Models for Energy Policy (Routledge Studies in the History of Economic Modelling, 1)

Models for energy policy

Energy policy is a key area in each of the world’s economies. The oil shocks of the 1970s emphasized how important energy had become. In recent years a growing awareness of environmental issues has had a major impact on perceptions of energy use, as growing numbers of people express concern at the relationship between energy and the greenhouse effect, acid rain and the depletion of the ozone layer. All of this has created a demand for more, and better, models of energy use. This book analyses a range of these models. It covers: ● ● ● ● ●

short-, medium- and long-term forecasting regional, national and international impacts models looking at individual sources of energy and the energy sector as a whole environmental issues formal models and non-formal approaches

The interaction between energy and the environment is singularly complex, and throughout the book integrates physical, technical, economic and social concerns. Jean-Baptiste Lesourd is at GREQAM, Universities of Aix-Marseille. Jacques Percebois is at the Faculté des Sciences Economiques, University of Montpellier-I, Montpellier. François Valette is at CNRS, Montpellier.

Routledge Studies in the History of Economic Modelling 1 Models for energy policy Edited by Jean-Baptiste Lesourd, Jacques Percebois and François Valette

Models for energy policy Edited by Jean-Baptiste Lesourd, Jacques Percebois and François Valette

London and New York

First published 1996 by Routledge 11 New Fetter Lane, London EC4P 4EE This edition published in the Taylor & Francis e-Library, 2005. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 © 1996 Association Econometrie Appliquée All rights reserved. No part of this book may be reprinted or reproduced or utilized 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. 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 Models for energy policy/edited by Jean-Baptiste Lesourd, Jacques Percebois, and François Valette. p. cm. —(The New international studies in economic modelling) Includes bibliographical references and index. ISBN 0-415-12975-3 1. Energy policy—Mathematical models. I. Lesourd, Jean -Baptiste. II. Percebois, Jacques. III. Valette, François. IV. Series. HD9502.A2M634 1994 333.79–dc20 95–813 CIP ISBN 0-203-98252-5 Master e-book ISBN

ISBN 0-415-12975-3 (Print Edition) ISSN 1 359-7973 (Print Edition)

In memoriam

Nikitas Deimezis, economist at the Directorate-General For Energy (D. G. XVII) of the EC, passed away on 12 May 1993. He held a degree in Engineering from the Polytechnical School of Athens, and a Doctorat d’Etat ès-Sciences Economiques from the University of Paris-Dauphine. In addition to being an excellent economist, well known for his modelling work in energy economics, he was a kind and amiable colleague.

The editors of this book are greatly indebted to him.

Contents

In memoriam—Nikitas Deimezis

v

List of figures

viii

List of tables

xi

Contributors

xiii

Acknowledgements

xv

Introduction J.B.Lesourd, J.Percebois and F.Valette

xvi

Part I New models and new modelling methodologies 1

Methodological advances in energy modelling: 1970–90 J.Griffin

2

International markets and energy prices: the POLES model P.Criqui

10

3

ERASME: a short-term energy forecasting model for the European Community N.Deimezis

23

4

The energy model MIDAS P.Capros, P.Karadeloglou, L.Mantzos and G.Mentzas

34

5

Biproportional methods and interindustrial dynamics: application to energy demand in France L.de Mesnard

54

6

Gas contract portfolio management: experiments with a stochastic programming approach A.Haurie, Y.Smeers and G.Zaccour

71

7

Modelling the European gas a market: a comparison of several scenarios J.Percebois and F.Valette

89

2

Part II Application to particular energy policy problems 8

World energy outlooks E.Vouyoukas

106

9

Energy shocks and the demand for energy P.S.Andersen and H.J.Bernard

116

vii

10

Modelling the petroleum spot market: a vector autoregressive approach W.C.Labys, V.Murcia and M.Terraza

133

11

Fiscal harmonization on oil products within the EC: problems and prospects J.-B.Lesourd and D.Meulders

152

12

Energy tax increases as a way to reduce CO2 emissions V.Detemmerman, E.Donni and P.Zagamé

168

13

Energy conservation and economic performance in Japan: an econometric approach K.Ban

185

14

Electricity futures markets E.Hope, L.Rud and B.Singh

202

Index

211

Figures

I.1 I.2 I.3 I.4 I.5 I.6 I.7 I.8 I.9 I.10 I.11 I.12 2.1 2.2 2.3 2.4 2.5 2.6 3.1 3.2 3.3 3.4 3.5 4.1 4.2 4.3 4.4 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10

POLES (Criqui) ERASME (Deimezis) MIDAS (Capros, Karadeloglou, Mantzos, Mentzas) Biproportional methods (de Mesnard) Gaz contracts model (Haurie, Smeers, Zaccour) SOSIE/gas market (Percebois, Valette) Tax harmonization in the EC (Lesourd, Meulders) HERMES (Donni, Zagame, Detemmerman) Energy conservation in Japan (Ban) Energy shocks (Andersen, Bernard) Electricity future markets (Hope, Rud, Singh) Modelling overall OPEC pricing 1985–8 Non-linear price adjustments Oil price function in the POLES model Hysterisis in oil price dynamics Import prices for natural gas and oil (CIF, 1975–90) Price of imported boiler coal in Europe and the OPEC price of oil (CIF, 1977–90) The ERASME model Real GDP (Index 1985=100): actual versus forecast Crude oil import prices: actual versus forecast Degree-days: difference from ‘normal’ Crude oil production: actual versus forecast Linear approximation of the load duration curve Allocation of plants in the electricity submode The representative refinery The coal supply curve Productive vertex: ‘Solid Mineral Fuels and Coke’ (Industry analysis) IIS (Industry) cumulated with households and rest of world (Ext.) Productive vertex: ‘Solid Mineral Fuels and Coke’ (Product analysis) IIS (Product) cumulated with households and rest of world (Ext.) Productive vertex: ‘Oil Products, Natural Gas’ (Industry analysis) Calculated evolution of value-added IIS (Industry) cumulated with households and rest of world (Ext.) Calculated evolution of imports Productive vertex: ‘Oil Products, Natural Gas’ (Product analysis) IIS (Product) cumulated with households and rest of world (Ext.)

xix xix xx xx xx xx xx xx xx xx xx xx 13 13 14 15 16 18 25 27 28 29 30 41 42 45 50 59 59 60 61 62 63 64 64 65 65

ix

5.11 5.12 5.13 6.1 6.2 6.3 7.1 7.2 7.3 7.4 7.5 7.6 7.7

Productive vertex: ‘Electricity, Gas and Water’ (Industry analysis) IIS (Industry) cumulated with households and rest of world (Ext.) Productive vertex: ‘Electricity, Gas and Water’ (Product analysis) Event tree Scenarios Base points Gas flows represented by the SOSIE-GAZ model Sub-model of the component elements of a country’s gas supply Diagram of all the modules comprising SOSIE-GAZ Seasonal demand in each country, in principle Seasonalized demand in the countries under consideration Functioning of the market in a stationary regime Rupture of deliveries of ex-USSR gas for six months, without a free market (France, Germany, Italy) 7.8 Rupture of deliveries of ex-USSR gas for six months, without a free market (Germany) 7.9 Rupture of deliveries of ex-USSR gas for six months, without a free market (France) 7.10 Doubling of consumption over ten years, set of hypotheses A 7.11 Doubling of consumption over ten years, set of hypotheses B 8.1 Model structure 8.2 Crude oil price (in constant US$) 8.3 World energy market shares (%) 8.4 Energy intensity (1975=100) 8.5 World energy shares (%) 8.6 World oil supply by region (MBD) 9.1 Relative energy prices: in relation to the GDP deflator (1972=100) 10.1 The VAR modelling approach A10.1(a) Crude oil price A10.1(b) Crude oil stocks A10.1(c) Crude oil consumption A10.1(d) Crude oil production A10.1(e) Crude oil exports A10.1(f) Crude oil imports A10.2(a) Oil prices (UVAR) A10.2(b) Oil stocks (UVAR) A10.2(c) Oil production (UVAR) A10.2(d) Oil consumption (UVAR) A10.2(e) Oil exports (UVAR) A10.2(f) Oil imports (UVAR) A10.3(a) Oil prices (RVAR) A10.3(b) Oil stocks (RVAR) A10.3(c) Oil production (RVAR) A10.3(d) Oil consumption (RVAR) A10.3(e) Oil exports (RVAR) A10.3(f) Oil imports (RVAR) A10.4(a) Experimental oil prices (UVAR)

66 68 68 73 74 76 94 95 96 97 98 99 100 100 100 101 101 107 108 109 111 112 113 126 134 144 144 144 145 144 145 145 146 146 146 147 147 147 148 148 148 149 149 149

x

A10.4(b) Experimental oil stocks (UVAR) A10.4(c) Experimental oil production (UVAR) A10.4(d) Experimental oil consumption (UVAR) A10.4(e) Experimental oil exports (UVAR) A10.4(f) Experimental oil imports (UVAR) 11.1 Evolution of proceeds with respect to taxation 13.1 Real GNP and energy consumption 13.2 Energy price and energy/GNP ratio 13.3 Consumption and demand for petrol 13.4 Impact on demand for petrol 13.5 Energy used to produce 1 tonne of crude steel 13.6 Impact on the iron and steel industry 13.7 Primary energy for electric utilities 13.8 Impact on electric utilities 13.9 CO2 emissions in Japan (276 million tons of carbon in 1988) 13.10 CO2 emissions 13.11 Impact on GNP growth 13.12 Welfare losses (reduction of GNP in 1985 prices) 14.1 The structure of the Norwegian power sector 14.2 Principal sketch of the electricity system

150 150 150 151 151 161 186 186 188 190 190 193 194 196 197 198 199 200 203 205

Tables

2.1 3.1 3.2 3.3 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 7.1 7.2 8.1 8.2 8.3 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 10.1 10.2 10.3 10.4 10.5

Simplified trade matrices, market share of natural gas and coal exports 20 STEO—Main features 24 STEO—Error analysis 29 Forecast report, actual versus forecasts: total apparent consumption in Mtoe and % 31 Segments and contract length 77 Other economic parameters 77 Base point netbacks 78 Base point bounds on total commitments 78 Optimal value 79 Past commitments 79 Results (selected) case 1: base case with past commitments 80 Results (selected) case 2: no past Commitments 81 Results (selected) for high-capacity expansion costs 82 Results (selected) for low-capacity expansion costs 82 Main flows of natural gas in Europe 91 Principal descriptors of consumption 92 World primary energy requirements (million tonnes of oil equivalent) 110 Deviation of $21 per barrel crude-oil price case from the $35 per barrel price case (1990 prices) 114 Impact in 2005 of an OECD-wide carbon tax on OECD energy consumption and emissions (% 115 deviation from reference case levels) Demand for energy by industry (in log levels, annual data, 1960–88) 118 Short-term equations of energy demand by industry (variables expressed as changes in log levels) 119 Demand for energy in commercial and residential sectors (expressed in log levels, annual data, 121 1960–88) Short-term demand for energy in commercial and residential sectors (expressed in changes in log 122 levels, annual data, 1960–88) Demand for energy in transportation (expressed in log levels, annual data, 1960–88) 124 Short-term equations of energy demand in transportation (variables expressed in log levels) 125 Demand for energy in industry in periods of rising energy prices (expressed in log levels) 127 Short-term adjustment equations for energy demand in industry in periods of rising energy prices 128 (expressed as changes in log levels) Selection of optimal lag order 136 Direction of influence confirmed by the UVAR(8) model 139 Selected optimal lag length for the RVAR model 139 Direction of influence confirmed by the RVAR model 141 RMSE values for the UVAR (8) and the RVAR models 141

xii

11.1 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on petrol (premium) 153 within the EC (1979–91) 11.2 Amounts of VAT (%) and of excise duties (ED) (National Currency Units) on unleaded petrol (Euro154 Super 95 RON) within the EC (February 1991) 11.3 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on diesel oil within 156 the EC (1979–91) 11.4 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on heating oil within 157 the EC (1979–91) 11.5 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on heavy fuel oil 158 within the EC (1979–91) 11.6 Evolution of excise duty proceeds on oil products (in real terms, deflator: consumers’ price index)159 within EC-9 (1975–86) 11.7 Evolution of excise duties between December 1985 and February 1991 within EC-12 159 11.8 Retail and wholesale prices (ECU, upper line) and tax-free prices (ECU, in parentheses) for the oil160 products under study (February 1991) 11.9 Apparent elasticities of demand (private consumption) 1995 162 11.10 Harmonization proposals of the EC Commission for excise duties on oil products (1987 and 1991)163 11.11 Percentage increases for excise duties on some oil products resulting from the 1991 harmonization164 proposals 11.12 Retail and wholesale prices (ECU, upper line) and tax-free prices for the oil products under study 166 after taking into account 1991 harmonization proposals (on the basis of VAT rates and international prices of February 1991) 12.1 Variation of the energy intensity in the production sector 171 12.2 Reduction of energy share in household consumption 171 12.3 Tax on final energy consumption 172 12.4 New behaviour towards energy-saving investments 175 12.5 Energy tax financing energy-saving investment 178 12.6 Energy tax combined with a reduction of social security contributions of employers 179 12.7 Energy tax combined with a reduction of personal income tax 180 12.8 Synthesis scenarios 182 13.1 Estimated price elasticities (coal, coke, fuel, oil, capital stock) 192 13.2 Estimated price elasticities (coal, oil, LNG, power plants) 195 13.3 Impact of an increase in nuclear energy 195 13.4 The impact on energy use 198 13.5 The impact on GNP components in 2000 199

Contributors

P.S.Andersen (Chapter 9) is a Senior Economist at the Bank for International Settlements, Basel. K.Ban (Chapter 13) is a Professor at Osaka University. H.J.Bernard (Chapter 9) is Senior Economist at the Bank for International Settlements, Basel. P.Capros (Chapter 4) is a Professor at the National Technical University of Athens. Patrick Criqui (Chapter 2) is a Research Professor at the Institut Economique et Politique de l’Energie (IEPE), Grenoble. Nikitas Deimezis (Chapter 3) was a Senior Economist, Directorate-General for Energy (D.G. XVII) of the Commission of the European Communities. V.Detemmerman (Chapter 12) is an energy expert based in Brussels. E.Donni (Chapter 12) is a Senior Economist, Directorate-General for Energy (D.G. XVII) of the Commission of the European Communities. James M.Griffin (Chapter 1) is Cullen Professor of Economics, Texas A&M University. Alain Haurie (Chapter 6) is Professor of Operations Research, Ecole des Hautes Etudes Commerciales (HEC) at the University of Geneva. Einar Hope (Chapter 14) formally Professor, Norwegian School of Economics, Bergen. P.Karadeloglou (Chapter 4) is a Senior Economist, Bank of Greece, Economic Research Department, Athens. Walter C.Labys (Chapter 10) is Professor of Resource Economics, West Virginia University. Jean-Baptiste Lesourd (co-editor; Chapter 11) is a Research Professor, Groupe de Recherche en Economic Quantitative d’Aix-Marseille (GREQAM), Marseille. L.Mantzos (Chapter 4) is a Researcher, National Technical University of Athens. G.Mentzas (Chapter 4) is a Professor, National Technical University of Athens. D.Meulders (Chapter 11) is Professor of Economics, Free University of Brussels. Louis de Mesnard (Chapter 5) is Professor in the Faculty of Economics at the University of Dijon. Véronique Murcia (Chapter 10) is a Researcher, University of Montpellier I. Jacques Percebois (co-editor; Chapter 7) is Professor of Economics, University of Montpellier I. Linda Rud (Chapter 14) is a Senior Researcher at the Foundation for Research in Economics and Business Administration, Bergen. B.Singh (Chapter 14) is a Senior Researcher at the Foundation for Research in Economics and Business Administration, Bergen.

xiv

Y.Smeers (Chapter 6) is a Professor at the University of Louvain-La-Neuve. Michel Terraza (Chapter 10) is a Senior Lecturer in Economics, University of Montpellier I. François Valette (co-editor; Chapter 7) is a Research Engineer, Centre National de la Recherche Scientifique (CNRS), Montpellier. E.Lakis Vouyoukas (Chapter 8) is a Senior Economist at the International Energy Agency, Paris. G.Zaccour (Chapter 6) is Professor of Mangement, Ecole des Hautes Etudes Commerciales (HEC), University of Montreal. P.Zagamé (Chapter 12) is Professor or Economics, Panthéon-Sorbonne University and Ecole Centrale de Paris, Paris.

Acknowledgements

We are greatly indebted to the following persons and organizations, who have been of great help during the preparation of this work: ● ● ● ● ●

the EC (Directorate-General XVII) (Kevin Leydon); the EC (Directorate-General XII) (Pierre Valette); the French Energy Agency, and the French Caisse Nationale de l’Energie (Jean Matouk); the City and the District of Montpellier, and the Conseil Général de l’Hérault (France); and the French National Research Center (CNRS), and the University of Montpellier.

Introduction J.B.Lesourd, J.Percebois and F.Valette

I.I MODELLING CONTEXT As this book is being published, an important invitation of tenders is made by the Directorate General XII of the European Community in order to specify the main orientations of its fourth Framework-Programme of Research and Development. As for its non-technological part, the titles of the main chapters of the ‘energy’ section of this programme are: ‘analysis of R&D policy options’; ‘socio-economic research for energy’; ‘modelling’; ‘energy-environment economic forum’; ‘research of synergy’ (between national policies in all these fields). We shall deal more or less directly with all these subjects in the following chapters. Of course such a convergence between the themes of two aspects of the same current events is not surprising. It merely confirms the need, felt by an increasing number of agents, to commit large and diversified funds to economic studies and research likely to clarify the decisions that remain to be taken in the energy field, a field that will obviously still be very ‘sensitive’ during the next decade with old and new problems concerning economics, the environment and employment. Beyond this consideration, we should point out that, through its budgetary choices, the invitation of tenders mentioned above shows an evolution of the social demand which by definition it tends to meet. Of course it confirms some important concerns of the most recent community programmes, such as the improvement of the performance of energy plants, and more generally the rational use of energy. It also expresses new tendencies, with some leading elements (new energies), others that stagnate (thus nuclear energy has apparently found its right place now, and its directions for the near future, even if for many people the debate is not closed concerning its security and wastes), and still others that are regressing (fossil fuels). But in this evolution, the change in the objectives of the contemplated studies is probably more essential than the popularity of the various possibilities of implementation of the suggested research and development. On this point, it should be noted that now people are expecting more ‘comprehensive’ evaluations of energy policies, including, in addition to financial considerations, thorough analyses of their impacts on employment and the environment, at various levels (particularly local, European and global—and no longer only national level). In this context, modelling now seems to have imposed itself, both as food for thought and as a vector of communication among the agents, thus enabling us to follow both the evolutions of social demand and technical data by adapting or anticipating them. On all these points, this book is remarkably up to date. The growth problems most industrial countries began experiencing in the early 1970s were aggravated by the oil shocks of the same decade; these difficulties seemed to entail all over the world the development of a

xvii

wide range of new models, called decision-aid models, which used various approaches to deal with the problems posed by the crisis into which the world was settling. Of course, in this new generation of models, the objective was often energy forecasting or prospective, to the request of governments and international institutions, large firms or research bodies entitled to act or give advice in this field. Beyond energy problems, in developed countries, the public was increasingly sensitive to environmental issues: great global dangers, in particular those linked to the greenhouse effect, acid rains and the depletion of the ozone layer, were then forecast by some scientists, before long, if present trends were maintained. Powerfully transmitted by the media, fears relating to these issues soon translated into demands for technical and economic evaluation by politicians, so as to better assess these risks and attempt to act as soon as possible—and as far as possible—in order to come up to people’s expectations. Initially developed for energy planning in industrial countries, many models concerning energy were thus more or less deeply altered so as to integrate at best the parameters liable to account for the relations between energy productions or consumptions and the emission of pollutants in development processes. On these bases, other specialized models were created too, particularly in order to refine the evaluation of demands, to reconsider the principles of criteria of supply optimization, or to integrate supplies and demands into more global representations of economies. Great attention was always devoted, in these papers, to CO2 emissions, the greenhouse effect problem and connected issues, especially the project of creation of a ‘carbon tax’, which sometimes seemed to hide many other environmental questions posed by ecologists from the main governmental or international authorities. At the same time as this energy-environment issue emerged, new events obliged model builders to reconsider the framework or the sources of their previous analyses, and sometimes endangered the reliability of some approaches that had been considered as reliable until then: 1 The emergence of new technologies and activities in production processes requires much time and effort to be detected with traditional statistics (especially owing to the difficulty in modifying the nomenclature these statistics are based on), often obliges analysts to resort to different information sources, which may not be as reliable as the previous ones, and to fill information gaps with estimates that may not be easily justifiable. Thus the decision-making problem over an ‘uncertain future’ appears as well, through facts, as a decision-making problem at a ‘present ill-defined time’. 2 The emergence of new actor strategies, in the context of general crisis and opportunities implying bitter competition at all levels, deeply distorts the so-called behavioural equations introduced in macroeconomic models on the basis of time-series data. This phenomenon probably largely explains the failures most large forecasting models regularly experienced for more than ten years, which lead many national decision-makers to consider prudently the recommendations resulting from their use. 3 Lastly, the new distribution of many skills among the institutions coordinating international economic decisions, governments and regions1 has markedly changed the power, hence decision, centres, to which these models were to be linked. If national models remain necessary, others become so at the regional or international level, in order to draw conclusions from these facts and give to each level of power the decision-making tools adapted to its real means (especially in terms of information and direct action). To the many ‘top-down’ approaches, which implicitly assume that we believe the strategies of the elementary actors in the economy are essentially determined by decisions at the top, many so-called ‘bottomup’ approaches are thus beginning to be added; without totally rejecting the former, the latter assume that

xviii

the strategies concerned are also largely determined by local or specific constraints to which the actors are confronted, with objectives possibly different from those of the whole within which they move. To the already rich corresponding debates, ‘transversal’ or sectoral approaches have added themselves, which convey the concerns of direct actors and observers, interested in what is taking place at the bottom as well as to what is being decided at the top: producers, processing agents and distributors, as well as research bodies, which a priori are neutral. Of course, the point is not making a decision among these forms of philosophies which, anyway, will sooner or later have to combine. It would be as grave to deny the interest of macro-econometric approaches, that cannot be ignored when one wants to assess beforehand the consequence of the implementation of the main instruments of economic policies, as to deny the interest of the other, more technical and/or more ‘micro’ approaches, equally inescapable in order to integrate new elements of reality which, in the long run, will certainly play an essential role. For the time being, what matters is merely to admit the coexistence of these micro-, meso- or macro-economic technical or socio-economic approaches, globally or directly motivated, by trying to manage it in the face of concrete problems, and being fully aware that none will come out unharmed or triumphant of this unavoidable confrontation of the interests of each type of actor. The main objective of this book, in which a dozen contributions presented in a recent conference2 are collected together, is to give an image as present and representative as possible of the ‘richness’ of the analyses and methods which are now being developed in the field of energy economics. 1 As regards time, the papers presented here cover the range of concerns of the main energy actors, from short- or medium-run forecasting to long-run analyses or projections. 2 As regards space, in the same way, they cover the main levels at which the same actors must play: i.e., of course, at the national and international levels, where most of the present general problems arise, but also at the sectoral level, in which equally determining interests are involved. 3 As regards the contents, they fairly well reflect the variety of technical, economic or ‘mixed’ approaches they impose: a) the models concerning a particular form of energy (as a resource, demand vector or object), b) the models representing the whole energy sector, so as to take into account the constraints of its internal coherence and be able to study the appropriateness of some substitutions, c) lastly, the models or analyses whose application focuses on environmental problems (even if they were originally developed with other aims in view). 4 Lastly, as for the approach to energy problems, the reader will later on find papers following three much different procedures: technico-economic or econometric modelling, and ‘non formalizing’ methods, i.e. methods not developed up to model-building, bearing in mind that they remain necessary and sometimes sufficient to pose upstream modelling problems. The richness of the whole can be appreciated quickly through a simple graphic representation, if it is associated to each classifying criterion (time, space, application) preceding any space dimension: then, in the selected general ‘energy-environment’ approach, the respective ‘locations’ of each model or analysis presented later on appear clearly, together with the complementarities and possible overlapping of objectives among these chapters. The following figures (I.1–I.11) numbered 1 to 11 carry out this process, so as to precisely figure out the content of each research:

xix

Figure I.1 POLES (Criqui)

Figure I.2 ERASME (Deimezis)

● abscissae show the intended time horizon (short, medium or long run); ● ordinates show the analytical framework (national, international, or sectoral); ● the heights show the scope or the orientation of the chapters, according as they concern only one or several forms of energy, or the environmental issues; ● lastly, the ‘colour’ of the volumes associated to each study conveys the nature of the chosen approaches: describe technically intensive models, the essentially econometric models, and the nonformalizing approaches.

The presentation in the same space of various figures is useful (Figure I.12). The scope of this test is of course limited: on the one hand, owing to the small number of chapters to which it may be applied, and, on the other hand, because the classification of some studies in the selected typology (e.g. time, space, objective) was not always easy. Thus we do not mean to draw hasty conclusions here. The process would undoubtedly deserve to be both enlarged and refined. However, if we assume that the field of analysis is the

xx

Figure I.3 MIDAS (Capros, Karadeloglou, Mantzos, Mentzas)

Figure I.4 Biproportional methods (de Mesnard)

energy-environment issue approached here, from the graph we obtain, we may argue that the chapters mentioned seem to cover the issue in question: 1 These chapters cover almost all the issue; they study it without much overlapping among them. Thus, redundancies seem limited, and there seem to be important functional complementarities among the papers concerned. This observation testifies to our previously mentioned initial concern, i.e. presenting here a range of works as representative as possible of the diversity of the ongoing research in the intended field. However, it shows as well that the economists’ community seems now to apprehend properly all the questions posed by energy. 2 Although no concentration of cubes appears here, concerning the figure, we shall stress, too, that the time horizon most often selected by the models is the medium term; that their most frequent field of analysis is the national level, for a given set of resources or energy demands; and, lastly, that the econometric approaches are by far the most numerous. In view of the importance of the forecasting and planning functions traditionally imputed to energy economics, these facts are by no means surprising.

xxi

Figure I.5 Gaz contracts model (Haurie, Smeers, Zaccour)

Figure I.6 SOSIE/gas market (Percebois, Valette)

Such a variety of approaches, all of them being necessary, leads to an equally large diversity of resulting recommendations, with which decision-makers themselves must deal with less simple criteria than in the future. If resorting to cost-benefit criteria remains the usual, logical, principle of all decisions made at all levels, it is no longer acceptable to make do with strictly monetary evaluations for each of the two terms thus connected: costs, as well as benefits, must in fact be now corrected so as to take into account new concerns, particularly about the consequences of environmental decisions, but also, as a result of the crisis, about employment. Thus, over less than twenty years, we have shifted from a quasi ‘monocriterion’ decision logic, where financial evaluations could be considered as sufficient, which integrated most of the elements perceived as important, to a more complex ‘multi-criteria’ logic, where the most profitable projects as regards strictly financial data are never certain to outmatch other projects implying external effects perceived as positive— even if these effects will appear in the long run, and if the financial impacts are still ill-estimated, with the blurred aspect typical of forecasts made in a crisis period. Thus the ‘impact on environment’ and ‘impact on employment’ headings are now integrated into most of the forms of project tendered to the authorities entitled to distribute research or development funds (the EC, ministries, local authorities), but the role they play at the decision-making level, as regards the ‘financial impact’ heading, remains uneasily delimited.

xxii

Figure I.7 Tax harmonization in the EC (Lesourd, Meulders)

Figure I.8 HERMES (Donni, Zagame, Detemmerman)

These implicit corrections of the selection methods aiming at integrating externalities are not easy, particularly because the latter are very diverse, and non independent. Beyond their inventory through impact analyses, which is not simple at all, and can but be limited to qualitative data at several levels, the most delicate question posed is that of assigning a value to these factors: if some of them can be appraised with some accuracy, in the space and over the time period where and when the decision has its most direct effects (that is ‘here and now’), others correspond to different impacts (i.e. ‘later on’, if, for instance, we want to integrate health problems or issues concerning the fate of future generations), or exported (i.e. ‘elsewhere’, if we want, for instance, to integrate employment or atmospheric quality issues). The so-called contingent evaluation techniques, developed in order to help decision-makers make better decisions, by trying to enhance the value of some of these factors in terms of willingness to pay or receive, obviously play this role only partially and in an unsatisfactory way. They do permit to put figures forward, as they are asked to—but there is some risk to see these figures over-exploited later on, outside the narrow framework of assumptions in which they are meaningful.3 This problem of ‘internalization of externalities’, whose name often seems funny to those who see it for the first time, is nonetheless important. In fact, this problem will obviously be solved only when public policies and commodity markets likely to meet the above-mentioned concerns have become realities, giving

xxiii

Figure I.9 Energy conservation in Japan (Ban)

Figure I.10 Energy shocks (Andersen, Bernard)

real price figures, at the present time, to all the factors that the complex arbitration of votes and consumption behaviours will have decided to keep on taking into account. Of course, this will not be done overnight, and other problems will undoubtedly have to be solved before we can get out of the current crisis. The existence of the new markets and policies previously mentioned will probably be an essential condition to this solution to the crisis, especially in order to create the expected jobs on the new themes our societies are interested in (e.g. environment, culture, health), while enabling, through new activities, the necessary redistributions of wealth among the still productive systems and the actors excluded by the crisis. Meanwhile, the elements required by these analyses must still be collected, and the methods to process them must also at the same time be perfected theoretically and practically. There is indeed much to say about environmental aspects of energy use, whether at industry level, or at end-consumers’ level; this is a concern which has become of paramount importance over the last ten years, that is, since about 1985. Whereas, during the 1970s or earlier, research on modelling energy demand, energy supply and, more generally, energy markets was triggered by the situation of comparatively high energy prices resulting from the ‘oil shocks’ of 1973–4 and 1979–80, since 1985, and since the ‘reverse oil shock’ of 1986, rather low energy prices prevailed, and, in many countries, economic growth has tended to

xxiv

Figure I.11 Electricity future markets (Hope, Rud, Singh)

Figure I.12 Modelling overall

remain quite intensive in energy and environmental goods, thus creating global concerns with long-term negative externalities. These externalities consist, more precisely, in atmospheric pollution by carbon dioxide and other gases thought to create long-term ‘greenhouse effects’, and by sulphur dioxide thought to result in ‘acid rains’ that is apt to damage some natural resources such as forests, while chlorofluorocarbides that may endanger ozone layers in the upper atmosphere are used in some appliances such as refrigerators. Except for chlorofluorocarbides, these pollutions may be described as productions with negative utility that are directly linked (most probably, through linear functions and dependences) to energy use in industry (and,

xxv

in particular, in oil and coal-fired power plants), and to fossil energy consumption by households. This means that emissions of these pollutants may be easily calculated in suitable models that lend themselves to long-term forecasting. This is what some of the models presented here, such as the POLES model developed by P.Criqui, and the econometric model developed by Detterman, Donni and Zagamé, tend to achieve. While the economic effect of these pollutions is, as noted previously, related to external effects that are quite difficult to assess, it is indeed possible to ‘internalize’ those external effects, that is, to induce economic agents to take them into account in their economic evaluations. It is well known that one of the ways to achieve such an internalization is to submit emissions of pollutants to so-called ‘Pigovian’ taxes (thus named after the British economist A.C.Pigou, who introduced this concept in 1920). One of the contributions presented here (Detemmerman, Donni and Zagamé) thus arms at assessing the economic effects of a so-called ‘carbon tax’, which is indeed a ‘Pigovian’ tax, on carbon dioxide emissions; but energy taxes need to be harmonized at an international level, especially within the EC (Lesourd and Meulders). In this volume, other concerns are still present, such as short-term concerns in the contribution by Capros, Karadeloglou, Mantzos and Mentzas, and the contribution by Labys, Murcia and Terraza, and medium-term concerns in the contribution by Andersen and Bernard: short-term instability of the international crude oil market, and in particular price instability, since the first ‘oil shock’, has indeed become one of the features of this market. Market structure and price formation, which are indeed at the heart of economic knowledge, are also present, especially in the contribution by Haurie and Zaccour, and in the contribution by Percebois and Valette. In the present context, reducing taxation on labour and increasing energy taxes in compensation, seems to be a profitable approach. Thus we may obtain a ‘double dividend’: as energy consumption decreases, the rate of pollution may be reduced. At the same time, the substitution of labour for energy and capital may reduce the unemployment rate. In order to assess the effects of such a policy, it is necessary to use models; not only global models but disaggregated models too. The consequences of a rise in energy taxes as a way of reducing CO2 emissions and of favouring labour-energy substitutions are not the same for each sector. It is the reason why some models presented here, such as the models of V.Detemmerman, E.Donni and P.Zagamé are useful for implementing new energy policies in relation to new fiscal policies. I.II DESIRABLE DEVELOPMENTS Each of these models tries to meet a particular concern, and in this way these models are complementary, but much remains to be done to integrate them into each other. Particular attention must be devoted to the micro- and macroeconomic relationships, and to taking into account uncertainty and irreversibilities in a field where scientific certainties are not generally admitted (e.g. the greenhouse effect). New databases will be necessary; progress in computing techniques is desirable, but, above all, model builders will have to be increasingly aware that they must choose a pluridisciplinary approach. In the field of energy and the environment, there are complex interactions among physics, technology, economics and social aspects, and approaching each phenomenon separately would lead to a dead end. NOTES 1 Over this period, most of the countries lost most of their technical, economic and social abilities, from a legal point of view as well as concerning their power and know-how. In fact these abilities were redistributed: some of

xxvi

them were attributed to organizations larger than nations (such as the EC) to set the new rules of the more open game resulting from recent events, the others were assigned to smaller entities, such as regions or towns, in order to partly define the policies that might fit into these regulations. 2 ‘International Energy Market Modelling’ Conference organized under the aegis of the Applied Econometrics Association in Montpellier, France, October 1991. 3 These figures are estimated through surveys, which are always questionable, in their questionnaires as well as in their implementation.

Part I New models and new modelling methodologies

Chapter 1 Methodological advances in energy modelling: 1970–90 James M.Griffin

1.I INTRODUCTION The 1970s were unique in that a micro-economic issue, energy economics, suddenly captivated the policy agenda. Not since the 1930s and the Great Depression, had the emphasis of policy makers fallen so squarely on economics. Generous energy research funding fuelled a virtual explosion of energy-related research, attracting a widely diverse group of researchers. Besides varying greatly in equality, the resulting literature varied greatly in breadth, reflecting the different perceptions of what the major policy questions were. Major topic areas included theoretical models of exhaustible resources, cartel theory applications to OPEC, environmental economics, as well as an assortment of energy supply/demand models designed to address some subset of a continually changing menu of policy questions. Just as oil prices reached their peak in the early 1980s and declined markedly for the remainder of the decade, so too did the energy policy agenda and research funds. Whether measured by the frequency of articles published on energy related issues, NSF funding for energy research, or the energy economics staff at Exxon or British Petroleum, evidence of a sharply reduced research effort is uniform. Perhaps it is time to look back over the last 20 years and ask, ‘What have we learned?’ The purpose of this chapter is to recount three major methodological advances which have profoundly affected energy modelling and have generated positive externalities for applied research in general. No attempt will be made to survey this voluminous literature or to summarize empirical findings. We begin by summarizing briefly in Section 1.II the status of energy modelling circa 1970. Section 1.III recounts three major methodological advances of the 1970s and 1980s. Section 1.IV recapitulates the major advances and offers my suggestions for a work agenda. 1.II THE STATE OF ENERGY MODELLING CIRCA 1970 1.II.1 Energy demand Energy demand research at Resources for the Future (e.g. Schurr et al. 1960) typified energy modelling. An implicit assumption underlying energy demand modelling was that energy was tied to output through a Leontief technology as follows:

METHODOLOGICAL ADVANCES IN ENERGY MODELLING: 1970–90

3

(1) According to equation (1), output (Q) was determined by a Leontief relationship depending on value added (VA(K,L)),1 energy (E), and materials (M) inputs. The technical coefficients γ1, γ2 and γ3 measure the Leontief recipe for the dollar value of any given input necessary to produce a dollars worth of output. Taking GNP as a proxy for output, energy demand was tied to GNP through a technically determined coefficient (γ2) as follows: (2) Thus energy demand forecasts depended entirely on GNP growth. Refinements of this methodology attempted to explain changes in γ2, linking it to changes in the technical efficiencies governing autos, railways, refrigerators and so forth. 1.II.2 Supply of individual fuels Having determined aggregate energy consumption in this manner, researchers then proceeded to allocate BTUs among the four primary fuels, hydroelectric, coal, natural gas and oil based on the supply of these fuels. Typically, hydroelectric and coal consumption were set equal to projected hydro and coal production. For example, coal production was viewed as constrained by existing mining capacity and the technical feasibility of adding capacity over a given planning horizon. Natural gas consumption was usually treated as constrained by pipeline capacity constraints and/or reserve levels, leaving oil to serve as the residual. By allowing oil to serve as the residual, all fuels were implicitly treated as perfect substitutes, despite the fact that the BTU equivalent prices of the four fuels were seldom equal. 1.II.3 Price determination In this system, prices were largely irrelevant both in the determination of aggregate energy demand or the supply/demand functions for individual fuels. Lacking a demand specification for individual fuels, price determination was handled in an ad hoc manner. The long-run supply schedule for coal was assumed perfectly elastic, given wages, so that coal prices could be determined as a markup over mining costs. Natural gas prices and crude oil prices in the US were administratively determined, obviating any need to model these prices. Because of the perceived unimportance of prices, the analysis tended to take on an engineering orientation with a focus on technical factors affecting both supply and demand. 1.II.4 Important caveats The above description of energy modelling circa 1970 is a stereotypical view based on modelling work at Resources for the Future, the US Bureau of Mines, and the leading oil companies. The problem with stereotypes is that they can be grossly unfair to certain individuals and organizations. Important academic work lead many industry practitioners to begin questioning the prevailing methodology. In a provocative paper in the Journal of Industrial Economics in 1967, F.G.Adams and his student, Peter Miovic, collected energy consumption data for 11 European (including six EEC) countries and showed that the energy/GNP ratio

4

NEW MODELS AND NEW MODELLING METHODS

varied significantly over time because of the differing thermal efficiencies of fuels. The transition in Europe from coal with very low thermal efficiencies to petroleum with high thermal efficiencies caused the energy/ GNP ratio to fall markedly. Adams and Miovic argued for adjusting fuels for their differing thermal efficiencies. In another development, looking at the discovery of oil reserves, Franklin Fisher (1964) found that oil prices impacted drilling rates, discovery rates and average discovery size. Fisher’s model offered a clearcut econometric alternative to the prevailing trend line analysis. On the question of interfuel substitution, the role of relative fuel prices as a determinant of demand was illustrated nicely in a famous paper by Balestra and Nerlove (1966). The Balestra/Nerlove paper was important not only for its findings of interfuel substitution between electricity and natural gas but also for its pioneering use of a panel data set. Besides these key academic papers, linear programming techniques made major inroads in refinery scheduling and expansion planning in the 1960s. Even though these models were non-econometric and engineering based, they provided a complete description of the production surface. Moreover, these optimization models were driven by prices—either cost minimization or profit maximization. Refinery optimization models played an important role in optimizing crude choice, in the selection of the optimal product mix, and in refinery expansion planning. 1.III THREE MAJOR METHODOLOGICAL ADVANCES: 1970–90 The Arab Oil Embargo of 1973 triggered a set of events that would see oil prices rising 10 fold by the end of the decade. The energy crisis of the 1970s became perceived as a complex problem involving geopolitical, economic and environmental aspects. To cushion the effects of price increases, an elaborate system of price controls and regulations were introduced in the US. Consequently, the menu of policy questions expanded exponentially. It became clear, that the existing set of energy models were not designed to answer many of these questions, providing a huge stimulus to funded research in the area of energy modelling. The decade of the 1970s witnessed major methodological advances in the estimation of more general production technologies and econometric techniques to utilize pooled data sets. Then in the 1980s, McFadden’s development of discrete choice models to appliance stock choices opened up yet another major avenue for applied research. 1.III.1 The translog revolution The initial appeal of a richer production technology

Even though the translog function (Christensen, Jorgenson and Lau 1973) is only one of a set of flexible functional forms, the fact that it can be thought of as a generalization of the Cobb-Douglas function made it the preferred choice of most researchers. The basic appeal of the translog and its sister functional forms is that they relax the range of substitution possibilities in a production technology. For example, instead of requiring a constant elasticity or unitary elasticity of substitution among all inputs, it became possible to estimate KLEM models in which some inputs could be substitutes and others could be complements. Q=f(KLEM) (3) Econometric applications centred on estimating the translog unit cost function dual to the linearly homogeneous production function: (4)

METHODOLOGICAL ADVANCES IN ENERGY MODELLING: 1970–90

5

where βij=βji. Because of the multicollinearity problems with direct estimation of (4), researchers utilized Shephard’s lemma2 to derive a set of cost share (Si) equations: (5) Note that because of restrictions on the βij (e.g. βij=βji and Σj βij=0) simultaneous estimation of the n–1 cost shares with the parametric constraints greatly increased the efficiency of the resulting estimates. Depending on the magnitude of the estimated βij coefficients and the cost shares, the Allen elasticities of substitution (σij) could take on any value. (6) Additionally, the long run cost function need not obey constant returns to scale. Rather by appending δo In Q +½δi(In Q)2+Σγj In Q In Pj to equation (4) and δj In Q to equation (5) a non-homogeneous, non-homothetic cost function follows. Tests for scale economies then depend on the elasticity of cost with respect to output (εCQ ) as follows: (7) Yet another advance was that the traditional time trend used to reflect a constant rate of Hicks’ neutral technical change could be generalized by introducing Øot+½Ø1(t)2+ΣΨjt In Pj into equation (4), which in turn resulted in the addition of Ψj t terms into the share equations. Consequently, technical change ( ) became measured as follows: (8) This specification allowed technical change to be both factor saving or using (if ) and proceed at differential rates over time (if ). Because of its generality, the translog cost function immediately raised a number of intriguing questions. Are energy and capital complements?3 Is technical change energy using? Answers to these questions had potentially important macroeconomic implications for explaining rates of capital formation and the productivity decline of the 1970s. Yet another advance in this literature arose in the nested subfunction approach to modelling interfuel substitution (Halvorsen 1977). The energy aggregate (E) in equation (3) can in turn be viewed as a function of four individual fuels, . Corresponding to the energy aggregate subfunction is a translog unit cost subfunction for energy (PH) dependent on the prices of the individual fuels (PFi) as follows: (9) Again by estimating the fuel share equations, a translog unit cost subfunction for energy can be estimated. By combining the estimates of the aggregate translog function with the energy subfunction, estimates of the own and cross price elasticities for any given fuel can be obtained. Subsequent refinements and reexaminations

Despite the numerous advantages of the translog function enumerated above, practitioners attempting to use translog models for forecasting and simulation analysis noted two serious flaws. First, the concavity conditions that assure factor demand functions are negatively sloped may be satisfied at the sample mean, but over other price regions yield perverse results. Concavity is satisfied globally only when the translog collapses to the special case of the Cobb-Douglas function. The price range over which the function obeys concavity depends critically on the values of the estimated parameters, so that for one set of parameter estimates, concavity is not a serious problem and for others it is. Without actually estimating the function

6

NEW MODELS AND NEW MODELLING METHODS

and performing a likely set of forecast simulations, one cannot determine a priori whether it is a problem. Diewert and Wales (1987) have developed a constrained estimation procedure to deal with the concavity problem. Second, the translog model implies static expectations and instantaneous adjustment to long-run equilibrium. In effect, the firm is assumed to instantaneously move from one long-run cost minimizing equilibrium to another as relative factor prices change. In energy applications, where long lags are known to exist, this characteristic is particularly troublesome. Attempts to remedy this difficulty have proceeded along three basic lines. In the first, researchers simply treated capital as a fixed input and proceeded to estimate a variable cost function instead of a total cost function. Using time series data to estimate the variable cost function, researchers could legitimately claim to have estimated a short-run relationship.4 The second vein of research is to develop a dynamic translog specification. In an attempt to provide a unified theoretical framework capable of describing the short-run equilibrium and the adjustment path to the long-run equilibrium. Berndt, Fuss and Waverman (1980) developed a model in which capital was costly to adjust. Besides yielding implausibly rapid adjustment to long-run equilibrium their model relied upon static expectations. Subsequently, Pindyck and Rotemberg (1983) incorporated rational expectations in a variable cost function with a capital stock equation that adjusts optimally over time. Because of difficulties in solving stochastic optimal control problems, the dynamic adjustment path can only be simulated for deterministic price scenarios, limiting its use as a prediction device. Yet a third response has been to abandon the translog in favour of other specifications more amenable to dynamic specifications. For example, Walfridson (1987) develops a dynamic model featuring a partial adjustment model coupled with a generalized Leontief function. Subsequently, Hogan (1987) adopts a similar partial adjustment model but utilizes the ‘symmetric generalized McFadden cost function’5 because of concavity problems. Even though the dynamics follow from an ad hoc specification, Hogan provides persuasive empirical support for its use. 1.III.2 Panel data methodology While conceptually the translog offered a major advance, the early empirical applications of it with time series data often lead to implausible estimates. For example, in Hudson and Jorgenson’s (1974) US energy model estimated from annual aggregate US time series data, the implied price elasticity for energy was perverse in the manufacturing sector. Furthermore, evidence for interfuel substitution was very weak, as the own and cross price elasticities were close to zero. Also, using aggregate time series data, Berndt and Wood (1975) interpreted their rinding of energy-capital complementarity as a long-run result rather than as a shortrun anomaly due to the fixity of the energy efficiency of the capital stock in the short run. In addition to translog studies, time series studies of petrol and electricity demand also tended to find rather price inelastic demand responses.6 In contrast, researchers utilizing cross section and panel data sets tended to report much larger price elasticities, reflecting the fact that interregional price differences were both quite substantial and persistent so that the between region demand variation tended to reflect long-run responses.7 Particularly for energy demand, the adjustment to the long run may take ten or more years because of the long-lived nature of the capital stock to which energy consumption is tied; consequently, one would expect intercountry crosssection data to elicit long-run responses while time series data would reflect incomplete adjustment. To illustrate, in Baltagi and Griffin’s (1983) petrol demand study, using time series data for the period 1960 to 1978 and a static specification, they found negative, statistically significant price elasticities for only 11 of

METHODOLOGICAL ADVANCES IN ENERGY MODELLING: 1970–90

7

the 18 OECD countries. Furthermore, the estimates ranged from −0.04 to −0.79. Then applying the same static specification to the same data but in a pooled context, the price elasticities were −0.32, −0.89, and −0. 96 for the within, OLS, and between models. Depending on the relative weighting of the between versus within variation, the elasticity estimates varied significantly. In principle, the reconciliation of the between and the within variation estimates can be achieved by the introduction of a dynamic specification so that the within variation defines not only the short-run contemporaneous effect of price, but the effects of lagged prices as well. Surprisingly, this study showed that a dynamic specification only raised the long-run price elasticity in the within variation model from −0.32 to −0.52 as contrasted to the between elasticity of −0.96. Consequently, in energy applications where the lagged price effects are quite large but occur over a long adjustment period, the pure cross sectional or between country variation becomes critical in isolating longrun responses. The 1980s witnessed significant increases in both the quality and number of panel data studies. Researchers increasingly perceived that intercountry and interregional price and income variation provided valuable information not usually available from simple aggregate time series data. Moreover, as energy consumption dropped sharply in the mid-1980s, researchers recognized that the larger price elasticities implied by the panel data sets provided better forecasts. Increasingly, empirical work tended to utilize dynamic specifications in the context of error component models. A striking attribute of these studies was the application of a battery of tests and a variety of feasible GLS estimators. Tests for one or two-way error components, poolability and specification error became commonplace. For example, Hausman’s specification error test is designed to test for correlation between the regressors and the country effects. Panel data sets offer the prospect of yielding superior measures of technical change. Baltagi and Griffin (1988) exploit the richness of a panel data set (across firms) and a translog framework to propose a method for estimating a general index of technical change, without the restrictive assumptions associated with a time trend. 1.III.3 McFadden’s application of discrete choice models Since Fisher and Kaysen’s (1962) analysis of appliance stocks and electricity consumption, economists have recognized that the demand for a given fuel depends first on the stock of energy consuming equipment (S), then on the efficiency (e) with which the fuel is utilized, and finally on the utilization rate (u) of the capital equipment: (10) Typically, for any given capital stock, the efficiency of fuel conversion is fixed by technical considerations, so that in the short run, the relevant decision variable affecting fuel consumption is the utilization of the capital stock. Interfuel substitution effects occurs only in the long run with the introduction of new capital. For example, the choice of electric, natural gas, or heating oil for home heating is dependent on the choice of capital equipment. Despite this intuitively appealing framework, most energy demand models made little use of this approach, largely because of inadequacies in capital stock data. Using a random utility maximization model, McFadden and Dubin (1984) developed a rigorous theoretical model amenable to empirical testing. Since the theory applied to the household, McFadden turned to household survey data for the period 1977–9, which contained detailed information on the economic and demographic characteristics of the households as well as their stock of various energy using appliances. Because of geographical location differences

8

NEW MODELS AND NEW MODELLING METHODS

individuals faced different marginal electricity price schedules and differing weather conditions. With these detailed micro data, McFadden and Dubin applied the multinomial logit model to the simultaneous discrete choice of appliance type for home heating and the amount of fuel consumed given that choice. With its massive micro data requirements, the multinomial logit model is likely to be of limited application to applied energy modellers at present. It nevertheless, represents a major conceptual breakthrough. Because of its superior theoretical underpinnings, empirical estimates from McFadden and others’ work will provide a useful cross-check against elasticities obtained from conventional models using aggregated data. 1.IV SUMMARY AND THOUGHTS ABOUT THE FUTURE Both the theory and practice of energy modelling has made phenomenal advances from a very humble background 20 years ago. Economic theory has provided the framework for empirical models consistent with received micro theory, while panel data sets have vastly enlarged the richness of econometric evidence. The results of these advances are that we now have much sharper estimates of the relevant elasticities and a greater understanding of the intersection between the energy sector and the macroeconomy. Energy economics has by no means been the sole beneficiary of these advances. Equally impressive has been the impact these advances have had in other areas of economics. Today, flexible functional forms are commonplace in a wide spectrum of applied studies. Today, much of the most exciting work in labour economics involves panel data and discrete choice models. Where will energy modelling go in the next 20 years? I cannot hope to answer the question, but I can offer a generalization and outline three areas that I would hope receive attention. While the advances of the last two decades are truly impressive, difficult conceptual and empirical problems remain. First, irrespective of the functional form, tough problems remain as to how best to model dynamics. Implicit in this question is issues of how best to model expectations and how to reflect the effect of different vintages of capital, given the limitations of existing capital measures. Second, we must devise better methods of using engineering information together with conventional data. Increasingly, energy modellers will be asked to respond to hypothetical questions covering longer time periods involving technologies known, but not yet adopted. Third, our understanding of cartels remains in its infancy. Notedly, absent from the list of methodological advances was any reference to the modelling of oil price determination. While game theory holds the promise of unlocking the mystery, to date it has not. If game theory is to retain its current prominence, it must surely affect what we know about cartels. Hopefully, new game theoretic insights into the durability and pricing policies of OPEC will be forthcoming. NOTES 1 The value added aggregate could in turn involve a Cobb-Douglas or CES relationship among capital (K) and labour (L). 2 Shepherd’s lemma requires that firms minimize costs and treat input prices as given. 3 See Berndt and Wood (1975) and Griffin and Gregory (1976). 4 A nice characteristic of the variable cost function is that the long-run equilibrium responses can be inferred by equating the shadow price of capital to the rental cost of capital. But knowing the short-run elasticities and the long-run elasticities does not describe the adjustment path to the long-run equilibrium. 5 See Diewert and Wales (1987).

METHODOLOGICAL ADVANCES IN ENERGY MODELLING: 1970–90

9

6 For surveys, see Bohl (1981). 7 For examples, see Griffin (1979) and Halvorson (1977).

REFERENCES Adams, F.G. and Peter Miovic, ‘On Relative Fuel Efficiency and the Output Elasticity of Energy Consumption in Western Europe’, Journal of Industrial Economics, November 1968, pp. 41–56. Balestra, P. and Marc Nerlove, ‘Pooling Cross-Section and Time Series Data in the Estimation of a Dynamic Model: The Demand for Natural Gas’, Econometrica, 34, 1966, pp. 585–612. Baltagi, B.H. and J.M.Griffin, ‘Gasoline Demand in the OECD’, European Economic Review, July 1983, pp. 117–37. Baltagi, B.H. and J.M.Griffin, ‘A General Index of Technical Change’, Journal of Political Economy, February 1988, pp. 20–41. Berndt, Ernst and David Wood, ‘Technology, Prices and the Derived Demand for Energy’, Review of Economics and Statistics, August 1975, pp. 259–268. Berndt, Ernst, Melvyn Fuss and Leonard Waverman, ‘Dynamic Adjustment Models of Industrial Energy Demand: Empirical Analysis for U.S. Manufacturing, 1947–74’, EPRI Research Report EA-1613 (Palo Alto, Electric Power Research Institute: 1980). Bohi, Douglas R., Analyzing Demand Behavior, (Baltimore, Johns Hopkins Press: 1981). Christensen, Laurits, Dale Jorgenson and Lawrence Lau, ‘Transcendental Logarithmic Production Frontiers’, Review of Economics and Statistics, February 1973, pp. 28–45. Diewert, W.E. and T.J.Wales, ‘Flexible Functional Forms and Global Curvature Conditions’, Econometrica, January 1987, pp. 43–68. Fisher, Franklin and Carl Kaysen, A Study in Econometrics: The Demand for Electricity in the United States, (Amsterdam, North-Holland: 1962). Fisher, Franklin, Supply and Costs in the U.S. Petroleum Industry, (Washington, Resources for the Future, 1964). Griffin, James M., Energy Conservation in the OECD: 1980 to 2000, (Cambridge, MA, Ballinger: 1979). Griffin, James M. and Paul Gregory, ‘An Intercountry Translog Model of Energy Substitution Responses’, American Economic Review, December 1976, pp. 845–57. Halvorson, Robert, ‘Energy Substitution in U.S. Manufacturing’, Review of Economics and Statistics, November 1977, pp. 381–8. Hausman, J.A., ‘Specification Tests in Econometrics’, Econometrica, 1978, pp. 1251–72. Hogan, William W., ‘Patterns of Energy Use Revisited’, mimeo, J.F.Kennedy School of Government, Harvard University, 1987. Hudson, Edward and Dale Jorgenson, ‘U.S. Energy Policy and Economic Growth, 1975–2000’, Bell Journal, Autumn 1974, pp. 461–514. McFadden, Daniel and Jeffrey Dubin, ‘Econometric Analysis of Residential Electric Appliance Holdings and Consumption’, Econometrica, March 1984. Pindyck, Robert and Julio Rotemberg, ‘Dynamic Factor Demands and the Effects of Energy Price Shocks’, American Economic Review, December 1983, pp. 1066–1079. Schurr, Sam, Bruce Netschert, Vera Eliasberg, Joseph Lerner and Hans Landsberg, Energy in the American Economy, 1850–1975, (Baltimore, Johns Hopkins Press: 1960). Walfridson, B., ‘Dynamic Models of Factor Substitution: An Application to Swedish Industry’, (Gothenburg Sweden, Gothenburg School of Economics, 1987).

Chapter 2 International markets and energy prices: the POLES model Patrick Criqui

Abstract The POLES model is a long-term (2010–2030) world energy simulation model, developed in connection with studies on Global Climate Change. Its goals are to construct scenarios for carbon dioxide emissions, to help in analysing the cost-effectiveness of emission control policies and, finally, to identify the impact of such policies on the international energy markets. It is a recursive simulation model, based on a hierarchical system structure linked at three levels: the international energy markets, national energy balances and energy production and consumption sectors. In the medium-long term (2010), the simulation of international energy prices is based, on one hand, on a target-capacity utilization rate model and, on the other, on the concept of the oil price being the leading price—with certain constraints— determining the price of other fuels. In the very long term (2010–2030), it is assumed that fossil fuel markets will become more independent, with the price of oil, coal and gas being determined by regionalized supply curves. This chapter presents the set of models dealing with international markets in the medium-long term. INTRODUCTION The purpose of the POLES1 model is to develop long-term world energy scenarios (2010–2030) and quantify corresponding CO2 emissions. It is designed to: ● identify uncertainties and operating margins in the long-term energy markets by integrating predetermined factors, the interaction of economic variables as well as variables related to sectoral policies; ● lay the foundations of a world-wide cost-effectiveness analysis by identifying the main emission sources in each region and the potential impact of different types of economic or technological measures; and ● analyse the effects of possible energy-environment policies on international prices and markets by taking into account inter-regional markets and simulating the markets for the main fossil fuels: coal, oil and natural gas. The structure of the model and its operating logic are based on two key concepts established by H.A.Simon: the concept of the hierarchical system to describe the structure of a complex artificial system2 and the concept of a procedural rather than substantive rationality to describe the behaviour of the economic players in a risky and unstable situation.3 It is thus a recursive simulation model, in which energy prices at time (t),

NEW MODELS AND NEW MODELLING METHODS

11

along with the exogenous variables (population, growth, etc.), determine supply and demand at time t+1, including adjustments for time lags or delays. The model operates on three interlinked levels: ● international energy markets for the three major fossil fuels, ● national energy systems, as represented in the accounting framework of the energy balance, ● technical subsystems of energy production, conversion and consumption within each national subsystem. The choice of a recursive simulation approach rather than an intertemporal optimization process, while the aim of the project is to help formulate energy policies for EC countries, also stems from the adoption of a procedural approach. The studies that can be conducted using the model will not be aimed at identifying the optimum policy, but rather at using contrasting scenarios to identify satisfactory policies. As regards the models for international markets, two different logics have been used to simulate price changes in the medium term (2010) and very long term (2030). In the medium term, there are still major uncertainties as to the role of the Gulf region in world supply. The greater the contribution of this region the greater the likelihood of oil being present in markets where it will compete with gas and coal. The relationship between oil, gas and coal prices must therefore be made clear. In the very long term, the central hypothesis is one of an increasing mobilization of Gulf reserves and an accompanying refocussing of oil on the markets of transport fuels and petrochemical feedstocks. This could result in international markets that are more separate and where prices are established according to the specific coal, oil and gas supply conditions. This chapter presents the entire market model for the medium-long term. The first two parts deal with fossil fuel price functions. In the first part, the oil price simulation model is described in relation to commonly encountered approaches, whether they be theoretical or applied. The second part of the chapter presents an analysis of the relationships between energy prices through the concepts of a leading price for oil and pricefloor and -ceiling for coal. The last part presents the module concerning the market shares of the major exporters, which is used to calculate the production of each country. 2.I A SIMPLE MODEL FOR SIMULATING THE OIL MARKET SHOCKS AND COUNTER-SHOCKS In his survey of oil price models, D.Gately4 makes a distinction between theoretical models based on the Hotellinian paradigm and applied models as identified in particular in the report of the Energy Modelling Forum No. 6.5 The price function adopted in the POLES model is clearly linked to the family of applied, or behavioural, methods. 2.I.1 Hotellinian models and applied models The Hotellinian theoretical model stipulates that ‘substantially rational’ behaviour on the part of owners of a non-renewable natural resource must result in an increase in the price of the resource corresponding to the discount rate. However, a simple analysis of the market price curve for oil over one hundred years raises serious doubts as to the relevance of the Hotellinian model. What is revealed is that strong initial fluctuations in oil prices were followed by variations of between 10 and 20 $/barrel in the first half of the century. In the second half of the century, there was a gradual stabilization at less than 10 $/barrel between

12

INTERNATIONAL MARKETS AND ENERGY PRICES: THE POLES MODEL

1950 and 1973, which was of course followed by sharp increases (45 $/barrel in 1980) and the subsequent plunge to the previous price of 20 $/barrel (1990 prices used throughout, unless otherwise indicated). There are two schools of thought concerning the failure of the Hotellinian model to describe and predict real changes in oil prices. M.A. Adelman completely rejects the Hotellinian system, claiming that the basic theory of rising prices is false because its two premises—the first that there is a fixed supply and the second that an oil field can be exploited at any rate until it is depleted—are themselves false. According to Adelman, ‘mineral costs and prices, including oil and gas, are the uncertain fluctuating result of two contrary forces: diminishing return versus increasing knowledge’.6 In his view, given the present state of reserves and prices, the price of oil could stabilize at 5 $/barrel for another few decades. D.Gately insists that it is impossible for the oil producing countries to adopt a scheme based on an optimizing rationality when they are faced with insurmountable uncertainties, particularly with regard to price-elasticity of nonOPEC supply and demand.7 Detailed econometric studies invalidate to all intents and purposes the hypothesis that the OPEC countries would behave like members of a cartel with each member adopting its own dynamic optimization behaviour.8 As for the applied oil price models, they use a system of rules—and thus a system of procedural rationality—to simulate oil price changes. The most commonly adopted of these rules is based on the hypothesis of a target capacity-utilization rate for the OPEC countries: seven of the ten models studied in EMF6 are in fact target capacity utilization (TCU) models. The US-DOE Oil Market Simulation model is undoubtedly the best known of these models.9 It links the relative variation in the price of oil over one year (● OP) to the capacity utilization rate of OPEC for the previous year (CU−1) according to an equation of the following type This type of model, which was developed in the early 1980s, can be used to describe in a satisfactory manner the crises which hit the oil market and the related price-ratchet effect: prices increase considerably when the utilization rate exceeds a threshold of 80 to 90% but decrease only very slowly afterwards. But this model, or at least this particular specification, is invalidated by the changes observed at the time of the counter-shock: the price of oil plummeted at a rate that was hardly consistent with the generic graph of the model (Figure 2.1). Like Hotellinian models, behavioural models do not come out unscathed from the confrontation with reality.10 However, the principle itself of a behavioural model is not completely rejected since other specifications can be used to better account for the reality observed. In fact, M.Rauscher proposes a more complete specification,11 which combines a sharp rise in prices when OPEC production goes beyond a certain level and a sharp decline when it drops below another level. Counter-shocks can then be simulated as well as shocks. 2.I.2 Price function in the POLES model The oil price function for the POLES model is based on the function proposed by M.Rauscher (see Figure 2.2), but two aspects have been improved: ● the objective is not formulated in terms of the desired production level, which compelled M.Rauscher to distinguish several sub-periods, butrather in terms of the desired capacity-utilization rate (CU*);

NEW MODELS AND NEW MODELLING METHODS

13

Figure 2.1 OPEC pricing 1985–8

Figure 2.2 Non-linear price adjustments Source: Rauscher (1989)

● calculation of the effective capacity utilization rate is based only on the Gulf countries (GCU, for Saudi Arabia, Kuwait, UAE, Iraq, Iran, Qatar) which are the real swing-producers, while the other OPEC countries have very little room to manoeuvre in terms of capacity. The estimate is then carried out for the function

14

INTERNATIONAL MARKETS AND ENERGY PRICES: THE POLES MODEL

Figure 2.3 Oil price function in the POLES model Source: POLES-IEPE

Calculations with different values of the desired utilization rate (GCU*) and the parameter n characterizing the function curvature, can then be used to obtain, for example, the following result:

The proposed specification thus corresponds to a desired utilization rate for the Gulf countries of 68%. It provides a good simulation of oil shocks and counter-shocks with strong price resistance when the utilization rate is between 60 and 80%. This model includes four recent important phases in the oil market: 1 2 3 4

adjustment by growth of the amount produced in the Gulf with low prices before 1973; oil crises and stabilization of production from 1973 to 1980; decline in output with high prices from 1981 to 1985; finally, rebound and market turnaround since 1986 (Figure 2.4).

From a theoretical point of view, this function reproduces the ‘response of a cyclical loading system’ and thus brings us clearly back to the concept of the hysteresis model.12 From an operational point of view, and with future studies in mind, it may be noted that different values assigned in an exogenous manner to the curvature parameter n can be used to simulate different hypotheses concerning price rigidity. Such hypotheses could in turn be linked to different patterns of relationship between the main players: with a high level of rigidity (low n coefficient) there is a corresponding price control, alternately by the consumers (low price) or the producers (high price); with less rigidity (high n coefficient) there may be more cooperative relationships in which prices are adjusted in a more regular manner.

NEW MODELS AND NEW MODELLING METHODS

15

Figure 2.4 Hysterisis in oil price dynamics Source: POLES-IEPE

2.II THE PRICE OF OIL AND OTHER FORMS OF ENERGY: A CONSTRAINED LEADING PRICE The main concepts used to characterize the relationships between the international prices of fossil fuels are that of the leading price, which involves indexing the price of coal and gas to that of oil, and that of pricefloor and -ceiling, which could have the opposite effect of disconnecting the price of coal in relation to that of oil. The system of relationships developed for the medium-long term in the POLES model links these concepts by making the price of oil a leading price but with certain constraints, that is within a range determined by a price-floor and -ceiling for coal. 2.II.1 The price of oil as the leading price for natural gas? J.Percebois and J.B.Lesourd have pointed out the existence of a linear relationship between the price of natural gas and the price of oil.13 They are thus able to demonstrate the dominant character of the indexation diagram, while taking into account a floor price for natural gas (the cost of bringing it to the port of shipment on LNGs) and clarifying the conditions for sharing the gas rent between producers and consumers.14 The study of the import price of gas (CIF) on the three main markets of Europe, Japan and the Americas, for the period 1975–90, confirms the overall findings of these analyses. However, an examination of oil price curves reveals considerable distortions in this relationship which result from the specific supply and demand conditions affecting each market (Figure 2.5). Two types of lesson can be learnt, concerning either the changes in relative prices or the individual dynamics of each market.

16

INTERNATIONAL MARKETS AND ENERGY PRICES: THE POLES MODEL

Figure 2.5 Import prices for natural gas and oil (CIF, 1975–90) Source: IEPE, BP data

In these three regions, and for the entire period, the relative price for imported natural gas is generally situated between 1 and ½. However, the price of gas may be greater than that of oil. This was the case in the United States at the end of the 1970s, in Japan between 1986 and 1989, and in Europe in 1986. The price of gas seems structurally higher in Japan: it was almost twice the European price in 1980 and twice the American price in 1990. As for the European price, it was generally lower than the American price in the 1970s but exceeded it in the 1980s. Each market is in fact subject to different changes. If the price of gas is high in Japan it is undoubtedly mainly because of transport costs, but also because of the specific advantages of natural gas—mainly in environmental terms—while the sources of supply remain concentrated despite considerable efforts at diversification in the 1980s. In Europe, the higher relative price in the mid-1980s reflects a sharing of the gas rent which was fairly favourable to the producing countries during this period. This advantage seems to have become less marked at the end of the 1980s, with the drop in oil prices and diversification of potential sources of supply. Finally, the high price of gas in the United States during the second half of the 1970s reflects the virtual-scarcity in the country at that time, while the subsequent collapse of prices was the result of deregulation which created a structural surplus of supply, the ‘gas bubble’. These analyses do not totally invalidate the choice of a linear relationship between the price of oil and that of natural gas, a relationship adopted a priori as a simplifying hypothesis in the POLES model. The analyses show, however, that within each market this relationship is not necessarily stable and that it depends mainly on the specific conditions governing the sharing of the revenue between the producers and consumers concerned.

NEW MODELS AND NEW MODELLING METHODS

17

2.II.2 Floor price and ceiling price of coal Is the price of coal today not related to that of oil? This is certainly the case if we consider that the equivalent oil barrel price of boiler coal was practically the same as that of fuel oil in the early 1970s, while today it is no more than half this price. Furthermore, according to P.N.Giraud, the international market for steam coal, after a period of strong growth, has entered a mature phase characterized mainly by a considerable degree of independence of prices in relation to those of other forms of energy.15 The price of steam coal would then fluctuate between an ‘Australian’ floor price and an ‘American’ ceiling price, representing respectively the price below which the cheap Australian mines could no longer operate at a profit or the price above which the market would be flooded by the output from marginal American mines. Between these limits, the price would change according to the specific fundamental variables of the coal market. The price function of coal used in the POLES model is based on these analyses, but without reintroducing the hypothesis of complete independence of oil prices and coal prices. Examination of the price curve for boiler coal imported into Europe (CIF) in relation to the price of oil reveals a distinct linear relationship, at least for the period 1977–90 (Figure 2.6). This is not a strictly proportional relationship, however, since the relative price fluctuates between ½ when the price of oil is low 1/3 and when it is high. A simple linear equation clearly illustrates the relationship between the CIF price of boiler coal in Europe (CP) and the price of oil (OP) for the period 1977–90: CP=0.24 OP+5.71 (9.81) R2c=0.88 DW=2.04 An examination of the same relationship in other markets (FOB price in the United States, CIF in Japan) provides much the same results, since price differences between the regions are less (of the order of 10%) and relative prices are more stable over time than in the natural gas markets (cf. 2.II.1 above). The existence of a linear relationship with the price of oil does not call into question the hypothesis of a pricefloor and -ceiling for coal. On the contrary, the variation range noted—between 9 and 15 $/bep, not including 1981 and 1982—corresponds to 35–60 $/t at production, taking into account an average freight price of $10/t. This range, although wider, is close to that put forward by P.N.Giraud16 of 40 to 65 Aus$/t (1986 prices), that is 34 to 54 $/t (1990 prices). There is thus a price function for coal, which has its boundaries defined by characteristics which belong to the coal industry, but which, within these boundaries, provides for fluctuations linked to those of oil. 2.II.3 Reconsidering oil prices The price function of coal, on the other hand, means that oil prices can be reconsidered, by introducing the concept of ‘policemen’ developed by P. Desprairies17 and J.M.Chevalier.18 This concept involves deducing oil price floors and ceilings not from conditions belonging to the oil industry itself but from conditions related to its competitivity with natural gas and especially coal: ● the price of oil cannot drop below a level at which its competitivity with coal starts to become overwhelming, the relative price being around 1;

18

INTERNATIONAL MARKETS AND ENERGY PRICES: THE POLES MODEL

Figure 2.6 Price of imported boiler coal in Europe and the OPEC price of oil (CIF, 1977–90) Source: IEPE, IEA data

● conversely, the price of oil cannot exceed the level at which large amounts of coal start to become available at a very low relative price The combination of the linear relationship mentioned above and the price floor and ceiling for coal (PFC and PCC) can be used to calculate the price floor and ceiling for oil (PFO and PCO):

and thus:

These limits in oil price variations are justified by analysis and are also absolutely conceivable. They can therefore be reintroduced in the oil price function developed above (cf. 2.I.3), where they would be very useful for setting limits to the changes simulated by the non-linear oil price function and thus ensuring consistent results. This would thus constitute a comprehensive system of price functions for the medium-long term in which the price of oil not only acts as a leading price for competing energy sources but is also constrained by the price of these other energy sources.

NEW MODELS AND NEW MODELLING METHODS

19

2.III DETERMINING THE MARKET SHARE OF THE MAJOR OIL-PRODUCING COUNTRIES OR EXPORT ZONES The international oil market can be considered as ‘one great pool’, where the suppliers are the world exporters and the buyers are the world importers, and with the breathing space formed by the inexpensive reserves of the Gulf region. This is of course a simplified picture, since, in reality, quality differentials for both supply and demand, along with transportation costs, create specific regional markets. However, the general idea of the large pool is sufficient in a model such as POLES, in that the production dynamics of the exporting countries outside the Gulf can be considered to be relatively independent of regional demand and are first and foremost a function of an average world price and the level of exploration and development in each country. Independent supply functions thus determine the capacity and production of all these countries, while the Gulf countries make up the balance of the world market. The same cannot be said of the gas and coal markets, where transportation constraints and costs have a much greater influence. One possible approach is to model world trade for these products by clarifying all the production and transportation costs and finding the inter-regional trade matrix to minimize these costs. Aside from the complexity of such a matrix, the number and nature of the data required, this approach is perhaps not the most effective for simulating medium-long term trade. In fact, it can be considered that the trade networks are relatively stable, that, in addition, they respond to the importers’ desire for a competitive market and enable them to insure against the risk of interruptions in supplies. With this in mind, the use of exogenous trade matrices, based on past profiles and taking into account long-term contracts—particularly for gas—and changes in export infrastructures, may be an effective way of simulating the world market and thus determining the actual output of the big exporters. It is this option that has been used in the medium-long term version of the POLES model. 2.III.1 Construction of simplified matrices of world trade in natural gas and coal Within the framework of the overall architecture of the model, the trade matrices must therefore help determine the actual output of the main exporting countries, under the constraint of production capacity determined country by country (by a supply function for natural gas and by exogenous hypotheses in the case of coal). Actual export levels being considered as dependent on regional demand, the world is divided into four large markets as far as importation is concerned: America (North and South), Western Europe and Africa, Eastern Europe (Central Europe and the former USSR), Asia. It is then a question of calculating the share in these markets of the big gas and coal exporters over the past years. The trade matrices published by UNO can be used to carry out this calculation: ● by including the ten biggest exporters exporting to the 22 biggest importers, 95% of world trade in natural gas in the 1980s can be reconstructed; ● the exports of the 8 biggest coal exporters to the 22 biggest coal importers represents 85% of world trade in this energy source for the same period. Simplified trade matrices are obtained when importing countries are included according to large markets (Table 2.1). These matrices highlight extremely varied structures, according to energy source and importing region.

20

INTERNATIONAL MARKETS AND ENERGY PRICES: THE POLES MODEL

It can be seen that, for natural gas, trade is extremely concentrated on the American market (from Canada to the United States) and the East European market (from the former USSR to Central Europe). Conversely, Table 2.1 Simplified trade matrices, market share of natural gas and coal exports Nat. Gas Americas

Europe/Africa 1989 E. Europe

Asia

1982 1985 1989 1982 1985 0.35 1982 1985 1989 1982 1985 1989

USSR

CAN

NOR

ALG

NETH

INDO

MAL

BRUN

IRAQ

UAE

— — — 0.28 0.27 — 1.00 1.00 1.00 — — —

0.93 0.97 0.97 — — 0.23 — — — — — —

— — — 0.27 0.23 0.20 — — — — — —

0.07 0.03 0.03 0.09 0.19 0.22 — — — — — 0.01

— — — 0.37 0.31 — — — — — — —

— — — — — — — — — 0.52 0.56 0.49

— — — — — — — — — — 0.17 0.20

— — — — — — — — — 0.33 0.19 0.16

— — — — — — — — — — — 0.08

— — — — — — — — 0.14 0.08 0.07

Coal Americas

Europe/Africa

E. Europe

Asia

1985 1989 1982 1985 1989 1982 1985 1989 1982 1985 1989

1982 0.04 0.06 0.08 0.19 0.14 0.00 0.02 0.06 0.42 0.46 0.49

AUS

USA

SA

USSR

CAN

POL

CHINA

COL

— 0.85 0.80 0.53 0.33 0.40 0.06 0.05 0.08 0.31 0.14 0.11

0.93 — — 0.21 0.25 0.19 — — — 0.07 0.13 0.11

0.00 — — 0.02 0.03 0.06 0.51 0.49 0.44 0.02 0.03 0.06

— 0.04 0.07 0.02 0.02 0.03 — — 0.00 0.15 0.19 0.18

0.02 0.07 0.07 0.14 0.17 0.09 0.43 0.44 0.41 — — 0.00

0.06 — — 0.00 0.00 0.03 — — — 0.04 0.04 0.04

— — 0.00 — — 0.08 — — — — — 0.01

—

imports to Western Europe are divided among four suppliers, each contributing more than 20% of the total. Finally, the Asian market (Japan) is dominated by Indonesia and the other Asian countries, but exports from the Middle East also play a certain role. It can be seen that, in every case, more than two-thirds of the supplies of a market come from countries belonging to the same geographical zone as the market itself and that, among the exporting countries, only Algeria and the former USSR supply more than one market. In this context, the coal market appears to be less geographically concentrated. All exporting countries supply several markets: Australia and the USA each supply four, while South Africa, on account of political constraints, supplies only two. Imports are also less concentrated. However, each market can be seen to

NEW MODELS AND NEW MODELLING METHODS

21

have a dominant supplier: the United States for the American and European-African markets, the former USSR for Eastern Europe, and Australia for the Asian market. Projecting market share matrices to the year 2010 must take into account the foreseeable emergence of new first order exporters: Qatar, Iran, Nigeria, Venezuela, Mexico and Thailand for natural gas; Indonesia and Venezuela for coal. Projections should also enable the simulation of a likely trend towards diversification of supply sources, especially in view of the political changes now taking place in Eastern Europe and South Africa. 2.III.2 The logic for closing regional and world markets by energy source Satisfaction of regional and world markets is based on the distinction made between three types of producers: ● second order producers, whose production is considered as given, in order to meet internal needs, and independent of international demand: ● first order producers, exporting to international markets and whose output depends on international demand, in accordance with the market share system and respecting capacity constraints; ● swing-producers, each associated with a market and providing the complementary production needed to maintain market stability.19 The process for calculating production levels is thus as follows: 1 Calculation of the primary consumption of each source considered, for each of the four main geographic zones (based on demand models). 2 Calculation of the volume of the international market of each zone, equal to primary consumption less the primary consumption of first order and swing-producers as well as the given production of second order producers. The object here is to isolate the market in which there is real competition between exporters. 3 Calculation of the theoretical output of first order producers based on their share in each of the four markets. 4 Downward readjustment of their exports to each market if the total exceeds production capacity, as indicated by figures in the supply models. 5 Finally, calculation of the output of make-up producers based on the difference between the volume of each market and the contribution of first order producers. All these procedures have now been developed within the model. They are currently being linked with the various supply and demand models and it will be possible to stimulate them once this step is complete. The model of medium-long term international markets thus constitutes a structured set of procedures for simulating energy prices and the production of the main exporters. It is based on behavioural functions for determining prices and on the import profiles of each main world region, thereby making it possible to combine both lessons from the past and hypotheses for future scenarios. The model therefore assumes that existing structures and behaviour have a certain significance for the future. Such an assumption appears well-founded and useful for analysing the medium-long term period but is probably much less so for the

22

INTERNATIONAL MARKETS AND ENERGY PRICES: THE POLES MODEL

very long term. For this reason, therefore, another model, based on the progressive mobilization of resources at increasing cost, has been developed for this more distant horizon. NOTES 1 Research programme developed at the IEPE and financed by the Direction Générate XII (Recherche) of the CCE and by the PIRSEM-CNRS. 2 H.A.Simon, ‘The Architecture of Complexity’, Proceedings of the American Philosophical Society, Vol. 106, No. 6, 1962, pp. 467–82. 3 H.A.Simon, ‘From Substantive to Procedural Rationality’, in S.Latsis (ed.), Method and Appraisal in Economics, Cambridge: C.U.P., 1976, 4 D.Gately, ‘A Ten-Year Retrospective, OPEC and the World oil Market’, Journal of Economic Literature, Vol. XXII, September 1984, pp. 1100–14. 5 Energy Modelling Forum No. 6, World Oil, Summary Report, Standford University, February 1982. 6 M.A.Adelman, Economic Theory of Mineral Depletion with Special Reference to Oil and Gas, in J.G.Rowse (ed.), World Energy Markets, Coping with Instability, IAEE 9th International Conference, Calgary 6–8 July 1987, pp. 43–51. 7 D.Gately, op. cit. 8 C.Dahl and M.Yucel, Dynamic Modelling and Testing of OPEC Behaviour, Oxford Institute for Energy Studies, 1990. 9 System Sciences Inc., The Oil Market Simulation Model, Model Documentation Report, Washington D.C.: USDOE, May 1985. 10 S.G.Powell, ‘The Target Capacity-Utilization Model of OPEC and the Dynamics of the World Oil Market’, in The Energy Journal, Vol. 11, No. 1, pp. 27–63. 11 M.Rauscher, OPEC and the Price of Petroleum, Theoretical Considerations and Empirical Evidence, Berlin: Springer Verlag, 1989. 12 B.Amable, J.Henry, F.Lordon and R.Topol, ‘Une tentative d’élucidation d’un concept flou: l’hystérésis’, in Méthodologie de l’économie théorique et appliquée aujourd’ hui, 1990 AFSE Conference, Paris: Nathan, 1990, pp. 135–46. 13 J.B.Lesourd, J.Percebois and J.M.Ruiz, ‘Equilibre et déséquilibre sur le marché international du gaz naturel’, in A.Ayoub and J.Percebois (eds), Pétrole: marché et stratégies, Paris: Economica, 1989. 14 J.Percebois, Economic de l’énergie, Paris: Economica, 1989. 15 P.N.Giraud, ‘Les logiques d’évolution des prix internationaux des charbons’, in Revue de l’Energie, No. 429, April 1991, pp. 226–33. 16 P.N.Giraud, ‘Le marché international du charbon vapeur; vers la maturité’, in Revue de l’Energie, No. 413, AugustSeptember 1990, pp. 741–52. 17 P.Desprairies, ‘La situation énergétique dans le monde’, in Revue de l’Energie, No. 401, April 1988, pp. 217– 20. 18 J.M.Chevalier, ‘Quels prix pour quelle énergie?’, in Sciences et Vie Economie, No. 32, October 1987, pp. 42–3. 19 Auxiliary producers of natural gas: Canada for the North American market, Norway for Western Europe, USSR for Eastern Europe, and Indonesia for Asia. Auxiliary producers of coal: USA for the North American market, South Africa for Western Europe, former USSR for Eastern Europe, and Australia for Asia.

Chapter 3 ERASME: a short-term energy forecasting model for the European Community Nikitas Deimezis

3.I INTRODUCTION The first ‘Short-Term Energy Outlook’ (STEO) was published by the Directorate General for Energy of the European Commission (DG XVII) in No. 0 of the review ‘Energy in Europe’ in December 1984.1 Since then 17 STEOs have been published, the latest in May 1991. From the beginning, the analysis and the forecast were done at the Community level, considered as a whole. This choice was made for both political and technical reasons. Practice has proved that this approach is not irrelevant and despite important differences among Member States, the energy markets in the European Community can be successfully analysed in a unique framework. The first nine STEOs, until mid-1987, were covering EUR-10, excluding Spain and Portugal. Starting from end-1987 an important change was made: The STEO now covers EUR-12 (excluding for the moment ex-GDR). At the same time, the first monthly econometric model (STEM, Short-Term Energy Model) has been replaced by a new quarterly econometric model called ERASME (Energy Relations in an Aggregate Shortterm Model for Europe). Table 3.1 presents the main characteristics of the STEO. The STEO, in its actual version presents an analysis of recent energy trends and quarterly forecasts for the following variables:2 ● ● ● ● ●

Average import prices of crude oil and steam coal in US dollars. Average final prices for 10 different forms or uses of energy in ECU. Energy demand by fuel and sector. Energy production by fuel, including electricity and refining. Stock changes and net imports by fuel.

The forecast is made as follows: A first version is obtained by running the ERASME model. Values for the main exogenous variables are partly defined on the basis of the latest macro-economic forecast of the Commission. There follows, within DG XVII, a discussion with experts of different fuels and then the final version is prepared and published.

24

ERASME: A SHORT-TERM FORECASTING MODEL FOR THE EC

Table 3.1 STEO—Main features No

EE Issue

Date

Coverage

Last known quarter

Forecast horizon

Forecast quarters ahead

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

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

Dec-84 Apr-85 Aug-85 Dec-85 Apr-86 Sep-86 Dec-86 Jul-87 Oct-87 Dec-87 Apr-88 Sep-88 Dec-88 May-89 Nov-89 Aug-90 Dec-90 May-91

EUR-10 EUR-10 EUR-10 EUR-10 EUR-10 EUR-10 EUR-10 EUR-10 EUR-10 EUR-12 EUR-12 EUR-12 EUR-12 EUR-12 EUR-12 EUR-12 EUR-12 EUR-12

2 Q 84 4 Q 84 1 Q 85 2 Q 85 4 Q 85 1 Q 86 3 Q 86 4 Q 86 1 Q 87 2 Q 87 3 Q 87 4 Q 87 2 Q 88 4 Q 88 2 Q 89 4 Q 89 2 Q 90 4 Q 90

1985 1985 1986 1986 1986 1987 1987 1987 1988 1988 1988 1988 1989 1990 1990 1991 1991 1991

6 4 7 6 4 7 5 4 7 6 5 4 6 8 6 8 6 4

3.II THE ERASME MODEL 3.II.1 General features The ERASME model is a quarterly econometric energy model. It contains a full description of the European quarterly energy balance sheet. It has 55 behavioural equations and more than a thousand identities describing the whole structure of the European Energy System, making the transformation from specific units by fuel to energy units (toe) and assuring the coherence between the monthly, quarterly and annual balance sheets. The main data source is the SIRENE statistical database of the Statistical Office of the European Community (SOEC) and in particular the monthly database, also published by Eurostat in the ‘Monthly Energy Statistics’. Data for final energy prices come mainly from the IEA. The database covers at the moment the period from the first quarter 1979 to the end of the third quarter 1990. In general, the model is fully re-estimated twice a year. Estimation is made by OLS except of the electricity block where 3SLS are used. The structure of the model (Figure 3.1) is recursive. In other words there is no feedback of the demand and the supply of energy on prices. This is a deliberate option for different reasons:

NEW MODELS AND NEW MODELLING METHODS

25

Figure 3.1 The ERASME model

1 The assumption is made that energy demand in the European Community can always be satisfied by supply, either domestically or imported. 2 Average final energy prices in the European Community are mainly defined by (exogenous) import prices, the taxation regime, national regulations and seasonal factors. In the short-term, final prices do not seem to be significantly influenced by demand changes. 3 Given the size of the model, the existence of a simultaneous block could seriously complicate the model’s behaviour. 3.II.2 General description of the model The general logic of the model is outlined below. Exogenous world oil prices, or import prices of steam coal (related, with a certain time lag to crude oil prices), the dollar exchange rate and changes in the fiscal regime are the main variables defining final energy prices. The model has endogenous price variables for 10 different forms of energy. Pf=f(PM* XR, T, SF,…) (1) where Pf PM XR

price of fuel f Price of imported energy (oil or coal) in US dollars Exchange rate ECU/US dollars

26

T SF

ERASME: A SHORT-TERM FORECASTING MODEL FOR THE EC

Taxation Seasonal factors

Energy demand by fuel and use is defined by exogenous macro-economic and sectoral variables (GDP, Private consumption, industrial production, etc.) and real energy prices. Substitutions among competing fuels are a function of relative prices. Efficiency gains due to technical progress and/or structural changes are introduced by a trend. Df=f(Q, Pf/P, Ps/P, t, SF, DD,…) (2) where Df Q Ps P t DD

Demand of fuel f Macroeconomic activity Price of competing fuel CPI Time trend Climatic conditions (degree-days)

Time lags are introduced mainly by Koyck lags. However the lag structure of the equations is redesigned in each new estimation of the model. The refining activity is linked to total oil demand and oil prices. However, this part of the model is not yet fully developed. The electricity sector is based on the assumption that total electricity generation is defined by demand. First the model calculates primary electricity (hydro, linked to hydrological conditions) and nuclear electricity (directly linked to the available nuclear capacity and seasonal considerations). Total generation of thermal electricity is then calculated by difference. In the next step, inputs to thermal power stations of solids and fuel oil are calculated on the basis of previous shares, relative prices and seasonal and climatic variables. If/TI=f([If/TI]−1, Pf/P, Ps/P, t, SF, DD,…) (3) where: If TI

Input of fuel f into thermal power plants Total inputs into thermal power plants

Natural gas use is determined as the difference to total generation needs. This part of the model is actually under review and will be soon replaced by a more sophisticated module. Production of primary sources of energy is treated in the following way: ● Production of hard coal is exogenous and it is based on national forecasts and expert opinions. ● Production of oil and gas is related to demand, prices and proxies for available resources. Stock changes by fuel are related to demand, production, prices and seasonal factors. Finally, net imports by fuel are calculated by difference to total demand. To produce a full quarterly energy balance sheet the model transforms variables from specific fuel units to oil equivalent units.

NEW MODELS AND NEW MODELLING METHODS

27

3.III FORECAST ERRORS To assess the quality of forecasts, an error analysis of previous quarterly forecasts, starting from those published in No. 9 of Energy in Europe important variables have been calculated by quarter, by forecast and (October 1987), has been conducted.3 Average absolute errors for 20 globally for the whole period.4

Figure 3.2 Real GDP (Index 1985=100): actual versus forecast

During the period under examination (December 1987 to May 1991, see Table 3.1) there have been at least four unexpected ‘shocks’, exogenous to the European energy system, that could a priori perturb the energy forecast: 1 The October 1987 stock exchange crash and the unforeseen economic growth in 1988 (Figure 3.2). 2 The Gulf Crisis and War and their impact on oil prices (Figure 3.3) and macro-economic activity (Figure 3.2). 3 Unexpected warm winters for three consecutive years (1988, 1989, 1990, Figure 3.4) and the resulting drought. 4 The accidents in North Sea oil production system (Piper Alpha, July 1988 and Cormorant Alpha, April 1989, Figure 3.5).

28

ERASME: A SHORT-TERM FORECASTING MODEL FOR THE EC

Figure 3.3 Crude oil import prices: actual versus forecast

Despite these ‘accidents’ the overall quality of the forecast seems to be satisfactory. Table 3.2 presents the average absolute error (AAE) over the whole period (44 to 47 forecast points) for the twenty important variables. The AAE are small except for oil prices (Gulf Crisis), primary electricity (drought) and crude oil production (accidents). Electricity demand and generation show the smallest error (1.7%). Total oil deliveries and total hard coal deliveries were forecast with an AAE of about 2.5%. Natural gas demand, which is more weather-dependent, shows an AAE of 3.4%. The AAE of the total energy demand (Total Apparent Consumption) forecast was 1.7%, which can be considered as an excellent result given the magnitude of the statistical error of the whole system. Table 3.3 shows the detailed error analysis for this variable. As it can be seen, 16 out of 44 forecasts have an error of less than 1%. The biggest errors were made in the forecasts of the third and fourth quarters 1988 (underestimate of demand, explained by the error on GDP) and the first quarters of 1989 and 1990 (overestimate, explained by weather assumptions).

NEW MODELS AND NEW MODELLING METHODS

Figure 3.4 Degree-days: difference from ‘normal’ Table 3.2 STEO—Error analysis

EXOGENOUS 1. GDP (*) 2. Crude oil price OIL 3. Deliveries, total 4. Motor petrol 5. Gas diesel oil 6. Heavy fuel oil 7. Crude production 8. Net imports NATURAL GAS 9. Consumption 10. Production SOLIDS 11. Hard coal deliveries 12. Deliveries to power plants

Unit

Average absolute error

In percent

Index, 100 USD/bbl

1.1 2.94

1.0% 14.2%

Mt Mt Mt Mt Mt Mt

2.9 0.5 1.7 1.3 3.0 6.2

2.5% 1.9% 4.0% 7.5% 10.4% 6.2%

Mtoe Mtoe

1.7 1.8

3.4% 5.7%

Mt Mt

1.9 2.1

2.4% 4.4%

29

30

ERASME: A SHORT-TERM FORECASTING MODEL FOR THE EC

13. Hard coal production 14. Coke production NUCLEAR 15. Nuclear heat ELECTRICITY 16. Demand 17. Generation 18. Primary electricity TOTAL ENERGY 19. Apparent consumption 20. Production

Unit

Average absolute error

In percent

Mt Mt

1.4 0.2

2.7% 1.9%

TWh

20.2

4.6%

Twh TWh TWh

7.0 7.6 7.7

1.7% 1.7% 20.3%

Mtoe Mtoe

4.7 4.7

1.7% 3.3%

Figure 3.5 Crude oil production: actual versus forecast

3.IV CONCLUSION During the last seven years the services of the Commission of the European Community has provided a quarterly Short Term Energy Outlook which is the only full energy forecast covering the whole EC and

NEW MODELS AND NEW MODELLING METHODS

31

published by a public administration. This forecast is important not only for internal purposes but also as a unique source of information for many users in the public and private sectors in Europe. In addition, experience shows that the STEO is providing forecasts of generally good quality on a field where surprises are frequent. Table 3.3 Forecast report, actua versus forecasts: total apparent consumption in Mtoe and % Forecast Quarter

Observed

1 Q 87 2 Q 87

302.0 241.3

3 Q 87

232.3

4 Q 87

287.4

1 Q 88

293.4

2 Q 88

239.8

3 Q 88

240.4

4 Q 88

294.7

1 Q 89

295.4

2 Q 89

251.6

3 Q 89

241.6

4 Q 89

299.6

1 Q 90

299.4

2 Q 90

257.5

3 Q 90

251.7

4 Q 90

302.5

Nov-87 EE 9

Feb-88 EE 10

Jun-88 EE 11

Nov-88 EE 12

Apr-89 S May 89

4 Q 87

1 Q 88

2 Q 88

4 Q 88

2 Q 89

240.4 −0.4% 228.2 −1.8% 279.5 −2.7% 298.1 1.6% 242.6 1.2% 228.8 −4.8% 283.5 −3.8%

232.4 0.0% 275.3 −4.2% 293.8 0.1% 243.8 1.7% 231.9 −3.5% 284.1 −3.6%

284.6 −1.0% 297.9 1.5% 248.8 3.8% 235.4 −2.1% 290.4 −1.5%

5

4

1 Q 91 315.9 2 Q 91 3 Q 91 4 Q 91 Average Absolute Error by Issue Forecast quarters 6

239.3 −0.2% 237.2 −1.3% 294.3 −0.1% 308.6 4.5% 254.6 1.2% 240.3 −0.5% 293.9 −1.9%

6

292.2 −0.8% 297.1 0.6% 251.9 0.1% 240.0 −0.7% 298.9 −0.2% 310.2 3.6% 258.5 0.4% 244.7 −2.8% 303.4 0.3%

8

32

ERASME: A SHORT-TERM FORECASTING MODEL FOR THE EC

Forecast Quarter

Observed

Mtoe % Per cent

Nov-87 EE 9

Feb-88 EE 10

Jun-88 EE 11

Nov-88 EE 12

Apr-89 S May 89

7.1 2.7%

7.1 2.6%

5.7 2.2%

4.5 1.6%

3.0 1.1%

Average Absolute Error by Quarter Oct-89 S Nov 89

Jul-90 S Jul 90

Dec-90 Unpubl

Apr-91 S May 91

4 Q 89

2 Q 90

4 Q 90

2 Q 91

251.5 0.0% 242.8 0.5% 302.0 0.8% 312.2 4.3% 260.4 1.1% 248.0 −1.5% 307.1 1.5%

6 4.6 1.6%

300.0 0.1% 298.6 −0.3% 256.8 −0.3% 249.2 −1.4% 307.6 1.7% 317.4 0.5% 263.4 251.8 313.5 5 2.3 0.8%

256.6 −0.3% 253.9 0.9% 301.3 −0.4% 306.6 −2.9% 260.3 249.9 306.9 3 4.2 1.4%

30 1.5 −0.3% 312.7 −1.0% 258.6 247.6 307.8 1 3.2 1.0%

Number of Forecasts

Mtoe Per cent

1 2 3 3 4 4 2 2

4.1 10.0 3.2 5.3 7.1 6.6 7.5 1.7

1.8% 3.5% 1.1% 2.2% 2.9% 2.2% 2.5% 0.7%

3

1.4

0.6%

3

2.9

1.0%

3

8.1

2.7%

3

1.5

0.6%

4

4.1

1.6%

4

3.0

1.0%

3

4.1

1.5%

44 4.7

NOTES 1 The STEO for the European Community is very close in its approach with the one published quarterly in the USA by the DOE. 2 For a detailed definition of the different variables, see STEO, May 1991, Annex I.

NEW MODELS AND NEW MODELLING METHODS

33

3 For an analysis of annual errors starting from 1984, see N.Deimezis, Short-term energy forecast: Analysis of the Forecasting Record of the Commission of the European Communities, presented at the ‘8th International Symposium on Forecasting’, Amsterdam, Netherlands, June 1988. An updated summary of this paper has been published in Energy in Europe, No. 12, March 1989. 4 For a discussion of error measurement of forecasts, see among others G. Keating (1985) and S.McNees (1978, 1981a and 1981b).

REFERENCES ‘Analysis of the Forecasting Record of the Commission of the European Communities’, Energy in Europe, Number 12, March 1989. ‘Short-Term Energy Outlook for the European Community’, Energy in Europe, Various issues. Deimezis, N., Short-term energy forecast: Analysis of the Forecasting Record of the Commission of the European Communities, presented at the 8th International Symposium on Forecasting, Amsterdam, Netherlands, June 1988. DOE/EIA, Short-Term Energy Outlook, Quarterly Publication and Annual Supplements, USA. Keating, G., The production and use of economic forecasts, Methuen London and New York, 1985. McNees, S., ‘The “Rationality” of Economic Forecasts’, American Economic Review, May 1978. —— ‘The Optimists and the Pessimists: Can We Tell Whose Forecasts Will Be Better?, New England Economic Review, May/June 1981a. —— ‘The Recent Record of Thirteen Forecasters’, New England Economic Review, September/October 1981b.

Chapter 4 The energy model MIDAS P.Capros, P.Karadeloglou, L.Mantzos, and G.Mentzas

Abstract MIDAS (Multinational Integrated Demand and Supply) is a large-scale system of country-specific energy models for medium-term energy planning developed for European countries under the auspices of CEC/DG-XII. MIDAS is a simulation model used for scenario analysis; each national application of MIDAS is a simultaneous system of about 1500 equations solved dynamically over a period of 10 years. The energy demand sub-model is a forecasting econometric system and uses relative energy prices to estimate energy demand and fuel substitutions by sector, at a disaggregated level. The energy supply sub-models combine econometrics with process analysis in formulating the energy supply behaviour. Energy prices are evaluated mainly as a function of supply costs, but they are also influenced by supply/demand disequilibrium pressures. The supply side consists of five interconnected sub-models, namely electricity generation, petroleum refining, natural gas and synthetic gas production and distribution, coal mining and coke-oven plants. The chapter also gives a short description of the linkage of MIDAS to the macro-economic model HERMES and to a sub-model evaluating CO2 emissions. 4.I INTRODUCTION Energy modelling, that is quantitative policy analysis of energy systems based on mathematical models, emerged as one of the consequences of the 1973–4 energy crisis. Early energy models merged operations research and economic analysis techniques and were used to support national or international energy policy formation. The initial focus of energy models was on investigating the strategic aspects of energy policy and on quantifying investment programmes and technological options that would reduce dependence on oil. Most of these models retained the normative paradigm in decision modelling and formulated network-based energy system representations as mathematical programming problems. The results provided by such models were generally used for intertemporal, long-range planning of the energy system. The main decision variables include energy flows and investment concerning both existing and new energy technologies, while energy supply and demand interactions are incorporated within a single programming framework. Energy prices are evaluated only indirectly, since they are induced from the shadow (dual) values of the mathematical programming problem. Numerous examples of such models can be mentioned: BESOM, TESOM of Brookhaven National Laboratory were the pioneers in USA, followed by PILOT and ETA of Stanford University, while the EFOM system is known mainly in Europe.

THE ENERGY MODEL MIDAS

35

Once strategic aspects and technological forecasting issues were sufficiently studied, the need to represent explicitly market adjustment mechanisms was identified. The incentive was President Carter’s synfuels programme at the end of the 1970s. The main issues were the investigation of the economic conditions for the success of the synfuels programme and its macro-economic impacts. The PIES (later renamed to MEFS) modelling system of the US DOE was built for this purpose. It was a system of linked sectoral energy models and used market clearing formulations. The modelling paradigm was general economic (market) equilibrium of independently deciding agents, the latter being refineries, coal suppliers, electricity producers etc. Hence energy market prices became the centre of model-based reasoning, since their adjustment induces equilibrium between supply and demand. Nevertheless, the central integrating module of PIES was still a linear programming model. Similar modelling efforts concentrated on the economic interactions of energy policy while retaining the general equilibrium paradigm; see e.g. Dale Jorgenson’s research work. Since 1982–3 strategic energy policy analysis has decreased in importance and new subjects for investigation emerged. Administrations put emphasis on short-/medium-term aspects of energy policy, including pricing, taxation and management issues. The growing concern about environmental impacts stimulated the modelling of economic and energy impacts of environment protection costs and investment. The descriptive paradigm is retained for such modelling efforts, and simulation techniques are frequently followed. Current models are generally formulated as systems of simultaneous non-linear equations and solved dynamically at discrete time intervals. In some cases, equations may be econometrically estimated, while details of energy process mechanisms may be included. Examples of short-/medium-term models are the following. The new modelling system of US DOE, namely IFFS (Murphy et al. 1984), retains the linkage architecture of PIES, but formulates market adjustment mechanisms in the centre of its functioning and abandons linear programming for the central integrating module. The latter implements a price-induced demand-supply equilibrium of markets. In addition, IFFS incorporates disequilibrium pressures, partial equilibrium, adjustments and anticipations that are consistent with the short-/medium-term perspective. Similar efforts in France have been undertaken with the Mini-DMS-Energie model, although at a significantly smaller size. In the Commission of the European Communities (CEC), the modelling gap is intended to be covered by the MIDAS energy model presented in this chapter. Notice that all three above mentioned models are designed in such a manner that they can be consistently linked both with macro-economic econometric forecasting models (DRI in the USA, HERMES in the CEC, DMS in France) and with environmental models (which evaluate emission of pollutants). Within the bulk of energy models that simulate the short-/medium-term evolution, we must make a fundamental distinction. It concerns the assumption about the energy market clearing mechanism. Assume that D=D(p) is demand as a function of prices and S=S(p) is supply also defined as a function of prices. If the market clearing price p* is calculated such that D=S, then the market is formulated as a price-adjusted one. Such a formulation corresponds to a perfectly competitive market. If the market has an oligopolistic or monopolistic character and competition does not operate perfectly, supplying firms may adopt a cost markup pricing policy. In this case, prices are a function of supply, i.e. p=p(S) while demand and supply are rationed by each other at the given level of prices, i.e. D=S=min{D(p), S(p)}. This market clearing formulation is called ‘quantity-adjustment’. If we further assume excess supply in the market, then demand is evaluated by D=D(p) prior to supply, which adjusts to demand S=D. Notice that in this case we may calculate a rate of capacity utilization as U=S/ where =S(p). This rate may influence prices, i.e. p=p(S, U), representing the disequilibrium pressures exerted on price formation. Models built for the USA energy system generally retain a price-adjustment formulation; this is the case of IFFS (Murphy et al. 1984) and IDES (Macal 1987). On the contrary, European energy models, such as Mini-

36

NEW MODELS AND NEW MODELLING METHODS

DMS-Energie and MIDAS retain quantity-adjustment formulations. This difference reflects the diverse conditions that prevail in energy markets. Nevertheless, within the perspective of the unified European market and the common energy carrier, the situation will change in Europe and models have to evolve towards the formulation of price-adjusting markets. Notice, however, that in spite of differences in formulation, both categories of models share common basic features, such as price-induced demand and supply behaviours, short-/medium-term representation of dynamics, etc. The MIDAS energy model project was initiated by N.Kouvaritakis (see ECOSIM 1986) and supported by the Commission of the European Communities. At that early stage, development was concentrated on the energy demand module. Recently, Detemmerman, Guillaume and Lecloux (1988) reformulated the MIDAS energy demand module and re-estimated econometric equations for four European countries. In parallel, the work presented here formulates the MIDAS energy supply and pricing modules and carries out the integration of the demand and supply MIDAS modules into a single model. The complete MIDAS model is available for United Kingdom, Italy, France and Germany. 4.II OVERVIEW OF MIDAS 4.II.1 Design principles Energy policy in the short/medium term has clearly an economic and management character, rather than a technological and planning one. In an eight to ten years’ perspective, plants under construction limit the alternative options while new technologies do not have enough time to penetrate; thus the degrees of freedom associated to investment programmes are limited. On the other hand, important economic analysis issues are identified in the short-/medium-term horizon. Examples include: evaluating the disequilibrium pressures in balancing energy demand and supply; regulating indigenous production and imports; satisfying peak demands for unstockable energy forms; evaluating average and marginal costs; assessing tarification and taxation policies that manage supply and demand integrity; and so on. The MIDAS—Supply and Pricing sub-model includes the following modules: ● ● ● ● ● ● ●

electricity production; petroleum refining; coal mining; natural gas distribution; synthetic gas and coke production; totals and equilibrium of energy balance; and energy prices.

The electricity sub-model constructs a linearized load duration curve using the results of MIDAS-Demand and allocates electricity plants within the load curve following a user-defined order for the satisfaction of each type of load; finally it approximates the marginal cost for each consumption category. Investment in new plants is assumed exogenous. The module evaluates electricity production, fuel consumptions, costs and tarification indices. The latter influence consumption prices. The petroleum refining sub-model uses an aggregate representation of a typical average refinery, including distillation, cracking and reforming. The desired production of each refining unit is compared to

THE ENERGY MODEL MIDAS

37

existing capacities in order to obtain estimations of required additional capacities and the rate of capacities utilization and to evaluate production costs by type of product (light, medium and heavy distillates). Refinery throughput and the flows within the refinery are evaluated by means of ex post econometric relations that consider profitability issues for major products and by means of engineering relations for lower importance products. Cost influences the competitiveness of national refineries in imports/exports and the formation of product prices. The production of hard coal is formulated by means of supply curves (involving reserves), which also serve in the evaluation of cost elements. The management problem concerning the operation of the different types of mines, when prices, costs and demand changes is represented. The formulation is based on the concepts of marginal long-range production costs and minimum acceptable selling prices. Competitive imports is estimated and protectionism policy is permitted by exogenous policy parameters. The supply of natural gas is analytically formulated. The model assumes that indigenous production of natural gas, as well as that of crude oil, are exogenous. The production of synthetic natural gas is connected to the operation of plants, depending on the exogenous investment programme. Natural gas imports by pipeline and liquefied natural gas imports are covering the gap between demand and supply. The formulation respects the constraints of pipeline throughput (min and max). The main task of the gas module is to evaluate costs. Gas supply in the medium term is mainly a load management problem (and a geographic one, but this is not considered here). Marginal costs of supply for each type of consumer are evaluated. They form the basis for the calculation of tarification indices by consumer. The latter are affected via the price module. The other transformations module includes coke-oven gas plants, blast-furnace plants, co-generation, geothermal energy and lignite. They have simple formulations, based on plant operation, production and cost estimations. Consistency of energy supply, transformation and demand flows is guaranteed, so that complete balance sheets constitute the output of the model. Energy prices evaluated by the supply submodel of MIDAS, enter the energy demand sub-model of MIDAS and induce changes on fuel demand and substitutions. Hence, a closed loop is initiated which is solved dynamically year-by-year. All costs and tarification indices that are resulting from the supply module, are inserted into the price equations of energy products. Consumer energy product prices are computed on the basis of ex-factory prices, the latter being evaluated within the MIDAS Supply sub-model. The formulation incorporates explicitly excise and VAT taxation. MIDAS evaluates international energy prices by using the dollar price of crude oil as the main exogenous variable. Most of the equations follow a rate of variation formulation connecting an energy product price with the price of a product considered as a leading one, and national cost elements derived from the energy supply modules. The MIDAS model has been installed in the TROLL (Time-series Reactive On-Line Laboratory) computer environment of MIT. The TROLL system is developed for quantitative research in economics and management science. TROLL, and consequently, MIDAS, are available both in mainframe computers and in personal computers, IBM AT/370 compatibles. The model is solved as a simultaneous system of about 1500 non-linear, dynamic, equations per country. The Gauss-Seidel algorithm is employed. The MIDAS data base comes from the following sources: ● ● ● ●

EUROSTAT detailed energy balances, OECD/IEA Energy Prices, EFOM Model DataBase, mining, refining and power generation investment statistics,

38

● ● ● ●

NEW MODELS AND NEW MODELLING METHODS

coal statistics, natural gas and crude-oil production and transmission statistics, technology parameters, and macro-economic indicators.

The use of the MIDAS model by decision-makers for the analysis of energy supply options in the mediumterm requires a step-by-step procedure. This procedure is outlined below. Once the user specifies the exact input characteristics of the scenario to be analysed (i.e. the scenario hypotheses) he/she should give values to the respective exogenous variables. Particular attention should, however, be paid to the values of all the directly affected variables in the case of exogenous investment programmes. The evaluation of the preliminary results is based on a set of output indicators that refer to rates of capacity utilization. The realization of this procedure requires a step-wise approach, since the first results of the scenario should be evaluated and used for the refinement of initial assumptions. The full circle closes with an in-depth analysis of model results and scenario assumptions. The following is a non-exhaustive list of policy analysis issues that can be studied with MIDAS: ● crude-oil prices, e.g. impacts of international crude-oil price shocks. ● fiscal policy, e.g. change of tax-rates on energy products and sectors. ● investment programmes, e.g. impact on energy demand and supply and on equilibrium energy prices of alternative technological and utilities investment programmes; impact of nuclear programmes. ● demand policy, e.g. direct and secondary impacts of incentives and measures in the energy demand side. ● impact of growth, e.g. energy systems impacts and secondary effects of macro-economic policy and forecasts. ● sectoral policy, e.g. energy systems impact evaluation of specific energy sectoral policies (natural gas, mining, refineries, etc.) ● environmental policy, e.g. carbon-tax policy, impact evaluation and potential assessment of pollution abatement equipment and reduced emission technologies. ● forecasts, e.g. consistent construction of future EUROSTAT balances and price-lists; e.g. ENERGY 2010 studies of CEC. ● EC issues, e.g. tax harmonization; EC unified market analysis; elimination of trade barriers: etc. Concerning future research on MIDAS, the aim will be twofold: 1 Given that the energy supply, demand and prices parts of MIDAS refer to the single-country level, a future research activity is to design and analyse the bilateral flows of energy products and networking connections between European countries. This will permit a consistent quantitative analysis of the impacts of the integration of the European energy market. 2 The second research activity refers to the representation of competitive price-adjusted energy markets for some of the energy products modelled by MIDAS and the transformation of MIDAS into a priceinduced equilibrium model. This is justified by the growing concern with the privatization of traditional energy monopolies and the opening-up of energy markets at the European single-market level.

THE ENERGY MODEL MIDAS

39

4.III ENERGY DEMAND SUB-MODEL Industrial energy demand is evaluated by six independent systems of equations, corresponding to the six sectors. Total useful energy demand for fossil fuels and for electricity are evaluated by econometric equations as follows: where Uf, Ue, RPf, RPe, Q denote useful demand for fossil fuels, demand for electricity, real aggregate price of fossil fuels and real price of electricity, respectively and i scans the set of industrial sectors. Demand for fossil fuels is further disaggregated in two demand components, namely liquid-gas and solid fuels demand, on the basis of relative prices. These demand components are further disaggregated in fuel demand. The generic equation is as follows: where Ej, Pjk denote energy demand for fuel j and price of fuel j relative to price of fuel k, respectively. Real aggregate energy price is evaluated by:

where PGDP is the GDP deflator, and relative fuel prices depend on the ratio Fuel prices are evaluated by the other components of MIDAS and, thus, a feedback mechanism is established. Function Φ applies distributed lags. Energy demand in transports and in the residential/tertiary sectors are more analytically evaluated and less price-dependent. Energy demand in transports is evaluated separately for road transport, air transport and navigation and depend heavily on economic activity, the car park and income. In the residential sector, energy demand formulae use variables, such as the number of dwellings, the rate of central heating penetration, the number of degree-days and the share of households holding electric appliances. Econometric equations were estimated by time-series (sample 1960–85) combined with cross-section data for the large European countries. 4.IV ELECTRICITY SUB-MODEL 4.IV Introduction Electric power planning has undergone significant changes during the past decade. The area of utility planning has expanded to cover more complicated research areas than cost minimization of electric utility operation and expansion planning. New modelling methodologies aim at covering the ‘integrated utility planning’ approach, which abandons the main assumption of given electricity demand and considers simultaneously demand and supply interactions. The purpose of the electricity sub-model of the MIDAS model is to support integrated utility planning for the short/medium term. Given this time horizon the issues related to long-run electricity capacity expansion

40

NEW MODELS AND NEW MODELLING METHODS

are neglected and investment programmes are considered given. The task of this sub-model is to regulate power supply to a given demand and compute electricity prices. The methodological choices available for the construction of the sub-model are limited, since the most widely applied technique is mathematical programming; see e.g. Bloom (1977), Bunn and Paschentis (1983), Caramanis (1983), Hutber (1973), Jenkin (1973) and Sherali and Soyster (1983). Another approach, first used by Griffin (1977), is to derive an econometric simulation model by applying LP techniques with pseudo-data. Unlike traditional approaches, the methodology retained in the following adopts the simulation approach; see e.g. the electricity market model of the US Department of Energy in US DOE (1984), and the electricity sub-model of the French Mini-DMS-Energie model described in Karadeloglou (1985) and Plateau (1985). The sub-model performs the following sequential tasks: construction of the load duration curve from variables evaluated by the demand sub-model; merit-order dispatching constrained on available plant capacities and evaluation of the electricity balance; computation of marginal costs and electricity prices by type of consumer. Electricity prices are re-injected into demand and, thus, a closed loop iteration is initiated. 4.IV.2 Technical description A fundamental characteristic of electric utilities is that they cannot inventory their product easily in order to meet fluctuations in demand. Thus, forecasting of demand-shifts and the way they affect capacity use and costs is an important feature in modelling the behaviour of electric utilities. The MIDAS electricity submodel starts from demand and computes a load duration curve. That is, it neglects particular features of demand as represented by continuous chronological load curves, and bases the analysis only on load duration. A step-wise linear discretization of the load duration curve is employed. Total electricity consumption per sector and type of end-use is given from the MIDAS Demand submodel. For each end-use of electricity, an annual load duration curve is defined. A linear approximation of the load duration curve is developed by slicing it into three points corresponding to 2000, 6000 and 8760 hours; see Figure 4.1. Power by time segment is evaluated in the way presented below. Notice that the main exogenous assumption concerns only the shape of the load duration curve for each end-use category. Demand-orientated policies may change these shapes. Assume that is the annual peak demand of the use j. Further assume that parameters aj, bj, cj are exogenously defined. Then, the other points of the polygon are computed by: Total electricity demand of the end-use j, , is given from the demand sub-model. Thus it is possible to compute from: This procedure is repeated for each electricity use of the MIDAS Demand model, i.e. for: ● ● ● ● ● ● ●

Iron and steel; Non-Ferrous metals; Chemical; Building Materials; Paper Pulp; Other Industries; Space heating;

THE ENERGY MODEL MIDAS

41

Figure 4.1 Linear approximation of the load duration curve

● ● ● ●

Cooking and water; Specific uses; Transport; and Exports.

It should be mentioned that a separate formulation of the curve has been considered for each use, so that engineering characteristics have been taken into account. The synchronization phenomenon is also taken into consideration. A summation of the separate curves for various electricity consumers gives the total annual load duration curve. That is:

where i: ø ,2, 6, 8. Having evaluated the demand-driven annual load duration curve, the next step is the utility management process which deals with the allocation of alternative capacities. Concerning this process, the model follows the standard merit-order dispatching procedure, i.e. it ranks equipment capacity in order of increasing operating costs and positions it against the load duration curve. The amount of electricity generated by each plant type is the area of the load curve covered by that plant type. The merit-order dispatching procedure operates, in the model, as follows: the load duration curve is divided in three regions; the dispatching is performed separately for each region; the dispatching is based on an ascending ranking of plants following their marginal costs; the latter are computed differently for existing and for potentially new plants, by considering short-term (excluding amortized investment costs) and long-term costing respectively; plant allocation is done along the load duration curve from the base towards the peak. The output of the procedure includes both the rate of use of existing plants and new investments. The latter updates the set of existing capacities in subsequent time periods. The main problem with which the electricity sub-model copes, is plant location by taking into consideration available capacities, certain technical constraints, technological characteristics of each plant, and production costs; see Figure 4.2, The sub-model considers various types of plants, including: nuclear, run-of-river, coal and lignite, lakes, fuel-oil, natural gas (boiler), peak devices and imports. Once the cost-

42

NEW MODELS AND NEW MODELLING METHODS

Figure 4.2 Allocation of plants in the electricity submode

order of the plants is defined, plants are allocated in four segments of the load curve: 0 to 500, 500 to 2000, 2000 to 6000 and 6000 to 8760 hours, so that the order of allocation is the same for each slice of the load curve. The allocation is such that supply matches demand separately for base demand and peak demand. The amount of energy provided by each plant is the area of the load curve covered by the plant. The above procedure continues until the whole range of the load curve is allocated. Finally, the sum of energy quantities that a power plant provides in different time-intervals gives the electricity produced (Ej) by each plant. By dividing effective energy produced by potential energy for each type of plant, the model evaluates the rate of capacity utilization. Fuel consumption in electricity production plants is estimated as Fij=Ej/EFj where EFj is the efficiency factor of the plant j and i is the type of fuel. Thus, the sub-model evaluates transformation inputs that enter the energy balance. 4.IV.3 Electricity prices Closely associated with the problem of capacity expansion and allocation is the electricity pricing problem. The major study in this issue is the work by Boiteux; see also Sherali and Soyster (1983). As Boiteux argues, at a solution which optimizes the welfare of the society, prices must equal the long-run marginal cost (which includes capital cost elements). This is desirable from the viewpoint that marginal costs reflect the actual cost of expanding service to meet additional demand, and hence prices tied to marginal costs are able to provide revenues from continuing service. In order to tackle the problem of evaluating electricity prices in the short to medium term, and not losing sight of the long-run perspective, the sub-model proceeds with a piece-wise approximation of the linear programming formulation. Assume that plant types are ordered according to increasing total cost and that electricity consumers are listed according to adequate criteria, which represent the importance of load factor

THE ENERGY MODEL MIDAS

43

and quantity demanded. This classification follows an order of decreasing load factor. Hence, the electricity demand of the first consumer is satisfied from the least cost power plant type. If the capacity of this plant type is not enough, demand is satisfied by assuming that the remaining part will be covered by the second plant type. In the case of excess capacity, the surplus will be used for the satisfaction of demand generated by the second consumer in the above mentioned order. Thus the evaluation of the pre-tax electricity price for each type of consumer takes into consideration the percentage shares of plant types used, the position of each consumer in load factor terms and the difficulty of providing electricity to each consumer. In this sense it can be considered compatible with the long-term objective. The MIDAS-Demand sub-model evaluates demand for five categories of electricity consumers. The price of electricity for each of these categories is based on the algorithm outlined below. For each of the major types of plants of the supply system the annual unit costs are computed by incorporating explicitly: specific capital costs annualized by using a discount rate based on long-term interest rates; interest payments during construction (allowing for different construction period lengths for different plant types) using current interest rates; annual fixed charges; and fixed operating and maintenance cost. The sum of these costs is then divided by 8760 hours multiplied by two load factors, the first represents the location of the plant type on the load duration curve and the second reflects the specificity of each category of consumer. Thus, the fixed cost per kWh by plant type and type of consumer is obtained. The variable non-fuel costs applicable to each plant type and the fuel costs are added to the above. The former are also affected by the category of consumer as the fuel efficiency depends on the load factor. From the above, a total cost Cij for each type of plant i and each consumer category j is derived. Moreover, the model uses the Wj consumption weights for each consumer category j. These weights are computed from variables coming from the Demand sub-model. The model computes also the available capacity shares Vi for each plant type (hydro, hard coal, lignite, fuel-oil, natural gas and nuclear). These shares are computed on the basis of the primary input (or equivalent) into power stations and, hence, take into account the location of each plant on the load duration curve. The elements are thus set for the resolution of the following linear optimization problem, which is simply a ‘transportation’ problem:

subject to:

The above linear programming problem is discretized, by applying an allocation algorithm which is based on a pre-ordering of plants and consumers. The ex-tax electricity price for each consumer type (PWj) is then computed by assuming that its growth rate is equal to that of the weighted sum of the total cost attributed to each consumer type.

44

NEW MODELS AND NEW MODELLING METHODS

4.V REFINERY SUB-MODEL 4.V.1 Introduction The last years have been a period of great change and unprofitable economics of refining, reflecting a significant drop in the demand for refinery distillation capacity and at the same time a major change in the relative demand for different refined products. A closure of many refinery units has been observed, while massive investment is required for the construction of products under new EC standards (e.g. for the production of unleaded petrol). The main management and strategic planning issues for European refineries are identified as follows: 1 increased intake flexibility is desirable to determine input mix as function of relative crude oil to feedstock prices; 2 increasing cracking and reforming capacities are necessary in order to cope with the increasing share of light distillates; notice that the closure of refineries concerned mainly traditional refineries which had limited cracking capacities; in the same time, an increase in cracking investment is observed, especially in modern refineries; 3 international competition is increasing, see Brondel (1984), especially concerning Middle East and African refineries; see also Caruso and Kramer (1984); 4 important investment is needed to apply the Community legislation for unleaded motor petrol and reduced sulphur content fuels. The modelling experience in the description of refining structures is practically limited to the application of linear programming techniques to detailed representations of the sector. Such models minimize total production cost when demand of petroleum products is exogenously given. Short-term models consider in addition that refining capacities are given, while long-term ones include both refining flows and investments as unknown variables of the optimization. Examples of refinery models include the well-known refining model of IFP (Institut Français du Pétrole) in France and the REMS model of the US Department of Energy; see US DOE (1984) and Garvin (1956). In order to incorporate a refinery sub-model into global energy models, two approaches have been identified: the first builds an aggregate version of broader dedicated refinery models, while the second adopts a simulation approach and usually generates econometric equations from pseudo-data constructed by multiple runs of the respective mathematical programming refinery models. Examples of applications of the latter technique are the refining sub-model of the French Mini-DMS-Energie model (see Babusiaux and Chaplon (1981) and Babusiaux and Valais (1981)) and the US Intermediate Future Forecasting System (see Murphy and Conti 1984). By adopting a simulation approach in MIDAS, it was necessary to formulate in simultaneous equations form, the complex refining structure and the related management issues. Pseudo-data generation was impossible due to the lack of a linear programming model. Hence, rather simple econometric techniques were adopted to formulate the refinery management behaviour.

THE ENERGY MODEL MIDAS

45

Figure 4.3 The representative refinery

4.V.2 Technical description The refinery sub-model of MIDAS concentrates mainly on the refining process characteristics, investment decisions and cost evaluations. As a single-country model (at least in its current version) it neglects issues related to international competition. The sub-model considers a typical representative refinery which comprises of three units, namely distillation, cracking and reforming; see Figure 4.3. In this respect, it is in accordance with data available from the Statistical Office of the European Community. The refinery distils crude oil and assimilated feedstocks and produces ten different petroleum products. It should be noticed that these assumptions imply important limitations to the accuracy of representation, since the aggregation problems in refining models are difficult to deal with; see also Babusiaux and Valais (1981), Babusiaux and Champion (1981) and Dembo and Zipkin (1983). These authors justify, however, the choice of a ‘high conversion’ refinery which is compatible to the one adopted as representative refinery type in MIDAS. The sub-model evaluates internal refining flows, crude-oil input, output, costs and optimal capacities of the three units. Simple relations ensure technological feasibility of the solution. After testing different econometric formulations, a supply-orientated representation of refining management was retained. Demand of petroleum products, classified in light, medium, heavy and gaseous distillates, is computed by the demand and the other sub-models of MIDAS. The refinery sub-model computes supply behaviour and then determines imports of each petroleum product by the difference between demand and supply. The level and structure of demand for petroleum products, as well as petroleum product prices and crude oil prices, influence a set of simultaneous behavioural equations which determine total refining throughput, the level and structure of production of petroleum products and the optimal capacities in distillation, cracking and reforming. Investment and variable costs, the degree of cracking capacities, the rates of capacities utilization and the international petroleum product prices are then used to compute the ex-factory prices of petroleum products. The latter transmits two types of feedback

46

NEW MODELS AND NEW MODELLING METHODS

effects: the first towards the behavioural equations of refining production and investment, as mentioned above, and the second towards final demand. Thus, two closed loop iterations are initiated. The following equations illustrate the functioning of the MIDAS refinery sub-model:

where K denotes capacity, D stands for demand; i scans the set of petroleum products; dist denotes distillation; Φ applies a lag structure;

where i denotes selected light/medium distillates; and j stands for cracking and reforming;

where Coil is the refinery throughput and Φ applies a distributed lag structure; Ur=Coil/Kdist where Ur is the capacities utilization rate;

where Oi, denotes refinery production of petroleum product i; where CCj is the unit production cost of conversion unit j (i.e. distillation, cracking and reforming), CKj is the capital cost depending on the unit investment cost, the lifetime, the interest rate, the construction time and the rate of utilization, CFj is the unit fixed cost and CVj is the unit variable cost;

and thus product prices are computed on the basis of costs and international competitive prices p*, by using also the allocation of conversion units to the production of the different petroleum products. The refinery sub-model of MIDAS evaluates all energy flows of EUROSTAT energy balances. Crude-oil price is assumed exogenous, while international petroleum product prices are evaluated by econometric equations in which the crude-oil price is leading. 4.VI GAS SUB-MODEL 4.VI.1 Introduction While the engineering economics of natural gas supply, especially for pipelines, have been extensively developed in the literature, the experience with macro-modelling of gas supply is rather limited. The Gas Analysis Modeling System of the US DOE (1984) is a noticeable exception. The model considers all gas

THE ENERGY MODEL MIDAS

47

related actors within a detailed disaggregated structure and adopts the price-induced equilibrium approach. Gas suppliers’ behaviour is represented by linear programming sub-models, while the links with demand are covered by means of a detailed geographical allocation model. Other modelling experiences include the TENRAC engineering and financial model which has a rather microeconomic character (see Brooks 1984) and the early work of Kent Anderson (1975) at Rand Corporation. The objective of the gas supply sub-model of MIDAS is twofold: to balance demand and supply of natural and synthetic gas; to evaluate costs and prices. The sub-model starts by considering as given the demand for natural and synthetic gas, which are evaluated in the demand and the other sub-models (e.g. electricity sub-model). By assuming as given the capacities of gas production, by means of exogenous investment rates, the gas sub-model computes the quantities of supply and then determines imports of natural gas by pipeline and in liquefied form, so that demand is satisfied. Hence, gas supply is an allocation problem, consisting in allocating the quantities of gas supply from different sources to a set of demand categories. A load management problem is in the centre of this allocation procedure, similar to that of the electricity sub-model. By considering explicitly load factor characteristics for each type of consumer, the gas sub-model applies an allocation algorithm, evaluates marginal costs of gas and computes ex-factory prices. The latter transmit feedback effects to demand. Capacity expansion issues, field exploration and indigenous natural gas production although vital for long range strategies and world-wide analysis are considered incompatible with the character of MIDAS; hence they are treated as exogenous. 4.VI.2 Technical description The gas sub-model considers the following suppliers: autonomous natural gas extraction; oil-associated natural gas extraction; natural gas imports by pipeline; imports of liquefied natural gas; coke-oven synthetic gas production; gas-works production; blast-furnace gas production; gas from coal gasification. For each supplier, the sub-model evaluates unit production costs, as follows:

where K, F, V are the capital, fixed and variable costs respective, ηj is the efficiency rate, Pgdp is the GDP deflator, Pwages is the wage rate, r is the interest rate, Tj is the life-time and pj is the price of the gas type. The latter is evaluated by means of econometric equations as a function of international gas prices and the crude-oil price. A load duration curve for gas demand is derived endogenously from the demand variables computed in other sub-models of MIDAS. By aggregating supply capacities in four categories (capacities are updated from exogenous investment rates), supply units are located on the load duration curve and their contributions are computed. A correction is applied on pipeline throughput, in order to respect maximum and minimum contractual obligations. Imports of liquefied natural gas are then evaluated so that supply (especially peak supply) matches demand. The load duration curve management provides energy balance figures, but also load factor estimates for each gas supplier. Demand for gas is allocated in three consumption categories following EUROSTAT definitions, which are based on the consumer’s load factor. The unit cost of gas supply for each gas source and each consumer is derived by applying load factors to the elements of cost evaluation presented above.

48

NEW MODELS AND NEW MODELLING METHODS

A linear ‘transportation’ problem is then formulated, as in the electricity sub-model, for the evaluation of consumption marginal costs. The optimization problem is piece-wise approximated. Results feed distributed lag formulae that evaluate ex-factory gas prices for the three consumption categories. These prices are used in the demand sub-model. 4.VII COAL SUB-MODEL 4.VII.1 Introduction The supply of coal, in terms of production and prices, is dependent on a set of factors including resource availability, coal mining profitability and competitiveness, new exploitable reserves, as well as governmental policy. The situation in Europe is not uniform, since some countries have a significant potential, while coal mining in some others is in decline. The experience in coal modelling is old and rich. Models are generally specific to the coal sector and incorporate disaggregated representations. Three categories of models were identified: 1 coal mining management models; 2 coal demand and supply market clearing systems; and 3 forecasting of microeconomic production costs. Some coal models include all three features, but most cover only one of them. Coal supply models adopt engineering analytical approaches and perform cost accounting. Most of them are influenced by the work of Zimmerman (see Lev, Murphy 1983), who proved the statistical dependence of costs on labour productivity which is affected by seam thickness. Supply curves have been used in the centre of large-scale models, especially in those of the US DOE (1978, 1979, 1982). Supply curves are linear approximations of supply conditions and their use is based on the concept of minimum acceptable selling price, which determines the profitability conditions for a mine to produce. This price is also connected to the long-run marginal production cost, which, in turn, is related to the depletion of reserves, as theory suggests. Demand and supply market clearing may be considered as an allocation problem which links a set of mines with different characteristics to a set of consumers with different demand levels and geographic locations. This problem is mostly formulated as a linear programming model (or even quadratic), as in US DOE (1978, 1979, 1982), the Coal and Electric Utilities Model (US FEA), the Los Alamos Coal and Utility Modeling System, the Argonne Coal Market Model, the DRI/Zimmerman Model and the Newcomb Coal Price Index-Quadratic Programming Model, all in USA. Some smaller scale attempts have been made also in UK, see UK DOE (1978). 4.VII.2 Technical description According to the short-/medium-term character of the model, MIDAS considers coal mining capacities and reserves as exogenous. The coal supply sub-model of MIDAS aims at simulating the adjustment rates of coal production and imports to the changing demand, pricing and resource availability conditions of the

THE ENERGY MODEL MIDAS

49

coal market. The sub-model output matches coal demand and supply and evaluates coal prices on the basis of marginal costing principles, so that pricing is consistent with supply and demand interaction. A closed loop cycle between demand, supply and prices is solved by MIDAS, as in the other sub-models. The coal sub-model considers three types of mines, namely surface mine, underground mine with medium seam thickness and underground mine at high depth. At this classification, it builds a coal supply curve and computes cost figures; these have two components, one related to average cost elements and the other depending on resource depletion (in other terms it approximates long-run marginal costs). For a given market price of coal, a marginal principle leads to optimal production levels: from each mine produce coal at a rate such that marginal cost equals the market selling price. The optimum management of the mine occurs at the point where short-term marginal cost equals the long-term marginal one. The marginal principle is implemented by computing the minimum acceptable selling price; the latter indicates the appropriate minimum prevailing market price which makes the operation of a given type of mine competitive. The share of coal imports within total demand depends on the rate of domestic to international coal prices. Once the optimal shares within demand of imports and national producers (by type of mine) are evaluated, the model applies a gradual adjustment process from actual to optimal values. Domestic coal prices are evaluated on the basis of production costs weighted by the market shares of the mines and governmental subsidizing policy. This price serves to re-evaluate the minimum acceptable selling prices, so that a closed loop within the coal sub-model is iterated. Coal prices also affect demand for coal, and thus a major closed loop cycle is iterated. Total coal demand is determined by other sub-models of MIDAS. The coal sub-model first updates reserves by type of mine and capacities on the basis of past production and exogenous rates. The analysis refers to a coal supply curve, as in Figure 4.4. The minimum acceptable selling price is estimated by using the long-run marginal cost for each type of mine and by taking into consideration a certain state policy on indigenous coal production. The long-run marginal cost, equivalent to production cost, is estimated by using the following equation:

where is the production cost-long run marginal cost (LRMC), Qj,max represents total available reserves and βj, nj are parameters. represents the total average unit production cost for each type of mine j. This is the sum of the unit investment cost (CCj), the unit fixed cost (CFj) and the unit variable cost (CVj): CTj=CCj+CFj+CVj The unit annual capital cost of investment is a function of the unit investment cost, the life time of the mine, the interest rate and the construction time. These are computed for each type of mine. The minimum acceptable selling price is estimated by additionally taking into consideration the state policy. If the market price Pm is greater than the estimated minimum acceptable selling price of mine type j, then is equal to the ex-ante full the mine j remains competitive and thus the maximum available supply capacity output determined by investment accumulation. When the minimum acceptable selling price Pj,t exceeds the market price Pm,t, the operation of mine is unprofitable and consequently a closure of mine of type j may be envisaged. By taking into consideration the inherent inertia in these decisions, concerning the short or medium term horizon, the closure of a mine operates gradually. However, a lower bound in coal production of mine of type j is maintained. This can be attributed to certain technicoeconomic restrictions.

50

NEW MODELS AND NEW MODELLING METHODS

Figure 4.4 The coal supply curve

A mechanism for allocating production of the three types of mines to demand is incorporated in the coal sub-model. This mechanism is based on the assumption that it is profitable that cheaper mines provide first coal to the market. Thus, the coal sub-model evaluates the optimal levels of coal production per type of mine and derives a first approximation to the capacity utilization rates. Then, it proceeds to the evaluation of the demandadjusted coal production levels per type of mine. This consists in applying the optimal market shares per type of mine, computed as mentioned above, to domestic demand for domestically produced coal. The latter is obtained by subtracting from total demand the share of coal imports. The relation between average production cost of indigenous coal and international price of coal determines the level of coal imports. The level of imports are evaluated by using the following formula:

where M stands for coal imports, D is coal demand, Pnat is the domestic price of coal, computed as a weighted sum of cost-derived prices per type of mine, and p* is the international competitive price of coal. The market price of coal is then computed by means of an econometric equation which links domestic cost-derived prices to international coal prices. where ShD and SHM are the shares of domestic demand for domestic coal and that of imports, respectively. The market price Pm transmits a feedback effect on the allocation of mines through the minimum acceptable selling price. REFERENCES d’Alcantara, G. and Italianer, A. (1982) ‘European project for a Multinational Macrosectorial Model’, CEC, MS11, XII/ 759/92.

THE ENERGY MODEL MIDAS

51

Anderson, P.Kent (1975) ‘A Simulation Analysis of US Energy Demand, Supply and Prices’, Rand Co., R-1591-NSF/ EPA. Babusiaux, D. and Chaplon, D. (1981) ‘La Modelisation du Secteur Raffinage dans Mini-DMS-Energie’, Institut Français du Petrole, Rapport No. 29585, October, pp. 1–22. Babusiaux, D. and Valais, M. (1981) Energy Modeling and Aggregation of Refining’, in W.Hafele (ed.), Modelling of large-scale Energy Systems, Pergamon Press, pp. 321–8. Bates, R. and Fraser, N. (1974a) ‘The electricity supply industry’, in Investment Decisions in the Nationalised Fuel Industries, Cambridge University Press. —— (1974b) ‘Investment Decisions in the Nationalised Fuel Industries’, The Natural Gas Industry, Cambridge University Press, London. Bloom, J. (1977) ‘Optimal Generation Expansion Planning for Electric Utilities Using Decomposition and Probabilistic Simulation Techniques’, in B.Lev (ed.), Energy Models and Studies, North Holland, pp. 495–511. Brondel, G. (1984) ‘The Refining Industry: Community Policy and Future Structure’, presented in the Energy Economics Group Symposium, ‘Economics of Refining’, Institute of Petroleum. London. Brooks, R.E. (1984) ‘The TENRAC Gas Pipeline Competition Model’, in B.Lev, J.A.Bloom, F.H.Murphy and A.S.Gleit (eds), Analytic Techniques for Energy Planning, North Holland, Amsterdam. Bunn, D.W. and Paschentis, S. (1983) ‘Economic dispatch of electric power by stochastic linear programming’, in B.Lev (ed.), Energy Models and Studies, North Holland, Amsterdam. Capros, P., Karadeloglou, P. and Mentzas, G.N. (1989) ‘Energy Policy Extensions of KLE-based Macroeconomic Models’, Journal of Policy Modeling, Vol. 11, No. 4, Winter, pp. 507–30. Capros, P., Karadeloglou, P., Mentzas, G.N. and Valette, P. (1989) ‘A New Modeling Framework for Medium-term Energy-Economy Analysis in Europe’, Energy Journal, October. Capros, P., Karadeloglou, P. and Mentzas, G.N. (1990a) ‘Short and Medium-term Modeling and Problems of Model Linkage’, Energy, The International Journal, Vol. 15, No. 3/4, March/April, pp. 301–24. —— (1990b) ‘Carbon-tax Policy and its Impact on CO2 Emissions’, paper presented at a Workshop organised by the Commission of the European Communities DG XII, April. Caramanis, M. (1983) ‘Electricity Generation Expansion Planning in the Eighties: Requirements and Available Analysis Tools’, in B.Lev (ed.), Energy Models and Studies, North Holland, Amsterdam, pp. 541–62. Caruso, G. and Kramer, M. (1984) ‘World Refinery Suppy/Demand Outlook’, presented in the Energy Economics Group symposium, ‘Economics of Refining’, Institute of Petroleum, London. Charnes, A., Cooper, W. and Mellon, B. (1952) ‘Blending Aviation Gasolines’, Econometrica, 20, pp. 13–18. Ciliano, R. (1984) ‘An Overview of the TENRAC Oil and Gas Supply Model’, in B.Lev, J.A.Bloom, F.H.Murphy and A.S.Gleit (eds), Analytic Techniques for Energy Planning, North Holland, Amsterdam. Data Resources Inc. (1984) ‘U.S. Energy Model Documentation’, Spring. Dembo, R. and Zipkin, P. (1983) ‘Construction and Evaluation of Compact Refinery Models’, in V.Lev (ed.), Energy Models and Studies, North Holland, Amsterdam, pp. 525–540. Detemmerman, M.V., Guillaume, Y. and Lecloux, M. (1988) ‘MIDAS Demand’, Report to the Commission of the European Communities, DG XVII. Ecosim, S. (1986) ‘The MIDAS Energy Model’, Report to the Commission of the European Communities, DG XVII. Eisenhauer, J.L. (1984) ‘Comparative Analysis of Coal-based Electric Energy Delivery Systems’, in B.Lev, J.A.Bloom, F.H.Murphy and A.S.Gleit (eds), Analytic Techniques for Energy Planning, North Holland, Amsterdam. Garvin, W. (1956) ‘Applications of Linear Programming in the Oil Industry’, Management Science, 3, pp. 1–22. Gordon, R.L. (1983) ‘The Evolution of Coal Market Models and Coal Policy Analysis’, in B.Lev (ed.), Energy Models and Studies, North Holland, Amsterdam. Griffin, J. (1977) ‘Long Run Production Modeling with Pseudo Data: Electric Power Cost Function’, Bell Journal of Economics, Spring, pp. 112–27. Hutber, F.W. (1973) ‘Modeling of energy supply and demand’, in Energy Modelling Special ‘Energy Policy’ Publication, presented at a special workshop organized by the US National Science Foundation and the Energy Research Unit, Queen Mary College, London, 15/19 October, pp. 44–56.

52

NEW MODELS AND NEW MODELLING METHODS

Italianer, A. (1986) ‘The HERMES Model: Complete Specification and First Estimation Results’, CEC, EUR 10669 EN. Jenkin, F.P. (1973) ‘Electricity supply models’, in Energy Modeling Special ‘Energy Policy’ Publication, presented at a special workshop organized by the US National Science Foundation and the Energy Research Unit, Queen Mary College, London, 15/19 October, pp. 44–56. Karadeloglou, P. (1985) ‘La Production d’Electricite dans Mini-DMS-Energie’, INSEE, Direction des Syntheses Economiques/Service des Programmes, no 320/ 151, Paris. Lev, B. and Murphy, H.F. (1983) ‘The sensitivity of the shape of the long-run supply curve of coal to different assumptions on seam thickness’, in B.Lev (ed.), Energy Models and Studies, North Holland, Amsterdam. Longley, R.C. (1984) ‘The Outlook for the Economics of Refining’, presented in the Energy Economics Group Symposium, ‘Economics of Refining’, Institute of Petroleum, London. Macal, C.M. (1987) ‘Overview of IDES: An Integrated Demand and Energy Supply Equilibrium Model’, in B.Lev et al. (eds), Strategic Planning in Energy and Natural Resources, North Holland, Amsterdam, pp. 17–30. Maingy, Y. (1967) Economic de l’énergie, DUNOD, Paris. Murphy, F. and Conti, J. (1984) ‘An Introduction to the Intermediate Future Forecasting System’, in B.Lev, J.A.Bloom, F.H.Murphy and A.S.Gleit (eds), Analytic Techniques for Energy Planning, North Holland, Amsterdam, pp. 255–64. Murphy, F., Conti, J., Shaw, S. and Sanders, P. (1984) ‘Modeling and Forecasting Energy Markets with the Intermediate Future Forecasting System’, Operations Research, Vol. 36, No. 3, May-June. O’Neill, R.P. (1984) ‘The Gas Analysis Modeling System’, in 13 B.Lev, J.A.Bloom, F.H.Murphy and A.S.Gleit (eds), Analytic Techniques for Energy Planning, North Holland, Amsterdam. Offen, M.J. (1984) ‘Overall Review of the Middle East and North African Refining Industry’, presented in the Energy Economics Group Symposium, ‘Economics of Refining’, Institute of Petroleum, London. Plateau, C. (1985) ‘Mini-DMS-Energie’, paper presented in ‘Incidences Macroeconomiques des Grandes Choix Energetiques: La Modelisation’, Seminaire du Centre de Geopolitique de l’Energie et des Matières Primaires’, Institute de Recherches Internationales, 11 Janvier. Price, J.P. (1984) ‘Coal supply models: The State of the art’, in B.Lev, J.A.Bloom, F.H.Murphy and A.S.Gleit (eds), Analytic Techniques for Energy Planning, North Holland, Amsterdam. Sarkes, L.A. (1983) ‘Natural Gas’, Chapter 8 in R.A.Meyers (ed.), Handbook of Energy Technology and Economics, John Wiley, Chichester. Sherali, H.D. and Soyster, A.L. (1983) ‘Analysis of Network Structured Models for Electric Utility Capacity Planning and Marginal Cost Pricing Problems’, in B.Lev (ed.), Energy Models and Studies, North Holland, Amsterdam, pp. 113–34. Sullivan, R.L. (1977) Power System Planning, McGraw-Hill. Tobin, L.R. (1984) ‘A network programming system for studying Coal transportation’, in B.Lev, J.A.Bloom, F.H.Murphy and A.S.Gleit (eds), Analytic Techniques for Energy Planning, North Holland, Amsterdam. Valette, P. and Zagamé, P. (eds) (forthcoming) ‘The HERMES Model’, Commission of the European Communities, DG/ XII, DG-XII. UK DOE (1978a) ‘Structure of the electricity/coal model’ (Annex 5), in Energy forecasting methodology, Energy Paper No. 29, Economics and Statistics Division, HMSO, London. —— (1978b) ‘Structure of the Gas Model’ (Annex 4), in Energy forecasting methodology, Energy Paper No. 29. Economics and Statistics Division, HMSO, London. US DOE (1984a) ‘Model Documentation: Electricity Market Module’, Energy Information Administration, Office of Energy Markets and End Use, US Department of Energy, Washington, DC. —— (1984b) ‘Model Documentation of the Gas Analysis Modeling System’, Vol. 1: Model Overview, Vol. 2: Model Methodology, Vol. 3: GAMS Software, Data Documentation and User’s Guide, DOE/EIA-0450, Energy Information Administration, Washington, DC. —— (1984c) ‘Refinery Evaluation Modeling System (REMS), Model Documentation’, Energy Information Administration, US Department of Energy, DOE/EIA-0460, October.

THE ENERGY MODEL MIDAS

53

—— (1986) ‘Model Documentation for the Oil Market Module of the Intermediate Future Forecasting System’, Energy Information Administration, US Department of Energy, DOE/EIA-MO15, April.

Chapter 5 Biproportional methods and interindustrial dynamics: application to energy demand in France Louis de Mesnard

Abstract One of the possible ways to analyse interindustrial dynamics is to compare temporally input-output tables, on the aim to detect the own change of the structure. To do that, we must take back what comes from scale variation of sectors, and what comes from margins of inputoutput tables. First, this chapter rapidly presents the method, the biproportional filtering, and its theoretical basis and mathematical properties. Then, it presents specific results for the energy activities (‘Solid Mineral Fuels and Coke’, ‘Oil Products, Natural Gas’, ‘Electricity, Gas and Water’, using the French definition) on the period including the first two oil shocks (1970–85). The cases of limited interindustrial table, interindustrial table with household, interindustrial table with household and foreign, are foreseen. The role of shocks (oil, politics, economics, etc.), substitutions of inputs and outputs, foreign trade, are indicated. INTRODUCTION Some studies measure the technological change in energy industries (Ostblom 1982), but not with a systematic analysis of input-output table for France. That is why it looks interesting to develop a methodology adapted to the intertemporal comparison of input-output tables, and apply it for the case of France. The chosen period was 1970–85: it reveals very well the evolution of energy and it shows an obvious break, because of oil shocks, development of crisis, European integration, etc. An extension beyond this period would be desirable, even if it causes some problems of joining statistical series. We shall study change in the French productive structure comparing annual input-output tables, with an original method of intertemporal comparison of matrices. A succinct exposition of the method is presented in this chapter providing a résumé of Mesnard (1990a, 1990b). We specify results for industries and products of the energy sector (‘Solid Mineral Fuels and Coke’, ‘Oil Products, Natural Gas’, ‘Electricity, Gas and Water’), for the period 1970–85. A detailed study for every sector could be seen in Mesnard (1988). The reader could see Mesnard (1990a, 1992b) for global and some detailed results for the same period.

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

55

5.I THE METHOD OF BIPROPORTIONAL FILTERING Technological change is measured here through the TES (Tableaux Entrées Sorties, i.e., input-output tables) of the French national account (we know that it is an initial bias, many other approaches could be used). Then, we must compare the terms of the input-output tables taken two by two at different dates, and we must elaborate an indicator of change deduced of a distance computation. However, it is not possible to compare directly two input-output matrices as input-output tables without providing a hypothesis: two different input-output tables have different margins, which remove any meaning to a measure of distance. 5.I.1 Comparing matrices The principle of comparison of two Input-Output Tables consists of: ● filtering the variation of margins, projecting the first table S on the margins of the second table M, that is to say, find the projected matrix P(S/M) nearest as possible of the initial matrix S, and endowed of the margins of the final matrix M; ● then computing a distance between the target matrix M (M furnishes margins) and the projection P(S/ M), which is the residue when we remove the effects mechanically bound to the variation of the margins and represented by the net structural change; it is noted: |(M) —P(S/M)|. This measure changes, taking into account the viscosity of the structure.

56

NEW MODELS AND NEW MODELLING METHODS

5.I.2 The projector The projector must be biproportional to take into account the simultaneous variation of the margins: it is the classical problem of filling in a table of which the margins are known. Many mathematical methods of projection may be foreseen. In a first group of methods, the additive correction methods consist for most of them to minimize a distance. There is euclidian projection, Friedlander’s method (Friedlander 1961), ASAM method (Durieux and Payen 1976). They generally violate the condition of positivity of the elements of the input-output tables (Thionet 1976), even if it is possible to take into account the positivity constraints (Froment and Lenclud 1976): this difficulty precludes such an approach. A second group includes the multiplicative correction methods, that drive for most of them to a biproportional form and respect the positivity condition of the elements of the matrix. There is VermotDesroches (1986): minimization of the quantity of information (Snickars and Weibull 1977), minimization of interdependence (Watanabe 1969; Giuasu 1979), maximization of entropy (Jaynes 1975; Wilson 1970), gravitational models (Choukroun 1975), RAS method (Paelinck and Waelbroeck 1963; Bacharach 1970; Lynch 1986), TAU and UAT methods (Snower 1990). We achieve a synthesis of the multiplicative methods, which allows to suggest a form called by us Synthetic Biproportional Projector method (SBP), written P(S/M)=ASB that is to say xij=AiSijBj, for every i, j. S is the structure we want to project, P(S/M) is the projected structure of which margins are known and identical to those of M. A and B are diagonal matrices assuring the respect of constraints, written (mi and mj are the margins of M, sij is the generic term of S):

The system of equations of SBP is an iterative process. Theorems we stated (Mesnard 1990) allow to demonstrate that the solution of this projector is unique, and the necessary and sufficient conditions of existence and convergence are determined: every vertex must face a potential demand superior or equal to its supply (and conversely). These conditions are satisfied in practice by input-output tables. The result of biproportionality is independent with the algorithm chosen to compute it (respecting the constraints of margins), that is to say, is independent with the form of Ai and Bj (accordingly to the central theorem in Mesnard (1990:130–2). This validates the appellation of synthetic, generalizes the result of Bachem and Korte (1979) and gives to biproportionality some rigorous theoretical basis. 5.I.3 The projections We must carry out a first group of sliding projections (measuring structural change from one year to another), P(t/t+1) P(t+1/t+2)

compared to (t+1) compared to (t+2)

and a second group of cumulative projections (reported to a fix basis),

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

P(t/t+1) P(t/t+2)

57

compared to (t+1) compared to (t+2)

Cumulative projections are not the sum of sliding projections, because SBP structure is non-linear: we must calculate it again. Actual changes may be only temporary or compensated, structural changes are longlasting, and they do not coincide. As the projection is not orthogonal and not linear, it is necessary to project on two possible paths, one prospective, sliding:

P(t/t+1) P(t +1/t+2)

compared to (t+1) compared to (t+2) cumulative:

P(t/t+1) P(t/t+2)

compared to (t+1) compared to (t+2)

and the other retrospective (some other cumulative retrospective projections could be made), sliding:

P(t+1/t) P(t+2/t+1)

compared to (t) compared to (t+1) cumulative:

P(t+1/t) P(t+2/t)

compared to (t) compared to (t)

Practically, results do not differ very much from prospective to retrospective path. First, we look to the interindustrial system limited to industries and products. Then, we widen it progressively to take into account the impact of household income and the impact of the rest of the world. Then, we compare the result of the enlarged system with those of the more restricted system, to measure the effect household and rest of the world. To simplify, we shall not use retrospective path for cumulative results, for households and for the rest of the world. 5.I.4 The indicators To measure structural change, it is necessary to settle indicators: an indicator of change in value, an indicator of change undergone by each vertex, and a contribution to the overall change in the economy. 1 Intensity of sectoral change handles every vertex in the same way, taking into account the change undergone by the vertex.

58

NEW MODELS AND NEW MODELLING METHODS

● Absolute intensity allows to compare a projection and the matrix providing margins, P(t/t+1) and (t +1), for example. It is a generalized euclidian distance between two matrices, or between two vectors (rows or columns) of these matrices. It measures, in currency units the structural change from one year to another. ● Relative intensity measures the ratio between the absolute intensity and the margin: it allows to draw back the effect of general growth of vertices. Cumulated, the same indicators correspond to the measure of change regarding to a year of reference (1970 or 1985). 2 Contribution of a vertex to the overall change measures what part a vertex takes into the overall change in the economy. The biggest vertices would generally contribute the more. Knowing the contribution of vertices to overall change allows to answer the question: what vertices have made the economy change? Technically, it is a contribution to a distance, calculated on squares: it is homogeneous to a ratio of variances, as in Data Analysis: deviations are amplified regarding to distances. We shall present each productive vertex of the energy sector, industry or product, in the case of interindustrial system, and then in the case of the system extended with households, and finally in the case of the system extended with the rest of the world. Results are given in millions of French francs (MF). 5.II ‘SOLID MINERAL FUELS AND COKE’ 5.II.1 Industry Interindustrial system

This industry is one of the most affected by structural change in the interindustrial system (noted ‘IIS’ on Figure 5.1): 16.33% and 10.95%. This is distributed over the period: cumulated evolution is regular, even if it is a little faster than in the beginning of the period (more than half of the structural movement is reached from 1976, starting from low structural movements along the period). It gives a small contribution to the global change, but it is strong mainly in 1972 (because of the increasing of the autoconsum0ption), 1976 (increasing of purchases of ‘Marketable Services to Enterprises’, of reduction of those of ‘Electricity, Gas and Water’), in 1981 and 1984 (autoconsumption increases). Evolution of the structure of supplies of ‘Solid Mineral Fuels and Coke’ is connected first to autoconsumption, including in 1976 where purchases to other sources of energy are lowering. With households

When households are taken into account (see Figure 5.2), change in this industry is the bigger among every industry: 35.08%, the first place. The level of change is fortified in the period 1971–5, then in 1979 and 1984, what is traduced at each time by a strong cumulated progression, creating extremely high levels. The cause is, at each time, a reduction of the value-added, to the essential benefit of autoconsumption. The industry takes up value-added to increase its autoconsumption. For a year like 1971, we measure an apparent decreasing in the value-added, in the autoconsumption and in the production: facing to a decreasing of its own production, and to try to maintain its autoconsumption, the industry takes up valueadded.

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

59

Figure 5.1 Productive vertex: ‘Solid Mineral Fuels and Coke’ (Industry analysis)

Figure 5.2 IIS (Industry) cumulated with households and rest of world (Ext.) With households and rest of the world

The rest of the world (noted ‘Ext.’) weaken the change from the beginning of the period, but mainly in 1981 (imports fall of 354MF, with, as a counterpart, an increasing in the value-added for 128MF and in the autoconsumption with 140MF). Though, this industry remains very changing with 20.47%.

60

NEW MODELS AND NEW MODELLING METHODS

Figure 5.3 Productive vertex: ‘Solid Mineral Fuels and Coke’ (Product analysis)

5.II.2 Product Interindustrial system

This product (see Figure 5.3) is one of the more variable over the period (10.55% and 9.33%). The years 1971 and 1972 bring the essential of the cumulated structural movement, because of the increasing of autoconsumption and sellings to ‘Mining and Ferrous Metal’ and of the decreasing of sellings to ‘Electricity, Gas and Water’: from the beginning of the period, dynamic is given for the essential. This movement is hindered in 1976, with the more important peak of instant change over the period, of what intensity exceeds 6%. It is caused by an inverse movement: increase of sellings to ‘Electricity, Gas and Water’, decrease of sellings to ‘Mining and Ferrous Metal’. Progression resumes mainly in 1978, and a little in 1979, with an increasing of sellings to ‘Electricity, Gas and Water’, and a decreasing of sellings to ‘Mining and Ferrous Metal’. Though slightly intense, these years 1978 and 1979 modify rather strongly the structure of outlets, as shown by cumulative quantity. However, reported to the initial situation, progression stops by 1980 and 1981 (new decreasing of sellings to ‘Electricity, Gas and Water’, and new increasing of sellings to ‘Mining and Ferrous Metal’; sellings to ‘Building Materials’ increase so). At the end, movement restarts from 1982, with a strong oscillation in 1984–5: decrease of sellings to ‘Electricity, Gas and Water’ and increase of sellings to ‘Mining and Ferrous Metal’. Some substitutions are clearly made between sellings to ‘Electricity, Gas and Water’ and to ‘Mining and Ferrous Metal’. With households

With households, the product of the industry ‘Solid Mineral Fuels and Coke’ acquires an important variability, of 23.31%, at the second place: instead of compensating one another, movements are additive in a same direction. Final consumption does not bring here supplementary conjunctural perturbations. To the contrary, it turns movement into a coherent direction, making it clearly perceptible (often because of the

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

61

Figure 5.4 IIS (Product) cumulated with households and rest of world (Ext.)

decrease of final consumption, essentially reported to an increasing in autoconsumption and in sellings to ‘Electricity, Gas and Water’). With households and the rest of the world

The effect of the rest of the world (see Figure 5.4) is very small over this product, however it becomes the more variable (22.41%). 5.III ‘OIL PRODUCTS, NATURAL GAS’ 5.III.1 Industry Interindustrial system

This vertex (see Figure 5.5) diverts from its initial position only at the end of the period, to reach 4.9% and 3.95% of intensity. Despite shocks, cumulated profile really increases only in 1980–1–2. It must be noted that the strong peak in 1972, measured in the sliding data profile, corresponds to a precocious attempt to return to the initial situation (clear increasing of autoconsumption, and decreasing of purchases of ‘Marketable Services to Enterprises’). The first oil shock can be measured only in 1975 (decreasing of purchases in ‘Electricity, Gas and Water’, increasing of purchases in ‘Transports’). However, change restarts from 1978 (with a decreasing of autoconsumption and an increasing of supplies in ‘Marketable Services to Enterprises’). It spins out to 1979 (autoconsumption increases), in 1980 (autoconsumption increases, with a decreasing of purchases of ‘Transports’, what drives far to the initial situation), in 1981 (always a decreasing in the purchases to ‘Transports’, combined with an increasing of purchases to ‘Marketable Services to Enterprises’), to reach a quasi maximum in 1982 (decrease of purchases to ‘Transports’, increase of purchases to ‘Marketable Services to Enterprises’).

62

NEW MODELS AND NEW MODELLING METHODS

Figure 5.5 Productive vertex: ‘Oil Products, Natural Gas’ (Industry analysis)

This denotes an important adaptation of the oil industry before and after the second oil shock, often by an increasing of autoconsumption. This corresponds to a strategy of folding up, to reduce costs: it is confirmed by the purchases of ‘Marketable Services to Enterprises’, the decreasing of purchases of ‘Transports’, quasitotally explaining the net change in the structure of supplies. With households

Oil shocks appear more clearly when a final vertex is taken into account (see Figure 5.6). So a very visible peak is formed in 1975, essentially because an increasing in the value-added for 818MF, which generates a decreasing in autoconsumption for 556MF and a decreasing of purchases of transports for 144MF. However there is no effect marked over the cumulated net situation: it corresponds to internal reorganizations of intermediate consumption. The second oil shock, in 1979, is also a decreasing of value-added for 425MF, of which effects are concentrated on an increasing autoconsumption. This continues in 1980, since value-added decreases of 376MF. Still there, the specific effect is low in cumulated terms. On the contrary, in 1983, an increasing of 300MF of value-added creates a significant impact over the cumulated curve regarding to the profile obtained before without households. In this view, we can trace the calculated evolution of value-added (it is the cell ‘value-added’ in the distance matrix). Oil shock in 1974 only follows previous oscillations. However occurring in a decreasing phase, it restraint consequences to bring on an important increasing of value-added, in 1975, probably because price effects. The second shock occurs in an increasing phase, that it contributes to brake durably, since it is only in 1983 that value-added takes some more raising, probably because the change rate of dollar. With households and the rest of the world

As awaited, and among others, ‘Oil Products, Natural Gas’ is an industry sustaining an important structural change induced by the rest of the world (to the detriment of coal), with 16.80%. It contributes to

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

63

Figure 5.6 Calculated evolution of value-added

change in the productive structure with households, for 3.82%, with the eighth place. Conversely, global movements of French imports are bound to a decreasing in imports of oil (for 6.56%). Cumulated change (see Figure 5.7) becomes continuous and quite regular over the whole period, to finish very far from its initial position. That could be the lagged effects of the first oil shock visible in 1976: imports considerably fall, −1060MF, to the benefit of an increasing of value-added for 619MF and of autoconsumption for 311MF. It is a consequence of the increasing of prices of energy and perhaps of first actions to save energy. However, it is difficult to know what come from refining for other countries and what come from needs for domestic consumption. A rather similar movement is repeated in 1980 and 1981 (imports decrease of 821 and 823MF, to the benefit of value-added (+481 and +448MF), and of autoconsumption (+187 and +251MF)). It is instructive to trace the curve of evolution of calculated imports of ‘Oil Products, Natural Gas’: it is the cell ‘imports’ of the matrix of distance, that is to say an absolute intensity (see Figure 5.8). We see clearly a constant decreasing of imports. The very strong decreasing of 1973 must be noted: it traduces the oil shock at the end of the year and difficulties of supplying that it implies immediately. Conversely, the second oil shock does not imply an important decreasing of imports because no quantitative rationing follows it: the fall was more laid out over the time (in 1980–1), and must be seen as the effect of policies to save energy. However, the fall in 1976 is explainable with more difficulty, but could be connected to a diminishment of exchange parity caused by depreciation of the French franc. 5.III.2 Product Interindustrial system

In the structure of sellings of the product ‘Oil Products, Natural Gas’, we see a period of change more intense between 1971 and 1973 (there is an increasing of sellings to ‘Electricity, Gas and Water’, to ‘Non Marketable Services’, a decreasing of sellings to ‘Transports’, a decreasing following by an increasing of

64

NEW MODELS AND NEW MODELLING METHODS

Figure 5.7 IIS (Industry) cumulated with households and rest of world (Ext.)

Figure 5.8 Calculated evolution of imports

autoconsumption; and also a decreasing of sellings to ‘Building Trade, Civil Engineering’ and to trade in 1972–3). The first two years contribute to the essential of the cumulated effect (see Figure 5.9). This movement is followed by a long lull, forming a slow comeback towards the original structure. However, the work of profound modification of the structure of outlets resumes from 1978: change of behaviour is caused by the return of attitude next to ‘Electricity, Gas and Water’: outlets to this industry cease to decrease. The period ends by a decreasing towards the initial structure in 1983 (sellings to ‘Electricity, Gas and Water’ start again to decrease), compensated in 1984 (there is a decreasing of sellings to ‘Electricity, Gas and Water’, to ‘Basic Chemicals, Synthesized Fibres’ and to ‘Building Materials’ and there is an increasing of autoconsumption). Oil shocks do not clearly influence oil products on the way they dispatch the production.

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

65

Figure 5.9 Productive vertex: ‘Oil Products, Natural Gas’ (Product analysis)

Figure 5.10 IIS (Product) cumulated with households and rest of world (Ext.) With households

A new peak, very clear, over the three years 1973, 1974 and 1975 characterizes the evolution. Final consumption of the product explains it (+922MF in 1973, −1307MF in 1974, +539MF in 1975), essentially autoconsumption for these three years and on the purchases made by ‘Transports’ in 1974 (+202MF, regarding to the situation where final vertex is omitted). This peak reflects of an oscillation in the value of final consumption received by the vertex, connected to the cumulated movement in 1973, contraried in 1974 and 1975. With households and the rest of the world

66

NEW MODELS AND NEW MODELLING METHODS

Figure 5.11 Productive vertex: ‘Electricity, Gas and Water’ (Industry analysis)

Taking into account the rest of the world (see Figure 5.10) increases the cumulated change over the product, mainly in 1976 (effect is indirect, because exports decrease of 70MF, even when final consumption decreases of 222MF, to the benefit ofautoconsumption for +311 MF) and in 1979 (export increases to the contrary of 309MF, final consumption decreases of 150MF). 5.IV ‘ELECTRICITY, GAS AND WATER’ We shall not study the effect of the rest of the world for this commodity, because it is imported and exported so little. 5.IV.1 Industry Interindustrial system

This industry (see Figure 5.11) moves from the beginning of the period, stabilizes it, then moves again at the end of the period. It has a quite important contribution (4.91% and 2.35%), but mainly an intensity among the stronger, with 15.8% and 15.21%. The cumulated curve shows: ● A first strong phase is up to 1973 (because of supplementary purchases of ‘Oil Products, Natural Gas’, and variable adjusting with ‘Solid Mineral Fuels and Coke”, and with ‘Marketable Services to Enterprises’). ● A peak is in 1976, without consequence over the cumulated curve. It is caused by an increasing of purchases of ‘Solid Mineral Fuels and Coke’ and of ‘Mining and non Ferrous Metal’ and a decreasing of supplies in ‘Oil Products, Natural Gas’, in ‘Building Trade, Civil Engineering’ and in ‘Marketable Services to Enterprises’. The industry tries to make savings against its increasing costs of energy

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

67

supplying: it shows the will to compensate the change, which is proved by the nil impact over the cumulated curve. ● A very strong resurgence, for more than 5% of intensity reflects cumulated change in 1984–5, after a stabilization. It is essentially caused by a restriction on purchases of ‘Solid Mineral Fuels and Coke’ (saving of coal), of ‘Oil Products, Natural Gas’, of ‘Building Trade, Civil Engineering’, but mainly to an increasing of supplies in ‘Mining and non Ferrous Metal’, with more than 0.7 billion French francs of supplementary purchases (prolonged in 1985). The development of the French electronuclear programme, essentially starting from 1974, does not drive the structure of supplies of the industry to turn away from its initial position up to 1983. However, the increasing in power of this programme begins only in the 1980s and accelerates quickly after. Probably this generates the purchases of nonferrous metals, seen in 1984 and 1985 (under the form of cables in copper to transport electricity). With households

Taking into account the levelling of value-added, the cumulated curve becomes quite regular and clearly less intense, with decreasing change in the beginning and in the end of the period (see Figure 5.11). The explanation of this movement is complex. In 1971, a decreasing of value-added is divided over numerous industries, which contributes very little to increase absolute change. In 1972, the final vertex brings very little to absolute change, which decreases relative change. In 1973, ‘Oil Products, Natural Gas’ and ‘Building Trade, Civil Engineering’ absorb the essential of the increasing of value-added. In 1976, ‘Solid Mineral Fuels and Coke’, ‘Oil Products, Natural Gas’ and ‘Building Trade, Civil Engineering’ profit from the decreasing in value-added. In 1984, the distribution of the decreasing of value-added in done over a big number of vertices, as in 1985 more intensely. 5.IV.2 Product Interindustrial system

Although the cumulated rhythm is quasi-linear, this vertex remains among the most precocious, because 50% of the total cumulated effect is reached within 3 years and six months. We can see three periods. In 1971–2, half of the evolution is done against base of 1970 (in 1971, more ‘Mining and non-Ferrous Metal’ is sold to ‘Smelting Works, Metal Works’, to ‘Mechanical Construction’ and to ‘Non Marketable Services’, and ‘Transports’ are less supplied; in 1972, autoconsumption decreases, sales to ‘Mining and Ferrous Metal’ decrease, sales to ‘Trade’ and to ‘Non-Marketable Services’ increase). In 1975, there are fewer sales to ‘Oil Products, Natural Gas”, to autoconsumption, to ‘Basic Chemicals, Synthesized Fibres’ and to ‘Non Marketable Services’; more sales to ‘Mining and Ferrous Metal’ and to ‘Trade’. Supplementary sales are delivered to ‘Mining and non Ferrous Metal’ against ‘Non Marketable Services’ in 1980. In 1981, sales decrease to ‘Trade’ and ‘Non Marketable Services’, sellings increase to ‘Car Trade and Repair Services’, ‘Hotels, Cafés, Restaurants’, ‘Marketable Services to Enterprises’ and ‘Hiring, Leasing for Housing’. With households

The product ‘Electricity, Gas and Water’ contributes quite considerably to change on the productive structure, for 4.04%, at the 8th place. The curve (see Figure 5.13) remains quite regular, restrained until

68

NEW MODELS AND NEW MODELLING METHODS

Figure 5.12 IIS (Industry) cumulated with households and rest of world (Ext.)

Figure 5.13 Productive vertex: ‘Electricity, Gas and Water’ (Product analysis)

1975 because a slight dynamic of final consumption, accelerates after, essentially because 1977 and 1983, where final consumption essentially contributes to change. A quite high level of change is reached, for 9. 26%. Moreover, this product contributes quite strongly to change in the structure of the final consumption, for 10.76%. CONCLUSION Perhaps some other approaches could be possible, but the adopted biproportional method allows intertemporal comparisons of input-output tables. It does not privilege the hypothesis of dominant demand,

BIPROPORTIONAL METHODS AND INTERINDUSTRIAL DYNAMICS

69

as made by the simple comparison, eventually econometric, of Leontief’s technical coefficients. Also, it does not privilege the alternative hypothesis of dominant supply (connected to Gosh’s outlet coefficients). No particular choice about demand-driven or supply-driven hypothesis is made. It traduces correctly interdependence between industries and products, what never could be a simply proportional method (as technical coefficients, outlet coefficients). It allows to measure change, taking into account the viscosity of the structure. Results show that, in the global structure, a strong change continues until 1976: the first oil shock contributes to maintain an important variability in the structure, what is without any doubt the sign of the beginning crisis. On the contrary, after 1976, external shocks, as oil shocks, involve a jamming of change: they obstruct every possibility to adapt agents (Mesnard 1990:200). We can deduce from results presented here that, in a global manner, productive vertices of energy contribute quite little to the variation of productive structure first delimited to productive vertices, then enlarged including households and the rest of the world. It is a surprising fact, perhaps, despite oil shocks: (which maintain a global disruptive role). These vertices contribute quite little to the change in the structure of the value-added, of the final consumption, of imports or exports, except ‘Oil Products, Natural Gas’ (which significantly influences domestic imports), and ‘Electricity, Gas and Water’ (which influences final consumption). Interdependence is strong among the three energetic vertices: it traduces substitutions between forms of energy and cross-deliveries. Industry and product of ‘Oil Products, Natural Gas’ are little variable. This may seem surprising, but shows that oil shocks do not influence oil itself: their action is indirect. On the contrary, ‘Electricity, Gas and Water’ is a little more variable, with the development of the electronuclear French programme. On the whole, ‘Solid Mineral Fuels and Coke’ is strongly variable: we see there well-known difficulties connected to exploitation of coal (costs too great implying closing of mines, and so forth). REFERENCES Bacharach, M. (1970) Biproportional Matrices and Input-Output Change, Cambridge University Press, Cambridge. Bachem, A. and Korte, B. (1979) ‘On the RAS-Algorithm’, Computing, No. 23, pp. 189–98. Choukroun, J.M. (1975) ‘A General Framework for the Development of Gravity Typetrip Distributions Models’, Regional Science and Urban Economics, Vol. 5, pp. 177–202. Durieux, B. and Payen, J.P. (1976) ‘Ajustement d’une matrice sur ses marges: la méthode ASAM’, Annales de l’INSEE, No. 22–23, avril-septembre 1976. Friedlander, D. (1961) ‘A technique for estimating a contingency table, given the marginal total and some supplementary data’, Journal of the Royal Statistical Society, Series A, vol. 124, pp. 412–20. Froment, R. and Lenclud, B. (1976) ‘Adjustement de tableaux statistiques’, Annales de l’INSEE, pp. 29–53, No. 22–23, Paris, avril-septembre 1976. Guiasu, S. (1979) ‘An entropic measure of connection and interdependence between the subsystems of a given large system’, Proceedings of the 3d formation symposium on mathematical methods for the analysis of large-scale systems, Academia, Prague. Jaynes, E.T. (1957) ‘Information theory and statistical mechanics’, Physical Review, Vol. 108, pp. 171–80. Lynch, RG (1986) ‘An assessment of the RAS Method for Updating Input-Output Tables’, in Readings in Input-Output Analysis: Theory and Applications (Ira Sohn, Ed.), pp. 271–94, Oxford University Press, New York. Mesnard, L. de (1988) Analyse de la structure interindustrielle française par filtrage biproportionnel. Thèse de Doctoral d’Etat es Sciences Economiques, Université de Paris I. —— (1990a) Dynamique de la structure interindustrielle française.Economica, Paris.

70

NEW MODELS AND NEW MODELLING METHODS

—— (1990b) ‘Biproportional Method for Analysing Interindustry Dynamics: the case of France’, Economic Systems Research, Vol. 2, No. 3, 1990. Ostblom, G. (1982) ‘Energy use and structural change’, Energy Economics, Vol. 4, No. 1, January 1982, pp. 21–8. Paelinck, J. and Waelbroeck, J. (1963) ‘Etude empirique sur l’évolution des coefficients input-output. Essai d’application de la procédure RAS de Cambridge au TES beige’, Economie Appliquée, No. 6. Snickars, F. and Weibull, J.W. (1977) ‘A minimum information principle theory and practice’, Regional science and urban economics, Vol. 7, pp. 137–68. Snower, D.J.(1990) ‘New Methods of Updating Input-Output Matrices’, Economic Systems Research, Vol. 2, No. 1, 1990, pp. 307–24. Thionet, P. (1976) ‘Construction and reconstruction de tableaux statistiques’, Annales de l’INSEE, pp. 5–27, No. 22–23, april-septembre 1976. Vermot-Desroches, B. (1986) Interdépendances spatiales and théorie moderne de l’information, Librairie de L’universite, Coll. de l’Institutee de Mathématiques Economiques, No. 30, Dijon. Watanabe, S. (1969) Knowing and guessing. A quantitative study of inference and information, Wiley, New York. Wilson, A.G. (1970) Entropy in urban and regional modelling, Pion Ltd, Monograph in spatial and environmental systems analysis, London.

Chapter 6 Gas contract portfolio management: experiments with a stochastic programming approach A.Haurie, Y.Smeers, G.Zaccour

Abstract This chapter presents the first experiments undertaken with a stochastic programming model of gas contract portfolio management. This model is intended to be the core of a decision support system for helping gas marketers of American or Canadian firms, facing the deregulated North American market, in their decisions to commit their reserves to different market segments. The data set used for these simulations is adapted from a previous paper in which the approach was based on an adaptation of the Markowitz model. The simulation results presented here permit a comparison between the open-loop and the closed-loop decision-making contexts. A closed-loop solution produces a policy adapted to the random changes of the state of the markets. This possibility to adapt is an important feature of the stochastic control model. Various perturbations of the base case run are considered for a first evaluation of the sensitivity of the model to variations in some key parameters. A discussion of the further developments needed for the construction of a decision-support system based on this methodology is also included. Finally the implementation of the model in the GAMS modelling framework is fully described. 6.I INTRODUCTION Since the advent of deregulation, a Canadian gas producer has to commit its production capacity (or reserves) to different market segments (e.g. long-term core vs spot market) which will generate, in the future, random netbacks that are not perfectly correlated. The gas marketer is thus faced with the difficult task of managing a portfolio of contracts so as to reach an efficient trade-off between expected return and risk. This situation is obviously reminiscent of the classical portfolio optimization problem in finance. In Haurie et al. (1990, 1991) a decision support system for managing a portfolio of gas contracts in the context of the American deregulated market has been formulated in two different configurations. In Haurie et al. (1990) an approach à la Markowitz (1985) has been proposed; the risk is modelled through the variance of the portfolio of contracts; the aim of the manager is to maximize the expected netbacks generated by the portfolio under a constraint on its variance. In order to compute the total netback variance the user has to provide the model with an estimate of the variance-covariance matrix of the random netbacks on different segments. The problem differs from a standard portfolio optimization problem since gas contracts can extend over very long periods (from one year to twenty years depending on the market segments) and this gives a dynamic structure to the problem. The illustration given in Haurie et al. (1990) used data provided by market research analysts of a leading gas company in Canada. These values are realistic; however, they are not

72

GAS CONTRACT PORTFOLIO MANAGEMENT

representative of a real life utilization of the model. In practice, the statistics concerning the variancecovariance matrix of the netbacks in different markets should come from a data-base of past observations or from a set of predictions. In a recent paper Konno and Yamazaki (1991) show the practical limitations of the Markowitz model or even its simplified variant involving ‘factors’ influencing the returns of different assets (here the market segments) (Sharpe 1963, Elton and Gruber 1987). These authors propose the use of a mean-absolute deviation criterion for measuring risk; this permits a reduction of the entire problem to a linear programming framework, eliminating the need to estimate the variance-covariance matrix. The approach proposed in Konno and Yamazaki (1991) is convincing and attractive for a classical (static) portfolio management problem. In the realm of gas contract portfolio management the dynamic structure poses a different challenge. Since the contracts may imply commitments for the distant future the problem has some features of a capacity expansion problem; the resemblance between the two types of problems can be even closer if the production capacity of the gas company can be increased through appropriate investment. For this class of problems the recent practice has been to use a scenario-based approach in conjunction with a stochastic programming formulation of the decision problem. This is particularly well illustrated in Eppenn et al. (1989), where a capacity expansion problem for the auto industry is presented. These considerations have led us to propose (in Haurie et al. 1991) a stochastic programming approach to gas contract portfolio management. The uncertainty concerning the future is organized in different scenarios which unfold according to an event tree. The advantage of the approach is threefold: 1 the scenarios can be built from market simulators, as e.g. the NARG model (Coad and Maerz 1989) or any similar model currently under development or in use in large gas companies; 2 the model is a linear programme which is amenable to a variety of large-scale optimization techniques that permits one to consider applying the method to real life situations (involving thousands of variables); 3 the approach permits a consideration of the so-called downside risk which only measures losses and not excess profits; this is important if the returns (netbacks) are not supposed to be normally distributed. The purpose of the chapter is to illustrate the use of the stochastic programming model. We shall still use imaginary data, since dealing with a complex market scenario generator as those in use in some gas companies would require the development of additional modelling capabilities as discussed in section 6.IV below. We hope that the illustration provided here will nevertheless convince decision-makers that the stochastic programming approach has great potential for providing insights on the efficient management of gas contracts. The chapter is organized as follows: in section 6.II we describe the event tree which will serve for the illustration and we discuss the structure of the portfolio management problem with a minimum of mathematics. In section 6.III we present the results of different runs and discuss the influence of different parameter settings. In section 6.IV we discuss the developments still needed to obtain a fully fledged decision-support system for gas contract portfolio management. In sections 6.V and 6.VI we recall the mathematical formulation of the closed-loop and open-loop models respectively. In section 6.VII we discuss the GAMS formulation of the model.

Figure 6.1 Event tree

NEW MODELS AND NEW MODELLING METHODS

73

GAS CONTRACT PORTFOLIO MANAGEMENT

Figure 6.2 Scenarios

74

NEW MODELS AND NEW MODELLING METHODS

75

6.II MODELLING UNCERTAINTY AND COMMITMENTS THROUGH SCENARIOS In this section we illustrate the scenario representation of uncertainty. The data we use are related to those already presented in Haurie et al. (1990); however, the probabilities and netback volatilities represented are somewhat arbitrary and have been chosen uniquely for illustration purposes. 6.II.1 An event tree representation of uncertainty In Figure 6.1 we have represented an event tree. Each node corresponds to a particular state of the market at a given time period. The source node is the present state. The tree extends over 30 time periods. A path from the source node to a terminal node, i.e. a node without successors, is called a scenario. In Figure 6.2 we represent the different scenarios which compose the event tree. Each scenario has a probability of occurrence. We can describe these probabilities on the arcs of the tree in which case, they are the conditional transition probabilities. Alternatively, we can specify the probability of each scenario. These two consistent descriptions of uncertainty are represented on Figures 6.1 and 6.2 respectively. In our model, detailed in Section 6.V, we represent the decision variables of the gas marketer at each node n as inm xnm In Kn

the new quantity of gas committed to segment m at node n the total quantity of gas committed to segment m at node n the physical investment (capacity expansion) undertaken at node n the total installed capacity at node n

These decision variables are linked through a set of constraints. The dynamic structure of the model comes from the fact that the current installed capacities and quantities of gas committed to different market segments are the result of past investments and contracts respectively. Each market segment m is characterized by a nominal length δm of the typical contract, (e.g. δ6=20 years for the cogeneration market segment); therefore a contract passed at time t in this segment will be counted in the total commitment for the next δ time periods. These typical contract lengths are listed in Table 6.1. The equations of the model detailed in Section 6.V show how this structure is represented in the model. The decision-support system will provide values for these decision variables in response to a specification of economic data and parameters. The key parameters of this model are ● the netbacks per unit of gas marketed on segment m at node n, denoted μnm; ● the upper bounds for the quantity of gas marketable on segment m at node n, denoted Other parameters and initial data will be discussed later on in this chapter, but these two sets of parameters convey most of the uncertainty plaguing this decision process. In practice, the definition of these parameter values will result from detailed scenario analysis, using the available market simulators and the long term forecasts on energy supplies and demands, transportation network expansions, industrial developments etc. In section 6.IV we discuss some possible developments permitting a relatively easy way to build such a data set

GAS CONTRACT PORTFOLIO MANAGEMENT

Figure 6.3 Base points

76

NEW MODELS AND NEW MODELLING METHODS

77

Table 6.1 Segments and contract length

Seg. 1 Seg. 2 Seg. 3 Seg. 4 Seg. 5 Seg. 6 Seg. 7

Segment

Contract length in years

spot fixed-price negotiated-price competing fuels gas to gas competition cogeneration long-term core

1 2 3 10 10 20 15

In Figure 6.3 we have represented the base point system which is used to generate the data set for our illustration. In order to avoid too lengthy a data introduction we have used a system of base points where we specify the parameters. These base points are such that all other nodes are on segments joining two base points. This permits us to obtain all other parameter values through linear interpolation. Only the base points are represented in Figure 6.3. Each base point refers to a part of Tables 6.3 and 6.4 where these economic parameters are given These data have been adapted from those used in Haurie et al. (1990). The other parameters used in the simulation runs are summarized in Table 6.2. 6.II.2 Closed-loop vs open-loop decision making A closed-loop solution corresponds to a decision-making context where the state of the gas market is known before any commitment. This corresponds also to the definition of a policy which tells what should be the action, given the observed state. In our model this is represented by the dependency of decision variables upon the nodes of the event tree. The stochastic programming approach has the great advantage of providing such a policy which could also be viewed as a contingency plan. By contrast the open-loop context is such that a complete stream of actions is taken, from the initial period up to the last one. The actions will only depend on the time periods and not anymore on the nodes. 6.II.3 Reserve and capacity commitments The total commitments are bounded above by an available capacity. This bound is modified endogenously through a capacity expansion programme. The capacity expansion modelling could either represent the evolution of Table 6.2 Other economic parameters Parameters

Value

Discount factor Threshold Maximum risk Low cost for capacity expansion

0.9 0 1000 5

78

GAS CONTRACT PORTFOLIO MANAGEMENT

Parameters

Value

Moderate cost for capacity expansion High cost for capacity expansion

10 20

Table 6.3 Base point netbacks Time period

Segment node

1

2

3

4

5

6

7

0 3 8 12 14 23 29 29 29 29 29 29

0 3 34 100 61 128 29 55 76 91 117 134

1.25 1.295 1.96 0.735 1.80 1.09125 1.59 2.5725 2.08125 1.7865 0.6075 1.09875

1.30 1.36 1.88 1.005 1.805 1.33125 1.69 2.3325 2.01125 1.8185 1.0475 1.36875

1.30 1.36 1.97 0.895 1.865 1.24875 1.69 2.5275 2.10875 1.8575 0.8525 1.27125

1.30 1.36 2.15 0.655 1.98 1.15625 1.69 2.6825 2.18625 1.8885 0.6975 1.19375

1.30 1.36 1.77 1.155 1.73 1.40625 1.69 2.1825 1.93625 1.7885 1.1975 1.44375

1.40 1.46 2.26 0.755 2.08 1.25625 1.79 2.7825 2.28625 1.9885 0.7975 1.29375

0.50 0.59 2.32 0.88 2.20 1.4175 1.895 2.88 2.3875 2.092 0.71 1.4025

Table 6.4 Base point bounds on total commitments Time period

Segment node

1

2

3

4

5

6

7

0 3 8 12 14 23 29 29 29 29 29 29 29

0 3 34 100 61 128 29 55 76 91 117 134 140

18.25 36.5 36.5 18.25 18.25 18.25 18.25 18.25 18.25 18.25 18.25 18.25 18.25

36.5 73.0 73.0 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75

36.5 73.0 73.0 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75

36.5 73.0 73.0 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75

36.5 73.0 73.0 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75 54.75

16.5 73.0 73.0 73.0 73.0 73.0 73.0 73.0 73.0 73.0 73.0 73.0 73.0

292 292 292 292 292 292 292 292 292 292 292 292 292

the reserve or the evolution of the delivery capacity. For the moment we refer to a loosely defined capacity which can be increased through an investment process.

NEW MODELS AND NEW MODELLING METHODS

79

6.III SIMULATION RUNS In this section we discuss the results obtained from the data described in previous section. We shall focus on the differences, for the decisions in the first early periods, between the closed-loop and the open-loop commitments under different data configurations. We shall only present the results pertaining to new commitments for the first five periods. In the closed-loop Table 6.5 Optimal value Optimal value Case

Case description

Closed loop

Open loop

Case 1 Case 2 Case 3 Case 4

Important past commitments moderate cost for capacity expansion Reduced past commitments moderate cost for capacity expansion Important past commitments high cost for capacity expansion Important past commitments low cost for capacity expansion

4966.13 5323.65 4690.55 5934.80

4835.47 5144.21 4579.78 5781.00

Table 6.6 Past commitments Period

Seg. 1

Seg. 2

Seg. 3

Seg. 4

Seg. 5

Seg. 6

Seg. 7

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

— — — — — — — — — — — — — — — — — — — — — — — —

3.65 — — — — — — — — — — — — — — — — — — — — — — —

3.65 3.65 — — — — — — — — — — — — — — — — — — — — — —

18.25 18.25 18.25 18.25 18.25 18.25 18.25 18.25 — — — — — — — — — — — — — — — —

14.60 14.60 14.60 14.60 14.60 14.60 14.60 14.60 14.60 14.60 — — — — — — — — — — — — — —

7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 7.30 — — — —

292.00 292.00 219.00 219.00 219.00 73.00 73.00 73.00 73.00 73.00 — — — — — — — — — — — — — —

80

GAS CONTRACT PORTFOLIO MANAGEMENT

Period

Seg. 1

Seg. 2

Seg. 3

Seg. 4

Seg. 5

Seg. 6

Seg. 7

25 26 27 28 29 30

— — — — — —

— — — — — —

— — — — — —

— — — — — —

— — — — — —

— — — — — —

— — — — — —

case this corresponds to the consideration of 0, 1, 2, 3, 4, 30, 92. The last three nodes (4, 30, 92) correspond to the first branching in the event tree. Table 6.5 shows the different scenarios that have been run and the differences in the objective function value. 6.III.1 Influence of past commitments Case 1: Important past commitments

In Table 6.6 we report the values of past commitments, i.e. commitments which result from contracts established before period 0 in the model. These values correspond to those already used in Haurie et al. (1990). They represent a situation in which an important part of available capacity is already committed at period 0, and in which these commitments will last a long time since most of them are in the long-term core segment. Of course this reduces the flexibility of choices. The simulation results, shown in Table 6.7, exhibit a noticeable difference in the new commitments for the open-loop vs closed-loop solutions. For Table 6.7 Results (selected) case 1: base case with past commitments Open-loop solution Period

New commitments i(t, m) Market segments m

0 1 2 3 4

Capacity K(t)

1

2

3

4

5

6

7

0 0 0 0 0

32.85 0.0 10.139 62.861 8.111

0.0 0.0 6.083 31.072 33.817

18.25 2.028 5.949 0.0 20.412

0.25 11.594 29.383 0.0 0.0

9.2 18.833 18.833 18.833 0.0

0 0 0 0 0

400.000 428.805 462.693 502.459 502.459

Closed-loop solution Node

0 1 2

New commitments i(n, m) Market segments m

Capacity K(n)

1

2

3

4

5

6

7

0 0 0

32.85 0.0 37.36

0.250 0.753 0.0

18.25 12.167 12.167

0.0 0.702 2.028

9.2 18.833 18.833

0 0 0

400.000 428.805 462.693

NEW MODELS AND NEW MODELLING METHODS

81

Closed-loop solution Node

New commitments i(n, m) Market segments m

3 4 30 92

Capacity K(n)

1

2

3

4

5

6

7

0 0 0 0

35.64 30.002 0.0 0.0

14.019 5.615 35.071 4.056

4.056 1.092 3.042 0.0

0.702 1.404 0.0 34.057

18.833 0.0 0.0 0.0

0 0 0 0

462.693 462.693 462.693 462.693

example, in market segment Number 5 the open-loop solution recommends an important commitment in period 2 whereas the closed-loop solution postpones the decision to period 4, where, depending on the event realization there is a large (node 92) or a small (node 4) or a null (node 30) commitment. Another interesting feature of this base case solution concerns the investment process for capacity expansion. Clearly, in the open-loop context, one increases the installed capacity more rapidly. Case 2: Reduced past commitments

In Table 6.8 we report the simulation results obtained if there are no past commitments. This situation permits an evaluation of the ‘cost’ of past decisions which are irreversible. It is noticeable that segment Number 7 loses much of its importance. The spot market (segment Number 1) is favoured. For this segment, the difference between open-loop and closed-loop commitment is important on periods 3 and 4. Table 6.8 Results (selected) case 2: no past commitments Open-loop solution Period

New commitments i(t, m) Market segments m

0 1 2 3 4

Capacity K(t)

1

2

3

4

5

6

7

18.25 24.333 30.417 30.417 34.472

36.5 12.167 48.667 24.333 46.639

36.5 12.167 12.167 48.667 10.139

36.5 12.167 12.167 0.0 0.0

36.5 12.167 12.167 0.0 0.0

16.5 18.833 18.833 18.833 0.0

72.083 0.0 0.0 0.0 0.0

400.000 428.000 462.000 443.167 443.167

Closed-loop solution Node

0 1 2 3 4 30 92

New commitments i(n, m) Market segments m

Capacity K(n)

1

2

3

4

5

6

7

18.25 24.333 30.417 18.25 10.763 6.083 22.306

36.5 12.167 48.667 24.333 47.965 48.667 46.639

36.5 12.167 12.167 48.667 11.465 12.167 10.139

36.5 12.167 12.167 0.0 4.446 9.125 0.0

36.5 12.167 12.167 0.0 4.446 3.042 0.0

16.5 18.833 18.833 18.833 0.0 0.0 0.0

42.583 0.0 0.0 0.0 0.0 0.0 0.0

400.0 400.0 400.0 401.5 401.5 401.5 401.5

82

GAS CONTRACT PORTFOLIO MANAGEMENT

6.III.2 Influence of capacity expansion The subsequent runs have been performed with some variations concerning the cost of capacity expansion. High cost of capacity expansion

When we double the investment cost (from 10/unit of capacity increase to 20/unit) there is no longer any capacity expansion. There are still important differences between the commitments of the open-loop and the closed-loop cases, in particular for market segments 3 and 4. Low cost of capacity expansion

When the investment cost is reduced to 5/unit, there is a much more important investment activity. Again the open-loop solution invests more rapidly than the closed-loop one. Also important differences in the commitments in segments 2 to 4 appear in the solution. Table 6.9 Results (selected) for high-capacity expansion costs Open-loop solution Period

New commitments i(t, m) Market segments m

0 1 2 3 4

Capacity K(t)

1

2

3

4

5

6

7

17.211 0.0 0.0 0.0 0.0

32.111 0.0 4.056 54.167 0.0

0.0 0.0 8.817 0.0 4.056

2.028 2.028 2.028 0.0 0.0

0.0 0.0 2.028 0.0 0.0

9.2 18.833 18.833 18.833 0.0

0 0 0 0 0

400 400 400 400 400

Closed-loop solution Node

0 1 2 3 4 30 92

New commitments i(n, m) Market segments m

Capacity K(n)

1

2

3

4

5

6

7

15.885 0.0 0.0 0.0 0.0 0.0 0.0

18.693 0.0 2.808 52.733 0.0 0.0 0.0

0.0 0.0 0.0 0.0 0.0 2.808 0.0

16.772 0.702 0.702 1.404 1.404 0.0 0.0

0.0 0.0 0.0 — 1.404 0.0 2.808

9.2 18.833 18.833 18.833 0.0 0.0 0.0

0 0 0 0 0 0 0

400 400 400 400 400 400 400

Table 6.10 Results (selected) for low-capacity expansion costs Open-loop solution Period

0

New commitments i(t, m) Market segments m

Capacity K(t)

1

2

3

4

5

6

7

0.0

16.472

32.85

2.028

0.0

9.2

0.0

400.00

NEW MODELS AND NEW MODELLING METHODS

83

Open-loop solution Period

New commitments i(t, m) Market segments m

1 2 3 4

Capacity K(t)

1

2

3

4

5

6

7

0.0 0.0 36.5 34.472

0.0 30.279 42.721 28.251

6.083 6.083 60.833 4.056

28.389 2.028 14.194 0.0

7.954 38.279 0.0 0.0

18.833 18.833 18.833 0.0

0.0 0.0 12.167 6.083

457.61 532.99 612.389 612.389

Closed-loop solution Node

0 1 2 3 4 30 92

New commitments i(n, m) Market segments m

Capacity K(n)

1

2

3

4

5

6

7

0.0 0.0 0.0 36.5 35.798 36.5 34.472

0.0 6.725 10.139 62.861 9.437 10.139 8.111

15.399 29.618 6.317 37.065 28.916 29.618 27.59

15.788 3.042 23.754 4.056 5.303 5.504 0.0

20.164 3.042 19.986 3.042 5.348 3.042 0.0

9.2 18.833 18.833 18.833 0.0 0.0 0.0

0.0 0.0 0.0 0.0 0.0 0.0 14.629

400.00 457.61 532.99 600.222 608.768 608.768 608.768

6.IV FUTURE DEVELOPMENTS In this section we discuss the additional elements still needed in order to develop a decision support system for the management of a portfolio of gas contracts. We organize this discussion along three topics dealing respectively with the linking of the model with a market simulator, the optimization techniques and the generation of reports. 6.IV.1 A Scenario generator A scenario is basically a description of the evolution paths of netbacks for different market segments and of the associated upper bounds for total commitments (portfolio proportions in market segments). The number of different scenarios can be extremely large if one decides to use a stochastic process representation of the unfolding of uncertainties. This is the case if, at each time period, a node has a given number of successors with some specified transition probabilities. A good interface between the model and the user should permit a rapid construction of a data set. This data set will be based on variations around some contrasted scenarios coming from comprehensive market simulations. For the construction of a data set one could use various techniques for representing the entire event tree in a convenient way. Some of these are listed below: ● graphic interface permitting a representation of nodes and arcs of the event tree; ● propagation rules permitting a recurrent or recursive construction of the event tree; ● input of a set of parameter values associated with a given mode, taken from the graphic display;

84

GAS CONTRACT PORTFOLIO MANAGEMENT

● propagation rules permitting a rapid input of data associated with whole subsets of nodes (e.g., the interpolation between base points as already explained in section 6.II). 6.IV.2 Optimizers In any real life application of this model, one has to expect a very large scale optimization problem. The event tree can be very rich if one takes into consideration not only the risk associated with the world price of oil, but also the uncertainty associated with transportation network capacity expansions. Recent advances in computer implementation of advanced optimization packages permit the desktop solution of quite sizeable linear programming problems. For bigger problems a decomposition technique, along the lines indicated in Rockafellar and Wets (1987) and Wets (1989) and, more recently Breton and El Hachem (1991), is implementable. 6.IV.3 A report generator The decision support system built around this model should have a complete report generation capability. It should be possible to select parts of the results and to display them either in tables or graphically. An essential feature of this system will also be to allow sensitivity analysis with respect to any parameter. 6.V THE STOCHASTIC PROGRAMMING MODEL In this section we recall the mathematical formulation of the stochastic programming model. This section is reproduced from Haurie et al. (1991). 6.V.1 List of variables and parameters N M

set of nodes set of market segments

T S π(n) a(n) ps qn xnm

planning horizon set of scenarios (complete path from root node to a terminal node) predecessor of node n N unique path of ancestors from root node to node n probability of scenario s probability of passing through node n total quantity of gas committed to segment m at node n upper bound to the quantity of gas committed to segment m at node n netback per unit of gas marketed on segment m at node n installed production capacity at node n physical investment in production capacity at node n investment cost at node n

μnm Kn In DC(Kn, In)

NEW MODELS AND NEW MODELLING METHODS

rn dn inm δm a(n, δm ) tn TENR EDSRt riskmax

85

net return at node n positive deficit with respect to threshold new quantity of gas committed to segment m at node n contract length in market segment m predecessor of ‘age’ less than δm of node n time period associated with node n discount factor at node n total expected net return over the planning horizon expected downside risk at period t threshold for downside risk at node n upperbound on downside risk at any node 6.V.2 Objective function

The gas producer wants to maximize the discounted stream of expected net returns over a long term planning horizon. (1)

6.V.3 Constraints The optimization is subject to the constraints outlined below. Expansion of production capacity

Gas production capacity can be increased through investment with a one-period lead time. (2) Total quantity of gas committed to market segments The total quantity of gas committed to a market segment at a given node is the sum of all contractual commitments established over a past history spanning a typical contract length. (3) Upper bound constraints The company management may want to maintain a diversification policy and so the total commitment to any market segment could be bounded from above. (4) Capacity constraints Total commitment to all market segments is bounded by the production capacity. (5) Net returns

86

GAS CONTRACT PORTFOLIO MANAGEMENT

Definition of the net returns. (6) Positive deficit with respect to threshold The deficit with respect to the threshold value , is the basic element of definition of the downside risk. (7) Expected downside risks The expected downside risk is to be defined at each period t. In order to compute it one considers the set {n|t(n)=t} of all nodes associated with the time period t. (8) Upper bound on expected downside risks The management wants to limit the expected downside risk at each time period. (9) Initial conditions The process starts with an initial capacity and a history of commitments already made for the different segments, given K0=K0 (10) and (11) 6.VI OPEN-LOOP APPROXIMATION In an open-loop formulation one commits all decision variables in future periods to values that will be identical for all nodes of a given period, and computed with the information available at the root node. We introduce the following new variables: x′tm i′tm

total quantity of gas committed at period t to segment m new quantity of gas committed at period t to segment m

The open-loop problem is thus defined as follows (12) such that (13) (14) (15) (16)

NEW MODELS AND NEW MODELLING METHODS

87

(17) (18) (19) (20) given K0=K0,

(21)

and (22) This problem is a restriction of the stochastic programming problem therefore its solution provides a lower bound for the optimal value of the objective function (Breton and El Hachem 1991). A better bound is obtained if one repeats sequentially the open-loop optimization procedure for the whole family of sub-problems, with root-node located at a given event-node n and with the decisions corresponding to the ancestors of that node ( a(n)) already specified by previous optimizations. This involves solving as many open-loop problems (of decreasing sizes) as there are branching nodes in the event tree. 6.VII CONCLUSION In this chapter we have shown that it is possible to implement a stochastic programming model for the management of a portfolio gas contracts over a long-term horizon. The implementation has been made possible by the use of a general purpose modelling system, GAMS, in association with a powerful optimizer, CPLEX, both running on a SUN-4 workstation. In this implementation we are already dealing with a quite large mathematical programming problem. The size of the problem is one more indication of the difficulty for the managers to get the proper insight about the tradeoffs between risk and returns in their management of gas contract portfolios. In a word, this is not a ‘toy problem’. The first experiments have exhibited an important difference between the open-loop and closed-loop solutions. This tends to indicate that the stochastic programming approach, through its ability to provide a decision rule or a policy instead of a single stream of decisions for the whole future, can give more insights to the decision makers than the simpler models which run on single scenarios. The model, presented in Elton and Gruber (1987) and in sections 6.V and 6.VI of the present chapter, is intended to be the core element of a decision support system which will permit a decision maker to run quickly a whole variety of simulations, dealing with different scenarios suggested by more comprehensive long-run market simulations. ACKNOWLEDGEMENTS We have benefitted from enlightening discussions with M.Breton and O.Janssens. D.Lavigne and E.Lessard have realized the programming of this model in the GAMS environment.

88

GAS CONTRACT PORTFOLIO MANAGEMENT

REFERENCES M.Breton and S.El Hachem, ‘Decomposition algorithm for stochastic dynamic programs, Cahier du Gerad, G-91-15, March 1991. L.A.Coad and D.H.Maerz, ‘Continental natural gas market Canadian export capacity in the 90s’, CERI, Calgary, Alberta, October 1989. E.J.Elton and M.J.Gruber, Modern Portfolio Theory and Investment Analysis, (3rd Ed.), John Wiley & Sons, Inc., New York, 1987. G.D.Eppenn, R.K.Martin and L.Shrage, ‘A scenario approach to capacity planning’, Operations Research, Vol. 37, pp. 517–27, July-August 1989. A.Haurie, Y.Smeers and G.Zaccour, ‘Toward a contract portfolio management model for a gas producing firm’, Cahier du GERAD, G-90-03, January 1990, to appear in INFOR. —— ‘Gas contract portfolio management: A stochastic Programming Approach’, in Advances in Operations Research in the Oil and Gas Industry, M.Breton and G.Zaccour, Eds., Editions Technip, 1991. H.Konno and H.Yamazaki, ‘Mean-absolute value deviation portfolio optimization model and its application to Tokyo stock market’, Management Science, Vol. 37, No. 5, May 1991, pp. 519–31. H.Markowitz, Mean-variance Analysis in Portfolio Choice and Capital Markets, Basil Blackwell, Oxford, 1985. R.T.Rockafellar and R.J-B.Wets, ‘Scenario and policy aggregation in optimization under uncertainty’, IIASA Working Paper WP-87-119, Laxenburg, Austria, December 1987. W.F.Sharpe, ‘A simplified model for portfolio analysis’, Management Science, Vol. 9, 1963, pp. 277–93. R.J-B.Wets, ‘The aggregation principle in scenario analysis and stochastic optimization’, mimeo, 1989.

Chapter 7 Modelling the European gas market: a comparison of several scenarios Jacques Percebois and François Valette

Abstract This chapter presents the SOSIE-GAZ model, which simulates the management of flows and stocks of natural gas in West Europe. The model is initially applied to the current gas market, elements of which are commented on. Then, as examples of the model’s application potentials, the occurrence of two types of scenario are envisaged. The first considers substantial disturbances, either in consumption (exceptionally cold or warm winters), or in supplies (rupture in the supply from a supplier’s source for technical or political reasons). The second scenario envisages longer-term changes in the market, resulting for example from a large-scale replacement of oil by gas, which is quite probable in the long term for environmental reasons, etc. Various producer and/or consumer strategies may be simulated, reacting to external disturbances or changes, taking account of the technical, financial and political constraints. These simulations may clarify actual options available now or in the future. 7.I INTRODUCTION Natural gas is currently the subject of considerable attention: in the present context of the recognition of environmental problems on a world scale, it is known that this fuel releases less CO2 than any other existing or foreseeable fuels available on a large scale. In addition, the recent discovery of large new deposits in the East makes it more certain that natural gas can make a secure contribution to Europe’s energy resources over the long term, provided, of course, that the choice of energy supply remains an open one. This new information reaches us at a crucial moment in political-economic terms, since the process of the opening of contact between the East and the West must find very specific items for trade if it is to be credible. In the absence of econometric or other models enabling us to proceed much further with these kinds of considerations, it has seemed to us that it would be useful to develop, for our research into energy savings and environmental questions, a tailor-made model which is simultaneously realistic on both the technical and economic levels, and designed on a European scale. The challenge on this model-creation is to be able to simulate the action of numerous strategies of nations, groups of nations, or producer associations or even of large purchasers, in the broad European context. The SOSIE-GAZ model precisely represents the European market for natural gas. It is a model, reproducing simulated monthly variations of supply and consumption, which enables us to evaluate the consequences of various disturbances or developments in the market, such as: a sudden interruption of the gas supply from one of the supplier sources; exceptional variations in the seasonal pattern of consumption; a spectacular development in natural gas consumption in the middle and long term, etc.

90

MODELLING THE EUROPEAN GAS MARKET

In the model, the function of the (annual) overall consumption of gas, an exogenous variable, is projected over a horizon of two, five or even ten years. This function may be estimated by econometric means. It can be subjected to short-term minor or major variations, reflecting, for example, the irregularities of certain types of demand such as those generated by caprices of the weather. Elements of the structure of the demand requirements for gas are known (in particular that industrial use is regular, but tertiary sector use fluctuates considerably). This is incorporated into the model as is the structure of supply (which is determined by long-term contracts, the degree of rigidity of which is an important factor). By using the currently available storage capacity, each country is able to achieve a permanent balance between supply and demand. We make the assumption that the producer countries have a limited capacity to increase exports and that the importer countries are supplied in a particular sequence. We also simulate the existence of a free market on which the exporters are able to sell their surpluses and the importers are able to find supplementary sources. Some of the interest in the model lies in the following issues: ● with respect to time, it brings out the difficulties of the reachievement of the supply-demand equilibrium, linked to forces of technical or contractual inertia; it is suggested in consequence that new types of contracts be negotiated. ● with respect to space, the model enables us to compare various strategies of obtaining supplies as functions of the capacities and proximities (in the broad sense) of potential suppliers. 7.II THE NATURAL GAS MARKET IN EUROPE 7.II.1 The relatively limited and variable share of gas as energy source in the EC The proportion of natural gas in the European Community’s total primary energy is less than its equivalent in the USA (19%, on average, against 25%). It is also very variable between one country and another: the share of natural gas in overall energy consumption varies between 0% (Portugal) to over 50% (Netherlands). Gas provides one third of the final consumption of energy in the residential and commercial sector in Europe, a quarter of the final consumption of industry and about 7% of the fuel requirements of the electric power stations. The geographical source of the gas consumed also varies strongly from one country to another (see Table 7.1): ● some countries obtain supplies basically or even exclusively from their own national deposits and/or cover the extra demand by imports originating from other countries in Western Europe (Netherlands, UK); ● other countries import the larger part of their gas from both Western and Eastern Europe (Germany, Belgium); and ● other countries obtain supplies not only from Western and Eastern Europe but also from countries outside Europe (France and, on a smaller scale, Spain). In total, the EC imports one third of its gas from outside the Community. Table 7.1 shows that the countries of the EC currently utilize four main sources of supply: the Netherlands, Norway, the former USSR and

NEW MODELS AND NEW MODELLING METHODS

91

Algeria. France is the only importer country which utilizes all these sources, at least in a significant fashion. The other importer countries obtain supplies generally from three sources at the most. 7.II.2 Long-term contracts Natural gas does not have any captive markets. It is not indispensable in any of its uses, unlike oil and electricity. Its penetration into the energy market Table 7.1 Main flows of natural gas in Europe Exporters Importers

Netherlands

Norway

Ex-USSR

Algeria

Belgium/Lux 4.66 2.03 — 3.68 France 4.33 5.56 8.38 8.96 Italy 5.65 — 11.43 11.16 Netherlands — 2.31 — — Spain — — — 2.71 Germany 18.8 8.22 21.1 — UK — 10.53 — 0.10 Total 33.44 28.65 40.91 26.61 Note: Figure in thousand million m3 (109 m3)=106 toe(toe=tonnes of oil equivalent).

Libya

Total

— — 0.30 — 1.15 — — 1.45

10.37 27.23 28.54 2.31 3.86 48.12 10.63 131.06

is therefore mainly dependant on its offered terms of sale, case by case. The sale price has to be sufficiently low for it to be competitive for the end user but must be sufficiently remunerative at the same time to cover its high cost of transportation. Below a given price floor, the exporter has no interest in selling it; above a given price, the importer has no interest in buying it—a successful transaction requires a compromise. The increase in environmental concern is likely to cause an increase in the share of gas in Europe, replacing coal and fuel oil, but there is no reason to expect a substantial change in the situation prior to the year 2000 or 2010. In the case of gas, supply contracts are relatively rigid. They are concluded for a long period (20 to 25 years) and provide for the delivery of a fixed quantity. ‘Take or pay’ clauses are generally contained in the contract, but this has little application in practice. A flexibility of the order of 10% to 20% around the contractual quantity is, in effect, introduced and, in the case of a serious problem, renegotiation of the contract is always possible. However, it is possible to imagine that a producer may be unable to meet his contractual commitments for political or technical reasons. It is also possible to envisage some substantial modification in European energy policy (the outlawing of nuclear power, an accelerated decline in coal) which may require the EC countries to require considerably more natural gas within a short period. In view of the rigidities found along the gas supply route-chain (both in transport and storage), what would be the consequences of an abrupt stoppage of deliveries from the ex-Soviet Union to West Europe (following, for example, a series of technical faults)?

92

MODELLING THE EUROPEAN GAS MARKET

7.II.3 A partially interconnected system The import side of the European natural gas market consists of a ‘club’ of several importer companies which know each other well and often negotiate side by side. They have, jointly, succeeded in providing overall security of supply by means of a system of substantially interconnected transnational gas pipelines. They have also succeeded in implementing the system of net-back (see below) within the framework of the purchase contracts. The gas pipeline system existing today has not completely connected up all the European markets: the UK, Ireland, Spain, Portugal and Greece are still not linked to central, continental Europe. The interconnection of the French and Spanish systems through a gas pipeline linking the deposits of Lacq and Serrablo is under research: this would enable Spain to import Norwegian (or even ex-USSR) gas via Belgium and France (or via Germany and France). In the future there will also be a pipeline linking the Algerian and Nigerian fields to the European network via Morocco. Table 7.2 Principal descriptors of consumption

Reserve s at 1.1. 90a Producti on in 1989a Share of gas in total energy consum ption in 1989 (%) Consum ption in 1989a Domesti c& Tertiary

Neth.

Norway

exUSSR

Algeria

Belg/ Lux France Italy

Spain Ger

UK

Libya

1725. 00

2295.00

52,000. 00

3250.00

—

30.00

317.00

22.00

188.00

560.00

1218.00

71.87

30.59

796.0

48.40

—

3.03

16.98

1.55

16.38

44.75

6.00

48.8

—

38.0

—

17

12

2.4

5.5

17.4

21.5

?

40.32

1.94

693.54

19.09

10.37

30.26

45.52

5.41

63.79

55.38

4.55

14.08

—

77.94

1.05

3.50

12.53

13.87

0.41

19.19

30.06

?

18.04

6.87

17.73

31.65

5.0

44.8

25.32

4.55

26.61

—

—

—

—

—

—

—

a

Industri 26.24 1.94 615.60 al & Producti ona Exports 33.44 28.65 40.91 Note: aFigures in cubic metres×109.

NEW MODELS AND NEW MODELLING METHODS

93

The European gas importers have been able to arrange that gas network in such a way that, after taking account of the applicable transport costs, the prices of all gases are virtually identical for the end user. This is called ‘reverse tolls’. Such an approach is logical for a fuel whose price is not the leading price on the market and which has to remain competitive with alternative products. The parity between the price of gas and that of competing oil products is calculated at the end user point (the net-back system). However, it is certainly possible that, in the future, the terms of purchase of gas may vary between one producer and another and that it will be relevant for consuming bodies to vary their supply policy. Another hypothesis is that a producer is unable to meet his commitments, for financial or technical reasons. 7.III THE SOSIE-GAZ MODEL 7.III.1 Context, constraints and objectives of the model At present there are only a small number of models representing or describing natural gas networks and markets, mostly developed by the institutions which have direct management concerns in this sector. These models can be classified as follows: 1 The first are models describing gas production and distribution networks, technically based. Their purpose is to aid the distributors to respond optimally to demand in all circumstances, particularly in the case of substantial fluctuations in this demand. The latter may be linked to weather/climate data (unexpected cold, or warm, spells) or to technical problems (leaks causing pipelines to be closed down; breakdowns or building work at gas holding stations). These models exist for all countries at the local, regional and national levels and are indispensable for designing and managing installations. Their objectives are purely technical and their economic content is small or non-existent. 2 The second are national models, describing the gas market from the viewpoint of economic and financial management, corresponding in particular to the concerns of the national level decision makers with regard to plant and installations and tariff policy. For our purposes they are inappropriate in that they are more detailed than necessary with regard to financial calculations, and unsuitable (because they are too tied into current data) to the simulation of substantial changes in structures and/or supply contracts. 3 The third are international or global models (which regard gas as a resource): these models forecast energy demand, or optimize the supply system. This type of model, with an econometric base, is more appropriate to short and medium term forecasting than to the time-spans under consideration here, and to our a priori more audacious conceptions in terms of possible changes in technical structures or contracts. These models, having a more complete technical basis, employ detailed information on plant and installations at the national level, but their interconnection with the corresponding sub-models is only partial, and they add little to scenarios which are truly ‘international’. It has been possible to use the general SOSIE software, which was developed by the Centre d’Ecotechniques to simulate the functioning of any complex structures, to serve in our context as the basis of the construction of a new model which enables us to tackle European gas questions and issues such as those mentioned above. A short presentation of SOSIE is supplied as an appendix below.

94

MODELLING THE EUROPEAN GAS MARKET

Figure 7.1 Gas flows represented by the SOSIE-GAZ model

7.III.2 Description of the model General description

The SOSIE-GAZ model describes the flow of trade of natural gas (see Figure 7.1) between: ● on the one hand, seven consumer countries or groups of consumer countries (France, Spain, Italy, Germany, Belgium and Luxembourg, Denmark and the Netherlands, the UK and Ireland); and ● on the other hand, six suppliers, either current ones (the Netherlands, Norway, ex-USSR, Algeria, Libya), or a potential one (for example a free market which might be organized by several countries). Each consumer country is described by its stock of gas, connected upstream to the suppliers’ reserves and downstream to the distribution networks linking to the final consumers (Figure 7.2). The current usable storage capacities amount to three or four months of consumption. The total capacity is approximately twice this size, but some is in the form of strategic reserves. Description of the structure (networks, plant and installations, institutions)

The SOSIE-GAZ model consists of 40 sub-systems, or ‘converters’ in the terms of the SOSIE programme, which, as a whole, represents Europe’s ‘gas system’ (see Figure 7.3): ee 1 Six suppliers of natural gas of which five are treated as reserves or stocks, corresponding to the countries currently exporting natural gas to Europe, and a free market (simulated in certain respects);

NEW MODELS AND NEW MODELLING METHODS

95

Figure 7.2 Sub-model of the component elements of a country’s gas supply

2 Seven gas stocks, associated with consumer countries or groups of consumer countries. These stocks are characterized (described) by their maximum capacity and by their minimum load (i.e. the minimum stock required by strategic constraints); 3 Seven modules (logic elements) of distribution, associated with these consumers, and a module of the same type corresponding to the fr market. In addition to the logic structure of the management of gas flows per se, all sorts of costs are taken into account (investment, maintenance, operating costs, possible leaks, losses of load, etc.) which may be associated with these modules; 4 Seven consumption countries or stations, generating the final demand for gas of the countries; 5 Seven logical operators (in the data-processing sense), describing the supply strategies of the countries; and 6 Finally, five global operators allowing logic data or instructions to be introduced which affect all the other elements of the system: management, developments, disturbances and forecasting. The last two of the above groups (both of operators) appear on the graph of the model (see Figure 7.3), but they are unlike any other senders or receivers of physical flows. Their sole purpose is to participate in the simulation of the trade in gas, by introducing the appropriate elements of logic, which are changeable from one scenario to another. Description of operation

For each time-step the SOSIE programme automatically deals with the information exchanged between the various converters which express, in particular, how much each can send or receive in the simulated period. The logic structure of supply applied in the basic version of the model is as follows:

Figure 7.3 Diagram of all the modules comprising SOSIE-GAZ Note: Only the flows corresponding to France are shown, to ensure the diagram is legible

96 MODELLING THE EUROPEAN GAS MARKET

NEW MODELS AND NEW MODELLING METHODS

97

Figure 7.4 Seasonal demand in each country, in principle

● demand is treated country by country, commencing with the satisfaction of the largest consumers and finishing with the smallest; ● for each consumer country, gas purchases are simulated starting with the largest suppliers and finishing with the smallest ones. It is apparent that this logic structure of supply is not necessarily the best, nor is it in conformity with reality. Rather, it has been established merely to provide a sequence for the calculations. It ispossible for it to be replaced from one scenario to another by others at will. These scenarios could respond, for example, to considerations of price, distance, political preference, etc., to study the effect of these variables on the financial quantities. The demand for gas is split into two components: industrial demand, assumed to be quasi-constant over the year, and tertiary demand, assumed basically to correspond to the heating of buildings as functions of the average monthly temperatures of each country (the ‘degree-day’ method). What is described here, for each country are realistic annual profiles of demand. Demand initially is actual current demand. Factors affecting demand may of course be modified at will, from one scenario to another, in particular to describe substantial changes in consumption over a long period (see Figures 7.4, 7.5). A noticeable contrast is observable between these demand profiles, according to whether the consumer countries are in the north or the south of Europe (this largely determines the climate) and according to the share of industrial demand in their total demand. The German, French and Italian profiles are contrasting in these two respects and, because of this and the large size of their consumption, it will be interesting to see how the scenarios are affected. 7.IV RESULTS With the options described above, the model may be used for the simulation of current trade with the storage capacity of each country acting as ‘blocks’. It may also be used to represent how the gas market will function following temporary disturbances, longer term changes or combinations of these. Below we will present, in the above order, various results obtained on these two types of hypothesis.

98

MODELLING THE EUROPEAN GAS MARKET

Figure 7.5 Seasonalized demand in the countries under consideration

7.IV.1 Simulation of current functioning As an exercise, the current functioning of the market in a stationary regime is simulated. We assume that the supply is in the form of constant monthly delivery contracts, equal to the average monthly consumption during the year. The corresponding results are shown in Figure 7.6, which indicates the variation in the quantities of gas in stock (curves 1, 2 and 3) and consumption (curves 4, 5 and 6) in Germany, France and Italy. It is noteworthy that stocks play an essential role in this hypothesis, allowing gas which is surplus to requirements in the summer months to be carried forward to the winter. The larger the share of tertiary consumption in total demand, the larger are these transfers. The security of supply appears to be plentiful in the sense that, beyond the ‘normal’ seasonal variations, the existing stocks seem sufficient to cope with the peaks in consumption reflecting exceptionally cold winters. Stock fluctuations have a maximum amplitude corresponding to about three months of average consumption in Italy and France, against 1.5 months in the case of Germany (in which industrial consumption is more predominant). 7.IV.2 Scenarios (simulation exercises) In view of the apparent robustness of the current functioning of the market, the scenarios envisaged correspond to rather ‘hard’ hypotheses, for only these are sufficient to ‘test’ the limits of the present equilibria.

NEW MODELS AND NEW MODELLING METHODS

99

Figure 7.6 Functioning of the market in a stationary regime Note: Scale represents cubic metres×109 Scenario 1: A rupture of the ex-USSR supplies for a six-month period

With the aid of this model it is possible to test a rupture in the supply coming from one of the suppliers. In the example which follows, the hypothesis is made that, for whatever reasons, the ex-USSR interrupts its deliveries to West Europe for six months. The simulated break in supplies takes place in late Spring, when the stocks are at their lowest levels (after the winter heating period). Problems of varying kinds then appear in the three importer countries represented (France, Germany, Italy) (see Figures 7.7, 7.8, 7.9): ● in Germany, the break in supply brings about a deficit from July to October, and then from December to April. Stocks are then reconstituted and demand can once again be fully met. But there is a danger that problems may reoccur in the following winters, to lessening degrees, which implies that Germany should purchase a little more gas in the following year if it wants to ensure it can cope with future winter peak demands. ● in France, the whole demand can be met in full from stocks during the period of the break, but a deficit appears in the following winter (between December and March). ● in Italy, the rupture of ex-USSR sales poses a problem from June to October, but the deficit is less than in Germany. However a larger deficit occurs during the winter, just as in Germany and France. The interpretation of these results is simple: if the logic element providing that management should satisfy consumers in order of size is not modified, Germany (which buys almost half of its gas from the ex-USSR) will see its stock of gas remaining at the minimal (i.e., strategic) level, instead of reconstituting it. Thus from the following winter it becomes impossible to meet demand there. In the following years, Germany can, however, rebuild its stock and thus return to the previous stationary regime. In contrast, France and Italy, served later in the logic of deliveries, have to reconstitute their reserves in the second year. Thereafter (slight) deficits occur in these countries during the following winters. These observations show the limits of the simplistic, simulated management logic structure.

100

MODELLING THE EUROPEAN GAS MARKET

Figure 7.7 Rupture of deliveries of ex-USSR gas for six months, without a free market (France, Germany, Italy) Note: Scale represents cubic metres×109

Figure 7.8 Rupture of deliveries of ex-USSR gas for six months, without a free market (Germany) Note: Scale represents cubic metres×109 Scenario 2: a doubling of demand over ten years

The hypothesis is made that demand increases at a rate of 7% per year during the coming ten years and one assumes at the same time that the producers are able to adapt themselves to requirements, with the exception of the Netherlands which cannot exceed its current production capacity (due to the relative exhaustion of its reserves).

NEW MODELS AND NEW MODELLING METHODS

101

Figure 7.9 Rupture fo deliveries of ex-USSR gas for six months, without a free market (France) Note: Scale represents cubic metres×109

Figure 7.10 Doubling of consumption over ten years, set of hypotheses A Note: Production potential capable of increasing by 10% through the free market, with a stock of 30×109 cubic metres, and with Netherlands production constant

We have compared two cases. In the first case the current stocking capacities are unchanged. Then, from year t+6 onwards, deficits appear to the extent that flexibility upwards, as well as downwards, is insufficient with respect to stocking (with the constraint of a strategic reserve of three months). The less than complete use of the stocking capacity thus becomes damaging. It is therefore necessary either to increase the physical stocking capacity, or to remove the constraint of strategic reserves. An alternative solution would be to find a profile of supply which follows variations in demand more faithfully. This in turn causes producers difficulties in adjusting supply and demand. The development of a free market could constitute a factor of flexibility in this respect (see Figure 7.10). In the second case the constraint of strategic reserves is removed, which has the effect of increasing the stocking capacity effectively useable. In this case the deficits disappear, but from year t+9 the situation

102

MODELLING THE EUROPEAN GAS MARKET

Figure 7.11 Doubling of consumption over ten years, set of hypotheses B Notes: a: 10% of production in the free market; release of strategic stocks b: Scale represents cubic metres×109

becomes critical since stocks are again insufficient and it is therefore necessary to establish supplementary stocking capacity or to resort to additional contracts, particularly to mitigate the deficiency in the Netherlands (see Figure 7.11). One of the interesting lessons observable in the model is the connection which exists between flexibility of contracts and stocking costs. In the long run, it will certainly be necessary to draw up more flexible contracts. In view of the substantial size of stocking costs, it will be preferable to choose more flexible contract formulas, even if it is necessary to compensate the producers for the extra charges which this flexibility may impose on them, by price increases. Also, in the long run one can foresee seasonal tariff systems, similar to those established in the electricity sector. 7.V CONCLUSIONS AND FUTURE DEVELOPMENTS The main original characteristics of the SOSIE-GAZ approach are: on the one hand, consideration of the entire European market, making it possible to have individual strategies of consumption and production; and, on the other hand, reproduction of the seasonal variations in demand by ‘chaining’ them over long periods. This model reveals the difficulties in adapting supply to demand in some situations, particularly in the case of the rupture of a part of supply and in the case of an increase in demand in a context where production is constrained. We therefore find ourselves forced to think about finding operational solutions to such difficulties: by creating a free market, by drawing up new contract formulas, and/or by revising the strategic criteria of reserve supplies, etc. The applications presented here are, no doubt, a weak subset of the capacity of the SOSIE-GAZ model. However, for this first presentation we can be content with verifying the technical equilibria, mainly linked to the describing of the quantities of gas traded. This could, without difficulty, be extended into the simulation of the associated financial flows, under various price hypotheses. Finally, the SOSIE programme, designed to manage several flows for each converter (in this case, country) would permit what is currently represented to be extended to the trade in electricity (introducing

NEW MODELS AND NEW MODELLING METHODS

103

the corresponding flows), and/or to other countries—for example those in the East (by introducing the corresponding converters). ACKNOWLEDGEMENTS The authors wish to thank M.Michel Jamet, Data Processing Engineer at the Centre d’Ecotechniques du CNRS and the developer of the SOSIE programme, for his decisive contribution to the construction of the model described here. REFERENCES H.Colas, ‘Modélisation intégrée bassin versant—activités humaines—milieux hydriques. Application au bassin versant du Lez des étangs palavasiens’. Doctoral thesis at the University of Montpellier II, 392 pp, Oct. 1991. M.Jamet, F.Valette, ‘Présentation d’un outil de simulation pour la gestion des ressources: Sosie 2’ (SOSIE II, Presenting a Simulation Tool for the Management of Resources). Centre d’Ecotechniques du CNRS, 55 pp, 1987. J.Percebois, ‘Gas market prospects and relationship with oil prices’, in Energy Policy, August 1986. —— ‘Le marché international du gaz naturel: contraintes et stratégies’ (The international Market of Natural Gas: Constraints and Strategies), in Energie Internationale 1987–1988, Editions Economica, Paris, 1987. —— ‘Economic de l’Energie’(Saving Energy), Editions Economica, Paris, 1989, 690 pages (See Chapter 7 ‘Le marché international du gaz naturel’, pp. 501–82). —— ‘Le marché européen du gaz naturel à l’aube du marché unique’ (The European Market for Natural Gas at the Dawn of the Single Market) in Energie Internationale. 1989–1990. Editions Economica. Paris. 1989. F.Valette, ‘Simulation et Optimisation de Systèmes Micro-énergétiques’ (Simulation and Optimisation of Micro-energy Systems), Doctoral thesis at the University of Paul Sabatier, Toulouse, 380 pp, June 1986. —— ‘L’ouverture européenne et ses besoins en méthodes: quelques exemples de prospective locale et régionale, autour du problème de la déprise des terres agricoles’ (The Opening up of Europe and Methodological Requirements: Various Local and Regional Examples Addressing the Problem of Set Aside of Agricultural Land), Communication to the Spanish-French-Italian Seminar on ‘Agricultures régionales, concurrences et politiques économiques’, Montpellier, 25–7 April 1988, INRA, 12 pp. —— ‘Présentation d’un nouvel outil de simulation: SOSIE 2’ (SOSIE II, Presenting a New Simulation Tool), Communication to the Franco-Vietnamese Seminar on Energy Planning, Training and Cooperation. Hanoi, Vietnam, 26–30 March 1990.

Appendix

The SOSIE (Simulation and Optimization of Integrated Systems) software was designed to facilitate the construction of simulation models of the operation of complex production structures. Its initial versions which focused on the optimization of multi-energy systems, were not easily convertible to other purposes. It was then written in a more general fashion, enabling it to be used to describe any system (modifiable at will, both in their structure and in their management), at the same time as real libraries of complex dataprocessing items were built up (sets of sub-programs, images, data, etc.) transportable without from one model to another. Functions such as management stocking operations, transfers or conversion of products, all subject to the same software treatment, may thus be described once for all, entered into libraries, and retrieved in the various models as often as necessary. The same thing may be done with the associations of converters, to simulate complex functions such as dams, networks, factories, etc. In this approach, any system is represented as an organization of several converters or sub-systems exchanging various flows between them. The system is considered to be fully described when: ● all its elements are placed in relationship to the others in space, and are given dimensions (either exogenously or endogenously). This requirement corresponds to the task of representation of the structure of the system. ● each flow of each type is described for each time-step on each trajectory that it can have in the system. This requirement corresponds to the task of representation of the functioning of the system. In practice, this very complete system is only easy to execute if the number of transformers and flows to be represented is high, in particular with respect to the functioning. An explicit description ‘of all’ is always possible (by specifying all the flow routes, without giving them any degree of freedom), but this is clumsy and may produce errors or omissions. The process may often be simplified by limiting it to the introduction of logic instructions of general application (such as hierarchies in the management of flows for each subsystem, or of circulation between the sub-systems), associated with a message service between sub-systems. A realistic management flow, and certainly a controllable one, is thus introduced. Numerous aides for the verification of the models, for following up the calculations and for expanding on the results are all integrated into this powerful tool. A broader diffusion of SOSIE is expected, going beyond the tests and improvements currently underway in various pilot-applications now under research.

Part II Application to particular energy policy problems

Chapter 8 World energy outlooks E.Lakis Vouyoukas

Abstract The IEA model is a medium-term global energy model. Its projections are based on econometrically estimated parameters, although adjustment factors are added where extraneous information is available. Under current baseline assumptions, the model projects a modest increase in OECD energy requirements to 2005 and significantly greater oil import dependence, especially from the Middle East. In non-OECD regions energy demand is projected to increase quite rapidly. Sensitivity analysis is carried out to examine the likely impact of stable oil prices and the imposition of carbon taxes. It is shown that even for modest emission reduction targets rather high taxes would be required over the medium term. 8.I MODEL STRUCTURE AND KEY ASSUMPTIONS 8.I.1 Key features of the model The IEA medium-term energy model has been developed over the past five years as an additional analytical tool for examining trends in energy markets and for carrying out sensitivity analysis.1 It has several key features: ● ● ● ● ●

It is an energy model, and it treats the macroeconomy exogenously. Its frequency is annual, currently running to 2005. Most of its parameters are econometrically estimated, usually over the 1965–89 period. Adjustment factors complement econometric forecasts in obtaining reference case projections. Its treatment of energy markets is especially detailed for the OECD regions, primarily because of the availability of high quality data. 8.I.2 Basic structure2

Figure 8.1 presents the basic structure of the model for each of the regions it covers. It can be seen that the model effectively consists of four interdependent sub-models and one relatively self-contained sub-model. The latter, the activity sub-model, effectively converts exogenous assumptions on GDP and population into

WORLD ENERGY OUTLOOKS

107

Figure 8.1 Model structure

the relevant activity variables for the sectors whose energy demand is endogenous within the model. Personal expenditure, industrial production by the iron and steel sector, and cars per head are examples of the output of this sub-model. A feedback effect from energy prices to GDP is possible and one has, in the past, been implemented between oil prices and GDP. However, the full endogenization of macro-economic variables is beyond the present design and objectives of the IEA model. Indeed, the IEA as an organization has traditionally relied on the OECD for macro-economic inputs. The other sub-models are strictly energy related. Demand is organized by end-use product, and the final demand sub-model solves for final energy demand on the basis of sector activity, the end user price and assumptions about other sector specific variables, such as saturation, technology etc. The model distinguishes three final demand sectors, namely, industry, the ‘other’ sector (a combination of residential and commercial demand), and the transport sector. The demand for final electricity, coal and gas is organized according to these sectors and, in the case of coal, a separate category for industrial coke is included. For final oil demand seven oil products are distinguished, namely, petrol, diesel, gasoil, kerosene (paraffinoil), heavy fuel oil, other products and bunkers. The demand for final electricity is then converted into primary fuel demand within the transformation or power generation sub-model given the structure of the electricity industry in each region, conversion efficiencies and assumptions about non-fossil fuels. Note that electricity generation from nuclear, hydro and brown coal is an exogenous input into the transformation sub-model. This is one of the reasons that the projection period of the model is limited to 2005. Power generation, together with a suitable aggregation of the output of the final energy demand sub-models, result in a set of primary fuel demands. The supply sub-model feeds off the output of the price sub-model and assumptions on reserves, discovery rates and other relevant variables, to produce a set of primary fuel supplies. These are then compared with the previously derived demands and, after allowing for trade, any surplus or shortfall feeds into the price sub-model. This takes the exogenous assumptions on crude oil price and inflation and solves for end-use and primary fuel prices by taking into account costs, any excess demand, and the competitive position of

108

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Figure 8.2 Crude oil price (in constant US$)

each product. The price sub-model is one of the most important elements of the model, as it captures one of the key aspects of energy markets, namely, the relatively low proportion of feedback from primary fuel prices to end user prices. Extreme cases of this phenomenon are petrol and electricity prices. Regionally, the model is divided into three OECD regions, North America, Europe and Pacific. The former CPEs (centrally planned economies) are divided into the CIS and Eastern Europe regions, while the LDCs (least developed countries) are divided into three continental regions plus the Middle East. China’s energy system is at present imposed on the model exogenously. The non-OECD regions are less detailed, as high quality disaggregated data is often non-existent or of very poor quality. 8.I.3 Oil price, GDP and other assumptions The reference case projections assume that crude oil prices in 1990 US$ settle at around $21 in 1992 and then gradually rise to about $35 by early in the next decade, remaining at that level thereafter (see Figure 8.2). To put this in context, a price of $35 per barrel is about equivalent to the average price that prevailed in the mid-1970s in real terms. This is significantly less than the 1980–1 peak, when prices were over $50 per barrel in constant 1990 terms, but higher than the pre-1973 level of about $10 per barrel. This oil price assumption, while reasonable, represents only one of a number of possible paths. It assumes that as excess production capacity outside the Gulf region is reduced, Middle Eastern production and capacity increase at a rate that will permit real price increases. An alternative crude oil price scenario, which is examined later on, assumes that crude oil prices, in real terms, average about $21 per barrel up to 2005. Economic activity over the period to 2005 is assumed to expand at an average rate of abut 2.7 per cent for the OECD, 3.1 per cent for the CIS and Eastern Europe, and 4.6 per cent for the developing countries.

WORLD ENERGY OUTLOOKS

109

Figure 8.3 World energy market shares (%)

A key feature of the reference case projections is the assumption that current energy and environmental policies will continue unchanged. Because of the obvious difficulties, no attempt was made to prejudge and incorporate into this outlook proposed policy measures, especially those aimed at slowing down or reducing the growth in energy-related greenhouse gas emissions. Many OECD governments are currently contemplating far-reaching changes to stem the growth of these emissions. This will be further discussed in the last section of the chapter. 8.II THE ENERGY OUTLOOK TO 20053 8.II.1 Major trends in energy demand Changing regional shares in energy consumption

World primary energy demand is projected to grow at an average rate of 2.3 per cent per annum over the next 15 years with wide divergences among the major regions (see Figure 8.3). Thus, while growth in the OECD is projected to be limited to 1.3 per cent per annum, total energy demand in developing countries is expected to grow at an average rate of 4.2 per cent per annum, and at a rate of 2.2 per cent per annum for Eastern Europe (EE) and the CIS. Consequently, developing countries are expected to account for a growing share of total world commercial energy consumption, and their share is expected to rise from 25 per cent now to about 34 per cent by 2005. The share of Eastern Europe and the CIS is expected to remain relatively constant at around 24 per cent, while that of OECD countries is expected to decline from 51 per cent to 43 per cent.

110

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

The projected rapid growth in energy demand among developing countries is due to a range of factors including the rapid growth in population and economic activity, to increasing industrialization and road transportation, and to the diminishing availability of non-commercial sources of energy. Developing countries consume significant quantities of non-commercial biomass fuel and the projections assume that incremental demand in developing countries will be met increasingly by commercial-type energy. Continued fall in energy intensity

Overall energy intensity in OECD countries (i.e. TPER per unit of real GDP) is expected to continue to decline, at an annual rate of about 1.3 per cent. This decline is due to continued technological advances, structural transformation of OECD economies toward less energy intensive sectors, and continued investments in energy. It also reflects the long-term price effects associated with the 1970s price increases through the continued roll over and replacement of old capital stock with newer, more energy efficient equipment. In non-OECD countries, by contrast, there has been no significant decline in overall energy intensity since 1973 (see Figure 8.4). In part this can be viewed as the result of population and economic pressures which, along with limited financial resources, tend to reduce the flexibility of developing countries and the Eastern European and former Soviet economies to take maximum advantage of improvements in energy production, distribution and consumption technologies. The projections for total energy demand for the developing countries as a whole imply only a very small or marginal decline in energy intensity for the period to 2005. A more pronounced decline is expected for Eastern Europe and the CIS. Consumer demand for more and better housing and energy-consuming durable goods such as household appliances and for more motor vehicles in those countries is likely to accelerate in the medium term. On the other hand, industrial reorientation away from energy-intensive sectors and above all price reforms (which invariably mean much higher real energy prices) should induce rationalization and energy savings in both Eastern Europe and the CIS. The net result according to this outlook is a decline in overall energy intensity of the order of 1.0 per cent per annum on average over the period to 2005, with most of the improvement occurring in the latter part of the 1990s and beyond. Changes in fuel shares

The share of oil in OECD countries is projected to decline from about 43 per cent in 1989 to 39 per cent in 2005, from 28 per cent to 26 per cent in Table 8.1World primary energy requirementsa (million tonnes of oil equivalent) Region OECDc

Total Solid fuels Oil Natural gas Nuclear energy Hydroelectricity and other CIS and Eastern Europed Solid fuels Oil Natural gas

1989b

1995

2000

2005

4134 970 1785 769 352 258 1929 594 537 662

4549 1022 1930 920 393 284 2188 602 616 818

4777 1127 1920 1016 407 307 2427 607 665 969

5090 1254 1974 1102 424 336 2740 642 710 1168

WORLD ENERGY OUTLOOKS

Region

1989b

1995

Nuclear energy 66 77 Hydroelectricity and other 70 76 Developing countries 2100 2725 Solid fuels 790 991 Oil 862 1107 Natural gas 237 351 Nuclear energy 24 31 Hydroelectricity and other 185 245 World total 8727 10041 Solid fuels 2354 2615 Oil 3183 3653 Natural gas 1669 2089 Nuclear energy 443 501 Hydroelectricity and other 514 604 Biomass (excl. OECD) 564 579 Notes: aNon-commercial fuels in non-OECD economies not included. bPreliminary for non-OECD countries. cExcluding eastern Germany. dIncluding eastern Germany.

2000

2005

103 82 3348 1199 1317 484 38 310 11140 2933 3902 2470 549 699 587

130 89 4073 1468 1462 713 38 393 12498 3363 4147 2983 593 818 594

111

Figure 8.4 Energy intensity (1975=100)

eastern Europe and the CIS, and from 41 per cent to 36 per cent in developing countries (see Figure 8.5). The bulk of the incremental oil demand growth is projected to originate within the transportation sector, especially in the OECD. The share of natural gas in total world energy requirements is projected to rise by

112

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Figure 8.5 World energy shares (%)

about 5 percentage points, from 19 per cent to 24 per cent, with the largest increase occurring in the developing countries (from 11 per cent to 18 per cent), and eastern Europe and the CIS (from 34 per cent to 43 per cent). An increase of only 2 percentage points in market share is expected in OECD countries. The projected growth in the share of gas is primarily due to the strength of the projected demand growth by the power sector. Recent technological advances, especially in gas turbines, have improved the economics of gas as a power generation fuel and the implied higher efficiency has increased the environmental appeal of gas.4 The contribution of solid fuels to total world energy requirements is expected to remain fairly constant at about 27 per cent, although with regional differences (the increased market share of coal in the OECD countries is expected to be offset by a lower share in non-OECD countries). On a world basis, the shares of nuclear and hydro are expected to remain relatively constant and satisfy a relatively small percentage of world energy requirements. 8.II.2 Oil supply outlook The oil production pattern over the next 15 years is expected to reflect closer the geographical distribution of reserves, with the Middle Eastern region accounting for an increasing share of world supplies (see Figure 8.6). The bulk of the 3.5 mbd OECD decline is expected to occur in the US, where oil production is projected to be little over 6 mbd in 2005, almost a third below current levels. Developing countries outside the prolific Middle Eastern region are rarely fully explored despite the fact that they include some of the most promising geological provinces. This has been, to a large extent, due to structural and political problems and to the frequent lack of resources to finance oil exploration and production. Favourable changes in recent years are reflected in the base case projection, which forecasts a 7 mbd increase in oil production by nonMiddle East developing countries in the period to 2005. Oil production in the CIS is expected to decline in

WORLD ENERGY OUTLOOKS

113

Figure 8.6 World oil supply by region (MBD)

the short to medium term recovering slowly towards the end of the decade but still failing to reach the 1980s peak even by 2005. The outlook for CIS oil over the next few years will depend to a large extent on two factors, namely, how much the oil industry and its suppliers are affected by the poor general economic conditions in the country and the resources that are effectively allocated to the sector. Total oil production outside the Middle Eastern region then is projected to rise by just over 3 mbd over the next 15 years, compared to 19 mbd increase in world oil demand. Consequently, the Middle Eastern region will be called upon to provide the bulk of the incremental oil demand or nearly double its level of production from current levels. There is little doubt about the geological potential of the region. Saudi Arabia, Kuwait and UAE alone, three countries with a population of less than 20 million, claim established reserves of more than 450 bn b or three times as much as the whole of OECD, CIS and China combined. The region as a whole accounts for nearly two-thirds of world proven reserves. While some current official reserve estimates may appear too high, it is likely that, with sufficient exploration, even higher reserves could be quickly established. One of the key issues then for the oil market outlook over the next 15 years is the extent to which these countries will be able and/or willing to expand their production capacity to meet the increase in world oil demand that was mentioned above. A very important outcome of the reference case is the projected increase in OECD oil import dependence from about 59 per cent to 70 per cent in 2005. This is not far from the level of import dependence in the early 1970s.

114

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

8.III SENSITIVITY ANALYSIS 8.III.1 Lower oil prices Table 8.2 illustrates the impact on world oil demand of the alternate crude oil price scenario of $21 per barrel constant to 2005. The comparison is made in terms of the difference or deviation in per cent from the $35 per barrel case. At the lower crude oil price of $21 per barrel, world oil consumption would be about 1 per cent higher in 1995 and about 15 per cent higher in 2005. As might be expected, oil consumption in the non-OECD regions tends to be less sensitive to price changes than OECD oil demand. The nature of the demand for oil in these regions is often much less elastic to price because it is used to satisfy basic needs and tends to be much less discretionary. Also, for structural reasons, it tends to be less amenable to inter-fuel substitution than in OECD countries. Moreover, non-OECD regions include both oil importing and oil exporting countries. While oil importing countries tend to benefit from lower oil prices and consume more oil, the converse is true for oil exporting countries which suffer an income decline as a result of lower prices for their main export commodity. Hence, because of these opposing trends, oil demand in these countries as a whole tends to be less sensitive to price changes than oil demand in the OECD region which on the whole is an oil import dependent area. Also, many of these countries tend to have administrative pricing regimes which do not fully reflect and/or respond to changes in world oil prices. As a result, consumer prices are often not fully adjusted to crude oil price changes and, therefore, demand is likely to be less price responsive. 8.III.2 Carbon taxes In recent years the international debate on possible global climate change, and what actions nations could take to avert it, has been gathering momentum. While any action is likely to consist of a combination of policies, Table 8.2 Deviation of $21 per barrel crude-oil price case from the $35 per barrel price case (1990 prices) Percentage difference

Oil consumption (average world) OECD Eastern Europe and CIS Developing countries

1995

2005

1.2 2.0 0.0 0.0

14.7 20.7 7.1 10.2

the measure that has been analysed and debated the most is a carbon tax or a charge on fuels according to their carbon content potential for producing carbon dioxide. Table 8.4 presents the results of two such cases, namely, a carbon tax of $65 and $US130 per tonne of carbon in all OECD regions. These taxes are very substantial and are equivalent to $8 and $16 per barrel of oil, $45 and $90 per tonne of coal and $1.00 and $2.00 per MBtu of natural gas.

WORLD ENERGY OUTLOOKS

115

As can be seen from Table 8.3, emissions of carbon in 2005 are about 8 per cent lower than the reference scenario for the $65 per tonne of carbon scenario while the $130 tax produces a deviation of about 13 per cent. However, even the $130 tax scenario would leave total OECD emissions around 7 per cent higher than present levels in 2005. Since the tax is applied differentially according to carbon content, and since coal is the most carbon-intensive primary fuel, solid fuel consumption is most severely affected. The next most affected fuel is petroleum. The impact on natural gas consumption is very small, and in fact slightly positive, despite the fact that a very substantial tax is applied on natural gas prices too. This is due to substitution away from coal and oil as the price competitiveness of natural gas is enhanced due to the tax differential. This substitution effect accounts for 30 per cent reduction in CO2 emissions resulting from the tax, the rest being due to the reduction in total primary energy demand. The relatively modest impact on energy demand and emissions of such high taxes is due, to a very large extent, on the small feedback from increases in primary fuel prices to final end-user prices. This is because end-use prices are often much higher due to high taxation or a very large non-fuel component. For example, a $65 tax would increase the price of crude oil by almost a quarter in 2005, within the model; it increases the price of petrol by less than 5 per cent in Europe and by just over 12 per cent in North America. Similarly, the price of domestic electricity increases by less than 6 per cent in Europe and by less than 10 per cent in the Pacific region. While some end-user prices, like that of coal, rise Table 8.3 Impact in 2005 of an OECD-wide carbon tax on OECD energy consumption and emissions (% deviation from reference case levels) Tax per tonne of carbon

Solids Petroleum Natural gas Total primary energy consumption CO2 emissions

$65

$130

−17.2 −3.0 1.0 −5.2 −7.5

−26.1 −7.0 0.1 −9.1 −12.7

significantly faster, the overall effect of high carbon taxes is much less pronounced on end use prices.5 Notes 1 The IEA model, in its present form, is mostly due to the work of Niko Kouvaritakis with the help of Jonathan Coppel. 2 For a more detailed description of the model the interested reader should consult ‘The IEA Medium Term Energy Model Presentation’ —an IEA document available on request. 3 Further details of the Energy Outlook presented here can be found in ‘Energy Policies of EEA Countries: 1990 Review’, Paris, 1991. 4 The IEA has recently completed a major study on gas which contains a large amount of regional detail on gas supply and demand trends. See ‘Natural Gas Prospects and Policies’, Paris, 1991. 5 For more details on the application and effectiveness of carbon taxes within the IEA model, see a recent paper by the author entitled ‘Carbon Taxes and CO2 Emissions Targets: Results from the IEA Model’.

Chapter 9 Energy shocks and the demand for energy P.S.Andersen and H.J.Bernard

Since 1945 the supply of oil has frequently been threatened by conflict. In 1951 Mohammed Mossadegh nationalised BP’s Iranian holdings. In 1956 the Suez Canal was blocked. In 1967 Israel attacked Egypt’s massing armies, and the Arabs tried to embargo supplies to America. Another Arab-Israeli conflict caused the ‘first’ oil shock in 1973, and the Iranian revolution led to the second in 1979. On each occasion the oil price jumped and the world economy shuddered. (The Economist, 12 January 1991) 9.I INTRODUCTION1 The Persian Gulf crisis and the resulting surge in the price of oil again raised the question of how energydependent developed economies are, especially at a time when a slowdown in economic activity (in some instances a recession) is being recorded. Indicators such as consumption of energy per unit of GDP show a general tendency for the energy intensity of output to be lower today than it was ten or fifteen years ago. Moreover, other sources of energy have helped to reduce the share of energy use met by oil as well as industrial countries’ reliance on energy imports, allowing their production systems to adjust more flexibly in periods of crisis. None the less, an increase in the price of oil still represents a deterioration in the terms of trade for most countries and risks sparking off a price-wage spiral unless appropriate policies are introduced. 9.II ENERGY DEMAND ELASTICITIES IN THE GROUP OF SEVEN COUNTRIES This section attempts to estimate the energy demand equations for the G-7 countries using statistical series supplied by the International Energy Agency and distinguishing between energy demand in three sectors: industry, commerce and households and transportation. In the early 1980s Mittelstaedt (1983) found that the long-term price elasticity of industrial demand for energy was −0.40 in the major seven countries and −0.62 in seven smaller OECD countries, with an adjustment lag which could be as long as seven years. In the IE A World Energy Outlook (October 1982) a long-term price elasticity of −0.65 was used. However, these estimates were based on observations confined to the period 1960–78, and since the last oil shock energy prices have fallen (even more so in relative terms) which might provide new evidence on these relationships. In fact, the estimates reported below tend to indicate that energy needs might be less price-sensitive than was estimated in the early 1980s but that adjustment lags might be shorter.

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

117

The econometric method applied in this section draws on the theory of co-integration, which, under certain conditions, allows a distinction between long-run relations and short-term adjustment. The basic idea behind co-integration is to search for a linear combination of individually non-stationary time series (selected on the basis of economic theory) that is itself stationary.2 In testing for co-integration we shall rely on three tests: The Durbin-Watson (DW) statistic, which gives a first but not very robust indication; the Dickey-Fuller test (DF) or augmented Dickey-Fuller test (ADF); and the sign and significance of the coefficient of the error-correction term in the adjustment equation.3 9.II.1 Energy demand by industry Long-run equations

The energy demand equation resulting from the theoretical discussion is non-linear and can only be estimated using a grid-search procedure. Data availability constitutes a further problem, and for the estimates to be discussed below a simpler specification was chosen. Thus, real value added was entered with a freely estimated coefficient and the energy price was only deflated by the GDP deflator. Moreover, gross fixed investment (measured as a percentage of GDP) was introduced as a separate variable and not linked to the relative energy price as a strict putty-clay model would require. Finally, to capture retrofitting as well as energy saving measures, which are difficult to quantify, all equations for industrial energy demand were estimated including a simple time trend with an expected negative coefficient.4 With these simplifications the long-run energy demand equation estimated for industry can be written as: (1) where the following apply: E Y Pe I t

energy consumed by industry, measured in millions of tons of oil equivalent total value added in constant prices (GDP)5 wholesale price of energy relative to the GDP deflator real non-residential fixed investment relative to GDP time trend

and all variables are expressed in log levels, so that the coefficients represent elasticities. Table 9.1 summarizes the main result which in all countries show that the long-term demand for energy in the industrial sector is dominated by output.6 However, there are wide differences in the estimated elasticities, reflecting variations in the share of industry in total output and in the cyclical variability of industrial output relative to aggregate GDP as well as in the energy dependence of industry across the seven countries. In the United States and Germany a 1% growth in GDP generates a rise in energy consumption of about 1¾%, while the output elasticity is 2–2½ in Japan and France and is close to or above 3 in the United Kingdom and Italy. Only Canada shows an output elasticity of unity. A negative time trend has been identified in all countries except Canada and might reflect energy-saving measures (which are not directly included in the specification) combined with shifts in the composition of output and demand towards less energy-intensive sectors and products. However, an

118

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Table 9.1 Demand for energy by industry (in log levels, annual data, 1960–88) Country

Constant

Real GDP

Prices

Nonresidential investment

Time trend

R2 DW

DF ADF

United States

−2.54

1.73

−0.18

0.26

−0.04

0.75

−2.50

(−1.8) −0.34 (−0.05) −1.76 (−2.1) −1.86 (−2.3) −6.61 (−7.2) −4.60

(4.0) 2.48 (22.4) 1.79 (11.6) 2.02 (14.7) 3.25 (14.8) 2.94

(3.3)−3 −0.21 (−5.8)−1 −0.18 (−2.4)−2 −0.32 (−4.0)−1 −0.31 (−4.5) −0.54

(1.1) −0.64 (−3.8) — — — — — — —

(−3.6) −0.10 (−17.2) −0.04 (−8.7) −0.05 (−11.5) −0.09 (−10.8) −0.07

0.91 0.99 2.00 0.91 1.27 0.91 1.09 0.96 0.85 0.88

— −4.91 — −3.38 –5.26 −2.96 −3.40 −2.69 −3.85 −2.69

(−2.3) 0.51 (3.9)

(7.3) 0.99 (22.7)

(−3.5)−3 −0.10 (−2.3)

— — —

(−8.1) — —

0.86 0.97 0.72

— −2.02 —

Japan Germany France Italy United Kingdom Canada

Notes: The critical values for the co-integration test (DF/ADF) as reported by Engle and Yoo for three variables at the 5% and 10% significance levels for a sample size of fifty are 4.11 and 3.73 respectively. t-Statistics are given in brackets with lags indicated as superscripts. For Canada, adding a time trend (coefficient: −0.04, t-statistic: −7.2) changes the price coefficient to 0.05 (t-statistic: 1. 5).

alternative interpretation is also possible and mitigates some problems concerning the size of the estimated output elasticities. Recalling the nature of the estimated relation, the output elasticities mean that, in the long run, a 1% rise in GDP is accompanied by a 2–3% rise in energy consumption. However, because GDP and the trend are highly correlated, the negative trend coefficients are likely to reflect the net effect of energy-saving measures and trend changes in output. In such a case, with the GDP trend being largely captured by the time trend, the output elasticities should be seen in relation to variations in de-trended output and as being closer to the short-term than to the long-term adjustment coefficient of energy demand. The problem may be solved by introducing and estimating an equation for trend GDP (Y*): Y*=f+gt (2) and substituting: (3) with Ŷ representing the deviation of actual from trend GDP. Equation (3) is no more than a linear transformation of (1) and when the coefficient to t was calculated using estimated figures for trend GDP growth (g), values much closer to 1 than those presented in Table 9.1 were found for five of the countries.7 Hence, when GDP is moving along its long-run growth path, there is almost a 1:1 relationship between industrial energy use and output. By contrast, for deviations from the long-run growth path, the much higher coefficients presented in Table 9.1 apply. Thus an increase in production to

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

119

above the long-run trend can be extremely expensive in terms of energy needs as factories or production facilities made obsolete by earlier changes in energy prices are put into operation. By the same token, and obviously important in periods of rising energy prices, short-run reductions in output can very quickly and effectively reduce energy demand. The response of energy demand to changes in the wholesale price of energy is of the correct sign in all countries, but again of differing magnitudes. In the United Kingdom a 1% increase in energy prices is met by a decline of one-half of one per cent in energy consumption, whereas in the United States, Japan, Germany, France and Italy, the response amounts to about one-fifth of one per cent and in Canada to only one-tenth.8 Despite the long-run nature of the equations, lags were necessary in some cases as substitution is extremely slow and is realized only as new equipment is put into place. In addition, in several countries administered energy prices have prevented the full and/or immediate adjustment of domestic oil prices to international levels, which may have reduced firms’ incentive to cut energy demand. Adding the ratio of private or total non-residential fixed investment to GDP9 to the equations does not provide much extra explanatory power. As noted, the process whereby energy use is replaced by other factors frequently requires changes in the capital stock, so that if total investment is dominated by energysaving investment the coefficient could be expected to be negative. In Japan, which imports almost all of its energy requirements, the equation shows a strong negative influence of non-residential investment, underlining major efforts undertaken by Japanese enterprises towards reducing energy dependence through capital/energy substitution and a shift towards lighter and less energy-intensive industries. In other countries, however, the coefficient was either positive or insignificant, probably reflecting the ad hoc nature of the specification as well as the fact that in many countries the gradual rise in the capital intensity of overall output has been accompanied by higher and not lower demand for energy.10 Short-run adjustments

The equations reported in Table 9.1 only attempt to identify long-run relationships and do not allow for cyclical fluctuations around the long-term trend. Moreover, as a further step in identifying stable relations, an error-correction equation is required to determine whether deviations from the long-term trend are being corrected, and (when convergence is found) at what speed. Table 9.2 Short-term equations of energy demand by industry (variables expressed as changes in log levels) Country

Constant

Real GDP

NRFI

Prices

Error-correction term*

R2 DW

United States

−0.05 (−4.7) −0.08 (−5.3) −0.06 (−6.8) −0.06 (−3.5) −0.08 (−5.9) −0.06

1.99 (7.0) 2.14 (7.9) 2.41 (9.5) 2.19 (4.5) 2.85 (9.2) 2.19

— — −0.40 (−1.7) −0.28 (−2.4) — — — — —

— — −0.20 (−4.8)−1 — — −0.30 (−2.9)−1 −0.09 (−1.8) —

−0.45 (−2.6) −0.94 (−4.3) −0.63 (−3.5)−1 −0.55 (−2.8) −0.48 (−3.9)−1 −0.33

0.72 2.12 0.87 1.83 0.82 2.62 0.68 1.69 0.76 1.88 0.53

Japan Germany France Italy United Kingdom

120

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Country

Constant

Real GDP

NRFI

Prices

Error-correction term*

R2 DW

Canada

(−4.4) −0.02 (−1.4)

(5.4) 1.45 (4.0)

— — —

— — —

(−1.9) −0.40 (−2.5)

2.14 0.55 2.02

Notes: Lags are reported as superscripts to t-statistics. * Lagged residual from main equation reported in Table 9.1.

Consequently, as a second step, the following equations were estimated: (4) where d denotes the first difference operator and ECM–1 is the lagged residual of the equations estimated in levels. Table 9.2 summarizes the results of the short-term equations and shows in all cases a significant error-correction coefficient of the correct sign (negative), with a large proportion (between 30% and 90%) of the error that appeared in the long-term equation being corrected within two years and in some cases within one year. Wholesale price changes influence the short-term demand for energy with a one-year lag in Japan, France and Italy, whereas the coefficients are not meaningful for the other four countries. Moreover, in both Japan and Italy the short-run elasticities are well below the long-run coefficients, so that in all cases the scope for short-run substitution of energy appears very limited. The dominating short-run influence is output and the coefficients are for most countries close to those found in the long-term equations, thus giving some support to the alternative interpretation discussed above. Non-residential investment has only a limited influence, with only Germany and Japan producing significant coefficients of the expected sign. 9.II.2 Demand for energy in the commercial and residential sectors Long-run equations

In deriving the demand equation for energy use in the commercial and residential sectors we have made the assumption that consumers attempt to maximize utility subject to overall income and wealth constraints. For a broad set of utility functions, this assumption generates an equation whereby aggregate consumption is a function of real disposable income and real wealth, whereas the demand for a particular commodity i depends on aggregate consumption and the price of i relative to the aggregate consumption deflator. When applying this model to the demand for energy some additional modifications were made, partly for theoretical reasons and partly reflecting the data available: 1 since the series on households’ use of energy are rather short, it was necessary to combine them with energy demand in the commercial sector. As a result (but also in the interests of obtaining elasticities comparable with those derived for the industrial sector), real GDP rather than real consumption was used as the activity variable, while the price of energy (retail prices) was deflated by the GDP rather than the consumption deflator; 2 since energy is normally used as an input to goods more directly serving consumer needs (heating installations; electrical equipment, etc.), the stocks of durable goods and single and multi-family houses may be more relevant than aggregate wealth. However, both variables are subject to the problem that while a rise in stocks of a given quality can be expected to boost energy demand, energy-conserving

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

121

measures incorporated in either additions to the stock or in replacements will have a negative effect, thus leaving the net effect undetermined. Moreover, data series on the stock of durable goods are for some countries either short or non-existent and probably capture quality improvements to a much smaller extent than those on residential construction. As a compromise, we have adopted the same procedure as for the industrial sector and included residential investment relative to GDP as a separate variable; 3 a number of other variables which might capture specific features of energy demand (such as the rate of unemployment and changes in the overall output share of the services and commercial sectors) were also tested, but in most cases did not yield any significant results. Given the compromises dictated by the aggregate nature of the data used and the relatively ‘stable’ development of energy demand in the residential and commercial sectors, identification of relevant parameters is likely to be more difficult than for the industrial sector. Generally, however, because substitution possibilities are less readily available and the cyclical sensitivity is much lower than in industry,11 both price and income elasticities may be expected to be smaller than those reported in Tables 9.1 and 9.2. As shown in Table 9.3, real GDP is also the major determinant of energy demand in the residential and commercial sectors, though again with very substantial differences across countries: the coefficients range from 0.2 in the United Kingdom to 1.6 in Italy, with an average elasticity of 0.9, or less than half of that shown for industry (2.0). There appears to be an inverse relationship between the size of the intercept terms and the estimated income elasticities, possibly suggesting that the former reflect basic or incompressible levels of demand. The table also reveals a rather weak price response of energy demand. The price12 elasticity is one-half in Canada and Italy but falls to one-third to one-quarter in Japan and France. In the United States and Germany the price coefficients are even lower and are meaningful only for the period following the second oil shock, and for the United Kingdom the estimated coefficient has a rather low t-value. In general, the price elasticities in the commercial and residential sectors are lower than those found in industry, except in France, where they are of the same magnitude, and in Italy and Canada, where the coefficients are actually much higher than the corresponding elasticities for industry.13 These differences in the behaviour of energy demand between the industrial and the commercial and residential sectors in the face of changing energy prices generally reflect different substitution possibilities. In some cases they might also be due to specific energy pricing Table 9.3 Demand for energy in commercial and residential sectors (expressed in log levels, annual data, 1960–88) Country

Constant

Real GDP

Prices

Residential R2 DW fixed investment

DF ADF

United States

1.75 (4.5) 1.19b (1.8) −0.13 (−0.5) −0.72 (−0.8)

0.78 (15.3) 1.10 (19.9) 1.06 (18.8) 1.07 (25.7)

−0.04a (−6.4) −0.33 (−3.1)−1 −0.04a (−6.6) −0.25 −2.0

−0.11 (−1.6) — — — — 0.33 (3.3)

−2.21 −2.22 −3.10 −2.48 −3.87 −3.57 −2.71 −3.39

Japan Germany France

0.91 0.68 0.99 1.14 0.94 1.55 0.96 0.95

122

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Country

Constant

Real GDP

Prices

Residential R2 DW fixed investment

DF ADF

Italy

−2.08 (−1.5) 5.70

1.55 (13.4) 0.19

−0.52 (−5.3) −0.11

0.41 (3.1) −0.32

0.98 0.67 0.85

−2.38 −2.26 −3.45

(8.2) 4.70 (6.4)

(4.8) 0.80 (21.1)

(−1.3) −0.50 (−7.4)

(−4.9) −0.32 (−2.3)

1.33 0.95 0.96

−3.66 −2.66 −3.16

United Kingdom Canada

Notes: aThe coefficients apply only to prices after 1979. bA dummy variable with 0 up to 1979 and 1 thereafter was entered with a coefficient of −0.10 (t-statistic: −2.1). The Dickey-Fuller tests reject the null hypothesis of no co-integration only at a low level of significance (see Table 9.1 for an indication of critical values), while the Durbin-Watson statistics combined with the existence of statistically significant error-correction terms reject this null hypothesis with more certainty.

policies which might have delayed or prevented the full adjustment of domestic energy prices to international levels. Residential fixed investment has a negative influence on energy needs in the United States, Canada and the United Kingdom, suggesting that the stock of housing has gradually become more energy-efficient. On the other hand, residential investment seems to have exerted a strongly positive influence on energy demand in France and Italy, whereas for the remaining two countries no significant coefficients were found. Short-run adjustments

The equations reported in Table 9.4 explain 50% or more of the short-term movements in energy consumption, except in Canada, for which the fit is very poor. The error-correction terms are meaningful and indicate that in the case of Germany, the United Kingdom and Japan, more than two-thirds of the error is corrected within one year. The short-term price elasticities are generally higher than those reported in Table 9.3 and the short-term ad1justment to price changes takes place Table 9.4 Short-term demand for energy in commercial and residential sectors (expressed in changes in log levels, annual data, 1960–88) Country

Constant Real GDP Energy prices Lagged dependent variable

Errorcorrection term

Residential R2DW(h) fixed investment

United States

— — −0.02a (−2.0) −0.02 (−1.1) — —

−0.18 (−1.8) −0.72 (−4.4) −0.94 (−4.0) −0.46 (−2.5)

— — −0.31 (−2.5) — — 0.39 (1.6)

Japan Germany France

0.35 (2.7) 1.15 (11.7) 1.16 (2.7) 1.01 (4.2)

–0.13 (−2.1) −0.58 (−4.8) — — −0.42 (−2.7)

0.32 (2.0) — — 0.23 (2.1) — —

0.56 0.24 0.91 1.69 0.52 0.14 0.56 1.45

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

123

Country

Constant Real GDP Energy prices Lagged dependent variable

Errorcorrection term

Residential R2DW(h) fixed investment

Italy

— — —

1.23 (6.6) 0.17

−0.32 (−3.1) −0.37

— — —

−0.31 (−2.2)–1 −0.79

— — −0.27

0.67 1.78 0.53

— 0.03 (4.0)

(1.1) — —

(−2.9) −0.26 (−1.8)

— — —

(−4.1) −0.38 (−2.4)

(−4.0) — —

1.99 0.20 1.89

United Kingdom Canada

Notes aApplies only after 1979. Lags are reported as superscripts to t-statistics.

without delay. Thus in all countries there appears to be some scope for short-run reduction of energy demand in response to higher energy prices.14 However, given the nature of the adjustments captured by the equations shown in Table 9.4, these results do not change the earlier conclusion concerning the much reduced ability of these sectors (compared with industry) to conserve energy through substitution. 9.II.3 Energy demand in transportation Long-run equations

Energy use in transportation is made up of individual and commercial use and one element common to both components is the almost complete lack of substitution possibilities.15 In such circumstances, the price elasticity of the demand for energy can be expected to be small and substantially below the levels of the sectors discussed above. On the other hand, GDP should remain a major determinant of energy demand as commercial transport is closely and positively correlated with trade and output and personal transport depends on personal income. In some instances investment may play a role but, like investment in industry and in the personal and commercial sectors, the net effect is ambiguous: net additions to the fleet of trucks and automobiles may lead to increased energy consumption whereas replacements of existing equipment could be accompanied by improved energy efficiency.16 Table 9.5 reports the main findings using the same specification as for the commercial and residential sectors, except that both wholesale and retail prices were tested (given the composite nature of the data) and investment was measured as purchases of machinery and equipment as a ratio to GDP. As expected, the price elasticities in the transportation field are low (except for Canada and Italy), and in some cases also have a low statistical significance. In the United States and France prices appear to have had no effect in the period preceding the second oil shock and no significant effects were found in the case of Germany and the United Kingdom. On the other hand, the activity/income effects, measured by real GDP, are well-determined and generally point to a 1–1⅓% increase in energy consumption for each 1% increase in output. Moreover, the coefficients are fairly homogeneous across countries, ranging from 1 for Canada to 1.4 for Italy. In Japan there is strong evidence that the capital stock has become more

124

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Table 9.5 Demand for energy in transportation (expressed in log levels, annual data, 1960–88) Country

Constant

Real GDP

Energy prices?

Investment in machinery and equipment

R2 DW

DF ADF

United States

−0.37 (−1.6) 0.87 (1.5) −0.73 (−3.8) −1.29 (−10.5) −1.83 (−3.6) −1.32

1.12 (20.0) 1.33 (14.8) 1.23 (71.3) 1.29 (44.1) 1.42 (14.2) 1.09

−0.03b (−5.3) –0.13 (−2.2) — — −0.01b (−1.7) −0.31 (−2.1) −0.02

— — −0.29 (−5.2) 0.13 (2.1) — — — — —

0.95 0.46 0.99 1.26 0.99 2.14 0.99 0.77 0.91 1.91 0.91

–1.4 −1.5 −3.4 −3.8 −5.0 −4.5 –2.5 — −4.9 −4.7 −6.1

(−2.0) 0.90 (1.5)

(16.9) 0.98 (17.1)

(−0.2) −0.46 (−7.9)–1

— −0.12 (−1.5)–1

2.36 0.97 0.31

−6.0 −1.2 −1.4

Japan Germany France Italy United Kingdom Canada

Notes: Lags are reported as superscripts to t-statistics. A time trend with a coefficient of −0.017 (t-statistic: −3.6) has been added in the case of Japan. aWholesale prices for France, Italy and the United Kingdom and retail prices for the United States, Japan and Canada. bThe coefficients apply only to energy prices after 1979. For Germany, Italy and the United Kingdom both the DW statistics and the Dickey-Fuller tests reject the null hypothesis of no co-integration whereas for the other four countries the Dickey-Fuller tests give very poor results.

energy-efficient—thus supporting the results observed for industry— whereas in Germany a rise in the investment/GDP ratio appears to have increased energy demand. Short-run adjustments

Similar to the results found for the commercial and residential sectors, some short-term equations have been adjusted, which confirm that there is some scope for short-run reductions in energy demand in response to higher prices. Except for Japan, for which no statistically significant short-term coefficient could be identified, all price coefficients are higher than in the long-run equations. On the other hand, the output elasticities are generally lower in the short run than in the long run. This was also found for the residential and commercial sectors, and probably reflects that increased energy demand partly depends on the possession of certain durable goods and pieces of equipment, the acquisition of which is subject to some lag and thus imparts a relatively low short-run income elasticity to energy demand. The speed with which deviations from the long-run trend are being corrected shows some variation from country to country. In France, Japan, Germany and Italy between one-third and 85% of the error is corrected

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

125

Table 9.6 Short-term equations of energy demand in transportation (variables expressed as changes in log levels) Country

Constant

Real GDP

Machinery and equipment

Prices

Error-correction term

R2 DW

United States

0.02a (3.1) — — 0.82 (1.0) — — — — — — −0.02 (−1.5)

0.55 (4.0) 0.95 (17.3) 0.93 (3.9) 1.18 (11.5) 1.07 (2.4)−1 1.01 (2.2) 1.12 (5.4)

— — — — 0.17 (2.1) — — — — — — −0.26 (−2.9)−1

−0.16 (−3.3) −0.11 (−2.2)–2 −0.11 (−2.1)–1 –0.09 (−1.7) −1.10 (−6.4) −0.18 (−0.6) — —

−0.24 (−3.3)–2 −0.58 (−4.0) −0.84 (−3.6) −0.34 (−2.3) −0.80 (−5.3) −0.18 (−6.0) −0.27 (−2.1)

0.77 1.78 0.94 1.76 0.59 2.07 0.85 1.74 0.72 1.49 0.58 2.09 0.81 1.57

Japan Germany France Italy United Kingdom Canada

Notes: Lags are reported as superscripts to t-statistics. aIn addition, a dummy variable with 0 up to 1979 and 1 thereafter has been entered with a coefficient of −0.02 (tstatistic: −3.2).

within one year while the adjustment process is slow in North America, with three years required to correct one-quarter of the error in the United States.17 In the United Kingdom the adjustment process is one of damped oscillations. 9.II.4 Energy demand in periods of rising and falling prices The elasticities reported so far have been estimated on the assumption that energy demand responds symmetrically to rising and falling prices.18 As can be seen from Figure 9.1, relative energy prices have been quite volatile over a period of estimation, with the volatility observed for individual countries depending on the share of imported energy in overall energy consumption, movements of exchange rates and the extent of administered pricing. However, in all seven countries there have been distinct phases of rising and falling prices, thus making it possible to test whether the assumption of a symmetric response is valid. A priori it may be argued that while in periods of rising prices firms and households have a strong incentive to reduce costs by installing more energy efficient capital equipment, few agents are probably prepared to scrap such equipment and replace it by less energy-efficient machinery in periods of falling prices. Thus the price elasticity is likely to be numerically lower during periods when prices are falling, and this is confirmed by Hunter and Rosenbaum (1991) in a study of US households’ demand for oil.19 At the same time, the finding of a numerically lower price elasticity in periods of falling prices does not necessarily imply that agents respond asymmetrically. In the first place, non-price conservation measures, which have mainly been introduced in periods of rising prices and are difficult to quantify in empirical estimates, may overstate the estimated response to a rise in energy prices. Secondly, to the extent that the

126

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

lag structure with respect to the influence of prices on consumption is not sufficiently long, the empirical estimates will again give a false impression of asymmetry. For instance, in a study of US oil demand Brown and Phillips (1989a) allow for a total lag of about ten years and find no asymmetries when the lag structure is determined optimally,20 whereas other studies using predetermined lags have found some evidence of asymmetric behaviour.

Figure 9.1 Relative energy prices: in relation to the GDP deflator (1972=100)

In order to preserve degrees of freedom we did not attempt to estimate energy demand functions separately for periods of rising and falling prices. Instead, we introduced a dummy variable (DUM) with a

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

127

value of 1 for years of rising prices and 0 otherwise. By adding the term e.DUM.Pe to the original equation (1) the nature and strength of possible asymmetries can then be evaluated on the basis of the sign and statistical significance of the parameter e. The asymmetry hypothesis was only tested for the industrial sector and the results for the level and corresponding error-correction equations are given in Tables 9.7 and 9.8. Comparing these with the earlier Tables 9.1 and 9.2 the following main features may be observed: 1 In virtually all countries the DF/ADF tests yield much more satisfactory results when the new term is added and in the two cases (United States and Canada) where the improvement is most pronounced the coefficient of the error-correction terms in Table 9.8 also increases significantly. 2 In three of the countries (United States, Germany and France) the inclusion of the additional price term raises the income elasticity as well as the numerical value of the trend term. In the remaining countries the income elasticity falls slightly or remains unchanged. 3 In the United States, Italy and Canada there is some evidence that the price elasticity is numerically lower in periods of rising prices, though for Italy the statistical significance of the additional price term is rather low. In the other four countries, the response to price changes appears to be higher in periods of rising prices, but only the equation for the United Kingdom yields a statistically significant coefficient for the additional price term. 4 The error-correction equations for France, Italy and Canada remain largely unchanged except for the larger error-correction coefficient for Canada. For the United States there is a clear improvement of the statistical properties and for the United Kingdom and Germany there is some—albeit weak—evidence of price change effects in periods of rising prices. Table 9.7 Demand for energy in industry in periods of rising energy prices (expressed in log levels) Countries

Constant Real GDP Prices, whole period

Pricesa rising

Non-resid. fixed investment

Time trend R2 DW DF ADF Periods of rising prices

United States

−3.92

2.30

−0.11

0.01

—

−0.06

0.83

−3.48

(−3.3) −0.32 (−0.5) −1.81 (−2.0) −2.96 (−2.2) –5.23 (−3.4) −3.09

(8.1) 2.48 (15.8) 1.81 (10.4) 2.20 (9.1) 3.02 (9.3) 2.64

(−2.5)–3 −0.21 (−5.5)–1 −0.18 (−2.5)–2 –0.22 (−1.7) –0.41 (−3.7) −0.61

(3.4) −0.00 (−0.2) −0.00 (−0.2) −0.01 (−1.0) 0.01 (1.2) −0.02

— −0.66 (−3.2) — — — — — — —

(−6.9) −0.10 (−12.6) −0.04 (−8.2) −0.06 (−6.9) −0.09 (−6.9) −0.06

1.34 0.99 2.00 0.91 1.29 0.91 1.24 0.96 1.06 0.89

−2.36 −5.00 −3.54 −3.47 −5.58 −3.26 −3.89 –3.11 −3.54 −2.75

(−1.5) 0.73 (6.4)

(6.2) 0.94 (26.0)

(−4.0)–3 −0.12 (−3.3)

(−1.8) 0.02 (4.1)

— — —

(−5.6) — —

0.88 0.98 1.13

−2.40 −3.21 −3.33

Japan Germany France Italy United Kingdom Canada Note:

1970–81

1973–82 1971–82 1973–85 1973–85 1973–85

1970–85

128

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Constant Real GDP Prices, Pricesa Non-resid. Time trend R2 DW DF ADF Periods of whole rising fixed rising period investment prices aPrices as defined in the previous column times a dummy with 1 in periods of rising energy prices as de fined in the last column and 0 otherwise. Countries

Table 9.8 Short-term adjustment equations for energy demand in industry in periods of rising energy prices (expressed as changes in log levels) Countries

Constant Real GDP Prices, whole period

Prices rising ErrorNon-resid. fixed correction term investment

R2 DW

United States

−0.06a (−4.7) −0.08 (−5.2) −0.05 (−5.2) −0.06 (−3.3) −0.08 (−5.9) −0.05

1.92 (6.7) 2.13 (7.9) 2.26 (8.3) 2.14 (4.5) 3.00 (8.9) 2.06

−0.13 (−1.7)−1 −0.20 (−4.8)–1 — — −0.27 (−2.7)–1 −0.09 (−1.8) —

— — — — −0.25 (−1.8)–1 — — — — −0.22

−0.59 (−3.2) −0.94 (−4.3) −0.68 (−3.2) −0.54 (−6.5) −0.52 (3.8)–1 −0.40

— — −0.39 (−1.6) –0.22 (−1.7) — — — — —

0.76 1.67 0.87 1.82 0.80 1.59 0.67 1.67 0.76 1.95 0.54

(−3.7) –0.02 (−1.3)

(5.1) 1.40 (3.9)

— — —

(−0.8) — —

(−2.2) −0.52 (−2.6)

— — —

2.16 0.56 2.13

Japan Germany France Italy United Kingdom Canada

Note: aAlso includes a dummy variable with 1 in periods of rising prices and a coefficient of 0.02 (1.8).

Overall, the results reported in Tables 9.7 and 9.8 are rather mixed. On the one hand, the inclusion of the additional price term improves the statistical properties of the estimated equations. On the other hand, the asymmetry effects are small and/or statistically insignificant, suggesting that the additional price term may not be capturing an asymmetric response of energy demand but rather phenomena which appear to have played a role in years of rising prices. However, the source of these influences has not been identified and would require further analysis of the timing of policy measures and of factors particular to the industrial sectors in each of the seven countries. 9.III SUMMARY AND CONCLUSIONS The main analytical and empirical findings of this paper may be summarized in five points. 1 Because of a fall in the share of net energy imports in GDP the industrial countries are now less exposed to energy price shocks than in 1973–4 and shocks has been significantly reduced. A further positive element in this respect is the very moderate real wage behaviour following the second oil 1978–

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

129

80. As a consequence, the risk to real output and inflation of such shock and the continuation of modest real wage increases in the 1980s. 2 However, the decline in imports has mainly resulted from the reduction in relative energy prices since 1985, whereas energy conservation and the development of indigenous sources of energy have been far less important. Moreover, in the absence of clear signs of fundamental changes in real wage behaviour a smooth adjustment of wages and other factor prices to externally induced real income shocks cannot be taken for granted. Hence, despite the decline in exposure policy makers need to remain alert to external disturbances. 3 The empirical results presented here point to total output as the key determinant of energy demand, whereas relative energy prices have only a small influence. This is especially the case in transportation and in the residential and commercial sectors, where substitution possibilities are less readily available than in industry. More specifically, for the Group of Seven countries on average a 1% fall in real output growth generates a fall in energy demand of somewhat more than 1% within the same year and of almost 1.5% in countries outside North America. For most countries long-run price elasticities are only around 0.2 for industrial energy demand and even lower for energy used for transportation and in the commercial and residential sectors. Moreover, a response of even this moderate size is only achieved in the course of several years and requires the installation of new and more energy-efficient capital equipment. 4 Despite the lower exposure, energy price shocks—and external price shocks in general—pose several problems and there are no easy ‘trade-offs’ for policy-makers. Firstly, the real income loss due to weaker terms of trade reduces aggregate demand and output in energy importing countries, while the rise in import prices increases inflationary pressures. Secondly, the most effective way of reducing energy demand and the dependence on oil is to lower output growth, but this is also the most costly solution in terms of foregone output and higher unemployment. Reducing energy dependence by encouraging substitution constitutes a more efficient solution but is bound to be a slow process given the low price elasticities and the crucial role of investment. Thirdly, while a quick reduction of energy demand and an effective containment of the inflationary risk call for restrictive monetary and fiscal policies, the accompanying rise in interest rates could reduce investment and thus postpone a more efficient long-run solution. Finally, measures to protect the environment often reduce the efficiency of energy use (especially in the transportation sector), thereby frustrating efforts to promote energy conservation. 5 Taking the experience gained from the first and second oil shocks as a guide,21 the containment of the inflationary risk should be assigned the highest priority in the very short run, with the degree of policy tightening depending on the response of domestic factor prices. This behaviour of nominal and real wages plays a crucial role in this respect, but the likely response of wages to possible future changes in energy prices is not easy to evaluate. On the one hand, the real wage adjustment to the second oil shock was remarkable and gives cause for optimism. Moreover, a number of countries introduced various labour market reforms in the course of the 1980s with a view to improving flexibility, possibly suggesting that real wage moderation may be even more pronounced than after the second oil shock. On the other hand, the lack of consensus regarding the rise of unemployment in the 1980s is disturbing and leaves some important aspects of real wage behaviour unresolved. Moreover, empirical studies which attempt to identify fundamental changes in real wage behaviour have so far failed to provide any firm evidence, except possibly for the United States. On balance, it thus appears that, based on the experience of the second oil shock and assuming a non-accommodating policy stance, the acceleration of wage and price inflation seen in the early 1970s is unlikely to be repeated in the future, especially if

130

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

the rise in oil prices is only temporary or is expected to be only temporary. On the other hand, to count on a more favourable response than after the second shock would seem too optimistic and difficult to justify on the basis of the empirical evidence. ACKNOWLEDGEMENTS We gratefully acknowledge comments by Dr Bockelmann, Dr Bisignano, Dr Hutchison and Mr H.Christiansen on earlier drafts of this paper. NOTES 1 Only a part of the paper is presented here (the editors). 2 Time series which are stationary only after differencing may have linear combinations which are stationary without differencing, i.e. they have similar long-term trends which offset each other in the combination. In such a case, those variables are said to be co-integrated. A co-integrated system can be represented in an error-correction structure which incorporates both changes and levels of variables such that all the elements are stationary. The error-correction structure captures short-term movements of the variables and also provides a framework for forecasting and for testing the co-integration conditions (Engle and Yoo 1987). 3 Engle and Granger (1987) presented a theorem showing that co-integrated series can be represented by an errorcorrection formulation. 4 As will be seen below, the coefficient was in fact significantly negative for six countries, probably as the combined result of specific energy-conserving measures and shifts in the composition of demand and industrial output towards less energy-intensive goods and sectors. However, an alternative and perhaps equally plausible interpretation of the negative time trends is also possible and will be discussed below. 5 In theory, manufacturing or industrial production should be used as the explanatory variable, but in practice the results were (except for the absolute size of the coefficients) very similar and total GDP was preferred in order to obtain comparable and more comprehensive coefficients. In the late 1980s the share of manufacturing production in total GDP ranged from 19% in the United States to 31% in Germany. 6 The results also give some support to the estimation procedure chosen. The Durbin-Watson statistics clearly reject the null hypothesis of no co-integration, whereas in some cases the Dickey-Fuller tests only reject the null hypothesis at a low level of significance. However, further error-correction equations reported in Table 9.2 tend to show that in most cases the null hypothesis of no co-integration can be rejected. 7 The actual coefficients were: United States: 0.54, Japan: 1.18, Germany: 0.78, France: 1.12, Italy: 2.77. For Japan, we used the 1963–83 GDP trend of 4.4% (rather than the 1968–88 trend of 5.5%) which is closer to the long-term trend currently observed. For Canada this transformation was not made, while for the United Kingdom the transformed trend coefficient was negative. This odd result derives from the very large negative trend observed in Table 9.1 combined with a relatively low trend growth of GDP (around 2%), and might be explained by the 0very sharp reduction in the share of industry in total GDP over the estimation period (some 13 percentage points for both manufacturing and industry [current prices] compared with 7–8 points for other Group of Seven countries on average). 8 With Canada being a major oil producer there is a risk that the long-run price elasticity reflects both demand and supply responses. A regression of industrial energy consumption on total domestic energy production yields an R2 of about 0.75. 9 For Japan, France and Italy total non-residential investment; for other countries private non-residential fixed investment. 10 One reason for the failure to identify any effects of the investment variable might be that because of the ratio form I is not an I (1) variable. However, except for the United States, the DF/ADF tests fail to reject the hypothesis of

ENERGY SHOCKS AND THE DEMAND FOR ENERGY

11

12 13 14 15

16 17 18

131

an I (1) process. On the other hand, using lagged prices in a long-term equation could in part be capturing the effects of capital stock changes. An additional reason for expecting a lower income elasticity might be that in industry energy is used as an intermediate input in generating total output, whereas in the commercial and residential sectors it is demanded as a final or semi-final consumption good. The energy component of the consumer price index was used in calculating relative energy prices. In the case of Canada this again suggests that the price elasticity for industrial energy demand is biased towards zero because of energy supply. For instance, by reducing room temperature, turning off lights earlier and, in the slightly longer run, putting in double glazing. Another common element is the lack of inter-fuel substitution possibilities. For instance, in 1988, for all seven countries taken together, petroleum products accounted for 99.3% of energy used by transportation (98.3% for the group excluding the United States), the rest being electricity, itself partly generated by oil. It should be borne in mind that in more recent years environmental policies may have reduced the energy efficiency of investment. The low coefficient should be seen against the poor results for the level equation. It may be recalled that the equations were estimated as log-level equations, so that the coefficients equal elasticities and are independent of the levels of energy demand, income and prices. Alternatively, the demand functions could have been estimated as linear equations:

for which the partial price elasticity would be cPe/E and thus numerically rising (falling) in periods of rising (falling) prices and falling (rising) quantities demanded. Similarly, the partial income elasticity, measured as bY/E, would be rising in periods when income is rising and E is falling because of rising prices. 19 Hunter and Rosenbaum rely on equations where the dependent variable is measured as household expenditure on energy and they estimate separate equations for periods of rising and falling prices respectively. 20 I.e. maximizing the R2. 21 For further discussion see Hutchison, op. cit.

REFERENCES Andersen, P.S. (1989): ‘Inflation and output: A review of the wage-price mechanism’, BIS Economic Papers, No. 24. Artus, J. (1984): ‘An empirical explanation of the disequilibrium real wage hypothesis’, IMF Staff Papers, pp. 1–36. Baily, M.N. (1981): ‘Productivity and the services of capital and labour’, Brookings Papers on Economic Activity, pp. 1–65. Bean, C., Layard, P.R.G. and Nickel, S.J. (1986): ‘The rise in unemployment: A multi-country study’, Economica, Supplement, pp. 1–22. Bean, C. and Gavosto, A. (1988): ‘Outsiders, capacity shortages and unemployment in the United Kingdom’, Centre For Labour Economics, LSE Working Paper, No. 1058. Blanchard, O.J. (1987): ‘Why does money affect output? A survey’, NBER Working Paper, No. 2285. Bradford De Long, J. and Summers, L. (1988): ‘On the existence and interpretation of a “unit root” in US GNP’, NBER Working Paper, No. 2716. Brown, S.P.A. and Phillips, K.R. (1989a): ‘An econometric analysis of US oil demand’, Federal Reserve Bank of Dallas, Research Paper, No. 8901. —— (1989b): ‘Oil demand and prices in the 1990s’, Economic Review, Federal Reserve Bank of Dallas, January, pp. 1–8. Bruno, M. and Sachs, J. (1985): The Economics of Worldwide Stagflation, Basil Blackwell, Oxford.

132

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Campbell, J.Y. and Shiller, R.J. (1988): ‘Interpreting co-integrated models’, Journal of Economic Dynamics and Control, pp. 505–22. Coe, D. (1990): ‘Insider-Outsider influences on industry wages’, Empirical Economics, pp. 163–84. Cross, R. (ed.) (1988): Unemployment, Hysteresis and the Natural Rate Hypothesis, Basil Blackwell, Oxford. Dolado, J., Jenkinson, T. and Sosvilla-Rivero, S. (1990): ‘Co-integration and unit roots: A survey’, Banco de España, Documento de Trabajo, No. 9005. Engle, R. and Granger, C.W.J. (1987): ‘Co-integration and error correction: Representation, estimation and testing’, Econometrica, Vol. 55, pp. 251–76. Engle, R. and Yoo, B.S. (1987): ‘Forecasting and testing in co-integrated systems’, Journal of Econometrics, Vol. 35, pp. 143–59. Fischer, S. (1977): ‘Long-term contracts, rational expectations and the optimal money supply rule’, Journal of Political Economy, pp. 191–206. Helliwell, J., Sturm, P. and Salou, G. (1986): ‘The supply side in the OECD’s macroeconomic model’, OECD Economic Studies, No. 6, Spring, pp. 75–137. Hendry, D.F. (1986): ‘Econometric modelling with co-integrated variables: An overview’, Oxford Bulletin of Economics and Statistics, pp. 201–12. Hubbard, G. (1986): ‘Supply shocks and price adjustments in the world oil market’, Quarterly Journal of Economics, pp. 85–102. Hunter, W.C. and Rosenbaum, M.S. (1991): ‘Supply shocks and household demand for motor fuel’, Economic Review, Federal Reserve Bank of Atlanta, March/ April, pp. 1–11. Hutchison, M. (forthcoming): BIS Economic Paper. Jarret, P. and Torres, R. (1987): ‘A revised supply block for the major seven countries in Interlink’, OECD Working Papers, No. 41. Katz, L.F. (1986): ‘Efficiency wage theories: A partial evaluation’, NBER Macroeconomics Annual, pp. 235–90. Lindbeck, A. and Snower, D. (1988): The Insider-Outsider Theory of Employment and Unemployment, MIT Press, Cambridge, MA. Malinvaud, E. (1977): The Theory of Unemployment Reconsidered, Basil Blackwell, Oxford. Malinvaud, E. (1982): ‘Wages and unemployment’, Economic Journal, pp. 1–12. Maurel, F. (1989): ‘Modèles à correction d’erreur: l’apport de la théorie de la co-intégration’, Economie et Prévision, No. 88–89, pp. 105–25. McDonald, I. and Solow, R.M. (1980): ‘Wage bargaining and unemployment’, American Economic Review, pp. 802–22. Mittelstaedt, A. (1983): ‘Utilisation des elasticités de la demande dans l’estimation de la demande de l’energie’, OECD Working Paper, No. 1, Table 9, p. 28. Nordhaus, W. (1980): ‘Oil and economic performance in industrial countries’, Brookings Papers on Economic Activity, pp. 341–88. OECD (1990): Economic Outlook, June, Paris. Rasche, R.H. and Tatom, J.A. (1981): ‘Energy price shocks, aggregate supply and monetary policy: The theory and the international evidence’, Carnegie Rochester Conference Series on Public Policy, No. 14, pp. 9–94. Solow, R.M. (1979): ‘Another possible source of wage stickiness’, Journal of Macroeconomics, pp. 79–82. Taylor, J.B. (1979): ‘Staggered wage setting in a macro model’, American Economic Review, pp. 108–13.

Chapter 10 Modelling the petroleum spot market: a vector autoregressive approach Walter C.Labys, Véronique Murcia and Michel Terraza

10.I INTRODUCTION Until 1973, pricing on the world oil market was largely dominated by the Major seven producers (Shell, British Petroleum, Mobil, Standard Oil of California, Gulf, Texaco, Exxon). The landmark change we recognize was on 16 October 1973 when OPEC decided to increase the price of crude petroleum first by 70% and later by several times over. This reflected a shift in market control from the ‘major seven’ to the major petroleum producing countries. One other market development was the instigation of a spot market for crude petroleum for transactions involving contracts of a longer term. The resulting prices formed on this market have became an independent indicator of crude oil market forces. These prices also have fluctuated widely, interesting consumers and producers alike in the prediction of future price movements. An agreement among producers and consumers in 1986 established a reference price of crude oil at US $18.00 and a production ceiling was established, based on a national quota system. Non-OPEC oil production also increased to the point where the Rotterdam price has become a price indicator for OPEC to follow. As a consequence, OPEC in 1987 adopted a sales contract with prices linked to the Rotterdam market. These events have made the modelling of oil spot market prices an important exercise. Several other several econometricians and energy economists have already produced a variety of such oil price models. However, the present modelling effort differs in its attempt to employ recent developments in time-series methods (see Labys et al. 1991b). 10.I.1 Vector autoregressive (VAR) models This class of econometric models has recently been applied to a variety of economic variables of time series nature. While the VAR approach adopted here models each equation separately (rather than simultaneously), it has the advantage of also producing efficient forecasts. Figure 10.1 summarizes the major steps and functions of this modelling approach. Exhaustive explanations of the development and application of this approach appear in different papers (Cromwell et al. 1994a and b; Hsiao 1979; Labys et al. 1991a, and Sims 1980). The empirical incentive for the present modelling effort stems from the work of Verleger (1982); he was one of the first economists to discover and to adopt the competitive characteristics of the crude oil spot

134

MODELLING THE PETROLEUM SPOT MARKET

Figure 10.1 The VAR modelling approach

market in the form of an econometric model (see Appendix 1). We follow his approach in developing equations for the major oil market variables as listed below. The purpose of the present study is to reformulate the original equations using the VAR approach and to compare the performance of the model by using a conventional or unrestricted VAR estimation procedure (UVAR) as well as one employing a restricted VAR approach (VAR). 10.II MODEL SPECIFICATION The model developed by Verleger concentrated on the simultaneous relation between oil prices and stocks. The present model is more open and includes imports and exports. The regional base for both models is that of the OECD countries. Data are quarterly and span the years 1974 to 1990; they are only of a preliminary nature and their description appears in Appendix 2. The specification embodies the following six variables: ● Rotterdam crude oil spot price (FOIL) ● Crude oil consumption (COIL) ● Crude oil production (QOIL)

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

135

● Crude oil stocks (INVOIL) ● Crude oil exports (XOIL) ● Crude oil imports (MOIL) Sources of data include ‘Energy Prices and Taxes’ and ‘Petroleum Quarterly’, both published by the International Energy Agency (OECD, Paris). 10.III MODEL ESTIMATION Following the major VAR modelling steps outlined in Figure 10.1, each of the included model variables have been subjected to the following tests: Nature of stationarity

This stage refers to the Dickey-Fuller Unit Root Test (1979): Let the random process Xt be defined in its autoregressive form, One tests the null hypothesis H0: ‘Xt has a unit root’ if

The coefficient ø1 is tested against a T-Student test, non standard because its distribution is non-regular (Dickey-Fuller 1979). Consequently, if the coefficient ø1 belongs to the null hypothesis space, the process Xt follows a random walk, and is non-stationary. If the process appears non-stationary, one seeks to find whether this characteristic is due to a seasonal influence or if it is generated by its own evolution of the process during time; accordingly two different procedures are employed: ● seasonal differentiation, and ● recurrent differentiation Presence of cointegration This stage is very important for the VAR model structure. Actually, if some variables selected for this kind of modelling are cointegrated, their effect will be more than proportional in the model’s equations. The cointegration relation, in the multivariate case, is: where Xt and Yt are two stationary processes and Zt measures the equilibrium deviation. The cointegration relation detects the similarity of evolution among one, two or three processes. Thus, if variables are cointegrated in the VAR model, they will influence strongly the specification and the forecast properties of the model. Direction of causality

The restricted VAR model is based on causal links between the selected variables. A VAR equation reveals the degree of causality between the endogenous variable and all the others of the model:

136

MODELLING THE PETROLEUM SPOT MARKET

with a (n, 1) vector the (aij) coefficient matrix (n, n)

Yit Aij under the normalization: ai0=1

the (n, 1) vector of innovations, white-noise process the optimal lag order

p

Selection of optimal lag lengths See discussion below. Results of the tests revealed stationarity in first differences for FOIL, XOIL and MOIL. Stationarity in fourth differences was discovered for COIL, QOIL and INVOIL. The transformed series employed for model estimation were DPOIL, DXOIL, DMOIL, INVOIL4, COIL4 and QOIL4. The estimation results for the unrestricted (UVAR) and restricted (RVAR) models are given below. 10.III.1 Unrestricted model (UVAR) Following Sims (1980), the same lag lengths are employed in each of the equations estimated. The forecast percentage error (FPE) criterion was used to determine the lag lengths. Table 10.1 Selection of optimal lag order

1 2 3 4 5 6 7 8

FPE

AIC

BIC

HQ

7.1244E+11 4.1692E+11 2.3516E+11 4.153E+11 4.2858E+11 5.5751E+11 3.4606E+11 1.765E+10

27.288486 26.726485 26.075027 26.468490 26.157384 25.781500 24.089567 18.489875

28.513134 29.196706 29.812312 31.494917 32.495634 33.454874 33.122004 28.905969

27.770146 27.696356 27.539704 28.434602 28.631580 28.770435 27.599880 22.528173

with T p Ω(p)

number of observations lag order variance-covariance matrix of residuals

The minimum FPE was obtained for a uniform lag length of 8 quarters, as shown in Table 10.1, where the different criteria result from the following equations: 1 Final Predictor Error (FPE)

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

where T P n Ω(P)

number of observations ordinary VAR model order number of endogenous variables of the model variance/covariance anticipated matrix of residuals

2 Akaike Information Criterion (AIC)

3 Bayesian Information Criterion (BIC)

4 Hannan-Quinn Criterion(HQ)

where c is a constant indicating the criterion consistency, normally above 1. The results of estimating the UVAR(8) model are:

137

138

MODELLING THE PETROLEUM SPOT MARKET

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

139

In the equations above numbers in parentheses represent the T or Student ratio. The resulting direction of influence of the right-hand variables on the left-hand variables can be consulted in Table 10.2. The nature of the goodness-of-fit in each of the above equations can be measured in the various graphs presented in Appendix 3; a major problem with imposing the lag length of eight periods uniformly in the unrestricted model is the necessity of estimating a large number of equation parameters, subject to a limited number of degrees of freedom. 10.III.2 Restricted model (RVAR) The restricted model employs the important constraint that the lag lengths for each of the variables employed has been selected individually. In this case, we have utilized on the minimum information criteria of Akaike (1974). The resulting lag lengths obtained are featured in Table 10.3. The results of estimating the RVAR model are as follows: Table 10.2 Direction of influence confirmed by the UVAR(8) model DPOIL

INVOIL4

DM

DX

QOIL4

COIL4

DX

QOIL4

COIL4

2

1 6

5 7

2 6

DPOIL INVOIL4 DM DX QOIL4 COIL4 Note: Y X corresponds to X explains Y. Table 10.3 Selected optimal lag length for the RVAR model Explanatory variables

DPOIL INVOIL4 DM DX QOIL4 COIL4

DPOIL

INVOIL4

6

3 4 1 7 2 8

1

5

DM 3 4 4 4

2 1

140

MODELLING THE PETROLEUM SPOT MARKET

The resulting directions of influence of the right-hand variables on the left-hand variables can be measured from Table 10.4. The nature of the goodness-of-fit in each of these equations can be examined in the various groups presented in Appendix 4.

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

141

10.III.3 Comparison of model performance The basis for comparing the UVAR and the RVAR models is the Root Mean Square Error (RMSE) of each of the included equations. where t+j is the value of the variable X forecast for the period (t+j). Results of computing the RMSE are presented in Table 10.5. An examination of those values indicates that the UVAR performed better than the RVAR for INVOIL, MOIL, QOIL and COIL. The explanation of oil spot prices was about the same for each model. The lesser performance of the RVAR is difficult to explain, because the RVAR represents a more precise model specification. Table 10.4 Direction of influence confirmed by the RVAR model DPOIL

INVOIL4

DM

DX

QOIL4

COIL4

DPOIL INVOIL4 DM DX QOIL4 COIL4 Table 10.5RMSE values for the UVAR(8) and the RVAR models RMSE

POIL INVOIL M X QOIL COIL

UVAR

RVAR

9.999 415.306 66.264 28.157 24.305 55.517

9.878 532.105 68.741 26.479 27.375 83.285

10.IV CONCLUSIONS Among the results that can be summarized, we base our conclusions on the RVAR model. Concerning oil spot prices, the present results confirm those of Verleger (1982) with respect to the close relation found between prices and stock changes. Explanation of oil prices over the sample period, of course, proved difficult, because of several jumps in oil prices, which were a result of OPEC policies. Oil imports were found to influence both stocks and prices. Production in the OECD countries would not be expected to be an important variable. However, consumption was, and this is reflected in its strong relation to prices.

142

MODELLING THE PETROLEUM SPOT MARKET

The VAR approach we have investigated has proved important regarding its ability to explain variations in the models’ endogenous variables over time. Better results could possibly have been obtained, if a greater number of quarterly observations had been available for model estimation. While the RVAR model gave superior predictions for the INVOIL and QOIL variables, overall it performed less well than the UVAR model (10.III.1). It is interesting to note differences in forecast performances over the three models elaborated: the two unrestricted models (10.III.1 and 10.III.3) succeed in forecasting price and exports, only with difficulty. The present modelling exercise however, was strictly of an experimental nature and requires further work, such as that of capturing the simultaneity between the important market variables. Further research could also be performed in extending the VAR equation results to forecast the variables of interest and, in particular, the crude oil spot price. REFERENCES H.Akaike (1974) ‘A new look at Statistical Model Identification’, International Electrical Engineering. IEEE Transactions on Automatic Control, 19, pp. 716–23. A.Ayoub and J.Percebois (1987) ‘Pétrole: Marchés et Stratégies’. Economica. J.Cromwell, W.Labys and M.Terraza (1994a) Univariate tests for Time Series Model, SAGE, Thousand Oaks, California. —— (1994b) Multivariate Tests for Time Series Model, SAGE, Thousand Oaks, California. D.Dickey and W.Fuller (1979) ‘Distribution of estimates for Autoregressive Time Series with unit root’, Journal of the American Statistical Association, 74, pp. 427–31. F.Diebold and S.Sharpe (1990) ‘Post-Deregulation Book-Deposit-Rate Pricing: The Multivariate Dynamics’, Journal of Business Economics and Statistics, 8, pp. 281–91. R.Engle and C.Granger (1987) ‘Cointegration and error correction model: representation, estimation and testing’, Econometrica, 55, pp. 251–71. C.Granger (1969) ‘Investigating Causal Relations by Econometric Models and Cross-Spectral Methods’, Journal of Econometrics, 37, pp. 424–38. C.Hsiao (1979) ‘Autoregressive modeling of Canadian money and Income Data’, Journal of the American Statistical Association, 74, pp. 553–60. —— (1979) ‘Causality Tests in Econometrics’, Journal of Economic Dynamics and Control, 1, pp. 321–46. P.Jacquet and F.Nicolas (1991) ‘Petrole: Crise, Marché, Politique. Dunod, Paris. M.Kaylen (1988) ‘Vector Autoregression Forecasting Models: Recent developments applied to the U.S. Hog Market’, American Journal of Agricultural Economics, 3, pp. 701–12. W.Labys, V.Murcia and M.Terraza (1991a) ‘Progrès économetriques et séries temporelles’. Centre d’Econométrie Pour l’Entreprise—Université de Montpellier I. —— (1991b) ‘Modélisation VAR du marché du pétrole’. Centre d’Econométrie Pour l’Entreprise—Université de Montpellier I. Communication du XXXIIo Colloque de l’Association d’Econométrie Appliquée. T.Mills (1991) Time Series Techniques for Economists. Cambridge University Press, Cambridge. C.Sims (1972) ‘Money, Income and Causality’, American Economic Review, 62, pp. 540–52. —— (1980) ‘Macroeconomics and Reality’, Econometrica, 58, pp. 1–48. P.Verleger (1982) Oil Markets in turmoil, an Economic Analysis. Ballinger, Cambridge, MA.

Appendix

Appendix 1 This contains 7 variables:

SVt It Qt

official crude price at the t period spot value at the t period product price to consumption at the t period stock demand at the t period anticipated oil production at the t period present oil production at the t period demand at the t period

The Verleger model is:

where the three last equations are accounting equations.

144

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Appendix 2

Figure A10.1 (a) Crude oil price Source: OPEC ‘Annual Statistical Bulletin’

Figure A10.1 (b) Crude oil stocks Source: ‘OECD Petroleum Quarterly’, IEA

Figure A10.1 (c) Crude oil consumption Source: ‘OECD Petroleum Quarterly’, IEA

APPENDIX

Figure A10.1 (d) Crude oil production Source: ‘OECD Petroleum Quarterly’, IEA Figure A10.1 (e) Crude oil exports Source: ‘OECD Petroleum Quarterly’, IEA

Figure A10.1 (f) Crude oil imports Source: ‘OECD Petroleum Quarterly’, IEA

Appendix 3

Figure A10.2(a) Oil prices (UVAR)

145

146

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Figure A10.2(b) Oil stocks (UVAR)

Figure A10.2(c) Oil production (UVAR)

Figure A10.2(d) Oil consumption (UVAR)

APPENDIX

Figure A10.2(e) Oil exports (UVAR)

Figure A10.2(f) Oil imports (UVAR)

Appendix 4

Figure A10.3(a) Oil prices (RVAR)

147

148

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Figure A10.3(b) Oil stocks (RVAR)

Figure A10.3(c) Oil production (RVAR)

Figure A10.3(d) Oll consumption (RVAR)

APPENDIX

Figure A10.3(e) Oil exports (RVAR)

Figure A10.3(f) Oil imports (RVAR)

Appendix 5

Figure A10.4(a) Experimental oil prices (UVAR)

149

150

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Figure A10.4(b) Experimental oil stocks (UVAR)

Figure A10.4 (c) Experimental oil production (UVAR)

Figure A10.4(d) Experimental oil consumption (UVAR)

APPENDIX

Figure A10.4(e) Experimental oil exports (UVAR)

Figure A10.4(f) Experimental oil imports (UVAR)

151

Chapter 11 Fiscal harmonization on oil products within the EC: problems and prospects J.-B.Lesourd and D.Meulders

11.I INTRODUCTION Fiscal barriers remain, together with physical and technical boundaries, one of the main problems to solve in order to complete the Single European Market of 1993. This is because, in spite of the disappearance of custom duties proper within the EC, there remain controls that are linked to national indirect tax systems (excise duties and Value Added Taxes (VAT)), which are often quite important; therefore, getting rid of fiscal boundaries requires adequate harmonization of indirect tax systems within the EC. Oil products are important in this respect, because they are submitted to indirect taxes which are in many cases, heavy. In their case, fiscal harmonization means harmonization of energy demand policies and of environmental policies. The present work proposes an analysis of several points related to fiscal harmonization of indirect taxes on oil products within the EC. 11.II THE PRESENT SITUATION This situation is rather complex, since there are several levels of prices, with, at every level, some specific regulations and policies for everymember country and every particular product. One may therefore define: 1 Tax-free prices, which (at the final stage, that of distribution) include all increments of prices, except indirect taxes (excise duties and VAT); at the final stage of distribution, a tax-free price may be defined as the theoretical price that would prevail in the absence of indirect taxes. This is the sum of first, the tax-free price charged by refiners to wholesalers (that is, in general, close to the international spot price); second, the wholesalers’ profit margin (which is the difference between the tax-free price charged to retailers and the tax-free price charged by refiners); and third, the retailers’ profit margin (which is the difference between the tax-free price at its final stage and the tax-free price charged to the retailers). 2 VAT-exclusive prices, that may be defined as the sum of tax-free prices as defined above (at final stage of distribution) and of excise duties: these constitute the basis on which VAT is calculated. 3 Final all-inclusive retail prices, that may be defined as the sum of VAT-exclusive prices and VAT. These are final prices paid by consumers to retailers. At the level of tax-free prices, regulations differ widely among EC member states, and the following situations may be encountered:

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

153

● prices may be completely deregulated for refiners, wholesalers and retailers (France, Federal Republic of Germany, the Netherlands and the United Kingdom), ● a ceiling may be imposed for final tax-free prices by some indexation scheme based on international spot prices; this amounts to regulating profit margins (Belgium, Denmark, Irish Republic and Luxembourg); ● ceilings may be imposed for both final tax-free prices and tax-free prices at the refiners’ level. In this case excise duties are used as a fiscal instrument for the regulation of wholesalers’ and retailers’ profit margins (Greece, Italy, Portugal and Spain). The evolution of VAT and excise duties between 1979 and 1991 for all EC member countries appears in Tables 11.1 to 11.5 for five oil products: premium petrol, unleaded petrol, diesel oil, heating oil and heavy fuel oil. These tables show that there is a great variety in situations as far as both excise duties and VAT are concerned. As far as VAT is concerned, the rule that prevailed for motor fuels, back in 1979, was that it used to be at its ordinary rate for all countries of EC-9; as far as heavy and light fuel-oils were concerned, situations were somewhat different, because VAT was at a reduced rate for some countries (Belgium, UK), at its ordinary rate for Denmark, the Federal Republic of Germany, France, Italy and the Netherlands; furthermore, there was no VAT at all in Ireland and Luxembourg. After 1979, the situation changes (Tables 11.1–11.5) and there is either an increase in the ordinary rate of VAT (Denmark, France, Ireland, Italy and the Netherlands) which applies to motor fuels (petrol and diesel oil) and, in several other countries, to heating fuels, or a shift from ordinary rate of VAT towards an increased rate (Belgium, Luxembourg). Greece applies an increased VAT rate to motor fuels and heating oil since it entered the EC. Spain and Portugal apply, respectively, ordinary and reduced VAT rates to all oil products under study since they entered the EC. However, we can conclude that, in most member countries, an important increase in VAT has taken place, for most products, between 1979 and 1991. Moreover, one can see that these rates vary enormously for motor fuels (8% to 36%), as well as for high fuel-oils (0% to 36%) and heavy fuel-oils (8% to 36%). Table 11.6 shows the evolution of excise duty proceeds, in real terms, for EC-9 member countries. One can see that these Table 11.1 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on petrol (premium) within the EC (1979–91) Petrol (premium (1,0001)

1979 1980 1981 1982 1983

VAT ED VAT ED VAT ED VAT ED VAT ED

Bel.

Den.** Ger.

Fra.

Irl.

Italy

Lux.

NL

UK

Greece Spain** Port.**

16% 8460 25% 8400 25% 9400 25% 10200 25% 11200

20% 1820 22% 1920 22% 1920 22% 2250 22% 2265

18.6% 1456 17.6% 1456 17.6% 1508 18.6% 1659 18.6% 1889

10% 95 10% 135 15% 165 18% 188 23% 227

12% 346382 18% 356970 18% 397530 20% 438300 20% 545040

— — — — 5% 6960 10% 8460 12% 8460

18% 489 18% 532 18% 577 18% 591 18% 628

15% 81 15% 100 15% 138 15% 155 15% 163

— — — — — — — — — —

13% 440 13% 440 13% 510 13% 510 13% 510

— — — — — — — — — —

— — — — — — — — — —

154

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

Petrol (premium (1,0001)

1984 1985

1986

1987

Bel.

Den.** Ger.

Fra.

Irl.

Italy

Lux.

NL

UK

Greece Spain** Port.**

VAT ED VAT ED

25% 11200 25% 11200

22% 2365 22% 2265

14% 510 14% 530

18.6% 3228 18.6% 2622

23% 238 23% 256

20% 632540 18% 641630

12% 8960 12% 8960

19% 726 19% 732

15% 172 15% 179

— — — 47074

— — 12% 27500

— — 8% 4036

VAT

25%

22%

14%

25%

18%

12%

20%

15%

—

12%

8%

ED

1120 0 25%

2700

530

18. 6% 2796

274

8960

733

194

55286

14%

12%

20%

15%

36%

2750 0 12%

3829

22%

7989 30 18%

2700

530

8960

844

194

35642

22%

14%

12%

20%

15%

36%

9960

854

204

31427

12%

18. 5% 861. 90 18. 5% 875. 22 18. 5% 881. 60

15%

36%

204. 40 15%

25869

VAT ED

1988

VAT ED

1989

VAT ED

1990

VAT ED

1991

VAT ED

1120 0 25% 1120 0 25%

2700

530

22%

14%

1240 0 25%

3653. 31 22%

650

1385 0 25% 1385 0

14%

3080

650

22%

14%

2900

670

18. 6% 2885 18. 6% 2979 18. 6% 3055. 30 18. 6% 3131. 10 18. 6% 3179. 60

25% 281 25% 295

8233 30 19%

25%

8474 40 19%

303. 50 25%

8230 93 19%

9960 12%

***

303. 50 23%

8719 98 19%

9960 12%

***

303. 50

9451 50

9960

221. 40 15% 224. 80

36% 39900 36% 64535

3700 0 12% 3700 0 12% 3700 0 12% 3970 8 12% 4850 0

8% 3525 8% 3525 8% 3525 8% 3920 8% 9700

Notes: *February 1991. **These countries impose special taxes other than VAT and excise duties (2.30% tax, until April 1990, and 2.50% after April 1990, in Denmark; ‘Renta’ in Spain, ‘I.S.P.’ in Portugal (these taxes have been integrated into the ED as of 1 January 1991). ***In Ireland VAT decreased from 25% to 23% on 1 July 1990, and from 23% to 21% on 1 March 1991. Source: EC DG XVII. Table 11.2 Amounts of VAT (%) and of excise duties (ED) (National Currency Units) on unleaded petrol (Euro Super 95 RON) within the EC (February 1991) Petrol (unleaded) (1,0001)

VAT ED

Bel.

Den.* Ger.

Fra.

Irl.

Italy

Lux.

NL

UK

Greece Spain

Port.

25% 12400

22% 2250

18.6% 2817.40

21%** 277.40

19% 882120

6% 8960

18.5% 811.50

15% 194.90

36% 63733

8% 8500

14% 600

12% 43500

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

155

Petrol (unleaded) (1,0001) Bel.

Den.* Ger.

Fra.

Irl.

Italy

Lux.

NL

UK

Greece Spain

Port.

Notes: * In Denmark, there is another tax of 2.50% on oil products. ** On 1 March 1991. Source: EC DG XVII

proceeds have strongly increased between 1975 and 1986 for most member countries (except Belgium, Italy and the Netherlands) in spite of fuel savings efforts after the ‘oil shock’ of 1979. However, the consumption of motor fuels, which are the most heavily taxed products as far as excise duties are concerned, has increased over that period. After 1985, excise duties have increased steadily in most countries, especially as far as motor fuels are concerned. Table 11.7 shows their evolution between December 1985 and February 1991 for all countries of EC-12 (including Greece, Spain and Portugal). It appears, from this latter Table, that reactions to the ‘reversed oil shock’ of 1985–6 differed widely within the EC, with an overall tendency to increased taxation, at least in nominal terms, except perhaps for Luxembourg, whose fiscality in oil products remains light throughout that period. In conclusion, we can see that most EC member countries reacted to the second oil shock through an overall increase in taxation of all oil products studied. This increase concerns both VAT and excise duties, but EC member countries’ reactions appeared to be quite diverse and, apparently, did not proceed from any general harmonization scheme for these indirect taxes. Therefore (Table 11.8) prices for all products under study vary widely. Tax harmonization on oil products within the EC is, consequently, a very important and topical issue for the implementation of the 1993 Single European Market, the more so that legislations and regulations concerning the markets for oil products are also quite diverse within the EC. In the next part of this work, we shall therefore discuss possibilities for tax and policy harmonizations concerning oil products within the Community. 11.III POLICY AND TAX HARMONIZATION CONCERNING OIL PRODUCTS: SOME FUNDAMENTAL ISSUES As far as regulations concerning markets for these products are concerned, the rational for the Single European Market will probably lead to complete deregulation of tax-free prices, as now practised by four member countries (Federal Republic of Germany, France, the Netherlands and United Kingdom). This deregulation is likely to lead to homogeneous tax-free prices, specially for refiners. As far as tax systems are concerned, harmonization leads to quite complex problems because, as discussed previously, indirect taxes on oil products vary considerably within the EC. Moreover, tax harmonization means harmonization of policy objectives. While some of these objectives (energy savings policies and budget equilibrium) underlie heavier taxation, other objectives (such as revenue policies, and inflation containment) may underlie lower taxation. Let us now examine the implications of these two groups of policy objectives.

156

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

Table 11.3 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on diesel oil within the EC (1979–91) Diesel oil (1,0001)

1979 1980 1981 1982 1983 1984 1985

1986

1987

1988

1989

1990

1991

Bel.

Den.** Ger.

Fra.

Irl.

Italy

Lux.

NL

UK

Greece Spain** Port.**

VAT ED VAT ED VAT ED VAT ED VAT ED VAT ED VAT ED

16% 3800 25% 3450 25% 4250 25% 4250 25% 5250 25% 5250 25% 5250

20% 300 22% 360 22% 360 22% 360 22% 360 22% 360 22% 480

13% 417 13% 417 13% 442 13% 442 14% 442 14% 442 14% 442

17.6% 764 17.6% 764 17.6% 815 18.6% 883 18.6% 1003 18.6% 1214 18.6% 1318

10% 39 10% 79 15% 117 18% 132 23% 161 23% 172 25% 189

14% 25050 14% 25050 15% 16300 15% 56390 15% 107650 15% 116350 18% 129128

— — — — 5% 2000 10% 2800 12% 4300 12% 4300 12% 4300

18% 186 18% 186 18% 192 18% 201 18% 200 19% 197 19% 197

15% 92 15% 100 15% 119 15% 133 15% 138 15% 145 15% 154

— — — — — — — — — — — — — 7800

VAT

25%

22%

14%

25%

18%

12%

20%

15%

36%

12%

8%

ED

5250

1850

442

18. 6% 1474

216

4300

198

164

7070

25%

22%

14%

12%

20%

15%

36%

1800 0 12%

—

VAT

2793 40 18%

ED

5250

1760

442

4300

288

164

7592

VAT

25%

22%

14%

12%

20%

15%

36%

ED

5250

1760

442

4300

288

173

8628

VAT

25%

22%

14%

12%

15%

36%

ED

6550

1760

173

6957

VAT

25%

22%

444. 10 14%

18. 5% 285. 60 18. 5% 402

15%

36%

1800 0 12%

187. 32 15%

1441 2 36%

2187 5 12%

190. 20

1476 9

3230 0

ED

8267

1760

VAT

25%

22%

444. 10 14%

*

ED Notes:

8600

1760

444. 10

18. 6% 1474 18. 6% 1532 18. 6% 1572. 50 18. 6% 1608. 25 18. 6% 1635. 80

25% 216 25% 223 25% 223 25%

3047 20 19% 3626 20 19% 3753 10 19%

4300 12%

***

223 23%

4943 90 19%

4300 12%

***

223

5601 40

4300

18. 5% 401. 40

— — — — — — — — — — — — — —

1800 0 12% 1800 0 12%

— — — — — — — — — — — — — —

8% — 8% — 8% — 8% — 8% 47730

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

157

* February

1991. countries impose special taxes other than VAT and excise duties (2.30% tax, until April 1990, and 2.50% after April 1990, in Denmark; ‘Renta’ in Spain, ‘I.S.P.’ in Portugal (these taxes have been integrated into the ED as of 1 January 1991). *** In Ireland VAT decreased from 25% to 23% on 1 July1990, and from 23% to 21% on 1 March 1991. Source: EC DG XVII. ** These

Table 11.4 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on heating oil within the EC (1979–91) Heating oil (1,0001)

1979 1980 1981 1982 1983 1984 1985

1986

1987

1988

1989

1990

VAT ED VAT ED VAT ED VAT ED VAT ED VAT ED VAT ED

Bel.

Den.**

Ger.

Fra.

Irl.

Italy

Lux.

NL

UK Greece

Spain**

Port.*

6% 450 6% — 17% — 17% — 17% — 17% — 17% —

20% 300 22% 360 22% 360 22% 360 22% 360 22% 360 22% 360

13% 17 13% 17 13% 17 13% 17 14% 17 14% 17 14% 17

17.6% 145 17.6% 145 17.6% 146 18.6% 146 18.6% 164 18.6% 272 18.6% 354

— 4 — 15 — 15 — 15 5% 18 5% 18 10% 18

14% 25050 14% 25050 15% 16300 15% 56390 15% 107650 15% 116530 18% 116350

— — — — 5% 380 5% — 6% — 6% — 6% —

18% 34 18% 34 18% 34 18% 34 18% 35 19% 35 19% 35

— 7 — 8 — 8 — 8 — 8 — 8 — 11

— — — — — — — — — — — — — —

— — — — — — — — — — — — — —

18. 6% 374

10%

18%

6%

20%

—

—

12%

—

18

—

35

11

16212

5600

—

18. 6% 383

10%

2146 40 18%

6%

20%

—

36%

12%

—

—

125

11

7070

5600

—

18. 6% 396

10%

2793 40 19%

6%

20%

—

36%

12%

—

—

121

—

8628

5600

—

18. 6% 405. 60 18. 6%

10%

3626 20 19%

6%

—

6%

12%

—

3753 10 19%

—

18. 5% 118. 80 18. 5%

11

6957

9000

—

—

6%

12%

—

VAT

17%

22%

14%

ED

—

1540

17

VAT

17%

22%

14%

ED

—

1850

17

VAT

17%

22%

14%

ED

—

1760

17

VAT

17%

22%

14%

ED

—

1760

58.20

VAT

17%

22%

14%

37

37

37.30 10%

6%

— — — — — — — — — — — — — 7800

****

158

1991

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

ED

—

1760

58.20

VAT

17%

22%

14%

ED

—

1760

58.20

*

414. 70 18. 6% 422

37.30 10% 37.30

4943 90 19%

—

5601 40

—

6%

122. 20 18. 5% 121. 70

11.67

14412

9417

—

—

8%

12%

—

11.80

14769

1000 0

—

Notes: * February 1991. ** These countries impose special taxes other than VAT and excise duties (2.30% tax, until April 1990, and 2.50% after April 1990, in Denmark; ‘Renta’ in Spain, ‘I.S.P.’ in Portugal (these taxes have been integrated into the ED as of 1 January 1991). *** In Portugal, light fuel oil and diesel oil are considered as being the same product. **** This VAT was increased from 6% to 8% on 1 May 1990. Source: EC DG XVII. Table 11.5 Evolution of VAT (%) and of excise duties (ED) (National Currency Units) on heavy fuel oil within the EC (1979–91) Heavy fuel oil (1,0001)

1979 1980 1981 1982 1983 1984 1985

1986

1987

1988

Bel.

Den.**

Ger.

Fra.

Irl.

Italy

Lux. NL

UK

Greece Spain**

Port.**

VAT ED VAT ED VAT ED VAT ED VAT ED VAT ED VAT ED

6% 100 6% 100 17% — 17% — 17% — 17% — 17% —

20.25% 340 22% 340 22% 410 22% 410 22% 410 22% 410 22% 540

13% 15 13% 15 13% 15 13% 15 14% 15 14% 15 14% 15

17.6% 1 17.6% 1 17.6% 1 18.6% 47 18.6% 53 18.6% 59 18.6% 165

— 5 — 16 — 16 — 10 5% 10 5% 10 10% 10

14% 1000 15% 1000 15% 1000 15% 1000 18% 1000 18% 1000 9% 10000

— — — — 5% 100 5% 100 6% 100 6% 100 6% 100

7% 7 — 8 — 8 — 8 — 8 — 8 — 8

— — — — — — — — — — — — — 3500

— — — — — — — — — — — — — —

VAT

17%

22%

14%

ED

—

2080

15

VAT

17%

22%

14%

ED

—

1980

15

VAT

17%

22%

14%

18. 6% 297 18. 6% 185 18. 6%

18% 15 18% 15 18% 15 18% 11 18% 11 19% 11 19% 11

— — — — — — — — — — — — — —

10%

9%

6%

20%

—

—

12%

8%

8

1000 0 9%

100

36

8

14017

100

121

6%

20%

—

6%

12%

8%

1000 0 9%

100

36

8

8986

100

—

6%

20%

—

6%

12%

8%

10% 9 10%

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

1989

1990

1991

ED

—

1980

15

129

9

VAT

17%

22%

14%

10%

ED

—

1980

VAT

17%

22%

131. 82 14%

ED

—

1980

VAT

17%

22%

ED

—

1980

18. 6% 131. 82 18. 6% 134. 57 18. 6% 136. 70

134. 57 14%

*

136. 70

7.96 10% 7.76 10% 7.66

1000 0 9%

100

40

8

14116

100

—

6%

—

6%

12%

8%

2208 3 9%

100

18. 5% 40

7.81

7304

1700

—

7.5%

6%***

12%

8%

7666 7 9%

100

8.28

8506

1700

—

15%

8%

12%

8%

9000 0

100

8.37

11793

1700

—

6%

6%

18. 5% 47 18. 5% 47

159

Notes: *February 1991. **In Spain and Portugal, there were special taxes other than VAT and Excise Duties until December 1990 (‘Renta’ in Spain, ‘I.S.P.’ in Portugal). ****This VAT was increased from 6% to 8% on 1 May 1990. In all countries, VAT on heavy fuel oil is deductible; the same is true of excise duties (1980 Danish Crowns) in Denmark. Source: EC DG XVII. Table 11.6 Evolution of excise duty proceeds on oil products (in real terms, deflator: consumers’ price index) within EC-9 (1975–86) Bel.

Den.

Ger.

Fra.

Irl.

1975 100 100 100 100 100 1980 85 117.8 142.5 101.4 141.1 1981 72.3 101.3 124.5 99.1 154.4 1982 74.5 99.5 127.6 97.6 134 1983 77.8 103.1 131.4 97.2 137.7 1984 75.3 101.3 137.7 98.2 133.4 1985 74.2 74.2 169.8 98.4 134.3 1986 77.2 77.2 184.6 104 147.1 Source: Bragard, Decourcy and Schwed 1989; OECD 1988.

Italy

Lux.

NL

UK

100 90.8 82.8 81.2 101.5 104.5 100.6 —

100 220.6 759.3 201.7 167.7 152.8 143.2 152.5

100 94.5 86.7 81.2 86.6 91.1 87.3 92.6

100 115.5 133.7 146.7 150.6 159.2 159.4 184.8

Table 11.7 Evolution of excise duties between December 1985 and February 1991 within EC-12 Premium petrol Bel. Den. Ger. Fra. Irl. Italy

Diesel oil

Heating oil

Heavy fuel oil

+23.6% +63.8% no change (no duty) no change (no duty) +28.0% +266.7% +266.7% +266.7% +26.4% +0.4% no change no change +21.2% +21.2% +19.2% −17.2% +18.6% +18.0% +107% −23.4% Retail prices are frozen with only period adjustments, so that ED are adjusted with variations of international prices, except for heavy fuel oil

160

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

Lux. NL UK Greece Spain Port.

Premium petrol

Diesel oil

Heating oil

Heavy fuel oil

+11.1% +21.9% +25.6% +37.5% +76.4% +2303%

no change +103.8% +23.5% +89.3% +79.4% No duty in 1985, goes up to 47730 ESC in 1991*

no change +267.7% +47.5% +89.3% +78.5%*

no change +327.3% +47.7% +236.9% 1600%* No duty since 1986

**

Notes: *These figures are of little interest since taxes other than VAT and ED existed in these countries until 1989 (Spain) and 1990 (Portugal). Furthermore, this evolution is between December 1986 and February 1991. **In Portugal, high fuel oil and diesel oil are considered as being the same product. Source: Bragard et al. 1989.

11.III.1 Energy demand management and environmental policies, and budgetary equilibrium constraint Indirect tax systems may be used to encourage energy savings and, as far as oil products are concerned, substitution of other energy sources to oil products. Indirect taxation is also a tool for environmental policies, because Table 11.8 Retail and wholesale prices (ECU, upper line) and tax-free prices (ECU, in parentheses) for the oil products under study (February 1991) Wholesale prices Retail prices

Belgium Denmark Germany France Ireland Italy

2000 to 50001

less than 2000 tonnes per month

Petrol (premium)

Petrol (unleaded) Diesel oil (Euro super 95 RON)

Heating oil

Heavy fuel oil

704.44 (235.05) 741.17 (224.28) 617.56 (214.80) 751.37 (177.73) 807.35 (261.98) 997.45 (230.42)

664.12 (237.18) 647.26 (231.72) 544.18 (184.53) 730.54 (212.13) 778.76 (272.02) 964.96 (244.07)

263.75 (225.32) 597.76 (254.35) 316.39 (248.97) 394.60 (215.78) 301.21 (225.39) 731.55 (250.76)

70.72 (70.72) 345.07 (85.15) 120.58 (105.93) 93.26 (73.67) 137.74 (127.79) 178.69 (120.21)

611.94 (285.57) 654.87 (299.99) 557.61 (272.41) 558.58 (236.50) 735.89 (308.36) 737.53 (255.78)

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

161

Wholesale prices Retail prices

Luxembourg Netherlands UK Greece Spain Portugal

2000 to 50001

less than 2000 tonnes per month

Petrol (premium)

Petrol (unleaded) Diesel oil (Euro super 95 RON)

Heating oil

Heavy fuel oil

526.55 (235.05) 710.27 (227.51) 619.79 (219.52) 638.20 (177.01) 671.64 (220.90) 833.12 (237.65)

474.37 (235.05) 689.04 (229.68) 577.02 (224.97) 592.61 (205.00) 648.21 (239.03) 777.57 (247.88)

241.93 (228.17) 364.45 (254.81) 218.67 (201.90) 305.42 (215.78) 359.25 (242.65) same as diesel oil

129.33 (126.96) 125.66 (105.30) 112.50 (100.61) 130.84 (68.97) 112.05 (98.77) 154.28 (131.56)

384.24 (240.98) 530.86 (273.88) 613.68 (263.43) 305.42 (215.78) 570.89 (257.47) 572.07 (264.59)

Notes: * VAT-free prices. Exchange rates (15.02.91): 1 ECU=42.1606 FB=7.8793 CD=2.0483 DM=6.9765 FF =0.7696 RL£=1358.92 Lire=42. 1606 Flux=2.3075 Fl=0.7037 UK£=219.366 Dr= 128.044 Pes. Source: EC DG XVII.

Figure 11.1 Evolution of proceeds with respect to taxation

taxes on energy are used to compensate negative externalities (such as pollutions and CO2 emissions) due to the use of fossil fuels. The same is true of the institution of lower indirect taxes on unleaded petrol, which should discourage the use of motor fuels which give rise to emissions of toxic. Here, the argument for taxation of lead compounds is similar to that used for justifying taxation of tobacco and alcohol, as a compensation for negative externalities resulting from the consumption of these products. As far as budgetary equilibrium and fiscal proceeds are concerned, if one accepts the hypothesis that these proceeds are

162

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

small with respect to total fiscal proceeds, one can argue in terms of partial equilibrium, so that the evolution of proceeds with respect to the rate of taxation is represented by an inverted U-shaped curve (Figure 11.1). It is clear from this graph that a given level of fiscal proceeds (R) may be obtained with two rates of taxation, one corresponding to a higher level of consumption and the other corresponding to a lower level of consumption. There also is a taxation rate (t*) corresponding to maximum fiscal proceeds (R*) since the first objective is not maximizing fiscal proceeds. Table 11.9 gives the apparent elasticities of demand VS-price, calculated from the MIDAS model, for gasoline and diesel oil. These elasticities, as derived from a macro-economic modelling point of view, take into account the elasticity of the number of vehicles with respect to price and the elasticity of consumption with respect to the number of vehicles. These Table 11.9 Apparent elasticities of demand (private consumption) 1995 Country

Petrol

Diesel oil

Liquid fuels

Germany France Italy The Netherlands Belgium UK

−0.20 −0.80 −0.57 −0.32 −0.47 −0.27

−0.61 −0.12 −0.83 −0.25 −0.16 −0.16

−0.72 −0.92 −0.79 −1.06 −0.94 −0.63

elasticities are smaller for petrol and diesel oil than for other liquid fuels, for which an energy conservation policy appears to be compatible with smaller duties. As far as budgetary equilibrium is concerned, it is clear that the level of budgetary deficit appears as an objective for government policies. Excise duties on oil products, since they are relatively small, will not, however, play a determinant part in this respect (Meulders 1988). 11.III.2 Other objectives Increasing prices for oil products is a burden for the lower revenue bracket, and a redistribution objective may be carried out through indirect taxation. However, this objective leads to taxation policies that are conflicting with energy demand management policies, because excise duties must be high in order to be effective in terms of incentiveness to carry out energy savings, because of the low price elasticity of energy products. 11.III.3 Possibilities for indirect tax harmonization on oil products within the EC In 1989 and 1991, the European Commission sent to the Council some new proposals for tax harmonization in terms of VAT and excise duties, following the recommendations of a task group composed by experts of the ministries of finance of the member countries, which was to evaluate previous proposals and amendments thereto. The conclusions of this task group were that member countries had to be supplied with

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

163

details concerning fiscal measures, in order to assess their financial, budgetary, economic and social consequences, and to be clearly informed for the final decisions. In June 1991, EC-12 members had reached an agreement concerning excise duty harmonization, while VAT harmonization was still under discussion. Table 11.10 presents the proposals of the Commission with regard to excise duties, and their evolution since 1987. At first sight, one can see that the proposals of 1991 differ little from the proposals of 1987, except, perhaps, for diesel oil and heating oils, for which more flexibility is proposed for excise duties, since those are supposed to vary within some interval. In 1987, on the contrary, no such interval was proposed, so that flexibility in the harmonization process was to be achieved through VAT, which are calculated on a broader basis, so that its effects on the budgets of member states are larger. In order to determine excise duties, the approach of the Commission had been, in 1987, to ensure some equity between member states, with changes as small as possible for each product (EC 1987a, 1987b, 1989). The method used by the Commission in order to determine excise duties on oil products had been the following: the Commission used to propose, for each of the Table 11.10 Harmonization proposals of the EC Commission for excise duties on oil products (1987 and 1991) Products (1000 litres)* Harmonization products of 1987 Harmonization proposals of 1991 Objectives 1991 (ECU) (ECU) (ECU) Premium petrol 340 Kerosene 340 Unleaded petrol 310 LPG 85 Diesel oil 177 Heating oil 50 * Heavy fuel oil 17 Notes: *1000 kg for heavy fuel oil. Source: EC DG XVII.

337 337 287 85 245–270 47–53 16–18

495 495 445 85 / / /

main product categories, a rate that minimized budgetary effects and economic effects in terms of industrial costs. For instance, for petrol, which is by far the most important product in terms of fiscal proceeds, the taxation rate chosen rate was simply the average of existing rates. However, for diesel oil, the Commission’s opinion was that a weighted average (on the basis of consumption) was more appropriate, because it minimized this effect of harmonization on industrial costs. In all cases, the 1987 proposals led to harmonization on the basis of average values, whether weighted or not. This means that motivations concerning energy and environmental policy harmonization were absent in the Commission’s proposals. At that time the Commission was essentially aiming at reaching a politically acceptable compromise (in terms of budgetary impacts) between the member states. The 1989–91 proposals seem, however, somewhat more elaborate, in as much as other motivations appear to be underlying them. These motivations are linked to some kind of a European environmental policy, but they appear practically only as long-term objectives in terms of some excise duty rates which appear to be (especially for premium and unleaded petrol) much higher than the rates proposed for immediate application. These higher rates for gasoline include an increase of some 150 ECU/10001, which is presented as a correction for inflation, while an increase of some other 50 ECU/10001 is presented as a

164

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

tax on CO2 emissions. Practically, some environmental motivation appears for immediate application, through an excise duty smaller (by 50 ECU/10001) on unleaded petrol than the excise duty on premium petrol. Table 11.11 shows the variations (in terms of percentage) implied by the Commission’s proposals for eleven member countries. This table is to be compared to Tables 11.6, 11.7 and 11.8, in which it appeared that the energy Table 11.11 Percentage increases for excise duties on some oil products resulting from the 1991 harmonization proposals Country

Premium petrol (Proposition: 337 ECU)

Belgium +2.5% Denmark −8.5% Germany +3.0% Greece +14.6% Spain −11.0% France −26.1% Ireland −14.5% Italy −51.5% Luxembourg +42.6% The Netherlands −11.8% UK +5.5% Note: *No excise duty before 1991.

Diesel oil (Proposition: 245 to 270 ECU)

Heating oil (Proposition: 50 ECU)

+20.1 to 32.3% +9.7 to 20.8% +13.0 to 24.5% +263.9 to 301.0% −2.9 to +7.0% +4.5 to 15.2% −15.4 to −6.9% −40.6 to −34.5% +140.2 to 170.0% +40.8 to 55.2% −9.4 to 0.2%

*

−77.6% +75.9% −25.7% −36.0% −17.3% +3.2% −87.9% *

−5.2% +198.2%

and fiscal policies conducted by EC countries had led to a great diversity in situations. One should, furthermore, notice that policies following decreases in prices that occurred in 1985 have led to results contrary to the principles of harmonization proposed by the Commission; this was the case in France and in Italy for three products, and, to a smaller extent, in the Netherlands for premium petrol and for heating oil. Even if environmental policies are not included in the discussion, one can see that, as was the case with the 1987 proposals, the recent proposals of the Commission are such that some countries, which used excise duties on energy policy and energy conservation policy tools will have to reduce their excise duties by amount which may be large; this may be a problem for these policies, and, in addition, this may lead to serious budgetary problems for these countries. If one takes into consideration the variations in the amount of excise duties, the importance of budgetary proceeds from excise duties and the general budgetary problems in various countries, one can see that the solution chosen by the Commission (harmonization on the basis of averages, in order to reduce budgetary effects) appears to be ineffective with regard to the Commission’s policy goals, since it implies severe budgetary problems for several member states. In the case of VAT, the proposals which have been done are part of a more general proposal concerning VAT harmonization between EC member states (EC 1987b, 1989). This VAT harmonization is a rather difficult political problem: a proposal on the basis of an average, or on the basis of the smaller VAT rates

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

165

practised by some member comities, would be politically very difficult to impose, since it implies direct taxation increases in some member states. This problem has not yet been solved through any politically acceptable compromise between member states. Some proposals which have, however, been discussed are orientated towards the implementation of two VAT rates only: a reduced rate between 5 and 7%, and a standard rate between 14 and 20%. In the particular case of oil products, the proposals which have been made would imply standard VAT rate for premium and unleaded petrols, for diesel oil and for heating oil, with no VAT on heavy fuel oil in these proposals. Since some discussion on VAT harmonization still appear to take place, while some definite proposals concerning excise duties have been elaborated, we only give here (Table 11.12) an estimate of retail prices that would prevail in the event of application of these harmonization proposals for excise duties, but keeping the present VAT rates (VAT-free prices being given for heavy fuel because VAT is deductible in that case). One can see that, despite the diversity in VAT rates applied, these harmonization proposals on the basis of excise duty harmonization alone lead to some convergence between retail and wholesale prices of the various oil products under study. 11.III.4 Discussion Even if the proposals that have been made are effectively apt to lead to a progress towards the solution of the problem of fiscal harmonization for the Single European Market of 1993, they may be criticized in as much as they are merely based on average or weighted average calculations for excise duties, the problem of VAT being, at the moment, not fully solved. In our opinion, it would be important if indirect tax harmonization could be based on an in-depth study of fiscal and budgetary effects and on some harmonization in energy conservation policies and in environmental policies (in as much as the intensive use of some energy sources, such as coal, fuel oil, and nuclear energy leads to environmental protection problems). As we have already noticed, several member countries, which used excise duties for their energy conservation policies, might be obliged to reduce their excise duties by important amounts, which might weaken their energy policies. This means that an overall study of this problem, in relationship with environmental and energy conservation policy concerns, might lead to propose, at least for some oil products, an harmonization of excise duties at a higher level than the one implied by an average, or even a weighted average, of the existing excise duties for various member countries. As far as VAT is concerned, the problem of its harmonization is largely a

166

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

Table 11.12 Retail and wholesale prices (ECU, upper line) and tax-free prices for the oil products under study after taking into account 1991 harmonization proposals (on the basis of VAT rates and international prices of February 1991) Wholesale prices Retail prices

Belgium Denmark Germany France Ireland Italy Luxembourg Netherlands UK Greece Spain Portugal

2000 to 50001

less than 2000 tonnes per month

Premium petrol

Unleaded petrol (Euro super 95 RON)

Diesel oil

Heating oil

Heavy fuel oil*

717.25 (235.05) 701.88 (224.28) 629.06 (214.80) 610.47 (177.73) 724.76 (261.98) 675.23 (230.42) 640.96 (235.05) 668.94 (227.51) 640.00 (219.52) 635.33 (177.01) 624.85 (220.90) 615.22 (232.65)

655.23 (237.18) 648.66 (231.72) 537.54 (184.53) 591.97 (212.13) 676.41 (272.02) 631.97 (244.07) 553.37 (235.05) 612.27 (229.68) 588.71 (224.92) 669.12 (205.00) 589.15 (239.03) 577.76 (247.88)

663.21 (285.57) 681.50 (299.99) 589.85 (272.41) 571.06 (236.50) 669.57 (308.36) 595.93 (255.78) 544.30 (240.98) 614.87 (273.88) 584.69 (263.43) 626.66 (215.78) 562.77 (257.47) 550.36 (264.59)

318.61 (225.32) 376.83 (254.35) 337.41 (248.97) 311.66 (215.78) 299.63 (225.39) 354.33 (250.76) 291.68 (228.17) 487.57 (254.81) 248.90 (201.90) 283.80 (215.78) 324.40 (242.65) 336.52 (264.59)

86.72 (70.72) 101.15 (85.07) 121.93 (105.93) 89.67 (73.67) 143.79 (127.79) 136.21 (120.21) 142.96 (126.96) 121.30 (105.30) 116.61 (100.61) 84.97 (68.97) 114.77 (98.77) 147.56 (131.56)

Notes: * VAT-free prices. Exchange rates (15.02.91): 1 ECU=42.1606 FB=7.8793 CD=2.0483 DM=6.9765 FF =0.7696 RL£=1358.92 Lire=42. 1606 Flux=2.3075 Fl=0.7037 UK£=219.366 Dr=128.044 Pes. Source: EC DG XVII.

macro-economic problem, so that VAT is less apt to become a tool for an energy conservation policy at the EC level. Finally, one should notice that for some observers, no tax harmonization should a priori be required, because market constraints will by themselves impose to the member states some harmonization of their indirect tax systems, on oil products as well as on any other products. This view has been held by some

FISCAL HARMONIZATION ON OIL PRODUCTS WITHIN THE EC

167

member countries, such as the UK, during the discussion. A similar approach appears to prevail in federal states such as USA and Canada. It is not, however, in our opinion, in contradiction with the fact that energy conservation policies and environmental policies of member states should reach some harmonization at the level of the EC. 11.IV CONCLUSION The present disparities between indirect taxes on oil products within the EC is, to a large extent, the consequence of the fact that energy conservation policies, environmental policies, and general budgetary policies of member countries differ widely. Under these conditions, the indispensable tax harmonization, which seems necessary for the development of the single European Market, should, quite logically, result from the harmonization of energy conservation policies and of environmental policies within the EC. Such an harmonization, if it leads to the definition of some coherent European policy in these areas, should then lean on excise duties as far as indirect taxes will be used as policy tools, VAT being more of a macro-economic policy tool. Taking into account these elements, it might be useful to study further possible orientation of energy conservation and environmental policies at the EC level. Within such a discussion, a definition of adequate tax levels on fuels and oil products would quite naturally take place. REFERENCES Bragard, G., de Courcy, P., Schwed, N. La fiscalité du tabac, de l’alcool et des huiles minérales dans la Communauté Européenne et son harmonisation dans la perspective du marché unique de 1992. École de Commerce Solvay, Brussels, 1989. Cnossen, S. (Ed.) Comparative Tax Studies. Essays in Honor of Richard Good. North Holland, Amsterdam, 1983. EC. Proposition de Directeur du Conseil concernant le rapprochement des droits d’accises sur les huiles minérales. Document COM (87)327, CEE, Brussels, 1987a. ——. Proposition de Directive du Conseil complétant le système commun sur la Valeur Ajoutée (rapprochement des taux de T.V.A.). Document COM (87)321, CEE, Brussels, 1987b. ——. Proposition de Directive du Conseil instituant un processus de convergence des taux de Taxe sur la Valeur Ajoutée et les accises. Document COM (87)324, CEE, Brussels, 1987c. ——. Proposition de Directive du Conseil concernant le rapprochement des taux d’accises sur les huiles minérales. Document COM (89)256, CEE, Brussels, 1989. Meulders, D. Les politiques ftscales menées dans les pays de la CEE suite aux variations des prix de l’énergie.Document de travail, DULBEA, Brussels, 1988. Neyrinckx, Y. L’harmonisation des droits d’accises. Évolution et comparaison du niveau de la CEE. École de Commerce Solvay, Brussels, 1988.

Chapter 12 Energy tax Increases as a way to reduce CO2 emissions V.Detemmerman, E.Donni and P.Zagamé

INTRODUCTION The present work is part of a broader research programme on options to reduce CO2 emissions, undertaken in the framework of the JOULE programme from the Directorate General for Science, Research and Development (DG XII). It deals mainly with energy and macro-economic impacts of a tax increase on energy as a means to reduce CO2 emissions. This work was performed with the help of a network of macro-economic models HERMES developed and managed by DG XII. The study covered four countries (Germany, France, Italy and United Kingdom) and it was extended to the description of the accompanying measures which could be implemented to increase the efficiency of the tax in particular. ● emissions reduction measures ● general economic measures with impact on growth, competitiveness ● employment, etc. As usual when dealing with new questions it was necessary to modify some mechanisms in the model. At present, no distinction has yet been established between the different energy sources and, it is the impact of a general tax equally applicable to all energy products which is analysed. Further work will concentrate on differentiated taxes and on the choice of fuel mix thus on the required fuel switching. But now, with this approach, the only way for reducing CO2 emissions is to reduce energy consumption. The magnitude of the shock considered in this study was enough to produce significant1 behaviour reactions, i.e. to result in noticeable energy savings. A tax of 20% on the price of energy was considered. It will be seen that a tax of this magnitude on its own does not spontaneously lead to a considerable energy savings or substitution. Nevertheless, this results in a tax revenue which, for some countries, can reach 1.5% of GDP. This will spontaneously induce a recession spiral with immediate and damaging impacts on production and employment. These impacts are even emphasized since macro-economic models do not include reaction functions to describe the spontaneous behaviour of the administration when faced with budgetary deficits or surplus. It is then necessary to simulate accompanying measures where the administration will be expending the increased resources brought by the energy tax. In a first stage and merely for didactic purposes no accompanying measures have been considered and only the effects of increased taxation have been analysed. This allowed to study the adaptation behaviour of

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

169

the agents faced with new energy prices and to evaluate the revenue effects, the substitution effects and the energy intensity change. On the other hand, macro-economic consequences were also evaluated in terms of deflation, competitiveness, employment, budgetary surplus, etc. The second part was completely dedicated to the analysis of the accompanying measures, i.e. the redistribution of the budget surplus. Two possibilities were considered: subsidies which induce a more voluntary behaviour in terms of emissions reduction and which can be focussed on actions introducing new energy savings; and tax deductions affected to the households (income tax of individuals) or to firms. In this last case different possibilities can be considered: a deduction of the general corporate taxes of the firms or a more oriented deduction affecting investments, a reduction in the VAT or in the social contributions of the employers to encourage the utilization of a production factor (labour) etc. For this reason this work could not cover the globality of possible measures so only the impacts of a couple considered more opportune were evaluated. 12.I ADAPTATION BEHAVIOUR TO THE INCREASE IN ENERGY TAXES In this section the adaptation behaviour to the tax increase are analysed as well as the freezing of about 1.5% of GDP by the administration. It is clear that for the scenario chosen—a 20% increase of the final price2 of energy—and from a macroeconomic point of view ‘income’ effects are significant and substitution consequences are secondary. The first analysis reports to the spontaneous mechanisms in the model: but these mechanisms are mostly based on past behaviour. This is the econometric approach of transposing the past into the future, which is always a rather conservative approach. Nevertheless, events occurring during the last fifteen years induced several changes including in particular the agents ability to face fluctuations in the (real) price of energy which cannot be neglected. But these need to be better analysed since history does not repeat itself and several structural changes based on behaviour adaptation took place, e.g. a voluntary behaviour in terms of energy saving: increase in energy saving investments, fast stripping of old equipment with high levels of energy consumption and emissions, adoption of new and less pollutant processes, etc. This section has thus been divided in two parts: spontaneous behaviour and more voluntary behaviour in the adaptation to an energy tax increase. 12.I.1 Spontaneous behaviour Substitution effects, income effects and macro-economic induced effects resulting from the model’s mechanisms will be described. A distinction will be made on their area of application (firms, households, etc.). These effects are not new since they have been thoroughly analysed during the various oil crises, the difference now being that transfers are made to the benefit of the administration. Impact on energy consumption

A distinction will be established between production sectors and households. Production sectors

170

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

In this case another distinction will be made: between industrial sectors (consumption goods, equipment goods, intermediate goods) and other production sectors. In fact, the model describes the first by means of a ‘putty-clay’ production function which integrates the energy factor. This means it considers an ex ante substitution (at the margin, for new equipment) and an ex post complementary once the equipments have been installed. This is not the case with the other production sectors. As soon as the average price of energy increases, industrial branches of firms undertake substitution among the production factors depending on their past complementarities and substitutions. In a couple of countries (such as Italy and France) a complementary has been found between energy and investment. In the present case, that leads to a simultaneous decrease both in energy utilization and in capital per unit of production driven by an increase in the price of energy and therefore leads to an increase in employment. Once this combination has been adopted, firms register an increase in their production costs (nevertheless lower than in the case should no substitution have taken place) and income and inflation effects will be noticed due to the upstream repercussion of these increased prices. In any case, a reduction in the specific consumption of energy can only be obtained by means of their substitution. For the other sectors, the production function is rather straightforward paying no particular attention to the energy factor. Any eventual substitution taking place is related with a change in the relative cost of capital and labour, this latter being dependent on the evolution of the relative prices structure in the macro economic sphere. These effects are rather weak maybe with the exception of the transport branch since the model already integrates a substitution mechanism in which energy specific coefficients depend on energy prices. Household

In the HERMES model, the classification of the household consumption has been done with some detail and fifteen consumption categories have been considered to reach the total consumption: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Food, beverages and tobacco Clothing and footwear Gross rent Fuel for domestic utilization Electricity Domestic services Furniture, household appliances and furnishings Personal transport equipment Fuels for personal transportation Transport services Communications Medical services and other health expenses Leisure, shows, education and culture Hotels, restaurants, coffee-houses, financial services and travelling Expenses of residents abroad

These categories define a certain number of functions. Those we are particularly interested in pertaining to the direct consumption of energy: fuel for domestic utilization; electricity; petrol and lubricants. The consumption evolution of each of these products depends on their price as well as the trade-offs among the different functions. The mechanism in the model relies on substitutions among both the price of these three functions and that of other functions. In other words an increase in the prices of energy will induce

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

171

substitutions in benefit of functions, with no or low energy expenses. Results on the relation between energy consumption and the households expenses are presented in Tables 12.1 and 12.2. Economic impact

Economic impacts are first of all due to the large income effect resulting from a deduction of 1.5 to 2% on total income. Income effects are obtained mainly through the price increase: the real households income is ‘ex ante’ andautomatically cut by about 2% since direct and indirect expenses with energy represent about 10% of their usable income. Similarly, firms will not be able to immediately transfer the energy Table 12.1 Variation of the energy intensity in the production sector (differences in %)

1991 Energy intensity in the productive sector 1995 Energy intensity in the productive sector

Germany

France

Italy

UK

−1.61

−1.57

−1.48

−2.21

−3.69

−2.65

−4.10

−4.24

Table 12.2 Reduction of energy share in household consumption (differences in %)

1991 Energy consumption/total household consumption 1995 Energy consumption/total household consumption

Germany

France

Italy

UK

−2.95

−2.65

−4.10

−4.24

−5.80

−4.13

−2.60

−3.25

price increase in their production cost and will suffer from a profit reduction in constant prices. On the contrary, the administration improves its situation by an amount equal to the overall deduction. This will induce a recessive effect since households will reduce their consumption and firms their investment. In some countries this last aspect is reinforced by the complementary between energy and investment. Inflation will progressively increase and it will stabilize at a level compatible with the income deduction and the evolution of unemployment. This will induce a deterioration of competitiveness with different effects, depending on whether the measure is introduced in a single country or simultaneously in all countries.3,4 This deterioration reduces the contribution of the external balance to growth although this effect is not very significant since energy imports decrease. All these phenomena are rather well known and HERMES has the advantage of determining coherent evaluations for each of them. The mechanisms where effects are more relevant to the final results are the following: 1 First of all the indexation process: in some countries income is automatically indexed on the basis of the evolution of the consumption prices. In this case the deduction will affect mainly the firms. In practice this is not done but countries can be differentiated by their indexation delay time and by the elasticity of their indexation and by that of the unemployment variable in the salary formula. The evolution of the real households income and therefore that of consumption will depend on all these factors.

172

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

2 The income effect, which in the indexation case, affects the firms leads to very different results in terms of investment. Depending on the country, investment is strongly or weakly related with profit. 3 Similarly, the magnitude of the income deduction for firms is very dependent on the delay of the transfer of costs into production prices and this delay changes from country to country. 4 The competitiveness effect will depend on this transfer and on the behaviour of both imports and exporters in terms of price definition. 5 Finally and in the short term, it should be noticed that the level of production will strongly depend on the level of consumption. The results on Table 12.3 clearly describe this situation. 1 Already in the first year, GDP decreases significantly. This decrease is due mainly to the compression of households consumption (real income of households is diminishing because of the indexation lag) and of investment but with a less important effect. Table 12.3 Tax on final energy consumption

1991 GDP Private consumption GFCF Employment Consumption prices Net bonus of government* Current balance** Final energy consumption* FEC by household 1995 GDP Private consumption GFCF Employment Consumption prices Net bonus of government* Current balance* Final energy consumption* FEC by household

Germany

France

Italy

UK

−0.87 −1.33 −1.04 −0.50 2.04 −1.55 0.15 −3.12 −4.27

−0.72 −1.00 −0.47 −0.25 2.63 −1.20 0.20 −2.86 −3.81

−1.10 −1.06 −1.33 −0.33 2.37 −1.29 0.23 −2.60 −2.62

−1.44 −1.87 −0.10 −0.43 2.65 −1.88 0.46 −3.87 −4.70

−0.97 −1.39 −1.59 −0.60 2.71 −1.30 0.15 −5.59 −7.07

−1.30 −0.50 −0.87 −0.80 5.46 −1.50 0.10 −4.31 −4.93

−1.34 −1.52 −1.29 −0.82 3.74 −1.03 0.45 −4.78 −4.12

−2.12 −1.94 −2.50 −1.55 4.01 −1.76 0.48 −6.08 −5.44

Note: Charge per baseline in %, except* in point of % of GDP.

2 After the first five years, France and the United Kingdom appear detached from the other countries having more difficulty in absorbing the inflation shock caused by the energy price. France registers a price differential of 5.4% and the United Kingdom of 4.6%. The French trade balance becomes

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

173

negative due to a reduction of competitiveness and in spite of the strong GDP contraction (1.3%). The magnitude of the inflation evaluated for France is compatible with recent studies on oil shocks. 12.I.2 More voluntary behaviour From now on we shall be working in the domain of reasoned hypothesis describing behaviours not modelled by HERMES. This does not mean that they have never been observed but only that they have never been translated by an econometric relation as those contained in the model. This was due to diversified reasons including the absence of long statistical data series for the production sectors (investments for energy conservation for example), the introduction of new technology which may significantly alter the combination and relevance of the different production factors and even some changes in behaviour induced by previous oil crises. All these factors favour the study of non modelled behaviours and even more if we consider that previous results demonstrated the need of a more voluntary behaviour of adaptation in order to reach significant energy savings. The measures required to promote this type of behaviour will be discussed later. For the moment we shall address the simulation of different combinations of the production factors to evaluate their impact in terms of energy saving and reduced emissions. These simulations will be carefully done and moreover, they do not rely on econometric evidence. Some of the quantitative values may be challenged and revised depending on the opinion of other experts during the presentation of the study. The different simulations will be firstly described and the energy conservation investments will be further discussed in detail. Other cases are object of a specific study and will be presented elsewhere.5 Some expected behaviour modifications

We will start with energy conservation investments: these reduce energy consumption while the production capacity remains constant and so does the productivity of the factors (apart from energy). Therefore these equipments appear as additional non-productive equipments, not affecting the initial production function and allowing some energy savings. From the macro-economic point of view, they appear as an expense in supplementary capital from the firms perspective and as equipment goods delivered to the different sectors without the corresponding production equivalent, i.e., demand is increased but offer remains constant. They will function as ‘Keynesian Multipliers’. However, this is rather simplified methodology to treat energy conservation investments which in fact imply much more complex phenomena: there are other ways to reduce the level of emissions associated with energy consumption. A first example is the scrapping of equipment with high levels of consumption. This can be promoted by an increase in energy price sufficient to make this equipment obsolete from an economic point of view, i.e. the production combination is no longer justified in view of the new relative prices of the factors. These mechanisms of internalization of scrapped equipment by the relative price of the factors were initially considered in the HERMES model6 but their econometric relation was not sufficiently strong to be kept in the final version of the model. Although substantial energy savings can be reached in this way, the scrapping of the equipment induces a reduction of production capacity which needs significant and additional capacity investments. Returning to the subject of energy conservation, this can be applied only to the new invest ments or to the investments stocks existing before the introduction of the measure. In these two possibilities it will be necessary to evaluate the complementary investment amount and energy savings.

174

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

In the long term, we can also imagine modifications of the production function itself due to some endogenous technical development resulting in energy saving. This is hard to simulate and cannot obviously be done by econometric methods. This shall again rely on the personal opinion and estimates of experts. The case of energy conservation investments

For the following exercise it will be considered that firms will increase their investment on energy conservation by 1% of GDP. Mechanisms considered within the model to analyse the impact on a tax on energy on industrial sectors Spontaneous and voluntary behaviour 1 Substitution of the margin so that the coefficient of marginal production will depend on the relative price of the factors by means of an energy production function of the putty-clay type. 2 Higher rate of scrapping of old equipment too expensive in terms of energy. 3 Increasing energy saving investments, which do not change production capacity nor the combination of the factors labour and capital, nor the productivity of these factors. These investments can be directed to all equipment goods or be restricted to new ones. 4 In the long term, a modification in the production function through technical progress. 1 is a spontaneous behaviour 2 and 3 have been done 1 and 3 are presented in this note This is a considerable shock since it represents about 8% from the total of the gross fixed capital formation of the firms. This value was chosen so that the variables described in this study can be clearly evaluated and result in clear impacts, even if it is later necessary to proportionally reduce the results to adjust them to reality. This increase in the firms’ expenses of about 1% of GDP, when kept during the whole period, results in yearly energy savings of 3%, i.e., 15% at the end of the simulation period. In fact the energy-saving investment is decreasing return of scale and a more precise analysis ought to take account of that; probably the figure of 3% for energy savings is important for 1% of investment, but the rate of return would be more important for figures less than 1%. Some orders of magnitude (For France in 1988 and in billions of FF) Productive Sector Investment 550 Energy expenses 220 Expenses increase in the 1st year 50 Yearly energy savings 7 Pay-off period (in years) 7 Rough estimates are sufficient to demonstrate that by the application of such a hypothesis to all the productive sections in a country like France, the initial increase in expenses is recovered after seven years. This in fact is rather long and justifies the analysis of incentive measures described in the second part of this study.

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

175

The mechanisms activated by this variant are easy to describe: the increase in the expenses of firms results in an increase of demand directed both to the equipment goods sector and to the services sector. This amounts respectively to 0.8% and 0.2% of GDP. We were here forced to adopt a hypothesis which somehow limits the relevance of the conclusions: the sector which produces equipment goods allowing energy savings and so reduction of emissions has not been differentiated from the remainder of the equipment goods sector. This means that it was assumed for this sector a production structure similar to the average structure of the branch (for imports, exports, productivity, prices) when in reality it is well known that the production of such equipment is very specific. We can take for instance the case of Germany where the advance of this specific sector is much more pronounced than the advance of overall equipment goods sector. A demand increase without a counterpart in terms of productivity or supply (apart from energy savings) functions as a short-term Keynesian multiplier. Nevertheless, the investment cost of firms increases during the first years without a compensation from energy savings due to their difference in magnitude. Inflation impacts will depend on the way this cost increase will be transferred to final consumers and all agents in general who will bear this charge. Ignoring the mechanisms in model, we can identify three limiting cases:7 ● investment is integrally covered by the firms without any price consequence; ● investment is integrally covered by the public administration without any price consequence; ● final users will bear all costs. A combination case: the progressive repercussion of costs In this case the situation spontaneously described by the model will be analysed: firms will progressively transfer their costs into production prices. This is an intermediate case between the limiting situation where firms cover all the required finances and the other extreme when final users bear all costs. The remaining situation where the Administration bears the costs will be described in the second part of this study when dealing with subsidies.8 On Table 12.4 results of this hypothesis have been displayed: the increase in energy savings investments was considered as 1% of GDP and this amount is progressively transferred into production prices. The following can be verified: Significant ‘outward flows’ (trade balance) are observed due to the inflation increase already mentioned, and to a tendency to import supplementary equipment goods, the demand for this type of products increases by 8%. This explains why impacts multipliers in terms of GDP are rather weak, between 0.3 and 0.7%, while traditional Keynesian multipliers reach an almost unitary value after the first year. It is only after five years that inflation impacts start being significantly compensated by energy savings which already represent, for firms, 6 to 7%. The differences among the various countries must not be considered as significant. In particular the better performance of France compared to that of Germany is rather irrelevant. Had we differentiated the level of penetration Table 12.4 New behaviour towards energy-saving investments

1991 GDP Private consumption

Germany

France

Italy

UK

0.50 −1.11

0.70 0.48

0.33 −0.26

0.42 −0.10

176

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

Germany GFCF 4.66 Employment 0.15 Consumption prices 0.38 Net bonus of government* −0.16 Current balance* −0.44 Final energy consumption −0.45 FEC by household −0.14 1995 GDP 0.52 Private consumption 0.26 GFCF 4.97 Employment 0.20 Consumption prices 1.22 Net bonus of government* −0.22 * Current balance −0.50 Final energy consumption −4.14 FEC by household −0.26 Note: Charge per baseline in %, except* in point of % of GDP.

France

Italy

UK

5.72 0.16 0.46 −0.20 −0.81 −0.15 −0.35

6.86 0.17 0.88 −0.43 −0.58 −0.78 −0.41

4.62 0.27 0.64 −0.09 −0.06 −0.61 −0.16

0.71 0.85 4.49 0.28 2.12 −0.67 −0.66 −3.22 0.53

0.31 0.29 5.68 0.21 1.75 −0.51 −0.37 −4.41 −0.26

0.38 0.41 4.60 0.32 1.04 −0.21 −0.20 −4.02 −0.16

of equipment goods saving energy and the result would certainly be different. In short, a voluntary behaviour in terms of energy conservation immediately results in an inflation increase and significant negative trade balance effects. On the other hand and in the longer term it can nevertheless achieve significant energy savings and so increase in competitiveness. To avoid the drawbacks of this type of behaviour, we can imagine a subsidies system to reduce the burden of firms as well as the implementation of harmonized policies within Europe which would prevent some negative trade effects. 12.II ENERGY TAXATION AND ACCOMPANYING MEASURES A tax of 20% on the price of energy results in an overall reduction of about 1.5% of GDP and it has been concluded in the previous section that measures to redistribute this value are required to avoid a reduction of the economic activity and an increase in unemployment. A wide variety of measures are available ranging from subsidies to a reduction in the social contributions of the employers. Subsidies are generally directed towards very specific areas and tend to promote different behaviours, thus their intervenient character. Tax reductions on the contrary characterize an Administration which does not interfere with the process. These measures have quite different effects on their own and depending on the way they are applied. To clarify these aspects we start from the situation of an energy tax and its main drawbacks: ● reduction of production and employment

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

177

● price increase ● lower purchase power from the households ● results in terms of energy consumption and emission reduction which do not compensate the previous aspects. Subsidies promoting energy conservation and in particular energy saving investments would eliminate this final inconvenient and we shall start by employers also seems a very promising measure and this is also tested. It has analysing this measure. The reduction in the social contributions of the the benefit of attenuating the cost increase due to the new tax, allowing some substitution favouring employment and limiting energy consumption. These are in short the HERMES model conclusions for the industrial sectors.9 All measures studied will be analysed by the following order: first, the subsidies to energy conservation investments. Then two cases of tax reductions, a reduction in the social contribution of the employers and a reduction in the general income tax of the households. At the end a synthesis scenario combining different measures and reaching a good balance between the energy plus emissions targets and the macroeconomic constraints (competitiveness and employment) is also discussed. This has the advantage of using more realistic simulation values. 12.II.1 Subsidies: the case of energy conservation investments Results are presented for the simultaneous application of two measures: a tax of 20% on the 1991 prices of all oil products and its integral redistribution by means of energy conservation investments. To simplify this analysis, we shall start from the variant used in the previous section to describe a non subsidized increase in investment costs; the difference now being the finances by the Administration of part or all of the firms supplementary expenses. In this case this should result in an almost insignificant impact on prices. Apart from this effect, the same types of results can be found. In this exercise things have been balanced so that when the Administration finances the totality of the energy conservation investment it uses about the same value as that resulting from the 20% energy tax. Obviously this should result in a quite significant amount of investments but this situation was studied to clarify the results presentation. The more complex scenarios presented at the end will adapt these hypotheses to more realistic situations. As Table 12.5 shows, the growth loss brought up by the new tax income is recuperated in most countries. This indicates that the multiplier effect of the expenses on energy conservation investments does increase beyond unity in case of government intervention while these effects used to be smaller (between 0.5 and 0.7) in previous occasions. These results might be slightly overevaluated since eventual effects of eviction following an interest rate increase are not taken into account. Nevertheless, the combination of both measures annuls all costs linked to growth and employment and allows substantial energy savings. They attain in certain countries nearly 11% if one takes the increase of growth into account. Weak points—which play mainly an important role in the short and medium term—are due to inflation, fall in competitiveness and the deterioration of the international account for certain countries. This disadvantage is supposed to shade off, once the projection horizon is stretched because of decreasing energy consumption reducing firm costs. The fuel bill of firms is reduced by more than 30%, representing on average a 2% price decrease.

178

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

Table 12.5 Energy tax financing energy-saving investment Germany 1991 GDP 0.16 Private consumption −1.21 GFCF 6.01 Employment −0.06 Consumption prices 2.16 Net bonus of government* −0.20 Current balance* −0.35 Final energy consumption −3.81 FEC by household −4.15 1995 GDP 0.11 Private consumption −0.95 GFCF 4.44 Employment −0.11 Consumption prices 3.05 * Net bonus of government −0.20 Current balance* −0.33 Final energy consumption −10.66 FEC byhousehold −6.64 Note: Charge per baseline in %, except* in point of % of GDP.

France

Italy

UK

0.84 0.44 6.37 0.09 2.47 −0.20 −0.80 −2.93 −3.37

0.18 −0.72 8.86 0.12 2.90 −0.26 −0.65 −2.69 −2.34

0.10 −1.53 8.75 0.25 2.81 −0.20 0.16 −4.54 −4.63

0.61 0.90 6.23 0.21 5.84 −0.30 −0.94 −8.41 −3.76

−0.36 −0.67 5.13 −0.20 4.22 −0.54 −0.04 −7.00 −3.42

0.18 −0.46 6.90 −0.12 4.56 −0.30 0.21 −11.64 −5.30

If the measures had been simultaneously applied in different countries, multiplier effects, expressed in relative values, would have increased by 10 to 20%, creating more growth and more employment. Meanwhile the commercial balance would have become less depressed, increasing slightly costs in terms of imported inflation from other European countries. 12.II.2 The reduction of taxes We do not intend to start a general discussion concerning the advantages and disadvantages of all different ways to reduce contributions. The aim is to focus on two measures which seem to be fairly well adapted to avoid the negative side effects of contributions concerning employment, costs of firms, prices and purchase power of households, namely the reduction of social contributions from employers and the reduction of the personal income tax. The reduction of social contributions from the employers

Taken independently, the consequences of this measure are rather known in terms of the underlying mechanisms to the models: The injected sums, resulting from the contribution’s reduction, acts as a Keynesian multiplier whose supply side has been improved by the reduction of production costs and prices:

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

179

In this respect, the multiplier impact is improved (compared to the pure Keynesian multiplier) by die positive effect on competitiveness. The subsequent regain of growth is granted by the price decrease, resulting ultimately in a rise of real income for households, a profit decrease for firms, stimulating thereby investment and competitiveness. The HERMES-specific substitution effects in production techniques at industry branch level account for a second order phenomenon: the decreasing wage costs induce an increase in employment and a decrease in the amount of energy used per production unit. The results on Table 12.6 refer to the exercise combining the increased energy tax income with a subsequent redistribution of the tax income by decreasing the social contributions from the employers. According to the results, initial growth is not restored by the end of first year, in spite of the redistribution of tax income throughout this period. The reason for this is, that taxes are redistributed entirely within the firm sector. But only a part of these (redistributed) taxes are paid by firms. In the end, a non neglectable part of the increased tax incomes is still supported by final users. Thereby firms make a profit out of these combined measures. The subsequent decrease in their costs will not necessarily have immediate repercussions in 1terms of prices for final users. Consequently, the consumer price index will be superior to what it is in the central account and households will suffer a prejudice in terms of real income, maintaining the Table 12.6 Energy tax combined with a reduction of social security contributions of employers Germany 1991 GDP −0.22 Private consumption −0.62 GFCF −0.47 Employment 0.04 Consumption prices 1.29 Net bonus of government* −0.30 Current balance* 0.09 Final energy consumption −2.89 FEC by household −3.80 1995 GDP 0.38 Private consumption −0.09 GFCF 0.00 Employment 0.52 Consumption prices 1.77 * Net bonus of government −0.80 Current balance* 0.02 Final energy consumption −4.90 FEC by household −6.00 Note: Charge per baseline in %, except* in point of % of GDP.

France

Italy

UK

−0.50 −0.84 −0.34 −0.20 2.27 0.00 0.20 −2.72 −3.65

−0.16 −0.54 −0.67 +0.06 2.17 −0.40 0.02 −2.23 −1.94

−0.45 −0.60 0.01 0.22 0.74 −0.50 0.16 −3.30 −5.55

0.38 0.18 1.13 0.07 2.16 −0.60 −0.15 −2.96 −4.09

0.66 1.06 0.46 0.39 2.44 −0.70 +0.04 −2.72 −1.44

0.08 1.07 0.29 0.48 3.10 −0.60 −0.30 −4.46 −2.55

180

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

private consumption on behalf of the level corresponding to the reference account. Employment does not deteriorate as much as one could expect because of a favourable substitution effect linked to the labour cost reduction of 3 to 4% (sometimes even more). After a period of five years growth will be restored on average in all the countries. This is a general feature of the multiplier reducing social contributions: It is superior to unity and acts in the medium run on costs, competitiveness and sometimes even investment. From the global scenario point of view one can observe that while joining together the two measures, price levels did not reach those of the central account. In fact, it is the consumption price which automatically incorporates the increase of the tax revenue, while the GDP price index is more representative of competitiveness and near to the one of the central account since it represents the decrease in wage costs for firms. In spite of a high price of consumption, the current account did therefore not deteriorate. Concerning energy conservation investments we noticed the opposite case. Notice that the substitution effects, due to a wage cost decrease, do positively influence employment (and sustain therefore economic activity). They also reduce energy consumption, although the extent of this effect remains much smaller as the initial 20% tax rate. The reduction of income taxes

Maintaining the 20% tax rate on energy consumption, all different measures seem to harm households in one way or another. The evolution of their real disposal income in the short run, after one year, stands for this deterioration. The starting point of the following exercise is to avoid the negative effects on household income of previous measures by choosing a scenario which aims to restore to households their purchase power by orienting the redistribution on the reduction of income tax. Table 12.7 shows the comparative advantages and disadvantages with respect to the previous solution: First of all, one observes a less important reduction of GDP at the beginning of the period, mainly due to the support by household consumption. This turns out to be a relatively expensive undertaking in terms of inflation and loss of competitiveness. As a matter of fact, the employment embodied growth becomes less favourable. Five years later, growth and especially employment become for all countries (except Italy) less favourable compared to previous cases. This is merely Table 12.7 Energy tax combined with a reduction of personal income tax

1991 GDP Private consumption GFCF Employment Consumption prices Net bonus of government* Current balance* Final energy consumption FEC by household 1995

Germany

France

Italy

UK

0.02 0.34 −0.26 −0.06 2.10 −0.42 −0.18 −2.28 −2.89

−0.07 0.15 0.50 −0.08 2.50 −0.12 −0.16 −2.38 −2.91

−0.41 −0.58 −0.35 −0.15 2.75 0.01 0.08 −1.85 −2.03

−0.30 −0.24 0.02 0.03 2.38 −0.06 0.23 −3.15 −4.32

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Germany GDP 0.52 Private consumption 1.48 GFCF −0.17 Employment 0.16 Consumption prices 3.12 Net bonus of government* −0.58 * Current balance −0.52 Final energy consumption −3.80 FEC by household −4.46 Note: Charge per baseline in %, except* in point of % of GDP.

France

Italy

UK

0.18 1.96 1.40 −0.05 5.66 −0.30 −1.00 −3.13 −2.81

0.58 1.24 −0.09 0.25 4.24 −0.31 −0.34 −2.12 −1.35

−0.08 1.35 −0.89 −0.25 4.57 0.67 0.07 −4.19 −4.58

181

due to the degradation of the balance of current accounts and the absence of second range substitution effects. Therefore the overall reduction in energy consumption will be less important. 12.II.3 Synthesis scenario All different redistribution scenarios of section two have their specific advantages and disadvantages: ● The subsidy for energy conservation investments is a very favourable measure in order to reduce CO2 emission, but creates (at least at the beginning of the period) problems in terms of costs, prices and competitiveness. Furthermore this measure weakens very much the purchase of households. At last, it does not seem to represent a very realistic scenario regarding the size of investments at stake. ● Reducing social contributions from employers has the advantage of decreasing considerably the costs of firms, improving the medium term competitiveness and, in a second instance, promoting employment and energy savings. Opposed to this, households are being harmed at the beginning of the period by a tax they cannot turn down on other agents and which is not compensated by any form of subvention. ● Finally, the reduction of income tax represents mainly an advantage for households by means of deteriorating prices and competitiveness. Consequently, these three measures can be viewed as follows: The first one mainly improves energy savings, the second one costs and competitiveness, the third one purchase power of households. In order to find a solution which preserves all three objectives, it appears absolutely necessary to take into consideration a combination of all these measures, e.g. a synthesis scenario supporting all three forms of redistribution simultaneously. It matters now to find a criterion allowing us to determine the optimal combination, e.g. the best ‘policy mix’. How to determine this policy is a deeper research matter which includes a comparative eva luation of all three different scenarios from the viewpoints of energy savings, competitiveness and employment. What we are doing right here is nothing else than to rebuild a scenario which can be supported by some of the findings made previously:

182

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

1 Since the initial objective of all policies was the reduction of CO2 emissions (and thereby energy savings), the measure aiming to redistribute energy tax income for energy conservation investments seems to have priority, especially if one considers that the GDP multiplier and employment behave very favourable once the measure is simultaneously implemented in all EC countries. The underlying idea of this scenario is the quest for the largest possible size of impact. If one considers that firm investment represents 12% of GDP, the figure of 0.5% of GDP for energy conservation investments appears to be a maximum since it stands for a little bit more than 4% of investment. 2 Opposed to this view, one knows the inconveniences of the previous measure lying mainly in the price increase, loss of competitiveness and losses through international trade. At this scanning stage of research work, the idea might be to choose the less inflationary measure, which (in a second instance) does also create favourable employment and energy substitutions: the reduction of social contributions from employers, combined with subsidies in such a way, that in the end all tax income is redistributed. At this stage of our research work (cf. Table 12.8) we obtain a relatively favourable scenario in terms of GDP competitiveness, employment and energy savings. A deeper research of other measures and refining selection criteria might reveal the existence of possible other scenarios (with other advantages and disadvantages). To conclude, we would like to draw out the principal lessons of this research work and mention perspectives for possible further research topics: Table 12.8 Synthesis scenarios

1991 GDP Private consumption GFCF Employment Consumption prices Net bonus of government* Current balance* Final energy consumption FEC by household 1995 GDP Private consumption GFCF Employment Consumption prices Net bonus of government* Current balance* Final energy consumption FEC by household Note:

Germany

France

Italy

UK

−0.09 −0.82 1.69 −0.04 1.53 −0.25 −0.08 −3.20 −3.92

−0.15 −0.67 2.46 −0.12 2.36 −0.10 −0.22 −2.79 −3.90

−0.02 −0.61 2.22 −0.09 1.98 −0.29 −0.31 −1.80 −2.06

−0.31 −0.84 2.31 0.23 1.31 −0.43 0.16 −3.63 −3.83

0.27 −0.31 1.70 0.20 2.26 −0.61 −0.11 −7.15 −6.59

0.46 0.41 2.83 0.10 3.39 −0.49 −0.41 −4.77 −4.80

0.18 0.33 2.67 0.21 2.81 −0.33 +0.01 −4.86 −2.43

0.11 0.63 2.13 0.32 3.51 −0.36 0.05 −6.45 −3.39

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

Germany Charge per baseline in %, except* in point of % of GDP.

France

Italy

183

UK

● The very first conclusion of this work is that, regarding past results, an increased energy price does not induce spectacular behaviours in terms of energy savings and reduction of CO2 emissions. In this respect, it is therefore necessary to induce a more voluntary energy saving behaviour. ● Furthermore, taxation of energy consumption does increase prices and depresses economic activity. Redistribution is therefore an essential feature. ● Promoting energy conservation investments is one possible way to induce a more energy-saving behaviour, although this appears to be costly (at least in a first instance) and deteriorates the external equilibrium. Opposed to this, economic activity is sustained and therefore subventions to energy conservation investments seem to be a convenient matter of redistribution which reveals to have an incentive character as well. ● But the whole tax income cannot be redistributed in such a way and it is strictly necessary to couple this measure with ways of decreasing firm costs, namely a decrease of social contributions from employers. The advantage of this measure is to induce favourable substitutions in the fields of employment and reduction of energy consumption. At last, considering the high costs of a voluntary energy saving behaviour from an international point of view, it becomes important that these prices are carried out simultaneously in all European countries. Further research in this field might go in more than one way: ● The analysis might be extended beyond the year 2000 in order to identify maturation effects which become effective only in a long run. ● Choices might be made on the best policy mix subventions, reduction of compulsory contributions etc. ● To end with, extension of research towards the choice of energy product combinations (fuel mix). ACKNOWLEDGEMENTS The study ‘Increase of taxes on energy as a way to reduce CO2 emissions’ is an example of a fruitful collaboration between CEC officials and national experts in the framework of the Research Programme ‘JOULE—Models for Energy and Environment’ of the Directorate-General for Science, Research and Development (DG XII) of the Commission of the European Communities. The HERMES model, developed in the same Research Programme, was used to realize all the evaluations presented here. So, this study was only made possible thanks to the work accomplished, since many years, by all the Hermes national teams. Moreover, without the substantial contribution of some national teams, and more particularly the French team, it would not have been possible to achieve all the simulations contained in this paper. We are furthermore indebted to Mr S.Standaert of the Maestricht University who has participated to the design of the study and is preparing new analysis on the same subject. We are also grateful to people from other DG and particularly Mr P. Buigues and Mr M.Mors from the Directorate-General for Economic and Financial Affairs (DG II), and Mr J.Delbeke and Mr L.De Nocker from the Directorate-General for the Environment (DG XI), who have followed the development and commented on the intermediary results of the study.

184

ENERGY TAX INCREASES AS A WAY TO REDUCE CO2 EMISSIONS

Let us finally acknowledge the helpful assistance of Mrs S.Protasio and Mr P.Bandilla who have translated the document into English. NOTES 1 2 3 4 5 6 7 8 9

In the case of France, for example, this represents 1000 FF per ton of CO2 i.e. about 150 ECU 1991 prices. This measure is attenuated by inflation. In this chapter in fact the different models are not linked and of course the results are identical. Even before the mechanisms in the model are activated and before any indexation. Note from S. Standaert ‘Quantifying the effects of non-prices environmental policies’ Country report Italy March 91. This way an increase in the rate of equipment scrapping could be seen as a spontaneous behaviour. For further details one should consult a study carried out by DG XII in 1989 (‘Internal Market and Environment’ report for the DG XI study). In theory 12% but 6 or 7% should be considered due to the adaptation delays. Another possible measure is the reduction of VAT; but we did not deal with this reduction to avoid interference with the VAT harmonization problem in the EC countries.

Chapter 13 Energy conservation and economic performance in Japan: an econometric approach K.Ban

Abstract In this chapter, some results of a newly developed energy macroeconometric model of the Japanese economy are provided. In this model, an increase in the real energy prices has played an important role in improving the energy efficiency of the Japanese economy since the first oil shock. High energy prices usually slow down the growth of the economy. However, they stimulate the substitution of energy with other inputs. We have observed that energy and capital were substitutes in the major industries in Japan. This means that business firms are encouraged to make active investments in order to conserve energy inputs, thus mitigating the depressing impacts of high energy prices on the Japanese economy. As a result, the energy/GNP ratio has decreased by 40 per cent since 1973. Two alternative policies for reducing CO2 emissions in Japan are evaluated. The possibility of stabilizing the level of CO2 emissions in 2000 within the 1990 level is investigated. It is shown that a carbon tax on fossil fuels is an efficient way to achieve this target. Even though imposing a carbon tax will lead to a reduction of GNP, its welfare losses will be smaller than those associated with legislative regulation of CO2 emissions. It is also shown that the impact of a carbon tax might be mitigated by substitution among factor inputs as compared with legislative regulation. 13.I INTRODUCTION Most countries have experienced two dramatic energy crises since 1973. The first crisis occurred from 1973 to 1974, and the second crisis from 1979 to 1980. After these crises, significant efforts have been devoted to energy conservation. Figure 13.1 depicts both real GNP and final energy consumption in Japan, which are normalized to be 1.0 in 1960. According to this figure, energy consumption grew faster than real GNP until 1973. On the other hand, energy consumption remained almost unchanged during the period from 1973 to 1985. During these periods, the growth rate of real GNP slowed down, but still grew at an annual rate of 4.2 per cent. As a result, the ratio of energy consumption to real GNP (energy/GNP ratio) declined at a rapid pace since the first oil shock as shown in Figure 13.2. Real energy prices almost triplicated, as is also illustrated in Figure 13.2. As Bruno and Sachs (1982) have pointed out, high energy prices usually stimulate the substitution between energy and other productive factors. However, it sometimes negatively impacts the economy, and reduces the rate of long-run economic growth. Energy consumption was stimulated during the periods of low energy prices before 1973. On the other hand, energy consumption was depressed during the period of high energy prices from 1974 to 1985.

186

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

Figure 13.1 Real GNP and energy consumption

Figure 13.2 Energy price and energy/GNP ratio

As a result, the energy/GNP ratio decreased by almost 40 per cent. These facts imply that the high energy prices have been very effective in stimulating savings of energy consumption in Japan. One area of interest concerns the impact of the decrease in real energy prices since late 1985. Energy consumption has accelerated again. According to the most recent data, the energy/GNP ratio increased in 1990 In recent years, there has been growing concern about the fact that human activities may be affecting the global climate through the emissions of greenhouse gases such as CO2. These emissions will lead to a rise in

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

187

global temperature over the next century, and such a warming is suggested as having a major impact on economic activity. As a result, policy makers have begun to consider various ways to reduce greenhouse gases. The Japanese government declared that emissions of CO2 in the year of 2000 should be stabilized at the 1990 level. There is a rapidly growing literature to evaluate the impact on the economy of policies to reduce greenhouse gases, where the impact on long-run economic growth is assumed to depend crucially on the degree of substitutability between energy and other goods. This assumption has also been investigated empirically in order to evaluate the impact of oil shocks on the economy. In this chapter, some results from an econometric model are used to evaluate the impact of energy prices on the Japanese economy. There are three producer sectors in the model, iron and steel, chemical, electric power and other industries. There are four consumers goods: fuel and power, automobile, house appliances, and other goods. Relative prices play a most important role in the model. The impact on the economy of an increase in energy prices is evaluated. According to some simulations, the recent decline in energy prices will probably inhibit any improvement in energy efficiency, which will result in an increase in greenhouse gas emissions. It is suggested that it is very urgent for the government to implement an energy policy to improve energy conservation and reduce greenhouse gas emissions. Alternative energy policies to reduce greenhouse gas emissions will be evaluated from the view point of economic welfare. In section 13.II, a model of household demands for automobiles and gasoline based on dynamic optimization behaviour is presented. The model of factor demands in the iron and steel industry is presented in section 13.III, and the model of the demands for primary energy in electric utilities is presented in section 13.IV, respectively. Some simulation results illustrating the impact of energy prices are provided in each section. In section 13.V, we evaluate two alternative energy policies to reduce CO2 emissions. 13.II HOUSEHOLD DEMANDS FOR AUTOMOBILE AND PETROL The share of energy consumed by transportation equipment is now more than 20 per cent in Japan. Household consumption amounts to almost half of this share. Despite energy conservation in industry since 1973, the most serious problem in Japan is the rapid increase in energy consumption in the household sector. Figure 13.3 graphs private consumption expenditure and the demand for gasoline both of which are normalized to 1.0 in 1973. The demand for petrol increased faster than private consumption during the period until 1978. This was mainly due to a rapid increase in income and a decline in new automobile prices. Though the growth rate of gasoline expenditures slowed during the period from 1978 to 1988, it accelerated again after 1988. It is well known that the price of petrol affects the demand for both automobiles and petrol. A decrease in both demands results in energy conservation. So, it is necessary to consider the impact of a gasoline tax on both demands at the same time in order to evaluate the long-run effects. On the other hand, a decrease in automobile demands might depress the economy through output reductions in the auto industry, which will be evaluated in the full simulations. In Japan, the petrol tax is 53.8 yen per litre, which is almost 50 per cent of the sale price. An increase in the tax rate is a very effective way to change the relative price of petrol. In this section, the impact of an increase in the petrol tax is evaluated. First, a dynamic demand model based on Tinshler (1983), where an individual is assumed to maximize a life-time utility function subject to budget constraints is estimated. The model is formulated as follows:

188

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

Figure 13.3 Consumption and demand for petrol

(1) such that

where the following apply: U(C1t, C2t, Kt) C1t C2t Kt It At PKt P1t P2t Rt Yt p δ

utility function other consumption the number of kilometres the number of automobiles the number of new automobiles assets price of new automobiles price of petrol price of other consumption fixed cost for holding cars interest rate income discount rate depreciation rate

It is assumed that the optimal path lies in the interior of the budget set and the representative individual agent follows his optimal path of consumption. Given these assumptions, the following two first-order conditions are obtained:

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

189

(2) (3) where Ui (i=1, 2 and K) are the partial derivatives of the utility function with respect to C1t, C2t, Kt respectively. The UCat defined by equation (2) is the user’s cost of an automobile. It is an increasing function of the price of a new automobile, the price of petrol and capital loss. If the utility function is assumed to be CES in form, the following estimable functions can be obtained: (4)

(5) Equations (4) and (5) were estimated using annual data obtained from the National Survey of Family Income and Expenditure for Japan over the period 1971 to 1989. Unfortunately, the restriction that the coefficients of price elasticity in the two equations are the same is rejected at the significant level of five per cent. As a result, the following interrelated system of unrestricted equations was estimated. The rationale for this specification is adaptive price expectations. (6) (7) Figure 13.4 illustrates the response of the number of kilometres and the number of automobiles to a permanent 10 per cent increase in the price of petrol. The number of kilometres falls by a significant amount in the first few years. It falls by 4.8 per cent in the first year and 7.8 per cent in the fourth year. Then it gradually recovers to two per cent below the initial level. On the other hand, the number of automobiles falls gradually. It falls by only 1.2 per cent in the first year, and by 4.5 per cent in the tenth year. As the price of petrol has doubled since 1973, this simulation suggests that the consumption of petrol should have fallen by 20 per cent as compared with the case of no oil shock. On the other hand, the recent decline in petrol prices might stimulate consumption in the future. 13.III ENERGY DEMANDS OF THE IRON AND STEEL INDUSTRY The iron and steel industry is one of the most energy intensive industries. In 1989, it consumed 13 per cent of the total final energy consumption in Japan. What is worse is that 17 per cent of the total CO2 emissions in Japan originates from this industry. On the other hand, the gross domestic product (GDP) of the iron and steel industry is 2.4 per cent of the total GDP. So a depression of this industry caused by high energy prices might depress the macro-economy. Fortunately, the iron and steel industry has been very active in saving on energy use since the first oil shock in Japan. Many blast furnaces were scrapped and rebuilt. As a result, the energy used in the iron and steel industry declined from 465 trillion (1012) Kcal in 1972 to 414 trillion Kcal in 1989, even though output

190

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

Figure 13.4 Impact on demand for petrol

Figure 13.5 Energy used to produce 1 tonne of crude steel

remained almost unchanged (103 million metric tons in 1972 and 108 million metric tons in 1989).1 Figure 13.5 illustrates the energy used in producing one metric ton of crude steel. It peaked at 4.89 million Kcal in 1975, and then declined sharply to 3.79 million Kcal in 1984. A two-stage optimization model of the production structure under the assumption of homothetic separability is developed in order to evaluate the impact of energy prices on factor input usage in the iron and steel industry. First, it is assumed that a firm maximizes its discounted presented value:

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

191

(8) such that where the following apply: Kt Lt Et It wt et qt τ p δ

capital stock labour input aggregate energy input gross investment wage rate price of aggregated energy price of investment goods rate of corporate income tax discount rate depreciation rate

It is assumed that capital is a quasi-fixed factor, but that labour and energy are perfectly variable factors. The presence of It in the production function reflects adjustment costs. The discrete maximum principle is used to solve the intertemporal maximization problem in (8). The discrete Hamiltonian is defined as follows: (9) The optimal trajectory is described by the following first-order conditions: (10) (11) (12) (13) In this model, conditions (11) and (12) determine the static equilibrium. If the production function is assumed to be homogeneous and of CES form, the following two estimable functions can be obtained as: (14) (15) where Qt is output. Solving the difference equation (13) indicates that λt is the discounted present value of the marginal product of capital. On the other hand, the optimal condition (10) describes the marginal cost of current investment. So, the optimal condition for investment is that the marginal discounted value of capital equals the cost of investment. Under the appropriate conditions, the following estimable investment function can be obtained. (16)

192

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

where πt is the discounted present value of the marginal cost of capital, and uct is the user’s cost of capital. The iron and steel industry chooses among five kinds of energies (E): coal, coke, heavy fuel oil, natural gas and electricity. If we impose homothetic separability in energy use, total energy can be treated as one homothetically aggregated factor in the production function as shown in equation (8). Next, we turn to the derivation of the energy sub-model. It is assumed that the firm minimizes energy cost, given total energy shown in equation (15). (17) such that Φ(E1, E2, E3, E4, E5)=E where ei and Ei are the prices of the i-th energy and its demand with i=1, 2, 3, 4, 5 indicating coal, coke, heavy fuel oil, natural gas and electricity, respectively. If the functional form of Φ is assumed to be CES, then the following estimable demand functions can be obtained. i=1, 2, 3, 4, 5

(18)

Equations (14), (15), (16) and (18) were estimated using the annual data from 1971 to 1989. Factors of both Hicks neutral and factor augmented technical progress are excluded because of statistical insignificance.2 This implies that energy conservation in the iron and steel industry is mainly due to a substitution among various productive factors that was caused by a change in their relative prices. Table 13.1 shows the price elasticities of the demands for coal, coke, fuel oil and capital stock. In this table, the rows denoted by (S) show the short-run elasticities where the capital stock does not adjust. On the other hand, the rows denoted by (L) show the long-run elasticities where the capital Table 13.1 Estimated price elasticities (coal, coke, fuel oil, capital stock)

PCOAL PCOKE POIL

(S) (L) (S) (L) (S) (L)

Coal

Coke

Oil

K

−1.812 −4.642 1.534 3.526 0.055 −0.064

−0.000 −0.027 −0.226 −0.903 0.085 0.026

0.002 −0.086 0.866 3.231 −0.258 −0.256

— 0.045 — 0.248 — 0.015

stock adjusts. According to these results, all of the own price elasticities of the demand for energies are negative both in the short run and in the long run. The price elasticity of the demand for coal is relatively large in magnitude, while those of coke and fuel oil are small. We can observe that coke and fuel oil are substitutes both in the short run and in the long run. However, coal and fuel oil are substitutes in the short run, while they are complements in the long run. The capital stock, which is denoted by K, responds strongly to the price of coke. This implies that the active investment behaviour of the iron and steel industry is mainly due to the response to increases in the price of coke. Figure 13.6 illustrates the per cent deviation of the simulated solutions, which was calculated using the estimated model of the iron and steel industry, from the historical values during the periods from 1974 to

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

193

Figure 13.6 Impact on the iron and steel industry

1986. In the simulation, the output of the iron and steel industry and the relative prices are fixed at their 1973 levels. The historical value of energy consumption is 397 trillion Kcal in 1985, which is 40 per cent less than the simulated value (650 trillion Kcal). So, the high energy prices forced the iron and steel industry to save on the energy consumption. According to the same figure, savings on coke consumption were 30 per cent in 1985, and savings on heavy fuel oil consumption were 44 per cent. Coke is one of the most basic materials used for blast furnaces, and is suggested as being difficult to save for technological reasons. However, the simulation results imply that changes in relative prices have played an extremely important role in technological progress towards improving energy efficiency. 13.IV PRIMARY ENERGY DEMAND OF ELECTRIC UTILITIES Substantial progress in the substitution of fuel oil with coal, LNG, and nuclear energy has been made since the mid-1970s as shown in Figure 13.7. Fuel oil used for electricity peaked at 71 per cent in 1973, and then sharply declined to 24 per cent in 1986. However, the target for reducing oil use has become difficult to achieve because of low oil prices since late 1985. As a result, fuel oil used for electricity has gradually increased again to 27 per cent until 1989. On the other hand, nuclear energy has rapidly increased from only 2 per cent in 1973 to 30 per cent in 1987. It certainly plays an increasingly important role in Japan. However, because of the recent acceleration of the economy and the slight reduction of load factors to nuclear plants, it has declined to 26 per cent until 1989. Even though the Japanese government encourages the electric utilities to construct new nuclear power plants, it has become more difficult to obtain the public acceptance. Electric utilities have to determine the best electric power supply mix to supply electricity stably at the lowest feasible cost. In our model, hydro and nuclear energy are treated as exogenous. Both energies have played an important role in electric power generation in the past, and will do so in the future. Hydro energy

194

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

Figure 13.7 Primary energy for electric utilities

accounts for thirteen per cent of total electricity. However, most of the economically viable sites for hydro electric generation have already been developed. On the other hand, it usually takes more than ten years to construct a new nuclear power plant. A model of primary energy demands for coal, fuel oil and LNG is developed, which is almost the same as that of the previous section. However, homothetic separability in energy use is not accepted statistically. So, the following production function is used to formulate the dynamic optimization model: (19) where Qt is the electricity generated by steam power plants. As a result, the energy frontier function is defined as follows: (20) where coal, fuel oil and LNG are assumed to be substitutes with each other in the same plant. Then, the following energy demand functions can be obtained: (21) Equations (14), (15), (16) and (21) were estimated using annual data from 1971 to 1989. Factors of both Hick neutral and factor augmented technical progress were also excluded. Table 13.2 shows the price elasticities of the demands for coal, oil, LNG, and power plants. In the table, the rows denoted by (S) show the short-run elasticities where the power plants do not adjust. On the other hand, the rows denoted by (L) show the long-run elasticities where the power plants adjust. According to these results, all of the own price elasticities of the demand for primary energies are negative both in the short run and in the long run. The price elasticity of demand for coal is relatively large in magnitude, while that of LNG is small. It can be observed that coal, fuel oil and LNG are substitutes with each other in the short run. However, coal and fuel oil are substitutes in the long run, while fuel oil and LNG are

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

195

Table 13.2 Estimated price elasticities (coal, oil, LNG, power plants)

PCOAL POIL PLNG

(S) (L) (S) (L) (S) (L)

Coal

Oil

LNG

K

−0.316 −1.628 0.049 0.005 0.008 −0.100

0.005 0.119 −0.134 −0.915 0.008 −0.014

0.001 −0.042 0.004 −0.227 −0.018 −0.212

— −0.018 — −0.300 — −0.141

complements. Power plants, which are denoted by K, respond strongly to the price of fuel oil, which is twice as high as that of LNG in magnitude. In the case of electric utilities, an increase in energy prices depresses investment in the power plants. In this model, hydro and nuclear energy are treated as exogenous. If the supply and demand for electricity are assumed to be in equilibrium, and hydro and nuclear energy increase more than anticipated, an increase in the supply of electricity will reduce its price. These facts imply that an increase in hydro and nuclear energy has the impact on the electric power supply mix. Table 13.3 shows the impact of an increase in nuclear energy by 10 per cent, where the level of income is fixed in the simulation. According to these results, total electric power supply increases by 1.59 per cent in the long run, while its price declines by 2.55 per cent. Coal and fuel oil consumption decline by 3.44 and 2.69 per cent respectively in the long run, but LNG consumption increases by 0.28 per cent. It can be concluded that nuclear energy is very effective in reducing coal and fuel oil usage. Figure 13.8 illustrates the per cent deviation of the simulated solutions of electric utilities from that of the historical values during the periods from 1974 to 1986, where GNP is fixed at its historical value and relative prices are fixed at their 1973 levels. The historical value of electricity generated was 604 billion KWh in 1985, which was 50 per cent less than the simulated value (1.2 trillion KWh). So, savings in high energy prices reduce the demand for primary energies. Coal and fuel oil have been almost 50 per cent. That is, the demand for coal and fuel oil would have doubled, were it not for the two Table 13.3 Impact of an increase in nuclear energy

Short-run Long-run

Coal

Oil

LNG

Electric power

Electric price

−0.52 −3.44

−0.30 −2.69

−0.03 0.28

1.34 1.59

−1.66 −2.55

energy shocks. These simulation results imply that changes in relative prices have also played an important role in the conservation of primary energies by the electric utilities. 13.V THE IMPACT OF A CO2 EMISSIONS TAX ON THE JAPANESE ECONOMY The greenhouse effect has evolved from a purely scientific issue to an important public policy debate. In this section, the costs of reducing carbon dioxide (CO2) emissions are evaluated. Many legislative proposals

196

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

Figure 13.8 Impact on electric utilities

have set physical targets for emission reductions. The Japanese government declared that CO2 emissions in 2000 should be stabilized at the 1990 level. The amount of CO2 released in Japan was 267 million tons of carbon in 1975; it peaked at 282 million tons in 1980, and declined to 270 million tons in 1985. That is, CO2 emissions stabilized during the period from 1975 to 1985. This was mainly due to stable energy consumption as illustrated in Figure 13.1. However, CO2 emissions increased again since 1986 because of a decline in energy prices. Figure 13.9 shows the amounts of CO2 emissions by major industries in 1988. Some 32 per cent of CO2 was released from the electric utilities, 20 per cent was released from the transport industry and 18 per cent

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

197

Figure 13.9 CO2 emissions in Japan (276 million tons of carbon in 1988)

was released from the iron and steel industry. These three major industries, which were investigated in the previous section, released almost 70 per cent of the total CO2 in Japan. In this section, two alternative policies to reduce CO2 emissions during the period from 1991 to 2000 are evaluated. One policy is a carbon tax on coal, fuel oil and LNG. The tax is fuel-specific, since CO2 emissions are generally proportionate to the carbon content in each fossil fuel. The tax is levied on imported primary fuels, and is assumed to change the relative price of each energy. In the case of Japan, almost all fuel oil and LNG are imported, and more than 90 per cent of coal is imported. So, even though the tax is applied to the import of primary fuels, distortions between domestic and imported fuels are avoided. The other policy is legislative regulation of CO2 emission. Each industry or household is endowed with initial quotas of emission rights. If the amount of CO2 emissions are limited at initial endowments, each industry has to slow down its activity to keep within this upper limit. Then, some industries will also develop the technologies to reduce the CO2 emissions, or buy emission rights from other industries, while some industries will sell their emission rights and withdraw. If emission rights were perfectly marketable, the price of emissions would be determined as the equilibrium carbon tax rate with emission constraints. However, the initial endowment is too arbitrary to be allocated. These regulations might cause huge welfare losses from an economic point of view. In order to evaluate these two alternative policies, the following two simulations were undertaken during the period from 1990 to 2000. The price of fossil fuels is expected to increase by four per cent at an annual rate in the baseline scenario. In the case of a CO2 tax policy, the rate of the carbon tax is assumed to be 30 per cent on coal, 24 per cent on fuel oils and 16 per cent on LNG, respectively. These carbon taxes have a major impact on relative prices and government revenue. On the other hand, in the case of legislative regulation, output of the iron and steel industry, electric utilities and the gross national product have to be slowed down to achieve the level of CO2 emissions under the limits. Figure 13.10 illustrates the impact of introducing a carbon tax on CO2 emissions. In the case of a baseline scenario, CO2 omissions are expected to increase from 292 million tons of carbon in 1990 to 362 million tons in 2000, that is, an increase at an annual rate of 2.2 per cent. On the other hand, if a carbon tax is introduced, CO2 emissions in 2000 will be 298 million tons of carbon. This means that the rate of carbon tax

198

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

Figure 13.10 CO2 emissions

to be imposed is enough to stabilize CO2 emissions at the 1990 level. The same path of CO2 emissions can be obtained by legislative constraints. Reducing CO2 emissions will mainly occur via the costly route of a policy-induced reduction in energy use. Table 13.4 shows the impact of the two alternative policies on energy use. A reduction in CO2 emissions is mainly due to a decrease in coal usage. Coal contains the most carbon among fossil fuels, and its price elasticity is the highest. Figure 13.11 illustrates the impact of the two alternative policies on the growth rate of real GNP. In the case of a carbon tax, the growth rate is reduced by 1.1 per cent in 1991 (from the baseline of 4.3 per cent to 3.9 per cent with the carbon tax), and by 0.3 per cent in 1992 (from the baseline of 4.6 per cent to 4.0 per cent). However, it recovers thereafter and overshoots for several years, then oscillating to converge with that of the baseline scenario. In the case of the legislative constraint, the impact is more drastic. Table 13.4 The impact on energy use

CO2 tax Regulation

Final energy consumption

Electricity

Coal

Oil

LNG

−6.0 −3.4

−4.3 −7.8

−24.0 −20.0

−6.2 −2.8

−3.5 −7.4

Real GNP shows a growth rate of zero per cent in 1991. However, it also overshoots the baseline by 0.5 per cent in 1992, and 1.5 per cent in 1993, then converges with that of the baseline scenario. The relatively moderate impact of the carbon tax is mainly due to price adjustments. Higher energy prices force business and households to substitute energy with other inputs. It also stimulates the business investments for conserving energy, because capital and energy are usually substitutes in major industries in Japan, as shown in section 13.III. The impact of the two alternative policies on the growth rate of the economy will apparently diminish in the late 1990s. However, the components of GNP will change drastically. Table 13.5 shows the per cent deviation of each component from the baseline scenario in 2000. A carbon tax changes relative prices,

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

199

Figure 13.11 Impact on GNP growth

which forces business firms and households to substitute energies with other goods and services. It also stimulates business investment. Of course, high energy prices have an inflationary impact on the economy, which results in a decrease in exports. A decline in imports is mainly due to the decrease in imports of fossil fuels. On the other hand, a Table 13.5 The impact on GNP components in 2000

CO2 tax Regulation

Consumption

Investment

Exports

Imports

−0.4 −1.7

3.2 −5.0

−2.6 1.7

−1.6 −5.6

policy of legislative regulation will only depress the economy. There is no substitutional effect. As a result, an increase in exports and a decrease in imports will offset the decrease in domestic demand. This is known as ‘the beggar-thy-neighbour’ action. Figure 13.12 illustrates the accumulated welfare losses in terms of reductions in GNP. In the case of a carbon tax, the accumulated reductions in real GNP will be 28 trillion yen in 1986 prices. On the other hand, it will be 41 trillion yen in the case of legislative constraint. The impact of a carbon tax on welfare losses will be only two-thirds as large as that of legislative constraint in magnitude. There are some shortcomings that remain at the conclusion of this work. As Edmonds and Reilly (1983) and many other authors point out, a global model is needed to investigate the problem of CO2 emissions. However, in this chapter’s analysis, the world outside Japan is treated as exogenous. As the next step, it is necessary to develop a feedback loop between Japan and the world in the model.

200

ENERGY CONSERVATION AND ECONOMIC PERFORMANCE IN JAPAN

Figure 13.12 Welfare losses (reduction of GNP in 1985 prices)

13.VI CONCLUSIONS In this chapter, some results of a newly developed energy macro-econometric model of the Japanese economy are discussed. In this model, an increase in real energy prices plays an important role in improving the energy efficiency of the Japanese economy since the first oil shock. High energy prices usually slow down the growth of an economy. However, they also stimulate the substitution of energy with other inputs. It is observed that energy and capital are substitutes in the major industries in Japan. This means that business firms are encouraged to make active investments in order to conserve energy inputs, which mitigates the depressing impacts of high energy prices on the Japanese economy. In the case of households, a reduction in expenditures for energy intensive goods forced business firms to invent energy efficient durable goods. As a result, the energy/GNP ratio has decreased by 40 per cent since 1973. However, improvements in energy efficiency have stagnated since 1986, and energy consumption for transportation has increased in the last few years. This is mainly due to the recent decline in energy prices. Two alternative policies for reducing CO2 emissions in Japan are evaluated. The baseline solution that predicts trends in the absence of control policies shows a growth of 2 per cent in CO2 emissions at annual rate until 2000. The possibility of stabilizing the level of CO2 emissions in 2000 within the 1990 level is investigated. It is shown that it is necessary to introduce a carbon tax of more than 30 per cent on fossil fuels in order to achieve this target. Imposing this carbon tax leads to a reduction in GNP, totalling 28 trillion yen during the period from 1991 to 2000. This welfare loss will be smaller than that associated with legislative regulation of CO2 emissions. The impact of carbon tax might be mitigated by substitution among factors inputs as compared with legislative regulation. NOTES 1 In this book one billion=109, one trillion=1012 and so on.

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

201

2 As Manne and Richels (1990) pointed out, we have to identify whether the remarkable improvement in the energy efficiency is attributable to the price mechanism, or to the autonomous time trend. According to our structural stability test using the data from 1986 to 1989, the autonomous time trend model was statistically not accepted.

REFERENCES Bruno, Michael and Jeffrey Sachs, 1982, ‘Input Price Shocks and the Slowdown in Economic Growth: The Case of U.K. Manufacturing’, Review of Economic Studies 49, pp. 679–705. Edmunds, Jae, and John Reilly, 1983, ‘Global Energy and CO2 to the Year 2050’, The Energy Journal 4, pp. 21–47. Manne, Alan S. and Richard G.Richels, 1990, ‘CO2 Emission Limits: An Economic Cost Analysis for the USA’, The Energy Journal 11, pp. 51–74. Tishler, Asher, 1983, ‘The demand for Cars and Gasoline: a Simultaneous Approach’, European Economic Review 10, pp. 271–87.

Chapter 14 Electricity futures markets Einar Hope, Linda Pud and B.Singh

Abstract The electricity sector in Norway is undergoing transition from a traditionally centralized organization to a more economically efficient decentralized structure. The key words in this reorganization are: market-based production and distribution systems, decentralization in decision making, deregulation/reregulation and risk management. The chapter focuses on perhaps the most important part of the reorganization of the system: the creation of well-functioning markets for electricity. A hydrosystem is exposed to various types of risk; consequently, risk management through market transactions becomes an issue. The paper discusses how risk markets in the form of futures markets can be organized for electricity, with specific proposals for implementation of futures market in the Norwegian electricity system. INTRODUCTION The electricity sector in Norway is undergoing transition from a traditionally highly centralized organization to a more economically efficient decentralized structure. The key words in this reorganization are: marketbased production and distribution systems, decentralization in decision making, deregulation and risk management. The formal document which provides the general guidelines for this reorganization is the new ‘Energy Law’ (Law of Production, Transformation, Transmission, Sale, and Distribution of Energy) passed by the Norwegian Parliament in June 1990. While the text of the new law is held in relatively general terms, the main emphasis in the background policy documents is on deregulation and competition with regards to the specific proposals for reorganizing the electricity system. The need for development of competitive markets for electricity to achieve the underlying objectives of the new law is obvious. This chapter draws on the background documents of the new ‘Energy Law’ and analyses the main issues related to the development of competitive markets in general and futures markets in particular, for electric power in Norway.1 The chapter is divided into four sections. Section 14.I gives a description of the present structure of the industry, and outlines the main proposals for creating a market-based system for the electricity industry in Norway. Section 14.II provides a discussion of issues related to the development of futures markets for electricity; an important element in the reorganized system. Section 14.III addresses the issues related to the designing of the futures contract; an instrument which is crucial to the success of the futures market stipulated in the new system. Section 14.IV provides the main conclusions of the chapter.

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

203

Figure 14.1 The structure of the Norwegian power sector

14.I THE STRUCTURE OF THE ELECTRICITY SECTOR IN NORWAY Figure 14.1 is a flow diagram of the structure of the Norwegian power sector. The arrows in the diagram indicate the flows of transactions (sale and purchase) in this sector. Thus, for example, while the producers sell forward contracts, they both buy and sell in the spot market. Similarly, distribution companies buy forward contracts, and sell to the final consumers. The Norwegian electricity sector is almost exclusively based on hydro-power generation. There are approximately 70 primary producers and 230 distribution companies in the system. A major producer is the state owned ‘Statkraft’ which accounts for nearly 30% of total power production in Norway and possesses exclusive rights for exchange of power across the national borders, with Sweden and Denmark as the main markets. The company has a relatively strong political position and is used by the government as an instrument for implementing parts of social and industrial policies via subsidized electricity prices. As regards ownership, most of the distribution companies are municipal or inter-municipal companies. There are strong vertical ties between production and distribution companies either through ownership or long-term contractual arrangements. Approximately 75% of total electricity demand from the household and commercial sector is covered by vertically integrated companies; an issue which has been a subject of discussion, both in itself and as part of the market orientation debate.2 Consumers in the Norwegian hydro-electric power system can be classified into two categories; the intermediary bulk consumers mainly comprising of distribution companies and energy-intensive industry firms with or without own production; and the final consumers which include households, agriculture, public services and other industry. The distinction between the final and intermediary bulk consumers and between those with or without own production capacity is important today with respect to access of these consumers to the existing power markets in Norway. Primary markets for electric power in Norway can be classified into three categories: the forward market, the spot market and the export market. The most important of these is the forward market, which has traditionally consisted of a large number of non-transferable bilateral contracts between producers and bulk

204

ELECTRICITY FUTURES MARKETS

consumers (distribution companies and energy intensive industries) for supply of ‘Firm Power’ referred to as ‘Fastkraft’. Over 90% of the Norwegian electric power production is sold under these contracts. The remaining 10% of the supplies is exchanged on the spot market which has till recently been the main source of ‘Interruptible power’ or ‘Utvekslingskraft’. These power exchanges are agreed upon on very short notice, i.e. on an hourly or a daily basis reflecting the current excess demand and excess supply of individual producers and bulk consumers with own generating capacity. The spot market transactions are carried out at a market-clearing price calculated on the basis of expected demand and supply schedules derived from point estimates by the producers. The market is managed by a power exchange company, ‘Samkjøringen’, the national clearing house for these transactions. This company, which is jointly organized by the power producers, is also responsible for operation of the transmission network in the country. More recently, in 1990, ‘Samkjøringen’ also introduced a market for what is defined as short-term firm power, called the market for supplementary power. In general, the main feature of these spot-markets has been limitations on market access for final consumers and for bulk consumers with own annual production less than 100 Gwh. In brief; the most important weaknesses of the present system can be summarized as follows: 1 There is no efficiently functioning market for firm power. Firm power contracts are negotiated individually, being predominately bilateral, non-standardized contracts between buyers and sellers. 2 There is no efficiently functioning second-hand market for contracts. Thus renegotiating contracts may be costly, making the greater part of the electricity market very inflexible. 3 Prices and other contract terms are generally set by administrative or political decree (e.g. the basic Statkraft price is set by the Norwegian Parliament as part of the regulation of the company). 4 While the short-term power exchange allows for market determined spot prices, the limitations on market access remain an important distortion in the functioning of the spot market. The structure of the electricity sector described above is now undergoing significant changes. The recent Energy Law of June 1990 allows for a deregulated and a competitive structure in the energy markets. Essentially, the Act takes a functional view of the electricity sector, where functions which have a natural monopoly character (transmission and distribution) are to be organized as regulated monopolies, while production will be open to competition in futures and spot markets. Figure 14.2 gives the basic features of the desired market based system for electricity. Transmission and distribution are essentially transportation activities, with cost characteristics which make them a natural monopoly. Realizing the natural monopoly character of these activities, the marketbased organization as given in Figure 14.2 assumes regulated monopolies to undertake these activities in the electricity sector. A publicly owned, pure transmission company which owns the national grid and is responsible for power transmission is proposed for the reorganized system. The main objective is to ensure independence, transparency and open access to the grid. As regards distribution, the reorganized system envisages regulated local monopolies for this activity. A regulatory regime for these distribution companies which ensures transparency with respect to tariffs and other conditions for power deliveries is essential. It may be noted that, in principle, the consumers may choose to buy power directly in the bulk-power markets and use the services of the distribution company for delivery of power to the consumers. In practice this opportunity may only be open to consumers with consumption above a certain quantity per unit of time. The intention expressed in the background documents to the new Energy Law is to keep this option open for the consumers. Similarly, vertical

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

205

Figure 14.2 Principal sketch of the electricity system

integration between production and distribution is considered undesirable for the functioning of the reorganized market based system. The core of the reorganized market based system is the power markets in which electricity is traded. Two power markets, a spot market and a futures market are considered essential for the new system. The need for an efficiently working spot market is clearly recognized. The spot market envisaged in the reorganized system is an extended counterpart of the existing spot market, although with open access for all participants who fulfil certain requirements for the purchase or sale of electricity in this market. An important distinction between the earlier system and the new system is the introduction of a futures market for power which provides instruments for risk management to producers and consumers in the Norwegian electricity sector. We now turn to a discussion of how to organize such a market. 14.II ORGANIZATION OF A MARKET-BASED ELECTRICITY SYSTEM: THE FUTURES MARKET Participants in the electricity industry are exposed to various types of risk. We can divide these risks into two categories; price risk and quantity risk. Price risk arises from the variations in input and output prices facing the producers, distribution companies and final consumers. For the hydro-power producers, price risk refers to uncertain prices for their production, while a distribution company mainly is concerned with the price risk of electricity as an input. Quantity or production risk arises mainly due to the hydro-power system’s dependence on access to water, which in turn is dependent on the variations in weather, such as rainfall, temperatures, etc. It is therefore important to have markets where the participants may be able to hedge against uncertainty according to their risk preference. Bilateral forward contracts in the existing system provide to some extent a means of risk management. The general inefficiency of forward contracts due to the related transaction costs, information problems and the absence of market-based pricing, state the case for a more flexible system. Futures markets offer the

206

ELECTRICITY FUTURES MARKETS

participants flexibility and liquidity allowing them to establish a position easily and to close it with equal ease. Although the futures contracts might not be perfect physical substitutes for the commodity relevant for the firms, being correlated to the firms’ risk, the futures market provides an important risk sharing instrument. By collecting and disseminating information via market prices, futures prices also constitute an important source of information. Consequently, the establishment of a futures market for standardized contracts in electricity is considered as an essential element of the reorganized electricity sector in Norway.3 Futures markets develop in response to the need for managing risks in a particular commodity market. Although efforts are underway in other countries, development of futures markets for electricity is still a recent phenomenon and there is limited experience to draw on in this respect. A basic question is: ‘Is electricity a suitable commodity for futures trading?’ Experience with commodities traded successfully on futures markets indicates that the following characteristics are common to these commodities: 1 2 3 4 5 6

homogeneity, storability, deliverability, the existence of a viable cash market in order to facilitate the delivery procedure, price variability, and the existence of speculators to take up the balance of open positions.

However, these characteristics are by no means neither necessary nor sufficient conditions for the suitability of a commodity for futures trading and would not explain the development of futures markets for nonstorable commodities such as live beef, fresh eggs, etc. or non-deliverable futures such as interest rate options or trading in share price indices. As regards electricity, it is a commodity which is physically homogeneous, (reasonable requirements as to standardization may be met), it is subject to delivery, and it is indirectly storable—by storage of fuels or water. Thus, given a spot market which ensures flexibility and variability in the price of electricity, a futures market for electricity seems a viable proposition. It may be mentioned that while the physical characteristics at best can indicate the suitability of a commodity for futures trading, the ultimate success will depend on a number of other factors. Of crucial importance is the futures contract and the existence of institutional arrangements for trading the contract. 14.III FUTURES CONTRACT: THE INSTRUMENT Designing the futures contract involves standardizing all details of the contract, except the price. The most important details of the contracts are size, length of the delivery period, length of the contract, delivery provisions, and settlement procedures. An institutional structure is also a prerequisite; a power exchange which will facilitate the clearing of these contracts. In many respects a futures market for electricity can be organized similarly to futures markets for most other kinds of commodities. In other respects, however, electricity is a somewhat special commodity, calling for special arrangements. This is especially due to peak-load problems, which require a closer a specification of both quantity and time-pattern of delivery. Given the fact that the majority of participants in a futures market for electricity will be producers, distribution companies and large users, whose main motive is to hedge against price risk, a futures contract for electricity must satisfy the hedging needs of these participants. A contract whose price movements are closely correlated with the movements in the commited spot asset is essential. Equally important is the need

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

207

to attract large participants from other energy industries and speculators in order to ensure a large trading volume and liquidity in the contract. 14.III.1 The size of the contract In general, the commodity in a futures contract is standardized with respect to grade, quantity and time of delivery. In the case of electricity futures, given the time variation problem in demand, a futures contract for electricity requires standardization both with respect to quantity, measures e.g. in kWh, and the delivery period. While standardization of the contract with respect to quantity is a question of deciding on the denomination of the contract, a decision about the delivery period is not so straightforward. The central issue here is the nature of price variation and the needs of the producers and consumers to hedge against these variations. Two aspects are particularly important in choosing the length of the delivery period: The futures contract must be suitable for hedging purposes, allowing the participants to hedge against the relevant price risk. Furthermore, the contract must allow the participants to acquire a portfolio of futures contracts adapted to variations in their energy requirements over a specified time period. This is essential whether the hedger uses the futures market solely as a financial tool or the market is actually used to take physical delivery of electricity on the futures contract. A delivery period of one year was initially considered in the Norwegian reorganization debate. As prices quoted for futures contracts apply to the whole delivery period, there will be only one futures price for the entire period. A contract having such a long delivery period is obviously incapable of capturing price volatility within a year. At contract maturity, i.e. at the beginning of the delivery period, the contract price is fixed for the whole year. The price is thus actually set long before all relevant information of demand and supply for the various periods within the year has emerged in the market. Also, correlation with the spot market will be close to zero, as the futures price in a way represents an average of the spot prices during the whole year. Consequently, we conclude that a contract with a delivery period of one year is not suited as an instrument for risk management. Furthermore, a contract with a standard specified time pattern of delivery over the year is poorly suited for physical delivery, as it does not match individual production and consumption patterns. These will typically have a high degree of variability during the year. A weekly futures contract is more useful if the objective is to hedge the intra-year or seasonal price risk. With several futures contracts within a year, hedgers can tailor their portfolio of futures contracts to their consumption and production patterns during the year. Seasonal variations in prices can be hedged with the help of a suitable portfolio, e.g. of weekly contracts. In addition, the market allows for an explicit determination of futures prices for different time periods within the year. For producers with water storage capacity, such prices provide useful information for production planning with the aim of maximizing the value of water in storage. With a functioning spot market, one would expect a high degree of correlation between the weekly futures price and the spot prices. Futures contracts with a delivery period of one week may lead to higher transaction costs compared with longer contract periods. This must be balanced against the gains which a relatively short contract period can give in terms of flexibility and adaptability, given the inherent nature of price variability in the electricity market. On balance, we conclude that a weekly contract seems to be a viable alternative.

208

ELECTRICITY FUTURES MARKETS

14.III.2 Length of contract The length of a futures contract is the length of the total trading period from the first trading day to the maturity day of the contract. Normally trading in futures contracts starts 3–18 months before the date of maturity. In the Norwegian debate concerning the development of a futures market for electricity, the electricity industry has expressed a need for opening up for trade in contracts several years ahead of maturity. This demand seems to be rooted in the traditional forward contracts in the electricity sector in Norway, where the role of such contracts was mainly as a planning and financial instrument for capacity investments; investments which may have an economic life of up to 40–50 years. While, in principle, there are no problems in opening up for trade in futures contracts several years before maturity, given the nature of uncertainty in the electricity sector, a major practical problem is that it might lead to a thin futures market. This would again affect the efficiency of the market. 14.III.3 Settlement procedures The settlement procedure is the procedure used to settle futures contract obligations between the buyer and the seller upon maturity of the contract. The settlement procedure is the mechanism whereby prices quoted on the futures market and the cash market are forced into alignment. There are two main alternatives with respect to contract settlement; cash settlement and physical delivery. Cash settlement has been increasingly used in the design of new futures contracts in recent years. Under a cash settlement arrangement the parties to the contract, for example a seller who has not offset his contract by the end of trading, compensates the buyer for the difference in the contracted value and the spot value of the contract on settlement. Physical delivery of the commodity is being arranged by a corresponding transaction on the spot market. The contracted value of the futures contract is determined by a formula specified in the contract. Another alternative for settlement is physical delivery. Upon maturity the commodity is delivered as specified in the futures contract and according to the delivery conditions. As an average for all commodity futures markets, less than one per cent of all futures contracts are actually held until maturity. Three important factors4 relevant in considering cash settlements versus physical delivery are: 1 the behaviour of prices (possible forms of arbitrage and effects on the stability of cash markets); 2 the effectiveness of hedging (the degree of correlation between futures prices and spot prices); and 3 the potential for manipulation. However, it is generally not clear which settlement procedure is best with respect to these issues. Experience indicates that cash settlement has proved favourable for commodities which have special requirements with respect to delivery, e.g. for commodities that are perishable, require special handling, or are geographically dispersed making physical delivery on futures relatively costly. For cash settlement contracts to work, traders of futures contracts must have confidence that the settlement is a reasonably accurate reflection of current values of the commodity. This is important in linking the futures price to the physical spot price and thus for the effectiveness of hedging. Problems may, however, arise in defining the settlement formula. In the case of electricity, a price index based on past prices is one alternative. On the other hand, it may be difficult to construct an index showing sufficient variability.

APPLICATION TO PARTICULAR ENERGY POLICY PROBLEMS

209

The existing interruptible power spot market and a well-developed national grid system in Norway, have resulted in delivery provisions which can be used for the settlement of futures contracts. The majority of the participants in the new futures market will probably be hedgers from the electricity industry. Traditionally, these actors are used to forward contracts in which physical delivery is obligatory. Physical delivery is thus a familiar settlement procedure in the Norwegian context. Common carriage principles are already applied in determining access to the grid. A potential problem in this respect is that physical delivery is not very suitable for pure speculators and hedgers from the non-electricity energy sectors. Failure to attract speculators to the market may affect a hedger’s costs through lack of liquidity and, in general, the efficient functioning of the market. Considering these aspects, both cash settlement and physical delivery may be possible for the settlement of electricity futures. However, the ability to make and take delivery on a futures contract appears to be crucial for initial success of a futures market. In the long run one would expect the futures market to become more financial in nature, with delivery occurring less frequently. Given the electricity industry’s needs, physical delivery seems to be the most appropriate settlement procedure for electricity futures, at least to begin with. 14.III.4 Power exchange A power exchange house is the market clearing unit which is essential for the functioning of the power markets in the reorganized electricity sector in Norway. The exchange which administers and supervises trade in the spot and futures markets should be organized as an independent company. Financial integrity and neutrality of the exchange in relation to the participants in the market should be emphasized. A system for effective internal and external controls and safeguards for such an exchange is under development. The current power exchange company (Samkjøringen) has valuable experience in operating existing interruptible power spot markets and can provide the required infrastructure for the establishment of the power exchange house in the reorganized system. However, Samkjøringen is presently owned by power production companies. In order to ensure the neutrality of the spot and futures exchange, the subject of ownership would have to be reexamined. New rules for membership are being considered by the industry. Equally important is the role of the transmission company in facilitating spot transactions and settlement of futures contracts by physical delivery. Close co-ordination of the power exchange and the transmission company is essential. However, legal separation of the power exchange corporation and the transmission company is deemed necessary to retain financial integrity and independence of the exchange. 14.IV CONCLUSIONS The electricity sector in Norway is undergoing transition from a traditionally highly centralized organization to a more economically efficient decentralized structure. The core of this reorganization is the spot and the futures markets which are stipulated in the reorganized system. While the existing spot market for interruptible power provides a starting point for the spot market part, designing a futures market is not so straightforward. The basic question is whether electricity is a suitable commodity for futures trading. Experience with futures markets for other commodities indicates that futures markets for electricity is a viable proposition. Ultimate success of electricity futures markets would, however, depend on whether the electricity futures contracts satisfy the hedging need of the actors in the market or not. Of particular

210

ELECTRICITY FUTURES MARKETS

importance are the considerations related to the size and length of the contract, settlement procedures and the establishment of a credible futures exchange to facilitate trading. On the assumption that settlement by physical delivery would be important in the initial phase, a close co-ordination between the futures exchange and the transmission company ought to be established. Norway is a major participant in the regional trade in electricity in Scandinavia. The common carrier discussions in the European Community would further extend the trade potential. Germany might become an important trade partner in this process. How these external markets should be integrated in the reorganized Norwegian electricity sector is a question that remains to be answered. A prerequisite for any such integration would be a market-orientated electricity sector in the partner countries. Integrating Norwegian electricity markets with export markets where national power companies dominate the system may distort the competitive structure of the Norwegian markets. This calls for a discussion and an analysis of how European electricity markets should be organized to obtain efficient use of resources in an international context. NOTES 1 The Centre for Research in Economics and Business Administration (SNF) in Bergen has been involved in a major research project, under a contract with the Ministry of Oil and Energy and the Ministry of Finance, with the main purpose of providing policy research related to the reorganization of the electricity sector in Norway. The views expressed in this chapter are those of the authors and do not necessarily reflect the view of the above mentioned ministries. 2 The Norwegian Water Resources and Power Board—a government directorate— had been advocating a plan for creating 20 country-wide, vertically integrated companies. As a consequence of the new ‘Energy Law’, the plan has been abolished. A fallback position taken by some in the industry is for vertically integrated companies, where production and distribution are divisionalized separately. 3 Singh, B. (1989): ‘Futures Markets for Non-oil Energy: The Hydro-Power Sector in Norway’. The paper examines the case for establishment of futures markets for electric power in Norway. A formal model of a futures market, where producers are subject to both production and output-price risk, and distribution companies face input-price risk, is presented. Optimizing equations for each of these actors are specified and market equilibrium in the futures market is described. 4 Paul, A.B.(1985): ‘The role of cash settlement in futures contract specification’, in A.E.Peck (ed.) (1985), Futures Markets: Regulatory Issues, Washington DC: American Enterprise for Public Policy Research.

Index

acid rain(s) xxii activity; economic; macroeconomic 33, 129 Africa 25–6, 53 Algeria 26, 106–8, 111 algorithm 45; Gauss-Seidel 45 America 26; North 25, 84; South 25, 84 Arabia (Saudi–) 135 Asia 25–7 Australia 22, 26–7

concavity; of translog cost functions 7–8 consumption see demand contingent evaluation (of environmental goods) xxx cracking (in oil refinery) 56 crisis; gulf 139; energy 139 decision; making 91; modelling 41 demand; coal 58–62, 130–33; crude oil 160–75; electricity 10, 48–53, 130–3, 239–42; energy xxxi, 3–4, 30–1, 33, 45, 65, 105, 130–3, 139– 59, 188–91, 200–17, 218–37; energy, management 188; final energy 234; natural gas 106–9, 118–19, 130–3; oil products 190–1; primary energy 105 Denmark 110, 177–90, 193, 195 deregulation 181 Dickey-Fuller test (DF) 140; augmented (ADF) 140 disequilibrium 41, 43 distillation (of crude oil) 54 dollar 22–4, 33, 45; Australian 23; US 22–4, 33, 45 double divided xxxii Durbin-Watson test (statistic) (DW) 18–19, 22–3, 140–1, 143–6, 147–9, 152–3 duty; excise (ED) 176–97

Belgium 106–8, 110, 113–14, 177–90, 193, 195 Brunei 26 Canada 25–6, 84–5, 139–59 carbon tax xxii, 136–8, 198–217, 231–5 cartel; of major seven oil companies (BP, Exxon, Gulf Oil, Mobil, Shell, Standard Oil of California, Texaco) 160; of OPEC countries 16, 160 China 26, 135 cointegration test 140–59 Community of Independent States (CIS, see also ex-USSR) 129–36 Colombia 26 coal 15, 22–3, 25–8, 37, 58–62, 69–73, 137; marginal cost (long run) 60; prices 60; sub-model 58–62; unit investment cost 60; unit fixed cost 60; unit variable cost 60 coke 69–73 211

212

INDEX

elasticity; GDP energy demand, or GNP energy demand 139–59; price (of demand) 6–10, 139–59, 190–1, 226–30; price (of supply, non-OPEC oil) 16; substitution 6–10 electricity 44, 48–53, 238–49; futures markets (in Norway) 238–49; production 44; emission (of pollutants) 14, 41, 104, 130, 189, 231–5; CO2 14, 41, 189, 218; greenhouse gas 130 electricity demand 49–50, 239–40 endogenization 128 endogenous 32; price variables 32 energy xxi–xxii, xxiv–xxv, xxix–xxxii, 3–13, 14–15, 20, 30–4, 37, 41–8, 51, 62–4, 65, 82, 105, 122, 127–35, 137–8, 139–55, 159, 172, 188–91, 196, 198–206, 208– 10, 212–16, 218–21, 224–30, 234, 236; fossil fuel xxi; new sources xxi; nuclear xxi; rational use of (see also energy conservation) xxi energy aggregate 7 energy balance 46 energy conservation 147, 188–91, 194, 196, 200–17, 218– 37; investments 200–17, 218–37; in European countries (France, Germany, Italy, UK) 200–17; in Japan 218–37 energy demand xxxi, 3–4, 30–1, 33, 45, 65, 105, 130–3, 139–59, 188–91, 200–17, 218–37; in iron and steel industry (in Japan) 224–8; in industry 140–4; in commercial and residential sectors 144–7; in transportation 147–50; in periods of rising and falling prices 150–4; in various industries or sectors 201, 224; of electric utilities (in Japan) 228–31 energy intensity 131–2 energy modelling xxi, 3–13 energy policy 44 energy price determination 4–5 energy products 45 energy production 30 energy saving(s) (see also energy conservation) 188

energy supply 4, 31, 44–8 energy, -environment (interactions) xxi–xxii, xxiv environment 3, 42, 46; -al policy 46; -al economics 3; -al impacts environmental see environment error correction (equations, models see also cointegration) 143–59 Europe 26, 27, 30–64, 104–23, 176–97; Central 25; Eastern 27; Western 26 European Community (EC, see also European Union, EU) 30–64, 104–23, 176– 97; commission of the (CCE) 37 exchange rate(s); ECU/US Dollar 32–3 externality (-ties) (external effects) xxx–xxxii, 189–90 excise duties (ED) 176–97 exogenous 30, 32, 33, 49, 56 feedback 62 fiscal 176–97; harmonization 176–97 forecast; errors 34–7 France 65–83, 106–8, 110, 111, 113–19, 139–59, 177–90, 193, 195 fuel; efficiencies 5; prices 32; (market) shares 132–4 gas (natural) 20–2, 25–8, 37–41, 84–103, 104–23; contracts (in North America) 84–103; market (in EC Europe) 104–21 game theory 11 Germany (Federal Republic of) 106–8, 110, 113–19, 139– 59, 177–90, 193, 195 Greece 107, 110, 113, 139–59, 177–90, 193, 195 greenhouse gas effect 130 gross domestic product (GDP) 33, 128–9, 140–5, 147–9, 151–4, 198, 203, 205–15 gross fixed capital formation (GFCF) 208–10, 212–23 gross national product (GNP) 218–20 group of seven countries (Canada, France, Germany, Italy, Japan, UK, USA) (G-7) 139–59

INDEX

gulf (Arabo-Persian) 15, 18, 24, 139 Hotelling’s rule 16 Indonesia 26, 27 input-output 65–83; tables 66 interindustrial system 69, 71, 78, 80 Iran 27 Iraq 27 Ireland 107, 177–90, 193, 195 Italy 106–8, 110, 113–16, 118–19, 139–59, 177–90, 193, 195 Japan 22, 139–59, 218–37 Keynesian multiplier 205, 207 koyck lag 33 Kuwait 135 lag (time-) 33, 47, 142–4, 147–9; distributed 47; koyck 33 lignite 52 liquefied natural gas (LNG) 45 load; duration curve (electricity) 49–50; duration curve (natural gas) 58; factor 52 Luxembourg 106, 108, 177–90, 193 Malaysia 26 market xxxiii, 14–29, 42–3, 84–5, 238–49; coal 22–4, 25–7, 58–62; deregulated 84; futures (forelectricity, in Norway) 238–49; energy 14–29; natural gas 20–2, 25–7, 84–103, 104–21; oil 16–20, 24–5, 160–75; oil products 176–97; Single European 176; steam coal (boiler coal) 22–3 market structure xxxii Markowitz (portfolio theory) 84 Mexico 27 Middle East 15, 53, 135 model; Argonne Coal Market 59; BESOM 42;

213

bottom-up xxiii; coal (supply) 58; coal and electric utility (of US DOE) 59; decision-aid xxi; discrete choice 10; DMS 43; DRI 43; DRI/Zimmermann 59; econometric energy xxv, xxxii, 30–40, 41–64, 139– 59; EFOM 42; ERASME 30–40; ETA 42; forecasting xxiii; gas analysis modelling system 57; gas contract portfolio 84–103; HERMES 41, 43, 198–217; Hotellinian (based on Hotelling’s rule) 16; hysteresis 19; IEA 127–138; IFFS 42; input-output 65–83; linear programming 53, 59; macroeconomic xxiii; methodologic aladvances in energy 3–13; MEFS 42; MIDAS 41–64, 190–1; national xxiii; oil market simulation (of US DOE) 17; Newcomb coal price index (quadratic programming) 59; PIES 42; PILOT 42; POLES 14–29; quadratic programming 59; recursive 30–40; short-term energy forecasting 30–40; SOSIE GAZ 104–23; stochastic programming 84–103; STEM 30; STEO 30; target capacity utilization (TCU) 16–17; technically intensive xxv; TENRAC 57; TESOM 42; Translog 6–8; vector auto-regressive (VAR) 160–75 Morocco 107

214

INDEX

natural gas see gas (natural) Netherlands 26, 106–8, 110–1, 113–14, 118–19, 177–90, 193, 195 Nigeria 27, 107 Norway 26, 106–8, 111, 238–49 OECD 127–38 oil 16–20, 22–5, 32, 37, 41, 56, 73–8, 160–75, 176–97; distillates (light, medium, heavy) 44; price 16–20, 22–4, 32, 160–75; spot market 160–75; products 37, 73–8, 176–97 OPEC 16–17 panel data 8–10 petroleum see oil planning; long-range 41 Poland 26 portfolio theory 84–6 Portugal 107, 177–90, 193, 195 power plant (s) 33 price; coal 15, 22–3, 58–62; crude oil 5, 16–22, 56, 129–30, 136, 160–75; electricity 51–3; energy (in general, or aggregate) 150–4, 219; fuel energy 32–3,47; natural gas (-ing) 20–2, 57, 84–103, 107; oil products 56, 177–97, 221 production function; Cobb-Douglas 7; Leontief 4; Translog 6; Putty-Clay 200 programming; linear 53, 59, 85; quadratic 59; stochastic 84–103 Qatar 27 rationality; procedural (after H.Simon) 16; substantive 16 recursive (model) 15, 30–40 refinery (crude oil) 53–6 refining see refinery reforming (in crude oil refinery) 54–6

Saudi Arabia 135 scenario 97–8, 104–21, 214–16 Shephard’s lemma 6 shock; counter 17; energy 139–59, 218; oil, 17, 34, 65, 155; price 155 solid mineral fuel (s) see coal South Africa 26 Spain 106–8, 113–14, 177–90, 195 supply; coal 44–5, 58–62; electricity 44, 48–53, 239–42; energy; 44–8; natural gas 44–5, 56–8, 91–9, 106–7, 116–18; oil (crude) 134–5, 160–75; oil (refined products) 44–5, 53–6 swing producer 18, 27 tax xxxii, 33, 45, 136–8, 176–97, 198–217, 231–5; carbon xxii, 136–8, 198–217, 231–5; indirect on oil products (in the EU) 176–99; Pigovian xxxii; value-added (VAT) 45, 176–81, 191–6, 199 tax harmonization (on oil products, in the EU) 176–99 taxation see tax test; augmented Dickey-Fuller (ADF) 140–1, 146, 148–52, 161; Dickey-Fuller (DF) 140–1, 146, 148–52, 161; Durbin-Watson (DW) 18–19, 22–3, 140–1, 143, 146, 147–9, 152–3, 161; student 118–19, 22–3, 141, 143, 146, 147–9, 152–3, 165–8 Thailand 27 three-stage least squares (3SLS) 31 time-series reactive on line laboratory (TROLL) system 45 top-down; approach, model xxiii trade matrices; for world trade in natural gas and coal 25–7 translog 6–8; cost function 6–8; production function 6 United Arab Emirates (UAE) 26

INDEX

United Kingdom (UK) 106–8, 110, 113–14, 139–59, 177– 90, 195 United States of America (USA) 22, 139–59 Union of Soviet Socialist Republics (USSR, ex-, former-, see also CIS) 25–7, 106, 108, 111, 116–19 utilities; electricity 49 value-added 4, 140 value-added tax (VAT) 45, 176–81, 191–6, 199 Venezuela 27

215