A model-based analysis of strategic consolidation in the German electricity industry

Energy Policy 29 (2001) 987–1005

A model-based analysis of strategic consolidation in the German electricity industry John Bower, Derek W. Bunn*, Claus Wattendrup Energy Markets Group, London Business School, Decision Technology Centre, Sussex Place, Regents Park, London NW1 4SA, UK Received 2 June 2000

Abstract When Germany liberalised its electricity market in April 1998, wholesale and heavy industry prices fell by as much as 60%. This was initially seen around the world as a success story in electricity market liberalisation as it involved a minimum of institutional interference. However, this approach may have become a victim of its own success. Facing a significant fall in profits, the largest generators adopted a strategic response with three sequential phases: (i) cut costs so as to be able to lower prices and maximise market share to deter entry, (ii) seek regulatory approval to acquire or merge with rivals to create four dominant vertically integrated firms, and (iii) close marginal plant to reduce overcapacity. Using a model-based simulation approach that has previously been successfully applied to strategic behaviour in the UK electricity market, this paper demonstrates that the process of strategic consolidation could result in average annual on-peak prices rising by 87% and average annual off-peak prices by 50%. Furthermore, the impact of an increase in industry concentration would be magnified as generators’ close their marginal plant. The creation of four dominant firms, who have the discretion to strategically withdraw capacity, appears to result in a substantial increase in market power in the wholesale electricity market and hence a significant rise in prices above the competitive levels that initially emerged. # 2001 Elsevier Science Ltd. All rights reserved.

When the German electricity market was liberalised in April 1998, implementing the European Union’s (EU) Electricity Market Directive (European Parliament, 1997), all consumers were immediately allowed, in principle, to choose their supplier. The results were remarkable. Wholesale market prices fell sharply and approached marginal production costs during 1999. Fig. 1 shows the development of wholesale prices in Germany and elsewhere during this period. The Central European Price Index (CEPI) index takes into account all reported day-ahead deliveries in (what used to be) the PreussenElektra (now E-ON) region in Northern Germany and is published by Dow Jones. The SWEP index reports day-ahead on-peak prices at the Swiss– German border near Laufenburg. While the CEPI and SWEP indices lack liquidity they do indicate that German prices converged to those in NordPool during

1999 and undercut those in the Amsterdam Power Exchange (APX). Within a matter of months, German wholesale price indices were much lower than those in the Pool of England & Wales, which had been seeking to provide competitive wholesale prices since 1990, but where generator market power had remained a significant problem (see for example Office of Electricity Regulation, 1998, 1999a–d).1 The German wholesale market for bulk electricity between generators, suppliers, and large consumers developed spontaneously through an informal bilateral market. Average industry tariffs were reduced by 27% from the beginning of the liberalisation to the end of 1999 as calculated by DJ/VIK Index (see Fig. 2). The lowest prices were seen in the transmission areas owned by the largest operators. Within this broad aggregate, price falls of up to 60% were reported for the largest industrial customers, according to Atkins and Taylor (1999).

*Corresponding author. Tel.: 0044-207-262-5050; fax: 0044-207-207724-7875. E-mail addresses: [email protected] (J. Bower), [email protected] (D.W. Bunn), [email protected] (C. Wattendrup).

1 All of the reports published by the UK regulator (Ofgem} Office of Gas and Electricity Markets) and its predecessor (Offer}Office of Electricity Market Regulation) can be found at http://www.ofgem.gov.uk

1. Introduction

0301-4215/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 3 0 1 - 4 2 1 5 ( 0 1 ) 0 0 0 3 4 - 9

988

J. Bower et al. / Energy Policy 29 (2001) 987–1005

Fig. 1. Comparative European electricity spot prices 1998–2000.

The rapidity and size of the price falls were a surprise and raise a number of issues concerning market design, industry structure and regulation for other countries considering or currently in the process of implementing liberalisation of their own electricity markets. The initial success of the liberalisation was achieved without an institutionalised market (two enterprising power exchanges began operation in 2000), and indeed without a dedicated regulatory body (the German Cartel Office and European Commission have been seeking to restrain anti-competitive behaviour). However, there were a large number of players in the market and this, together with over capacity, has generally been seen as the main reason for the competitive pressure. However, the initial price reductions might also be explained as a deliberate strategic move by incumbents aimed at retaining market share by preventing new entry. The marketing strategy of the incumbents was generally one

of matching or undercutting the best prices advertised elsewhere, with an inevitable downward price spiral. As such it ensured that there was no new physical entry during this period and relatively little consumer switching, especially at the household level. The consequent fall in prices reduced industry profitability, for example RWE Energie (RWE) announced a rise in sales of 25% but profits down 15% in the year to June 2000. As a result all of the fall in profitability, all eight of the major vertically integrated generating companies, and many smaller ones, were involved in merger negotiations by the beginning of 2000. Although these did attract concern at the European Commission level, the German electricity market was eventually transformed from a fragmented highly competitive market structure at the beginning of 1999 to one where four dominant vertically integrated firms with a combined market share of over 90% controlled the market by the beginning of 2001.

J. Bower et al. / Energy Policy 29 (2001) 987–1005

989

Fig. 2. Dow Jones/VIK industrial electricity price index for Germany.

Within a couple of years therefore, Germany faced the same regulatory dilemma as many other countries that have liberalised their electricity markets, namely how to control the exercise market power by a concentrated oligopoly of firms operating in the generation sector. In the remainder of this paper we address this question using an agent-based simulation (ABS) technique that captures the complexity of the strategic and regulatory choices facing the German electricity generation sector. This approach allows the strategic behaviour of generators to be simulated, and wholesale price scenarios to be investigated under a variety of industry structures. The paper is organised as follows. In the next section, the structure of the German ESI, the liberalisation process, and the legislative and regulatory framework that underpinned it are described. The strategic response of the industry to falling prices is also examined and a preliminary analysis of market power is carried out using traditional measures of industry concentration. Then the ABS simulation approach is presented, followed by results for a number of strategic scenarios.

2. The German electricity liberalisation Germany is the largest producer of electricity in the European market with more than 520 TWh of total generation output from an installed capacity of over 100 GW. Generation technology is diverse, with load being served by nuclear (31%), conventional thermal (54%), hydro (4%), industrial co-generation (6%) import (5%) capacity during 1999. Transmission and distribution are still (in 2001) vertically integrated with

generation and retail supply businesses in Germany, but companies are obliged to keep separate accounts for these. Eight large companies operated the high-voltage grid and controlled more than 80% of generation capacity, in 1999 there were also approximately 80 regional electricity companies, together with more than 800 municipalities and smaller suppliers. One of the most unusual features, as shown by Drasdo et al. (1998), was the complex cross-ownership with the eight large firms being financially interlaced and controlling more than half of the 50 biggest generators. Heavy industry accounts for 50% of demand with the remaining 50% equally split between commercial and domestic users. The daily load profile in Germany is typical of a western industrialised economy with peak demand during mid morning and late afternoon with lowest demand during the early hours of the morning. System peak demand usually occurs in December or January and is at a low in August. 2.1. Legislative programme The liberalisation of the German electricity market, starting in 1998, ended more than 100 years of local monopoly supply. Germany implemented the EU Electricity Market Directive from 1996 into a new energy law, the Energiewirtschaftsgesetz (EnWG), on 19 April 1998, its first fundamental change since 1935. The EnWG combined the negotiated third party access model with an optional single buyer approach for small municipalities in order to protect their local interests (e.g. using profits from sales of electricity to fund public transport). With these legal changes, Germany, in contrast to most of its neighbours, opened its market

990

J. Bower et al. / Energy Policy 29 (2001) 987–1005

fully to competition at once, ending an era of regional monopolies protected by demarcation agreements. Suddenly every consumer was able to choose from a wide range of different suppliers. Germany rejected the idea of an independent system operator and left questions like the detailed regulation of grid access and transmission pricing to be negotiated by different associations in the electricity industry itself and the German heavy industry. The results of these talks were the so-called associations agreement or . Verbandevereinbarung (VV) in May 1998, and the grid code by the grid operators organisation. However, practical problems still remained, mainly because of insufficient regulation of transmission. High transmission prices, several cases of transmission access being refused and the vertically integrated structure of the German ESI with 8 large companies owning the grid and most of the generation capacity provoked criticism, leading to a revision of the first VV. The VV2 came into effect on 1 January 2000 and was designed to overcome major problems, abolishing distance-based tariffs apart from a transmission fee between two newly established zones (North–South) and facilitating access for small customers. 2.2. Strategic response The response of firms to liberalisation was rapid, both inside and outside Germany. Many foreign energy companies already had experience of electricity market liberalisation and moved quickly to start their own retail activities, like Fortum from Finland (with Hansestrom) and Vattenfall (with VASA Energy), or buy substantial shareholdings in German generators. Notable were the . 25.1% of Energie Baden-Wurttemberg (EnBW) that Electricit!e de France (EdF) bought from the city of Stuttgart, the 26% of Berliner Licht- und Kraft AG (BEWAG) that was acquired by Southern Energy (US) and the 25.1% and 21.8% of Hamburgische Electricit.ats-Werke (HEW) that were bought by Vattenfall and Sydkraft, respectively. In order to compensate for the drop in earnings most of the incumbent German firms announced cost cutting measures, principally consisting of reductions in the workforce, e.g. BEWAG declared a workforce cut of . 50% over 3 years (Knodler, 1999). This is on top of average reductions of the 30% that energy firms implemented between 1991 and 1998 as part of the general restructuring of German industrial conglomerates. As a result of intense competition, one-third of the smaller utilities entered into merger or joint venture discussions (Atkins and Taylor, 1999) during early 1999. This corporate activity was however dwarfed by a series of mergers between the large vertically integrated firms which were instigated in part by their desire to become significant internationally, for example RWE stated

publicly its aim to increase its share in the European energy market from 2.3% to 10% by 2010 (see Wintermann, 1999; Atkins, 1999). The first major move to consolidate the industry came in September 1999 when VEBA and VIAG, two German conglomerates with electricity subsidiaries PreussenElektra and Bayernwerk, respectively, revealed plans for the biggest merger in German economic history. This created an interconnected utility (E-ON) that stretches the entire North–South length of the former West Germany. One month later, RWE announced its intention to merge with Vereinigte Elektrizit.atswerke (VEW), its smaller regional counterpart to form a dominant group in the North West of Germany. Together, the VEBA/VIAG and RWE/VEW mergers meant that more than 50% of generation capacity and output was under the control of just two groups. The anti-trust law, Gesetz gegen Wettbewerbs. beschrankungen (GWB), was modified in January 1999 to include the electricity sector. Before then, regional monopolies in the energy sector were tolerated and excluded from competition control. Despite this, the price war induced a benign regulatory attitude within Germany, though the European Commission sought to moderate the VEBA/VIAG and the RWE/VEW mergers with the imposition of special conditions, involving the divestiture of some plant (Financial Times, 2000a). The VEBA/VIAG and RWE/VEW mergers were authorised in early 2000 (Wetzel, 2000) after agreement was reached on (i) divestment of shares in commonly owned generators, especially in BEWAG and the East German generator Vereinigte Energiewerke AG (VEAG), (ii) scrapping of the transmission tariff between North and South Germany, and (iii) an agreement to sell or auction cross-border transmission capacity where constraints appeared (Harnischer and Hargreaves, 2000). Various foreign energy suppliers subsequently announced their interest in VEAG, especially Southern Energy that already had a stake in BEWAG. However, Vattenfall eventually gained control of BEWAG via its German HEW subsidiary (Financial Times, 2000e). It seems likely that Southern Energy and Vattenfall will eventually agree terms for merging BEWAG and VEAG. This will create a third large energy group in the northeast of Germany.2 The fourth merger began when EnBW increased its shareholding in its neighbour Neckarwerke Stuttgart (NWS), Germany’s biggest regional electricity company, to become the majority shareholder. The city of Stuttgart agreed to sell at least 17.5% of its 42.5% stake in NWS to EnBW which 2 HEW supply covers Hamburg and as such is embedded in the supply area of VEBA/VIAG but is also close to the supply area of BEWAG/VEAG so it remains, in 2001, unclear as to which group it will eventually be merged with.

J. Bower et al. / Energy Policy 29 (2001) 987–1005

already owned 25% of NWS (cf. Die Welt, 1999). Combined with the stake that EdF holds in EnBW this creates a fourth large player in the southwest of Germany (Financial Times, 2000b). The degree to which the structure of the German ESI has been transformed by liberalisation is illustrated in Fig. 3. This compares the structure of the industry before mergers took place, as on 31 March 1999, and as it might look by the end of 2001 assuming that all the mergers discussed above are eventually authorised. 2.3. Conventional analysis of market concentration Many regulatory bodies throughout the world look at mergers using the Hirschman–Herfindahl Index (HHI), and this is also employed by the Monopolkommission, set up in 1973 by the German government to report, every 2

991

years, on mergers and competition policy. The HHI is the sum of the squared market shares with a minimum of zero for perfect competition and a maximum of 10,000 for a total monopoly. Fig. 4 shows the HHI for the German electricity market before and after the VEBA/VIAG and RWE/VWE mergers. Additionally the HHI is calculated for four mergers, taking into account the former two, an acquisition of NWS by EnBW with a controlling stake held by EDF and a fusion of BEWAG and VEAG. This analysis shows a considerable increase in concentration after the first two mergers of VEBA/ VIAG and the RWE/VEW and more after the completion of BEWAG/VEAG and NWS/EnBW/EDF. The calculation of the HHI is based on data evaluating the market share in generation capacity by the 10 biggest German generators and their shares of production

Fig. 3. Capacity shares in German ESI before and after proposed mergers.

Fig. 4. Herfindahl Index calculated from generation capacity before and after mergers.

992

J. Bower et al. / Energy Policy 29 (2001) 987–1005

capacity in jointly owned plants. VDEW data is extracted from Vereinigung deutscher Elektrizita¨tswerke (VDEW) (1998) that takes shared plants into account as well. Own data evaluates the output of each company’s own plants plus the output from other utilities in which these companies have a share of at least 60%. Depending on the database the HHI before the mergers is between 500 and 800 and after the mergers between 800 and 1400. The US Department of Justice regards markets with HHI below 1000 as not concentrated and between 1000 and 1800 as moderately concentrated. So, on this conventional criterion, these four mergers would not appear to present a major effect on concentration. The German anti-trust law, GWB defines market control when a single company has a market share of at least one-third, or a group of not more than three companies has a market share of at least 50%. A merged VEBA/VIAG would have a share in electricity generation of over 20% and of about 25% in generation capacity, while a merged VEBA/VIAG and RWE/VEW would have together a capacity share of approximately 50% even after divestiture of VEAG. Given the GWB guidelines and wholesale market prices that were close to marginal cost during summer 1999 it is not surprising that these mergers, with some marginal divestiture, met with institutional approval. However, recent experience of other wholesale electricity markets in Europe such as England and Wales and Spain, as well as in North America, show that measures such as HHI do not capture the true potential for generating firms to exercise market power. For this reason we have sought to analyse the impact of the recent mergers in the German electricity industry by carrying out a full industry simulation of the strategic bidding behaviour of firms in the short-term bilateral wholesale market.

3. Modelling strategic behaviour The fact that German wholesale electricity prices fell to industry marginal cost in summer of 1999 suggests that generating firms were, at least initially, competing in a manner which is consistent with theoretical economic model of perfect competition. In this case, the dominant competitive strategy for all firms is to attempt to maximise market share by undercutting their rivals with the result that market prices fall to industry marginal cost. In contrast, where an oligopoly of firms exists, as was the case in the German ESI from the beginning of 2001, each firm has to consider the impact of its actions on the competitive behaviour of the other firms in the industry. The increased concentration of the generation sector since market liberalisation therefore has the potential to significantly impact the way that firms compete. In other words, the actions of firms are likely to become more interdependent due to the

mergers that have taken place and some or all of the newly merged firms will have an incentive to compete strategically. If they were successful in finding a mechanism to implicitly or explicitly coordinate their actions then market prices would be expected to rise to a level above industry marginal cost. The extent to which the industry is able to achieve this outcome in future will be determined by the degree to which firms can exercise market power. 3.1. Agent-based simulation philosophy Traditional equilibrium economic models tend to look at an industry in aggregate, from the ‘top down’. In contrast, the evolutionary economic approach seeks to understand the behaviour of economic systems from a ‘bottom up’ perspective. Though this brings advantages in terms of realism, the inclusion of learning effects, and the potential for heterogeneity between firms, it also brings with it all the complexity and analytical intractability of a real economic system. The agent-based simulation (ABS)3 methodology addresses these difficulties by modelling explicitly each of the individual economic agents, and their interactions with each other, using a discrete event simulation platform. In particular the ABS approach analyses an economic system by modelling: (i) the actions and interactions of individual economic agents through time, not as a one-shot static equilibrium; (ii) agents as continually adopting and adapting new strategies in response to their changing environment, as well as the actions of other agents in the system; (iii) agents as highly boundedly rational, with a very limited capacity to sense and process information from their environment; (iv) system complexity, at the macro-level, emerging endogenously through the repeated interaction of agents, using ‘rules of thumb’ (routines), rather than imposing exogenous complexity during the model building process. At its simplest, an ABS model consists of an economic environment (market), populated by agents (firms) that are each individually represented in the model with their own unique stock of assets, skills, knowledge, and even behavioural routines. Like real firms, the agents are typically capable of gathering (sensing) and storing (remembering) limited amounts of information from their environment as the simulation progresses, to which they apply simple logical rules (routines) in order to 3

ABS has also been applied in many biological, chemical, physical and social situations where understanding the underlying complexity or emergent behaviour of the system is the key objective.

J. Bower et al. / Energy Policy 29 (2001) 987–1005

come to decisions about pricing, inputs, outputs, investment, etc. The role of the modeller/analyst is limited to assigning initial resources, and capabilities to the agents and determining what objectives they will seek to achieve such as profit, turnover, market share, etc. Put simply, an agent is free to adopt any strategy, within the constraints imposed by the environment and the actions of other agents, providing that strategy helps the agent achieve its objectives. There are some precedents for using ABS approaches in the study of electricity markets, most notably CurzonPrice (1997), who used a genetic algorithm to study the strategic bidding behaviour of a generating duopoly, and H.am.al.ainen (1996) who modelled the individual behaviour of electricity consumers in demand-side management schemes. Day (1999) used the supply function equilibrium approach of Klemperer and Meyer (1989) with each agent making the conjecture that all other agents would compete as they did in the previous period. When played through time, this ‘best response’ model allowed agents to learn highly complex strategic behaviour strikingly similar to that in the England and Wales Pool. It revealed a high degree of tacit collusion and market power in the industry. In the next section an ABS model is presented based on work by Bower and Bunn (2000) that analyses the impact of recently proposed changes to the electricity trading arrangement in the Pool of England and Wales. The model is used to measure the impact of industry concentration on market power and strategic behaviour in the German ESI. The model contains a set of trading arrangements, a set of agents, and a demand schedule. The trading arrangements consist of a daily repeated auction with hourly bidding, representing short-term bilateral trading in Germany. Each agent represents a single competing firm, endowed with a portfolio of generating plants, characterised by capacity, fuel type, efficiency, availability, and marginal costs. In contrast to the supply side of the market, all the agents on the demand side are assumed to be price takers with no ability to influence the market through strategic behaviour. For simplicity, these are modelled as an aggregate demand curve. The behaviour of individual agents, in the market milieu, is determined by simple decision rules based on fundamental processes: *

*

*

Mutation}agents can select randomly from an infinite set of potential bidding strategies which they could potentially follow through time; Feedback}agents can observe the results of their bidding strategies; Selection}agents can measure the success of their bidding strategies against both a profit and capacity utilisation benchmark, then repeating any successful strategies; and

*

993

Competition}agents continually respond to the competitive behaviour of other agent.

3.2. Agent structure As the German ESI has over 500 firms involved in electricity generation, operating over 3000 gensets, it has been necessary to simplify the industry structure for the purposes of creating a tractable model. Table 1 summarises the industry structure and agent characteristics based on VDEW (1998) and UCTE (2000) data. The 10 largest generating firms are modelled separately, but with genset capacity of a similar type being aggregated and represented as a single plant. All plants where the generator has a stake of at least 60% have been considered under the control of the majority shareholder. The remaining small local and regional utility firms that operate generation capacity have been aggregated and modelled collectively as if they had merged into 14 regional operating companies. In fact, there have been a significant number of mergers in this segment of the German ESI recently and it seems likely that consolidation will continue. Given this trend, and that these operators own approximately 15% of the industry’s total capacity, thus limiting each agents capacity to approximately 1%, this modelling approximation is unlikely to create a significant distortion in the analysis of market power. There are over 40 (UCTE, 2000) separate crossborder transmission interconnections between Germany and neighbouring countries totalling over 40,000 MW of capacity. The interconnection with France is the most important one. As an aggressive exporter of low cost nuclear power to neighbouring countries, including Germany, EDF is capable of undercutting any fossil fuel power plant and it has been the largest exporter of power to Germany during 1999 with a net export of 13.5 TWh and peak export flow rate of approximately 2500 MW (UCTE, 2000). Overall Germany was a net exporter to other countries during 1999 especially the Netherlands. Though the import of electricity into Germany from neighbouring countries could potentially constrain the exercise of market power by incumbent generators, in practice it is unlikely to play a significant role, at least in the short term, because physical transmission constraints within Germany already curtailed planned imports into Germany during 1999. Even though imports only represent 5% of total demand, vertical integration of generation and transmission means that incumbent firms have no incentive to make the necessary investments to relieve transmission constraints. In addition to physical transmission constraints, significant political difficulties still exist over the implementation of the EU Electricity Directive that has led to a failure to agree an EU-wide system

994

J. Bower et al. / Energy Policy 29 (2001) 987–1005

Table 1 Summary of agent structure in German ESI model Plant number

Plant owner

Plant type

Operational capacity (MW)

Starting bid (Euro/MWh)

Unplanned outage (%)

Planned outage

Available capacity (MW)

Marginal cost (Euro/MWh)

Bayernwerk Bayernwerk Bayernwerk Bayernwerk Bayernwerk Bayernwerk Bayernwerk Bayernwerk Bayernwerk Bayernwerk

Run-of-river Nuclear Lignite old Hard coal S Natural gas CCGT Natural gas Conv. Pumped storage Natural gas Peak Fuel oil L Fuel oil Peak

920 3600 880 1250 110 1950 160 90 1225 340

1.00 5.00 12.00 13.00 14.00 18.00 18.00 25.00 30.00 40.00

3 3 15 5 8 5 3 5 5 5

Winter Summer Summer Summer Summer Summer Summer Summer Summer Summer

892 3492 748 1188 101 1853 155 86 1164 323

0.00 0.00 12.00 13.00 14.00 18.00 16.00 25.00 30.00 40.00

2

BEWAG BEWAG BEWAG BEWAG BEWAG BEWAG BEWAG

Hard coal L Lignite old Natural gas CCGT Natural gas Conv. Natural gas Peak Fuel oil L Fuel oil Peak

1360 570 280 150 70 300 600

11.00 12.00 14.00 18.00 25.00 30.00 40.00

10 15 8 5 5 5 5

Summer Summer Summer Summer Summer Summer Summer

1224 485 258 143 67 285 570

11.00 12.00 14.00 18.00 25.00 30.00 40.00

3

EdF

Interconnector

4000

5.00

1

Summer

3960

0.00

4

EnBW EnBW EnBW EnBW EnBW

Run-of-river Nuclear Hard coal L Fuel oil L Fuel oil Peak

210 2620 2340 280 140

1.00 5.00 11.00 30.00 40.00

3 3 10 5 5

Winter Summer Summer Summer Summer

204 2541 2106 266 133

0.00 0.00 11.00 30.00 40.00

5

HEW HEW HEW HEW HEW

Run-of-river Nuclear Hard coal S Natural gas Conv. Fuel oil Peak

100 800 940 1030 670

1.00 5.00 13.00 18.00 40.00

3 3 5 5 5

Winter Summer Summer Summer Summer

97 776 893 979 637

0.00 0.00 13.00 18.00 40.00

6

Local Utility A

Fuel oil L

1000

30.00

5

Summer

950

30.00

7

Local Utility B

Fuel oil Peak

1500

40.00

5

Summer

1425

40.00

8

Local Utility C

Hard coal L

1200

11.00

10

Summer

1080

11.00

9

Local Utility D

Hard coal S

1500

13.00

5

Summer

1425

13.00

10

Local Utility E

Hard coal S

1500

13.00

5

Summer

1425

13.00

11

Local Utility F

Natural gas CCGT

500

14.00

8

Summer

460

14.00

12

Local Utility G

Natural gas Conv.

1200

18.00

5

Summer

1140

18.00

13

Local Utility H

Natural gas Conv.

1200

18.00

5

Summer

1140

18.00

14

Local Utility I

Natural gas Conv.

1200

18.00

5

Summer

1140

18.00

15

Local Utility J

Natural gas Peak

1500

25.00

5

Summer

1425

25.00

16

Local Utility K

Run-of0river

500

1.00

3

Winter

485

0.00

17

Local Utility L

Pumped storage

1200

18.00

3

Summer

1164

16.88

18

Local Utility M

Pumped storage

1200

18.00

3

Summer

1164

16.88

19

Local Utility N

various

1500

15.00

5

Summer

1425

15.00

1

995

J. Bower et al. / Energy Policy 29 (2001) 987–1005 Table 1 (continued) Plant number

20

21

22

23

24

25

26

Plant owner

Plant type

Operational capacity (MW)

Starting bid (Euro/MWh)

Unplanned outage (%)

Planned outage

Available capacity (MW)

Marginal cost (Euro/MWh)

NWS NWS NWS NWS NWS NWS

Run-of-river Nuclear Hard coal L Natural gas Conv. Natural gas Peak Fuel oil peak

100 2170 1150 230 110 470

1.00 5.00 11.00 18.00 25.00 40.00

3 3 10 5 5 5

Winter Summer Summer Summer Summer Summer

97 2105 1035 219 105 447

0.00 0.00 11.00 18.00 25.00 40.00

PreussenElektra PreussenElektra PreussenElektra PreussenElektra PreussenElektra PreussenElektra PreussenElektra PreussenElektra PreussenElektra PreussenElektra PreussenElektra RWE

Nuclear Hard coal L Hard coal L Lignite old Hard coal S Natural gas CCGT Natural gas Conv. Pumped storage Natural gas peak Fuel oil L Fuel oil Peak Run-of-river

3350 3000 2860 760 1230 250 1810 870 120 550 180 140

5.00 11.00 11.00 12.00 13.00 14.00 18.00 18.00 25.00 30.00 40.00 1.00

3 10 10 15 5 8 5 3 5 5 5 3

Summer Summer Summer Summer Summer Summer Summer Summer Summer Summer Summer Winter

3250 2700 2574 646 1169 230 1720 844 114 523 171 136

0.00 11.00 11.00 12.00 13.00 14.00 18.00 16.88 25.00 30.00 40.00 0.00

RWE RWE RWE RWE RWE RWE RWE RWE RWE RWE RWE RWE

Nuclear Nuclear Lignite new Hard coal L Lignite old Lignite old Lignite old Hard coal S Natural gas CCGT Natural gas Conv. Pumped storage Fuel oil Peak

2500 2620 130 1050 3000 3000 3890 190 280 1200 150 120

5.00 5.00 10.00 11.00 12.00 12.00 12.00 13.00 14.00 18.00 18.00 40.00

3 3 5 10 15 15 15 5 8 5 3 5

Summer Summer Summer Summer Summer Summer Summer Summer Summer Summer Summer Summer

2425 2541 124 945 2550 2550 3307 181 258 1140 146 114

0.00 0.00 10.00 11.00 12.00 12.00 12.00 13.00 14.00 18.00 16.88 40.00

Shared Nuclear

Nuclear

2400

5.00

3

Summer

2328

0.00

STEAG STEAG STEAG STEAG

Hard coal L Hard coal L Hard coal S Fuel oil peak

2000 1680 820 110

11.00 11.00 13.00 40.00

10 10 5 5

Summer Summer Summer Summer

1800 1512 779 105

11.00 11.00 13.00 40.00

VEAG VEAG VEAG VEAG VEAG VEAG VEAG

Run-of-river Lignite new Lignite new Hard coal L Lignite old Pumped storage Fuel oil peak

610 3000 2610 550 1130 1170 420

0.00 10.00 10.00 11.00 12.00 18.00 40.00

3 5 5 10 15 3 5

Winter Summer Summer Summer Summer Summer Summer

592 2850 2480 495 961 1135 399

0.00 10.00 10.00 11.00 12.00 16.00 40.00

VEW VEW VEW VEW VEW VEW

Nuclear Hard coal L Hard coal S Natural gas conv. Natural gas conv. Fuel oil peak

1310 1710 460 1570 500 200

5.00 11.00 13.00 18.00 18.00 40.00

3 10 5 5 5 5

Summer Summer Summer Summer Summer Summer

1271 1539 437 1492 475 190

0.00 11.00 13.00 18.00 18.00 40.00

103415

of tariffs (Financial Times, 2000d) on existing cross-border transmission lines. Furthermore, environmental pressures will continue to make the authorisation of new merchant transmission lines a

96668

lengthy process. Finally, even if cross-border import capacity were to be made available to importers, the German government is unlikely to authorise a significant increase in imports from France unless

996

J. Bower et al. / Energy Policy 29 (2001) 987–1005

Fig. 5. Summary marginal cost supply curve for German ESI model.

German firms are allowed to compete with EDF in the French market (Financial Times, 2000c). For simplicity, we have therefore assumed that the maximum potential import capacity is limited to 6000 MW operated and controlled exclusively by EDF that is approximately 225% above the peak flow reported by UCTE during 1999. Demand created by export flows to other countries, and that served by industrial auto-producers is assumed to stay at current levels and excluded from the model by taking only internal demand on the public transmission system into account. Gross geneset capacity is reduced by planned outages for maintenance, and unplanned outages due to plant failures, estimated from data gathered from a variety of industry sources. Outages on transmission interconnectors are assumed to be negligible at 1% that reflects the multiplicity of transmission routes and inherently low failure rate on transmission lines. The marginal cost supply curve for the industry is shown in Fig. 5 estimated from German spot market prices for fossil fuels at the end of 1999 and plant heat rates based on comparisons with similar plants in USA and other European countries. Nuclear plant is assumed to have a marginal cost of zero, to reflect the inflexible nature of its operation, as has run-of river hydro plant. Fig. 6 shows the evolution of on-peak and off-peak marginal cost of production throughout the year taking into account plant outages. This shows that, soon after liberalisation took place, prices briefly fell below marginal production costs.

3.3. Agent bidding A key feature of the ABS approach is that it avoids the necessity to make the usual restrictive assumptions and simplifications, which are required by traditional economic analysis of imperfect competition. Instead, the agents use simple internal decision rules, summarised in Table 2, that allow them to ‘discover’ and ‘learn’ strategic solutions which satisfy their profit and market share objectives over time. Taken together, these rules constitute what is essentially a naive reinforcement learning algorithm4 that seeks out and exploits successful bidding strategies while discarding unsuccessful bidding strategies. As a result, the behaviour of the simulated market is almost entirely emergent as it is created endogenously by the aggregate interaction between agents and their environment. A general empirical criticism levelled at classical and game theoretic models is that they assume economic actors not only assimilate information perfectly, but that they also have the power to observe it perfectly. This hyper-reality seems unrealistic and the ABS model developed here assumes, instead, that agents know everything about their own portfolio of plants, bids, output levels, and profits, but nothing about other agents or the state of the market. Their ability to capture and retain data is very limited, they have no powers of strategic reasoning, and hence they exhibit a high degree 4

See Sutton and Barto (1998) for a definition and fuller discussion of the many different forms of reinforcement learning which have been developed.

J. Bower et al. / Energy Policy 29 (2001) 987–1005

997

Fig. 6. Seasonal marginal cost versus actual CEPI prices during 1999–2000.

of bounded rationality. Moreover, the agents’ bidding strategies are not specified exogenously by the modeller but are developed by the agents themselves. The model also has the advantage of allowing bidding strategies to be observed for asymmetric bidders, right down to the individual plant level. This reflects the ‘bottom-up’ approach of ABS, focussing on individual strategic decision making behaviour, rather than the top-down aggregate market behaviour. Strategic learning is driven by each agent striving to satisfy two objectives: (i) continuously increasing its own overall profitability, from one period to the next; and (ii) reaching a target utilisation rate on its plant portfolio in every period. To reach these objectives, agents may follow either a ‘price raising’ strategy, by adding a random percentage to the bid(s) they submitted in the previous trading day or a ‘price lowering’ strategy, by subtracting a random percentage.5 The agents may raise or lower bid prices to any level, between zero and 1000 Euro/MWh, but plants with high marginal costs of production must always bid higher prices than plants, in the same portfolio, with lower costs of production. It is assumed that forward contracting reflects each generator’s desire to guarantee itself a minimum level of market share, or output, in a 5 In all the simulations discussed here, agents draw their random percentage values from a uniform distribution with a range  10% and a mean of 0%. Other distributions have been tested with little apparent effect.

given period. For each agent, a minimum target rate of utilisation is exogenously assigned for its plant portfolio, which is expressed as a percentage of expected total available MW of capacity. From the point of view of the simulation, if an agent failed to reach its target utilisation rate on the previous trading day, then it lowers the bid price(s) on all of its plants for the current trading day. Though this disregards the impact on profitability, and the success of previous strategies, the target utilisation rate is attached to an agent’s portfolio, not to particular plant(s), so they can still explore a wide range of bidding strategies that satisfy their profit and utilisation objectives. Finally, an agent can transfer a successful bidding strategy from one of its plants, to all other plants in its portfolio. Agents with portfolios containing a large number of plants benefit from this learning effect as they naturally have more opportunities to experiment with, identify and adopt successful bidding strategies than a single plant operator. This is achieved by allowing agents to automatically raise the bid price on any plant to the level of the next highest bid price submitted, if it sold its output for less the marginal sales price achieved in the portfolio on the previous trading day. In practice, each agent is continuously updating its profit objective, as the simulation progresses, always using the previous trading day’s profit as a benchmark against which it compares the current day’s profits. By updating continuously their profit objective at the end of each trading period, agents are forced to compete continuously against each other. As in the real world, not all the agents can increase their profits indefinitely and, at some point, a profit increase by one agent will cause a profit

998

J. Bower et al. / Energy Policy 29 (2001) 987–1005

Table 2 Summary of agent bidding rules and objectives Rule 1. Self awareness Agents receive feedback data from their own trading activities for the previous two trading days: (i) plant avoidable costs of production; (ii) plant bid prices; (iii) plant sales prices; (iv) plant and total portfolio expected available capacity; (v) plant and total portfolio sales volume; (vi) plant and portfolio rate of utilisation; (vii) plant and portfolio profit; (viii) portfolio target utilisation; (ix) portfolio profit. Rule 2: Information restrictions Agents do not know the past, current, or future, actions of other agents or the state of the market. Rule 3: Objective functions Agents have common objectives for each new trading day which are to achieve: (i) at least their target rate of utilisation for their own plant portfolio; and (ii) a higher profit on their own plant portfolio, than for the previous trading day. Rule 4: Strategy selection Agents submit bid price(s) for each plant in their portfolio, at the beginning of the current trading day, using decision criteria in the following order of precedence: (i) if the target rate of utilisation was not reached across the portfolio, on the previous trading day, then randomly subtract a percentage from the previous day’s bid price for each plant in the portfolio; (ii) if any plant sold output for a lower price than other plants across the portfolio, on previous trading day, then raise the bid price of that plant to the next highest bid price submitted; (iii) if total profit did not increase across the portfolio, on the previous trading day, then randomly add or subtract a percentage from the previous day’s bid price for each plant in the portfolio; and. (iv) if profit and utilisation objectives were achieved across the portfolio, on the previous trading day, then repeat the previous trading day’s decision Rule 5. Strategy restrictions Agents can follow any strategy on condition that the bid prices in their plant portfolio are always: (i) no less than Euro 0.00; (ii) no more than Euro 1000.00; (iii) rounded to two decimal places; and (iv) higher for high marginal production cost plant than for low marginal production cost plant in the portfolio.

decrease for another agent. When an agent suffers a profit decrease it is prompted to abandon its current bidding strategy and randomly look for a more successful one. When it eventually finds a better strategy, which might mean taking profit from another agent, this would trigger a new strategy search by the affected agent, and so on. 3.4. Aggregate demand The demand side is modelled as an aggregate demand curve rather than as individual agents. At the time of writing, daily demand data was not yet available for 1999, however, 1998 demand data for the third Wednesday of each month was obtained from the UCTE. This was used to create a stylised time series of working day load profiles which excludes loads served by industrial and railway autoproduction but includes demand created by pump storage facilities. Demand on each

working day, in any given month, is therefore assumed to be identical to all the other working days in that month and corresponds to actual demand on the third Wednesday of that month in 1998 as shown in Fig. 7. Weekends and holidays are excluded from the data and the year is assumed to contain 12 standard months of 21 days (i.e. 248 working days). There is insufficient data available yet to allow an accurate estimate of the price elasticity of demand in Germany as the market has only been operating for a relatively short time. However, as an industrialised Western economy it can be assumed that the load response in Germany will be relatively insignificant, especially at low prices, we therefore assume that demand falls by 50 MW per unit rise in prices above 125 Euro/MWh. No demand response is assumed to occur below 125 Euro/MWh, (and indeed despite the substantial drop in prices during 1999, no increase in overall demand was observed).

J. Bower et al. / Energy Policy 29 (2001) 987–1005

999

Fig. 7. Stylised annual weekday demand profile for Germany.

3.5. Plant despatch The bilateral market that has developed in Germany since April 1998 can be characterised by two different features, a long-term contract market and an informal short term wholesale market, which is reflected by indices such as CEPI and SWEP. With the evolving spot markets and exchanges and a growing number of trading companies it can be assumed that the latter will gain importance very quickly as it did in the markets of Scandinavia and Netherlands. Therefore, the German system has been modelled by assuming that agents trade in a short-term bilateral market submitting 24 separate bid prices for each of their plants, one bid for each hour of the coming day (i.e. an agent with ten plants submits 240 separate bids). The working day on which plant is to be despatched is assumed to run from the beginning of the hour ending at 5.00 a.m. when demand is generally at its lowest point in the day. All agents are assumed to submit their bids to the market simultaneously, and on a day-ahead basis (i.e. trading is assumed to be completed on the day before actual despatch takes place). Once submitted, bids cannot be changed and it is assumed that agents will offer all of their operationally available capacity to the market. The market is cleared separately for each hour of the working day by stacking plants in strict bid price merit order, lowest to highest. Plants are despatched centrally by allocating demand up to each plant in turn either up to the total available capacity of the plant, or residual unallocated demand, whichever is lowest. Agents are paid on a Pay Bid basis by multiplying the demand allocated to each of their plants by the bid price of that plant less its marginal production cost.

The model does not take into account any aspect of the ownership structure of the transmission grids in Germany, there are no physical transmission constraints, and the cost of transmission is zero. The plant is despatched purely on the basis of hourly bid prices, there are no plant operational constraints in the model, and it is assumed that plant may be ramped up or down within an hour at zero cost.

4. Results The ABS model described above has been used to run multiple simulations, each corresponding to the different industry structure under the different merger proposals and also under a range of plant closure scenarios. The results are summarised in Table 3. 4.1. Simulating strategic bidding Market prices initally seen in the German market indicated that firms were willing to bid at or even below marginal cost to gain or retain market share. Advertising, marketing campaigns, and press announcements by firms themselves suggested that most of the industry was following a market share dominance strategy, and this is what brought about the significant price cuts. Generating firms in the German ESI were therefore initially behaving as if they had no market power and the market as a whole was operating in a manner that is close to the classical economic model of perfect competition with prices falling to the short-term marginal costs of production. If this is the case, each firm can assume that its competitive actions have no influence on the

1000

J. Bower et al. / Energy Policy 29 (2001) 987–1005

Table 3 Summary simulation results Scenario

Marginal cost Base case 1999/00 Two merge Four merge Two merge+oil close Two merge+nuclear close

Mean annual price

Difference versus marginal cost

Off peak Euro/MWh

On peak Euro/MWh

Off peak Euro/MWh

On peak Euro/MWh

Off peak

On peak

11.59 15.12 15.54 21.01 19.42 22.46

13.39 19.79 22.65 31.47 30.89 36.88

} 3.53 3.95 9.41 7.83 10.87

} 6.40 9.27 18.09 17.51 23.50

} 30.44 34.06 81.19 67.54 93.75

} 47.83 69.22 135.12 130.80 175.54

Fig. 8. Comparison of simulated prices with different plant utilisation rate objectives.

actions of the other players. In other words, no firm can adopt a bidding strategy which will allow it to raise prices significantly otherwise other firms will undercut it with spare unused capacity. To test the actual and potential levels of market power that could be exercised in the German wholesale electricity market the strategic bidding behaviour of German generators has been simulated using the ABS model set-up described above. In the first simulation, all firms are assumed to only be interested in gaining market share regardless of price, and hence have a constant 100% plant utilisation objective. In other words, no generator has any interest in bidding strategically to raise prices. In the second simulation, firms are assumed to be willing to restrain their competitive to the extent that they lower their utilisation objective to 60%, with the exception of the municipalities that are too small to bid strategically, thus keeping their utilisation rate at 100%. In essence, this second model assumes that the industry is able to coordinate upon, and enforce, a tacitly collusive agreement where

firms give up a limited amount of market share in return for higher prices and profits. The results as presented in Fig. 8 show that the more players ignore the strategic impact of their behaviour, by attempting to sell more of their output, prices remain at or even below marginal cost. Where firms are willing to curtail their utilisation objectives, and hence market share targets, then marginal prices rise. There is no economic theory to show why real firms in the German ESI should coordinate upon any particular utilisation rate. However, as Fig. 9 shows, when the simulated on-peak and off-peak prices are superimposed onto the actual CEPI price data there is a high degree of correlation in both the level and seasonal pattern of actual and simulated prices for 1999. Comparing simulated and actual price data in this way has therefore allowed us to calibrate the model to replicate the otherwise unobservable strategic objectives of generating firms in the German market. Using this model result, containing the 1999 industry structure and a 60% target utilisation rate, as a base case therefore allows us to

J. Bower et al. / Energy Policy 29 (2001) 987–1005

1001

Fig. 9. Comparison of simulated and actual prices (CEPI) for Germany.

Fig. 10. Comparison of simulated market prices before and after two mergers.

compare results from simulations with alternative industry structures. 4.2. Simulating industry mergers To test the impact of the industry consolidation, as described in Section 2, market prices have been simulated by assuming that some or all of the proposed mergers were allowed to go ahead. The simulated prices resulting from either two or all four described mergers

are presented in Figs. 10 and 11, respectively, and compared with the prices from the base case simulation. A 60% target utilisation rate is assumed to apply in all cases excluding the municipalities. The result is that mean on-peak annual market price rises by approximately 21% and off-peak prices rise by approximately 4% for the VEBA/VIAG and RWE/VEW mergers alone. For all four mergers, mean on-peak annual market prices rise by approximately 87% and off-peak prices rise by approximately 50%. However, it is

1002

J. Bower et al. / Energy Policy 29 (2001) 987–1005

Fig. 11. Comparison of simulated prices for base case and after four mergers.

Fig. 12. Comparison of simulated market prices after marginal oil plant closures.

interesting to note that even after four mergers have taken place, simulated prices are still below those seen for the England and Wales Pool reported in Fig. 1. 4.3. Simulating plant closures A strategic consequence of consolidation through mergers and acquisitions would normally be a process of asset rationalisation and an emphasis upon delivering shareholder value. In this section, therefore, the impact

of closing some of the less economic plant across the industry is considered. Assuming that all four of the mergers discussed above are allowed to go ahead and that four dominant strategic players emerge in the industry it seems likely that they will be at the forefront of any efforts to curtail excess capacity. The simulation results in Fig. 12 compare market prices in the base case versus those after the VEBA/VIAG and RWE/VEW merger and then close their oil plant. This plant closure is assumed to only involve peaking oil plants or about

J. Bower et al. / Energy Policy 29 (2001) 987–1005

1003

Fig. 13. Comparison of simulated market prices pre- and post-nuclear capacity closure.

6% of total system capacity but seems to have a disproportionate effect with a 83% increase in mean annual on-peak price and 37% rise in mean annual offpeak price. In light of the decision made by Sweden to begin closing its nuclear plants, and the change in government in Germany in 1998 with the Green Party forming a coalition with the Social Democrats, nuclear power has been the subject of much political debate in Germany. At the time of writing, no firm decision had yet been made but the impact of closing nuclear plants has been modelled with the results shown in Fig. 13. Even under the limited assumption that only 25% of the nuclear capacity will eventually be closed, and that firms keep all their other existing plants open, a 128% increase in mean annual on-peak price and 63% rise in mean annual off-peak price is seen.

5. Discussion Taken together, the results suggest that firms in the German ESI were capable of exercising a limited amount of market power before further horizontal integration, because they have been able to keep prices above marginal cost in the winter months of 1999. To achieve this, the simulations show that firms must have been bidding strategically and been willing to curtail their capacity utilisation, and hence market share objectives, because otherwise prices would have remained at marginal cost throughout the year. By calibrating the ABS model to market prices achieved in 1999 we have been able to abstract a target utilisation

rate of 60% for the agents. This represents a measure of the extent that firms are willing to curtail their overall market share objective and cooperate with each other to maintain market prices above marginal cost. We also show that if the agents were able to more effectively coordinate their actions, by reducing their minimum target utilisation rate, they would be able raise prices even further. Clearly, it is impossible to know what utilisation rates targets firms are really pursuing in the German ESI, and it even less clear whether or not they will be able to achieve higher prices by showing more restraint in future. However, what is clear is that if firms are allowed to merge with their competitors it will be far easier to achieve a tacitly collusive outcome, with greater price discipline, with two or four large generators rather than four or eight, respectively. The estimations and assumptions made in this simulation regarding transmission constraints, plant despatch and cost data, are limiting the quantitative validity of the results, so it is impossible to derive predications from them. However, the approximations and simplifications inherent in the model remain constant in all the simulation scenarios, making it at least possible to gain comparative insights. When the simulations were carried out, with an increasingly concentrated industry structures, market clearing prices increased significantly, even without any reduction in target utilisation rates by the merged firms. Given that the agents in the ABS model rely on pure learning process to coordinate their actions, this result suggests that the agents find the task of reaching a collusive outcome far easier than under a fragmented industry structure. After the VEBA/VIAG and

1004

J. Bower et al. / Energy Policy 29 (2001) 987–1005

RWE/VEW mergers, firms seem to be able to exercise market power to some degree, especially in winter months, and this increases even further if all four mergers are considered. However, these mergers alone do not appear to be the main threat to prices, since the number of firms competing in each segment of the market will still remain sufficiently high to ensure that no single firm can gain a dominant share. Most importantly, no firm will be in a position to act as residual monopolist even after mergers have taken place. In fact, the simulation results show that firms must accompany any merger with restraint in terms of their overall market share. If firms continue to aim for 100% utilisation of their plant, regardless of the strategic impact on other firms, then prices will remain low. The real impact on market power and market prices comes when combined with the additional strategic opportunities it opens up for the merged firms. By also cooperating in the coordinated closure of capacity, firms would not only raise utilisation rate on remaining plant but also increase returns on capital, thereby reducing both the means and the incentive for firms to fight for market share. The simulation results show that once they have merged, firms could engage in strategic withdrawal of marginal capacity and reap significantly greater prices for the same level of output. With high levels of excess capacity and low prices, it is difficult to see how the German Cartel Office could argue that old inefficient plant should be forced to stay open on economic grounds. This is really the crucial issue that should guide thinking upon acceptable levels of consolidation. Once merged, firms will be free to make other strategic choices that will be much harder to regulate and which, when combined with the latent impact of an increased industry concentration, may result in unforeseen levels of market power. In this respect, firms may change their current bidding strategies, reduce their market share aspirations, or close plant that is currently operating only at peak loads. A similar result can be observed in the case of nuclear capacity closure with the side effect of an increase in output of older fossil fuel plant and an obvious potential impact of the environment. The simulation results show clearly the large impact, plant closures can have on market prices and as well founded justifications for reducing capacity should be easy to find at current, this might represent a major problem for the regulator. Moreover, recent developments in the Netherlands show that even small amounts of import transmission constraint, which could be regarded as mere capacity closures, can have an enormous effect on market prices. In the England and Wales market, strategic capacity withdrawal, especially of marginal plant, has been a major regulatory problem and Ofgem has over the years launched a number of investigations

into this kind of behaviour by the largest fossil fuel generators PowerGen and National Power (Offer, 1998). More recently, the regulator ordered that any firm wishing to close plant had to prove that it was uneconomic to operate them at current market prices which, in effect, would force generators to put spare capacity up for sale to competitors before allowing them to close it down. In conclusion, this paper identifies the value of undertaking a thorough strategic simulation of the implications of incremental consolidation of the industry. Although the question of how concentrated the industry should be allowed to become through mergers remains an unresolved one, it is clear that mergers will allow prices to rise, especially as prices remained below a sustainable level during 1999. Furthermore, given that the German government, in common with many other countries, wishes to promote the interests of German firms in international markets, and with a relatively benign outlook for prices, the internal pressure to curtail mergers for competitive reasons may, therefore, be outweighed by political considerations. Thus, the German minister for economic affairs was reported to prefer the mergers to go ahead despite objections from the Cartel Office (cf. Die Welt, 2000). Furthermore, it seems inevitable, that once started, industry consolidation will be followed by asset rationalisation, and the consequences of this second stage in the strategic evolution of the industry could effectively double annual average prices. The dramatic price reductions of 1999 could, in retrospect, appear to have been short-lived, and if, indeed, they created the circumstances for a benign endorsement of industry consolidation, a longer period of high prices could follow. As other countries have found, this would be much harder to reverse.

References Atkins, R., 1999. RWE set to spend £ 20bn on energy groups. Financial Times 23 September 1999. Atkins, R., Taylor, A., 1999. Germany shows the way for Europe. Financial Times 21 October 1999. Bower, J., Bunn, D.W., 2000. A model-based comparison of pool and bilateral market mechanisms for electricity trading. Energy Journal, 21(3), 1–29. Curzon-Price, T., 1997. Using co-evolutionary programming to simulate strategic behaviour in markets. Journal of Evolutionary Economics 7(3) 219–254. http://link.springer.de/link/service/journals/00191/tocs/t7007003.htm. Day, C., 1999. A computational approach to modelling strategic behaviour in electric power pools. Ph.D Thesis, London Business School, Sussex Place Regent’s Park, London NW1 4SA, United Kingdom. Die Welt, 1999. Stuttgarts Neckarwerke-Paket geht an EnBW. Die Welt 18 December 1999. . Die Welt, 2000. Mullers Energiepolitik schaft kritisiert. Die Welt 26 January 2000. DJ/VIK Index, 2000. Dow Jones/VIK Strompreisindex. www.vik-online.de, 15 January 2000.

J. Bower et al. / Energy Policy 29 (2001) 987–1005 Drasdo, P., et al., 1998. Konzentration und Wettbewerb in der . deutschen Energiewirt-schaft. Munchen, Oldenbourg Verlag. European Parliament and the Council of the European Union, 1997. Directive Concerning Common Rules for the Internal Market in Electricity. Official Journal of the European Communities 30 January, No. L 27/20 [EN]. http:www.europa.eu.int/en/comm/ dg17/gazel en.htm. Financial Times, 2000a. Germany’s two utility mergers in jeopardy. London, April 15. http://www.ft.com. Financial Times, 2000b. Brussels to probe French bid for German utility. London, October 3. http://www.ft.com. Financial Times, 2000c. Power struggle. London, October, 20. http:// www.ft.com. Financial Times, 2000d. Free power flow setback for Brussels. London, November 22. http://www.ft.com. Financial Times, 2000e. Vattenfall ready to take helm at Veag. London, December 12. http://www.ft.com. H.am.al.ainen, R.P., 1996. Agent-based modelling of the electricity distribution system. Proceedings of the IASTED International Conference Modelling, Identification and Control, Innsbruck, Austria, February 19–21, 1996, pp. 344–346. Harnischer, U., Hargreaves, D., 2000. German power market faces more competition. Financial Times, 7 May 2000. Klemperer, P.D., Meyer, M.A., 1989. Supply function equilibria in oligopoly under uncertainty. Econometrica. 57, 1243–1277. . Knodler, T., 1999. Fast jeder zweite Job f.allt bei der BEWAG weg. Die Welt 8 July 1999. Office of Electricity Regulation, 1998. Independent assessor’s report on plant closure. April 1998. Office of Gas and Electricity Markets,

1005

Stockley House, 130 Wilton Road, London SW1 V 1LQ. http:// www.open.gov.uk/offer/pubsmain.htm. Office of Electricity Regulation, 1999a. The new electricity trading arrangements: Volume 1. July 1999. Office of Gas and Electricity Markets, Stockley House, 130 Wilton Road, London SW1V 1LQ. http://www.ofgem.gov.uk/elarch. Office of Electricity Regulation, 1999b. The new electricity trading arrangements: Volume 2. July 1999. Office of Gas and Electricity Markets, Stockley House, 130 Wilton Road, London SW1V 1LQ. http://www.ofgem.gov.uk/elarch. Office of Electricity Regulation, 1999c. The new electricity trading arrangements: a draft specification for the balancing mechanism and imbalance settlement. Version 1.2. July 1999. Office of Gas and Electricity Markets, Stockley House, 130 Wilton Road, London SW1V 1LQ. http://www.ofgem.gov.uk/elarch. Office of Electricity Regulation, 1999d. The new electricity trading arrangements: Ofgem/DTI conclusions. October 1999. Office of Gas and Electricity Markets, Stockley House, 130 Wilton Road, London SW1V 1LQ. http://www.ofgem.gov.uk/elarch. Sutton, R.S., Barto, A.G., 1998. An Introduction to Reinforcement Learning. MIT Press, Cambridge, MA. UCTE, 2000. Statistical Yearbook 1999. http://www.ucte.org/Publikationen/English/Default Pub E.htm. VDEW, 1998. Leistung und Arbeit 1997. Frankfurt am Main, Verlagund Wirtschaftsgesell-schaft der Elektrizit.atswerke. Wetzel, D., 2000. Deutschland braucht die vierte Kraft auf dem StromMarkt. Die Welt, 26 January 2000. Wintermann, J., 1999. RWE geht in Europa auf Einkaufstour. Die Welt, 7 September 1999.