Competitive Electricity Markets and Sustainability

Competitive Electricity Markets andSustainability

CompetitiveElectricity Marketsand Sustainability

Edited by

François Lévêque

Professor of Law and Economics, Centre of IndustrialEconomics (CERNA), École des mines de Paris, France

Edward ElgarCheltenham, UK • Northampton, MA, USA

© François Lévêque 2006

All rights reserved. No part of this publication may be reproduced, stored ina retrieval system or transmitted in any form or by any means, electronic,mechanical or photocopying, recording, or otherwise without the priorpermission of the publisher.

Published byEdward Elgar Publishing LimitedGlensanda HouseMontpellier ParadeCheltenhamGlos GL50 1UAUK

Edward Elgar Publishing, Inc.William Pratt House9 Dewey CourtNorthamptonMassachusetts 01060USA

A catalogue record for this bookis available from the British Library

Library of Congress Cataloguing-in-Publication Data

Competitive electricity markets and sustainability / edited by François Lévêquep. cm. –

Includes bibliographical references and index.1. Electric utilities. 2. Competition. 3. Electric Utilities—Finance.4. Demand-side management (Electric utilities) I. Lévêque, François, 1957–

HD9685.A2C577 2006333.793�23–dc22

2006011132

ISBN-13: 978 1 84542 921 8ISBN-10: 1 84542 921 4

Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall

Contents

List of figures viList of tables viiList of contributors viiiPreface by Jean Syrota xiiiAcknowledgements xv

1 Investments in competitive electricity markets: an overview 1François Lévêque

PART I INVESTMENT IN GENERATION

2 Investment and generation capacity 21Richard Green

3 Generation technology mix in competitive electricity markets 54Jean-Michel Glachant

PART II INVESTMENT IN TRANSMISSION

4 Problems of transmission investment in a deregulated powermarket 87Steven Stoft

5 Patterns of transmission investments 131Paul Joskow

PART III COORDINATION BETWEEN INVESTMENTSIN GENERATION AND TRANSMISSION

6 Long-term locational prices and investment incentivesin the transmission of electricity 187Yves Smeers

7 Compatibility of investment signals in distribution,transmission and generation 230Ignacio Pérez-Arriaga and Luis Olmos

Index 289

v

Figures

2.1 The determination of electricity capacity and prices 232.2 How the capacity mix affects revenues 272.3 Investment in England and Wales 422.4 Investment in Finland 432.5 Investment in Norway 442.6 Investment in Sweden 442.7 Investment in the United States 452A.1 The determination of electricity capacity and prices 493.1 Present-day cost of generating electricity in the UK,

2003/04 693.2 CCGT cost of entry by country in Europe in 2005 703.3 Spark spread in Texas, 1999–2002 783.4 Finnish comparison of generation costs 804.1 Defining congestion rent and congestion cost 894.2 Cost to consumers compared with congestion cost and rent 904.3 Relationship of congestion to a transmission-cause

reliability problem 924.4 A positive present value is not sufficient 974.5 Lumpy technology may not exhibit returns

to scale in the long run 1014.6 Option rights reduce the feasible set of rights 1114.7 Optimal investment in lump technology may be preferable 1144.8 Optimal investment eliminates congestion 1164.9 Investors should not capture full social benefit 1177.1 Process of computation of locational signals 2527.2 Average L and G tariffs in Europe 2657.3 L nodal tariffs in Europe 2667.4 G nodal tariffs in Europe 2667.5 Original and new L tariffs in Spain for the IEM-13 system 2697.6 Original and new G tariffs in Spain for the IEM-13 system 2707.7 Comparison between the transmission tariff and the

net inter-TSO payment for 17 European countries 2717.8 Evolution of the energy price of several power exchanges

belonging to Europex, January 2000 to November 2004 2717A2.1 Proportionality principle in average participations 285

vi

Tables

3.1 Generation capacity in the USA, 1990–2002 573.2 Generation capacity in California and Texas 583.3 Generation capacity in Norway, England and Wales,

Spain and Italy 593.4 Generation fuel mix in the USA, 1998–2002 603.5 Generation capacity changes in the USA, 1990–2002 633.6 Generation capacity changes in California and Texas 653.7 Generation capacity changes in England and Wales,

Spain and Italy 663.8 1996 forecast costs of producing electricity, 2000 and 2015 693.9 2004 forecast costs of producing electricity, 2010 and 2025 703.10 Nuclear generation costs in the early twenty-first century 723.11 Nuclear generation costs in the 2003 MIT study 743.12 Nuclear versus gas CCGT cost of capital analysis 764.1 Three views of congestion 905.1 Reliability upgrade projects: New England regional

expansion plan 2004 1435.2 Schedule of transmission network use of system

generation charges, 2004/2005 1605.3 Schedule of transmission network use of system demand

charges and energy consumption charges, 2004/2005 1615.4 E&W system operator incentive mechanism under NETA 1625.5 PJM interconnection charges: proposed Erie West HVDC 1705.6 Market window ‘economic’ transmission projects

in PJM as of November 2004 1745.7 Examples of transmission congestion mitigated by

reliability investments in PJM 1787.1 Impact of different factors on the total generation

capacity needed to supply a 384-MW load, locatedclose to a main consumption center, from two differentlocations, one close to the load center and the otherclose to an entry point for LNG 272

7.2 Comparison of the cost savings involved in supplyinga 384-MW load located close to a main load center 274

vii

Contributors

Jean-Michel Glachant is tenured professor and Head of the Departmentof Economics at the University of Paris Sud (France) where he createdthe Network Industry Research Group (GRJM). Prior to joiningUniversity Paris Sud in 2000, throughout the 1990s he was deputy direc-tor or director of the leading French institutional economics researchcentre (ATOM) at the Sorbonne University. His work focuses on the insti-tutional economics of competitive reforms in the European networkindustry. His current work focuses on the creation of a single energymarket in the European Union extended to 25 member states. He hasadvised the European Commission (DG Energy and DG Competition)on electricity reforms. He has been a member of the Economic AdvisoryCommittee at the French Energy Regulatory Commission. He is amember of the board of the International Society for New InstitutionalEconomics (ISNIE) and of the Faculty of the European School forInstitutional Economics (ESNIE); and a partner of the Electricity PolicyResearch Group at the University of Cambridge, and of the EuropeanEnergy Institute (EEI). He received his PhD in economics from theSorbonne University.

Richard Green is professor of economics at the University ofBirmingham, UK and Director of the Institute for Energy Research andPolicy. He has been studying the economics and regulation of the elec-tricity industry since 1989, just before the industry in England and Waleswas privatised. With David Newbery, he was responsible for the mostinfluential study of competition in the British electricity spot market. Hehas written two books, and more than 40 articles and book chapters,mostly on the electricity industry and its regulation. He is an associateeditor of the Journal of Industrial Economics, and on the Editorial Boardof the Journal of Regulatory Economics. He has spent a year on second-ment to the Office of Electricity Regulation, and has been a visitingFellow at the World Bank Institute, the University of California EnergyInstitute and the Massachusetts Institute of Technology. He has been aspecialist advisor to the House of Commons Trade and IndustryCommittee, and is on the academic advisory panel to the staff of the UK’sCompetition Commission.

viii

Paul Joskow is Elizabeth and James Killian Professor of Economics andManagement at the Massachusetts Institute of Technology and Directorof the MIT Center for Energy and Environmental Policy Research. Hereceived a BA from Cornell University in 1968 and a PhD in economicsfrom Yale University in 1972. Professor Joskow has been on the MITfaculty since 1972 and served as Head of the MIT Department ofEconomics from 1994 to 1998. At MIT he is engaged in teaching andresearch in the areas of industrial organization, energy and environmen-tal economics, competition policy and government regulation of industry.He has published six books and more than 120 articles and papers in theseareas. His papers have appeared in the American Economic Review, BellJournal of Economics, Rand Journal of Economics, Journal of PoliticalEconomy, Journal of Law and Economics, Journal of Law, Economics andOrganization, International Economic Review, Review of Economics andStatistics, Journal of Econometrics, Journal of Applied Econometrics, YaleLaw Journal, New England Journal of Medicine, Foreign Affairs, EnergyJournal, Electricity Journal, Oxford Review of Economic Policy and otherjournals and books. He is a Director of National Grid plc, a Director ofTransCanada Corporation and a Trustee of the Putnam Mutual Funds.He previously served as a director of New England Electric System. Hehas served on the US Environmental Protection Agency’s Acid RainAdvisory Committee and on the Environmental Economics Committeeof the EPA’s Science Advisory Board. He is a member of the ScientificAdvisory Board of the Institut d’Économie Industrielle (Toulouse,France) and the Scientific Advisory Board of the Conservation LawFoundation. He is a past-President of the International Society for NewInstitutional Economics and a Fellow of the Econometric Society and theAmerican Academy of Arts and Sciences.

François Lévêque is professor of law and economics at École des mines deParis and visiting professor at the University of California at Berkeley. Heis Director at Cerna, the research centre of the École des mines de Paris inindustrial economics. He has published several books on antitrust eco-nomics (Antitrust, Patents and Copyright, Edward Elgar, 2005; MergerRemedies in American and European Union Competition Law, EdwardElgar, 2003), on the economics of regulation (Économie de la réglementa-tion, Éditions La Découverte, 1999 and 2005; Transport Pricing ofElectricity Networks, Kluwer Academic, 2003) and on the economics ofintellectual property rights (Economics of Patents and Copyright, BerkeleyElectronic Press, 2004). He is the author of 50 articles in the same areas. Hehas coordinated several large European research programmes on electric-ity reforms and energy policy.

Contributors ix

Lévêque taught economics of natural resources at the École des mines deParis (1984–90), and environmental economics at École des Hautes Etudesen Sciènce Sociàles (EHESS) (1997–2001) and at Pavia University(1999–2002). In 1999 he created a new major in law and economics at theÉcole des mines. He has taught industrial economics at the École des minessince 1996 and energy economics since 2004. He has also taught EUCompetition Law at the Boalt School of Law, University of California atBerkeley, since 2002.

He has regularly been commissioned by the French government, theOrganization for Economic Cooperation and Development (OECD) andthe European Commission to undertake consultancy and to participate inadvisory committees. He founded Microeconomix, a Paris-based boutiquespecialising in the economic analysis of legal disputes. He is also a memberof the French Environment Accounting Commission and of the Councilon Intellectual Property.

Luis Olmos was born in Madrid, Spain, in 1976. He received an ElectricalEngineering degree and a PhD degree from the Universidad PontificiaComillas (UPCO) in 2000 and 2006, respectively. Currently, he is aresearcher at the Instituto de Investigación Tecnológica. His interestsinclude areas such as the regulation of electricity markets and planning ofpower systems. He has worked on several aspects of the operation of powersystems, such as the provision of ancillary services (load-frequency regula-tion). Currently, he is working on transmission pricing issues in the contextof regional markets with a special focus on the problems of congestionmanagement, sunk costs recovery, tariff design and grid expansion.

Ignacio Pérez-Arriaga was born in Madrid in 1948. He received a degree inelectrical engineering from Comillas University, Madrid, Spain, and MSand PhD degrees in electrical engineering from the Massachusetts Instituteof Technology (MIT), Cambridge, USA.

He is Director of the BP Chair on Sustainable Development and FullProfessor of electrical engineering at Comillas University, where he wasFounder and Director of the Instituto de Investigación Tecnológica (IIT)for 11 years, and has been Vice-Rector for Research. For five years he servedas a commissioner at the Spanish Electricity Regulatory Commission. He isa life member of the Spanish Royal Academy of Engineering and Fellow ofthe Institute of Electrical and Electronic Engineers (IEEE). He is Directorof the annual Training Course of European Energy Regulators at theFlorence School of Regulation within the European University in Florence.He was the author of the White Paper on the Spanish electricity sector com-missioned by the government in 2005. He has been principal researcher in

x Contributors

more than 40 projects and has published more than 100 papers in nationaland international journals and conference proceedings. He has worked andlectured extensively on power system dynamic analysis, monitoring anddiagnosis of power system devices and systems, intelligent computer designof industrial systems, planning and operation of electric generation and net-works, and regulation of the electric power sector. In this last topic he hasbeen a consultant for governments, international institutions, industrialassociations and utilities in more than 30 countries. His current researchinterests are centered on regulation of the electric power industry, the designof regional electricity markets and energy sustainability.

Yves Smeers is the Tractebel Professor of Energy Economics at theUniversité Catholique de Louvain in Belgium where he is affiliated with theDepartment of Mathematical Engineering and the Center for OperationsResearch and Econometrics. He received an engineering degree from theUniversité de Liège in 1967, and a degree in economics from the UniversitéCatholique de Louvain in 1969. He also obtained an MS degree in indus-trial administration and a PhD in operations research from CarnegieMellon University in 1971 and 1972, respectively. His current researchinterests concentrate on computational equilibrium models and riskmanagement in restructured electricity and gas systems. His experience inthe area extends from operational to strategic market simulation models.He has published extensively in the area and acted as project leader onmany projects for the European Commission, the World Bank, the OECDand the Belgian government. He has also conducted various assign-ments for major European gas and electricity companies, as well as forregulators. He is currently scientific adviser at the Department of Strategyof Electrabel/Suez where he works on market simulation models and riskmanagement. He has recently published several articles in OperationsResearch, the Journal of Network Industries, Networks and SpatialEconomics and Utilities Policy.

Steven Stoft is an economist and independent consultant with 12 years’experience in power market analysis and design. He is the author of PowerSystem Economics, Designing Markets for Electricity (IEEE, 2002) and alsoof many published article on electricity market design. He has advisedPJM’s Market Monitoring Unit since 1999, was an expert witness forCalifornia’s Public Utility Commission and Electricity Oversight Board(EOB) in their litigation over long-term contracts before the Federal EnergyRegulatory Commission (FERC). Beginning in 2004, he has worked withISO New England in designing their Investment Capacity Payment (ICAP)market and was their expert economic witness before FERC. He is also

Contributors xi

working with the EOB on installed capacity markets for California.Previously he was a Senior Research Fellow at the University of CaliforniaEnergy Institute, worked on regulatory and restructuring issues at theLawrence Berkeley National Laboratory and spent a year in the Office ofEconomic Policy at FERC. He received his BS in engineering math and hisPhD in economics from the University of California at Berkeley.

xii Contributors

PrefaceJean Syrota, Chairman (2000–2006), Commission deRégulation de l’Energie

At the beginning of the liberalisation of electricity markets during the1990s, the electricity industry was in a situation of plentiful generation andtransmission. Then in order to monitor the liberalisation process, muchattention was paid to the regulatory mechanisms that encourage the actorsto improve the efficiency of their short-term operations and to deal with thetechnical complexity that was faced in substituting daily free energymarkets for the traditional centralised dispatch without reducing the safetyof operation or downgrading the economic efficiency of the previouslyintegrated systems.

Accordingly, a large number of researchers and practitioners workedactively to define consistent rules dealing with daily markets, redispatching,balancing mechanisms and ancillary services procurement in parallel withgrid-use tariffs and ex ante congestion management methods. Despite thework done, improvements are still possible and desirable.

Meanwhile, with the increase of the overall demand and the reduction ofgeneration margins notably entailed by the retirement of the oldest gener-ators, significant spot price increases have been observed and the gridbecame congested more frequently, especially at interconnections betweenthe European countries. This situation raised some new issues that werelargely ignored in the first phase of the process.

Especially in a liberalised context, these issues are strongly related to thedefinition of appropriate incentives for investment in generation and trans-mission. In the electricity sector there are difficulties that are not presentelsewhere. As far as the grid is concerned, there is lumpiness in a number ofcomponents and in a significantly meshed network, a very common situ-ation in the real world, there are numerous externalities between individuallines, making it difficult or even impossible to compute their incrementalindividual benefits. Moreover, even if generation and transmission areoften complementary goods, there are also many situations where they arereplaced, for example when grid improvements can reduce or delay the needfor new generation and conversely.

xiii

Thus generation and transmission are theoretically coupled and the ver-tically integrated structure that prevailed for a long time was supposed toallow a joint optimisation. When transmission and generation are properlyunbundled, as required by European law, the joint optimisation can nolonger be achieved, but some kind of coordination must be found to main-tain an acceptable level of efficiency without hindering the competitionbetween generators. Additional difficulties come from the length of timenecessary to achieve new generation and transmission infrastructures withthe associated risks for security of supply in the case of late commission-ing. The regulator must assess to what extent coordination mechanisms arecompatible with the requirements for the achievement of a fair and efficientcompetition in supply.

To deal with such complex issues, the French Regulatory Commissionfor Energy (Commission de Régulation de l’Energie: CRE) found it usefulto promote academic research by prominent specialists of the liberalisationprocess in electricity markets. This book is the result of that research. TheCRE is indebted to the authors for their very interesting analyses of theinvestment decision-making process and of the possibility of reconcilingthe perhaps diverging interests of two major parts of the electric industry.I am sure that these analyses will also be of interest to those who want tobetter understand the multiple aspects of the ongoing liberalisationprocess.

xiv Preface

Acknowledgements

The origin of the book is based on discussions between the authors whomet in January 2005 in Paris at the Commission de Régulation del’Electricité, the French Regulatory Electricity Authority. The authors aregrateful to the CRE and especially to Michel Massoni for nurturing thedebates. The book would not have been written without their support. Theusual disclaimers apply.

xv

1. Investments in competitiveelectricity markets: an overviewFrançois Lévêque

1. INTRODUCTION

Over the course of the past 20 years, most of the countries in the OECDhave engaged in a competitive opening of their electricity markets. Theincumbents were stripped of their legal monopolies, wholesale marketswere formed, and dedicated organisations assumed management of thetransmission grid. Large consumers acquired the ability to choose theirelectricity supplier. This opening to competition brought about a profoundchange in terms of the investment in both generation and transmission.Decisions concerning the construction of new power plants, in particularthe timing and the technology mix (that is, the proportion of hydro elec-tricity, nuclear, thermal and so on) now depend on decentralised initiativesof investors, and not on public authorities. As to transmission, whichremained a monopoly, the reinforcement and expansion of high-tensionpower lines are no longer directly controlled by the generators. Systemoperators have greater leeway for initiative. Depending on the specific case,they can sell financial transmission rights, submit investment programmesto the regulatory authority, or invest as they see fit.

In a word, investments in an electricity system that is open to competi-tion will no longer be coordinated by the same mechanisms as in the past.The planning that enabled a monopolistic and vertically integrated pro-ducer to adjust base- and peak-load capacities, as well as generation andtransmission capacities, has been replaced by a series of decentralised deci-sions partly based on prices. This new decision set – which involves manyagents and combines market signals with regulation – must be understoodin detail. A thorough understanding is necessary to reveal to what extent,and under what conditions, competitive opening will result in an invest-ment level that is consistent with the public interest. Only this will allowidentification and evaluation of solutions to situations of investment short-fall or oversupply such as those we have seen arise on several occasions (forexample, underinvestment in interconnection capacity in California, and

1

overinvestment in independent gas-powered plants in the United Statesduring the 1990s.) This is the spirit in which this book was prepared.

This first chapter contains seven sections. Following this introduction,Section 2 reviews the new terms of investment in generation and transmis-sion. Sections 3 and 4 address investment in generation and in transmission,respectively. Section 5 resumes the discussion of the interface between invest-ments in generation and transmission that we briefly began in Section 2.Section 6 provides a preview of some essential points that the co-authors ofthis book raise in subsequent chapters, and that were not mentioned in thepreceding sections. Finally, Section 7 concludes.

2. THE ISSUE

Ideally, an optimal level of investment in the electricity system wouldinvolve joint optimisation of investments in generation and transmission. Infact, the goal is to minimise the cost of electricity to consumers. From aneconomic perspective, generation and transmission are complementarygoods; if the price of one decreases, the quantity sold of the other increases.The mechanism underlying this phenomenon is simple: consumers are sen-sitive only to the total price of electricity since they do not consume the gen-erated electricity and the transmission service separately. Consequently, ifthe price of a KWh falls, ceteris paribus, they will consume more electricityand demand a greater quantity of the transmission service. Consequently,investments in generation and transmission complement each other.

Sometimes, however, investments in generation and transmission aresubstitutable. For example, in an isolated region with limited interconnec-tion with the grid, a rise in local demand can be satisfied by either rein-forcing the line or building a new power plant within the zone. If bothinvestments occur simultaneously, then neither will be profitable. Whenboth activities are combined within a single firm, joint optimisation ofinvestments is deemed self-evident, since the stockholder or manager max-imises overall profits. In an electricity system that is open to competition,the visible hand of the manager fails to ensure coordination betweengeneration and transmission. Transmission is separated from generation inone way or another (that is, accounting, managerial or legal unbundling)in order to ensure that rival generators have equitable terms of access tothe grids.

According to Steven Stoft, this new situation opens the door to strategicbehaviour on all sides. In order to provide for future investments in trans-mission, the transmission system operator (TSO) must be informed offuture investments in generation. Conversely, to plan these investments in

2 Competitive electricity markets and sustainability

generation, producers require forecasts of the TSO’s future investments inthe grid. To escape from this deadlock, one of these stakeholders must‘draw first’ by revealing its intentions and proceeding with the investment.However, the first to invest becomes hostage to the other, since it is impos-sible to move a power plant, or pylons, without forfeiting the bulk of theirvalue. This is the classical economic problem of the hold-up occasioned bystranded costs. The upshot is generalised underinvestment: each party,knowing that it may be taken hostage ex post, reduces investments ex ante.Thus, we cannot apply the idealised rule for investment in transmission,which would have the system operator (SO) plan investment by optimisingtransmission and generation as a function of future demand and then laythe power lines in the hope that the market will induce generators to investaccording to plan.1

None the less, it is necessary to avert a profusion of waste by findingsome way to coordinate investments in generation and transmission.Various instruments, such as financial transmission rights and a zone-based rate structure for the grid, have been proposed in the recent economicliterature. These are described and discussed in the chapters by StevenStoft, Yves Smeers, and Ignacio Pérez Arriaga and Luis Olmos. Beforeexamining them more closely, it will be useful to examine the optimisationof transmission and generation separately. Although this simplifies theissue considerably, these two issues taken individually are far from trivial.We shall now examine how to optimise the utilisation and size of the gridwhen generating capacity is optimal, and how to optimise the utilisationand volume of generating capacity when the grid is optimal.

3. INVESTMENT IN GENERATION

Apart from grid constraints, what obstacles must the market mechanismcontend with to yield a socially efficient level of investment in generation,that is, a level that satisfies the users’ needs at the lowest cost?

The optimal investment in electricity generation is precisely determinedby the theory, which addresses both total capacity and its distributionamong power plant types. These latter, in fact, differ in terms of bothvariable costs, which are usually linked to the price of fuel, and fixed costs,which essentially reflect expenditures on construction. For a nuclear powerplant, the former are low and the latter very high; for a gas turbine this isinverted. Consequently, nuclear plants should be used throughout the yearto meet base-load requirements, while gas turbines should only be called onto meet peak-load demand at times of the year when there are spikes indemand.

Investments in competitive electricity markets: an overview 3

In Chapter 2, Richard Green presents a simple model of optimal levels ofgenerating capacity comprising only these two types of plants. Emphasisinga graphical approach, he demonstrates how to identify the load-durationcurve for the 8,760 hours in a year and how to translate it into an hourlyprice curve. Naturally, the highest price is found when demand is greatest.As this demand exceeds available capacity, the equilibrium price is not set atthe marginal cost of the last unit generated, but rather at a higher level equalto the marginal opportunity cost of consumption (that is, above which thelast consumer prefers to forgo rather than consume). The gap between thesetwo marginal costs thus allows the peak-load plant that operates for theshortest period during the year to cover its costs. Note that this result con-tradicts the conventional wisdom that the electricity market is incapable ofensuring that plants’ fixed costs are covered. This confusion arises from anoverly hasty equating of the equilibrium price with the marginal cost ofgeneration. In the presence of congestion, as during extreme peaks inthis case, the shortage must be managed and resources allocated to thoseeconomic agents on whom the lack of access imposes the greatest cost.

Furthermore, as Green reminds us, economic theory demonstrates thatif the peaking plant that is used least covers its total costs, and if theallocation among the various means of generation is efficient, then all otherplants can cover their total costs with market prices that are based on mar-ginal costs.

We note that investors clearly had no doubts regarding the ability ofelectricity markets to render new investments in generation profitable. Inthe United States, as in England, there was even talk of a boom in the con-struction of new power plants, in particular those based on combined-cycle gas turbine (CCGT) technology. At the end of his chapter, Greenexamines trends in gross and net investment (the latter accounts for thedecommissioning of old plants) and of the capacity margin in those twocountries, as well as in Finland, Norway and Sweden. He particularlynotes two phenomena. First, the capacity margin is shrinking. This resultis consistent with the expected and desired results of the electricity systemreforms, in the sense that the previous regime was characterised by excesscapacity attributable to cost-plus regulation. Second, beyond a certainthreshold, the shrinking of the capacity margin serves as a trigger tostimulate the resumption of investment. In his chapter, Jean-MichelGlachant also draws attention to the shift in the energy mix towardsgas-based electricity generation. He measures and comments on it in thecase of several US states, England, Italy and Spain. This evolution isconsistent with developments in the relative performance of the differenttechnologies, as the total cost of CCGTs has fallen below that of nucleartechnology.


The preceding economic model assumes that there is no uncertainty interms of demand.2 However, consumers’ reactions to price changes are verypoorly understood. Except in the case of certain large consumers, whoadjust their consumption to variations in the real-time prices on the spotmarket or accept compensation for forgone consumption, information onthe price sensitivity of demand is inadequate. Most consumers are not con-fronted with hourly, or even daily, fluctuations in the price of electricity.Their consumption is measured on a monthly or quarterly basis, and theyare charged a rate per KWh that is independent of the hourly distributionof their consumption. Shielded thus from real-time price volatility, theyhave no need to hedge against the risk of high prices. Furthermore, mostdomestic consumers cannot be disconnected individually. And yet, there isno reason to believe that residents of residential neighbourhoods will facethe same opportunity cost of not consuming. However, since they are allhooked into the same distribution network, creating a market of inter-ruptible contracts cannot be readily envisaged. Consequently, there is nomechanism for revealing households’ willingness to pay during peak hours.

Note that the underlying problem of short-term price inelasticity ofdemand did not originate with the opening of electricity systems to com-petition. Under the previous arrangement, estimates of the value of elec-tricity lost in the event of a service interruption (value of loss load, orVOLL) were simulated by the planner in order to decide when generationcapacity needed to be boosted. When the cost of the new investment waslower than the benefit of the averted service interruption – VOLL multi-plied by the reduction in risk of blackout attributable to the increasedcapacity (loss of load probability, or LOLP) – the investment was deemedworthwhile. To fix an order of magnitude, if VOLL is €10,000 per MWh,then the public interest is served by the construction of a power plant thatwill reduce the risk of interruption by approximately five hours over thecourse of a year. Today, with electricity systems that are open to competi-tion, VOLL can also serve as a reference value. For example, during criti-cal periods, a system operator may decide to purchase power at a priceequal to VOLL. In this event it is acting in the name of, and on behalf of,consumers.

However, it is quite unusual for the regulatory authorities to authorisesuch an astronomical price on the spot market, even during critical periods.The very potential of prices to reach that level provides a powerful incen-tive to generators to withdraw some capacity from the market so as to driveup the price – that is, to exercise their market power during periods oftension between supply and demand. Thus, for reasons of social accept-ability and market power, the spot price is often capped by regulation at alevel far below the VOLL. This type of intervention inevitably distorts the


market signal towards underinvestment, and the plant with the shortestperiod of operation during the year can no longer cover its fixed costs. Theentire cascading structure for covering the fixed costs of the various plantscollapses.

When real-time market prices are capped, undercutting investments, itbecomes necessary to invoke other instruments to provide economicagents with a signal for the optimal capacity level. One elegant approachis based on the notion that generators do not supply a single good, elec-tricity, but rather two goods, energy and capacity. Consumers value twoservices, the power itself when they want to watch television or turn on alight, and also an option value for being able to do this at any time. Fromthis perspective, generators should be compensated for the capacity theysupply regardless of their utilisation. In practice, two systems have beenimplemented: obligation capacity and capacity payments. In the former,retailers (suppliers to the end-users) are obliged to maintain a capacity thatexceeds their expected peak load. To meet this requirement, they acquirepurchasing rights from generators on a capacity market created for thatpurpose.3 In principle, the required capacity level must be determined bycomparing VOLL with the cost of the supplementary obligation capacity.Provision must also be made to penalise retailers for failure to comply withthe mandatory supplementary capacities imposed on them. We observethat this penalty establishes a de facto ceiling on the capacity market;retailers will prefer paying it to buying capacity at a higher price.Consequently, the amount of this fine must be linked to the cost to gener-ators of making capacity available. In the United States, the utility PJM(Pennsylvania–New Jersey–Maryland) has enforced this type of obliga-tion capacity market for several years. The required level represents about20 per cent of peak load and the penalty corresponds to the fixed costs ofa peaking plant ($7.4 per MWh).

During the 1990s, the English Pool established a capacity paymentssystem. Here, the compensation to generators for the capacity they suppliedwas directly integrated into the electricity spot price. Unlike under theprevious system, there was no dedicated capacity market on which supplyand demand met directly. The capacity payment is also determined fromVOLL. In the case of England, it was set equal to VOLL minus the higherof the station bid and marginal price (SMP), this difference being multi-plied by LOLP.

Whether the selected system is obligation capacity or capacity payments,it is essential to bear in mind that the signals sent to investors originate atleast as much from public authorities as from private agents. On the side ofthe invisible hand of the market: all the decentralised consumption andgeneration decisions that propel the evolution of the price; on the side of


the visible hand of public intervention: identifying and setting the price capand calibrating VOLL. We shall see this hybridisation recur in the case ofinvestments in transmission.

4. INVESTMENT IN TRANSMISSION

Like other network infrastructures, electricity transmission grids presenttechnical and economic characteristics that are quite challenging from theperspective of resource allocation. Like highways and airport runways,electrical transmission lines are congested. As a result, use of this infra-structure by one agent may degrade the quality of service available toanother. In economic jargon, this is known as a negative externality. In thecase of electricity, congestion may even result in the complete collapse ofthe system. If the current is not cut, the lines may stretch and melt! Again,like in the case of highway and airport infrastructures, investment occursin discrete units, leading to discontinuous jumps in capacity. To expand ahighway or an airport, a lane or a runway must be added in a single stroke.Smaller, fractional investments are impossible. In electricity, the line typefor the high-voltage grid cannot be modulated by a single KV at a time. Forexample, either 220 or 400 KV must be chosen. Similarly, the gauge of thecable is not available in increments of a millimetre – the choice is limited.

These two technico-economic characteristics, congestion and indivisibil-ity (or lumpiness), are sometimes evoked in defence of misguided concepts.First misconception: investment must proceed until congestion is elimin-ated. In fact, if it were necessary to reinforce electricity transmission linesto the point that their capacities would be able to carry any and all trans-actions between generators and consumers at all times, the grid would bebloated and astronomically expensive. If, during a single hour in one year,a plant that is remote from a consumption zone is €10 per MWh cheaperthan a local, more expensive generator, and if one MW of that generationcannot be transmitted to the consumers because of an inadequate linerating, then that line is congested during that hour. The cost of this con-gestion is €10 per year. Clearly, adding one MW of capacity to that linewould be much more expensive! Eliminating all congestion would onlymake sense if grid construction costs were nil. Obviously, this is not thecase, and consequently the economically optimal level of congestion is notzero. In fact, it is found at the point at which the cost of reinforcing the gridis equal, at the margin, to the savings it makes possible, that is, electricitythat can be bought from farther away at lower cost. The second miscon-ception is that investment should be undertaken as soon as the new lineconstruction project is profitable. It may, indeed, be preferable to wait and


opt for a much more profitable project later – one which will add far greatercapacity at a single stroke. Stoft uses a numerical example to illustrate howit could be better to construct a 1,000 MW line in two years than a 600 MWline today. This is attributable to the lumpiness of the investment, whichdoes not allow demand growth to be matched by an increase in lockstep ingeneration.

As with any infrastructure, it is worthwhile to distinguish betweenefficient use and efficient size of the network. In the first case, capacity istreated as a given. Economic optimisation is thus a matter of allocating itsuse to the economic agents who value it most highly. The theory reveals thatthe key to accomplishing this lies in setting the access price equal to theshort-term marginal cost. In electricity, this cost has two components. Thefirst is due to ohmic losses that make it necessary to inject more electricitythan is withdrawn at the other end of the line. The second component iscongestion, which makes it impossible to replace local, high-cost electricitywith less-expensive power from a more distant plant. Note that both ofthese elements of the marginal cost can be expressed as a function of theprice of the transmitted power itself, for example in €/kWh. This allows usto establish an equivalence between the marginal cost of transmission andthe marginal cost of generation. Between two local competitive markets,the equilibrium transmission price will equal the difference in marginal pro-duction costs, so that a buyer will be indifferent between buying from aseller who is closer but sells at a higher price and one who is farther awayand sells more cheaply. The energy pricing system that corresponds tosetting electricity transmission fees equal to the short-term marginal cost iscalled nodal pricing, or marginal locational pricing. These terms reflect thefact that the electricity price is different at each node of the network. It alsovaries across time since demand, and by extension congestion, fluctuatesbetween the nodes. For example, the systems operator of PJM, the largestelectricity market in the United States, computes the price at the 3,000nodes several times per hour. The issue of efficient network size is an issueof optimal investment. The goal is to achieve the equilibrium size, that is,expand capacity to the point at which marginal cost rises above the benefityielded by continuing. In electricity, we have seen that this benefit amountsto displacing local generation with more remote, cheaper generation.

The distinction between efficient use and efficient investment arisesbecause of a discrepancy between short- and long-term marginal costs –the former being lower than the latter – and between marginal and averagecosts – the former again being the lower. These gaps are explicable in termsof contingencies as well as by the presence of lumpiness and economies ofscale. For historical reasons, the current network is far from its optimal size.As Paul Joskow points out, the electricity transmission system we have


inherited today reflects historical institutional arrangements, the limits ofcorporate activity, political boundaries and historical patterns of urbanand industrial development. He states: ‘We can change the institutions butwe cannot erase the existing infrastructure in place at the time sector liber-alization reforms are implemented but only change it gradually over time’.As a rule of thumb,4 networks that predate the competitive opening arebloated. Governmental, and especially regulatory, intervention favouredcapital expenditures and provided for broad margins of safety to accom-modate growing demand and counter the risk of blackouts. Furthermore,investment in tiers is incompatible with the notion that installations erectedfor a 20- or 30- year lifespan can reflect the optimal network size duringeach year. Inevitably, it will be under- or oversized, depending on thetiming. Once again, overinvestment wins out because of economies of scale(that is, the greater the investment in capacity, the lower the cost of capitalper unit of capacity).

The essential result of the realities described above and the discrepanciesthey give rise to is that a price equal to the short-term marginal cost ensuresefficiency in use, but does not fully cover the investment expenditures nec-essary to construct an optimally sized grid. In other words, the nodal pricingsystem does not compensate the fixed costs of investments in transmission.As Pérez-Arriaga and Olmos emphasise, ‘[C]ost recovery by nodal energyprices typically does not exceed 20 percent of total transmission costs’.

This consequence keeps the market from operating efficiently. ForJoskow, ‘Transmission networks do not and will not evolve through theworkings of the invisible hand of competitive markets’. We note that, ifthere were no gap between the short-term marginal cost and the averagecost, then a decentralised mechanism leading to an optimal level of invest-ment might have been feasible. Such a mechanism has been conceptualised.The underlying principle is to allocate transmission rights that yieldcongestion rents to the owners of each line as they are generated. In hischapter, Stoft describes this mechanism – of which we have provided a bareoutline here – in detail, establishing the link between the level of congestionrents, lumpiness, and economies of scale. Decentralised investments intransmission lines (called merchant lines by convention) are thus confinedto modest growth. This conclusion recurs in the contributions of Joskow,Stoft, and Pérez-Arriaga and Olmos. These authors envisage merchantlines only as a complement to investments regulated by public bodies. Tocite Joskow again: ‘Most transmission investment projects are being devel-oped today and will be developed in the future by regulated entities’. Or,according to Pérez-Arriaga and Olmos: ‘[R]egulated investment . . . mustplay a predominant role in the future development of almost every real-world transmission network’.


5. . . . AND BACK TO THE COORDINATIONBETWEEN INVESTMENTS IN GENERATIONAND TRANSMISSION

In a perfect world, in which demand reacts to the price of electricity andcompetitive local markets are linked by incrementally extensible transmis-sion lines, the combination of nodal electricity prices and financial trans-mission rights ensures a decentralised joint optimisation of investments ingeneration and transmission. Prices exactly cover the costs of the efficientmix, in terms of both the generation technologies and the distributionbetween power lines and power plants. Consequently, from a theoreticalperspective, perfect coordination of investments in generation and trans-mission in an electricity regime that is open to competition is not impossi-ble. The problem resides in the unrealistic nature of the assumptions – allof which are needed to generate this result. Indeed, in our imperfect elec-trical world, consumers’ willingness to pay is not known, some generatorspossess market power, and transmission technologies feature lumpinessand economies of scale. And yet, the theory is not ready for the scrapyard.On the contrary, it suggests solutions for approaching the optimum andminimising market failure.

In light of the failure of financial transmission rights to cover the fixedcosts of transmitting, other methods must be envisaged and implementedto complement signals of short-term grid use with long-term signals todrive investment. A first theoretical method is suggested by Smeers. It isbased on designing a rate structure that captures several components. Theinvestment model he elaborates succeeds in inducing an optimal level andlocation of generating capacity as well as in providing an incentive to theTSO to efficiently manage congestion and develop infrastructure despitethe fact that investments are indivisible. Smeers draws on the work ofO’Neill et al. (2004) who expand the definition of goods, energy in our case,to their spatial dimension. His model is more in keeping with the institu-tional environment prevalent in Europe than that in the United States.Network management is performed by an owner–operator of the infra-structure who integrates dispatching, maintenance and renewal of theinfrastructure. Unlike in the situation in which ownership and dispatchingare separated, here it is necessary to ensure that the system operator doesnot curtail investments in order to increase revenues by creating congestion.

In the Smeers model, the system operator receives instructions from theregulator concerning how to set the long-term component of the price. Italso receives monetary transfers as an incentive to select an appropriategrid configuration. The regulator is assumed to know electricity generators’costs, consumers’ willingness to pay and the set of possible network


configurations. Finally, markets are competitive and all agents – includingthe systems operator – take prices as given.

The work of Pérez-Arriaga and Olmos also deals with pricing that isbased on several components, combines short- and long-term signals, andcovers fixed costs. However, their procedure takes a more operationalapproach. Like Smeers, they focus on the European context. A numericalapplication of their model of long-term transmission costs has been com-puted for all European grids. From a practical perspective, it allows levelsto be set for payments between system operators for use of the grid in othermember countries. We observe that the work of Pérez-Arriaga and Olmosis more relevant to cost allocation than to optimisation. It can be sum-marised as follows. To ensure that all transmission costs are covered, asecond component of revenues must be added to the fee structure based onnodal electricity prices. Two cases can be distinguished. The first deals withhighly integrated networks: consumption centres and generation units aremore or less evenly distributed throughout the territory, and no systemiccongestion is foreseeable at any specific location. In this case, there is noneed for localisation signals, especially since the beneficiaries of invest-ments in transmission would be difficult to identify and allocating individu-alised costs and benefits impracticable. The other element of the price, tocover fixed costs, must be computed by applying the Ramsey rule (that is,the size of the mark-up is inversely related to the consumer’s price elastic-ity of demand). In the second scenario, the additional component mustcapture as nearly as possible the costs and benefits to the grid of decisionsrelating to the siting of the new plant or large energy consumer. Among theseveral available algorithms that are based on some measure of electricityuse, Pérez-Arriaga and Olmos recommend simple and robust schemes thatare based on the average network use and, in those circumstances when itis essential to send to new network users signals reflecting their responsi-bility on new network reinforcements, they propose some new ideas on howto modify standard algorithms to achieve this purpose.

Joskow emphasises the importance of consistency in the organisation ofenergy markets and the institutions that govern transmission. ‘Organizingpower markets and transmission institutions as if a clear separation existsinevitably leads to serious problems’. He analyses two cases: one on eachside of the Atlantic. The aforementioned PJM is characterised by a systemoperator who does not own the grid. The grids are owned by electricity util-ities that are vertically integrated in generation, distribution, and wholesaleand retail operations. However, it is the PJM system operator who runs theday-ahead and balancing markets. It also operates the capacity market.Load-serving entities are, in fact, subject to capacity obligations computedon the basis of their monthly peak requirements. As Joskow explains, these


supply requirements play an important role in the process of investment intransmission and in providing siting incentives to generators. The othercase he examines is the Anglo-Welsh system. Until March of 2001, thewholesale market was organised into a mandatory pool. Generators werecompensated separately for power and for capacity. An energy-only marketfollowed with the implementation of the NETA (New Electricity TradingArrangements). We observe that the price on this market is not capped. Thesystem operator, National Grid Company (NGC), is integrated. It func-tions as the system operator, oversees maintenance of the grid, and makesthe investments. The transmission price is regulated by Ofgem (Office ofGas and Electricity Markets) and includes an element that depends onlocation. Generators in the north of the country pay more than those in thesouth. In matters of investment, NGC is bound by obligations specified inthe network code and by various standards. To comply with them, it con-ducts studies based on regional demand and supply estimates. When a vio-lation of a standard is identified, NGC determines which investmentprojects should proceed. Their size determines whether they requireapproval from the regulator. Joskow emphasises how well the Anglo-Welshsystem has performed since the mid-1990s. He considers this to be the mostsuccessful experience in market liberalisation anywhere in the world.

6. OVERVIEW OF THE BOOK AND SYNOPSIS OFTHE CONTRIBUTIONS

In addition to this introductory chapter, the book consists of three parts.Each of these comprises two chapters, where pure theory alternate withpractical application or empirical study. Part I (Chapters 2 and 3) is devotedto investment in generation, while Part II (Chapters 4 and 5) addressesinvestment in transmission. Part III (Chapters 6 and 7) examines the issueof coordination between investments in generation and transmission.

Green’s contribution (Chapter 2) deals with the theoretical mechanismsthat determine the choice of the level and mix of electricity generationcapacity. A broad outline of these mechanisms was briefly presented above.We shall underscore some of the contributions of this chapter in moredetail. Green reminds us that, in the final analysis, investments in genera-tion are not only about increasing capacity to satisfy growing demand.Even with constant demand, new capacity is required to replace plants thatare inefficient – owing to technological obsolescence – and those that are atthe end of their lifespan. It is just as important to examine the economicdeterminants of plant decommissioning as of new construction, especiallysince some plants can be mothballed before being definitively shut down.


They can be called on to meet exceptional needs. We note that there is acertain parallelism between decommissioning an old plant and commis-sioning a new one: both actions are irreversible. In the presence of demanduncertainty, this implies that it may sometimes be preferable to delay thedecision rather than act immediately, since time may yield better informa-tion. Thus, investment is triggered, not when the price rises above marginalcost, but when it exceeds the marginal cost plus the option value.Conversely, decommissioning of a plant occurs when the price falls belowmarginal cost minus the option value. Green’s contribution also discussesthe cyclical character of investments in electricity generation. He notesthat, in contrast with other commodities, the possibility of keeping plantsin reserve and committing to long-term contracts should smooth the cycles.These latter operate in two ways. First, they mitigate the uncertainty facingentrants by allowing them to fix a price or a margin of their sales price overthe cost of fuel. Second, long-term contracts can function as a sort of coor-dination mechanism for investments in generation – a mechanism that isstarkly lacking after the transition from a monopolistic to a competitivemarket structure.

In Chapter 3, Glachant presents a descriptive and applied economic por-trait of the changes to the technology mix induced by the competitiveopening. Did the reforms to the electricity sector have an impact on thechoice of generation technologies? Does competition create new incentivesthat are biased towards certain technological developments? Or, conversely,does competition marginalise certain technologies that prospered in thecontext of a regulated industry? Drawing on extensive data, Glachantobserves that, in the United States as in many European countries, elec-tricity reforms were accompanied by a technology shift towards generationwith CCGTs. To a lesser extent, an expansion of renewable energies canalso be detected. On the other hand, the construction of new nuclear reac-tors came to a halt and the amount of electricity generated by this tech-nology is declining. The conventional explanation for this dual trend is asfollows: liberalisation created competition among technologies, allowingthe efficiency of gas to come to light, while the reforms also put an end togovernment subsidies to the nuclear option. Rather than any simple intrin-sic superiority of gas, or the withdrawal of government support fromresearch and development into nuclear technology, Glachant demonstratesthat nuclear power is handicapped by much higher capital costs than thoseof electricity generated from gas. This differential is attributable to muchgreater financial risks associated with the choice of nuclear technology.Construction costs and the operational performance of these plants (par-ticularly capacity availability and lifespan) are imprecise and highly vari-able. Glachant also reveals that capital intensity, the size of the minimum


unit of capacity, randomness in the construction schedule owing to anti-nuclear mobilisation, and the absence of any correlation between the priceof the fuel and the price of electricity, are all factors that increase the risk-iness of the investment. According to a MIT study that was extensivelyreviewed by Glachant, this set of factors gives gas an edge over nuclear interms of the gearing rate (40 percent equity, versus 60 percent for nuclear)and a lower-yield requirement for these funds (8 versus 15 per cent). Asillustrated by the credit arrangement of the Finish nuclear project TVO, theyield to investments in nuclear power is undoubtedly to be found in long-term contracts between generators and future buyers, reducing the risksand, by extension, the cost of capital.

In Chapter 4, Stoft applies a pedagogical approach to elements of theeconomic theory that shed light on investments in transmission and theobstacles that undermine market efficiency. Here the reader will finddefinitions of essential concepts, such as congestion (or redispatching)costs, congestion rent and the cost of congestion to load. These costs areuncorrelated and should not be confused. Stoft also takes care to distin-guish between two concepts that are often linked because they both under-lie fixed costs and violate a basic assumption of the invisible hand of themarket; to wit, the convexity of the cost function. These concepts arereturns to scale and lumpiness. As a final pedagogical item, Stoft debunkstwo misconceptions that are currently in vogue: it is not true that the levelof congestion should be reduced to zero; and it is not true that marketpower is required to recover fixed costs. In his contribution, Stoft comparesthree different approaches to investment in transmission: the traditionalplanning approach, the merchant approach and the performance-basedregulation (PBR) approach. He discusses the last of these approaches atlength. Here the reader who is unfamiliar with the theory of incentivesapplied to natural monopoly will find developments that shed light on theunderlying principles of the price cap and on the dilemma confronting theregulator seeking to encourage the system operator to cut costs while notleaving an excessively high rent that will penalise consumers. Stoft pointsout two major difficulties associated with establishing incentive regulationin the case of the electricity transmission grid. The first is linked to thelength of the delay in benefits accruing to the investor. The timeframe ofthese investments may, in fact, look as follows: considerable sums must becommitted over several years, which are followed by several more yearsduring which the return is nil, or minimal, and only 10 or 15 years after thebeginning of the project does it truly begin to pay off. The second difficultyarises from the tight linkage between the investment and security of supply(reliability). In the United States, most of the major blackouts thatoccurred during the past 35 years were attributable to problems with trans-


mission rather than generation. In light of this, regularly pruning treesgrowing beside power lines, updating computer systems and installing line-trip relays are all essential. Thus, incentive mechanisms that cover activitiesother than the construction or reinforcement of power lines are needed. Inthe words of Stoft, PBR for transcos will be useful for shorter-term incen-tives, but it cannot yet be relied on to solve the long-term investment prob-lems’. Since the development of merchant lines is bound to be constrainedby issues surrounding the recovery of fixed costs, as we saw above, there is,in the final analysis, no alternative for government authorities but to pursuetraditional regulation.

In Chapter 5, Joskow sketches out, in some detail, the various existinginstitutional arrangements that govern operation of the grid, inform theregulatory framework and provide incentives to invest in transmission. Hedemonstrates how these arrangements depend on the historical, economicand physical characteristics of the network and examines their perfor-mance. Joskow’s contribution is too rich to be summarised here. Forexample, it contains an exhaustive list of the various components of thenetwork that play a role in reinforcing its capacity. Economic models tendto focus too exclusively on the construction of new lines, for there are manyother ways to reinforce a network. Too illustrate, from Joskow’s list: newrelays and switches, reconductoring existing lines, and new remote moni-toring and control equipment. He also proposes a classification system fordifferent types of investment and discusses it in detail. We also draw yourattention to two original observations by Joskow. The first pertains to thegulf between the viewpoints of economists and engineers on the subject ofinvestments in transmission. The models of the former have little incommon with the manner in which investments in transmission are actu-ally programmed and developed, or in how the associated services arepriced. They do not account for the engineering reliability criteria on whichengineers base decisions to reinforce a network. Of course, there cannot betwo disjoint types of investment, one based on economic calculations andthe other on reliability. Joskow vigorously argues that these two approachesneed to be reconciled. The second observation concerns the distinctionbetween inter- and intra-SO transmission grid investments. Each TSO willfirst tend to deal with congestion issues on its own grid independently, andthen facilitate residual economic exchanges with other grids. This policyresults in congestion being pushed across borders and in reduced economicefficiency. Joskow suggests that inter-TSO investment opportunities can beaddressed more effectively through interconnected zones using the samereliability criteria and standards of evaluation, as well as by integratingwholesale markets and harmonising pricing practices across countries. Hestrongly recommends the creation of regional transmission operators.


In Chapter 6, Smeers addresses a knotty and so-far unsolved problem –finding price signals that will motivate system operators to invest opti-mally and allow them to recover their costs. In other words, is it possible todecentralise investment decisions when they are lumpy? The solution sug-gested by Smeers is a price comprising several components and incorpo-rating access and congestion fees. The reader unfamiliar with optimisationmodels, in particular non-linear models, may benefit from reading theintroduction (Sections 1 and 2) and the discussion (Sections 8 and 9) of thischapter, where the author’s approach and results are summarised in a non-technical manner. One result that merits comment here deals withEuropean regulation of interconnections. It stipulates that rates mustcomply with three principles: economic efficiency, cost causality and non-discrimination. Smeers begins by building a model with no linkage betweenagents’ localisation decisions and the structure of the network. Thus, hismodel does not respect the principle of cost causality. Nevertheless, theproposed price structure is efficient because it is based on price discrimina-tion. It is well known in economics that price discrimination provides aneconomically optimal way for fixed costs to be recovered. Next, Smeersintroduces cost causality, which allows discrimination to be reduced but noteliminated. This can be accomplished without endangering the balancedbudget of the system operator, but only at the cost of partially sacrificingthe goal of economic efficiency. The prohibition on price discriminationmust be juxtaposed with the loss of social surplus it entails. When subsidiesto investments in interconnections are precluded, an arbitrage between theallowed level of discrimination and the tolerable amount of economic lossbecomes necessary. Nothing in the European texts or discussions providesfor this arbitrage.

In Chapter 7, Pérez-Arriaga and Olmos address the same issue asSmeers, long-term siting signals and covering the fixed costs of transmis-sion networks. However, they take a different approach – their perspectiveis practical and their process operational. This compels them to makecertain concessions, notably in adopting cost-allocation methods thatsometimes owe more to accounting than to economics, and also in sim-plifying the physical functioning of the grid. Their contribution nicelyrounds out the preceding contributions. In addition, they examine howlocational signals that are derived from the existence of the transmissionnetwork – differences in energy prices due to losses and congestion, plustransmission charges with locational differentiation – compare numeri-cally among themselves and also with other non-electrical locationalsignals, such as potential charges for the use of gas infrastructures ordifferences in the efficiency of thermal power plants because of thealtitude over sea level.


It is recognised that the agents who make the decisions on transmissioninvestments strongly depend on the specific regulatory paradigm that isadopted in each country: system operators, regulators, coalitions of networkusers and merchant investors – alone or in different combinations – can bethe responsible parties. Accordingly, the economic signals that may provideincentives to make correct decisions on new transmission investmentsdepend of the adopted regulatory paradigm. Although all the consideredparadigms are useful ones, not all of them would result in a well-developednetwork.

Under a competitive regulatory framework it is essential for the success-ful development of both generation and transmission to minimise theuncertainty that the decisions of generators create for the network plannerand, conversely, that complete and reliable estimates of future transmissionconditions be facilitated to generators by the system operator. Several regu-latory instruments can be applied to reduce the unavoidable level of uncer-tainty that surrounds the decision making process of generators andtransmission planners.

Pérez-Arriaga and Olmos remind us that there is more to the electricitynetwork than the high-voltage transmission grid. The structure andrenewal of the distribution grid must also be considered. However, thesetwo components of transmission fulfil different functions. Consequently,regulatory approaches and investment criteria must differ as well. Aseries of practical considerations are proposed in this contribution inorder to ensure compatibility of signals for investments in transmission anddistribution.

7. CONCLUSION

After reading this introductory chapter, the reader may be amazed at thelength of the road to be travelled on the way to ideal investment conditions.It should not be forgotten that this difficult task springs from a very ambi-tious goal. Investment is an issue of dynamic economics. This is more com-plicated than problems of static efficiency, and the corresponding economictools are less robust. Moreover, in this case the duration of investments ismeasured, not in years, but in decades. Seeking to know the optimal gen-erating and transmission capacity of the electricity system is no less ambi-tious than attempting to build the cities of tomorrow and design thenetwork of highways and byways that will link them. We must accept thatthe ideal of an electrical utopia will elude us, but instead we can elaborateprinciples of urban and land-use planning that will make decentraliseddecisions more efficient. Such is the hope of this undertaking.


NOTES

1. Note that application of this idealised rule not only runs up against the opportunism ofgenerators. It also assumes that the SO (or the competent regulator) acts in the publicinterest, is able to forecast future energy demand, and is able to define precisely theoptimal level of capacity (that is, the number, type and location of plants) and the gridconfiguration that will satisfy that demand efficiently.

2. It also assumes risk neutrality of investors. Risk aversion leads to underinvestment inpeaking plants – some of which are only profitable, in principle, if they operate severalhours per year on average.

3. If they are vertically integrated, they can arrange this supply internally.4. With the notable exception of interconnections between countries. In Europe, these were

built for security rather than business considerations. The opening to competition andburgeoning trade soon made their inadequacy clear.

REFERENCE

O’Neill, R.P., P.M. Sotkiewicz, B.F. Hobbs, M.H. Rothopf and W.R. Steward Jr(2004), ‘Efficient market-clearing prices in markets with non convexities’,European Journal of Operations Research, 164 (1), 269–85.


PART I

Investment in generation

2. Investment and generation capacityRichard Green*

1. INTRODUCTION

This chapter considers the questions of how much generation capacity, andof what types, is required, and whether the market will provide it. It startswith a model in which there are two generating technologies, and showshow to obtain the optimal level of each kind, before focusing on the ques-tion of the overall level of capacity.

Note that the discussion so far has been in terms of the level of capacity,rather than of investment. In an ideal world, this focus on capacity is appro-priate, since net investment would equal the difference between the desiredlevel of capacity and the current level. This is defining net investment as theamount of capacity added to the industry, less the amount taken out ofservice, rather than by deducting any kind of depreciation charge fromgross investment. The amount of plant taken out of service can be endo-genous, of course, since far more plants are retired because they havereached the end of their economic life than because they have become phys-ically incapable of (safely) generating any more electricity.

In the real world, investment (and closure) decisions are more compli-cated than a simple comparison of actual and desired capacity wouldsuggest. In particular, many capital-intensive industries are prone to cap-acity cycles, and there is a danger that the electricity industry could godown the same path. Investments in power stations are also typically lumpyand irreversible, and the theory of real options has potentially importantimplications for decisions of this type. This chapter considers these issues.

2. A SIMPLE MODEL OF CAPACITY

It makes little sense to think of investment without thinking about theoverall level of capacity in an industry. We therefore start with a model thatdevelops the optimal level of generation capacity. The presentation here isgraphical,1 while there is a mathematical version in the appendix. To keepthe model as simple as possible, we shall assume that there are just two types

21

of generation capacity available to the industry. Peaking plants have rela-tively low fixed costs per kW of capacity, but relatively high marginal costsper MWh generated. Open-cycle gas turbines are typical peaking plants,since they are relatively quick and cheap to build, but have low thermalefficiency, and generally run on expensive distillate fuels. When only a smallamount of energy is required, a diesel plant may be the most efficientoption. In many systems, much of the peak demand is met by output fromold plants. Their avoidable fixed costs are relatively low, since the costs ofbuilding them have been sunk, and the choice is between closing the plantand keeping it open. Since the cost of staffing, maintaining and insuring theplant, and paying fees to the owner of the transmission system can gener-ally be avoided if the plant is closed, it would be wrong to say that theseplants have no fixed costs – the term ‘going forward costs’ can be used todescribe costs that have to be incurred if the plant is kept open. Because theplants are relatively old and inefficient, their marginal fuel costs will behigh, and their age may also increase their variable operations and mainten-ance costs.

Base-load plants, in contrast, have relatively high fixed costs per kW, butlower marginal costs per kWh generated. These are generally new plants,and the key decision would be whether to build more of them, incurringcapital costs. The ultimate example of a base-load plant would be a nuclearstation, where very high capital costs are (sometimes) offset by low runningcosts – the French seem to have achieved a favourable trade-off, the Britishnot. Coal- and gas-fired stations generally have lower capital costs thannuclear stations, but are also generally built in order to run at high loadfactors, at least when new.

The top panel of Figure 2.1 (panel 1) thus shows how the total costs (in€ per MW over the year) of our two types of capacity depend upon thenumber of hours a year for which the station is operated. The vertical inter-cept shows the station’s fixed costs, while the slope of the line gives the costper MWh. If the station is required to run for T* hours or less per year, thenit is cheaper to use a peaking station, while if it is required to run for morethan T* hours, then a base-load station has lower total costs. The bold,kinked, line gives the lower envelope of the two linear total cost functions,showing the efficient cost of meeting a demand lasting for any givennumber of hours.

How can we tell how many hours a particular station will be needed for?Moving down to panel 2, this gives us a load-duration curve. The hours ofthe year are ranked in order of the demand for electricity, so that the hourwith the highest demand is placed at the left-hand end of the panel, and thehour with the lowest demand at the right-hand end. The vertical axis thenshows the demand in that particular hour. The demand during the T*th

22 Investment in generation

Investment and generation capacity 23

Notes:(1) Total costs by plant type.(2) Load-duration curve.(3) Reflecting line.(4) Marginal cost and demand.(5) Price-duration curve.

Figure 2.1 The determination of electricity capacity and prices

Hours/year

Hours/year

Hours/year

€/MW-year

GWGW

GW

GW

€/MWh €/MWh

B K D

P T*

P T*

T*

(2)

Peaking Baseload

(1)

(5)(4)

(3)

B K D

B

D

CB

CP

PR

K

Maxdemand

highest hour is thus B GW. We take this to be a gross demand for electri-city, including transmission losses and the amount of plant that has to bekept part loaded, or available at short notice, for reserve. Reading the graphthe other way round, we could say that there are T* hours in which thisgross demand for electricity is B GW or more.

This therefore implies that if we want base-load plants to meet all thedemands for electricity that last for T* hours or more of the year, then weshould ensure that B GW are available. In the electricity industry’s tradi-tional model, this might have been the end of the process, since prices wereoften set, by a regulator or a self-regulating state-owned industry, to ensurethat the industry was able to recover its average costs. Sometimes, however,prices were related to marginal costs (Electricité de France was a pioneer inthis), and our framework can be used to obtain marginal costs. In a marketsetting, sufficient competition implies that prices will equal marginal costs,increasing the importance of calculating them.

Panel 3 is simply a reflector, allowing us to move the capacities shown onthe vertical axis of panel 2 to the horizontal axis of panel 4. If the marginal(operating) cost of base-load plant is equal to CB, then this will be the mar-ginal cost of the industry whenever demand is equal to or less than thecapacity of this type of plant – assumed to be B GW at the optimal solu-tion. At somewhat higher levels of demand, then the marginal cost willequal CP, the marginal cost of peaking plant.

We must now discuss the total level of capacity. The load-duration curvein panel 2 has been drawn on the assumption that the price in the first T*hours of the year is equal to CP, and that the price for the rest of the yearis CB. At a price of CP, the maximum demand for electricity would be equalto D GW. If the industry had this much capacity, then that demand couldbe met in full, but the price would never exceed the marginal cost of thepeaking plants, and those plants would not be able to cover their fixed costs.This is clearly not sustainable in a market which investors are free to leave,and will not enter without a clear expectation that they will cover theircosts, including an appropriate return on capital.

What would happen if capacity were not sufficient to meet a demand ofD GW in full? This depends in part on the details of the market arrange-ments for the industry. The system operator may be able to reduce theamount of reserve plant that it is carrying, so that customers are not imme-diately affected, although this increases the risk of failures disruptingsupply to larger numbers of customers. The cost of such a failure, multi-plied by its probability, gives the value of additional generation at thesetimes. With a shortage of capacity, generators would be able to bid up tothis expected cost, and it would be rational for the system operator to payit, even when it is greater than the generator’s marginal operating cost. It is


more likely, however, that the system operator will ask for load manage-ment, and will start to pay some customers to reduce their demand. Bidsfrom these customers might set the price in a real-time market directly, orgenerators might raise their bids above their marginal cost, knowing thattheir competition now comes not from one another (since all are needed)but from the demand side.

We discuss ways in which actual markets pay for capacity in Section 5.For the time being, we can assume that there is a relationship between theactual level of available capacity and the price that will be paid for elec-tricity in the peak hour, and that this is shown by the downward-slopingline in panel 4. It shows that if there were D GW of capacity available,the price could be CP, which was the price assumed in drawing the load-duration curve with its peak demand at D GW. With less capacity, however,the price must rise to clear the market. If there is only K GW of capacityavailable, then the price must rise to PR in order to ration demand to thelevel of capacity. This can in fact be defined as the marginal cost of powerat that time, set not by variable operating costs but by the opportunity costof a consumer that has decided to reduce its demand.

This is the price level shown at the top left of panel 5, which shows aprice-duration curve. At the highest-demand hour, the price must be PR,but lower prices are possible in hours with lower demands. After P hours,the load-duration curve of panel 2 shows that demand at a price of CP hasfallen to K GW, and this can of course be met in full by that level of cap-acity. In terms of the load actually served, the load-duration curve thushas a flat segment at K GW, and the area above this represents ‘unserved’load, rationed either by price or some other means. Between hour Pand hour T*, therefore, the price in a perfectly competitive market wouldbe equal to CP, the marginal cost of the peaking plants which are at themargin, but not fully employed. From hour T* onwards, none of thepeaking plants is needed, and the marginal cost falls to the marginal costof the base-load plants, at CB. In a competitive market, this would also bethe price.

What about the total revenue received by any power station? For this, weneed to return to panel 1. Remember that the marginal cost of a plant(reflected in the level of the lines in panels 4 and 5) is given by the slope ofits line in panel 1. Since the price between hour P and hour T* is equal tothe marginal cost of peaking plants, the slope of a total revenue line (perMW per year) would be identical to the slope of the total cost line forpeaking plants over this period. (The price is normally given in €/MWh,while the axes of panel 1 are in €/MW per year and hours per year, so thata line in this panel has a slope measured in €/MWh.) Similarly, from hourT* onwards, the slope of a total revenue line would be given by the marginal


cost of a base-load plant, and this is also the slope of the efficient cost envel-ope over these hours.

If the market price never exceeded CP, then the total revenue line wouldstart from the origin and run parallel to the efficient cost frontier through-out its length. Peaking plants would cover their variable costs but make nocontribution towards their fixed costs. Base-load plants would make a con-tribution towards their fixed costs, since the price exceeds their variable costsfor the first T* hours of the year, but there would be a shortfall equal to thefixed costs of the peaking plants. This shows that if we had D GW of cap-acity, it would not be able to cover its costs in a competitive market system.

With less capacity, however, the price will exceed CP for the hours withthe highest levels of demand, and this will allow all the stations to make anadditional contribution towards their fixed costs. The slope of a totalrevenue line will be greater than the slope of the efficient cost envelope, andone possible line is shown rising from the origin to meet the efficient envel-ope at P hours. The slope of this line is given by the level of prices in panel 5.The implication is that the prices of panel 5 would be just sufficient for apeaking plant to cover its fixed costs, plus its variable costs of running forP hours. Since the load-duration curve shows that with a total capacity ofK GW, all the peaking plants will need to run for at least this number ofhours, this means that they would all be able to cover these costs.

The total revenue line would then continue upwards at a rate of CP €/MW per hour, which is the market price between hour P and hour T*, andalso the slope of the efficient cost envelope. In other words, the totalrevenue line is now superimposed on the efficient total cost line. After hourT*, the slope of the total revenue line falls to CB €/MW per hour, just likethe slope of the efficient cost envelope. This shows that if the peaking plantswith the shortest periods of operation are able to cover their total costs, andif the plant mix is efficient, then all the other plants will just be able to covertheir total costs from market prices based on marginal costs. (See Box 2.1)

We can summarise the links explained in Box 2.1 between the amount ofcapacity and the profitability of that capacity. If there is too little capacityin total, then peaking capacity will be making supernormal profits, while ifthere is too much capacity overall, then peaking plants will be makinglosses. If we have the right amount of base-load capacity, then these sta-tions will be making the same amount of profit (per MW-year) as peakingstations. If there is too much base-load capacity, then base-load stationswill make less profit per MW-year than peaking stations. If there is too littlebase-load capacity, then these stations will make more profit per MW-yearthan peaking stations.

In this simple model, there are three reasons for investment in new cap-acity. The first is that the actual level of (a given type of) capacity is less than



Notes:(1) Optimal capacity mix.(2) Right total, too little base load.(3) Right total, too much base load.(4) Total too large, right base load.(5) Total too small, too much base load.

Figure 2.2 How the capacity mix affects revenues

Hours/year

€/MW-year

T*

Peaking Base load

(1)

(3)(2) €/MW-year

T*

Peaking Base load

€/MW-year

T*

Peaking Base load

Hours/yearHours/year

(5)(4)€/MW-year

T*

Peaking Base load

€/MW-year

T*

Peaking Base load

Hours/yearHours/year


BOX 2.1 WHAT IF THERE IS THE WRONG LEVEL,OR MIX, OF CAPACITY?

The profitability of peaking plants depends on the total amount ofcapacity, rather than on the amount of peaking plant in the indus-try. To see this, consider what would happen if the total capacityremained at K GW, but that there was less than B GW of base-loadcapacity (and hence more peaking capacity than in the optimalplant mix). This is the situation shown in panel 2 of Figure 2.2.(Panel 1 repeats the top panel of Figure 2.1, showing the situationwith the optimal level and mix of capacity.) For the highest-demandT* hours of the year, the situation is exactly the same as in Figure2.1, and so the total revenue line rapidly rises to join the total costline for peaking plants, and then runs on top of it. Things changeafter T * hours, however.With too little base-load capacity, some ofthe peaking plants will be running for more than T * hours, whendemand is still too high to be met entirely by the reduced numberof base-load plants. In those hours, the price will be equal to CP,and the total revenue line (in bold) continues with the same slopeas before. Only once demand has fallen to the level of base-loadcapacity, at some point to the right of hour T *, does the price fallto CB, and the total revenue line flatten to run parallel to the totalcost line for base-load plants. Because the total revenue line isabove the total cost line, all of the base-load stations are making asupernormal profit, over and above their costs. Because the linesare parallel, the profit earned by a base-load plant does notdepend on how many hours it runs (as long as it runs in all thehours when the price is above CB).

What If the Mix of Capacity is Wrong?

We might have a situation with the right amount of capacity in total,but too many base-load plants (and hence too few peaking plants),for example. This is shown in panel 3 of Figure 2.2. Once again,the prices for the hours with the very highest demands areunchanged, and the total revenue line meets the total cost line forpeaking plants while all of those plants are running. Base-loadplants become marginal before the demand has fallen to B GW,however, and so the price falls to CB at a point to the left of hourT *. The total revenue line then runs parallel to, but below, the totalcost line of base-load plants.The peaking plants are covering their


costs in full, but the base-load plants are making losses. Onceagain, in this simple example, the scale of the losses does notdepend on exactly how many hours the base-load plant runs,within the range where the model shows them to be marginal.

What If the Level of Capacity is Wrong?

Panels 2 and 3 thus show that having the wrong mix of base-loadand peaking plants does not affect the profitability of peakingplants, as long as the industry has the correct overall level ofcapacity. If the industry has the wrong level of capacity, however,then the peaking plants’ profits will be affected. In panel 4, there areB GW of base-load plants, but total capacity is assumed to bebetween K GW and D GW. This means that there are some hoursin which capacity is insufficient to meet demand in full at a price ofCP, but fewer than P of them. The price exceeds CP during thesehours, and so the total revenue line starts with a steeper slope, butthe revenues (over and above variable costs) are not sufficient tocover the fixed costs of a peaking plant. At the point where capacityis sufficient to meet demand in full at a price of CP, the total revenuecurve becomes less steep, running parallel to, and below, the totalcost line for the peaking plant. At hour T *, base-load plantsbecome marginal, and the total revenue line flattens again. Thisimplies that both types of capacity make the same loss, measuredin €/MW per year. Even though we have the right amount of base-load capacity, the surplus of total capacity forces them into loss.

What If the Level and Mix of Capacity are Wrong?

Finally, panel 5 shows that it is possible for peaking capacity tomake a profit, but for base-load capacity merely to break even, ifwe have the wrong amount of both total capacity and base-loadcapacity. In this case, a shortage of total capacity means thatprices exceed CP for more than P hours, and so the total revenueline moves above the total cost line for peaking plants. It then runsparallel to that line for some hours, until demand falls to the levelof base-load capacity, and those plants become marginal. Thispanel has been drawn so that this occurs just as the point wherethe total revenue line meets the total cost line for base-load plants.This is to the left of hour T *, which implies that there is too muchbase-load capacity. That would normally drive all the base-loadplants into loss, because prices would be at CB for too much of the

the optimal level. The second is that some capacity has reached the end ofits physical working life and must be replaced. The third reason is becausechanging relative costs make it economic to replace older capacity with amore efficient station. This could be the result of technological progress,and is more properly a matter for Chapter 4. The model implies that allthree types of (optimal) investment will be profitable for the firm thatundertakes them. In a competitive market, we should therefore expect thatfirms would be willing to make these investments.

When should capacity be reduced? In this model, if we have too muchcapacity, then it will not be able to cover its full economic costs includingan appropriate return on capital. That is certainly a signal that no newinvestment should be made. In the absence of sunk costs, it is also a signalthat capacity should be withdrawn from the market. Electricity generation,however, is subject to significant sunk costs, in that power stations arecapital intensive, have no real alternative use, their output can only bemoved to a new market if transmission capacity is available, and the costsof physically moving a power station will generally be prohibitive. In thepresence of sunk costs, it is no longer optimal to close a station as soon asit is unable to make a full return on the capital invested, as long as it is atleast covering its variable costs. This drives a wedge between the prices atwhich entry becomes profitable and those at which exit is sensible. The con-sequences of this wedge, and more generally of the need to make irre-versible investments in an uncertain world, are considered in the nextsection.

3. IRREVERSIBILITY AND UNCERTAINTY

We have just pointed out that electricity generation involves significantsunk costs. The electricity industry is also subject to considerable uncer-tainty, over fuel prices and the level of demand. While fuel prices mostlyaffect the choice between different types of power station, the level of


year. As the panel has been drawn, this is exactly cancelled out bythe higher prices in the very peak hours. It should be obvious thatthis is a rather special case. If there was a slightly lower proportionof base-load plants (within the given total capacity) then they wouldalso be making a profit, while a slightly higher proportion (implyingfar too little peaking plant) would mean that the base-load plantswere actually making losses.

demand feeds in to the total capacity required. Two types of demand uncer-tainty are relevant here. First, the short-term level of demand, relative to itstrend, depends on the weather, and the state of any interconnected powersystems – net demand in an area will be higher if an adjoining area that nor-mally exports to it is short of capacity, as California found to its cost whenhydro-electric generators in the Pacific North West were short of water in2000. Second, the trend itself is uncertain, depending on consumers’ reac-tions to prices, the level of economic activity and technological change. Tosome extent, short-term fluctuations in demand can be hedged. A warmwinter will reduce demand and lower prices relative to their expected level,but generators and retailers can sign contracts that fix their revenues (forthe contracted level of output) at the expected price level.2 A cold wintermay be more of a problem, in case it raises demand above the level of avail-able capacity, but contracts can once again largely insulate retailers fromthe financial consequences of this uncertainty.

Uncertainty over the trend level of demand creates more problems. Ifcapacity is built to meet a need that does not arise, then prices may bedepressed for several years, until demand rises or other plant is retired.Retiring plant, however, is an irreversible decision, and so generators willnot lightly close a plant for good. Even mothballing a plant in the hope ofreopening it later involves some costs, and so generators will not withdrawcapacity from the market as soon as prices start to fall below their variablecosts. The implication of this is that it is not sensible to make investmentsat the first signs of an increase in demand, in case it is not sustained.

Dixit and Pindyck (1994) consider the theory of irreversible investmentunder uncertainty. Their key point is that a firm that has not made an irre-versible investment has the option of making the investment at some latertime, and it gives up this (so-called ‘real’) option once it makes the invest-ment. Since the firm is giving up the value of the option as well as the sunkcost of the investment itself, it is best to wait until the expected value of theinvestment exceeds the sum of its direct cost and its option value.

Consider a firm which is considering whether to build a power station.In the first period of its life, it can make a profit before fixed costs of €100.In all subsequent periods, we expect the station to make a profit of €100 aswell. The discount rate is 10 per cent, and so the value of this expected profitstream is €1,000 (we get a simpler present value if we ignore the fact thatthe station will not actually last for ever). If the cost of the station is €800,then building it now will lead to an expected profit of €200. On the tradi-tional analysis, this is clearly the sensible decision.

Assume, however, that our expected profit of €100 per period after thefirst period is actually the average of a profit of €50, and of a profit of €150,each expected with a probability of 0.5. Furthermore, assume that if the


generator waits until the end of the first period until it invests, it will knowwhich of these will be the case. In the low-demand state, with a profit ofonly €50 per period, investing in the second period would not be profitable –a revenue stream of €500, discounted to the second period, is clearly lessthan the investment cost of €800, incurred at that time. In the high-demandstate, investment is profitable, since the investment of €800 brings in rev-enues with a present value of €1,500. The firm’s expected profit from asecond-period investment is €350 – there is a 50 per cent chance that it canmake an investment with a profit of €700, while it will lose no money ifdemand is low and it does not invest.

Discounting this expected profit to the first period, we get an expectedprofit of €315. This is greater than the expected profit from investing atonce, by some €115. This figure is the option value of waiting one perioduntil the uncertainty is resolved. If the firm follows the traditional analysisand invests at once, it is giving up this option value, and hence the chanceto wait and avoid the mistake of investing too early.

This does not mean that waiting is always optimal. If the firm had to wait6 periods until it knew whether demand was going to be high or low for therest of time, then its expected profit from making the investment decisionat this time will still be €350, but this must now be discounted by a factorof 1.16. This gives a present value of €350/1.16, or €197.57, which is lessthan the present value of investing now.

Alternatively, assume that the expected value of revenues in each periodis €140, with an equal chance of revenues of €70 and €210 after the firstperiod. In the second period, it would still be optimal not to invest ifdemand turns out to be low, and so investing only in the case of highdemand brings expected profits of €650 (equal to €0.5 (€2,100 – €800)).Discounting this to the first period gives an expected profit of €590.91. Ifthe firm invests at once, however, it gets an expected profit of €600, whichis greater. The option value of waiting is now outweighed by the immedi-ate profits to be gained.

This leads us towards the kind of investment rule that Dixit and Pindyckderive – investment is optimal if conditions are at the level where thecurrent profit to be had from investing now outweighs the benefit of waitingfor more information. They express this as a trigger price investment rule –the firm should invest only once the variable reaches a particular level. Thisis a feature of the stochastic processes that they use, in which the futurepath of a variable depends only upon its current level. A more complex sto-chastic process, in which the future depended on the path taken by the vari-able, would require a more complicated investment rule.

Note that the trigger conditions for investment will generally be wellabove the minimal conditions at which investment would just be profitable,


if the expected future conditions were sure to be realised. This is not onlythe result of a market-based system, however, for Dixit and Pindyck showthat the social welfare-maximising decision rule is identical to the rule thatwould be adopted in a competitive market. Furthermore, at least for arange of problems, the competitive solution is to invest at the same triggerpoint as in the monopoly solution. The calculations involved are different –a competitive firm has no option value of waiting, because if it does notmake a profitable investment, it can safely assume that some other firm willdo so. With enough potential entrants, the price will never rise above thetrigger for further investment, but random movements in demand can beexpected to depress the price below this level. The trigger price must there-fore be high enough for the firms to expect to just cover their costs if theyinvest at this price, realising that it will be a maximum and they will obtaina rather lower average price. This clearly implies that the trigger price mustexceed their average costs. Interestingly, the perfectly competitive firm canperform some of its calculations as if there would be no further entry (Dixitand Pindyck, 1994, p. 291).

Real options theory can also be used to determine the point at whichcapacity should leave the system. This can happen in two ways. First, cap-acity can be mothballed, with the possibility of returning it to service later.Mothballing generally involves an ongoing maintenance cost, togetherwith costs when the plant is prepared for mothballing and then returned toservice. Second, a plant can be scrapped, which is an irreversible decisionfor that plant, but may not preclude the possibility of making a completelynew investment if conditions improve.

As long as putting a plant in mothballs and then taking it out involvessome fixed costs, it should be obvious that it will not be optimal to moth-ball a plant as soon as the price falls below its variable costs (minus the costof maintaining it in mothballs). Similarly, an even lower price is requiredto make total scrapping optimal. To the extent that withdrawing capacityfrom the market is likely to raise prices, this means that prices are likely tostay depressed for longer than if there were no sunk costs. On the otherhand, the option to mothball a plant, once it has been built, reduces thepotential losses from a period of depressed prices, and therefore raises thevalue of making the initial investment. This tends to encourage investment,compared to a situation in which mothballing is not possible.

So far, we have assumed that decisions can be implemented as soon asthey are taken. Bar-Ilan and Strange (1996) consider the impact of invest-ment lags – it normally takes several years to build a power station, and thepre-construction formalities, such as obtaining planning permission(zoning permits) can add considerably to this time.3 When these lags exist,uncertainty can actually encourage investment, in the sense of reducing the


trigger price at which it is optimal to start construction. The intuition is thatfor a given level of uncertainty and of the current price, the expected profitfrom starting operation in, say, four years’ time is greater than the expectedprofit from starting now. The price might rise over the intervening period,which would raise profits, while the downside risk from a price fall is limitedby the option of abandoning the project. While waiting to start the invest-ment still has a value, in that the firm learns more about the evolution ofprices, it involves a cost, in that if prices rise strongly, the firm will not beable to take advantage of these immediately, and will have missed out onsome potential profits. A higher level of uncertainty increases the expectedvalue of the benefit from high prices, while the losses from low prices arestill capped by the option of abandonment. Bar-Ilan and Strange show thatfor some parameter values, the overall effect of an investment lag is to lowerthe trigger price at which investment is started to below the price that wouldtrigger investment in a world of certainty. For a time to build of three orfour years, however, which should be more than sufficient for a combined-cycle gas turbine, uncertainty still raises the trigger price at which it isoptimal to invest.

Bar-Ilan et al. (2002) obtain similar results in a model which explicitlystudies the problem faced by a (traditional) electric utility. The utility facesa cost of carrying excess capacity, and a (possibly different) cost ofinsufficient capacity, and must decide when, and how much, capacity tobuild in order to meet its stochastically evolving demand. The utility mustincur a fixed cost each time it adds capacity, as well as the per-unit cost ofthe capacity. If construction lags are short, then uncertainty leads theutility to wait longer before adding capacity (that is, it will not invest untilthe level of demand is higher, relative to its existing capacity), although itwill then add a larger increment of capacity. If construction lags are long,however, the utility may optimally choose to invest when it has a higherlevel of spare capacity (with growing demand, this effectively means invest-ing earlier), and to build less capacity each time. Once again, however, avery long lag is required for the utility to find it optimal to invest earlierthan it would do in a world of certainty.

Finally, we should perhaps ask how applicable the model of a stochasticelectricity price is. The impact of the real options approach to investmentdepends upon the level of uncertainty – if there is very little uncertainty,then option values are low. If the level of capacity adjusts quickly to thelevel of demand, then this would produce an electricity price that wasalways close to the cost of generation. In technical terms, the price wouldbe strongly mean reverting. This again reduces the level of uncertainty, andhence the value of real options. The extent and speed of mean reversion is,in principle, an empirical question, and one that requires rather more data


than we have available, given the relatively recent spread of electricitymarkets across the world. Pindyck (1999), however, shows that while manyfuel prices have been mean reverting over the long term, their short-termmovements can be considered as Brownian motion. Furthermore, a modelwith mothballing and entry will tend to put a floor and a ceiling on the priceof electricity, but the price can still vary over an economically significantrange. Our conclusion is that the real options approach does have lessonsfor investment in power stations. This conclusion will be strengthened whenwe consider the propensity of many industries to invest in waves, creatingcycles of over capacity and low prices, followed by inadequate capacity andhigh prices.

4. INVESTMENT CYCLES

Many capital-intensive industries are prone to investment cycles. At first, theindustry may be short of capacity and prices will be high. This acts as asignal to investors, who start to add capacity. In the absence of a coordina-tion device, however, they are in danger of over-reacting – too manyinvestors read the high prices as a signal that their own investment will beprofitable, and somehow fail to take the likely actions of others intoaccount. Once the new capacity comes on stream, it will depress prices. Thiswill be sufficient to halt most new investments, but the existing capacity islikely to stay in service. Scrapping decisions are irreversible and will not betaken unless the price falls sufficiently below the variable costs of staying inoperation. Furthermore, in a competitive industry there can be coordi-nation problems – all firms can agree that some capacity should be taken outof service, but the firm that actually does so will incur a loss that it wouldprefer others to bear. Eventually, however, capacity will fall as plants areretired, or demand will rise. As the margin between capacity and demandnarrows, prices will rise again, up to the point where investment is again per-ceived as profitable. The industry is then in danger of repeating the cycle.

Industries such as oil tankers and copper are well known for exhibitingthese kinds of cycle (Hawdon, 1978; Brennan and Schwartz, 1985). Therehave also been fears that the electricity industry could be vulnerable tosimilar problems. Bunn and Larsen (1992, 1994) discuss investment inEngland and Wales under the Pool system, which linked prices explicitly tothe degree of spare capacity. In each half-hour, a capacity payment equalto the calculated loss of load probability, multiplied by the value of lostload (see Box 2.2) less the bid-based price of energy, was paid to all sched-uled generators, and a similar payment was made to all those generatorsthat were available but not scheduled to operate. These payments rise



BOX 2.2 THE VALUE OF LOST LOAD AND THELOSS OF LOAD PROBABILITY

The value of lost load was used by system planners as an estimateof the amount that consumers would be willing to pay in order toavoid a power cut, and hence as an input into deciding the valueof additional capacity. It can be estimated by surveys, or by observ-ing actual behaviour, such as the point at which customers on real-time tariffs start to reduce load, or the amount that companies arewilling to spend on their own back-up generators. One importantissue is that the value of lost load varies from customer to cus-tomer, and depends upon the length of an interruption – the costper kWh of a short failure may be significantly less than the cost ofa long failure.

The loss of load probability is the probability that demand(including reserve) will exceed the available capacity, and that itwill be necessary to shed load. This depends upon the level andvariability of demand, and the level and reliability of capacity.Integrated over time, it gives the expected number of power cuts.A related concept is the expected amount of unserved energy –the number of power cuts multiplied by their length and duration.Traditionally, system planners were required to minimise the costof the system while keeping either the expected number of outagesor the amount of unserved energy below target levels.

The value of lost load gives the gross value of additional powerwhen the system is short of capacity and customers are beingrationed. Since the variable cost of production must be incurred toprovide this power, the net value of additional power (and hence ofthe capacity that could produce it) is equal to the gross value, lessthe variable cost. The expected value of capacity is thus the valueof lost load, less the plant’s variable cost, multiplied by the loss ofload probability.This, using the plant’s bid in place of variable cost,is the unscheduled availability payment that was given to all plantthat was available but not actually generating in the Electricity Poolof England and Wales. A plant that was scheduled to generatereceived a capacity payment equal to the value of lost load, minusthe marginal price of power, the system marginal price, multipliedby the loss of load probability. In theory, these payments repre-sented the true value of capacity in each half-hour, and the amountthat consumers should pay towards its costs.

sharply when the margin of spare capacity falls, for the loss of load prob-ability is very non-linear in spare capacity. If investors are over-sensitive tothe current level of capacity payments, compared to a long-run equilibriumlevel, then too much capacity will be ordered during times of shortage whencapacity payments are high, leading to a surplus of plant and negligiblepayments in a few years’ time.

Ford (1999, 2002) discusses similar issues in the context of the UnitedStates, and particularly the western states. Writing before the Californiacrisis, he predicts (1999) that if generators had to rely on the (now-defunct)Power Exchange’s energy price for all of their revenues, then they would beable to cover their full costs only at times of relative shortage. The sensitiv-ity of revenues to the level of spare capacity, coupled with delays in the per-mitting and construction of power plants, would lead to alternating periodsof high and low prices. His model suggested that these could be largelyeliminated by introducing a fixed (in US$/MWh) capacity payment, paid toall generators. A payment of US$5/MWh, for example, would stabilise gen-erators’ earnings and allow them to cover their full costs without requiringthe Power Exchange price to rise in shortage conditions. In the long-runequilibrium, there would be little difference between the average price withthe capacity payments and without, although prices would be higher in theshort run with the capacity payment scheme. These higher prices presum-ably encouraged sufficient extra investment to create a cushion of sparecapacity that allows the energy price to stabilise at the level of marginalcost, with little or none of the scarcity element shown to the top left ofpanel 5 of Figure 2.1.

Ford (2002) argues that investors’ overbuilding is typically the result ofnot taking into account the effect of others’ investment decisions on themarket price. They may well have accurate expectations about the price andhow it will respond to their own investments, but are not able to keep trackof other investors’ decisions during an investment boom. As a result, toomuch capacity is added, sowing the seeds for a following bust and futureshortage – Ford predicted that California could suffer further shortages ofpower in 2007, despite the large amount of capacity added just after thecrisis of 2000–01.

Are electricity markets fated to suffer from such cycles? Mothballing andlong-term contracts may help to smooth them. In England and Wales,nearly 10 per cent of the industry’s capacity has at times been mothballed,unavailable for short-term use (and freed from the requirement to pay trans-mission charges), but capable of being returned to service should marketconditions change. Putting older plant into mothballs as a wave of newcapacity enters the market can help to put a floor under prices. It should benoted, however, that since mothballing a plant and then returning it to


service involves sunk costs, the real options approach implies that this pricefloor will be below the level of the plant’s variable costs. Returning the plantto service can help to slow the increase of prices once demand rises relativeto capacity. Since the costs of taking a plant out of mothballs are much lessthan those of a new plant, the price needed to trigger new investment willbe higher than the price needed to return a mothballed plant to service. Thisimplies that by the time investors are rationally contemplating new invest-ment, there should be few mothballed plants to cushion further price rises.4

Long-term contracts may help to reduce investment cycles in two ways.First, they can reduce uncertainty for the entrants, allowing them to lockin a price, or a margin between their fuel costs and their selling price. Thiswill lower the price needed to trigger investment. First, it can reduce theentrants’ cost of capital, by allowing higher gearing, since their cash flowsare less uncertain. Second, less uncertainty reduces the value of waitingbefore investing, and lowers the trigger price for investment.

The second way in which long-term contracts may help to reduce invest-ment cycles is as a coordination device. The demand for long-term con-tracts comes from electricity retailers, and will depend on their estimates ofthe future demand for power. While one cause of overinvestment may bethe independent decisions made by investors who are not fully aware ofeach other’s plans, the retailer knows its own demand for power. Investorswho are unable to find a retailer to contract with will not be able to taketheir project forward on that basis. This does not mean that they will haveto abandon the project, however, because it is still possible to invest as a‘merchant plant’, selling in shorter-term markets. With higher risks, theywill need less gearing, which will raise their cost of capital. Furthermore,the higher uncertainty raises the trigger price at which investment becomesoptimal. If retailers are signing a lot of long-term contracts, there may beenough capacity to keep prices below this trigger, so that merchant plantsare neither needed nor observed. If retailers are reluctant to sign long-termcontracts, however, then there would not be enough capacity without mer-chant plants. Prices will have to reach levels that can trigger investment inthese plants, and without the coordination device of contracts, there is adanger that the reaction to high prices could be excessive.

Why would retailers be reluctant to sign long-term contracts? In the past,such contracts acted as a hedge on their purchase costs and reduced theirrisks. If retailers face competition, however, then long-term commitmentscan increase their risks (Newbery, 2002). If the wholesale market price ofpower falls, then entrants to retailing can take customers from the incum-bents, unless the incumbents reduce their prices in line with the wholesaleprice.5 If the retailers’ selling prices depend upon the current wholesaleprice, but their costs depend upon their long-term contracts, then their


profits become more risky. The best response in this situation is to reducethe proportion of their demand that they buy on long-term contracts.Green (2004) models this process, in a setting where the wholesale marketis imperfectly competitive and the volume of contracts affects the mark-upover marginal costs, rather than the amount of capacity. He shows that itcan lead to a quantitatively significant impact on contracting and on whole-sale market prices.

5. PAYING FOR CAPACITY IN PRACTICE6

The analysis so far has been based on the idea that prices in the highest-demand periods can rise above the marginal operating cost of the peak gen-erators, and that they will rise by more when capacity is relatively scarce.How do markets around the world put this idea into practice?

The key distinction is between markets that are ‘energy only’ and thosein which there are explicit arrangements to pay for capacity. Australia hasan energy-only market with a price cap set at A$10,000/MWh. If the systemoperator ever has to shed load because of a lack of capacity, the price is setto this level. At other times, the market price is set by the marginal genera-tor’s bid. When capacity is scarce, but sufficient to meet demand in full, gen-erators are likely to raise their bids above their marginal operating costs,recovering part of their fixed costs. In practice, most generators and retail-ers are likely to hedge most of their expected trades in advance. This sta-bilises the generator’s revenues, and minimises the retailer’s exposure toprices close to the cap. The price at which trades are hedged will depend onhow often spot prices are expected to approach the cap, which will dependon the level of spare capacity relative to normal levels of demand.

The level of the cap is critical to the success or failure of the scheme. Ifthe cap is too low, generators will be unable to cover their costs unlesspower shortages are frequent, which is unlikely to be politically acceptable.If the cap is too high, the problem will not be one of cost recovery but ofexcessive prices. In many markets, a price cap has been imposed as a way ofmitigating market power. The problem then is that one instrument is tryingto do two tasks – holding prices down when there is enough capacity butnot enough competition, and allowing generators to recover their fixedcosts when capacity is short. Not surprisingly, this is too much. In many USmarkets, regulators have enforced relatively low price caps and recognisedthat they will need to find other ways of recovering fixed costs. In Australia,the government actually raised the price cap from A$5,000/MWh toA$10,000/MWh in 2002, in response to fears that the lower figure wouldnot provide sufficient incentives to generators to keep capacity available.


The calculations which Australian market participants make if theychoose to hedge are similar in spirit to those which underlay the capacitypayment in the former Electricity Pool of England and Wales. As men-tioned above, this payment was based on the net value of lost load, multi-plied by the loss of load probability. This is effectively the expected value ofpayments under a strict version of the Australian system, in which pricesrose above marginal operating cost only when load was actually lost, andassuming that the price cap was set at the value of lost load. The Poolbecame discredited in practice (Offer, 1998), but in principle, capacity pay-ments should have given the correct incentives for investment and closuredecisions – a plant’s expected capacity revenues (which could easily behedged for at least a year in advance) depended on the value of the powershortages that it would prevent. The value of lost load and the loss of loadprobability were discussed in Box.2.2, above.

The main alternative to an energy-only market is to have a market forenergy and a separate market for capacity. The system operator, or a regu-lator, defines the amount of capacity that every retailer must have access to,relative to the load that it serves. Typically, this requirement for installedcapacity will exceed the retailer’s expected peak demand, since the actualpeak may be higher than expected, and not all capacity will be available atthe peak time. In order to meet the requirement, generators can sell, andretailers buy, capacity credits. A retailer that does not have enough creditshas to pay a penalty. This naturally caps the price that would be paid forcapacity credits, and so the penalty should be related to the cost of makingcapacity available. To the extent that the average price of capacity creditswill be less than the penalty, the penalty should exceed the fixed cost ofcapacity; to the extent that generators will have other revenues, the penaltycan be reduced. In the PJM (Pennsylvania–New Jersey–Maryland) market,the penalty is set to equal the fixed costs of a peaking plant.

An installed capacity market obviously provides incentives to makecapacity available. If capacity is relatively plentiful, then retailers can buythe full amount of credits required, and the price will be the avoidable costof keeping plants open. (This assumes that plants with insufficient revenuesfrom capacity credits would close.) If capacity becomes short, so that someretailers are facing the penalty, this will set the price of unsold capacitycredits. The price of capacity credits can thus vary from the cost of keepingold plant open to the cost of building new plant – as long as the penalty hasbeen set at an appropriate level. If so, the market can send appropriatesignals about investment and closure decisions – investment should beprofitable when the market falls short of its installed capacity requirement.

The level of the installed capacity requirement is therefore critical in deter-mining the amount of capacity which the industry will be given incentives to


provide. It is an administrative decision, just like the administrative decisionon the level of a price cap in an energy-only market. The advantage of choos-ing an installed capacity market is that the connection between the decisionand the risk of power cuts should be clear, as opposed to the rather indirectlink between a price cap and the resulting level of capacity.

In practice, however, administrative decisions are common throughoutelectricity markets. System operators have to decide how much reserve tobuy from day to day. Generators, and consumers able to reduce load atshort notice, may form the supply side of an active market, but the demandside is missing. Even if consumers knew their willingness to pay to avoidhaving their own load cut off, there is a public good aspect to reserves, inthat inadequate levels of reserves can result in a system-wide failure. If themarket grows short of reserves, system operators will reduce their require-ments to a minimum, raise prices and start to ration consumers throughrolling blackouts, aiming still to avoid a system breakdown. It is not clearhow prices are set in these circumstances. Economists need to learn moreabout how electricity system operations determine prices at peak times.Joskow and Tirole (2006) set out some of the key issues involved, but thereis room for further study.

6. INVESTMENT IN LIBERALISED MARKETS

We conclude this chapter with a brief look at investment patterns in anumber of liberalised electricity markets. Figure 2.3 shows investment inEngland and Wales since the reforms of 1990. The top line shows thecapacity margin – the excess of generation capacity over peak demand(weather adjusted), as a proportion of that peak demand. This started theperiod at more than 30 per cent, above the planning margin of 28 per centused by the nationalised Central Electricity Generating Board. With anapparently high level of spare capacity, we might expect little investment,but the second line in the figure clearly shows a stream of new investmentduring the 1990s, amounting to nearly half the peak demand. Most of thiswas in combined-cycle gas turbine (CCGT) stations, offering much greaterthermal efficiency and lower emissions than the industry’s existing coal-and oil-fired stations. At the time of the restructuring, nearly 80 per cent ofthe industry’s capacity was owned by just two generating companies, andthese built some CCGT stations in order to modernise their portfolios andreduce their emissions of sulphur dioxide. Most of the new capacity wasbuilt by entrants to generation, however. The first stations were mostly builtby the regional electricity companies, the former distribution and retailingmonopolies. The stations gave them some insurance against the market


power of the major generators, by diversifying their purchases, and allowedthem to earn some unregulated income. Both these stations and a later waveof unaffiliated entrants expected to benefit from the major generators’market power, in that prices in the wholesale market remained above thelevels needed to remunerate new entrants for most of the 1990s.

The bottom line of Figure 2.3 shows that net investment has been wellbelow gross investment – in other words, many plants have been closed overthe period. Most of these closures were by the major generators, makingway for the new entrants and keeping plant margins from rising. Capacitymargins actually declined steadily, but not excessively, through the 1990s.The small amount of net investment was not sufficient to keep up withdemand growth. The last major additions to capacity came in the financialyear 2000/01, and with little investment and some closures since then, thecapacity margin has fallen sharply. The margin between wholesale pricesand generators’ costs also fell in 2000. A period with little investment wasan appropriate response to these prices, and wholesale margins have risenrecently, reflecting the tighter market. It remains to be seen when invest-ment will recover, although two of the largest companies have recentlyannounced projects.

Figure 2.4 shows investment in Finland, which liberalised its market in1996.7 There have been very few plant closures, so that gross and net invest-ment are almost equal. There is a significant amount of investment inthe years immediately after liberalisation (which will have been plannedbefore liberalisation took effect, but during a period when it was becoming


Source: National Grid Company (NGC).

Figure 2.3 Investment in England and Wales

�0.1

0

0.1

0.2

0.3

1991/92 1994/95 1997/98 2000/01 2003/04

Proportion of peak demand

Capacity margin

Gross investment

Net investment

increasingly likely), and very little plant has been commissioned since then.Once again, capacity margins have been falling, but from a relatively highlevel. However, more plants will be needed as demand grows and old stationsretire, and a consortium of electricity companies and large industrial con-sumers is building a 1,600 MW nuclear reactor (work started in the spring of2005). This is believed to offer the lowest cost option for the base-load sup-plies of power that the industrial users need. Since the consumers are part ofthe consortium building the station, they will be insulated from the way inwhich market prices in Nord Pool depend on the level of rainfall.

Figure 2.5 shows the situation in Norway, which has also had little invest-ment. To some extent, this reflects the lack of suitable sites for new hydro-electric schemes. A national debate on whether to build gas-fired powerstations, which would increase the country’s carbon dioxide emissions (butmight reduce those of the world), has not yet reached a conclusion.Capacity margins fell sharply in the first half of the 1990s, after liberalisa-tion in 1991. They have risen since 1997, but this is because the peakdemand has been falling, rather than because of significant increases incapacity. As a hydro-dominated system, Norway will always require a rela-tively high margin of generation capacity over peak demand, because theaverage hydro-electric power station can only store enough water to operatefor about half the year. If the average level of demand is 70 per cent of thepeak demand, building enough hydro storage capacity to meet the annualdemand for energy will imply enough generating capacity to meet 140 percent of the peak demand. The level of trade among the Nordic countries


Source: Nordel.

Figure 2.4 Investment in Finland

�0.1

0

0.1

0.2

0.3

1992 1994 1996 1998 2000 2002/03


Capacity margin

Gross investment

Net investment

has risen over the 1990s, and Norway has made up for a decline in its cap-acity margins by importing increasing amounts of power in dry years.

There has been very little investment in Sweden, shown in Figure 2.6.Again, capacity margins were very high at the start of the 1990s – the countryhas a mix of hydro and thermal resources, so the appropriate margin will be


Source: Nordel.

Figure 2.5 Investment in Norway

Proportionof peakdemand

0

0.1

0.2

0.3

0.4

0.5

1992 1994 1996 1998 2000 2002/03

Capacitymargin

Grossinvestment

Netinvestment

Source: Nordel.

Figure 2.6 Investment in Sweden


0

0.1

�0.1

0.2

0.3

0.4

1992 1994 1996 1998 2000 2002/03

Capacity margin

Gross investment

Net investment

lower than for Norway. Shortly after liberalisation in 1996, a significantamount of old capacity was retired, and capacity margins fell. As a memberof Nord Pool, Sweden may be able to import power if its own supplies areinadequate, and so the desirable level of reserve capacity within its bordersfell after 1996. To the extent that the Nordic countries share commonweather patterns, however, the same shock that makes one country want toincrease its imports may be affecting its neighbours. This will reduce, but byno means eliminate, the insurance benefits of the multi-country system.However, transmission constraints can mean that capacity in anothercountry is not able to respond to failures, and more localised reserve isneeded. The Swedish system operator, Svenska Kraftnät, was concerned bythe low margins, and has been paying for stations in a capacity reserve. Thisis intended as a temporary measure, but no permanent solution has yet beenagreed. Plants entering the reserve (mostly being brought back into serviceafter mothballing in 1998 and 1999) accounted for nearly three-quarters ofthe capacity added between 2001 and 2003.

Finally, we come to the United States, shown in Figure 2.7. Note that thefigures given are nationwide averages, and the picture in individual regionsmay be quite different – in particular, low capacity margins in the westernstates contributed to the California crisis of 2000–01. On the national scale,however, capacity margins were high in the early 1990s, and there was accord-ingly little net investment. The lack of investment lasted until 1998, whilegrowing demand eroded the margin of spare capacity. Investment took off atthe end of the decade, however, with new capacity equal to over 20 per cent


Source: US Energy Information Agency.

Figure 2.7 Investment in the United States


0

0.1

�0.1

0.2

0.3

1992 1994 1996 1998 2000 2002/03

Capacity margin

Gross investment

Net investment

of peak demand added between 1999 and 2002. Three-fifths of this wasCCGT capacity, and most of the other plants were peaking combustion tur-bines. Building combustion turbines is generally the quickest way to dealwith a shortage of capacity, while the high thermal efficiency of CCGTs hasmade them the technology of choice in many markets. In the mid-1990s, moststates were debating whether to create electricity markets, and this regulatoryuncertainty caused most investors to wait until policy had been decided.

Once the uncertainty was resolved, a combination of low capacitymargins and the technological opportunity presented by CCGTs made asharp rise in investment optimal. In many markets, the CCGTs were expect-ing to replace stations burning the same fuel (natural gas) but at a thermalefficiency one-third lower. It is likely, however, that the dramatic increaseactually seen went well beyond the optimal level of investment. Electricitycompanies saw their share prices soar at the end of the 1990s, and invest-ment decisions were made in the atmosphere of a stock market bubble thatsubsequently collapsed. The incumbent utilities chose not to close the olderplants that the entrants were expecting to displace, and the price of naturalgas rose sharply, depressing the stations’ load factors as those burning otherfuels became more competitive. As in England and Wales, a number of thenew plants have suffered from financial problems.

How can we sum up these experiences? Clearly, each country is affec-ted by its own specific background, making generalisation dangerous.Liberalisation has often been accompanied by a reduction in capacitymargins, but these margins have not generally fallen to a level that posed adanger to security of supply.8 When capacity margins become very low,investment has followed. A high capacity margin is generally a signal thatlittle investment is needed, and so we should not be worried about theexamples that combine little investment with high margins. It is worthnoting, however, that we still have relatively little experience, given the longasset lives of power stations, and so it is early to draw firm conclusions. Inparticular, we do not yet know how willing investors will be to re-enter themarkets of the UK and the US when new capacity is needed, given thelosses some of them have suffered over the last few years.

7. CONCLUSIONS

This chapter has presented a theoretical model showing how the optimallevel and mix of capacity can be derived for an electricity industry withoutuncertainty. In the model, investors will receive normal profits when theindustry has the correct level of capacity, while a surplus will depress theirrevenues, sending a signal that some capacity should leave the market. If


there is a shortage of capacity, which is the more worrying case for our pur-poses, then prices will rise and existing plants will make supernormalprofits, which should act as a strong signal for new investment. In otherwords, the market is capable of sending the correct signals for the sociallyoptimal level of capacity.

The real world is not certain, of course. The real options approach toinvestment shows that it is rational to delay investment in the face of uncer-tainty, whether the investor is a private monopolist, a competitive firm or awelfare-maximising social planner. If demand evolves stochastically, theninvestment should be held back until it is clear that the new capacity willreally be needed, given the extent of sunk costs in the electricity industry.This implies that the price at which investment becomes attractive is higherthan the price needed for the plant to just cover its costs, since there is a riskthat the price will fall back in future. Similarly, it is not rational to closeplants as soon as the price falls below the level of their variable costs. Aprofit-maximising firm and a welfare-maximising social planner shouldreact to uncertainty in exactly the same way, implying that there is no marketfailure. As long as the market mechanism reveals the true value of capacityat times when there is a shortage, investors should receive the correct signals.

How will investors respond to these signals? We do not yet have enoughexperience to draw definitive conclusions. In the markets shown in the pre-vious section, there does seem to be an inverse correlation between invest-ment levels and the margin of spare capacity, which is desirable. Capacitymargins seem to be stabilising at lower levels in liberalised markets than atthe start of the 1990s, which is also probably desirable – holding too muchspare capacity is expensive. The US probably had an excessive level ofinvestment at the end of the 1990s, feeding through to a rapidly risingcapacity margin, and financial problems for many investors, in the earlyyears of this century. The UK experienced some similar problems on asmaller scale. Overall, the experience is mixed. When judging the ability ofmarkets to coordinate investment, however, we must remember that thealternative is not the outcome of a perfect social planner, but of imperfectregulation. On that basis, market-driven investment in electricity wouldseem to have an acceptable record, and prospects.

APPENDIX 2A A MATHEMATICAL MODEL OFCAPACITY AND PRICING

Let us consider the problem of finding the welfare-maximising levels ofcapacity, and period-by-period outputs. The dual variables to this problemare the electricity prices in each period. Assuming that a competitive


electricity market has rules that will set these prices, we can then rely on thefundamental theorem of welfare economics to show that these are the quan-tities of output and capacity that the competitive market would produce.

We shall work in discrete time, with T time periods, and I types of plant.Period 1 has the highest demand, and T the lowest, working from left toright of the load-duration curve in panel 2 of Figure 2A.1. Plant type 1 hasthe lowest operating costs, and type I the highest, with the others ranked inorder, working up the stack of capacity in the same panel. The operatingcost of plants of type i is ci per unit of output per period, and the fixed cost(for the whole planning horizon) is di per unit. Without loss of generality,we only list capacity types that will be used in equilibrium, which has toimply that d1�d2� . . . �dI. The output from plant type i in period t is qi,which must be less than its capacity of ki. The gross benefit from consum-ing an output of Qi in period t is Bt(Qi). Our problem is thus:

(2A.1)

Alternatively, we can write it as a Lagrangean, which gives us:

(2A.2)

If we then differentiate with respect to our choice variables and theLagrangean multipliers, we obtain:

(2A.3)

(2A.4)

(2A.5)

This gives us a first-order condition for output:

(2A.6)0 � qit � B�t��

iqit� � ci � �it � 0.

�L��it

� qit � ki.

�L�kit

� �t

� it � di

�L�qit

� B�t��i

qit� � ci � �it

Maxqit, ki

L � �t

Bt��i

qit� � �i�

tci qit � �

idiki � �

i�

t�it(qit � ki).

s.t. qit � ki � i, t

Maxqit, ki

W � �t

Bt��i

qit� � �i�

tci

qit � �i

di ki.



Notes:(1) Total costs by plant type.(2) Load-duration curve.(3) Reflecting line.(4) Marginal cost and demand.(5) Price-duration curve.

Figure 2A.1 The determination of electricity capacity and prices

GW

Hours/yearGW PT*– s T*+uT* B K D

CB

CP

PR

Hours/year

Hours/year

€/MW-year

GWGW

€/MWh €/MWh

B K D P T*

T*(2)

Peaking Baseload

(1)

(5) (4)

(3)

B

D

K

Maxdemand

B+D

With �it always (weakly) positive, this states that the output from planttype i will be positive if the marginal benefit of consumption equals orexceeds its variable cost. Furthermore, whenever the plant is producing, �itwill equal the difference between the marginal benefit of consumption andthe plant’s variable cost. From the second set of derivatives, we have:

(2A.7)

This tells us that the sum of those differences (between the marginalbenefit of consumption and its variable cost, while the plant is operating)will equal its fixed cost. In other words, the total cost of the plant is equalto the marginal benefit of its output, summed over all the periods in whichit is running. The third set of derivatives reminds us that the output fromeach plant type cannot exceed its capacity:

(2A.8)

At times when the plant is not running at full capacity, �it is zero – relax-ing the capacity constraint in those periods would not raise social welfare.These equations can be solved for the optimal capacities and pattern ofoutput. In a market system, the market price in each period should be equalto the marginal benefit of consumption, . We can get a better feelfor the pattern of outputs and capacities if we introduce some notation forthe number of periods of operation over which two (adjacent) plant typeshave equal total costs:

(2A.9)

We then know that we want plant of type i to be marginal at time ti, sothat all the plant of type i1 should stop running before this time. Thatallows us to calculate the total capacity requirement of types 1 to i, as equalto the level of output that will equalise the marginal benefit of consump-tion at time ti with its marginal cost:

(2A.10)

Note that this formulation includes the Lagrangean multiplier for themarginal type of plant, since that plant type might be capacity constrained.In terms of panel 4 of Figure 2A.1, we might be on one of the vertical seg-ments of the industry marginal cost curve. If we ignore that possibility, then

B�ti��i

j�1kj� � ci �iti.

ti �di � di1ci1 � ci

.

B�t(�

i qit)

0 � � it � kit � qit 0.

�t

� it � di.


the requirement is that the marginal benefit of running all the capacity upto and including type i should just equal the variable cost of plant type i.

We determine the amount of peaking capacity (type I) in a slightlydifferent way. We need to find the time, tr, at which the marginal benefit ofconsumption just equals the variable cost of peaking capacity, assumingthat all the plant types are running at capacity:

(2A.11)

We can then obtain an equation for the amount by which the marginalbenefit of consumption exceeds the variable cost of peaking capacity, ineach period up to period tr. This is the net benefit of the marginal unit ofcapacity in that period. The sum of these net benefits should equal the fixedcost of peaking capacity, dI:

(2A.12)

Solving equations (2A.11) and (2A.12) together gives us the solution forthe total amount of capacity, and hence the amount of peaking capacity.

One simplification in the main text is that the load-duration curve hasbeen drawn as a smooth curve, despite the apparently large drop in pricethat occurs at T*. Such a fall in price would normally lead to an increase indemand, of course. However, the load-duration curve has to be monotonic,by construction. Figure 2A.1 has been drawn to show what would actuallyhappen. We would expect some demand curves to pass through the verticalsection of the marginal cost curve in panel 4, and the price at those times isbetween the variable cost of base-load plant and the variable cost ofpeaking plant. In our mathematical notation, �it is between zero and (CP –CB) in those hours.

If we had enough peaking plant that some of it was still running untiljust before hour T*, then the base-load plant would already have recoveredits fixed costs by this time. With several periods just after T* in which theprice exceeds its variable cost, the base-load plant would earn supernormalprofits. Equation (2A.7) shows that this is not a feature of the optimal solu-tion. Instead, we need to have less peaking plant, and more base-load plant(within the same total capacity), so that the price starts to fall below CPbefore hour T* (at, say, T* – s), and is still above CB until some point afterhour T* (say, T*u). This gives us the downward-sloping segment in theprice-duration curve of panel 5. Because we are moving down a vertical

�tr

t�0B�t��I

j�1kj� � cI � dI.

B�tr��I

j�1kj� � cI.


segment of the marginal cost curve in these hours, the level of output isconstant, and we therefore have a horizontal segment in the load-durationcurve, from T* – s to T*u, at a height equal to the (new) level of base-load capacity, BD. In practice, however, most electricity industriescontain so many plants that the marginal cost curve has few large discon-tinuities, giving a smooth load-duration curve.

NOTES

* Support from the Commission de Régulation de l’Énergie is gratefully acknowledged.I would like to thank the Department of Applied Economics, University of Cambridge,for its hospitality. I have benefited from helpful comments by François Lévêque, Jean-Michel Glachant, Paul Joskow, Steven Stoft and participants at a seminar at the CRE.

1. This model draws on the presentation in Stoft (2002), although he does not put the dia-grams on a single page.

2. Depending upon the market arrangements, these could be contracts for physical delivery,or financial hedging contracts for differences around a spot market price.

3. The pre-construction stages obviously involve some sunk costs, but as long as thecompany does not sign binding contracts until these stages are completed, it is the lengthof the construction period that mainly determines the lag between committing to the irre-versible investment and being able to earn revenues from a power station.

4. If all plants were identical, it would not be optimal to build a new plant while an existingplant was in mothballs, but it is conceivable that investment in an efficient new plant couldbe profitable while an inefficient old plant was best kept out of the market (but not yetretired).

5. The incumbent may be protected by switching costs, so that few customers will switchunless the incumbent is significantly more expensive than its rivals, but even in this case,the incumbent’s optimal retail price varies with the current wholesale price.

6. This section owes a lot to comments from Steven Stoft, although he should not be heldresponsible for the way in which I have interpreted them.

7. The data on peak demands, obtained from Nordel, are not adjusted for weather, and athree-year moving average has been used to smooth out fluctuations in Figures 2.4–6.

8. It is far from clear that the power cuts in California in 2001 were caused by a physicalshortage of capacity, as opposed to a shortage of capacity being offered to the market(Blumstein et al., 2002).

REFERENCES

Bar-Ilan, A. and W.C. Strange (1996), ‘Investment lags’, American EconomicReview, 86 (3), 610–22.

Bar-Ilan, A., A. Sulem and A. Zanello (2002), ‘Time-to-build and capacity choice’,Journal of Economic Dynamics and Control, 26 (1), 69–98.

Blumstein, C., L. Friedman and R.J. Green (2002), ‘The history of electricityrestructuring in California’, Journal of Industry Competition and Trade, 2 (1–2),9–38.

Brennan, M.J. and E.S. Schwartz (1985), ‘Evaluating natural resource investments’,Journal of Business, 58, 135–57.


Bunn, D. and E. Larsen (1992), ‘Sensitivity of reserve margin to factors influencinginvestment behaviour in the electricity market of England and Wales’, EnergyPolicy, 20 (5), 420–29.

Bunn, D. and E. Larsen (1994), ‘Assessment of the uncertainty in future UK elec-tricity investment using an industry simulation model’, Utilities Policy, 4 (3),229–36.

Dixit, A.K. and R.S. Pindyck (1994), Investment Under Uncertainty, Princeton, NJ:Princeton University Press.

Ford, A. (1999), ‘Cycles in competitive electricity markets: a simulation study of thewestern United States’, Energy Policy, 27, 637–58.

Ford, A. (2002), ‘Boom and bust in power plant construction: lessons from theCalifornia electricity crisis’, Journal of Industry, Competition and Trade, 2 (1–2),59–74.

Green, R.J. (2004), ‘Retail competition and electricity contracts’, CMI WorkingPaper EP33, University of Cambridge–MIT Institute.

Hawdon, D. (1978), ‘Tanker freight rates in the short and long run’, AppliedEconomics, 10 (3), 203–17.

Joskow, P.L. and J. Tirole (2006), ‘Reliability and competitive electricity markets’,Rand Journal of Economics, forthcoming.

Newbery, D.M. (2002) ‘Problems of liberalising the electricity industry’, EuropeanEconomic Review, 46, 919–27.

Offer (1998), Review of Electricity Trading Arrangements: Proposals, July 1998,Birmingham: Office of Electricity Regulation.

Pindyck, R. (1999), ‘The long-run evolution of energy prices’, Energy Journal,20 (2), 1–27.

Stoft, S.E. (2002), Power System Economics: Designing Markets for Electricity,Chichester: Wiley.


3. Generation technology mix incompetitive electricity marketsJean-Michel Glachant

1. INTRODUCTION

Competitive electricity reforms began to emerge at the beginning of the1990s. We now have a certain distance, especially in Europe and the UnitedStates, for observing behaviour in terms of investments in generationwithin a competitive framework. The oldest of these reforms currently have15 years of experience, as in England (in effect since 1990) and Norway(since 1991). The first European Directive on creating a single electricitymarket (Eu,1996) was adopted a decade ago and enacted as of 1997 (likein Spain) and 1999 (like in Italy). In the United States, the beginning of thecompetitive reforms can be traced to the creation of independent systemoperators (ISOs) between 1995 and 1997 covering Texas, California andPJM (Pennsylvania–New Jersey–Maryland), but a true opening of com-petitive wholesale markets cannot be said to antedate 1998.

Did these competitive reforms of the electricity industry have an impacton the choice of generating technologies? Do the new competitive pressurescreate an incentive for producers to select new methods for generating elec-tricity (like the combined–cycle gas turbine: CCGT) and to abandon oldtechnologies adopted under their previous status as utility monopolies,these former technologies favoured by policies and government subsidiesand financed by a guaranteed sales price? Would there be a new trend thatthe most capital-intensive technologies (for example, nuclear power) wouldbe avoided? What link can we trace from actual competitive market reformto actual generation technology choice?

Chapter 2 discussed the economic theory of capacity investment and itscycles, and the actual patterns in liberalised markets; this chapter is a moreempirical study of actual generation technology choice. We shall addressthe potential impact of the competitive reforms on the choice of produc-tion technology in three stages.1 The first stage (Section 2), examineswhether the main competitive reforms in the United States and Europefostered an evolution in the volume and technology of investments in

54

generation. If a technology change occurs with low volumes of investmentwe may doubt its significance. Therefore we have to establish whether thesereforms were indeed often accompanied by high levels of investment, andwhether a technological change particularly favoured new gas-basedCCGT technology.

Having verified both the importance of the investment wave and of thetechnological shift, we have to ask, in a second stage (Section 3), why thistechnology change occurred in that wave of capacity investment. Bylooking at the traditional method of comparing electricity generation costs,‘levelised cost methodology’, we shall find that it provides a simple eco-nomic explanation for this technological shift by establishing the CCGTsystem as the least expensive among existing technological alternatives.However, in contrast to cost analysis performed in the UK and the US, themain French analysis of the cost of electricity generation always finds infavour of nuclear technology. Then, in Section 4, the third and final stageis dedicated to understanding why cost analysis can diverge that much. Inparticular, we shall focus on the economic determinants that have beenaccounted for in the comparative analysis of technology costs under a com-petitive framework, taking as a benchmark the cost study made at theMassachuselts Institute of Technology (MIT) in 2003. As already sug-gested by Chapter 2, we shall see that the core of this debate on the actualcosts of new plants is on evaluating the effects of competitive powermarkets and of vertical disintegration of generation on the cost of capital.Since nuclear power is extremely capital intensive, two to three times morethan its alternative technologies, it is much more sensitive to the way thefinancial market and the banking industry actually take into account therisks and uncertainty of generation investment.

Section 5 concludes.

2. ONE DECADE OF INVESTMENT INGENERATION IN A COMPETITIVEFRAMEWORK

The multitude of US and European experiences in competitive electricityreforms is characterised by a vast range of timings and a broad diversity ofmodalities. These reforms do not all present the same investment profile ortechnological choice. However, two broad traits can be seen to prevail. Onthe one hand, when these reforms were accompanied by investments in gen-eration, very large volumes of new capacity were created (up to 30 or 40per cent). This simple fact means that it has been a wave of significantinvestment. On the other hand, the technology principally used in this

Generation technology mix 55

important wave of ‘competitive’ investments is relatively new: gas-basedsystems (up to 90 per cent of new capacity). This shows that the new waveof investment is characterised by a kind of technology shift.

A Large Volume of Investment in Capacity

The evolution of generating capacity in the United States between 1990 and2002 shows a pronounced change in pace during the recent ‘competitive’period of 1998–2002. During the eight years preceding the pivotal yearof 1998, total production capacity increased by only 44 GW, or about 5 percent, while during the four years following 1998 this capacity rose by155 GW – about 20 per cent. This represents a genuine leap forward in gen-eration capacity (Hunt, 2002).

However, since a number of US states either did not embark on thesecompetitive reforms, or did so belatedly, there is some room for doubt con-cerning the link between the evolution of total volume of capacity and thereforms. This is why we must look at the distribution of these capacitiesbetween the electricity utilities (being the traditional regulated companiesfrom before the reforms) and the independent power producers (the IPPs).These ‘independent’ generators, though they may have been created beforethe competitive wave of the late 1990s and owned by the utilities, none theless represent one of its key features, because such generators work outsidethe traditional regulatory framework that has been applied to utilities fordecades.2

In the eight years following 1990, total capacity owned by the utilitiesdeclined by 7 GW (–1 per cent) while that owned by the IPPs increased by28 GW (300 per cent, either by construction or by acquisition from theutilities). However, during the four years after 1998, which was the truestart of the competitive era, the capacity owned by the utilities fellsubstantially, by 132 GW (–18 per cent), while that owned by the IPPs shotup 271 GW (700 per cent, either by construction or by acquisition asnew subsidiaries of the utilities). The strong growth in investments duringthe early competitive period, 1998–2002, which appears to be withoutprecedent in terms of volume, is thus particularly characterised by theactions of the IPPs. They not only accounted for the bulk of investmentsin new capacity, but also substantially dismantled the installed base ofpower plants managed by the utilities. When these four exceptional yearsended in 2002, the IPPs found themselves owning a generating capacity ofnearly 310 GW, or over half as much as the traditional utilities (597 GW)(Table 3.1). It is true that these IPPs can be subsidiaries of the utilities, par-ticularly from other states. But this does not negate the special role theyhave played in this massive wave of recent investment in generation.


If we now concentrate on only two of the most extensive competitivereforms in the United States, California and Texas (representing a total ofnearly 130 GW in capacity in 1998), we again find evidence of intenseinvestment in generation (20 GW overall) between the years 2000 and2002. The increase in capacity attains 10 per cent in California, and 30 percent in Texas where we observe a ‘boom’ in capacity (Table 3.2).

In Europe, two countries also witnessed considerable investments in gen-eration under the competitive reforms (totalling approximately 45 GW).First, in England and Wales the equivalent of 40 per cent of the initialcapacity was added during the first decade of the reform, while Spain addednearly 30 per cent in five years. Italy also created nearly 6 per cent in newcapacity in the first four years. Only Norway’s capacity had scarcelychanged (3 per cent) 10 years after the competitive reform, (Table 3.3).

A Profound Change in Technology: The ‘Dash for Gas’

While in the first years of the competitive era in the United States, thegrowth of capacity investment was unprecedented, the evolution of the fuel


Table 3.1 Generation capacity in the USA, 1990–2002

Producer type Nameplate capacity (gigawatts)

1990Total industry 781

Utilities 735IPP 9

1998Total industry 825Change total industry 1990–1998 44

Utilities 728Change Utilities 1990–1998 �7

IPP 37Change IPP 1990–1998 28

2002Total industry 980Change total industry 1998–2002 155

Utilities 597Change Utilities 1998–2002 �131

IPP 308Change IPP 1998–2002 271

Source: Own calculations – data from IEA.

mix too presents a break during this post-1998 competitive period (Tables3.4 and 3.5) Previously, between 1990 and 1998, coal and nuclear tech-nologies dominated generation in the United States with 56 per cent (in1990) and 54 per cent (in 1998), while gas (gas only or dual fuel-oil and gas3)accounted for about one-quarter (23 and then 27 per cent). During the firstcompetitive period, 1998–2002, gas-based systems jumped to 38 per cent ofcapacity (of which approximately 20 per cent was gas only), while coal andnuclear fell well below half at 45 per cent. In fact, in absolute value, coaland nuclear generation capacity remained unchanged over these four yearsof technological change, with 338 GW and 105 GW, respectively. It is ratherthe expansion of gas (113 GW) and dual fuel (38 GW) that opened thetechnological shift. New capacity in these two gas-based technologies rep-resented 97 per cent of incremental output in the United States between1998 and 2002.

The comparative evolution of the capacity and fuel mix of the utilitiesand the IPPs is also very significant (see Tables 3.4 and 3.5). In 1990, theIPPs generated 90 per cent of their output from hydro and other renewableenergy sources. Between 1990 and 1998, increased capacity of the IPPs inthe two gas-based technologies (gas only and dual fuel) represented 49 percent of total growth in these technologies in the United States, while thecorresponding number for the utilities was only 11 per cent4. Consequently,by 1998 the two gas technologies accounted for 52 per cent of the IPPs’


Table 3.2 Generation capacity in California and Texas

State Installed capacity (gigawatts) Annual variation (%)

CaliforniaYear 1998 54.3Year 1999 54.1 �0.6Year 2000 54.1 0Year 2001 57.2 5.7Year 2002 59.5 4.0Total change 1998–2002 5.2 9.6

TexasYear 1998 78.2Year 1999 79.9 2.2Year 2000 86.8 8.6Year 2001 94.0 8.3Year 2002 101.7 8.2Total change 1998–2002 23.5 30

Sources: Various years: EIA-DOE; PJM reports.

installations, while hydro and other renewables amounted to only 28 percent. However, at 10 GW, the ‘gas only’ capacity of the IPPs was only one-quarter that of the utilities. Between 1998 and 2002, the increase in theIPPs’ capacity in the two gas technologies represented 99 per cent of thetotal growth in these technologies in the United States. While the utilities’additional ‘gas only’ capacity amounted to only 8 per cent of that total,their capacity in dual-fuel fell by 16 per cent (the utilities converted or soldto the IPPs some 24 GW of their dual-fuel capacity). After this short periodof intense technological change, the IPPs’ ‘gas only’ capacity was 99 GWin 2002 and nearly twice as large as the utilities’ ‘gas only’ capacity. Overall,55 per cent of the total generation capacity of the IPPs consists of gas ordual fuel, as opposed to 26 per cent in the case of the utilities.

As a result, the first years of the competitive era opened a sweeping andrapid technological switch to gas in the United States, and it is actually closelylinked to the competitive reform. The technological evolution of two big


Table 3.3 Generation capacity in Norway, England and Wales,Spain and Italy

Country Installed capacity (gigawatts) Variation (%)

NorwayYear 1991 27.1Year 2002 28.0Change 1991–2002 0.9 3.3

ItalyYear 1998 75.0Year 2002 79.2Change 1998–2002 4.2 5.6

England & WalesYear 1990 63.9Year 2000 71.6Change 1990–2000 7.7 12Capacity additions 1990–2000 25.0 39Capacity closures 1990–2000 �17.3 �7Capacity mothballed in 2000 5.4 8

SpainYear 1998 49.1Year 2003 63.6Change 1998–2003 14.5 29.5

Sources: Various years: Statistics Norway and NVE; British Electricity Association;GRTN; REE.


Table 3.4 Generation fuel mix in the USA, 1998–2002

Year Producer Energy Energy mix Generations Capacitytype source (%) capacity change

(megawatts) 1998–2002(%)

1998 Total electric Coal 41.0 337,800power industry

1998 Total electric Petroleum 5.5 45,300power industry

1998 Total electric Natural gas 10.0 82,100power industry

1998 Total electric Dual fired 17.2 142,100power industry

1998 Total electric Other gas 0.0 0power industry

1998 Total electric Nuclear 12.7 104,800power industry

1998 Total electric Hydroelectric 11.6 95,500power industry

1998 Total electric Other 2.0 16,400power industry renewables

1998 Total electric Other 0.1 900power industry

Total industry (year 1998)� 100.0% 824,900

1998 Electric generators, Coal 44.0 320,600electric utilities

1998 Electric generators, Petroleum 5.9 42,800electric utilities

1998 Electric generators, Natural gas 5.8 42,400electric utilities

1998 Electric generators, Dual fired 17.0 124,000electric utilities

1998 Electric generators, Other gas 0.0 0electric utilities

1998 Electric generators, Nuclear 14.4 104,800electric utilities

1998 Electric generators, Hydroelectric 12.5 91,100electric utilities

1998 Electric generators, Other 0.3 2,200electric utilities renewables

1998 Electric generators, Other 0.0 200electric utilities

Total utilities (year 1998)� 100.0% 728,100

1998 Electric generators, Coal 17.7 6,600IPPs

1998 Electric generators, Petroleum 2.1 800IPPs


Table 3.4 (continued)


(megawatts) 1998–2002(%)

1998 Electric generators, Natural gas 27.6 10,300IPPs

1998 Electric generators, Dual fired 24.4 9,000IPPs

1998 Electric generators, Other gas 0.0 0IPPs

1998 Electric generators, Nuclear 0.0 0IPPs

1998 Electric generators, Hydroelectric 8.5 3,200IPPs

1998 Electric generators, Other 19.7 7,300IPPs renewables

1998 Electric generators, Other 0.0 0IPPs

Total IPP (year 1998)� 100.0% 37,200

2002 Total electric Coal 34.5 338,200 0.1power industry

2002 Total electric Petroleum 4.4 43,200 �4.6power industry

2002 Total electric Natural gas 19.9 195,000 137.4power industry

2002 Total electric Dual fired 18.4 180,200 26.8power industry

2002 Total electric Other gas 0.2 2,200 **power industry

2002 Total electric Nuclear 10.7 104,900 0.2power industry

2002 Total electric Hydroelectric 9.8 96,300 0.9power industry

2002 Total electric Other 1.9 18,800 14.9power industry renewables

2002 Total electric Other 0.1 800 �12.3power industry

Total industry (year 2002)� 100.0% 978,800 18.8

2002 Electric generators, Coal 43.7 260,600 �18.7electric utilities

2002 Electric generators, Petroleum 4.3 25,800 �39.7electric utilities

2002 Electric generators, Natural gas 9.2 54,600 28.8electric utilities

2002 Electric generators, Dual fired 16.6 99,200 �20.0electric utilities

competitive reforms in the United States (California and Texas) thus corrobo-rates the overall American data. In the additional capacity that came online,around 30 GW was powered by natural gas. In California and Texas virtuallyall post-1998 capacity addition was in gas-based technologies (Table 3.6).




(megawatts) 1998–2002(%)

2002 Electric generators, Other gas 0.0 100 **electric utilities

2002 Electric generators, Nuclear 11.3 67,400 �35.6electric utilities

2002 Electric generators, Hydroelectric 14.7 88,000 �3.5electric utilities

2002 Electric generators, Other 0.2 1,000 �54.3electric utilities renewables

2002 Electric generators, Other 0.0 0 –electric utilities

Total utilities (year 2002)� 100.0% 596,700 �18.1

2002 Electric generators, Coal 21.7 66,900 915.8IPPs

2002 Electric generators, Petroleum 5.0 15,400 1,900.5IPPs

2002 Electric generators, Natural gas 32.1 98,800 861.3IPPs

2002 Electric generators, Dual fired 22.8 70,100 671.6IPPs

2002 Electric generators, Other gas 0.0 0 **IPPs

2002 Electric generators, Nuclear 12.2 37,500 **IPPs

2002 Electric generators, Hydroelectric 2.4 7,200 130.0IPPs

2002 Electric generators, Other 3.9 11,900 62.4IPPs renewables

2002 Electric generators, Other 0.0 100 *IPPs

Total IPP (year 2002) � 100.0 307,900 728.0

Note: (**): one cannot do a division by 0. A third producer category (CHP: combinedheat & power) is not presented.



Table 3.5 Generation capacity changes in the USA, 1990–2002

Producer type Mix of the Capacity change As % of thechange (in %) (in megawatts) industry change

Capacity change total 100.0% 43,700 100.0%industry 1990–1998

(Industry capacity change 16.8 7,300 100.0in coal 1990–1998)

(Industry capacity change �20.4 �8,900 100.0in petroleum 1990–1998)

(Industry capacity change 49.0 21,400 100.0in gas 1990–1998)

(Industry capacity change 42.1 18,400 100.0in dual 1990–1998)

(Industry capacity change �7.4 �3,200 100.0in nuclear 1990–1998)

Capacity change total �100.0% �6,400 �14.6utilities 1990–1998

(Utilities capacity change �27.2 �1,700 �23.7in coal 1990–1998)

(Utilities capacity change �156.7 �10,000 112.5in Petroleum 1990–1998)

(Utilities capacity change �67.5 �4,300 �20.2in gas 1990–1998)

(Utilities capacity change 138.1 8,800 47.9in dual 1990–1998)

(Utilities capacity change �50.5 �3,200 100.0in nuclear 1990–1998)

Capacity change total 100.0% 28,600 65.4IPP 1990–1998)

(IPP capacity change 22.7 6,500 88.6in coal 1990–1998)

(IPP capacity change in 2.2 600 �7.1petroleum 1990–1998)

(IPP capacity change 35.5 10,100 47.4in gas 1990–1998)

(IPP capacity change 30.2 8,600 46.9in dual 1990–1998)

(IPP capacity change 0.0 0 0.0in nuclear 1990–1998)

(IPP capacity change in other 5.4 1,500 n.arenewable 1990–1998)



Producer type Mix of the Capacity change As % of thechange (in %) (in megawatts) industry change

Capacity change total 100.0% 154,700 100.0%industry 1998–2002

(Industry capacity change 0.3 400 100.0in coal 1998–2002)

(Industry capacity change �1.3 �2,100 100.0in petroleum 1998–2002)

(Industry capacity change 72.9 112,800 100.0in gas 1998–2002)

(Industry capacity change 24.6 38,000 100.0in dual 1998–2002)

(Industry capacity change 0.1 200 100.0in nuclear 1998–2002)

Capacity change total �100.0% �131,500 �85.0utilities 1998–2002

(Utilities capacity change �45.6 �60,000 �15,326.9in coal 1998–2002)

(Utilities capacity change �12.9 �17,000 816.9in petroleum 1998–2002)

(Utilities capacity change 9.3 12,200 10.8in gas 1998–2002)

(Utilities capacity change �18.9 �24,800 �65.2in dual 1998–2002)

(Utilities capacity change �28.4 �37,300 �21,199.2in nuclear 1998–2002)

Capacity change total 100.0% 270,800 175.0%IPP 1998–2002

(IPP capacity change 22.3 60,300 15,398.5in coal 1998–2002)

(IPP capacity change in 5.4 14,600 �703.5petroleum 1998–2002)

(IPP capacity change 32.7 88,600 78.5in gas 1998–2002)

(IPP capacity change 22.5 61,000 160.5in dual 1998–2002)

(IPP capacity change in 13.9 37,500 21,299.2nuclear 1998–2002)

(IPP capacity change in other 1.7 4,600 n.arenewable 1998–2002)

Note: A third producer category (CHP: combined heat & power) is not presented.


In Europe, the technological evolution of the three competitive reformsthat had invested in generating capacity, England and Wales, Spain andItaly, mirrors that of the United States. Overall, over 32 GW of gas-basedgenerating capacity was built in these three countries. This exceeds theirtotal expansion in capacity (27 GW). The dominant technology of newBritish and Italian development is gas, at between 75 and 95 per cent of thetotal capacity change. In Spain, a strong programme of support for renew-able energies (especially wind power) has limited new gas capacity to 43 percent of the total capacity change (Table 3.7).

3. ECONOMIC DETERMINANTS OF THE NEWDOMINANCE OF GAS-BASED TECHNOLOGY

The fact that the fuel mix could change subsequent to a major upheaval likethe competitive reforms was an open question before the reforms. However,given the technological changes seen after liberalisation and privatisationin the airline industry (with the expansion of the ‘hub and spokes’ model)


Table 3.6 Generation capacity changes in California and Texas

State Capacity changes Changes fuel(gigawatts) mix (%)

CaliforniaNuclear 1998–2002 0 0Coal 1998–2002 0 0Oil 1998–2002 0 0Gas 1998–2002 6.3 121Dual fuel 1998–2002 �1.4 �27Hydro 1998–2002 0 0Renewable & others 1998–2002 0.3 6Total capacity change 1998–2002 5.2 100

TexasNuclear 1998–2002 0 0Coal 1998–2002 0.1 0.4Oil 1998–2002 0.3 1.3Gas 1998–2002 22.8 98.7Dual fuel 1998–2002 �1.2 �5.2Hydro 1998–2002 0 0Renewable & others 1998–2002 1.1 4.8Total capacity change 1998–2002 23.1 100

Sources: Various years: EIA-DOE.

and in the telecom industry (with the expansion of digital technology andwireless network) (Bailey et al., 1985; Vickers and Yarrow, 1988); an energytechnology change was not seen as impossible. Notably coal, as a heavilysubsidised fuel in Europe, was seen as going to suffer. None the less, pre-dicting the sequence of changes that would follow the competitive reformswith any precision was rendered difficult because of the wide array of pos-sible inefficiencies under the former regime where cost and technologieswere submitted to public regulation and public energy policy, and resultedin prices being imposed to customers through franchised monopolies(Joskow and Schmalensee, 1983; Beesley and Littlechild, 1994). Severalinefficiencies might have distorted the choice of generation technologiesunder the old regime of regulated production differently (Joskow andSchmalensee, 1983, chs 7 and 12).

On the purely theoretical front, Averch and Johnson demonstrated in1962 that regulated firms could have a ‘rational preference’ for the most


Table 3.7 Generation capacity changes in England and Wales, Spainand Italy

Country Capacity changes As % of the total(gigawatts) capacity change

England & WalesNuclear 1990–2000 1.2 15Coal 1990–2000 �9.6 �124Oil 1990–2000 �4.2 �54Others 1990–2000 �1.8 �23Gas 1990–2002 22.1 286Total capacity change 1990–2000 7.7 100

SpainNuclear 1998–2003 0.2 1.4Coal 1998–2003 0.8 5.6Gas & dual fuel 1998–2003 6.2 42.7Hydro 1998–2003 0.1 0.7Renewable 1998–2003 7.2 49.6Total change 1998–2003 14.5 100

ItalyGas & dual fuel 1998–2002 3.8 73.0Hydro 1998–2002 0.7 13.5Renewable 1998–2002 0.7 13.5Total capacity change 1998–2002 5.2 100

Sources: British Electricity Association (2001); GRTN; REE – various years.

capital-intensive technologies because their ‘Cost ’ based price regulationwas giving a guaranteed ‘rate of return’ (ROR) to their invested capital.While the standard microeconomics was still asserting that the marginalpricing rule was able to rationally frame the management of existingmonopoly (Mishan, 1968; Turvey, 1968; Turvey and Anderson, 1977),further economic literature showed that economic theory has not improvedmuch in the four last decades of the twentieth century in explaining howthe regulated framework influences the electrical utilities in choosing theirgenerating technologies (Pollitt, 1995; Ishii, 2004; Ishii and Yan, 2004). Inpractice, however, many countries had entrenched a fuel mix by promoting‘national’ combustibles (such as coal and lignite) through providing subsi-dies on their prices, or financing a national variant within a given technol-ogy (such as nuclear) (Newbery and Green, 1996). In each of the US states,regulators have also used a variety of means to influence the choice oftheir utilities. In particular, US regulators have the right to set regulatedtariffs ex post (that is, after the facts) by giving or not giving their approvalto the evolution of the cost of generation (for example, oil and gas pricescould vary considerably from one period to another, without being totallypassed through by the regulator) (Joskow and Schmalensee, 1986; Joskow,1989, 2003).

Quite apart from these general distortions, which can logically be laid atthe feet of imperfect regulation, or the regulators’ preferences, or govern-ment energy policy, empirical studies have also discovered a whole universeof hidden inefficiencies of all kinds. These have been concealed as much bydifferences in construction costs as by the range of the operational perfor-mances of the main technologies (nuclear or fossil fuel), even within asingle country, and even when account has been taken of the impact of theage and size of the plants and the precise characteristics of the fuel(Joskow and Rose, 1985; Joskow and Schmalensee, 1987; Joskow, 2002;Wolfram, 2003).

Even within this imprecise framework, it is reasonable to expect thatextending competitive mechanisms and eliminating government fundingwould impose a strong competitiveness constraint on producers, and thuspromote the adoption of those technologies that truly are the mostefficient in generation (Littlechild, 1994; Pollitt, 1995). The spread of gas-based technologies observed in Great Britain as of 1992–93 is thus attrib-utable to the fact that this emerging technology proved to be the cheapestat the point in time when the competitive reforms were coming into theirown (Newbery and Green, 1996; Newbery, 2000; Bower, 2004b). In theUnited States, the same wave, occurring later, marked both a technologyshift and the end of regulatory uncertainty (Hunt, 2002; Ishii, 2004; Ishiiand Yan, 2004).


The Comparative Competitiveness of Gas and Coal

The economic benefits of new gas-based generating technologies have beendemonstrated many times and in various places, usually with the tradi-tional method for comparing electricity generation costs known as levelisedcost methodology. This methodology has been in use for decades and pre-dates the competitive reforms. It was, and continues to be, used as much byinvestors as by institutions and economists. Based on an estimate of thevarious types of costs (capital costs, operating and maintenance costs, fuelcosts) and annual energy generation over the lifespan of a typical plant fora given energy system, it discounts the flows of costs and energy pricesusing reference rates (traditionally, 5 and 10 per cent). The result is a com-parable series of electricity generation cost levels that allow a ‘ranking byeconomic merit’ to be established for choosing between alternative tech-nologies (DGEMP, 1997 and 2003; IEA, 1999, 2002 and 2003a; EIA-DOE,2000 to 2004; RAENE, 2004).

The new competitiveness of gas relative to coal was clearly of great eco-nomic importance in countries such as the United States and Great Britain,where coal dominated the fuel mix (60 per cent coal in British electricitycapacity in April 1990; 44 per cent coal capacity for US utilities in 1990and 1998).5 This competitive edge of gas was simultaneously substantialand enduring, since the most recent study of Britain’s Royal Academyof Engineering in 2004 continued to describe CCGT plants as the least-expensive method available for generating electricity (Figure 3.1).

In the United States, studies by the Department of Energy (EIA-DOE)in 1996 predicted economic dominance for gas that would persist beyondthe forecast horizon (2015) and featured a cost advantage of over 20 percent in the average scenario (approximately $8 per MWh). For the first timein a long while, these same forecasts, conducted in 2004, foresee that agreater competitiveness for coal could emerge between 2010 and 2025 (witha cost advantage of about 2 per cent, or $1.2 per MWh, $0.48 having beenabsorbed by higher network development costs explain this inclusion ofnetwork costs). Confirming this recent change, the February 2005 forecastsforesee one-third of the 2005–25 coming new plants in the United Statesbeing coal fired with about 90 GW new capacity installed and 1,000 TWhof supplementary coal generation in 2025. However, new gas plants are stillconsidered as providing about 60 per cent of the expected 280 GW gener-ation investment for 2005–25 (Table 3.8 and 3.9).

Of course, such studies are based on numerous assumptions, variants andscenarios that cannot be discussed in detail here. The economic dominanceof gas over coal is thus relative, not absolute. There are indeed several regionsin which coal’s competitiveness was never challenged by that of gas. In



Note: PF 5 pulverised fuel; CFB � circulated fluidised bed combustion; IGCC �integrated gasification combined cycle; OGCT � open cycle gas turbine; BFB � bubbling-fluidised-bed combustion.

Source: RAENG (2004).

Figure 3.1 Present-day cost of generating electricity in the UK 2003/04

0

1

2

3

4

5

6

7

8

Coal-fi

redPF

Coal-fi

red

CFB

Coal-fi

red

IGCC

Gas-fir

ed

OCGT

Gas-fir

ed

CCGTNuc

lear

Poultry

-litter

BFBOns

hore

wind fa

rmOffs

hore

wind fa

rmW

ave &

marine

Cos

t of g

ener

atin

g el

ectr

icity

(p/

kWh)

Standbygeneration cost

Cost ofgeneratingelectricity

Table 3.8 1996 forecast costs of producing electricity, 2000 and 2015

2000 2015

Conventional Advanced Conventional Advancedpulverised combined pulverised combined

Item coal cycle coal cycle

1994 mills per kilowatthourCapital 26.41 11.24 26.18 7.00O&M 10.72 4.82 10.72 4.82Fuel 13.58 22.35 7.42 24.38

Total 50.72 39.41 44.32 36.20

Btu per kilowatthourHeat rate 9.840 7.300 8.142 5.687

Source: EIA-DOE (1996).

European studies conducted by investment banks, such as those of the CréditSuisse First Boston, cost calculations are individualised country by countryto better account for local conditions (Figure 3.2). For the same CCGT tech-nology and the same forecast entry into service in 2005, a difference of $5 perMWh (or 13 per cent) is found to exist between Belgium and Italy.


Table 3.9 2004 forecast costs of producing electricity, 2010 and 2025

Costs 2010 2025

Advanced Advanced Advanced Advancedcoal combined coal combined

cycle cycle

2002 mills per kilowatthourCapital 33.77 12.46 33.62 12.33Fixed 4.58 1.36 4.58 1.36Variable 11.69 32.95 11.74 37.91Incremental 3.38 2.89 3.26 2.78

transmission

Total 53.43 49.65 53.20 54.38

Source: EIA-DOE (2004).

Source: Credit Suisse First Boston (2004).

Figure 3.2 CCGT cost of entry by country in Europe in 2005

0

5

10

15

20

25

30

35

40

45

50

CCGT – Belgium CCGT – UK CCGT – Spain CCGT – Germany CCGT – Italy

Cos

t of n

ew e

ntry

(E

uro/

MW

h)

Fuel cost Variable Fixed cash cost Capital costs

The future competitiveness of these technologies will also depend on the‘carbon price’, trading in which has only begun in Europe. In scenariosdeveloped in Oxford by John Bower, the price of a tonne of CO2 is the keyto the relative competitiveness of gas and coal, but also to that of existingpower plants and investment in new British plants. With the price of carbonat 15 €/CO2 ton, old nuclear energy is seen as having a marginal cost in2008–12 smaller than £15/MWh, old CCGT plants around £25, and oldcoal plants around £31 permitting them to temper the interest of investingin new CCGT plants (Bower, 2004a).

The Comparative Competitiveness of Gas and Nuclear Power

Since the beginning of the US and British ‘dash for gas’, many studies usingthe levelised cost methodology have established that nuclear power hasbecome the most expensive base-load thermal technology. The same con-clusion is found in studies by investment banks, such as those from theCrédit Suisse First Boston in 2004 (approximately 50 per cent productioncost gap to the detriment of new investment in nuclear power). Thus in theUnited States, during the past 10 years the DOE’s Annual Energy Outlookhas not foreseen any resumption of investment in nuclear power, regardlessof the timeframe considered (1995–2025). Similarly, in its most recentWorld Energy Investment Outlook (IEA, 2003b) the International EnergyAgency does not foresee any new investment in nuclear power in the UnitedStates or Europe, except in France and Finland, throughout the 30-yearperiod, 2000–30. A MIT study in 2003 set the cost spread of nuclear, gasand coal at 25 to 50 per cent (MIT, 2003).

In France the comparative analysis of the competitiveness of gas versuscoal has made considerably less of an impression than the position of gasrelative to nuclear energy. In a 1997 study by the Ministry of Industry(DGEMP, 1997), new CCGT technology successfully undercut the nucleardomination of the base load by assigning a price of €0.032 per kWh to thelatter (for a series of 10 nuclear reactors) compared to €0.029 for CCGTunder several scenarios (a fall in the price of gas, increased thermal efficiency,lower construction costs). This reversal of technological prospects is quiteabrupt relative to 1993 (only four years earlier), when the median cost advan-tage of French nuclear energy over coal (the only alternative envisaged at thetime) was 30 per cent. However, to date all French studies have consistentlydemonstrated the economic superiority of nuclear technology over thosebased on coal or gas in the median scenarios. The most recent study of ref-erence costs by the Ministry of Industry (DGEMP, 2003), describes a nuclearPressurized Water Reactor (PWR) operating at €28.4 per MWh in 2015(using 2001 prices, or €29.9 at 2004 prices), that is 20 per cent cheaper than


CCGT (€34.5). Since French calculations for gas and coal tend to yieldapproximately the same results as studies from other countries, thisdifference is entirely attributable to how the French ministry analyses thecosts of nuclear power. Several other recent studies (Santaholma, 2003;AREVA, 2004; RAENG, 2004) (See Table 3.10), whether they concur withthe French conclusions or not, have revealed that the discrepancies resultfrom methodological differences and reflect divergences in the understand-ing of the impact of the competitive reforms on the comparative economicadvantages of various generation technologies.

4. THE COST OF CAPITAL AS A KEY ECONOMICDETERMINANT OF TECHNOLOGY CHOICE INA COMPETITIVE FRAMEWORK

It is common practice to rely on the traditional levelised cost methodologyto evaluate the relative competitiveness of alternative technologies for gen-erating electricity. This assumes, at least implicitly, that the costs andbenefits of technologies can be computed with no regard for the context inwhich they are to be implemented (regulated monopoly or competition).This working hypothesis, even when implicit, is surprising. Indeed, theeconomic analysis of regulated monopolies suggests that their generationcosts could have been poorly understood or controlled, especially since therisks inherent in the choice of technology and capacity were not borne bythe producers, but rather by consumers (Joskow, 2000). Conversely, in thisnew competitive framework these risks are borne by the producers them-selves, a priori, and make up an essential element of the constraintsguiding their behaviour (Joskow, 2003). One essential feature of competi-tive electricity regimes is that these new constraints are activated by theabsence of guarantees on the demand addressed to each firm and the levelor evolution of market and input prices. These unconstrained movementsin volumes and prices, which could lead to extreme volatility, introduce a


Table 3.10 Nuclear generation costs in the early twenty-first century(per MWh)

Belgium – Finland France – DGEMP UK – RAE USA – MITAmpere 2000 2001 2003 2004 2003

€ 30 € 24 € 28.4 € 33.8 $ 67

Source: Areva (2004).

new risk for producers. That volatility is more consequential for generatorsthat are not vertically integrated and sell their output mainly or entirely inwholesale markets with neither established customer base nor contractualprice guarantee (like ‘Merchant Plants’, which are IPPs that have not con-tracted for their output). Furthermore, in a competitive electricity gener-ation regime, producers have access to neither government subsidies norgovernment capital (both of which are ‘State Aid’ in the EU competitionframework). Therefore, the ‘cost of capital’ for investments in generationbecomes one of the producers’ key decision variables (MIT, 2003), particu-larly given the high level of capital intensity of the electricity industry(IEA; 2003b).

Costs and Risks Specific to Nuclear Power in a Competitive Regime

The traditional framework for economic analysis of production costs, theso-called levelised cost methodology, encounters a major snag when com-paring very different technologies in the new competitive framework. Torank these technologies by cost, this method assumes that it is easy, or atleast not too difficult, to translate the risk profiles of the various techno-logies into discounted cost levels.

In the traditional methodological framework, a first, simple way ofexpressing the technological component of the risks is to conduct sensitiv-ity analysis. Thus, a sensitivity analysis of nuclear power’s vulnerability tofuel cost variations, as in the 2004 Royal Academy of Engineering study,confirms that nuclear technology insulates the generation process from ran-domness affecting the fuel. The cost of generating gas-fired plants is shownto be sensitive up to 30 per cent to their fuel costs, with extreme gas pricescenarios making nuclear power cheaper than CCGT (RAENG, 2004).

However, since the principal costs and risks of nuclear power lie outsideof fuel costs, it remains to adapt the sensitivity methodology to this tech-nology. The MIT study published in 2003 features a very interesting adap-tation of this methodology to account for the new competitive environmentin the United States that will frame any potential investment in Americannuclear power. The reference cost for nuclear power in the MIT study isvery high ($67 per MWh), compared with €28.4 in the DGEMP’s Frenchstudy published the same year.

None the less, the MIT study indicates how this high reference cost couldbe reduced in the United States by a concerted series of voluntary actions(Table 3.11). Two primary levers for lowering costs are identified. First, thetotal cost of nuclear power could be reduced by $12 per MWh (–18 percent) if construction costs ($2,000 per kW in the reference scenario) couldbe brought down to $1,500 (bringing them below the €1,700 of the Royal


Academy and near the €1,300 level of the DGEMP study for a series of10 reactors – at the exchange rate €1 � $1.2). Second, total costs could alsobe reduced by $9 per MWh (–13 per cent) if the cost of capital in nuclearpower could be brought to the same level as capital costs for gas and coaltechnologies.

The underlying economic reasoning is particularly relevant for ourpurpose. It underlines that private investors run much higher risks whenchoosing nuclear technology over gas or coal. Furthermore, investors cannotmitigate their own nuclear plant construction or management hazards byimproving their knowledge or exploiting any series effect. Also, the compet-itive electricity market is much more unpredictable and volatile than the oldregulated monopoly market. Merchant plants (independent generators) areparticularly dependent on their wholesale market prices exposure. Thus, interms of risk taking, nuclear power runs completely counter to the choice of‘sensible’ private investors. In the MIT study this is true for several reasons,all of which are adding their negative impact on the investors’ decision.

● The real construction costs of a nuclear plant are poorly defined,since they have not been the subject of a controllable experiment fora very long time and were difficult to control in the past.

● The capital intensity of a nuclear plant is three to four times greater($1,500–$2,000/kW) than that of a CCGT plant, and the unit size ofa plant is two to four times higher. Thus, the unit cost of the minimalinvestment could be 10 to 15 times greater. Moreover, if reduced con-struction costs are attained by investing in multiple reactors (often 10units), the minimum size of a nuclear investment programme is 10 to16 GW (depending on the reactor technology), somewhere between$15 and $30 billion.

● The timeframe for building a nuclear plant is inevitably long (at leastfive years in a best-case scenario, versus two years for a CCGT plant),


Table 3.11 Nuclear generation costs in the 2003 MIT study

Total reference With reduced With reduced With reduced With reducedcost ($ / MWh) construction construction O&M cost cost of capital

cost (–25%) time (5 years (down to 13 (down to gas� 4 years) $ / MWh) or coal cost

of capital)

67 $ 55 $ 53 $ 51 $ 42 $��12 $ (�18%) ��2 $ (�3%) ��2 $ (�3%) ��9 $ (�13%)

Source: MIT (2003).

but also uncertain. This makes the timing of entry into the marketconsiderably more random, while the per-unit generating capacity isalso much greater than for a gas-fired plant.

● There is no possible correlation between the evolution of the marketprice of electricity and that of the principal components of the costof nuclear power. This contrasts with the case of gas, for example,while this correlation, though imperfect, is effective (a substantialincrease in the price of gas will have a positive impact on the whole-sale price of electricity).

● Finally, the operational performances of nuclear power plants arealso varied and stochastic. This is particularly true for the capacityfactor, where a minimum factor of 85 per cent is required for financialequilibrium of the nuclear project. In England the five advanced gas-cooled reactors (AGRs) units operated from 1990 to 1995 at a meanload factor of 61 per cent, with the best unit at 78 per cent and theworst at 36 per cent (Mac Kerron, 1996). Another source of ran-domness affects operating and maintenance costs, with a $5 perMWh gap existing between current US mean performance, at $18(fuel included), and its best quartile, at $13.

Each of these specific components of the risk associated with nuclearpower are detrimental to the choice of this technology by an investor. TheMIT study expresses this risk gap with a structure and cost of capital thatdiffers from that of gas-based technology. For a CCGT plant, the Americaninvestor needs only to supply 40 per cent of funding from equity (compen-sated at a rate of 12 per cent) and can borrow the remaining 60 per cent(compensated at 8 per cent). For a nuclear plant, the investor needs tosupply 50 per cent of funding from equity (compensated at a rate of 15 percent) and borrow the remaining 50 per cent (at the same 8 per cent rate).

This analysis of cost of capital specific to nuclear power plants is a majorsource of the cost differences between MIT and other studies (the Belgian,British, Finnish and French shown in Table 3.10). However, a recent Frenchstudy (IGF-CGM, 2004) has shown that the current financial managementof Electricité de France (EdF) now counts on a 13.7 per cent yield to equityin its nuclear generation. This is particularly striking because all existingFrench nuclear assets have been financed in the former framework of aregulated monopoly backed by the financial guarantee of the French state.This suggests that a new nuclear investment which would rely purely on‘market-based’ private financing could not expect a lower return. Then alower average cost of capital cannot come from a lower yield to equity andwill crucially depend on the amount and rate of long-term debt or long-term bonds that the nuclear lender can find in the market. None of the


non-MIT studies distinguish between a nuclear and a non-nuclear invest-ment, between a rate of return on equity and on debt (Table 3.12).

In conclusion, the 2003 MIT study suggests that competitive reformshave a real impact on the choice of generating technology and that theyparticularly weigh in on the choice between nuclear and gas. They primar-ily impact on the cost factor most characteristic of nuclear power – itscapital intensity. Nuclear technology is extremely capital intensive, requir-ing $2,000/kW in construction costs in the United States (apart from thecost of capital and interest over the construction period) and yielding anannual energy volume of 7.4 MWh with median annual sales of approxi-mately $320. The sensitivity of nuclear power to investors’ financial behav-iour is thus not comparable to that of other technologies. If, in addition tothis core characteristic, nuclear technology also features a large number ofuncertainties in terms of costs and yields, the US capital markets cannotdeal with it on the same footing as the other technologies.

However, since the functioning of capital markets is neither unified inter-nationally nor stable over time, the MIT study does not allow any preciseconclusions to be drawn concerning the financing of future nuclear invest-ment in France. The 8 per cent borrowing rate and 15 per cent yield tostockholders used by the MIT study may appear too high given currentfinancial conditions in France. A recent French report already mentioned(IGF-CGM, 2004) places the cost of debt at under 5 per cent (with a goodcredit rating), while the new standard for yield to equity adopted by EdF iscurrently 13.7 per cent. However, there is no more guarantee that EdF willbe able to easily find long-term loans (30 years and more) or the requiredequity. The former period of unlimited financing for nuclear power at thebond-market rate with the guarantee of the French government is a thingof the past. A 50/50 financing split between equity and debt will constitutea challenge for the French nuclear champion, which would need tomobilise about €40 billion in equity to recreate its current nuclear base,


Table 3.12 Nuclear versus gas CCGT cost of capital analysis

Belgium – Finland France – UK – RAE USA – MITampere 2000 2001 DGEMP 2003 2004 2003

All capital at All capital at All capital at All capital at CCGT at5% discount 5% discount 8% discount 7.5% discount 9.6% (40% equityrate rate rate rate at 12 %) (60 %

debt at 8%) Nuclear at 11.5%(50% equity at 15 %)(50% debt at 8%)

after 2017 – the foreseen date of the first plant closure (Glachant andFinon, 2005) – at the current French ministry’s construction costs. Anotherrecent official report to the French minister of the economy, finances andthe industry suggested that the actual amount of equity at EdF is probablynil, or even negative, for the fiscal years 2004 and 2005 (Roulet, 2004).

Comparison of Costs and Risks Specific to Nuclear Power and Gas in aCompetitive Regime

If nuclear generation technology appears so sensitive to the competitivereform, then gas technology is as well, but in a different fashion. The burst-ing of the American electricity ‘bubble’ in 2002–03 revealed that the massiveflow of IPP investment in gas-based generation capacity did not represent asustainable financing or investment regime (Joskow, 2003). Clearly, both theIPPs and their bankers had underestimated the risks of independent gener-ation remunerated exclusively on the wholesale market’s terms and unableto rely on either a portfolio of end clients or a long-term sales contract.

A retroactive computation of the so-called ‘spark spread’ (that is, the gapbetween the wholesale price of electricity and the fuel cost of the gas nec-essary to generate it) reveals the evolution of producers’ ‘net’ revenues andhow the IPPs experienced the floundering of the wholesale electricitymarket under the impact of the new overcapacity. In Texas, CCGT capacityincreased by 23 GW between 1999 and 2002 while the on-peak spark spreaddecreased from $23 to only $6 a MWh (Figure 3.3) to face about $11 capitalcosts and about $5 operating and maintenance (O&M) costs!

More generally, during the quarters following the Enron bankruptcy inthe autumn of 2001, the business model consisting of a merchant plant togenerate electricity and trading platforms to sell it on the wholesale market,which typified the first period of the US competitive reform, collapsed inthe United States. The stock market values of the main pioneers of thereforms (AES, Williams, Calpine, El Paso, Mirant, Reliant, Dynegy and soon) fell 90 to 95 per cent between the spring of 2001 and the spring of 2003,resulting in a $130 billion loss in stock value (excluding Enron), while theircredit ratings fell to between BB– and B–. Securities representing theirnumerous bond issues were negotiated at below face value, by –15 to –75per cent. New flows of bank and bond financing into the US electricitysector fell by 70 per cent – (that is, US$30 billion) between 2001 and 2002(de Luze, 2003). Consequently, many of the IPPs’ assets have ended up inthe hands of their bankers, while any future financing of IPPs that do nothave sales contracts for their output appears to be out of the question.Thus, the business model of the American electricity sector is reverting toone of vertical integration between generation and sales, either in the


traditional form of a utility, or in the surrogate form of an IPP covered bya long-term sales contract (Joskow, 2003; CERA, 2004). Newbery (2000)showed how this ‘vertically integrated’ competitive model differs from apure ‘vertical unbundling’ framework when it comes to investing in gener-ation. The entry of new generators in England and Wales has been favouredin the ‘dash for gas’ by the hedging effect of the long-term contracts (15years) these generators signed with the regional distribution monopolies(the regional electric utilities (RECs)).

None the less, despite the severity of the financial correction that struckthe IPPs in the United States, their stockholders and their bankers, no shiftin technology choice has been induced. Gas-based generation has remainedthe norm. It is not the technology itself, but rather the business model thathas shifted in the United States. The US competitive reforms are evolvingtowards different forms of vertical integration between generation and thefinal sale, with no ‘technological correction’ for the excesses of the gasbubble until the rise of gas prices prompts such a change. The main cor-rection needed after the US ‘gas bubble’ was a capacity adjustment to drythe existing overcapacity (see Chapter 2, this volume).

The constancy of technological choices after a bubble and a shock ofthis magnitude may surprise. Under precisely these conditions, nuclear


Source: de Luze (2003).

Figure 3.3 Spark spread in Texas, 1999–2002

ERCOT On-Peak Spark Spreads

$23.11$22.33

$11.93

$6.15

$0.00

$5.00

$10.00

$15.00

$20.00

$25.00

1999 2000 2001 2002

US$/MWh

technology remains a candidate for a radical rethinking of cost compari-son methodologies that would allow the option value it represents to becaptured. The fact that nuclear costs are not correlated with the standardmarket costs has to be properly valued. Several interesting academic studieshave been working along these lines. Newbery et al. (2004) seek to substi-tute Monte Carlo simulations for the levelised cost methodology tocompare the ‘complete’ risk profiles of alternative technologies (coal, gasand nuclear), and then generate a measure of the option value of nuclearpower as a hedge against higher gas and coal prices. Gollier et al. (2004)propose a measure of the option value of a new, and much more ‘modular’,nuclear technology in which the investor could decide to successively build1, 2, 3 or 4, 300 MW modules and where the investment cost diminisheswith each additional module. But none of these studies calls for any link toan existing or coming investment.

Thus, the current Finish example of a group of electricity distributorsand large industrial clients co-financing a nuclear reactor (to be built by theFrench firm Areva) remains an isolated empirical instance to account for.According to Santaholma (2003), President of the Finnish EnergyIndustries Federation, this investment could represent 10 per cent of thecapacity used during peak hours in Finland. While this project has thestatus of a national project (publicly supported by the government, parlia-ment and labour unions), it is financed by TVO, a cooperative of local util-ities and large industrial consumers. It will resell at cost price a volumeprorated to reflect each participant’s share in the investment. Given thethree assumptions that total equipment costs will be €1,750 per kW (includ-ing all cost of capital fixed at 5 per cent), that reactor operation is expecteda capacity factor of 91 per cent (or 8,000 h annually), and that the includedO&M fuel cost will be €10.2 per MWh, the mean cost per MWh isannounced at €24, versus €32 for a CCGT plant – all values for year 2000(Figure 3.4).

This Finnish example illustrates that, within the framework of a verylong-term contract (40 years) financed by the future energy buyers andtransferring all risks to these buyers along with a very low cost of capital,it is possible to end up with very low values for capital cost and other costs(construction and O&M cost uncertainty being transferred to the energybuyers). In some ways, the cost analysis associated with the nuclear invest-ment in Finland expects to reproduce the conditions of the French nuclearprogramme of the 1970s to 1980s within the framework of the existingEuropean competitive reform.

Today, in France, other paths, more or less similar, are being explored tobind industrial clients to EdF with long-term contracts (IGF-CGM, 2004)or to establish durable links between other electricity concerns in Europe


and France’s nuclear capacity (Glachant and Finon, 2005). It is hoped thatthe dire predictions emanating from the United States regarding the struc-tural handicap impeding nuclear technology in a competitive investmentenvironment can be countered by the vertical integration of generationwith final sales or using long-term contracts, the construction of a largenumber of reactors and plants by very big concerns, the accumulated exper-tise of a world nuclear leader, very favourable credit ratings for this typeof borrower and long-term debt (up to 30 years), as well as considerationof the option value of nuclear power against a possible rise in the price offossil fuels and of carbon emission permits. The actual results of a newFrench nuclear case will be tested as soon as prototypes of the new nuclearreactor EPR are built in France (in the coming years), but large-scaleclosure of existing nuclear plants is not foreseen before 2020 and could bedelayed for more than a decade.

5. CONCLUSION

We have observed that the investment phases of the current competitivereforms in the electricity sector have been accompanied by a strong prefer-ence for gas-based generation technology, in particular for combined-cycle(CCGT) plants: 150 GW of capacity have been built in the United Statesand 32 GW in England, Spain and Italy.


Source: Santaholma (2003).

Figure 3.4 Finnish comparison of generation costs

14.922.8

13.97.6 5.3

10.2 13

40.17.2

7.4

1.5

6.58.2

10

3 17.1

23.7

15.8

18.4

0

10

20

30

40

50

60

Elspot price2000

Finland

Elspot price2001

Finland

Nuclear Coal Gas Peat Wood Wind operationhours=2200 h/s

Eur

o/M

Wh

Fuel costsO&M costsCapital costsElspot prices

Operation hours = 8000 hours/yearGeneration costs without

investmentgrant and electricity tax rebate

R. Tarjanne & K. Luostarinen 12.2.2002

Real interest rate = 5.0 %November 2001 Prices

Many studies attribute the new dominance of this gas technology to itslower cost compared to coal and nuclear technologies. These studiesassume that the electricity reforms only accelerated the recognition of thisgreater efficiency of gas and forced the hands of generators by introducingcompetition and curbing government subsidies or other state aids. Thisanalysis of the reversal of the ranking of total coal and gas technologycosts was broadly accepted during the 1990s and could, in fact, be over-turned again in the United States due to the current gas prices. However,the ‘ranking by economic merit’ between CCGT and nuclear technologiesremains contested in France by the ministry of industry and by Britain’sRoyal Academy of Engineering.

The MIT study has allowed two causes of the divergence in the costanalysis of nuclear to be clearly identified. First, the baseline costs ofnuclear technologies (construction and dismantling costs, O&M costs) andtheir operational performance (in particular the availability of plants andtheir lifespan) remain imprecise and highly variable. Second, characteristicsthat are intrinsic to investment in nuclear (especially: fixed R&D costs,capital intensity per kW, reactor and plant minimal size, construction time,lack of correlation between input costs and the price of electricity) increasethe risks assumed by the investor in a competitive framework. Whilenuclear power has a potential value as a hedge against the current rise offossil fuel prices and of carbon emission costs, the setbacks experienced bythe UK nuclear generator (British Energy, which had to be saved by directstate aid) demonstrate that the commercial survival of nuclear power in acompetitive environment is not assured even when plants are already built.It is therefore understandable that in a competitive environment the cost ofcapital for investments in nuclear power could be driven up relative to thatfor gas or coal, especially when the investment is not to be made in an oldand large vertically integrated utility which has a long record of excellentnuclear performance. All this undercuts the competitiveness of nucleartechnology – which is at least three times more capital intensive than gas.

In the United States, the bursting of the CCGT investment bubble andthe financial crisis confronting independent power producers, as well as thebankruptcy of Enron and the collapse of the ‘merchant plant trading’model, have not succeeded in rehabilitating nuclear technology on a puremarket base. They have, rather, exacerbated the typical nuclear problems byraising the cost of capital to all electricity generation projects. In Europe,however, or at least in Finland and France, nuclear professionals claim tobe able to fully manage the risks and costs specific to nuclear power. Allthey need now is for capital markets to share that analysis. The privatisa-tion of the French nuclear champion EdF could provide an occasion forthese markets to give their first feedback.


NOTES

1. I shall deal separately with the particular cases of peak-load technologies and renewableenergy. Electricity generation is characterised by the coexistence of several technologiesresponding to different parts of the load; notably a specific technology can be used to gen-erate at peak times (like a gas turbine having no ‘combined cycle’ to recycle the heat pro-duced by the ignition of the fuel). Renewable energy is another particularity in technologychoice, since the growth of renewable technologies’ adoption still depends on subsidisa-tion by the public energy policy and is not driven by independent decisions taken by theelectrical companies.

2. Note that in the United States expressions like ‘independent generators’ or ‘independentpower producers’ could refer to really new entrants as well as to new subsidiaries createdby the old utilities under the umbrella of new laws.

3. Until 1995 gas plants in the United States favoured the old gas technology existing beforethe CCGT.

4. The balance was accounted for by a third type of actor: combined heat and power pro-ducers – CHPs.

5. In turn, the French Ministry of Industry has recognised the competitiveness of the newCCGT in the 1997 update of its Reference Costs for Power Generation – representing abreak with the position taken in 1993. CCGT replaced coal as the traditional thermalenergy benchmark. It dominated the mid-base and, in some scenarios, even the base-loadgeneration.

REFERENCES

Areva (2004), Competitiveness and Sustainability: the Respective Trumps of Nuclearand Gas Power Generation, Paris.

Averch, H. and L. Johnson (1962), ‘Behavior of the firm under regulatory con-straints’, American Economic Review, 52, 1052–69.

Bailey, E., D. Graham and Kaplan D. (1985), Deregulating the Airlines, Cambridge,MA: MIT Press.

Beesley, M. and S. Littlechild (1994), ‘Privatization: principles, problems and pri-orities’, in M. Bishop, J. Kay and C. Mayer (eds), Privatization and EconomicPerformance, Oxford: Oxford University Press, pp. 15–31.

Bower, J. (2004a), Energy Trading Solutions in a Carbon Free World, Oxford EnergyInstitute, June.

Bower, J. (2004b), ‘Price impact of horizontal mergers in the British generationmarket’, in D.W. Bunn (ed.), Modelling Prices in Competitive Electricity Markets,Chichester: Wiley, pp. 99–120.

British Electricity Association (2001), Electricity Industry Review, London, January.CERA (2004), North American Electric Power Watch, Cambridge, MA: CERA,

Winter.Crédit Suisse First Boston (CSFB) (2004), Electricity Handbook, Boston: CSFB,

March.de Luze, G. (2003), ‘Investment in power generation: a banker’s perspective’,

IEA/NEA (International Energy Agency and Nuclear Energy Association)Workshop on Power Generation Investment in Liberalised Electricity Markets,Paris, March.

Direction Générale de 1’Energie et des Matières Premières (DGEMP) (1997), Coûtsde référence de la production d’électricité, Ministère de l’Industrie, Paris.


Direction Générale de 1’Energie et des Matières Premières (DGEMP) (2003), Coûtsde référence de la production électrique, Ministère de l’Économie, des Finances etde l’Industrie, Paris.

EIA-DEO (Energy Information Agency and Department of Energy) (1996–2004),Annual Energy Outlook, Paris.

(EU) European Union (1996), Common Rules for the Internal Market in Electricity,Directive 96/92/EC of the European Parliament and the Council of 19 December1996.

Glachant, J.-M. and D. Finon (2005), ‘A competitive fringe in the shadow of a stateowned incumbent: the case of France’, Energy Journal, Special Issue ‘EuropeanElectricity Liberalisation’, 26, 181–204.

Gollier, C., D. Proult, F. Thais and G. Walgenwitz (2004), ‘Comment valoriser la mod-ularité dans les projets d’investissement électronucléaire?’, IDEI & CEA (Institutd’Économie Industrielle, Toulouse, and Centre d’Energiè Atomique) February.

GRTN (Gestore Rete Transmissione Nazionale) (2000–04), Statistical Data andAnnual Reports, Rome, Italy.

Hunt, S. (2002), Making Competition Work in Electricity, Chichester: Wiley.IEA (International Energy Agency) (1999), Electric Power Technology.

Opportunities and Challenges of Competition, Paris.IEA (International Energy Agency) (2002), Security of Supply in Electricity

Markets. Evidence and Policy Issues, Paris.IEA (International Energy Agency) (2003a), Power Generation Investment in Elec-

tricity Markets, Paris.IEA (International Energy Agency) (2003b), World Energy Investment Outlook

2003, Paris.IGF-CGM (Inspection Générale des Finances & Conseil Général des Mines) (2004),

Étude sur la formation des prix de l’électricité dans un marché concurrentiel, Rapportau Ministre de l’Économie, des Finances et de l’Industrie, Paris, September.

Ishii, J. (2004), ‘Technology adoption and regulatory regimes: gas turbine electricitygenerators from 1980 to 2001’, CSEM Working Paper, 128, Center for the Studyof Energy Markets, University of California Energy Institute, Berkeley, CA.

Ishii, J. and J. Yan (2004), ‘Investment under regulatory uncertainty: U.S. electricitygeneration investment since 1996’, CSEM Working Paper, 127, Center for theStudy of Energy Markets, University of California Energy Institute, Berkley, CA.

Joskow, P. (1989), ‘Regulatory failure, regulatory reform and structural change in theelectric power industry’, Brookings Papers on Economic Activity: Microeconomics,125–298.

Joskow, P. (ed.) (2000), Economic Regulation, Cheltenham, UK and Northampton,MA, USA: Edward Elgar.

Joskow, P. (2002), ‘Electricity sector restructuring and competition: a transaction-costperspective’, in E. Brousseau and J.-M. Glachant, The Economics of Contracts.TheoryandApplications,Cambridge,MA:CambridgeUniversityPress,pp. 503–30.

Joskow, P. (2003), ‘The difficult transition to competitive electricity markets in theUS’, Cambridge Working Papers in Economics and CMI Working Paper, 0328,University of Cambridge–MIT Institute.

Joskow, P. and N. Rose (1985), ‘The effects of technological change, experience andenvironmental regulation on the construction costs of coal-burning’, RandJournal of Economics, 16 (1), 1–27.

Joskow, P. and R. Schmalensee (1983), Markets for Power. An Analysis of ElectricUtility Deregulation, Cambridge, MA: MIT Press.


Joskow, P. and R. Schmalensee (1986), ‘Incentive regulation for electric utilities’,Yale Journal on Regulation, 4 (1), Fall, 1–49.

Joskow, P. and R. Schmalensee (1987), ‘The performance of coal-burning electricgenerating units in the United States: 1960–1980’, Journal of AppliedEconometrics, 2 (2), 85–109.

Littlechild, S.C. (1994), ‘Competition, monopoly, and regulation in the electricityindustry’, in M.A. Einhorn (ed.), From Regulation to Competition: New Frontiersin Electricity Markets, Boston, MA: Kluwer, pp. 125–50.

Mac Kerron, G. (1996), ‘Nuclear power under review’, in J. Surrey (ed.), The BritishElectricity Experiment. Privatization: The Record, the Issues, the Lessons,London: Earthscan, pp. 138–63.

Mishan, E. (1968), ‘A survey of welfare economics’, in American EconomicAssociation and Royal Economic Society (eds), Surveys of Economic Theory,Vol. I, London: Macmillan, pp. 154–222.

MIT, (Massachusetts Institute of Technology) (2003), ‘The future of nuclear power.An interdisciplinary MIT study’, MIT, Cambridge, MA.

Newbery, D.M. (2000), Privatization, Restructuring and Regulation of NetworkUtilities, Cambridge, MA: MIT Press.

Newbery, D.M. and R.J. Green (1996), ‘Regulation, public ownership and privati-zation of the English electricity industry’, in R.J. Gilbert and E.P. Kahn (eds),International Comparisons of Electricity Regulation, Cambridge, UK:Cambridge University Press, pp. 25–81.

Newbery, D.M., F. Roques and S. Connors (2004), ‘Nuclear as a hedge against gasand carbon prices uncertainty. Preliminary results’, CMI Research Seminar,Cambridge, UK, 6 November.

NVE, (Norwegian Energy Regulatory Authority) (1999–2004), Statistical Data andAnnual Reports, Oslo.

PJM (Pennsylvania–New Jersey–Maryland) (1999–2004), Statistical Data andAnnual Reports.

Pollitt, M. (1995), Ownership and Performance in Electric Utilities: The InternationalEvidence on Privatization and Efficiency, Oxford: Oxford University Press.

RAENG (Royal Academy of Engineering) (2004), ‘The costs of generatingelectricity’, London, March.

REE (Rede Electrica de Espana) (1999–2004), Statistical Data and Annual Reports,Madrid.

Roulet, M. (2004), ‘Rapport au Ministre de l’Économie, des Finances et del’Industrie sur le projet industriel et financier d’EDF’, Paris, November.

Santaholma, J. (2003), ‘Nuclear power investment, case Finland’, IEA/NEA(International Energy Agency and Nuclear Energy Association) Workshop onPower Generation Investment in Liberalised Electricity Markets, Paris, March.

Statistics Norway (1996–2004), Statistical Data, Oslo.Turvey, R. (1968), Public Enterprise. Selected Readings, Harmondsworth: Penguin

Books.Turvey, R. and D. Anderson (1977), Electricity Economics, Baltimore, MD: Johns

Hopkins University Press.Vickers, J. and G. Yarrow (1988), Privatization: An Economic Analysis, Cambridge,

MA: MIT Press.Wolfram, C. (2003), ‘How might restructuring affect the efficiency of electricity gen-

eration in the US?’, CSEM Working Paper, 111, Center for the Study of EnergyMarkets, University of California Energy Institute, Berkeley, CA.


PART II

Investment in transmission

4. Problems of transmission investmentin a deregulated power marketSteven Stoft

1. INTRODUCTION

From the earliest days of commercial power production, transmission hasgrown steadily in importance. New wholesale power markets have sparkedinterest in distributed generation, but trade between these markets has onlyincreased the need for transmission investment. As with generation, amarket for the use of existing assets is not difficult to imagine, but a marketto supply these assets is more problematic. In fact, the discrepanciesbetween the properties of transmission costs and benefits and the assump-tions of competitive economic theory are so substantial that a market solu-tion is probably not desirable. Even incentive regulation may prove sodifficult to design and so inaccurate that a planning solution may be prefer-able, at least until wholesale power markets are functioning efficiently andthe generation–investment problem has been solved.

Transmission, an exceptionally heterogeneous product, can be both asubstitute for and complement to generation, and suffers from returns toscale and lumpiness,1 as well as major externalities, both positive and neg-ative. This chapter investigates the extent to which these problems can beovercome. To simplify, this chapter ignores transmission losses becausethey play a relatively small role in the investment problem and one similarto the role of congestion, which is considered. After introducing someproperties of congestion prices and transmission costs, three basicapproaches to transmission investment are explored.*

Three Approaches to Investment

Three approaches to investment stand out as relatively distinct, althoughmany mixtures of these are possible. A planning approach refers to a systemthat does not include any incentives specifically tailored to the long-runtransmission investment problem. Such an approach would be carried out

87

* Definitions and notations are given in the glossary at the end of the chapter.

by a group of engineers and economists under instructions to build anefficient system. In practice, it would be backed by rate-of-return regulationwith a requirement that investments be ‘used and useful’. A merchantapproach would allow any private company to modify the transmissionsystem, subject to certain restrictions, and would reward (or punish) suchmodifications by allocating transmission rights to investors. A performance-based regulation (PBR) approach would induce investment by a for-profitowner of the transmission system (a transco) by adjusting its profit level onthe basis of the cost and performance of the system.

Congestion: The Opportunity Cost of Using the Grid

All three approaches will be assumed to exist within the framework of awholesale power market based on publicly known nodal prices (Hogan,1992; Harvey et al., 1996). That is, at each relevant point (node) in thenetwork, a price is established, and these prices together clear the market.They may be purely competitive prices or they may be distorted by marketpower, but in any case there is one price at each node and all energy trans-actions at a given node take place at that price. These prices are adjustedeach time there is a change in supply or demand.

Such a pricing system automatically prices the use of the transmissionsystem even though it applies directly only to energy transactions. If theprice at node A is $20/MWh and at node B is $30/MWh, then the price totransmit energy from A to B is $10/MWh, while the price to transmit it fromB to A is negative $10/MWh. Although nodal prices have some peculiarproperties, it is important to understand that they are simply the result ofthe normal forces of supply and demand constrained by the physical limitsof the transmission system. When these limits restrict flow, the system is saidto be congested. When supply and demand are both competitive, nodalprices are simply the standard competitive market prices and have all theproperties expected of such prices. Except when prices are determinedsomewhat arbitrarily because a vertical supply coincides with a verticaldemand curve, competitive nodal prices are unique. Although they are oftencalculated from bids in a centralized auction, they are not the product of anyspecial rules of calculation but are the prices at which a well-arbitragedbilateral market would arrive if the transmission constraints were enforced.

Three distinct costs associated with congestion are often confused, con-gestion rent (CR), congestion cost (CE), and the cost of congestion to load(CL). Economists focus on the first two, while consumers react to the third.Consider a load pocket with 1,000 MW of load and a 300 MW line into theload pocket from a large system that could supply 800 MW at $20/MWh andmuch more at $40/MWh. These costs are represented as a ‘remote supply’

88 Investment in transmission

function for the load pocket in question. Suppose that local load is fixed at1,000 MW and that the pocket contains 600 MW of $30/MW generationand 200 MW of $50/MWh generation. The ‘net local demand’ curve showsthe demand for imported power net of what would be purchased locally(Figure 4.1). For example, at a local price of $40/MWh, the load pocketwould consume 1,000 MW and supply 600 MW, leaving it with a netdemand of 400 MW. In other words, net local demand accounts for localsupply as well as local demand, and it is local supply that provides the pricesensitivity of this ‘demand curve’ and not actual demand responsiveness.

The congestion cost is also called the, ‘redispatch cost’ because it is theextra cost of dispatching more expensive generators than would be neededif the transmission system had ample capacity and did not constrain powertransfers. In the present example, 100 MW of local $50 generation and400 MW of local $30 generation must be used in place of 500 MW of remote$20 generation that would have been used had there been no transmissionconstraint. Congestion cost is a deadweight loss, not a transfer payment.

Congestion rent is the amount collected by the owners of the rights to thetransmission line. In a one-line network these rights would typically pay theowners an amount equal to the line’s capacity times the difference betweenthe prices at the two ends of the line. In the case of a load pocket, this is thedifference between the internal price and the external price. Congestion rentis a transfer payment from line user to line owner, as using the line has noactual cost.

Problems of transmission investment in a deregulated power market 89

Figure 4.1 Defining congestion rent and congestion cost

200 400 600 800 1,000 MW

Transmission constraint

Remote supply

$50

$30

$20Net

local demand

Congestion rent (CR)

Congestion cost (CE)

Finally, there is the cost of congestion to load (CL) (see Figure 4.2). Withample transmission, the load in the pocket would import 800 MW of powerand use 200 MW from internal generators. The price would be set by theintersection of supply and demand at $30/MWh, so the total cost of powerto load would be $30/MWh�1,000 MW, or $30,000/h. Because of the con-gestion, the price in the pocket is $50/MW and so load must pay $50/MWh �1,000 MW, or $50,000/h, which makes the cost of congestion to load$20,000/h. The three costs are shown in Table 4.1.

As can be seen there is no particular relationship between these threeconcepts. Frequently consumers find it unfair that congestion can costthem far more than the ‘congestion cost’. It does not help that it can alsocost them more than congestion cost and rent combined with the excessrevenue accruing to generators that seem to benefit from the constraintwithout reason. Although the matter is beyond the scope of this chapter, itshould be noted that if the transmission and generation markets satisfy theaxioms of perfect competition, nodal prices will just cover the long-runcosts of the efficient mix of generation and transmission. In other wordsthere is nothing inherently unfair or inefficient about the distribution ofrevenues under nodal pricing in a congested system. Of course this does notindicate that either transmission or generation is, or can be supplied by, acompetitive market, only that any problems with these costs and pricesarise from non-competitive features of the markets and not simply from themethod of nodal pricing or the effects of transmission congestion on prices.


Table 4.1 Three views of congestion

Congestion (redispatch) cost CE $7,000/hCongestion rent CR $9,000/hCost of congestion to load CL $20,000/h

Figure 4.2 Cost to consumers compared with congestion cost and rent

$/MW

MWh

RentCost

Constraint

Demand

Supply

$/MW

MWh

Cost toconsumers

Constraint

Supply

The Zero-congestion Fallacy

Because transmission congestion imposes costs, one recurrent view holdsthat it should simply be eliminated. This is now the policy of the Alberta gov-ernment (see Box 4.1). Although examples can be manufactured for whichthis is the least-cost solution, in real power markets such situations neverexist. If there is one hour per year in which a remote generator is $10/MWhcheaper than the most expensive local generator in use, and if 1 MW of thatgenerator’s output cannot reach local load, then the line is constrained duringthat hour and the cost of the constraint is $10/year. Adding 1 MW of capac-ity to that transmission path would cost far more than $10/year. Eliminatingall congestion – allowing every last megawatt of trade – is simply not efficient.When transportation is expensive, it is often cheaper to consume the localproduct than to transport a slightly cheaper product from a distance.

BOX 4.1 ALBERTA REGULATION #174/2004,ELECTRIC UTILITIES ACT

Alberta’s Electric Utilities Act took effect on January 1, 1996. Fromthe beginning, the Electric Utilities Act has imposed uniform prices(forbidden competitive prices) throughout the province. With itsnew requirement (shown below) to overbuild the grid, uniformprices will become competitive prices.

From Section 8(1) of the Electric Utilities Act:

(e) taking into consideration the characteristics and expectedavailability of generating units, plan a transmission systemthat(i) is sufficiently robust to allow for transmission of 100 per

cent of anticipated in-merit electric energy referred to insection 17(c) of the Act when all transmission facilitiesare in service, and

(ii) is adequate to allow for transmission, on an annualbasis, of at least 95 per cent of all anticipated in meritelectric energy referred to in section 17(c) of the Actwhen operating under abnormal operating conditions;

(f) make arrangements for the expansion or enhancement ofthe transmission system so that, under normal operatingconditions, all anticipated in merit electric energy referred toin clause (e)(i) and (ii) can be dispatched without constraint.


The fallacy of eliminating all congestion may arise from confusionbetween congestion and unreliability. Unreliability is the result ofhaving too little local generation to meet local demand net of imports.This can result from congestion, but in almost all cases, congestion isnot associated with unreliability. It is simply the result of having morethan enough local generation, but at a cost higher than the cost of remotegeneration that could be accessed with a larger transmission line (seeFigure 4.3).

In other cases the desire to eliminate congestion may result from a desireto increase local supply and thereby lower the local market price. For sometime, this can save consumers money even if it raises the long-run cost ofproducing and delivering power. But if expanding transmission lowers con-sumer costs without lowering total cost, it is a form of monopsony powerwhich essentially expropriates some of the sunk costs of generation in theimport-constrained area.

Suppose that generation fixed costs in an import-constrained regionare $40,000/MW-year higher than in the surrounding area, and there are1,000 MW of such generation installed in the load pocket. A 200 MWexpansion of the import line may virtually eliminate extra fixed (sunk) costrecovery needed by reducing the local price to the external price. This maysave consumers $200,000/year per MW of new line. This is probably muchmore than the cost of the line. This appears to be a saving, largely becauseit expropriates $40 million per year in generation fixed costs. It also pro-vides a legitimate saving by providing genuinely less-expensive externalpower to the expensive constrained zone.

If capacity produces power in half of all hours and if the new line’scapacity is half used, then each MW of line provides a real saving of$40,000/MW, one-fifth as much as the initial saving through the expropri-


Figure 4.3 Relationship of congestion to a transmission-cause reliabilityproblem

$/MW

MWh

Constraint Netlocaldemand

Remotesupply

$/MW

MWh

Constraint

Remotesupply

Loss of load

Net localdemand

Congestion

ation of generation sunk costs. In the long run, internal capacity will retireand eventually the generating capacity in the load pocket will againrecover its fixed costs. Perhaps the savings from the line will amount to$240,000/MW for the first five years and $40,000/MW after that. If this hasa present value of $1,250,000/MW, then it might be thought that this is thebreak-even point for building such a line, but that omits the impact of regu-latory risk.

Investors take account of the cost of expropriations by factoring thatpossibility into future investment decisions. This demonstration effect willnot be confined to the load pocket in question. It is impossible to predictwhat the cost of this will be, but the resulting investment risk premiums willaffect the rate of return on all capital in the affected load pockets, not justthe new investment. This is a consequence of a market-clearing electricityprice. Probably the most sensible guess is that all of the money transferredto load from existing generators will be lost to load through higher riskpremiums.

In conclusion, it is inefficient to eliminate all congestion, and it is wrongto focus on short-run consumer cost reductions when planning transmis-sion investments. As with any type of productive investment, the goalshould be to minimize the total cost of production and delivery. If themarket is competitive or the regulation effective, these cost savings will bepassed through to consumers.

Optimal Transmission (Static)

The optimal amount of transmission minimizes the total cost of produc-ing and delivering electricity (Box 4.2). It can be determined in simpleexamples by using the standard first-order condition of calculus. At theoptimum, the value of an additional kW of transmission equals the cost ofbuilding it. Because realistic transmission cost functions can be quitecomplex, the optimal design may need to be found by evaluating manyoptions rather than applying calculus.

BOX 4.2 OPTIMAL TRANSMISSION

● Transmission investment should not eliminate congestion.● Transmission investment should not minimize short-run con-

sumer costs.● Transmission investment should minimize the total cost of

transmission and the production cost of power.


A simple example will illustrate the main points. Consider a city (loadpocket) that can produce power at a cost of $50/MWh but can buy it for$30/MWh over a transmission line. Suppose the line can be built for a rentalcost of $6,000/h plus $5/MWh. How large a line should be built? (See Box4.3 on defining the rental cost of transmission lines.)

To solve the above problem, the city’s load must be specified. Suppose thepeak load is 800 MW and its minimum load is 400 MW and it takes onintermediate values according to a uniform probability distribution. Thesavings from the line will be $20/MWh for the first 400 MW of line capacityand will then decline linearly to zero for additional capacity up to 800 MW.


BOX 4.3 DEFINING THE RENTAL COST OFTRANSMISSION LINES IN $/HOUR

The cost of a particular transmission line might be $500,000,000,but when analysing power systems it is much more convenient tothink of renting capital than buying it outright. Rental cost is natu-rally expressed in $/MWh, and these units are particularly conve-nient for computation. This is not how engineers calculate costs,but it is very convenient for economic calculations.

Given the one-time cost of the line, its capacity, and a savingsper MWh of energy transported, it is generally impossible to tellwhether building the line saves money or not. This is because onemust know how long the line will last, its maintenance costs, andthe debt and equity costs associated with the project. All of theseare properly taken into account by a rental cost, and so can beignored once the rental cost is specified.

The one-time cost can be converted to an amortized annualcost, of say, $50,000,000/year. Adding maintenance of say2,000,000 year and dividing by 8,760 hours/year gives a rentalcost of $5,936/h. If the line is a 1,000 MW line, then the cost is$5.94/MWh. This is the cost of renting 1 MW of the line for 1 hourand is independent of whether the line is used or not.

Simple investment problems often involve a transmission costfunction such as C�cbQ, where Q is the MW capacity of theline. In this case it will be convenient to specify C, the line’s rentalcost, in $/h and c in $/h, and b in $/MWh. Sometimes c will becalled the fixed cost of the line meaning that it does not depend onthe choice of capacity.

As long as additional capacity saves more in energy production costs thanthe cost of the additional capacity, the line should be larger. Since the costof additional capacity is $5/MWh, the line should be expanded until thesaving falls to $5/MWh on average. When the capacity is three-quarters ofthe way from 400 MW to 800 MW, load will be great enough to use the lastMW of capacity only a quarter of the time. Thus the savings of the lastMW will be only one-quarter of $20/MWh or $5/MWh. Hence 700 MW isthe optimal capacity of the line.

This conclusion is actually a bit premature. It is necessary to check thatthe net benefit from the line is positive when the fixed cost of the lineis included. The total benefit is $(20�40012.5�300)/h, which is$10,750/h, while the total cost is $(6,0005�700)/h, which is $9,500/h.The line is worth building, but if the fixed costs had been $8,000/h, it wouldnot have been.

This example is the basis of the conclusion that eliminating congestionis almost never the right answer. All transmission lines are used to varyingdegrees at different times of the day and year. Their maximum potential usewill occur for only a few hours. To eliminate congestion it is necessary tobuild enough capacity to accommodate this maximum, but that means thelast megawatt of capacity is used only a tiny fraction of the time and it isalmost never economical to build capacity for such infrequent use. It mightbe argued that the lumpiness of transmission investment will naturallycause a choice between a much-too-small line and a too-large line, and thatthe economic choice will turn out to be the too-large line and no conges-tion. Besides the fact that this is unlikely to happen for all lines, there is adeeper problem. As load grows, every line reaches a point where its capacitywould be exceeded without redispatch for just a few hours per year. Toavoid this, a new line, or at least an expansion will be needed, and con-sumers will have to pay for it. Due to the fixed costs this will come to at least$6,000/h in every hour of the year in the above example.

Turn next to the general static optimization problem. This asks whatcapacity line should be built given the way congestion changes with linecapacity. A small line will be congested frequently and congestion rents willbe high, while a large-enough line will never be congested. What is the con-dition for the optimally sized line?

For simplicity assume that the line connects a local region where the totalcost of energy production is CL(Q) to a remote region where it is CR(Q),and assume that power is cheaper in the remote regions so the flow is intothe local region. A 1 kW expansion of the line will increase production inthe remote region and decrease it in the local region by 1 kW. If the localmarginal cost of power is MCL and the remote marginal cost, MCR, thenthe savings is the difference. In a competitive market, prices equal marginal


costs so the savings is (PL – PR) times 1 kW. The line is worth expanding upto the point where the marginal cost of expansion is equal to the averageprice differential.

Static optimal transmission result (one line) In an optimal one-linenetwork, the marginal cost of line expansion equals the all-hour averageabsolute nodal price differential between the ends of the line. In otherwords, the marginal cost of expansion equals the average congestion rent.

In a network, things are a bit more complex. If we have two lines from A toB and power tends to flow equally on each, but A has much less capacitythan B, when line A becomes congested it will limit the flow on line B. Thislimitation is not a physical restriction, rather the system operator will beforced to limit the total flow on both lines to protect the weaker line. If lineA is expanded by 1 MW, this will increase the usefulness of line B by 1 MW.Consequently the value of expanding A is twice what would be computedfrom the flow on A times the congestion price of A.

Static optimal transmission result (network) Let d be the power distri-bution factor on line A–B calculated as the fraction of power flowing onthat line when power is injected at A and withdrawn at B. Then, in anoptimal network, the marginal cost of expanding the constrained lineequals the average congestion rent divided by d.

Optimal Transmission (Dynamic)

Power systems are dynamic. Load grows; generators are built andare retired. This dramatically increases the complexity of the optimaltransmission–investment problem. To illustrate this, consider the simplestexample of a dynamic investment problem. Suppose transmission linescome in two sizes, 600 MW at a rental cost of $5/MWh and 1,000 MW at acost of $4.20/MWh. Suppose that the power transmitted over the path inquestion starts at zero and grows by 100 MW per year indefinitely and thatthe saving from transmitting this power is always $6/MWh. Obviously it willbe economical to build power lines.

Building the 600-MW line would cost $3,000/h, and when load hasgrown to 500 MW, this would save $3,000/h. This is the break-even point,and if the small line is to be built it should be built at this point in time. Thealternative is to wait another two years until load has grown to 700 MW, atwhich point the cost and savings of the 1,000-MW line would be $4,200/h.The smaller lines start saving money sooner, but larger lines will save$1.80/MWh in the long run, while smaller lines will save only $1.00 in the


long run. The choice between the two strategies depends on the discountrate, but as can be seen from Figure 4.4, it does not require a very low dis-count rate to make the larger lines the more economic choice.

The conclusion to draw from this dynamic example is that, just becausea line is worth building, it should not necessarily be built. It may be betterto wait a while and build a larger and more economical project. This is aresult of the lumpiness of the investment decision. Note that waiting has nooption value in this example because the future is known with certainty.

Optimal Transmission (Option Value)

In the above dynamic example, two sets of projects were considered; (i)build small lines at optimal intervals, and (ii) build large lines at optimalintervals. Even if there had been uncertainty in this example, each set ofprojects could have been evaluated to find its expected net present cost oftransmission and generation. If all reasonable sets of projects are evalu-ated in this manner and the one with the lowest cost is chosen, this consti-tutes a complete dynamic analysis. Unfortunately the selected projectmay not be the best choice. This is not because expected net present cost is


Figure 4.4 A positive present value is not sufficient

4 8 12 16 20 24 28 32

$600/h

$1,200/h

$1,800/h

$2,400/h

$3,000/h

$3,600/h

Years

Savings from building600-MW lines

Savings from building1,000-MW lines

the wrong criterion, but because the set of choices was unnecessarilyrestricted.

Besides sets of projects, there are also investment strategies. One suchstrategy might be to build a small line now and then wait until loadincreased by 600 MW and build another small line if the wait was eightyears or more and build a large line if it was eight years or less. A strategyis different from a specific ‘set of projects’ because it waits for more infor-mation and then chooses one project or another. Quite often, some strat-egy will be more cost effective than any specific plan.

For example, consider a system with no transmission, one city and a pos-sible remote coal plant with a 50 per cent chance of being built. If it is built,the most efficient transmission project would start now and have a netpresent value of $200 million. If the coal plant is not built, this projectwould have a net present value of minus $100 million. Because there is a 50per cent chance that the coal plant will be built, building the line would havean expected net present value of $50 million ((200 – 100)/2).

If the line were started a year later it would have a (current) net presentvalue $180 or minus $90 million depending on whether the coal plant is oris not built. Waiting a year and then building the line thus has an expectednet present value of $45. Other projects could be considered that delay theline for different amounts of time or build lines of different capacity. But itis quite plausible, and this example will assume, that of all specific projects,building the line now is most valuable. In spite of this, there may be a morebeneficial strategy for selecting projects.

If in one year, we shall know whether or not the coal plant will be built,the strategy of waiting a year and then selecting the best project is morevaluable than any particular project. The expected present value of thisstrategy is 50 per cent of $180 million (if the coal plant is built) plus 50 percent of $0, if the coal plant is not built. This strategy has an expectedpresent value of $90 million which is $40 million greater than the value ofbuilding the transmission line now, the most valuable project given today’sinformation. This $40 million is called the option value of waiting a yearto decide.

There is often a cost of delaying the start of projects with a positiveexpected net present value, but there is also generally an option value todelaying the decision to go forward with a project. It is only when this costof delaying is greater than the option value of waiting to decide that acommitment should be made. Option value is difficult to estimate and itshould be estimated for various waiting periods. In short, consideringoption value greatly expands the number of possibilities that must beconsidered.


2. COST RECOVERY FOR OPTIMALTRANSMISSION INVESTMENTS

The previous section demonstrated that the optimal grid will suffer conges-tion which will result in the collection of congestion rents and that these rentsare related to the cost of the grid. This raises the question of to what extentcongestion rents on the optimal network will cover the cost of that network.To consider this question it is useful to expand the notion of congestionrent. The congestion rent collected on a one-line network is CR(LAB)�Q (PB – PA), where Q is the power flow from A to B on the line between Aand B. This is the trading surplus collected if Q is sold at A and purchasedat B. Expanding this concept to the entire grid results in defining the con-gestion rent for the grid to be the revenue from selling all energy injections attheir nodal prices and purchasing all energy withdrawals at their nodalprices. If Wi is the net energy withdrawal at node i and Pi is the price at node i,then the congestion rent for the network G is CR(G)��Wi Pi.

Considering only lossless networks, it is possible to decompose the set ofnet energy withdrawals into a set of bilateral trades each with one injection(negative withdrawal) and one withdrawal of equal magnitude. Each bilateraltrade from node A to node B, which can be any two nodes on the network,has associated with it a congestion rent, CR(BAB)�Q (PB – PA), where Q isnow the magnitude of the injection and withdrawal of bilateral trade B.

Note that one possible decomposition of the net nodal energy with-drawals into bilateral trades corresponds exactly to the power flows on theindividual lines. Doing so associates with each line a congestion rent equalto the line’s flow times the price at the withdrawal node minus the price atthe injection node.

The lossless congestion-revenue result If the set of all bilateral trades, B,sums to the total net energy withdrawals from the network, then the totalcongestion rent is the same whether computed by node for the entire grid,G, as the sum of congestion rents on all lines, L, or as the sum of con-gestion rents on all trades.

Proof: Since the injections of bilateral trades are paid the nodal priceand withdrawals are charged the nodal price, the net revenue collected ata node is the nodal price times the sum of withdrawals minus the sum ofinjections. Since these two sums add up to the net withdrawal, the netrevenue collected from bilateral trades at node i is just Wi Pi, and overall nodes bilateral trading revenues sum to the congestion rent calculated

CR(G) � �All Lines

CR(L) � �All Trades

CR(B).


for the entire network. That congestion rents on lines sum to the samevalue depends on these power flows summing to the net withdrawals onthe network. This follows from conservation of energy in a losslessnetwork. The net power flows on lines into a node must sum to the netwithdrawal from the grid at that node.

This result demonstrates that the trading revenue that is collected frombuying and selling power in a congested network with nodal prices (even ifthese prices are not the competitive prices) will exactly cover the congestedrents calculated on a line-by-line basis. This assumes that there is no powerloss, so that the power that flows out of one end of a line equals the powerthat flowed into the other end. (In reality losses are typically well under 5per cent on a high voltage transmission system.)

It would be desirable if the trading surplus in an optimally built networkwere to cover the cost of the transmission network. The above result makesit reasonable to investigate this question by looking at the cost-recoveryproperties of congestion rent from a single line.

Congestion Rents Recover Linear Line Costs

Suppose the cost of a transmission line is strictly proportional to themegawatt capacity of the line, so that cost is given by C�c Q, where c isin $/MWh and Q is in MW. In this case, according to the static optimaltransmission result (one-line) given above, the marginal cost of the line, c,should equal the average congestion rent per MWh, P. As a result, therevenue from the line, PQ, will equal the cost of the line c Q. In the optimalone-line (lossless) network with linear costs, congestion rents will cover thecost of transmission capacity exactly.

Not surprisingly this result extends to networks. If the power distribu-tion factor on the Q-MW line from A to B is d, then when line A–B is con-gested, it controls the power flow of Q/d. If the congestion price from A toB is P, then the congestion rent associated with this constraint is P Q/dand the benefit of increasing the line’s capacity is P �Q/d. In this way thecost recovery of constraints in a network can be seen to be analogous to thecost recovery of a single line in a one-line network. An optimally con-structed network with linear cost functions and no losses, will recover itsfixed costs through marginal-cost (competitive) pricing.

Congestion Rents and ‘Fixed’-cost Recovery

When investing in transmission, some costs are roughly independent of thecapacity of the line and some are roughly proportional. Those that are inde-


pendent will be called ‘fixed’ costs in this chapter. The linear-cost functionjust considered has no fixed-cost component, so to add realism, consider oneof the form C�fc Q. In this case, the same first-order conditions willdetermine the optimal transmission system, so congestion rents will recoverc Q, but not f. In other words all costs that are proportional to capacity willbe recovered, but none of those that are independent of the line’s capacitywill be recovered from congestion rent in an optimally sized system.

The cost function just considered is one example of returns to scale.Another model of returns to scale has a cost of line capacity that is pro-portional to the square root of capacity, C�a K1/2. In this case the mar-ginal cost of capacity is MC�(a/2) K –1/2. If the line is built to the sociallyoptimal level, MC will equal the average congestion rent, which is paid onthe whole line capacity, so revenue is R�(a/2) K1/2�C/2. In other words,at the socially optimal level of investment the line provides a congestionrent of exactly half of what it costs. This is true regardless of how large orsmall the optimal line is.

Lumpy technology also exhibits fixed costs and, at least, limited returnsto scale. But the two concepts, returns to scale and lumpy technology, canbe usefully distinguished. Figure 4.5 provides the basic intuition.

Both types of technology violate the ‘non-convexity’ assumptionrequired for perfect competition. Both types have a fixed-cost component.The lumpy technology shown above requires a fixed cost to provide its firstmegawatt of capacity, but the marginal cost of capacity is then zero up to


Figure 4.5 Lumpy technology may not exhibit returns to scale in thelong run

Totalcost

Quantity

Returns toscale

‘Lumpy’technology

100 MW

100 MW. In this sense, the cost of this technology is all fixed. On a smallscale, up to 100 MW, there are returns to scale; over large changes in cap-acity, there are no returns to scale.

3. TRANSMISSION PLANNING WITH STRATEGICGENERATION INVESTMENT

Under a planning approach, no performance-based incentive mechanismsare applied to the problem of deciding on long-lived transmission invest-ment, mainly lines and transformers. The planning might be carried out bythe independent system operator (ISO) or by a transco, and payments forthe cost of investment will be carried out under rate-of-return regulation.To simplify discussion, it will be assumed that the planning is done by atransco that owns and operates the grid, but that there is an ISO whichoperates the wholesale energy market. Because PBR is not used for long-run investments does not mean it would not be used for the day-to-dayoperation of the grid, but the problems of efficient grid operation will notbe discussed in this chapter (Joskow, 2004).

As illustrated in the previous section, planning transmission investmentsis a complex problem, and is in fact far more complex than these illustra-tions suggest because of the complexity of the network. This is widely rec-ognized. An additional complication is also recognized. Because generationis not planned, transmission planners facing a wholesale power market mustforecast the location of generation and load many years in advance.2

These problems are common to all three approaches mentioned in theintroduction (planning, merchant and PBR). Under each approach, thedynamics of the combined transmission/generation system and the optionvalue of waiting must be taken into account. Under each approach theinvestor will not have control or direct knowledge of future generationinvestment, and under each, the complex cost structure and network exter-nalities cause additional difficulties. This section does not focus on thecommon problems but on strategic issues peculiar to the planning processwhen generation investment is deregulated.

Strategic Manipulation of the Zero-congestion Policy

The possibility of strategic manipulation is particularly acute in the case ofa zero-congestion policy, such as Alberta’s. In the example for static trans-mission optimization, the optimal line size was 700 MW, and if an 800 MWline had been built there would never have been any congestion. This makesit appear that building for zero congestion would not be too expensive. In


this case it would require only a 14 per cent increase in transmission capac-ity. But consider the case of a wind farm that can be located at various dis-tances from load. If it is close to load, the cost of transmission may be only$1/MWh while if it is in the most remote location it might be $20/MWh.Suppose that the most remote location is the windiest and it is no moreexpensive to build wind generators there than closer in. What is the effectof a zero congestion policy on the location of this wind farm?

Clearly, it is most profitable for wind generators to locate in the mostremote location whether or not the benefit of more wind comes close tooffsetting the extra cost of transmission. Moreover, since wind power hasnearly zero marginal cost, it is always ‘in merit’, and consequently trans-mission capacity must be built to accommodate the windiest hour of theyear. The last megawatt of such transmission capacity will have almost novalue. (One can be sure that a zero-congestion regulation will be violated inpractice simply because adding capacity to a remote location to capture onehour of supply makes so little sense.) This illustrates the fundamental pointof the strategic generation investment problem.

Strategic generation investment problem The policy of the transmissionplanners influences the distribution of installed generation. Hence, plan-ning policies cannot be appropriately compared on a static model of gen-eration and transmission but must be compared using a game-theoreticapproach that recognizes generation investment strategies.

This problem complicates transmission planning in a market environmentbecause the problems of predicting generation investment and planningtransmission for that investment can no longer be separated. For example,one might think that planning could proceed by having two groups of plan-ners, one of which forecasts generation investment, and the other of whichtakes the forecast and plans the ‘best power lines’ for that predicted invest-ment. The only communication necessary between the two groups would bethe transfer of forecasts from the first group to the second. Unfortunately,the strategic generation investment problem shows that this will not work.

First, the strategic investment problem tells us that the forecaster will needto know what policy the transmission planners will follow. This is necessarybecause generation investors will react to this policy and the forecasters needto understand this reaction to make good forecasts. Second, the plannerscannot simply build the best lines for the predicted generation. They mustchoose an investment policy that induces good generation investment, sothey must understand how generation investment will respond to theirpolicy.


The Difficulty of Implementing Simultaneous Optimization

There is one obvious hope for resolving this chicken-and-egg problem. If theplanners choose the ‘truly best’ transmission investment policy – call this theideal transmission planning policy – this rule should induce generators tomake the best possible investment decisions. In this case, transmissioninvestment policy should be optimal for the implied generation investmentincentives, and these should be optimal for the transmission policy. But thisdoubly optimal system deserves a closer look as it involves more than tech-nical difficulties.

At the opposite end of the spectrum from the zero-congestion policy, isthe ideal planning policy. This requires the planner to estimate future loadgrowth and then plan both generation and transmission simultaneously tominimize the total expected present cost of delivered power.3 Ideal plan-ning is extremely difficult, but, besides the technical difficulty, anotherproblem blocks our path.

The ideal transmission planning policy Plan transmission and genera-tion together and optimize both for expected load growth, then buildthat transmission and hope the market induces suppliers to invest inoptimal generation.

In principle, this policy works because, given optimal transmission,investors will find the co-optimized generation to be their optimal strategy.But, both joint optimization and any real markets for generation are subjectto error. Consequently, there will be times when ideal optimization directsthat generation should be built in location X, and a corresponding trans-mission line be built to serve that generation. If such a line is built and themarket decides not to build generation at location X, the planners will beseverely embarrassed by their line to nowhere.

The problem to focus on is not the error, but the ‘embarrassment’. Errorsare taken into account by our theory of minimizing expected cost. The‘embarrassment’ causes a more fundamental problem. It prevents plannersfrom adopting the ideal transmission planning policy.

Planners will not undertake a project that can lead to such an embar-rassing situation. Instead, they will simply attempt to optimize their trans-mission for generation that has been built, is being planned, or at the least,appears to be an obvious extension of an existing trend. They will notpredict optimal generation and build transmission for that prediction. Itmight seem that this will make no fundamental difference, but it does. Thestrategic generation problem tells us that as soon as the generators realizethe actual planning policy is no longer the ideal planning policy, they will


no longer invest optimally for that policy. Instead they will invest optimallyfor the new policy, and that is an entirely different matter.

Transmission planners will not build for theoretically determined futuregeneration investments but will instead build for generation as determinedby generation investors. This means that generation investors can manipu-late the transmission planners by their selection of generation investmentsites. They will learn what policy the planners are using and game thatpolicy. This is the strategic generation investment problem.

Strategic Manipulation of Optimal Transmission Planning

Having given up on the ideal transmission planning policy, the questionbecomes, what is the best realistic planning policy? There are many choices,but only a few can be stated simply. One of these, the zero-congestionpolicy, has already been ruled out. Can we find a better one? The next-bestpractical alternative will be considered. This specifies that the plannershould optimize transmission taking generation and anticipated generationas given. This will be called the ‘practical planning policy’.

The practical planning policy Build the transmission system that isoptimal given actual and anticipated generation.

If planners follow the practical planning policy, and generation decidesmistakenly or perversely to locate in a remote region where fuel is cheap buttransmission is so expensive that the combination is uneconomical, theplanners may well have to build accommodating transmission.4 Althoughthe result may not be optimal, it will be far more sensible than building tothe point of zero congestion. To analyse the planner’s choice, consider thenet social benefit of a transmission and generation project.

This project consists of remote generation which ‘exports’ power over atransmission line to the central market. Net social benefit consists of (i) thenet benefit to consumers, (ii) generation profits and (iii) transmissionprofits. To simplify the calculation, assume that the entire project is smallcompared with the central market and that long-run supply in the centralmarket is very elastic. This implies that the remote generation projectwill not change the price paid by central consumers, and as a consequenceit will be of no net benefit to them. The project, if efficient, will dis-place central production with cheaper remote production and delivery, butthe savings will be entirely captured by the investors. As a consequence netsocial benefit reduces to generation profits plus transmission profits:

Net Social Benefit�Generation ProfitsTransmission Profits.


Assume that remote generators are paid one price and central consumersare charged a higher price which creates congestion rent. From this, the costof the transmission investment can be subtracted to find transmissionprofits, which can be negative or positive:

Net Social Benefit � Generation Profits (Congestion Rents – Line Costs).

In effect, remote generation is paying the congestion rents because thecentral price is determined by the central market cost of supply. If remotegeneration is required to pay the full cost of the transmission line throughcongestion rents, then net social benefit equals generation profits, and profitmaximization by generators will maximize net social benefit. In this casesuppliers will invest optimally in generation and if the planners follow thepractical planning policy, the combined project will be optimal.

For linear (proportional) transmission costs, congestion rents do coverline costs for the optimal line. Consequently if transmission costs are linear,the practical planning policy will induce optimal generation investmentand the optimal transmission investment for generation. The combinedsystem will be optimal.

The corollary of this result is that when transmission costs are not linear,investment is unlikely to be optimal. For example, consider again a windfarm. Under the zero-congestion policy, the wind farm could locate as faras it liked from the central market with complete impunity. Under the prac-tical planning policy, the planner will build only the optimal transmissionline and the further from the central market that the wind farm locates, thesmaller will be that line. The marginal cost per MW of line capacityincreases roughly in proportion to the length of the line for any line capa-city. Since the optimal line capacity occurs at the point where the marginalcost of line capacity equals the average congestion rent, the longer theoptimal line, the greater the average congestion rent. Since the wind farmwill pay these rents, it will care how long the line is.

This is good news. The practical planning policy should not only buildmore economic lines for existing generation, but also induce the existenceof a more efficient spatial distribution of generation. The combined savingshould be significant and perhaps enormous. But this does not imply thatthe induced generation investment will be optimal under the practical plan-ning policy.

Suppose the cost of transmission is which implies that for an optimalline, congestion rent is exactly half the total cost of the line. The wind-farminvestors will realize that they must pay only half the cost of the line since,under the example’s assumptions, the only transmission cost they pay is the

c√k,


congestion rent. When they differentiate profit with respect to distance, x,from the central market in order to optimize their location, their value forthe dCR/dx will be half what it should be and this will cause them to locatetoo far from the central market. The planners will then be forced to build aline that is optimal for generation located too far away, but the result will bea combined generation-transmission project that is suboptimal.

The obvious remedy for this problem is to charge generation thedifference between the congestion rent and the cost of the line. This impliesa charge that varies by location and some measure of a generator’s size, butthat is quite different from a congestion charge. The need for such chargeshas long been recognized (Brunekreeft et al., 2004; Vogelsang, 2004) and hasbeen the focus of many economic schemes, few of which have been informedby legitimate economic analysis. Many of these go under the name of ‘MW-mile’ charges, though the most elaborate one was an Alberta scheme calledSERP. It based locational charges on an analysis of line-by-line power flowsthat would be caused by a short circuit at the location in question.

Without such a scheme, the fallback position is to charge every MWdelivered to the system a fixed (independent of location) charge per MWh.This has the advantage of preserving short-run efficiency by leaving the dis-patch unaffected. It will, however, leave the practical planning policysending inefficient long-run signals for generation location.

These insights define a fundamental problem for transmission planningin the context of a wholesale generation market that includes market-driven investment:

The fundamental transmission-planning problem Is there a mechanismfor collecting transmission fixed costs which, when coupled with the prac-tical planning policy and nodal pricing based on competitive energyprices, will improve the total efficiency of the power system? Both short-run dispatch efficiency and long-run incentives for generation investmentmust be considered. The standard of comparison is a per-MWh chargeon all supplied energy independent of location.

Although it seems unlikely that any charging mechanism can be found thatwill be both short- and long-run optimal, it does seem likely that someimprovement is possible at least for realistic networks. Because so manyerroneous collection mechanisms have already wasted time and money,none should be seriously considered until some proof is given that they willlead to improved efficiency on some well-specified collection of powersystems. These test cases can be gross simplifications of real power systems,but they must include the possibility of locational choices for generationinvestment.


4. MERCHANT TRANSMISSION INVESTMENT

If transmission is not built by a public or regulated entity, it will be built byprivate investors. This happened before the power industry was regulatedand it happens today. But the question of interest is how much will be builtand how efficiently it will be built. Certainly there are no easy assurancesfrom theoretical economics that an unregulated market will perform well,and to date there is no empirical evidence for this proposition. There are,however, several theoretical reasons for concern.

Returns to scale suggest that investors will need market power to recovertheir fixed costs, but market power, in this industry as in all others, leads tounderinvestment. Externalities generated by the use of the grid for tradesuggest there will be free-rider problems, which will exacerbate the under-investment problem. Transmission investment provides two other external-ities, the value of which private investors will not easily capture. First itreduces market power in the generation market (Stoft, 1997; Borensteinet al., 2000), and second it provides reliability. Generation market power isa particularly knotty problem because energy suppliers are known to lobbygovernment bodies in an attempt to block transmission investment that isnot in their interest. Any attempt to reduce market power with transmis-sion is likely to be the target of supplier lobbying.

In spite of these difficulties, transmission investment is not always asdifficult to finance as many assume. Often congestion rents are viewed asthe sole source of remuneration to merchant transmission. In fact, linesmay be built without any assumption of congestion income. If there is acheap but remote area for generation investment, the suppliers that locatethere may build lines simply to bring their power to market. They will stillhave free-rider problems and the like, but they will be motivated by otherthan future income from congestion rents. Similarly a city may find localsupply too expensive and may build transmission out to the larger networksimply to access cheaper power with no thought of future congestion rents.Alternatively all three motivations may coincide.

Although congestion rent may not be the primary motivation for build-ing lines, rights to new lines should be given to investors to encourage suchinvestment and internalize the line’s benefits to the extent possible.Transmission rights can protect an investor from congestion cost thatwould otherwise have been paid if the investor used the line and can provideincome to the extent others use it when it is congested. Another benefit oftransmission rights is to cause negative externalities to be internalized.Before turning to the difficulties of merchant investment, it is worthwhileunderstanding how transmission rights should be granted in return forinvestment and what useful role they can play.


Transmission Rights

Because alternating current (AC) transmission lines are so thoroughly inte-grated with the power grid, their owners generally cannot be given physicalcontrol or the right to charge anyone who uses the line. For example, almostany power flow from one point to another on a power system causes somepower to flow on most transmission lines in the grid, although the amountthat flows on remote lines is too small to matter. AC power lines are in someways like a set of connected water pipes without valves between them.Pushing water from one point to another affects the flow in almost every pipe.Because of such physical complexities, the standard proposal is to reward theinvestor in a transmission line with a set of financial rights, not physical rights.

The standard financial transmission right is a congestion revenue right(CRR) which is defined by a quantity, Q, source and sink, A and B, and aset of time intervals, T. At any point in time during T, the CRR pays (PB –PA)Q, which is the congestion price from A to B times the megawatt quan-tity of the right. The payment has nothing to do with actual power flowsassociated with the owner of the right.

Although CRRs can, in principle, be privately issued, they are generallyissued by the ISO and that will be assumed throughout this discussion.Consequently, at any point in time, there is a well-defined set of CRRs, R,that have been issued. An important property of R is its feasibility, which isdefined as follows. Corresponding to every CRR there is an imaginary powerflow of Q MW from A to B, during time intervals T. This imaginary powerflow has nothing to do with actual flows on the grid. Since every CRR in thisset corresponds to an imaginary power flow, we define R to be a feasible setof rights if the corresponding set of imaginary power flows could take placeon the system without violating any reliability constraint. This has nothingto do with load or generation and concerns only the transmission system.

The following procedure can be used, at least in principle, to rewardinvestors in transmission upgrades. First, sell a set of CRRs, in an auctionthat does not withhold any feasible CRR. This should leave no valuableCRR unallocated. When a transmission upgrade is completed the systemshould be able to accommodate more power flow reliably, and this shouldexpand the feasible set of CRRs. The investor is allowed to claim any set ofCRRs which, combined with the existing set, forms a feasible set in theupgraded system. This allows the investor a certain amount of choice andit accounts, to some extent, for positive external affects of the upgrade.There is one more rule. If the ‘upgrade’ has actually reduced the feasible setof CRRs, the investor must take counter-flow CRRs such that the new allo-cated set is feasible. These will have negative financial value and will prop-erly discourage any system downgrades, provided that the initial set of


CRRs matched the actual flows on the system (Bushnell and Stoft, 1996and 1997). Unifying these two rules, the reward for a modification of thetransmission system is this:

Feasible CRR allocation rule for rewarding transmission investment Themodifier of a transmission system must take from the ISO a set of CRRssuch that together with the pre-existing CRRs the new complete set is afeasible set on the modified system.

This approach to rewarding investment has several advantages. First, if thepre-existing set of CRRs matches the flows on the system, it makes itunprofitable to damage the system. Second, it gives the investor themaximum possible congestion rent while treating others fairly. Third, forfeasible sets of rights, the cost to the ISO of paying CRRs is never morethan the congestion rent collected.5 Unfortunately there are also a numberof drawbacks. First, the awarded CRRs do not adequately compensateinvestors. Second, to live up to its potential, the set of CRRs that the ISOmakes available needs to be quite complex. For example, the investor maywant north–south rights at some times and south–north rights at others.Third the results assume investors have no market power in the energymarket.

Other Styles of Rights

There are a number of other styles of transmission rights (Hogan, 2002;Gribik et al., 2004). PJM uses financial transmission rights (FTRs) which intheir original (obligation) form are identical to CRRs except that therevenue associated with the entire set of rights is adjusted to equal the con-gestion rent collected from the entire transmission system. This correctionis generally small. The FTR option (as opposed to obligation), introducedinto PJM in 2003, is a fundamentally different form of transmission right.At any point in time an FTR option from A to B pays its holder themaximum of what a similar FTR obligation pays, or zero. Unlike FTRobligations, FTR options, like other options, have only non-negative values.

Using FTR options instead of obligations results in a smaller feasible setof rights because cancellation of rights in opposite directions is not allowedunder the normal meaning of ‘feasible’. For example, consider a three-nodetriangular network with generation at A and B and with load at C, and witha 100-MW line from A to B and two 500-MW lines to C (Figure 4.6). Withthe standard, obligation-style, CRRs, two 500-MW FTRs, one from A toC and the other from B to C, are simultaneously feasible. With FTRoptions, only two 300-MW FTRs are feasible because if one option were


not exercised, the other would correspond to a flow that would load theA–B line to its limit.

Now, the power flows used to compute the feasibility of a set of FTRsare purely imaginary, so the same feasibility rule could be used with optionsas with FTR obligations. But in this case, two 5,000-MW options, one fromA to C and the other from C to A would be feasible (because these flowscancel). The option in the congested direction, from A to C, would earn apayment of 5,000 MW times the congestion price while the ISO wouldcollect only a congestion payment of, at most, 600 MW from A to C. Sincethe counter-flow option would not have a negative value and would pay theISO nothing, the ISO would find itself short of congestion rent by 4,400MW. For this reason PJM computes feasibility of its options rights asfollows.

The FTR auctions maximize the quote-based bid value of a set of simul-taneous feasible FTRs awarded in the auction. To ensure feasibility,counterflow created by an FTR option bid must be ignored when the FTRbids are tested for feasibility.6

There may be no economic justification for the ISO to create optionrights. The system operator is in a good position to create standard CRRsbecause it can back them with congestion revenue without risk and conse-quently does not need to charge any risk premium for them. But becauseadding options to the mix reduces the total amount of hedging available, itmay be better to let a private derivatives market supply CRR options ifthere is a demand for them.

The motivation for these alternative styles of rights, and for that matterthe motivation for transmission rights in general, is mainly to allow thehedging of energy transactions. For example, the main FTR page of the


Figure 4.6 Option rights reduce the feasible set of rights

A B

C

100 MW

500 MW500 MW

PJM website states that ‘Financial Transmission Rights (FTRs) [are pro-vided] to assist market participants in hedging price risk when deliveringenergy on the grid. . . . The FTRs provide a hedging mechanism that canbe traded separately from transmission service. This gives all market par-ticipants the ability to gain price certainty when delivering energy acrossPJM’.7 No mention is made of transmission investment.

Market power is another area of concern with transmission rights(Joskow and Tirole, 2000). This also is not closely related to transmissioninvestment. If an energy supplier has market power in a load pocket, it canenhance its power by purchasing transmission rights into the pocket. Theserights will pay more when it raises the local price of energy, which makesits exercise of market power more profitable.

To date, the main use of financial transmission rights has been as a sub-stitute for prior rights held by transmission owners. This has been quiteuseful because of the compatibility between CRRs and nodal pricing. Thissubstitution and the more general use of transmission rights has also pro-vided a useful hedging mechanism for nodal price differences, that is, con-gestion rents. The value of CRRs in this regard is still not well documented,but they seem to be well accepted in this role. Although their use as a partialincentive for transmission investment has long been advocated, and at leastPJM and NYISO have rules in place to this effect, there does not seem tobe any documented instance of CRRs playing a significant role in any mer-chant transmission project.

The Paradox of Transmission Rights

The appeal of rewarding transmission investment with CRRs comes in partfrom their properties in an idealized world of perfect competition. In thiseconomic model, marginal investments are always possible and their cost islinear. Consequently a line may be upgraded by one megawatt for 1/100 thecost of a 100 megawatt upgrade. Moreover, any investor can upgrade anyline; there is no ownership of the transmission path. This brings perfectcompetition to each line in the system. In such a system the congestion rentthat would be earned by a marginal upgrade of a transmission path wouldexactly equal the value of the upgrade in reducing the redispatch costcaused by congestion. For example, if a line is congested for 1,000 hoursper year with a price differential of $10/MWh, then a 1 kW (1/1,000 of aMW) upgrade of the line would save $10�1/1,000�1,000, or $10 per yearin redispatch cost by allowing cheaper generation to substitute for moreexpensive generation. Similarly, if the investor is granted a CRR for 1/kWin the direction of the congestion, the investor will earn $10 per year,exactly what the line is worth.


Investment on every path will proceed as long as the marginal investmentcosts less than the congestion rent earned by that investment. Because thisequals the value of line expansion, investment will proceed exactly to thepoint where it no longer pays to continue investing. Suppose a 100-MW lineneeds a 50-MW expansion because of a new load. Some investor may build30 MW of the expansion, but at that point further investment may becomeunprofitable because it lowers the congestion rent on the 30 MW of trans-mission rights received for the initial investment. In the real world thiswould most likely stop investment before the optimal transmission capac-ity is achieved, but in the idealized world of perfect competition, someother investor will continue the investment, perhaps for another 10 MW.Then as this investor’s stake in high congestion rents discourages furtherinvestment, yet another investor will take over the project. In this way everylast kilowatt of economic investment will be made.

Under the heroic assumptions of perfect competition, rewarding invest-ors with all of the congestion rents provides the ideal incentive for invest-ment (provided that the allocation of rents is also ideal). Under realisticassumptions, which include market power, paying investors more whenthere is more congestion on their line results in withholding of investmentand too much congestion. As will be discussed in Section 5, the exact oppo-site payment scheme has merit when the investor is a monopoly transco. Inthis case, charging the investor the amount of the congestion rent insteadof paying the amount of the congestion rent results in an ideal investmentincentive.

Returns to Scale and ‘Iumpiness’

Returns to scale, as discussed above, means that optimal transmission invest-ments will simply not generate enough congestion rent to pay for themselves.Obviously, this means that merchant investors will build less than the sociallyoptimal level of investment (Joskow, 2003; Joskow and Tirole, 2004).

The same does not hold for lumpy technology. Consider Figure 4.7,which shows transmission investment coming in lumps. If the rental cost ofa lump of transmission is $8/MWh, it would not be worth building thesecond lump because the redispatch ‘cost’ triangle that would be eliminatedaverages less than $8/MWh.8 Thus the social optimum is to build only onelump and the rent on this lump will be $16/MWh which pays for the linetwice over. This shows that with lumpy technology, socially optimal invest-ment may produce more than enough congestion rent to pay the cost ofthat investment.

If we call the lumpy technology in Figure 4.5, ‘linear-lumpy technology’,indicating that it has constant returns to scale over a range of lumps, we


can ask the question: would merchant investors underinvest in linear lumpytechnology? Figure 4.7 shows that they might invest optimally and earnwell above a normal rate of return. If demand were a bit greater in thefigure, so that the intersection of supply and demand were to the right of‘2 lumps’, the social optimum would be two lumps. However, merchantinvestors would fail to build the second lump because it would earn almostno congestion rent. The result would be underinvestment and excess profits.Of the three possibilities, underinvestment, optimal investment and over-investment, they will avoid the third. A static analysis would conclude thatthis should lead to too little investment on average but to excess profits. Thedisconcerting aspect of this conclusion is that it appears that on average, infact almost always, they would earn above-market rates of return. Thisseems unlikely.

In a dynamic market this result appears even more suspicious. Investorswill anticipate the possibility of high rates of return and will invest a littleearly, which will increase their average investment and lower their rate ofreturn. Considering this dynamic effect, there seems little reason to suspectthat lumpy technology will lead to systematic underinvestment, or even thatinvestment will be wrong on average. This is an optimistic result, but onemore effect needs consideration.

Lumpy investments pay least when they are first made and most justbefore the next investment is completed. In fact, many optimal transmis-sion investments lead to a protracted period with virtually no congestionand hence no congestion rent when they are first completed. This effect maybe dramatic. When a lumpy upgrade is first made it is common for the lineto be almost completely uncongested and this situation may last for years.The island of Nantucket off the coast of Massachusetts is served by a singledirect current (DC) cable which is likely to become congested in the coming


Figure 4.7 Optimal investment in lump technology may be preferable

$/MW

MWh

Rent

1 Lump

Demand

Supply

2 Lumps

30

14

Cost

few years for a few hours in August. This might cause a very partial black-out during the tourist season which is unacceptable, so a second identicalcable will soon be added. This one will not be congested for perhapsanother 20 years. Even when a lumpy transmission investment can expectfull recovery of its costs from congestion rents over the long run, the costrecovery may well not begin for years and will be very slow when it starts.This ‘back-end loading’ of the revenue stream creates grave risks for theinvestor. What if a new technology, such as cheaper high-voltage DC linesor aluminum–zirconium wires, comes on the market before high levels ofcongestion kick in? What if load growth is slower than anticipated? Whatif gas pipelines are built to fuel new generation that competes with powerimported on the transmission line (Barthold, 2003)?

This investor’s payment stream does not mirror the stream of socialbenefit which results from the elimination of, or reduction in, previous con-gestion rents. Because that benefit stream starts out at the rental cost of theline, it is far less risky than the stream of congestion rent. Risk is costly, somerchant investment based on collecting congestion rents from CRRsissued in return for the investment will be much more costly than a sociallysponsored investment in the same project. Low-risk investing is cheaperthan high-risk investing.

A simple example may help explain the relationship between the socialbenefit stream and the congestion rent stream on a lumpy transmissioninvestment. Suppose load in a load pocket takes on values between X andX200 MW with a uniform probability distribution. Suppose the pricedifference between supply from the load pocket and external supply is$20/MWh. Suppose additional transmission costs $5/MWh and comes in100-MW lumps. When should transmission be built?

Only when the present line is congested would a new line add value.When the line is congested half the time, as shown in Figure 4.8, the valueadded will range from zero when it is just barely congested to $20/MWh ofnew-line capacity when the load is at its peak value and the new line is fullyutilized. On average, under these conditions the new line would provide$10/MWh of line capacity in social benefit while congested and $0/MWhwhile not congested. Consequently when the existing line is congested halfof the time, it will provide $5/MWh of benefit on average. Since this is whatthe line costs to build, this is the break-even point. When the line is con-gested less often, investment is not warranted.

If new lumps of transmission are added at the socially optimal times, con-gestion rent just before the new addition will be $10/MWh on average, andjust after the new addition it will fall to zero. When the new line is first putin place the line will earn nothing, but its earnings will grow until the nextlump of transmission is built. If the growth in X is linear, then, on average,


the line will earn $5/MWh, exactly enough to cover its cost. (This exampleassumes a zero discount rate.) At least in this case of lumpy investment,optimal investing will be rewarded with exactly the right level of congestionrents.

Note the difference between the social benefit from a transmission upgradewhich starts out covering the rental cost of the line on day one and the streamof congestion rents which flow to the merchant investor. These start out atzero and only reach the break-even point half way to the point in time whenthe next investment will be made and rent will again drop to zero. Also notethat, as shown in Figure 4.9, the social benefit of building the line is muchgreater than its cost. It is normal to find such ‘consumer surpluses’ in a well-functioning market. Obviously, this is a very narrow result, but it disprovesa common view which holds that lumpy investments inherently underpayinvestors similar to the way in which increasing returns to scale (a decliningmarginal cost curve) underpay optimal investing. The main problem withlumpy investments is that they pay off merchant investors very late, whichmakes them extremely risky for a merchant investor even though risk insocial benefit is low. This can greatly increase the cost of merchant lines rel-ative to their cost if built under rate-of-return regulation.

Free Riders?

When large merchant transmission projects are contemplated it is oftennoted that many will benefit from such a project in the initial years but allwill attempt to avoid paying for it. As soon as the line is completed certain


Figure 4.8 Optimal investment eliminates congestion

Load duration 1

X

X +200

Existing line

Use of new line

loads will find their prices reduced and certain suppliers will find they cansell at a higher price. Both will want the line to be built, but all will wantothers to pay for it. Figure 4.9 illustrates such a situation.

In the first year after the line is built, market participants will benefit by$5/MWh of line capacity, yet the investors will be paid next to nothing incongestion rent. This observation usually leads directly to the conclusionthat all those who benefit without paying are free riding. But, in thisexample, that is not the case. The investor will be paid in full and the gapbetween social benefit and congestion rent is not the result of free ridingbut is simply normal consumer surplus.

Of course, transmission projects are likely to suffer from the effects ofreturns to scale as well as lumpiness, so it is likely that investors will beunderpaid if they depend on congestion rents. In this case free riding is acorrect diagnosis, but it will be extremely difficult to assess the extent of thefree riding. In particular it is wrong to believe that investors should capturethe entire consumer surplus even at the beginning.

An investor with market power may be able to capture some of the valuethat would otherwise accrue to free riders. Similarly, an investor withmarket power may be able to capture part of the normal consumer surplusthat would be provided by optimal investing and complete fixed-cost recov-ery. Both reasons probably help to explain the many proposals to allow theexercise of market power by merchant transmission investors. Some marketpower would help cover investment costs, so there is some legitimacy to thesuggestion. But once started down this path, merchant investors may wellask for more market power than needed to break even. Proposals to basethe fundamentals of cost recovery on market power fail to provide a ration-ale for a market that must rely largely on market power to cover costs. Theinvestment efficiency of such a market is unclear at best.


Figure 4.9 Investors should not capture full social benefit

$20/MWh

time

Congestion Rent

Social benefit

$5 Costof line

The principal argument for efficiency in such a market is that supplierswith market power still minimize costs, given their output. But this argu-ment also holds for suppliers who do not decide what transmission isneeded but simply respond to an auction held by the system operator forthe provision of certain transmission capacity. Until some means can befound of tailoring the exercise of market power to provide the right level offixed-cost recovery only on efficient transmission investments, the argu-ment for the deliberate introduction of market power as a method of induc-ing investment is weak to non-existent.

Mixing Planned and Merchant Transmission

The possibility of planned transmission both discourages and threatensmerchant transmission. Before a merchant line is built, potential sub-scribers to the project would prefer to induce the planners to build the lineand spread the cost over the broader market. This discourages participa-tion in the project by those who should buy a long-run contract for the useof the line. Essentially, this exacerbates the free-rider problem.

Once a merchant line has been built, those who have not pre-paid for itsuse will still wish to encourage the planners to build a competing line, asactually happened in Australia (Firecone, 2003; Littlechild, 2004). As themerchant line was a DC line, it could directly charge those who used it(Brunekreeft, 2004b), but had it been an AC line those wishing to use thepath would still have reason to lobby the authorities to overinvest by build-ing a second line, thus driving down congestion rents and providing acheaper alternative than use of the merchant line. This possibility is a threatto merchant investment.

Because merchant transmission has so far proven itself entirely inade-quate, some have suggested a mixture of merchant and planned transmis-sion (Hogan, 1998; Rotger and Felder, 2001; Chandley and Hogan, 2002).For this to succeed, the level of discouragement and threat must bereduced. This can be accomplished if the role of planning can be definedwith enough clarity. If merchant investors know which lines will and willnot be built by the planners, then planning should not discourage invest-ment in lines that will be needed for merchant purposes, but which the plan-ners will not build.

One suggestion for the bright line between planned and merchant linesis that planners will build only lines which would require the cooperationof ‘many parties’ because in this case the free-rider problem is thought tobe particularly severe. While this would give merchants clear guidance atthe two extremes, there would inevitably be an important middle ground ofambiguity, perhaps encompassing most transmission expenditures.


Another proposal for separation was made several years ago in thecontext of the Alberta market (Stoft, 2002). That market, to a greaterextent than most, fails to send adequate locational signals for genera-tion investment and, like all power markets, lacks proper real-time demandelasticity. Consequently, transmission investment is occasionally requiredfor reliability purposes. It was proposed that the planners build only forreliability, and that when such a project is undertaken, merchant invest-ment be allowed to expand the project for the incremental cost of theexpansion, thus avoiding significant fixed costs. To further facilitatemerchant investment, it was proposed that the transmission administrator(planner) also facilitates joint investments by joining merchant projectsunder certain circumstances when lumpiness is a problem. The transmis-sion administrator would buy a part of the line and keep it out of use untila new party decided to purchase it. This proposal was not viewed as ideal,but only as a better alternative than the rule Alberta eventually did imple-ment, requiring that congestion be completely eliminated. It also has theadvantage of not depending on or blessing the exercise of market power.

5. PERFORMANCE-BASED REGULATION FORTRANSMISSION MONOPOLIES (TRANSCOS)

The planning process provides non-directive and generally weak incentives.Planners know that if they do a demonstrably poor job, they may findthemselves out of work. This provides an incentive and most engineers areactually quite motivated by this and by professional pride and a desire forprofessional recognition. Consequently, it is a mistake to believe that theplanning approach lacks good incentives. However, these incentives maydiffer from a pure incentive to minimize the total cost of delivered powerand may put too much weight or not enough weight on reducing com-plaints about the occasional outage or about congestion that inhibits trade.Consequently, it may be better to design explicit formulas that determinemonetary rewards for minimizing total cost. These rewards cannot easilybe applied to individuals, so the standard approach is to apply them to theprofits of a regulated monopoly, a transco.

Any regulation of a monopoly provides financial incentives, but oftenthese have not been explicitly designed or even considered. The incentivesof cost of service regulation are usually poorly thought out and derivemainly from unintentional lags in rate setting and the subjective applica-tion of rules such as the requirement that investments must be ‘used anduseful’. When financial incentives are explicitly designed, the regulation iscalled ‘performance-based regulation’ (PBR), or ‘incentive regulation’.


A Direct Approach to PBR for Transmission Investment

If it is assumed that power is available to all customers at the competitiveprice, then the objective of transmission investment is the minimization ofthe total cost of delivered power. Without the assumption of availability, costcould be minimized by maximizing blackouts, and in the limit, delivering nopower at all. Because of occasional blackouts (load-shedding events), thecost-minimization framework must be maintained by assigning a cost to‘unserved load’. Assuming that this assignment of cost can be accomplished,the goal of transmission investment is total cost minimization.

This goal is easily translated into a theoretical scheme for incentive regu-lation (Gans and King, 2000; Léautier, 2000). A monopoly transco shouldbe paid a fixed but generous sum, R, per megawatt of delivered power lessthe cost of congestion (CE, the redispatch cost) and less the cost ofunserved (lost) load, CLL. In most power systems, $10/MW hour of deliv-ered power would be more than sufficient for R.9 The transco’s profitswould then be:

Profit�R – CE – CLL – CT,

where CT�the rental cost of the transmission system.As with a single transmission line, the rental cost of the system includes

the cost of capital as well as maintenance. Note that the transco does notkeep the congestion rent. If the transco can reduce the sum of CE and CLLby more than $1 by investing and thereby raising CT by $1, it will find itprofitable to do so and this will be beneficial to society. Hence this incent-ive mechanism aligns the transco’s incentives with social welfare. Anyreduction in CECLLCT increases the social surplus by the sameamount, and this amount goes into the pocket of the transco.

Note that this incentive scheme properly rewards ‘effort’ which has anon-monetary cost to the transco and is consequently not observable by theregulator. If the transco can increase its monetary profit, as defined above,by $10, but only by expending $9 worth of unobservable effort, it will payit to do so, as it should, since this is socially beneficial. As will be seenshortly, this property is shared only by what are called ‘high-powered’incentive mechanisms.

This scheme presents three difficulties, measuring CE and CLL, andsetting R. Congestion costs, CE, are the difference between the actual pro-duction cost of energy and the lower cost that could be achieved withoutany transmission limits. This difference can be fairly well approximated inany system with centralized bidding and nodal pricing. Occasionally theremay be some difficulty with knowing how much certain generators could


produce were they not constrained by the transmission system, but mostsystems have on record a realistic estimate of each generator’s outputcapacity and this should serve as an adequate proxy for the true value.(Losses are quite easily estimated.)

The cost of unserved load is far more problematic. The standard error forsuch an estimate is probably a factor of three. In other words, if it is esti-mated to be $15,000/MWh, there is probably only a 68 per cent chance(roughly) that the true value is between $5,000/MWh and $45,000/MWh.(This is a purely subjective conjecture.) More problematic is the fact that thiscost occurs erratically. Major cascading blackouts may happen about onceevery 20 years and result in half the load being lost for six hours. The costof such a blackout would be 6�0.5�$15,000 per MW of load which isequivalent to about $7/MWh for an entire year. When such a blackout doesoccur it could cost the transco a year’s revenue and result in bankruptcy.

Perhaps a solution to this is to undervalue lost load by a factor of 10 or20 and use a nominal value such as $1,000/MWh. This would still result ina rather erratic cost stream, but it might be tolerable. Certainly, consumerswould pay a significant risk premium to a transco under such an incentive.Sometimes it will also be difficult to tell whether loss of load is due to gen-eration or transmission problems, and this will lead to litigation costs andother inefficiencies.

The problem of setting R is the most fundamental problem of regula-tion. The regulator always has less knowledge (information) of the cost-minimizing solution than does the regulated firm. Because of this, theregulated firm can extract some ‘information rent’ from the regulator. Ingeneral the stronger the incentive, the better the performance of the regu-lated firm, but the more rent it can extract. The present scheme provides thestrongest incentive; the transco keeps every dollar of cost that it saves. As aconsequence it will be able to extract considerable rent. Specifically, the reg-ulator knows that if it sets R below total cost, CECLLCT, the transcowill go out of business, a result that must be avoided, but it does not have agood estimate of total cost. Its only reasonable choice is to set R three orfour standard deviations above the cost-minimizing investment level of CECLLCT. Given the level of uncertainty in estimating this total, this islikely to result in an extremely high rate of return for the transco.

Two Approaches to Reducing Information Rent

The advantage of the PBR scheme just described is that it strongly motivatesthe supplier to expend cost-saving effort that is difficult if not impossible forthe regulator to observe. The disadvantage, as we have just seen, is that itrequires the payment of high information rents. In the case of transmission


investment, these rents could be extremely high and well beyond any accept-able level.

Because of such high information rents, it is desirable to reduce thepower of the incentive mechanism. Many types of PBR are forms of price-cap regulation including the mechanism just described, though it is a rathernon-standard price-cap incentive. In that scheme, R can be thought of asthe fixed part of a two-part transmission price, the other part being the con-gestion price. There are two standard ways of reducing the power of aprice-cap incentive. First, the cap, R in this case, can be reset periodically.The more frequently it is reset, the lower the power of the incentive it pro-vides. Second, profits under the mechanism can be shared between themonopolist and the consumers. The smaller the share kept by the mono-polist, the lower the power of the incentive. In both cases, lower power willcorrespond to lower information rents paid to the transco.

Consider the periodic resetting of the price cap. When the price cap isreset, the objective is to provide the monopolist with a certain allowed rateof return during the next period. To this end, the values of CE , CLL and CTwill be estimated for that period. If the period is short, most of the transco’scosts (CT) will be correctly anticipated and covered by the allowed rate ofreturn. The longer the period, the more cost will be saved or incurred unex-pectedly. This will lead to unexpected changes in CE , CLL which will changerevenues. These intra-period changes in expenditure (CT) and revenue (R –CE – CLL) result in profit deviations from the targeted allowed rate of returnand this provides some incentives for both cost minimization and beneficialinvestment. The longer the period between rate cases, the greater the pro-portion of expenditures for which the price-cap mechanism can provide anincentive. At one extreme lies pure price-cap regulation and at the otherpure rate-of-return regulation. In between we find actual rate-of-returnregulation in which price caps are reset roughly every three years.

Timing is the major problem with using periodic price-cap setting toachieve a lower-powered incentive and lower information rents. Investmentcosts are often incurred over a much shorter period of time than thebenefits from the investment. Hence the reset period may be long relative tocosts but short relative to benefits. In this case a lower proportion of coststhan revenues will be captured in the resetting process. This will tend to dis-courage efficient investment. This is related to the well-known incentiveproblems that occur shortly before a rate case. At this time, it becomesadvantageous to make costs appear high and revenues appear low. Also theincentive for investment is diminished shortly before a rate case, because theregulator may view such expenditures as already paid for.

Profit sharing avoids these timing issue because it does not make peri-odic adjustments, but instead shares profit in some fixed proportion on a


continuous basis. Economic profit, for example ��R – CE – CLL – CT,accounts for a normal rate of return on investment (in CT), so profitsharing can never prevent a supplier from achieving a normal rate ofreturn, it will only bring it closer to that level.10 If unshared profit, �, ismaximized by a certain investment strategy, then half of the unsharedprofit, �/2, will also be maximized by exactly the same investment strategy.Consequently, if there were no information problem with computing �,profit sharing would leave the transco’s behavior unchanged. This would beideal. The excess wealth transferred by the high profits of pure price-capregulation could be reduced by any amount simply by setting the sharingparameter appropriately and this would cause no loss of efficiency. Butthere are information problems, and the fundamental trade-off of regula-tion ensures that, if the power of the incentive is greatly reduced, efficiencywill suffer. That is true whether the incentive’s power is reduced throughperiodic resetting of the cap, as previously described, or through continu-ous profit sharing, but it is easier to explain the effect in the continuousprofit-sharing context.

To glimpse the contradiction inherent in ignoring the informationproblem, consider, the case in which the dollar-valued economic profit, �,divided by the invested capital needs to equal 50 per cent for the initialdetermination of R in order to avoid any significant probability of bank-ruptcy. If the normal rate of return on equity is 15 per cent, then themonopolist will start out making a 65 per cent rate of return on equity.11

To reduce this, consider a profit-sharing ratio of 1 per cent for the transcoand 99 per cent for load. This reduces � to �/100 and brings the initialreturn on equity to 15.5 per cent. If the transco raises � to 100 per cent witha superb effort, it will receive 16 on equity and if it performs terribly, letting� fall to 0 per cent, it will still receive 15 per cent on its equity. Even thoughit prefers 16 to 15 per cent, this limited reward is not likely to induce theeffort level required to raise � divided by invested capital from 0 per centby 100 per cent.

The important point about effort is that it is a real cost that is notincluded in CT because it is not monetized. It is a cost that does not appearon the books. It will, of course, affect costs that do appear on the books.Effort will reduce these costs for the same level of transmission perfor-mance or increase performance for the same level of monetary cost. Lackof effort raises monetary cost relative to performance. Lack of effort, likeeffort, can take many forms. It can take the form of ‘gold-plating’ officesand equipment or ‘shirking’ by management and workers. Inefficientexpenditures can purchase tickets to the San Francisco Giants’ baseballgames as was done by a major California utility. ‘Graft’ is another type oflack of effort; for example, the transco can subcontract a job to someone


who pays a kickback to someone at the transco. As the rewards for effort(good behavior) fall, all types of unwanted behavior will increase.

Profit sharing on a 50/50 basis takes away half of �, which means it takesaway half of revenues minus costs. If effort had a monetary value, then a$10 effort that produced an $11 revenue would increase � by $1 beforeprofit sharing and by only $0.50 after profit sharing. In either case the effortis worthwhile and will be induced. But because the effort goes unobserved,its cost is not shared by the profit-sharing mechanism, and so the result ofthe $10 effort is an after-sharing revenue of $5.50 which fails to compen-sate for the $10 effort. Consequently, with 50/50 profit sharing such effortswill not be undertaken. A $10 effort that produced a $30 increase in revenuewould still pay off even after profit sharing allowed the transco to keep only$15 of the revenue. So profit sharing does not eliminate the incentive toprovide unmeasured but costly effort, it only reduces it. As the share ofprofits kept by the transco decreases towards zero, the incentive to expendeffort decreases towards non-existent. This explains why profit sharing islimited as a means to control information rents.

Difficulties with PBR for Transcos

The price-cap mechanism just described is most likely impractical becauseit suffers from at least two severe difficulties. First, just as with merchanttransmission the transco’s investments will pay off with very long lag timesand are consequently very risky (Brunekreeft and McDaniel, 2005). Typi-cally, a large cost must be incurred over a period of several years, then forseveral more there will be little or no return on the investment and finally,10 or 15 years after the start of the project, significant payback will begin.This is just one possibility, but a very plausible one. The risk of this delayedpayback contrasts sharply with society’s risk, which is much less becausethe societal payback starts immediately upon completion of the line at arate equal to the rental cost of the line.

Second, transmission investments are tightly linked to reliability (Crewet al., 2004; Sun et al., 2004), and the reliability part of this mechanism,CLL, introduces severe risks and promises to be extremely controversial ifattempted. Most of the major blackouts in the United States over at leastthe last 35 years have been linked more to transmission problems than togeneration problems. This has tended to involve operations and invest-ments other than major line and transformer investment, such as tree trim-ming, computer systems for state estimation, and the setting of line-triprelays. It should be possible to design a separate mechanism for perfor-mance in these areas, but that will not completely disentangle reliabilityfrom line investment.12


The motivation for upgrading a transmission path, which typicallyinvolves several lines, is congestion on that path and the congestion cost,CE, that it causes. But this cost can be reduced in two ways, first by aphysical upgrade, and second by re-rating the path to a higher capacity. Thesecond approach is far cheaper, in fact is almost free, but it decreasesreliability. Unfortunately path ratings are not easily audited as they aresomewhat controversial even among engineers. This is because they are notbased primarily on hard data, such as the temperature at which a wiremelts. Instead, ratings must couple hard data with somewhat subjectivedata, including probabilities of contingencies such as line and generatoroutages.13 Any powerful incentive to upgrade lines will also be a powerfulincentive to cut corners on contingency ratings. Consequently if there arestrong incentives to upgrade lines, there must also be heavy penalties forcutting corners on path ratings. Since path ratings are too difficult for regul-ators to monitor, these penalties must instead be applied directly to black-outs which are very costly but occur very rarely. Imposing the cost of lostload, CLL, is such a penalty, but as explained, it introduces severe risks andwould be extremely controversial. The danger of degrading reliabilityappears to create severe difficulties for designing a useful PBR incentive forupgrading lines.

A third difficulty, less severe than the first two, faced by any transco pro-posal is the fundamental planning problem described above. Any realistictransco incentive will induce the transco to invest in the lines that areoptimal for existing generation, not for optimal generation. The conse-quence will be that generation investors will build generation based on thetransco’s response. For this circularity to produce the efficient outcome, thecharges used to supplement the congestion rents and provide the transco’srevenues must be allocated in a way that induces the correct location ofgeneration.

A number of PBR alternatives for transcos have been suggested, includ-ing some by Vogelsang based on the price-cap tradition, and which arereviewed by Rosellón (2003) and Vogelsang (2004). Vogelsang’s most recentproposal is related to these and the social surplus mechanism describedabove. Unfortunately this proposal is still not able to solve the investmentproblem taking into account reliability and load growth. His current con-clusion is ‘Long periods mark the limits of regulatory commitment and arestill short relative to network investments. As a result, incentives should befurther weakened by adjustments based on rate-of-return regulation with a“used and useful” criterion.’ (p.141). PBR for transcos will be useful forshorter-term incentives (Joskow, 2004), but it cannot yet be relied on tosolve the long-term investment problems.


6. CONCLUSION

Currently, wholesale power markets are undergoing a slow and inefficientdevelopment process marked by such events as the California meltdown,the overbuilding of gas-fired generation in large parts of the eastern UnitedStates, the largest blackout in US history, and the complete redesign of theBritish market. In particular the generation investment problem seems toremain far from solved, though some reasonable incentive mechanismsseem to be on the drawing board.

Moreover, it should be recognized that transmission investment is crucialto the functioning of the new energy markets. The less congestion, the lessmarket power in wholesale energy markets (Stoft, 1997; Borenstein et al.,2000; Gilbert, et al., 2002). For example, San Francisco and New York bothsuffer from serious market power because both have limited transmissionand must rely for significant portions of their energy on local suppliers. Themore transmission into these cities, the more competition in the wholesaleenergy market. Moreover, in every market in the United States, there arenumerous examples of generation units under ‘regulatory must run’ con-tracts. These give the market administrator the right to require the plant torun and in return provide regulatory payments which are often substantial.Such contracts exist largely where these generators have extreme marketpower during some hours of the year because of transmission limitations.Such situations have been numerous and problematic from the beginningand show no signs of disappearing.

Fortunately it is extremely cheap to overbuild the transmission system asmall amount and thereby reduce market power below the level that wouldbe found under an optimized network. This is because, at optimal invest-ment, the derivative of total system cost with respect to increased capacityis zero. That is the first-order condition for optimality.

Neither a merchant approach nor a PBR approach is conducive to alle-viating wholesale market power problems by overbuilding the network.Both have biases towards underinvestment, and both are likely to be erraticin their behavior during the decades it will take to tame the likely flaws intheir designs. Rate-of-return regulation is more adaptable. Provided theregulator declares in advance that a line will be considered ‘used-and-useful’, it should not be difficult to get the transco to build it.

Neither a merchant approach nor a transco/PBR approach has yet beendeveloped to the point where it could be considered useable in practice.Both appear to be in rather early stages of theoretical development.Moreover, it appears that any application of these approaches will requirea level of understanding and subtlety that is not yet apparent among regu-lators, at least in the United States. Given these difficulties and the poor


record of the energy market deregulation process, it appears to be too earlyto begin any policy initiatives based on either of these approaches. Withtransmission investment costs amounting to only 3–8 per cent of the retailcosts (Joskow and Tirole, 2002), it is better to rely on a relatively safeapproach to transmission investment, even though it is a bit less efficientthan results promised by some poorly understood theoretical approaches.This is not to say that merchant transmission investment should be dis-couraged. It should be allowed and regulated only lightly (Brunekreeft,2004a), but it cannot be depended on and should not be the focus of thoseconcerned with ensuring sufficient transmission capacity.

Rate-of-return regulation applied to a transco or to wire companiesunder the direction of an ISO is not without its drawbacks. Besides the fun-damental planning problem discussed above, there is the additionalproblem in the context of a deregulated generation market that both gen-erators and load will constantly lobby regulators for more or less transmis-sion. If the fundamental planning problem is not solved, some generatorswill lobby for more transmission to get a free ride to market, while otherswill argue for less to keep their local price up and allow them to exercisemore market power. Load will argue for more transmission from cheapregions to expensive ones, not just to save production costs and dampenmarket power (legitimate reasons), but also to exercise monopsony poweragainst generation in the high-cost regions. For years to come, transmissioninvestment appears to be the knottiest problem in the deregulation process.

NOTES

1. Returns to scale imply that transmission investment costs are non-convex. Lumpinessrefers to having to buy an integer number of transmission lines selected from a small setof available capacities, but it is better understood as simply referring to a cost functionwith a fluctuating slope.

2. This is not as bad as it sounds, because with growth, almost any line that saves more thanits rental cost at the time it goes into service will continue to be economic in the long run.

3. In fact it requires even more. Load is to a small extent determined endogenously by theprice of power, so it should be determined simultaneously, but this complication will beignored.

4. This section concerns strategic manipulation by generation, but it assumes that thewholesale generation market is in every respect competitive. Manipulation is not theresult of market power, but of generation’s first-mover advantage in the optimizationgame.

5. This assumes that the set of feasible transmission flows is convex, which is not quite true,and that the feasible set of rights is based on average feasible power flows and not thepower flows that are possible only under ideal conditions.

6. From PJM’s document ‘FTR market frequently asked questions’, updated February 1,2005, question 49.

7. Found on www.pjm.com/markets/ftr/ftr.html, the website of PJM Interconnection, aregional transmission organization (RTO) that coordinates wholesale power trading in


a region of the US from Michigan to Washington DC, to Kentucky. It includes44 million customers and 135,000 MW of generating capacity.

8. On the left is the cost of $16/MWh, but it falls linearly to zero before reaching the rightend of the lump.

9. For completeness the cost of replacing losses should be included in CE.10. For example if profit�1, then net operating revenue, R – CE – CLL, equals CT1, which

means that transmission investments are earning more than a normal rate of return. Ifprofit is cut in half, operating revenues still cover CT1–

2, and investments still earn more

than a normal rate of return.11. Economic profit is profit above the normal rate of return on invested capital.12. Wilson (1997) addresses reliability in a franchise transco context.13. Stability ratings, though more firmly based in physics, also involve subjective judgments

as to how close is too close to the point of instability.

REFERENCES

Barthold, L.O. (2003), ‘Whither EHV? Distributed generation reverses the trend’,IEEE Power and Energy Magazine, 88, 85.

Borenstein, S., J. Bushnell and S. Stoft (2000), ‘The competitive effects of transmissioncapacity in a deregulated electricity industry’, Journal of Economics, 31 (2), 294–325.

Brunekreeft, G. (2004a), ‘Regulatory issues in merchant transmission investment’,Cambridge Working Papers in Economics, CWPE 0422, and Cambridge–MITWorking Paper, 38.

Brunekreeft, G. (2004b), ‘Market-based investment in electricity transmission net-works: controllable flow’, Utilities Policy, 12 (4), 269–81.

Brunekreeft, G. and T. McDaniel (2005), ‘Policy uncertainty and supply adequacyin electric power markets’, TILEC Discussion Paper, DP 2005-006. Tilburg Lawand Economics Center. Forthcoming in revised and shortened version: OxfordReview of Economic Policy.

Brunekreeft, G., K. Neuhoff and D. Newbery (2004), ‘Electricity transmission: anoverview of the current debate’, Cambridge Working Papers in Economics,CWPE 0463, and Cambridge–MIT Working Paper, 60.

Bushnell, J. and S. Stoft (1996), ‘Electric grid investment under a contract networkregime’, Journal of Regulatory Economics, 10, 61–79.

Bushnell, J. and S. Stoft (1997), ‘Improving private incentives for electric grid invest-ment’, Resource and Energy Economics, 19, 85–108.

Chandley, J.D. and W.W. Hogan (2002), Independent Transmission Companies in aRegional Transmission Organization, Cambridge, MA: Center for Business andGovernment, John F. Kennedy School of Government, Harvard University.

Crew, M., P.R. Kleindorfer and M. Spiegel (2004), ‘Reliability, regulation and trans-mission investment’, Risk Management and Design Processes Center, WorkingPaper 04-20-PK, Philadelphia.

Firecone Ventures Pty Ltd (2003), ‘Regulatory and institutional framework fortransmission (Australia): Final Report’, Victoria, November.

Gans, J. and S. King (2000), ‘Options for electricity transmission regulation inAustralia’, Australian Economic Review, 33 (2), June, 145–61.

Gilbert, R., K. Neuhoff and D. Newbery (2004), ‘Allocating transmission tomitigate market power in electricity networks’, Journal of Economics, 35 (4),691–709.


Gribik, Paul R., D. Shirmohammadi, J.S. Graves and J.G. Kritikson (2004),‘Transmission rights and transmission expansions’, draft to IEEE Transactionson Power Systems, 2004.

Harvey, S., W. Hogan and S. Pope (1996), ‘Transmission capacity reservationsimplemented through a spot market with transmission congestion contracts’,mimeo, Harvard University.

Hogan, W.W. (1992), ‘Contract networks for electric power transmission’, Journalof Regulatory Economics, 4, 211–42.

Hogan, W.W. (1998), ‘Transmission investment and competitive electricitymarkets’, mimeo, Harvard University.

Hogan, W.W. (2002), ‘Financial transmission rights formulations’, mimeo, HarvardUniversity.

Joskow, P.L. (2003), ‘Remedying undue discrimination through open access trans-mission service and standard electricity market design’, AEI-Brookings JointCenter for Regulatory Studies, Regulatory Analysis, 03-1, Washington, DC,February.

Joskow, P.L. (2004), ‘Performance based regulation’, May (slides), Washington, DC.Joskow, P.L. and J.J. Tirole (2000), ‘Transmission rights and market power on elec-

tric power networks’, Rand Journal of Economics, 31 (3), 450–87.Joskow, P.L. and J.J. Tirole (2002), ‘Transmission investment: alternative institu-

tional frameworks’, mimeo, Toulouse.Joskow, P.L. and J.J. Tirole (2004), ‘Merchant transmission investment’, Journal of

Industrial Economics, 53 (2), 233–64.Léautier, T.-O. (2000), ‘Regulation of an electric power transmission company’,

Energy Journal, 21 (4), 61–92.Littlechild, S. (2004), ‘Regulated and merchant interconnectors in a Australia: SNI

and Murraylink revisited’, Cambridge Working Papers in Economics, CWPE0410, and Cambridge–MIT Working Paper, 37, May.

Rosellón, J. (2003), ‘Different approaches towards electricity transmission expan-sion’, Review of Network Economics, 2 (3), 238–69.

Rotger, J. and F. Felder (2001), ‘Promoting efficient transmission investment: therole of the market in expanding transmission infrastructure’, sponsored byTransÉnergie US Ltd, November, Westborough, MA.

Stoft, S. (1997), ‘The effect of the transmission grid on market power’, LawrenceBerkeley National Laboratory Working Paper, LBNL–40479, Berkeley, CA.

Stoft, S. (2002), ‘A proposal for long-run and short-run congestion managementin Alberta’, Before the Alberta Energy and Utilities Board, File No. 1803-4,March 4.

Sun, H., M. Sanford and L. Powell (2004), ‘Justifying transmission investment inthe markets’, Electricity Transmission in Deregulated Markets, December 15–16,Carnegie-Mellon University, Pittsburgh, PA.

Vogelsang, I. (2004), ‘Transmission pricing and performance-based regulation’,Electricity Transmission in Deregulated Markets, December 15–16, Carnegie-Mellon University, Pittsburgh, PA.

Wilson, R. (1997), ‘Implementation of priority insurance in power exchangemarkets’, Energy Journal, 18 (1), 111–23.


GLOSSARY AND SYMBOLS

CE Congestion cost. Excess cost of energy from dispatching out of meritorder because of transmission constraints.CR Congestion rent. Revenue from energy injections at nodal prices lessrevenue from energy withdrawals at nodal prices.CL Congestion cost to load. Excess cost of energy to load due to trans-mission constraints.CLL Cost of unserved (lost) load.CT The rental cost of the transmission system paid by a regulatedtransco. Includes the annualized costs of capital and maintenance.CRR Congestion revenue right which pays (PB – PA) Q, during timeperiods T.FTR PJM’s financial transmission right. The obligation variety is thesame as a CRR except that the revenues are adjusted for any revenuesurplus or insufficiency in total congestion rents. The option variety omitsthe negative payments possible with an obligation.G The transmission grid.ISO One of several ‘independent system operators’ that run markets inthe US. They have now switched status to become RTOs, (regional trans-mission organizations) in keeping with the Federal Energy RegulatoryCommission’s changing terminology.K The capacity in MW of a transmission line.NYISO The New York Independent System Operator.PA , PB Nodal energy prices at nodes A and B.PL Price in the local regionPJM The ISO now covers Michigan, Pennsylvania, Washington DC,Tennessee and more, with 44 million consumers and 135,000 MW of gen-erating capacity.PR Price in the remote regionQ The megawatt flow named in a CRR, or on a line.R The regulated price paid to a transco per MWh of delivered power.RTO Regional transmission organization. The new name for an ISO inthe US.T The time period covered by a CRR, for example, peak hours during 2010.Transco A regulated monopoly transmission company.Wi The net power withdrawal at node i.


5. Patterns of transmission investmentsPaul Joskow

1. INTRODUCTION

A transmission network with good performance attributes is essential tosupport well-functioning competitive wholesale and retail markets for elec-tricity. The transmission network allows decentralized generators, mar-keters, distributors and large consumers to trade power in competitivemarkets. It can expand the geographical expanse of competition amongpower suppliers, giving consumers access to lower-cost energy and operat-ing reserves. By expanding the geographic expanse of competition thetransmission network can increase the effective number of competitorsand reduce market power and thus prices. A well-functioning transmissionnetwork facilitates the entry of new generators to match demand andsupply efficiently at different network locations to achieve economic andreliability goals and supports the development of demand response optionsfor wholesale and retail market participants.

Electricity sector liberalization has not changed the physical constraintsor physical laws that govern reliable transmission network operation orits role in supporting economical supplies of electricity. The network muststill satisfy the same physical parameters and constraints (frequency,voltage, stability, coordination with interconnected networks) and providefor operating reserves to respond to uncertain realizations of demand andunplanned outages of equipment to maintain reliability and avoid majorlosses of load or a widespread network collapse. However, electricity sectorliberalization has necessitated changes in the organization of the electricpower sector and the tools available to operate the network economicallyand reliably and to stimulate investment in the network to reduce conges-tion and maintain the physical integrity of the network.

Implementing effective transmission investment policies has proven to beespecially challenging as countries liberalize their electricity markets. In theUS, transmission congestion has increased and barriers to needed trans-mission investment are perceived to be a growing problem. Transmissionline relief orders (TLRs) in the Eastern Interconnection have grown bya factor of five since 1998. Congestion charges in the traditional PJM

131

(Pennsylvania–New Jersey–Maryland) area grew by a factor of 10 between1998 and 2003. Congestion charges in the New York independent systemoperator (ISO) have more than doubled since 2001 (Joskow, 2005).Congestion has grown rapidly in Texas, Southern California, New YorkCity, and New England as well. At the same time, investment in new trans-mission capacity has lagged the growth in electricity demand and thegrowth in new generating capacity (Hirst, 2004). In Europe, as wholesalepower trading has grown, transmission congestion limits the geographicexpanse of competition, limits opportunities fully to exploit generatingcapacity with the lowest operating costs, has led to concerns about genera-tor market power within several countries (Newbery, 2004) and has createdreliability challenges. As market liberalization proceeds, there has been verylittle investment in inter-transmission system operator (TSO) transmissioncapacity in Europe or the US. Intra-TSO congestion is a growing problemin some European countries as well. Policy makers in many countries withcompetitive power markets are increasingly concerned about reliabilityproblems and reliability considerations, and associated engineering operat-ing and planning criteria are playing an increasingly important role at theinterface of wholesale market design, transmission pricing and transmis-sion investment policies.

In this chapter I discuss a number of issues associated with the creationof an institutional environment that supports the identification of andefficient investment in transmission infrastructure. I illustrate how thewholesale market and transmission investment frameworks have addressedthese issues in England and Wales (E&W) since 1990 and in the PJMregional transmission organization (RTO) in the US since 2000. I am led tothe following conclusions:

1. The simple models of transmission network congestion and invest-ment that are used by economists have little to do with the way trans-mission investment is actually planned and developed, and theassociated transmission services priced within the boundaries ofindividual TSOs today. Economic models and analysis need to beexpanded to better capture the factors that TSOs and regulators con-sider when they identify transmission investment needs, especially asthey relate to the implementation of reliability criteria used fornetwork investment planning and system operations.

2. The application of a set of complex electric power network models,engineering reliability criteria, and simulation studies using thesemodels guide almost all intra-TSO transmission investment that istaking place around the world today. Commonly used economicmodels of transmission networks and transmission investment


opportunities do not capture these reliability criteria or their applica-tion adequately, if they do so at all.

3. Policy makers in a number of countries have sought to distinguishbetween ‘reliability’ and ‘economic’ transmission investments. Theformer are conceptualized as being needed to meet engineering reli-ability criteria while the latter are conceptualized as being developedto reduce congestion costs (and losses). These two categories ofinvestment are often treated as if they are distinct and independent.This is nonsense. ‘Reliability’-driven transmission investments are notindependent of the variables thought to create the need for ‘economic’transmission investments. Reliability investments can have significanteffects on current and forecast locational marginal prices (LMPs) forenergy and operating reserves, can have significant effects on intra-TSO congestion and losses, and can affect inter-TSO transmissioncapacity, congestion and losses as well.

4. Changes in network operating practices, TSO discretion in theprocedures used to evaluate whether and when reliability criteria willbe violated, and TSO discretion in the implementation of reliabilitycriteria in actual operating practices can have significant effects onlocational prices for energy and operating reserves, congestion costsand rents, the cost of losses, and incentives to invest in transmissioncapacity to reduce congestion. System operators need discretion tooperate transmission networks reliably. However, discretionary deci-sions affect the level and locational distribution of wholesale marketprices and the associated incentives to invest in both generating andtransmission capacity.

5. There are major asymmetries between the way intra- and inter-TSOtransmission investment planning, evaluation and pricing are imple-mented. Differences in inter- and intra-TSO transmission investmentframeworks reflect organizational and political boundaries, as wellas the attributes of the legacy networks controlled by incumbentTSOs, rather than the physical attributes of the larger synchronizednetwork, portions of which are controlled by individual TSOs. Inter-TSO investment opportunities can best be addressed through widerarea planning using a common set of reliability criteria and evalu-ation principles and by integrating wholesale power markets and har-monizing the principles for setting transmission service prices acrosscontrol areas to support either regulated or merchant transmissioninvestment.

6. Horizontal integration of previously independent TSOs can havesignificant effects on network operations, generator dispatch andLMPs for energy and operating reserves, congestion costs and

Patterns of transmission investments 133

incentives to invest in transmission facilities to meet reliability andeconomic goals by internalizing inter-TSO inefficiencies under alarger geographic TSO umbrella.

7. Merchant transmission investment has and is likely to make a verysmall contribution in the overall portfolio of transmission investmentprojects that will be made in the future. The merchant model thatseems to be evolving is one in which either regulated entities (andultimately their customers) take on the risk of entering into long-termperformance contracts with a developer of a high-voltage directcurrent (HVDC) transmission link to expand ‘interconnection’ capac-ity between TSOs or in situations where there are very limited inter-connections between TSOs with large sustained differences in prices,where market participants are willing to enter into long-term trans-portation contracts in return for firm rights and where the intercon-nector is sized so as to have a small impact on the difference inlocational prices.

8. In addition to the problems with relying primarily on a merchanttransmission investment model discussed in Joskow and Tirole(2005b, forthcoming), the sensitivity of locational prices for energyand operating reserves and associated congestion rents and costs toregulated investments in ‘reliability’-driven transmission projectscombined with discretionary changes in TSO implementation ofoperating reliability rules create significant additional barriers tointra-TSO merchant transmission investment.

9. The interconnection rules and associated cost responsibilities govern-ing the interconnection of new generators and interconnections ofnew inter-TSO transmission links to a TSO’s internal network havesignificant effects on locational incentives faced by new generatorsand on both the economic attractiveness and the economic efficiencyof merchant transmission investment projects. ‘Deep’ interconnectionrules that require market participants to pay the relevant intercon-nections costs at different locations, and the associated allocation ofcost responsibilities or interconnection service prices, provide super-ior locational incentives to ‘shallow’ interconnection rules and inter-connection prices that do not vary by location.

10. Most transmission investment projects are being developed today andwill be developed in the future by regulated entities. Accordingly, thecreation of a sound, stable and credible regulatory framework togovern regulated transmission investments is very important. Theabsence of such a framework for the identification of transmissionneeds, for transmission cost recovery, for mechanisms to aligninvestor incentives with public interest goals, and for efficient pricing


of the associated transmission service is a major barrier to the efficientmobilization of transmission investment. An attractive regulatoryframework will accommodate but not rely on merchant transmissioninvestment.

11. There exists no single mechanical ‘silver bullet’ incentive regulationmechanism that can be developed to govern transmission investment.A practical regulatory framework will inevitably include a mix ofcost-of-service regulation with an overlay of performance-basedregulation mechanisms based on benchmarking, profit sharing(sliding scale) and ‘ratchets’ (see Chapter 4 by S. Stoft). The develop-ment and application of performance norms, formal investment cri-teria, as well as considerable regulatory judgment is an inevitablecomponent of a sound regulatory process. One component of such aregulatory framework is a transparent regional transmission invest-ment planning process with clear rules for achieving defined reliabil-ity and economic goals.

12. The bifurcation of regulatory responsibilities in the US between thestates and the federal government (Federal Energy RegulatoryCommission: FERC) and in Europe between Brussels and the memberscountries creates significant potential disincentives to transmissioninvestment in what is only a partially liberalized sector. Full unbundlingof transmission service and the transfer or harmonization of regula-tory responsibility for all transmission service to federal authoritieswould be very desirable.

13. In order to implement an effective regulatory process, regulators willneed more information about the performance of the transmissionnetwork, will have to establish performance norms and criteria, andapply performance-based regulatory (PBR) systems that align TSOincentives with public interest performance goals. These incentivemechanisms must satisfy firm viability/participation constraints andreflect rent extraction goals in the context of information asymme-tries between the regulator and the firms it regulates. Chapter 4 dis-cusses incentive regulation in more detail.

14. TSOs that are also vertically integrated into generation and market-ing activities create additional regulatory challenges because of theconflicts of interest between operating and investment decisions madeby the TSO and their impacts on the profitability of generation andmarketing businesses that make use of the same transmissionnetwork. Regulatory rules requiring ‘functional’ separation eliminateany benefits of vertical integration if they are followed while provid-ing imperfect protection against abusive self-dealing behavior by theTSO. The creation of truly independent TSOs reduces the regulatory


burdens and creates entities whose management is focused on thetransmission business.

15. Separating SO and TO functions may be a second-best response tovertical integration between transmission, generation and power mar-keting, but it also is likely to lead to some inefficiencies.

2. ATTRIBUTES OF TRANSMISSIONINVESTMENTS

Transmission investment policies must respond to a number of interde-pendent questions. What are the societal goals that a transmission invest-ment framework should seek to achieve? What are the respective roles ofeconomic goals, reliability goals and other potential public policy goals?What are the physical and economic attributes of different types of trans-mission investments? How are transmission investment needs identified?What entities are expected to develop the new facilities? How are the asso-ciated costs expected to be recovered through transmission charges or pricearbitrage profits resulting from transmitting power from a relatively highwholesale price location to a lower wholesale price location? Which entitiesthat make use of the network should pay for its various components?Where does ‘transmission’ end and ‘distribution’ begin?

While policy makers talk about ‘transmission investment’ in general, inreality those responsible for identifying investment needs and opportunitiestypically divide transmission investment into a number of different cate-gories. If we are going to make progress in understanding the transmissioninvestment problem from a theoretical and empirical perspective, we needto better coordinate economic analysis with the conceptual framework thatgoverns the consideration of transmission investment by system operators,transmission owners and policy makers.

Let me note as well that there is no uniform definition of the facilitiesthat make up the high-voltage transmission network that is subject to thecontrol of the system operator. In E&W, the transmission network licenseincludes only facilities with voltages of 275 kV and 400 kV. In the US, trans-mission facilities typically, though not always, include lines that operate at66 kV and above with various exceptions based on differences in networktopology, legacy ownership and regulatory arrangements. In France,Réseau de Transport de l’Electricité (RTE) transmission network includesfacilities with voltages similar to those in the US.1 Different rules may beapplied for defining which ‘side of the fence’ interconnection facilities lie.I shall ignore these differences in the definition of the facilities that com-prise the transmission network in different countries in what follows.


However, we should keep in mind that these differences complicate com-parisons of transmission network performance, since performance indicialike congestion costs, losses, network component availability, unservedenergy (loss of load), operation and maintenance costs and so on, willdepend on which network components are included in the definition of‘transmission’ and which are not.

Categorization of Transmission Investments

Different TSOs also categorize (and characterize) transmission investmentsin a variety of different ways, use a wide range of very different methods toassess charges to cover the capital and operating costs of transmission facil-ities, to cover the costs of congestion and losses, and to assign responsibil-ity for payments to cover the cost of investments in new transmissionnetwork facilities. In the discussion that follows, I shall make use of the fol-lowing categorizations which are broadly consistent with those used in thetransmission planning and investment frameworks in the US and the UK.

Generator interconnection investmentsWhen new generators are constructed they must have interconnections tothe transmission network in order to sell energy and ancillary networksupport services in the wholesale market. Some minimal level of invest-ment is required merely to connect the generator to the closest point ofinterconnection to the network and to allow the generator to deliver itsmaximum generating capacity to the network at this point of intercon-nection. At a minimum, these investments will include new (or reinforced)transmission lines between the generating plant’s switchyard and the firstpoint of interconnection to the high-voltage network and investments intransformer capacity at the point of interconnection to the network toaccommodate the reliable injection of additional power into the networkat the proper voltage. The investments required will vary directly withthe generator’s maximum capacity, the maximum capacity of proximategenerating facilities that share an interconnection point on the network,the voltage at which the power is delivered to the network, and the relia-bility of the interconnection facilities as measured by their planned (formaintenance) and unplanned outage rates under different system condi-tions.

Interconnection investments alone do not assure the associated gen-erator that there will be adequate transmission capacity to transmitthe power from the point of interconnection to the network on to otherlocations on the network without curtailments or additional charges dueto congestion. As a result, as I shall discuss in more detail presently, if the


generator expects to be able to utilize fully its generating capacity todeliver power to serve demand nodes dispersed around the networkwithout experiencing curtailments or incurring congestion charges,investments ‘deeper’ into the network are likely to be required.Alternatively, the generator or its customers will have to secure transmis-sion rights to utilize the scarce transmission capacity that already existsfrom others.

Distribution network and large retail customer interconnection investmentsDistribution networks and large customers who take power directlyfrom the transmission network must also have transmission facilities thatinterconnect them to the high-voltage transmission network. These inter-connection investments are the flipside of generator interconnection invest-ments except that distribution networks typically have multiple points ofinterconnection with the transmission network and individual loads’ loca-tional decisions will, in most cases, be insensitive to interconnection costs.At a minimum, these investments will include new (or reinforced) trans-mission lines between the distribution network’s facilities and the first pointof interconnection to the high-voltage network and investments in trans-former capacity at these points of interconnection. The investmentsrequired will vary directly with the distribution network’s maximum coin-cident demand, the number and attributes of interconnection points, thevoltage at which the power is delivered to the distribution network beforebeing further stepped down by the distributor, and the reliability of theinterconnection facilities as measured by their planned (for maintenance)and unplanned outage rates under different system conditions. Inter-connection investments per se do not assure the distributor that there willbe adequate ‘upstream’ transmission capacity to transmit all of the powerit needs to meet its end-use customers’ demand because there may be con-gestion between the distributor’s point of interconnection and generationnodes on the network under some operating conditions. However, a distri-bution company will not add interconnection capacity unless it can fill thatcapacity with energy drawn from the transmission network by securing, inone way or another, the network capacity ‘deeper’ into the network neededto gain access to enough energy to meet the demand of its distributionservice customers. Of course, the distributor may also be able to balancesupply and demand with generation embedded in the distribution network(distributed generation) and with load reduction programs, including theimpacts on consumer demand of real-time pricing or priority rationingcontracts (Chao and Wilson, 1987). As already noted, there is no well-accepted firm line between ‘distribution’ and ‘transmission’.


‘Intra-TSO’ economic transmission network upgrade investmentsBy intra-TSO, I mean investments made within the footprint of a specificTSO. The TSO may cover only a portion of a larger synchronized ACnetwork as in the US and Europe. Economic models of transmissionnetwork operations and investment focus on the effects of transmissioncapacity (whether in the context of a simple two-node network or a multi-node network with loop flow) on congestion costs and congestion rents.Congestion costs, congestion rents, differences in locational prices causedby congestion and the prices of congestion are discussed in Chapter 4.

Economic models of transmission expansion should, in principle, alsoinclude the cost of losses (Joskow and Schmalensee, 1983, pp. 36–7) in bothlocational prices and investment planning. And loss cost considerationsplay a significant role in traditional engineering–economic system planningmodels.2 However, perhaps for convenience, many contemporary economicmodels have ignored losses, although in the wholesale markets operation inNew York and New England, LMPs reflect both the marginal costs of con-gestion and the marginal cost of losses. Indeed, marginal losses lead tosignificant differences in LMPs in these markets even when there is no con-gestion. In what follows, when I refer to congestion costs I am using theterm to encompass the cost of losses as well.

So-called ‘economic transmission investments’ (whether intra- or inter-TSO) are motivated by the opportunity for such investments to reduce thesocial costs of congestion. Optimal economic investment involves a trade-off between investing in additional transmission and the associated reduc-tion in congestion (and loss) costs. That is, the incremental cost oftransmission investment should be compared to the incremental reductionin the cost of congestion (sometimes referred to as the ‘redispatch cost’) onthe network (Joskow and Tirole, 2005b). In the absence of ‘lumpy’ invest-ments,3 and assuming that all nodal prices reflect the relevant marginalsocial opportunity cost at each node, it is optimal to make expenditures on‘economic’ transmission capacity up to the point where the marginal costof transmission investment is equal to the (expected) reduction in trans-mission congestion and loss costs. Since transmission investment is anexpenditure today that creates a long-lived asset and congestion is a flowthat depends on future supply and demand conditions in both the electri-city and input markets (for example, fuel prices), the benefits of an eco-nomic transmission investment are necessarily uncertain at the time theyare made and are realized over a period of many future years.

‘Inter-TSO’ economic investments (interconnectors between TSOs)By ‘inter-TSO’, I am referring to investments that are designed to increasetransfer capacity between two (or more) TSOs and to reduce congestion


between them. When TSOs operate portions (‘control areas’) of the samesynchronized AC network, the differences between intra- and inter-TSOeconomic transmission upgrades are primarily institutional, reflecting his-torical ownership structures, political boundaries and differences in whole-sale market design and regulatory mechanisms. The underlying physicalattributes of investments at different locations on the larger AC networkcontrolled by multiple TSOs are basically the same as would be the case ifthere were a single TSO for the entire network. That is, with a single TSO,inter-TSO investments would by definition become intra-TSO transmissioninvestments governed by the same market, regulatory and transmissioninvestment frameworks. However, differences in the market designs andtransmission investment frameworks of the multiple TSOs controlling por-tions of the same synchronized network, incompatibilities between theinstitutions governing interconnected TSOs, and various transaction costsresulting from horizontal separation that affect wholesale market pricesand congestion on both networks, are likely to affect transmission invest-ment decisions. It is frequently the case that intra- and inter-TSO trans-mission investments are treated – even conceptualized – very differently dueto these institutional differences rather than to basic physical and economicrealities.

Differences in market design and coordination between interconnectedTSOs on the same synchronized AC network can affect the economicattributes and evaluation of opportunities to expand transmission cap-acity to reduce congestion both between the TSOs’ networks and evenwithin their individual networks. This is the case, in part, becausedifferences in market design and network operating practices can affectlocational prices and dispatch decisions within both of the individualTSOs’ control areas. These effects are exacerbated when multiple TSOsadopt operating protocols that are based on fictional physical character-izations of the interconnected free-flowing AC network – for example,that a large synchronized AC network is really several separate networksconnected by radial lines with no loop flow and no congestion within eachTSO. Individual TSOs first tend to resolve congestion inside their net-works and then facilitate residual economic trades between networks.These policies tend to push congestion out to the borders between TSOsand reduce economic efficiency. While this is not a necessary result ofhaving multiple TSOs on the same free-flowing network, it appears to bea practical reality of the decentralized operating protocols adopted byindividual TSOs. A good example of this is the experience following theintegration of PJM West (Allegheny Power Systems) with the incumbentPJM system to the East in April 2002 when significant changes in con-gestion patterns and locational prices occurred.4


This suggests that the horizontal consolidation of TSOs into a singleTSO covering the larger geographic footprint of the real physical networkcould lead to very different evaluations of and incentives for economictransmission investments. By ‘internalizing’ wholesale market and trans-mission institutions under a single TSO, both the locational price and con-gestion patterns that drive economic transmission investments are likely tochange. Transmission upgrade evaluation policies as they relate to inter-TSO transmission investments are likely to change as well. As we shall see,inter-TSO economic network upgrade opportunities and intra-TSO trans-mission network opportunities may be evaluated very differently by TSOson the same AC network. The internalization of transmission investmentdecisions and the integration of wholesale market institutions are two ofthe primary motivations in the US for FERC’s efforts to create large RTOsthat consolidate the multiple control areas that now exist. Consolidatingpreviously separate control areas is expected to transform inter-TSO eco-nomic transmission investment opportunities into intra-TSO transmissioninvestment opportunities governed by a single transmission investmentframework, a common wholesale market design, and wider market areawith a set of fully coordinated locational prices.

There are, of course, situations where inter-TSO economic transmissioninvestments involve the creation or expansion of interconnections betweentruly separate AC networks. For example, by building HVDC interconnec-tors between two separate networks, opportunities to increase trades ofpower from high- to low-price areas can be exploited. The HVDC linkbetween England and France, the HVDC links between Quebec and NewEngland, and the HVDC link being constructed between Tasmania andVictoria, Australia are examples.5

Interconnection investments to support inter-TSO transmission linksBuilding or expanding an inter-TSO transmission facility (an ‘intercon-nector’ in European parlance) is only the first step in increasing tradebetween two TSOs whether they are on the same synchronized network orgovern independent networks. The new interconnector will withdrawpower under the control of one TSO and deliver it to the network con-trolled by the other. Facilities need to be constructed to affect the inter-connection with each network, just as would be the case for a generatorwith equivalent capacity located at the point of interconnection at thedelivery end or a large load located at the point of interconnection atthe withdrawal end of the new inter-TSO link. Moreover, just as in thecase of new generators, whether or not the interconnector capacity can befully utilized to deliver power to serve load depends on network conges-tion beyond the point of interconnection to each network and how scarce


transmission capacity on the rest of each network is allocated. Inter-connectors may also have reliability implications, especially when they arerelatively large and become binding contingencies that affect the evaluationof whether the network is meeting established reliability criteria. However,unlike a generator seeking to locate on a single network, a proper evalua-tion of the value of and incentives to invest in an interconnector justifiedby the cost reductions realized by expanding use of low-cost power to dis-place the use of higher-cost power (plus the change in total net surplusresulting from lower prices and increased demand on the importingnetwork) will depend as well on the compatibility of the interconnectioninvestment policies and the ‘deeper’ network upgrade policies on both net-works. Some TSOs ‘socialize’ the costs of these deeper network upgradesinto a general ‘postage stamp’ transmission service tariff rather than requir-ing generators or interconnectors, causing the need for additional ‘deep’network investments to pay for them. This is called a ‘shallow interconnec-tion’ pricing policy. In other TSOs, the costs of deeper network investmentsrequired to restore reliability parameters and/or relieve congestion arecharged to generators and interconnectors at the locations where powerflows cause the need for these deeper network investments. This is called a‘deep interconnection’ policy. As discussed further below, PJM has a defacto deep interconnection policy while most other TSOs in the US haveshallow interconnection policies. In E&W, the use of system charges varyby location and are, effectively, a deep interconnection pricing policy.

Reliability transmission network investmentsThese are transmission investments that must be made to restore exo-genously specified TSO planning reliability criteria that may be violated asa consequence of changes in demand patterns, generation investment andgeneration retirements. (Planning and operating reliability criteria are nottypically the same.) TSO reliability criteria have generally been carried overfrom the old regime of regulated vertically integrated monopolies. As Ishall illustrate with several examples below, virtually all of the transmissioninvestment underway today in the US and, effectively, in E&W are eitherdirect interconnection investments as discussed above or some type of ‘reli-ability’ investment. I am informed that this is the case in many other coun-tries as well. One’s first reaction might be that this is a terrible situation. Itsuggests that current transmission investment frameworks consider onlyreliability and ignore the economic costs of congestion! However, while‘reliability’ and ‘economic’ transmission investments are often treated as ifthey were distinct and independent types of transmission investments, thisis a complete fiction. Investments made to restore engineering reliability cri-teria can have very significant impacts on congestion and locational prices


and, accordingly, on the valuation of and incentives for ‘economic’ trans-mission investments. Similarly, ‘economic’ investments can have impacts onreliability parameters.

Neither reliability transmission investments nor the interrelationshipbetween reliability criteria and economic parameters are given much atten-tion in the literature on competitive electricity markets or transmissioninvestment. Yet so-called reliability investments are playing an increasingrole in the overall intra-TSO investment profile and exacerbate incompati-bilities between inter- and intra-TSO transmission investment. The engin-eers and the economists interested in transmission investment issues clearlyneed to be introduced to each other. These issues will be discussed in moredetail after the case studies of E&W and PJM are presented in Section 7,below.

Physical Attributes of Transmission Network Components

The standard metaphor for transmission investment is the construction ofa major new transmission line on new rights of way. While major new trans-mission lines can cost hundreds of millions of dollars, many socially desir-able projects are relatively inexpensive and do not require expanding thegeographic footprint of the network. These latter investment opportunitiesare especially important in a world where the construction of major newlines is constrained by ‘Nimby’ (‘not in my backyard’) constraints. The dis-tribution of project costs for transmission investment projects identified ina recent New England ISO transmission plan is indicative of the patternsof transmission investment opportunities. The 2004 transmission planincludes 245 projects with a total expected cost of $2.1 billion. The five mostexpensive projects are projected to cost $1.4 billion and the remaining 240projects a total of about $700 million (ISO New England, 2004). The fulldistribution of ‘reliability’ project costs in the New England transmissionexpansion plan is displayed in Table 5.1. Of the roughly 50 transmission


Table 5.1 Reliability upgrade projects: New England regional expansionplan 2004 ($ million)

Projects Total cost Average cost

Top 5 1,388 277.6Next 5 322 64.4Next 15 296 19.7Remaining 220 132 0.6

Source: ISO New England, (2004).

projects listed in PJM’s ‘economic’ transmission investment market windowin November 1994 (discussed in more detail below) estimated investmentcosts vary from $20,000 to $39 million. These investments all seem to belumpy in the sense that they mitigate the congestion identified completelyand could not be financed out of the residual congestion rents.

Projects to enhance transmission networks include a wide range of physi-cal components that are to be added to the network or to replace compo-nents that are already in the network. They include:

● new relays and switches;● new remote monitoring and control equipment;● transformer upgrades;● substation facilities;● capacitor additions;● reconductoring of existing links;6

● increasing the voltage of specific sets of transmission links;● new transmission lines on existing corridors; and● new transmission lines on new corridors (above or underground).

In addition, the effective capacity of the network may be increased atlittle or no cost with the adoption of better remedial action schemes orspecial control schemes that increase the speed with which other transmis-sion links or generating plants can respond to unplanned equipmentoutages. Changes in operating practices and the way contingencies areevaluated and handled when they occur can also magically increase (ordecrease) effective transmission capacity.

The diversity of network components that can be added to or substitutedfor existing network components reflects in part the factors that limit trans-mission capacity. On most networks, transmission limitations are driven byreliability criteria and associated assessments of the ability of the networkto physically balance supply and demand without shedding load involun-tarily or violating network voltage, frequency or stability criteria that wouldincrease the probability of a network collapse. These reliability criteria typ-ically reflect the objective of keeping the probability of involuntary loadshedding to a very low level and the probability of a widespread networkcollapse to zero. The limitations on utilization of the network are fre-quently one or more sets of ‘contingency’ constraints evaluated under avariety of system conditions (‘study conditions’) that maintain the proba-bilities of load shedding or network collapse to acceptable levels ratherthan binding pre-contingency thermal limits on particular lines. Thebinding constraint limiting transmission capacity could be the reliability ofa breaker, the speed with which a switch can be pulled, or the ability accu-


rately to monitor line sag in real time. Better or faster communicationsbetween system operators controlling portions of the same synchronizedAC network can also relax contingency constraints and increase theeffective capacity of the network. Accordingly, when we think aboutexpanding transmission capacity we should have in mind the full range ofphysical and behavioral options as well as the importance of engineeringreliability criteria and associated contingency studies and constraints.

Note that the discussion in this subsection also implies that measuringtransmission ‘capacity’, or changes in transmission capacity, using meas-ures of the length of transmission lines – for example, MW miles – is notappropriate. Especially in light of the difficulties of siting major new trans-mission lines, increases in transmission capacity are likely to focus on ‘deep-ening’ the existing transmission infrastructure and minimizing theexpansion of its geographic footprint. When new lines are necessary, sitingdifficulties will also lead to more underground links and the use of morecostly routes to avoid environmentally and politically sensitive areas.

Legacy Infrastructure Considerations

It is important to recognize that electricity sector liberalization reforms takeplace with an existing infrastructure composed of long-lived assets with par-ticular attributes. The attributes of the legacy infrastructure reflect histori-cal institutional arrangements, corporate boundaries, political boundaries,historical patterns of urban and industrial development, and historical eco-nomic and technological opportunities. The attributes of this legacy infra-structure will affect the behavior and performance of the system for manyyears into the future. We can change the institutions but we cannot erase theexisting infrastructure in place at the time sector liberalization reforms areimplemented but only change it gradually over time.

For example, in the US the electric power sector evolved with a largenumber of vertically integrated utilities serving geographic areas thatvaried widely in size. This structure was significantly influenced by federaland state laws passed during the 1930s that sharply restricted mergers ofproximate utilities, especially when they served more than one state.Infrastructure development focused most intensively on the geographicareas served by individual utilities with transmission networks developedto link generators owned by the utility with the load centers within theutility’s geographic franchise area. The strengthening of the transmissioninfrastructure connecting vertically integrated utility control areas pro-ceeded later and more slowly. In many cases it was motivated primarily byreliability considerations rather than with the goal of importing largeamounts of power from neighboring vertically integrated utilities (Joskow


and Schmalensee, 1983). So, for example, New England has only limitedtransmission interconnections with New York State (about 1,500 MW con-necting two networks with peak loads of about 29,000 MW and 35,000MW, respectively). This reflects much more the ownership structure of util-ities in this area of the US during the last half of the twentieth century(there was no common ownership between utilities serving areas in bothNew York and New England while some utilities in New England had oper-ating companies and generating facilities in two or more New Englandstates) and historical political boundaries (the New England states joinedtogether to form the New England Power Pool in 1969 while New Yorkcreated its own power pool at about the same time) than it does any naturaleconomic and technological attributes. Similarly, large integrated utilityholding companies like AEP and Southern developed strong transmissionnetworks covering several states in which they had operating companieswhile small independent vertically integrated utilities in other areas of thecountry have weak interconnections with their neighboring utilities and, asa result, enter the liberalization with weak regional networks.7

In Europe, where several countries relied on one or a small numberof vertically integrated utilities, or as in Spain, consolidated responsibilityfor a ‘shared’ high-voltage transmission network, there tend to be muchstronger ‘intra-country’ than ‘inter-country’ transmission networks. Thishas led European transmission policy to focus on expanding ‘interconnect-ors’ between countries rather than on intra-country wholesale marketdesign, locational pricing and transmission policies, sometimes using theargument (almost certainly wrong) that the national networks are so strongthat there is no internal congestion. In Italy, for example, there are severalcongested interfaces, in addition to the congested transmission interfaceswith France, Switzerland, Austria and Slovenia. Clearly the attributes ofthe legacy infrastructure are likely to have significant implications for theneed for additional transmission investment to support competitive whole-sale power markets.

Dimensions of Transmission Network Performance

While I am focusing here on transmission investment, transmission net-works have multiple and interrelated performance dimensions. The designof supporting organizational, regulatory and market institutions and judg-ments about the overall performance of the transmission network shouldtake them all into account. These performance attributes include:

● Costs of congestion, losses and ancillary network support services.● Network operating and maintenance costs.


● Availability of network components and efficiency of outage restor-ation in response to congestion and loss costs.

● Reliability of the network – involuntary losses of load and networkcollapse.

● Costs of market power and other market inefficiencies affected by theoperation of and investment in the network.

● Efficiency with which the investment framework mobilizes invest-ment to expand the ‘intra-SO’ network to meet reliability and eco-nomic goals.

● Efficiency with which the investment framework mobilizes capital toexpand ‘inter-TSO’ transmission capacity to meet reliability and eco-nomic goals.

● Efficiency with which innovations in ‘software’ and ‘hardware’technologies are adopted for improving all aspects of networkperformance.

3. TRANSMISSION NETWORK ORGANIZATION

Transmission network organizations have both vertical and horizontaldimensions and we see a variety of different vertical and horizontal struc-tures across countries. These organizational differences are likely to affectthe incentives to make transmission investments as well as how transmis-sion opportunities are evaluated. These include:

● Full vertical integration This model is characterized by vertical inte-gration between system operation, transmission ownership and main-tenance, generation, retail and wholesale marketing. In these situations,regulations governing non-discriminatory access to the network,non-discriminatory transmission pricing, and non-discriminatoryevaluation of and investment in transmission facilities are extremelyimportant but very difficult to implement satisfactorily. The fully inte-grated TSO has an inherent conflict of interest because its transmissionnetwork operating, maintenance and investment decisions affect thevalue of its generation portfolios and marketing businesses. Moreover,in such companies, the transmission business is likely to represent asmall fraction of the income of the enterprise as a whole, and, as aresult, transmission is less likely to be the primary focus of managementattention.

● Independent transco This model is characterized by the separation oftransmission network functions (SO and TO functions) from genera-tion and power marketing functions. This is the independent transco


model that has been adopted in England and Wales, Spain, NewZealand, Italy (soon) and France (if we ignore the Electricité deFrance (EdF) holding company affiliation). System operation,network maintenance and network investment are vertically inte-grated and can be managed in a coordinated manner by the transco.The conflict of interest inherent in an organization where the TSO isnot independent of market participants no longer exists and the firm’smanagement is now focused on the provision of transmission services.

● ISO This model is characterized by the separation of systemoperations from transmission facility ownership, investment andmaintenance, as well as from ownership of generation and marketingbusinesses. The independant system operator (ISO) does not own ormaintain transmission assets, but is responsible for scheduling and dis-patching generation and load in coordination with operating reliabilitycriteria and market rules, managing and enforcing procedures and rulesfor allocating scarce transmission capacity, interconnection arrange-ments, administering tariffs governing transmission service prices, andworking with TOs and other stakeholders on the coordination of main-tenance schedules and planning for new transmission investments tosupport changes in the demand for and supply of generation services.This is the model that has been or is being adopted in large portions ofthe US, Alberta, Argentina, Norway and other countries.

There are several rationales for creating a separate independent systemoperator rather than an independent transco. It may not be politically fea-sible to force the separation of transmission ownership from generationownership and marketing activities. An ISO is created to sit on top of thevertically integrated utilities to provide an independent network managerand tariff administrator to govern relationships between market particip-ants and the vertically integrated owners of the transmission network’s facil-ities. There may be geographically balkanized ownership of transmissionassets (either regulated or unregulated) and the horizontal integration oftransmission assets is deemed to be politically infeasible or undesirable,especially if merchant investment is expected to play an important role inthe system. The ISO can then manage a larger physical network with multi-ple transmission owners more efficiently than would be the case if each TSOoperated its own control area. Finally, it is sometimes argued that gener-ation and transmission ‘compete’ (that is, they are horizontally as well asvertically related) with each other, that even a transmission owner with nogenerating assets cannot be truly independent and will have incentives todiscriminate against generators on the network. In this case, an ISO that hasno direct interest in the financial performance of the owners of any of the


assets that comprise or utilize the transmission network will be ‘unbiased’.This naturally leads to the question of what the ISO’s objectives are andwhat incentives influence the monopoly ISO’s behavior and performance.

Other things equal I would expect different organizational arrangementsto have different performance attributes and to create different regulatorychallenges. I offer the following hypotheses:

● Vertical integration among transmission, generation and marketingcreates significant regulatory challenges to mitigate incentives to dis-advantage generation and marketing rivals. Moreover, since theregulatory response to vertical integration is typically to require func-tional separation of the SO/TO functions from generation and mar-keting and to apply regulations that are designed to force the firm tooperate as if its SO/TO functions are not affiliated with generationand marketing businesses, there are no remaining social benefits tovertical integration between SO/TO functions and generation, mar-keting and other unregulated lines of business that make use of theaffiliated transmission network. What is the point of continuingcommon ownership of entities that regulators are trying to ensurebehave completely independently?

● Vertical separation of system operations from ownership andmaintenance of transmission facilities is likely to make coordinationbetween system operations, network maintenance and outage restora-tion, and investment more costly than if the TO/SO functions werecombined. Moreover, to the extent that transmission owners also owngeneration and are engaged in power-marketing activities, it will bedifficult for the SO and the regulator to ensure that TO behavior, espe-cially related to maintenance, interconnection investment and invest-ment to reduce congestion, will not be affected by their impacts onaffiliated generation and marketing companies. Both the SO and theregulator will have imperfect information about the TO’s cost oppor-tunities, efforts and incentives. For example, one of the easiest thingsto accomplish is to fail to get a permit to build a new transmission linkthat will reduce congestion into an area where an affiliate owns gener-ating capacity. Indeed, I suspect that PJM’s hostility toward regulated‘economic’ transmission investment (more below) is not unrelated tothe fact that all of the transmission owners under the PJM umbrellaare also vertically integrated into generation and marketing.

● Limited horizontal expanse of SO functions is likely to createinefficiencies. The more control area operators there are on thenetwork, the more conservative will reliability criteria be, reducingthe availability of inter-TSO transmission capacity, and the more


difficult it will be for separate market areas efficiently to coordinatewholesale trading of power and the allocation of scarce transmissioncapacity. As I shall discuss presently, there are significant asym-metries between the framework governing intra-TSO transmissioninvestment and inter-TSO investment. Internalizing inter-TSO linksthrough horizontal integration is likely to lead to less congestion andmore transmission investment.

These are, of course, only hypotheses that should be verified throughempirical analysis.

4. PRINCIPLES TO GUIDE TRANSMISSIONINVESTMENT REGULATORY FRAMEWORKS

A sound transmission investment regulatory framework must address severalinterrelated issues. The following discussion reflects my view that the bulk ofintra-TSO transmission investment will be mediated through a regulatoryprocess of some type and that so-called merchant investment will play alimited role.8 Merchant investment of one type or another may play a largerrole in mobilizing investment for expansion of inter-TSO transmission facil-ities (interconnectors) as a result of various institutional and political con-straints. Merchant opportunities may emerge as well if incumbent TOs arepermitted to develop unregulated merchant projects on their own networks,exploiting the market power that they possess. I shall also assume that all ofthe TSO’s revenues come from entities that use the network; there are no gov-ernment subsidies, and a viable TSO, SO or TO must balance its budget.

1. Objectives and performance norms The regulatory framework mustspecify clearly what the regulator’s objectives are for the TSO (or SOand TOs if they are separated) – that is, what the TSO is expected toaccomplish – how the TSO’s performance will be measured, whatnorms and benchmarks will be applied to evaluate its performance,and what instruments the TSO may use to achieve these performanceobjectives. In the case of an organizational structure that separatesSO and TO functions, the division of responsibilities and mechanismsfor coordinating relationships between the SO and the TOs under itmust be clearly defined. As I shall illustrate presently, integratingso-called reliability goals and criteria with economic goals and per-formance norms is especially important.

2. TSO participation or viability constraints The regulatory frameworkmust recognize that there is a firm viability or participation constraint


that any regulatory mechanism must adhere to (Tirole and Laffont,1993). This ‘budget balance’ constraint can be defined simply as therequirement that any acceptable regulatory mechanism must have theproperty that expected revenues from the provision of transmissionservices must at least cover the costs that the regulated firm incurs toprovide these services. Private firms cannot be expected to offer tosupply services if they do not expected to be compensated for theassociated costs. State-owned firms cannot satisfy hard budget con-straints (no government subsidies) unless they can recover the cost ofproviding transmission services from transmission service revenues. Iftransmission service costs have non-convexities (for example, scaleeconomies), actual prices for transmission service must depart fromefficient prices. We are in the world of second best.

3. Rent extraction goals The flip side of the firm viability or participa-tion constraint is the impact of higher prices on consumers. Thehigher are the prices charged by the regulated firm the lower is thesurplus left to consumers and, where prices exceed their efficientlevels, the lower is aggregate welfare. In a world with asymmetricinformation, where the regulator has less information than does theregulated firm about its costs, it is well known that there is a trade-offbetween providing the firm with incentives to supply efficiently (costand quality dimensions) and rents left to the regulated firm fromcharges to consumers that exceed the firm’s costs of production(Laffont and Tirole, 1993). Over time, we would like to see the benefitsof lower costs flowing through to consumers as lower prices.

4. Incentive alignment Regulators have imperfect information about aregulated firm’s cost opportunities, service quality, managerial effort,consumer demand and other factors that influence the cost andquality of services provided by the regulated firm. Regulatory mech-anisms should be designed to reflect the asymmetry of informationavailable to the regulated firm and the regulator while making efficientuse of the information that is available to the regulator. The goal ofeffective regulatory mechanisms is to align the incentives faced by theregulated firm with the performance goals established by the regula-tor. This can be accomplished by (partially) tying the regulated firm’sprofits to its ability to meet or beat performance goals established bythe regulator. The power of such incentive schemes is necessarilylimited by the information that the regulator has about the basic costand demand conditions faced by the regulated firm as well as by firmviability constraints and rent extraction goals. Under all realistic situ-ations, the second-best regulatory mechanism will partially tie a regu-lated firm’s revenues to the actual costs that it incurs and partially


place the regulated firm’s profits at risk for variations in performance.This can be accomplished with a sliding-scale mechanism (profit-sharing formula), and/or with periodic ‘ratchet’ mechanisms thatrealign the firm’s revenues with its costs from time to time (seeChapter 2 in this book).

5. Incomplete contract considerations Regulatory frameworks can beviewed from a contractual perspective in which regulatory rules definethe terms and conditions of an incomplete contract between the regu-lator and the regulated firm. The regulatory contract also defines arenegotiation framework that allows the terms and conditions of thiscontract to be adjusted over time as supply and demand conditionschange (Joskow and Schmalensee, 1986). Investments in transmissionfacilities are long-lived assets that provide services for many years intothe future. While the costs of investments are incurred up front, therevenues that the firm will receive from these assets will be realizedfrom transmission service revenues extending over the life of the asset.On the one hand, once the investment is made, the regulated firm mustbe concerned that it may be subject to a ‘regulatory hold-up’ aimed atconfiscating the ex post quasi-rents created by the investments.Investors in regulated assets will seek a credible commitment thatsuch hold-ups will not occur. A credible full-contingent claim con-tract negotiated ex ante would be ideal from this perspective. On theother hand, the regulator is not in a position to define an efficient full-contingent claim contract ex ante that also satisfies a budget balanceconstraint. Over the life of regulated transmission investments supplyand demand conditions are likely to change considerably, affectingboth the profitability of the regulated firm’s investment and the rentsextracted from consumers. Moreover, the regulator will learn moreabout the attributes of the regulated firm, its costs, revenues and thequality of service over time as well. An effective regulatory process islike a good incomplete contract (Joskow, 1988). It defines initial termsand conditions, performance norms, formula adjustments to reflectchanging economic conditions, and an adjustment process that pro-vides an efficient framework for adjusting these terms and conditionswhen they fall outside of a ‘self-enforcing range’.

6. Transmission service price structures It is convenient to think aboutthe components of the regulatory framework above as establishingthe aggregate revenues (or profits) that the regulated firm can earnunder various contingencies. These ‘allowed revenues’ reflect firm via-bility, rent extraction and incentive alignment considerations or, tooversimplify, the regulated TSO’s current budget constraint is deter-mined first. Prices must then be established for the various services


that the TSO provides. These prices should provide efficient signals totransmission system users so that their behavior can adjust to reflectthe (marginal) costs of the services provided to them in the short andlong runs. They must also be set at levels that produce the aggregaterevenues (or profits) that the regulated firm is allowed to earn basedon the terms and conditions of the regulatory arrangements discussedabove.

7. Other terms and conditions of network access In addition to thespecification of the prices for using the transmission network, otherterms and conditions of service must also be defined. This is especiallyimportant when the TSO or the TO is not independent of market par-ticipants. These terms and conditions include the rules governing theprocess through which interconnection requests by generators or mer-chant transmission projects will be processed, specification of costresponsibility for interconnection and network reinforcements, theapplication of reliability criteria to evaluate the availability and costof providing transmission service, the specification and allocation ofphysical or financial transmission rights, and the mechanisms for allo-cating scarce transmission capacity in the short and long runs. Someof these terms and conditions are ultimately linked to the attributesof the wholesale markets that are supported by the transmissionnetwork.

8. Relationships between transmission and wholesale market institutionsIn the early years of electricity sector liberalization in the UnitedStates, Europe, Japan, Australia and other countries it was oftenargued by policy makers that there was a natural and fairly simple‘separation’ between competitive power markets and the transmissionnetwork that is necessary to support these markets. It is quite cleartoday that no such simple separation exists. Organizing powermarkets and transmission institutions as if a clear separation existsinevitably leads to serious problems. Efficient power markets, efficienttransmission operation and investment behavior, and the satisfactionof reliability goals at the lowest reasonable cost are all fundamentallyinterdependent. Competitive market prices for power (spot andforward) are signals of the value of both energy and transmissioncapacity at different locations. These price signals can be used to allo-cate scarce (congested) transmission capacity to highest-valued(lowest-cost) users, can enable consumers to express their willingnessto pay for ‘reliability’ and express their risk preferences regardingprice volatility, can allow generators to factor locational and timeseries differences in power prices into operation and investment deci-sions, can allow transmission networks to incorporate the costs of


congestion, the value of reliability and other factors into maintenanceand investment decisions and so on.

However, the social value of these price signals and the costs andbenefits of agents responding to them are only as good as theefficiency of the markets that produce them. Moreover, some of theattributes of electric power networks – for example, the possibility ofnetwork collapses – can make investments in ‘reliability’ a public good(Joskow and Tirole, 2006). Other market imperfections (for example,generator market power, lumpiness in investments, imperfectlydefined property rights) and regulatory interventions (for example,price caps, SO procurement behavior, non-price rationing – ibid.)affect the prices for generation and the value of scarce transmissioncapacity in both the short and the long runs and can distort ratherthan improve transmission operating and investments decisions.Accordingly, a well-functioning transmission network depends on thedesign and implementation of sound wholesale market institutions aswell as a sound regulatory framework (economic, reliability, networkplanning) for transmission network owners.

9. Transmission planning Transmission networks do not and will notevolve through the workings of the invisible hand of competitivemarkets. Even if one were to believe that all transmission investmentsshould be ‘market driven’ and developed by merchant investors, theimpacts of proposals for new transmission links on the network must,at the very least, be evaluated by the SO to define the attributes of theincremental network capacity that a merchant project creates and thecombinations of any incremental transmission rights that are consist-ent with the changes in the feasible set of power flows anticipated tobe created by the investment, whether the operation of the new facil-ities would lead to conflicts with existing transmission rights, and thespecific allocation of transmission rights that will be conveyed to thetransmission developer (Joskow and Tirole, 2005b). As I shall discusspresently, in the real world, entry (and exit) of generating plants andchanges in demand patterns affect both network congestion asreflected in simple economic models of transmission networks(Joskow and Tirole, 2000) as well as reliability constraints as definedby system planners and operators. Investment opportunities drivenby economic criteria and investment needs driven by reliabilitycriteria are highly interdependent. At least in the current state of play,a transmission planning process is required to evaluate someaspects of regulated reliability-driven transmission investments, regu-lated congestion cost-driven transmission investments, merchanttransmission investments and generator interconnection investments.


Transmission planning processes should be transparent, provide forstakeholder input, and reflect the objectives and norms defined byregulators for the transmission network.

10. Merchant transmission investment The regulatory framework,including the transmission planning process, should accommodateproposals for ‘merchant’ transmission investments. Merchant trans-mission investment was initially conceived as unregulated trans-mission investment projects that would be developed on anentrepreneurial basis in response to congestion (differences in loca-tional prices) between points on the same network or to differences inelectricity prices on different networks that the merchant project con-nects. Basically, merchant investors would recover their costs bybuying power at one end of a link where it is cheap and reselling it atthe other end where it is expensive; or selling the rights to use the mer-chant link to third parties to engage in this type of trading behavior.That is, the merchant investor makes money by arbitraging pricedifferences between the locations to which the merchant investmentcreates new transmission rights to buy and sell wholesale power.

The volume of talk about merchant investment far exceeds theinvestment activity of merchant investors, despite the fact that thetransmission frameworks in Australia, New England, New York, andPJM were designed to accommodate ‘market-driven’ investments for‘economic’ transmission investment opportunities. Two small mer-chant links have been developed in Australia which intended to earnrevenues and profits by arbitraging spot-price differences between thenetworks in adjacent states. Both projects have now applied for reg-ulated transmission status. A merchant underwater HVDC trans-mission link is being constructed between Tasmania and Victoria.However, the project is being developed in response to a request forproposals (RFP) from the state-owned electric power company inTasmania and will be supported by a long-term contract (whose costscan be recovered from the sponsor’s customers) between the sponsorand the developer. A competitive RFP process initiated by the munic-ipally-owned Long Island Power Authority (LIPA) supported by a20-year long-term contract governs the completed Long IslandSound HVDC link between Connecticut and Long Island, NewYork. A similar 20-year contractual arrangement is supporting aproposed HVDC project between PJM and Long Island. HVDCprojects linking PJM with New York City and between UpstateNew York and New York City (recently cancelled due to the failureto obtain financing) have been discussed for several years. That’sabout it.


The merchant model that seems to be evolving is one in which reg-ulated entities (and ultimately their customers) take on the risk ofentering into a long-term performance contract with an HVDC trans-mission link developer to expand ‘interconnection’ capacity betweennetworks with no or limited interconnections and large sustaineddifferences in prices that are not affected significantly by the additionof the link.9 Perhaps a better term for this model is ‘private initiative’transmission investments. It should be recognized as well, that thefinancing costs for a merchant project are significantly higher thanthose for an equivalent regulated project. A recent analysis of thefinancing costs for a $100 million merchant transmission project10

indicated that the cash flow required to finance a regulated projectdeveloped by a utility and subject to traditional cost of service regu-lation would be $9.4 million per year. The annual cash flow for thesame merchant project with a long-term contract (taking on con-struction cost and performance risk but not market price risk) usingproject financing was estimated to be $13.9 million per year. Theannual cash flow for the same merchant project without a long-termcontract (taking on, in addition, market price risk) using projectfinancing was estimated to be $16.5 million per year. Thus, thefinancing costs for a traditional merchant project that relies on varia-tions in spot market prices would be about 70 per cent higher than aregulated utility financed project. This capital cost variation suggeststhat the efficiency benefits of merchant versus regulated projectswould have to be quite large to justify relying on merchant investment.

Joskow and Tirole (2005b) identify ‘lumpiness’ as one barrier toefficient investment under a merchant transmission investmentmodel. ‘Lumpiness’ is a relative not an absolute size concept. That is,whether an investment project is lumpy or not must be measured rel-ative to the impact of the most efficiently sized project on the con-gestion rents that it would reduce. The post-investment congestionrents are the source of the revenue that a merchant investor wouldcount on to support the investment. Regardless of the absolute costof the project, if an efficient (benefits greater than costs) project ofoptimal scale were to eliminate congestion completely, for example,there would be no way for it to be financed under a merchant invest-ment framework. Similarly, a large project of optimal size (forexample, a 1,000-MW HVDC link to New York City) may not havesuch a large effect on price differences as to make the investmentuneconomical. Some commentators have suggested that the ‘lumpi-ness’ problem can be addressed by treating very large projectsdifferently from small projects. This policy prescription reflects a


misunderstanding of what ‘lumpiness’ means in this context. Indeed,this policy advice is likely to get it backwards. As we shall see in thediscussion of PJM’s transmission investment policies below, thereare many small projects that completely mitigate congestion and,accordingly, would not be financed on a merchant basis. At the sametime, the merchant projects that are attracting the most attention arelarge projects that link market areas with demands that are muchlarger than the scale of the projects and have significant sustainedcongestion and the associated locational price differences. Theselarge projects are small relative to the size of the markets that arebeing linked and, as a result, their completion is not expected to havea large effect on differences in locational prices.

While I view the opportunities for merchant transmission projectsas being limited primarily to inter-TSO investments that fall outsideof TSO regional planning procedures, where there exist large sus-tained price differences and where a regulated entity is willing toprovide long-term contract support for the project, there is no reasonwhy such projects should not be accommodated in the regulatory andplanning process. A practical model for doing so has emerged in PJMand I shall discuss it further below.

5. TRANSMISSION REGULATION ANDINVESTMENT FRAMEWORK IN ENGLANDAND WALES

In 1990, the electricity sector in E&W was privatized and restructured tocreate competitive wholesale and retail markets for power. The state-ownedgeneration and transmission company (Central Electricity GeneratingBoard: CEGB) that historically had provided wholesale power to distribu-tion entities (area boards) and large industrial customers in E&W wasbroken into three generating companies and a single regulated transmissioncompany (National Grid Company (NGC)). NGC owns the E&W high-voltage transmission network (400 kV and 275 kV facilities), maintains thenetwork and is responsible for making investments in it to meet its obliga-tions specified by various license conditions. It also is a joint owner withRTE (the French transmission operator) of a 2,000-MW HVDC transmis-sion link between France and E&W.11

There has been much written about the design and performance of thewholesale power markets in England and Wales (for example, Henney,1994; Wolfram, 1999; Sweeting, 2000). Accordingly, I shall provide only abrief description of these wholesale market arrangements. From 1990 until


March 2001, the wholesale market for power was built upon a mandatorybid-based pool which determined the economic dispatch and associateduniform market-clearing price for energy (and where applicable capacitypayments) for each of 48 thirty-minute periods each day. Generators wereeffectively provided with firm transmission service in the sense that if NGChad to dispatch generators out of bid merit order to deal with congestionand other network operating constraints it had to pay generators either toreduce their scheduled generation or to increase it. In March 2001, the NewElectricity Trading Arrangements (NETA) was introduced. NETAreplaced the mandatory pool with a new wholesale market design that wasstructured to encourage generators and load to enter into bilateral con-tracts and to minimize the amount of trade going through a ‘centralizedpool’. NETA requires generators and loads to submit generation anddemand schedules up to a short period before actual dispatch. These sched-ules became financial commitments on the part of generators and loads.NGC is then responsible for balancing the system using offers to buy andsell increases and decreases in real-time generation supplies mediatedthrough a pay-as-bid ‘balancing market’. NGC’s balancing responsibilityincludes real-time balancing of demand and supply for energy and man-agement of network congestion and other network operating constraints.Generators or load that voluntarily deviates from their schedules must(effectively) buy or sell energy in the balancing market. As before, genera-tors paying interconnection and use of system charges (below) areeffectively buying firm transmission service and must be compensated ifNGC needs to increase or decrease their output from the pre-scheduledlevels to manage congestion and other network constraints.

Among other things, NGC’s license conditions and associated codes andstandards specify the operating procedures and principles governingNGC’s relationships with all users of the transmission system (generators,distributors and retail electricity suppliers). Under its transmission license,NGC must operate the network in an efficient, economical and coordinatedmanner and offer its services based on non-discriminatory terms and con-ditions. Transmission System Security and Quality of Service Standardshave been developed to govern NGC’s responsibilities. These codes andstandards define reliability criteria that are to be used by NGC to planneeded enhancements to the transmission system and to identify transmis-sion investment requirements. NGC evaluates transmission investmentneeds and alternatives to meet these obligations on an ongoing basis.It publishes an annual Seven-Year Forward Statement12 which providesforecasts of demand, supply, approved transmission enhancements andexpected transmission enhancements that would be needed to accom-modate additional generation at various locations on the E&W grid. The


Seven-Year Statement is made available to provide information to new gen-erators regarding the capabilities of the network to accommodate new gen-erating capacity at various future dates and the network enhancements thatNGC has identified as being required to accommodate new generatingcapacity of various amounts at different locations on the network. TheConnection and Use of System Code (CUSC) specifies a contractualframework for interconnection to and use of the network. NGC is also thesystem operator for E&W and thus is vertically integrated into all aspectsof transmission operation, maintenance and investment.13

NGC is subject to regulation by the Office of Gas and ElectricityMarkets (Ofgem). Separate but compatible incentive regulation mech-anisms are applied to the transmission owner (TO) and system operatingfunctions (SO). These regulatory mechanisms effectively yield values forthe revenues NGC is permitted to earn from charges for transmissionservice and system operations. Transmission customers (generators andretail suppliers) pay NGC for transmission service pursuant to a regulatedtariff. The tariff has two basic components. The first is a ‘shallow’ connec-tion charge that allows NGC to recover the capital (depreciation, return oninvestment, taxes and so on) and operating costs associated with the facil-ities that support each specific interconnection (now using the ‘plugs’methodology). The second component of the transmission tariff is com-posed of the transmission network use of system (TNUoS) charges.

The general level of charges is set to allow NGC to recover its cost-of-service-based ‘revenue requirement’ or budget constraint as adjustedthrough the incentive regulation mechanism that I shall discuss presently.The structure of the TNUoS charges provides for price variation by locationon the network based upon (scaled) differences in the incremental costs ofinjecting or receiving electricity at different locations as specified in theinvestment cost-related pricing methodology. So, for example, generatorspay significantly higher transmission service costs in the North of Englandthan in the South (where the prices may be negative) because there is con-gestion from North to South and ‘deep’ transmission network reinforce-ments are more likely to be required to accommodate new generation addedat various locations in the North but not in the South. Similarly, load in theSouth pays more than load in the North because transmission enhancementsto increase capacity from constrained generation export areas benefit cus-tomers in the South more than those in the North. The locational TNUoScharges for 2004/2005 for generation and demand are displayed in Tables 5.2and 5.3. The stated objective of this pricing mechanism is as follows:

[E]fficient economic signals are provided to Users when services are priced toreflect the incremental costs of supplying them. Therefore charges should reflect


the impact that Users of the transmission system at different locations wouldhave on National Grid’s costs, if they are to increase or decrease their use of thesystem. These costs are primarily defined as the investment costs in the trans-mission system, maintenance of the transmission system and maintaining asystem capable of providing a secure bulk supply of energy. (NGC, 2004a, p.12)

Finally, in its role as system operator, NGC has an obligation to balancethe supply and demand for energy in the system in real time (energybalancing) and to meet operating reliability criteria (system balancing).These costs include the net costs NGC incurs to buy and sell power in thebalancing market (or through short-term bilateral forward contracts) tobalance supply and demand at each location, including to manage conges-tion, provide ancillary services, and other actions it must take to meet thenetwork’s operating reliability standards. These costs are recovered from


Table 5.2 Schedule of transmission network use of system generationcharges (£/kW), 2004/2005

Generation Zone Generation Short-term generation tariff (£/kw)zone area tariff (£/kW)

STTEC STTECC STTECperiod � period � period �28 days 35 days 42 days

1 Northern 9.009237 1.891940 2.364925 2.8379102 Humberside 5.767201 1.211112 1.513890 1.8166683 North West 6.222266 1.306676 1.633345 1.9600144 Pennies & North 4.121912 0.865602 1.082002 1.298402

Wales5 Dinorwig 10.715347 2.250223 2.812779 3.3753346 Anglesey 7.011370 1.472388 1.840485 2.2085827 East Anglia 2.889748 0.606847 0.758559 0.9102718 West Midlands 2.032089 0.426739 0.533423 0.6401089 South Wales & �2.150590 0.000000 0.000000 0.000000

Gloucs10 Oxon & Bucks 0.004330 0.000909 0.001137 0.00136411 Estuary 1.733641 0.364065 0.455081 0.54609712 Central & SW �6.604821 0.000000 0.000000 0.00000

London13 South Coast �1.507146 0.000000 0.000000 0.00000014 Wessex �3.829097 0.000000 0.000000 0.00000015 Peninsula �6.836065 0.000000 0.000000 0.000000

Note: STTEC � short term tariff of electricity.

Source: NGC (2004b).

system users through an ‘uplift’ charge based (mediated through an incen-tive regulatory mechanism discussed further below) on the quantities ofenergy supplied to or taken from the network.

The regulatory framework for determining the revenues that NGC canrecover through the use of system charges and the energy and system bal-ancing charges is based on a set of incentive regulation mechanisms. Thesemechanisms have a cost-of-service base, a performance-based incentive, anda ratchet that resets prices from time to time to reflect NGC’s realized orforecast costs. A base annual aggregate ‘revenue requirement’ for use ofsystem charges is established at the beginning of each five-year ‘price review’period (though the latest period is being extended to seven years). The start-ing revenue requirement is determined based on a fairly standard cost ofservice principles. A rate base (regulatory assets value: RAV) is defined thatis composed of the carrying value for the existing assets that make up thetransmission system plus the forecast cost of incremental capital expendi-tures budgeted for the next five years to meet NGC’s interconnection andsystem security criteria described above. The final investment budget isdetermined by Ofgem through a public consultation process. Depreciationrates and a cost of capital (allowed rate of return) are defined and appliedto the RAV to yield allowed capital charges for the starting year. Currentallowable operating and maintenance (O&M) expenditures are defined and


Table 5.3 Schedule of transmission network use of system demand charges(£/kW) and energy consumption charges (p/kWh), 2004/2005

Demand zone Zone area Demand tariff Energy consumption(£/kW) tariff (p/kWh)

1 Northern 4.940866 0.6565852 North West 8.325173 1.1002543 Yorkshire 8.455923 1.1716114 North Wales 8.709914 1.107068

and Mersey5 East Midlands 10.771600 1.4794246 Midlands 12.600874 1.7334137 Eastern 11.007104 1.3949348 South Wales 16.130442 2.2280759 South East 14.321101 1.773924

10 London 16.761568 2.43027711 Southern 15.679987 2.07648912 South Western 17.798154 2.198679

Source: NGC (2004b).

added to the year one capital charges. A target rate of productivity improve-ment in O&M charges – the ‘X’ factor – is then defined. The value of X isdetermined through a regulatory consultation process based on NGC’s fore-casts of O&M requirements, wage escalation, and various benchmarkingstudies performed for Ofgem by independent consultants. The starting valuefor allowed capital charges is then adjusted each year for budgeted incre-mental capital additions and changes in an inflation index while allowancesfor O&M costs are escalated based on a general price index minus ‘X’.Unbudgeted capital expenditures during the price review period can be con-sidered in the next price review, though NGC may be at risk for amortiza-tion charges during the period between reviews. Underspending on capitalmay also be considered in the next price review and adjustments made goingforward. After a five-year (or longer) period another price review is com-menced, the starting price is reset to reflect then-prevailing costs, and newadjustment parameters defined for the next review period.14

In its role as the E&W system operator, NGC has also been subject to aset of incentive regulation mechanisms. Each year forward targets areestablished for the costs of energy and system balancing services. A sharingor sliding scale formula is specified which places NGC at risk for a fraction(for example, 30 per cent) of deviations from this benchmark (up or down)with caps on profits and losses. Table 5.4 displays the attributes of the SOincentive mechanism in effect since NETA went into operation. However,Ofgem and NGC were unable to agree on a new SO incentive mechanismfor 2006/2007 and a traditional cost of service recovery mechanism wasadopted as a default. Ofgem has also applied a new incentive regulationmechanism that would apply to network outages that lead to variations inthe fraction of ‘lost energy’ resulting from transmission network outages(Ofgem, 2004b).

This brings us finally to the transmission investment framework. NGChas the obligation to identify transmission investments required to meet itsobligations under the Grid Code, the Transmission System Security and


Table 5.4 E&W system operator incentive mechanism under NETA

Parameter First year Second year Third year

Target expense (£m) 484.6–514.4 460 416Upside sharing (%) 40 60 50Downside sharing (%) 12 50 50Cap (£m) 46.3 60 40Floor (£m) �15.4 �45 �40

Source: Ofgem (2003b).

Quality of Service Standards, and the Connection and Use of SystemCode. The Transmission System Security and Quality of Service Standardsare engineering reliability criteria used for planning purposes that havelargely been carried over from the pre-restructuring era. They are of fun-damental importance for transmission investment planning purposes. Thetransmission planning process is built around a set of reliability criteriadesigned to meet these security and quality of service standards.

The Standards specify criteria (to oversimplify) for defining a set of‘boundary circuits’ and associated power flows over which the generatingcapacity on one side of the boundary must be able to flow reliably (thermal,voltage and stability) over the boundary to serve demand there if any twocircuits are out of service. NGC performs power flow studies based on fore-casts of demand and generating capacity at various locations to identifyboundaries (individually or collectively) where reliability criteria may beviolated during the forecast period. Transmission investment projects arethen identified which will restore the relevant reliability criteria when andif they are expected to be violated. Depending on the nature and magnitudeof the transmission investments identified, various ‘siting’ approvals mustbe obtained for proceeding with actual investments. NGC will also seek toinclude these projects in the investment case for the subsequent price review.These planning criteria do not take the economic cost of congestiondirectly into account. However, the reliability criteria effectively providefirm transmission service to system users under the study conditions usedfor transmission planning purposes and necessarily mitigate congestionunder the study conditions in the process of meeting reliability criteria.However, variations in supply and demand conditions, as well as outagesof transmission facilities, can lead to congestion in real-time operations.Through the balancing incentive mechanisms, NGC must pay for a shareof the costs of balancing the system in the face of congestion that may arisein real-time operations. This provides additional incentives to NGC tomake transmission investments with short paybacks that were not includedin the plan upon which the price control was based or to advance invest-ments in the base plan to reduce congestion and other system balancingcosts. It also provides incentives for NGC to maintain the network andspend resources on restoration of outages when these expenditures arevaluable because they reduce system balancing costs. As with all incentiveregulatory mechanisms, these mechanisms reflect a balancing of the incen-tives to reduce costs and meet quality standards and capturing the ‘rents’from cost reductions for consumers (Tirole and Laffont, 1993, Chapter 1).

Investments in interconnectors with other networks are not covereddirectly by NGC’s license. The existing interconnector with France is noworganized as a separate ‘merchant’ business and the associated capacity is


allocated by auctioning physical rights of various durations. In principle,both NGC and third parties are free to propose adding interconnec-tors between NGC’s network and, for example, France or Belgium. Theregulatory treatment of such facilities can be negotiated with Ofgem,though the assumption has been that these facilities would be built on amerchant basis. No interconnectors have been added since the CEGB’srestructuring in 1990, so how this would play out in practice is unclear.Moreover, the UK’s interconnector policies are in the process of being har-monized with the ‘regulated third party access regulations’ specified byrecent EU directives (Ofgem, 2003b, 2004a).15

The liberalization program in E&W is, in my view, the most successful inthe world. An important component of this successful restructuring wasthe creation of a single independent transmission company that combinestransmission ownership, maintenance and system operating responsibili-ties in an organization that spans the entire transmission network coveringE&W. The utilization of and continuous improvement in benchmarkingand incentive regulation have also contributed to the excellent performanceexhibited in the transmission segment since the mid-1990s when thesemechanisms were first introduced. While the fact that England and Walesis (almost) an electrical island, eased the successful reliance on a singlenational network compared to the loop flow and related issues thatnational networks must face in continental Europe, the E&W experiencecontains many lessons for successful operation of and investment in high-voltage transmission networks in liberalized electricity sectors.

6. MARKET, REGULATORY AND TRANSMISSIONPOLICIES IN PJM

It is difficult to describe or evaluate transmission investment policies in theUS in a simple way. This is the case for several reasons. First, transmissionpolicy in the US has been in a constant state of change for the last decade.Second, the regulatory responsibility for important aspects of transmissionpolicy is split between the federal government and the states and reflects thelegacy of vertically integrated utilities regulated primarily by the states.Third, different states have taken very different approaches to liberalizationof the electricity sector (Joskow, 2005). No federal laws have been enactedclearly to promote wholesale and retail competition or the changes in sup-porting institutions required to help to make these competitive initiativesachieve their goal of providing long-term benefits to consumers. Fourth,the availability of consistent data on transmission prices, investment, andnetwork performance is extremely limited (US EIA, 2004). Accordingly, I


shall focus here on transmission pricing and investment policies in PJMwhere FERC’s vision for the ideal model for wholesale market design andtransmission institutions (the so-called ‘standard market design’: SMD)has been implemented and for which we now have several years of experi-ence. A more detailed discussion of US transmission pricing and invest-ment policies can be found in Joskow (2005).

Industrial Organization of PJM and Wholesale Market Design

PJM entered the electricity liberalization era as a multi-state power pool(‘tight pool’) which centrally dispatched the generating facilities for verti-cally integrated utilities in Pennsylvania, New Jersey, Maryland, Delawareand Washington DC based on the marginal costs of the generating unitsowned by PJM’s member utilities. PJM’s origins and experience in eco-nomic generator dispatch, management of network reliability, and systemplanning can be traced back to the 1920s when it began to be created by theprivate vertically integrated electric utilities in this area. In 1998, the PJMagreement was restructured to turn the cost-based power pool into a set ofbid-based wholesale spot power markets and supporting institutions,including transmission pricing and investment protocols.

PJM is now an ISO and has been qualified as an RTO by FERC pursuantto Order 2000. It is structured as a for-profit limited liability company withan independent board of directors, though it presently operates de facto asa non-profit organization. PJM is not a market participant, does not owngeneration, transmission and distribution assets and is not engaged inwholesale or retail marketing.16 It is responsible for system operating reli-ability and for applying reliability rules and criteria developed by regionalreliability councils (Mid-Atlantic Area Council (MAAC) in the case of theoriginal PJM footprint). PJM’s geographic footprint has expanded in thelast couple of years to include transmission owners in portions ofPennsylvania that were not previously in PJM, and utilities covering por-tions of West Virginia, Ohio, Kentucky, Indiana, Virginia and Illinois.17

(The Midwest ISO (MISO) includes the transmission owners covering therest of these Midwestern states.)

The transmission owners in PJM are all vertically integrated utilities thatalso own generating capacity and distribution companies, and have unregu-lated wholesale and retail marketing affiliates. They continue to have trans-mission operating functions, including transmission maintenance, outagerestoration and investment responsibilities, subject to various agreementsbetween the transmission owners and PJM and supporting FERC regula-tions. The prices for ‘unbundled’ transmission service made available bythese transmission owners to third parties (generators, retail and wholesale


marketers, and unaffiliated distribution companies) is regulated by FERC.The prices for ‘bundled’ transmission service that the vertically integratedtransmission owners make available to their own retail customers (thosewho have not agreed to be supplied by competitive retail suppliers in thosestates with retail competition) are effectively regulated by each state as partof the overall regulation of the prices for bundled retail service. Thus, thesame transmission facilities are compensated through two regulatedrevenue streams, one (unbundled) governed by FERC regulation and one(bundled) governed by state regulation. The prices for transmission servicesare set based on traditional cost-of-service or rate of return principles (asdiscussed in more detail in Joskow (2005) and below) applied to each trans-mission owner’s facilities. Although FERC Order 2000 encourages it, thereare no formal incentive regulation mechanisms applicable to costs orquality of service that is applied by FERC or the state regulators to eitherthe TOs in the PJM area or to PJM itself.

PJM operates (voluntary) day-ahead and real-time (adjustment or bal-ancing) bid-based markets for energy and ancillary services. Market partic-ipants submit bids and offers to the day-ahead and real-time markets. LMPsthat balance supply and demand at each location on the network and theallocation of scarce transmission capacity are performed together using aleast-cost bid-based security-constrained dispatch (state-estimator) modelthat incorporates the physical topology of the network and reliability con-straints. The LMPs reflect equilibrium marginal energy costs and the mar-ginal cost of congestion at each location (marginal losses will be includedsoon, as in the LMP systems in New York and New England). Participationin day-ahead and real-time markets is voluntary in the sense that generators,loads and marketing intermediaries may submit their own day-ahead sched-ules for energy and ancillary services to the RTO and can (try to) use bilat-eral arrangements to stay in balance in real time. However, bilateralschedules are still liable for congestion and loss charges and any residualimbalances are settled at the real-time prices. Congestion is priced based onthe difference in LMPs between the designated delivery and receipt pointsof generation supplies chosen by a transmission service customer.

Load serving entities (LSEs – distribution companies or competitiveretail suppliers which have responsibility for supplying retail consumers) inPJM have forward generation ‘capacity obligations’ based on their expec-ted peak loads in each month and must contract forward for capacity orpay penalties. PJM operates capacity markets, but it appears that bilateralarrangements govern the allocation of most qualifying generating capac-ity. These capacity markets are in the process of being restructured as this iswritten. Generators must meet certain transmission ‘deliverability’ require-ments to qualify as capacity resources. As discussed further below, these


deliverability requirements play an important role in the transmissioninvestment process and in providing locational incentives to generators.

Transmission Pricing and Related Policies

PJM administers an open access transmission tariff that requires the trans-mission owners in PJM to offer transmission services at non-discriminatorycost-based prices. This tariff (along with the PJM Operating Agreementand the PJM Reliability Assurance Agreement which are interdependent)establishes prices for various categories of transmission service available tothird party transmission users;18 defines how the associated revenues aredistributed to TOs; specifies interconnection rules and obligations for gen-erators, merchant transmission owners (none yet) and regulated TOs;specifies the definition, allocation mechanisms, accounting and settlementsfor financial transmission rights (FTRs); and establishes a process for iden-tifying and approving regulated transmission expansion projects and theallocation of the associated costs and FTRs.19 These transmission pricingarrangements are being revised as this is written.

Transmission Investment Framework

Transmission investments in PJM are grouped into several categories.

Direct interconnection investmentsWhen a new generating unit or merchant transmission projects seeks toconnect to the PJM network, the TO in whose transmission zone the projectwill be located performs a study of the direct capital and operating costsassociated with the new transmission facilities required to make the directconnection to the network. The proposed generating project is responsiblefor 100 per cent of these direct interconnection costs. About $304 million ofinvestments that appear in PJM’s July 2004 ‘Regional TransmissionExpansion Plan’ (RTEP) (PJM Interconnection, 2004f) update fall in thiscategory, out of a total approved projects of about $785 million.20 Directinterconnection costs are therefore treated similarly in PJM and E&W.

Interconnection network reliability investmentsPJM and the TO in whose transmission zone the facility is located alsoevaluate the impact of the proposed project on network reliability. A seriesof engineering studies are performed to assess whether the proposedproject, as an increment to the existing facilities on the network, will leadto any violations of PJM’s reliability criteria. These criteria are much morecomplex than the simple N – 1 operating reliability criterion that is often


discussed in the literature. The reliability assessments involve a set ofassumed study conditions under various contingencies: when all facilitiesare operating; N – 1; N – 2; multiple contingencies; and delivery to loadcriteria. These criteria and their application have not changed significantlysince before the new PJM markets were created and take no account of theLMP mechanisms or of the associated market mechanisms for allocatingscarce transmission capacity. If the engineering studies indicate that relia-bility criteria are violated, the expected costs of network investmentsrequired to restore the reliability parameters are identified. The proposedgenerator will be required to pay for these costs, though they may be sharedwith other generators in the construction pipeline that benefit from thesenetwork enhancements (the cost allocation mechanism is fairly compli-cated). The generator will receive its proportionate share of any new FTRs(or the revenues produced when these rights must be auctioned) created as aconsequence of the network facility enhancements it is required to pay for.

It is important to note that these reliability assessments are based on aset of engineering assumptions and study conditions that may bear littlerelationship to the way the network would actually operate if the networkenhancements were not made and increased congestion was realized. Thatis, if the generators were built and these ‘deep’ network enhancements werenot made, the network would not necessarily suffer a violation of its oper-ating reliability criteria. Instead, redispatch would have to be used tobalance the system.

Generator deliverability investmentsIf a generator or HVDC merchant transmission project wants to qualify asa ‘capacity resource’ under PJM’s Reliability Assurance Agreement andwholesale market Operating Agreement, as is typically the case since thereis significant ‘capacity value’ in the PJM market, they must meet a final‘reliability’ criterion called ‘generator deliverability’. Engineering studiesare performed to determine whether (oversimplifying a complex process)the full power that the proposed generator can produce can be reliablydelivered outside of its transmission zone under a set of engineering studyconditions that assume all existing generators are dispatched first to meetload.21 If the generator deliverability condition is not satisfied the genera-tor must either pay for any necessary network enhancements (and receiveany incremental FTRs) or purchase firm transmission service from a thirdparty that supports deliverability. Interconnection network enhancementsand deliverability network reliability enhancements together account forabout $207 million of investments in PJM’s 2004 RTEP update (as of July2004). These obligations are conceptually most similar to the generatorcomponent of the locational TNUoS charges in E&W. Thus, generators are


obligated to pay for about $511 ($304 million direct interconnection + $207million ‘deep’ network upgrade investments) of the roughly $785 millionof transmission investments approved through the 2004 PJM planningprocess or about 75 per cent. Thus, PJM effectively has a ‘deep’ intercon-nection pricing policy.

It should be noted that interconnection network investments and deliv-erability network investments provide potentially powerful locationalincentives to new generating projects. The network upgrade costs at somelocations may be zero (or even negative) and at other locations these costsmay be substantial (as are the generator TNUoS charges in E&W). Newgenerators can reduce their investment costs by selecting a location wherethese network upgrade obligations are low rather than high. It is likely thatthese interconnection network upgrade cost obligations play a more impor-tant role in generator location decisions than do variations in LMPs.

Other network reliability investmentsThe PJM RTEP process may indicate that one or more PJM/MAAC reli-ability criterion is expected to be violated for other reasons, for example,load growth or generator retirements at specific locations. PJM can directTOs to make the necessary investments required to restore the reliabilityparameters. The associated costs are then recovered from charges to theload that benefits from the investments. These costs amount to about $274million in the 2004 RTEP. This appears to be the fastest-growing categoryin the RTEP planning process and would include network upgradesrequired as a consequence of retirements of existing generating facilities.

Merchant transmission investmentsThe original design of the PJM system was predicated on the assumptionthat any ‘economic’ transmission investments that were not required for‘reliability’ would be made on a merchant basis. The costs of merchanttransmission projects would be borne by the developer and the developerin turn would receive the financial transmission rights created by the invest-ment. The incentive for merchant investment would then be the marketvalue of the transmission rights created by the project. The associatedexpected value of the transmission rights created is then the expecteddifference between the LMPs between the affected delivery and receiptpoints times the incremental transmission capacity between these pointscreated by the investment (Joskow and Tirole, 2005b). In the case of ACfacilities, a merchant investor would receive any incremental FTRs result-ing from the investment. HVDC merchant transmission facilities aretreated like generators and effectively create physical import or exportrights to the AC network.


Merchant transmission projects must also pay for direct interconnectionand ‘deep’ network upgrade costs in essentially the same way as do new gen-erators. Table 5.522 illustrates the results of the PJM interconnection studyprocess and the estimated costs of direct interconnection and ‘reliability’network upgrade costs for a proposed merchant HVDC project under LakeErie connecting Ontario with Pennsylvania (now cancelled). The totalinterconnection costs for this project were estimated to be $102 million ofwhich about 10 per cent were direct interconnection charges and 90 per cent‘deep’ network upgrades to restore a long list of reliability problemsexpected to be created by the project.

PJM’s ‘deep’ interconnection pricing policies for new generators and mer-chant investment projects are not typical of the pricing of interconnectionand transmission use of system services elsewhere in the US. A ‘shallow’interconnection policy is more typical. Generators pay direct interconnectioncharges as in PJM. The costs of network upgrades deeper in the network arethen typically rolled in with the legacy network costs to create use of systemcharges that are identical at all interconnection points on an individual TO’snetwork. FERC’s most recent interconnection rule provides for shallowrather than deep interconnection charges.23 As RTOs have grown, FERC has


Table 5.5 PJM interconnection charges: proposed Erie West HVDC

Estimated cost ($m)

Direct connection facilities 9.5‘Deep’ network upgrades 91.5

Single contingencySecond contingencyMultiple facility contingencyGenerator DeliverabilityOther

Total cost interconnection cost 1023.5 year construction time

Note: Transmission Interconnection Quene #G00_MTX3 is a TransEnergie U.S. Ltd.request to connect the southern terminal, of the Erie West to Nanticoke HVDC intertie, tothe Erie West 345 kV substation. TransEnergie proposes to construct an HVDC converterstation in the vicinity of Erie West, and a double circuit 345 kV line to connect Erie West tothe converter station. The northern terminal of the intertie will be connected to Nanticokesubstation in the Ontario Hydro system. The interconnection request is nominally rated at1,000 MW net of losses on the HVDC system. The developer has requested Firm (Capacity)Transmission Injection Rights in the amount of 1,000 MW and Firm TransmissionWithdrawal Rights in the amount of 1,060 MW at the HVDC terminal in PJM. Project#G00_MTX3 is scheduled for commercial operation in 2004.

Source: PJM Interconnection (2003).

endeavored to (effectively) reallocate these costs to eliminate ‘pancaking’ andto shift network use charges to load from generators (Joskow, 2005). Thesereallocations of transmission costs have been quite controversial.

Several merchant transmission projects have been proposed through thePJM interconnection and regional transmission planning process, primar-ily DC interconnectors with neighboring control areas. Two transformerupgrades have been made by a TO in PJM as merchant projects in returnfor FTRs. None of the proposed DC interconnectors has yet gone intooperation and several have been cancelled. The most active projects areHVDC interconnections between PJM and New York City and LongIsland. The farthest along is a project that has been awarded a long-termcontract for transmission between PJM and Long Island by LIPA, a muni-cipal utility which can pass on the associated costs to its regulated cus-tomers without approval of a state or federal regulatory agency. LIPAalready has a long-term contract for all of the 330 MW capacity of theCross Sound Cable connecting New England with Long Island, the only‘merchant’ project completed so far in the US.

HVDC links to New York City and Long Island are especially attrac-tive for a number of reasons. The LMPs in New York City and LongIsland are consistently significantly higher than those in neighboringareas – about $20/MWh on an annualized basis. In addition, these areboth very difficult places to find sites for new power plants and haveextremely high construction costs. In addition, HVDC links from PJMand New England can be brought in under water where Nimby issuesshould be less of a problem (though this did not mute the controversy overthe Cross Sound Cable process). Finally, on Long Island there is a munic-ipal distribution utility that is willing and able to sign long-term contractsfor the transmission capacity developed in this way. This means that thedeveloper does not have to rely on differences in spot market LMPs toproduce the revenues for the project, reducing financing costs and oppor-tunism problems.

Economic planned transmission facilitiesPJM resisted doing any analysis of ‘economic’ transmission investmentopportunities or including such potential investments in its regional trans-mission plan and requiring TOs to proceed with them if merchant investorsdid not show any interest in them. As before, by ‘economic transmis-sion’ investment opportunities I refer to transmission investments whoseexpected economic benefits arise from reducing congestion (and losses).When the expected incremental reduction in congestion and loss costsexceeds the incremental cost of the network enhancement then the invest-ment is ‘economical’.


PJM’s dream that the invisible hand would lead merchant investors tocome forward to make intra-TSO investments in response to congestionrents has not been matched by reality. After a contentious regulatoryproceeding, in 2003 FERC issued an order that required PJM to includepotential ‘economic’ transmission investments in its planning process. PJMhas now developed a process to identify transmission constraints thatcreate ‘unhedgeable congestion’ and to assess the benefits and costs ofpotential network enhancement projects that would mitigate this conges-tion. When projects that mitigate unhedgeable congestion are identifiedand pass certain cost/benefit thresholds they are included on a ‘marketwindow’ list. The projects on this list are then open for one year to propos-als from merchant investors. If satisfactory proposals are not forthcoming,PJM may direct incumbent TOs to build the projects as regulated projectsand include them in the PJM tariff for cost recovery. The process iscomplex, still evolving, and the phrase ‘unhedgeable congestion’ somewhatmisleading.

The process for identifying so-called unhedgeable congestion actuallyyields an estimate of the costs of congestion after netting out congestionrents. To oversimplify,24 PJM defines unhedgeable congestion as congestionwhich cannot be hedged with the existing portfolio of FTRs. The best wayto think of PJM’s unhedgeable congestion concept is as an approximationto the social cost of congestion. And this appears to be the number that oneactually would want to use in order properly to evaluate potential ‘eco-nomic’ transmission investment opportunities. For the 14-month periodfrom August 2003 to September 2004 there was $1.6 billion of ‘gross’ con-gestion in PJM (including congestion rents), of which $336 million wasdefined as ‘unhedgeable’.25

Where unhedgeable congestion is identified, a set of simple cost–benefitassessments of transmission upgrades are then performed by PJM. Theactual unhedgeable congestion values attributed to each constraint over theprevious 12-month period is divided by the estimated cost of a transmissionupgrade that would mitigate the congestion costs identified.26 This is definedas the ‘benefit/cost ratio’, though it is actually a measure of the simplepayback period for each identified investment opportunity assuming thatcongestion rents do not change in the future. When these assessments yieldbenefit/cost ratios that exceed certain specified thresholds a project is put ona list of potential regulated ‘economic’ transmission projects. Market par-ticipants are then given a year to propose alternative ‘market solutions’ tothe identified projects. If market solutions are not forthcoming, the projectsare added to the PJM RTEP and the incumbent TOs in whose transmissionzones the projects are located are directed to make the investments. Theresulting costs, net of revenues from the auctioning of FTRs created by the


investments, are then recoverable through the PJM Open Access Tariff fromthe customers of the LSEs who are expected to benefit from the investments.The responses to the first ‘market window’ open for proposals to resolve thiseconomic congestion were set in 2005.

Roughly 50 potential ‘economic’ transmission investment projects havebeen identified since this evaluation process was implemented in March2004 and ‘market windows’ are now open for merchant projects to fill theseneeds before regulated transmission projects are added to the RTEP.27 Thecost–benefit analyses indicates that seven of the identified projects havesimple paybacks of three months or less (again, assuming that unhedgeablecongestion does not change in the future). Another 12 have simple pay-backs of less than four years (see Table 5.6). If FERC had not forced PJMto examine ‘economic’ transmission investment projects, all of these wouldhave been left on the table in the hope that merchant investment wouldeventually come forward. It should also be noted that in several cases, fairlysmall investments completely eradicate the congestion so that they are notconducive to being supported by merchant investments. Apparently threemerchant transmission network upgrade proposals have been made so farin connection with the first market window, though studies have not beencompleted and they may be affected by other transmission investment pro-jects that have been proposed.28

Inter-TSO (interconnector) investmentsThe expansion of interconnections with neighboring control areas is notincluded in the PJM planning process, though procedures are in place thatgovern the rules governing pricing of the costs of interconnecting inter-TSO facilities to the PJM network. Accordingly, by default, inter-TSOtransmission investments are left to merchant developers. As already dis-cussed, a few merchant HVDC links with New York City and Long Islandhave been proposed and at least one is likely to move forward, supportedby a 20-year contract with LIPA. There is little if any additional merchantinvestment activity on the horizon. However, by incorporating neighboringTSOs into PJM by expanding its boundaries, it is effectively internalizinginter-TSO transmission investment opportunities (as well as integratinggenerator scheduling, wholesale market and congestion managementmechanisms) into the intra-TSO transmission investment planning process.As these additional TSOs are integrated into PJM, the PJM generatorinterconnection, reliability and economic investment protocols will applyto what were previously inter-TSO opportunities that have largely beenignored due to the balkanization of transmission ownership and systemoperations. Just as a fairly large number of ‘economic’ transmission invest-ment opportunities popped up once PJM actually looked for them, I expect



Table 5.6 Market window ‘economic’ transmission projects in PJM as ofNovember 2004

Monitored facility *Unhedgeable Limit Cost to Cost/congestion ($) relieve limit ($) benefit

LINE 230 KV 1,091,588 Circuit 200,000 �0.25ADA-BRUX switcher

LINE 500 KV 1,607,237 Wavetrap 75,000 �0.25BED-BLA

BED-BLA 83,999,705 Voltage 5–25 �0.25million

LINE 230 KV 4,146,221 Circuit 200,000 �0.25ADA-BENX switcher

LINE 138 KV 1,134,130 Circuit 200,000 �0.25BRU-EDI switcher

LINE 69 KV 3,397,773 Conductor 500,000 �0.25SHI-VIN

LINE 500 KV 307,337 Disconnect 45,000 �0.25FTM-PRU switch

PJMW500 3,284,457 Voltage 5–25 0.25–4million

LINE 230 KV 2,739,456 Conductor 1,000,000 0.25–4NWA-WHI

EAST 2,264,606 Voltage 5–25 million 0.25–4

JACK ME 230 2,454,986 Transformer 2,500,000 0.25–4KV 4 BA-P

YORKANA 230 1,647,801 Transformer 2,500,000 0.25–4KV 1A BANK

LINE 230 KV 709,851 Disconnect 50,000 0.25–4CED-CLIK switch

LINE 230 KV 654,222 Cable 2 million 0.25–4BER-HOB

LINE 138 KV 499,774 Circuit 200,000 0.25–4EDI-MEAR switcher

LINE 500 KV 112,364 Wave trap 300,000 0.25–4ELR-HOS

LINE 69 KV 47,120 Disconnect 20,000 0.25–4EDG-NSA switch

that many more ‘reliability’ and ‘economic’ projects will emerge as PJM’stransmission planning footprint grows to incorporate what were previouslyseparate TSOs. Two major new regulated transmission projects betweenwhat were previously separate TSOs were proposed in 2006.

Despite the investment in new intra-TSO facilities in PJM, congestioncharges in PJM continue to grow (see Table 5.1). Moreover, the prospect ofa growing number of generation retirements is also leading to a need fornetwork reliability investments. Since there are no exit fees, these chargesare likely to be paid for by the TOs in the areas where the retiring gener-ators are located (PJM, 2004c).

7. ‘RELIABILITY’ VERSUS ‘ECONOMIC’TRANSMISSION INVESTMENT

All economic models of transmission investment that I am aware of focuson transmission investment as a mechanism to reduce the costs of conges-tion (Joskow and Tirole, 2000, 2005b). Some (properly) include the cost oflosses as well. When transmission capacity is congested, high-cost genera-tion must be substituted for low-cost generation to balance supply anddemand. The incremental cost of the high-cost generation that must be dis-patched due to transmission capacity constraints plus any deadweight lossassociated with reduced demand resulting from higher locational prices isthe cost of congestion. Transmission investment should then optimally bemade (ignoring lumpiness, market power and other market imperfections)up to the point where the incremental cost of transmission capacity is equalto the incremental reduction in the expected present discounted value ofcongestion and loss costs. These models bear little if any relationship to theway intra-TSO transmission investments are actually evaluated by TSOs inthe US and E&W.



Monitored facility Unhedgeable Limit Cost to Cost/congestion ($) relieve limit ($) benefit

LINE 230 KV 200,355 Wave trap 200,000 0.25–4BRA-FLA

JACK ME 115 9,272,381 Transformer 2,500,000 0.25–4KV 5 BA-S

Source: PJM Interconnection (2004f).

As we have seen, in E&W and PJM, virtually all of the transmissioninvestments that have been approved have been justified either by directinterconnection costs or by ‘reliability’ considerations.29 The E&W systemdoes not even appear to have a transmission investment concept akin toeconomic transmission investments that are justified by savings in conges-tion costs aside from the incentives to reduce congestion costs embodied inthe SO incentive mechanism. In New England, with a similar market designto PJM’s, the New England ISO manages a very detailed regional trans-mission expansion planning process that examines needs and opportunitiesfor both ‘reliability’ transmission investments and ‘economic’ transmissioninvestments. This process includes models that forecast congestion. The2004 update to the New England ISO’s regional transmission expansionplan identified $2 billion ($1.5 to $3.0 billion) of transmission investmentprojects and essentially all of them are justified as ‘reliability’ investments(ISO–NE, 2004). Not a single project was identified which could be sup-ported by congestion cost savings alone.

In fact, many network upgrade investments that are justified on ‘reliabil-ity’ grounds could just as well be categorized as ‘economic’ transmissioninvestment opportunities. In many cases, if the investments were not made,the network could still be operated ‘reliably’, but there would be more con-gestion, more controlled load shedding, and much higher prices in someareas. Moreover, many reliability investments affect the future trajectory ofLMPs and incentives for generation and transmission investments. On theother hand, ‘economic’ transmission investments can also often confer‘reliability’ benefits as well. Thus, in my view, at the very least, reliabilityand economic transmission investments are interdependent. At worst, thedistinction between them is analytically flawed. Moreover, the distinctionsbetween reliability-driven and congestion cost-driven transmission invest-ments creates a very significant asymmetry between the treatment of intra-TSO network investments and inter-TSO network investments. The formerare evaluated and priced as reliability investments while the latter must bejustified and paid for based on congestion cost savings alone, by default ona merchant basis.

It is fairly clear that transmission investments driven by reliability crit-eria have significant effects on LMPs and network congestion. In addition,discretionary changes in system-operating practices, including changes inthe ways that operating reliability criteria are applied or evaluated, can havedramatic effects on the ‘capacity’ of portions of the network and on theresulting congestion rents and congestion costs.

In the studies underlying the New England ISO’s 2004 regional expansionplan it is quite evident that reliability investments get triggered well beforelocational prices or congestion are allowed to rise anywhere close to the


value of lost load (ISO–NE, 2004). In PJM, the data that have been madepublic regarding ‘economic’ transmission opportunities also make it clearthat reliability investments can have a very significant impact on transmis-sion congestion and the incentives for transmission investment to reducecongestion costs. As already mentioned, roughly 50 projects have beeninitially listed in the ‘market window’ for potential regulated ‘economic’transmission investment. Some 32 per cent of these projects subsequentlywere tagged with the notation ‘reliability upgrade expected to mitigate con-gestion’. One of these projects had 12-month unhedgeable congestion costsof $192 million. The full list is contained in Table 5.7. Two additional pro-jects were designated as benefiting from changes in operating practices. Oneof these projects had 12-month unhedgeable congestion costs of $90 million.These example are, of course, only indicative of the more general observa-tion that so-called reliability transmission investments, as well as discre-tionary changes in operating practices and study assumptions, can mitigatea lot of congestion that would otherwise emerge on the network well beforeit is actually revealed. This in turn has implications for the consideration ofeconomic transmission investment models that are driven by the trade-offbetween transmission investment and the costs of congestion. In particular,for a potential merchant investor, the possibility that reliability-driven trans-mission upgrades and discretionary changes in operating practices and theimplementation of operating reliability criteria will significantly reduce oreliminate congestion, is likely to be a significant deterrent to investment thatmust be supported from congestion rents.

This discussion should not be read as implying either that reliability cri-teria are unnecessary (in Joskow and Tirole, 2006, we explain why operat-ing reliability criteria are necessary due to the threat of network collapsesthat make reliability a public good) or that they have been set incorrectly.It does imply two things: (a) we need to better understand the economicjustification (costs and benefits) for these reliability criteria and (b) eco-nomic models of transmission investment need to take into account thefactors that create a need for administrative reliability criteria and theimpacts of reliability criteria that are applied in practice.

8. CONCLUDING THOUGHTS

Major questions have been raised about whether and how efficient levels oftransmission investment can be mobilized in liberalized electricity sectors.Significant barriers to efficient transmission investment continue to exist inmany countries with liberalized electricity sectors. These barriers are pri-marily institutional rather than fundamental. The experience in England



Table 5.7 Examples of transmission congestion mitigated by reliabilityinvestments in PJM

Monitored *Unhedgeable Limitfacility congestion ($)

LINE 230 KV 268,024 Line trap RTEP reliability upgradeGRE-POR expected to mitigate congestion

WYLIERID500 6,797,499 Transformer RTEP reliability upgradeKV TRAN 5 expected to mitigate congestion

CEDAR 5,480,787 Voltage RTEP reliability upgradeexpected to mitigate congestion

BRANCHBU500 192,863,356 Transformer RTEP reliability upgradeKV 500–1 expected to mitigate congestion

BRANCHBU500 3,556,256 Transformer RTEP reliability upgradeKV 500–2 expected to mitigate congestion

NORTH PE 1,841,999 Voltage RTEP reliability upgradeexpected to mitigate congestion

LINE 138 KV 383,541 Line trap RTEP reliability upgradeLAN-MIN expected to mitigate congestion

LINE 69 KV 180,726 Stranded bus RTEP reliability upgradeLEW-MOT2 expected to mitigate congestion

LINE 230 KV 61,392 Wavetrap RTEP reliability upgrade priorMAR-MRP to spring of 2004

LINE 138 KV 1,738,983 Conductor RTEP reliability upgradeGLA-MTP expected to mitigate congestion

LINE 69 KV 536,976 Conductor RTEP reliability upgradeBEC-PAU expected to mitigate congestion

HUDSON 230 138,865 Transformer RTEP reliability upgradeKV HUDSON2 expected to mitigate congestion

WYEMILLS138 316,952 Transformer RTEP reliability upgradeKV AT-2 expected to mitigate congestion

SICKLER 230 592,446 Transformer RTEP reliability upgradeKV SICK #1 expected to mitigate congestion

LINE 69 KV 209,335 RTEP reliability upgradeCED-SAN expected to mitigate congestion

LINE 69 KV 30,141 Conductor RTEP reliability upgradeTAL-TRA expected to mitigate congestion

Note: *Previous 12 months.

Source: PJM Interconnection (2004f).

and Wales demonstrates, however, that liberalization does not necessarilylead to depressed levels of transmission investment. The experience in PJMillustrates that regional planning mechanisms and transmission investmentcriteria can be used effectively to identify transmission investment needsand to price transmission services to provide good locational incentives.The PJM experience also illustrates some of the problems of separating SOand TO functions, vertical integration of TOs with generation, and thebifurcation of regulatory responsibilities between incompatible state andfederal regulatory processes. Let me supplement the summary of my con-clusions contained in the Introduction to this chapter with the followingobservations.

● Industrial structure Many countries have failed to fully restructuretheir electricity sectors to support competition. The creation ofindependent regulated TSOs with system operations, transmissionnetwork ownership, maintenance and investment responsibilities withadequate geographic scope is the foundation of efficient operationsand investment programs. The full unbundling of transmissionservice prices subject to a single regulatory regime is a natural com-plement to the creation of such TSOs. The structure adopted inEngland and Wales is superior to the RTO structure being promotedin the US. However, both are superior to structures with no ISO at all.

● Geographic scope TSOs typically span only portions of larger syn-chronized AC networks. The mobilization of investment for intra-TSO transmission enhancements is much better developed than is themobilization of inter-TSO transmission investments. This was aproblem (perhaps not perceived) before liberalization and it is a con-tinuing problem today. In the US, the effort to consolidate controlareas under larger RTOs provides one path to reducing the ‘seams’problems at the boundaries between TSOs. The creation of a singleTSO for Great Britain that covers Scotland, as well as England andWales, reflects a similar motivation. However, there are practical andpolitical limits on the consolidation of TSOs in many countries. Thisimplies that new cooperative mechanisms need to be developed toharmonize reliability criteria, economic criteria, transmission pricingand investment policies, and wholesale market mechanisms to betterintegrate inter-TSO behavior so as to smooth out the seams as muchas is feasible.

● Regulatory framework Most of the transmission infrastructure thatis in place and future investments in it are likely to be governed bysome regulatory framework. A clear, credible and transparent regula-tory framework that specifies the TSO’s responsibilities, performance


norms and regulatory mechanisms consistent with these objectivesand performance norms is essential. All regulatory frameworksare imperfect. However, there is no choice but to draw on availableexperience and regulatory tools to develop and to apply the best fea-sible regulatory frameworks. A practical regulatory framework willinevitably include a mix of cost-of-service regulation with an overlayof PBR mechanisms based on benchmarking, profit sharing (slidingscale) and ‘ratchets’. The development and application of perfor-mance norms, formal investment criteria, as well as considerable regu-latory judgment is an inevitable component of a sound regulatoryprocess. One component of such a regulatory framework is a trans-parent regional transmission investment planning process with clearrules for achieving defined reliability and economic goals. The regu-latory framework in E&W has many attractive properties. The bifur-cation of regulatory responsibilities in the US and the failure to fullyunbundled transmission service prices create significant disincentivesto efficient transmission investment.

● Reliability versus economic investments The liberalization programsin most countries carried along with them the transmission networkplanning and reliability rules and evaluation criteria from the era ofregulated vertically integrated monopolies. Transmission investmentactivity today is driven almost entirely by reliability criteria. Wheredid these criteria come from? Why are they the right criteria? Littleeffort has been made to review these rules and criteria in light of thedevelopment of markets that both provide information that can beused to evaluate the costs and benefits of these reliability standardsand provide market mechanisms that can be used to achieve reliabil-ity criteria more effectively. Intra-TSO reliability-driven transmissioninvestments and intra-TSO congestion cost-driven investments are,at the very least, interdependent. At worst the distinctions betweenthem are not particularly useful. Clearly, economic and reliability cri-teria need to be better integrated into the transmission investmentplanning and regulatory arenas. Modest steps to do so are nowtaking place in the RTOs in the Northeastern and Midwestern US.

● Investment characteristics Transmission investment opportunitiesinvolve much more than the construction of major new transmissionlinks. Because many transmission limitations reflect contingencylimits and associated reliability rules (which should be re-evaluatedas noted above), there are often investment opportunities of modestcost that can increase significantly transmission capacity. The insti-tutions and regulatory mechanisms to identify and undertake theseopportunities need more attention. This is especially important in an


era when it is difficult to obtain permission to build new transmissioncorridors.

● Merchant transmission investment Market-driven transmissioninvestment may be a complement to regulated transmission invest-ment but it is not a substitute. Merchant transmission investment hasand is likely to make a very small contribution in the overall portfolioof transmission investment projects that will be made in the future.The efforts to debate its role have been a distraction from more pro-ductive initiatives.

● Wholesale market design Efficient transmission network operationand investment decisions are necessarily interdependent with thedesign, operation, incentives and price signals generated by the whole-sale markets for power and ancillary services. Good wholesale marketdesign, the efficient allocation of scarce transmission capacity, andefficient investment programs go hand in hand and cannot easily beseparated.

● Economic models of transmission investment The simple models oftransmission network congestion and investment that are used byeconomists have little to do with the way transmission investmentis actually planned and developed, and the associated transmissionservices priced within the boundaries of individual TSOs today.Economic models and analysis need to be expanded to better capturethe factors that TSOs and regulators consider when they identifytransmission investment needs, especially as they relate to the imple-mentation of reliability criteria used for planning and system oper-ations. Economists and network engineers need to develop betterways to work together.

We have made a lot of progress in understanding the challenges associ-ated with stimulating efficient levels of transmission investment in liberal-ized electricity markets but there is still a lot of work to do.

NOTES

1. Some 63 kV and above. See www.rte-france.com/htm/an/qui/qui_reseau_lignes.htm.2. I am grateful to Ignacio Pérez-Arriaga for reminding me not to forget the costs associ-

ated with losses.3. See Chapter 4 for a discussion of ‘lumpy’ investments.4. PJM (2004b, Ch. 6).5. These links can also support bi-directional economic power trading opportunities. For

example, New England typically imports from Quebec over a DC link during the dayand exports power to Quebec at night so that Quebec’s hydroelectric dominated systemcan store water at night when prices are relatively low in New England and sell it backduring the day when prices in New England are relatively high.


6. Reconductoring with new conductor technology can also increase effective transmissioncapacity without adding new transmission corridors or towers. ‘Reconductoring withgap-type conductors allowed the company [NGC] to increase the capacity of a criticaltransmission line by 24 per cent without requiring changes to the transmission towers’,Electric Transmission Week, March 14, 2005, p. 14, SNL Energy.

7. There were a few major interregional transmission facilities developed in the US toallow high generation cost areas to access less costly power in remote areas. The Pacificintertie projects (AC and DC) linking the Pacific Northwest and British Columbia withCalifornia began to be developed in the 1960s with support from the federal government,federal and municipal power entities (Bonneville and Los Angeles Department of Waterand Power) and cooperative agreements with the three vertically integrated investor-owned utilities in California. Two HVDC interties were developed in the 1980s to linkQuebec with New England. These projects were promoted by Hydro-Quebec (the low-cost power supplier) and were supported by a cooperative agreement involving all of themajor vertically integrated utilities in New England (the high-cost power buyers). Whenvertically integrated utilities took ownership interests in generating facilities outside oftheir traditional service areas they developed transmission facilities to allow them to gainaccess to the power generated by these facilities. Most of the transmission infrastructurelinking Southern California with Arizona, Nevada and New Mexico was developed inthis way, as was the very high voltage network in PJM.

8. For a similar view, see Chapter 4.9. Merchant opportunities may emerge as well for incumbent TOs seeking to exploit their

market power if the regulatory framework permits it.10. Presentation of Gary Krellenstein, JP Morgan, December 16, 2004, CMU Transmission

Conference, Carnegie Mellon University.11. There is also a six-circuit AC interconnector between Scotland and England. The costs

of this interconnector and associated facilities are included in the TOs’ use of systemcharges (except that there is a separate charge for use of non-firm capacity above the850 MW of firm capacity that existed in 1990). The interconnector’s capacity ispresently allocated using an administrative procedure that involves pro-rata allocationswhen requests for capacity reservations exceed capacity. After the British Trading andTransmission Arrangements (BETTA) went into effect the assets forming theScotland–England interconnector have been subsumed into the Great Britain transmis-sion system. The regulator is developing new mechanisms to allocate scarce capacityacross this interface (Ofgem 2003d).

12. See www.nationalgrid.com/uk/library/documents/sys_04/default.asp?action=&sNode=SYS&Exp=Y.

13. Under the recently enacted reforms, NGC’s system operating functions have beenexpanded to cover Scotland as well (BETTA). However, in Scotland the incumbent ver-tically integrated companies will remain the transmission owners.

14. There is also an incentive regulation mechanism that governs network losses thatinvolves annual adjustments in the benchmark.

15. The Dutch government recently granted permission to TenneT, the manager of the high-voltage grid in the Netherlands, to finance a regulated transmission interconnectionbetween Norway and the Netherlands, after taking direct economic, reliability and com-petition considerations into account. ‘Decision on the Application of TenneT for per-mission to Finance the NordNed Cable in Accordance with section 31(6) of theElectricity Act of 1998’, 23 December 2004 (English translation).

16. In theory an independent transco could qualify as an independent system operator orRTO as well, but this would require substantial ownership restructuring in the UScontext.

17. The APS network (PJM–West) was integrated into PJM in April 2002; theCommonwealth Edison network in May 2004; and the AEP network in October 2004.Virginia Electric Power (Dominion) became part of PJM in 2005.

18. The incumbent regulated transmission owners, all of whom were previously (and mostof whom still are) vertically integrated utilities providing generation, distribution and


transmission services to retail customers (‘native load’) do not actually purchase trans-mission service under the PJM open access transmission tariff to use their own trans-mission networks to serve their retail customers. Instead they provide the transmissionservice ‘internally’ and the associated costs are included (recovered) in the regulatedbundled prices they charge to their retail customers. However, they are subject to all ofthe other terms and conditions of the PJM Tariff, PJM Operating Agreement and thePJM Reliability Assurance Agreement.

19. A more detailed discussion of the structure of this transmission tariff can be found inJoskow (2005).

20. The $785 million figure covers projects completed since 2000 as well as future projects thatare scheduled for completion over the next few years. The rate of investment is significantlylower than in E&W from 1990 to 2001, though the systems had similar peak loads.

21. New generator deliverability criteria were recently proposed.22. All interconnection studies performed through PJM’s RTEP Process can be found on the

PJM website: www.pjm.com/planning/rtep-baseline-reports/baseline-report.html.23. FERC Order 2003, ‘Standardized Generator Interconnection Procedures,’ July 23, 2003.24. For a detailed discussion of the procedures that were recently adopted by PJM, see PJM

FERC Filing in Docket Number RT-01-2-01, dated April 21, 2004, www.pjm.com,accessed June 15, 2004.

25. PJM congestion spreadsheet downloaded from www.pjm.com on December 4, 2004.26. Unlike the New England ISO, PJM has refused to include congestion forecasts in its

planning process.27. PJM FERC Filing in Docket Number RT-01-2-01, Appendix 1, dated April 21, 2004 and

PJM ‘market window’ spreadsheet downloaded December 4, 2004. Available on the PJMwebsite www.pjm.com.

28. See www.pjm.com/planning/project-queues/merch-queue-o.jsp accessed on May 27,2005 and ‘Developer Argues Exelon-PSEG projects could disrupt role of merchanttransmission’, Electric Transmission Week, May 30, 2005, pp. 9–10.

29. In E&W an unknown portion of additional transmission investments or planned reli-ability investments that were moved forward to an earlier date were driven by the annualSO incentive scheme. As previously discussed, PJM has adopted a new framework forregulated economic investments.

REFERENCES

Chao, H.-P. and R. Wilson (1987), ‘Priority service: pricing, investment and marketorganization’, American Economic Review, 77, 899–916.

Henney, A. (1994), A Study of the Privatisation of the Electricity Supply Industry inEngland and Wales, London: EEE Limited.

Hirst, E. (2004), ‘US transmission capacity: present status and future prospects’,Edison Electric Institute, Washington, DC, June.

ISO New England (2004), ‘Regional Transmission Expansion Plan (RTEP04)Technical Report’, October 21.

Joskow, P.L. (1988), ‘Price adjustment in long-term contracts’, Journal of Law andEconomics, 31, 47–83.

Joskow, P.L. (2005), ‘Transmission policy in the United States’, Utilities Policy, 13,95–115.

Joskow, P.L. and R. Schmalensee (1983), Markets for Power: An Analysis of ElectricUtility Deregulation, Cambridge, MA: MIT Press.

Joskow, P.L. and R. Schmalensee (1986), ‘Incentive regulation for electric utilities’,Yale Journal on Regulation, December, 1–49.


Joskow, P.L. and J. Tirole (2000), ‘Transmission rights and market power on electricpower networks’, Rand Journal of Economics, 31 (3), pp. 450–87.

Joskow, P.L. and J. Tirole (2005a), ‘Retail electricity competition’, Rand Journal ofEconomics, (forthcoming), http://econ-www.mit.edu/faculty/download_pdf.php?id=918.

Joskow, P.L. and J. Tirole (2005b), ‘Merchant transmission investment’, Journal ofIndustrial Economics, 53 (2), 233–64.

Joskow, P.L. and J. Tirole (2006), ‘Reliability and competitive electricity markets’,Rand Journal of Economics, (forthcoming), http://econ-www.mit.edu/faculty/download_ pdf.php?id=917.

Laffont, J.-J. and J. Tirole (1993), The Theory of Incentives and Procurement inRegulation, Cambridge, MA: MIT Press.

Newbery, D. (2004), ‘The benefits of electricity deregulation: Europe’, presentation,MIT Center for Energy and Environmental Policy Research, Cambridge, MA,December.

NGC (2004a), ‘The Statement of the Use of System Charging Methodology’,Effective from 1 November.

NGC (2004b), ‘Interim Great Britain Seven Year Statement’, November.Ofgem (2003a), Annual Report 2002–03, 14 July.Ofgem (2003b), ‘NGC System Operator Incentive Scheme from April 2004, Initial

Consultation Document’, December.Ofgem (2003c), ‘Scotland–England Interconnector: Access Criteria from 1 April

2004, Consultation Paper’, December.Ofgem (2004a), ‘Proposals for the Amendment of the Licensing Application

Regulations’, (Interconnectors) Consultation Document, June.Ofgem (2004b), ‘Electricity Transmission Network Reliability Incentive Schemes,

Final Proposals’, December.PJM Interconnection (2003), ‘PJM Economic Planning Implementation Stake-

holder Process Meeting’, August 29.PJM Interconnection (2004a), ‘PJM Economic Planning Implementation

Stakeholder Process Meeting’, February 10.PJM Interconnection (2004b), ‘State of the Market Report 2003’, March 4.PJM Interconnection (2004c), ‘Open Access Transmission Tariff’, updated to May 6.PJM Interconnection (2004d), ‘Amended and Restated Operating Agreement’,

updated to June 15.PJM Interconnect (2004e), ‘Transmission Expansion Advisory Committee Meeting’,

June 21.PJM Interconnection (2004f), ‘Regional Transmission Expansion Plan’, July,

www.pjm.com, December 14.PJM Interconnection (2004g), ‘Generator retirement analysis’, presentation,

Transmission Expansion Advisory Committee, undated.Sweeting, A. (2000), ‘The wholesale market for electricity in England and

Wales: recent developments and future reforms’, MIT Center for Energy andEnvironmental Policy Research, WP-2000-007, November.

US Energy Information Administration (EIA) (2004), ‘Electricity Transmission ina Restructured Industry: Data Needs for Public Policy Analysis’, December.

Wolfram, C.D. (1999), ‘Measuring duopoly power in the British electricity spotmarket’, American Economic Review, 89 (4), 805–26.


PART III

Coordination between investments ingeneration and transmission

6. Long-term locational prices andinvestment incentives in thetransmission of electricityYves Smeers*

1. INTRODUCTION

The economic principles that rule capacity expansion under constantreturn to scale are well known (for example, Crew et al., 1995). One investsso as to equalise short- and long-run marginal costs. If these convenientassumptions prevailed in the transmission of electricity, one would investso as to equalise the marginal transmission capacity cost with marginalcongestion and loss costs. This is not possible. The transmission of elec-tricity suffers from many undesirable economic properties that make thedirect application of these principles impossible. It combines both eco-nomies of scale and lumpy investments which render the definition of long-run marginal cost illusory. It is also plagued by considerable externalities.Because the external benefits of transmission investments are not appro-priable by those who are at the origin of these investments (Bushnell andStoft, 1996), there is no hope of fully internalising these externalities.

The objective of this chapter is to present a model of network invest-ments in a competitive electricity system based on the idea of long-termlocational signals introduced in the European Regulation 1228/2003 onCross-border Exchanges in Electricity, hereafter referred to as the‘Regulation’ (European Parliament and Council, 2003). We conduct thisanalysis under the following drastic simplifying assumptions:

● The architecture of the market essentially consists of spot energy andtransmission markets. These are coordinated by the transmissionsystem operators (TSOs). Such architectures have been extensivelyexplored in the literature and are now commonly implemented.Current proposals of the association of European TransmissionSystem Operators (ETSO) and power exchanges (Europex) also go inthat direction (see Section 2).

187

● We ignore the distinction between day-ahead and real-time marketsand assume only one spot market that integrates both. This simpli-fication is acceptable if the day-ahead and real-time markets are wellarbitraged, a currently heroic assumption in the European context.We justify this assumption by our objective to concentrate on the mixof long- and short-term locational signals covered by the Regulation.This requires that the detail of short-term operations be somewhatdisregarded.

● We assume no market power and disregard contracts. The reader willthus find no reference to physical and financial contracts, or bilateraland centralised markets. This simplification is important becausethese contracts influence the incentive of agents in imperfect com-petitive markets. This influence disappears if we assume that there isno market power, as we do in this chapter. The contract issue shouldbe taken up as soon as one departs from the assumption of price-taking agents.

● We assume perfect information. This implies, in contrast with theassumption underlying the ‘new’ theory of regulation, that the regu-lator perfectly knows the demand functions of the consumers as wellas the production sets and costs of both the transmission and gener-ation activities. If reliability considerations need to be invoked, thenall agents foresee the same contingencies with the same probabilitiesof occurrence.

Except for taking network indivisibilities explicitly on board, we essen-tially make all the assumptions of the perfect competition model that weparticularise, as described above, to accommodate some idiosyncrasies ofthe electricity system. We then concentrate on network indivisibilities byadapting and slightly generalising O’Neill et al. (2004), which deals withindivisibilities in the unit commitment problem.

The chapter is organised as follows. Section 2 introduces the notion oflocational prices in the context of the Regulation. Section 3 casts thechapter in the economic literature on electrical networks. Section 4 presentsa reinterpretation of O’Neill et al. (2004) in terms of network pricing.Section 5 relates the interpretation of these results to notions appearingin the Regulation, namely ‘cost reflectiveness’ and ‘non-discrimination’.Section 5 also presents a transposition to the network problem of com-ments made by Hogan and Ring (2003) on O’Neill et al. Section 6 extendsthe model to address cost causality. This requires an extension of O’Neillet al.’s Theorem 2: it is presented in the Appendix, using the theory of conicduality. Section 7 takes up the question of non-discriminatory prices. Itindicates that contrary to the common wisdom underlying the European

188 Coordination between investments

debates preceding and following the introduction of the Regulation, non-discriminatory prices may entail welfare losses in a decentralised marketwith indivisibilities. Section 8 briefly discusses institutional matters. Section9 lists remaining questions that should be addressed in future work. Theconclusions in Section 10 can be summarised as follows: long-term loca-tional prices are fraught with difficulties that are largely unexplored. Theirintroduction in the Regulation may have helped to achieve a political com-promise. The task remains to make it workable.

2. INVESTMENT, LOCATIONAL SIGNALS ANDREGULATION

The concept of long-term locational prices as a signal for guiding invest-ment and location of generation plants in restructured electricity marketsappears in the Regulation on ‘Conditions for Access to the Network forCross-border Exchanges in Electricity’ that came into force in July 2004.This law is essentially an outgrowth of the work of the Florence RegulatoryForum,1 set up by the European Commission to find means to facilitateelectricity exchanges between Member States in the so-called ‘internal elec-tricity market’. The Regulation contains five parts:

1. Articles 1 and 2 set the scope of the law and introduce importantdefinitions.

2. Articles 3 and 4 state the conditions for accessing the network. Article3 presents a system of cross-border compensations between TSOs.Article 4 introduces the notion of locational prices, which is the mainfocus of this chapter.

3. Articles 5 and 6 discuss congestion management at the interconnec-tions. Article 5 makes general informational recommendations. Article6 requires that congestion management be market based.

4. Article 7 presents conditions for exemption of new interconnectorsfrom the general rules.

5. Finally, Articles 8 to 15 elaborate on procedural but crucial matters andin particular the role of Comitology (Article 13).

This chapter concentrates on long-term locational prices that it considerstogether with congestion management. The Regulation requires that loca-tional prices be efficient, cost reflective and non-discriminatory (Article 4,paragraph 1). At the time of the adoption of the Regulation (June 2003),none of the documents elaborated by the Florence Regulatory Forum hadproposed any method for computing locational prices that would satisfy

Long-term locational prices and investment incentives 189

these three criteria. The Forum and related studies (for example, Pérez-Arriaga et al., 2002) certainly looked at long-term locational signals. Butthey did so in purely accounting terms, that is, by allocating network costsamong network users. Cost allocation methods reflect costs only in a veryweak sense that does not imply any causality: agents are not charged for thecost that they cause; they only collectively pay for the total costs incurred.Cost-allocation methods are also discriminatory in an economic sensebecause they do not rely on marginal costs. The fact that the marginal cost ofthe electricity network is difficult to define, makes the combination of thesetwo criteria difficult if not impossible to implement in a non-ambiguous way.Finally, cost-allocation methods are not economically efficient in the senseof inducing investments in the right location (see Curien, 2003; Green, 2003;and Lévêque, 2003). The lack of economic efficiency is particularly danger-ous in the restructuring context: it distorts the incentive to invest and henceendangers the security of electricity supply. Lack of efficient locationalsignals and economic incentives to invest did not matter much in the oldregulatory days. The regulatory obligation to satisfy demand implied someso-called ‘optimal’ mix of capacities and locational decisions (taking con-straints such as site adequacy and environment into account). This is nolonger true in a competitive system where long- and short-term locationalprices are the sole market instruments that can induce both the right mix ofequipment and their proper location. It is thus of the essence that the pricesbe right and produce the good incentive to invest. Unfortunately, economictheory tells us that allocated costs offer no guarantee in that respect.

The absence of any precise reference to efficient long-term locationalsignals in the work of the Florence Regulatory Forum should not come asa surprise. We do not know at this stage how to construct efficient long-termlocational signals, let alone efficient, cost-reflective (in the strong sense ofcost caused) and non-discriminatory long-term price signals. The reason issimple: a generator locates on some site or it does not. Location and thechoice of a plant type are discrete decisions, and we know very little abouthow to induce the right discrete decisions through prices. It is indeed a basicprinciple of economic theory that discrete decisions are difficult if notimpossible to drive through price mechanisms because of non-convexityphenomena (see Scarf, 1994, for a clear statement of the problem andPérez-Arriaga and Smeers, 2003, for a discussion of the question in elec-tricity networks). The problem becomes particularly acute if one notes thatthe impact of these decisions generally covers a time period of several years.

While long-term locational signals are a rather unexplored area, short-term locational signals for dealing with congestion are a well-known buthighly controversial subject in Europe. These are treated in Articles 5 and6 of the Regulation. Congestion management involves continuous decision


variables so that one understands much better how to decentralise througha market process. The consequence is that different market-based conges-tion methods exist and others are proposed. It is certainly surprising thatthe Regulation recommends introducing long-term locational prices thatwe do not know how to construct and implement but avoids suggestingwell-understood short-term locational prices that have proved effective inseveral systems.

3. RELATED LITERATURE

Different approaches to grid investments exist in the literature. They arediscussed in Rosellón (2003), which puts our work in perspective. We referthe reader to that survey for further information and a guide to the litera-ture. A global and in-depth analysis of investment in transmission can befound in Woolf (2003).

Hogan (2002 and 2003) and Pope and Harvey (2002) extended the theoryof nodal prices and financial transmission rights (FTRs) to the problem ofinvestments in the grid. Their objective is to provide a market-driven mech-anism for expanding the grid. This analysis underlies the notion of mer-chant lines found in the Federal Energy Regulatory Commission (FERC)standard market design (SMD) proposal, which eventually also found itsway into the Regulation (Article 7). We discuss neither merchant lines northeir compatibility with the rest of the Regulation but instead focus on howour model relates to Hogan, Pope and Harvey’s theory. The basic idea ofthese authors is that it is possible to decentralise (at least some) investmentsin the grid to economic agents (generation companies, consumers, investorsand so on) provided that these receive long-term FTRs that guarantee thepayment of congestion rents over the life of the project. The system opera-tor (SO) grants the long-term rights in an auction. Because investments innew lines can destroy existing transmission rights, some restrictions on theallocation of the long-term rights are necessary to keep the set of grantedrights physically feasible for the network. This latter process is rathercomplex (Hogan, 2002) but the bottom line is that congestion rents remu-nerate the investors who invest in the network. Joskow and Tirole (2005)argue that the theory suffers from several shortcomings and Hogan repliedto some of these comments (Hogan, 2003). We shall not elaborate on thesediscussions, which can be found in summarised form in Rosellón’s paper, butconcentrate on one of the points unveiled in Joskow and Tirole (2005) andclearly recognised in Hogan (2002 and 2003), namely investment lumpiness.

FTRs were primarily introduced as hedges against random congestioncharges (Hogan, 1992). FTRs are forward contracts on spot transmission


prices. They are special forward contracts though, in the sense that theyneed to be traded through the TSO in order to guarantee that they are com-patible with the physical possibilities of the network. Under the usual no-arbitrage assumption, FTRs satisfy the standard property derived fromfinance theory that the forward price is the expectation, in some risk-neutral probability, of the congestion price. In other words, the economicsignal embedded in long-term FTRs contains only congestion charges. Inorder to see this more clearly, consider a deterministic world. There is noneed to hedge and an FTR is then exactly equivalent to a payment of thecongestion charge in real time. Resorting to long-term FTRs for inducinginvestments in the grid is then equivalent to using congestion charges asincentives to invest in the grid. Investment lumpiness limits this potentialincentive. We elaborate in the following on the consequences of this pointand argue that, notwithstanding investment lumpiness, capacity expansionof the grid can still be decentralised provided that one introduces moreprices than the sole congestion charges. In Section 8 we shall return to theinterpretation of that result in terms of Hogan’s theory.

The transco model is a second approach to investments in the grid.2 Atransco is a company that builds and operates the network for profit. In theterms of this chapter and taking the concepts of the Regulation on board,the transco is remunerated with both long- and short-term locationalcharges. It then develops the grid by solving a network capacity expansionproblem with a view to maximising its profit. The network is a naturalmonopoly that gives market power to the transco. If not regulated, thecompany will set long- and short-term charges in a way that maximises itsprofit but degrades welfare. The question is thus to find charges thatprovide the incentive to be efficient and permit the firm to recover its costs.This has different aspects.

Perhaps more than any other regulatory question, investments in the gridhave a strong flavour of asymmetry of information. The electric network isindeed a complex technology that may be difficult to grasp from outside theindustry. The problem of incentivising the transco will then require theregulator to offer a menu of contracts. Models of this type usually assumean explicit demand function for the services of the regulated company (forexample, Vogelsang, 2001). For the sake of analytical convenience, theseanalyses also resort to simplifications that neglect the multi-product natureof transmission services. We depart from these considerations and see thetransco as an ordinary profit-maximising firm. We also consider that thedemand for transmission services is not explicitly given by a demand func-tion but derived from the interactions between profit-maximising genera-tors and consumers. Last, we also assume an electricity network that offersdifferent point-to-point services. Our objective, in following that path, is


not to overlook the important problem of the market power of the transco,but to attempt to construct a partial equilibrium model with price-takingagents that we intend to expand in future developments.

Our objective can be stated as follows: consider a world of perfect infor-mation with price-taking agents (including the transco). The question is tofind the price signals that induce the transco to invest in an efficient networkand allows one to recover the cost of building this network. The need foradequate price signals for a transco derives from the now well-recognisedfact that a (price-taking) transmission company that receives only conges-tion charges and an additional fixed revenue has an incentive to underinvestin order to increase congestion. It is also admitted that the sole recovery ofcongestion charges allows the transco to recover only a relatively small frac-tion of the total costs of the grid (Pérez-Arriaga et al., 1995). One can thusexpect to see multi-part tariffs emerge as necessary instruments for achiev-ing both the efficiency and cost-recovery objectives. Non-linear tariffs arecommonly encountered in the regulatory literature for dealing with asym-metry of information. They also appear in non-convex economies (seeBjorndal and Jörsten, 2004, for a recent treatment through dual price func-tions). This latter context is the one adopted in this chapter.

In short, and possibly in contrast with the existent literature on transcos,we disregard the question of market power, but concentrate on the exist-ence of price signals that induce a price-taking transco to manage conges-tion and develop the network in an efficient way even though investmentsare lumpy. We do this by constructing a model that fully accounts for themulti-product nature of the transmission infrastructure, and develop anequilibrium framework where the demand for transmission services resultsfrom other agents’ actions. This was the approach formerly adopted instudies of operations and market design. We hope to pave the way to similardevelopments in two directions: on the engineering side, we introduce thecapability to accommodate realistic grid representations; on the economicside, we derive price-taking equilibrium conditions susceptible of beingextended to accommodate market power.

4. A MODEL OF LOCATIONAL AND CONGESTIONPRICING

The mix of discrete decisions (that one does not known how to decen-tralise through prices) and continuous decisions (that one knows very wellhow to decentralise through prices) found in the electricity grid is notunique. It also appears in another area of electricity restructuring, namelyunit commitment and plant dispatching. As mentioned above, Hogan and


Ring (2003) and O’Neill et al. (2004) address the question of finding pricesignals for inducing plants to start up and shut down. We transpose andslightly extend their reasoning to the questions of locational and conges-tion pricing.

The Grid Equilibrium Model

The following primal mixed-integer program (PIP) was considered byO’Neill et al. (2004) for studying price-based decentralisation of start-updecisions in unit commitment problems:

(6.1)

s.t.(6.2)

(6.3)

(6.4)

This model is a standard mixed-integer program.3 It is easily interpretedin the context of optimal centralised investments in a regulated electricitysystem. We accordingly assume a system where investment decisions in thegrid and in power generation are under the supervision of a benevolent,perfectly informed and perfectly rational (with unlimited computationalpossibilities) regulator. Let j�1, . . . , J denote the nodes of the electricalnetwork. We cast O’Neill et al.’s model in the locational pricing contextthrough the following interpretation.

● Let k�1, . . ., K designate an agent, consumer or generator, thatdecides to connect to the network.

● zk is the vector of binary variables representing locational decisionsof agent k. A component of zk, let zkj, is equal to 1 if k connects tothe network at location j�1, . . ., J. It is zero otherwise. All compo-nents of zk are 0 if agent k does not connect to the network. The deci-sion to connect to the network is a simplified view of a more complexdecision: an agent, whether a consumer or a generator, connecting tothe network indeed builds a plant of a certain type, a feature that isnot represented yet in the variable z. From a welfare point of view,the decision to connect therefore also implies the decision to build

xk 0, zk � {0, 1}n(k).

Bk1xk Bk2zk � bk � k

�k

Ak1xk �k

Ak2zk � b0

max �PIP � �k

ck xk �

kdk

zk


and hence a cost. We limit ourselves in this section to the sole loca-tional decisions. We shall later introduce different technologies, i�1,. . ., I, and extend the definition of zk by considering zkij�1 whenagent k connects a plant of technology i to location j of the network.

● xk is the vector of power injection/withdrawal of generator/consumerk. There are potentially as many components of xk as there are nodesin the network. We replace xk by signkxk in O’Neill et al.’s expressions(6.1) and (6.2) in order to distinguish between injections and with-drawals. In this updated expression, all xk are assumed positive andsignk is a diagonal matrix whose components are –1 for injections and1 for withdrawals; signkxk is thus the vector of the net nodal with-drawals of agent k.

● dk is the vector of fixed costs or benefits before any payment for loca-tional charges incurred by agent k when it connects to the network.dkj is thus the fixed cost of building a plant of (currently) unspecifiedtechnology in location j. Alternatively dkj can be interpreted as avector of fixed benefits or costs accruing to a consumer k when it con-nects to the network in location j.

● ck is the vector of the marginal costs of the generators. Alternatively,ck is the marginal utility of the consumer k. Both marginal costs andutilities are constant as in O’Neill et al.’s model. A slight complica-tion of the notation makes it possible to approximate non-linearconcave utilities and convex costs.

● Given the interpretation of the x variables, relation (6.2) representsthe load flow-based constraints that limit the injections and with-drawals in the network. Ak1 is thus identical for all generators andconsumers k and can be noted A. The matrix A consists of two parts.One row of the matrix expresses the balance between injections andwithdrawals in a lossless network. This is written:

(6.5)

where 1– is a row vector of 1. The other rows of A and the associatedconstraints express the limitations on the injections and with-drawals imposed by thermal limits on the lines. These are expressedusing the negatives of the power transfer distribution factors(PTDFs) that give the flow in each line of the grid as a result of theinjections and withdrawals signkxk. We keep the inequality formu-lation from O’Neill et al., even though relation (6.5) is an equality.This facilitates the notation and can easily be justified.

1 · �k

signk xk � 0,


● Ak2 is taken as 0 in this application and will no longer be used in thissection.

● The first component of b0 is the right-hand side of (6.5) and is thuszero. The other components of b0 are the thermal limits on the linesof the network (that are written in only one direction in order to sim-plify notation). Thus (6.2) represents the usual DC load flow approx-imation of network flows.

● Bk1 is an identity matrix and Bk2 a diagonal matrix of the capacities(noted m) installed by generator/consumer k; bk is identically 0.

Using these notations, PIP can be restated as program PIPLOC:

(6.6)

s.t.

(6.7)

(6.8)

(6.9)

Extensions and Restrictions

This program involves both investment (z) and operation (x) variables. Oneimmediately notices that PIPLOC boils down to the usual dispatch/welfareoptimisation problem extensively studied in the congestion managementliterature when the zk variables are fixed. This reduced problem is the basisof the study of short-term locational signals. PIPLOC therefore embeds thequestion of congestion management covered in Articles 5 and 6 of theRegulation. The aim of this chapter is to free the zk variables so as to alsocover the long-term signals of Article 4.

The model can easily be extended to accommodate several time seg-ments and contingencies. Let be the power generation/withdrawal ofgenerator/consumer k in time segment or contingency �. The model can berewritten as program (where mp stands for multi-period)

s.t.

max �PIPmpLOC

� �k�

�

c�k˛signk˛

˛x�k �

kdk˛signk˛

˛zk

PIPmpLOC

x�k

xk 0, zk � {0, 1}n(k).

xk � mkzk � 0

A�k

signk xk � b0

max �PIPLOC � �

kck˛signk˛

˛xk �k

dk˛signk˛

˛zk


.

As mentioned above, the model can also be easily extended to allow zk torepresent the decision to locate at a certain node and to build a certaintype of capacity. This extension only makes sense in a multi-period (timesegment) model. This model would thus consider variables zkij and where i is the type of plant. In order to simplify the discussion and to facil-itate the establishment of the correspondence between this chapter andO’Neill et al., we first limit ourselves to the simple formulation (6.6) to (6.9).

Program PIPLOC should be interpreted as the problem to be solved by aregulator operating under the following ideal conditions:

1. Perfect information: the regulator knows the marginal willingness topay of the consumers and the marginal cost of the generators. The reg-ulator also knows the locational fixed costs (the fixed cost of buildinga plant of given technology at some location).

2. Perfect competition: agents are price takers.3. Perfect congestion management: congestion is managed by nodal prices.4. Simplified electrical assumption: the network is lossless and its structure

is described by a PTDF matrix.

Assumptions (6.2), (6.3) and (6.4) are common in studies of restruc-tured electricity systems. Assumption (6.1) may look harsh, but is notmore stringent than the implicit assumption that prevailed in the formerregulatory regime where the regulator was also assumed to decide oraccept investments according to a perfect cost-minimisation model. Amain difference is that the regulator of the former monopoly regime didnot have to interpret the results of this optimisation model in terms oflocational prices. The regulator could be wrong with less damaging conse-quences. It is on these prices that we concentrate, assuming all the abovesimplifying assumptions.

Locational Charges

O’Neill et al. suppose that one can solve problem PIPLOC and we followsuit. The assumption is not unrealistic given the astonishing numericalprogress made these last 20 years in mathematical programming in generaland mixed integer programming in particular. Suppose this has been done

z�kij

x�k 0, zk � {0, 1}n(k)

x�k � m�

k zk � 0 � �

A��k

signk x�

k � b�0 � �


and let zk � k be the optimal location of the generators/consumers, we statea primal linear program PLIP(z):

(6.10)

(6.11)

(6.12)

where u0, uk and wk are the dual variables of the constraints (6.10), (6.11)and (6.12), respectively.

The dual DLIPLOC(z) of that problem is written as:

(6.13)

(6.14)

(6.15)

(6.16)

Note from the above discussion that signk in relation (6.14) multiplies thenodal components of u0A by –1 or1 depending on whether agent k is agenerator or a consumer at that node. The dual variables of the problemcan then be interpreted as follows:

● The first component of u0 is the price of electricity at the hub node.The other components of u0 are the values of the capacities of thelines (flowgates). These components of u0 are indeed the dual vari-ables of the thermal capacities of the lines and should be interpretedin the same sense as in usual discussions of congestion management.One can easily verify that this implies that u0A is the vector of thenodal prices of electricity.

● uk is the vector of scarcity premium on tight generation and con-sumption capacities. The constraints (6.14) express that the nodalprice at some active generation node (the corresponding nodal

u0 0, uk 0, wk unconstrained � k.

� uk mk wk dk signk � k

u0A signk uk ck signk � k

min u0b0 �k

wkzk

xk 0,

zk � zk (wk) � k

xk � mk zk � 0 (uk) � k

A�k

signk xk � b0 (u0)

max �k

ck˛signk˛

˛xk �k

dk˛signk˛

˛zk


component of u0A) is equal to the variable cost (the correspondingnodal component of ck) of a generator at that node which operatesat a positive level, but below capacity. This price is equal to the vari-able cost plus a scarcity premium if the generator operates at fullcapacity at that node (note that signk�–1 implies (u0A)k – uk�ck). Asimilar interpretation holds for consumers.

● wk are locational prices in the sense that a generator/consumer k payswkj to locate at j. Note that dk are costs for generators. A constraint(6.15) expresses that the locational price is greater or equal to thecapacity margin minus the fixed cost of the plant. The equality holdsfor a plant that is effectively located at that node. In that case the loca-tional price and the capacity margin pay for the cost of locating theplant at that node. Note that wk is unconstrained: its components canbe positive or negative. If negative, a component of wk should beinterpreted as a payment to be given to the agent to locate.

Efficiency Properties of Locational Charges

We now transpose O’Neill et al.’s result in the interpretation of locationaland congestion pricing.

Definition Efficient short- and long-term locational prices are prices (congestion) and (location) such that agents (generators and con-sumers) when charged ( , ) behave efficiently in the following sense.Suppose that these agents maximise their profits under the sole capacityconstraints:

In other words, agent k solves the following profit-maximisationproblem:

(6.17)

s.t.(6.18)

. (6.19)

Then the solutions , that they obtain are those desired by the regu-lator who solves PIPLOC. Moreover the injections and withdrawals balance and comply with the network constraints:

x�z�kx�k

xk 0, zk � {0, 1}n(k)

xk � mk zk � 0

max(ck � u0A) signkxk (dk signk � wk)zk

xk � mk zk � 0.

wu0wk

u0


. (6.20)

Note that this definition refers only to economic efficiency. It corresponds tothe notion of market-clearing price in O’Neill et al. but does not embed thenotions of cost reflectiveness or non-discrimination also foreseen by theRegulation. We examine these notions in subsequent sections of this chapter.

Proposition 1 Suppose that the regulator successively solves prob-lems PIPLOC and PLIPLOC( ) where is the solution to PIPLOC . Let and be part of the optimal dual solution of PLIPLOC( ). Then and are respectively efficient short- and long-run locationalprices. Let and be the decision of agent k. One has

Proof The proposition is a direct application of Theorem 2 in O’Neillet al., which is recalled in generalised form in the Appendix.

In this system, the regulator or the system operator receives the congestioncharges signk and the location charges . Because the state-ment of the problem does not contain any information on the cost of thenetwork, Proposition 1 contains no result on the balance of the TSO’sbudget. Similarly, there is no information on how the cost of the networkdepends on the locational decision of the agents; hence the result cannotexpress any cost causality.

It is remarkable that generators that decide to locate and produce makea zero profit. Similarly, consumers that decide to locate and consume have azero net welfare gain. This is the standard result of a perfectly competitiveequilibrium market which is recovered here by the application of O’Neillet al.’s Theorem 2. All other agents would incur a loss if they were to locateon the network and generate/consume electricity.

5. DISCUSSION

The above congestion and locational prices satisfy one objective of theRegulation, namely economic efficiency. They leave the decision to locate,generate and consume to the agents who pay these prices. The prices and are also quite distinct. The payment results from the decision tolocate on the network and can be interpreted as the long-term signals ofArticle 4 of the Regulation. These payments must be made irrespective ofthe consumption or generation levels, provided that the decision is made to

wku0wk

�k˛wk˛z�kx�ku0˛A˛�k

(ck � u0˛A) signk x�k (dk signk � wk˛) z�k � 0.

z�kx�k

wk

u0zwk

u0zz

A �k

signk˛

˛˛x�k � b0


connect to the network. The nodal prices entail congestion charges thatare directly proportional to consumption and generation. They complywith the obligation to use market-based congestion management methodsof Article 6 of the Regulation. Besides offering the two price signalsrequired by the Regulation, and also form true two-part tariffs in thesense commonly understood by economists.

As in O’Neill et al., the approach extends the commodity space beyondthe sole energy space. Both energy and the locational rights are priced. Oneshould note that it is of the essence not to convert the fixed part of thetariff into a proportional charge that would come on top of . Doing thiswould indeed convert the two-part tariff into two single-part linear prices,which cannot sustain the equilibrium in the presence of indivisibilities. Thisimportant remark appears neither in the regulation nor in the allocationmechanisms discussed by the TSOs and regulators in the Florence Forum.While Proposition 1 of Section 4 states that these non-linear tariffs satisfythe efficiency objective of Article 4 of the Regulation, none of the otherdesired properties (cost reflectiveness and non-discrimination) required bythe Regulation is achieved. This is not specific to this construction as two-part tariffs are often discriminatory in two senses. By construction they leadto average prices that are decreasing with quantity. This second-degree (forexample, Tirole, 1998) price discrimination is accepted by Courts. Two-parttariffs are also first-degree (ibid.) discriminatory prices. They differ by agentin the sense that, except for their proportional part which is equal to themarginal cost, the fixed part does not reflect the cost incurred by the gen-erator but the willingness to pay of the agent. This is unlikely to be acceptedby competition authorities without further justification. We therefore elab-orate on this discrimination:

● The are first-degree discriminatory in the sense that two genera-tors that locate at the same site may not necessarily pay the samelocational price. The only possible justification of that discriminationis that, in contrast with the common wisdom of the RegulatoryForum, price discrimination is sometimes necessary in order toachieve economic efficiency. This is what happens here: the signals aremuch more than simply locational; they are individual signals, tar-geted at each candidate in each location.

● The are not cost reflective. They bear no relation to the cost of thenetwork which is completely absent from the statement of theproblem.

These prices also lack other important desirable properties. First, thereis no guarantee that the sum of the payments accruing from congestion and

wk

wk

u0Awk

wu0A

u0A


long-run locational prices recovers the cost of the network, let aloneinduces an optimal development of the network. Second, the locationalprices do not reflect the incremental cost that the TSO would incur becauseof the decision of an agent to locate at some node of the network. This isnot surprising as the model PIPLOC does not contain any representation ofthe cost of the network.

In short, efficiency is achieved at the cost of sacrificing non-discrimination,cost reflectiveness and other desirable properties. One may thus search for amore involved tariff structure that exhibits more components, some of themrelated to the cost of the network. Before turning to this point, it is useful toreview Hogan and Ring’s (2003) comments on O’Neill et al.’s model and tosee whether they can be transposed to the problem of locational and con-gestion prices.

Hogan and Ring (2003) note that there may be many wk capable of sup-porting the dispatch equilibrium in a unit commitment problem. We wantto show that the same holds for the locational equilibrium. To see this,recall first that is the scarcity rent captured by the optimally locatedplants when operating in a perfectly competitive system. There may beseveral but take one of them. Transposing Hogan and Ring’s reason-ing to locational pricing, we first consider the minimal payment that should be paid to agent k in location j in order to induce a decisioncompatible with the desired by the regulator. Using Hogan and Ring’snotation and considering the case of a generator as an example, onedefines:

(6.21)

(6.22)

is the profit collected by agent k at location j if it behaves accordingto the regulator’s plan. This profit consists of a capacity scarcity margin

minus the incurred fixed cost . is the profit that the sameagent would collect if it made its own decision to invest in location j on thebasis of the expected capacity scarcity margin and its investment costdkj. The profit is what would be driving the generator locational deci-sion in a pure nodal system. Define:

. (6.23)

Suppose is strictly negative, then one can verify that a paymentof – to generator k at j) compensates this agent for:w�

kj(u)w�

kj(u)

w �kj(u) � min(0; Pkj � �

kj )

�kj

ukj

�kjdkj˛zkjukj˛mkj

Pkj

Pkj � max(�dkj ukj

mkj)zkj, zkj � {0, 1}.

Pkj � �dkj zkj ukj

mkj

z

wkj(u)uk

uk


● the opportunity cost of not following its own optimal strategy;● the cost of following the strategy of the regulator.

Any higher payment would overcompensate the agent.Conversely, one may also be interested in computing the maximal posi-

tive charge that can be levied on the generator that decides to invest beforeit modifies its decision. Define:

(6.24)

One can verify that a levy will not induce a generator k thatspontaneously decides to invest in location j to change its decision. Onecan thus imagine that a regulator could try to finance (part of) its pay-ments from charges . More generally, let KJ and KJ– be,respectively, the pairs (k, j) for which and have been defined. A w�

supports the locational equilibrium if:

(6.25)

(6.26)

This implies that:

(6.27)

is the minimal net payment that the regulator must be prepared to make inorder to induce the different agents to abide to its locational strategy .Nothing guarantees that this minimal payment is negative, that is, that theregulator will break even. Any deficit should be covered otherwise.

A similar reasoning applies to the consumers with a like conclusion. Theoverall conclusion in terms of covering revenue requirements is more posi-tive as soon as one considers generators and consumers together. Supposingthat the total surplus of the market, after paying the cost of the network, ispositive (a very mild assumption), then it is always possible by playing onthe w according to relations (6.25) and (6.26) to satisfy the revenue require-ment. The drawback is that this will increase price discrimination.

The obligation for the regulator to break even may thus lead one toincrease the discrimination between agents with respect to the original w.This suggests introducing constraints on the w that are directly inspiredby commonly found regulatory objectives. One may for instance wish toachieve:

z

� �(k,j)�KJ

wkj (u) �

(k,j)�KJ�

w�

kj (u)

w�kj � wkj (k, j) � KJ

w�kj � w�

kj (k, j) � KJ �

w�kjw

kj

wkj(u)w�

kj(u)

wkj � wkj

wkj (u) �

–�

kj.

�kj


● a budgetary balance. This would imply:

(6.28)

or

(6.29)

● Non-discrimination constraints. These impose some equalitiesbetween the negative and/or other equalities between the posi-tive .

Needless to say there is no guarantee that imposing these regulatory con-straints on top of (6.25) and (6.26) will lead to feasible . However, it isalways possible to try to get close to regulatory objectives by minimising theviolation of the constraints. We shall not elaborate any further on thesevariations in this section and return to the problem of cost causality.

6. COST CAUSALITY AND NON-LINEARLOCATIONAL PRICE

Section 5 concludes that the derived from problem PLIPLOC are dis-crimatory and not cost reflective. This is not surprising. Non-discriminatoryprices are well known not to be able to support perfect competition equilib-rium when there are indivisibilities and economies of scale. The absence ofcost reflectiveness could also be expected as the model PIPLOC used tocompute the does not link agents’ locational decisions to the grid struc-ture and hence to the grid costs. Also, note that the above tariff prices onlyenergy and location, while tariffs that include a capacity charge and hence athird component in the tariff are common in practice. The following devel-ops a richer set-up with the view of introducing cost causality into themodel.

Representation of the Network

Consider an extension of problem PIPLOC, constructed as follows:

● There exists a set of possible network configurations n�1, . . ., Neach of cost and PTDF matrix PTDFn. The TSO selects one ofd

n0

wk

wk

w�k

w�kj

w�kj

�k�

mw�km

zkm � Fixed Network Cost.

�k�

mw�km

zkm � 0


these configurations. Let z0 be an N-dimensional vector with �1 ifthe TSO selects the nth network configuration.

● The locational price takes the form of a two-part tariff, namely afixed charge for accessing the network and a proportional charge forreserving capacity (a capacity or demand charge). This reservedcapacity can be a single variable valid for the whole year, or a vectorif one implements seasonal reservations, for example, summer andwinter (see Pérez-Arriaga and Smeers, 2003). We first work with asingle annual capacity reservation in order to facilitate the discus-sion. With this interpretation xk plays both the role of energy gener-ated or consumed and reserved capacity.

● One introduces some causality between the decision to locate, thereserved injection and withdrawal capacity and the network struc-ture. This causality is complex in the real world. Strictly speaking, itderives from a network expansion planning problem. Because ourmethodology uses optimisation-type techniques to explore economicproblems, we adopt an approach theoretically justified in optimisa-tion terms but practically still to be elaborated and represent causal-ity through two types of constraints:

(a) The ‘skeleton network constraints’:

(6.30)

imply that the decisions of the agents to locate (the zk, � k�0)impose a certain minimal structure on the network. This resultsin a set of allowed network structures z0 (there may be several z0for a given set of location decisions zk).

(b) The ‘incremental network constraints’:

(6.31)

imply that the capacities reserved by the different agents(the xk, � k�0) further restrict the set of allowable z0 of thenetwork:

● The selection constraint:

(6.32)

implies that the TSO can select only a single network configuration.

�N

n�1zn

0 � 1

��k�0

G1k xk G20z0 � 0

F0z0 � �k

Fk�0zk � 0

zn0


Together these constraints restrict the set of possible confi-gurations of the network compatible with agents’ decisions to locateand reserve capacities. This might be sufficient to fully determine theconfiguration of the network, but need not be so. We therefore alsoleave the possibility that the TSO optimally (that is, through costminimisation) selects the configuration. One could obviously arguethat the constraints (6.31) can be merged with (6.30) and hence thatone needs to write only one set of constraints. A reduction of thenumber of constraints serves no useful purpose whether computa-tionally or economically. From a computational point of view, theformulation that explicitly retains the first two constraints is likely tobe tighter in an integer programming sense (see Wolsey, 1998) andhence more efficient. From an economic point of view, the two con-straints express different cost causalities and hence should be madeexplicit. We recognise that the construction of the constraints (6.30)and (6.31) may be difficult in practice, but leave it to further researchto investigate how they can be inferred from both engineering prac-tice and mathematical programming models. Before leaving thispoint, we note that the absence of any agent connecting to the grid(zk�0, xk�0, � k) eliminates the need for a network and henceallows the solution z0�0. This justifies setting the right-hand side ofthese constraints to zero. This also means that these constraintsdefine a cone when the integer restrictions are relaxed.

● Constraints (6.30) and (6.31) delineate the set of acceptable networkconfigurations. To each of them, one associates a matrix of PTDFcoefficients, PTDFn, and a vector of thermal limit constraints bn. Wedefine the set of acceptable injections and withdrawals in networkconfiguration n as:

(6.33)

where 1– is a row vector of 1 of appropriate dimension. Note thatwe ignore the constraint – bn�PTDFn in order to simplify thepresentation. Kn is thus the set of balanced injections and withdrawalsthat result in flows that do not exceed the thermal limits of the linesin configuration n. We also introduce a cone Cn associated with thenetwork configuration n as:

(6.34)Cn � �(xn0, z

n0) |

xn0

zn0

� Kn, 1 xn0 � 0 if zn

0 � 0, xn0 � 0 if zn

0 � 0�

xn0

Kn � {xn

0 | PTDFn xn0 � bn, 1 xk

0 � 0}


(recall that as defined above is equal to 1 if one selects networkconfiguration n and is zero otherwise).

Define:

(6.35)

and the cone:

(6.36)

It is obvious that (6.34) defines a convex cone. We also know thatthe Cartesian product of convex cones is a convex cone. (x0, z0)� C0 is thusa convex constraint that we shall impose on the network company.

The TSO selects a single �1 that minimises the cost and satisfies(6.30) – (6.31). Given this selection, the TSO can offer injection and with-drawal services x0 such that (x0, z0) � C0. The agents k request injection andwithdrawal services xk such that:

(6.37)

(6.38)

(6.39)

Relation (6.37) equalises the production and consumption of thenetwork services (injection/withdrawal). It also transforms the differentxk into a single non-zero vector x0 of injection/withdrawal services forthe selected network configuration. Relation (6.38) expresses that theTSO can only offer network services that are compatible with theselected configuration of the grid. This is the convex part of the produc-tion set of the transmission company. In contrast, (6.39) describes itsnon-convex part, namely that the TSO can select only one configurationof the network. In order to simplify the notation, we rewrite relation(6.37) as

E0x0 � �k�0

Ek xk � 0.

�N

n�1zn

0 � 1, z0 � {0, 1}N.

(x0, z0) � C0

�k

signk xk � �

nxn

0

d n0zn

0

C0 � {(x0, z0) | (xn0, zn

0) � Cn0, n � 1, . . ., N}

x0 � (xn0, n � 1, . . ., N), z0 � (zn

0, n � 1, . . ., N)

zn0


In order to fully use conic duality, we also define the cone Ck for each gen-erator or consumer k:

The Regulator’s Global Problem

Introducing a diagonal matrix sign0 with only –1 components, the exten-sion of the PIPLOC problem can be stated as:

Problem EPIPLOC

(6.40)

(6.41)

(6.42)

(6.43)

(6.44)

(6.45)

. (6.46)

In this model, (6.41) expresses the equality between produced andconsumed network services; (6.42) and (6.43) relate the selection of thenetwork configuration and the locational decisions of the agents;(6.44) states that only a single network configuration is allowed; (6.45)describes the production sets of all agents, TSO, generators and con-sumers. Specifically, C0 imposes the flows to be feasible for the selectednetwork configuration and Ck requires that an agent first needs to locateand reserve some injection or withdrawal capacity before generating orconsuming.

Note that in contrast with problem PIPLOC which is a linear mixed-integer program, the conic constraint (6.45) makes problem EPIPLOC anon-linear mixed-integer program. This non-linearity should not be a real

zk � {0, 1}n(k)

(xk, zk) � Ck, � k

1 · z0 � 1

��k�0

G1k xk G20z0 � 0

F0x0 � �k�0

Fkzk � 0

E0x0 � �k�0

Ekxk � 0

max �k�0

ck˛signk˛

˛xk �k

dk˛signk˛

˛zk

Ck � {(xk, zk) ̌ | xk � mk zk � 0, xk 0, zk 0}.


concern though as EPIPLOC can easily be restated as a linear mixed-integer program of the type precedingly developed by power engineersfor network capacity expansion (see Latorre et al. (2003) for a survey ofthese models). The non-linear version is used here because of its tract-ability for extending O’Neill et al.’s Theorem 2.4 We therefore againassume that this problem can be solved (for instance by solving the linearmixed-integer version of it) and note and , � k an optimalsolution.

The Continuous Version of the Regulator’s Problem

Let be the locational vector selected by the regulator after solvingEPIPLOC. Define:

Problem EPLIPLOC( )

(6.47)

s.t.

(6.48)

(6.49)

(6.50)

(6.51)

(6.52)

(6.53)

Note that c0�0, G10�0 and G2k�0 for k�0. We slightly generaliseO’Neill et al.’s Theorem 2 by relying on the extension of standard linearprogramming (LP) duality to (convex) conic programming duality (BenTal and Nemirovski, 2001) presented in the Appendix. Program EPLI-PLOC( ) is a conic program which comprises the conic constraint (6.53).As usual we write the dual variables of the constraints at the right of eachof them. The dual variables are associated with xk and zk appearingin the conic constraint (6.53). The conic dual of EPLIPLOC( ) may bewritten as:

zx*k, z*k

z

(xk, zk) � Ck, �k (x*k, z*k)

zk � zk � k (wk)

1 · z0 � 1 (�0)

� �k�0

G1kxk G20z0 � 0 (�g)

F0x0 � �k�0

Fkzk � 0 (�f)

E0x0 � �k�0

Ekxk � 0 (u0)

max �k�0

ck˛signk˛

˛xk �k

dk˛signk˛

˛zk

z

z

(xk, zk)(x0, z0)


Problem EDLIPLOC( )

(6.54)

s.t.

(6.55)

(6.56)

(6.57)

(6.58)

(6.59)

(6.60)

where 1– designates a vector of 1 of appropriate dimension.

Our interest in this extended model is to find prices that induce the TSO toinvest in the appropriate network configuration and the generators andconsumers to invest in the right location and operate efficiently.

Short-term Locational Charges

Before getting into this development, we first interpret the different con-straints of the dual of EPLIPLOC( ) and particularly those involving dualcones. Consider first the dual of the cone C0 associated with the TSO.Recalling the relations (6.34) to (6.36), one can write:

The following lemma characterises the dual cone :

Lemma 1

for some and some 0}.�n�nz*n0 �n bn,

C*0 ��n

C*0 n ��

n{(x*n

0 , z*n0 ) | x*n

0 � �n 1 � �n

PTDFn,

C*0

if zn0 � 0, xn

0 � 0 if zn0 � 0.

C0 � �n

{(xn0, z

n0) | PTDFn

xn0 � bnzn

0, 1 · xn0 � 0

z

�f , �g 0 u0, �0, wk � k unconstrained

(x*k, z*k) � C*k,�

��f Fk �0 wk � dksignk � z*k

�f F0 �gG20 �0 w0 d0 � z*0

� u0Ek � �gG1k � ck signk � x*k

u0E0 � x*0

min �0 �k

wk˛˛zk

z


Proof Consider the following linear program:

(6.61)s.t.

(6.62)

(6.63)

unconstrained, (6.64)

where the dual variables of the constraints are �n and �n, respectively.The dual of this problem can be restated as:

max 0s.t.

(6.65)

(6.66)

belongs to iff the minimal value of the primal problem isattained and is zero. This happens iff there exists a dual solution �n0,�n unconstrained, such that ��n1 - �n PTDFn and . Thiscompletes the proof.

The proof of the lemma immediately suggests an interpretation of the con-straint � . The condition (6.65) indeed implies that is a setof nodal prices where �n is the electricity price at some hub and �n are theprices of the flowgate capacities in network configuration n. Condition(6.66) implies that is bounded below by the merchandising surplus.Because the value of the primal and the dual are equal, one also has:

and by duality when ; is then exactly equal to themerchandising surplus. The above can be summarised by saying that

iff is a vector of nodal prices associated with flow inconfiguration n and is the corresponding revenue of the TSO accruingfrom congestion charges.

z*n0

xn0x*n

0(x*n0 , z*n

0 ) � C*n0

z*n0zn

0 � 0z*n0 � �n

0b

xn0 x*n

0 zn0 z*n

0 � 0

z*n0

x*n0C*n

0(x*n0 , z*n

0 )

z*n0 �nbnx*n

0

C*n0(x*n

0 , z*n0 )

�n 0, �n unconstrained

zn0�nbn � z*n

0

��nPTDFn �n1 � x*n

0 xn0

zn0 0xn

0

1 · xn0 � 0 �n

�PTDFn xn

0 bnzn

0 0 �n

min x*n0 xn

0 z*n0 zn

0


The TSO’s Problem

We are now equipped to interpret the role of the TSO. Denote withan upper – the dual variables that are solutions of the dual problemEDLIPLOC( ). Consider the TSO’s problem CIP0:

(6.67)

(6.68)

(6.69)

Note by an upper ‘–’ and a ‘�’ prime the optimal solution of CIP0 (forexample, ). We briefly discuss the interpretation of the different expres-sions appearing in this problem.

Note first that c0 is identically zero because the TSO does not have vari-able operating costs. By (6.55), is equal to x*0 and hence, as shown inLemma 1, is a vector of nodal prices in network configuration n. As aresult, is the merchandising surplus innetwork configuration n, and hence the congestion revenue of the TSO inthat configuration. d0 is the vector of investment costs of the differentnetwork configurations. is a vector of payments/charges imposed by theregulator to induce the TSO to select the adequate network configuration.Finally, and are locational payments, respectively, resultingfrom the generator’s/consumer’s decision to locate and reserve certaininjection and withdrawal capacities. These two last terms do not carrymuch intuitive interpretation. This is not really surprising: one couldnot expect to find an easy interpretation of cost causality when the causalrelationships that drive the development of the grid are themselves murky.We shall come back to these terms later in the discussion. In conclusion,relations (6.67) to (6.69) represent the problem of a TSO that maximisesthe profit accruing from short- and long-term payments when investing inthe grid.

The Generator’s and Consumer’s Problem

We now turn to the sets Ck. Recall that Ck is defined as:

We have the following lemma:

Ck � {(xk, zk) ̌ | mkzk � xk 0, xk 0, zk 0}.

�lgG20�lf F0

w0

�u0E0x�0 � �x*n0 x�n

0 � z*n0 z�n

0 � �nbn

u0E0

x�0

1 · z0 � 1; z0 � {0, 1}N.

(x0, z0) � C0

max �CIP0 � �u˛0E0x0 � (d0 lf F0 lgG20 w0)z0

z


Lemma 2

Proof Consider the following linear program:

(6.70)

s.t.

(6.71)

. (6.72)

The corresponding dual can be stated as:

max 0

s.t.

�kmk�z*k

��k�x*k

�k0

iff the primal problem has a solution and the objectivefunction value is equal to zero. This requires that the dual problem alsohas a solution and hence that:

.

Note that zk�0, xk�0 implies by the complementarity conditions that:

�kmk�z*k and ��k�x*k and hence that .

We are now equipped to interpret the actions of agent k. This latter solvesthe problem:

s.t.

(xk, zk) Ck.�

max �CIPk� (ck˛signk˛

˛ lgG1k u0Ek)xk (dk˛signk˛

˛ lf Fk � w*k)zk

x*k z*kmk

� 0

x*k

z*kmk

0 and z*k 0

(x*k, z*k) � C*k

zk 0, xk 0

mkzk � xk 0 �k

min z*k zk x*k

xk

C*k � {(x*k, z*k) ̌ | z*k 0,

z*kmk

x*k 0}.


Denote again with an upper ‘–’ and a ‘�’ the optimal solution of problemCIPk (for example, ). The different expressions appearing in this problemcan be interpreted as follows. ck signk is the generation cost or the will-ingness to pay of agent k. is the revenue collected/paid by agent kfor injection/withdrawal of electricity in a nodal price system. isthe demand charge component due to the long-term locational prices. Itcan be positive or negative depending on the matrix G1k. dk signk is thecost incurred by agent k in order to locate on the network and to build theequipment. is the charge accruing from the locational decision. Thischarge can also be positive or negative depending on the matrix Fk. isa charge/payment levied by the regulator in order to induce agent k to selectthe right location. Again CIPk represents the problem of an agent that max-imises the profit accruing from short- and long-term payments when decid-ing to locate and to operate on the grid.

Compatibility between the TSO and Other Agents’ Decisions

We need to show that the decisions of the generators and consumers, whensolving their respective subproblems, combine to give energy flows thatsatisfy the constraints of the configuration selected by the TSO. We alsoneed to explore whether financial flows balance. Finally, we also hope thatthe obtained behaviours do not depart from those found in the usual theoryof congestion management. These properties are obtained from the solu-tions of the primal and dual problems EPLIPLOC( ) and EDLIPLOC( ) aswe shall now show. We begin with the less usual result coming from conicduality.

The following lemmas cast our model in the standard theory of nodalpricing and hence directly relate to congestion management.

Lemma 3 Let and be respectively part of the primaland dual optimal solutions of EPLIPLOC( ) and EDLIPLOC( ). We have:

. (6.73)

(6.74)

s.t.

(6.75)

(6.76)1 · xn0 � 0

PTDFnxn0 � bn

max x*n0 xn

0

If z�n � 0, then x�n0 solves:

x�n0 � 0 if z�n

0 � 0

zz(x�*0 , z�*0

)(x�0, z�0)

zz

w*kz�k

lf Fkz�k

z�k

lgG1k x�k

u0 Ek x�k

x�k

x�k


where is a vector of nodal prices in network configuration .

Proof Let be the network configuration selected by the TSO ( in(6.52)). Then (6.73) follows immediately from the definition of C0. Theresult for n � follows by noting that the problem (6.74), (6.75) and(6.76) is identical to the problem (6.61) to (6.64) of Lemma 1 after replac-ing by 1. can be interpreted as a vector of nodal prices as shownin Lemma 1.

Lemma 3 recovers a fundamental result of congestion management innodal pricing (Hogan, 1992). Given a structure of the network, is thevector of nodal prices and the TSO selects an offer of injection andwithdrawal services that maximises the value of the network. Then (6.73)requires that the energy flows can only be different from zero in the selectedconfiguration . and (6.75) requires that the thermal limits of the lines inthe selected configuration are not violated. Altogether problem(6.74)–(6.75) requires that the system operator maximises the value of thenetwork capacities, using the nodal prices .

Lemma 4 is the congestion charge paid by agent k.

Proof As argued in the discussion following Lemma 1, the constraint(6.55) at optimum boils down to:

.

The result then follows by noting that Ek�signk.

Combining Lemmas 3 and 4, one sees that the congestion managementpart of the problem is identical to the one described in the usual theory ofnodal pricing when the network configuration is given.

In contrast, the remaining expressions and donot have any standard interpretation. These are the ‘cost causal’ parts ofthe tariffs that ‘price’ the impact of the decision to locate and to reservecapacity on the network. They reflect the cost incurred by the TSObecause of these decisions. The lack of interpretation comes from the factthat constraints (6.30) and (6.31) are of a combinatorial nature and do nothave any standard economic meaning. But we shall see that they inducethe right behaviour on the TSO and the agents k. In no way do these‘locational prices’ satisfy the criterion of transparency demanded by theRegulation. But that should be expected, as the network expansionprocess is itself a complex problem that is not transparent. Before

lTg G1klT

f F0, �Tf Fk, lgF20

x*n0u0 �

u0Ekxk

x*n0

n

x�n0

x*n0n

x*n0zn

0

n

zn0 � 1n

nx*n0


exploring the global compatibility of the behaviour of the differentagents, we note that complementarity slackness implies at least someaccounting relations between the charges imposed on the agents and therevenue of the TSO.

Lemma 5 (i) . The sum of the locational chargespaid by the generators and consumers is equal to the locational

revenue received by the TSO. (ii) . The sumof the locational charges paid by the generators and consumersis equal to the locational revenue received by the TSO.

Proof Obvious from (6.49), (6.50) and complementarity conditions.

Decentralisation of the Decisions

It remains to show that the decentralised solutions of the TSO problem andof the different agent k problems solve the regulator’s problem, that is, thatthey induce the TSO to select the network configuration desired by theregulator and similarly that they incentivise generators and consumers tolocate and develop as desired by the regulator. Finally, we want to be surethat the electricity and transmission markets clear and that all agents,including the TSO, break even.

The following proposition is a direct consequence of Theorem 1 statedin the Appendix:

Proposition 2 Suppose all agents k, when solving problem CIPk, arecharged/pay the price found in the solution of EPLIPLOC( ).Then the solutions found by these agents are identical to the solution ofproblem EPLIPLOC. Moreover energy markets and the transmissionservice market clear.

Proof The result immediately derives from Theorem 2 in Appendix 6A.

This implies the following corollary:

Corollary 1 Lemmas 3, 4 and 5 hold if the solutions of problemEPLIPLOC( ) are replaced by those of the CIPk.

This implies that prices and allow one to fully decentralisedecisions among agents. Congestion management boils down to nodalpricing and retains its usual properties. TSO maximises the value of thenetwork and collects the merchandising surplus. Also the demand charges

wku0, lf, lg, w0

z

zu0, lf, lg, wk

lgG20z0

lgG1kx�k

lgG20z�0 � � klgG1kx�klf F0z�0

lf Fkz�k

lf F0z�0 � �k�0lf Fk z�k


components of the location prices paid by the users of the network andreceived by the TSO balance. Most remarkable, the budget of the TSO,including receipts from long- and short-term prices and taking investmentcosts into account is balanced.

7. NON-DISCRIMINATORY PRICES

The Regulation and the work of the Florence Regulatory Forum requirenon-discriminatory prices in order to induce competition. The above tariffsare only partially non-discriminatory. The line congestion charges in thevector are the same for all agents k and hence are non-discriminatory.This is the usual result of congestion management by nodal prices. Thesame is true for the demand charges of the long-term locational signals, atleast if one accepts the description of causality given by relations (6.30) and(6.31). In contrast the wk are truly discriminatory. This discrimination isjustified by the objective of economic efficiency. In contrast with thecommon wisdom underlying both the Regulation and the work of theForum, there are indeed situations where discrimination is necessary forachieving efficiency. Ramsey pricing is the best-known illustration of theusefulness of discriminatory prices in the network industries: its justi-fication is that it makes everyone better off. Still it remains a discriminationand hence may be unlawful. One can obtain non-discriminatory prices, butat the cost of some economic inefficiency. This is achieved by constructinga model where one explicitly requires that the wk are equal for all genera-tors and consumers. We briefly turn to that question.

Recall that the model of Section 6 leads to three-part locationalprices that are efficient and cost reflective. The congestion and demandcharges are non-discriminatory; the locational price is cost reflective andnon-discriminatory but the locational price (the w) remains both non-costreflective and discriminatory. The question we address here is the removalof the discrimination. We first take up the question on the simple model ofSection 4 and then extend the discussion to the more involved model ofSection 6.

Consider the problem PIPLOC and the dual problem DLIPLOC( ) ofproblem PLIPLOC( ). We modify DLIPLOC( ) into CDLIPLOC( ) (con-strained dual LIP problem) by imposing the non-discrimination constraintthat all locational charges are identical. This leads to:

(6.77)min u0b0 w�k

zk

zzzz

u0


s.t.

(6.78)

(6.79)

. (6.80)

The only difference between CDLIPLOC( ) and DLIPLOC( ) is theimplicit introduction of the non-discrimination constraint:

that expresses that locational charges are identical for all agents.The corresponding primal problem CPLIPLOC( ) is written as follows:

(6.81)

s.t.

(6.82)

(6.83)

(6.84)

(6.85)

Note that the constraint xk – mkzk�0 guarantees that zk0. It is obviousthat problem CPLIP( ) is a relaxation of PIPLOC( ) because it aggregatesthe capacity objectives zk� k � k of the regulator into the single constraint(6.84). Strictly speaking, this does not solve the regulator’s problem butonly approximates it. Nevertheless, it may be a very good approximation aswe discuss now. The solution of PLIPLOC( ) aims at selecting the optimalchoice of the capacities for each agent in different locations. It is unlikelythat the regulator has enough information to differentiate between the coststructure of the different agents. Suppose this is indeed so and the regula-tor cannot differentiate between agents. He/she is then able to select onlythe best mix of technologies and capacities at the different locations. Thevector describes the total capacity of each technology built at each�

kzk

z

zzz

xk 0.

�k

zk � �k

zk

xk � m zk � 0 � k

A �k

signkxk � b0

max �cksignkxk �dksignkzk

z

wk � w � 0 � k

zz

u0 0, uk 0,�w unconstrained

�ukmk w dk signk � k

u0A signk uk cksignk � k


location. The price signal therefore depends on both location and tech-nology but not on the agent. As a locational signal, it may still be consid-ered as discriminatory to the extent that it differentiates the locational priceas a function of the technology. This may be forbidden by law except if itcan be argued that this differentiation is justified by cost casuality. Forinstance, gas and coal plants are not expected to operate in the same wayand hence could imply different investments in the network. At least, theapproach does not differentiate the locational signal by agent, somethingthat would certainly be forbidden.

The problem of this formulation is that the prices found by solving prob-lems (6.77) to (6.80) do not guarantee that the constraint (6.84) will be met inthe decentralisation process. Agents may still build too much or too littlecapacity. This is the welfare loss implied by the non-discrimination constraint.

This discussion can be extended to the model of Section 6. We considera slightly different (but equivalent) formulation of the non-discriminationproblem based on that model:

� 0

� 0

� 0

�1

This model is obtained from PLIPLOC( ) by two operations:

1. one dualises the constraint (6.52) zk� k, and2. one imposes that the prices associated with these constraints are the

same for all generators and consumers.

One immediately sees that the second equation imposes the non-discrimination constraint on it that requires that all w are identical in theobjective function of the problem.

z

z

zk �{0, 1}n(k).

(xk, zk) � Ck

1 · z0

��k�0

G1k xk G20z0

F0x0 � �k�0

Ekxk

E0x0 � �k�0

Ekxk

infw w0 z0 w �k�0

zk max�k

ck signk xk � (d0 w0)z0 �

k�0(dk signk ˇ w)zk

w


This formulation will guarantee non-discriminatory locational prices butit will not ensure that the decentralisation process will lead to locations andinvestments that meet the objective or of the regulator. This isthe same phenomenon as mentioned for the preceding model. Generatorsand consumers may over- or underinvest, therefore entailing a loss of eco-nomic efficiency. It is impossible to foresee a priori the extent of the loss ofefficiency that would result from this non-discrimination constraint on thesole basis of the model formulation. But it is certainly quite possible toinvestigate the question numerically.

Finally, note that an obvious improvement of both models would beto differentiate the locational charge depending on whether agent k injectsor withdraws at some node. This implies defining w as the locationalcharge for all generators and w� for all consumers. This extension isstraightforward.

8. INSTITUTIONAL DISCUSSION

The above discussion offers an abstract setting that is useful to try to relateto current practical proposals. We again rely on the discussion provided byRosellón (2003) that has already been invoked in Section 3.

It should be clear that our formulation of the TSO problem best fits thetransco model. In this model a single company is in charge of both theinvestments and the operations of the transmission system. This proposal,which is favoured in Joskow and Tirole (2000) is implemented in the UnitedKingdom. Combined with an appropriate incentive regulation, this experi-ence is credited to be very successful. The current European institutionalframework makes no recommendation in favour of that system and someof the language of the Regulation (Article 6, paragraph 6) even seems notto recommend it. What is certain is that it is impossible to impose a transco-like solution throughout Europe, let alone to adopt a single transco for thewhole of Europe. Nevertheless it remains useful to retain this interpreta-tion both because it is implemented in practice and also because that imple-mentation is successful.

Our model obviously differs from the exact implementation of thetransco in the UK.5 In this chapter, the TSO receives instructions on howto set the long-term signals; it also receives some money transfers throughthe w0 and �0 that induce it to select the right network configuration. Themodel also supposes a well-informed, quite knowledgeable and very intru-sive regulator. However, the regulator does not need to solve problemPIPLOC him/herself. The transco, or an independent consultant can do soon the basis of commonly agreed data and assumptions. But the regulator

zk � zk� kzk


is assumed to be comfortable with the whole process and to agree that the(or in case of non-discriminations) are desirable objectives.The merchant line is the other approach to grid investments recalled in

Section 3. It is analytically grounded in the theory of nodal prices and theirextension into long-term financial transmission rights as hedging instru-ments. The ISO model, where operations and ownership of the grid areseparated, provides the institutional background of the approach. Hogan’s(2002) theory of merchant lines essentially extends the role of financialtransmission rights from operations to both operations and investments.Joskow and Tirole (2005) argue that this can only be done under drasticassumptions that are violated in practice. In this chapter we concentrateon one of these assumptions, namely the lumpiness of investments, andexplore its possible consequences. Our conclusion is that it is still possibleto decentralise lumpy investment in the grid provided one invokes a morecomplex set of prices that covers not only congestion but also accesscharges. In other words, there is a combination of access and congestioncharges that provides the necessary incentives for investments. It is remark-able that this is exactly what the Regulation foresees, but without indicat-ing how this can be done. The reality is that the derivation of these chargesis a very demanding task. It indeed requires first, extending the role of theISO from the sole operation of the existing system, to include both theoperation and investments in the grid. This extension raises several ques-tions that we do not discuss here; however, it retains a key property of theISO model in the sense that this latter does not need to own the grid. Ourabstract model is thus also fully compatible with the notion of a largeregional transmission operator (RTO) (for example, PJM6). The creationof RTOs is still a long way off in the European context, but it is definitelymore possible than a single transco. The task of the RTO becomes formi-dable, however. Instead of auctioning only physically feasible financialtransmission rights that only cover congestion charges, this global RTOshould now auction two types of long-term contacts, correspondingrespectively to the access and congestion charges. In the same way as con-gestion charges need to be carefully tuned in order to ‘get the prices right’,the access charges also need to be well tuned in order to induce the rightinvestments in the grid. Needless to say, ownership of access rights on topof congestion rights by generators as a result of the auction would exac-erbate the generator’s market power in the sense of Joskow and Tirole(2000) and Gilbert et al., (2004). The answer may lie in Hogan’s suggestion(Hogan, 2002) that transmission companies would probably be the mainowners of these rights.

� kzkzk


9. FURTHER QUESTIONS

The locational prices obtained in Section 6 satisfy two objectives of theRegulation. They are economically efficient and cost reflective. They alsoallow for the separation between short- and long-term locational pricesforeseen by the Regulation. Also all the theory of nodal price/flowgate con-gestion management remains unchanged.

Revenue adequacy is a key element of the theory of congestion manage-ment. It states that the TSO will normally receive a non-negative profit fromcongestion management. This non-negative profit occurs in a short-run equi-librium problem, that is for given network capacities, provided that certainconditions related to ex ante and ex post contingencies are met. We retain thestandard result of revenue adequacy in congestion management operatedunder nodal pricing in our more general set-up. Moreover, by Theorem 17 allagents make zero profit at equilibrium. This would suggest that revenue ade-quacy can be extended to encompass both long- and short-term signals. Butthere is a difference between this extended revenue adequacy and the stan-dard congestion management result. Congestion payments between TSOsand agents k are in balance as can be seen by multiplying relation (6.48) by

. Revenues and expenses from demand charges also balance as can be seenfrom the complementary condition of constraints (6.42) and (6.43).

is the total residual charge paid by the TSO and the other agentsto the Regulator. Assuming that the electricity system has a positive socialvalue, this charge can only be non-negative. We therefore recover a revenueadequacy property where the residual profit is this time with the regulatorand not with the TSO. We discussed in Section 5 how Hogan and Ringexpanded the scope of prices capable of sustaining the equilibrium. Theirreasoning can be extended to the more general model of Section 6. All thisshould be studied further, probably numerically.

Still, these locational prices leave several open problems. We brieflymention some of them.

The realism of the models can be improved. The model of Section 6assumes that the system is operated during a single time period, thereforeassimilating the demand charge to an additional energy charge. Thisrestriction can easily be removed. It has been a tradition among power engi-neers to plan the network for the peak period. This is unlikely to be correctbecause transmission flows are not necessarily highest during the peak.This suggests that a set of reference periods (seasons) should be introducedthat are critical for designing the network and hence that can be used fordetermining the demand charge. This can easily be done by extending themodel of Section 6. This would allow for a richer description of the costcausality and hence would lead to a richer set of non-linear tariffs. The

l*0 � kwkzk

u0


same can be said about the inclusion of the choice of a technology. The zkvariables have been defined to refer only to locational decisions. As arguedin Section 4, it is easy to expand their interpretation to encompass a tech-nological choice. This would also lead to a richer set of non-linear tariffs.

Recall that the models discussed in this chapter fail to achieve fully non-discriminatory tariffs. Bjorndal and Jörnsten (2004) were able to constructnon-discriminatory dual price functions for the unit commitment problemtreated by O’Neill et al. (2004). The possibility of extending their analysisto the grid problem constitutes another area of interest.

The following issues may be more challenging. The current model sup-poses that the cost causality can be expressed through relations of the (6.42)and (6.43) type. This is certainly true in principle but does not say anythingabout how to derive these relations. This problem raises both power engi-neering and optimisation questions. Consider first the electrical engineer-ing issue. A main feature of model EPIPLOC is to embed a description ofthe production set of the TSO. This description involves both causal rela-tions ((6.30), (6.31)) as well as a description of the transportation possibil-ities of the different network configurations contemplated by the TSO((6.33) and (6.34)). The representations of the production set of the TSOin economic models is unusual but not totally absent from the literature(Vogelsang, 2001). Our relations ((6.30), (6.31) and (6.33)) are stylised fromthe following considerations. We assume that the TSO selects its invest-ments using a capacity expansion model. There are several examples ofthese models in the electrical engineering literature (for example, Latorreet al., 2003, for a survey of these models). Many of these models are of themixed-integer programming type. We suppose that one of these models wasselected. Using the language of combinatorial optimisation, relations(6.30) and (6.31) are valid inequalities, derived from this mixed-integerprogram, that describe the polyhedron of feasible solutions of EPIPLOC.The construction of these inequalities is currently an area of intenseresearch in combinatorial optimisation. It is not clear, however, that theseinequalities have so far been derived for the type of model involved innetwork capacity planning. Also, current valid inequalities are limited tolocal description of the feasible polyhedron. In short, the introduction ofrelations (6.30) and (6.31) makes a lot of sense from the point of view oflogic, but their construction remains to be explored.

Consider now the following economic issue: the current model is estab-lished under assumptions of perfect knowledge and absence of marketpower. Relaxing each of these assumptions creates a whole set of additionalcomplexities. Consider the assumption of perfect knowledge first. Followinga traditional assumption of the ‘old’ theory of regulation, the regulator issupposed to know the cost of the generators, the willingness to pay of the


consumers and the set of possible network configurations of the TSO. Thelimitations of this type of assumption have been extensively explored inthe new theory of regulation, using the notion of asymmetry of informa-tion. Embedding this machinery in extensions of Theorem 2 of the appen-dix would add a whole new dimension to the work.

The absence of market power is another limitation of the model. An exten-sion of the model discussed here to one that would accommodate marketpower is straightforward to formulate: the problem EPIPLOC should bereplaced by a mathematical program subject to equilibrium constraint(MPEC). This model would request that the regulator selects the long- andshort-term locational prices in order to induce agents operating in an oli-gopolistic market to behave in such a way that they maximise welfare. Butit is not clear that this problem has a solution. Again, the real difficulty is toextend the decentralisation result of Theorem 2 to that more general set-up.

The static nature of the model is certainly one of the major shortcom-ings of the above discussion. Static models are common when discussingequilibrium problems but the simplification is a real drawback when itcomes to implementation. Bushnell and Stoft (1996) have initialisedthe discussion of the dynamics of investments in the electrical gridsand their consequences on the validity of existing financial transmissionrights. This has since been extensively elaborated in various papers, amongthem the discussion of merchant lines. It is totally absent from this set-up.

10. CONCLUSION

The idea of separating long- and short-term locational signals in theRegulation on Cross-border Exchanges in Electricity is probably a good one.It offers the opportunity to depart from simple linear tariffs that are notsufficient to induce the right investments and locational decisions in a systemplagued by discrete decisions. But this separation leaves open how to con-struct the long-term locational signals. It is easy but misleading to pretend tosolve the problem by cost allocation rules. Unless proved otherwise bynumerical experiment, there is no reason to believe that these will generatethe right signals. This chapter considers the goal of finding an alternativeapproach or at least to identify ideal abstract conditions that would allow forsuch an alternative. It retains three criteria referred to in the Regulation,namely economic efficiency, cost reflectiveness and non-discrimination. Itfinds that the three criteria cannot be achieved simultaneously, even underideal conditions. But one can at least trade non-discrimination for the othercriteria. It also finds that transparency of the long-term signals is probablyhopeless. But this is not surprising as the network expansion process is itself


a murky one. Last and probably more important it provides a frameworkwhere both the long- and short-term signals can be cast, and the usal theoryof short-term signals remains completely unaffected.

APPENDIX METHODOLOGICAL BACKGROUND

The following provides a small extension of the main theoretical result ofO’Neill et al. The extension is appropriate for introducing cost causality inthe network problem.

Consider the following mixed integer conic problem

(6A.1)

s.t.

(6A.2)

(6A.3)

(6A.4)

where is a convex cone. The only difference between this CIP problemand O’Neill et al.’s PIP is the replacement of the constraint by(xk, zk) � Ck. Because also defines a convex cone CIP generalises PIP.Let be the dual cone of Ck.One knows that (strictly feasible) conic programs satisfy the same dualityproperties as linear programs. We then extend O’Neill et al.’s formalisationas follows. Let be the value of zk dual variables in an optimal solution toCIP, define the following primal conic program:

s.t.


�k

Ak1xk �k

Ak2zk � b0 � k

PCIP (z*) max �PCIP � �k

ckxk �k

dkzk

z*k

Ck � {xk, zk, xkxk zkxk 0 � (xk, zk)� Ck}xk 0

xk 0Ck

(xk, zk) � C0;�zk � {0, 1}n(k)


�k

Ak1xk �k

Ak2zk � b0

max �CIP � �k

ckxk �k

dkzk


Using duality theory in conic programming, one states the following dualproblem:

s.t.

wk unconstrained

Theorem 1 of O’Neill et al. is readily extended into:

Theorem 1

where * indicates the optimal solution value for the respective problems.

The proof is identical to that of O’Neill et al.Consider now a set of price vectors . O’Neill et al.’s agent

problem can be adapted into:

.

We then define a competitive equilibrium as a set of prices { } forall k and allocations { } for all k such that at the prices { }, theP*0, Pz*

kx*0, x*k

P*0, Pz*k

(xk, zk) � Ck,�zk � Zk

Bk1xk Bk2zk � b1

CIPk max �CIPk� (ckxk dkzk) � P0(Ak1xk Ak2zk) � Pz

kzk

(P0, Pzk)

�*CIP � �*PCIP � �*DCIP

(xk, zk) � Ck.

y0 0,�yk 0

y0Ak2 ykBk2 � dk wk � z*k�y0Ak1 ykBk1 � ck � x*k�

DCIP(z*) min �DCIP � y0b0 �k

ykbk �k

wkz*k

(xk, zk) � Ck

zk � z*k � k


allocations { } solve PIPk for all k, and the market clears:

O’Neill’s Theorem 2 can then be generalized as:

Theorem 2 Let { } be the solution to CIPk and PCIP(z*) and let{ } be the solution to CDIP(z*). If and thenthe prices { } and allocations { } for all k is a competitiveequilibrium.

Proof The proof is almost identical to the one reported by O’Neill et al.The only difference is the replacement of the two complementarityconditions:

by the generalised complementarity condition:

which states that:

.

These generalised complementarity conditions are used to proveby noting that:

because:

(ck � y*0Ak1 � y*kBk1)x**k (dk � y*0 Ak2 � y*kBk2 � w*k)z**k � 0

�**CIPk� �*CIPk

(y*0Ak1 y*kBk1 � ck)xk (y*0Ak2 y*kBk2 w*k � dk)zk � 0

� y*0Ak1 y*kBk1 � ck

y*0Ak2 y*kBk2 w*k � dk� � �xk

zk� � Ck

�xk

zk� � Ck

�xk

zk��xk

zk� � Ck,Ck

0 � (y*0Ak2 y*kBk2 w*k � dk)�zk 0 �k

0 � (y*0 Ak1 y*kBk1 � ck)�xk 0 �k

x*k, z*�ky*0, w*k

w*k � Pzky*0 � P0y*0, y*k, w*k

x*0, z*�k

�kAk2zk � b0.�kAk1xkx*0, x*k


�

and the definition of .

NOTES

* Support from the French Commission de Régulation de l’Energie (CRE) is gratefullyacknowledged.

1. The Electricity Regulatory Forum of Florence was set up to discuss the creation of a trueinternal electricity market. The participants are national regulatory authorities, memberstates, the European Commission, transmission system operators, electricity traders,consumers, network users and power exchanges. See http://europa.eu.int/comm/energy/electricity/florence/index_en.htm.

2. See Stoft’s chapter in this book:3. See any textbook in integer programming, for example, Wolsey (1998).4. The theorem is recalled in the Appendix.5. For a description of the electricity system in England and Wales, see Joskow’s chapter in

this book.6. See Joskow’s chapter.7. See the Appendix.

REFERENCES

Ben-Tal, A. and A. Nemirovski (2001), ‘Lectures on modern convex optimization:analysis, algorithms, and engineering applications’, Society for Industrial andApplied Mathematics (SIAM), Philadelphia, PA.

Bjorndal, M. and K. Jörnsten (2004), ‘Equilibrium supported by dual price func-tions in markets with non-convexities’, Department of Finance and ManagementScience, Norwegian School of Economic and Business Administration, Bergen,Norway.

Bushnell, J.B. and S.E. Stoft (1996), ‘Electric grid investment under a contractnetwork regime’, Journal of Regulatory Economics, 10, 61–79.

Crew, M.A., C.S. Fernando and P.R. Kleindorfer (1995), ‘The theory of peak-loadpricing: a survey’, Journal of Regulatory Economics, 8, 215–48.

Curien, N. (2003), ‘Cost allocation methods’, in F. Lévêque (ed.), Transport Pricingof Electricity Networks, London: Kluwer Academic, pp. 73–101.

European Parliament and Council (2003), ‘Regulation (EC) No. 1228/2003 of theEuropean Partiament and of the Council of 26 June 2003 on Conditions forAccess to the Network for Cross-border Exchanges in Electricity 26.6.2003’,Official Journal, L 176, 15/07/2003, 0001–0010.

Gilbert, R., K. Neuhoff and D. Newbery (2004), ‘Allocating transmission to mitigatemarket power in electricity markets’, Rand Journal of Economics, 35 (4), 691–709.

Ck

�xk**zk**� � Ck

� y*0Ak1 y*kBk1 � ck

y*0Ak2 y*kBk2 w*k � dk� � Ck


Green, R. (2003), ‘Cost recovery and the efficient development of the grid’, inF. Lévêque (ed.), Transport Pricing of Electricity Networks, London: KluwerAcademic, pp. 137–53.

Hogan, W.W. (1992), ‘Contract networks for electric power transmission’, Journalof Regulatory Economics, 4, 211–42.

Hogan, W.W. (2002), ‘Financial transmission right incentives: applicationsbeyond hedging’, Presentation at Harvard Electricity Policy Group, May 31,www.whogan.com.

Hogan, W.W. (2003), ‘Transmission market design’, Conference ‘Electricity deregu-lation: where to go from there’, Paper presented at Texas A&M UniversityApril/4, www.whogan.com.

Hogan, W.W. and B.J. Ring (2003), ‘On minimum-uplift pricing for the electricitymarket’, Kennedy School of Government, Harvard University, www. whogan. com.

Joskow, P. and J. Tirole (2000), ‘Transmission rights and market power on electricpower networks’, Rand Journal of Economics, 31 (3), 450–87.

Joskow, P. and J. Tirole (2005), ‘Merchant transmission investment’, Journal ofIndustrial Economics, 53 (2), 233–64.

Latorre, G., R.D. Cruz, J.M. Areiza and A. Villegas (2003), ‘Classification of pub-lication and models on transmission expansion planning’, IEEE Transactions onPower Systems, 18 (2), 938–46.

Lévêque, F. (2003), ‘Legal constraints and economic principles’, in Lévêque (ed.),Transport Pricing of Electricity Networks, London: Kluwer Academics, pp. 3–33.

Mas-Colell, A., M.D. Whinston and J.R. Green (1995), Microeconomic Theory,Oxford: Oxford University Press.

O’Neill, R.P., P.M. Sotkiewicz, B.F. Hobbs, M.H. Rothkopf and W.R. Steward Jr(2004), ‘Efficient market-clearing prices in markets with non convexities’,European Journal of Operations Research, 164 (1), pp. 269–85.

Pérez-Arriaga, I., L. Olmos Camacho and F.J. Rubio Odériz (2002), ‘Report on costcomponents of cross border exchanges of electricity’, http://europa.eu.int/comm/energy/electricity/florence/index_en.htm.

Pérez-Arriaga, J.I., F.J. Rubio, J.F. Puerta Gutiérrez, J. Arceluz and J. Marin (1995),‘Marginal pricing of transmission services: an analysis of cost recovery’, IEEETransactions on Power Systems, 10 (1), 546–53.

Pérez-Arriaga, I. and Y. Smeers (2003), ‘Guidelines on tariff setting’, in F. Lévêque(ed.), Transport Pricing of Electricity Networks, London: Kluwer Academic,pp. 175–203.

Pope, S.L. and S.M. Harvey (2002), ‘TCC awards for transmission expansions’,www.pjm.com/services/trans/rtp-meetingnote3.html.

Rosellón, J. (2003), ‘Different approaches towards electricity transmission expan-sion’, Review of Network Economics, 2 (3), 238–69.

Scarf, H.E. (1994), ‘The allocation of resources in the presence of indivisibilities’,Journal of Economic Perspectives, 8 (4), 111–28.

Tirole, J. (1998), The Theory of Industrial Organization, Cambridge, MA: MIT Press.Vogelsang, I. (2001), ‘Price regulation for independent transmission companies’,

Journal of Regulatory Economics, 20, 141–65.Wolsey, L.A. (1998), Integer Programming, New York: John Wiley & Sons.Woolf, F. (2003), Global Transmission Expansion: Recipes for Success, Tulsa, OK:

Penn Well.


7. Compatibility of investment signalsin distribution, transmission andgenerationIgnacio Pérez-Arriaga and Luis Olmos

1. INTRODUCTION

Classical centralized optimization of expansion planning in electric powersystems was a conceptually easy task, regardless of the many practicaldifficulties in its actual implementation. All relevant decisions remained inthe hands of the vertically integrated utility (Pérez-Arriaga et al., 1987;Joskow and Tirole, 2002). This is true, in particular, for the expansion ofgeneration and the transmission grid, which are closely related. Generationexpansion planning took place first, typically assuming a perfect networkwithout losses or capacity limits. Then the utility planned the transmissiongrid, taking the location and characteristics of any new generation invest-ment as input data. In some rare cases the expansion of both generationand transmission was planned jointly, with the single objective of mini-mizing the cost of electricity supply. Distribution network planning tookplace separately, where the main input data were the individual or aggre-gated demands of consumers – both load profile and location – as well asthe substations connecting the high-voltage distribution grids to the trans-mission network where generation is feeding the electricity supply (Brown,2002).

In a liberalized regulatory environment a complete new paradigm for theexpansion of power systems becomes necessary. Now, the planning of gen-eration, transmission and distribution take place independently from oneanother. Generation expansion and operation is now left to the initiative ofprivate investors with the adoption of any regulatory instruments topromote adequacy of generation capacity and security of supply remain asopen issues (see Chapter 3 of this book). Distribution expansion hasbecome more sophisticated with new ideas, such as performance-basedremuneration and benchmarking or the stimulating perspective of wide-spread distributed generation, but the change is not necessarily associated

230

with the liberalization of the power sector. The accumulated experience inthe restructuring of the sector shows that transmission expansion criticallydepends on the regulatory paradigm that is adopted for the activity. Bytransmission planning regulation we mean the choice of the entity respon-sible for preparing and approving the plan, the decision on who will buildthe transmission facilities, the remuneration of the transmission servicesand the economic signals and incentives used to encourage any involvedentity to perform these tasks efficiently.

There are different approaches to make compatible, in an efficient way,the two worlds of competitive generation and monopolistic transmission inthe context of a liberalized regulatory framework. One possible option isleaving some specialized public-service type of entity to keep performingcentralized transmission expansion in the classical manner. Under thisscheme, there are several aspects that now become really important anddeserve to be discussed specifically. A clear and objective process isneeded – the so called ‘regulatory test’ – to determine which new invest-ments must be made. This process must avoid being discriminatory, that is,it should not benefit some agents at the expense of others. Also relevant ishow to allocate the cost of new lines among market agents. The advisabil-ity of establishing incentives for the system operator to plan the expansionof the grid in the most efficient way must also be investigated.

Another option is to let market agents participate actively in the decision-making process of transmission investment. There are several possibilities.One has to decide up to what extent to leave transmission investment inprivate hands and how to make private investment compatible with somelevel of centralized decision making or regulatory control. Now the role oftransmission rights in this process usually becomes very relevant (seeBushnell and Stoft, 1996; Joskow, 2005).

In order for the development of the system to be efficient it is necessarythat economic signals are sent to coordinate the investment decisions madeby the institution in charge of the network expansion – or the coalitions ofnetwork users, depending on the adopted approach – and generationinvestors. A major unsolved problem is the interaction between transmis-sion and generation expansion.

The links between transmission and distribution expansion also need tobe studied. First, the reasons for regulating distribution and transmissionindependently have to be examined and it must also be evaluated whetherthey are still justified in the new regulatory environment. One has to bringhere the experience that has been obtained from existing sound regulatoryschemes of transmission and distribution network planning (see Khatorand Leung, 1997 and Brown, 2002).

This chapter examines a number of issues that are frequently faced by

Compatibility of investment signals 231

transmission planners in the new regulatory environment and that still donot have a conclusive answer:

● Whether there should be an underlying coincidence between theoutcome of transmission investment under the traditional and themarket-based regulatory approaches.

● The role that transmission-related economic signals and incentives –when separately applied to the activities of generation, distributionand transmission – can fulfill in achieving joint efficiency in invest-ment and operation of the complete power system.

● Whether the set of transmission-related economic signals and incen-tives that are useful under a given regulatory paradigm can also makesense under a different one.

● The relative strength that transmission-related locational signals canhave when compared to other competing economic signals or to thedistortions that are introduced by lack of regulatory harmonization intariff setting. In particular it will be examined whether transmission-related locational signals – as they are currently used in some powersystems – could be enhanced in an efficient and cost-responsiveway so that they may result in more efficient location decisions by thenetwork users.

Following this introduction, Section 2 reviews and classifies the severalparadigms that have been proposed to regulate transmission investment.This section also introduces and discusses the regulatory test, a constructto provide conceptual support to any decision related to the regulatoryapproval of a new transmission investment. Section 3 examines which loca-tional signals make sense to encourage good practices in transmissionplanning under different regulatory paradigms. Section 4 analyses the com-patibility of economic signals for generation and transmission and howthese signals may complement one another. Section 5 examines this verysame issue but now concerning distribution and transmission. Section 6presents several case examples, based on actual power systems, whichhelp to illustrate the principles that have been presented in the precedingchapters. Section 7 concludes.

2. REGULATORY PARADIGMS IN TRANSMISSIONINVESTMENT

With the introduction of competition in generation and supply, the regula-tory environment in which the planning of transmission takes place has


changed radically. Even the organization model of the transmission activityis under scrutiny. Most systems maintain traditional planning approachesthat correspond to a regulated monopoly, although some incentive schemesmay be added. And there are those who favor the restructuring of this activ-ity by leaving some responsibility in the expansion of the grid to the privateinitiative. Conceptually this can be done either by allowing coalitions ofusers to promote or even decide the construction of some lines – subject tosome sort of regulatory approval and the application of the same generalrules of open access, remuneration and tariff setting as for any other lines –or by allowing private investors to build a line and sell its transmission ser-vices, by buying electricity at one end of the line and selling it at the other.

It is important to realize from the outset that the incentives to invest intransmission do not exist per se. They depend on the regulatory frameworkthat has previously been adopted for the transmission activity as a whole.Ideally, those agents that benefit from the expansion of the grid, that is, themarket agents, should be left to decide, or at least propose, the constructionof new lines. However, due to the fact that market benefits are frequentlyvery difficult to identify and allocate or are very widely distributed, such anapproach may turn out to be problematic. One cannot say much on thelocational signals that are derived from transmission until the several regu-latory paradigms for transmission investment have been carefully charac-terized. This is why this section summarizes the main regulatory schemesfor transmission investment previously presented in Chapters 2 and 5, in aformat that is suitable for the discussion on investment signals in transmis-sion, distribution and generation that will take place in the next sections.One can distinguish several cases.

First, the system operator is responsible for proposing a reinforcementplan, which has to be authorized by the regulator. Then, the constructionof this reinforcement may be assigned by competitive bidding or (more fre-quently) to the incumbent transmission company – which in most cases isalso the system operator and in European countries is thus named thetransmission system operator (TSO) under cost-of-service remuneration.

A more detailed description: The system operator must propose a plan forreinforcements of the transmission network, after taking into considera-tion (or rejecting, with due justification) any proposals made by networkusers. The regulatory authorities approve the plan (or not) and, if it isapproved, they authorize the construction of individual new facilities. Herethe system operator can be termed ‘passive’, because it proposes plans forthe expansion of the transmission network, but does not make the finaldecisions regarding approval of the plan and construction (for example,REE in Spain1). Both the system operator, when proposing the construc-tion of a line, and the regulator, when deciding upon the approval of the


project, need some kind of rule to apply in order to determine whether theinvestment is justified and what is the net benefit associated with it. Thisrule is called the ‘regulatory test’. See below for further details regardingthe concept of the regulatory test. Appendix 7A1 presents the mathemati-cal formulation of the problem of optimizing the expansion of the gridwithin the traditional framework, and how it is conceptually equivalent tothe regulatory test to be applied by the system operator and the regulatorin the new competitive environment. However, one cannot ignore that boththe regulator and the system operator have a natural inclination to acceptoverinvestment in grid facilities, since both can be held responsible for theoverall security of the power system in one way or another. Therefore themore prominent the role of these institutions in making the final decisionson transmission investments, the greater is the degree of overinvestment tobe expected with respect to what would be optimal with the assumed utilityfunction of consumers. It is debatable whether this is a positive feature ofthis method.

Construction, operation and maintenance of each facility are here theresponsibility of the incumbent transmission company (usually the TSO).An alternative is competitive bidding. If competitive bidding is adopted,the winner of the bidding process is paid as bid, and the bid becomes theregulated cost of the line or lines built. Since the contract will have a limitedduration (the economic life of the asset, typically), the operation and main-tenance of the facility may be auctioned again at the end of the period.

Availability targets may be set for each facility and penalties or creditscan be applied, depending on the actual performance.

Second, a for-profit (private, usually) company is awarded the transmis-sion license and it is regulated as a monopoly: investment and operation issubject to a prescribed grid code, while the remuneration is based on someprice-control scheme of the type RPI-X.2

A more detailed description: A private company is awarded the transmis-sion license and regulated as a monopoly. The transmission company mustfollow prescribed design requirements (a mandatory grid code) and it willbe subject to incentives to meet performance targets for the transmissionsystem. This is similar to the more customary regulation of distributionnetworks. This private company can be called an ‘active’ TSO because it isultimately responsible for the expansion of the network (for example,National Grid Company (NGC) in England and Wales).

In this approach the scheme that is typically adopted to determine theglobal remuneration of the complete transmission network is of the typeRPI-X where, given that transmission network investments can easily beexamined individually, the final remuneration should also take explicitly intoaccount the actual new investments, the economic lives and the depreciation


of existing investments, the financial situation of the transmission companyand the expected improvements in efficiency. The combination of a globalapplication of RPI-X to the entire network plus consideration of individualinvestments and economic incentives to promote better performance makestransmission remuneration more of an art than anything else.

One concern with this otherwise sound approach is that optimality ofinvestments will not be attained in general. It is not a trivial task to developremuneration mechanisms that encourage the system operator to pursuethose investments that are most beneficial for the system, while pursuing itsown benefit, since the incentives may have implications in different aspectsof the security and economics of the power system. For instance, economicincentives to reduce congestion costs might result in reduced securitymargins. Or economic incentives to reduce network losses might result inoperating measures that could be detrimental to the efficiency of the gen-eration dispatch. A carefully designed combination of moderate economicincentives and regulatory supervision might be successful in practice.Experiences in countries where this type of approach has been adoptedshow that certain well-justified reinforcements – interconnections in par-ticular – are not attractive to the planner because they would not earn anyincome for their construction (see CEER, 2003).

Third, some coalition of network users proposes a reinforcement, whichhas to be authorized by the regulator. In order for the regulator to autho-rize it, and according to what has been explained before, both the positiveand negative effects of the line on the outcome for all the market agentswould matter. The construction is assigned by competitive bidding, whichdetermines the total regulated cost to be recovered by network tariffs. As isthe case for alternative 1, where a passive system operator is responsible forproposing new investments that must be approved by the regulator, a regu-latory test is needed for the regulator to be able to decide whether a new lineis justified (see Littlechild and Sherk, 2004).

A more detailed description: Here the initiative to propose – and in somecases also to build – network reinforcements corresponds to coalitions ofnetwork users. Several options are possible and some have been tried inthe Argentinean system. In a first option the coalition builds and pays forthe reinforcement, which needs authorization from the regulator, has toprovide open access and receives the proceeds of general transmissioncharges resulting to the users of the facility. In a second option, coalitionsfor and against the building of the line have to go through a quasi-judicialprocess, and the regulator decides whether the reinforcement is justified.If justified, its construction is assigned by competitive bidding and paid-as-bid to the winner of the auction. Availability targets may be set andpenalties or credits will be applied, according to the actual performance.


The license will have limited duration and will be auctioned again for oper-ation and maintenance for the next period. The regulated cost of the rein-forcement will be charged to all network users by means of transmissiontariffs. The third option is known as ‘investments at risk’. As in the secondoption there is a quasi-judicial process. If the reinforcement is finallyfound to be justified and the proprietary coalition is properly defined, con-struction is assigned by competitive bidding. The regulated remunerationfor the line is determined by the auction. A fraction of it, which may berelated to the fraction of the line that is actually used, is charged to allnetwork users by means of regulated transmission tariffs. The remainingfraction is covered by the proprietary coalition, who also receives financialrights (firm transmission rights, FTRs) for part of the congestion rents ofthe line.

Fourth, merchant lines, whereby their unregulated remuneration is basedon the market value of their transmission services.

A more detailed description: The basic idea is to regulate the transmis-sion activity as any other competitive business. Therefore, the owner of theline could use it to buy energy at one end of the line (where the energy ischeaper) and sell it at the other (where it is more expensive) at market-determined prices. Network constraints – most frequently line conges-tions – are at the origin of the differences in energy prices between nodes.The remuneration of a merchant line comes from the difference in pricesbetween the end nodes of the line or from the revenues resulting from thesale of the corresponding transmission rights (in some DC lines it has evenbeen proposed that the network capacity be bid in a short-term market).Here FTRs may be seen not only as a risk-hedging mechanism, but also asa way to finance investments in the grid. Promoters considering investingin a new line or corridor will have more certainty of recovering the invest-ment if they are able to sell transmission rights for part of the capacity ofthe line in advance of its construction. Some of the difficulties in general-izing this approach to a significant number of lines are: (a) insufficiency ofmarket-driven revenues to pay for the total cost of a well-developednetwork,3 resulting in the fact that only those lines where high congestionrents are expected will be built under this approach; (b) high exposure torisk in revenues for the network investor, in particular if regulated lines canalso be built which reduce the amount of congestion on the line; (c) under-investment in a merchant line is a way of abusing a market power situation,given the highly discrete nature of transmission investment. Once a mer-chant line is built that is smaller than what is advisable, according to thegeneral interest, it may not be justified that the central planner, supposingthere is one, installs a new line, since the total investment in the corridormay be excessive.


Therefore, when merchant lines are allowed as a possibility for buildingreinforcements they should always be just an alternative to anotherapproach such as the first or second case above, since they are capable ofbuilding any line that is justified because of economic and security reasons.If some kind of centralized network planning and merchant investmentscoexist, rules for priority between both approaches have to be established.In some systems the installation of a merchant line must be subject to theregulatory authority giving its consent to it. The consent might be subjectto verification that the proposed merchant line is not detrimental to thesystem and, it also seems appropriate, that it is not included in the plan ofregulated investments that has been prepared by the responsible centralizedinstitution. Note that it should be more favorable for the network users topay the regulated cost of a line rather than its congestion rents, which sup-posedly are expected to be higher, since the merchant investment in the lineappears to be attractive. See CEER (2003) for details on this type of regu-lation, in the context of the process of formation of the internal electricitymarket of the European Union.

Broadly speaking, the drawbacks of most of these methods stem fromthe difficulty in aligning the interest of the grid promoter, whoever it maybe, with that of the system as a whole. The regulatory test, to be discussednext, makes it possible to verify whether any proposed transmission invest-ment increases the global welfare.

The Regulatory Test

The regulatory test is the rule that should be applied by the systemoperator – or the institution with the responsibility of proposing anetwork expansion plan – when preparing transmission expansion plansand by the national regulators (if this is the case, depending on thespecific domestic regulation) when deciding whether any proposednew investments should be authorized. Conceptually, the rule shoulddistinguish a justified investment from another one that is not justified.What is more, if the expansion of the grid is centrally planned, itshould help the regulator to identify the most efficient investment amonga set of possible ones. In order to do so the test should also evaluate theopportunity cost corresponding to the alternatives to the investmentunder consideration (see Australian Competition and Consumer Com-mission, 2003).

Under traditional regulation the rule is as follows: ‘One should invest intransmission network assets only while the additional network investmentcost is still smaller than the additional saving in system operation costs’.Obviously, as has been mentioned above and expressed in simplified terms,


when alternative investment possibilities exist, the network planner shouldchoose the combination of investments that maximizes the ratio operationsavings/additional network cost.

Appendix 7A1 proves that this definition is consistent with the one thatshould be applied in a context of competition: ‘Invest while the additionalnetwork investment cost is smaller than the net aggregated benefits (oncenetwork charges are included) of all network users (that is, all generatorsand all consumers)’. The net benefit of a generator, because of a new trans-mission investment, is the increment in its margin of market revenues overoperation costs, minus the network charges corresponding to the newinvestment. The net benefit of a consumer, because of a new transmissioninvestment, is the increment in the margin of its utility function over its costof purchasing electricity, minus the network charges corresponding to thenew investment.4

Note that, conceptually at least, the regulatory test should allow thejustification of so-called ‘reliability lines’, that is, lines whose justificationis mostly because of a general improvement of reliability conditions in thepower system. Reliability can be quantified in economic terms, even if thebeneficiaries whose reliability conditions are improved might be widely dis-persed. This may be done by making use of the utility function associatedwith the consumption of electricity. However, most frequently, transmis-sion planning is subject to hard reliability constraints in network designthat are imposed by some sort of national or international ‘grid code’,regardless of their economic justification.

Note, also, that reinforcements in the transmission network may bringsubstantial benefits to large aggregations of distributed generation that areor plan to be connected to distribution networks, since these reinforcementsmay facilitate – or even make possible – the secure injection into the lower-voltage grid of significant amounts of power that otherwise could not bereliably absorbed by the local demand.

Most required network investments are not large ones, such as long linesor complex substations, but minor reinforcements, replacements or modi-fications. The nature of actual transmission investments is explained inChapter 5 of this book, so it will not be repeated here.

Therefore, although conceptually sound, the regulatory test is verydifficult to apply in practice in strict terms. Different kinds of approxima-tions and simplifications are used in practice throughout the world byregulators.5 Note that, in order for the planner to strictly assess whether anew line is beneficial or not – or the best option to meet the new needs inthe system – one should try to compute, at least approximately, the trueeconomic benefits that are produced by the line.


3. ECONOMIC SIGNALS RELATED TOTRANSMISSION INVESTMENT

We are now interested in designing incentives that could promote anoptimal expansion of the grid. Quite logically, these incentives should varyaccording to which entity is chosen under the adopted regulatory schemeto be responsible for deciding on the lines to be built and where they shouldbe located. In this section, the network planners and potential investors foreach regulatory approach are identified, as well as the most appropriateeconomic incentives for each of them.

Potential Network Planners and Investors

There are three main categories of potential grid promoters:

● The system operators (either integrated or not with the transmissionfirm or firms); here we shall assume that this integration exists, sincethis is the common practice in Europe, where most system operatorsare also TSOs, subject to some degree of regulatory supervision.These TSOs may be ‘active’ or ‘passive’, as explained in the previoussection.

● Coalitions of network users, with their proposals subject to regula-tory approval and execution by the regulator, or just authorizationand execution by the coalition itself (investment at risk).

● Merchant investors, who have unregulated remuneration and may besubject to regulatory authorization. Merchant investors will collectjust the congestion rents of their lines, or their expected values, viacapacity contracts or firm transmission rents of some kind.

Incentives for Transmission Network Investment

For the three categories of grid promoters that have just been presented, weshall now examine which are the most suitable incentives to promote trans-mission investment.

First we shall consider the case where the TSO is in charge of making theplan to develop the grid. Two situations can be distinguished, dependingwhether the system operator is fully responsible for the development of thegrid or the final decisions on investment rest with the regulator:

● Active TSOs: The remuneration of the transmission activity forthese TSOs should refer to an efficient and well-adapted network –not the actual one. Economic incentives could also exist that are


related to the actual performance of the transmission system regard-ing quality of supply. Then these TSOs will invest in transmission atleast in order to comply with any mandatory grid codes and onlybeyond this if the remuneration scheme (typically a mix of RPI-X,cost-of-service and performance-based incentives) appears to makethe new investment profitable. Compliance with the minimum levelsof quality of supply can be guaranteed by the use of administrativepenalties. Assuming that the adopted remuneration scheme is suchthat the income of the TSO is not directly related to its actual invest-ment, the system operator will try to cut down its expenditure in newgrid equipment as much as possible. This avoids the risk of overin-vestment that is typical of more regulated approaches, but it isdifficult that an optimal level of investment may be achieved this way.

● Passive TSOs: TSOs propose plans for the expansion of the trans-mission network, but they do not make the final decisions regardingapproval of the plan and construction of the facilities. It is difficultto design correct incentives for a TSO that is not fully responsible forthe expansion of the grid that finally takes place. Here the finalresponsibility for the new network reinforcements actually corre-sponds to the regulatory authorities. The remuneration of the trans-mission activity for these TSOs should refer to the existing networkand, in principle, any incentives – either penalties or credits – shouldonly depend on the availability record of the network equipment.Typically, these TSOs will tend to propose plans with some excessinvestment, since this will reduce the potential problems of securityof supply that may be derived from lack of transmission capacityand maintaining a secure operation is the main responsibility of theTSO. Besides, if the TSO builds more lines, the size of the companywill grow with assets that have a guaranteed and attractive remu-neration (if the allowed rate of return is reasonable). Therefore,under this approach one should not expect that the TSO will proposethe most efficient grid that it is possible to build. But it must beremembered that here the ultimate responsibility for transmissionplanning corresponds to the regulator. However, it may also be pre-sumed that it will be difficult for the regulator to reject investmentsthat have been proposed by the TSO, therefore resulting in somelevel of overinvestment.

Coalitions of network users constitute the second category of potentialnetwork investors, with their proposals subject to regulatory approval andassignment of constructors, or just authorization (investment at risk).Nodal energy prices and transmission tariffs with locational signals, when


they exist, will provide incentives to the network users to organize coali-tions to promote new transmission investments that are beneficial for them,if the regulation allows this participation in the process. The success of thisscheme critically depends on the existence and correctness of the locationalsignals in nodal prices and/or transmission tariffs.

Lines that are economically justified but whose benefit is widely dis-persed (for example, lines whose main purpose is to improve the overall reli-ability of the power system) will never get a coalition of users that isinterested in promoting or building them. In a well-meshed network, as isgenerally the case in Europe, the economic benefits of new transmissionreinforcements will probably be quite dispersed. Therefore this approachhas to be supplemented by a centralized back-up scheme that makes surethat these lines are built.

It remains a controversial issue whether a line should be accepted whenit is proposed by a coalition whose benefits because of the line are higherthan the cost of the line, even if the construction of the line harms otherusers and the line is not beneficial for the system as a whole. If it is agreedthat only those lines that have a net benefit for the system should getapproval for construction, then the coalition should bear a fraction of thecost of the line, so that the net benefit of the line for all users is positive.Alternatively, transmission charges could be designed so that agentswould pay not only for the cost of the line they are benefiting from butalso for the harm the construction of the line is doing to other agents inthe system.

The third category of network investors comprises the merchantinvestors: their income will consist of the congestion rents of their lines, ortheir expected values, via application of nodal energy prices or the sale ofsome kind of firm rights of transmission capacity. Open network access bybuyers and sellers is obviously needed. The expectation of these congestionrents is the only driver behind network investments by merchant grid pro-moters. Merchant lines will only be built if there is an underlying structureof energy prices that reflect differences between locations because ofnetwork constraints. These differences in energy prices may take placewithin a single power system or in between power systems.

Some Initial Conclusions

We therefore conclude that regulated investment, which may adopt avariety of formats, must play a predominant role in the future developmentof almost every real-world transmission network. This is especially truewhen the system network is already well meshed, as it is mostly the case inEurope (at least within each country).


When TSOs are responsible for decision making in transmission expan-sion and construction of the new facilities, some performance-based incen-tives may be given to the TSOs so that network reinforcements go beyondthe minimum that is required by any mandatory grid code. When TSOs onlypropose the expansion plans and the regulatory authorities make the finaldecisions about the realization of all investments, the natural incentive ofthe TSO will be towards overinvestment, since this creates a comfortablesecurity margin for the operation of the power system, and also the oppor-tunity to grow in size. The regulator must be aware of this tendency,although the regulator itself is not free from the same impulse towards moresecurity.

If economic signals are provided by nodal prices and transmissiontariffs with locational content, coalitions of network users receive thecorrect economic incentives to promote the construction of new transmis-sion facilities. However, this sound mechanism can only be trusted forthose lines with clearly defined and relatively few economic beneficiaries.Also interesting is the idea of coalitions of TSOs who could propose thedevelopment of interconnections or other facilities of interest for cross-border trade. The facilities could be authorized by the corresponding regu-lators after using some kind of extended regulatory test. These regulatorsshould make sure that these cross-border investments are properly remu-nerated. This appears to be a pragmatic approach to getting some linesbuilt, when they are mainly justified in terms of their impact oncross-border trade. This is particularly true of those lines that are builtwithin a country that does not happen to be the main beneficiary of theirconstruction.

Nodal prices and/or congestion rents are natural incentives for merchantinvestors who want to appropriate these price differences. However, theinvestment that would maximize the profits of a merchant investor is typic-ally of a lower capacity than the investment that the regulator would havechosen. A larger investment would reduce too much the remaining conges-tion rents. Besides, if merchant and regulated network investments coexist,then the merchant investor always runs the risk of losing his congestionrents because of a new regulated investment. This is why merchant invest-ment can contribute to the development of a transmission network only ina few specific instances.

One major problem faced by the grid promoters in most liberalizedsystems is that the certainty about generation investment and the differencein time scales in construction time of transmission and generation duringthe ‘good old times’ of traditional central planning no longer exist. Thisadds a new level of complexity to transmission network expansion, whichwill be discussed next.


4. COMPATIBILITY OF INVESTMENT SIGNALS INTRANSMISSION AND GENERATION

The Difficulties

The objective now is to make compatible, in an efficient way, the two worldsof competitive generation and monopolistic transmission in the context ofa liberalized regulatory framework. Strong evidence supports the claimthat, despite being independently managed, transmission and generation(and to a lesser extent also demand) influence each other. As a consequenceof this, the optimal development of the system can only be achieved if coor-dinating signals are employed that indicate to network users what theimpact – in terms of network costs – of new generation or demand beinginstalled at each point of the system will be and, vice versa, how the agentswill be affected by the addition of a new line.

Contrary to what happened in the traditional regulated environment,new investment in generation does not result from a centralized process butfrom the non-coordinated decisions of a multiplicity of investors. There-fore, the central planner no longer knows with certainty what the develop-ment of generation in the system will be.

The construction of new lines may have a significant influence on thecompetitive position of some generators in the wholesale markets, there-fore guiding the installation of new generators in each area. For instance,some reinforcement may eliminate a network constraint that was forcingthe operation of an expensive plant that, otherwise, would remain idlemost of the time. Or, on the contrary, it may eliminate a bottleneck andallow a generator to increase its output and export its power to anotherregion. Thus, for example, the reinforcement of the corridors connectingthe Iberian system to the rest of Europe could allow less-expensive genera-tion in a very wide area to access the future Iberian market and theaverage energy price in this local region should presumably fall below itspresent level.

Locational transmission signals (a combination of losses, congestion andtransmission network tariffs) may be influential in the efficient election ofnetwork connection sites for new investments in generation and, perhaps,large consumers as well. In order for the agents to make efficient investmentoptions, they should face the real cost that the system will incur because ofthe installation of a new generation or consumption facility at a given pointin the network. This can be achieved by sending locational signals in theform of nodal energy prices on the one hand, and transmission tariffs withlonger-term locational content on the other. Although the existence oftransmission-related charges is just one among several considerations when


choosing a site for a new investment, in some cases it may have a decisiveinfluence. This is particularly true when there are actual choices to be made,such as building a plant near a convenient fuel supply (a gas field or a high-pressure pipeline) and shipping the power via the transmission network tothe load center, or building the plant near the load center and bringing thegas to the plant via a pipeline. It would be clearly inefficient to apply thesame transmission-related charges to two identical new power plants suchthat: (a) one would be located near the major load center in the system, sothat losses and congestions are small and no network reinforcements areneeded and (b) the other would be located far from the main load centers,so that losses will increase, congestion problems will appear and the networkwill need to be reinforced. A numerical case example will be presented inSection 6, showing some of the major considerations to be included whendetermining transmission-related locational signals.

Last but not least, and mainly due to growing environmental concernsraised by local and regional administrations and organizations in general,the process for obtaining the environmental permits and rights of wayneeded to begin the construction of a new line is significantly longer thanwhat was usual before. Consequently, construction time for the prevalentgeneration technologies in most countries (combined-cycle gas turbines,CCGTs, typically) is often shorter than the construction time of new trans-mission lines. As a consequence of all this, the transmission planner findsit increasingly difficult to predict where the new generation will be installedwith a sufficient time margin. For instance, according to the Spanish systemoperator, the average time to get a line built is about five years now, far morethan the one to one-and-a-half years strictly needed to build the line andalso more than the two or three years that the construction of a CCGTsplant ideally takes (in practice at least three and sometimes even up to fiveor six years). In some cases the construction of highly controversial lineshas been delayed by more than 20 years or even indefinitely.

Signals Derived from Transmission Pricing

Short- and long-term signalsTransmission pricing is the allocation of the regulated annual revenues ofthe transmission activity to the network users. The first attempt to designthese prices should be to resort to nodal energy prices, since nodal pricesare perfectly efficient short-term signals, that is, geographically differ-entiated short-term marginal costs of energy (see Schweppe et al., 1988).Nodal prices of energy implicitly include the effect on prices of losses andcongestion in the network. They send adequate signals for decisions con-cerning the economic operation of generators and loads.


Strict application of nodal prices to generators and loads results in a netamount of revenues, which should be applied to partly pay for the cost ofthe network. Under ideal circumstances, impossible to find in practice,these revenues would suffice to pay the total network costs fully. However,as indicated before, these revenues are usually very insufficient to cover thetotal network costs (cost recovery by nodal energy prices typically does notexceed 20 per cent of total transmission costs) (see Pérez-Arriaga andRubio, 1995).

Thus, additional signals are needed to recover the remaining transmis-sion network costs. These costs have to be assigned to the network users sothat distortion of economic efficiency is minimized. Therefore, in the firstplace, these signals must be long term, so that they will not interfere withthe nodal prices. This can be achieved by designing them as annual charges(although they may be distributed monthly, for instance). Ideally theselong-term signals should be consistent with the underlying cost function ofthe transmission activity, so one should design them according to the driverbehind transmission investment. This implies that cost causality, that is,cost responsibility, should be applied as much as possible, (see Chapter 6 ofthis book by Smeers, and Pérez-Arriaga and Smeers, 2003). In the new com-petitive regulatory framework, investment in a new transmission line isjustified whenever the present value of the aggregated benefits of all thenetwork users (generators and consumers) is larger than the present valueof the cost of the line. No existence of market power is assumed.

Transmission cost allocation criteriaThen, conceptually, the solution is to charge the network cost that is notrecovered through nodal prices in proportion with the benefits that thetransmission network (either globally or line by line) provides to each oneof its users.6 The resulting long-term economic signals have no purpose inthe operation (that is, short-term) timeframe; they are only meant toprovide locational signals to new generators and loads – or to those con-sidering retirement – that is, to inform them about the transmissionnetwork costs that are incurred because they locate or have located in onepart of the network instead of another one. In the long term the responseof the potential new network users to the transmission charges (that is,whether they will decide to install) will depend on their expected profits,after transmission charges are duly included.

Unfortunately, allocation of transmission costs to the economic bene-ficiaries is plagued with difficulties in practice. Most of the problems arisefrom the lack of adequate information about the generators in a competi-tive setting and the need to estimate the future behavior of the system, butalso because it is difficult to evaluate the economic impact on the market


agents of each individual line in a well-developed network with some levelof reliability-driven redundancy. This is why some measure of electrical usehas frequently been adopted as a pragmatic approximation to benefits (andit is also much easier to compute) (see Pérez-Arriaga, 2002; Pérez-Arriagaand Smeers, 2003). If we refer to the case of the internal electricity marketin Europe, this is the prevalent line of thought in the Florence Forum, aseries of meetings where representatives of the regulators and the maingroups of stakeholders in the European system gather to discuss and agreeon the regulation that will be applied at European level. Several algorithmshave been proposed to compute the electrical use that each agent makes ofthe system grid (see Rubio, 1999). At present, in most systems transmissionnetwork costs are socialized and charged uniformly to all network users, orto generators and consumers separately, according to some prescribedratios of the total network cost.

Non-transaction-based transmission chargesAn important practical conclusion that is derived from the criteria of alloc-ation of the long-term signals is that transmission tariffs should not betransaction based. Indeed, the adopted criterion of cost allocation hasnothing to do with the commercial transactions that the agents are engagedin at a given moment in time, under the assumption of a working marketthat is competitive and with perfect information. Transmission tariffs maydepend on the connection point to the network, on the nature of the agent –producer or consumer – on the amount of power injected into or retrievedfrom the network and on the time of injection or withdrawal, even on theeconomic benefits that ideally a market agent could obtain because of thedevelopment of the network, but not on whether the agent, in a particularmoment in time, is buying from or selling to a power exchange or via a bilat-eral contract, be it with a local or with a foreign agent.

Avoid tariff pancakingIn the context of a regional market it is very important to recognize thatwhat intuitively appears to be a fair transmission pricing rule may lead tocompletely wrong results. This is the case of the still prevalent rule world-wide of charging to an international power transaction that ‘crosses’ Ncountries the corresponding charge of each country ‘as if it were a nationaltransaction’, therefore leading to a piling up or ‘pancaking’ of transmissiontariffs. This seems a fair treatment from each individual country’s view-point, but it results in a tarification system that critically depends on theshape of political borders, rather than on the physical reality of networksand flows. This pricing rule has two major defects: (a) it is transactiondependent; and (b) the transmission tariff that is applied to a cross-border


transaction is the accumulation of the tariffs of all the countries that havebeen ‘crossed’ (the ‘pancaking’ effect referred to above), instead of somekind of average regional tariff which would have been applied in a trulyopen regional market without political borders. The correct approach to anefficient system of regional transmission pricing is ‘the single system para-digm’, that is, a pricing scheme that tries to get as close as is practicable tothe transmission tariffs that would be applied if the entire region were con-sidered as a single country.

Transmission rights and market powerThere is the general belief among some experts in transmission regulationthat transmission rights over transmission capacity should be allocated tothose who pay this transmission capacity, in proportion to their networkcharges. We are of the opinion that transmission rights should never beexplicitly allocated to those agents who contribute to the recovery of analready sunk cost. They should only initially be assigned to a private party,in proportion to its share in the payment of the cost of the line or lines,when this party is promoting the construction of the new line and as a wayof helping the agent to partly finance its investment.

Even when congestion rents – or the revenues resulting from the sale oftransmission rights – contribute to the reduction of the fraction of the sunkcost of a line to be recovered by other means, there is the possibility thatmarket agents enjoying market power get transmission rights that providethem with an incentive to exercise this power (see Gilbert et al., 2002). Ifthere are congestion rents associated with a line and these rents are used toreduce the transmission tariff of those agents that pay for the line, then thenetwork users are already receiving the same economic benefits as if theywere holding the same transmission rights. Therefore, it is not advisable touse the congestion rents from a line whose owner has the recovery of itsinvestment guaranteed, for any other purpose than to reduce the amountof network charges in the system as a whole. However, this topic is stilllargely unexplored.

Recommendations for Practical Implementation

Some recommendations can be derived from the basic principles that werepresented in the preceding subsection. If the transmission network is wellmeshed and there are no clear locational signals to be sent, because gener-ation and load are more or less evenly distributed and no systematic con-gestions are likely to occur, then the beneficiaries (or major users) of thenetwork cannot be clearly identified on the basis of their location.According to economic theory, in the absence of a clear indication from the


underlying transmission cost function, it makes sense to refer to the inverseprice elasticity rule (that is, the concept behind Ramsey pricing) in order tominimize the loss of efficiency. This rule must provide an indication of howto split the global charge between generators and consumers and then alsohow to charge individual consumers on the one hand and generators on theother. Regarding allocation to the individual consumers, the inverse elas-ticity rule would advocate charging more to the least elastic consumers.7

Note, however, that this may be considered to be an unacceptable discrim-ination (see Chapter 6 in this book, and Pérez-Arriaga and Smeers, 2003).

Assuming that there is strong competition on the generation side andthat generation is perfectly adapted to the system, due to totally free entryand exit by generators, so that whenever a generator is no longer profitableit will leave and accordingly generators will enter the market whenever thereare business opportunities to take advantage of, the rule advocates charg-ing transmission costs mostly to consumers, since generation will be veryelastic to prices and in the long run the large elasticity of generators willresult in a complete transfer of the charges to the consumers (energy priceswill rise or decrease as necessary in order for the generators to recover thetransmission charges they are paying from consumers so that they willmake no profits in real terms, once the average rate of return for an invest-ment in a sector with the same level of risk is deducted). Note that this isnot a trivial or universal rule, although it is a common misperception that:‘consumers always pay all network charges in the end’. A simple examplewill illustrate this fact: consider a project to build a new generator with non-contested access to an inexpensive energy source in a remote location.Assume that transmission pricing rules are such that the generator ischarged a large fraction of the transmission line that would connect it tothe major load centers without the project becoming unprofitable. In thiscase the network charge will not ultimately be transferred to consumers,since it will be fully absorbed as a cost of the new generation project.

If the transmission network is such that long-term locational signals areneeded and they can be more clearly identified – because of systematicstructural limitations of the network – then the allocation of transmissioncosts should pay attention to location. Note that these long-term signals areno longer useful for existing generators and loads (except for those consid-ering retirement because of economic reasons); they are meant to encour-age new facilities to select adequate sites and to fully recover network costs.8

However, for the sake of simplicity and to avoid any appearance of dis-crimination, most regulators choose that both the existing and the newnetwork users must be subject to the same charges. Following on from whatwas said in the previous point, if generation is perfectly adapted due to freeentry and exit, then it is not very important how much is recovered through


generators and how much through consumers, since consumers will end uppaying the entire bill. What would matter in this case would be thedifferences in charges among generators when they are placed in differentlocations, so that they have the right incentive to locate in the network and,similarly, the differences in charges among consumers. However, if the con-ditions presented before are not fulfilled, the absolute value of transmissioncharges will matter, since generators will not be able to completely transferthese charges to consumers and the absolute value of tariffs could alsoaffect the decision of a generator to install.

Both situations may take place in the individual system or at nationallevel. In those countries where it is deemed that there is little need forlong-term locational signals in transmission, transmission costs maybe allocated to generators and consumers without any geographicaldifferentiation. This seems to be the case in most European countries. Onthe other hand, in those countries where long-term location signals appearto be necessary (for example, England and Wales, Norway or Sweden inEurope; Chile, Argentina and Australia are also good examples), transmis-sion charges could have geographical differentiation.

These criteria are equally valid in a regional or multinational context. Ifgeographical differentiation of the long-term signals is not a majorconcern, then uniform regional transmission charges for generators andconsumers could be applied in strict application of the single-system para-digm. However, this would require a very high level of regulatory integra-tion and a pragmatic alternative could be to let each country charge itsnational tariffs to its network users, who in this way would automaticallygain access to the entire region.

However, the opposite situation may also be possible. At regional level,one may also want to send long-term signals in order to indicate the mostappropriate and inappropriate zones to locate new generation and load. Ifthe locational problem is a serious one – that is, the economic utilization ofgeneration resources at regional level to meet the regional load causes muchstress in the existing transmission network – then the long-term locationalsignals are needed. A rigorous approach would consist of assigning the costof each one of the lines in the region to those agents that use it (or benefitfrom it) while ignoring any political borders. However, this regionaltarification scheme may only be possible in markets with a very high levelof integration. Less radical alternatives are possible, such as replacing thenodal allocation of transmission costs at regional level by compensationmechanisms among countries, which would be based on how much eachcountry uses (or benefits from) the networks of other countries.

As indicated above, complete socialization of transmission tariffs (thatis, the postage stamp tariff) does not contain any locational signal. More


cost-reflective approaches, based on either the computation of the electri-cal use – such as the ‘areas of influence’ method in Chile or Argentina – orsome estimate of the responsibility in network investment – such as the‘investment cost-related pricing’ (ICRP) method in the UK or Colombia –allocate the annual transmission network charge among all network users,according to some procedure that typically involves a high level of costsocialization among all users. A stronger locational component could beobtained if the transmission tariff could reflect the positive or negativeincremental cost impact that the new network users impose on the trans-mission network. Thus, we may want to develop a more ‘aggressive’ set oftransmission network charges, that is, one with a stronger locational com-ponent. This topic is discussed in the next subsection.

A Proposal to Compute Incremental Transmission Charges for NewNetwork Users

The basic conceptThe emphasis here will be to identify the incremental grid cost that a gen-erator causes the system to incur when locating at a given node, whilekeeping in mind that the final driver behind the network charges should becost causality (see Pérez-Arriaga and Smeers, 2003).

One should be aware of the frequent discrepancy between networkcharges and the volume of network usage or benefits that are obtained fromthe network or individual lines by its users. On one hand, the total amountof income that has to be collected to remunerate the owner of any of thelines must equal the regulated annualized cost of the line, which is the samefor every year. However, due to the existence of economies of scale and thelumpiness of investments in the grid, the lines that are actually built maynot coincide with the volume of investment that is strictly needed at a giventime. In principle, one should avoid making new agents responsible for thetransmission capacity in excess of what is necessary just because of theirconnection, since this excess capacity is actually meant for future networkusers that obviously cannot be charged now for it.

For the purpose of this analysis grid reinforcements will be classified intothree types: (a) connection lines, that is, lines that are built for the exclusiveuse of a generator (or consumer) so that it can be connected to a suitablepoint of the common network; (b) lines that are built to meet the needs oftwo or more network users; and (c) investments that are needed to extendthe useful life of transmission facilities (since they are almost alwaysreplaced but not retired).

Only a reduced subset of all the lines in the system are connection lines,their length is typically small when compared to the remaining lines and


their use – and therefore their cost – can easily be allocated to specific gen-erators or demands. Network reinforcements of type (c) are meant toreplace existing facilities that arrive at the end of their physical life; newagents should not be assigned more responsibility in the investment forthese lines than already existing agents. The cost of these lines could beallocated in proportion to some measure of the average use of the line byeach agent.

It is more difficult to assign the responsibility for lines in the secondgroup (b). Here it is argued that the allocation mechanism should reflect thefact that new generators are more responsible than the older ones for theconstruction of those lines that have recently been built in response tothe need for new network expansion in the system. In the same way, a gen-erator or a load leaving the system (in generators this is typically becauseof retirement) may alleviate a network constraint or, on the contrary,increase its severity, and this should also be reflected in a transmissioncredit or charge.

The decision by a generator to install in a mainly importing area (or ademand to install in a mostly exporting one) may reduce the need for newinvestment in the grid. In this case, a sound transmission pricing mech-anism may result in a negative charge (that is, a credit) for that generator ordemand, thus making it attractive for new generators (loads) to install inimporting (exporting) areas, since this will promote a more efficient use ofthe existing grid.

The proposalThe method that is proposed here employs the installation time of lines andgenerators and some measure of how much each agent uses – or benefitsfrom – the network as the input to compute transmission charges. Themethod is applied only to lines of type (b) above. Any agent requesting con-nection to the grid should pay the cost of its direct connection line. Whenthe connection line has extra capacity so that future agents may connect tothe same node, the cost of this extra capacity can be socialized to demand.As for those network reinforcements of type (c) that are built to extend theuseful life of a given line, the same transmission pricing rules that wereapplicable to the original investment should be applied now.

Figure 7.1 graphically shows the process of computation of the loca-tional signals that result from the allocation of the regulated revenue of thelines of type (b) in the system to the network users. Several choices areoffered to the regulator regarding the share of the cost to be jointly borneby consumers and generators. General criteria to decide on how to makeuse of these choices in each particular case have been presented before,although in practice this has a certain degree of arbitrariness. A first choice


concerns the possibility of socializing to demand the cost of the fraction ofthe transmission capacity of each line that is not actually used,9 accordingto some agreed definition. In Figure 7.1, � is the used fraction of the lineand 1 – � the unused part.

Next, the cost of the used fraction of each line is divided into a fraction� to be paid by generators and the remaining fraction 1 – � to be paid bydemand. Again this is a choice to be made by the regulator using the generalcriteria that were discussed before. The cost of each one of these two partsis initially allocated to the corresponding market agents in proportion totheir share in the utilization of the line (CGi for generator i and CDj forload j). What is new in the method being proposed here, is that these par-ticipation factors are modified according to how long ago the line and theagents (generators and loads) were connected to the system. This is accom-plished through the use of coefficients representing how old the line andeach one of the agents are (kl for the line, kGx for generators, where x is thegenerator number, and kDy for loads, where y is the load number). Thus, theparticipation factor of an agent in the recovery of the corresponding part


Figure 7.1 Process of computation of locational signals

....

....

.... ....

Tot

al c

ost o

f lin

e l (

CT

)

Cos

t of

the

used

fra

ctio

n of

the

line

Paid

by

dem

and

Paid

by

gene

ratio

n

1–�

1–�

�

�

Cost of the unused fraction of the line is socialized(preferably to demand)

Allocation of the usemade of the line

Cgi: usage factor forgenerator i

Scale downparticipation

factors

Modifyparticipation

factors accordingto how old is the

line and eachgenerator

.......Same process

for loads

G2 G2 G2 CS

CS

CG2

Gng–1 Gng–1Gng–1CGng–1

1 + kl kG2

1 + kl kGng–1

Gng

Gng

Gng

D1

D2

Dnd–1

Dnd

G1G1 G1

of the cost of line l (where this corresponding part is �.(1 ��) CT1 for gen-erators and �.�.CT1 for loads) comprises three components. The first oneis simply the agent’s share in the use made of the line: CGi for generator i,for instance. The second component is a function of the time for which theline, and the agent itself, have been in service (kl and kGi, respectively, forline l and generator i, where these are input parameters whose value has tobe defined a priori).10 Finally, these participation factors are scaled downso that the sum of the participation factors of all the generators is 1 andthe same for the loads. Therefore, the expression for the total usage charge,corresponding to line l, that generator i must pay is:

(7.1)

where CTl is the total cost of line l, � is the used fraction of the line, � isthe fraction of the cost of the used part of line l to be recovered from con-sumers (and therefore 1 �� is the fraction to be recovered from generators),CGi is the usage factor corresponding to generator i, kl is a factor repre-senting how old line l is, kGi is a factor representing how old generator i isand CS is the scale factor used to modify participations so that all of themadd up to 1. The expression for the scale factor is:

(7.2)

We are proposing here that cost causality should be interpreted in an‘incremental’ rather than a plain ‘average’ way, so that new agents areassigned more responsibility for the most recent network reinforcements.This results in the use of coefficients kl and kGi so that new generators andloads pay a larger part of the cost of new lines than those market partici-pants who have already been operating for a number of years. Thus, thefactor used to compute generator i’s contribution to the recovery of the costof a line l results from increasing generator i’s usage factor (CGi) by the pro-portion represented by the product of coefficients kl and kGi.

DiscussionThe major difference between ‘marginal’ methods of transmission pricingand ‘average’ ones is that the former result in both positive charges (whenthe installation of the agent creates the need for new grid investments) andnegative ones (when the agent reduces the amount of network investmentthat is needed). ‘Marginal’ methods send stronger locational signals typic-ally, although they require a higher level of arbitrary assumptions than

CS � 1

�ng

i�1

(1 kl kGi)CGi

.

PGi � CT1· � · (1 � �) · CGi · (1 k1kGi) · CS,


‘average’ methods (see for details: Pérez-Arriaga, 2002; Pérez-Arriaga,et al., 2002). The approach that is proposed here can be used with any typeof method for network cost allocation, either marginal or average, but itintroduces the notion of temporality in the allocation, so that the link ofcausality between recently installed network facilities and new agents isemphasized. The method is supposed to produce stronger locationalsignals, where ‘strong’ does not mean that the numerical value is larger, butthat the relationship of causality has been reinforced. The method caneasily be adjusted to emphasize more or less this temporal factor.

The method can equally be applied to compute transmission charges fornew generators and consumers. However, it is not expected that consumerswill be very responsive to these economic locational signals. Besides, manycountries have decided to apply uniform electricity tariffs for every class ofend consumer. If this is the case, flat charges may be applied to consumersand the proposed method could be applied only to generators.

The concept of introducing the temporal dimension in the causality rela-tionship when computing transmission charges can be found in the regula-tion of PJM in the United States (see Joskow, 2005). Transmission chargesfor generators corresponding to new grid investments may differ widely inPJM from one area to another, thus sending important locational signalsto new generators. However, unlike the scheme of signals proposed here,grid charges to generators in PJM are computed once (at the moment thegenerator is installed). Furthermore, charges paid by a new generator coveronly the cost of facilities built because of the construction of the generatoras well as those lines built shortly before the installation of the generatorfrom which it is benefiting. In other words, new generators are not heldresponsible for the cost of lines built after the installation of these genera-tors unless these lines were already projected when the generators decidedto install. Responsibilities in network cost are evaluated by computing theincremental impact of the installation of each new generator on the flowover congested lines or corridors that must be reinforced.

Temporal differentiation plays an important role as well in the proposalfor the computation of grid charges for the Peruvian power system by agroup of international consultants that included the first author of thischapter (see Mercados Energéticos, 2005). An interesting feature of theapproach that has been proposed for Peru is that it employs two differentnetwork cost-allocation schemes, one for the existing lines and the exist-ing network users (method A) and a second method for the new lines andnetwork users (method B). The point here is there is little use in sendingeconomic signals to consumers and generators that are already connectedto the grid, so there is not much interest in perfecting any current trans-mission pricing method A that is already in use for existing lines and


network users. However, one could aim even for the most adequate methodB – allocation to the economic beneficiaries – when it concerns only the newlines and the new network users. The major difficulty with the method ofallocation to beneficiaries – going deep into the past and the future guess-ing what the economic benefits were or will be – disappears if there is noneed to go into the past and if the exploration into the future is limited intime. Limiting the exploration into the future can be achieved by returningto method A once some years have passed and what were new lines andnetwork users can be considered to be existing ones. Another approach –the one that was adopted in the Peruvian study – consists of computing thefuture transmission charges of the new lines for the new users once and forall, not subject to future revision. This also has the positive effect of elimin-ating any uncertainty of the new users with respect to any possible changesin transmission charges in the future.

Locational signals for retiring generatorsIn order to encourage market agents to make the right decisions in the longterm we must send them economic signals so that market agents realize thecost that the system incurs as a consequence of either their installation ortheir retirement. The previous subsection discussed a system of economicsignals aimed at encouraging market agents to install in the best locationfrom the system point of view. This subsection presents a scheme of signalsthat is meant to induce those agents considering the possibility of retiringa power plant to take into account the grid costs that may result from thisdecision.

Computing the effect that the decision of retiring a plant has on thenumber and nature of future grid investments is very difficult. Besides, theapplication of economic signals that are based on costs not yet incurred bythe system is very problematic from a regulatory point of view. Therefore,it seems more appropriate to derive signals to market agents from theimpact that their decision to leave the system would have on their currentelectrical usage of the existing grid.

It is proposed here to follow for retiring plants exactly the same basicapproach that has been proposed for new entrants. There are two majorpossibilities. The first one is to hold the agent responsible only for thechange in the use of any congested lines, which will probably have to bereinforced in the near future if the plant is retired. The second one is to con-sider that the agent is responsible for the change in the use of any line in thesystem, whether congested or not. It seems better to compute the gridcharges taking into account only the congested lines, since there is no evi-dence that other lines will have to be reinforced because of the decision ofthe agent to leave.


In summary, the same method presented above to send ‘strong’ loca-tional signals to new generators is also proposed to compute the gridcharges associated with the retirement of an existing plant. The retiringgenerator would be regarded as new as well as the congested lines and thelines that have actually been installed recently. The signal should be imple-mented as a single payment to be made when the agent retires the plant.

Other Locational Signals for Generators and Network Planners

The previous subsections have discussed the design of a system of trans-mission charges that could send meaningful economic signals to existingand future network users, while recovering the total network costs. How-ever, there are other locational signals of a different nature that generatorsand also system planners could send and receive. These signals are mostlyrelated to information that helps in removing the uncertainty that accom-panies investment decisions in both network and generation.

As we have seen, the development of the grid must be planned in con-junction with that of the generation and the demand in the system. Thelocation of new power plants and loads influences the decision by grid plan-ners on which are the most interesting reinforcements. Likewise, marketagents should take into account the limits that are imposed by the existinggrid and the investments in new lines or other transmission facilities thatare planned for the future. Both the network planners and the potentialinvestors in new generation need to know what the other party will do withas much certainty as possible.

Therefore, it is important on one side for the network planner to have areliable indication of the level of commitment of new agents that ask forconnection to the grid, since network expansion should be speciallyaddressed to meet the needs that result from new network users. Consideras an example the case of the Spanish power system, with a peak load of43,000 MW in 2005 and where the system operator has standing requestsfor new connection of more than 50,000 MW of wind generators and morethan 50,000 MW of CCGT power plants. Most of these requests will notmaterialize, since their purpose is just to make sure that a particular site hasmet all the administrative requirements to install new capacity – a proce-dure that usually takes several years. A regulatory measure that can helpthe system operator to sort out the more serious requests from the merelyspeculative ones is to require from the potential investor a financial guar-antee, in order to start the administrative process of authorization of theconnection to the grid plus any environmental or other kind of permits thatmay be necessary. The guarantee will be returned only if the power plant isactually built within a prescribed time limit.


On the other hand, the system operator should provide the market agentsthat are considering the installation of a new plant, with as much informa-tion as possible regarding the future development of the grid including theprojected importing and exporting capacity with neighboring systems, theexpected value of transmission charges at each node as well as the lossfactors and the nature, the location and severity of the restrictions that thegrid is anticipated to impose on the operation of the system. An adequatetime horizon for these estimations should be about five or ten years. Thisinformation should be made available in some kind of ‘N-year-ahead trans-mission report’ to be issued annually.

It is also very important that the regulator and the system operator makeclear the network access rules to all potential generation investors. Accessrules encompass the management of network constraints as well as the cri-teria to grant connection to the transmission network. Ill-conceived accessrules may deter new generation investment by creating too much uncer-tainty or, at the other extreme, they may provide excessive advantages to thefirst new entrants with respect to future ones. For instance, a reasonableaccess rule for generators may specify that connecting first to a given nodeof the transmission network does not provide any advantage regarding pri-ority of access to the network, in case the network cannot accommodate allthe power that the generators that are connected to that node are willingto inject. An investor that considers building a new power plant and toconnect it to this particular node should expect competition with the gen-erators already connected just for the use of the limited capacity of injec-tion at the node. FTRs may be used by the existing and the future powerplants to manage the economic risk of being displaced from producing ata node by more competitive generators, therefore creating more favorableconditions for investment. Note that, in general, if there is a bottleneck, thenetwork planner will decide to reinforce the grid – if it is economically andtechnically justified – so that the congestion is conveniently reduced or eveneliminated. Therefore one should not expect this type of situation to be fre-quent, unless some of the existing generators at the node are very inefficientand network reinforcement in this case is not justified. Each power systemmay choose different access rules from the ones that have been presentedhere as an example. The message is that access rules are an essential ingre-dient of the information that potential generation investors need to makesensible decisions – whether to invest in a given power system and whetherto connect at a given node of the transmission network.

Access rules and transmission charges can be used to guide the processof connection to the grid of new loads in a meaningful way. The applicablelegislation in many countries allows agents to apply for connection at anyvoltage they choose and the network planner is then forced to comply.


Some agents may take advantage of this situation to get connected tovoltage levels where they know they will pay fewer network charges or haveother types of advantages, but which may not be convenient for the overalldesign of the distribution or transmission network. One possibility to guidethe location of new agents is to charge them according to the voltage levelwhere they should be connected – depending on their peak demand andtotal consumption, for instance – instead of the voltage level where they areactually located.

5. COMPATIBILITY OF INVESTMENT SIGNALS INTRANSMISSION AND DISTRIBUTION

The separation between transmission and distribution networks is not aclean one. Most European countries have adopted the simple and, up to apoint, arbitrary decision to define transmission as encompassing allnetwork facilities operating at 220 kV and higher (400 kV, typically) and,also frequently, all interconnection lines with other countries. Distributionwould correspond to all other networks at lower voltages. Obviously, thereare some interactions at the interface between both networks. For instance,depending on the physical configuration of each particular network, somefurther development of the 220 kV grid may help the underlying high-voltage distribution network to avoid some investments. Similarly, a stronghigh-voltage 132 kV subtransmission grid may replace some investments at220 kV level.

Traditionally, the regulatory approaches to the transmission and distrib-ution activities have been very different. There are some good reasons forthem to remain so.

First, transmission and distribution grids perform different functions.Transmission makes sure that major generators are well connected to majorload centers. Flows over the lines in the transmission grid change directionfrequently. Transmission has a great impact on the implementation of thewholesale market. On the contrary, if we leave aside the existence of dis-tributed generation – which in some areas may outweigh local demand sothat the corresponding distribution network is a net exporter – power overthe distribution grid always flows in the same direction: downstream fromtransmission to the end consumers. The operation and planning of distrib-ution networks as a whole are mostly influenced by the location and loadprofiles of final consumers. On the other hand, the internal configurationand operation of distribution networks has a negligible impact on thewholesale market and the configuration of the transmission network.However, at higher-voltage levels distribution grids tend to be meshed and


their configuration can be adapted to the particular operating conditionsby opening and closing the corresponding devices. In some cases the powerflows in the transmission network cannot be fully understood without con-sidering the impact of the subtransmission network. Besides, more andmore generation is connecting to the distribution grid at all voltage levels.Therefore, one can easily conclude that, in many cases, those parts of thetraditionally considered distribution grids that are immediately belowtransmission substations are really performing transmission grid functions.The difference between transmission and distribution grids, especially atthe border between them, is becoming increasingly blurred.

Second, the number of lines and other network facilities in distributiongrids is much higher than that of transmission grids. In transmission eachindividual investment can be considered separately for authorization andremuneration whereas in distribution, particularly at lower voltages, thiswould be an extremely tedious and meaningless task. Performance-basedregulation is a preferable choice for distribution networks.

Third, transmission has an important role in maintaining the overallsecurity of the power system. Failures at transmission level are very infre-quent. However, when they occur, there is usually a large disruption inpower supply. Failures at distribution level are much more frequent. Thequality of supply of most consumers mostly depends on the distributionnetwork. Therefore, the regulation of distribution should heavily relate toquality of supply considerations.

The above-mentioned broad differences that exist between transmissionand distribution in many aspects should result in a specific treatment ofremuneration, planning of the expansion, operation and, quite import-antly, the short- and long-term signals to be sent in each one of the twocases:

● Regarding the remuneration of the activity, there are implicationsboth for the computation of the regulated revenues of the owner ofthe grid and for the split of these revenues into the part to be collectedfrom consumers and the one to be collected from generators. In theabsence of distributed generation, consumers should pay 100 percent of the distribution network costs. According to cost causality,generators connected to transmission should not pay distributionnetwork costs, since the design of distribution networks is notinfluenced at all by the nature or location of the generation powerplants. On the other hand, the costs of the transmission networkshould be shared by generators and consumers (here we mean bothlarge consumers that are directly connected to the transmissionnetwork and smaller consumers that obtain their power from the


distribution networks). When distributed generation (that is, genera-tion that feeds into the distribution network) cannot be absorbedlocally and feeds into the transmission network, it should also paytransmission network charges.

● Given the ambiguity that exists when defining both the borderbetween transmission and distribution and the responsibilities inexpanding the network, the approaches used to compute the remu-neration of the transmission and distribution companies at leastshould be consistent with each other.

As said before, performance-based regulation should be employed todetermine the revenues of the owner and developer of the distribution grid.In the case of transmission, currently most of the countries compute theincome of the transmission company or companies through some sort ofcost of service regulation. This lack of coherence may result in a subopti-mal development of both grids. We have already mentioned the naturalinclination of system operators and regulators to overbuild the transmis-sion network. Moreover, when faced with a remuneration scheme that isbased on the cost of individual transmission facilities or on the winningbids in a public auction, transmission companies or TSOs will be happythat as many lines as possible are built. On the contrary, provided that theremuneration of the high-voltage distribution company is some kind ofRPI-X, they will prefer that the needs of reinforcement of the subtrans-mission network could be met – whenever possible – by reinforcements attransmission level, in order to reduce costs without hurting the overallnetwork service. As a consequence of this, too much transmission grid maybe built that will perform a function that would have been better accom-plished by the distribution grid. Regulators should be aware of these dis-tortions in order to minimize them by adjusting the correspondingremuneration mechanisms. However, a complete elimination of these dis-tortions does not seem to be possible because of the discontinuity of theregulatory treatment at the border between the two types of network.

Ideally, planning transmission and distribution grids should be coordi-nated. For instance, the decision on the location and number of the trans-mission/distribution substations at the border between both grids is criticalin this regard. An iterative process, similar to the method that is used insome cases to determine the location of substations and transformersbetween different distribution grid levels, could provide the location andnumber of the substations between transmission and distribution as well(see Peco, 2001).

In the absence of any joint planning in practice, coordinating economicsignals might be useful. These locational signals could be applied to the end


consumers and also to the distribution utilities. As shown before, in generalnetwork charges for consumers in the distribution grid should include bothtransmission and distribution costs, whereas those aimed at consumers andgenerators at transmission level could ignore the distribution grid. This istrue as long as distribution grids do not contribute to transporting the poweroutput from distributed generation to load centers located in other distrib-ution networks. Locational signals in relation to transmission investmenthave already been discussed in Section 3. The next paragraphs explain hownodal prices could be employed to send incentives for investment both to endconsumers and to the distribution companies. Since the distribution andretailing activities should be unbundled in a sound regulatory framework,the distribution utility does not purchase electricity. However, the point hereis that regulatory incentives for the distribution utility should be designed sothat the price of acquiring at each moment in time from the transmissionnetwork the power that is needed to physically supply the end consumers ofthe distribution company is as low as possible. A regulatory mechanism toaccomplish this is to make the distribution company responsible for thedifference between the actual cost of purchasing electricity at hourly nodalprices from the corresponding transmission nodes and the amount of elec-tricity consumed by the end consumers of this distribution company multi-plied by some average loss factor and some average hourly energy systemprice. This is a generalization of the regulatory mechanism that is frequentlyused to provide incentives to distribution utilities to reduce network losses.

The first level of implementation of this scheme of locational signals,with effects over the development (and also operation, see below) of the dis-tribution grid, would consist in letting the distribution company assume theeconomic difference between some reference purchasing hourly price forthe energy demanded by the consumers of the distribution company(affected by some correction loss factor) and the pretended purchase (sincethe power is actually bought by the retailing company) of the energy fromthe transmission system at the corresponding nodal price, which would bedependent on the point of interface with the transmission grid. Under thisscheme, the distribution company would react by planning the develop-ment of the grid in order to minimize not only the cost of losses over thegrid, which has traditionally been its major objective, but also the cost ofpurchasing wholesale power.

Nodal prices can also encourage efficient operation of the distributionnetwork. Similarly to what happens with the planning of the grid, pretend-ing that the distribution company buys the energy for its customers at thecorresponding transmission hourly nodal prices would lead the distributionutility to operate the grid in order to minimize the cost of purchasing thewholesale power at any given time.


By operation of the grid we mean here the possibility of modifying theconfiguration of the subtransmission grid. Normally, this grid is wellmeshed because of security, quality of supply and stability reasons.However, at least part of this grid is operated as a radial network by suit-able switching. Changing the network configuration during the operationallows the distribution company to modify the incurred losses and, if theybuy power from the transmission grid at nodal prices, to modify also theamount paid for this power as well. Some indication of the quantitativerelevance of these locational signals can be found in Carillo-Caicedo andPérez-Arriaga (1995).

A second level of implementation of a scheme of nodal prices wouldbe to pass through the wholesale power prices to the end consumers.Computation of nodal prices at the end consumer nodes of the distributiongrid has to be discarded because of the high volatility these prices mayexperience, which would make them unacceptable (see Outhred and Kaye,1996). It would be possible, however, to charge each end consumer a zonalprice based on the nodal prices at the node or nodes of the transmissiongrid where the distribution company buys its power. These zonal pricescould be obtained by averaging, over the period of time for which the pricesare computed, the nodal prices at the different nodes of the transmissiongrid weighted by the amount of energy that the considered distributionzone is obtaining from each transmission node.

As mentioned before, the massive introduction of distributed generationwill likely blur the differences between transmission and distribution regu-lation, especially if this generation ends up being partially fed into thetransmission network. Then, the development of transmission and distrib-ution grids will become much more dependent on one another. The plan-ning of transmission and high-voltage distribution grids would probablyhave to be carried out jointly or, at least, would involve a much higher levelof coordination. Transmission and distribution grids would perform muchmore similar functions and, therefore, the regulatory treatment of trans-mission and distribution would have to evolve accordingly.

6. CASE EXAMPLES

This section presents numerical examples of the most relevant concepts thathave already been presented. The first case shows how nodal transmissiontariffs can be computed, both at national and multinational levels, and alsohow the same technique can be employed to determine economic compen-sations between countries or power systems because of the use that agentsin one country make of the networks of other countries. The potential of


these locational signals to guide transmission and generation investment isexamined. The second case example shows how nodal transmission tariffsare just part of an ensemble of other locational signals and what theexpected weight of each one of them is.

Nodal Transmission Tariffs in the Internal Electricity Market of theEuropean Union

This subsection presents numerical results that help to illustrate some ofthe ideas that have been discussed in the chapter within the context of theinternal electricity market (IEM) of the European Union. As mentionedbefore, nodal (that is, with locational content) transmission tariffs, togetherwith nodal energy prices, may influence the decisions by market agents onwhether to install new generation and consumption capacity (or retireexisting capacity) and where to do it. In addition, nodal transmission tariffsand nodal energy prices can also encourage coalitions of network users(generators and loads) to promote the construction of new network facili-ties. This may happen in different ways, depending on who is ultimatelyresponsible for the development of the grid and the role that the prevailingnetwork regulation assigns to the several market agents in this process.

In Europe, the adoption of a common system of transmission tariffsappears to be too ambitious for the time being. To start with, the regulatorof each country has approved regulated transmission costs using widelydifferent methods, and also has adopted widely different procedures tocompute the transmission tariffs for the network users. The end result is atotal lack of harmonization in the computation of transmission charges(see Pérez-Arriaga et al., 2002). However, a common procedure of inter-TSO payments has been established to compensate each country for theexternal use of its network by agents in other countries. The net value ofthe compensations must then be used to modify the local tariffs calculatedby the regulator of each country (ibid.).

A method that is able to compute nodal transmission tariffs at regionallevel must also be used to determine the net volume of inter-TSO paymentsthat each country has to receive. In order to do so, one only has to aggre-gate the resulting nodal transmission tariffs at country level, thus directlyobtaining how much the agents within each country use their own grid andthe grid of others.

After exploring different methods of allocation of grid costs to individ-ual network users (that is, of computation of transmission tariffs), all ofthem based on some measure of electrical usage, we have reached the con-clusion that the method of average participations (AP) is the best oneamong those currently available for this specific application (see ibid. for a


description of the AP method and a comparison with other approaches).Appendix 7A2 provides a succinct description of the AP method, since thisis the one that has been used in the numerical examples that will be shownbelow. However, one must bear in mind that this method is one amongseveral electric usage-based methods that are currently available tocompute transmission tariffs or inter-TSO payments and that other novelinteresting methods may be proposed at any moment (see, for instance,Florence School of Regulation, 2005). We have used the AP method toobtain nodal transmission charges for each of the 3,965 nodes of a modelcomprising 13 countries that belong to the IEM. Twenty-four different sce-narios, corresponding to the real operation of these 13 countries through-out the year 2002, have been used to compute the annual tariffs. The datahave been provided by the association of European Transmission SystemOperators (ETSO).

Some other data have been used to obtain the results here presented. Wehave had to estimate the annual cost per km of a 400-kV line in order toexpress both compensations among countries and nodal transmission tariffsin monetary terms. First, we computed the number of kilometers of equiva-lent 400 kV lines within the horizontal network of 11 European countries:Austria (A), Belgium (B), France (F), Germany (D), Italy (I), Portugal (P),Spain (E), Switzerland (CH), the Czech Republic (Z), the Netherlands (NL)and Slovenia (SLO). The figures obtained were 3,896 km for Austria, 3,930for Belgium, 4,477 for Switzerland, 5,116 for the Czech Republic, 34,811 forGermany, 21,511 for Spain, 27,916 for France, 13,019 for Italy, 3,737 for theNetherlands, 3,627 for Portugal and 696 for Slovenia. Second, we dividedthe regulated cost of the horizontal network of each country (values werethose used in the provisional method for the year 2002) by the number ofkilometers of equivalent 400-kV lines within the network of the country.Thus, a regulated cost per kilometer of equivalent 400-kV line was obtainedfor each country. Finally, the standard unit cost of a 400-kV line used toexpress all the results in monetary terms resulted from computing theaverage of the regulated unit costs obtained for the different countries. Anannual value of €0.0352 million per km of 400-kV line was obtained.

From these nodal tariffs and for each agent and scenario an averageannual tariff has been obtained for every agent. In order to calculate theaverage tariff for an agent, its nodal tariff in each scenario has been weighedby the amount of power either produced or consumed by the agent. Bydefault the AP method equally allocates the total flow over each line to allthe generators and to all the loads in the system. Therefore 50 per cent ofthe use of the line (and consequently of its cost) is allocated to generatorsand 50 per cent to loads. However, the method easily allows for any othersplit of the cost of the grid between generation and load.


Figure 7.2 shows the average nodal L and G tariffs for the 13 countriesthat have been considered in the study. For each country, the L tariff isrepresented by a light column and the G tariff by a dark one. Figures 7.3and 7.4 show the nodal L and G transmission tariffs, respectively, for all thenodes in the IEM-13 network. The average values for G and L for eachcountry have been represented by a dotted line, whereas the average inter-nal tariffs for each country (resulting from dividing half of the cost of thenational grid by the total amount of load or generation in the country; thisis the internal average transmission tariff for each country before the appli-cation of inter-TSO payments) are represented by a dashed line. Instead ofapplying a system of pan-European tariffs like this, the main stakeholdersin the IEM have agreed to establish a system of compensations amongcountries (inter-TSO compensations) for the use each country makes of thegrids of others. The difference between both horizontal lines in eachcountry indicates the significance of the impact of the inter-TSO paymentmechanism on the transmission tariffs of each country. Numbers areexpressed in €/MWh.

We can appreciate from the figures that the distribution of the averagetransmission tariffs per country is not too dispersed. The average values forG range between €1.03/MWh and €2.89/MWh, with an average value forthe IEM of €2.02/MWh. In the same way, the average values for L range


Note: A�Austria; B�Belgium; CH�Switzerland; CZ�Czech Republic; D�Germany;E�Spain; F�France; H�Hungary; I�Italy; NL�Netherlands; P�Portugal;SLO�Slovenia; SK�Slovakia.

Figure 7.2 Average L and G tariffs in Europe

0

0.5

1

1.5

2

2.5

3

3.5

A B CH CZ D E F H I NL P SLO SK

nodal_L_tariff nodal_G_tariff

between €1.35/MWh and €2.89/MWh, with an average value for the IEMof €2.33/MWh. The significant differences among the average transmissioncharges in different countries are not only due to the different pattern offlows existing in each part of the grid but also to the fact that the amount


Note: For country abbreviations, see Figure 7.2.

Figure 7.4 G nodal tariffs in Europe

0

2

4

6

8

10

A CHB CZ D E F H I NL P

Nodal tariff Average nodal tariff Average internal tariff

Note: For country abbreviations, see Figure 7.2.

Figure 7.3 L nodal tariffs in Europe

0

2

4

6

8

10

B CZCH D E F H I

NL P

Nodal tariff Average nodal tariff Average internal tariff

A

of transmission assets per MWh varies widely from one country to another.Inter-TSO payments among countries account for the difference betweenthe average transmission charge that would exist in each country if theimpact of cross-border exchanges of electricity were ignored (dashed linein every figure) and the average transmission charges per country when theAP method is employed (dotted line). For instance, in this example Austriamust receive a positive inter-TSO payment of €26 million. This is the netamount that other countries have to pay Austria, because of the differencebetween the cost of the Austrian grid (represented by the average internaltariffs on G and L) and the total cost of the whole grid used by the Austriangenerators and consumers (represented by the actual average G and Ltariffs when cross-border flows are accounted for).

The distribution of transmission charges within every country shows alarge dispersion with respect to the average value. This is related to thedifferent ways in which the same pattern of flows can affect the transmis-sion charge computed with the AP method for two agents located not veryfar from each other. Due to the non-linear nature of the method, theunitary participation in the use made of the grid by an agent may criticallydepend on the size of G and L at the nodes. The AP method provides veryreasonable results at the macro level (inter-TSO payments among coun-tries) but at the micro level (nodal charges) the values are somewhat volatileand require some interpretation and further treatment.

Now we shall focus on a single country – Spain for instance – in order toexamine this issue. In general, generation and/or load nodal tariffs aresimilar in nodes located nearby (in the same geographical area). Outlierswell above the average can be explained as corresponding to agents thateither produce or consume a small amount of power and are connected tothe rest of the grid through a dedicated line or lines (therefore, a large frac-tion of the cost of this line is charged to a small amount of power). Valueswell below the average are normally generators in a node where demand ispredominant (mostly importing nodes) or conversely loads in a node wheregeneration is dominant (mostly exporting nodes). One possibility to avoidthe problem of excessive dispersion of the tariffs is to compute and applyzonal tariffs. In order to compute them, each nodal tariff in a given zonewould be weighted by the amount of power either consumed or producedby the corresponding agent and an average for predefined geographicalregions would be obtained. In this way, those extreme values would have asmall influence in the resulting tariff to be applied to every generator or loadin the area (see Figure 7.4).

It is also possible to design methods to smooth out the individual nodaltariffs by simply taking into account the values of nearby nodes. This is theapproach that has been adopted here. The next paragraphs explain the


method that has been used to get rid of outliers and to make the geographicdistribution of transmission tariffs smoother.

The basic idea is to compute the new transmission tariff for a generationor a load node by taking into account both the original tariff of this nodeand the ones of the nodes of the same type (either generation or load) sur-rounding it. Therefore, only the tariffs of the surrounding generation nodesare considered to modify the tariff of a generation node and the same fordemand nodes. A sound method for allocation of the sunk grid costs mayprobably result in tariffs that are quite different for the generation and theload in the same region, depending on whether the region has a deficit ofgeneration or load. G tariffs in a mainly importing area will probably besmaller than L tariffs and vice versa for a mainly exporting area, thus sig-naling the advisability of installing more generation in the former case andmore demand in the latter. The objective is not to distort these signals bysmoothing the G and L tariffs together. Therefore the smoothing process iscarried out for generators and loads independently. The new tariff for agiven node A is computed as the weighted average of all the nodes of thesame type as node A. The tariffs of its neighbors are weighted using a func-tion that decreases with the distance to node A. The weighting factor thatis used for the original tariff of node A is 1 and it decreases the further awaywe move from this node. Therefore, nodes that are electrically closer tonode A will have a stronger influence in the new tariff of this node thanthose which are located farther from an electrical point of view. The elec-trical distance between two different nodes is measured in terms of theequivalent electrical impedance between these two nodes. Figures 7.5 and7.6 show, for the same IEM-13 system referred to above, the detail of theoriginal G and L tariffs for the different provinces within Spain. This is rep-resented by a thick line. The same figures show the corresponding newtariffs once the smoothing process is applied. This is represented by a thinline. Numbers are again expressed in €/MWh.

Finally, Figure 7.7 compares, for 17 European countries, the averagetransmission tariff of a typical industrial customer to the net inter-TSOpayment resulting from the application of the AP method to the same 24scenarios referred to above (see Pérez-Arriaga et al., 2002; Pérez-Arriagaand Olmos, 2003). Numbers are expressed in €/MWh.

According to this figure, there can be a difference of up to €7/MWhbetween the transmission tariffs in two different countries (see thedifference between the transmission tariff in Sweden and Spain). Thisdifference is of the same order of magnitude as that between the energyprices in two power exchanges (see Figure 7.8), which has been provided bythe Spanish power exchange OMEL. Since transmission tariffs are long-term signals that should not interfere with system operation, it is advisable


to give transmission tariffs the format of a capacity (€/MW per year)charge, or even a lump-sum charge (€/year), but never an energy charge(€/MWh), since this could introduce non-negligible distortions in the gen-eration dispatch in the region.

Figure 7.7 also shows that the net inter-TSO payment each country mustreceive (expressed again in €/MWh) is much smaller than the typical trans-mission tariff. Therefore, it is very unlikely that these inter-TSO paymentswill be able to send any significant locational signal in the presence ofwidely different transmission tariffs in the considered countries.

One must be aware of the fact that any method used to compute inter-TSO payments is, at the same time, allocating the cost of any line amongthe countries in the system. The importance of this feature should not beunderestimated, since the adopted method will automatically allocate thecost of any new network investment among the different countries, accord-ing to the use that each one of them is making of this facility. This has beenimplicitly accepted when adopting the inter-TSO payment mechanism andit appears to be an excellent way of solving a problem that typically hasproven to be a contentious one.


Figure 7.5 Original and new L tariffs in Spain for the IEM-13 system

0

2

4

6

8

10

Ala

va

Bad

ajoz

Bar

celo

na

Bur

gos

Các

eres

Cad

izS

anta

nder

Ciu

dad

Rea

lC

ordo

baG

uada

laja

raG

uipú

zcoa

Hue

lva

Hue

sLa

Cor

una

La R

ioja

León

Mad

rid

Mal

aga

Mur

cia

Nav

arra

Ore

nse

Pal

enci

a

Sal

aman

ca

Sev

illa

Tar

rago

naT

oled

o

Val

enci

aV

allo

dolid

Viz

caya

Zam

ora

Zar

agoz

a

13.6 10.46

Province ORIG_T New_T

Case Example of Transmission-related Locational Signals

This subsection compares the locational economic signals correspondingto two different areas in the transmission network of a fictive system wherea potential investor might consider installing a new CCGT power plant.The first location (L1) is close to a liquefied natural gas (LNG)regasification facility on the coast. The second location (L2) is close to amain load center. In both cases, the power plant is meant to meet thedemand growth in the aforementioned load center.

The most important locational factors can be grouped into three cate-gories:

● Locational signals related to the impact of losses, in both the elec-tricity and the gas networks. Network constraints are ignored heresince systematic congestions are infrequent in well-developed net-works. Most systematic congestions are usually removed by newnetwork reinforcements in the short or medium term since they typic-ally have a significant impact on the system operating costs.


Figure 7.6 Original and new G tariffs in Spain for the IEM-13 system

0

2

4

6

8

10

Province ORIG_T New_T

23.45

Ala

va

Bar

celo

na

Bur

gos

Các

eres

Cad

iz

Ciu

dad

Rea

lG

rana

da

Gui

púzc

oa

Hue

lva

Hue

s

La C

orun

aLa

Rio

ja

León

Mad

rid

Mur

cia

Nav

arre

Ore

nse

Pal

enci

a

Sal

aman

caS

evill

a

Tar

rago

naT

oled

o

Val

lodo

lidV

izca

ya

Zar

agoz

a

Alb

acet

e

Ast

uria

s

Can

tabr

ia

Lugo

Pon

teve

dra


Note: Figures are expressed in €/MWh. Per unit inter-TSO payments for Italy and Franceare too small to be visible since these are large countries with small transits. They are almostpurely exporting (France) or importing (Italy) countries. No data were available for Greeceand Luxembourg.

Figure 7.7 Comparison between the transmission tariff and the net inter-TSO payment for 17 European countries

–2

0

2

4

6

8

10

12

Austri

a

Belgium

Denm

ark

Englan

d an

d W

ales

Finlan

d

Franc

e

Germ

any

Greec

e

Irelan

dIta

ly

Luxe

mbo

urg

Nethe

rland

s

Norway

Portu

gal

Spain

Sweden

Switzer

land

Trans. tariff Net inter-TSO payment

Figure 7.8 Evolution of the energy price of several power exchangesbelonging to Europex, January 2000 to November 2004

70€/MWh

2000 2001 2002 2003 2004

60

50

40

30

20

10

0

Date

1º T

rimes

tre

2º T

rimes

tre

3º T

rimes

tre

4º T

rimes

tre

1º T

rimes

tre

2º T

rimes

tre

3º T

rimes

tre

4º T

rimes

tre

1º T

rimes

tre

2º T

rimes

tre

3º T

rimes

tre

4º T

rimes

tre

1º T

rimes

tre

2º T

rimes

tre

3º T

rimes

tre

4º T

rimes

tre

1º T

rimes

tre

2º T

rimes

tre

3º T

rimes

tre

4º T

rimes

tre

APX NORDPOOL LPX POWERNEXT OMEL GME

● Those signals related to the cost of new network facilities – both forgas and electricity – that are built because of the installation of newagents in the system.

● Other locational signals that may exist because of various reasons. Inthis case we have to consider signals derived from the effect of thealtitude over sea level on the useful generation capacity of the CCGTpower plants.

With the purpose of facilitating the comparison of both options, thepower plant has been assumed to provide the same amount of usable powerto the system regardless of whether it is installed in one place or the other.Table 7.1 shows the effect of transmission losses, the provision of ancillaryservices and the altitude over sea level on the total generation capacity thatis needed to supply an increment in demand of 384 MW in the load center.In both cases, the power plant must have the same useful generation capa-city (see first row of Table 7.1, ‘Useful power delivered’), which equals theassumed increase in the power demand in the consumption centre. Thisfigure is then modified to take into account the different increments in elec-tricity transmission losses that the system incurs when the same extraamount of usable power is produced in the two locations (see second row,‘Net power output’). Including the ancillary services that each power plantmust provide leads to a 4 per cent increase in its generation capacity (seethird row, ‘Nominal capacity’). Finally, we consider the effect of the alti-tude of each location. Whereas the plant near the LNG facility is locatedat sea level, the power plant close to the main load center is supposed to be


Table 7.1 Impact of different factors on the total generation capacity neededto supply a 384-MW load, located close to a main consumptioncenter, from two different locations, one close to the load centerand the other close to an entry point for LNG

Close to main Close to an LNGload center (L1) facility on the coast (L2)

Useful power 384.0 384.0delivered (MW)

Net power output (MW) 385.1 407.8(transmission losses)

Nominal capacity (MW) 401.1 424.8(ancillary services)

Nameplate capacity (MW) 422.2 424.8(altitude over sea level)

located considerably higher, thus experiencing a reduction of 5 per cent inits total generation capacity (20 MW for a plant with a nameplate genera-tion capacity of 400 MW). Therefore, in order to be able to produce 401.1MW, that is, the required nominal capacity of the plant in L2, the name-plate capacity of the power plant must be 422.2 MW. On the other hand,the power plant in L1 does not experience any efficiency losses related to itslocation because the effect of the altitude is negligible there (see last row,‘Nameplate capacity’).

Table 7.2 shows, for both locations, the different cost componentsinvolved in meeting an increase of 384 MW in the demand for electricity ofthe consumption centre close to L2. The useful life of the power plant hasbeen assumed to be 20 years and the cost of capital per annum 6 per centreal. The price of electricity production with natural gas has been assumedto be €17/MWh for the entire time horizon of the study (20 years) and thethermal efficiency of the power plant has been assumed to be 50 per cent inboth cases. We estimate that in both locations the power plant would befunctioning 5,000 hours a year at its full capacity. The per-unit investmentcost of the considered CCGT power plant is €1.208 million per megawattof installed capacity. This figure includes the cost of some dedicated facil-ities assumed to be needed to connect the power plant to the main electric-ity and gas systems (both the electrical transmission line and the pipelineconnecting the plant to the substation and the main pipeline system, respec-tively, have been supposed to be 40 km long).

Electricity transmission losses and the provision of ancillary serviceshave an impact on the amount of energy produced as well as on the cost ofconstruction of the power plant. The power plant must produce some extraamount of energy in order to supply the transmission losses and ancillaryservices allocated to it. This amount varies with the location of the plant.Moreover, the required capacity of the generation facility also depends onthe loss of capacity because of transmission losses and ancillary services(see Table 7.1).

The altitude over sea level has an effect on the required nameplate capa-city of a CCGT power plant because it limits the maximum power outputof the generator. Therefore, if this power output is prescribed from theoutset, the higher the power plant is located above sea level the larger thenameplate capacity of the power plant must be. Consequently, all otherfactors being equal, generation investment costs increase with the altitudeover sea level. Since the final amount of energy produced and the fuel con-sumption are the same regardless of the altitude, production costs aretherefore not affected by this factor.

Regarding the losses incurred when transporting the gas from the pointof injection to the place where it is consumed, it has been assumed that


a fraction of the transported gas is lost. The assumed percentage can onlybe regarded as an estimate for the purpose of this example, since we do notknow about consistent statistics on the relative importance of losses in gaspipeline systems. No gas losses are considered for the plant located at L1,since the plant is very close to the LNG regasification facility. On the otherhand, losses incurred when the power plant is located near the load centerat L2 are assumed to account for 0.7 per cent of the total gas consumed.


Table 7.2 Comparison of the cost savings involved in supplying a 384-MWload located close to a main load center

Market value Investment Investment inof energy in generation transmission

produced (€m) facilities (€m) facilities (€m)

L2 L1 L2 L1 L2 L1

Annual impact of 0.2 4.3 0.1 2.4 – –electricity transmissionlosses

Annual impact of 2.9 3.1 1.6 1.7 – –the provision ofancillary services

Annual impact of – – 2.1 0 – –the altitude oversea level

Annual impact of 0.5 0 – – – –losses in the gaspipeline system

Annual impact of – – – – 0 7.9the use of theelectricity gridinfrastructures

Annual impact of – – – – 3 0the use of the gasgrid infrastructures

Overall annual 72.7 76.5 42.0 42.3 3 7.9value (total annualvalue of the energyproduced or totalannual cost of thefacilities built)

We have taken into account both the price of gas and the thermal efficiencyof the power plant in order to compute the cost of these losses, which affectthe production cost of the power plant.

As for the locational signals related to the cost of those electric trans-mission lines that the new power plant must be held responsible for, wehave considered only new lines here. The cost of those lines that alreadyexist will probably be allocated according to some ‘average’ method whichresults in a socialization of their cost thus not giving rise to any locationalsignal.11 In order to assign the cost of new lines the novel method pre-sented in Section 4 has been used here, with the following simplifyingassumptions:

● The flow pattern in the transmission grid is such that area L1 exportspower to area L2 where the large consumption center is located.

● Lines connecting areas L1 and L2 are already congested or close tobeing congested.

Then, installing a new power plant in L1 to serve the load located nearto L2 would imply that a new line reinforcing the connection between bothareas must be built. If the new CCGT plant is the only new generator usingthe line and if a split of the cost of transmission lines of 50 per cent to gen-eration and 50 per cent to load is assumed, application of the methodexplained in Section 4 would result in the new plant at L1 having to pay halfthe cost of the line (the entire fraction of the cost to be recovered from gen-eration). The line has been assumed to be a typical 400-kV line with acapacity of 100 MW and a cost per kilometer of line of €0.35 million. Ifthe length of the line is assumed to be 550 km, the annual transmissioncharge to the plant at L1 would be €7.9 million. This figure is of the sameorder of magnitude as the locational signals that result from the existenceof losses in the transmission grid or from the effect of the altitude on thegeneration capacity of the power plant. Therefore transmission charges inthis example could influence the decision on whether to install the plant inone place or the other.

Alternatively, we could have used some ‘average-use’ method like AP toallocate the cost of both the old and the new lines. In this case, thedifference between total transmission charges in both locations (L1 and L2)would probably not have been very large and locational signals related tothe allocation of the cost of electric transmission facilities would have beenweaker than those considered in this case example with the new proposedscheme for transmission pricing.

Likewise, the cost of infrastructures of the gas grid that are needed bythe power plant at L2 have to be accounted for. Here we shall assume that


the gas consumed by the CCGT plant at L2 is coming from L1 through asingle pipeline whose cost is shared with gas consumers. It has been sup-posed that the infrastructures that are used to transport gas from L1 toL2 include 90 km of 20� pipeline, 450 km of 26� pipeline and two com-pression stations, both in the 26� pipeline. Given the amount of gas con-sumed by the power plant and the typical amount of gas transportedannually through the different gas pipelines, it has been estimated that aCCGT power plant at L2 would be using between 10 and 15 per cent ofthese infrastructures on average. This results in an annual charge of€3 million for the use of gas grid infrastructures by the plant at L2. Per-unit construction costs have been taken from Spanish Ministry ofIndustry, Tourism and Trade (2005).

Locational signals relating to the existence of congestion in the grid havenot been taken into account in this example since congestion is typicallyinfrequent within any system where the grid is sufficiently meshed.Assuming that the lines between L1 and L2 become congested in the futureand different energy prices are computed for both areas, locational signalsresulting from this congestion would amount to the difference in energyprices between both areas times the total power production of the powerplant over the period of time during which the lines are congested.Considering a difference in energy prices of 10 per cent of the averagesystem price (supposed to be €36/MWh) and assuming that the line is con-gested 30 per cent of the time, annual revenues from the sale of energywould be about €2 million higher at L2 than at L1. This figure is compar-able to the difference between the annualized effect of the altitude on theinvestment cost of the power plant in both locations (which is also about€2 million) but clearly smaller than the difference between the annual valueof the losses assigned to the generator in both cases (which is about €6.4million). However, numbers depend critically on the assumptions that havebeen made.

From the numbers shown in Table 7.2 one can conclude that the netpresent value of the cost savings that are obtained annually wheninstalling the power plant close to the consumption center, instead ofinstalling it close to the LNG regasification facility, would amount to €9million. This figure is the difference between the total annual costs asso-ciated with each one of the two locations (see last row in Table 7.2).Numbers in this row correspond to either the total annual value of theenergy produced in the two locations or to the total annual cost of thefacilities built in the two cases (9�(76.5 – 72.7)(42.3 – 42)(7.9 – 3)).Note that, when computing these numbers, we have assumed a certainpattern of electricity flows in the grid that remains the same during theentire time horizon of the study. Therefore, transmission charges and


losses also remain the same. The breakdown of costs in Table 7.2 showsthe relative importance of each one of the locational factors for this case.Obviously, these results may be different for other cases, but the approachto be followed in the evaluation should be basically the one that has beenshown here.

7. CONCLUSIONS

Restructuring and liberalization of the electric power industry has resultedin the unbundling of the activities of generation, transmission, distributionand supply. Investment decisions for generation and network infrastruc-tures are no longer in the same hands. The ensuing loss of efficiency due tounbundling may be partly compensated by coordinating locational eco-nomic signals at the different interfaces.

Efficient investment in generation can be encouraged by the use ofnodal energy prices and transmission tariffs with locational content.These signals should be designed on the basis of responsibility on theincurred network costs. The usefulness of network-related locationalsignals for transmission investment critically depends on the adoptedregulatory approach for transmission and, in particular, on which one isthe entity in charge of the expansion of the grid: the regulatory author-ity, the transmission system operator, merchant investors or the networkusers themselves. Other useful locational signals of a non-economicnature, whose purpose would be to decrease the level of uncertainty thatboth potential generation investors and transmission network plan-ners must face, consist in providing reliable information on future invest-ment decisions (generators) and expected network conditions (systemoperator).

Economic locational signals that are based on cost causality are alwaysuseful to guide the siting in the network of new generation and load facil-ities. The issue here is whether these locational signals may or may not bestrong enough to have any actual impact on the location of new networkusers or even on the decision to retire a power plant or a load from agiven site. The chapter has provided criteria and numerical examplesshowing the nature of these signals and what can and cannot beexpected from their application. A novel scheme to design transmissioncharges with a special emphasis on cost causality has been proposed.Locational signals may also be useful at the interface between trans-mission and distribution networks in order to minimize the distortionsthat result from having separate network operators under widely differentregulatory rules.


APPENDIX 7A1 MATHEMATICAL FORMULATIONOF THE PROBLEM OF CENTRALLYPLANNING THE EXPANSION OFTHE GRID BOTH WITHIN THETRADITIONAL AND THECOMPETITIVE REGULATORYFRAMEWORKS

Centralized Planning of Network Investments in a Traditional RegulatorySetting

The objective of transmission planning in the traditionally regulated envir-onment is to maximize the global social welfare resulting from the expan-sion of the grid. Given that generation was traditionally planned togetherwith transmission, maximizing the social welfare was equivalent to mini-mizing the cost incurred when supplying the system load while complyingat the same time with certain technical and reliability criteria. In order todo so one has to take into account the investment, maintenance and fuelcosts of the power plants, as well as the investment, operation and mainten-ance costs of the grid.

When deciding how to supply the demand, it was possible to consider thejoint cost of the transmission and generation involved in a number ofdifferent possible expansion plans and to choose the most efficient alterna-tive. The central planner had to consider a number of uncertain inputs tothe planning process, which basically were the demand forecast, the esti-mation of hydrological conditions, the costs of fuel and capital, and theexpected availability of the already existing and new generation and trans-mission equipment. From all this information the central planner had todecide where and how much to invest in new transmission and generationfacilities. In other words, contrary to what happens now that competitionhas been introduced, the central planner knew about the generation expan-sion plans and could modify them.

It is possible to write the mathematical formulation of this optimizationproblem. For the sake of simplicity, a ‘static’ optimization model is pre-sented, where the optimal mix of generation and transmission is deter-mined for a single given future year, ignoring the trajectory of investmentsalong the planning horizon. The objective function to be minimized is:

(7A1.1) CEN�N

i�1rs,i

�l��

�Kl

k�1hlk�

Mlk

m�1Xlkm �

n��Wn

w�1hnwYnw D�

s��

ps�N

i�1�

Ei

c�1cie

gs,ie


where � is the subset of branches (that is, corridors) where new investmentsare allowed, hlk the investment cost of a line of type k in corridor l, Kl is thenumber of different types of line (220 or 400 kV, single or double circuit,number of conductors per phase, overhead or underground) that areallowed as expansion options in corridor l, Mlk is the maximum number ofnew lines of type k that can be added to corridor l, Xlkm is a binary variablewhose value is 1 if line m of type k is built in corridor l and it is 0 otherwise,� is the subset of nodes where new investment in generation can be located,hnw is the investment cost of a power plant of type w at node n, Ynw is abinary variable that is 1 if a plant of type w is built at node n and it is 0 oth-erwise, D is the duration of the planning horizon, � the subset of consid-ered production cost scenarios, ps is the probability of occurrence ofscenario s, N represents the total number of areas, Ei is the total number ofgeneration blocks in area i, cie corresponds to the variable production costof generation block e in area i, gs,ie is the power output of generation blocke in area i for scenario s, CEN is the cost of unserved energy and rs,i is theunserved power in area i for scenario s.

The objective here is to minimize the total cost of supplying the demand,while taking into account the fixed costs of the newly installed generation,the variable power production costs and the transmission costs. Ohmiclosses can, for instance, be taken into account by adding a fictitious load atboth ends of each line with every load being equal to one half of the ohmicloss of the line.

A number of constraints have to be considered: maximum total numberof lines that can be installed in each corridor (7A1.2), maximum numberof lines of a given type that can be installed in each corridor (intrinsic tothe formulation), maximum total investment per corridor (7A1.3), or forthe entire network (7A1.4), maximum output for every thermal genera-tion block and for the equivalent hydroelectric unit in each area (7A1.5),maximum unserved energy in every area (7A1.6), electric networkequations (7A1.7) and the transmission capacity limit for every corri-dor (7A1.8).

(7A1.2)

(7A1.3)

(7A1.4)�l��

�Kl

k�1hlk�

Mlk

m�1Xlkm � TI

�Kl

k�1hlk�

Mlk

m�1Xlkm � Il l � �

�Kl

k�1�Mlk

m�1Xlkm � Ll l � �


(7A1.5)

(7A1.6)

(7A1.7a)

(7A1.7b)

(7A1.8)

where Il is the maximum investment in corridor l, Ll is the maximumnumber of lines to be installed in corridor l, TI is the maximum total invest-ment in the network, is the vector representing the ohmic lossesfor each one of the areas, � represents the vector of ohmic losses for eachone of the corridors, g is the vector of active outputs from every thermalgeneration block and every hydroelectric equivalent unit in each area, |g| isthe vector of maximum active power outputs from every thermal genera-tion block and every hydroelectric equivalent unit in each area. This vectoris a function of the investment variables in generation Ynw. Finally, r is thevector of area load curtailments.

Given that the construction times and the investment costs of powerplants used to be much higher than those for transmission lines, generationplanning normally took precedence over transmission planning. Once theexpansion of generation had been planned, investments in the transmissiongrid were decided while considering the investments in generation as given.In this case, the mathematical statement of the problem can be simplifiedsince the investment variables in generation Ynw are now input data.

As for the planning of the distribution grid, the objective remains thesame as that used for transmission. The planner must minimize the costincurred by supplying the forecast demand while complying with certainreliability criteria. Reliability can be incorporated into the objective func-tion to be minimized, it can be taken into account through reliability con-straints or it can be included both in the objective function and the set ofconstraints. Traditionally, planners developed the distribution grid inde-pendently from the transmission grid. Despite the fact that the planning oftransmission and distribution grids deserve a specific treatment, some levelof coordination between them is also advisable, as has been explained. Herewe leave aside the fact that the volume of distributed generation is growingvery fast in many countries. The cost factors that play a role in the opti-mization process are mainly investment in distribution grids, ohmic lossesand the cost of non-supplied energy (for details, see Peco, 2001).

� � 12ST�

| fs| � fs

fs � DST�s � 0

� Sfs gs rs � ds �s

0 � rs � ds

0 � gs � gs


Differences and Similarities in Centralized Transmission Planning UnderTraditional and Competitive Regulatory Frameworks

When competition in the generation activity is introduced, the entity incharge of optimizing the development of the grid no longer knows what theevolution of generation in the system will be. This fact adds much morecomplexity to the task of deciding which lines to build, due to the higherdegree of uncertainty that the network planner must face. It is generallyagreed that, also within a competitive framework, the ultimate objective ofthe process of planning the expansion of the grid should be to maximizethe global social welfare. At the same time, generators are competingamong them with the purpose of maximizing their profit, like consumersalso do. Therefore, assuming that there is a central network planner, his/herobjective will be to maximize the joint net profit made by generators andconsumers because of the utilization of the grid. This means that both thebenefits and the adverse effects of any type that are incurred by the con-struction of a line must influence the decision by the planner on whether toconstruct a line.

If it is accepted that generation has precedence over transmission –both in the traditional and the liberalized setting – then the theoreticalequivalence of the outcome of transmission planning under both regu-lations can be proved, when the former objective function is replaced bythe new one. Therefore, it will be shown that it is possible to provide asound conceptual answer to the problem of the objective function to bepursued by transmission planning in the new competitive environment.This has practical implications in the definition of the regulatory test(see Section 2). The simple proof of this statement follows: within thetraditional approach, investment in transmission and generation isjointly optimized. Since generators are paid their cost of service, theobjective is just to maximize the consumers’ welfare, that is, the utilitythey obtain from the use of electricity minus the cost of producing anddelivering electricity. Therefore, the objective function can be simplyexpressed as:

(7A1.9)

where U(D) is the utility function of consuming a demand D, CFG repre-sents the generation fixed costs, CVG are the generation variable costs andCT are the transmission costs (capital plus operation and maintenancecosts, which altogether can be considered as fixed costs).

Assuming an inelastic demand (in the range of prices that may be modi-fied by transmission investment) and if generation planning is also

Max{U(D) � CFG � CVG � CT},


prescribed from the outset, transmission planning becomes the typicalminimization of generation operation costs via network reinforcement:

(7A1.10)

Within the competitive regulatory framework, the entity in charge oftransmission planning (the system operator, typically, under regulatorysupervision) must apply the following optimization criterion in order toidentify the network reinforcements that must be proposed to the regula-tory entities for authorization:

(7A.11)

where the total cost of any justified investments is implicit in these netbenefits as network charges to consumers and generators. In general, it is agood sign in the design of the rules for competitive markets that the idealoutcome coincides with the one that the traditional approach wouldproduce under the same circumstances. This is exactly what has beenaccomplished here, as is shown next.

In a competitive wholesale market, the following expression holds:

(7A1.12)

where PD is the total payment (at wholesale level) of consumers, IG is thetotal income of generators (net of any network payments), IVT is theglobal variable income of the transmission network (based on the applica-tion of nodal prices to both consumers and generators) and RNC is theresidual network charge of the transmission network (that is, the part ofthe total network cost CT that is not recovered by IVT).

The preceding expression allows one to replace the objective function ofthe maximization problem in the traditional approach by this one that isentirely equivalent:

(7A1.13)

which shows that the maximization problem in the traditional approachcan be replaced by the following equivalent problem in the context of thecompetitive approach:

(7A1.14)Max{Net benefit of consumers Net benefit of generators}

{U(D) � PD} {IG � CVG � CFG} {IVT RNC � CT},

PD � IG � IVT � RNC � 0,

Max{Net benefit of consumers Net benefit of generators}

Min{CVG CT}.


since the transmission network is regulated so that CT�IVTRNC. Notethat, embedded in the net benefit of consumers and generators, are thecomplete payments for any justified investment in transmission facilities.

One must be aware that, in order for the proof to be valid, the assumedentity in charge of the centralized development of the grid in a competitiveenvironment must pursue those grid expansions that maximize the globalsurplus resulting from them. The same outcome in general will not beattained under different regulatory schemes, such as leaving the initiative tobuild any new network facility to coalitions of network users willing to payits cost or letting private promoters invest in new lines.

APPENDIX 7A2 THE METHOD OF AVERAGEPARTICIPATIONS

Background

The basic intuition behind the AP method is that the sources of the supplyto loads and the destination of the power injected by generators, as well asthe responsibility for causing the flows in all lines, can be assigned byemploying very simple heuristic rules that only make use of the actualpattern of network flows. Although this procedure does not intend tocapture the details of the physics of the problem, one could argue that, inan electricity market that works reasonably well, the power flows fromnodes where it is less expensive towards nodes where it is more expensive.Thus, using the actual network flow pattern may be a way of assigningsources and sinks to loads and generators, respectively, in a reasonable andeconomically meaningful manner. It is not the only possible way, but it isintuitive, and simple to explain and to compute. The earliest reference thatthe authors of this chapter know about of the AP method is a handbookon transmission pricing by the New Zealand transmission company,Transpower, dated late 1980s.

The AP algorithm is simple and robust, and it uses as its basic input thedata of historical or computed network flows, where there are obviously nosimplifications, and the actual network topology. The method is based ona proportionality assumption: the inflows to a node are distributed pro-portionally among the outflows. Causality, or attribution of responsibility,directly results from this assumption which allows one to trace the flowsupstream or downstream. Robustness, that is, little volatility with respectto input data, or the absence of arbitrary decisions, such as the choice of aslack bus, which critically influence the results in other more sophisticatedmethods, is a very desirable characteristic of this method. This is even more


so when the application of the method has significant economic implica-tions for the network users. The AP method is obviously not free fromassumptions – such as the rule of proportional allocation of flows at eachnode that is not fully supported by engineering principles (and it cannot be,as power does not really flow in the networks as a fluid in a pipeline) – butthis rule is of a physically intuitive nature, very much associated with therobustness qualities of the method, and it basically guarantees the ration-ality of its results, which cannot easily be challenged.

The algorithm of average participations has been used, with minor vari-ations, in real systems such as New Zealand (in the late 1980s), Poland orSouth Africa, as well as in specific transmission pricing studies in at leastChile, Central America, Romania, Spain and the IEM of the EuropeanUnion. It can be programed easily and it has been studied thoroughly in thetechnical literature. It is not free from objections, but it appears to be a rea-sonable procedure whose level of sophistication is well adapted to theproblem under consideration.

Description

The method requires as its basic input data a complete snapshot of thenetwork power flows corresponding to the specific system conditions ofinterest. The algorithm is based on the assumption that electricity flows canbe traced – or the responsibilities for causing them can be assigned – by sup-posing that, at any network node, the inflows are distributed proportion-ally between the outflows. Implicit in this rule is the additional assumptionthat the physical flow of electricity in a line is not the combination of manyflows in opposite directions, since only flows in the same direction as thefinal physical flow are assumed to exist. Under these assumptions, themethod traces the flow of electricity from individual sources to individualsinks; that is, the model identifies, for each generator injecting power intothe network, physical paths starting at the generator that extend into thegrid until they reach certain loads where they end. Symmetrically, the pathsfrom loads to generators can also be found, yielding exactly the same resultsof allocation of responsibility of flows to generators and loads. Then, thecost of each line is allocated to the different users according to how muchthe flows starting at a certain agent have circulated along the correspond-ing line.

This is how the method works: for every individual generator i, a numberof physical paths are constructed, starting at the node where the producerinjects the power into the grid, following through the lines as the powerspills over the network, and finally reaching several of the loads in thesystem. An analogous calculation is also performed for the demands,


tracing upstream the energy consumed by a certain user, from the demandnode until some generators are reached. One such physical path (with asmany branches as needed) is constructed for every producer, and for everydemand. In order to create these paths, a basic criterion is adopted: in eachnode of the network the inflows are allocated proportionally to theoutflows. This makes the method easier to implement. A simple example isshown in Figure 7A2.1.

According to the proportional distribution rule, generator G1 wouldcontribute 15�20/(2040) MW to the flow over line 1, 35�20/(2040)MW to the flow over line 2, as well as the total of 20 MW flowing over line3 and nothing to the flow over line 4. Similarly, demand D2 would con-tribute 20�35/(101535) to the flow over line 3, 40�35/(101535)to the flow over line 4, all the 35 MW flowing over line 2 and nothing tothe flow over line 1. The demand D3 of 10 MW has a contribution of20 � 10/(101535) to the flow over line 3, 40�10/(101535) to theflow over line 4 and no contribution to the flows over lines 1 and 2.

The participation of agent i in the use made of line j is obtained as thefraction of the flows starting at agent i that passes through line j. Themethod implicitly results in a 50/50 global allocation of costs to generatorsand demands. However, if desired, an ad hoc weighting factor could beused to modify this percentage.

NOTES

* The material in this chapter has benefited from countless discussions with manycolleagues in different parts of the world and for many years. We should mention atleast the members of the Infrastructure Working Group of the Council of EuropeanEnergy Regulators (CEER), the many participants in the discussions leading to the


Figure 7A2.1 Proportionality principle in average participations

G1

G2

D1

D2

15 MW

35 MW

D3 = 10 MW

20 MW

40 MW

L3

L4

L1

L2

transmission regulation for the Central American Electricity Market (SIEPAC project),the professionals from Red Eléctrica de España (the Spanish grid company) with whomwe have had discussions for several years about transmission investment issues and whohave provided the data for the second case example on locational signals, the par-ticipants in the meetings of the Cambridge–MIT Consortium, the members of theWorking Group on Transmission Tariffs of the French Electricity RegulatoryCommission (CRE) that for two years met regularly to discuss transmission issues, thestaff of the Spanish national energy regulator from whom we learned about the tariff-setting process in the gas sector and our colleagues Javier Rubio, Carlos Vázquez,Michel Rivier, Tomás Gómez, Juan Rivier and Jesús Peco from the Instituto deInvestigación Tecnológica (IIT).

1. REE stands for Red Eléctrica de España, the Spanish Electricity Grid.2. The alternative regulatory frameworks that are presented here are representative

cases of the existing possibilities, since an exhaustive enumeration is out of question.For instance, a combination of the first and second cases is possible: the TSO may haveto present a global transmission expansion plan to the regulator to be approved(under a technical viewpoint) as in the first case, but the TSO will be responsible forthe actual implementation of the plan and be subject to RPI-X remuneration, as in thesecond case.

3. This is a well-known fact that has proved to be true in numerous actual networks andthat also has a theoretical support (see Pérez-Arriaga and Rubio, 1995).

4. This is how The Victorian Supreme Court has expressed the regulatory test: ‘A new inter-connector or transmission system augmentation satisfies this test if it maximizes the netpresent value of the market benefit, having regard to a number of alternative projects,timings and market development scenarios. Market benefit here means the total netbenefits to all those who produce, distribute and consume electricity in the electricitymarket’. See Littlechild (2004, p. 20).

5. For instance, ‘benefits’ are frequently replaced by some measure of ‘electric usage’ andvery often the traditional version of the rule is used even in the context of competitivewholesale markets.

6. Since nodal prices can generally only recover a small fraction of the total transmissioncosts, the problem of determination of transmission tariffs that pay for transmissioncosts will be considered from now on to be tantamount to the problem of determinationof the long-term signals, regardless of whether nodal prices of energy are actuallyapplied in a system or not.

7. In most European countries it is probably true that most domestic consumers are amongthe least elastic ones, while large industrial consumers are typically very elastic to elec-tricity prices. However, this statement may be wrong in many developing countries,where industrial consumers generally subsidize domestic consumers.

8. Remember that signals that are derived from losses and congestions are short-term ones;they cannot generate the complete revenues for the required investment since: (a) ingeneral they will be too weak for that, due to the typical overinvestment in transmission;and (b) these signals will typically be much reduced – even almost disappear – once thereinforcement is built.

9. Assume a line with a nominal transmission capacity (in MW) that is defined as C1according to some agreed procedure. However, because of system security reasons, themaximum real power that the line can transmit under normal operating conditions is C2.Then, if the actual real power flowing through the line is P, the fraction of the line � thatis actually used is simply P/C2.

10. For instance, we could use coefficients that are much larger than 1 (like 5, for instance)at the time the corresponding line (in the case of kl) or generator (in the case of kGi) entersinto operation. Then, these coefficients would gradually decrease with the passing oftime until they become zero after five years for generators and 10 years for lines, forinstance.

11. Locational signals are caused by the difference among the transmission charges paid bygenerators placed at different locations.


REFERENCES

Australian Competition and Consumer Commission (2003), ‘Review of the regula-tory test’, www.accc.gov.au/content/index.phtml/itemId/142.

Brown, R.E. (2002), Electric Power Distribution Reliability, New York: MarcelDekker.

Bushnell, J. and S. Stoft (1996), ‘Grid investment: can a market do the job?’, TheElectricity Journal, 9 (1), 74–9.

Carrillo-Caicedo, G. and I. Pérez-Arriaga (1995), ‘Optimal reconfiguration of dis-tribution networks for a diversity of regulatory frameworks’, IEEE/KTHStockholm Power Tech Conference, Stockholm, Sweden, August 19–21.

Council of European Energy Regulators (2003), ‘CEER Guidelines on RegulatoryControl and Financial Reward for Infrastructure’, www.ceer-eu.org/.

Florence School of Regulation (2005), ‘Study on the inter-TSO compensationmechanism pursuant to article 3 of Regulation (EC)’, n.1228/2003, www.iue.it/RSCAS/ProfessionalDevelopment/FSR/.

Gilbert, R., K. Neuhoff and D. Newbery (2002), ‘Allocating transmission capacityto mitigate market power in electricity networks’, Massachusetts Institute ofTechnology, Cambridge, MA, www.econ.cam.ac.uk/electricity/.

Joskow, P.L. (2005), ‘Transmission policy in the United States’, Utilities Policy, 13,95–115.

Joskow, P. and J. Tirole (2002), ‘Transmission investment: alternative institutionalframeworks’, in Wholesale Markets for Electricity, November 22–3, Toulouse,France.

Khator, S.K. and L.C. Leung (1997), ‘Power distribution planning: a review ofmodels and issues’, IEEE Transactions on Power Systems, 12 (3), 1151–8.

Littlechild, S. (2004), ‘Regulated and Merchant Interconnectors in Australia: SNIand Murraylink revisited’, CMI Working Paper 37, Cambridge, UK, January.Transactions on Power Systems, 9 (4), 1886–94.

Littlechild, S.C. and C.J. Sherk (2004), ‘Regulation of transmission expansion inArgentina. Part I: State ownership, reform and the Fourth Line’, Department ofApplied Economics, CMI electricity project, University of Cambridge, WorkingPaper ep 61, www.econ.cam.ac.uk/electricity/publications/wp/, p. 75.

Mercados Energéticos for Osinerg (Peruvian Energy Regulatory Agency) (2005),‘Plan Estratégico para Modernización del Marco Regulatorio’, www.osinerg.gob.pe/.

Outhred, H.R. and R.J. Kaye (1996), ‘Incorporating network effects in a com-petitive electricity industry: an Australian perspective’, in M. Einhorn andR. Siddiqui (eds), Issues in Transmission Pricing and Technology, Boston, MA:Kluwer Academic, pp. 207–28.

Peco, J.P. (2001), ‘Modelo de Cobertura Geográfica de una Red de DistribuciónEléctrica’, Universidad Pontificia Comillas, Doctoral Thesis, Instituto deInvestigación Tecnológica, Madrid.

Pérez-Arriaga, J.I. (2002), ‘Methodology for cross-border tarification in the inter-nal electricity market of the European Union’, Power Systems ComputingConference (PSCC), Seville, June 24–8.

Pérez-Arriaga, J.I. and L. Olmos (2003), ‘Extension of the project on cost com-ponents of cross border exchanges of electricity’, European Commission,Directorate-General for Energy and Transport, http://europa.eu.int/comm/energy/electricity/publications/index_en.htm.


Pérez-Arriaga, J.I. and F.J. Rubio (1995), ‘Marginal pricing of transmission ser-vices: an analysis of cost recovery’, IEEE Transactions on Power Systems, 10 (1),546–53.

Pérez-Arriaga, J.I. and Y. Smeers (2003), ‘Guidelines on tariff setting’, in F. Lévêque(ed.), Transport Pricing of Electricity Networks, Boston, MA: Kluwer Academic,pp. 175–204.

Pérez-Arriaga, J.I. et al. (1987), ‘Situación del estado del arte en la planificación deredes de transporte de energía eléctrica’, Report prepared for Red Eléctrica deEspaña (in Spanish).

Pérez-Arriaga, J.I., L. Olmos and F.J.R. Odériz (2002), ‘Cost components of crossborder exchanges of electricity’, European Commission, Directorate-Generalfor Energy and Transport. http://europa.eu.int/comm/energy/electricity/publications/index_en.htm.

Rubio, F.J. (1999), ‘Metodología de Asignación de Costes de la Red de Transporteen un Contexto de Regulación Abierta a la Competencia’, Universidad PontificiaComillas, Doctoral Thesis, Escuela Técnica Superior de Ingeniería, Madrid.

Schweppe, F.C., M.C. Caramanis, R.D. Tabors and R.E. Bohn (1988), Spot Pricingof Electricity, Boston, MA: Kluwer Academic.

Spanish Ministry of Industry, Tourism and Trade (2005), ‘Orden ITC/102/2005 porla que se establece la retribución de las actividades reguladas del sector gasista,www6.mityc.es/energia/archivos/Orden%20Retribuciones%202005.pdf.


AC interconnectors 182AC power lines 109, 118access charges 205, 221access rights 221access rules 153, 233, 235, 241, 257–8‘active’ TSOs 234, 239

incentives for 239airline industry 65Alberta

mixing planned and merchanttransmission in 119

MW-mile charges in 107transmission network organization

in 148zero-congestion policy in 91, 102,

119algorithms 246alternating current (AC) power lines

109, 118altitude over sea level 273–3, 274aluminium–zirconium wires 115amortization charges 162ancillary services, cost of provision of

146, 272, 273, 274, 275Anderson, D. 67arbitrage 16, 155, 188areas of influence method 250Areva 72, 79Argentina

transmission investment in 235transmission network organization

in 148transmission pricing in 249, 250

association of European TransmissionSystem Operators (ETSO) 187,264

asymmetry of information 121–4,151–2, 192, 193, 224

attributes of transmission investments136–47

auctions 88, 109, 111, 118, 163, 168,191, 221, 233, 234, 235, 236

Australiaenergy only market in 39–40HVDC link with 141, 155transmission investment in 155transmission pricing in 249

Australian Competition and ConsumerCommission 237

Austrianet inter-TSO payment of 267, 271transmission pricing in 265, 266, 271

average cost of transmission 253–4average participation (AP) method 11,

263–4, 267, 268, 275, 283–5Averch, H. 66–7

Bailey, E. 66balancing market 158, 160, 163, 166Bar-Ilan, A. 33–4Barthold, L.O. 115base-load plants 3–4, 22–30, 43, 49Beesley, M. 66Belgium

cost of capital in 76generation investment in 70, 72net inter-TSO payment of 271transmission pricing in 265, 266, 271

Ben-Tal, A. 209benchmarking 135, 162, 164, 180benefit/cost ratios 172, 174–5, 238, 245BETTA (British Trading and

Transmission Agreements) 182Bjorndal, M. 193, 223blackouts see power cutsBorenstein, S. 108, 126Bower, J. 67, 71Brennan, M.J. 35British Electricity Association 59, 66British Energy 81Brown, R.E. 230, 231Brunekreeft, G. 107, 118, 124, 127budget-balance constraint 151, 152,

159

289

Index

‘bundled’ transmission services, pricesfor 165–6

Bushnell, J.B. 110, 187, 224, 231

Californiacongestion in 132generation investment in 57, 58, 62,

65independent system operators 54power shortages in 1, 31, 37, 45, 126

capacity charges 204, 205, 214, 217,222, 269

capacity expansion model 223capacity factor 75, 79capacity margins 4, 41–6, 47, 199capacity obligations 6, 11–12, 166capacity payments 6, 35–7, 40, 158capacity reserves 45capital costs 13–14, 55, 68, 69, 70,

72–80, 81, 120, 161–2, 273capital intensity 13–14, 55, 66–7, 74, 81carbon price trading 71, 80categorization of transmission

investments 137–43CCGT see combined-cycle gas turbine

(CCGT) plantsCEER (Council of European Energy

Regulators) 235, 237, 285Central Electricity Generating Board

(CEGB) 41, 157, 164Chao, H.-P. 138Chile, transmission pricing in 249, 250circulated fluidised bed combustion

technology 69coal-fired plants 219

technological switch to/from 58, 60,61, 62, 63, 64, 65, 66

factors affecting 67, 68–72, 81coalitions of network users, investment

by 233, 235, 239, 283incentives for 240–241, 242

combined-cycle gas turbine (CCGT)plants 4, 41–2, 46, 54, 244, 256

factors affecting location of 270–277technology change to/from 13–14,

54–5, 69, 70, 71–2, 73, 74–5, 79,80, 81

combined heat and power producers62, 64

Comitology 189

competitive bidding see auctionscomplementary goods 2computer systems 15, 124congestion 4, 8, 10, 16

economically optimal level of 7, 14EU regulation on management of

189, 190–191, 196, 201growth of 131–2pushed across borders 15, 140revenue adequacy in congestion

management 222social cost of 172three distinct costs associated with

14, 88–90see also congestion costs;

congestion rents; cost ofcongestion to load

unhedgeable 172, 173, 174–5, 177,178

zero-congestion policy 14, 91–3, 95,106, 119

strategic manipulation of 102–3congestion charges

bilateral schedules liable for 166efficiency properties of 199–202FTRs as hedge against 191, 192, 221incentive effect of 193, 221in New York ISO 132non-discriminatory 217in PJM 131–2, 173–4and revenue adequacy 222

congestion costscompared with congestion rent and

cost of congestion to load 14,88–90

definition of 89economic models focusing on 139LMPs reflecting 166measurement of 120–121, 172trade-off between security and

reduction in 235trade-off between transmission

investments and reduction in113, 139, 175–6, 187

ways to reduce 125congestion rents

compared with congestion costs andcost of congestion to load 14,88–90

definition of 89

290 Index

economic models and 139and merchant investment 9,

110–118, 156, 171, 177, 236,237, 241, 242

optimal transmission investment and95–6, 99–102, 106–7, 112–16

PJM projects financed from 144transmission rights yielding 9, 89,

110–113, 115, 191, 236, 247congestion revenue rights (CRRs)

109–10, 111, 112, 115constant returns to scale 113–14, 187construction costs

electricity transmission lines 275gas grid infrastructures 276location and 171plant type and 3, 67, 71, 73–4, 76,

79, 81consumer surplus 116, 117control areas 140cost–benefit analysis 172, 173cost causality, locational signals based

on 16, 188, 189–90, 200, 201, 202,204–17, 222, 223, 224, 225, 245,250, 253, 254, 259, 277

cost minimization 120–121, 122, 197,206, 230, 278, 279, 280, 282

cost of capital 13–14, 55, 68, 69, 70,72–80, 81, 120, 161–2, 273

cost of congestion to load 14, 88, 90cost of service regulation 119, 135, 156,

161, 166, 179, 233, 240, 260cost of unserved load 120, 121, 122,

125cost reflectiveness see cost causality,

locational signals based onCouncil of European Energy

Regulators (CEER) 235, 237, 285Credit Suisse First Boston (CSFB) 70,

71credits 234, 235Crew, M.A. 124, 187Curien, N. 190CUSC (Connection and Use of System

Code) 159Czech Republic, transmission pricing

in 265, 266

day-ahead markets 166, 188DC power lines 114–15, 118, 171, 236

see also high-voltage direct current(HVDC) transmission links

de Luze, G. 77‘deep’ interconnection policy 134, 142,

168–70delivery to load criteria 167–8demand charges see capacity chargesdemand schedules 158demand uncertainty 30–31Denmark

net inter-TSO payment of 271transmission pricing in 271

Department of Energy see DOE(Department of Energy)

depreciation rates 161DGEMP (Direction Générale de

l’Energie et des MatièresPremières) 68, 71, 72

digital technology 66direct current (DC) power lines

114–15, 118, 171, 236see also high-voltage direct current

(HVDC) transmission linksdiscount rate 97, 116discrete decisions 190, 193, 224distributed generation 230, 238, 258,

259, 260, 261, 262, 280distribution networks

coordination between transmissioninvestment and 17, 138, 231,258–62, 277, 280

definition of 258regulation of 234

Dixit, A.K. 31–3DOE (Department of Energy)

Annual Energy Outlook 71see also EIA-DOE (Energy

Information Agency andDepartment of Energy)

dual fired plants 59–66dual price functions 193

economic efficiency principle 16,189–90, 199–202, 217, 220, 222,224, 245, 248

economic models of transmissioninvestment 15, 132–3, 139, 154,181

‘economic’ transmission investments15, 133, 139, 142–3

Index 291

in PJM 143–4, 171–3, 174–5, 176–7versus ‘reliability’-driven

transmission investment 133,175–7, 180

economies of scale 8, 9, 10, 14, 87, 187compared with lumpiness 101–2,

113–16and market power 108transmission pricing with 151, 204,

250efficient size of network 8–9efficient use of network 8effort, reward for 120, 121, 123–4EIA-DOE (Energy Information

Agency and Department ofEnergy) 58, 65, 68, 69, 70

see also DOE (Department ofEnergy)

electric power network models 132–3Electric Transmission Week 181Electricité de France 24, 75, 76–7, 79,

81, 148electricity utilities, investment by

56–64, 67emissions 41, 43emissions trading 71, 80energy balancing costs 160–161, 162energy charges 269energy only markets 12, 39–40engineering reliability criteria

in England and Wales 158, 160,162–3

factors affecting 154ignored by economic models 15,

132–3, 181ISO responsible for applying 148,

165minimizing cost while complying

with 278, 280PJM’s 165, 167–9reliability investment to restore 142transmission investments driven by

134, 144–5, 162–3, 176, 177, 180England and Wales

capacity payments system in 6, 35–7,40

cost of capital in 76generation investment in 4, 35–8,

41–2, 47, 54, 57, 59, 65, 66, 67,68, 69, 70, 72, 78, 80

net inter-TSO payment of 271regulatory framework for

transmission in 159–62, 180transmission investment in 132, 142,

158–9, 162–4, 175, 177transmission network facilities in

136transmission network organization

in 147–8transmission pricing in 12, 37, 142,

158, 159–60, 161–2, 168, 169,182, 249, 271

wholesale market arrangements in12, 157–62, 179, 220

Enron 77, 81Erie West HVDC 169–70ETSO (association of European

Transmission System Operators)187, 264

EU see European Union (EU)European Commission 189, 228European Parliament and Council 187European Regulation 1228/2003 on

Conditions for Access to theNetwork for Cross-borderExchanges

in Electricity 187–92, 196, 200, 201,215, 217, 220, 221, 222, 224

European Union (EU)Directives 54, 164inter-TSO payments in 263–4, 265,

267, 268, 269, 271internal electricity market of 54,

189, 237, 246nodal transmission tariffs in

263–70, 271Regulations see European

Regulation 1228/2003 onConditions for Access to theNetwork for Cross-borderExchanges of Electricity

regulatory responsibilities in 135transmission investment policy in

146exit fees 175expected amount of unserved energy

36expected net present cost 97–8, 104expected value of capacity 36externalities 7, 87, 102, 108, 187

292 Index

feasible sets of transmission rights109–11

Federal Energy RegulatoryCommission (FERC) 135, 141,164, 165–6, 170, 172, 173, 191

Felder, F. 118FERC (Federal Energy Regulatory

Commission) 135, 141, 164,165–6, 170, 172, 173, 191

financial guarantees 256financial transmission rights (FTRs) 3,

10, 110–112, 167, 168, 169, 171,172, 191–2, 221

fines see penaltiesFinland

cost of capital in 76, 79generation investment in 42–3, 71,

72, 79, 81net inter-TSO payment of 271transmission pricing in 271

Finon, D. 77, 80Firecone Ventures Pty Ltd 118fixed costs of generation 3, 22, 23, 40,

195, 197recovery of 4, 6, 24, 26, 39see also construction costs

fixed costs of transmission 14, 94, 95,119

recovery of 9, 10, 11, 14, 15, 16,92–3, 100–102, 107, 108,117–18, 144

see also construction costsFlorence Regulatory Forum 189–90,

201, 217, 246Florence School of Regulation 264Ford, A. 37forward contracts 160, 192France

cost of capital in 75, 76–7HVDC link with 141, 157transmission network facilities in


in 147–8transmission pricing in 265, 266, 271

free riders 108, 116–18, 127fuel costs

congestion influenced by 139technology choice influenced by 30,

68, 69, 70, 71, 73, 78, 79, 80, 81

correlation between electricity priceand 14, 75

fuel mix see generating technologies,choice of

functional separation 135–6, 147–8future grid development, information

on 257

game theory 103Gans, J. 120gas-based electricity generation 41–2,

43, 126, 219, 244, 256factors affecting location of 270–277peaking plants 3–4, 22, 46technological switch to/from 13–14,

54–6, 57–65, 66, 80factors affecting 65–72, 73, 74–5,

77–80, 81gas bubble 78, 81gas grid infrastructures 16, 274, 275–6gas losses in pipeline 273–5gas only plants 58, 59, 60, 61, 62, 63,

64, 65, 66gearing rate 14, 38, 75, 76generating technologies, choice of

13–14, 30, 54–81generation, investment in 3–7, 12–14,

21–52, 54–81coordination between transmission

investment and 10–12, 16–17,102–7, 125, 187–228, 230–231,243–58, 277, 27–83

case examples 263–77generation capacity

investment cycles and 35–9investment in liberalized markets

41–6irreversibility and uncertainty and

30–35, 47modelling optimal level of 21–30,

47–52paying for capacity in practice 39–41

generation investment strategies 102–7generator deliverability investments

166, 168–9generator interconnection investments

137–8, 167geographic scope of TSOs 179Germany

generation investment in 70

Index 293

net inter-TSO payment of 271transmission pricing in 265, 266,

271Gilbert, R. 126, 221, 247Glachant, J.-M. 77, 80‘gold-plating’ 123Gollier, C. 79government subsidies 13, 16, 54, 67,

73, 81, 82, 150, 151graft 123–4Greece


Green, R.J. 39, 67, 190Gribik, P.R. 110grid codes 162, 234, 238, 240, 242grid equilibrium model 194–6GRTN (Gestore Rete Transmissione

Nazionale) 59, 66

harmonizing transmission pricingpractices 15, 133, 232, 263

Harvey, S.M. 88, 191Hawdon, D. 35hedging 39–40, 52, 78, 79, 81, 111–12,

191–2, 221, 236unhedgeable congestion 172, 173,

174–5, 177, 178Henney, A. 157Hirst, E. 132Hogan, W.W. 88, 110, 118, 188, 191,

193–4, 202, 215, 221horizontal integration of TSOs 133–4,

150Hungary, transmission pricing in 265,

266Hunt, S. 56, 67HVDC (high-voltage direct current)

transmission links 134, 141,155–6, 157, 169–71, 173, 182

hydro-electric generators 31, 43, 44, 58,60, 61, 62, 65, 66, 69

Hydro-Quebec 182

Iberian System 243ICRP (investment cost-related pricing)

250IEA (International Energy Agency) 57,

62, 64, 68, 71, 73IGF-CGM (Inspection Générale des

Finances & Conseil Général desMines) 75, 76, 79

incentive alignment 151–2incentive regulation see performance-

based regulation (PBR)incomplete contracts 152independent power producers (IPPs)

73investment by 56–64, 77–8, 81

independent system operators (ISOs)congestion revenue rights (CRRs)

issued by 109, 110, 111creation of 54FTR option rights issued by 111model of 148, 221planning by 102rationale for 148–9regulatory challenges associated with

149–50see also PJM (Pennsylvania–New

Jersey–Maryland)independent transco model 147–8, 179,

192–3, 220–221indivisibility see lumpinessinformation rent 121–4installed capacity markets 40–41integrated gasification combined cycle

technology 69inter-TSO investments 15, 132, 133,

134, 139–42, 147, 150, 157, 163–4,176, 179

in PJM 173–5inter-TSO payments in the EU 263–4,

265, 267, 268, 269, 271interconnection charges 158, 159interconnection rules 134, 167, 170internal electricity market 54, 189, 237,

246nodal transmission tariffs in 263–70,

271International Energy Agency (IEA) 57,

62, 64, 68, 71, 73intra-TSO investments 15, 132, 133,

134, 139, 147, 150, 176, 179, 180in PJM 171–3

inverse price elasticity rule 248investment at risk 236, 239, 240investment cost-related pricing (ICRP)

250investment cycles 35–9

294 Index

investment lags 33–4investment strategies 2–3, 98, 102–7Ireland


irreversible investment underuncertainty 13, 31–5, 47

Ishii, J. 67ISO New England 143, 176ISOs see independent system operators

(ISOs)Italy

generation investment in 4, 57, 59,65, 66, 70, 80

net inter-TSO payment of 271transmission network organization

in 147–8transmission pricing in 265, 266, 271

Johnson, L. 66–7Jörsten, K. 193, 223Joskow, P.L. 41, 66, 67, 72, 77, 78, 102,

112, 113, 125, 127, 132, 134, 139,145–6, 152, 154, 156, 164, 165,166, 169, 170, 175, 177, 183, 191,220, 221, 230, 231, 254

Kaye, R.J. 262Khator, S.K. 231King, S. 120

Laffont, J.-J. 151, 163Larsen, E. 35Latorre, G. 209, 223Léautier, T.-O. 120legacy infrastructure considerations

145–6Leung, L.C. 231levelized cost methodology 55, 68, 71,

72, 73, 79Lévêque, F. 190lignite 67line type 7linear lumpy technology 113–16linear mixed-integer programs 208–9Littlechild, S.C. 66, 67, 118, 235, 286load-duration curve 4, 22–5, 48, 49,

51–2load reduction programs 138lobbying 108, 118, 127

locational signals see locationaltransmission signals, long-term;nodal energy pricing

locational transmission signals, longterm

information-related 256–7, 277transmission charges 187–228,

240–250access rules and 257–8case examples 262–77incremental charges for new

network users 250–255for retiring generators 255–6

LOLP (loss of load probability) 5, 6,35–7, 40

Long Island, HVDC links with 155,171, 173

Long Island Power Authority (LIPA)155, 171, 173

long-term contracts 13, 14, 38–9, 78,79, 80, 134, 155–6, 157, 171,173

two types of 221long-term locational signals see

locational transmission signals,long term

loss of load probability (LOLP) 5, 6,35–7, 40

lump-sum charges 269lumpiness 7–8, 9, 10, 14, 87, 95, 97,

154, 187, 188compared with economies of scale

101–2, 113–16merchant transmission investment

with 119, 156–7transmission pricing with 191–2,

201, 204, 221, 250Luxembourg


MAAC (Mid-Atlantic Area Council)165

MacKerron, G. 75marginal cost of generation 195

equilibrium energy price comparedwith 4, 13, 22–6, 37, 39, 40,95–6, 165, 166

geographically differentiated 8, 95,244

Index 295

known to regulator 197wind power 103

marginal cost of transmissioncompared with congestion cost 139,

175, 187compared with congestion rent 100,

101–2, 106, 113compared with nodal price

differential 8, 95–6transmission prices reflecting 153,

253–4marginal locational pricing see nodal

energy pricingmarginal opportunity cost of

consumption 4, 25market design 140, 181market power

absence of 188, 193, 224, 245diversification as insurance against

41–2and fixed cost recovery 14, 108,

117–18of local suppliers due to

transmission limitations 119price distortions caused by 5, 39, 42,

88transmission rights and 110, 112,

247underinvestment as a method of

abusing 113, 236market windows 172, 173, 174–5,

176–7Massachusetts Institute of Technology

(MIT) 55, 71, 72, 73–6McDaniel, T. 124mean load factor 75mean reversion 34–5measuring transmission capacity 145Mercados Energéticos for Osinerg

(Peruvian Energy RegulatoryAgency) 254

merchant plants 38, 73, 74, 77, 81merchant transmission investment 14,

88, 102, 108–18, 236, 239congestion rents and 9, 110–118,

156, 171, 177, 236, 237, 241,242

financing costs for 156long-term contracts undertaken by

134

mixing planned and merchanttransmission 9, 118–19,180–181, 237, 242

in PJM 155, 169–71, 173policy initiatives based on 126–7regulatory framework

accommodating 150, 155–7,191, 221, 237

Mid-Atlantic Area Council (MAAC)165

Midwest ISO (MISO) 165Mishan, E. 67MISO (Midwest ISO) 165MIT (Massachusetts Institute of

Technology) 55, 71, 72, 73–6monitoring and control equipment 15,

144mothballing 12–13, 31, 33, 35, 37–8, 45

National Grid Company (NGC) 12,42, 157–64, 234

Nemirovski, A. 209net investment

definition of 21in selected countries 42, 43, 44, 45

net local demand curve 89net power output 272net present value 98, 276net social benefit 105–6NETA (New Electricity Trading

Arrangements) 12, 158, 162Netherlands

net inter-TSO payment of 271transmission pricing in 265, 266, 271

network operating practices 133, 140,144

network upgrade costs 142, 168–70,172

New Electricity Trading Arrangements(NETA) 12, 158, 162

New England, HVDC links with 141,171, 182

New England Power Pool 146New England regional expansion plan

143, 176New York City

HVDC links with 155, 156, 171,173

market power in 126New York ISO 112, 132

296 Index

New Zealandaverage participation (AP) method

used in 283, 284transmission network organization

in 147–8Newbery, D.M. 38, 67, 78, 79, 132NGC (National Grid Company) 12,

42, 157–64, 234‘Nimby’ constraints 143, 171nodal energy prices

congestion managed by 197, 198–9,201, 211, 212, 214–15, 216, 217,222

definition of 8financial transmission rights and 10,

112, 191, 221locational signals in 11, 240–245,

261, 262, 263, 277and optimal investment mix 10, 88, 90in PJM 8

nominal capacity 272–3non-discriminatory prices 16, 188–90,

200, 201, 202, 204, 217–20, 223,224, 248

non-linear mixed-integer programs208–9

Nord Pool 43, 45NordNed Cable 182Norway

generation investment in 43–4, 54,57, 59

net inter-TSO payment of 271transmission network organization

in 148transmission pricing in 249, 271

nuclear power plantsbase-load plants 3–4, 22, 43technological switch to/from 13–14,

54, 55, 58, 60, 61, 62, 63, 64, 65,66

factors affecting 67, 69, 71–2,73–80, 81

nuclear Pressurized Water Reactor(PWR) 71–2

NVE (Norwegian Energy RegulatoryAuthority) 59

O’Neill, R.P. 10, 188, 194, 223obligation capacity 6, 11–12, 166Ofgem (Office of Gas and Electricity

Markets) 12, 162, 164, 182, 159,161, 162, 164

oil-fired plants 60, 61, 62, 63, 64, 65, 66oil tankers 35Olmos, L. 268OMEL 268open access 233, 235, 241open-cycle gas turbines 22, 69operating and maintenance costs 68,

69, 75, 79, 81, 146, 161–2opportunity cost of investment 237optimal transmission

cost recovery for 99–102dynamic 96–7option value 97–8static 93–6, 102–3strategic manipulation of optimal

transmission planning 105–7option rights 110–111option value 6, 13, 31, 32, 33, 34, 79,

80, 97–8, 102Outhred, H.R. 262

participation constraint, TSO 150–151participation factor 252–3‘passive’ TSOs 233, 235, 239

incentives for 240path ratings 125payback period 172, 173, 174–5peaking plants 3–4, 22–30, 40, 46, 49,

51–2Peco, J.P. 260, 280penalties 6, 40, 125, 166, 234, 235, 240Pérez-Arriaga, J.I. 190, 193, 205, 230,

245, 246, 248, 250, 254, 262, 263,268, 286

perfect competition 33, 90, 112, 113,188, 197, 200, 204

perfect information 188, 193, 197,223–4

performance attributes of transmissionnetworks 146–7

performance-based regulation (PBR)distribution networks 259, 260transmission investment 88, 102,

119–25, 135, 180, 240difficulties for transcos 14–15,

124–5direct approach to 120–121NGC revenues 161–2

Index 297

policy initiatives based on 126–7two approaches to reducing

information rent 121–5performance norms for TSOs 135, 180permits for new transmission links

149physical attributes of transmission

network components 143–5Pindyck, R.S. 31–3, 35PJM Interconnection 127–8, 167, 170,

175, 178, 181PJM (Pennsylvania–New

Jersey–Maryland)capacity markets operated by 6, 11,

40, 166congestion charges in 131–2, 173–4creation of 54expansion of 140, 165, 173financial transmission rights (FTRs)

used by 110–112, 167, 168, 169,171, 172

HVDC projects involving 155,169–70, 171

industrial organization andwholesale market design in11–12, 165–6

nodal energy pricing by 8Operating Agreement 167, 168Regional Transmission Expansion

Plan (RTEP) 167, 168, 169, 171,172, 173

Reliability Assurance Agreement167, 168

reliability criteria in 165, 167–9reports 58transmission investment in 132,

143–4, 167–79transmission pricing in 142, 164–6,

167, 168–70, 173, 254PJM West (Allegheny Power Systems)

140planning approach, traditional

distribution investment 230, 280generation investment 230, 278–80transmission investment 14, 15,

87–8, 102–7, 118–19, 230,278–80

compared with competitiveregulatory framework 234,237–8, 281–3

plant decommissioning see retiringgenerators

plant size 79, 81Pollitt, M. 67pool system 6, 12, 35–7, 40, 43, 45,

146, 158, 165Pope, S.L. 191Portugal

net inter-TSO payment of 271transmission pricing in 265, 266,

271postage stamp tariff 142, 249power cuts 14–15, 36, 41, 115, 120,

121, 124, 125, 126Power Exchange 37power flow studies 163power transfer distribution factors

(PTDFs) 195, 197, 204, 206practical transmission planning policy

105–7pre-construction formalities 33price capping 5–6, 7, 12, 14, 39–41,

125, 154reducing power of incentive

mechanism 122–5price discrimination 16, 188–90, 200,

201, 202, 204, 217–20, 223, 224,248

price-duration curve 4, 23, 25, 49, 51price elasticity of demand 5, 10, 11,

119price-taking agents 193, 197primal linear program (PLIP) 198primal mixed-integer program (PIP)

194privatization 65production set 223productivity improvements 162profit maximization 2, 47, 106, 192,

199, 214, 242, 281profit sharing 122–4, 135, 152, 180property rights 154proportionality principle in average

participation 283–4, 285public consultation 161public goods 41, 154, 177public interest 1, 5, 18, 134, 135pulverised fuel technology 69

Quebec, HVDC links with 141, 182

298 Index

R&D costs 81RAENG (Royal Academy of

Engineering) 68, 69, 72, 73Ramsey rule 11, 217, 248ratchet mechanisms 122, 123, 135, 152,

161–2, 180rate of return (ROR) on investment

guaranteed 67high 114investment risk premiums affecting

93normal 123regulation 88, 102, 122, 125, 126,

127, 166rationing 25, 36, 41, 138, 154RAV (regulatory assets value) 161‘real’ option 31real-time markets 138, 166, 188reconductoring existing lines 15, 144Rede Electrica de Espana (REE) 59,

66, 233redispatching costs see congestion

costsreference periods (seasons) for network

design 222regional transmission investment

planning process 135regional transmission operators

(RTOs) 127–8, 141, 165, 179, 221regulatory assets value (RAV) 161regulatory framework for distribution

261regulatory framework for transmission

134–6, 179–80, 230–231in England and Wales 159–62, 180European see European Regulation

1228/2003 on Conditions ofAccess to the Network forCross-border Exchanges inElectricity

principles to guide 150–157regulatory paradigms 17, 232–8,

278–83in US 135, 165–6, 167, 170, 180

regulatory hold-up 152regulatory inefficiencies 66–7regulatory risk 93regulatory test 231, 232, 234, 235,

237–8, 242, 281relays and switches 15, 124, 144

reliability assessments 167–8reliability criteria see engineering

reliability criteria‘reliability’-driven transmission

investments 15, 133, 142–3, 154,241

in PJM 167–9, 175, 178regulatory test allowing 238versus ‘economic’ transmission

investment 133, 175–7, 180reliability of supply see engineering

reliability criteria; ‘reliability’-driven transmission investments;security of supply

remote monitoring and controlequipment 15, 144

remote supply function 88, 89renewable energies 13, 58, 59, 60, 61,

62, 63, 64, 65, 66, 69, 82see also hydro-electric generators;

wind powerrent extraction goals 151rental cost of transmission lines 94,

113, 115, 116, 120, 124request for proposals (RFP) 155Réseau de Transport de l’Electricité

(RTE) transmission network 136,157

retailers, long-term contracts with 38–9retiring generators

economic factors determining 12–13,21, 31, 33, 47

growing number of 175locational signals for 255–6

returns to scale see constant returns toscale; economies of scale

revenue adequacy 222revenue requirement 161Ring, B.J. 188, 194, 202risk

attitudes to 18, 153regulatory 93and technology choice 13–14, 73–80,

81Rose, N. 67Rosellón, J. 125, 191, 220Rotger, J. 118Roulet, M. 77Royal Academy of Engineering

(RAENG) 68, 69, 72, 73

Index 299

RPI-X remuneration 234–5, 240, 260RTEC (Réseau de Transport de

l’Electricité) transmission network136, 157

RTOs see regional transmissionoperators (RTOs)

Rubio, F.J. 245, 246, 286

Santaholma, J. 72, 79, 80scale factor 253scarcity rent 202Scarf, H.E. 190Schmalensee, R. 66, 67, 139, 146, 152Schwartz, E.S. 35Schweppe, F.C. 244Scotland–England interconnector 182second-best regulatory mechanism 151security of supply 147

capacity margins and 46transmission investment and 14–15,

108, 119, 124–5, 132, 137, 138,234, 235, 240, 242, 259

see also engineering reliabilitycriteria; ‘reliability’-driveninvestments

sensitivity analysis 73SERP 107Seven-Year Forward Statements

(NGC) 158–9‘shallow’ interconnection policy 134,

142, 170share prices 46, 77Sherk, C.J. 235shirking 123simulation studies 132single system paradigm 247Slovakia, transmission pricing in 265Slovenia, transmission pricing in 265SMD (standard market design) 164–5,

191Smeers, Y. 190, 205, 245, 246, 248, 250SO procurement behaviour 154social cost of congestion 172social welfare maximization 278, 281software technologies 147Spain

generation investment in 4, 57, 59,65, 66, 70, 80

net inter-TSO payment of 271requests for connection in 256

transmission investment in 146transmission network organization

in 147–8transmission pricing in 265, 266,

267, 268, 269, 270, 271Spanish Ministry of Industry, Tourism

and Trade 276spark spread 77stability ratings 128standard market design (SMD) 164–5,

191standards 12, 158, 162–3State Aid 73, 81static models 224stock market values 77Stoft, S.E. 52, 108, 110, 119, 126, 187,

224, 231stranded costs 3Strange, W.C. 33–4strategic behaviour 2–3, 98, 102–7strategic generation investment

problem 103, 104–5subcontracting 123–4subsidies 13, 16, 54, 67, 73, 81, 82, 150,

151substation facilities 144, 230, 260Sun, H. 124sunk costs 30, 31, 33, 38, 47, 92–3, 247,

268supernormal profits 26, 28, 47Svenska Kraftnät 45Sweden

generation investment in 44–5net inter-TSO payment of 271transmission pricing in 249, 268,

271Sweeting, A. 157Switzerland

net inter-TSO payment of 271transmission pricing in 265, 266, 271

system balancing costs 160–161, 162

tariff pancaking 246–7Tasmania, HVDC link with 141, 155technology, locational price as function

of 218–19, 223technology changes see generating

technologies, choice oftelecommunications industry 66TenneT 182

300 Index

Texascongestion in 132generation investment in 57, 58, 62,

65, 77independent system operators

covering 54thermal efficiency 41, 46, 71, 273, 275thermal limits 195, 198, 206, 215thermal power plants 16, 44timeframe of investments 14, 74–5, 81,

124, 242, 244Tirole, J.J. 41, 112, 113, 127, 134, 139,

151, 154, 156, 163, 169, 175, 177,191, 201, 220, 221, 230

total revenue, generating plants 25–6transaction costs 140transcos

performance-based regulations forsee performance-basedregulation (PBR)

planning by 102see also independent transco

modeltransfer payments 89transformer upgrades 144, 171transmission

investment in 7–9, 14–15, 87–127,232–42

coordination between distributionnetwork investment and 17,138, 231, 258–62, 277, 280

coordination between generationinvestment and 10–12, 16–17,102–7, 125, 187–228,230–231, 243–58, 277, 278–83

case examples 263–77see also regulatory framework for

transmissiontransmission charges

England and Wales 12, 37, 142, 158,159–60, 161–2, 168, 169, 182,249, 271

harmonization of pricing practices15, 133, 232, 263

locational signals in 187–228,240–250

access rules and 257–8case examples 262–77incremental charges for new

network users 250–255

for retiring generators 255–6non-transaction-based 246PJM 142, 164–6, 167, 168–70, 173,

254transmission cost functions 14, 93, 94,

100, 101–2, 106, 112, 245, 248transmission licences 158, 163, 234,

236transmission line relief orders (TLRs)

131transmission losses 8, 24, 146, 166

in distribution grids 280economic incentives to reduce 235in economic models of transmission

investment 139, 175locational signals related to 270magnitude of 87, 100, 273, 274mechanisms to account for 279

transmission network facilities,definition of 136–7

transmission network organization147–50

transmission planning process,requirements of 154–5

transmission rights 9, 88, 89, 108–10,138, 153, 163, 236, 241, 257

congestion revenue rights (CRRs)109–10, 111, 112, 115

financial transmission rights (FTRs)3, 10, 110–112, 167, 168, 169,171, 172, 191–2, 221

and market power 110, 112, 247paradox of 112–13planning incorporating 154, 231

transmission service price structures152–3

transmission signals, locational seelocational transmission signals,long term

Transmission System Security andQuality of Service Standards 158,162–3

transparency of long-term signals215–16, 224–5

Transpower 283tree pruning 15, 124trigger price investment rule 32–3Turvey, R. 67TVO 14, 79two-part tariffs 201

Index 301

uncertaintyirreversible investment under 13,

31–5, 47signals to reduce 255, 256–8, 277types of 30–31

underwater links 155unhedgeable congestion 172, 173,

174–5, 177, 178unit commitment problem 193, 202,

223United Kingdom see England and

Wales; Scotland–EnglandInterconnector

United Statesblackouts in 124, 126cost of capital in 75, 76generation investment in 2, 4, 13–14,

37, 45–6, 47, 53–65, 67, 68, 71,72, 73–4, 77–8, 80, 81

legacy infrastructure in 145–6market power in generation in 126obligation capacity in 6regulatory framework for

transmission in 135, 165–6, 167,180

reliability of supply in 14–15transmission congestion in 131–2transmission investment in 142,

164–5, 180see also PJM (Pennsylvania–New

Jersey–Maryland)transmission network facilities in


in 148, 179transmission pricing policy in 164–5,

170see also PJM (Pennsylvania–New

Jersey–Maryland)unreliability 92unserved load, cost of 120, 121, 122,

125upgrading lines 125US Energy Information

Administration (EIA) 164

use of system charges in England andWales 142, 158, 159–60, 161, 168,169, 182

regulatory framework fordetermining 161–2

useful life of a power plant 273useful power delivered 272

value of lost load (VOLL) 5–7, 36, 40,176

vertical integration model 147regulatory challenges associated with

149viability constraint, TSO 150–151Vickers, J. 66Victoria (Australia), HVDC link with

141, 155Vogelsang, I. 107, 125, 192, 223VOLL (value of lost load) 5–7, 36, 40,

176voltage increases 144voltage level, transmission charges

related to 257–8

wholesale market prices 38–9, 42, 74,77, 120, 133, 140

willingness to pay 5, 10, 41, 153, 197,201, 214, 223–4

Wilson, R. 128, 138wind power 65, 69, 103, 106–7,

256wireless network 66Wolfram, C.D. 67, 157Wolsey, L.A. 206, 228Woolf, F. 191

Yan, J. 67Yarrow, G. 66yield to equity 14, 75–6

zero-congestion policy 14, 91–3, 95,106, 119

strategic manipulation of 102–3zonal prices 3, 262, 267

302 Index

Competitive Electricity Markets and Sustainability

Documents

Transcript of Competitive Electricity Markets and Sustainability