions and policies, it is most likely that similar successful performances would occur in Ethiopia. The chapter is organised as follows. Section 2 considers a select group of countries whose development pattern will be used to outline the Ethiopia roadmap. In Sections 3, and 4 a model is formulated and applied for planning basic parameters

Normed planning  259 of the educational system such as enrolments, teachers, recurrent and capital costs, and the breakup of finance into government and private resources. The model is estimated from data of the select group, applied to Ethiopia, and alternative transitional paths simulated. Section 5 comments on qualitative features in HRD that made achievements in the select countries feasible, and which are prerequisites for the realisation of the modelling results. The section also examines how prepared are the various regions in Ethiopia towards adopting the roadmap, and how to bridge the lags. Ultimately, any transformations along the lines of the roadmap have to take place in the regions, and regional conditions are known to vary significantly. Section 6 treats matching aspects between demand and supply of human resources in the medium term. Section 7 emphasises that there are non-HRD aspects that have been critical for the socio-economic development of the select countries, and if the roadmap is to be followed, its success requires that links with the non-HRD aspects be specified, measured, and realised. The section reflects on forecasts of related economy-wide imbalances in Ethiopia. Section 8 concludes.

2  The targeted select group of countries Several criteria were formulated and used in the selection of the targeted countries. The criteria aim at selecting the best performing countries in HRD and economic development at large. The countries should (a) belong to the mediumlevel development group, (b) have a ranked HDI higher than the ranked index of GDP per capita in dollars of purchasing power parity, $ppp, (c) have achieved a high economic growth in the last decade and be expected to continue, and (d) have a population size comparable to Ethiopia or bigger. These criteria have assured an optimal selection of top performers. The selection resulted in eight countries. These are China, Vietnam, Indonesia, Philippines, India, Egypt, Brazil, and Russia. On request of MOE authorities, two more Asian countries lying much further ahead are added to the select group; these are Korea and Japan, resulting into ten countries. The ten countries show big jumps in their ranked HDI over the last decade. It is not surprising that six of the ten countries come from broader East Asia, which is the world region with the highest socio-economic performance.2 The modelling approach is to use data of the ten countries over the period 1990–2005 in estimating a long-range conditional forecasting model that can then be applied to Ethiopia towards the year 2030. Since this database is incomplete for some countries and (or) years the forecasts may not fully reflect the experiences for the missing countries and (or) years. Once selected, a compact databank was then set up for the selected ten countries for the period 1990–2005, which was then used to estimate the equations of the model.

3  Long-range targeting model for HRD The model aims to display destination structures of the educational system over the long term. The elements in these structures cover the mix between primary,

260  Normed planning secondary, and higher education, as regards quantity (that is, relative enrolments), quality (that is, relative teacher/student ratios), and balanced costs, as well as the financing of these levels by government and private sources. The basic assumption made is that HRD structural variables follow paths that are interdependent with the socio-economic development phase, as represented by GDP per capita (GDPpc ) in $ppp. The HRD variables are specified as dependent on GDP per capita. The model contains seven reaction equations with regression coefficients to be estimated from the select countries databank, and seven definitional equations that hold the educational system together and allow formulating a consistent and comprehensive development plan for the educational sector as a whole, and its balanced financing. Together, the model consists of 14 sets of equations in 14 sets of endogenous variables that have to be solved, assuring thus a determinate model. Some of the endogenous variables—that are explained in regressions—are structural ratios such as the gross enrolment ratios by educational level, denoted by GERe.. Other structural variables are student/teacher ratio, STRe, and recurrent cost per student, RCSe. The equations in Box 1 can be briefly described, making use of the notations in Table 13.1. All the regressed equations are specified below in a general linear form, but when estimated, the best-fitting form of the equation was selected which need not be the linear form. Annex Table, 13.8 gives the selected best fits. The model is very simple and is solved by consecutive substitutions. Take first eqs. 1 to 6. Regressions of reaction eqs. 1, 3, and 5 allows predicting for Ethiopia from these three equations the GERe, STRe and the RCSe, respectively. Next, inserting these outcomes in eqs. 2, 4, and 6 results in the number of enrolments, ENRe, number of teachers, TCHe, and the recurrent expenditure costs of education by level, REEe. Reaction eq. 7 specifies the dependence of the capital cost: recurrent cost ratio, KRC, on GDPpc. This is formulated for all educational levels taken together. Turning to the equations for governmental financing of education, there are here two reaction eqs. 8 and 9 which determine, consecutively, the size of the overall government budget for spending on all sectors, GTS, and that part of which that goes to education, GTE.3 In eq. 10 the capital cost of education is deducted from total spending on education to give government recurrent expenditure on all education, GRE. Eq. 11 is a regression equation which estimates the distribution of this GRE on the three educational levels to give GREe. Eq. 12 is a definitional equation. GRE does not add up to the sum of GREe because there is the spending on overheads and other activities in education, GRO, which has to be accounted for. This is obtained as a residual. This variable occurs only in this equation and has no impact on other variables in the model, but it is an item which is highly informative for the work of budget planners in balancing the various components of the educational expenditure. To find the non-government recurrent expenditure on education by level, NREe, deduct the government recurrent spending by level from the total recurrent expenditure per level, as in eq. 13. Furthermore, the model sums all educational

Normed planning  261 Table 13.1 Notations Index e denotes educational level, where e = 1 for primary, e = 2 for secondary, and e = 3 for tertiary Endogenous variables and (structural ratios) ENRe Number of enrolled students by level e TCHe Number of teachers by level e REEe Recurrent expenditure cost of education by level e Government total spending on all sectors GTS Government total expenditure on education GTE Government recurrent expenditure on education GRE GREe Government recurrent expenditure on education by level e Government recurrent expenditure on education in overhead and other GRO activities NREe Non-government recurrent expenditure on education by level e Total government and non-government expenditure on capital and recurrent TGN costs of education GERe Gross enrolment ratio by level e STRe Student / teacher ratio by level e RCSe Recurrent cost per student by level e expressed relative to resource base, i.e. recurrent cost per student / GDP per capita, both are measured in national currency Ratio of capital cost/ recurrent cost KRC Share of TVET (technical and vocational education training) enrolment in TVS total secondary enrolment Share of natural sciences students in total tertiary enrolment NSA Exogenous variables GDPpc Gross domestic product per capita in $ppp, in constant prices of year 2000 Ethiopia’s gross domestic product in birr millions, in constant prices of 2000 GDP Ethiopia’s total population POP Ethiopia’s share of the age group e, corresponding with level e, in the total SAGe population

expenditure in eq. 14 (that is, total government and non-government expenditure on education, including all recurrent and capital costs), to give total government and non-government spending on education, TGN. Simple as it is, besides comprehensively projecting the various components of a plan for education, the model provides checks on important key relations such as the distribution structures of enrolments, teachers and spending, and the comparative gross enrolment ratios and student/teacher ratios in the three levels. The model allows also for evaluating the cost per teacher per level and its relation to the GDP per capita. For policy appraisal purposes, the total expenditure on education can be calculated as a proportion of the GDP, that is, TGN / GDP.

262  Normed planning Box 1 Model specification Educational structure of enrolments, teachers, and cost GERe = α1e + β1e GDPpc ENRe = GERe . SAGe . POP STRe = α3e + β3e GDPpc TCHe = [1/STRe ] ENRe RCSe = α5e + β5e GDPpc REEe = RCSe . ENRe KRC = α7e + β7e GDPpc Public and private financing of education GTS / GDP = α8 + β8 GDPpc GTE / GTS = α9 + β9 GDPpc GRE = [1 / 1 + KRC] GTE GREe / GRE = α11e + β11e GDPpc GRO = GRE − GRE1 − GRE2 − GRE3 NREe = REEe − GREe TGN = GTE + (1 +KRC) Σe NREe

e = 1, 2, 3 (1) e = 1, 2, 3 (2) e = 1, 2, 3 (3) e = 1, 2, 3 (4) e = 1, 2, 3 (5) e = 1, 2, 3 (6) (7) (8) (9) (10) e = 1, 2, 3 (11) (12) e = 1, 2, 3 (13) (14)

As the roadmap model contains some basic structural relationships within each level of education and between the various levels of the education sector, the roadmap results have the additional advantage of providing systematic and comprehensive sets of monitoring indicators to check the pace of achievement and lags for specific items of HRD at specific future periods.

4  Roadmap results and transition paths to destinations 4.1  Estimation, results, and apparent gaps For estimation of the regression equations a compact databank was set up that brings together data for the above variables for the ten countries in the period 1990–2005. The databank relies heavily on the World Bank Development Indicators Databank, next to additional material from UNESCO, UN Population Division, and UNDP Human Development Reports. The number of observations covered from the ten countries varied per estimated relational equation depending on data availability and validity. The shapes of the relational equations fitted included linear, logarithmic, powered, and polynomial, as shown in the Annex Table 13.8. An indication of the extent to which the observations of the ten countries fit together in case of a specific estimated relational equation is provided by calculation of R2 for that equation. The average R2 for all the estimated relational equations based on data from the ten countries is R2 = 0.73. The figure of 0.73 is reasonable by conventional standards, given the data diversity of the ten countries and non-uniformities in definitions and data coverage. It is also important to note that the model has the advantage of including several checks and balances via definitional equations that correct for off-line regression estimates that may creep in.

Normed planning  263 The first use of the estimated model is to fill in GDP per capita of Ethiopia for 2005 in the equations and solve for the predicted HRD positions of Ethiopia for 2005. These can be checked with the actual position of the country in 2005. This is done in Table 13.2. The gaps tell a lot about the current imbalances assessed in the framework of the select countries. Take for instance primary gross enrolment rate in primary education, its actual value is higher than the expected, which is not surprising given the very high priority that Ethiopia assigns to this target variable. But apart from this advantage, Ethiopia shows a lag in all other target variables including student/teacher ratios and recurrent cost per student in all three levels, next to significant gaps in gross enrolment roles in secondary and tertiary education reaching 57 per cent and 48 per cent, respectively. Although the actual total and expected total government financing of education are close to each other, the distribution of government spending differs, and reflects Ethiopia giving greater preference to primary at the cost of secondary and tertiary. The roadmap assessment suggests that the government has overspent in primary and underspent in secondary and tertiary HRD.

Table 13.2 HRD positions: gaps between required levels and actual levels for Ethiopia in 2005

GER = Gross enrolment ratio Primary Secondary Tertiary STR = Student/teacher ratio Primary Secondary Tertiary RCS = Recurrent cost per student/GDPpc Primary Secondary Tertiary Other parameters KRC = Capital/recurrent cost ratio GTS/GDP, Share government spending all services GTE/GTS, Share government total expenditure on education GREe /GRE Share government recurrent expenditure Primary Secondary Tertiary

Actual

Required

Gap

 0.86  0.19  0.031

 0.85  0.45  0.06

 0.01 −0.57 −0.48

62 49 36

30 36 27

 1.08  0.36  0.33

 0.154  0.484  4.987

 0.107  0.268  3.4

 0.44  0.81  0.47

 0.33  0.31  0.18

 0.25  0.31  0.148

 0.32  0.00  0.22

 0.59  0.18  0.16

 0.34  0.28  0.19

 0.74 −0.37 −0.15

264  Normed planning 4.2  Roadmap results of the whole model All equations of the model are then solved as a whole system to give solutions for the years 2015 and 2030. To do this, exogenous variables of Ethiopia are required for these future years. These are the population growth projections for Ethiopia as published by the UN, and the population and its breakup by eligible enrolment age groups, as in Table 13.3. For sources see the endnotes.4 These are consistent with the estimates of the Ten Year Development Plan up to 2015. The other exogenous variable for future years is the GDPpc in $ppp, in constant prices of 2000. This amounted to $965 in 2005.5 We take for 2005–15 an annual growth of the GDPpc at 5 per cent (7.5 per cent GDP growth less 2.5 per cent population growth); this gives a total growth over the ten years of 163 per cent. Starting from the value of $965 in 2005, the projection is $1573 for 2015. For the next 15 years, 2015–2030, the annual growth of GDPpc can rise by 0.5 per cent every five years in view of the greater momentum with economic development and reductions in population growth that are feasible in view of their correspondence with UN population projections. An annual growth of GDPpc of 5.5 per cent in 2015–20 would result in $2056 in 2020, followed by an annual growth 6.0 per cent in 2020–25 resulting in $ 2751 in 2025, and finally an annual growth of 6.5 per cent in 2025–30 giving $ 3770 for 2030. The equivalent of the GDPpc in 2005 in birr amounted to 1144 birr, both assumed to grow at the same rates and in constant prices of 2000.6 Multiplying by population gives the GDP in national currency. These data are found in Table 13.4. Table 13.3 Ethiopia: projected population by age groups and total Age group

Eligibility for education

7–14 Primary 15–18 Secondary 19–22 Tertiary 22 Total population

Share of age groups, SAGe , % Population in millions 2005

2015

2030

2005

2015

2030

0.1996 0.0915 0.076 0.6324 1.0

0.2025 0.0910 0.0760 0.6306 1.0

0.1934 0.0950 0.0794 0.6322 1.0

14.75  6.76  5.65 46.74 73.9

19.13  8.60  7.18 59.59 94.5

24.27 11.93  9.97 79.34 125.5

Table 13.4 Ethiopia: projected GDP and GDP per capita 2005 GDPpc in $ppp thousands in constant prices GDPpc birr thousands in constant prices Total population million GDP birr million, constant prices

0.965 1.144 73.9 84553

2015 1.573 1.865 94.5 137826

2030 3.770 4.470 125.5 330326

Normed planning  265 Applying these exogenous data to the model gives roadmap requirements (or projections) as in Tables 13.5 and 13.6. The model generates the structure of the educational system for 2030, based on projections of variables relating to number of enrolments, teachers, and recurrent and capital expenditure for the three levels and the division of these into government and non-government sources. Together, these variables form the ingredients of a comprehensive plan for the education sector. Table 13.5 gives a summary structure of the education sector, and Table 13.6 gives the details. The roadmap expects the growth of enrolments in primary education to be lower than that of secondary and tertiary so as to fulfil expected targets of gross enrolment ratios at percentages 100:71:18, for the three levels, respectively, in 2030. The enrolment structure consistent with these targets is a distribution of students along the following percentages 70:25:05 for 2030. Student/teacher ratios are required to converge in the three levels by 2030, thus resulting in a distribution of teachers similar to that of enrolments, at 74:20:6 for primary, secondary, and tertiary levels in 2030. While the recurrent cost per student divided by GDPpc is supposed to remain stable for primary education, the roadmap path expects the relative cost to increase for secondary and decrease for tertiary, but nevertheless keeping significant distances from each other. When taking unit cost and the other costing aspects mentioned above the recurrent expenditure structure is expected to be more or less equally divided into one-third for each level by 2030, precisely in the percentages of 35:33:34. The structure of financing education shifts over time towards relatively less from government and more from non-government resources. The ratio moves from 71 per cent government to 29 per cent non-government in 2005, to 61 per cent to 39 per cent in 2015, and to a change in major roles towards 41 per cent to 59 per cent in 2030. Table 13.6 shows that for 2030, government would have to allocate its educational expenditure among the three levels in the percentages 36:30:15, with the remainder in others and overheads. In contrast, non-government sources are supposed to fund primary and secondary education in equal proportions of 30 per cent and 31 per cent, and devote more to tertiary education, 39 per cent, which is generally the case in the world at large, as user charges and returns to tertiary education are tied to the private beneficiaries. What can be said about the height and pattern of the gaps shown in Table 13.5? Taking 2005 as a benchmark, the gaps between the actual and expected are significant, as can be gathered from focusing on the ten variables of enrolments, teachers, and recurrent expenditure for the three levels and total expenditure on education divided by GDP. The average of the summed gaps for these ten variables is 28 per cent, while the average of the absolute gaps is 38 per cent, see Table 13.6. The striking thing about the 2005 gaps is that the total expenditure on education as a proportion of the GDP is about 6.5 per cent for both the actual and expected, and is thus well balanced; nevertheless, significant gaps arise in the subdivisions of the educational system, due to various factors that are examined below. Of course, the fact that the development profile of Ethiopia differs from the composite of the select countries determines the gaps. Apart from this obvious statement, why are the gaps so large? This is due to generally non-controllable factors and controllable factors, which are discussed overleaf.

266  Normed planning Table 13.5 Ethiopia: roadmap structure in terms of the three educational levels (primary: secondary: tertiary) Structure Primary: secondary: tertiary

Roadmap destinations 2005

2015

2030

Gross enrolment ratios, % Student/teacher ratios Recurrent cost student/GDPpc Government exp, share %a Share government in total educ. exp., %

85:45:06 30:36:27 0.11:0.27:2.30 34:28:19 71

89:54:08 27:32:25 0.11:0.28:2.00 34:28:18 61

100:71:18 21:27:19 0.12:0.33:1.30 36:30:16 41

Enrolments structure, % Teachers structure, % Recurrent cost structure, % Non-gov. exp., share % Share non-government in total educ. exp.

79:19:02 81:16:02 41:25:35 38:05:57 29

22:21:03 79:18:03 40:28:33 36:18:46 39

70:25:05 74:20:06 35:33:34 30:31:39 59

Note a Government expenditure on overheads and others form the remaining share

The initial situation of Ethiopia in terms of a relatively large school-age populations and a low GDP are given and cannot be controlled. Ethiopia is at the top of its demographic transition, when population growth is at its highest rate. With a school-going population comprising almost 40 per cent of the population, the targeted enrolment rates would require massive spending on education that cannot be maintained with the low GDP resources available. The absolute levels of population and GDP are at opposite ends, thus causing significant gaps in enrolments, teachers, and expenditures. To demonstrate the influence of the demographic factor, results can be reported from a simulation applied to the model. The simulation fixed population for 2005 at a reduced level of 10 per cent, 66.5 instead of 73.9 millions. At the expected gross enrolment ratios, required enrolments at all three levels were diminished, and accordingly for required teachers and finances. The average summed gap fell from −0.28 to −0.22, a reduction of 6 points or 27 per cent. Regarding controllable factors causing the gaps, these lie in (first) pursued educational policies aiming at universal enrolment in primary education, and (second) the generally very costly unit costs at all levels in relative terms. First, more resources have gone to raising enrolment in primary education so as to meet the millennium goal of universal coverage by 2015; this has gone beyond the required paths as depicted by the select countries. As a result, much fewer than expected resources have gone to secondary and tertiary education. The results are enrolment imbalances in all three levels. The greater focus on enrolment numbers coincided with a lower than expected quality of education, as represented by a higher than expected student/teacher ratios, STR, which apply for all three levels. In primary education the actual STR is double the expected STR. In secondary and tertiary education the gap is half as much.

Normed planning  267 Second, unit recurrent costs in all three levels in relation to the GDPpc are much higher than expected following the select countries. This is also true if other countries than those selected were taken as the reference point. Unit costs (that is, teacher salary) relative to GDPpc in Ethiopia are remarkably high. Because the non-teacher costs vary between 6 per cent and 15 per cent of the recurrent costs, and because of the very low number of teachers per student that varies between 1:60 to 1:35, depending on the level of education, the fault lies in the relatively very high salary scales of teachers compared to GDPpc. As an example, the average annual salary for a primary school teacher in the reported period was 8191 birr for the country as a whole, against a GDPpc of 1380 birr. This gives a ratio of about six times as much. In the select countries the ratios vary around four times. In relative terms, there is thus 50 per cent more pay for primary teachers than the expected norm. For secondary and tertiary teachers the pay beyond the expected norm runs at levels reaching 100 per cent.7 The high cost is also due to a high non-academic staff-teachers ratio, especially in tertiary education.8 Ten years later, in 2015, the Ethiopia Social Development Plan (ESDP IV) plans a more balanced educational system that is closer to the expected paths of the selected countries. If the plan for 2015 is realised, Table 13.6 shows that the average summed gaps would fall significantly from the high level of −0.28 to −0.12. In terms of the average absolute gap, the improvement is from 0.38 to 0.23, which is a reduction in gaps of 40 per cent. The causes behind the gaps in 2015 are similar to those for 2005. The large population combined with a low GDP is mainly a non-controllable cause. A simulation of a 10 per cent smaller population for 2015 diminishes the average summed gap to only −0.03, down from a value of −0.12. The effect in 2015 is bigger than in 2005 due to the larger schoolgoing population in 2015, which is reduced accordingly. Among the controllable policies, the continued over-investment in the primary level as planned for 2015 causes gaps in all three levels. On the other hand, if the plan figures for reductions in recurrent unit costs, implying lower teacher salaries, were to materialise in 2015, then a major cause of the gaps would be eliminated. (The table shows some reduction in the recurrent unit costs between 2005 and 2015, and a tendency to approach the roadmap pattern.) One point on the realisation prospects of ESDP IV and the related gap that needs emphasis is the following. Table 13.6 shows the planned total expenditure from government and non-government sources as percentage of the GDP to be 7.3 per cent in 2015. This is 0.6 per cent higher than 2005, and 0.3 per cent higher than the roadmap pattern for 2015. While government provisions can be assumed to happen, realisation of the non-government provisions would require increased efforts to mobilise domestic sources, that is, parents and graduates paying user charges, and local communities and business circles paying back for external benefits from a more skilled labour force, as well as foreign assistance. 4.3  The transition phase: flexible routes to roadmap destinations The gaps between the observed situation in 2005 and the trajectory are supposed to be bridged during a transition phase and the duration of this phase is flexible,

2015

2005

200 4245

1.109 21.22 47 451534

557 2019

0.4213 3.623 40 90574

0.031 0.1739 36 4.85 4.987

0.0551 0.396 28 14133

Tertiary educational level

0.19 1.314 49 26.93 0.484 554 728

Secondary educational level

0.86 12.66 62 203.0 0.154 176 2228

Primary educational level

Plan

Actual

Gross enrolment ratio, % Enrolments millions Student/teacher ratio Teachers, thousands Recurrent cost per student/ GDPpc

Gross enrolment ratio, % Enrolments millions Student/teacher ratio Teachers, thousands Recurrent cost per student / GDPpc Recurrent cost per student in birr Secondary recurrent exp. in Bm

Gross enrolment ratio Enrolments, millions Student/teacher ratio Teachers, thousands Recurrent cost per student / GDPpc Recurrent cost per student in birr Primary recurrent exp. in Bm

Variable

0.06 0.337 27 12.5 3.4

0.45 3.042 36 84.5 0.268 306.6 933

0.85 12.538 30 417.9 0.107 122.4 1535

2005

0.08 0.575 25 23.0 2.7

0.54 4.644 32 145.1 0.281 524.1 2434

0.89 17.031 27 630.8 0.110 205.2 3494

2015

Roadmap 2030

0.18 1.794 19 94.4 1.5

0.71 8.465 27 313.5 0.326 1457.2 12335

>1.0 24.272 21 1155.8 0.120 536.4 13019

Table 13.6 Ethiopia: future development paths of enrolments, teachers, and expenditure by educational level

−0.48 −0.48 0.33 −0.61 0.47

−0.57 −0.57 0.36 −0.68 0.81 0.81 −0.22

0.01 0.01 1.08 −0.51 0.44 0.44 0.45

2005

Gapa

0.31 0.31 −0.12 0.39 1.00

0.22 0.22 −0.25 0.38 1.00 −0.06 0.17

−0.25 −0.25 −0.74 0.28 1.00 0.03 −0.21

2015

7863 3111

5705 992

0.0558 0.0104 0.0662 0.84 0.16 (b) (b)

0.0465 0.0261 0.0726 0.64 0.36

0.31 0.15 0.53 0.164 0.251 0.055

Gov total exp. on education/GDP Non-gov. total exp. on educ./GDP Total gov., non-gov. exp. on educ/GDP Share education finance by gov. Share of ed. finance by non-gov Average of summed shaded gaps Average of absolute shaded gaps

National budget/GDP Share gov. exp edu in national budget Share gov. rec. exp. on primary educ. Share gov. rec. exp. on secondary educ. Share gov. rec. exp. on tertiary educ. Share gov. rec. exp. others, overhead

Total recurrent expenditure Capital/recurrent cost ratio Total recurrent and capital expenditure

Recurrent cost per student in birr Tertiary recurrent expenditure in Bm

Variable

0.0459 0.0187 0.0646 0.71 0.29

0.31 0.148 0.34 0.28 0.19 0.19

3779 0.25 4723

3890 1311

2005

0.0428 0.0268 0.0696 0.61 0.39

0.31 0.138 0.34 0.28 0.18 0.2

8821 0.22 10762

5036 2893

2015

Roadmap

0.0336 0.0487 0.0823 0.41 0.59

0.28 0.12 0.36 0.3 0.16 0.18

37380 0.145 42800

6705 12026

2030

−0.22 0.44 0.03 0.19 −0.46 −0.28 0.38

0.00 0.22 0.74 −0.37 −0.15 −0.62

0.04 0.32 0.11

0.47 −0.24

2005

Gapa

−0.09 0.03 0.04 0.04 0.07 −0.12 0.23

0.00 −0.09 −0.56 0.41 −0.39 0.73

−0.06 −0.44 −0.15

−0.56 −0.08

2015

Notes a Gap 2005 = (actual-roadmap) / roadmap, Gap 2015 = (plan-roadmap) / roadmap, Actual and Plan figures are from Education Statistics Annual Abstract 2006, and MOE Ten Years Education Sector Plan 2006–07 to 2015–16, respectively. b The average of the absolute gap removes the negative signs before averaging. The average of the summed gap does not. Both are indicative of the required adjustments to bring the educational system to its preconceived balanced pattern. All birr values are in constant prices of 2000. Bm = birr millions

PM

0.31 0.18 0.59 0.18 0.16 0.07

Public and private financing

3948 0.33 5250

9375 0.316 12337

2015

2005

All levels

Plan

Actual

270  Normed planning as long as the course of development links with the trajectory. Of course, the costs are less and the benefits are more the shorter is the transition period. Obviously, the indicated gaps are too big to be resolved in ten or 15 years. (This is not only because of the size of the gaps, but also because any decision to adjust the current policies may not take place before 2010 or 2015 given the already drawn-up plans and their implementation.) The obvious way out is to take the whole period of 2005 to 2030 as one of a gradual transition towards the roadmap destinations, which is the surest way of reaching the destination in time. The annual growth rate over the 25 years for each element of the roadmap is a helpful guide in applying the required adjustment. These growth rates are displayed in Figure 13.1 for the main variables. Annex Table 13.9, at the end of the chapter, gives a larger coverage of the transition details, and the growth paths of all variables in the model along the years 2015, 2020, 2025, and their destination in 2030. The transition paths shown in Annex Table 13.9 are flexible, and what cannot be bridged for some variables in one period can be done in more periods. This applies especially for shaded variables with annual growth rates of 10 per cent and above, that is, enrolments and teachers in secondary and tertiary education, and non-government expenditure on all three levels. Especially in this area incentives need to be instituted to mobilise greater efforts from the private sector in financing HRD. This approach is a most flexible one, and has the additional advantage of avoiding the simulation and appraisal of alternative strategies to bridge this or that gap. Such alternative strategies often reflect arbitrary choices, advisor preferences, and untested schedules which are not convincing to policy makers. Furthermore, the annex table specifies the target values of structural indicators that relate the various parts of the education sector to each other for various

0.0 Primary enrolments Primary teachers Primary rec exp

2.0

4.0

6.0

8.0

10.0

2.6 7.2 7.3

Secondary enrolments Secondary teachers Secondary rec exp

7.7

Gov total spending on education Non-gov total spending on education

10.3 12.0

Tertiary enrolments Tertiary teachers Tertiary rec exp Total recurrent expenditure Total capital expenditure

12.0 14.0

9.8 10.5

12.6

9.4 4.5 3.5 8.7

Figure 13.1 Ethiopia: anticipated annual growth rates of main variables of the education system 2005–30 according to the roadmap, percentages

Normed planning  271 periods until 2030. These structural indicators and their targets are especially suited to monitor the pace of implementation towards the roadmap destinations. 4.4  More roadmap specifications: technical, non-technical, gender and literacy The model contains additional equations that specify the HRD profile in more detail. For example, one equation gives the share of technical enrolments in total secondary enrolment. The equation is formulated as in the other regressed equations: the share is made dependent on GDPpc, and the data employed to estimate the equation are those of the select countries. Applying the estimates of the equation to the Ethiopian GDPpc gives the anticipated roadmap share of technical secondary enrolments for Ethiopia. From this share, and the anticipated roadmap target for total secondary enrolments, the latter is then disaggregated in its two components: general and technical. A similar equation has been estimated to split tertiary education into its two components of enrolments in natural sciences and in social sciences. The gaps between the actual distribution and the roadmap distribution are depicted in Figure 13.2. The gaps indicate that Ethiopia has a positive balance in technical secondary education of +16 per cent, and a negative balance in general secondary education of −59 per cent. The gap for total secondary education is −57 per cent. The situation for tertiary education shows about the same negative balance for natural and social sciences. To assess the anticipated gap in gender performance the actual versus predicted performance for the girl to boy ratio in primary and secondary enrolments, a regression was run to explain this ratio GBR in terms of the GDPpc. The data for the select group show that an equal gender ratio of 1 was realised at a GDPpc of $4000 $ppp. Observations falling below this point belonged to Egypt, India, Indonesia, Vietnam, and some early years in China.9 For Ethiopia, the required GBR is 77 per cent for 2005, which is exactly equal to the 77 per cent actually

Adult literacy Gender access Natural sc share in tertiary Social sc share in tertiary Total enrolment tertiary TVET share in secondary General share in secondary Total enrolment secondary –70%

–60%

–50%

–40%

–30%

–20%

–10%

0%

10%

20%

Figure 13.2 Ethiopia: roadmap gaps in various HRD positions, benchmark 2005

272  Normed planning reached in 2005. This is clearly shown in Figure 13.2. Eliminating the remaining gender gap as the GDPpc of Ethiopia approaches the $4000 level in 2030 is thus a target within very easy reach. The positive performance regarding gender stands in contrast to an excessive negative gap regarding adult literacy. An equation outlining the adult literacy ratio, ALR, as a long-term function of the GDPpc was tested. Application of the regressed equation10 to Ethiopia gave results implying that it would require that Ethiopia should have an ALR of 57 per cent in 2005, 61 per cent in 2010, and increasing to 65 per cent and to 88 per cent in 2015 and 2030, respectively. The situation is that the actual adult literacy ratio for 2005 is only 36 per cent against the required 57 per cent, see Figure 13.2. This is a gap of 21 per cent. Although the size of the literacy gap is moderate when compared to others, the required finance for its elimination surpasses all other gaps calculated so far. The gap is financially too huge to be closed by conventional literacy campaigns. The situation suggests a need to employ non-conventional policy approaches towards greater participation and integration of the least literate art of the population.

5  Additional refinements The quantitatively targeted structure of the educational system as was displayed above in the roadmap cannot be considered in isolation from the qualitative educational features that circumvent the educational system. These features are also broadly defined as the educational institutions along which the educational system is activated and runs. Accordingly, the policy model was supplemented with a separately implemented comparative descriptive review of qualitative HRD features of the educational system in the ten select countries that have been pivotal for their relatively high performance. As these qualitative considerations fall outside the focus of the book they are left out. Notwithstanding, it is important to acknowledge that one cannot take over and transplant the quantitative results from a normed planning model that is estimated from the database of select countries, and hope that these performances can be realised elsewhere, without giving due consideration to the qualitative foundations in which the quantitative results were formed. While the roadmap sketches and focuses on the national contours of HRD, actual implementation is for the greatest part always at the regional level. How prepared are the various regions in Ethiopia for adopting the roadmap? Which regions are furthest from the roadmap, and would thus require more attention and resources in the coming years? Since the policy modeller is also a policy advisor he or she should attempt to answer these questions which are central for policy making. Use is made of regional data on gross enrolment ratios (GER), and on student/ teacher ratios (STR), in both the primary and secondary levels of education, for specifying and quantifying a Regional Index of Educational Development. The two indicators and the Index shown in Table 13.7 are sufficiently broad to cover quantity and quality aspects of education. Of course, the education sector contains many other aspects relating to graduation, dropouts, efficiency, cost, returns, and differing stakeholder interests that cannot be dealt with at the regional level, often because of limited data. Fortunately, performances regarding (GER) and

Normed planning  273 (STR) tend to correlate with most of the other aspects, so that the index can give reasonable answers to the posed questions. When applied to Ethiopia the results show Addis Ababa to be at the top with an index value of about 0.9, followed by Harari and Gambella with index values of around 0.8. Next, there are Dire dawa and Bennishangul-gumuz at around 0.65, Tigray and Afar at around 0.55, and Amhara, Oromiya, and Snnpr at the lower end of 0.45, and Somali at only 0.19. The HRD status of the last four mentioned regions, namely Amhara, Oromiya, Snnpr and Somali, is at about half or less than that achieved in Addis Ababa, suggesting that major efforts are needed in these regions to bring them to par with the rest. Strengthened effort applies also to Tigray and Afar but to a lesser extent. Table 13.7 supplements regional performances with regional data on the share of government enrolment in total enrolment at the primary and secondary levels. This gives an idea on the self-help efforts and dependence on government versus non-government resources in the various regions. The high performances in Addis Ababa and Dire dawa are made possible by a greater role of the nongovernment sector, as compared to other regions. It is also worth noting that the higher development ranks of Harari, Gambella, and Bennishangul-gumuz correspond with the highest shares of government enrolments. On the other hand, Afar Table 13.7 Regional indicators (2006/07) and Regional Index of Educational Development Region

Gross enrolment ratio, GER Primary

Tigray Afar Amhara Oromiya Somali Bennishangulgumuz Snnpr Gambella Harari Addis Ababa Dire Dawa Total

Student/teacher Regional Regional Share government ratio, STR Index of rank enrolment in total Educational Secondary Primary Sec. DevelopPrimary Sec. per cent per cent ment.a

104.8  22.2  93.1  91.4  38.6 127.9

28.2  4.6 22.05 21.15  3.65 26.55

42 31 58 64 137 47

43 27 53 50 92 36

0.58 0.52 0.46 0.45 0.19 0.62

 6  7  8  9 11  5

96 76 95 98 96 99

 97  77  99  96 100 100

 97.8 181.4 116.8 146.6

18 28.8 51.6 63.9

68 35 30 28

53 31 31 31

0.44 0.78 0.81 0.92

10  3  2  1

96 97 91 46

 98 100  96  76

 80  91.7

40.05 21.4

33 59

36 48

0.67 0.47b

 4

78 95

 84  96

Notes a Index is obtained by converting STR to its inverse so as to work with increasing values as a development measure. Indicators of columns 1 to 4 are to be indexed in terms of the highest regional value, and averaged over the four columns. b Simple average of the column gives 0.59. Source: ESDP 2008 Annual Review Meeting.

274  Normed planning ranks as seventh out of 11 regions, in spite of a relatively lower involvement of the government sector. As far as the optimal mixture of government and nongovernment resources in education, as depicted by the roadmap, is concerned, it can be observed that Addis Ababa, Dire dawa and Afar are closest to the roadmap features.

6  Matching of the labour market in the short and medium terms While the roadmap is focused on reaching a balanced structure of HRD in the long term in conformity with selected criteria, it is not meant to answer questions on detailed supply and demand of the labour market in the short- and mediumterms. Such questions as how to tackle the problems of temporary and enduring unemployment11, how many technicians for the various trades should be trained, or how many engineers, physicians, managers, accountants, and the like should exit from tertiary education, in a couple of years or so, to meet market requirements are very important, but they are supposed to be resolved at more detailed policy levels than at the roadmap level, although within its boundaries. In a mature economy, the matching of demand and supply of HRD is resolved and decided upon by the stakeholders in the labour market. Stakeholders generate and use signals of labour market information, LMI, to help them in deciding where to enrol in education and training and where to apply for appropriate jobs that match their qualifications. In a developing country, as is very typical of Ethiopia, the labour market is embodied in barriers that prevent its fluid functioning. The barriers are fourfold; the labour market is highly segmented, labour market information is lagging, absorption of LMI by agents is minimal, and even if the above conditions are fulfilled, conditions for an elastic mobility of agents between the segments are usually unfavourable. Moreover, established mobility barriers between regions substantiate the segmentation tendencies further. On the other hand, the labour market matching tasks of the government in Ethiopia are facilitated by the fact that currently major shares of the employment of secondary and tertiary graduates are in government activities, while at the same time public education is the dominant supplier. With both ends in the same hands matching exercises by government are relatively simple for now. But, as the share of the private sector increases in both the employment and supply of education and training, the government’s ability in matching in the labour market gets more complex, and reliance on an LMI system becomes indispensable. In this regard, a few comments can be added on the state of the LMI system in Ethiopia The labour force survey (LFS) is the major source of LM information. The detailed elaboration of HRD programmes by manpower type requires much more data than those contained in LFS. Such programming is supposed to be done making use of (a) training needs assessments; (b) manpower forecasts; (c) analysis of educational returns; and (d) labour market signalling, within the broad destination structures that are depicted in the road map. Regarding (a) and (b), ministerial departments engaged in the employment and provision of training are well equipped to make training needs assessments at a meaningful and usable detailed level. But

Normed planning  275 little or nothing has been done in the area of skilled manpower forecasts in Ethiopia, though such forecasts face many weaknesses that limit their effective use in HRD programming. Regarding (c), there are calculations of educational returns to the three educational levels in Ethiopia. An elaborate review is found in a World Bank (2005). There are various studies that show the returns to primary education are lower than to secondary or higher, that is, 10.6 against 15.9 and 15.1, respectively, which is in variance with many developing countries. However, the results of the roadmap model which suggest an over-investment in primary education at the cost of the secondary and tertiary education tend to justify the empirical findings on low comparative returns to primary education in Ethiopia. Notwithstanding, calculations of educational returns can hide serious misunderstandings since the activity segment in which they were calculated conditions the educational returns, and there can be multiple segments. Besides, when positive externalities such as contribution to awareness and innovation are considered the returns to primary education would become higher. But that will also apply for the other levels if the analysis is extended. Moreover, the level of aggregation in calculations of educational returns is very aggregated and is not helpful for any detailed HRD programming for specific skills. Finally, with regard to (d), labour market signalling in forms of gauging the number of vacancies and rates of remuneration by skills is more useful for making decisions on education and training than manpower forecasts or educational returns. The Federal Civil Service (FCS) appears to have updated market information on vacant positions and remuneration rates for a large number of skilled occupations and this information is used flexibly by the FCS for scaling new recruits and upgrading the salary scales of skilled personnel in public sector.12

7  The roadmap as part of a sustained development trajectory: anticipated economy-wide imbalances HRD in general, including the education sector in particular, do not work in isolation from the rest of the economy, polity, and society. We pinpoint four crucial conditions in overall development that the select countries have fulfilled and which can serve as guidelines for a balanced path of socio-economic development for Ethiopia. These are (I) quick demographic transition, (II) speeded industrial transformation, (III) greater outward orientation, and (IV) restraints in income inequalities. Quick demographic transition: UN sources project that an annual population growth of Ethiopia would fall to 2.06 in 2015 and to 1.72 by 2030. These projections have been used in the model so far. How do these trends match with the past experience of the select countries? We tested a logarithmic function that gave significant results in which average annual population growth for five-year periods: 1975–80, and so on up to 2005–10 for the select countries which is made dependent on GDPpc for initial years of the five year periods, that is, 1975, and so on until 2005, see Annex Figure 13.4 at the end of the chapter. Periodical annual growth of population =   2.236 e −0.0946 Periodical GDPpc R2 = 0.78 (15)

276  Normed planning Application of the estimated equation to Ethiopia predicts an annual population growth in 2025–30 of about 1.6 percent. The UN projected path for Ethiopia is 1.72 per cent, which is fractionally higher than the select path. The difference is 1.72 − 1.6 per cent = 0.12 per cent, which is about 7 per cent more. It can be argued that some slight additional effort to restrain population growth may be required to bring population growth in Ethiopia into line with that in the select countries. Speeded industrial transformation: All select countries have passed, or are passing, through a swift industrialisation process in which the shares of the labour force and of the value added move from agriculture to the non-agriculture sectors. The share of the labour force in industry is the best indication of the concentration and interaction of agents, and the more agents there are in industrial activities the greater is the industrialisation process. Furthermore, the share of the value added that is generated in industrial activities is an expression of the extent of industrial transactions taking place and the pace of industrialisation. These two indicators are transformed into an industrialisation index by summing the two shares and regressing the outcome against the GDPpc for the select countries between 1980 and 2005. This gave the following well-performing estimated equation, Figure 13.4. Industrialisation Index = 14.644 ln GDPpc + 51.821

R2 = 0.9 (16)

The select countries already reached values of the index ranging between 97 and 72 in 2005. Applied to Ethiopia, the equation would recommend that Ethiopia should have an index value of 52, while the actual value is only 36 in 2005. The gap is about 30 per cent, which is in line with other gaps reviewed so far for 2005. The required value for 2030 is 72. The transition path from the actual 2005 to the required value for 2030 over a period of 25 years implies an annual change in the index of 3.6 per cent. This adjustment growth rate is lower than the adjustments required in mainstream educational variables, suggesting that it may be easier to cope with broader conditions of development sustainability than with education. Greater global orientation: In all select countries, and in the world at large, economies become more outward oriented with more economic development over time. Capturing this process in an analytically operational way is not easy. The indicators of foreign direct investment FDI as a percentage of GDP and foreign merchandise trade FMT as percentage of GDP are often used to indicate the degree of outward orientation of the economy. The regressing of the FDI/GDP indicator for the select countries did not deliver reportable results. The results for FMT/GDP were better but not fully satisfactory, see Figure 13.4. A tested equation gave the following rather weak results. (FMT/GDP) % = 23.65 + 2.3684 GDPpc R2 = 0.69 (17) The required value following the tested estimates should be 26 per cent in 2005, and rising to 33 per cent in 2030. Ethiopia is ahead of the norm path regarding outward orientation, having an actual FMT/GDP in 2005 of 44 per cent. This can be seen as a positive sign. But it can be also judged as an imbalance that feeds

Normed planning  277 and maintains other imbalances such as the backlog regarding income disparities discussed below. Given the weak estimates, more refined analysis may be needed to validate the above assertions. Restraints in income disparities: It is generally established that reductions in regional disparities and strengthened processes of national integration and political stability are essential for sustainable development. This is manifested in the select countries and the world at large. Taking the select countries as the norm, an equation was tested that regresses an equality indicator to GDP per capita. The equality indicator is the income share of the richest 20 per cent of the population divided by the income share of the poorest 20 per cent of the population. A value of 1 is total equality, higher value indicate degrees of inequality. The fitted equation is a sort of Kuznets curve. As generally acknowledged, it takes the shape of an S-curve: there is a tendency for inequality to increase during take-off, followed by more equality in the middle range, and more inequality in the highest range, see Figure 13.4. Equality index = 0.008 GDPpc3 − 0.327 GDPpc2   + 3.3891 GDPpc − 0.6896

R2 = 0. 61 (18)

The expected value of the equality index for Ethiopia in 2005 is 2.2, while the actual value for 2000 was 4.3, almost the double.13 The roadmap approach suggests that there are established grounds for reducing current inequalities should the profile of the select countries be consistently followed. Following the anticipated path of the select countries the implications for Ethiopia, drawn in Figure 13.3, indicate a combination of backlogs in the demographic transition, industrialisation pace, and income equality, with a better than the expected outward orientation of the economy. Though tentative, the results suggest that some adjustments may be required on all four fronts, and especially

50% 40% 30% 20% 10% 0% –10% –20% –30% –40% –50% Quick Speeded demographic industrial transition transformation imbalances

–7%

–31%

Greater global integration –41%

Restrained income disparities –49%

Figure 13.3 Ethiopia: anticipated imbalances economy wide elsewhere

278  Normed planning on industrial transformation and reduction of income disparities to resolve possible contradictions between them, and to lead the socio-economy along the norm path.

8  Summary and conclusions Lifting human resource development (HRD) in Ethiopia to appreciable levels comparable with middle-level countries would take more than a couple of decades. A roadmap model towards that end can only be specified and estimated by making judgmental use of the past experience of selected countries that are known to have performed better in HRD, and whose realised structure of the education sector can be projected backwards so as to apply to Ethiopia. Several selection criteria emphasising highest HRD performance produced ten select countries. A model was subsequently formulated for planning basic parameters of the educational system such as enrolments, teachers, recurrent and capital costs, and the break up of finance into government and private resources. This was done for the primary, secondary and tertiary levels, with some detail with regard to general and technical education. When applied to Ethiopia, the results show for benchmark year 2005 that the total expenditure on education as a proportion of the GDP is about 6.5 per cent for both the actual and expected, and is thus well balanced; nevertheless, significant shortfalls in the subdivisions of the educational system occur, reaching an average gap of 28 per cent from the roadmap norms. Ethiopia’s high population growth explains the gaps in part, but also the pursued policy for universal primary education and the very high unit costs at all levels in relative terms (read a too high salary per teacher in relation to GDP per capita) are major causes of the gaps. These obstacles cause imbalances elsewhere in the education system resulting in shortfalls towards reaching the enrolment norms in secondary and tertiary levels, but also teachers, The official plan for 2015 was simulated. If realised, the average summed gaps would fall significantly from the high level of 28 per cent in 2005 to 12 per cent in 2015. A simulation of a 10 per cent smaller population for 2015 diminishes the average summed gap to only 3 per cent. This is not feasible, however. A policy shift of focus to secondary and higher education and a relative reduction in the current cost per pupil are essential to get Ethiopia’s HRD into line with the roadmap. Other features of the roadmap model that can be highlighted are the ability of the model to stylise the optimal distribution of educational expenditure between government and non-government sources, for the whole sector as well as to the three levels separately, and the proposed flexibility for closing the gaps by taking the whole period of 2005 to 2030 as one of a gradual and flexible transition towards the roadmap destinations. Since ultimately any transformations along the lines of the roadmap have to take place in the regions, and regional conditions are known to vary significantly, it was necessary to formulate and apply a Regional Index of Educational Development to identify which regions are closest to the roadmap profile, and which are

Normed planning  279 most distant, and would thus require a greater effort to align them with the roadmap profile. Finally, the educational sector is part of a whole socio-politico economic system. There are qualitative and non-educational economic aspects that have been critical for the socio-economic development of the select group. Following the norm path of the select countries, Ethiopia is shown to have a combination of backlogs in the demographic transition, industrialisation pace, and income equality, with a better than expected outward orientation of the economy. Adjustments are required on most of these fronts to bring the country’s development in line with the profile of the select countries.

Annex Table 13.8 Estimated equations and detail assessment of estimated variables Eq.

Dependent variable

 (1) Gross enrolment ratio Primary Secondary Tertiary  (3) Student/teacher ratio Primary Secondary Tertiary  (5) Recurrent cost per student/ GDPpc Primary Secondary Tertiary  (7) Capital/recurrent cost ratio  (8) Share gov. spending all services  (9) Share gov. total exp. on education (11)

Share gov. recurrent exp. by level Primary Secondary Tertiary

Estimated equation GERe = α1,e + β1,e GDPpc GER1 = 77.5 + 0.0075 GDPpc GER2 = 44.4 + 19.47 ln GDPpc GER3 = 0.9 + 4.57 GDPpc STRe = α3,e + β3,e GDPpc STR1 = 29.7 − 6.16 ln GDPpc STR2 = 34.6 − 5.99 ln GDPpc STR3 = 27.3 − 6.10 ln GDPpc RCSe = α5,e + β5,e GDPpc RCS1 = 0.1029 + 0.0045 GDPpc RCS2 = 0.2481 + 0.0208 GDPpc RCS3 = 3.7005 GDPpc−0.7394 KRC = −0.0122 + 0.0509 GDPpc− 0.0024 GDPpc2 GTS/GDP = α8 + β8 GDPpc GTE/GTS = 15.15 − 1.2826 GDPpc + 0.1326 GDPpc2 − 0.0035 GDPpc3 GREe/ GRE = α11,e + β11,e GDPpc GRE1/ GRE = 0.3306 + 0.0087 GDPpc GRE2 / GRE = 0.2689 + 0.0088 GDPpc GRE3 / GRE = 0.1967 + 0.0101 GDPpc

0.91 17.448 40.34 432.5 0.14 259.7 4531 0.33 2.807 38.51 72.9 0.41 772.0 2167 0.06 0.450 27.82 16161 3.08 5747 2584

Secondary educational level Gross enrolment ratio Enrolments millions Student/teacher ratio Teachers, thousands Recurrent cost per student / GDPpc Recurrent cost per student birr const. prices Secondary recurrent expenditure Bm

Tertiary educational level Gross enrolment ratio Enrolments millions Student/teacher ratio Teachers, thousands Recurrent cost per student / GDPpc Recurrent cost per student birr const. prices Tertiary recurrent expenditure Bm

0.19 1.314 49 269.3 0.484 554 728

0.031 0.1739 36 0.0048 4.987 5705 992

2015

Primary educational level Gross enrolment ratio Enrolments millions Student/teacher ratio Teachers, thousands Recurrent cost per student / GDPpc Recurrent cost per student birr Primary recurrent expenditure Bm

Variable

0.86 12.66 62 203.0 0.154 176 2228

2005

Actual

Annex Table 13.9 Roadmap transition path

0.09 0.719 24.5 29350 2.4 5904 4245

0.423 4.089 34.21 119.5 0.38 933.2 3816

0.94 19.642 32.45 605.2 0.132 322.9 6343

2020

Roadmap transition path

0.13 1.141 21.6 52876 1.9 6210 7085

0.548 5.910 30.39 194.5 0.35 1155.0 6826

0.97 21.933 26.10 840.2 0.126 411.2 9019

2025

0.18 1.794 19 0.094 1.5 6705 12026

0.71 8.465 27 313.5 0.326 1457.2 12335

1 24.272 21 1155.8 0.120 536.4 13019

Destination 2030

continued overleaf

7.30 9.79 −2.51 12.61 −4.70 0.63 10.50

5.32 7.74 −2.34 10.32 −1.55 3.95 12.00

0.61 2.64 −4.26 7.21 −0.99 4.56 7.32

An. growth % 2005−30

0.0558 0.0104 0.0662 0.84 0.16

0.31 26211 0.18 4718 3539 0.59 0.18 0.16 0.07 2089 625 571 253 138 103 421 662 881 5599

3948 0.33 5250

2005

Actual

Financing by public and private sources Gov. total spending on all services/GDP Gov. total spending on all services Bm Share gov. exp. edu. in gov. total spending Gov total expenditure on education Bm Gov recurrent expenditure on levels Bm Share gov. recurrent exp. on primary educ. Share gov. recurrent exp. on secondary educ. Share gov. recurrent exp. on tertiary educ. Share gov. recurrent exp. others and overhead Gov. recurrent exp. on primary educ. Bm Gov. recurrent exp. on secondary educ. Bm Gov. recurrent expenditure on tertiary ed. Bm Gov. spending on others and overhead Bm Non-gov. recurrent exp. on primary ed. Bm Non-gov. recurrent exp. on secondary ed. Bm Non-gov. recurrent exp. on tertiary educ. Bm Non-gov. total recurrent exp. on educ. Bm Non-gov. capital, recurrent exp. on educ. Bm Non-gov. total expenditure on education Bm PM Government total exp. on education/GDP Non-gov. total exp. on education/GDP Total gov. and non-gov. exp. on educ./GDP Share education financing by government Share of education financing by private

All levels Total recurrent expenditure Capital/recurrent cost ratio Total recurrent and capital expenditure

Variable

0.0455 0.0215 0.0670 0.68 0.32

0.30 52436 0.15 8016 6477 0.48 0.22 0.16 0.14 3138 1412 1042 885 1393 755 1542 3019 3784 12859

9282 0.24 11488

2015

0.0411 0.0332 0.0743 0.55 0.45

0.291 74417 0.141 10485 8725 0.439 0.242 0.161 0.16 3829 2114 1401 1381 2515 1702 2844 7060 8484 18969

14404 0.202 17309

2020

Roadmap transition path

0.0371 0.0408 0.0779 0.48 0.52

0.286 107295 0.130 13931 11895 0.397 0.269 0.160 0.17 4728 3202 1908 2058 4291 3623 5177 13091 15332 29263

22929 0.171 26853

2025

0.0336 0.0487 0.0823 0.41 0.59

0.28 157010 0.120 18841 16455 0.36 0.30 0.16 0.18 5924 4937 2633 2962 7090 7394 9389 23872 27334 46175

37380 0.145 42800

Destination 2030

−0.41 5.17 −1.62 3.48 4.11 −1.96 2.13 −0.03 3.76 2.08 6.34 4.08 8.03 18.44 19.82 13.72 16.38 15.69 8.67

9.41 −3.23 8.75

An. growth % 2005−30

eq. 15 y = periodical annual population growth, %

eq. 16 y = industrialization index 100 90

2.5

y = 2.236e -0.0946x R 2 = 0.7768

2 1.5 1 0.5

Industrialisation index

annual population growth %

3

y = 14.644ln (x) + 51.821 R2= 0.8994

80 70 60 50 40 30

0 0

2

4

6

0

8 10 12 14 16 18 20 22 24 26

2

4

6

8 10 12 14 16 18 20 22 24 26 28

GDP per capita %ppp, 1000s

GDP per capita 1000 ppp$ Notes. Average annual population growth for 5-year periods: 1975-80, and so on up to 2005-10, is from UN http://www/un.org/esa/population/publications. These are regressed against GDPpc for initial years of the 5-year periods, i.e. 1975, and so on until 2005. Russia is excluded, having had a different demographic transition much earlier in the 20th century. The industrialisation index is the average of share of population in non-agriculture and share of GDP in non-agriculture.

eq. 17 y = foreign merchandise trade FMT/GDP, %

eq. 20 y = income shares richest 2% / poorest 20%

120

y = 2.3684 x +23.649

110 100

R2 = 0.6882

90 80 70 60 50 40 30 20 10 0 0

2

4

6

8 10 12 14 16 18 20 22 24 26 28 30

GDP per capita $ppp in 1000s

Income share richest 20% / poorest 20%

foreign trade merchandise/GDP, %

130 3

2

y = 0.008x − 0.327x + 3.3891x − 0.6896

13 12 11 10 9 8 7 6 5 4 3 2 1 0

2

R = 0.6057

0

2

4

6

8 10 12 14 16 18 20 22 24 26 28

GDP per capita $ppp, 1000s

Notes: FMT/GDP as % is for years in 1991–2005. Outliers outside range indicated by dashed lines are left out. For the income disparity indicator Brazil is left out as an outlier, having about the highest income inequality in the world.

Annex Figure 13.4 Ethiopia: regression diagrams for four equations (their application give Figure 13.3)

14 Labour market imbalances and adjustments Forecast model with RAS component

1 Background The origin of this chapter goes back to advisory work on labour markets and manpower planning in the 1980s in several countries.1 While these manpower planning assignments have had an impact on decisions of the responsible ministries in various ways, it became also increasingly recognised that it is very difficult to forecast the future in with regard to the labour market due to untraceable interactions that take place between ex ante states (expectations of skills demanders, that is, employing firms; and expectations of skills suppliers) and ex post states (realisations of employment and unemployment). The tasks are made more difficult due to the absence of an analytical framework that could separate the ex ante from the ex post state of the labour market; besides, there are no data for tracing the adjustment processes between these two states. The supplementation of forecasting models of labour demand and supply by subsequent modelling of adjustment processes and the clearance of labour markets presented themselves as challenges that require resolution, testing, and applied use. The challenges were relevant from the viewpoints of both academic inquiry and policy-making concerns. The objective of this chapter is to present a modelling framework that incorporates these challenges and remedies to them, show how these remedies work, and investigate results and implications in the context of three countries: Colombia, Korea, and Pakistan. The chapter is organised as follows. Section 2 will present ex ante forecasts of demand and supply of labour skills by sector of activity.2 Section 3 extends the forecasts of demand and supply to occupational qualifications and educational levels. Many details which are specific to the individual countries for which the modelling framework was developed are avoided on purpose in the current presentation, so as to keep to a minimum the number of generally applicable equations to all three countries3, and to assure that the same classifications of occupations and education are followed consistently. Section 4 departs from forecasts of imbalance and enters into a simulation of the labour adjustment process that ultimately results in resolving forecasted ex ante imbalances and clearance of the labour market. The approach followed is to specify additional equations that describe the adjustment processes between demanders and suppliers of occupations and education and the generation of the ex post values of employment and unemployment along the lines of bi-proportional fitting in statistics, known by

Labour market forecasts and adjustments  285 RAS method. This is an iterative algorithm for updating cell values of a prospective matrix on the basis of initial values and given totals. Section 5 validates the approach and demonstrates results of these applications. Section 6 presents a later development and refinement of the modelling framework: most labour market imbalances and their subsequent adjustments (expressed in numbers of workers) have their counterpart in earning imbalances and subsequent revisions in earnings expectations. The section explores the formulation of rates of return that diverge between employing firms and labour which offers the skills, whereby rates of return based on ‘job competition’ theory are seen to represent the expectations of firms, employers, and rates of return based on ‘human capital’ theory are seen to represent expectations of skill suppliers. Section 7 is an empirical application on these diverging returns. Section 8 includes concluding remarks.

2  Aggregate demand and supply at the sector level A main feature of the framework is the disaggregation of the economy into modern and non-modern sectors. In the modern sectors employers and employees work under relatively fixed labour contracts and working conditions, enterprises have access to credit and modern equipment, the firms are usually registered and regular statistics are collected on them. Labour can be assumed to work commercially at productive norms, which are technically prescribed. Underemployment can therefore be generally excluded. In the nonmodern sector, the primary mode of employment is that of the self-employed and family workers, where non-commercial motives are present and underemployment feasible. Although clear-cut criteria for dividing the economy into modern and non-modern sectors are usually lacking, a useful rule of thumb for demonstrative purposes is to take the non-modern sector as consisting of household production units with a majority of self-employed and family workers. This adds to considering most of agriculture as the non-modern sector and non-agriculture as the modern sector. Rules that clear quantities in the labour market differ for the two sectors. In the modern sectors employment is determined by the demand of employing enterprises, in the non-modern sectors employment is determined by the supply of the active population. In the non-modern sector the supply of labour (total labour force less open unemployment less modern employment) manages to find work in self-employment, as paid and unpaid apprentices or as family workers, so that supply equals employment. It is also well established that the (technical) demand for labour is below the (residual) supply of labour in this sector: supply is greater than technical demand. As a result, some underemployment is inevitable; this is measurable by employment less technical demand. Stable and low rates of open unemployment have been observed in most developing countries for many years. The approach followed assumes a stable and predictable open unemployment rate. For the purposes of policy making, the open unemployment rate can also be interpreted as a policy target. The above statements are expressed in eqs. 1 and 2. Eq. 1 determines employment in the modern sectors LDj making use of yearly labour input rates per unit

286  Labour market forecasts and adjustments of the gross value added of modern sector λj and planned or projected yearly gross domestic product by modern sectors Vj. Eq. 2 determines employment in the lump-summed non-modern sector j′, denoted by LDj′, as a residual: obtained by deducting the above mentioned sum of labour requirements in the modern sectors, ΣjLDj , and the given unemployment rate, LU, from the exogenously given supply of the labour force, LS. LDj = λ1 Vj

j (1)

LDj′ = (1 − LU) LS − Σj LDj (2) where LDj and LDj′ are employment in the modern and non-modern sectors, respectively. To facilitate comparisons across countries, relevant data for Colombia, Korea and Pakistan were reclassified and aggregated. For practical purposes, and because of a lack of directly comparable disaggregated data, the application takes the agricultural sector as a close approximation of the non-modern sector. The non-agricultural sectors constitute the modern sectors. Table 14.1 summarises the growth rates obtained for the above variables for the three countries for two ten-year forecast periods ending 1980 and 1990.

Table 14.1 Colombia, Korea, Pakistan: annual growth rates of GDP, employment, and labour productivity in modern and non-modern sectors, and labour force for the periods 1970 –80 and 1980 –90 Colombia

Korea

Pakistan

1970−80 1980−90 1970−80 1980−90 1970−80 1980−90 Modern sectors Value added, Σj Vj Employment, Σj LDj Labour productivity

 7.0  4.8  2.2

 6.7  5.5  1.2

8.5 4.9 3.6

14.0 5.7 8.3

5.7 1.9 3.8

7.7 3.4 4.3

 4.1  3.7  0.4  1.8

 4.0 −1.0  5.0 −3.0

4.6 2.6 2.0 1.6

7.6 2.0 5.6 2.6

2.6 2.1 0.5 3.2

4.9 2.5 2.4 1.9

Whole economy Total labour supply, LS  4.5 Unemployment rate LUa 14.0

 3.0  7.0

3.8 4.0

2.7 4.0

3.0 2.0

3.0 2.0

Non-modern sector: GDP − Σj Vj Employment LDj′ Labour productivity Added underemploymentb

Notes a Unemployment rate at end of period: 1980, 1990, in per cent. b Added underemployment in the non-modern sector is realised employment in the non-modern sector less demand for labour based on labour input rates that apply for the modern sectors.

Labour market forecasts and adjustments  287

3  Demand and supply by occupation and education: forecast model With the above segmentation of the labour economy settled the forecast model of demand and supply for labour focuses on the modern sectors. The forecast model contains six equations and makes use of two indexes: Index q is for the occupational qualification, distinguishing the following categories, basically along the International Standard Classification of Occupations lines: 1 = professional, technical, 2 = administrative and clerical, 3 = sales personnel, 4 = farm workers, 5 = production workers and 6 = service workers. Index e is for the educational classifications; following UNESCO practice, four levels are distinguished: 1 = higher, 2 = secondary, 3 = primary, and 4 = none. Eqs. 3 and 4 are demand equations. Eq. 3 converts employment demand by sector j, LDj , into employment demand by occupation q, LDq , via labour input coefficients λqj. LDq = Σj λ qj LDj

q (3)

Eq. 4 converts these occupational demands into educational demands. In doing this, use is made of distributive coefficients obtainable from cross tabulations by occupation and education, denoted by λeq,o. These are available with frequencies of several years as part of labour or population surveys. Since the coefficients of such a matrix are subject to changes and adjustments over time, as will be discussed later, the forecast is no more than an indication of the ‘hypothetical’ demand by educational level. Because educational demand is ‘hypothetical’ the variable is denoted by LDeΘ. Thus, eq. 4 assumes that the most recent matrix of e by q reflects in an initial way the conversion rates which business would use, ceteris paribus, to convert their occupational employment into a hypothetical demand by educational levels. LDeΘ = Σq λeq,o LDq

e (4)

where LDeΘ = hypothetical demand for labour by educational level e, λeq,o = proportion of labour by educational level e in occupational qualification, q, for the most recent period, o. Eqs. 5 and 6 are supply equations. Eq. 5 permits calculation of current enrolment in year course c, denoted by NTc, (in many countries, c amounts to 18 annual courses distributed evenly on primary, secondary, and higher education levels, with each level consisting of six annual courses). Current enrolment in year t in course c is the result of (a) flows of students from the previous stocks of enrolment NTc′,t − 1; these flows are affected by attrition due to death, as represented via coefficients σc′, repetitions via ρc′, and graduation via γc′, and are distributed over the various courses via δcc′; and (b) new entrants in the educational system, denoted by NEc. These usually take place in the first-year course in primary education and are dependent, among other things, on government enrolment policies and available finances. They are exogenously given. NTc,t = Σc′ δcc′ (1 − γc′ − ρc′ ) σc′ NTc′,t − 1 + NEc,t

c (5)

288  Labour market forecasts and adjustments where NTc = enrolment in course c, NEc = new entrants in course c, δcc′ = rates distributing promoted pupil from course c to course c′, always c > c′, γc = rate of graduating and leaving the educational system at the end of course c, ρc = rate of repetition applicable to course c, σc = rate of demographic survival (that is, 1 − rate of mortality) applicable to course c. In eq. 6, labour supply distinguished by educational level e, is a function of its lagged variable and the inflow of school leavers from the educational system that participates in the labour force. The inflow of school leavers are certified from course c and assigned as having attained educational level e via attainment coefficients αce . LSe,t = σe LSe,t − 1 + Σc αce πc γc NEc,t − 1

e = 1, 2, 3 (6)

where LSe = labour supply by educational level, πc = rate of participation of school leavers of course c in the labour force, αce = rates converting course c into educational level e. In eq. 7, labour supply with no education, indexed as e = 4, is obtainable as the difference between the just obtained labour supply by the other educational levels and the exogenous total labour supply. LS = Σe LSe (7) In eq. 8, as in eq. 4, a transposed matrix of q by e for a recent period o can be assumed to reflect, initially, the pattern of conversion which, ceteris paribus, guides the labour force by educational levels to reassemble into a hypothetical supply by occupational groups; hence the use of LSqΘ to denote such a hypothetical supply. LSqΘ = Σe λqe,o LSe

q (8)

where LSqΘ = hypothetical supply of labour by occupational qualification q, λqe,o = proportion of labour by occupation, q, in labour with education attainment, e, for the most recent period. The outcomes of eqs. 4 and 8 are independent forecasts of demand and supply by occupation and education. Table 14.2 shows the occupational imbalances forecasted, and Table 14.3 shows the educational imbalances forecasted. The results will be commented upon in the respective sections below. The produced forecasts in eqs. 4 and 8 cannot claim to be more than an initial round. The realisation of consecutive rounds is treated in the next section on labour market adjustment.

Labour market forecasts and adjustments  289

4  Labour market adjustment: RAS iterations A major bias of planning exercises in manpower and education is the assignment of a predominant role for policy makers in the balancing of supply and demand by occupation and education. The real situation is that the labour market has its own mechanism of adjustment which holds, irrespective of whether government intervenes or not. When there is potential for large discrepancies between demand and supply of manpower types it is inevitable that both demanders and suppliers of these manpower types adjust their positions and move towards a balanced situation; this may involve adjustment in other types as well, given the presence of substitutability between the types. In this section we incorporate the simulation of these interactions. One simple and validated way of incorporating a balancing mechanism via the labour market would be to allow the coefficient matrix of the occupational/ educational mix λeq,o (and λqe,o ) to become endogenous fully (or partly) in such a way that a realistic balance is reached between demand and supply of occupations and demand and supply of education. In fact, there are long and short-run labour market adjustments which shape this occupational/educational distribution cf. Zymelman (1980). Regarding the long run, adjustments towards more capital and skill intensive production processes are an inherent part of economic development and imply regular rises in the skill level of most occupations. Furthermore, younger members of the labour force have more formal education than older members because of the general tendency to increase years of compulsory education. As a result, there is a built in upward trend to shift the median years of formal education in occupations. Regarding the short run, in times of labour shortage employers accept workers with less formal education. But in the context of a labour surplus in a given occupation, employers are more strict in their recruitment criteria and demand more years of educational attainment and higher certificates. Thus, an occupational shortage tends to induce a downgrading of educational requirements for entering the concerned occupation, while a surplus tends to induce an upgrading. Seen from the point of view of the work force, a shortage of labour at a particular educational level often induces the respective graduate to join the labour force instead of taking up further education. A surplus of labour at a particular educational level tends to induce the respective school leaver to postpone entry into the labour force and instead continue further education. In theory, the differential between a desired wage in a desired job with a particular sector/occupation/educational specification and the actual wage in a directly available job with another sector/occupation/education specification plays a dominating role in the decision of the individual to take up an offered job or postpone decision. The modelling of the behavioural framework within which such decisions are taken is hardly possible in view of the multiplicity of options and the absence of crosstabulated data on wages by sector/occupation/education. (Sections 6 and 7 will explore some aspects of the difference between desired and actual wages.) As a result, simulation of labour market adjustment in a probabilistic manner, albeit mechanistic, makes more sense. This is outlined below in eqs. 9 to 13.

290  Labour market forecasts and adjustments Eq. 9 states that in the final analysis, the resolution of the occupation/educational imbalance can be described as a trial and error search by the suppliers of occupations q and demanders of education e, resulting in certain upgrading and downgrading. Both parties, starting from a (past) initial occupation/educational mix, as represented by λeq,o, would try to adjust the mix so as to exhaust their excesses and satisfy their shortages as implied by the hypothetical forecasts of demand and supply. The newly installed mix is an equilibrium mix and can be indicated by λeq,r. This new mix assures the clearance of the labour market subject to the overall open unemployment constraint. This trial and error adjustment process can be simulated by applying the RAS method. Although the RAS method was designed to update input-output tables, it can be considered as a general estimation aid for the unknown interior of matrix tables with an initial probability matrix as a point of departure and with prescribed column and row totals. Every column of the initial matrix is multiplied by a specific factor chosen such that the column total becomes equal to the prescribed total; the same is done for the rows. However, while normalising rows to their prescribed totals the column totals are disturbed again. Thus a new normalisation of columns will be necessary; and of rows, and of columns again, and so on. In the present context the application proceeded as follows. We want to predict an adjusted or a new occupational/education matrix λeq,r in 1980 based on the initial matrix λeq,o that belonged to 1970; the prescribed row totals are the predicted supply by education e, while the column totals are the predicted demand by occupation LDq plus a given provisional allowance for unemployment among q based on recent figures or eventually fixed as policy targets; so that the grand total of the predicted supply by education and the grand total of the predicted supply by occupation are made equal, to allow for application of the RAS method. The definite unemployment rates by q and e will be among the outcomes of the RAS application. λeq,r = function of [λeq,o , LSe , (LDq + unemployment q   )] (9) The RAS method in the present context can be explained in terms of a search for the appropriate row multiplicators RMe and column multiplicators CMq, as in eq. 10; so as to satisfy equalisation of forecasts of supply by e and of demand by q (plus provisional figures of unemployment by q), as in eq. 11. λeq,r = RMe λeq,o CMq (10) LSe = λ eq,r [LDq + unemploymentq ]

(11)

In the process of realising adjustment in the labour market, it is feasible that the assumed magnitudes in the demand and supply models be revised by employers and workers as part of the adjustment process. For instance, in the case of shortages employers may revise their plans and cut down the predetermined production levels or the predetermined unemployment level may undergo adjustment downwards. However, if the predetermined values for production and unemployment were correctly forecasted, their revision is, by definition, excluded. The ability of the model to produce realistic results is the ultimate test of its viability. Fortunately, the predicted occupational/educational matrix for Korea,

Labour market forecasts and adjustments  291 Colombia and Pakistan for 1980 corresponded very closely to the observed one for 1980, which encouraged us to proceed further with the analysis. Once found, the adapted λeq.r are utilised to convert demand LDq into new equilibrium values of realised employment LDe as in eq. 12. Application of a transposed matrix λqer to forecasts of LSe gives equilibrium values of the realised labour force, LSq, as in eq. 13. LDe = Σq λeq,r LDq (12) LSq = Σe λqe,r LSe

(13)

These solutions assure the following approximate balances: LDq ~ LSq and LDe ~ LSe. Relative discrepancies can be expected to be in the neighbourhood of the general unemployment rate, as will be seen from the applications in Tables 14.4 and 14.5. These results are discussed in the next sections. Note that the model can go straight to simulating labour market adjustment, instead of implementing the two stages of (first) solving for the forecasted imbalances (via eqs. 4 and 8) and (second) apply the RAS to simulate market clearance (via eqs. 9–13). Going straight is to skip eqs. 4 and 8. We follow here the phasing of the model in two stages, which is helpful in emphasising an expected ex ante state of imbalances based on diverging expectations of demanders and suppliers of the labour force, as distinguished from the observed ex post equilibrium state. The differences provide important information on the size and pattern of the necessary labour market adjustments, which policy makers can use in designing development strategy. Before reviewing results of the application it is worth noting that use of the RAS method in manpower forecasting is not new, but the form in which the method is applied here is new and economics-wise meaningful. Evans and Lindley (1973) used the RAS method in what amounts to generating new values of the education occupation cells after modifying the values of some specific cells. Their concern was to obtain statistical consistency. Our approach is different. We use the RAS method to simulate the interaction in the labour market between demanders of occupations and suppliers of education so as to clear the market.

5 Applications The model is calibrated and solved to give forecasts of demand supply imbalances for the two periods of 1980 and 1990. Required values for exogenous variables such as Vj are either plan targets or semi-official projections. LS, NE and unemployment estimates are either semi-official or own projections, depending on the country. Required estimates of the labour input coefficients λqj are mostly projections based on in depth analysis of various labour force surveys, conducted in the individual countries. Estimates of coefficients in eqs. 5 and 6, were collected from ministerial sources and applied for or around 1980. They were kept at their values of around 1980, for obtaining solutions for 1980 and 1990. Finally, as was stated earlier, the matrices giving λeq (and λqe) are from are the most recent

292  Labour market forecasts and adjustments cross-tabulation of occupation by education in the 1970s. These matrices were used in eqs. 4 and 8 to generate demand forecasts by education e and supply forecasts by occupation q, respectively. The same matrix of education by occupation is employed in eq. 9 but in a different context and for a different use. As eq. 9 explains, the matrix is fed with manpower forecasts of (educational) row totals and (occupational) column totals for 1980, and iterated in a number of runs to give RAS solutions for 1980. The RAS-ed matrix for 1980 was considered to represent the labour market adjustment processes that took place in 1980 to clear the labour market imbalances within given unemployment margins. The obtained RAS-ed matrix for 1980 was fed in turn with forecasted row and column totals for 1990 and subjected to several iteration rounds to give RAS solutions for 1990. In similarity with the above, the RAS-ed matrix for 1990 was considered to represent the labour market adjustment processes that took place in 1990 to clear the labour market imbalances within given unemployment margins. Simulation of the labour market adjustment (LMA) in this manner, and along the lines of eq. 9, needs to be validated. This behaviour can be validated by comparing the RAS-ed λeq.n cells for 1980 with the observed data of λeq,o for 1980, where available. Such a validation was feasible for all three countries. as their statistical sources provided the matrix for both 1970 and 1980. The RAS-ed matrix of 1970 with the row and column totals for 1980 year showed general correspondences with the observed matrix for 1980. There were other indications that the results for 1980 and 1990 are also generally valid. For instance, the RAS-ed adjustments were found to be concentrated in similar cell clusters in all the three countries. The pattern of adjustment is highly consistent among the countries and over time. The size of the adjustment in the cell clusters with the highest changes rarely exceeded 15 per cent over 10 years. Altogether, the simulated results indicate the use of the RAS to simulate adjustment of the labour market is operational and meaningful. Table 14.2 displays forecasts of demand supply imbalance by occupation, and shows how these imbalances are cleared in the process of LMA, which is approximated by the RAS model. With little exception, the imbalances forecasts across countries show growth in supply to surpass growth in demand for the groups of professional/technical, administrative/clerical and sales personnel. In contrast, the primarily manual groups including farm, production and service workers experience a somewhat higher or equal growth in demand than in supply. Defining the surplus occupational imbalance rate as (LSΘq − LDq)/LSΘq, third column, Table 14.2, the results show that the surplus imbalance rates are highest for professionals, with the exception of Pakistan4. The surplus imbalance can be interpreted as a ‘preference’ of the labour force for non-manual jobs as opposed to a ‘preference’ of the employing economy for manual jobs. The occupational imbalances are forecasted to be generally higher for 1990 than for 1980. Application of the RAS model, eqs. 9 to 13, ends up in adjusting the positions of demanders and suppliers to each other. As job preferences and searches by the employing business and the labour force iterate to each other in the adjustment processes, they converge towards definite unemployment rates that are lower than the imbalance

Labour market forecasts and adjustments  293 rates, and also vary within a narrower range than the imbalance rates. The unemployment rates are defined as (LSq − LDq ) / LSq in the fourth column in Table 14.2. Table 14.3 presents corresponding forecasts of demand supply imbalances by education and their subsequent clearance via the LMA-RAS model. The forecasts indicate higher growth rates in demand for and supply of manpower with more advanced educational levels as compared to those with less education. Furthermore, educational imbalance rates LSΘe − LDe ) / LSΘ, generally show surpluses for manpower with higher and secondary education (again, with the exception of Pakistan in 1980). The imbalance rates for higher and secondary education are generally higher in the 1990s than in the 1980s, which is evidence of a heightened gap between the economy and education. Application of RAS method shows outcomes for unemployment rates by education, (LSe − LDe ) / LSe at lower levels, and diverge less than the ex ante imbalance rates. Higher unemployment rates are still predicted for manpower with advanced as compared to lower educational levels. Comparative results in the two tables support the following. In economies that develop slowly, demand for and supply of labour tend to keep track of each other. Imbalance rates are close to unemployment rates and there is a diminished role for labour market adjustments. In contrast, a more rapidly developing economy, that is, Korea is characterised by major structural changes which are manifested, among others, in vigorous adaptations of labour demand and supply to each other, with regard to both occupations and education. The more rapid the growth of the economy, the greater the gap between (i) forecasted imbalance rates and (ii) realised unemployment rates, and the more intensive are labour market adjustments. These tendencies are apparent in the results, showing the more rapidly growing Korea to have greater imbalances and greater LMA than Colombia or Pakistan. Further analysis of the LMA results can be done making use of Theil’s coefficient of inequality, which provides a systematic measure of the intensity of LMA. Theil’s coefficient would relate (i) to (ii): the higher the value of the coefficient the more intensive is the LMA. Table 14.4 gives Theil’s coefficient by occupational LMA and educational LMA, across countries and periods. There are two types of conclusions. First, more rapidly changing economies show higher intensities of LMA. Across countries, the Republic of Korea has highest growth rates of value-added and labour productivity. Korea also has the highest values of Theil’s coefficient. Over years, the 1980s, which have higher growth and productivity, are predicted to have greater labour market adjustments than the 1970s. In general, the quicker the pace of economic development the greater are the educational requirements of specific occupations and the less job-oriented is the educational system. Second, the value of Theil’s coefficient is lower for manpower by occupational groups than by educational levels, which indicates a less intensive LMA among occupations than among education. The reason is that imbalances in different occupations are often shifted to a particular education while imbalances in a particular education are often distributed over many occupations. For example, occupation q = 4 (farm workers) which is highly concentrated in education e = 4 (no education), as against the same level of education e = 4 (no education), which is spread over more occupations q = 1,…,6.

Korea

 5.6  5.6  5.0 −0.8  5.1  5.6

1980−90 Prof/technical Admin./clerical Sales personnel Farm workers Production worker Service workers

 8.3  5.8  5.8 −2.5  4.0  3.6

 9.6  6.2  1.1  1.6  2.0  8.7

 42  17  17 −10   1   −6

 25  16  14  14  11  13  6  6  6  8  7  7

12 13 13 14 14 14  4.6  5.2  4.5 −2.0  6.6  4.6

 3.8  5.8  2.0  2.0  5.4  2.4  7.0  6.5  4.0 −2.7  5.8  3.9

 6.5  9.3  5.1  0.7  6.0  2.8

Note: Imbalance rates = (LSΘ q − LDq ) / LSΘq. Unemployment rates= (LSq − LDq ) / LSq

 7.9  6.0  6.0  1.3  2.2  8.4

1970−80 Prof./technical Admin./clerical Sales personnel Farm workers Prod’n worker Service workers

Annual growth Imbalance Unemploy- Annual growth ment LSΘq LDq LSΘq LDq

Colombia

 41  34   6 −26   1   0

 26  24  10 −10   9   8 5 5 4 3 4 4

5 5 4 3 4 4 3.0 3.3 3.9 2.5 3.4 3.2

4.4 6.3 2.3 2.3 5.2 6.1

4.5 5.0 4.1 2.2 3.4 3.2

6.0 4.7 2.2 2.6 5.2 6.2

Imbalance Unemploy- Annual growth ment LDq LSqΘ

Pakistan

 17  20   5   −2   1   1

−14 −14   1   4   3   3

5 5 2 1 2 2

4 5 2 1 1 1

Imbalance Unemployment

Table 14.2 Occupations: forecasts of periodical growth rates per annum of demand and supply, imbalance rates, and unemployment rates after labour market adjustment for 1980, 1990.

5.4 5.3 3.7 2.4

1980−90 1 Higher 2 Secondary 3 Primary 4 None

12.3 5.5 2.3 −2.0

8.1 5.0 2.9 3.0

LSe Θ

68 22 −12 −11

38 20 2 31 8 8 6 5

12 12 15 14 4.9 4.5 2.4 −0.2

5.3 5.7 1.5 3.4 9.1 4.2 0.4 −1.2

8.5 9.3 0.3 0.4

Imbalance Unemploy- Annual growth ment LDe LSe Θ

Korea

Note: Imbalance rates = (LSΘe − LDe ) / LSΘe ; Unemployment rates = (LSe − LDe ) / LSe

4.2 4.0 4.1 0.6

1970−80 1 Higher 2 Secondary 3 Primary 4 None

LDe

Annual growth

Colombia

52 24 −31 −52

29 27 −8 −36 5 4 3 2

5 5 3 2

3.1 3.2 3.2 2.9

7.7 5.4 5.0 1.8

4.6 5.4 5.0 2.3

4.8 3.4 5.0 2.3

Imbalance Unemploy- Annual growth ment LDe LSe Θ

Pakistan

19 24 18 −5

−34 −21 −1 5

6 7 3 1

5 6 3 1

Imbalance Unemployment

Table 14.3 Education: forecasts of periodical growth rates per annum of demand and supply, imbalance rates, and unemployment rates after labour market adjustment for 1980, 1990

296  Labour market forecasts and adjustments In retrospect, as more standard certification for specific jobs takes place over the years, specific occupations and a specific education can be expected to associate more exclusively with each other. As a result, the intensity of LMA among occupations and education would tend to converge with development. This is also borne out by Theil’s coefficient which shows on average a value for educational LMA which is 2.7 times the occupational LMA in 1980, as compared to the lower value of 1.7 times in 1990. Table 14.4 Intensity of forecasted labour market adjustment as measured by Theil’s coefficient

1  Occupational LMA 2  Educational LMA 3  Educational LMA/occupational LMA

Colombia

Republic of Korea

Pakistan

1980

1990

1980

1990

1980

1990

0.4 1.3 3.3

2.4 4.5 1.9

2.9 6.0 2.1

 5.0 10.4  2.1

2.0 3.9 2.0

2.4 2.7 1.1

Analysis of the pattern of adjustment is facilitated by examining the row multiplicators and column multiplicators, RM and CM, implied in the conversion5 of λo into λr. Row multiplicators give the factors by which λεq,o should be appropriately multiplied to give λεq,r. RM represent adjustments which the employer undertakes to restructure the educational composition of his occupational requirements to meet long-term technological trends and future restrictions on the labour market. Table 14.5 gives a summary of results for row and column multiplicators. For 1970–1980, last two columns, the average RM for higher education is 1.28, meaning that employers upgrade their requirements for higher education by 28 per cent in less than one decade. The upgrading is less for required manpower with secondary education, i.e. an upgrading of 5 per cent during the period (RM2 = 1.05). This upgrading happens at the cost of reduced requirements for primary education, less 13 per cent, and no education, less 2 per cent (RM3 = 0.87 and RM4 = 0.98). The average row multiplicators are generally higher for 1980–1990 than for 1970–1980, except for manpower with no education for whom requirements are forecasted to be squeezed appreciably. In figures, RM for 1980–1990 should be 1.95, 1.24, 0.97 and 0.77 for the four educational levels respectively. On an annual basis an upgrading is predicted for employment of manpower with higher education by 6.9 per cent, an upgrading of secondary education by 2.2 per cent, a slight downgrading of primary education by −0.3 per cent and an appreciable downgrading of no education by −2.6 per cent. These shifts are associated with an average annual growth rate of GDP in the non-agricultural sectors of 9.5 per cent in the 1980s. Predicted elasticities of the educational LMA with respect to GDP in the 1980s suggest the following values for the four educational levels respectively: 0.7, 0.2, −0.03, and −0.3. At an average annual growth rate of labour productivity in the non-agricultural sector of 4.6 per cent, predicted elasticities

Labour market forecasts and adjustments  297 Table 14.5 Row and column multiplicators for 1970 –80 and 1980 –90 Colombia

Korea

Pakistan

Average

1970–80 1980–90 1970–80 1980–90 1970–80 1980–90 1970–80 1980–90 Row multiplicators 1 Higher 1.60 2 Secondary 1.16 3 Primary 0.85 4 None 1.23 Col. multiplicators 1 Prof./technical 0.96 2 Admin./clerical 0.95 3  Sales personnel 0.95 4  Farm workers 1.02 5 Production 1.21   workers 6  Service workers 1.76

2.60 1.24 0.88 0.57

1.72 1.39 0.83 0.63

1.79 1.01 0.81 0.83

0.53 0.62 0.94 1.07

1.47 1.48 1.22 0.91

1.28 1.05 0.87 0.98

1.95 1.24 0.97 0.77

0.64 0.84 0.84 1.29 1.09

0.68 0.74 0.96 1.22 1.02

0.70 0.79 1.02 1.16 1.10

1.42 1.47 1.02 0.95 1.00

0.83 0.82 0.97 1.06 1.04

1.02 1.05 0.98 1.06 1.08

0.72 0.82 0.94 1.17 1.08

1.51

0.99

1.07

1.03

1.08

1.26

1.22

of the educational LMA with respect to labour productivity are at about twice the above values. Regarding column multiplicators, their lower values imply that the occupational LMA is milder than the educational LMA. Nevertheless, the prediction is that for the 1980s the labour force may be forced, or ready, to solicit less for nonmanual jobs, as apparent from CM1, for the professional/technical group = 0.72. LMA for administrative/clerical and for sales personnel shows similar but less pronounced trends, CM2 = 0.82 and CM3 = 0.94. Opposite behaviour is expected from the labour force with regard to the primarily manual occupations. More of the labour force is to be channelled into the occupational groups of farm, production and service workers, with CM > 1.0. The policy implications of the above trends for general education and vocational training are obvious. More specifically, as an aid for policy making at the individual country level, one may examine the predicted adjustments to be brought about in the individual elements for λeq,r (and λqe,r ). Table 14.6 lists the largest adjustments in λqe,r for Colombia, Korea and Pakistan for 1980–1990. Each change in the educational row e is neutralised by compensating changes in the nearest educational row and within the same occupational column q. The table predicts systematic shifts in the educational/occupational mix which three countries at different levels of development are likely to undergo in the 1980s. The detailed results on which the table is based constitute essential information for an effective planning of the development and utilisation of human resources. As stated earlier, governments have instruments they can use in controlling parameters of demand and supply. But once the future occupational demand and the future educational supply positions are set, anticipated changes in λeq would occur as a result of the free play of market forces. In principle, government policy

298  Labour market forecasts and adjustments Table 14.6 Predicted major adjustments in the educational/occupational mix for 1980 – 90 Colombia (C), Korea (K), Pakistan (P) Education

Occupation 1 Professional/ technical

1 Higher C+, K+ 2 Secondary P+ 3 Primary 4 None

2 Admin./ clerical

3 Sales

C+, K+ P+

C+, K+

4 Farm worker

5 6 Production Service worker worker

K+ P−

C−, P−

C+, K− P−

K− C−, P−

has very little control on these market forces. This is generally valid even though the anticipated changes may take place more smoothly if government encourages dissemination of information on the future demand and supply conditions and encourages or discourages upward and downward wage revisions by occupation and education, as the case may be. The more significant use of the model lies in the fact that knowledge by policy makers of the nature of anticipated changes in λeq is essential for an effective use of the instruments which policy makers can use in setting the right demand and supply positions in the future. Effective planning of manpower demand and supply requires knowledge of how labour markets adjust to overlapping between supply and demand.

6  Earnings imbalances: human capital versus job competition We dealt so far exclusively with LMA in terms of quantities of labour, but any allocation of quantities implies a certain set of prices. The shift from matrix λeq,o to λqe,r implies changes in remuneration patterns by occupations and educations. Given rigidities in wage changes for certain occupations and educations, the simulation of the adjustment process is supposed to observe such restrictions. The model by itself treats only labour adjustment and cannot predict the adjustment in earnings which accompanies labour reallocation. To do that requires the estimation of earning functions by occupation and education. In this section an exploratory attempt is made to connect earnings imbalances with quantity imbalances in labour market adjustments. To generate imbalances in earnings expectations, we need to have two contrasting models of earnings; this is more or less in similarity with forecasting imbalances in labour quantities. The two opposite earnings models applicable here are based on human capital theory and job competition theory. Conventional methods for estimating returns to education rely on human capital theory, which equates earnings to the marginal productivity of the worker and explains the latter in terms of the education attained by the worker. Job competition theory, which is equally legitimate in many contexts, downgrades the impact of education and assigns, instead, a primary

Labour market forecasts and adjustments  299 role for occupations in determining wages. Job competition theory asserts that it is the marginal productivity of the job, or the occupation, that determines the wage rate to which a worker will be matched, as was first formulated by Thurow (1969). Wages are paid on the basis of the characteristics of a job or an occupation. Occupations differ in their intensities of using capital, handling information, and practicing leadership. More demanding occupations are paid higher wages. Productivity is considered to be an attribute of occupation. In the job competition model workers are matched to occupations by specific worker characteristics and job requirements. While in the human capital model education has a direct link with productivity and wage, in the job competition model the role of education in influencing wages is indirect and secondary, namely via occupational upgrading. More education gives access to better occupations with higher pay. Investment by students in schooling, and their expected wages, are linked to each other in the job competition model by the expectations of an occupational upgrading, and hence, a higher occupational wage. Employers undergo significant transaction costs in matching the right selection of persons to jobs. Educational certificates can serve as a screening device in the selection process, and can save transaction costs. Entrants to the labour force, aware of employers’ behaviour, expand on and use higher educational certificates as a signalling device. This often results in an inflation of certificates of higher education. In human capital theory the basis for calculating the rate of return to investment in educational level e is estimated directly from earning functions or indirectly by employing age-earnings profiles by educational levels to estimate future benefits and finding the discount rate that equates costs to benefits. Under simplified assumptions that all earnings can be explained by education, a constant stream of benefits, foregone earnings incurred in the first year, and no schooling costs, the rate of return is obtained as a benefit cost ratio, eq. 14. (LRe − LRe − 1 ) / (LRe − 1,o . ne )= RHCe (14) where RHCe = rate of return to educational level e following human capital theory, LRe = annual average earnings of workers with completed educational level e, LRe − 1,o = initial average earnings of workers who forego educational level e, or simply starting wage ne = foregone years of earning during education e, or simply duration of education e. In job competition theory, the impact of additional education on enhanced earnings is found by considering that more education to give access to an upgraded occupational mix with higher labour productivities and earnings. Returns to education in this case are captured by the internal rate of return RTCe, in eq.15. The left-hand terms give the discounted net benefits over the working years, t = 1…w. The right-hand terms give discounted foregone earnings during the educational course, t = 1…n. The discounting periods are different in the two terms. The equation allows for variable occupational wages over time and for changing compositions over time of the occupational and educational matrix of

300  Labour market forecasts and adjustments which λqe are elements. Changing compositions over time in the present context are equivalent to occupational mobility. In job competition theory, wages are coupled to occupations in a manner that reflects the relative productivities of different occupations. More productive occupations are remunerated with higher wages. Wage formation is assumed to be more dependent on the specifics of the job and less on the personal characteristics of the particular job holder. Σ t =1.w [ Σq (λqe LRq − λqe − 1 LRq ) (1+ RJCe )−t] = Σt = 1…n [( λqe − 1 LRq,o ) (1+ e RJCe )−t] (15) where RJCe = rate of return to educational level e following the job competition theory, LRq = annual average earnings of occupation q, LRq,o = initial average earnings of occupation q or simply starting wage, λqe = proportion of the labour force with education e in occupation q, ne = duration of education e w = working years after completing education e, As was stated earlier, in many countries there are cross-tabulations of occupational and educational distributions q × e that can be used in combinations with earnings data to apply the above rates of returns. In the long run, there are persistent substitution tendencies towards more capital and skill-intensive production processes. These tendencies are part of economic development and imply regular rise in the skill level of most occupations. In the shorter run, and in times of labour shortages, employers will accept people with less formal education. In times of labour surplus, employers become stricter in their recruitment criteria and may demand higher certificates. The longer-run tendencies are more relevant for our purpose.

7 Application We apply both approaches to one of the three countries studied, namely Pakistan, under simplified assumptions to give the rates of return to education under both models. We let RHCe and RJCe , represent the rates of return under the human capital model and the job competition model, respectively, with e denoting the educational level, that is, primary, secondary, and higher levels. The rate of return RHCe is calculated from earning functions for the different educational levels. In contrast, the rate of return RJCe is based on wage functions by occupations and cross-matrices that convert occupational to educational mixes. Remuneration, educational and occupational data of Pakistan are used for calculating the rates of return. The data are from the Labor Force Surveys and Household Income and Expenditure Surveys of Pakistan. These are supplemented by data on schooling costs from ministerial sources. Figure 14.1 and Table 14.7 show the rates of return based on the two approaches considering three educational levels: primary, secondary, and higher, and making use of the distinction between seven major occupational categories. Rates of

Labour market forecasts and adjustments  301 return based on human capital RHCe are consistent with and fall within the range of previously estimated rates via more elaborate methods for Pakistan, Khan and Irfan (1985). Comparative studies for developing countries usually find returns to primary education to edge above returns to secondary or higher education, cf. Psacharopoulos (1980). Pakistan’s results show generally lower rates of return, and a reverse ranking pattern, that favours secondary and higher education to primary education, being one of the few exceptions. Pakistan’s rates of return based on the two models differ in two aspects. First, RJCe are generally lower than RHCe. The average returns on all educational levels following RJCe are about 15 per cent lower. Results are also consistent with the expectation that the reduction in returns is greater for higher educational levels, and is moderate for lower educational levels. The human capital model emphasises the remuneration premium of additional education. The job competition model eliminates the premium and assigns all earnings to the job irrespective of the education of the employee. Consequently, the job competition approach tends to downgrade the returns to education across the board, and particularly for higher educational levels. As a result, there is a tendency that RHCe > RJCe for all e. Secondly, as reductions in the returns are more pronounced the higher the educational level, the job competition model gives substantive support to the hypothesis of decreasing returns to increased education. The model also suggests that greater returns to education would occur for the type of education that leads to greater upward occupational mobility, and especially towards the higher paid jobs. Primary education, being a crucial threshold for occupational mobility, scores greater returns in comparison to secondary or higher education.

Table 14.7 Pakistan: results of different frameworks for estimation of rates of return to education Rates of return

Primary

Secondary

Higher

RHCe RJCe

0.087 0.117

0.093 0.075

0.123 0.069

0.14 0.12 0.1 0.08 0.06 0.04 0.02 0

Pakistan Primary Secondary Higher

Human Capital

Job Competition

Figure 14.1 Pakistan: educational returns under human capital and job competition

302  Labour market forecasts and adjustments In how far do these earnings imbalance correspond with the quantity imbalances found earlier under LMA in Table 14.3? RHCe showing highest values for higher and secondary education fits with the LMA forecast for 1970–80 showing highest shortage imbalances for higher and secondary education. RJCe results are a reversal of rates of return favouring primary education over upper levels. This is consistent with a corresponding reversal of forecasted imbalances for 1980–90 which show a shortage imbalance for primary education and surpluses for upper education. It is encouraging to note that the results obtained from the quantity side reinforce those from the earnings side and are consistent with each other. Furthermore, the predictions of the combined results that for 1980–90 a reversal in relative returns to the advantage of less educated labour and to the disadvantage of the more educated happens to be in conformity with the changing scarcities as predicted by the manpower forecasts.

8  Concluding remarks An open ended forecast of demand for and supply of manpower by sector, occupation and education offers a more flexible tool in the planning of manpower and education than other approaches, i.e. manpower requirements. Besides simulating balancing policies by government, the open-ended framework can be extended to simulate adjustments in the labour market (LMA). Resolution of occupational and educational imbalance can be described as a trial and error search by the demanders and suppliers of educational skills and occupational titles. The RAS method is shown to be a helpful tool in simulating this adjustment process. In general, analysis is more realistic and useful in a framework that considers interactions in the labour market. Theil’s coefficient of inequality is helpful in measuring the intensity of LMA, while row and column multiplicators derived from application of the RAS method are effective means of reflecting on the pattern of LMA. The imbalances in terms of labour quantities have their counterpart in converging expectations on earnings. This is demonstrated by formulating and applying earnings expectations based on two contrasting expectations as viewed by suppliers and by employers. One is based on human capital theory, which tends to overestimate educational returns and underestimate occupational returns. The other is job competition theory, which predicts more returns for occupations and less for education. The results of the educational returns are consistent with the forecasted scarcities and surpluses in 1970–80, and their reversal in 1980–90.

15 Privatisation decisions during transition A CBA model applied to Poland

1 Introduction One of the salient features of the transition period from a planning-oriented economy to the market-oriented economy is the process of privatisation of state enterprises. There are many different opinions concerning how this process should be organised (Sarcevic (1992); Ajani et al. (1992); Lee and Nellis (1990)), but most of them agree on the point that national entrepreneurship in the transiting economies may not have enough resources and (or) experience to finance, and (or) lead the privatisation process; an attractive and realistic solution to the problem may have to involve foreign firms and foreign capital in managing or in taking over parts, or the whole, of the privatised company. Foreign firms are interested also in directly investing in transiting economies given the size of their purchasing power. The potential foreign investor would appraise his returns from investing in Poland, Hungary, or other transiting economies with returns from investing in their own established markets. Having found it is worthwhile to invest abroad, the foreign investor would select a destination based on a comparison of expected rates of return for investments in alternative activities in alternative transiting economies. The aim of this chapter is to develop a modelling framework for an ex ante valuation of alternative privatisation transactions from the points of view of the foreign buyer, say a multinational company, but also from the seller’s viewpoint, which is the government of the economy in transition. The modelling approach we follow can be described as a double cost-benefit analysis from the points of view of buyer and seller, giving thus two rates of return to a specific privatisation venture. The closer the gap between the two rates of return, the more likely it is that the transaction takes place. The outcomes will also show in which economic activities more privatisation can be expected to occur.1 What are the main features of the model and how is the chapter organised? The privatisation transaction for a particular industry is examined in terms of the transaction value for the foreign buyer, K F , and the transaction value for the seller, that is, the government KG . The margin for negotiation would depend on the relative bargaining powers of buyer and seller, reflecting demand and supply conditions in the world market for the industry in question. Section 2 will

304  Privatisation decisions summarise notations used in the model and Section 3 will specify the transaction values for buyer and seller. The model is based on the idea that the potential investor calculates for each sector (activity) the price of the transaction he is going to pay for access to the market of interest to him (including property rights of real assets). The price of the buyer is dependent on (a) benefits, that is, discounted present value of future sales, and (b) costs of additional investment to be made, and the desirability of meeting competitive returns on investment. Since the transaction may take place only if it satisfies both sides, a buyer’s offers are set against a seller’s values of transactions, which are based on government’s calculations of gains and losses from the privatisation decision. Section 4 will specify benefits, Section 5 specifies costs and Section 6 treats the impact on government revenue. The model will be demonstrated with parameters for several sectors of the Polish economy in Section 7. Empirical results are reported in terms of ranks assigned to these sectors by the buyer and the seller, indicating their preferences for buying and selling real assets located in different sectors, and values of several exogenous parameters managed by buyer and seller in order to ensure that the transaction takes place. Section 8 concludes.

2  Time horizon and notations used in the model The model is determinate with 17 equations and 17 unknowns. Table 15.1 summarises the notations employed in the model. The time horizon in the model considers several periods: the period of recession, transition, and privatisation which ends with a turning point, denoted by ^ , a period when the operation of the firm is normalised, indicated by T; and a longer term period T + n. These are worked out in Figure 15.1 and will be highlighted in the equations of the model.

ST So

Sa

KT Kd

Ka

K0 0

^

Kc + Kb

T

T+n

Figure 15.1 Main phases in a privatisation venture: ^ = turning point; K = capital; S = sales

Table 15.1 Notations Indices: o = period before privatisation; t = year; T = period of transition of several years during which operation of the firm is normalised, n = age of the oldest vintage operating in period o Endogenous variables Buyer’s valuation to seller’s valuation of the privatised activity, in short: buyer:seller ratio BS Scrapped part of the capital stock that is, inefficient vintages Ka Direct investments required to increase capacity output to produce expected sales in T, Kb that is, ST Investment required in delivering sectors Kc Replacement investments Kd Transaction value for the foreign buyer KF Transaction value for the seller government KG Capital stock in the privatised activity at end of transition period T KT Labour requirement for producing output So (or labour input for operating Ko) LDo LF, LI Fall in employment, and increase in employment End result in unemployment associated with privatisation transaction LU Profits P Capacity lost due to scrapping of inefficient vintages (= loss of sales occurring at the Sa beginning of the transition period St , ST, Expected level of sales in year t, and at the end of the transition period T, respectively Unemployment benefits paid by government for displaced employees of privatised UB activity Exogenous Capital stock book value at the beginning of the transition period, that is, base year o Vintage of capital goods installed in past period t and still in use in base year o Buyer’s desired rate of return on the privatisation transaction Seller’s (government) desired rate of return on the privatisation transaction Sales capacity in base year o Real wage rate

Ko Kt R F* RG* So W

Coefficients Annual rate of depreciation of real assets Coefficient of inefficiency, particular to the transferred assets Dummy variable defined on the basis of a scrapping rule before ‘technical death’

δ ε ϕ γ ι κt κ′ λt λ′ π ρ τ σo , σT σa ω

Long-term rate of growth of sales (represents demand for the produced good) Interest rate for discounting future benefits and costs Capital output ratio of a vintage Kt Coefficients of capital augmenting technical progress Labour capital intensity coefficient for a vintage Kt Coefficient of labour augmenting technical progress Profit rate, defined as profits divided by sales, thus P / S Revenue per unit of sales Profit tax rate on the privatised activity Market shares of concerned activity in base year o and period T Competitive increase in the market share due to introducing higher quality by the investor Unemployment benefit paid by government as proportion of the real wage rate

306  Privatisation decisions

3  Transaction values for foreign buyer and seller government Let us consider a (buyer) multinational firm that has decided to invest in Poland, and is searching for privatised companies with the highest returns. The privatised company would go through various time phases that are displayed in Figure 15.1, and the buyer would be interested in knowing the costs and benefits in each time phase. His ultimate aim is to calculate the transaction value of the privatisation venture from the foreign buyer’s point of view. First take the lower curve of cumulated investments costs, noted by K. At the beginning of the transition period, o, the capital stock has a book value of Ko , which, after scrapping a part of it, Ka , results in a productive stock of Ko − Ka. To this value, renovating investments, Kb , is added in order to increase the capacity output of privatised activity to produce sales expected at the end of transition period T. There is also a required investment of amount Kc in delivering sectors, which will bring the capital stock at the end of the transition period T to KT = Ko − Ka + Kb + Kc. Finally, after the transition period ends, the capital stock in the privatised activity KT is increased each year in small amounts of Kd sufficient to meet replacement purposes. Second, take the upper curve of sales revenue, denoted by S. The curve gives the sales revenue of the privatised activity. It is assumed that sales revenue falls from So at the beginning of the transition period by Sa then gradually recovers to reach, at the end of the transition period, the expected level of sales ST (produced by assets KT ). Then sales continue to grow in line with demand conditions (absorbing replacement investments Kd ). The profit curve P can be derived from the sales curve S via application of a profit rate π defined in terms of sales, thus P = πS. Both the investment costs and the profits are discountable at international interest rate ι. The buyer’s rate of return on the privatisation transaction, R F*, can now be specified as in eq. 1, where τ is the rate of profit tax on the privatised activity. In order to meet a given target R F*, the foreign buyer calculates the present value of K F from the given parameters of eq. 1. [Σt((1 − τ) P)t (1 + ι)–t] / [K F + Σt (Kb + Kc + Kd )t (1 + ι)−t] =                  benefits / costs =R F* (1) Next is to consider the transaction value from the seller government’s point of view, denoted by KG. A definition of a seller’s gain from the privatisation transaction is not straightforward, since the government, who is the seller, has to consider a lot more of the indirect implications of the decision than the buyer. These implications cover such aspects as future growth, state budget, balance of payments, unemployment, environment, and national wealth as well. To study these entire issues one should use a model of the national economy, designed for the purpose of analysing a privatisation processes. Instead, we shall consider a partial approach to approximate various elements of the seller’s gain from a privatisation transaction and compare them with the opportunity cost of foregoing the transaction.

Privatisation decisions  307 The privatisation transaction for the government consists in the conversion of real productive assets into financial assets. If the proper measure of real assets is their market value, the role of the government is to get in exchange for privatised assets their market value. The government would aim at value equivalency in negotiating the transaction, and thus minimise the reduction of national wealth. Because of the absence of capital markets in planning-oriented economies and, otherwise, perfect market conditions in the world at large, the government in such an economy usually does not know the market value of the candidate assets and the government’s estimate of their value may differ substantially from the one calculated by the foreign buyer. Usually the government’s value of privatised real assets is closer to the book value of these assets (or replacement value) than to the buyer’s evaluation based on a valuation of cumulated expected profits from the sales of output produced by these assets. Since the privatisation of an enterprise is very likely to increase unemployment, the government will reduce its offer price by minimum income transfers payable to the persons unemployed as a result of layoffs following a restructuring of the enterprise. Unemployment benefits are denoted by UB. But the act of privatisation, which is meant to cause an increase of economic efficiency, will help the government to finance such transfer payments, and usually the government can expect higher budget inflows from the taxation of a higher profit from the privatised enterprise than was the case without privatisation. On the basis of this reasoning one can define a government-based measure of the efficiency of a privatisation transaction by relating current and future discounted public financial flows to the value of real assets, which the government desires to sell. This measure is defined in eq. 2. [KG + Σt (τ P − UB)t (1 + ι)−t] / Ko = RG* (2) The transaction value for government, KG, can be found from this equation given a desirable value of RG*, which in turn is derived from the community’s valuation of the transition process and its patience towards achieving it. In the extreme case, if the government is very cautious in approaching the transition and is freely disposing of its public enterprises, the denominator of eq. 2 will be the book value of the public assets, and RG* will be set close to 1. At the other extreme, advocates of quick transitions will put RG* at a fraction of 1. Of course, the optimal value of RG* could be fixed at the ratio of market to book value, but as was stated earlier, the market value is difficult to determine in a situation where the number of privatisation transactions taking place is small and the contents of the transactions are not transparent. The situation changes as more transparent transactions take place leading to the formation of reasonable expectations on RG*. Agreeing on the terms of settlement is best described as a negotiation process aiming at reaching a transaction value for the buyer K F and for the seller KG which is close enough. The buyer/seller ratio, denoted by a BS ratio, can be simply expressed as in eq. 3. BS = K F / KG (3)

308  Privatisation decisions The BS ratio would reflect the tightness of the private foreign investment market. Values of the BS ratio of above 1 suggest a tight market where there is a greater demand than supply for privatised ventures. It is more likely that the opposite may prevail, as will also be shown in the empirical application of the model. In particular, stronger appetites by more Eastern European governments towards privatisation via the channels of private foreign investment will push BS downwards, which is comparable in the terminology of Bhagwati to a process of immiserisation, cf. Bhagwati and Brecher (1982). It can be mentioned too that there is likely to be a greater degree of coordination between candidate buyers than among governments of the selling Eastern European countries, which may result in a further deterioration of the value of the BS ratio to levels which shun the seller—that is, the government.

4  Expected sales and profits In determining profit expectations from a takeover, the potential investor will consider first of all the size of the market for his products—measured by the value of sales—which is expected at the end of the transition period, ST. To determine the value of expected sales the buyer takes into account an evaluation of the following factors: (i) transformation of the initial demand structure towards a demand structure typical for free market economies, which leads to a change of the market for the concerned good’s share from σo to σT at the end of the transition period; (ii) a factor σa representing a competitive increase in the market share due to introducing higher quality by the investor. The competitiveness factor σa is dependent on the scale of modernisation Kb + Kc; (iii) long-term growth rate of demand, γ; and (iv) length of the transition period, T. The value of expected sales at the end of the transition period, ST is expressed in eq. 4. ST = So (σT / σo ) (1 + σa ) (1 + γ)T (4) This same equation can be elaborated to give sales after the transition period if the sales are expected to continue their growth at the same long-term rate, γ, thus ST + t = (1 + γ)t ST For the transition period it is assumed that the value of sales falls at the beginning of the period from initial value So to So − Sa as a result of the necessary scrapping of a part of privatised real assets and a recessionary transition, and then gradually recovers to reach the expected value ST at the end of the transition period. St = [(ST / (S0 − Sa )1/ T]t (S0 − Sa )

t = 0, 1, 2 ,…, T − 1 (5)

Future profits are calculated on the basis of the expected value of sales, St. Pt = π ε St

t = 1, 2,…, T, T + 1,… (6)

where π = profit rate expressed in terms of sales, and ε = the coefficient of inefficiency, particular to the transferred assets. The current value of the cumulated discounted future profits determines the market value of the privatised productive

Privatisation decisions  309 real assets. This value is a sum of two geometric series. The first one concerns the transition period, the second is for the period after transition.2

5  Costs: direct, associated, and replacement investment costs As a result of the privatisation transaction the foreign buyer becomes the owner of a part of productive real assets located in a particular sector of activity. This gives him access to the particular market. His next step is to reorganise and invest, in order to create capacities sufficient to produce at the end of the transition period the value of sales ST. To avoid details concerning capacity utilisation and inventories we assume that sales = output = capacity output. We distinguished between three types of investments: Kb , Kc , and Kd , for respectively, direct investment outlays, Kb , necessary to renovate and generate capacity output, ST , associated investment, Kc , and replacement investment, Kd . Direct investment: In determining Kb, which denotes investment outlays necessary to generate capacity output ST in period T, we describe the production process in the privatised activity in terms of an empirically desirable vintage model of production. But first it is essential to make an inventory of the capital vintages that form the given capacity in the base period o. This is given and can be described as follows: S0 = (1/κ1 ) K1 + (1/κ2 ) K2 + … + (1/κn ) Kn ; where Kt is a vintage of capital goods installed in past period t and still in use in period o in the privatised activity; κt is the capital-output ratio of a vintage Kt, and n is the age of the oldest vintage operating in period o. Labour inputs that are associated with the operation of these capital vintages Kt and their corresponding capital-output ratios κt , are known data via labour-capital intensity coefficients, denoted by λt . (Note, this is an unnumbered equation, and is outside the model, as the parameters it contains are already known and constitute given data.) In the calibration runs, empirical values of Kt , κt , and λt for the past were estimated on the basis of 1989 data concerning the average capital-output ratio for state enterprises in the privatised activity; the empirical age structure of state-owned assets in the privatised activity, and the average labour-capital intensity coefficient for state-owned enterprises in the privatised activity; and experimentally chosen values of the capital-output ratio κ1 and labour capital intensity λ1 of the newest vintage in use before privatisation. A search procedure was applied to find values of capital-augmenting and labour-augmenting technical progress coefficients, κ′ and λ′, respectively, on the basis of the following relations: κt + 1 = κt (1 + κ') and λt + 1 = λt (1 + λ')

t = 2, 3,…, n − 1

The inefficient vintages that are scrapped, denoted by Ka, are specified in eq. 7. Ka = ϕ1 K1+ ϕ2 K2+…+ ϕn Kn (7) where ϕ is a dummy variable defined on the basis of a scrapping rule before ‘technical death’, which amounts to keeping a vintage of equipment in use as long as its quasi rent is positive, that is, above the wage rate that is denoted by W.

310  Privatisation decisions Equipment will remain in use, thus ϕt = 1, if ρ (κt / λt ) > W. Otherwise they will be scrapped, thus ϕt = 0. The quasi rent is formed by ρ (κt / λt ). We take here ρ for the revenue per unit of sales, and (κt / λt ) is in fact the quantity of production per unit of labour3 applicable to vintage t. The capacity lost due to the scrapping of inefficient vintages is defined by eq. 8. Sa = (ϕ1 /κ1 ) K1+(ϕ2 /κ2 )K2 + … +(ϕn /κn ) Kn

(8)

The value of direct investments needed and necessary to produce ST is denoted by Kb , and is specified in eq. 9, where κb is the capital-output ratio for a new vintage of capital goods Kb installed in period o to assure, after scrapping of inefficient vintages, a sufficient increase of production capacity required for reaching sales ST . Kb = κb [ST + Sa − S0 ] (9) Associated investment costs: Modernisation of equipment in privatised enterprise increases both capital and labour productivity but full utilisation of the modern capacity may be limited by an underdeveloped economic environment affecting the enterprise. In order to eliminate external limitations regarding full capacity utilisation in the privatised enterprise, direct investments in the privatised activity have to be followed by some additional infrastructural investment in delivering activities. For example, very modern equipment installed in a privatised car-producing factory has to be accompanied by additional capital investments in a factory producing modern tyres. As Kc is the associated investment in infrastructure necessary to assure production of ST + Sa − So , we have thus (ST + Sa – So ) = min [(1/κb ) Kb , (1/κc ) Kc ]. Since Kc should not be the limiting factor we have eq. 10, therefore. Kc = (κc / κb ) Kb (10) In analogy with the pattern of growth of sales, as in eq. 6, a uniform distribution of investment outlays Kb + Kc is adopted in the transition period. The yearly outlays are discounted by the international discount rate ι referred to earlier.4 Replacement investment costs: Capital stock in the privatised activity in period T is shown in eq. 11. KT = Ko − Ka + Kb + Kc

(11)

In order to continue the growth of sales after the transition period by the rate γ (see eq. 5), KT should grow each year by an amount sufficient to meet replacement purposes, this category of investment is labelled replacement investment and is denoted by Kd. The size of replacement investments depends on three factors: (i) the capitaloutput ratio of new vintages κ; (ii) the expected rate of growth of sales, γ, and (iii) the annual rate of depreciation of real assets, δ. If these factors are treated as constants then the necessary replacement investments in a year t is calculated from eq. 12. K d,t = (γ + δ) Kt − 1 (12)

Privatisation decisions  311 Note that when assuming that the capital stock required to maintain a γ per cent growth of sales will have to grow by the same rate γ, then the replacement investment for a year t after period T can be derived along eq. 12 and written as Kd,T + t = (γ + δ) (1 + γ)t − 1 KT. Finally, the total discounted value of replacement investments can be found along the lines of the endnote to eq. 6 discussed earlier.5

6  Impact on government revenue To compute the transaction value for the seller government, KG , in eq. 2, the values of innings from profit taxes, and unemployment benefits, UB, have to be ascertained. Net innings for the government from profit taxation are calculated as the difference between the budget incomes from profit taxes imposed on expected profits P, (eq. 8), and the incomes the government would have had if the activity was not privatised (status quo profits and innings). The status quo profits and innings are calculated using the following assumptions: (i) Sales grow by a rate γ' which is less than γ ; (ii) The share of profits in sales is lower than in the privatised activity; (iii) The profit tax is paid from profits diminished by reinvested profits; (iv) Profits are discounted at the rate ι; (v) Reinvested profits are recalculated, whereby the rate of growth of sales γ is replaced by a lower γ’ and the depreciation rate δ by a lower δ'. As regards the unemployment benefits, these are obtainable from five related equations specified below. In the first place, the total amount of labour required, denoted by LDo to produce output of the base year So consists of labour inputs allied to the vintage of past capital goods Kt in period o in the privatised activity; they are linked via labour capital-intensity coefficients tied to the capital vintages, denoted by λt. This gives eq. 13. LD0 = (1/λ1 ) K1 + (1/λ2 ) K2 + … + (1/λn ) Kn

(13)

The scrapping rules described earlier (see eqs. 7 and 8) lead to a fall in employment, denoted by LF, which can be specified as in eq. 14. In contrast, new investments Kb lead to an increase in employment, denoted by LI, depending on the labour capital intensity of the new vintage λb in eq. 15. The ultimate result in unemployment associated with a privatisation transaction amounts to LU, in eq. 16. The eventual number of unemployed persons LU is multiplied by a transfer payment per person unemployed, defined as a part ω of the wage rate W*, to give the cost of unemployment benefits, UB, associated with the privatisation transaction, as in eq. (17). LF = LD0 − [(1 − ϕ1 ) /λ1 ]K1 − [(1 − ϕ2 ) / λ2 ] K2–… − [(1 − ϕn ) / λn ] Kn (14) LI = (1 / λb ) (1/κb ) Kb

(15)

LU = LF − LI (16) UB = ωW. LU

(17)

312  Privatisation decisions

7  Empirical results The proposed model of a privatisation transaction was tested empirically on examples of five Polish industries: food processing industry, Food, light industry, Light, wood and paper industry, Wood, chemical industry, Chem, and electro-engineering industry, Elen. Table 15.2 contains the list of basic economic indicators which characterise these industries. Values of these indicators were derived from statistical data for 1989 and 1990 concerning the Polish state industry. The subject of the privatisation transaction is a hypothetical state enterprise with the level of sales in 1989 corresponding to 10 per cent of the market share of the sector. The value of sales, assets, and employment are found in Table 15.2. A few parameters are the same for all five industries. These are: • • • • • • • •

Expected length of transition period T is 5 years; The annual rate of economic growth γ is 2 per cent; The structure of the market is not changed in the transition period; Annual discount rate for future profits ι is 10 per cent; The annual real wage rate per employee in the privatised activities, denoted by W, is 5.174 million at the end of transition period, expressed in 1989 zlotys prices (equivalent to US dollars 545); Profit tax rate τ is 25 per cent; Proportion of unemployment benefits, ω, is equal to 60 per cent of the wage rate W. Exchange rate of the zloty to US dollars is 9500 zlotys per 1 dollar.

Table 15.3 reports solutions that were produced assuming a desirable rate of return for the foreign buyer, R F*, equal to 20 per cent of total discounted costs. The rate of return for the seller government was defined according to eq. 2, where the denominator is the book value of privatised assets. The desirable seller rate of return, RG*, was set at the level of 80 per cent of the book value of privatised assets. The analysis of the buyer and seller preferences is based on comparison of the rates of return, which would result if the offer from the other side were to be accepted. The results presented in Table 15.3 show the similarity of buyer and seller ranking of the assessed industries. The analysis of the feasibility of the transaction recalls the buyer/seller ratio, BS, in eq. 3. The BS ratio measures the relative discrepancy between the offers of the buyer and the seller. A buyer bidding at a value lower by more than 20 per cent of the offer of the seller indicates that the partners are not going to reach the agreement without deep revisions of their status. In such circumstances, only transactions concerning the chemical and the electro-engineering industries seem to be feasible. The last rows of the table present the results in a reversed form. For given K F* and KG*, the targeted rates of return for buyer and seller are reversed in status and become unknowns as R F and RG. The results suggest that the gaps between the expectations of buyer and seller are very wide for all sectors, though the ranking of the

Table 15.2 Economic and technical characteristics of selected Polish industries

Whole sectors Sales 1989 (US dollars m) Real assets end of 1989 (US dollars m) Employment (thousands) Selected samples Sales 1989 (US dollars m) Real assets end of 1989 (US dollars m) Employment (thousands)

Food

Light

Wood

2081.5 3256.4  410.6

1168.1 2146.7  633.9

 460.9  900.9 2526.9 1486.0 3769.0 8524.1  211.0  276.7 1300.7

 208.2  116.8  290.4  197.8   36.6   58.4

  46.1  141.4   20.1

Chem

Elen

  90.1  252.7  362.7  789.4   26.6  120.5

Table 15.3 Poland: main characteristics of the privatising enterprise and privatisation transactions Food Buyer calculations Capital-output ratio, κ Labour productivity in US$, λ Depreciation rate, δ Expected profit sales ratio, π Efficiency index, ε Sales growth, Ka /Ko Scrapping rate, (ST − So ) / So Renewal rate, (Kb + Kc ) (Ko − Ka) Discounted profits Transition value in US$ million, K F Seller calculations Ultimate change in unemployment in thousands, LU LU/employment, per cent Discounted gains from profits, US$m Transition value US$ million, KG Confrontation of buyer to seller Offer buyer/offer seller, KF/KG = BS ratio Buyer and seller ranking PM Endogenous R F for given KG (Buyer preference) Buyer ranking Endogenous RG for given K F (Seller preference) Seller ranking

Light

Wood

Chem

Elen

1.4 5686.0 0.12 15 0.7 13.6 0.35 0.68 593.4 114.9

1.7 1999.5 0.10 15 0.8 15.6 0.54 0.74 287.3 106.5

3.1 2294.6 0.07 15 0.8 15.9 0.38 0.61 166.4 13.9

4.0 3382.7 0.04 25 0.7 23.6 0.42 0.66 218.5 222.9

3.1 2097.6 0.08 25 0.7 16.8 0.36 0.64 73.2 468.8

0.6

7.2

1.3

0.2

3.0

1.6 38.5 194.1

12.3 23.0 138.1

6.4 13.1 100.5

0.7 38.0 252.3

2.5 90.3 542.4

0.59

0.77

0.14

0.88

0.86

4 18.6

3 19.0

5 15.2

1 19.3

2 19.4

4 52.7

3 64.0

5 18.8

2 71.9

1 70.7

4

3

5

1

2

314  Privatisation decisions sectors is remarkably consistent for buyer and seller. Furthermore, the results suggest that more privatisations can be expected in the sectors of chemicals and electro-engineering industries, where the gaps are smaller than in the other sectors. The process of renegotiating for the other three industries, which is not reported in the results, would mean that both sides would have to weaken their initial positions. For example, the seller may accept a lower rate of return than the 80 per cent from the solution. The buyer, in turn, may revise his evaluation and assign higher weights to immediate incurred costs than to replacement costs, which consist of about one half of total discounted costs, see Table 15.4.

Table 15.4 Poland: structure of discounted buyer costs, percentage

Transaction value Direct investments Associated investments Replacement investments Total costs

Food

Light

Wood

Chem

Elen

 10.7   8.2  20.6  60.5 100.0

 17.0  12.3  18.9  51.8 100.0

  5.1  17.9  19.9  57.1 100.0

 27.1  16.8  21.0  35.1 100.0

 21.1  12.5  17.8  48.6 100.0

In this study we assumed values of factor productivity coefficients at levels observed for 1989 because for this year the Central Statistical Office (CSO) provided data for fixed assets evaluated in the current prices of December 1989. Use of these data and calculations in constant prices of 1989 in our opinion permits us to omit the problem of an inflationary increase in the prices of capital goods. Results can be sensitive to capacity utilisation. At the present stage of our research, differences in the levels of capacity utilisation were ignored, which results in a bias of the capital-output coefficients. Given average factor productivities, such factors as the rates of capital and labour-augmenting technical change, which are crucial for the determination of scrapping and the level of direct necessary investments, are important in influencing results. Some arbitrary values were fixed for factor productivity in the newest vintage of assets as well. Though some arbitrariness in fixing parameter values is unavoidable in such a model as followed here, their reliability can be evaluated by simulation experiments designed for checking the sensitivity of the solution to parameter changes. Such an evaluation produced stable results.

8  Concluding remarks The model formulated conditions under which privatisation between a seller government and a foreign buyer takes place making use of a two-sided cost-benefit

Privatisation decisions  315 analysis by seller and buyer and considering the typical features of the transitional phase in which the privatisation is supposed to take place. An empirical application to five industries in Poland showed that the chemical and electro-engineering industries are better situated than other industries such as light manufacturing, wood and paper manufacturing, and food processing for privatisation transactions.

16 Economic policy solutions to social queuing problems A random sampling model

1 Background Problems of social queues and their resolution are typical occurrences in the use of social services. For example, for various reasons there is a tendency for the social demand for health services to exceed the effective supply of medical facilities. This gap is often manifested in the formation of queues of patients waiting for surgical treatment. Although the generally publicised rule that applies to resolving the queues is that of ‘first come first treated’, the physician may decide otherwise based on his or her own personal judgement. The social queues can be observed in other areas as well: specific education, subsidised housing, unemployment retraining, and who to serve first in poverty alleviation campaigns. In principle, there are no queues for private goods that are commonly bought and sold in a private market. The social queue problem occurs in the context of activities that are socially accepted as merit goods, are collectively provided by the public sector, but their supply is short of meeting demand, hence the use of the term ‘social’ to emphasise the ingredients of the problem and its general applicability to more areas. Can the problem of the social queue be resolved using economic criteria? How and what are the consequences for social welfare? Is an application of economic criteria superior to current practices in terms of social welfare? These are questions that this chapter attempts to answer at the hand of an applied model with demonstrative figures for health services, and that makes use of random sampling from a representative stock of patients. Although parameters of the model were applied to the case of medical facilities in the Netherlands at around 2000, they are general enough to apply to other countries and periods and can be easily replaced by other parameter sets while keeping the model intact. As background to the social queue problem and our approach to resolving it, the following applicable context is recalled. Although total expenditure in health care, and in particular the government’s health care budget, have increased significantly in rich countries in the past four decades, hospitals are often not able to increase their capacity satisfactorily in the face of varied and rising demand, increasing treatment costs and constrained budgets.1 The results sometimes include long waiting lists of patients requiring medical treatment, and a growing force of less-able people, which costs the state large sums of money.2 In response

Queuing problem  317 to long waiting lists, it is sometimes proposed to give preferential treatment to the active workforce. In defending this discriminative policy it is argued that the economy needs these productive employees most. Other arguments are added. For example, employees contribute to the government’s budget via tax payments on their earned income. When employees are obstructed from work due to sickness, this causes a strain on a firm’s resources and on the government’s revenue. Such proposed policies provoke resistance on the grounds of inequity and tend eventually to be voted down. The model we develop and apply studies the efficiency and equity effects of discrimination in medical treatments requiring surgical operations, the criterion for positive discrimination being the income earnings of the incoming patient. The application of the model makes use of Monte Carlo methods where the subject patients for treatment are sampled randomly from a representative population of patients. The questions addressed are: Can a discriminatory approach based on income earnings increase the capacity and productivity of the health-care sector? Can waiting lists be shortened or eliminated? Can equity and fairness be enhanced? Can policies be designed in such a way that trade-offs between conflicting goals be minimised or eliminated? It will be seen that increasing microeconomic efficiency would release resources that can be used for treating more patients and thus reduce waiting times, and financial cost. The results will show a convergence between the efficiency and fairness goals. The chapter is further divided as follows. Section 2 deals with quality adjusted life years (QALY). The QALY approach provides an indication of the effectiveness of medical attention in terms of gained life years in good health after treatment. In this way medical authorities can be guided as to which activities to undertake within their budgetary resources that contribute most to a healthy community. In this approach, no attention is given to the income effects caused by the increased quality of life of a patient. Section 3 links the income effects to the QALY approach. Section 4 presents the basic model. Section 5 discusses the quantification of the model and the simulation methods used. Section 6 discusses the results of the simulations, Section 7 performs additional random sampling simulations, and Section 8 concludes.

2  Quality adjusted life years (QALY) Private markets allocate efficiently only if the standard assumptions hold, that is, perfect information, perfect competition, and no market failures such as externalities or collectivities. Because health care conforms only minimally to the assumptions necessary for market efficiency, there is the view that many significant health services belong to the public domain. On the other hand, there is the contrary view that declaring health care as a public good does little to resolve problems of free riding, limited non-rivalry, mounting health-care budgets, and persistent social queues. Both views are supported by empirical observation. In the light of these limitations it is generally recognised that the design of holistic systems of health care is fruitless. Instead, health analysts have focused more on developing accounting frameworks and performance measures for

318  Queuing problem health that are rational and reasonably operational. Such measures need to incorporate a maximand that values improved health, allow establishing causal links between sickness condition, health care and improved health, appraise the comparative advantage of private and public delivery mechanisms, and apply greater economy in the treatment of patients. It may seem, at first sight, that given the normative nature of value judgements, there is no way of reaching a consensus on the most desirable performance measure. However, on second thoughts, performance measures can be so designed that their outcomes are shown to be consistent with the application of a more than one system of value judgements, and hence approach a consensus. A fairly neutral starting point in designing a multijudgment performance measure is provided by a quality adjusted life years (QALY). Literature on QALY is abundant. The QALY approach was first considered in the 1970s and has remained an important research topic since then; see for example, Torrance (1986), Wagstaff (1991), Barr (1998), Dolan (2000), and Culyer (1980). The policy objective underlying QALY literature is maximisation of the community’s health subject to a cost utility analysis. The premise of the QALY approach is that benefits from medical care should not only be measured in terms of the quantity of extra life years it produces, but also of its quality. Extra quantity and quality of life after treatment define utility benefits an individual receives from medical treatment. The QALY approach looks at the extra life years adjusted for quality resulting from a specific treatment, and appraises these extra years against the specific treatment costs. The QALY approach is less concerned with the individual’s need and utility for medical care; it is more concerned with the maximisation of society’s health status at least cost.

1.0 0.8 Quality of life index

Level 2

0.6 0.4 0.2 0

Treatment costs in USD

USD

Figure 16.1  QALY

Level 1 x

x + y Years

Queuing problem  319 In the QALY accounting framework normal health is assigned 1, with the score going down according to the poorer the health of the person. One QALY is one year equivalent of normal health. Figure 16.1, level 1, shows QALY over time for a reporting patient with particular sickness conditions and with a QALY of 0.4, who does not undergo treatment and dies in year x. If treated, the person recovers doubly to a higher health score of 0.8 and lives longer, y more years, as shown in curve 2. The area of the benefit is given by the difference between levels 1 and 2, or simply the difference ∆QALY. Figure 16.1 shows also the costs incurred for the treatment that can be expected to concentrate in the beginning and stabilise thereafter. The expression (∆QALY/cost) is a benefit/cost ratio that can be calculated for various diseases or sickness conditions. The reciprocal expression of cost per QALY gained per disease treatment is more comprehensive for policy makers in allocating health provisions. There are presently such estimates for treatments of cancer, heart, kidney diseases, and various body injuries.

3  Linking QALY to earnings On decisions where to spend the public health care budget, the QALY approach suggests to choose the medical procedures that generate the highest cost-efficient gain in quality of life. Based on the QALY approach, a life will be saved or enhanced as long as there is a significant increase in quality of life to be expected. Although health maximisation is the focus of the doctrine, and not the utility derived from health, an increase of the sum of quantity and quality of life can in this context be described as utility benefits. For approaches towards the measurement of health capital see Nordhaus (2002). In principle, procedures can be developed for ranking activities according to the gains to health they create for every dollar of resources spent. Priority can then be given to those activities that generate more health per dollar spent. There are four gains from application of such procedures. First, internal efficiency is increased. Second, the society’s sum of health (sum of QALYs) is increased, which lessens the strain put on public expenditures. Third, the productivity of treated persons will also be increased. When people receive extra QALY due to treatment or cure, their future earnings and the national income are increased. Fourth, now that there is more income, future collectable tax payments increase too. And, in principle, a portion of the additional tax revenue could go directly to the health care budget, and augment it further. However, the QALY method stops short of treating some urgent policy issues. For instance, QALY analysis does not suggest how to improve the capacity and productivity of the health care sector or how to eliminate long waiting lists. What if, on the other hand, treatment should become income-related? The decision on which patient or which disease type to treat would then be based on the height of an individual’s income combined with his or her QALY score. What if priority in treatment is given to people who earn higher incomes and, correspondingly, have higher labour productivity? When this group of people receives treatment earlier, they recover sooner, rejoin the labour force, and generate a greater value added than other groups. Furthermore, regained higher incomes

320  Queuing problem allow tax payments that contribute to raising state allocations to the health sector. The greater health care budget can be used to treat more persons, and so the general standard of health for the whole population can be shifted upwards and to a higher level than when selection of patients would have been done on a first come first treated, or other basis. The above proposition, in which allocation of medical resources is guided by the height of patient income can certainly be condemned on ethical grounds. However, when economic considerations in the selection of patients will show a significant increase in national income, tax payments, and recycled resources for health resources, leading to an increase in the number of treated and cured patients per year, this proposition could prove to be more efficient and more just in the long run than the random application of the QALY approach. We describe below a simple demonstrative model in five sets of equations that links QALY of patients to earned incomes of patients. To illuminate the basics of the model we take the patient population as being employed in the labour force, their total number is given at 1000 patients who are all in the waiting list for medical treatment. The composition of the patient population is structured along three dimensions: disease type, age group, and educational level. A patient has one of two diseases d (either cancer c or a heart problem h). Patients fall in two age groups g (either middle age m or senior s); and into two levels of educational attainment e (intermediate i and university u). Age and education determine income earnings, and it is the intention to link QALY and income earnings to each other. Hence, it is important to specify the dimensions of age and education in the profile of the patient population. The combination of g and e gives four groups in a descending order of discounted lifetime income: um followed by im, us, is. Since patients combine all three attributes each patient L is identifiable in terms of the three indices g, d, and e. This will apply also to variables belonging to the individual patients such as income earnings and future discounted income. The distribution profile of the patient population on the three attributes (from which the model runs Monte Carlo drawings), should be representative of the real world so as to avoid bias in a random drawing of patients for treatment. Since the application is meant to approach, as much as possible, the case of the Netherlands we shall design a composition of the patient population that matches labour force survey data. Cost of treatment differs by disease and age. Material returns to the national economy, and eventually the recycling of these gains to the health sector via tax payments and increased health budgets, depend on the earning capacities of recovered patients, which are a function of their disease, age, and education.

4  The model The model consists of five sets of equations. A fixed population of 1000 diseased individuals whose characteristics are selected so as to be representative of the whole population is assumed to be treated. When a person is not treated, he or she dies after exactly one year. Furthermore, it is assumed that the available annual health-care budget will be spent entirely. Table 16.1 summarises notations used.

Queuing problem  321 Table 16.1 Notations Indices: e = Educational level index; i and u for an intermediate and a university educational level; g = Age group index; m and s for middle and senior age; d = Disease type index; c and h for cancer and heart disease; n = Number of years between current age A of the patient and age on retirement T; p = Policy index to denote the policy or simulation which is run, p = 1, 2, 3 and 4 Endogenous variables Lged Number of treated patients specified by disease type d, age group g, and educational level e Achievable QALY giving probable remaining life years after treatment; Qged specified by disease type d, age group g, educational level e Yged Earnings in Dutch guilders (NLG) specified by disease type d, age group g, and educational level e A person’s total discounted future income received when he or she is cured, Xged in NLG W Welfare utility in terms of gained QALYs and saved patients L Exogenous variables Age of retirement, given at 65 years T Profile of incoming patient by age, education and disease Aged The government’s initial health care budget in NLG; given variable B Remaining funds in the budget, which is to be used fully in curing additional R patients during the one year term, R tends to 0 and is hence quasi-given Coefficients Cost of treatment by disease type and age in NLG κgd Tax rates progressively set on basis of educational levels attained, reflecting τe earnings ability Allocated share of tax revenue to health care η Intercept of the earnings functions αge Slope of the earnings functions βge Discount rate, set at 6 per cent δ Assigned values to QALY ϕgd Number of patients willing to pay fully for priority received treatment πged

Eq. 1 specifies the objective function, which is to maximise the product of QALY (denoted by Q), and the number of treated patients (denoted by L), after one year of treatment. W = Max (Σg Qged Lged ) (1) Eq. 2 is a generally formulated technical function that expresses Q prospects for a treated patient L. The degree of recovery will depend on depth of the treatment, as reflected in the parameter of treatment cost. In principle, the function is specifiable by d, g and e. Qged = ϕgd (κgd, Lged )

d, g, e (2)

322  Queuing problem Eq. 3 specifies the income earnings of each patient in the representative population stock as being dependent solely on age and education. In this equation the set α are intercepts and β are slope coefficients of earning functions. Yged = αged + βged Aged

g, e (3)

Eq. 4 combines earnings Y and the generated Q after treatment to give future income. This is discounted over years n (= T − Ag) using discount rate δ, set at 6 per cent per annum, to give total discounted future income X via the following set of enhanced and realisable earnings functions. The recovered patients are assumed to pursue their productive life after one full year of treatment: Xged = Σn Qged {Yged,n / (1 + δ )n − 1}

n = Ag+1, … , T (4)

The fifth equation is crucial for the approach. It deals with alternative queuing procedures for treatment of the patient population within one and the same given budget restriction. Eq. 5 is formulated in four versions allowing the running of four different policy simulations p = 1, 2, 3, 4. In the first simulation p = 1, treatment of patients will be done according to the first come first treated principle, randomly picked from the patient population. There is no income discrimination here and the inflow of patients is randomly determined. Furthermore, the government’s health care budget will remain constant at level B during the simulation and there will be no link between this budget and the incomes of patients as is the case today in the Netherlands. Remaining funds R will be used in the same period to take in and treat more patients, thus increasing L, until R is reduced to approximately 0. This results in the following budget restriction: Σg κgd Lgedp + Rp = B

p = 1, R1 → 0 (5.1)

Next, the second simulation, p = 2, holds to the same principle of random draw from the stock of patients as described above, but goes further to introduce the revenue effects of progressive tax rates τe imposed on all individuals who have received treatment. The tax rates amount to 40 per cent and 60 per cent for earners who have attained intermediate and university education, respectively. Furthermore, the public health care budget is allowed to grow by the allocated share of the tax revenue to the health sector, which applies to the nation as a whole, η. The allocated share is 10 per cent. In this way the health budget is automatically increased and more patients can be treated. If patient treatment will allow patients to be productive again, earn an income, and pay a tax then the realisation of these benefits must be due to the intervention of the health sector and the economic rationale would argue for ploughing back part of these revenues to the health sector. Σg κgd Lgedp + Rp = B + η Σg τe Xge Lgedp

p = 2 , R2 → 0 (5.2)

The third simulation, p = 3, will be one in which the economic rationale is further accentuated. Discrimination is introduced on the basis of income, so that patients with a higher discounted lifetime income are treated first, and those with a lower discounted lifetime income will be treated later. The tax rates and the

Queuing problem  323 allocated share to health care are the same as in the previous policy simulation. The restriction are obviously the same as in the previous policy, except for the crucial difference that the composition of L gives priority to the treatment of the highest income earning group, um, followed by im, us, and is. Σg κgd Lgedp + Rp = B + δΣg τe Xge Lgedp

p = 3, R3 → 0 (5.3)

The fourth simulation, p = 4, goes a step further in the economic rationale. It considers privatised health care in which immediate treatment is to be given to patients who have the willingness to pay the full cost of their treatment. The number of those willing to pay will be based on the accumulation of income and savings in past years. Here, those willing to pay are represented by πgde and what they pay is the cost of treatment depending on their age and the disease they have κgd. Once this stream of priority patients is exhausted, further selection of patients is based on the previous policy, in descending order of income earning groups um, us, im, and is. The same progressive tax rates and the recycling back of a portion of the tax revenue to health care apply as in the previous policy; furthermore, the collected cost payments will supplement the health budget. The budget restriction is as follows. Σg κgd Lgedp + Rp = B + Σd κgd πgde + η Σg τe Xge Lgedp

p = 4, R4 → 0 (5.4)

To summarise, the model consists of five equation sets in five unknown variable sets: L, Q, Y, X, and W. It is a determinate model which is solvable to give the maximand W. Note that the maximisation of the objective function W automatically implies that for each of the given budget restrictions the rest variable R is to be minimised.

5  Quantification Quantification of the model requires a representative design of the patient population so as to correspond as closely as possible to the configuration of the employed labour force and its encountering of the two disease types. Next are the parameter estimates in eqs. 2 to 5. The design of the patient population: It was stated earlier that the composition of the patient population of 1000 has to correspond to the actual distribution of the employed workforce on disease type, age group, and educational level. This is done by constructing a composition along the lines of a simple human resources matrix3 of two age groups by two educational levels based on the labour force data of the Netherlands in 1999, as in Table 16.2. Table 16.2 A representative human resource matrix by age and education Patient force L

e = intermediate education (i)

e = university education (u)

Total

g = middle age (m) g = senior age (s) Total

493 357 850

 87  63 150

 580  420 1000

324  Queuing problem This table is extended step-wise to include the classification of the patient population into the two disease types, giving Table 16.3. The disease types’ proportions are fixed at 75 per cent for cancer and 25 per cent for heart-related diseases, which are close to the actual occurrences of the two disease types found in figures provided in CBS (1999). Table 16.3 The representative population of 1000 patients by age, education, and disease types Patient force L

e= intermediate education (i) d = cancer c

g = middle age (m) 370 g = senior age (s) 268 Total 638

e = university education (u) Total

d = heart h d = cancer c d = heart h 123 89 212

 65  47 112

22 16 38

 580  420 1000

The costing of QALY: Estimates of costs per patient for treatment and cure of the two diseases in eq. 2 were derived from figures published by NZI (1998). Average costs per hospital admission are estimated at about NLG 25000, and an average admission lasts for about ten days. However, for the specified disease types multiple admissions are necessary and treatment costs are not solely based on these admissions alone. Medicinal treatment, monitoring of progress, and repeated efforts to cure one patient constitute a significant cost item. Technical consultations with medical experts resulted in the cost figures κgd in Table 16.4. Table 16.4 Values of κgd

g = middle age (m) g = senior age (s)

d = cancer (c)

d =heart (h)

NLG 150000 NLG 200000

NLG 125000 NLG 175000

In the best of all cases such a cost figure as indicated above will bring the patient back to full recovery, but the situation of partial recoveries may be more prevalent. To allow for both variations the general formula of eq. 2 will be interpreted in both senses. In the first sense the assigned indicator for Quality Adjusted Life Years, Q, is held at unity implying that each cured and recovered patient will be assumed to have a quality of life of 100 per cent. Thus, the functional term ϕgd is assigned the value 1, and hence Qged = 1. In the second sense, although patients are cured, they do not recover fully after treatment. The quality of life will therefore not reach a level of 100 per cent. This is the principle of the QALY approach. The severity of the two disease types is usually equally high and it is the age of the patient that is a significant factor in recovery. We assign φgd and Qged the values4 in Table 16.5.

Queuing problem  325 Table 16.5 Values of ϕgd and Qged g = middle age m

g = senior s

Age 45–49

Age 50–54

Age 55–59

Age 60–64

ϕgd = Qged = 0.8

ϕgd = Qged = 0.7

ϕgd = Qged = 0.6

ϕgd = Qged = 0.5

The determination of earnings: The model is based on the assumption that people, who earn higher incomes, could contribute more to social welfare. When these people are favoured and treated first, the government could collect more funds for improving the health care sector. If people with higher incomes could all return to their jobs earlier, then the sum of expected future earnings would increase, which in turn would increase expected future tax payments. Of course this applies to all individuals with a steady income, but the effects are greater for the above-mentioned group. For applying the model, expected future earnings functions of the patients need to be calculated. The earnings functions depend on several determinants covering such aspects as educational qualifications, work experience, physical conditions, and psychological characteristics of the earning patient, in addition to the job access situation that is largely influenced by the economic outlook. It is assumed that individuals in the sample have all reached the age of 45 to 65, have obtained a university or an intermediate educational level, have a job, and have gained work experience as reflected by the age reached, and are struck by one of two possible diseases: cancer or cardiovascular disease. These are the minimal assumptions necessary to deduce the income functions Y and the total discounted future income functions X. This implies that we take the earnings functions as fully dependent on education and work experience as reflected by age, which is coherent with human capital theory. Other determinants of earning such as psychological characteristics and job access are considered randomly distributed. The earnings functions are displayed in Figure 16.2. The horizontal line represents the starting wage that can always be earned. There are two curves representing the age-earnings profile of labour that attained intermediate education and university education, respectively. At an early age a person receives schooling, occasional training, and gains experience. After schooling, training on the job is undergone. Due to received schooling, training, and experience, income increases rapidly towards middle age, slows somewhat, tends to flatten at senior age and eventually falls. The slopes of the curves for earners with a university education are usually steeper than that of the intermediate education in the growth phase of earnings towards and through middle age. In contrast, the slopes of the curves during senior age tend to reverse and show a greater tendency to fall among the intermediate education than the university education. Working with linearised portions of the earning curves is sufficient for the purpose in mind. Working with two educational levels and two age groups

326  Queuing problem requires identifying four tangents. We denote the four tangents by mu and su for middle age university-educated and senior age university-educated, and mi and si for middle age intermediate-educated and senior age intermediate-educated.5 There is a large amount of empirical literature on dependence of earnings on education and age which can in principle be used to provide estimates indirectly of intercepts and slopes of these tangents, or in combinations of them. However, the classifications by education and age in this empirical literature do not readily fit with those used in the chapter. Being an exploratory exercise, use is made of CBS research results on the wage structure to determine these linear functions.6 Two points of intersection linking income earnings to age and educational levels are used to calibrate each function7. The α and β coefficients in Table 16.6 are the outcomes, and they correspond neatly with the expected shape of the tangents in Figure 16.2. Other parameters: there are four more parameters to treat. The height of the progressive income tax τe was already mentioned, amounting to 60 per cent and 40 per cent on earnings of recovered patients with university and intermediate educational attainments—reflecting their different earning capacities—respectively. The allocated share of tax revenue to health care, η, is approximated at 10 per cent. The recycled funds are considered to be available for the public health-care budget in the same year, and that these additional funds will be entirely used to increase capacity and intake. In the fourth simulation, where privatised health care will be considered, we introduce a parameter giving the number of patients willing to pay the full costs of medical treatment if they are immediately treated. It will be assumed that only some wealthy patients in the senior age group with a high educational level have such a willingness to pay. Other patients may not have accumulated sufficient means in their past years to be able to pay for their treatment. Based on findings in the empirical literature on welfare provided CBS (1999), we consider that from the patients in the senior age group with a university education only Earnings

su mu

u = university education earnings curve si

ml

i = Intermediate education earnings curve

Starting wage Costs of education Years in education

m= middle age

s= senior age

Figure 16.2 Earnings functions by age group and attained education

Queuing problem  327 Table 16.6 Estimates of α and β coefficients i = intermediate education

u = university education

m = middle age

m = middle age

s = senior age

Ymi = αmi + βmi  Ami Ysi = αsi + b si Asi Ymu = αmu + βmu Amu = 29550 + 105 Ami = 39956 − 84 Asi = 43143 + 455 Amu

s = senior age Ysr = αsu + βsu Asu = 52872 + 278 Asu

15 individuals are able to pay full costs for immediate treatment, giving πgde the value of 15. Note that after treatment of these 15 patients, the fourth simulation continues further in the form of the third simulation. Patients who have university education and are middle aged, group mu, have the highest discounted future income and will expect to be treated first. Hence, they may not be willing to pay for treatment knowing that they may be treated as first anyway. This reduces the prospects of many volunteers for immediate treatment, and diminishes marginal returns of a privatisation strategy. Finally, the initial health care budget, B, will be fixed at NLG 25 million for the stock of 1000 patients, so that given the costs per patient only a portion of the patient population could be helped within this budget.

6  Findings and discussion In order to test whether economically oriented discriminatory policies towards patient selection for medical treatment could increase medical capacity and reduce waiting time, 100 simulations were run for each of the four policies in the context of two different appraisal situations. The four policies were identified above in terms of the composition of patients as P = 1, 2, 3, and 4. The composition is drawn randomly from a population of 1000 patients, subject to restrictions imposed by the respective policies, where applicable. While policy 1 represents the current practice of first come first treated, and is a pure random draw from the population, policies 2, 3, and 4 are more economically oriented in an ascending order. The 100 simulations run for each policy are meant to guarantee that the application can produce a plausible range of stable outcomes. As for the appraisal situations, appraisal situation I postulating that Q = 1 actually means there is a fully sustained recovery of the treated patient until retirement. Welfare utility W is reduced simply to the number of treated patients L. In appraisal situation II, Q is attained at Q = 0.8 and falls to Q = 0.5 as age increases. Here, W depends on treated patients L and their QALY. Table 16.7 gives for each of the four policies, under the two appraisal situations, the average W and its standard deviation over the 100 simulations8. The standard deviations of policies 1 and 2 are significantly higher than those of other policies.9 Recall that the principle of policy 1, just as in policy 2, is first come first treated. Therefore, the random element is strongly present in this simulation. Policies 3 and 4, by

328  Queuing problem discriminating on the basis of income of patients, restrict randomness and manifest less variation regarding treated patients. Table 16.8 presents three ratios that compare the outcome for W under each of the proposed policies of 2, 3, and 4 to that of the currently followed policy 1, as well as comparative ratios of relative performances among the proposed policies. These ratios are helpful in indicating which policy is more efficient than the other policies. Each policy simulated, in ascending order, is able to treat a larger number of patients than the previous one and give greater welfare utility under both appraisal situations I and II. This ascending improvement shows that a discriminatory economic approach is effective in solving social queues. Policy 1 is able to treat on the average 152 patients. Policy 2, by recycling part of the gained revenues to health care makes possible a treatment of 166 patients that is 3 per cent more than in the previous policy. Policy 2 is an application of what can be called primary economic principles. When the health sector cures patients it increases human capital, leading to positive externalities that require internalisation. Policy 3 goes a step further towards what can be called the secondary economic rationale. By discrimination of patients on the basis of future discounted earnings the policy treats on average 1.28 times and 1.40 times more patients, L, than policies 2 and 1, respectively. And if L is weighed with Q the favourable effects are even more dramatic giving for policy 3 scores of 1.41 times and 1.45 times those of policies 2 and 1, respectively. This shows that if policy 3 would be implemented, the expected increase in health status W would be significantly higher than if current practices are followed, and would ultimately lead to a swift end to Table 16.7 Average W and standard deviation for alternative policies obtained from sets of 20 simulations per policy P=1

P=2

P=3

P=4

Appraisal situation I: W = L Average σ

151.8   2.0

165.8   2.0

212.8   2.0

212.8   2.0

Appraisal situation II: W = QL Average σ

108.9   2.5

108.9   2.5

108.9   2.5

108.9   2.5

Table 16.8 Ratios of relative performance of alternative policies P2/P1 P3/P1 P4/P1 P3/P2 P4/P2 P4/P3 Appraisal situation I: W = L ratios Appraisal situation II: W = QL ratios

1.09 1.03

1.40 1.45

1.51 1.54

1.28 1.41

1.39 1.50

1.08 1.06

Queuing problem  329 waiting times and health queues. Two comments need to be made on this policy. First, policy 3 contains an element of unfairness, but this can be easily repaired. For instance, there is always a number π(poor) of patients with very low perspective earnings Y(poor) who are treated in policy 1 or policy 2, and these may never be treated under policy 3. The unfairness can be removed when repeated simulations will identify and fix this number π (poor), and when this number is randomly drawn from the group Y(poor) for direct treatment. Ultimately, this ‘adjusted’ version of policy 3 will cure fewer patients than the ‘unadjusted’ version. The difference is the trade-off between efficiency and fairness. A utilitarian, who values the life of all persons equally, will choose the unadjusted policy. An egalitarian or Rawlsian may choose the adjusted policy. Second, policy 3 should not be misunderstood as favouring currently rich patients. The emphasis is on expected discounted lifetime income in the future, which means that a low-income young earner can win from the richest person close to retirement age in heading the waiting list. Next is policy 4, which can be called the tertiary economic rationale. Here, the wealthier patients among the senior age group, who have a higher educational level, who are willing to pay the full costs of their treatment receive an immediate treatment. This is the policy that maximises most the stated objective function.10 Although the capacity available for the health sector increases, when compared to the previous policy 3, only a limited number of additional patients will benefit from the paid contributions of wealthy individuals who are willing to pay. The marginal benefit to the national health status is not improved significantly by the privatisation incentive. Of course, this policy could be more effective when the payment for priority treatment is reduced to a portion of the actual cost, and consequently more patients may be willing to pay. On the other hand, when one takes into account that most of the wealthy patients may not be willing to pay if they know that policy 3 is the next policy option to policy 4, and that consequently they stand high in the waiting list and will most probably be among the first to be treated, then the favourable effects of the privatisation incentive can be expected to diminish significantly, and may not even become operational. Notwithstanding the ethical objections towards a VIP status in health care, our tentative conclusion is that policy extremes such as privatised bidding for medical treatment will not significantly reduce health queues and improve health capital, and they may not be operational either in view of the calculating behaviour of bidding patients. This statement does not deny that there exist positive complementarities between privatised health care with no waiting, and public health care with waiting lists, that need to be identified and exploited. Such complementarities are studied in Hoel and Saether (2001).

7  Additional random sampling and policy simulations We ran 100 additional Monte Carlo simulations on which we report briefly. In Figures 16.3 and 16.4, we give results of simulations in graphic forms following appraisal situation I, that is, W = L; and following appraisal situation II, that is, W = QL, respectively. These are done under increasing allocations to the health budget B, that is, beyond the NLG 25 million reported so far.

330  Queuing problem It is shown that at a budget of 75 million, policy 3 (and policy 4) perform at their maximum when compared to other policies. As the health budget increases beyond 75 million, the number of treated persons converges among all the policies. At a budget of 165 million, all patients will be treated and there is no sense in discrimination based on economic rationales. Furthermore, the ranking of policies is the same under the two appraisal criteria of I and II.

8  Concluding remarks The chapter developed and applied a model to demonstrate the effectiveness of an economically discriminative policy in the elimination of waiting lists in health care. The model specified linkages between QALYs, patient earnings, recycling of tax revenues to health care, and a likely greater intake of patients. The model was supplemented by drawing patients at random from a representative population for medical treatment, and the economic impacts of the treatment calculated by the model. Simulated applications gave numerical outcomes of alternative policies. The tentative conclusions presented here are based on raw figures and crude calibrations and there is a lot of space for improvement. With due acknowledgement of the need for such refinements, the general conclusion remains valid that applying economic criteria in solving queuing problems is effective in the selection of more productive lifetime patients, raises the national health capital, treats more patients, reduces health queues, and ultimately achieves a greater sense of fairness in relieving patients. We drew also borderlines by demonstrating that an extreme interpretation of economic criteria, such as outright privatisation or bidding for the first place in a queue, may not appreciably improve the outcomes obtained. It can also be corruptive in view of calculating behaviour of involved medical authorities. It is important to emphasise that the problem of health queues and economic solutions to the queuing problem is not unique for health services. Social queues and economic solutions to them are applicable to many scarcely available provisions in the educational sector, subsidised housing, poverty alleviation, admission of queuing migrants from poor to rich countries, and many related phenomena. Our approach can be readily adapted and effectively applied to these policy issues.

0

0

P=2

P=3

99 132 166 199 232 265 298 332 365 398 431 466 497 532 565 597 630 663 696 730 763 796 829 863 896 930 961 995 1000100010001000

15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160 165 170 91 121 152 182 212 242 272 303 334 364 394 426 455 487 517 546 576 607 637 667 698 728 759 789 819 850 880 911 941 971 999 1000

96 135 174 212 251 290 329 368 406 445 484 523 562 595 623 651 677 704 730 757 784 811 837 864 891 917 943 970 996 1000100010001000

66

10 61

67 113 152 191 229 268 307 346 385 423 462 501 540 579 611 640 666 693 719 746 773 799 826 852 879 906 932 959 985 10001000100010001000

49

33

5 30

Initial budget (*1.000.000)

P = 4 17

0 0

P=1

0

200

400

600

800

1000

1200

Results of 100 simulations of appraisal I: W = L (averages)

Figure 16.3 Performance under more allocations to health. Appraisal situation I: W = L

Number of treated patients

34

44

0

P=3

P = 4 10

23

0

P=2

5 22

0 0

P=1

0

100

200

300

400

500

600

700

800

91 115 138 160 183 206 229 252 275 298 323 343 368 391 413 435 459 481 504 528 550 573 597 619 643 664 688 711 716 717 716

20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160 165 170 87 109 131 151 174 195 217 239 260 283 305 325 349 371 391 412 435 456 478 500 521 543 565 587 609 630 652 674 695 717 716

98 128 158 188 218 247 277 307 337 366 396 426 456 476 492 508 524 539 555 571 587 603 618 634 650 665 681 697 713 716 716 716

69

15 65

78 108 138 168 198 227 257 287 317 347 376 406 436 466 485 502 517 533 548 564 580 596 612 627 643 659 675 690 706 716 716 716 716

69

46

10 44

Results of 100 simulations of appraisal II: W = QL (averages)

Figure 16.4 Performance under more allocations to health. Appraisal situation II: W = QL

Number of gained QALYs

17 Modelling convergence in economic growth between rich and poor countries

1 Introduction The modelling applications in this chapter are stylistic and analytical. A specific hypothesis is formulated to explain stylised facts, the hypothesis is then measured, analysed, and its implications studied. The stylised facts relate to catching-up tendencies in economic growth between poor and rich countries. Among the questions considered are the following. Is there a convergence between rich (or developed) countries and poor (or developing) countries? How do we separate the effects of different economic systems from the effects of levels of economic development? How do we assess the roles of such exogenous forces as the world market and government policy in economic systems and levels of economic development? In this context, an opening sentence from Haavelmo (1964) is a good starting point for the chapter. He wrote: In the world economic picture that we can piece together from current international statistics, perhaps the most striking feature is that of economic dissimilarities … The ultimate question to be expected from the thinking citizen is as plain as it is scientifically formidable. It is the question why certain areas or peoples are economically ‘backward’ while others are ‘advanced’. Haavelmo (1964) pp.1–2 It took a quarter of a century after Haavelmo’s plea for understanding long-range economic development before the study of income convergence patterns between rich and poor countries could take off. The take-off has been greatly facilitated by the well-known data set of real GDP for 130 countries over 35 years, compiled by Summers and Heston (1988), which opened the way for various empirical studies.1 Taking all rich versus all poor countries together, the statistical material shows that there is a slight catching-up tendency. Further disaggregation has highlighted a convergence of income levels within richer countries but divergence within poorer countries, with some even falling behind the rest and becoming relatively poorer.2 However, economic theorising on these tendencies and empirical testing emphasised supply factors. Little attention was given to demand factors that could lie behind convergence tendencies.

334  Convergence in economic growth We offer in this chapter a demand-side model for formulating the relationship, which is tested, analysed, and interpreted. After a brief review of the supplyside theory and evidence in Section 2, the chapter will lay out the demand-side approach in Section 3, the evidence in Section 4, the simulations in Section 5 and elaborations thereon, with concluding remarks in Section 6.

2  The convergence hypothesis: supply-side theory and evidence Two economic models—both emphasising the supply side—have been invoked to explain the mixed convergence tendencies between poor and rich countries referred to above: Solow’s growth model, which predicts convergence, Solow (1956) and Krugman’s divergence model, Krugman (1981). The mechanism behind Solow’s growth model is diminishing returns to reproducible capital. A poor country characterised by a low capital-labour ratio, has a higher marginal productivity of capital and thereby tends to grow at a higher rate than a rich country with a higher intensity and lower marginal productivity of capital. Furthermore, there is a tendency for capital to move from rich to poor, thereby accelerating the convergence process. The contrary model of Krugman stresses increasing returns to capital, technological edges, and learning, in assuring higher levels of more competitive capital and industrial exports in the rich country. Endogenous growth is seen to work to the advantage of the rich country that grows at a higher rate than the poor country. Capital flow tends to reverse from poor to rich, aggravating income gap between rich and poor, further. Lucas (1993), Barro (1991), and others have elaborated Krugman’s model along endogenous growth theory lines to show basically the same thing: an increasing income gap between countries that invest in human resources and are able to capture the public goods character of those investments, and countries that do not (are unable or unwilling) to invest sufficiently in human resources, learning, and innovation. A synthesis is found in Mankiw et al. (1992) who develop a model that combines mechanical growth theory as represented by Solow and endogenous growth theory as represented by Krugman, Lucas, and others. They test their model with the data set of Summers and Heston and find that countries with similar technologies and rates of accumulation and population growth should converge in income per capita. Yet this convergence occurs more slowly than the Solow model suggests. More generally, the results indicate that the Solow model is consistent with the international evidence if one takes account of (dis)advantages of individual countries with respect to human and physical capital endowments. Another empirical paper which contributes to a synthesis is by Barro and Lee (1993). They explain the growth performance of 116 economies from 1965 to 1985. They find a conditional convergence effect, whereby a country grows faster if it begins with lower real GDP per capita in relation to its initial level of human capital, next to other stimulating factors, such as high ratio of investment to GDP, small government, and political stability.

Convergence in economic growth  335

3  The convergence hypothesis: demand-side theory and evidence The models mentioned above emphasise supply factors in the determination of economic growth. The debate has so far been unbalanced, as it excluded models of economic growth that emphasise demand factors. If we wish to elaborate the demand side, then the fittest demand framework is undoubtedly the circular flow model based on the social accounting matrix (SAM). Several chapters in this book have treated the SAM in depth. It is sufficient to mention here that—being a circular flow matrix of transactions between agents—the SAM can be written as a model of the economy wherein a vector of endogenous variables y is linked to a vector of exogenous variables x via a coefficient matrix A, which when inverted gives multipliers M, as in eq. 1. The endogenous variables include production and value added by sector, and income, consumption, and investment, among many others. The exogenous variables in such a model are those of government and rest of world. y = (I − A)−1 x = Mx

(1)

Our concern is with a tiny part of eq. 1. We focus on the multiplier effect of a weighted composite of exogenous demand injections sector activities j through public spending and (or) exports on national income (approximately the GDP); we focus on this income multiplier effect for rich and poor countries and examine their growth tendencies to shed light on the convergence hypotheses. Denote the exogenous demand injection belonging to vector x, by X (this is a million units of the currency of the concerned country that is distributed on the injected sectors j in accordance with the sector shares in GDP). Denote the affected endogenous variable of national income by Y, belonging to vector y. Eq. 1 can be written specifically as eq. 2. The multiplier effect that we focus on is m, which is the weighted sum of relevant cells3 in the multiplier matrix M. Y = m X

(2)

Eq. 2 can be rewritten as eq. 2.1 Y = m (X/Y) Y

(2.1)

This can be re-expressed in growth rates using index g as eq. 2.2. g

g

Yg = mg + (X/Y) + Y

(2.2)

Note here that we treat the three growth rates on the right-hand side as hypothetical values, in the sense that these growth rates are either assumed or forecasted and are consistently estimated in relation to each other. The combined effect of the three growth rates results in the realisable growth rate of the national income on the left-hand side, which can differ from the hypothesised growth rate. If a hypothesised growth rate is denoted by h and a realisable growth rate by o, eq. 2.2 can be rephrased as eq. 3.

Y go

> gh m + (X/Y) gh + Y gh
Y p < mpgh + (X/Y) p + Y p

and

gh go > gh Y r < m rgh + (X/Y)r + Y r



(3.1)

The hypothetical and realisable values of the growth rate of income, Ygh and Y go, respectively, are generally different due to the independent determinacy of gh mgh and (X/Y) . If it can be shown for the groups of the poor and rich countries that starting from the same hypothetical growth rates, Ypgh = Yrgh, the following holds, gh

m pgh + (X/Y)p > mrgh + (X/Y)rgh

(3.2)

go

go

then it follows that realisable growth rates will show Yp >Yr , which is an indicagh tion of catching up. We treat first the growth of the exogenous share (X/Y) and show that this can be expected to be higher for poor than rich countries, and take up later the prospects for mgh. gh We start with (X/Y) . An interesting feature of the accounting system is that the row element of government expenditure and exports X can be divided by the row of total national income Y to give the exogenous share, X/Y. We have defined X to consist of government expenditure and exports. Our hypothesis is that the share of these items in the national income tends to grow rapidly during the early stages of economic development but ebbs down and stops growing at higher stages of economic development. This hypothesis is in Figure 17.1 that shows the relationship between X/Y and income per capita, Y/N, this being the conventional expression for the stage of economic development. The quasi-logistic curve in Figure 17.1 can be formulated as eq. 4. This is also the form in which the hypothesis will be empirically tested.

X/Y =

β (Y/N)

(4) Wagner’s Law predicts that at higher levels of economic development, that is, as income per capita grows, the relative share of the public sector in national income will grow. Although the basis of the statement of Wagner’s Law was the empirics of the nineteenth century, the theoretical foundations behind the phenomenon were developed later by Peacock and Wiseman, Musgrave, Baumol and others, using various public choice arguments. More recent experiences in the balancing of budgetary deficits in rich countries directed attention to fiscal, monetary, and incentive limits to the further growth of the government share in total expenditure. So the share of the public sector grows as income per capita grows, up to a certain limit. This share has a tendency to stabilise at the higher levels of income per capita. (Y/N) + α

X/Y

Convergence in economic growth  337

Y/N

Figure 17.1 Relationship between the exogenous share in national income X/Y and income per capita Y/N

A similar tendency applies to the share of exports in income, which share is very much dependent on economic development, location, and population. As per capita income grows, there is a tendency for the economy to become more open and to attain a higher share of exports up to a point where the share value levels off as more open-economy countries get their portions of world exports. It is also established that the larger the country is, in terms of population and economy, the lesser the share of exports in income.4 The conclusion is that as far as the exogenous share is concerned, and this applies to both constituents of government expenditure and exports, the growth of this share for poor countries is higher than for rich countries: gh

gh

(X/Y)p > (X/Y)r

We go now to mgh. Recalling eq. 2, we have: 1/m = X/Y. Seen as a definition, a rise in X/Y should lead to a proportional fall in m. The relationship between m and X/Y is depicted in Figure 17.2, and can be put down more generally as eq. 5 which will be empirically tested in the next section. m = γ(X/Y)−δ

(5)

In Figure 17.2, curve m1 is obtained for values γ = δ = 1, while curve m2 corresponds with our empirical estimation, which results in γ having a slightly lower value than 1, and δ < 1 indicating that the fall in m is somewhat moderated. The underlying relationship behind empirical curve m2 is that m is higher at the low shares of exogenous demand that correspond with an early stage of economic development, that m declines rapidly with higher X/Y as economic development takes off, but that m tends to stabilise with values of X/Y of around 0.6 or higher. The circular flow effects fall with a rise in the exogenous share, this decline is

338  Convergence in economic growth 12.0 m 10.0 m1 m2

8.0 6.0 4.0 2.0 0.0 0

0.2

0.4

0.6

0.8

1

X/Y

1.2

Figure 17.2 Relationship between multipliers m and exogenous share in national income X/Y

rapid in the beginning but takes a lower rate than the rise in the exogenous share at higher values of the exogenous share. The reasoning is as follows. When the size of the exogenous part is relatively small, as happens in the early stages of development, then the size of the inverted endogenous part will be relatively large, resulting in high multipliers. As the exogenous share of final demand increases, there is relatively a lesser endogenous part to invert and the multipliers are bound to fall. But the proportionate fall in the multipliers is less than the proportionate increase in the exogenous part. The incomeexpenditure-production linkages in the economy, that have been accumulated throughout the past, result in significant increases in the density of the input-output relationships beyond which further increases are only marginal. Once the endogenous linkages are built over time, their multiplier effects will not be proportionally written off with an increased exogenous share of the circular flow of the economy. There is an economic growth advantage here for the rich versus the poor country. The above means that the relative decline in the multiplier m with an increasing exogenous share of the circular flow can be expected to be higher for the poor than for the rich country, mpgh < mrgh. This contributes to a widening of the gap between rich and poor. With the above contrary implications of the tendencies in X/Y and m for convergence and divergence in economic growth, it is then up to empirical testing to verify which effect is greater. As will be empirically shown in the next sections, the widening tendency in economic growth due to m is not strong enough to gh gh countervail the catching-up tendency due to, (X/Y)p > (X/Y)r , so that in the longer run convergence in economic growth will occur, nevertheless. It is important to emphasise that the well-known empirical findings and theoretical elaborations by Kuznets, Clark, Fischer, Chenery, Syrquin and others form the backbone of the demand-side explanation to catching-up tendencies. They highlighted the fact that as higher economic development is achieved there

Convergence in economic growth  339 is a strong shift in final demand and value added from primary goods to manufacturing and services, and that at still higher income levels the share of manufacturing declines and of services increases. The findings are consistent with earlier works of Clark (1940) and Fisher (1939) that gave a demand-side explanation based on relatively higher income elasticities of demand for services as compared to non-services, and that predicted the emergence of the service economy at the cost of a deindustrialisation process at the higher end of economic development. In the SAM analysis, the changes in final demand are driven by higher levels of exports and government spending that favour, at a later stage, services more than non-services. The result is a rapid growth in X/Y at low-income levels that strengthens convergence. The findings are also consistent with the work of Chenery and Watanabe (1958) who found that during the process of development, the total use of intermediate inputs relative to gross output increases and its composition shifts as the importance of primary products declines and of heavy industrial products and services increase. Multiplier effects tend to diminish with development due to increases in the density of the input-output matrices as the economy evolves from relatively simple handicraft production to more complex systems of fabrication and delivery. In the SAM analysis presented in this paper, the Chenery-Watanabe explanations support high levels of m in low-income countries but falling levels of m as the exogenous share of final demand rises with economic development, and stabilising at a higher and stable exogenous share of final demand as is typical for rich countries.

4  Empirical results This section will report on selected results from cross-country comparisons of SAM models applied to ten developing countries (India, Pakistan, Sri Lanka, Indonesia, Iran, Kenya, Colombia, Egypt, South Korea and Surinam), two centrally planned economies (Poland and Hungary) and four developed market economies (the Netherlands, Italy, Germany and Spain). The classification of activities in these SAMs had to be limited to three large groups of sectors: agriculture, industry, and services; industry includes mining, manufacturing, and energy utilities; and services includes construction and transport, among other private and public services. Distinguishing more sectors would reduce the uniformity and comparability of the 16 SAMs reported here. The disaggregation of households in the SAMs of the developing countries emphasises dualities in the location of population in urban and rural areas, and the differentiation within urban and rural groups by level of income earned. This differentiation is done by a categorical split among urban households leading to the distinction between the three groups of employers, employees, and the selfemployed; and a split among rural households by size of land ownership leading to the three groups of large landowners, medium landowners and small or landless households. As a result, there are six groups of households. For a couple of countries a seventh residual group was incorporated so as to accommodate classifications that did not fit the standardised six categories. The SAMs of the

340  Convergence in economic growth European countries distinguish household groups by income classes obtainable from personal income distributions. Testing eqs. 4 and 5 requires data by country on the exogenous share of government and exports in national income, X/Y, and the income multiplier, m, which are obtainable from the SAMs and the matrix inversions, respectively. Data on a third variable is needed, this is GNP per capita, Y/N, expressed in US dollars for the 16 countries and their related years. These are obtainable from published tables of the World Bank Atlas, which are especially suitable in our context as they are based on conversions that smooth the impact of annual fluctuations in exchange rates. Table 17.1 brings these data together. Note that the value of X/Y varies from a lower value of 0.12 for India (poor country) to the highest value of 0.89 for the Netherlands (rich country). The income multipliers start from 7.06 for a poor country and fall to 0.85 for a rich country. Table 17.1 SAM features and GNP per capita of sixteen countries Country

GNP per capita 1000$

Exogenous share = X/Y

SAM income multipliers following sectoral spending injections Average value = m

Rank of sectorsa

Highest/ lowest b

 0.09  0.17  0.17  0.21  0.22  0.24  0.34  0.35  1.51  2.21  0.55

0.12 0.24 0.23 0.37 0.13 0.45 0.22 0.43 0.43 0.76 0.34

7.06 6.11 2.32 2.90 2.82 1.28 2.47 1.15 1.79 0.95 2.89

ASI ASI ASI ASI ASI ASI SAI ASI ASI SAI ASI

1.20 1.24 1.24 2.15 1.40 2.03 1.18 1.86 1.66 1.48 1.54

Rich countries (East Europe) Poland 1987  1.93 Hungary 1990  2.59 Average  2.26

0.40 0.49 0.45

0.92 0.77 0.85

SAI SAI SAI

1.57 1.46 1.52

Rich countries (West Europe) Spain 1980  5.40 Italy 1984  6.42 Germany 1984 11.13 Netherlands 1987 11.86 Average  8.70

0.29 0.43 0.57 0.89 0.54

1.53 1.50 1.32 0.85 1.30

ASI SAI SAI SAI SAI

1.26 1.42 1.47 1.13 1.32

Poor countries India Pakistan Sri Lanka Indonesia Iran Kenya Colombia Egypt South Korea Suriname Average

Year

1968–69 1979 1970 1975 1970 1976 1970 1976 1979 1979

Notes a ASI = Agriculture–Services–Industry; SAI = Services–Agriculture–Industry b For example in case of India dividing the average income multiplier of agriculture by that of industry gives 1.20.

Convergence in economic growth  341 The regression results of eqs. 4 and 5 are found in Table 17.2. Eq. 4 describes a quasi-logistic function that makes the level of the exogenous share dependent on income per capita. To account for a particularly low share in the case of a large rich country, for example Germany, Italy, and Spain, and too high a share of exports for a few particularly foreign trade-oriented small countries, for example the Netherlands, Surinam, and Kenya, a dummy variable is included that takes the value of 1.0 for the first group and −1.0 for the second group. The equation is estimated by non-linear least squares. The regression performs very well in terms of the signs of the coefficients, their t-values, and goodness of fit as indicated by R2 (above 0.8). The predicted highest value of the exogenous share in the observed sample, disregarding the dummy, can be calculated at 61 per cent for the richest country. The predicted and observed lowest values of the exogenous share are the same, at 12 per cent for the poorest country. Table 17.2 Regression results of equations (4) and (5) Item

Explained, explanatory variables and coefficient estimates

Eq. 4 Coefficient t–value Eq. 5 Coefficient t–value

X/Y =

ln m =  

β (Y/N) / 0.632 (12.95) ln γ + −0.077 (−0.58)

[α + (Y/N)] 0.369 (3.25) δ ln (X/Y) −0.619 (5.41)

+ φ4 D4 −0.201 (−4.84) + φ5 D5 −0.799 (−6.39)

R2 0.813

0.890

Because Y is determined by the whole system including eq. 4, the question is raised on possible correlation between the explanatory variable, income per capita Y/N, and the disturbance term yielding a biased non-linear least squares estimator. Note that the explanatory variable is expressed as Y/N and not in terms of Y only. Furthermore, the residuals in eq. 4 were found not to correlate with the explaining variable of national income per capita (r = 0.34), giving no ground for applying more sophisticated regression methods than the one followed, namely the non-linear least squares method. Eq. 5 describes a convex function between the income multiplier and the exogenous share. For estimation purposes the equation is formulated as ln (m) = ln(γ) + δ ln (X/Y) and tested by ordinary least squares. There is a need to introduce dummies in the equation given the composition of the observations. One dummy needs to be introduced to account for a high income multiplier bias in the SAMs of India and Pakistan: the available SAMs of India and Pakistan do not register complementary imports to the full extent, or at all, and hence they underestimate the leakage and overestimate the multipliers. Another dummy is required to account for the differential impacts of economic systems, for example Poland and Hungary. Although one should expect higher multipliers for the relatively poorer Eastern Europe countries (Poland and Hungary) as compared to the richer Western European ones; instead, they have

342  Convergence in economic growth about the same levels. This underperformance of Poland and Hungary is due to the presence of institutions that do not make full use of the potential internal leakage effects within the system. The variation of income multipliers among the West European countries as represented by the ratio of the highest to lowest sectoral multiplier can be calculated as 1.44. For Eastern European countries the variation is higher. It is noted too that Poland has a wider variation (1.57) than Hungary (1.46) has, which reflects a more balanced and well-knit economy in this respect. Eq. 5 was tested with the two above-mentioned separate dummies as well as with one dummy carrying the value of –1.0 for India and Pakistan and 1.0 for Poland and Hungary. The results are very similar so that we can work as well with the simpler case of one dummy, which is reported in Table 17.2. The regression performs very well in terms of all prerequisites. If the mean of m over the 16 countries in eq. 5 was anything meaningful, we would have obtained values of 1 for γ and for δ (curve m1 in Figure 17.2), but we do obtain values of ln(γ) = –0.077; and δ = 0.619, (curve m2 in Figure 17.2). These results are not due to whether the values of m are calculated as weighted or unweighted sectoral impact multipliers, but they are due to the shapes and significance of linkages changing with economic development as was previously stated. We calculated m as an unweighted sectoral average as well. It can be readily seen from Table 17.1 that when m is calculated as a weighted sectoral average the curve of eq. 5 falls more steeply and flattens earlier with values of ln(γ) and δ even further away from γ = δ = 1. Note also that in Table 17.1 agricultural multipliers score the highest and they have the highest share in poor countries. Weighting sectoral multipliers by sectoral shares results in higher aggregate values of m for the poor countries as compared to rich countries causing the curve to shift further away from curve m1 in Figure 17.2.

5 Demonstration With the estimates of α, β, γ, and δ we are now in a position to predict for a poor gh and a rich country respectively, such growth rates as (X/Y) and mgh for assumed gh values of Y . Inserting these in eq. 3 for the average poor and average rich country separately and solving gives the realisable growth rates of income of the poor and g go rich countries Yp and Yr . Recall eq. 3 for the poor and rich country:

Y pgo > mpgh + (X/Y)pgh + Y pgh >

Y rgo > mrgh + (X/Y)rgh + Y rgh >

The above simulations are done in Table 17.3 where we start from an initial income, population, and income per capita for a poor and a rich country (poor and rich as was indicated by the averages in Table 17.1). We assume for both types of countries the same annual rates of growth of 2 per cent per income, 1 per cent for population and 1 per cent for income per capita. Using the estimates of α, β, γ, and δ we obtain the predicted values of growth rates of X/Y and m. These

Convergence in economic growth  343 are used in solving for the realised growth rates of income of the poor and rich country in the last column. The calculations show that the realised growth rate of income of the poor country will exceed that of the rich country. The poor country would achieve an annual growth rate of 2.17 per cent while the rich country would grow annually at 2.02 per cent. Another scenario is run with assumed growth rates of income per capita for the poor and rich at 3 per cent, this scenario results also with a higher rate of realised growth for the poor than the rich: 3.19 per cent compared to 3.05 per cent. More generally, Table 17.4 simulates the annual growth rate of income for rising levels of income per capita. The table shows lower growth rates of income at higher levels of income per capita, the growth rates diminishing slowly and approaching stability as income per capita goes beyond 20,000 US dollars.

Table 17.3 Selected simulations: initial runs for rich and poor countries Year

Assumed Income pc Y/P ($)

Rich 0 1 1 Poor 0 1 1

8703 8790 8877  551  557  557

Predicted Solution Population P

growth (mln.) rate 0.01 0.02 0.01 0.01

220.0 222.2 222.2 880.0 888.8 897.6

Income Y

Exog., share X/Y

growth (mln. $) growth (%) rate rate 0.01 0.01 0.01 0.02

1914660 1953145 1972483 484880 494626 499523

growth rate

0.606 0.0201 0.607 0.00040 0.0302 0.607 0.00080 0.379 0.0201 0.380 0.00399 0.0302 0.380 0.00399

Multiplier m

Income

Value

growth rate

1.2612 1.2609 1.2606 1.6879 1.6838 1.6838

growth rate

−0.00025 0.02025 −0.00049 0.03051 −0.00246 0.02163 −0.00246 0.03173

Table 17.4 Selected simulations: alternative runs assuming different income levels Year Assumed

Predicted Solution

Income pc Y/P ($) 0 1

  100   101

0

 1000

1

 1010

0

10000

1

10100

0

20000

1

20200

Population P

Income Y

growth (mln.) growth (mln. $) rate rate

0.01

220.0 222.2

0.01

220.0 0.01

222.2

0.01

220.0 0.01

222.2

0.01

220.0 0.01

222.2

0.01

Exog. share X/Y Multiplier m growth (%) rate

  22000   22442 0.0201

0.135 0.136

 220000

0.462

growth rate

0.00785

value 3.1977 3.1822

 224422 0.0201

0.463 0.00268 1.4903 0.610

2244220 0.0201

0.610 0.621

4488440 0.0201

0.621

growth rate

−0.00483 0.02312

1.4928

2200000 4400000

growth rate

Income

−0.00165

0.02113

1.2571 0.00035 1.2568 −0.00022 0.02023 1.2432 0.00018

1.2430

−0.00011 0.02017

344  Convergence in economic growth The convergence tendency, Ypgo > Yrgo, is thus decomposed into a part due to (X/Y)g and a part due to mg. The positive but diminishing contribution of (X/Y)g standing for an exogenous growth potential at a lower level of economic development, and an exhaustion of possibilities for exogenous growth at higher levels of economic development, dominates the negative effect of mg, standing for the diminishing multiplier effects.

6  More convergence through transfer mechanisms The analysis concentrated so far on the income multiplier effect of exogenous demand injections to sectors originating from government and the rest of the world. How significant are the income multiplier effects of exogenous changes in income transfers originating from government and rest of world and received by households, say T for transfers, and in which direction do they act? In principle, the demand injection analysis can be repeated but with reference to transfer injections to households. It is also possible to demonstrate the effects via shortcuts. Recalling eq. 2 that specified income as dependent on exogenous demand and its multiplier effect, the equation can be extended to include exogenous transfers T and their multiplier effect, m′. For example, Y = m X/Y + m′ T/Y. Dividing throughout by Y gives: 1 = m X/Y + m′ T/Y This equation can be specified for poor and rich countries as: 1 = mp (X/Y)p + m′p (T/Y)p (6.1) 1 = mr (X/Y)r + m′r (T/Y)r (6.2) Inserting average values of the above parameters for the two country groups gives the following results:5 1 ≈ (2.545) (0.356) + (2.423) (0.078) ≈ 0.83 + 0.17 1 ≈ (1.3) (0.543) + (1.648) (0.242) ≈ 0.63 + 0.27

Estimated (6.1) Estimated (6.2)

These results show the effect of X to be about 2.5 to 5.0 times that of T in determining economic growth. At higher levels of economic development, the relative strength of the X and T effects shifts from X to T. This happens via an increase in the share of T/Y (and a lower share of X/Y) as well as fewer reductions in m´ (as compared to m). At still higher levels of economic development, the increases T/Y are restricted by the same constraints that apply to X/Y. First, the T/Y share for individual rich countries had reached its ceiling in the late 1980s or early 1990s and is falling in others, c.f. Bayens and Cohen (1993). Second, the growth effect of transfers in rich countries, m´r , forms 68 per cent of that for poor countries, m´p. Therefore the conclusions reached on converging tendencies due to the X effects apply to the T effects as well. In this chapter we investigated whether the gap in the income per capita between rich and poor countries is widening or diminishing. While most empirical literature relied mainly on supply-side models of economic growth, we offered a

Convergence in economic growth  345 SAM-based demand-side model, estimated for 16 countries. Evidence from the supply side finds a conditional convergence in income per capita. Our demandside results also show conditional convergence. The SAM model predicts, after adjusting for peculiarities of economic systems, higher economic growth at lower as compared to higher levels of income per capita, which is indicative of a convergence tendency. The main cause behind this convergence is the ability of a poor country to increase significantly its shares of exports and government in its GDP, and reap growth benefits from early changes in the compositional pattern of demand; rich countries are, in relative terms, near satiation levels. The convergence tendencies apply irrespective of whether the exogenous demand impulses are done via spending in sector activities or via income transfers to households. The two mechanisms reinforce the convergence.

18 Modelling of distinctly behaving economic systems Theory and applications

1 Introduction In the past century, the traditional framework used for the analysis of economic systems reflected the ideological divide and the Cold War between two opponent regimes of capitalism and communism. There was in general more interest in the descriptive and comparative analysis of capitalism and communism. Little attention went to theorising on the formation, evolution, and differentiation of economic systems, with the exception of several basic contributions in Eckstein (1971). It is not surprising, therefore, that once the choice between capitalism and communism was no longer relevant, after 1990, the subdiscipline of economic systems came to face an identification crisis. Recently, however, there has been a revived interest in economic systems stimulated by the recognition that (a) next to convergence tendencies in some areas there are significant institutional diversifications among the OECD countries, see Jackson and Deeg (2006); (b) behavioural traits of the communist regime are re-emerging in former Soviet countries, see Beck and Laeven (2006); (c) the quest for healthy relationships between economy and polity in the context of developing countries, see Barro (1996); (d) new attempts to interpret historical developments, as in Acemoglu and Robinson (2005), and North et al. (2006). Finally, (e) the search for effective state regulation of the banking sector on the eve of the financial meltdown of 2008 has emphasised the urgency of reforming the economic system, see Stiglitz (2008). This chapter follows that revived interest. It aims at contributing to theories and empirics on the formation, evolution, and differentiation of economic systems. The models specified emphasise location and interaction of agents in different kinds of behavioural settings. In the long term, agents, and the economic system in which they interact, become aligned with and manifest a dominant behavioural type, organisation, and outcome that associate with the most dominant behavioural setting. At the macro level, different socio-economic systems can be observed to have different dominant behavioural settings and correspondingly different dominant behavioural types. We focus on three distinct kinds of behavioural settings namely those typical of the household, firm, and state settings (state settings include governmental settings at various levels and related settings). We treat economic transformations

Models of economic systems  347 within these settings, as well as inter-transactions, inter-communications, and inter-mobility of agents between these settings. We derive four broadly defined socio-economic systems, or simply social systems, that associate with these settings, display the mechanisms involved, and examine why and how different regimes evolve in different parts of the world. In separate sections we study empirical validations of the theoretical foundations and reflect on the future of prospective economic systems, and governance policy at the global level.

2  Behavioural types and behavioural settings Following Fox (1984) and Fox and Miles (1987), a behavioural setting b is defined as a physical site populated by interacting persons who have become members of the setting by accident and (or) choice. Behavioural settings relevant for economic analysis are those that generate for their participants added value from the transformation of some activity. Agents inhabiting such a setting engage in a value added transformation of goods and services, subject to institutional rules, information flows, and physical and technological boundaries. The most common examples of behavioural settings of interest for economic analysis are the household, firm, and state settings, to be denoted by b= {h, f, s}. There are more behavioural settings that are not engaged in economic transformations, and other behavioural settings that have significant bearings for economic transformations, as will become clear. A distinct behavioural type, B, characterises each of the three behavioural settings, b. Transformation processes in the household, firm, and state settings are driven by intrinsically distinct behavioural motives that are typical of the given environment that circumscribes the setting. In terms of motivation, social sharing and reciprocal exchanges are the underlying motives in household settings. Profit maximisation is the intrinsic motive in firm and market settings. Political returns and rent seeking are the intrinsic motives in state and related settings. The external environment is the basic determinant of the nature of a behavioural setting. Different external environments generate typical coordination structures that coincide and fit with typical behavioural settings. For example, a closed world, homogeneous population, strong kinship, severe scarcity of resources, and low levels of material welfare characterise the external environments of household settings. This external environment promotes sharing behaviour and solidarity structures. Agents in these settings become conditioned towards adopting institutions of familial altruism, brotherhood attitudes, and income sharing among households linked by kinship and location. Households unrelated by kinship or locations tend to be excluded from the sharing tradition. Familial altruism is thus double edged, with inclusive and exclusive dimensions. In contrast, the external environment typical of firm settings is materially better off, and is characterised by an open world with frequent changes, product discoveries, and choice opportunities, and a high mobility of agents. Finally, the external environment typical of state settings is characterised by highly skewed human endowments and rank among differentiated population groups, often

348  Models of economic systems generating conflicting interests and requiring authoritarian rules to resolve them. It contains also barriers that obstruct mobility, access, and choice. The three distinct behavioural motives can be modelled as is done in eqs. 1, 2, and 3, below.1 In the household setting, the agents lump together their benefits and costs in an effort to make total benefits exceed total costs. In eq. 1, V h stands for the value added in the household setting, while rewards R and efforts E of agents i and i′ are lumped together and somehow shared among all i. The agents would strive to distribute these benefits and costs between i and i′ in ways that contribute to a maximum value added, or at least a positive result for the whole setting. The resulting distribution can be affected by personal and relational circumstances. V h ≡ Ri + Ri′ − Ei − Ei′ ≥ 0

(1)

In the firm setting, each agent would like to realise the highest positive returns to him or herself. In eq. 2, Vf stands for the value added in the firm when agents i and i′ maximise their relative returns, defined as benefits less costs per unit of capital invested; the latter can be approximated by taking multiples of the total costs, or to simplify things we set the total costs as the denominator. The resulting income distribution is likely to show returns of one agent higher than the other. Vf ≡ Σ (Ri − Ei )/Σ Ei ≥ ρ

(2)

To model the state setting, we employ variables R, E, and V, with subscript ps for the pre-state setting, and with subscript s when the state setting sets engage. We also employ k = 1,…,K to represent agents with state authority. Eq.3.1 shows a higher value added in the state setting compared to the pre-state setting. This is due to a reorganised transformation with intervention of state agents k that results in Σ Ri > Σ Rpsi and (or) Σ Ei + Σ Eik < Σ Epsi. Part of Σ Ei is a privately incurred cost and the other part is the collectively invested expenditure that allows for the higher value added transformation. Vps ≡ Σ Rpsi − Σ Epsi ≤ 0,    Vs ≡ Σ Ri − Σ Ei − Σ Eik ≥ 0

(3)

Agents in the state setting, k = 1,…, K, acquire an authority to extract a remuneration from all other agents, expressed in terms of their effort, and denoted by Eik , such that the average remuneration for k is higher than the average level of net benefits left over for agents i,..,I. Distribution of incomes will manifest, on average, a higher level for the authority agent k than for subordinate agents i. The coordination mechanisms in the three settings are distinctly different. The coordination mechanism in households is typically sociological in character, in firms coordination is economically motivated, and in state settings coordination is politically motivated. In any country there are households, firms, and state settings coexisting in large numbers side by side. The same agents can be members of more than one setting simultaneously. Agents communicate with agents within their own settings and other settings. The interactions are given shape in Figure 18.1. The squares, triangles, and circles refer to the three behavioural settings, each with

Models of economic systems  349 its own members; the engagement lines linking them indicate transformation and mutual exchanges taking place among agents in or between the organisations, as well as communicated behaviours. Each engagement line can be interpreted as consisting of a large number of bits of exchanged transformations and communicated traits. Such an engagement line can be denoted by Gibi′b′. As these bits are not uniform in intensity in terms of time, effort, or effects, they can be normalised making use of some scale of engagement intensity, which can be denoted by N, and expressed in terms of time, effort, or effect. The engagements weighted by intensities can be expressed by Σib (Gibi′b′ . Nibi′b′ ). This term can be eventually divided by the sum of all engagements, weighted by intensity in the whole economy, to give a relative measure of the strength and dominance of the engagement lines. They can be drawn lightly or heavily so as to reflect relative strength, as shown in the figure. It can be added also that settings can relate to each other horizontally and vertically, just as organisations do. A setting generates material and immaterial outcomes that are distributed as material and immaterial rewards to its members. The distributed rewards in competing settings are crucial for an evaluation that participating agents regularly do, and which guides them in their decision to continue in the setting, voice, or exit, and enter another setting.2 The propensity to move and participate in alternative settings satiates when the marginal utility of the agent from shifting a unit of effort between settings is equal to the marginal cost of the shift. The engagement lines in the figures can accordingly be given an additional meaning: they express agent mobility between alternative settings. Processes of exchanged transformations, communicated traits, and agent reallocations over lengthy periods lead to relatively greater concentrations of agents in one kind of behavioural setting b rather than others b′, opening the way for the spread and dominance of the behavioural type B that coincides with behavioural setting b. Once a threshold is reached with regard to accepting a specific behavioural type B, this B can be expected to gain momentum in view of network externalities, and will spread further and subordinate other B′. The adoption and spread of a particular behavioural type among more agents was studied in many contexts, and there are well-known relating mechanisms in the literature.3 A conglomeration of interacting behavioural settings having a common dominant core is very likely to evolve into, and be identifiable as, a social system. The structure, conduct, and outcomes of the social system will tend to be relatively homogeneous and reasonably predictable. Each social system has its own integral economic aspects that together describe its economic system. Similarly, each social system has its own integral polity. It is logical to expect high degrees of consistency and correlation between the economy and polity of a specific social system.4 What makes a network of interactions comprehensible as a distinct system is the prevalence of common revealed preferences and typical coordination tracts, structures, and performances as can be observed in countries considered to have adapted to that system, as compared to other groups of countries with a different behavioural focus. Five factors are behind how the common is formed, and how the prevalence is caused: (a) sharing of common external environment and past history fosters convergence towards a common behavioural type; (b) intensive

350  Models of economic systems and extensive interactions and communications of agents participating in more settings extend the prevalence of the advantaged behavioural type; (c) agents observe the transformation outcomes in alternative settings, and move to the advantaged setting or copy its behaviour, thus resulting in the prospect that the typical behaviour of the advantaged setting becoming prevalent; (d) network externalities enforce convergence towards the advantaged behavioural type. Further, (e) when a behavioural setting b happens to stand higher than b′ in the hierarchy of settings, then b is also able to set behavioural rules typical of b that other settings b′ would follow. In this way, behavioural type B overrides B′, allowing a further dominance of B on B′. However, the pulling forces encounter spatial, cultural, and political limits that hinder the emergence of one integral system for the whole universe of interacting entities. The interplay between pulling and pushing forces can result into the coexistence of differentiated equilibriums.

3  Prototypes of dominant behaviours in economic systems Agents adopt the behaviour of the setting that they inhabit most and in which they interact most, and spread it to other settings they communicate with, or partially inhabit.5 Sharing common external environments among related behavioural settings, communication in and between settings, mobility among settings, and network externalities, tend to end up in a social system with an integral economic system allied to it that would manifest one dominant behavioural type B over others B′. Given the three prominent behavioural settings examined, large numbers of agents inhabiting each of them, vast volumes of transactions and communications in each and between them, and the pulling forces as well as the pushing forces, it is conceived that over long periods of time the three broad prototypes of economic systems would become dominant in different parts of the world. The first prototype, and the oldest, is the economic system that circles around households and in which all settings have adapted to household behavioural traits. This can be called the household-intensive system, HIM, as in Figure 18.1a. In the real world, many rural regions within developing countries would qualify as HIM. At the country level, there are limited examples that fully operate along the lines of HIM. The second prototype, as in Figure 18.1b, is the economic system where agents adopt a firm-like behavioural type, that is, maximisation of material returns at least material cost. The firm-intensive system, FIM, has many copies in the real world; the best example is the United States. The third prototype, as in Figure 18.1c, is the economic system where agents have adapted to a state-like behavioural type that is guided by rent appropriation and political returns. In the real world, Russia is a close example of countries that operate along the state-intensive system, SIM, though this was more so during the communist regime. The ability to predict the conduct and performance of the economic system of a specific country is greater the more uniform and integrated is the economic system in that country. For example, the modelling and analysis of conduct and

Models of economic systems  351 Households

Firms

State

(a)

Households

Firms

State

(b)

Households

Firms

State

(c) Figure 18.1  Configurations of three socio-economic systems: (a) HIM, (b) FIM, (c) SIM

performance in FIM relating countries along lines of profit maximisation, and in SIM relating countries along lines of rent appropriation, can be seen as workable approximations made possible by over majorities of the agents behaving uniformly along these two distinguishable lines in the two systems, respectively. To drive home the point, at the cost of some exaggeration, it can be maintained that in the ‘US’ the high concentration of agent interactions in firms pushes intrinsic motivations in the household and state settings aside and become replaced over time by profit maximisation, typical of the firm settings. In contrast, the same processes oblige agents in household and firm settings in ‘Russia’ to follow a politicised motive, typical of state settings. As a result, all three settings in ‘US’ behave in

352  Models of economic systems ways typical of firm settings, while in ‘Russia’ they manifest behaviour typical of state settings. In the extreme, a comparison between the economic systems of ‘US’ and ‘Russia’ is a comparison of two contrasting behavioural types that involves the political, and the whole social system. In ‘US’ the economic motive dominates, and the polity can be described to have adapted itself to the economic motive. Next to constitutional checks and balances, and an independent judiciary system, that keep state discourse in control, profit-maximising firms and agents have installed more institutions for controlling state conduct, and in some cases bending the polity to realise economic interests. In contrast, in ‘Russia’ the polity can be seen as exogenous to, and determining the economy.6 There are arguments for drawing up a fourth configuration. Specific conditions exist that may hinder convergence towards one dominant behavioural type. Where the absorption of agents from households in firms or state is limited because of the sheer large numbers involved, as in China or India and several others, the result is a loosely linked multipoles system, MPM; see Figure 18.1d for a sketch of MPM. These countries have vast rural populations that are bound to household settings, but also significant urban populations manifesting subcultures relating to the firm and state behavioural types. The distribution of agents on the three settings has been historically stable, and given the involved magnitudes the distribution may not change much in the future.7 In this multipolar environment, the need to streamline the vast heterogeneity of agents has enhanced the significance of what can be called persuasion settings. These include political congresses, judiciary courts, and religious, intellectual, and

Urban households

Firms

Urban/rural administration

Rural households

Figure 18.1(d)  A sketch of the multipoles system (MPM)

Models of economic systems  353 media channels and, not least, combined firm-state ventures that play a crucial role in the streamlining of consensus and development in a segmented system. Persuasion settings are exclusive settings. Participating agents are highly talented leaders who are able to place themselves as leaders in various contexts: household, firm, state, and religious, intellectual, and judiciary settings. They are the so-called wise men, and they are able to obtain the support of leaders who lead different settings. They have the natural authority to affirm the status quo and anticipated changes. To the extent that agent distribution among these segments in these countries will continue to keep a stable balance in the future, it can be expected that persuasion settings will increase their leverage significantly in this environment.8

4  The start and the long-range development of economic systems If at the end of the road three, or four, broadly defined economic systems show up, then how did each start at all at the very beginning, and develop further? The starting point is conveniently the situation where household settings are already there. Which settings followed: firm settings or state settings? Hicks (1969) is among the many economists who contend that tribal, feudal, army, and state settings historically preceded firm and market settings. Most anthropologists, like Firth (1967), maintain that economic organisation and the transformation and exchange of products within and between neighbouring primitive societies came first and preceded political organisation. There is documented history that can be used to support both scenarios in different regions in the very far past.9 Our model of long-range systemic development allows for both scenarios. The model is basically microeconomic, and is based on several premises. 1

Agents inhabiting household settings—and interacting with each other in a communal environment—experience specific needs, or can be attended to uncovered needs, and they are ready to embrace these needs. Agents are also innovative and capable of finding solutions and creating new transformation settings that satisfy these needs. 2 The needs of an agent can be personal or collective.10 Personal needs of one agent can be directly satisfied by the innovative and responsive transformation offered by another agent, opening ways for exchange and firm settings. Collective needs are indivisible and their satisfaction would require a joint effort of agents and a third-party intervention to coordinate the task at a higher level; which opens ways for authority and state settings. 3 In communities where personal needs happen to have an overweight, the firm setting will emerge and prosper. In communities with an overweight of collective needs, the state setting will emerge and prosper. 4 The model assumes, tentatively, that the external environment is kept constant and refines this stepwise. Communities where personal needs had overweight, thus pushing the economic system towards adopting a firm-like behaviour, happen to be more

354  Models of economic systems of the open type. In an open community, agents search more frequently for new products. These are matched with talented and skilled agents who are capable of innovating and achieving a value added transformation in the form of a new product that satisfies some felt personal need. The newly introduced transformation is by itself a new setting in formation. The larger the community and the more open it is, the greater is the probability that more personal needs, talented transforming agents, and new transformation settings will emerge in the form of firms and markets. As demand for personal goods diversifies and increases, more agents reallocate from households to transforming firms and markets, and thus setting into motion the decline of the income-sharing behavioural type in favour of the profit-maximisation behavioural type. The case of communities where collective needs have overweight allows for the emergence and prevalence of state settings. Collective needs consist of urges to curb uncertainties and externalities that the agents experience. Unable to resolve their uncertainties and externalised divisions, agents i and i′ call on a ruling agent k to organise security and to execute public actions to satisfy collective needs. In all societies there are talented agents who are well equipped to take the role of ruler k, lead politically, find compromising solutions to collective needs, and govern effectively towards realising these solutions. Having a governing monopoly, agents k pursue an economy of measured exploitation in fixing their compensation for the services they provide to agents i, i′. If all the counteracting i, i′ agents calling for collective action count I and each produces y, and the production is liable to fall to x if no collective action is taken, but may increase to z if the state intervenes, whereby x < y < z, then I (z − x) represents the maximum amount that agent k can collect from agents i, i′ for the collective actions rendered, including the bureaucratic apparatus required to implement these actions. If the number of agents k involved in the collective actions is K, then their average reward is I (z − x)/K. Under crude but realistic assumptions agent k can be shown to end with about double the income of agent i, i′ or more.11 Under leadership of politicians and bureaucrats, governments have evolved over time into natural monopolies.12 Anywhere, there is only one government to run state affairs. This monopoly position seduces state officials to appropriate rent in the process of organising and executing collective actions in the future. State officials are also politically motivated to create and expand community needs for collective goods. In this way, the received rent is extended to more activity areas, and the monopoly is sustained and broadened. To create community needs for collective actions, political man would take measures to discourage private solutions to emerging needs. The above does not mean that all state agents behave along rent-appropriating motives. Many, or most, state agents will pursue benevolent motives as when the state is embedded in a household-intensive system, HIM. Similarly, many, or most, state agents will seek no more than their opportunity cost if they are embedded in a firm-intensive system, FIM. But if they function within a system that is intensively dominated by the state, SIM, then state agents will excel in rent seeking and political behaviour, and cause this behavioural type to spread to other settings and converge towards SIM.13

Models of economic systems  355 Finally, persuasion settings play an important role in affirming the principles of a newly installed socio-economic order, whether it is one where firm settings dominate or one where state settings dominate. If persuasion settings, being the highest, affirm the socio-economic order, then prevalence of a stipulated behavioural type in the economic system is complete.14 To summarise, alternative scenarios are feasible regarding sequence of regimes. It follows therefore that it is perfectly plausible that the state (firm) settings emerge first and firm (state) settings follow later if collective (personal) needs were felt to be more urgent than personal (collective) needs. The opposite evolutionary path is conceivable elsewhere. So both routes are possible, and these can vary over time and place. The breakthrough towards a predominance of firm or state settings would then depend on the cumulative relative weights of personal versus collective needs. In the real world and over time, external environments have changed appreciably and have brought forces that promoted the dominance of FIM or SIM, at the cost of HIM. A few examples suffice: (1) the age of discovery has allowed a major diversification in demanded products at the personal level, a huge jump in personal needs of the newly settling migrating populations, and a push for firm settings; and (2) technological advance has enhanced demand for new personal products, and their divisible production encouraged firm settings. Both examples explain the early inclination of the United States and Europe towards FIM. On the other hand, (3) frustrations with an unfair and insecure social order encouraged calls for authoritarian rules, and (4) military conflicts with neighbouring countries and defeat promoted growth of state settings. Both examples explain the inclination of Russia to SIM after WWI, and a stronger state in Germany between WWI and WWII. In principle, there are conceivable conditions under which the two orders can temporarily (or permanently) reverse. If in specific circumstances collective needs overshadow personal needs there will be a strong demand for state settings. Similarly, disappointment with achievements of state settings and rising demand expectations based on better attainments in firm settings can shift the overweight from collective to personal needs, giving some push towards firm settings. A pronounced example is that of Russia that experienced both reversals in the twentieth century. Other examples of modest realignments between the two orders are inherent in deregulation and regulation waves in the United States and other FIM-related countries: deregulation of the utilities and communication sectors in the 1990s, and regulation of the finance and banking sectors following the financial meltdown of 2008. With the above intervening factors in consideration, the emergence and prevalence of specific behavioural types in terms of the satisfaction of personal and collective needs can be conceptually modelled below, other things remaining the same. The probability that a behavioural type identifiable with setting V b prevails over all other V b′ is expressed as in eq. 4. V b prevails if (ω1 Ab + ω2Cb ) ≥ ν (4)

356  Models of economic systems In this equation, there are two share parameters that affect prevalence. Ab is the share of agents in setting b, with respect to all agents in all settings. Cb is the share of commodities demanded that are most suitably transformed in setting b, with respect to all demanded commodities. The personal commodities are most suitably transformed in firm settings. The collective needs are most suitably transformed in state settings. Eq. 4 proposes that the greater the shares of those agents and commodities associated with a particular setting, the greater is the probability that the behavioural type underlying this setting prevails over other behavioural types. In this equation, ω1 and ω2 are weights applying to these two shares, whereby ω1 + ω2 = 1.15 In this equation ν is a proportion that represents a critical mass. Once the practice of a particular behavioural type reaches this critical mass, this behavioural type can be expected to benefit from network externalities and to extend its maintenance to practically the whole population. There are different views concerning the likely value of the critical mass. Values of two-thirds and threequarters are among the most quoted in the literature relating to a critical mass.16 There is thus justification for fixing the value of ν at around 0.7. The two shares and their weights, as depicted in eq. 4, form an Index of Interactive Influence, which is indicative of the assertive power of setting b over all the other settings. Quantification of this index for individual countries would give the systemic orientation of the countries as to which setting is most dominant. We apply the index later in Section 6 to explore for the world at large the relative dominance of prospective economic systems that associate with leading countries. An important question is what are the effects of an otherwise neutral economic development process on the likely paths of the share of demand for commodities, Cb, and the shares of agents Ab for each setting b? The likely paths would give extra shape to the economic system. Attention can now be directed to examining the likely paths of the two shares. To start with the concentration of commodities, define Cb as the proportion of demanded commodities that are most suited to be transformed in setting b, (call them Db), in total demanded commodities, (call them D). As a result Cb = Db/D. Applied to the three settings we have therefore the share of demanded personal commodities that happen to be most suitably transformed in firm settings Cf = Df / D, and the share of demanded collective commodities that happen to be most suitably transformed in state settings Cs = Ds / D, so that the share of demanded domiciled commodities that happen to be most suitably transformed in household settings becomes Ch = 1 − Cf − Cs. The evolutions of the commodity shares of Cf,t and Cs,t over time t depend on the growth rates of per capita demand for the f and s commodities: call them γf and γs, respectively.17 As we stated earlier, the shares can shift abruptly due to exogenous changes in the external environment. On the evolution of the commodity shares Cb, the following is noted. Ch tends to fall with development over time in favour of Cf and Cs, as demands for personal and collective commodities overtake the demand for homemade household production. However, the takeover is tempered if the per capita growth rates of demand for commodities suited for production in f and s settings, γf and γs, are low, due to a high population growth and (or) slow economic growth.

Models of economic systems  357 Next, considering the concentration of agents, define Ab as the proportion of the labour population in setting b, (call them Lb) in the total labour population, (call this L), so we have Ab = Lb / L. Applied to the three settings we have thus the share of agents in firm settings Af = Lf / L, and the share of agents in state settings As = Ls / L, so that the share of agents in household settings is found as residual, Ah = 1 − Af − As. The evolution of shares of agents in firm and state settings Af,t and As,t over time t will depend on the growth rates of per capita demand for commodities relating to personal and collective needs, γf, γs, the growth rates of labour productivity in firm and state settings where commodity transformations take place, denoted by λf , λs, and the growth rate of the total labour force, or the population at large, denoted by π.18 The following can be noted on the evolution of the agent shares Ab. 1 The share of agents in household settings Ah tends to decline, and those of firm and state settings, Af and As, tend to increase with development over time. However, if the per capita growth rates of demand for personal and collective commodities γf and γs are low, due to a high population growth and (or) low economic growth, then the effects just mentioned are tempered. 2 While there is disputed evidence that γf > γs, to the extent that this is true then firm settings have a greater probability to spread than state settings. 3 It is generally established that the growth rate of labour productivity is higher in the transformation of personal goods as compared to collective goods λf > λs; this tendency favours a relative rise in the agent share of state settings as compared to firm settings. 4 Finally, a high population growth rate π sustains the prevalence of Ah, while low population growth enhances the shares of Af and As.

5  Empirical validation Empirical validation of the outlined theory of economic systems is a matter of bits and pieces. First, there are no real-world observations of economic systems in the conventional sense. There are countries and groups of countries that can be assigned as having this or that economic system prototype; but these representative observations are short of the absolute observations that are commonly used in most other empirical validations; and thus, tight proofs are not feasible. In this section, the empirical validation will start with positioning of various country regions across the three axes of economic systems. Second, the Index of Interactive Influence measured across alternative settings for a particular country would indicate which setting is dominant in that country. However, construction of the index would require micro data on the intensity of agent interactions and the size of commodity transformations by setting, which are not readily available, although there are some macro data that can be utilised. In these circumstances, the verification will have to be indirect in the sense of showing that a particular country manifests typical features of the economic system to which the country is assigned.

358  Models of economic systems First, Figure 18.2 lays out a positioning of country groups along the discussed axes of economic systems. The study of a large number of empirical indicators on firm and state settings, Cohen (2009), shows the United States to be closest to FIM, while Japan and West European countries, also identifiable as FIM, show differing inclinations to the other two poles. The indicators show Russia to fit most to SIM, with the former Soviet Republics and East European countries also manifesting SIM but showing differing inclinations to the other two poles. As can be expected, developing countries show significant differentiations by region. Second, we display several indicators that support the above positioning of countries, and explain their conduct and performance in terms of the displayed theory. For more indicators and analysis see Cohen (2009). To start with, the World Value Survey, taken around 2000, showing country differences in attitudes of agents towards the household and the family, as well as towards firms and the state, gives support to the proposed positioning of countries. Figure 18.3 gives the percentage of responses that see family ties as ‘very important’. The average for ‘very important’ of all country groups is 89.1 per cent. FIM and SIM countries are below this. The developing regions are above this average supporting their positioning nearer to the HIM system.19 The same World Value Survey reflects on attitudes towards firm and state settings. Figure 18.4 gives on the left side the percentage of respondents that believe private ownership of business should be increased, and on the right side the percentage of respondents in the interviewed sample that believes government ownership of business should be increased. The contrasting responses in FIM- and SIM-related countries are obvious, and are consistent with our positioning of countries along the axes of the distinguished economic systems. The empirical validation can be elaborated further. With respect to FIM countries, various indicators can be recalled that support the central positioning of the United States on the dimension of the firm-intensive economic system, while placing Japan and Western Europe respectively to the left and right of the United States. The positioning reflects a sharing-inclined economy in Japan, and a control-inclined economy in Europe. In Table 18.1, an index based on ten

South Asia SA HIM sub-Saharan Africa SSA Middle East Nr. Af. MENA Ex-Soviet republics EXSR

US

Japan Latin American Caribbean, LACA

FIM

India China SIM Russia

Western Europe WE Baltic, Central, Eastern Europe CEEE

Figure 18.2  Positioning of economies along axis of dominant systemic interactions

FIM

87.7

SIM

67.9

EAP

88.9

SA

93.7

MENA

95.0

GCC

94.7

SSA

94.4

LAC

90.6 0

10

20

30

40

50

60

70

80

90

100

Percentage respondents reporting family as very important, average dashed line = 89.1%

Figure 18.3 Importance of household settings: country groups ranking of importance of family Source: Compiled from World Value Survey 1999-2000, at. http://www.worldvaluessurvey. org. Country data were averaged to give the level for regional groups. Abbreviations: EAP = East Asia and Pacific; SA = South Asia; MENA = Middle East and North Africa; GCC = Arab Gulf countries; SSA = African countries; LAC = Latin America and Caribbean

18%

82%

US+

24%

76%

WE Japan+

70%

CEEE EXSR

50%

30%

43%

MENA

55%

GCC SSA

36%

45%

55%

60%

40%

49%

51%

47%

53%

50%

Russia EAP SA

64%

57% 45% 38%

62%

LAC 0

10

20

30

40

% for private

50

60

70

80

90

100

% for public

Figure 18.4 Pro-firm and pro-state attitudes Source: Compiled from World Value Survey 1999-2000, at http://www.worldvaluessurvey. org. Results are for national sample surveys, around 2000. Definitions: (% for private) = Accumulated percentage of respondents in the interviewed sample that believes private ownership of business should be increased (% for public) = Accumulated percentage of respondents in the interviewed sample that believes government ownership of business should be increased.

360  Models of economic systems indicators of business competitiveness (including business, trade, fiscal, government control, monetary, investment, financial, property rights, corruption, and labour mobility) shows for 2006 a greater ability of agents to compete in the United States than in Europe, 82.4 compared to 70.9. The index is evidence of a more firm-intensive economic system in the United States than in Europe. The index shows also Japan to be closer to the United States than Europe. Table 18.1 contains another indicator that shows lower public shares in the United States and Japan than in the European Union, being maintained over the last four decades. The figures support the proposed differentiation in the positioning of FIM related countries.

Table 18.1 Indicators reflecting degree of firm intensity in FIM-related countries

Index of business competitiveness Government expenditure/GDP, 1960–2000, %

United States

European Union

Japan

82.4 31.1

70.9 44.4

74.5 27.3

Sources: The Heritage Foundation (2000): Index of Economic Freedom, published by The Wall Street Journal and The Heritage Foundation, New York. And, OECD (1964–94, supplement): Expenditure and Revenue Statistics of OECD Member Countries, OECD, Paris

Turning to variations within SIM-related countries, Russia was positioned as closest to the state-intensive economics, with former East European satellites, BCEE, slightly leaning to firms, and the other former Soviet republics, specially the Asian Islamic republics, linked more to household settings. This orientation is supported by the indicators in Figures 18.3 and 18.4, which show less positive attitudes of agents towards firms in Russia compared to BCEE, and more sympathy with state settings in Russia compared to BCEE. These results are particularly significant in the context of the economic transitions through which these countries are passing. The micro foundations of system formation which was displayed in the previous section suggest that the scope of the required change in agent behaviour, functional for a transition from a state dominated system to a firm dominated system, requires much more than introducing market-oriented institutions and policy reforms. It comes to converting the most commonly practised rent-appropriation behaviour of agents in the SIM environment into a profit-maximising behaviour that fits with a FIM environment. An increasing intensity of agent interactions in firm settings over state settings is an absolute requirement for a shift in agent behaviour from rent appropriation to profit maximisation. The implication of our premises is that for those transition countries, where relatively more agents were inclined to behave more along profit-maximisation lines, these countries would be more able to achieve the conversion from SIM to FIM, than transition countries where relatively more agents are absorbed in state settings and generally speaking practise rent-appropriating behaviour. Our

Models of economic systems  361 hypothesis is that if the BCEE group shows more ability to convert than Russia and other EXSR, then this is reflective of and is primarily due to a FIM > SIM orientation of agent behaviour in BCEE, as compared to a SIM > FIM in Russia and EXSR. The above is borne out in several indicators in Table 18.2. First, the greater fall in the GDP and the slower pace of transition and recovery in Russia and EXSR compared to BCEE (fall in GDP of 47 per cent and 55 per cent compared to 32 per cent, respectively) can be interpreted in terms of a greater dominance of state structures in Russia and EXSR than in BCEE. The repercussions of the breakup of the centrally planned system are least in BCEE. Second, the greater state dominance explains the lower score on the index of business competitiveness for Russia and EXSR compared to BCEE; that is, 50.3 and 52.3 compared to 61.5. Third, the same explains the higher occurrence of insider governance in privatised companies in Russia compared to BCEE. Fourth, an index of global integration shows less success in Russia and EXSR than in BCEE. Two other relevant indicators show the significance and greater influence of the state in running the economy in Russia: one is the extent of institutionalisation of the rule of law versus state discretion, and the other is the share of firms paying bribery tax. All these show the SIM to be more prevalent in Russia and EXSR than in BCEE. Table 18.2 Structure, conduct, and performance in transiting economies reflecting differentiated dominance of state settings BCEE EXSR Russia GDP decline = lowest year − 1988/1988, per centa Index of business competitiveness, per centb Share of privatised firms ending with insider governance, % c Index of global integration 2000–05, per cent d Rule of law index (10 = best, 0 = worst)e Share of firms paying bribe tax, per centf

−31.5  61.5  72  54   7.2  42.4

−54.9  52.3  40   4.4  72.4

−46.5  50.3  90  23   3.7  64.6

Sources: a = International Monetary Fund, International Financial Statistics, World Economic Outlook, b = The Heritage Foundation (2000) Index of Economic Freedom. c = Adapted from World Bank (1997), p.53. d = The index is obtained from two indicators: share of foreign merchandise trade in GDP, and share of FDI in GDP. For each indicator country figures for 2000 and 2005 are combined to give country average values for the period 2000–05, which are then expressed as proportions of the highest country average value. The resulting proportions of the two indicators are then averaged to give a final score (data from World Bank at http://devdata.worldbank.org/query). e = Wall Street Journal. The figures are also quoted in Hoff and Stiglitz (2002). f = WB/EBRD The Business Environment and Enterprise Performance Survey (BEEPS) 2002.

6  On the future outlook for economic systems After theory and validation, one can ask what next? Given the ongoing globalisation and increasing intensity of agent interactions across countries and systems, are there grounds for expecting an evolution towards one system or a synthesis of systems? The question can be approached from two angles. The

362  Models of economic systems first approach—winner takes all—would identify the prototype that has the highest economic performance and would predict for the long run a convergence of competing systems towards the winner prototype. The second approach that can be called dominance calculus states that the answer to this question is not independent of the future outlook for contemporary leading countries that align with particular economic systems. The first approach asks which system performs better. While it is generally agreed that a household-intensive economic system is less capable than firmcentred or state-centred economic systems in producing larger valued transformations, the question whether a firm-intensive or a state-intensive economic system, in their pure forms, would perform better economically has been controversial for more than a century. One line of thinking is that competing firms in the firm-centred economy would maximise their profits to the point where inefficient profits would be eliminated by more entry and exit of firms. As they maximise their profits, the firms lead the economy to its highest performance. The other line of thinking is that if the state in a state-centred economy would calculate its exploitation margin well, it would impose the right margin for itself; that is to say, it may not reduce the size of the cake and the ultimate margin as well, but instead enhance both. In the real world there are no pure forms of FIM or SIM, and the two systems are confronted by market failures and polity failures, respectively. The validity of either of the two standpoints depends on (a) the condition that market failures due to indivisibilities, uncertainties, externalities, and maldistribution, are collectively remedied in FIM-related countries, and (b) the condition that polity failures due to uninformed or exploitative rulers are collectively avoided in SIM-related countries. There is an empirical record over several decades on the comparative performance of FIM- and SIM-related countries. The empirical conclusion, in the light of the above considerations, is that there has been a conditional edge for FIM-related economies to perform better than SIM-related economies over longer periods of financial stability and sustained growth. The empirical evidence on economic growth for the years 1950 to 1990 for FIM- and SIM-related countries shows that the growth in FIM countries was higher than in SIM countries, with the exception of the first ten years.20 There is a fourth system, MPM, driven by consensus decisions in persuasion settings. What is the score of this system regarding performance? Because decision making in the MPM system is not done by one dominant actor but by consensus and persuasion among affected and benefiting parties (for example, firm(s), community and government leaders confer), it is more likely that more attention is given to direct and indirect effects in making decisions; and can thus lead to better decisions and more rewards to all. However, there is an involved risk of emerging closed clubs; but this risk is minimised as the reputation of participants in persuasion settings matters and their further mobilisation depends on their past success. A less risky performance of MPM compared to FIM is also empirically plausible.21 If the MPM is hypothesised to be superior to FIM, the implication is obvious: agents in the FIM will

Models of economic systems  363 be better off if they adopt or imitate MPM, predicting an evolution of FIM towards MPM. It remains true, and is generally recognised, that establishing empirically the superiority of a specific socio-economic system via evaluation of system-related country performance is full of identification difficulties, because of external interventions and disruptions that distort a comparative framework. The second approach centres on dominance calculus. This approach is well equipped in reflecting on futures economic systems. Systems are carried, defended, and disseminated by particular countries. If country x dominates the world economy, then the economic system of country x is more likely to spread and prevail in other countries. In other words, what is the future outlook for the relative influence of competing economies? We have postulated that the dominant setting is determined by: (1) the relative distribution of agents, and (2) the relative distribution of transformed value added in competing settings. The two relative distributions are the shares denoted by Ab and Cb, respectively. We suggested taking an average of the two shares, thus using equal weights, in applying an Index of Interactive Influence. Trustworthy projections of both distributions for a few coming decades are readily available and are used to calculate the index in Table 18.3 (see the sources below the table). Table 18.3, projecting the relative influence of the main leading countries in the year 2050, shows two main results. First, there would be a reduction in the interactive influence of the United States, the European Union, and Japan, who represent the firm-intensive system; and some marginal increase in the interactive influence of Russia that is closest to the state-intensive system. But the significant gainers in interactive influence are China and India, representing a multipoles-based system.22 The expected changes in relative influence will not pass unmarked on the other interacting economic systems. Some typical features of the social systems in China and India will influence, be passed to, or be adopted in other social systems. Second, the table shows that the Index of Interactive Influence varies at around 20 percentage points for any country and its related system, confirming that a dominance of one any one country-system is excluded. The FIM configuration consisting of the United States, the European Union, Japan, and a few others is reduced to some 23 points. China holds at 21 points and India at 17 points. The index predicts a balanced distribution of power between leading countries and their relating systems. The probabilities of dominance or convergence towards one global system are very remote given the values of the Index of Interactive Influence that do not exceed the 20 per cent for any particular country system in any year. The future situation would most likely lead to a new systems competition that is already taking shape. The question is then whether the new systems competition will be non-cooperative or cooperative. Mixed signals are observable on this issue. At a more abstract level, it is generally valid that when the contending parties have influential powers that are more or less equal, and perceive the situation as such, the parties are more inclined to use reason and knowledge and adopt cooperative attitudes in resolving frictions between them. Under a skew distribution of influential powers it is more likely that a non-cooperative attitude emerges. Table 18.3

Table 18.3 Future outlook of major countries as reflected by the Index of Interactive Influence based on country shares in the world totals with respect to population and GDP 2000

2050

Population GDP USD Index of Population GDP US million billion Interactive million dollars Influence billion World total

6,124

31,800

Per cent distribution United States 4.7 European Union 6.9 Japan 2.1 Russia 2.4 China 20.7 India 17.1 Rest of the world 46.1 World 100.0 FIM: US, EU, Japan 13.7

30.7 25.3 14.6 1.2 3.4 1.5 23.3 100.0 70.6

17.7 16.1 8.4 1.8 12.1 9.3 34.7 100.0 42.2

9,191

170,721

4.4 4.8 1.1 4.2 15.3 18.0 52.1 100.0 10.3

20.6 10.4 3.9 3.4 26.0 16.3 19.4 100.0 34.9

Index of Interactive Influence

12.5 7.6 2.5 3.8 20.7 17.2 35.8 100.0 22.6

Sources: Population figures are from UN Population Division at http://esa.un.org/unpp GDP figures for 2000 are from World Bank at http://devdata.worldbank.org/query GDP projections for 2040 and 2050 for the individual countries, expressed in constant price of 2003, are from Wilson and Puroshothaman (2003). We used their projected aggregated growth path for France, Germany, Italy and the United Kingdom to obtain the projections for the European Union, which consists of the 15 Western European countries. The projections for the Rest of World Group are from Fogel (2007). The projected world total for the GDP is thus obtained by summing the regions, and the percentage distribution by region is calculated. The Index of Interactive Influence in column 3 = (col. 1 + col. 2) /2. Column 6 = (col. 4 + col. 5) /2.

45 40 35 30 25 20 15 10 5

US

EU

China

2050

2025

2000

1975

1950

1925

1900

1875

1850

1825

1800

1700

1600

1500

1400

1300

1200

1100

1000

0

India

Figure 18.5 Displacement tendencies with significant effects on the future dominance of economic systems Note: The vertical axis denotes the percentage share of a country’s share in the world GDP. The horizontal axis denotes years. Years 1000 to 1975 are reported in Maddison, A. (2003) The World Economy: Historical Statistics, OECD, Paris. The year 2000 and forecasts for 2040 and 2050 are from Table 18.3.

Models of economic systems  365 predicts a future world in 2050 with a much more equal balance of power than in 2000, and thus feeds the expectation that the new systems’ competition ahead will be more of the collaborative than the non-collaborative type with a greater role for persuasive settings in coordinating and streamlining global governance.

7  Concluding remarks The distribution of agents in contrasting behavioural settings, such as household, firms, and state settings, and their interactive participation, produce over long periods highly contrasting economic and political systems that coincide with distinct behavioural prototypes, observable in many countries. The chapter laid down a micro–macro conceptual framework that outlines the formation of prototypes of economic systems, and ways for measuring dominance and convergence towards behavioural prototypes, Making use of differing compositions of personal and collective needs, the chapter explored the paths of different economic systems since early development that associate with satisfaction of personal and collective needs. The economy-polity configurations of countries like the United States and Russia are reasoned to be outcomes of firm-intensive and state-intensive economic and political transformations, with overweight of personal and collective needs, respectively. The positioning of more countries across the axes of household, firm, and state settings is followed by empirical validation. The remarkably big sizes of the segments relating to rural households, urban households, firms, and state settings in countries like China and India assures us that these segments keep a stable balance to each other. Persuasion settings play a greater, crucial, role in coordinating the economic and political systems in these countries than elsewhere. The analysis predicts a greater role of persuasion settings in economic systems globally, given the future outlook of an increasing influence of both countries.

Notes

1 Introduction 1 The division of models into these two categories may have its origin in the distinct orientations of Marshall and Quesnay. 2 Krugman (1993) and Sugden (2002), among others. 3 In this context, mention can be made of a recent study by Estrada (2011) that aimed at developing an analytical tool called the ‘Policy Modelling Consistency (PMC) Index’ for the purpose of evaluating policy modelling. The evaluation involves checks on the use of input-output tables, and classification of variables, and identification of parameters, among others. Estrada suggests that various possible effects of economic policies can be shown using multidimensional graphical means. 4 See, for instance, Tinbergen (1962). 5 In the weekly common room lunches of Tinbergen with his teammates—these took place in Rotterdam on Thursdays—one could notice the concerns and doubts of Tinbergen at the time. He admitted that the development effort in the Indian subcontinent is obstructed by the feudal system, and that the topic of agrarian reform is neglected by development economists. He emphasised the need and urgency for policy studies on the institution of lump-sum land taxes and crop-purchase schemes by the state. 6 I was approached a year later by Tinbergen on whether I could undertake the funded research on agrarian reform. The research took more time to finish than was originally planned, due to obligations in teaching and other studies. 7 Cohen (1978); Cohen (1977); Cohen (1981). 8 The model, which formed the subject of my dissertation, bears the influences of a two-year period of employment at the United Nations Research Institute on Social Development (UNRISD), subsequent employment at the Netherlands School of Economics, and of Jan Tinbergen as teacher and supervisor, and later as my colleague. The model was first published as an article in Cohen (1972). The dissertation was published later in Cohen (1975). 9 Was there any policy interest in the modelling application from the side of Korean authorities? It is interesting to recall the following. In 1975 the Korean Minister of Finance and Development, on his way to a WB/IMF meeting in Washington, made a stopover in The Hague for a lunch meeting to discuss the study with me and with Jan Tinbergen. It was a Sunday. Tinbergen opted for a soup, excusing himself that it was Sunday and that Mrs Tinbergen had already prepared the evening meal; and I followed his choice. For at least one hour I explained how the model was applied to Korea and emphasised the positive results obtained for both growth and redistribution; while watching the minister’s handling of a full three-course lunch. After finishing his meal, he gave his opinion at the end, saying that the Korean government knew all the way through that Korea has no problem combining growth with redistribution, and that he was glad to hear that the Dutch government believes that too!

Notes  367 The minister apart, there was much more academic interest in the book coming from the Korea Development Institute, and Seoul National University, where a couple of related articles appeared in their economic journals. 10 This work, done in collaboration with Sanjaya Acharya, constituted his PhD. I am grateful for the cooperation and contribution. I have implemented some modifications. Any error or omission in the process is mine. 11 This work, done in collaboration with Eisa Abdel Galil, constituted his PhD. I am grateful for the cooperation and contribution. I have implemented some modifications. Any error or omission in the process is mine. 12 In that sense, the former Soviet Union and allied countries are rightly described as transiting economies (rather than transition economies): transiting from a stateplanned economic system to a market-based economic system. 13 The achievements of development economics and development policy were heavily criticised by Hirschman (1981), among others. 14 Work of the author on centrally planned economies and transiting economies dates back from collaboration with Janos Kornai in the early eighties. Major involvement started in the nineties with EU support for the construction of SAMs for East European countries, and was bolstered by giving leadership to several projects funded by EU/TACIS that aimed at introducing economics teaching and research at the Russian Higher School of Economics. 15 The application was done in collaboration with Rini Braber. I am grateful for the cooperation and for his contribution, without which this chapter was not possible. I have implemented some modifications in the model and rerun a few computations. Any error or omission in the process is mine. 16 The study is the result of collaboration with Adam Czysewski. I am grateful for the cooperation and his contribution. The presentation of the model has undergone important changes since then. 17 Some examples of the common uses that the book follows are summarised here. Indexes are expressed in small letters: j = sectors of production, c = commodities, h = household groups, q = skill types, among others. Variables expressed in values that belong to national accounts statistics and that often occur in economy-wide models will be denoted by one capital letter: for example, X for gross production (output), V for value added, Y for income, C for consumption, I for investment, E for exports, and M for imports, and so on. Capital letters P and Q will be reserved for price and quantity, and will be attached to the above variables where applicable. For instance, XPj is the price index of output in sector j. There are other variables with notations of three letters that signal the meaning of the variables: for example, FXR is foreign exchange rate, FCF is foreign capital flow. Furthermore, it is noted that throughout the book t is the index used to denote year t.

2  Some essentials in economy-wide policy models  1 The CES production function, Arrow et al. (1961), exhibits constant elasticity of substitution between the factors of production in terms of their factor proportions. Other production functions mostly used are special cases of the CES production function. Where the substitution elasticity approaches 1, we have the Cobb-Douglas function; where it approaches 0, we have the Leontief (perfect complements) function. The Harrod-Domar production function falls into the same category as the Leontief function. Most of the economy-wide models in this book use a combination of Cobb-Douglas and Leontief production functions, while in Chapter 12 the CES and Leontief production functions are jointly used. It goes without saying, that where justifiable, the production functions of different sectors may be specified differently.   2 Installed investment is fixed investment. However, the word ‘fixed’ may bring misunderstandings as we use the same word in the sense of fixing an exogenous value of the variable. To avoid such overlapping we shall refer to fixed investment as installed investment or simply investment.

368  Notes 3 A general discussion of closure rules, in this case of the savings investment balance, without specific reference to a specific model or a country application is a woolly undertaking with a doubtful contribution. Nevertheless, the literature contains many such discussions. For example, there is the neoclassical closure in which savings dominate investment, but the mechanism(s) for realising this closure and the adjustors involved are seldom defined, and discussed. At the opposite end, there is the Johansen closure which attributes a decisive role to an autonomously determined investment, implying a subordinate role for savings. Taxes and subsidies are proposed to be the adjustors. In another version, in models that incorporate money and prices, real savings are the subordinate adjustors. The loanable funds/Fisherian closure assumes that the interest rate brings investment and savings into equilibrium. It is not established that the interest rate is empirically strong enough to accomplish the adjustment. The neo-Keynesian-Kaldorian closure looks for adjustors in the factor market (labour is not paid its marginal productivity). The Keynesian closure allows for adjustments in unemployment. The Kaleckian closure is conceived to be one among various structuralist closures. With respect to CGE models, there is no one default closure that CGE models follow. Most choose a dominance of savings over investment; less so for dominance of investment over savings; and rarely for a configuration in which both ends meet half way. In most deliberations, there is no mention of a role for inventory changes or capacity utilisation on the investment side or hoarding and borrowing on the savings side. 4 Simon (1953). The general rule for establishing causal ordering is: endogenous variables of the first order are those which are influenced by parameters (and predetermined variables) and (or) by each other mutually. Endogenous variables of the second order are those which are influenced only by parameters (and predetermined variables), and (or) endogenous variables of the first and second orders, and so forth. 5 This elaboration may require subsequent elaboration in the equations of exports and imports, that is, eqs. II.13 and II.14. The offered price of exports and the domestic price of imports need to be adjusted upwards with the level the taxes. As the price of imported goods is raised, the composite supply price of commodity c is also bound to rise, so that the adjustment has to be carried forward to eq. III.11.1 above. These aspects are worked out in Chapter 7.

3  Socio-political regimes and economic development 1 Myrdal (1968). The Asian Drama presented a pessimistic and a rebuffing perspective in which development planning, plans, and targets did not matter, and in which the benefits of economic development went primarily to the wealthy and the feudal landlord. 2 Cohen (1978). See also Cohen (1981) and Cohen (1977). I am greatly indebted to Jan Tinbergen for giving me the opportunity to undertake the study. Thanks go also to former colleagues for their useful comments, among whom F. Bishay, H. Bos, P. Cornelisse, L. Mennes, J. Sandee, A. Waarts, as well as S. Chakravarty, who was visiting professor in Rotterdam in 1975–76. 3 The model does not specify explicitly the cash held, but it can be derived as a residual from the following identity F1 = (cash held) +ϕf 21F2. 4 See note 3. 5 In the Harrod-Domar production function the minimal of labour and capital supply determines production. If labour is assumed to be in abundance, (incremental) capital becomes the constraining factor that determines (growth of) output. This setup was standard use in early models for developing countries and in UN modelling of the 1st, 2nd and 3rd development decades’ policy programs. 6 In eqs. 2 and 9, the maintenance of agricultural gross production at the stipulated levels X1 and X2 would require investments I1 and I2 along the lines of a simple HarrodDomar production function. This combines with the input-output technology that

Notes  369 enters in eqs. 18, 20, and 21 to give a two-level production technology, as is common of economy-wide models.  7 It was empirically handy to make use of given projected values for the foreign capital flow FCF; this led to investment I33 being determined in the savings investment balance, eq. 24. When I33 is then fed into the product market balance (nonagriculture), eq. 23, the solution for net exports (non-agriculture), E3, is obtained. The same outcomes could be secured if instead the ordering is reversed: Take E3 as given, which is then fed into eq.23 to give I33; that is then entered in the savings investment balance, eq. 24, thus solving for an endogenous FCF. Most of the models in this book follow the latter, where FCF is made endogenous. In this respect, the difference between this model and the others is only formal.   8 In a circumscribed and interdependent accounting system of n rows, knowledge of n − 1 rows is sufficient to solve the system. The nth row is not required and is left out. Correspondingly, Walras Law states that in general equilibrium, specification of market clearance for n − 1 markets is sufficient, and the nth market clearance can be dropped, this being automatically guaranteed.   9 The causal ordering of a system can be decomposed along the lines of Simon (1953). Endogenous variables of the first order are those which are influenced by parameters (and predetermined variables) and (or) by each other mutually. Endogenous variables of the second order are those which are influenced only by parameters (and predetermined variables), and (or) endogenous variables of the first and second orders, and so forth. 10 Of course, it is an oversimplification to maintain that the peasantry does not resist rich farmers in India. There are regions where growing awareness has slowly and steadily eroded the dominance of landlords. Agrarian studies in a vast country like India require a multiregional analysis as suggested by Joshi (1975) p. 96.

4  Social economic development goals in economy-wide policy models   1 The UNRISD list of basic needs indicators was applied to developing countries in Drewnowski and Scott (1967). The application was extended in Cohen (1968). This application was later updated to cover the period 1920–75, and was published in Cohen (1986).  2 It can be rightly argued that the group of employers, self-employed, and family workers is a very heterogeneous group and cannot have common interests. The group usually contains the richest and some of the poorer parts of the population. Its distribution between rural and urban areas is even more subtle. Disaggregated data on the constituents of this group are very scarce for the moment, however.   3 Poverty groups, in the development context, occur in household groups headed by wage earners and among some fractions of the self-employed and family workers. It is operationally more plausible to incorporate the wage-earner group in policy models than fractions of the self-employed and family workers. As poverty among both groups tends to correlate, our restricted specification can be given a more general interpretation.   4 The national survival rate is 1 minus the national mortality rate.   5 Quoting from Adam Smith, more than two centuries ago, when he stated: ‘The liberal reward of labour, as it encourages the propagation, so it increases the industry of the common people. The wages of labour are an encouragement of industry, which, like every other human quality, improves in proportion to the encouragement it receives. A plentiful subsistence increases the bodily strength of the labourer, and the comfortable hope of bettering his condition, and ending his days perhaps in ease and plenty, animates him to exert that strength to the utmost …’ Smith (1776), reproduced in Cannan, ed. (1961), p. 91.  6 Fisher et al. (1965).

370  Notes   7 The oldest attempts seem to have appeared in Japan (1965), and by Byung-Nak Song (1972).   8 The index q, stands for skill types, wherein l = occupational categories comprising ISCO Major Groups 3/9, which is briefly called ‘low skill’; while 2 = occupational categories comprising ISCO Major Groups 0/2, which is briefly called ‘high skill’.   9 As a result, Y h becomes in million won per person, WC1 is in thousand calories per person, and WD1 is rooms per person. 10 As it is assumed that high-skilled labour has already reached BNA > 1, specification of a basic needs attainment productivity effect in the case of high-skill labour is not applicable. The specification is limited to low-skill workers only. 11 Although the distribution of destination proportions is assumed fixed for a period t, the coefficients need to be updated for distributional changes between periods. The basis for the specification of the coefficient is the relative investment requirement of a sector in relation to all sectors: κpj = (1 / κj ) (X − Xj, t − 1 ) / Σj′ ((1 / κj′ ) (Xj′ − Xj′, t − 1 ), where κj stands for the incremental capital output ratio. The destination proportions should be consistent with the weights of relative growth among the sectors. 12 The reader is directed to equations 6 and 5, in Chapter 14. There, the formation of NEe, NTe and LSe is specified in detail for each year. These equations contain transition rates on demographic survival, promotion of students from one class to another, repetition, graduation, certification by educational level, and participation in the labour force. When the three variables over t more years are approximated linearly to give their values at the end of six year periods, we obtain equations 6 and 7 as they are expressed in the current model. For more details see Cohen (1975). 13 For example: Japan (1965) and Byung-Nak Song (1972). 14 See, for instance, Grier and Tullock (1989) and Easterly and Rebelo (1993). 15 The supply elasticities of production for low skill and high skill amount to 0.24, and 0.006, and for capital to 0.754, in the Cobb-Douglas production function of the agriculture sector. The three elasticities are 0.34, 0.08, and 0.58 for the extended industry sector. 16 The calibration of plan requirements to observed employment performed econometrically very well with R2 = 0.98 and high t-value. The results show a 1 per cent increase in planned requirements of manpower skills has been implemented with 1.1 per cent to 1.2 per cent increase in actual employment. For the past period, the derivation of manpower input rates from planned requirements can be described to underestimate the actual manpower absorption, the discrepancy between planned and realised being higher in the case of the high skill than in that of the low skill. 17 For derivation of the transition rates see endnote 10, above. 18 R2 = 0.97 and 0.99 for regressed labour supply equations for h = 1, 2, respectively. 19 The equation explains that the remuneration rate for wage earners performed better than that for salary earners, with R2 = 0.83 and 0.67, respectively. In the case of wage earners, the variance is explained for 80 per cent by labour productivity and for 20 per cent by the unemployment rate. 20 Compared with other regressed equations, the proposed private investment equation performed least well. It had to be modified to a simple dependence on the GDP, which did very well with R2 = 0.96, t-value high, and serial correlation low. Transformation of the slope coefficient into an elasticity at mean values gave an elasticity of private investment to GDP of 2.75, which is characteristic of the accelerating effect of economic growth on investment, and likewise a high feedback effect of investment on future economic growth, both typical for Korea. 21 As Korean data on the daily calorie consumption available were insufficient for testing the equation, we have relied on mixed data from cross sections (from countries of the ECAFE), and time series (between 1957 and 1967). In total, 20 observations were employed. The following regression results were obtained: (Ch1 + Cg1 ) / (π1P. WC1 ) = 0.0051 + 0.0498 (Y1 / π1P). Estimates are highly significant with R2 = 0.97.

Notes  371 22 Estimation of the equation for housing gave an intercept at 0.0032, and a highly significant slope coefficient at 0.0427, with R2 = 0.91. 23 The more general employment ratio, meant to include all earners, employers, selfemployed and family workers divided by the labour force is, by definition, lower than WM as used here. 24 These conclusions are in agreement with results of the Bari Loche world model for Latin America obtained in a different context; see Herrera et al. (1974). 25 GBD and FCF were also positioned in the highest orders in the CEM prototype model, Chapter 2. They do not influence other variables.

5  Growth and distribution in SAM models   1 The commodities account has been also named the wants account. The commodities can be classified along the same lines of activities, or distinctly.   2 The classification into endogenous and exogenous variables can have different forms, depending on focus and circumstances, as will be apparent from the application of the SAM to developed countries, in Chapter 11.   3 For the sources of the ten SAMs analysed in this chapter, see Cohen (2002a).   4 India and Pakistan are shown to have the highest average output multipliers, 5.9 and 8.8 respectively, which is partly due to a treatment of imports in their SAMs that does not fully consider leakages due to imports. At the other extreme, Suriname and Egypt show the lowest output multipliers, 1.9 and 2.5 respectively, reflecting low degrees of interdependence.  5 GLI was first discussed and applied in Cohen (1988) and Cohen (1989). In these publications the index was named the relative distributive measure, RDM. Reflecting on it ex post, it is sounder to call it a gainers and losers index, GLI.   6 See Cohen (2002a).

6  Simplified statics and dynamics in the CGE model   1 See Thorbecke (1990) for a broad discussion of Indonesian CGE models.   2 Different household groups are denoted by index h = 1, 2, and so forth. We also use a subset of h to represent firms. There is no need to introduce a separate index for firms. Household groups and firms have in common that they receive a profit income, pay taxes, and engage in transfer payments. Otherwise, they differ: households earn wage income, firms do not; households consume, while firms do not have consumption expenditures.   3 Installed investment is fixed investment. However, the word ‘fixed’ may bring misunderstandings, as we use the same word in the sense of fixing an exogenous value of the variable. To avoid such overlapping we shall refer to fixed investment as installed investment or simply investment. This note is the same as note 2, Chapter 2.   4 See Chapters 2, 3, and 4 where causal ordering following Simon was followed in displaying the structures of the models presented.   5 By defining a low-skilled wage rate as an index, factor payments to the low skilled become equal to LR1 LD1j , which is a readily available figure in the SAM. For highskilled labour and capital the same approach can be followed. However, for reasons of converting the static model into a dynamic model later on, a different approach was followed regarding these production factors. In the dynamic model, low-skilled and high-skilled labour must add up to total labour supply. Thus high-skilled labour has to be expressed in the same productive units as low-skilled labour. To obtain this result, it is assumed that differential productivity is reflected by differential earnings, and that a unit of high-skilled labour earns four times the wage of a low-skilled worker in the benchmark period, which is empirically supported in the Indonesian context. This implies that in the benchmark LD2j is equal to LR2 /4 LD2j. Similarly, in the dynamic model new investment has to be added to the existing capital stock,

372  Notes so capital has to be expressed in the same physical units as investment. The way in which this is done is elaborated in Section 5. 6 Aggregate demand consists of demand categories with a negative unitary elasticity (these are consumption and investment), and with a zero elasticity of demand (these are intermediate deliveries, exports, competitive imports). As a consequence, the aggregate demand elasticity lies between −1 and 0. 7 Take ε′q for the unit cost of educating labour of skill type q, ε′′q for the part of educational costs charged to the user, and ε′′′2 for foregone earnings during n years of educational upgrade from low to high skill. The total cost for skill type 2 can be abridged into ε2 = ε′2 + ε′′2+ ε′′′2. The total cost for skill type 1 is ε1 = ε′1 + ε'′1. The equation makes use of the aggregated costs ε2 and ε1.

7  Growth with redistribution through liberalisation with restructuring 1 Installed investment is fixed investment. However, the word ‘fixed’ may bring misunderstandings, as we use the same word in the sense of fixing an exogenous value of the variable. To avoid such overlapping we shall refer to fixed investment as installed investment or simply investment. This note is the same as note 2, Chapter 2. 2 In principle, estimation of a private investment function should be econometrically tested. This is not feasible, given the available data for Nepal. Under the circumstances a Cobb-Douglas type equation for investment demand offered a plausible solution with a profit effect and a crowding-in public investment effect. Various studies suggest elasticity values for the public investment effect ranging between 0.8 and 0.4. We assume initially equal values for both elasticities ıg = ık = 0.5. For more details see Acharya (2011). 3 This is partly due to the strong growth of industry and strong input-output linkages that push the composite price of agricultural commodities upwards. 4 For a review of adaptation of the model to these considerations, application, and results see Acharya and Cohen (2008) and Acharya (2011). 5 Taking ε′q for the unit cost of educating labour of skill type q, ε′′q for the part of educational costs charged to the user, and ε′′′2 for foregone earnings during n years of educational upgrade from low to high skill. The total cost for skill type 2 can be abridged into ε2 = ε′2 + ε′′2+ ε′′′2. Similarly, the total cost for skill type 1 is ε1 = ε′1 + ε′′1. The equation makes use of the aggregated costs ε2 and ε1. The content of this note is the same as note 7, Chapter 6.

8  Sustained development of land resources 1 The Sudan is a typical country in which the four resource-degrading factors apply, see Abdelgalil (2000). For example, the low level of farmers’ education and training leads to overexploitation of arable land. Next, price controls imposed by the government on irrigated agriculture have discouraged farmers from adoption of long-term sustainable cultivation practices. Land tenure insecurity has pushed farmers in mechanised agriculture to ‘mine’ arable land in pursuit of short-run gains. Finally, poverty in subsistence agriculture has compelled farmers to exploit arable land unsustainably. Furthermore, open access to woodland is a major cause behind the excessive clearance of forestland, and lack of well-defined property rights over grazing land has left no incentives for livestock owners to invest in improving the prevailing conditions in grazing land. 2 Irrigated land NQ1 represents about 24 per cent of the total cultivated area in Sudan, it is generally fixed and its expansion requires significant public investments in irrigation works. Mechanised NQ2 and subsistence NQ3 are rain-fed and they form 33 per cent and 43 per cent, respectively. Rain-fed land can shift between subsistence and mechanised cultivation, very much governed by market forces, depending on the relative value of land in both alternatives. The rest of rain-fed land is used by forestry NQ4 and livestock NQ5. In 1990, NQ4 was twice NQ5. One feddan equals about 0.42 hectare or 1.038 acre.

Notes  373   3 This applies to cases of one and the same crop that is being produced in all three sectors, as well as when the aggregate productivity value per feddan is compared for the three sectors.   4 Unable to satisfy basic needs, and having a limited time horizon, subsistence farmers often overexploit the land during their lifetime, cf. Barbier (2000). It has been regularly observed that the negligence of the fallow period, crop rotation, and soil conservation and improvement measures in subsistence rain-fed agriculture reflect both the poverty and the lack of necessary farming knowledge in this sector, cf. Kevane (1997). Subsistence farmers, being the poorest, are driven by this factor to cultivate the land more frequently and, hence, the fallow period becomes shorter and shorter, or ignored completely. Ignoring crop rotation reflects the limited options they have when it comes to the issue of crop diversification. They mainly cultivate food crops (sorghum and millet) on which they depend for their survival, and it happens that these food crops are more soil erosive than others, cf. Coxhead and Jayasuriya (1994). The lack of crop rotation leads to soil nutrient depletion and soil erosion and consequently declining yield.   5 There can be rare, but desirable, situations from the environmental viewpoint where Tgj can function as a tax imposed by the government on households, that is, Tgj < 0. If such a tax is levied on households, then wood cutting and livestock slaughter will have to increase so as to maintain their living standard as it was before the tax.   6 Ideally, a private investment function should be econometrically regressed from the country’s data. The Sudanese data do not allow this. Instead, we followed here a Cobb-Douglas type equation for investment demand, and plugged in an appropriate value for the crowding-in public investment elasticity based on Taylor (1991), p.166, and Ibrahim (1995), p.54, at ıg = 0.82, so that the price effect elasticity can be set at ıp = 1 − 0.82 = 0.18.   7 Installed investment is fixed investment. However, the word ‘fixed’ may bring misunderstandings as we use the same word in the sense of fixing an exogenous value of the variable. To avoid such overlapping we shall refer to fixed investment as installed investment or simply investment. This note is the same as note 2, Chapter 2.   8 We incorporate an upper limit to the permissible level of AFL exports, beyond which a government role is implied as a subsidiser of additional exports where certain costs have to be incurred to provide the additional supply.  9 FCF is assumed to grow at 5 per cent per annum. 10 See Chapter 2 for more discussion of Walras Law. 11 The cost of reclaiming a degraded feddan of arable land of an average quality is found to be about 10 per cent of its foregone yield, cf. Gigengack et al. (1990). Note that arable land degradation can be seen as a process that ranges from mild degradation to extreme desertification. 12 Relative prices between the AFL sectors have moved very closely with those of the relative indices based on world market prices, so that working with the latter is justified. The strong inflationary tendencies in the Sudan were neutral in effect as far as these relative prices are concerned. The price elevation is about 9 times for each of the AFL sectors. Taking the irrigated price index as numeraire, the relative price indices for all AFL sectors are calculated accordingly, and set to grow exogenously, based on the past growth path of relative world market prices. Note that, in contrast, the price index for the ROE sector is endogenously solved. For more details, see Abdelgalil (2000). 13 The level of desired basic needs income, YBN, for the poor population engaged in subsistence agriculture, about 26 per cent of the population, is calculated as being in the range of 13 per cent of GDP, and is assumed to grow annually at 5 per cent, same as that of GDP in the past. Government transfers to the poor population in subsistence and forestry sectors, Tgj, for the purpose of meeting basic needs is set to zero in the year 1990, but is simulated at positive levels in the policy runs. Agricultural labour force is assumed to grow by an annual growth rate of 2.7 per cent, as it did in the previous decade.

374  Notes 14 The relative price indices are projected to move within a narrow range. For instance, while the numeraire price index of irrigated grows at 0.048 per cent per annum, that of mechanised grows at 0.041 per cent, giving a declining relative price index of mechanised at XP2 = 1.041/1.048 = −0.9933. The magnitudes of the other indices are XP3 = 0.9924, XP4 = 0.9952, XP5 = 0.9943. These projections show XP4 and XP5 standing for forestry and livestock prices, to experience slightly more favourable price trends than those of agricultural produce XP2 and XP3. 15 More results are discussed in Abdelgalil and Cohen (2001). 16 Livestock output per feddan increases because both non-agricultural output and the GDP work in the same direction to increase final consumption and intermediate demand for livestock products. This improves sustainable productivity, as vegetation yield rises while degradation costs decline.

10  Transiting from fixed- to flexible-price regimes   1 The SAMs were constructed in the context of an EU funded research cooperation between West and East European countries on structural policies for transition economies; see Cohen (ed.) (1993) for more results.   2 The other way to think of modelling planners’ behaviour is to assume that they optimise some social welfare function. A discussion of different modelling approaches to central planning can be found in Bennett (1989).   3 The adjustments are the following: 1. Capital income in Hungary is not distributed directly to the households as a total proportion of capital income, but flows to the firms and is distributed over the households as a fixed proportion of firm income; 2. the government sector also comprises social security; 3. firms are allowed to buy consumption goods; and 4. indirect taxes are levied not only on production, but also on consumption and on imported investment goods.

11  Public spending multipliers in extended SAM models for a developed economy   1 The SAM applications to the Netherlands were implemented in the context of an EU research grant on comparing and integrating national accounts systems in European countries. Other countries for which SAMs were constructed and analysed were Germany, Italy, Spain, Hungary, and Poland. See Cohen (1993). In this context it was possible to construct a 10-year series of SAMs for the Netherlands starting from 1978 up to and including 1987. In this chapter we shall make use of those for 1978, 1981, 1984, and 1987. The series was supplemented later with SAMs for 1995 and 2000.   2 The exchange rate that was formally employed to convert Dutch guilder into euro was 1.0 euro = 2.27 Dutch guilders.   3 We have made attempts to convert SAMs from current to constant prices, with some success. These are reported in Cohen and Tuyl (1992).   4 See Cohen (2002b).   5 See Berg and Meer (1988).

12  Fiscal policy simulations in adapted CGE models   1 A fiscal policy rule is defined, in a macroeconomic context, as a permanent constraint on fiscal policy, typically defined in terms of an indicator of overall fiscal performance. Ideally, a fiscal rule should be well defined, transparent, adequate, consistent, simple, flexible, enforceable, and efficient. See Kopits and Symansky (1998).   2 For the benchmark year of 1981, this was 920 million Dutch guilders.   3 In general, when there are n interconnected accounting balances forming a coherent accounting system, solutions obtained from any combination of n − 1 balances, and thus leaving out one balance, are guaranteed to fit into that balance which is left out.

Notes  375   4 The two sources are known as the FREIA model, which was instituted in the Central Planning Bureau (1983), and RASMUS, which was instituted in Erasmus University Rotterdam, De Groene et al. (1984).   5 CPB (2010).

13  Normed planning of human resource development   1 HDI is an equally weighted index of indicators of health (life expectancy) and education (the adult literacy rate and combined primary and secondary enrolment ratio).   2 It is understandable that the selection resulted in the absence of sub-Saharan African countries in view of the sub-performance of the region. Even though South Africa has a relatively advanced educational state, the progress in the HDI is significantly behind that of the index of GDP per capita, −22 points in 2004; the HDI ranking has deteriorated, and GDP growth is moderate.   3 This is consistent with conventional decision making that is based on first knowing the size of the government spending budget on all sectors, GTS, and second, is the decision as to what part of the budget will be expended on the education sector, GTE.   4 The percentage distribution of age groups 7 to 14 and 15 to 18 for 2005 and 2015 is derived from the Ten Years Education Sector Plan 2006/07 – 2015/16, backed by the simulation model of enrolments and other educational variables at the MOE planning department. As no data are classified for the age group 19 to 22, this is taken as 80 per cent of age group 20 to 24, for which data are available. For 2030, and due to the projected diminished growth rate of the population, the share of the 7 to 14 age group is projected to decrease between 2015 and 2030 by about 4.5 per cent. The shares of the age groups 15 to 18 and 19 to 22 are assumed to increase by the same margin of 4.5 per cent. Population projections to 2015 are as in the plan. For later years we apply UN/ Ethiopia projected average annual growth for consecutive periods of five years. The annual growth rates behind the projections are 3.2, 2.6, 2.3, 2.08, 1.92, and 1.72 for the periods 2000–05, 2005–10, 2010–15, 2015–20, 2020–25, and 2025–30, respectively.   5 Note that $965 is the GDP per capita in $ppp in 2005, in constant prices of 2000. This is different to the GDP per capita in US dollars in 2005 in the current prices of 2005. While not relevant for the analysis here, this later figure amounted to only $150. The difference between the two figures lies in expressing the GDP per capita in $ppp and in constant prices dating back to five years earlier.   6 The deflation index to convert 2005 birr back to 2000 is 0.8585, from the World Bank Development Indicators database.   7 The problem is acknowledged by authorities as reflected in documented policies that aim to bring the ratio of teachers’ salary to GDP per capita across all education towards 4 times its current levels of 6 to 8 times.   8 The ratio of non-academic to academic staff in Ethiopia is 2:1; this is 3:1 in African countries, ESDP III, p.20.   9 The regression gave the following weak results for the equation.   GBR = 0.074 GDPpc + 0.7012  R2 = 0.63 This implies that GBR reaches its maximum level of 1.0 at GDPpc of 4000 $ppp and above. 10 The magnitude of the backlog can be measured from eq. 16, which relates the adult literacy rate, ALR, to GDPpc. The tendencies show a 100 per cent adult literacy at GDPpc of 8000 $ppp and above. For the observation period 1991 to 2005, Korea, Japan, and Russia had 100 per cent throughout, and are thus left out.  ALR = 21.615 ln GDPpc + 54.746  R2 = 0.69 11 The LFS of 2001 shows the unemployment rate for first-time entrants in the labour force to jump from 1.6 per cent for entrants with no schooling to 6.9 per cent for primary graduates, 30.6 per cent for secondary graduates, and 9.1 per cent for tertiary graduates, reflecting interactions between search time and job opportunities. For workers of all ages, the unemployment rates take a more balanced profile at 3.7

376  Notes per cent, 5.7 per cent, 19.9 per cent, and 7.1 per cent for the four educational levels, respectively. In 2001 the unemployment rate for all ages was 5.2 per cent. In 2005 this was 5.0 per cent, with about the same educational distribution. 12 For example, shortages in ICT skills have been compensated by salary adjustments upwards for ICT graduates by up to 20 per cent above non-ICT graduates with equivalent years of education. The variations in the salary scales are rich sources of information that can be fruitfully used in programming education and training activities, and in resolving LM mismatches by stimulating graduates to pursue the jobs and (or) training types that are in more demand. 13 The height of income disparity can be also gathered from the 6 to 10 times discrepancy between average teacher salaries and the GDP per capita, discussed earlier in the roadmap projections.

14  Labour market imbalances and adjustments   1 Related assignments in various countries resulted in several reports that prepared the way for a concerted effort towards resolving the problem of forecasted imbalances and the shaping of labour market adjustments. The reports are listed here: (a) Cohen and Sarmiento (1982). (b) A Study of the Occupational and Educational Manpower Requirements and Supply of the Fifth Plan, Manpower Division, Government of Pakistan, Islamabad. 1981. (c) Projection of Human Resource Development in the Framework of the Fifth Kenyan Development Plan 1984–88, report for the European Commission, Brussels, and Government of Kenya, Nairobi, 1983.(d) Structural Changes in the Indonesian Employment Situation, commissioned report by the World Bank Seventh Education Project and Government of Indonesia, Washington and Jakarta, 1988. This chapter makes use of the applied work for Colombia and Pakistan; and supplements the results with similar applications for Korea.   2 The number of sectors distinguished varies in the applications between 9 and 32 depending on the country.   3 The mode takes the variables of the GDP per sector to be exogenously given by policy makers and official plans, for all four countries. But when first applied in Pakistan and Korea, these variables were embedded also within an economy-wide model which includes more aspects of the circular flow.   4 Pakistan in 1980 shows low or some negative imbalances for professional but also for manual occupations; which is due to labour migration to the Gulf countries.   5 The adjusted λeq,r is related to the initial λeq,o, via RMe, in the following way:   λeq,r = RMe . λeq,o / Σe RMe . λeq,o

15  Privatisation decisions during transition   1 Our modelling approach which allows for separate valuations by buyer and seller of a privatisation venture stands in contrast with OECD (1993) where one and the same valuation for buyers and sellers is assumed.   2 The first geometric series is based on P1 + P2 + … + PT and consists of a finite number of elements, dependent on the expected length of the transition period: {1 + [ST / (So − Sa )]1/T} (1 + ι)−1. The second geometric series consists of profits over an infinite number of years after period T, thus PT + 1 + PT + 2 and so on. The value of this series is calculable for a ratio of (1 + γ) / (1 + ι) equal to less than 1, which is normally the case, cf. Cohen (1993).   3 The ratio (κ/λ) is (capital/output) / (labour/capital), which gives quantity of output per unit of labour.  4 The total value of discounted investment expenditures in the transition period is found as,  (Kb + Kc )discounted = [(Kb + Kc ) / T] [(1 + ι)−1 + (1 + ι)−2 + … .+ (1 + ι)−T].

Notes  377   5 For instance, (Σ Kd )discounted = [Kd,T + 1 / (1 + ι) ] + [Kd,T + 2 / (1 + ι) ] + and so forth. Substituting for Kd,T + 1 the right-hand side of eq. 18 we get the geometric series with a ratio (1 + δ) / (1 + ι) less than 1. Hence (ΣKd )discounted can be written as [(γ + δ) KT / (1 + ι)T] [1/(ι − δ)]. T+1

T+2

16  Economic policy solutions to social queuing problems   1 See Oxley and MacFarlan (1995) for a review   2 Prismant, a research institute, estimated that on 1 May 2001 the waiting list for treatment in hospitals in the Netherlands amounted to about 161,000 patients, of whom 30,000 were waiting for longer than one year; reported in Volkskrant of 26 June 2001. The problem has become a major political issue, especially in periods of parliamentary elections. For a comparable situation in the United Kingdom see Tudor Edwards (1997).   3 For a discussion of human resource matrices and their use, see Cohen (1994) pp.37–47.  4 This is a crude way of estimating the QALY indications. The QALY indications should be based on a linear function relating costs of treatment per disease type and age to gains in QALY. In principle, these functions would provide a better picture of actual cost-utility analysis.   5 It is assumed here that the number of persons by different ages is evenly distributed within each age group. Otherwise, the four tangents would not provide an accurate estimation of the earning parameters.   6 CBS (1979) Loonstructuuronderzoek (that is, wage structure research), different years.   7 We acknowledge that this is a very crude way of estimating these earning functions. Empirical literature contains many more valid estimates based on many more data sets, and in principle such estimates should be used in future applications. Note also that we attach disease d to the coefficients for the sake of convenience, but we do not consider the association of specific earnings with specific diseases.   8 Recall that W is the product of the QALY indicator Q and the number of treated patients L.   9 The five lowest and the five highest values are omitted in the calculations; these are assumed not to be representative due to the random nature of the simulations. 10 The maximisation of the objective function can be interpreted as the maximisation of the health care capacity, as we assume that payments obtained from treated persons are used to supplement the health budget.

17  Modelling convergence in economic growth between rich and poor countries   1 Works using the data set include those of Baumol (1986), Dowrick and Gemmell (1988), Barro (1991), Mankiw et al. (1992), Sprout and Weaver (1992), and Theil and Seale (1994), among others.   2 If one considers economic welfare variables other than income per head, such as attainment in education or health, then the catching-up trends are very significant and beyond doubt.  3 m is a weighted sum of multipliers from injections in sector j, that is, m = Σj mj sj , where mj are the relevant cells in the multiplier matrix M, and where sector shares in the GDP, sj, are employed as weights.   4 Among the rich countries we study, those of Germany, Italy, and Spain fall into this class. Besides, a small population country, centrally located like the Netherlands, has higher shares of foreign transactions with the rest of the world.   5 The sums in eqs. 6.1 and 6.2 do not add to 1 because of unweighted values over sectors and countries.

378  Notes 18  Modelling of distinctly behaving economic systems 1 The following are basic notations used in the chapter in order of occurrence. B = behavioural type b = behavioural setting, with specifications of types h, f, and s for household, firm, and state settings, respectively; ps for pre-state setting i = agent, with specification of k for agents with state authority m = economic system Ri = rewards of agents, Ei = efforts of agents, V b = value added from transformation in a behavioural setting G = engagement lines of transactions and communications of agents, N = intensity of engagement lines of agents Ab = share of agents participating in setting b, Cb = share of commodity transformed in setting b Db = demanded commodity, specified by behavioural setting b in which their transformation is most suited. D = total demand Lb = labour in setting b, L = total labour 2 The notions of loyalty, voice, exit, and entry, emphasised by Hirschman (1970) help explain the dynamics of social systems. Entry and loyalty lead to growth of settings and organisations. Voice and exit lead to their decline. Where two communicating economic systems m1 and m2 exist, with performance of m1 higher than performance of m2, then this would lead to tension in m2 (voice and exit) and pressure on m1 (hostility and entry). 3 Literature relating to logarithms of convergence lays emphasis on mechanisms of integration causing the spread and dominance of particular behavioural types and that give support and background to our hypothesis. Mention can be made of the following mechanisms: imitation, convention, focal points, information cascades, reciprocal behaviour, group learning, and Markov chain inversions. 4 This means, for instance, that the economy and polity in the social system of the United States are consistent with and reinforce each other so as to speak of an integrated whole of economy and polity. Likewise, Russia has a different social system, with its own consistently integrated economy and polity. A fruitful comparison of the United States and Russian economies would require giving due consideration to their corresponding polities. 5 For example, as agents in the United States are intensively engaged in firm settings where the profit motive dominates, the agents adopt this inclination when they participate in household and state settings, and over time these settings converge towards the same profit motive. In Russia, the significant preoccupation of agents with state settings that pursue political motives tends to reorient agents’ behaviour in most settings towards a rent-appropriating political behaviour. 6 North et al. (2006) take the position that interactions between economy and polity are inseparable, which stops the analysis short of reaching fruitful conclusions. We argue here that there are proofs for the supremacy of the economy on the polity, and likewise the opposite. While in the United States the focal points are the economy, firms, profits, and exchange, with a subordinated polity to the economy, in the Russian context the focal points are the polity, state, rent, and authority, and hence, a subordination of economy to polity. 7 Historical records for India and China show that distribution of agents between the three distinguished settings has been intact for many centuries, in spite of significant regime changes in the polity and economy of both countries, cf. Maddison (1971). 8 Persuasion settings are not restricted to developing countries; their active role in advanced economies is discussed in Murphy and Shleifer (2004). 9 North et al. (2006) see a substantive degree of physical violence in primitive society as the precondition for the creation of state institutions, which they call a limited access order. This view is narrow and restrictive. They describe also a transition from

Notes  379 the limited to an open access order in which the Western industrial democracies are classified, but they do not enter into the micro behavioural mechanisms that characterise the transition, and how features of the new order become typical and prevalent. Although the limited and the open-access regimes have some correspondences with the state-intensive and firm-intensive economic systems that we have modelled in this chapter, our approach and the implications thereof are basically very different. 10 Satisfaction of these two types of needs corresponds with the distinction between private and public goods. 11 Take y equal 100 and x and z fixed at 10 per cent below and above y, and take I/K to be 100, implying 100 citizens per one state official, the transaction can then create in the extreme case an income for agent k of 200 as compared to 90 for agent i or i′. To secure a sustainable income, the state agents would pursue a measured exploitation. Too much extraction would cut into the earning base of agents i, i′, and threaten a reduction in the income of k. 12 For brevity’s sake, we formulated state agents k as a unified party. Of course, an elaborate treatment would distinguish between governing agents and those aspiring to govern, bureaucrats, and various intermediary groups, and clients who are aligned to and are materially supported by the state. In spite of the tension between these categories of state and quasi-state agents, the presumption of self-enforcing arrangements among them to preserve their monopoly status and role in state matters is generally valid. 13 Rent seeking can occur also in firm settings, but in a FIM the counteracting forces towards profit maximisation would mobilise more firms to act for greater competition and free entry, which would make the realisation of rent seeking only temporary and neutrally distributed over rent takers. Rent seeking in SIM is permanently institutionalised, and is tied to the office of the authorised rent taker, whoever holds this office. Finally, the demarcation line between rent seeking and profit maximisation is not arbitrary, since profit maximisation abides by the institutional rules that hold for all firm agents, while rent seeking in a state setting strives strategically to adjust and reinterpret the institutional rules towards the benefit of the state authority in the current and future periods. 14 Compare checks and balances restricting power of the state that were built by the Founding Fathers in the US Constitution and passed by the Congress in 1789–91, with the unlimited powers legitimised to the state by the Russian Revolutionary Council in 1917–21, about 130 years later. 15 The two shares are not independent of each other. For instance, Cf affects Af positively in the long run, while when relatively more agents go in f, thus increasing Af , the potential for demanding and producing the f type of commodities is enhanced, and thus Cf is influenced positively. In spite of the interdependence, the two shares stand for different aspects that feature the identification of the economic system. It can be expected that the two shares correlate, which is an argument for following a simple aggregation equation and giving them equal weights in eq. 3. With weights at ω1 = ω2 = 0.5, eq. 3 gives an Index of Interactive Influence that can be fruitfully applied in various contexts to assess the extent of dominance; see Section 7 for an application. 16 See Simon (1993). 17 Commodity shares of firm and state settings are Cft = Dfo (1 + γf )t / {Dho + Dfo (1 + γf )t + Dso (1 + γ)t}; and Cst = Dso (1 + γs )t / {Dho + Dfo (1 + γf )t + Dso (1 + γs )t}. The case for household settings is obtained as a residual Cht = Dho / {Dho + Dfo (1 + γf )t + Dso (1 + γs)t}. 18 The agent share for firm settings can be written as Aft = Lft / Lt = [(Lfo / Dfo) (Dfo (1 + γf )t (1 − λf)t )] / Lo (1 + π)t . Substitution and elimination gives Aft = Afo (1 + γf )t (1 − λf )t / (1 + π)t. The share for state settings is Ast = Aso (1 + γs )t (1 − λs )t / (1 + π)t. The share for household settings is the residual Aht = 1 − Af − As.

380  Notes 19 The EAP region shows a high variability, with China and Vietnam reporting family focus below 82 per cent and other countries reporting above 98 per cent, which can be due to the stronger influential role of the state in these two countries. Such a decline in family focus is consistent with results for SIM countries. 20 See Cohen (2009), among others. 21 Recent empirics indicate that FIM countries, in periods of heavy financial turmoil, can abruptly lose a significant proportion of accumulated wealth over previous years; the MPM countries are more immune. 22 China and India are recorded to have been leading economies in the world until about the eighteenth century. After two centuries of downfall, their economies have risen again and are forecasted to regain their leading positions by 2040 and 2050. The forecasts assume the absence of economic calamities at the foreign front that are caused by world recessions, credit crunch, trade protectionism, and inelastic supply of energy resources; and at the domestic front caused by financial mismanagement, severe epidemics, ethnic conflicts, inequality divides, poverty extent, civil disorder, and polity shake-up.

References

Abdelgalil, E. (2000) Economic Policies for Sustainable Resource Development: Models Applied for the Sudan, PhD thesis, Erasmus University, Faculty of Economic Sciences, Rotterdam. Abdelgalil, E. and Cohen, S. I. (2001) Policy Modelling of the Trade-off between Agricultural Development and Land Degradation–the Sudan, Journal of Policy Modeling, 23(8): 847–874. Acemoglu, D. and Robinson, J. (2005) Economic Origins of Dictatorship and Democracy, Cambridge University Press, Cambridge. Acharya, S. (2011) Pro-poor Growth and Liberalization in Developing Economies: The Case of Nepal, Routledge, New York. Acharya, S. and Cohen, S. I. (2008) Trade Liberalisation and Household Welfare in Nepal, Journal of Policy Modeling, 30(6): 1057–1060. Adelman, I. and Robinson, S. (1978) Income Distribution Policy in Developing Countries: A Case Study of Korea, Oxford University Press (for the World Bank), Oxford. Ahmed, M. and Sanders, J. (1998) Shifting from Extensive to Intensive Agricultural Systems: A Case Study in the Sudan, Agricultural Systems, 58(2): 253–268. Ajani, G., Dallago, B. and Grancelli, B. (1992) Privatization and Entrepreneurship in Post-socialist Countries: Economy and Society, MacMillan, Basingstoke, UK. Arrow, K. J., Chenery, H. B., Minhas, B. S. and Solow, R. M. (1961) Capital Labor Substitution and Economic Efficiency. Review of Economics and Statistics, 43(3): 225–250. Ballard, C., Fullerton, D., Shoven, J. B. and Whalley, J. (1985) A General Equilibrium Model for Tax Policy Evaluation, University of Chicago Press, Chicago. Bandara J. S. and Yu W. (2003) Desirable is the South Asian Free Trade Area? A Quantitative Economic Assessment, The World Economy, 26(9): 1293–1323. Barbier, E. (1990) The Farm-level Economics of Soil Conservation: The Uplands of Java, Land Economics, 66(2): 199–211. Barbier, E. (2000) The Economic Linkages between Rural Poverty and Land Degradation: Some Evidence from Africa, Agriculture, Ecosystems and Environment, 82: 355–370. Bardhan, P. K. (1974) Redistribution with Growth, Some Country Experience, in Chenery, H., Ahluwalia, M., Bell, C., Duloy, J. and Jolly, R. (eds) Redistribution with Growth, Oxford University Press, Oxford, 265–262. Barr, N. (1998) The Economics of the Welfare State, Oxford University Press, Oxford. Barro, R. J. (1991) Economic Growth in a Cross Section of Countries, Quarterly Journal of Economics, 106(2): 407–443. Barro, R. (1996) Democracy and Growth, Journal of Economic Growth, 1(1): 1–27. Barro, R. J. and Lee, J. W. (1993) Losers and Winners in Economic Growth, Proceedings of the World Bank Annual Conference on Development Economics, 267–298.

382  References Baumol, W. J. (1986) Productivity Growth, Convergence, and Welfare: What the Long-Run Data Show, American Economic Review, 76(5): 1072–1185. Bayens, R. and Cohen, S. I. (1993) Dependency Rates in the Dutch Economy, in Hermkens, P. L. T. (ed.), Solidarity of Generations, Thesis Publishers, Amsterdam, 315–350. Beck, T. and Laeven, L. (2006) Institution Building and Growth in Transition Economies, Centre for Economic Policy Research, Discussion Paper No. 5718. Bennett, J. (1989) The Economic Theory of Central Planning, Blackwell, Oxford. Bennett, M. K. (1948) International Disparities in Consumption Levels, American Economic Review, 41(4): 632–649. Berg, L. van den and Meer, J. van der (1988) Dynamics of Urban Systems: General Trends and Dutch Experiences, Environmental Planning, 20: 1471–1486. Bhagwati, J. N. and Brecher, R. A. (1982) Immiserizing Transfers from Abroad, Journal of International Economics, 13(3–4): 353–364. Bleaney, M. and Greenaway, D. (1993) Long-run trends in the Relative Price of Primary Commodities and in the Terms of Trade of Developing Countries, Oxford Economic Papers, 45: 349–363. Bluffstone, R. (1995) The Effects of Labour Market Performance in Developing Countries Under Open Access: An Example from Rural Nepal, Journal of Environmental Economics and Management, 29(1): 42–63. Biro Pusat Statistik Indonesia (BPS) (1982) Social Accounting Matrix Indonesia 1975, Vol. I and II, Indonesian Central Bureau of Statistics, Jakarta. Bovenberg, L. (1985) Dynamic General Equilibrium Tax Models with Adjustment Costs, in Manne, A. (ed.), Economic Equilibrium: Model Formulation and Solution, Mathematical Programming Study, 23, North-Holland, Amsterdam, 40–55. Byung-Nak Song (1972) Planning Model Combining Interindustry and Aggregate Models, Working Paper No. 7211, Korean Development Institute, Seoul, 1972. CBS (1979) Loonstructuuronderzoek (Wage Strucure 1979 and Later Years), CBS, Voorburg. CBS (1999) Jaarboek Welvaartsverdeling (Yearbook of Welfare Distribution), Kluwer, Voorburg. Central Planning Bureau (CPB) (1983) FREIA, A Macro Economic Model for the Medium Term, CPB, The Hague (in Dutch). Central Planning Bureau (CPB) (2010) CPB and Dutch Fiscal Policy in View of the Financial Crisis and Ageing, Report prepared by Bos, F. and Teulings, C., Document No. 218, CPB, Voorburg. Chenery, H. B. and Watanabe, T. (1958) International Comparisons of the Structure of Production, Econometrica, 26(4): 487–521. Chenery, H., Ahluwalia, M., Bell, C., Duloy, J. and Jolly, R. (eds) (1974) Redistribution with Growth. Oxford University Press, Oxford. Chitiga, M., Kandiero T. and Mabugu, R. (2005) Computable General Equilibrium Micro-Simulation Analysis of the Impact of Trade Policies on Poverty in Zimbabwe, Poverty and Economic Policy (PEP) Research Network (www.pep-net.org), Working Paper 01, 2005. Clark, C. (1940) The Conditions of Economic Progress, Macmillan, London Cockburn, J. (2001) Trade Liberalisation and Poverty in Nepal: A Computable General Equilibrium Micro Simulation Analysis, Working Paper, Centre for the Study of African Economics and Nuffield College (Oxford University), and CREFA. Cohen, S. I. (1968) The Level of Living Index in the UK, Netherlands, Japan, and Czechoslovakia, in 1920–1965, Discussion Paper, UNRISD, Geneva. Cohen, S. I. (1972) A Social Planning Model for a Less Developed Economy, Annals of System Research 2: 1–21.

References  383 Cohen, S. I. (1975) Production, Manpower and Social Planning, Rotterdam University. Cohen, S. I. (1977) Development Models with Different Decision-Makers, in The Polish Academy of Science and World Future Studies Federation Future Research, Planning and Decision-making, Conference Proceedings, Jablonna, Poland Cohen, S. I. (1978). Agrarian Structures and Agrarian Reform, Kluwer, Boston. Cohen, S. I. (1981) Distribution of Power and Agrarian Reform, an Explanatory Model Applied to India and Chile, Journal of Policy Modelling, 3(1): 37–59. Cohen, S. I. and Sarmiento, A. (1982) Sistema de Planeacion de Recursos Humanos, Sena Press, Bogota. Cohen, S. I. (1986) A Comparative Study of Needs and Their Satisfaction, Bulletin of the European Association of Development Institutes, 1: 103–120. Cohen, S. I. (1988) A Social Accounting Matrix Analysis for the Netherlands, De Economist, 136(2): 253–272. Cohen, S. I. (1989) Multiplier Analysis in Social Accounting and Input-Output Frameworks; Evidence for Several Countries, in Miller, R., Polenske, K. P. and Rose, A. (eds), Frontiers of Input-Output Analysis, Oxford University Press, New York, 79–99. Cohen, S. I. and Tuyl, J. M. C. (1992) Social Accounting Matrices for the Netherlands in Constant Prices. Discussion paper series no. 9213/G, Institute for Economic Research, Erasmus University Rotterdam (in Dutch). Cohen, S. I. (ed.) (1993) Patterns of Economic Restructuring for Eastern Europe, Ashgate Publishing, Aldershot, UK. Cohen, S I (1994) Human Resource Development and Utilization, Avebury, Aldershot. Cohen, S. I. (2002a) Social Accounting and Economic Modelling for Developing Countries, Ashgate, Aldershot, UK. Cohen, S. I. (2002b) Social Accounting for Industrial and Transition Economies, Economy-wide Models for Analysis and Policy, Ashgate, Aldershot, UK. Cohen, S. I. (2009) Economic Systems Analysis and Policies: Explaining Global Differences, Transitions, Development. Palgrave MacMillan, Basingstoke, UK. Comite Interamericano de Dessarrollo Agricola, CIDA (1966) Tenencia de la Terra y Dessarrollo Socio-Economico del Sector Agricola, Santiago de Chile. Coxhead, I. and Jayasuriya, S. (1994) Technical Change in Agriculture and Land Degradation in Developing Countries: A General Equilibrium Analysis, Land Economics, 70(1): 20–37. Cromwell, E. and Winpenny J. (1993) Does Economic Reform Harm the Environment? A Review of Structural Adjustment in Malawi, Journal of International Development, (5)6: 623–49. Culyer, A. J. (1980) The Political Economy of Social Policy, Martin Robertson, Oxford. Decaluwe, B. and Martens, A. (1988) CGE Modelling and Developing Economies: A Concise Empirical Survey of 73 Applications to 26 Countries, Journal of Policy Modelling, 10(4): 529–568. De Groene, J. H., Siebrand, J. C. and Windt, N.v.d. (1984) RASMUS Annual Model for the Dutch Economy, Erasmus University Rotterdam. Delft, A. van (1995) Spatial Ordering and Regional Policy in the Netherlands, in Velden, W. van der and Wever, E. (eds), The Economic Emancipation of the Wider Netherlands, Van Gorcum, Assen (in Dutch). Dervis, K., Melo, de J. and Robinson, S. (1982) General Equilibrium Models for Development Policy, Cambridge University Press, Cambridge UK. Dixon, P. B. and Parmenter, B. R. (1996) Computable General Equilibrium Modeling for Policy Analysis and Forecasting, in Amman, H. M. Kendrick, D. A. and Rust, J. (eds), Handbook of Computational Economics, Volume I, North-Holland, Amsterdam, 3–85.

384  References Dixon, P. and Rimmer, M. (2002) Dynamic General Equilibrium Modelling for Forecasting and Policy: A Practical Guide and Documentation of MONASH, NorthHolland, Amsterdam. Dolan, P. (2000) The Measurement of Health-related Quality of Life, in Cuyler, A. J. and Newhouse, J. P. (eds) Handbook of Health Economics, Elsevier, Amsterdam, 1723–1760. Dowrick, S. and Gemmell, N. (1988) Industrialisation, Catching Up and Economic Growth: A Comparative Study Across The World’s Capitalist Economies, Economic Journal, 101(405): 263–275. Drewnowski, J. and Scott, W. (1967). The Level of Living Index, Discussion paper, UNRISD, Geneva. Easterly, W. and Rebelo, S. (1993) Fiscal Policy and Economic Growth: An Empirical Investigation, Journal of Monetary Economics, 32(3): 417–458. Eckstein, A. (ed.) (1971) Comparison of Economic Systems: Theoretical and Methodological Approaches, University of California Press, Berkeley. Estrada, M. A. R. (2011) Policy Modelling: Definition, Classification and Evaluation, Journal of Policy Modelling, 33(1): 523–536. Evans, G. J. and Lindley, R. M. (1973) The Use of RAS and Related Models in Manpower Forecasting, Economics of Planning, 13(1–2): 53–73. Faki, H., Gumma, Y. and Ismail, M. (1995) Potential of the Sudan Irrigated Sector in Cereal Grains Production: Analysis of Various Policy Options, Agricultural Systems, 48(4): 457–483. Firth, R. W. (1967) Themes in Economic Anthropology, Association of Social Anthropologists of the Commonwealth, St. Anthony’s College, Oxford University, Tavistock Publications, London. Fisher, A. (1939) Production, Primary, Secondary and Tertiary, Economic Record, 15(1), 24–38. Fisher, F., Klein, L. and Shinkai, Y. (1965) Price and Output Aggregation in the Brookings Econometric Model, in Duesenberry, J., Fromm, G., Klein, L. and Kuh, E. (eds), The Brookings Quarterly Econometric Model of the US, North-Holland, Amsterdam, 653–679. Fogel, R. W. (2007) Capitalism and Democracy in 2040: Forecasts and Speculations, NBER Working Paper Series No. 13184, June. Fox, K. A. (1984) Behaviour Settings and Eco-Behavioral Science, Mathematical Social Sciences, 7(2): 117–138. Fox, K. A. and Miles, D. G. (1987) Systems Economics: Concepts, Models and Mu1tidisciplinary Perspectives, Ames, Iowa State University Press, Iowa. Fulpen, H. V., Herwaarden, F. V., Hooijmans, E., Kam, F. D., Pommer, E. and Wieberns, C. (1985) Accountable Policy, State Press, The Hague (in Dutch). Gigengack, A. R., Jepma, C. J., MacRae, D. and Poldy, F. (1990) Global modelling of dryland degradation, in Dixon, J. A., James, D. E. and Sherman, P. B. (eds) Dry Management: Economic Case Studies, Earth Scan Publications Ltd., London, 330–349. Graafland, J. J. and Mooij, R. A. De (1998) Analyzing Fiscal Policy in the Netherlands: Simulations with a Revised MIMIC, Research Memorandum No. 140, CPB, The Hague. Grier, K. G. and Tullock, G. (1989) An Empirical Analysis of Cross-National Economic Growth, 1951–1980, Journal of Monetary Economics, 24(2): 259–279. Haavelmo, T. (1944) The Probability Approach in Econometrics, Econometrica 12 Supplement (July), iii-iv; 1–115. Haavelmo, T. (1964) A Study in the Theory of Economic Evolution, North-Holland, Amsterdam.

References  385 Herrera, A., Scolnik, H., Chichilnisky, G. and Gallopin, G. (1974) Latin American World Model, Fundacion Bariloche, 8 volumes, International Institute of Applied Systems Analysis, Luxembourg. Hicks, J. R. (1969) A Theory of Economic History, Oxford University Press, Oxford. Hirschman, A. O. (1970) Exit, Voice and Loyalty: Responses to Decline in Firms, Organizations and States, Harvard University Press, Cambridge MA. Hirschman, A. O. (1981) Essays in Trespassing: Economics, Politics and Beyond, Cambridge University Press, London. Hoel, M. and Saether, E. M. (2001) Public Health Care with Waiting Time: The Role of Supplementary Health Care, Working Paper No. 562, CESifo, Munich. Hoff, K. and Stiglitz, J. E. (2002) After the Big Bang? Obstacles to the Emergence of the Rule of Law in Post-Communist Societies, Working Paper No. 9282, NBER, MA, October. Ianchovichina, E., Nicita, A. and Soloaga, I. (2001) Trade Reform and Household Welfare: The Case of Mexico. Working Paper No. 2386, The World Bank, DECRG-Trade. Ibrahim, O. A. (1995) Constraints of Development in Sudan: A Three-gap Model Analysis, MA thesis, Institute of Social Studies, The Hague. Jackson, G. and Deeg, R. (2006) How Many Varieties of Capitalism? Comparing the Comparative Institutional Analyses of Capitalist Diversity, Max Planck Institute for the Study of Societies, Discussion Paper No. 06/2. Japan (1965) Economic Models for Medium Term Economic Plan 1964–1968. Economic Planning Agency, Government of Japan, Tokyo. Johansen, L. (1960) A Multi-sectoral Study of Economic Growth, North-Holland, Amsterdam. Joshi, P. C. (1975) Land Reforms in India: Trends and Perspectives, Allied Publishers, New Delhi. Kemper, N. J. and Pellenbarg, P. H. (1993) Business Mobility in the Netherlands, Economisch Statistische Berichten, March 1993: 249–252; April 1995: 380–384 (in Dutch). Kevane, M. (1997) Land Tenure and Rental in Western Sudan, Land Use Policy, 14(4): 295–310. Khan, S. R. and Irfan, M. (1985) Family Background and Other Determinants of Earnings and Rates of Return to Education in Pakistan, Annual Meeting of the Pakistan Society of Development Economics, Islamabad. Knudsen, M. B., Pedersen, L. H., Petersen, T. W., Stephensen, P. and Trier, P. (1998) Danish Rational Economic Agents Model–DREAM, Version 1.2, Statistics Denmark. Koopmans, T. and Montias, J. M. (1971) On the Description and Comparison of Economic Systems, in Eckstein, A. (ed.) Comparison of Economic Systems, University of California Press, Berkeley CA, 27–28. Kopits, G. and Symansky, S. (1998) Fiscal Policy Rules. IMF Occasional Paper, No. 162 Krugman, P. R. (1981) Trade, Accumulation and Uneven Development, Journal of Development Economics, 8: 149–161. Krugman, P. (1993) How I Work, The American Economist, 37(2), 25–31. Kuipers, S. K., Jongbloed, B., Kuper, G. and Sterken, E. (1980) Sector Structure and Economic Growth: A Simple Growth Model with Six Sectors of the Dutch Economy, State Press, The Hague (in Dutch). Kumhof, M., Muir, D., Mursula, S. and Laxton, D. (2010) The Global Integrated Monetary and Fiscal Model (GIMF)–Theoretical Structure, Discussion Paper No. 41, International Food Policy Research Institute (IFPRI), Washington DC; Working Paper No. 10/34, International Monetary Fund (IMF), Washington DC.

386  References Lee, B. and J. Nellis, J. (1990) Enterprise Reform and Privatization in Socialist Economies, World Bank, Washington. Löfgren, H., El-Said, M. and Robinson, S. (1999) Trade Liberalization and Complementary Domestic Policies: A Rural-Urban General Equilibrium Analysis of Morocco. Discussion Paper No. 41, International Food Policy Research Institute (IFPRI), Washington DC. Lofgren, H., Harris, G. and Robinson, S. (2001) A Standard Computable General Equilibrium (CGE) Model in GAMS, Discussion Paper No. 75, Trade and Macroeconomic Division, International Food Policy Research Institute (IFPRI), Washington DC. Lucas, R. E. (1980) Methods and Problems in Business Cycle Theory, Journal of Money, Credit and Banking, 12: 696–715. Lucas, R. E. (1993) Making a Miracle, Econometrica, 61(2): 251–272. Lysy, F. and Taylor, L. (1977) A Computable General Equilibrium Model for the Functional Income Distribution: Experiments for Brazil, 1959–71, World Bank, Washington DC Maddison, A. (1971) Class Structure and Economic Growth: India and Pakistan Since the Moghuls, Allen and Unwin, London. Mankiw, N. G., Romer, D. and Weil, D. N. (1992) A Contribution to the Empirics of Economic Growth, Quarterly Journal of Economics, 107(2): 407–437. McKitrick, R. R. (1998) The Econometric Critique of Computable General Equilibrium Modeling: The Role of Functional Forms, Economic Modelling, 15: 543–573. Meester, W. J. (1994) Recent Changes in the Locational Preference of Dutch Entrepreneurs, paper presented at RSA European Congress, Groningen. Mitra-Kahn, B. H. (2008) Debunking the Myths of Computable General Equilibrium Models, Schwab Center for Economic Policy Analysis, SCEPA Working Papers, No. 2008-1. Morgan, M. S. and Knuuttila, T. (2012) Models and Modelling in Economics, in Mäki, U. (ed.) Handbook of the Philosophy of Economics, Elsevier, Amsterdam, 49–87. Murphy, K. and Shleifer, A. (2004) Persuasion in Politics, NBER Working Paper Series No. 10248. Myrdal, G. (1968) Asian Drama: An Inquiry into the Poverty of Nations, Pantheon, New York. Nieuwenhuis, A. (1985) Consumption Spending Distributed by Sector of Origin, Central Planning Bureau (in Dutch). Nordhaus, W. D. (2002) The Health of Nations: The Contribution of Improved Health to Living Standards, Working Paper No. 8818, National Bureau of Economic Research, Cambridge, MA. North, D. C., Wallis, J. J. and Weingast, B. R. (2006) A Conceptual Framework for Interpreting Recorded Human History, NBER Working Paper Series No. 12795. NZI (1998) De Intramurele Gezondheidszorg in Cijfers, (Health statistics), Institute for Health Care Management (NZI), Utrecht. OECD (1993) Valuation and Privatisation, OECD, Paris. Oxley, H. and MacFarlan, M. (1995) Health Care Reform: Controlling Spending and Increasing Efficiency in OECD Economic Studies, 24(1995/1): 7–55. Persson A and Munasinghe M. (1995) Natural Resources Management and Economywide Policies in Costa Rica: A Computable General Equilibrium Modelling Approach, The World Bank Economic Review, 9(2): 259–285. Priemus, H. et al. (1995) Housing Construction and Spatial Developments, in Velden, W. V. D. and Wever, E. (eds) The Economic Emancipation of the Wider Netherlands, Van Gorcum, Assen (in Dutch). Psacharopoulos, G. (1980) Returns to Education: An Updated International Comparison, Education and Income, World Bank Staff Working Paper, No. 402, World Bank, Washington DC.

References  387 Pyatt, G. and Roe, A. (1977) Social Accounting for Development Planning, Cambridge University Press, Cambridge. Ratto, M., Roeger, W. and Veld, J. (2006) Fiscal Policy in an Estimated Open-economy Model for the Euro Area, Economic Papers, 266, EU-DSGE, Brussels. Reed, D. (ed.) (1992) Structural Adjustment and the Environment, Westview Press, Boulder. Robinson, S., Yu´nez-Naude, A., Hinojosa-Ojeda, Lewis, R. J. and Devarajan, S. (1999) From Stylized to Applied Models: Building Multisector CGE Models for Policy Analysis, North American Journal of Economics and Finance, 10(1): 5–38. Sarcevic, P. (1992) Privatization in Hungary, Past, Present and Future, Graham and Trotman, London. Scarf, H. (1969) An Example of an Algorithm for Calculating General Equilibrium Prices, American Economic Review, 59(4): 669–677. Scarf, H. and Shoven, J. B. (eds) (1984) Applied General Equilibrium Analysis, Cambridge University Press, Cambridge UK. Sen, A. K. (1963) Neo-classical and Neo-Keynesian Theories of Distribution. Economic Record, 39, 46–53. Sen, A. K. (1982) Choice, Welfare and Measurement, MIT Press, Cambridge Mass. Simon, H. A. (1953) Causal Ordering and Identifiability, in Hood, W. C. and Koopmans, T. C. (eds): Studies in Econometric Method, Cowles Commission Monograph No. 15, New York. Simon, H. (1993) Altruism and Economics, American Economic Review: Papers and Proceedings, 83(2): 156–161. Smith, A. (1776) An Inquiry into the True Nature and Causes of the Wealth of Nations, 1961 edn, Cannon, E. (ed.) London, Solow, R. M. (1956) A Contribution to the Theory of Economic Growth, Quarterly Journal of Economics, 70(1): 65–94. Sprout, R. V. A. and Weaver, J. H. (1992) International Distribution of Income: 1960– 1987, Kyklos, 45(2): 237–258. Stiglitz, J. (2008) We Aren’t Done Yet: Comments on the Financial Crises and Bailout, The Economists’ Voice, 5(5): 11. Stone, J. R. N. (1947) On the Interdependence of Blocks of Transactions, Journal of the Royal Statistical Society (Supplement), 9(1): 1–45. Stone, J. R. N. (1978) The Disaggregation of the Household Sector in the National Accounts, paper presented at the World Bank Conference on Social Accounting Methods in Development Planning, Cambridge. Sugden, R. (2002) Credible Worlds: The Status of the Theoretical Models in Economics, in Mäki, U. (ed.) Fact and Fiction in Economics: Models, Realism, and Social Construction, Cambridge University Press, Cambridge, 107–136. Summers, R. and Heston, A. (1988) A New Set of International Comparisons of Real Product and Price Levels, Estimates for 130 Countries, Review of Income and Wealth, 34(1): 1–25. Taylor, L. (1990) Structuralist CGE Models, in Taylor, L. (ed.), Socially Relevant Policy Analysis, MIT Press, Cambridge MA, 1–70. Taylor L. (1991) Income Distribution, Inflation and Growth: Lectures on Structuralist Macroeconomic Theory. MIT Press, Cambridge MA. Theil, H. and Seale, J. L. (1994) The Geographic Distribution of World Income, 1950– 1990, De Economist, 142(4): 387–419. Thorbecke, E. (1990) Adjustment, Growth and Income Distribution in Indonesia, OECD Development Center, Paris. Thorner, D. (1956) The Agrarian Prospect in India, University Press, Delhi.

388  References Thurow, L. (1969) Poverty and Discrimination, Brookings Institution, Washington DC. Tinbergen, J. (1956) Economic Policy: Principles and Design, North-Holland, Amsterdam. Tinbergen, J. (1962) Shaping the World Economy: Suggestions for an International Economic Policy, The Twentieth Century Fund, New York. Tinbergen, J. (1963) On the Theory of Economic Policy, North-Holland, Amsterdam. Tinbergen, J. (1970) Planning and Economic Forecasting, in Research Priorities in Quantitative Socio-Economic Development Planning, International Institute of Quantitative Economics, Montreal. Torrance, G. W. (1986) Measurement of Health Status Utilities for Economic Appraisal, a review, Journal of Health Economics, 5(1):1–30. Tudor Edwards, R. (1997) NHS Waiting Lists: Towards the Elusive Solution, Office for Health Economics, London. United Nations (1966) Preliminary Report of Targets on Social Development, ECOSOC, 1966. United Nations (1971) Second United Nations Development Decade, A System of Overall Review and Appraisal of the Objectives and Policies of the International Development Strategy, ECOSOC, 1971. Wagstaff, A. (1991) Health Care: QALYs and the Equity-Efficiency Tradeoff, Journal of Health Economics, 10(1): 21–41. Wilson, D. and Puroshothaman, R. (2003) Dreaming with BRICS: The Path to 2050. Goldman Sachs, Global Economics Working Paper No. 99, October, at https://www. gs.com World Bank (2005) Education in Ethiopia, World Bank, Washington. World Bank (1997) The State in a Changing World: World Development Report 1997, Oxford University Press, Oxford. Zymelman, M. (1980) Forecasting Manpower Demand, Education Department, World Bank, Washington DC.

Index  389

Index

Abdelgalil, E. 159, 372–4 Acemoglu, D. 346 Acharya, A. 136–7, 372 Adelman, I. 9, 91 adjustor variables 21–2, 28–9, 32–, 42–3, 86–8, 79, 97, 122–3, 139, 146, 171–2, 243–55, 368 agriculture: models, policies, performances 44–257 Ajani, G. 303 Allende regime 6–7, 44, 58, 60 Armington function 141, 144–5 Arrow, K.J. 250 Ballard, C. 128 Bandara, J.S. 136 Barbier, E. 159 Barr, N. 318 Barro, R.J. 334, 346, 377 Baumol, W.J. 336, 377 Bayens, R. 344 Beck, T. 346 behavioural settings 17–18, 346–50, 365 Bennett, M.K. 65 Bennett, J. 374 Berg, L.vd 374 Bleaney M. 112 Bluffstone R. 159 Bovenberg, L. 128 Braber, R. 367 budget deficit 14, 26, 30, 37–8, 75, 120–2, 139, 143, 148–9, 156–8, 208–15, 241–3, 254–7 Byung-Nak Song 370 causal ordering 8–9, 29–39, 51–2, 84–7, 96–7, 123, 146, 368–9, 371 CET function 141, 145 Chenery, H.B. 9, 91, 338–9 Chitiga, M. 137

Clark, C. 338–9 clearance mechanisms 28–9 Cockburn, J. 138 Cohen, S.I. 9, 53, 91, 197, 201, 344, 358, 366–72, 374, 376–7, 380 conditional convergence 17, 334, 345 consumption functions 44, 50, 71, 75, 80, 112, 141, 206, 247, 251 cooperative, non-cooperative 363, 365 coordination mechanisms 308, 347–9 Coxhead, I. 373 Credit: crunch 380; policies 45, 53–7, 60–1 critical mass 356 Cromwell, I. 159 Culyer, A.Y. 318 Czysewski, A. 367 Decaluwe, B. 115 decomposition of: exogenous stimulus 236; growth performance 14, 224, 233–6, 238; multiplier effects 99, 102–5, 197; planning model 65 Deeg, R. 346 degradation cost 163, 172–84, 374 De Groene, J.H. 375 Delft, A. van 231 Dervis, K. 115 development strategies 6, 9, 91, 96, 291 discounted value, discount rate 47, 56–7, 164, 169, 299, 310–12, 311–12, 377 distribution of: agents 17, 59, 320, 352–3, 363–5, 269, 375, 378; assets 54, 60; capital 41, 127, 131, 209–10, 310, 370; income 9, 13, 58, 82, 86, 98–9, 104–6, 109–15, 125–7, 132, 136–9, 148–50, 156, 186, 196–9, 203, 214–15, 218–20, 226–8, 232–3, 249–56, 277, 340, 348; labour 40, 153, 289, 300, 323; land 49, 52, 54, 57, 59, 169, 184; power 5, 51, 53, 62, 363; production 86, 98, 228, 232

390  Index Dixon, P. 128 Dolan, P. 318 Dowrick, S. 377 Drewnowski, J. 369 dualism, formal informal sectors 9, 109–10, 285–6 Dutch miracle 240 earning functions 298–300, 322, 377 Easterly, W. 370 EBRD 361 Eckstein, A. 346 economic transformation 202, 275–8, 308, 346–50, 353–4, 357, 362, 365, 378 education: labour markets 284–302; roadmap planning 256–83; and system modeling 70–90, 129–30 effects: transfer, open-, closed-loop 99, 102–6; productivity 19, 67–9, 82–6, 90, 370; trickle down 5, 13, 109, 111 employment, unemployment: 5, 8–9, 14, 17, 21, 26, 30, 40, 64–75, 80–1, 86–8, 90, 118, 130, 133, 136, 153, 234, 239, 241–3, 250–7, 274, 284–96, 306–7, 311–13, 368–371, 375–6 environmental sustainability 6, 10, 159 equilibrium: see markets Estrada, M.A.R. 366 European Union 12, 241–2, 360, 364–4 firm settings 347, 351–7, 360, 378–9 Firth, R. 353 Fisher, A. 339 Fisher, F. 369 Fogel, R.W. 364 food: consumption 122, 194, 218–19; subsidies 67, 84; food processing 312–14 forestry: modeling, policies, results 159–84, 372, wood processing 312–14 Gemmell, N. 377 Gigengack, A.R. 373 goods: private, public 17, 317, 334, 379 government revenue 14, 26, 40, 70, 75, 120, 140–6, 148, 156, 165, 170–1, 181–2, 207–8, 242–6, 249–56, 304, 311 government spending 8, 67, 93, 113, 212–18, 221, 240, 246, 249, 252, 261–3, 339, 375 Graafland, J.J. 241 Greenaway, D. 112 Green GDP 12, 163, 167, 177–8, 183–4 Grier, K. G. 370

growth and distribution:5–6, 9–12, 83, 91–2, 99, 106, 109–12, 136, 220–3, 228–32, 366 Haavelmo, T. 2, 18, 333 health: modeling, policies, results 17, 66–90, 149, 254, 316–32 Heritage Foundation 360–1 Herrera, A. 371 Heston, A. 333 Hicks, J.R. 91, 353 Hirschman, A. 367, 378 Hoel, M. 329 Hoff, K. 361 household settings 347, 352–3, 356–60, 379 housing 66–70, 77–8, 80–4, 90, 231, 316, 371 human capital: policies 12, 152, 178–84, 241, 328, 334; theory 285, 298–302, 325 human resource matrix 323 Ianchovichina, E. 136 Ibrahim, O.A. 373 IMF 11–12, 136, 242, 366 index of: attained sustainable productivity (ASP) 19, 162–8, 180–2; basic needs attainment (BNA) 64–72, 76–7, 85, 88, 370; business competitiveness 360–1; economic freedom 361; educational development 273, 278; equality 277; gainers and losers (GLI) 105–9, 195–8, 225–33, 237–8; Gini 81; global integration 361; human development (HDI) 77, 258, 375; industrialisation 276; interactive influence (III) 19, 356–7, 363–4, 379; Paasche price 228; policy modeling consistency 366; rule of law 361 industry: models, performances, policies 44–257; privatisation 311–14 investment goods: installed, inventories 24–6, 30–1, 35–6, 70–1, 75, 79, 86, 118, 120–2, 126, 131, 139, 143, 146, 170–2, 367, 371–2; in transition 304–11, 314 investment functions: private 2–6, 71, 80, 120, 141, 372–3; public 31, 50, 55, 60–1, 67, 71, 75, 83, 86, 141–3, 70, 372–3 Jackson, G. 346 Jayasuriya, S. 373 job competition 285, 298–302 Johansen, L. 36, 114, 368 Joshi, P.C. 369

Index  391 Kemper, N.J. 231 Kevane, M. 373 Knudsen, M.B. 128 Knuuttila, T. 2 Koopmans, T. 113, 186 Kopits, G. 374 Krugman, P. 18, 334, 366 Kumhof, M. 242 Kuznets Curve 109, 277 labour imbalances by occupation, education: 288–92. See also employment, unemployment Laeven, L. 346 land: reform 54–6, 60–1; tax 47, 54–7, 366 landlord–peasant relationship 44–62 leakages 6, 195, 371 Lee, B. 303 Lee, J.W. 344 liberalisation reforms 6, 10–12, 150–2. See also international trade linkages 6, 9, 109, 128, 150, 219, 225, 231, 330, 338, 342, 372 livestock: modeling, policies, results 159–84, 372–4 Lofgren, H. 137–8 Lucas, R.E. 2, 18, 334 Lysy, F. 113 Maddison, A. 378 Mankiw, N.G. 334 market failures 6, 317, 362 market incentives 162–6 markets: factors 21, 25, 41–3, 72. 93, 116, 123, 132, 138, 142, 152, 207, 243–8, 368; financial 27, 79, 96, 123, 139, 146–7, 172, 244; labour 15–16, 81, 142, 207, 241–2, 274, 284–302; land 46–7, 169 markets: products, commodities 26–7, 31–40, 43, 78–9, 95, 120–2, 137–9, 144–5, 162, 174, 184, 208–11, 242–3, 247–9, 252, 369, 373 Marshall, A. 1 Martens, A. 115 McKitrick, R.R. 115 Meer, J. vd 374 Meester, W.J. 231 Mitra-Khan, B.H. 115 model closures 18–19, 23, 28, 32–9, 87–8, 113, 123, 138–41, 146–8, 151, 157, 160–2, 174–7, 210–11, 223, 243, 248, 368; with leaders, followers 43–5, 50–1, 62–3 model structures 62, 89, 137

models: analytical, planning forms 84–90; combined econometric multisector CEM 22–9, 31–4, 42–2, 69, 114; computable general equilibrium CGE 4, 6, 10–14, 19, 22–8, 31–42, 114–84, 202–22, 240–57; cost benefit 303–15; iterative fitting 284–302; normed planning 258–83; random sampling 316–32; social accounting SAM 4, 6, 9–23, 41–2, 91–113, 185–96, 223–39 Monte Carlo methods 17, 317, 320, 329 Montias, J.M. 113, 186 Mooij R.A. de 241 Morgan, M.S. 2 multipliers: Leontief versus SAM 102–4, 189; see also decomposition of multiplier effects Munasinghe, M. 159 Murphy, K. 378 Musgrave, R. 336 Myrdal, G. 7–8, 44, 53, 368 needs: basic 8–9, 46–7, 50, 64–72, 76–8, 85, 88–90, 161, 164, 166, 176, 178–84, 369–70, 373; personal, collective 353–6 negotiations 16, 136, 252, 307 Nellis, J. 303 Nordhaus, W.D. 319 North, D.C. 346, 378 nutrition: modeling, policies, results 66–72, 76–7, 80–1, 86 OECD 346, 360, 364, 376 Pareto optimality 186 Parmenter, B.R. 128 Peacock, A.T. 336 Pellenbarg, P.H. 231 Persson, A. 159 persuasion settings 352–5, 362, 365 poverty reduction 18, 111, 138, 186 prices: composite 39–40, 121, 138, 143–9, 151, 157, 205, 208–9, 244, 247, 372; fixed, flexible 202–20; relative 13, 19, 32, 42, 69, 112, 119, 121, 138, 150, 160–8, 174–8, 184, 202–3, 228, 236, 239, 373 privatisation decisions: applications 311–15; modeling 301–11, production functions: CES 25, 31, 33; Cobb-Douglas 25, 31–2, 68, 71–4, 116–19, 142–3, 165–7, 170, 206–8, 245, 250–1, 255, 367; Harrod-Domar 25, 43, 47, 49, 167, 367–8; Leontief 26, 41, 49, 73–4, 118, 142, 167, 203–7, 367

392  Index property rights 11–12, 159–61, 167, 178–84, 304, 360, 372 pro-poor economic growth 10–11, 137, 150, 158 purchasing power parity PPP 259–61, 264, 271, 283, 375 Puroshothaman, R. 364 Pyatt, G. 9, 91 QALY 317–30, 377 Quesnay, F. 1, 91 RAS method 19, 95, 152, 191, 224, 285, 290–3, 302 Ratto, M. 241 Rebelo, S. 370 Reed, D. 159 rent: appropriation 351; seeking 347, 354, 379 Rimmer, M. 128 Robinson, S. 9, 113, 115 Robinson, J. 346 Roe, A. 91 Saether, E.M. 329 Sarcevic, P. 303 Scarf, H. 115 Scott, W. 369 Seale, J.L. 377 Sen, A.K. 28 services: models, policies, performances 44–257 Shleifer, A. 378 Shoven, J.B. 115 Simon, H. 8, 29, 51, 69, 84, 90, 123, 368–9, 371, 379 Smith, A. 368 social indicators 64–5 social queues 16–19, 316–17, 328, 330 soft state 5, 7, 44, 53 Solow, R.M. 334 Sprout, R.V.A. 377 state settings 346–8, 351–61, 365, 378–9 Stiglitz, J. 346, 361 Stone, J.R.N. 9, 91, 103 Summers, R. 333 Symansky, S. 374 Syrquin, M. 338 systems competition 363, 365

Tableau Economique 1, 91 taxes: direct 26, 75, 86–9, 93–5, 118–20, 143, 207, 242, 246, 252–6, 305–7, 311, 322–6; indirect 25, 40, 67, 72–3, 75, 94–5, 118, 124–5, 139, 144, 165, 170, 204, 207, 252, 256, 361, see also land tax Taylor, L. 62, 113, 115, 373 Theil, H. 377 Theil’s coefficient 249–51, 253, 256, 293, 296, 302 Thorbecke, E. 9, 91, 371 Tinbergen, J. 2, 7, 8, 18, 44, 89, 366, 368 Torrance, G.W. 318 trade: foreign 17, 32, 37, 43, 67, 116, 121–2, 126, 170–2, 247, 252, 276, 297, 360–1, liberalization 11, 136–58 trade-offs: efficiency and equity in health care 317, 329; growth and distribution 9–10, 92, 109, 219, 223; growth and environment 159–60, 163, 173, 180–4; short and long-run benefits 83 transaction value 304–7, 311, 314 Tudor Edwards, R 377 Tullock, G. 370 Tuyl, J.M.C. 374 UN/CDP 5, 7, 9 UNDP 15, 78, 258, 262 UNESCO 262, 287 UNIDO 258 UN Population Division 262, 364 UNRISD 8, 65, 366, 369 Walras Law 28, 79, 147, 172, 211, 248, 369, 373 Wagner’s Law 336 Wagstaff, A. 318 Watanabe, T. 339 Weaver, J.H. 377 Wilson, D. 364 Winpenny, J. 159 Wiseman, J. 336 World Bank 9, 11–12, 114–15, 136, 190, 262, 275, 340, 362, 364, 375–6 Yu, W. 136 Zymelman, M. 289