Structure and
Performance of Six
European
Wholesale
Electricity Markets
in 2003, 2004 and
2005
Executive
Summary
Presented to DG
Comp 26th February
2007
Prepared by London
Economics in
association with
Global Energy
Decisions
February 2007
Structure and Performance of Six
European Wholesale Electricity Markets
in 2003, 2004 and 2005
Executive Summary
Presented to DG Comp 26th February 2007
DG Comp
Prepared by London Economics in association
with Global Energy Decisions
February 2007
© Copyright London Economics. No part of this document may be used
or reproduced without London Economics’ express permission in
writing.
Contents Page
1 Executive Summary 1
1.1 Introduction, background, and overview 1
1.2 Overview of methodology and data 7
1.3 Results 15
1.4 Conclusions and summary 31
London Economics Page i
February 2007
Tables & Figures Page
Table 1.1: CR(n) and HHI average across countries 15
Table 1.2: Price Cost Margins 18
Table 1.3: Contribution to Power Price (€/MWh) 20
Table 1.4: Contribution to Fixed Costs for largest companies
(€'000) 25
Table 1.5: Summary of country RSIs for largest companies
(2003-2005) 27
Table 1.6: Regression results relating margins (PCMU) to
market structure (RSI) for largest generators 30
Figure 1.1: Estimated impact of RSI on margins: Spain 29
London Economics
February 2007
Section 1 Executive Summary
1 Executive Summary
This chapter presents a summary of the report, including results, across the
countries studied, namely Belgium (BE), Germany (DE), Spain (ES), France
(FR), the Netherlands (NL), and Great Britain1 (GB). The purpose is to give a
high-level overview of the work carried out and to facilitate comparisons
across countries for certain key elements. The chapter gives an overview of
the study and presents summary key results for traditional market structural
measures, electricity-specific measures of market structure, market outcome
measures, and regression analysis. Significantly more discussion and results
in relation to a number of sensitivity and alternative scenarios are contained
in the detailed country specific chapters of the report.
1.1 Introduction, background, and overview
This report studies a selection of the EU’s major electricity wholesale markets
with the aim to determining how competitive markets were in the three years
up to 2006.
The EU is in the midst of creating a fully competitive internal market for
energy and electricity. There are concerns and doubts as to how much has
been achieved along these lines. In 2005, the European Commission’s DG
Competition launched a sectoral inquiry into the European electricity and gas
markets. The final report of the inquiry was published on 10 January 20072.
Among many findings made by the sector inquiry, the two main findings that
concern this report were:
Market concentration remains very high in a number of geographical
and product markets; primarily national wholesale/generation
markets.
Large energy consumers doubt that prices on spot and forward
wholesale markets result from fair competition.
1 The market comprising England, Scotland and Wales (i.e. the United Kingdom without
Northern Ireland).
2 Further information on the sector inquiry, including its Final Report, are available at:
http://ec.europa.eu/competition/antitrust/others/sector_inquiries/energy/
London Economics Page 1
February 2007
Section 1 Executive Summary
The many findings of the final report, and especially the two findings above,
suggest that competition in wholesale electricity generation is not functioning
properly. While there is little overall informed dispute about the high levels
of concentration in some EU countries (e.g., BE and FR), the reliability of
traditional concentration measures as an indicator of market structure and
market outcomes varies by situation, time, and in response to other factors
such as interconnection. Determining how competitive electricity wholesale
markets are is thus quite complex. Market concentration measures merely
relying on market share may not give a good indication of how competitive
electricity markets are likely to be. This provides the impetus for a more
detailed study into the electricity markets using more innovative and
advanced techniques.
Even more contentious (than the usefulness of market concentration as an
indicator) are other issues, especially whether recent spot and wholesale
prices are reflective of full and fair competition. This may be for a variety of
reasons; but many of the reasons that make it difficult to determine if
electricity prices are fair, flow from the physics of electricity. Electricity
cannot be stored (cheaply); the transmission system must be balanced in real
time; there are a vast variety of generation technologies, capital vintages, fuel
types; trading occurs in many markets including spot, exchange, over-thecounter,
to name just a few. These factors (and others) combine in complex
ways to impact on cost, available capacity, and other factors, which interact in
complex ways. The result is that observation of electricity prices and costs,
without further in-depth study, can give sometimes little insight into how
competitive electricity market outcomes have been.
The EU Commission, and the DG Competition in particular, therefore have
been endeavouring to shed additional light on the above contentions, i.e., that
current levels of concentration are leading, or are likely to lead, to poor
market outcomes and that wholesale prices are not as keen as they could be.
London Economics Page 2
February 2007
Section 1 Executive Summary
In the course of these endeavours, the European Commission DG
Competition (DG Comp) commissioned London Economics (LE) in
association with Global Energy Decisions (GED) and Professors Natalia
Fabra, Jean-Michel Glachant, and Nils-Hendrik M. von der Fehr to carry out
an in depth and detailed study into the structure and functioning of the EU’s
electricity markets. The project was led by LE who formed a consortium
specifically for the project. LE developed and proposed the methodology and
ultimately analysed the results and drew conclusions. GED provided the
software, despatch modelling, data, and a technical project coordinator. GED
also ran the software and provided support on the software, including a
specific applet for price calculation. The associated academic advisors,
internationally recognised experts in EU electricity market competition
economics, were asked to review the proposed methodology and key
assumptions involved in the modelling and analysis, and propose
alternatives if the methods undertaken were not consistent with
recognised best practice in electricity competition economics. These experts
agreed with the methodology and their comments were taken on board. The
whole team, including the DG Comp project team, contributed to many parts
of the project along the way, but the conclusions are ultimately LE's.
The study was to be a largely quantitative and fact-based assessment of
market structure and market performance in selected European electricity
markets and with the ultimate objective of uncovering the details any
potential relationship between the two.
The study is unique in several ways, most significantly it is based on actual
data returned by the generation companies and utilities on output, cost, and
unit characteristics from the generation units. The study involved simulation
of the electricity markets for the six countries and relating the simulations to
market outcomes via econometric (statistical) techniques. While a complex
subject (and many of the detailed results involve quite technical nuances of
statistical and electricity despatch simulation modelling), we have
endeavoured to present the results (especially in this chapter) as accessible as
possible to a more general audience.
The study of this very difficult problem of relating electricity market structure
to market outcomes (and uncovering evidence of either competitive or uncompetitive
market outcomes) requires the use of a detailed methodology.
Our approach was therefore several-fold.
London Economics Page 3
February 2007
Section 1 Executive Summary
First, we investigated the more traditional structural measures of market
concentration (those based on market shares) such as the concentration ratio
and the Hirfindahl-Hirschmann index (HHI). Since market conditions in
electricity markets can change hour-to-hour, season-to-season, year-to-year,
we estimated these measures on an hourly basis over the 2003-2005 sample
period. Forced outages, maintenance events and deratings, among other
issues, are all accounted for in the calculation of the total capacity available to
the market in each hour. Market shares were also calculated based on
generation output. A number of sensitivities were also included to estimate
the impacts of interconnectors, long-term contracts, and capacity reserved for
system balancing commitments.
We next undertook to estimate electricity-specific measures of market
structure, the residual supply index (RSI) and the pivotal supplier index (PSI).
These measures, of how ‘pivotal’ a given supplier is, are designed to deal
more effectively with the changing nature of electricity markets, and are
calculated hourly.
The next element of the analysis was the calculation of market outcome
measures. Lerner Indices (LI) and Price-Cost Mark-Ups (PCMU) were
estimated as standard economic measures of the competitiveness of market
outcomes. To calculate these measures first requires the estimation of the
marginal cost of energy on the electricity grid for each hour of the 2003-05
sample period. To do this we used GED’s Prosym™ 3 market despatch
simulation model, with inputs on unit characteristics, fuel prices, thermal
efficiency, planned and forced outages, available capacity, plus other relevant
factors collected by DG Competition as part of the recent Sector Inquiry from
the generators and transmission system operators (TSOs). This allowed for a
simulation of the market, based on the actual generator data, of the optimal
system despatch and resulting marginal cost.
3 Nearly 200 companies worldwide (including many European generators, traders and TSOs)
use the ProSym engine to do long term electricity price forecasting, conduct resource
planning studies and optimize generation despatch.
London Economics Page 4
February 2007
Section 1 Executive Summary
The final element of our analysis was to relate the calculated market outcome
measures to the electricity-specific market structure measures, the RSIs. This
was done using econometric regression (statistical) analysis. The econometric
approach is both useful and necessary as it allows one to control for many
random factors (weather, temperatures, outages) and more economically
benign factors (scarcity) can combine to raise prices and margins. These
factors may have little to do with market power, and so thus must be
controlled for. Controlling for random and deterministic variables was thus
done via the econometric approach.
It is noteworthy, before going further, to highlight the scale and extent of our
analysis, as well as the degree to which we believe we have advanced the
debate and state of knowledge of the electricity sector in the EU’s major
markets. The study involved data collection and analysis on a scale that, to
our knowledge, is unprecedented in the electricity-economics field globally.
The study involved the hourly observation, simulation, and relating to price
outcomes of output and marginal cost, and market structure for almost every
generation unit in every hour (8,760 hours/year and 8,784 in 2004) for three
years ((8,760x3)+24=26,304) for six countries (x6=157,824). With countries
having between 60 to over 350 units, and over 25 variables either collected or
estimated, the study involved the collection of about 500 million data points
and manipulation and analysis of close to 1 billion data points, and resulting
in an approximate total of 75GB of data inputs and outputs. Further, while
previous studies have looked at market outcome measures in selected
markets in the USA, and also related the RSI to price cost margins, no study
to our knowledge has comprehensively related market structure measures
and market outcome measures in a manner similar to that advanced in this
report. Previous studies either focused on a single country, a limited time
period, or did not account for many factors. Thus, our multiple regression
results can be seen as groundbreaking.
London Economics Page 5
February 2007
Section 1 Executive Summary
In the course of this study and as part of the wider sector inquiry, the DG
Comp and the LE team have undertaken to collect a very large amount of
data from the electricity generators in the EU. Data has been collected on a
great number of electricity market details, including; unit characteristics and
operational details (installed capacity, thermal efficiency, planned and forced
outages, flexibility in coming online), outputs (total generation), costs
(running cost, fuel costs, fuel prices paid, operational costs), as well as
contractual obligations (long-term contracts and system balancing
commitments), in relation to almost all companies and generation units in the
six countries studied. These data were necessary for, above all, the
simulation of the electricity markets. We acknowledge the efforts and
cooperation of the EU generators in the study and the collection of these data
by the DG Comp team throughout the course of the study. We also note that
the utmost confidentiality of the data was maintained through the most
rigorous procedures4.
Finally, the results of our extensive analysis present a detailed assessment of
the structure and outcomes of the EU six electricity markets studied, as well
as an investigation of the structure-outcome relationship. However, our
analysis is still not without limitations. Ultimately, our analysis supports the
two key points of the sector inquiry report; namely, that the current market
structure in the EU electricity markets (the six markets studied) in a
significant number of hours is likely to be conducive to anticompetitive
behaviour. And secondly, that price outcomes on the EU wholesale
electricity markets may have been less keen, than they otherwise would have
been, had the markets been structured more competitively. The general
caveat of the analysis is that uncontrolled for factors may have caused the
appearance of market power, and that the existence of market power is not
necessarily evidence of its abuse. However, our regression results have
controlled for a great number of factors.
We turn now to a summary of the methodology and of the results; summary
conclusions end this chapter.
4 All data from the generators was processed and analysed on site at the DG Competition in a
locked data room on designated laptops without internet or outside access to the room or
data. Company names were anonymised as appropriate during the exercise and for the
report.
London Economics Page 6
February 2007
Section 1 Executive Summary
1.2 Overview of methodology and data
1.2.1 Methodology adopted
Electricity as a commodity has many special features, and thus analysis of
electricity market competition is more complex than other markets. Among
the most important are that it is generally difficult/costly to store. It can be
shipped but only on the HT transmission system (which can be congested);
and supply and demand must be matched on the system within very tight
margins in real time. Electricity market interactions can also involve complex
commercial bidding strategies. Finally, random factors like weather and
outages impact supply and demand. The special features of electricity
markets suggest that standard tools of competition analysis should be used in
concert with measures designed specifically for electricity market analysis.
Further, more advanced techniques such as regression analysis are required
to relate market structure and market outcomes and control for random error.
HHI and CRn
Our analysis first focused on the standard tools of competition economists
and competitions authorities such as Herfindahl-Hirschman Indices (HHI)
and concentration measures (CR(n)). In the context of electricity markets it is
important to make use of detailed data and modelling. Thus our analysis
involved adjusting the measures to account for hourly changes in capacity of
the system and market shares of players: despatch changes, as market
conditions (e.g., hydrological, fuel prices, etc) change, as players take planned
maintenance outages, schedule outages, forced outages and as plants run for
reliability-must-run conditions, as well an many other factors.
Another important issue is market definition. The analysis and modelling
considers the relevant geographic market to be the wholesale generation
market within a particular country, with each country assessed seperately.
Therefore, the role of interconnectors5 is limited to a sensitivity analysis
carried out for each of the markets to assess the potential impact of
interconnector linkages with neighbouring countries on the degree of
concentration in a particular market.
5 We did not have data allocating actual flows or capacity reservations to companies.
London Economics Page 7
February 2007
Section 1 Executive Summary
HHI and CR(n) measures were constructed for a selected number of capacity
measures (MW), as well as generation output (MWh).
Electricity-specific market structure indicators: RSI and PSI
Standard concentration measures based on market share may still be
insufficient as indicators of market power in electricity markets. If capacity is
tight relative to demand, even a small firm could exercise market power,
while if capacity is in excess relative to demand, even a large firm might
typically have to engage in tough price competition. In light of the difficulties
with standard market structure measures, electricity-specific market structure
indicators, namely the Residual Supply Index (RSI) and the Pivotal Supplier
Index (PSI), were estimated to account for the specificities of electricity
markets. Rather than focussing on market shares of the companies, as is the
case with the traditional concentration measures, these measures identify the
indispensability (or ‘pivotalness’) of companies to meeting the hourly
demand on the system. The more indispensable/pivotal a company is the
more market power that company is considered to have.
At the most basic level the PSI answers the question as to whether a company
can be considered to be pivotal to meeting demand in a particular hour. A
company is considered pivotal if the available capacity capable of being
supplied to the market, less the available capacity of the company of interest
is less than demand in that hour, then that company’s involvement in the
market is necessary to meeting demand.
The RSI is a generalised continuous form of the PSI. The RSI calculates the
indispensability of a specific company relative to the load in the particular
hour and thus provides for a continuous measure of market power.
Therefore, with the RSI one not only determines whether a company is
indispensable in a given hour but also one can assess the degree to which the
market is relying on this company’s available capacity to meet demand.
London Economics Page 8
February 2007
Section 1 Executive Summary
Market outcomes: price cost margins
To assess market outcomes we estimated price cost margins6. This approach
is regularly applied by competition economists and competition authorities
and has previously been applied to electricity markets. Lerner Indices (LI)
and Price-Cost Mark-ups (PCMU) were both calculated to assess the relative
difference between price and cost and as such allow for an assessment of the
likely competitive nature of the market that produced the result. The
calculation of both measures allows for a direct comparison with results
found previously in other studies, which may have only applied a single
methodology to calculating market outcome measures. The Lerner Index (LI)
relates the difference between price and cost to the price in the market,
whereas the Price-Cost Mark-up (PCMU) relates the price cost differential to
the cost, where the cost is the marginal cost of the system in both cases7.
6 These are the margins of the market exchange price of each country over the estimated
system marginal cost in each county. These were calculated hourly for each country. The
system marginal cost is the estimate of the perfectly competitive price. Thus, market price
above the perfectly competitive price estimate is an estimate of how competitive markets
have been.
7 It may facilitate one’s understanding of these measures if one keeps in mind that the value of
both the Lerner Index and the Price-Cost Mark-up in a perfectly competitive market is zero,
as in a perfectly competitive market price equals marginal cost.
London Economics Page 9
February 2007
Section 1 Executive Summary
In order to calculate market outcome measures, one first requires an estimate
of the marginal cost of the system. In electricity markets, due to demand
variations, marginal costs are likely to change on an hourly basis. To estimate
hourly marginal cost for the system, one that reflects the system’s technical
characteristics and dynamic constraints on plants, each of the six electricity
markets were modelled separately8 using GED’s Prosym software—a state-ofthe-art
dynamic electricity market simulation package.
This approach yielded an optimal system despatch, subject to system
constraints and dynamic constraints on plants, and an estimated marginal
cost. The marginal cost returned by the model was obtained by applying the
Marginal Average Cost approach in MARKETSYM. Under this approach the
model simulates the cost-minimizing commitment and despatch to serve
loads. Hourly generation and generation cost values for each station in the
model are reported and the system marginal cost equals the fuel cost of the
highest per MWh unit producing in each hour, albeit with some exceptions.
In estimating the system marginal cost, some stations were excluded from
setting the price as they are not truly marginal. Stations were excluded from
the ability to set price by prior specification based on the information
provided by the companies. Primarily, these units were stations that are
specified as Must Run, as these units are forced to run out-of-merit
operations, and units that have simulated generating levels at their reported
minimum stable generation level9 . Of the remaining units, the unit(s) with
the highest average cost, in each hour, is identified as marginal; their average
costs is the system Marginal Average Cost.
Start costs10 were excluded as they were determined to be small, between 0.1
and 0.2% of average fuel costs in each market.
8 Importantly, the model takes into account imports and exports which effectively took place during the
period. It means in particular that, whenever a market was exporting (this is especially true for France
and Germany), the effect of price increases created by the exports which took place during the period
has been factored in by the model.
9 Units at minimum stable generation in the simulation are units forced to run out-of-merit
operations in the simulation such as some Combined Heat and Power stations which are
not economic to run only for electricity purposes or plants called in to serve reserve
purposes in the simulation.
10 Generators typically incur a fuel cost to start their units. For some units, this can be
significant.
London Economics Page 10
February 2007
Section 1 Executive Summary
The result of the modelling was to produce, on an hourly basis, an optimal
system despatch and hourly marginal cost for each of the six markets. This
marginal cost was used in all of the outcome measures.
The price chosen to relate to the modelled marginal cost is also of
fundamental importance to the result of the analysis. In order for a set of
hourly prices to be deemed representative of the market and the prevailing
market conditions, it is important for these prices to adjust to market
conditions. During periods of relative scarcity of available installed capacity,
associated with periods of peak demand under normal conditions, one would
expect to observe higher prices. Therefore, it is important for this purpose to
use a price series that includes sufficient detail and variation to reflect the
market and its likelihood for change on an hourly basis. Prices from dayahead
electricity exchanges were assessed and found to reflect required
criteria and therefore were used: APX (NL), EEX (DE), PowerNext (FR),
OMEL (ES), and UKPX (GB). No such price was available for Belgium11.
Electricity market assessment prices from Platts were also used as a check in
all markets.
1.2.1.1 Relating the outcomes to market structure
A primary goal of the analysis was to assess how market structure impacts on
competitive outcomes. Since electricity-specific structural indicators of RSI
and outcome measures (price cost margins) were calculated hourly for three
years, a further and more general aspect of competition in electricity markets
could also be assessed; the link between the structure and outcomes of the
market. One would expect to see decreases in the RSI value (indicating an
increase in indispensability of the relevant company and thus an increase in
their market power) to be correlated with an increase in the margin or markup
earned in the market. An assessment of this relationship was undertaken
through the use of econometric analysis.
11 In Belgium, the BPI is a daily price for which both price and quantity are set by the largest
operator. See the Belgium chapter for details.
London Economics Page 11
February 2007
Section 1 Executive Summary
The optimal despatch modelling has also allowed us to calculate a number of
additional variables. The results of the GED modelling of each system have
allowed for a detailed assessment of the economic cost of carbon since the
introduction of the EU ETS in 2005. It has also allowed one to compare the
actual despatch with the results of the modelled optimal despatch for each
unit. It furthermore enabled us to calculate an estimated contribution to fixed
cost by company. This was mainly performed as a model check, to show that
the estimated competitive price (the optimal system marginal cost estimates)
were not so low that generators would not earn substantial sums towards
there past investments.
1.2.2 Data description
A major and unprecedented element of this study was the extent and
coverage of the data. The data for this study was largely taken from a DG
Comp database of responses to the questionnaires sent as part of the Sector
Inquiry into electricity markets. This database provided, for the purpose of
the measures and indicators calculated, all12 of the unit-specific data used for
each company included in the study. The highest levels of confidentiality
were maintained throughout. Crosschecks using GED’s European database
and other public sources were also performed.
At the most basic level the data provided in response to the sector inquiry
questionnaires included the name, location and owner13 of each generation
unit, for the period January 2003 to December 2005. Data are included on the
normal maximum operating capacity (MW) and technical characteristics of
each unit greater than 25MW. Units below 25MW were aggregated by
technology. Hence, the simulation took into account virtually all plants
relevant for the price-formation mechanism.
12 There were of course occasional problems with missing or inconsistent data. This was either
rectified with further questionnaires, resorting to public data, or other assumptions.
Details are in the chapters and annexes.
13 In the case of co-owned units, the ownership share of the company was provided.
London Economics Page 12
February 2007
Section 1 Executive Summary
The output of each listed unit was provided as the hourly net electrical
production (MWh) to be transported to the high voltage grid. These data
were analysed on a unit-by-unit basis in order to address issues such as
daylight-savings time, missing data points and inconsistent data points
(generation far in excess of the normal maximum operating capacity,
incorrect unit of measurement applied), which could affect the relative
consistency of the series. This process resulted in 26,304 hourly observations
on the net electrical production of all units greater than 25MW and a similar
number of observations for smaller units, less than 25MW, aggregated by
company and technology type.
The planned and actual maintenance schedules of each unit has also been
provided, and thus planned and forced outages were included in the
database and incorporated into concentration measures and the despatch
modelling.
Data on the heat rate (GJ/MWh) of each unit were provided by companies.
From this data standard heat rate curves were derived based on technological
profile and reference to GED’s worldwide database of plant performance
data. The average monthly cost of fuel (€/GJ) for each unit was also
provided.
The TSOs provided information on the interconnector links of each system,
capacities (MW) and flows (MWh). This data was not company-specific,
however, thus preventing an analysis of the reservation and use of capacity
over interconnectors as part of this report.
Data on power prices were taken from the following day-ahead power
exchange markets in each of the countries;
• Belgium – Belgian Power Index, an index reported by Electrabel of
daily prices of electricity it has agreed to buy and sell with
counterparties.
• France – Powernext day-ahead hourly price series.
• Germany – European Power Exchange (EEX) day-ahead hourly price
series.
• Netherlands – Amsterdam Power Exchange (APX) day-ahead hourly
price series.
• Spain - Compañía Operadora del Mercado Español de Electricidad,
S.A. (OMEL) day-ahead hourly price series.
• Great Britain – UK Power Exchange (UKPX) hourly price series.
London Economics Page 13
February 2007
Section 1 Executive Summary
The marginal cost used in the calculation of the outcome measures in the
country-specific chapters of the report are the result of the GED modelling of
the optimal despatch of the systems. The hourly returned marginal cost on
the system reflects the average generating cost of the most expensive
generating unit (€/MWh), excluding those units whose generation is
constrained. In the modelling scenarios for 2005 where CO2 emission
allowances were fully factored in, the data on the cost of CO2 was taken from
the EEX and applied in all countries, while the basic emission rates of units,
(data that was not readily available), was based on a benchmarking exercise
undertaken by GED. Individual units with known emission rates were
compared with the interpolated emission rates assigned and were seen to
compare favourable with the benchmark approach in all cases. Further
discussion of this is contained in the report and the Appendices.
London Economics Page 14
February 2007
Section 1 Executive Summary
1.3 Results
1.3.1 Summary of CR(n) and HHI
The first part of our analysis calculated concentration ratios and HHIs on an
hourly basis. The market shares, which form the basis of these market
structure indicators, were calculated based on both capacity and generation
bases. Results for the HHI across countries show large contrasts.
Table 1.1 below shows the traditional measure of market concentration for
each country. These are average figures based on hourly market shares based
on market shares formed on an available installed capacity basis.
Table 1.1: CR(n) and HHI average across countries
Country CR(n)14 HHI
BE Belgium 90.7% 8,307
DE Germany 54.1% 1,914
ES Spain 71.4% 2,790
FR France 92.6% 8,592
NL Netherlands 57.7% 2,332
GB Great Britain 32.6% 1,068
Source: LE
The respective chapters provide histograms of the CR(n) and HHI values to
show the impact of hourly variations.
14 The number of companies for the CR(n) calculation is different for each country. This is
because the concentration ratio of say the biggest company in FR and BE already raised the
concentration to a very high level. Therefore, BE and FR, n=1, Spain and NL n=2, and DE
and UK n=2.
London Economics Page 15
February 2007
Section 1 Executive Summary
The cross-country results with sensitivities showed that some countries are
highly concentrated (BE and FR), some countries range from moderately
concentrated to highly concentrated (DE, ES and NL), and GB ranges from
borderline unconcentrated to moderately concentrated15. There is some
variation over time and by sensitivities. Many sensitivities were carried out16,
but in general, the qualitative conclusions were not sensitive to the changes in
assumptions about market definition and how market shares were formed,
with the main exception being the impact of interconnection in the
Netherlands.
15 Commonly used but somewhat arbitrary thresholds are: HHI< 1000; unconcentrated;
1000<HHI<1800; moderately concentrated; HHI>1800, highly concentrated.
16 For more details see the individual country reports.
London Economics Page 16
February 2007
Section 1 Executive Summary
1.3.2 Market Outcome Measures
Our analysis also calculated price cost margins for each country where
relevant or possible17 using both the price-cost mark-up (PCMU) and the
Lerner Index (LI). Results on the calculated PCMU in each market are
presented in Table 1.2. We note that the results for Belgium and France are
not reported as they come with a strong caveat18, (as does their comparison
with other countries).
17 This was not done for BE. Our opinion was that the BPI was not a relevant comparator with
which to calculate margins as it is not an hourly exchange traded price. It further was
found not to be related to scarcity, and its quantity and price are set by the largest operator
in the market. The price is not likely to reflect market conditions. Margins for Belgium are
indicative only. We also found it useful to breakdown the Belgian price to include the cost
of carbon.
18 In France, the first issue that arose was that reported total available capacities for many
nuclear plants were substantially higher than the maximum actual generation over
substantial periods of time for many such plants. This led the simulation model most likely
to under-estimate the marginal cost in France, as nuclear is generally a very low marginal
cost technology, with high capital costs. The reality of this apparent data discrepancy
could be either explained by technical reasons on the operations of nuclear plant or by
some other less benign explanations (see the chapter of France for more details). We could
not resolve the issue further. The second issue was the amortisation of fixed costs in the
context of the particularly flat merit curve of France: we lacked data to address that further
issue. All in all, we judged the marginal cost results for France to be less reliable than other
countries. In Belgium, the difficulty with the data was that an economically meaningful
and hourly marginal or exchange price did not exist over the sample period. It was
therefore our opinion that estimating hourly price cost margins for Belgium could not be
done on a comparable footing relative to the other countries.
London Economics Page 17
February 2007
Section 1 Executive Summary
Table 1.2: Price Cost Margins
Period 2003-05 2003 2004 2005
Including carbon cost
based on weighted averages of GED SMC & Exchange prices
Germany 27% 59% 22% 15%
Spain 21% 26% 5% 28%
Netherlands 6% 33% -2% -5%
Great Britain 11% NA 4% 13%
Excluding carbon cost
based on weighted averages of GED SMC & Exchange prices
Germany 51% 59% 22% 72%
Spain 35% 26% 5% 66%
Netherlands 14% 33% -2% 13%
Great Britain 31% - 4% 42%
Source: LE
The cross-country comparison shows a variety of margins19 across countries
and time. On average the lowest margins are in the GB and the Netherlands
(11% and 6% respectively). Germany and Spain have the highest and second
highest margins on average about 27% and 21% respectively. Negative
margins in the Netherlands are likely a result from CHP and other thermal
units (including Must-Run units) producing in reality more (to serve heat or
must-run purposes and thus "dumping" such electricity on the market) than
what could be assessed from the data supplied by the generators20. Also, in
Great Britain in 2005, margins may have been overestimated due to the fact
that very high gas prices might have given the incentive to generators to sell
their gas rather than generate. The opportunity cost of such sales was not
modelled (and thus not reflected in the cost), as we relied on the reported fuel
costs. These results are broadly in line with what one might expect given
the results of the concentration measures, HHI and CR(n).
19 We note that the margins presented here are the margins of the load weighted-average of the
price and cost, with the weights being formed by the share of total annual load reported in
each hour.
20 See chapter on the Netherlands for more details.
London Economics Page 18
February 2007
Section 1 Executive Summary
One can easily assess the impact of the introduction of the EU ETS on the
PCMU values in these markets from the table above. For 2005, we ran a
scenario "with carbon" (whose results are shown in the first part of the table
above) and a scenario "without carbon" (whose results are shown in the
second part of the table above). In the scenario "with carbon", the cost of a
given plant includes the value of carbon for that plant whereas in the scenario
"without carbon", the cost of the plant does not include the value of carbon.
The value of carbon for a given plant (per MWh) is estimated by multiplying
the carbon emissions of that plant (tonne/MWh) by the value of a tonne of
carbon (€/t), as provided by the market of CO2 allowances. Each scenario
("with carbon" and "without carbon") led to a marginal cost of the system in
each hour21 and the impact of ETS is equal to the difference between those
two marginal costs22. . From the table, it can be seen that without carbon, all
countries studied had significant margins above cost. Even with carbon fully
priced in, only the Netherlands did not have substantial margins. The
presence of carbon changes the rank order of the countries’ margins in 2005.
Germany had the highest margins without carbon, while Spain had the
highest margins with carbon. This is a function of the carbon intensity of the
technologies in the countries.
1.3.3 Breakdown of power price including margin, fuel
cost and carbon
The discussion above in the immediately preceding section discusses
margins. We now turn to the total breakdown of the exchange price into its
constituent parts, including carbon pricing/marginal cost impact23. Table 1.3
presents annual load weighted values of each of these variables.
21 A different unit may set the price in each scenario.
22 Please note that this is the maximum possible impact of ETS as the scenario "with carbon" relies on the
assumption that companies fully price in the value of CO2 allowances. It is unclear how far electricity
generators have explicitly factored in the prices of CO2 allowances into their electricity prices and to
which extent they should be allowed to do so from a policy perspective. In a pending German
competition case regarding this issue, the Bundeskartellamt judges that at maximum 25% of the costs
for CO2 certificates should be factored in.
23 In this section, we present France and Belgium along with the other four countries. But we
recall the caveat that direct comparisons across countries’ margins should be made with
extreme caution.
London Economics Page 19
February 2007
Section 1 Executive Summary
Table 1.3: Contribution to Power Price (€/MWh)
Country 2003 2004 2005
BE Belgium24
Sys Modelled MC € 29.75 € 31.70 € 50.40
Carbon € 0.00 € 0.00 € 10.11
Mark-Up € 11.31 -€ 0.70 -€ 10.23
Total € 41.06 € 31.00 € 50.28
BE INDEX Price € 41.06 € 31.00 € 50.28
DE Germany
Sys Modelled MC € 19.46 € 24.27 € 28.17
Carbon € 0.00 € 0.00 € 13.86
Mark-Up € 11.42 € 5.36 € 6.39
Total € 30.88 € 29.63 € 48.42
EEX Price € 30.88 € 29.63 € 48.42
ES Spain
Sys Modelled MC € 23.95 € 27.51 € 33.65
Carbon € 0.00 € 0.00 € 10.12
Mark-Up € 6.29 € 1.39 € 12.20
Total € 30.24 € 28.89 € 55.97
OMEL Price € 30.24 € 28.89 € 55.97
FR France
Sys Modelled MC € 11.09 € 12.92 € 15.63
Carbon € 0.00 € 0.00 € 3.65
Mark-Up € 18.96 € 15.98 € 28.85
Total € 30.05 € 28.90 € 48.13
POWERNEXT Price € 30.05 € 28.90 € 48.13
NL Netherlands
Sys Modelled MC € 36.26 € 34.64 € 50.50
Carbon € 0.00 € 0.00 € 9.52
Mark-Up € 11.99 -€ 0.63 -€ 3.09
Total € 48.24 € 34.01 € 56.93
APX Price € 48.24 € 34.01 € 56.93
24 We have included Belgium and France in this case because we felt it was worthwhile to
document the estimated impacts of carbon. Please recall the previous (footnoted) caveat of
the comparison of Belgian and French margins to the others; The French data on nuclear
plants (availability, costs) raised several issues; the Belgian prices, the BPI and the Platts
prices were not sufficiently detailed nor sufficiently related to hourly supply and demand
outcomes and scarcity for explicit analysis of the margins.
London Economics Page 20
February 2007
Section 1 Executive Summary
Table 1.3: Contribution to Power Price (€/MWh) (Continued)
Country 2003 2004 2005
GB/Great Britain25
Sys Modelled MC - € 33.33 € 39.06
Carbon - € 0.00 € 10.00
Mark-Up - € 1.25 € 6.35
Total - € 34.58 € 55.41
UKPX Price - € 34.58 € 55.41
Note: all values in this table are load weighted average values.
Source: LE
From the data in the table, one first notices a range of values in the marginal
impact of carbon, from €3.65 to €13.86. This result is largely in accordance
with what one would expect given the different generation portfolios of each
of the countries concerned. The carbon impact was estimated by taking the
emissions of the marginal plant in each hour (based on generation, efficiency
and fuel type) and multiplying by the market price of carbon. In terms of
cost, the marginal impact of carbon depends on how much carbon intensive
technology was on the margin in each hour in any particular market, as well
as its thermal efficiency. Therefore, the impact of carbon on France’s
estimated marginal cost is low (weighted average €3.65/MWh). This is
because technology in France tends to be less carbon intensive26. The carbon
intensity of nuclear and hydro (prevalent in France) is zero. Conversely, for
countries such as the UK and Germany where coal and gas are the marginal
plant in many hours, these are carbon intensive technologies (and likely, the
marginal plants will have lower than average thermal efficiency), the impact
is more pronounced. In these cases, the marginal impact of carbon in 2005 is
substantial—Germany having the highest cost of carbon at €13.86/MWh.
25 Hourly price data from the UKPX is only available from July 2004 onwards. Therefore, there
is no result for 2003 under this approach and the result for 2004 should be viewed in the
light of the data availability issue.
26 Nuclear is setting the price in a significant number of hours in the modelled optimal
despatch. This is effectively because pumped storage was modelled by subtracting its
energy from the estimate of demand.
London Economics Page 21
February 2007
Section 1 Executive Summary
We also see a variety of margins. Some countries, such as the Netherlands,
show an average negative margin in 2004 and 2005 (recall this is mostly due
to CHP and other plant not shutting down and operating out-of-merit).
Others, such as France show quite high margins (over 100%-although a
strong caveat once again applies; in addition, the fixed cost of nuclear would
require higher margins). Note here, the margins as calculated are including
carbon at full market cost (i.e., the amounts reported are the maximum
possible impact of the ETS if generators fully factor in the price of CO2
certificates in a competitive environment), this therefore represents a
somewhat conservative approach to 2005.
It is possible that recent prices in the power exchanges studied are not
reflective of the opportunity cost of carbon, since companies received their
emissions trading allowances free. In this case, the mark-ups presented
above would be higher. Whether companies’ should price in carbon fully is
another question, which we did not address.
London Economics Page 22
February 2007
Section 1 Executive Summary
1.3.4 Contribution to fixed cost
In addition to estimating the price cost margins, we also estimated the total
value by country (DE, ES, NL, and GB)27, by company and by year of the
contributions to fixed cost. This figure is found by multiplying the quantity
each unit produces under an optimal despatch by the difference between the
estimated modelled system marginal cost and the average fuel cost of the
unit, as reported28. The sum over each company’s units in every hour is then
the annual total. This was mainly done as a crosscheck of the credibility of
the marginal cost estimates. The rationale was to check whether the
estimated marginal cost from the optimal despatch was credible as a perfectly
competitive price such that it would still allow substantial contributions to
fixed costs29. The ability of companies to contribute to and cover fixed costs is
a strong determinant on future investment decisions in the sector, as well as
on the sustainability of current operations for individual companies.
These figures are found in Table 1.4 for the top 4 companies in each country.
We note that all of the top four companies in the largest markets (DE, ES, and
GB) and the largest (by this measure) company in the NL would have earned
(had they traded all their power at the simulated competitive price)
contributions to their fixed costs in excess of a billion euro over the three-year
period. The smaller companies in the Netherlands each would have earned a
little more or less than €500m.
27 As we did not estimate regressions for France and Belgium and the reliability of margin
estimates for these countries must come with a strong caveat, we did not estimate fixed cost
contribution figures for these countries either.
28 Algebraically, this can be represented by; CFC = Qm
ix(SMC – ACi)); where CFC is
contribution to fixed cost, Qm
i s is electricity generated by operators M and plant i and SMC
is system marginal cost and ACi are the average costs for plant i.
29 These calculations were not meant as a study of the investment decisions or incentives for
investment in each country. This was beyond the scope of this study.
London Economics Page 23
February 2007
Section 1 Executive Summary
This analysis is useful in that it highlights a number of factors. The analysis
indicates that the estimated competitive prices/marginal cost estimates are
not generally so low that companies would not earn an operating profit30. It
also suggests a consistency with the incentive to invest, but this result is
merely indicative, as one should note that investment decisions should be
analysed on a more micro basis.
Furthermore, one should note that for existing market participants, a
substantial proportion of their portfolio is likely to be partially or fully
amortised already thus reducing the need to cover such costs.
30 Further analysis shows that on a per MW basis, these numbers are in light with the
amortized cost of a generic new entrant for all countries studied except Spain. We note that
there will be a variety of vintages of capital stock, and so this represents a relatively high
hurdle in terms of the fixed costs needed for commercial operations.
London Economics Page 24
February 2007
Section 1 Executive Summary
Table 1.4: Contribution to Fixed Costs for largest companies (€'000)
Company 2003 2004 2005 Total
DE Germany
1338-S-DE 1,818,142.0 2,521,370.0 2,782,871.0 7,122,383.0
0436-S-DE 1,761,631.0 2,094,008.0 3,084,305.0 6,939,944.0
1681-S-DE 784,310.1 1,078,262.0 1,227,499.0 3,090,071.1
0569-S-DE 635,197.4 817,859.9 1,358,937.0 2,811,994.3
ES Spain
0577-S-ES 853,472.1 1,029,273.0 1,394,689.0 3,277,434.1
0875-S-ES 387,150.9 580,013.1 671,181.4 1,638,345.4
1646-S-ES 110,710.3 183,125.7 351,760.2 645,596.2
0850-S-ES 135,484.0 121,964.2 126,943.5 384,391.7
NL Netherlands
0712-S-NL 375,394.9 251,917.0 633,023.8 1,260,335.7
0511-S-NL 256,825.5 211,884.0 142,804.9 611,514.4
1193-S-NL 223,846.0 166,949.7 286,358.6 677,154.3
0439-S-NL 210,725.3 149,863.3 79,092.7 439,681.3
GB Great Britain
0242-S-GB 1,706,199.0 1,808,226.0 2,995,477.0 6,509,902.0
0453-S-GB 606,512.1 643,986.9 685,716.3 1,936,215.3
1340-S-GB 422,153.5 448,385.1 732,872.3 1,603,410.9
1477-S-GB 326,060.2 451,280.3 375,499.3 1,152,839.8
Source: LE
London Economics Page 25
February 2007
Section 1 Executive Summary
1.3.5 Electricity-specific measures of market structure:
RSI for top companies
Our analysis also included estimating electricity market specific measures of
market structure. These measures included the residual supply index (RSI)
and the pivotal supplier index (PSI) in relation to the biggest two to four
firms in each country. These indicators are electricity-specific indicators used
to give a richer measure of market structure in electricity markets than
standard concentration measures. They measure the degree to which an
individual supplier is ‘pivotal’ and thus required to meet demand. If a
supplier is pivotal, it is hypothesized that they have the power to raise price
unilaterally and profitably. The RSI is a continuous measure of pivotalness
and the PSI is a zero-one measure. An RSI<110% indicates the supplier is
near pivotal in that hour31. The summary results across countries for the RSI
are contained in Table 1.5.
The results (in % of hours RSI<110%) show a wide range across countries, as
well as across companies within a country. The result of comparing the RSI
measures across countries shows some suppliers are indispensable (required
to meet demand) in all hours. In the most concentrated countries, some
suppliers are near pivotal or close to being required to meet load in up to
100% of hours. In other countries, such as GB, the number of hours is less
than 5% that any of the top companies are in breach of the threshold and
therefore determined to be indispensable to meeting the load. Finally, it is
noteworthy that for both Spain and Germany, only two of the top four
suppliers are seen to be pivotal in a large percentage of hours.
31 One should note that this threshold value is not the result of a rigorous assessment of the
suitability of such a value to each market but rather represents a broad guiding principle
rather than a steadfast rule on indispensability based on studies previously carried out in
California by the CAISO.
London Economics Page 26
February 2007
Section 1 Executive Summary
Table 1.5: Summary of country RSIs for largest companies (2003-2005)
Country Company % hours RSI<110%
BE Belgium 0513-S-BE 100.0%
1469-S-BE 5.0%
DE Germany 0436-S-DE 47.7%
0569-S-DE 4.6%
1338-S-DE 77.1%
1681-S-DE 3.8%
ES Spain 0577-S-ES 41.1%
0850-S-ES 0.0%
0875-S-ES 49.2%
1646-S-ES 0.6%
FR France 0340-S-FR 0.5%
0472-S-FR 100.0%
1449-S-FR 0.0%
NL Netherlands 0439-S-NL 3.5%
0511-S-NL 32.8%
0712-S-NL 44.6%
1193-S-NL 22.7%
GB Great Britain 0242-S-GB 1.2%
0453-S-GB 1.7%
1340-S-GB 1.2%
1477-S-GB 2.3%
Source: LE
London Economics Page 27
February 2007
Section 1 Executive Summary
1.3.6 Regression analysis
1.3.6.1 Single variable models
The next and more groundbreaking part of our analysis was the regression or
econometric analysis of the relationship between market structure variables
to market outcomes, while controlling for random error and other factors.
This is a test of the hypothesis that firms that are pivotal have market power.
We estimated the impact of market structure on both market outcome
measures (PCMU and LI). For this reason, we undertook regression analysis
of PCMU and LI on RSI. Given the RSI is a continuous variable of the
pivotalness of an individual operator measured hourly, it can be seen as a
suitable candidate explanatory variable for the price cost margin using
regression analysis32. We started this with simple (single explanatory
variable) regression analysis—by regressing the price cost mark up on the RSI
relative to the largest companies in each market. A graphical depiction of the
results from the estimation of one of the specified simple regression equations
is presented in Figure 1.1.
32 Previous studies have used the HHI as the measure of market structure in regression
analysis in electricity markets but the variable has been found to perform poorly and is
generally not considered to be appropriate to such analysis.
London Economics Page 28
February 2007
Section 1 Executive Summary
Figure 1.1: Estimated impact of RSI on margins: Spain
-10123
PCMup5/Fittedvalues
.5 1 1.5 2 2.5
RSI_C01
PCMup5 Fitted values
Source: LE
The figure represents the estimated impact of the RSI for company 0577-S-ES
on the price-cost mark-up in Spain for the full sample period. The line
represents the predicted value, while the blue points the actual outcomes.
Deviations from the line are thus modelled as random error. The results of all
the univariate regressions are presented in the table below for the regressions
of PCMU on RSI33.
33 The results from the table above are the marginal impacts of RSI on the price cost mark up.
The RSI is a measure that is near one when a large supplier is exactly pivotal. So a coefficient
estimate of –1 would mean that essentially a 1% change in the RSI would lead to a 1% change
in the price cost margin (mark up). The expected sign is negative, since more capacity means
the market is less tight, and thus the RSI will rise, and margins would then be expected to fall.
The t-values are the test statistic for regression coefficients. The t-statistics are all statistically
significant for each RSI coefficient and for each company and country. The R-squared values
range from about 11% to 35%, which is a reasonable fit for this type (single explanatory
variable) of regression. The significance of the single variable regression models can be
interpreted as follows: there is a near zero percent chance that the margins are not related to
the RSI variables, even when not holding other factors constant.
London Economics Page 29
February 2007
Section 1 Executive Summary
Table 1.6: Regression results relating margins (PCMU) to market structure
(RSI) for largest generators
Company-Country Variable Name Coefficient Std. Err. t R-squared
0577-S-ES RSI -1.242 0.010 -120.0 35.6%
0875-S-ES RSI -1.385 0.012 -118.5 35.1%
0453-S-GB RSI -0.90 0.015 -58.8 20.8%
1340-S-GB RSI -0.87 0.015 -58.4 20.6%
1477-S-GB RSI -0.87 0.015 -56.4 19.5%
0436-S-DE RSI -2.36 0.034 -69.1 15.4%
0569-S-DE RSI -2.00 0.030 -66.7 14.5%
1338-S-DE RSI -2.43 0.042 -57.5 11.2%
1681-S-DE RSI -1.92 0.029 -67.0 14.6%
0511-S-NL RSI -1.22 0.021 -57.2 11.1%
0712-S-NL RSI -1.22 0.021 -57.2 11.1%
Source: LE
London Economics Page 30
February 2007
Section 1 Executive Summary
1.3.6.2 Multiple variable models
To test and develop our models and understanding further, we built and
tested additional regression models. We tested the sensitivity of the models
to violations in the classical linear regression assumptions and we tested a
variety of specifications. We included dummy (zero-one) variables for peak
and off peak, summer, winter, spring, and allowed the estimated RSI slope
coefficients to vary with daily peak and off peak periods.34 Additional
explanatory variables were included to account for differences between
modelled and actual generation by fuel type, competitive constraints posed
by competitors and relative scarcity of available installed capacity. A number
of further specifications included RSIs of more than one company as
regressors explaining the margins. In general, with few exceptions, the RSI
for large companies remained a statistically significant explanatory variable
for price cost margins in all four countries.
1.4 Conclusions and summary
This study has been a comprehensive and quantitative investigation into the
structure and functioning of six major EU electricity markets. Our study has
been groundbreaking in a number of ways, but especially in that, it is the first
in the EU to utilise generators’ own data to estimate a system marginal cost
and to subsequently construct market outcome measures based on these costs
and relate them to detailed electricity specific structural measures.
34 Additional regressions and model output details can be found in the country chapters.
London Economics Page 31
February 2007
Section 1 Executive Summary
The concentration of the markets was measured by traditional (CRn and
HHI) and more innovative measures (RSI and PSI) of market structure.
Concentration, in general as measured by traditional concentration measures,
shows marked differences across country, but little variation over time or by
method within countries. Some markets, such as GB, are borderline
unconcentrated except in some hours where it can become concentrated (see
chapter); others such as Spain, Germany and the Netherlands are moderately
to highly concentrated, while France and Belgium are highly concentrated.
Our results were, in general, not sensitive to a variety of factors and
sensitivities, such as the allocation of the interconnector. In most cases, our
qualitative conclusions were not sensitive to changes in assumptions. There
were notable exceptions, such as the allocation of interconnection for the
Netherlands (who controls what shares over the interconnector will likely
change the overall picture of whether the Netherlands is a concentrated or
unconcentrated market, however current domestic regulation can be seen to
moderate this impact).
Results from the RSI threshold test show that broadly similar outcomes occur.
The most concentrated markets fail the RSI threshold test in a very large
percentage of hours, with respect to the largest companies (e.g., 100% in
France and Belgium). On the other hand, in Great Britain the threshold test is
breached in only a small number of hours across all of the companies
assessed. Spain, the Netherlands and Germany have two companies with
significant market positions approximately 20% to 77% of the time.
London Economics Page 32
February 2007
Section 1 Executive Summary
Comparing the market outcome measures (LI and PCMU) and the actual
breakdown of costs also shows interesting results and differences across time
and country. The differences across time were most interesting. All countries
had large mark-ups in 2003, while countries diverged after that, with some
country’s margins falling (NL, DE) others rising (ES). Countries with lower
costs seem to have higher margins for example, but no broad consensus
correlation can be observed. GB had low margins, while the Netherlands had
the lowest margins; but GB is less concentrated than the Netherlands. France
had the highest margins but we view the results as not comparable to the
others due to data difficulties. There are also important differences in
margins across time. It is difficult to interpret these results, so the additional
analysis of RSIs and regressions is needed. We also note again that the
finding of negative margins is not uncommon in these types of studies35.
Difficulties in the data created by reported available capacity and actual
running capacity created challenges in modelling the marginal cost in France,
and the lack of a good hourly price for Belgium meant that estimates of
margins in these countries should be considered less reliable than in the other
four. In addition, further analysis such as regressions, was not carried out for
these countries.
35 Negative margins are common because it is often the case that factors such as CHP and
must-run units will continue to run even when the market price is below marginal cost.
This may cause a near excess of energy on the system. In addition, many thermal units
with long starting and stopping times will be willing to pay a premium to avoid shutting
down and being thus unavailable to generate when prices are more profitable.
London Economics Page 33
February 2007
Section 1 Executive Summary
Of potentially greater interest is the breakdown of the power exchange price
into components including carbon. We do not fully know whether
companies have passed on the full cost of carbon or whether they have
“raised” margins in response to carbon. In spite of the fact that utilities
obtained their emissions allowances for free, one would expect them to price
in carbon costs fully, unless they believed doing so would lead to reduced
carbon allowances in future rounds. However, we take no particular view on
what utilities should in fact do with their carbon allowances. By assuming
they are priced in, we essentially take the most conservative approach to
finding high margins (i.e., the amounts reported are the maximum possible
impact of the ETS if generators fully factor in the price of CO2 certificates in a
competitive environment). What we can conclude from our results is that
seemingly differential factors have occurred across countries, and the size of
the margin changes across countries with the introduction of EU ETS, while
the total marginal cost of carbon as estimated also changes.
We also compared estimates of the contribution to fixed costs. This section
indicated that had the largest companies traded at the theoretically
competitive price (equal to the estimated optimal marginal cost), they would
have earned billions of euro towards their investment costs. This confirms
the robustness of the model in terms of the overall level of estimated
competitive price.
Finally, we estimated regression models for four countries using a variety of
explanatory factors and specifications. The regression results included the
estimation of the impact of RSI on margins and estimates including a variety
of explanatory variables. The results showed that the RSI, a continuous
measure of the indispensability of individual large suppliers, significantly
explains market outcomes for almost all companies considered in all markets.
This result is apparently robust to controlling for a number of factors,
including model specification, changes in the assumptions of the statistical
model, and inclusion of other explanatory factors such as scarcity, year,
seasonality, and the modelled differences between actual and modelled
generation of coal and gas. The results are consistent with the hypothesis that
pivotalness gives firms market power in electricity markets, but does not
necessarily prove this.
Of the evidence found, the most important is the detailed regression results.
These results suggest that prices in the selected EU wholesale markets have
been raised due to the pivotalness of particular suppliers. This suggests
prices have not been as keen as they might have been.
London Economics Page 34
February 2007
Section 1 Executive Summary
It is difficult to say with 100% certainty whether any one country’s market is
exhibiting the exercise of market power. The regression results suggest that
market structure is significantly related to market outcome measures in all
the markets studied (DE, ES, NL and GB). In conclusion we can say that this
report represents a considerable advancement in the state of knowledge
about how electricity markets function (have been functioning) in Europe as
previously, no such relation had been studied in such detail in the EU’s major
electricity markets.
London Economics Page 35
February 2007
Section 1 Executive Summary
Glossary
Available capacity
The capacity a generation plant has available to produce in any given hour
or time period.
BPI Belgian Power Index
CR(n) Concentration ratio for the n largest firms
CHP Combined heat and power
EEX German hourly electricity spot price and market
EU ETS The EU Emissions Trading Scheme
FERC The US Federal Energy Regulatory Commission
Fixed cost costs that do not vary with the amount of power produced
GED Global Energy Decisions
GJ Giga Joule
GWh Gigawatt hour
Heat rate
Measure of the thermal efficiency of an electric generator in terms of thermal
fuel input and electric energy output
HHI Hirschman-Herfindahl index
Installed capacity The capacity a plant is technically able to produce based on technology
kWh Kilowatt hour
LE London Economics
LI Lerner Index: (P - MC)/P
Load Hourly electricity demand on the system
Merit curve
The curve showing the unit cost (€/MWh) as a function capacity (MW) of
each power plant on a system; the supply curve in a power system.
Merit order
The rank order of power plant on a system in terms of lowest to highest unit
cost €/MWh
MW Megawatt (capacity)
MWh Megawatt hour
OMEL Spanish hourly electricity spot price and market
PCMU Price cost mark-up: (P - MC)/MC
Platts An energy data company
PowerNext French hourly electricity spot price and market
Prosym
An electricity market despatch simulation model based on GED’s
MARKETSYM software.
PSI Pivotal supplier indicator
Regression
An econometric estimation technique used to find the impact of a number of
independent variables on a dependent variable
RSI Residual supply index
Start cost Costs incurred by the generation station when starting
TSO Transmission system operator
UKPX UK hourly electricity spot price and market
London Economics Page 36
February 2007